text
stringlengths 1
1.05M
| meta
dict |
|---|---|
Published: July 5, 2018
Introduction {#sec1}
============
The small GTPase RAS family proteins (KRAS, NRAS, and HRAS) are controlled through the exchange of the GDP-bound form for the GTP-bound one, which then allows RAS to bind various effectors, such as RAF, phosphoinositide 3-kinase (PI3K), and RALGDS ([@bib6], [@bib10], [@bib26], [@bib42], [@bib44]). RAS-associated signaling pathways play important roles in multiple cellular functions, such as cell growth, migration, adhesion, survival, and differentiation. The mutations at hotspots (G12, G13, and G61) in the *RAS* genes cause accumulation of the GTP-bound form due to defective intrinsic GTP hydrolysis activity and resistance to GTPase-activating proteins ([@bib34]). These oncogenic mutations in the *RAS* genes are observed in approximately 30% of all human cancers. *KRAS* is one of the most widely known oncogenes and is frequently found to be mutated in colorectal, pancreatic, and lung cancers ([@bib1]).
Oncogenic *KRAS* has been reported to play a significant role in stem cell activities in some types of cancers. For example, it has been shown that oncogenic *KRAS* in colon cancers enhances the embryonic stem (ES) cell-like program during human colon cancer initiation from adenoma to carcinoma, and activates cancer stem cell (CSC) properties in *APC*-mutated cells through the MAPK pathway ([@bib20], [@bib27]). In addition, oncogenic *KRAS* has been reported to enhance stemness in CSCs in pancreatic cancers through the PI3K/AKT/mammalian target of rapamycin pathways ([@bib24]).
The mutations in the RAS pathway are known to be involved not only in cancers, but also in other disorders including a series of congenital diseases and an acquired hemato-immunological disease, namely, RAS-associated autoimmune lymphoproliferative syndrome (ALPS)-like disease (RALD). RALD has been reported as a disease affecting the hemato-immune system, caused by a somatic *KRAS* or *NRAS* mutation in hematopoietic lineage cells. RALD patients exhibit ALPS- and/or juvenile myelomonocytic leukemia-like symptoms, including autoimmune cytopenia, lymphadenopathy, and hepatosplenomegaly ([@bib29], [@bib38], [@bib39]). Moreover, a RALD patient exhibiting intestinal Behcet\'s disease-like phenotypes was reported ([@bib28]). In RALD, individual patients have clones with *KRAS* or *NRAS* mutation and wild-type clones together in hematopoietic lineage cells in a mosaic state, allowing the generation of a set of isogenic induced pluripotent stem cell (iPSC) clones from the same patients. RALD patient-derived iPSCs therefore represent a unique experimental tool that is useful for studying basic RAS biology, particularly the roles of KRAS on stemness maintenance in the context of PSCs.
In the culture of human embryonic stem cells (ESCs) and iPSCs, basic fibroblast growth factor (bFGF) is essential to maintain their stemness through activating the MAPK and PI3K pathways. If human ESCs and iPSCs are cultured without bFGF, they lose their stemness and start to differentiate ([@bib8], [@bib12], [@bib19], [@bib21], [@bib22]). These observations clearly demonstrate the importance of bFGF-mediated signaling for the maintenance of human iPSCs and ESCs. However, it remains largely unknown how the status of effector molecules including KRAS located downstream in bFGF signals affects stemness maintenance in human iPSCs.
Here, we investigated the roles of KRAS on stemness maintenance in the context of human iPSCs by using isogenic *KRAS* mutant (G13C/WT) and wild-type (WT/WT) iPSCs, generated from two RALD patients with the same somatic *KRAS* mutation. By genome-editing techniques, we succeeded in generation of "gene-corrected" wild-type iPSCs (WT^ed^/WT) and heterozygous knockout iPSCs (Δ^ed^/WT), both of which could serve as relevant controls for the experiments. Using this series of isogenic iPSCs, we determined how the status of *KRAS* could impact upon stemness maintenance in human iPSCs and differentiation propensity under permissive conditions.
Results {#sec2}
=======
Establishment of iPSC Clones from RALD Patients {#sec2.1}
-----------------------------------------------
We generated iPSCs from CD34^+^ hematopoietic stem/progenitor cells of two RALD patients with the same somatic G13C heterozygous mutation in the *KRAS* gene ([Tables S1](#mmc1){ref-type="supplementary-material"} and [S2](#mmc1){ref-type="supplementary-material"}). We obtained mutant (G13C/WT) and isogenic wild-type (WT/WT) iPSC clones from each patient as confirmed by direct sequencing ([Figure 1](#fig1){ref-type="fig"}A). The presence of oncogenic mutations other than *KRAS* was excluded by whole exome sequencing ([Table S3](#mmc2){ref-type="supplementary-material"}). Karyotyping showed that all RALD patient-derived iPSC clones exhibited a normal 46XY karyotype ([Figure 1](#fig1){ref-type="fig"}B). All iPSC clones expressed the markers, OCT4, NANOG, TRA-1-60, and SSEA4 ([Figure 1](#fig1){ref-type="fig"}C).Figure 1Establishment and Characterization of iPSC Clones Generated from Two RALD Patients(A) *KRAS* sequences of wild-type (WT/WT) and mutant (G13C/WT) iPSC lines derived from two RALD patients (cases no. 1 and no. 2). A position of mutation (G to T) is indicated by red letters.(B) Karyotypes of WT/WT and G13C/WT iPSC lines derived from case no. 2 (RALD patient).(C) Immunocytochemistry of iPSC markers (OCT4, NANOG, TRA-1-60, and SSEA4) in WT/WT and G13C/WT iPSC clones. Ho, Hoechst 33342. Scale bar, 100 μm.See also [Tables S1](#mmc1){ref-type="supplementary-material"} and [S2](#mmc1){ref-type="supplementary-material"}.
To assess how the status of KRAS affects stemness and differentiation in these iPSC clones, we investigated changes in gene expression levels of stemness and lineage markers after *in vitro* embryoid body (EB)-mediated differentiation for 16 days. In the suspension culture to induce differentiation, all iPSC clones formed EBs ([Figures 2](#fig2){ref-type="fig"}A and [S1](#mmc1){ref-type="supplementary-material"}A). However, RNA sequencing (RNA-seq) analysis showed that there were clear differences in stemness and lineage marker expression between WT/WT and G13C/WT genotypes in case no. 2 ([Figure 2](#fig2){ref-type="fig"}B). Following 16-day differentiation, mRNA levels of stemness markers decreased over time in the WT/WT genotype, whereas they seemed to remain high in the G13C/WT cells, especially R2-1 clone. In general, mRNA levels of lineage markers were elevated upon differentiation in both WT/WT and G13C/WT cells. The expression levels of endodermal and early mesodermal markers were higher in the G13C/WT cells than in the WT/WT counterparts, whereas those of ectodermal markers in the G13C/WT genotype were lower. The same trend in expression differences of stemness (*POU5F1* and *NANOG*), mesodermal (*EOMES* and *T*), and ectodermal markers (*PAX6* and *ASCL1*) was confirmed in qRT-PCR analysis of case no. 2 ([Figure 2](#fig2){ref-type="fig"}C). However, mRNA levels were comparable in two genotypes regarding endodermal markers (*FOXA2* and *SOX17*) after differentiation. Although *PAX6* mRNA levels after differentiation were comparable between groups, the clear difference in two genotypes was also observed on *POU5F1*, *NANOG*, and *ASCL1* in qRT-PCR analysis of case no. 1 ([Figure S1](#mmc1){ref-type="supplementary-material"}B). Thus, we focused on the marked differences of stemness and ectodermal markers in the two genotypes.Figure 2Different Differentiation Propensity between WT/WT and G13C/WT iPSCs Generated from Two RALD Patients(A) Embryoid body formation of WT/WT and G13C/WT iPSC clones from case no. 2. Scale bar, 200 μm.(B) RNA-seq data showing gene expression levels of stemness and lineage markers from case no. 2 samples before and after 16-day *in vitro* differentiation. Undiff. and Diff., undifferentiated iPSCs and differentiated cells, respectively.(C) qRT-PCR analysis of 16-day *in vitro* differentiated cells from WT/WT and G13C/WT iPSC clones derived from case no. 2 (n = 3 independent experiments; mean ± SEM).(D) Immnunocytochemistry of βIII-Tubulin and MAP2 in 16-day *in vitro* differentiated cells from WT/WT and G13C/WT iPSC clones derived from case no. 2. Scale bar, 50 μm.See also [Figures S1--S3](#mmc1){ref-type="supplementary-material"}.
Considering that both *PAX6* and *ASCL1* are transcription factors essential for neurogenesis ([@bib2], [@bib7], [@bib31]), we compared expression of neuronal markers βIII-Tubulin and MAP2 between WT/WT and G13C/WT iPSCs after 16-day differentiation ([Figure 2](#fig2){ref-type="fig"}D). In differentiated cells harboring the WT/WT genotype, many βIII-Tubulin- and MAP2-double-positive neurons with long extending neurites could be seen. In contrast, there were fewer βIII-Tubulin- and MAP2-double-positive neurons differentiated from G13C/WT iPSCs. There were some neurites positive for βIII-Tubulin but negative for MAP2 in clone R2-1. In the case of clone R2-3, neurites of βIII-Tubulin- and MAP2-double-positive cells were few in number and very short in length. Similar observations were also made from immunostaining of case no. 1-derived iPSC line differentiated cells. βIII-Tubulin-positive neurons were seen in differentiated C1-1 cells, whereas no positive cells were detected in differentiated R1-1, R1-2, or R1-3 cells ([Figure S1](#mmc1){ref-type="supplementary-material"}C). These results indicate that oncogenic *KRAS* impacts upon self-renewal capacity and differentiation propensity in human iPSCs.
Rescue of the *KRAS* Mutation by Genome Editing {#sec2.2}
-----------------------------------------------
To validate whether the phenotypes described above were attributable to the specific *KRAS* genotype, we attempted to generate "gene-corrected" wild-type iPSCs from a G13C/WT clone. For genome editing, we used the CRISPR/Cas9 system for one G13C/WT iPSC clone from RALD patient no. 1 (clone R1-2) ([Figure S2](#mmc1){ref-type="supplementary-material"}A). Through homologous recombination, we were able to obtain genome-edited wild-type homozygous clones (clone; C8 and H2, WT^ed^/WT; "ed" meaning genome-edited). In addition, we could obtain heterozygous knockout clones (Δ^ed^/WT) from the same genome-editing experiment (clone; A1 and D2, [Figure S2](#mmc1){ref-type="supplementary-material"}B). We checked the top five off-target candidate sites of the single guide RNA used in this procedure and observed no off-target cleavage ([Figure S2](#mmc1){ref-type="supplementary-material"}C). Western blotting analysis confirmed that KRAS expression levels were comparable in WT^ed^/WT and G13C/WT clones (clone; F4 and G4, which are non-edited clones). KRAS expression was lower in Δ^ed^/WT clones compared with the other two cell types, consistent with the haploinsufficient state ([Figure S2](#mmc1){ref-type="supplementary-material"}D).
These clones were subjected to 16-day *in vitro* differentiation and examined for the expression of stemness genes and three-germ layer markers. In the suspension culture to induce differentiation, all clones formed EBs ([Figure S3](#mmc1){ref-type="supplementary-material"}A). Analysis by qRT-PCR showed that expression levels of *POU5F1* and *NANOG* remained high in G13C/WT cells (F4 and G4) after 16-day differentiation, although they decreased in WT^ed^/WT (clone; C8 and H2) and Δ^ed^/WT cells (clone; A1 and D2). Transcription of *ASCL1* mRNA in G13C/WT cells was not induced upon differentiation, whereas it showed clear induction in WT^ed^/WT and Δ^ed^/WT cells ([Figure S3](#mmc1){ref-type="supplementary-material"}B). Expression of the neuronal lineage markers, βIII-Tubulin and MAP2, also became more intense in WT^ed^/WT clones in comparison with G13C/WT mutant cells ([Figure S3](#mmc1){ref-type="supplementary-material"}C). These results demonstrated that gene-correction in the *KRAS* allele was enough to normalize the phenotypes inherent to iPSCs carrying the G13C/WT genotype.
Enforced Retention of Self-Renewal by G13C/WT iPSCs in the Absence of bFGF {#sec2.3}
--------------------------------------------------------------------------
Based on the results of *in vitro* differentiation assays, we speculated that there could be different expression patterns of genes relating to stemness, self-renewal, and pluripotency, between WT/WT and G13C/WT iPSCs. RALD patient-derived iPSCs were cultured with or without bFGF for 5 days, after which global gene expression was monitored by microarray. Scatterplots between the two combinations of iPSC lines and culture conditions are depicted in [Figure 3](#fig3){ref-type="fig"}A. In the presence of bFGF, there was a high degree of correlation in gene expression pattern between WT/WT and G13C/WT iPSCs ([Figure 3](#fig3){ref-type="fig"}A, upper left). However, the extent of correlation between WT/WT and G13C/WT cells decreased when they were cultured without bFGF for 5 days ([Figure 3](#fig3){ref-type="fig"}A, upper right). Comparison in gene expression for WT/WT iPSCs between two conditions, i.e., with and without bFGF, showed marked changes, characterized by downregulation of stemness genes (*POU5F1* and *NANOG*) and upregulation of ectodermal genes (*PAX6*) in the absence of bFGF ([Figure 3](#fig3){ref-type="fig"}A, lower left). In sharp contrast, there were no changes in gene expression observed in G13C/WT cells even when they were cultured without bFGF ([Figure 3](#fig3){ref-type="fig"}A, lower right).Figure 3Microarray Analysis of Isogenic WT/WT and G13C/WT KRAS Mutant iPSCs from a RALD Patient and Those Cultured without bFGF for 5 Days(A) Scatterplots with coefficients of correlation (*R*) for whole genes in WT/WT and G13C/WT iPSCs cultured with (w/) or without (w/o) bFGF for 5 days. Two clones per genotype were used: C2-1 and C2-2 for WT/WT; R2-1 and R2-2 for G13C/WT, derived from case no. 2. In this analysis, data of the same genotypes were averaged. Positions of *POU5F1* (two probes), *NANOG*, and *PAX6* (two probes) are indicated. The green lines indicate the diagonal and 2-fold changes between the two samples.(B) A heatmap with hierarchical clustering for stemness and lineage marker expression in the same samples as described above.
[Figure 3](#fig3){ref-type="fig"}B is a heatmap of stemness and linage markers. Stemness markers, including *POU5F1*, *NANOG*, and *MYC*, were expressed at a high level in both WT/WT and G13C/WT iPSC lines cultured in the presence of bFGF (four lanes in the middle).
When cells were cultured without bFGF, the expression of these genes diminished in the case of the WT/WT genotype (two lanes on left, C2-1 and C2-2; w/o). On the contrary, they remain highly expressed, almost unchanged in the G13C/WT genotype (two lanes on right, R2-1 and R2-2; w/o). The dendrogram indicated that the expression pattern was more closely related between G13C/WT samples cultured with bFGF and those without bFGF than between G13C/WT iPSCs and WT/WT iPSCs both in culture with bFGF. This was also reflected in the results of the scatterplots in [Figure 3](#fig3){ref-type="fig"}A. Among linage markers, ectodermal marker expression tended to be elevated even after 5-day culture without bFGF in WT/WT cells (two lanes on left, C2-1 and C2-2; w/o), but it did not in bFGF-depleted G13C/WT cells (third and fourth lanes from left, R2-1 and R2-2, w/o). Taken together, these microarray data demonstrate enforced retention of self-renewal of G13C/WT iPSCs in the absence of bFGF.
Correlation between *KRAS* Genotypes and Stemness Maintenance Potentials Quantified by an Imaging Approach {#sec2.4}
----------------------------------------------------------------------------------------------------------
To analyze the enforced retention of self-renewal at the protein level, the expression of stemness markers was confirmed by immunocytochemistry staining. Upon bFGF removal, cell morphology changed to a more flattened appearance and the immunocytochemical expression of stemness markers (OCT4 and NANOG) dramatically decreased in WT/WT iPSCs ([Figure 4](#fig4){ref-type="fig"}A, left panel, C1-1, C2-1, and C2-2). In contrast, cell morphology and stemness marker expression in G13C/WT iPSCs remained almost unaffected (R1-2, R2-1, and R2-3). Regarding other stemness markers, TRA-1-60 also decreased in expression, whereas SSEA4 did not do so much in WT/WT iPSCs by the depletion of bFGF for 5 days ([Figure 4](#fig4){ref-type="fig"}A, right panel). Quantitative imaging analysis of bFGF-deprived cells showed that OCT4^+^ rates of G13C/WT clones (R1-1, -2, and -3) remained higher than 75% of vehicle treatment, whereas those of the WT/WT clone C1-1 reduced to levels less than 25% ([Figure 4](#fig4){ref-type="fig"}B). Similar results were obtained in iPSC clones derived from case no. 2 ([Figure 4](#fig4){ref-type="fig"}C). Combining the data of the three RALD iPSC clones for each genotype from case no. 2, we examined the effects of genotype and bFGF depletion on OCT4 expression by using two-way ANOVA, and obtained statistically significant results ([Table S4](#mmc1){ref-type="supplementary-material"}).Figure 4Enforced Retention of Self-Renewal of *KRAS* G13C/WT Mutant iPSCs in the Absence of bFGF, Revealed by Immunocytochemical Reactivity for Stemness Markers, Colony Formation, and Alkaline Phosphatase Staining(A) Immunocytochemistry of stemness markers (OCT4, NANOG, TRA-1-60, and SSEA-4) in WT/WT and G13C/WT iPSC clones from RALD patients cultured without bFGF for 5 days. Scale bar, 100 μm.(B and C) Quantitative imaging analysis for OCT4 in iPSC clones from RALD patients, cases no. 1 (B) and no. 2 (C), respectively (n = 8 independent experiments; mean ± SEM; ^∗∗∗^p \< 0.001; two-way ANOVA followed by Bonferroni\'s multiple comparison test).(D) Whole six-well (left) and magnified images (right) of alkaline phosphatase (ALP)-stained WT/WT (C2-1) and G13C/WT (R2-1) cells. Scale bar, 100 μm.(E) Quantification of ALP-positive (ALP^+^) colony number in (D) (n = 3 independent experiments; mean ± SEM; ^∗∗^p \< 0.01; ^∗∗∗^p \< 0.001; two-way ANOVA followed by Bonferroni\'s multiple comparison test).See also [Figure S3](#mmc1){ref-type="supplementary-material"}; [Table S4](#mmc1){ref-type="supplementary-material"}.
We next compared oncogenic *KRAS* genotype-associated changes in OCT4^+^ rates in genome-edited iPSCs after removal of bFGF. The OCT4^+^ area of WT^ed^/WT cells was drastically reduced compared with G13C/WT cells. Interestingly, the OCT4^+^ area of Δ^ed^/WT cells was significantly reduced compared with that of WT^ed^/WT cells ([Figure S3](#mmc1){ref-type="supplementary-material"}D). The detailed summary of statistical analysis is shown in [Table S4](#mmc1){ref-type="supplementary-material"}.
We then investigated if mutant iPSCs retaining stemness marker expression after bFGF depletion also maintained pluripotency. To this end, the iPSCs pre-depleted of bFGF for 5 days were reseeded as single cells and cultured for another 7 days back in bFGF-replete culture. As shown in [Figure 4](#fig4){ref-type="fig"}D, G13C/WT cells formed many iPSC colonies which were alkaline phosphatase (ALP)-positive, whereas only a few colonies appeared in WT/WT cells ([Figure 4](#fig4){ref-type="fig"}D). ALP^+^ colony numbers in G13C/WT genotype were significantly higher than those in WT/WT genotype ([Figure 4](#fig4){ref-type="fig"}E). These results confirmed oncogenic *KRAS* genotype-associated enhancement in retention of self-renewal capacity in iPSCs.
Biochemical Analysis on KRAS Activity and Downstream ERK and AKT Pathways {#sec2.5}
-------------------------------------------------------------------------
To determine the mechanisms underlying the *KRAS* mutant-specific phenotypes as described above, we first examined the existing amount of KRAS-GTP by GST-RAF1 pull-down assays. We observed significantly larger amounts of KRAS-GTP in G13C/WT iPSCs compared with WT/WT iPSCs under both conditions with and without bFGF ([Figures 5](#fig5){ref-type="fig"}A and 5B). Importantly, KRAS-GTP levels still remained high in G13C/WT iPSCs, but not in WT/WT iPSCs, even in the absence of bFGF, indicating constitutive activation of KRAS in G13C/WT iPSCs ([Figures 5](#fig5){ref-type="fig"}A and 5B). Consistent with these results, the same pattern was observed when genome-edited WT^ed^/WT clones were compared with G13C/WT clones ([Figures S4](#mmc1){ref-type="supplementary-material"}A and S4B). Furthermore, KRAS-GTP levels in Δ^ed^/WT clones were significantly reduced compared with those in WT^ed^/WT clones in the absence of bFGF ([Figures S4](#mmc1){ref-type="supplementary-material"}A and S4B).Figure 5Biochemical Analysis on the ERK and AKT Pathways Activity in Enforced Retention of Self-Renewal of *KRAS* G13C/WT iPSCs(A) GST-RAF1 pull-down assays of RALD patient-derived iPSC (WT/WT and G13C/WT) clones from case no. 1 and no. 2 cultured with (w/) or without (w/o) bFGF for 3 days.(B) Densitometric analysis of western blotting results shown in (A). Each point indicates an individual clone\'s value (n = 4--6 independent clones; mean ± SEM; ^∗∗∗^p \< 0.001; Student\'s t test or Mann-Whitney test).(C) Western blot analysis of WT/WT and G13C/WT iPSC clones stimulated for the indicated time course after the removal of bFGF for 3 days. ERK and AKT were analyzed for their phosphorylation. β-Tubulin was used as an internal control.(D) Western blot analysis of WT/WT and G13C/WT clones cultured w/ or w/o bFGF for 2 days. Phosphorylation of ERK and AKT was analyzed as in (C).(E and F) Densitometric analysis of ERK (E) and AKT (F), respectively, in (D). Each point indicates an individual clone\'s value (n = 4--6 independent clones; mean ± SEM; ^∗∗∗^p \< 0.001; Student\'s t test or Mann-Whitney test). GST, glutathione S-transferase; RBD, Ras-binding domain.See also [Figure S4](#mmc1){ref-type="supplementary-material"}.
Next, we analyzed the phosphorylation levels of ERK and AKT in G13C/WT and WT/WT iPSCs to determine downstream signaling activity in KRAS-signaling pathways. Upon bFGF stimulation for the indicated times after its removal, the ERK pathway was prominently upregulated in both cell types, whereas it showed only slight upregulation of pAKT ([Figure 5](#fig5){ref-type="fig"}C). To see if sustained activation of the downstream pathways from KRAS occurred in G13C/WT iPSCs even after bFGF deprivation, we compared phosphorylation levels of ERK and AKT after culturing cells with or without bFGF. Although there was only a marginal increase in pERK in G13C/WT iPSCs compared with WT/WT counterparts in the presence of bFGF (p = 0.0936), significantly higher pERK levels were maintained in the mutant clones than the control in the absence of bFGF (p \< 0.001) ([Figures 5](#fig5){ref-type="fig"}D and 5E). On the other hand, there was no significant difference in pAKT levels between WT/WT and G13C/WT iPSCs in either case with or without bFGF ([Figures 5](#fig5){ref-type="fig"}D and 5F). Using genome-edited iPSC clones, retention of pERK expression in the absence of bFGF was also confirmed for G13C/WT clones when compared with WT^ed^/WT clones. Notably, Δ^ed^/WT clones showed pERK levels that were significantly lower than those in WT^ed^/WT clones in the absence of bFGF (p \< 0.05), which is similar to aforementioned results and consistent with the idea that these artificial control clones exhibit haploinsufficiency and thus "loss-of-function" phenotypes ([Figures S4](#mmc1){ref-type="supplementary-material"}C and S4D). These results indicate that bFGF stimulation appears to act through the ERK signaling pathway and that there is sustained activity of KRAS-ERK signaling in G13C/WT iPSCs despite bFGF withdrawal.
Pharmacological Analysis on the Involvement of the MEK-ERK and PI3K Pathways in the Enforced Retention of Self-Renewal and Suppressed Neuronal Differentiation Propensity of G13C/WT iPSCs {#sec2.6}
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Finally, to investigate which downstream pathway from KRAS was responsible for the retained OCT4 expression in the mutant clones, we performed pharmacological intervention using kinase-specific inhibitors targeting the RAF-MEK and PI3K pathways ([@bib14], [@bib17], [@bib25], [@bib33], [@bib37], [@bib43]) ([Figure S5](#mmc1){ref-type="supplementary-material"}A). As expected, the retention of OCT4 expression after bFGF removal in G13C/WT iPSCs was reversed by treatment with MEK inhibitors (PD184352 and U0126) ([Figures 6](#fig6){ref-type="fig"}A and [S5](#mmc1){ref-type="supplementary-material"}B). Western blot analysis showed that pERK levels were diminished by MEK inhibitors at the same concentrations that induced the inhibition of OCT4 expression, while pAKT levels were not affected ([Figure 6](#fig6){ref-type="fig"}B). As for RAF inhibitors, AZD628 decreased the OCT4^+^ rate, whereas ZM336372 led to an increased OCT4^+^ rate ([Figures 6](#fig6){ref-type="fig"}C and [S5](#mmc1){ref-type="supplementary-material"}C). Consistent with these results, AZ628 decreased pERK levels, but ZM336372 treatment led to an increase in pERK ([Figure 6](#fig6){ref-type="fig"}D). Clone R2-1 also displayed the same results as clone R1-2 ([Figures S5](#mmc1){ref-type="supplementary-material"}E and S5F).Figure 6Pharmacological Analysis on the Involvement of the RAF-MEK-ERK and PI3K-AKT Pathways in Enforced Retention of Self-Renewal of *KRAS* G13C/WT iPSCs(A and B) Effects of MEK inhibitors (PD184352 and U0126) on OCT4^+^ area (A) (n = 3 independent experiments; mean ± SEM) and phosphorylation of ERK and AKT (B), respectively, in G13C/WT iPSCs (clone R1-2).(C and D) Effects of RAF inhibitors (ZM336372 and AZ628) on OCT4^+^ area (C) (n = 3 independent experiments; mean ± SEM) and phosphorylation of ERK and AKT (D), respectively, in G13C/WT iPSCs (clone R1-2).(E) Representative fluorescent images of G13C/WT iPSCs (clone R1-2) treated with the compounds at indicated concentrations. Vehicle, 0.1% DMSO. Scale bar, 100 μm.See also [Figure S5](#mmc1){ref-type="supplementary-material"}; [Table S5](#mmc1){ref-type="supplementary-material"}.
When these compounds diminished OCT4^+^ areas, a part of OCT4^--^ cells exhibited flattened morphology especially in the treatment of MEK inhibitors, which resembled that of WT/WT iPSCs cultured without bFGF ([Figure 6](#fig6){ref-type="fig"}E). Regarding PD184352, the concentrations that induced 20%, 50%, and 80% (IC~20~, IC~50~, and IC~80~) inhibition relative to vehicle were determined for all clones ([Figure S5](#mmc1){ref-type="supplementary-material"}G; [Table S5](#mmc1){ref-type="supplementary-material"}). Among PI3K inhibitors, LY294002 showed a partial decrease in the OCT4^+^ rate, whereas wortmannin did not affect that, although it prominently decreased pAKT levels ([Figures S5](#mmc1){ref-type="supplementary-material"}D, S5H, and S5I).
To ask whether the MEK-ERK pathway was also responsible for the suppressed neuronal differentiation observed in G13C/WT iPSCs, mutant iPSCs (clones R1-2 and R2-1) were treated with PD184352 during 16-day *in vitro* differentiation. qRT-PCR analysis showed that mRNA expression of *ASCL1* increased following treatment with PD184352 in a concentration-dependent manner ([Figure 7](#fig7){ref-type="fig"}A). Immunoreactive βIII-Tubulin appeared in greater density in differentiated cells from G13C/WT iPSCs following treatment with PD184352 in a concentration-dependent manner ([Figure 7](#fig7){ref-type="fig"}B). These data support the notion that *KRAS* mutant-specific phenotypes depend mostly on the RAF-MEK-ERK pathway, but not the PI3K-AKT pathway.Figure 7Effects of a MEK Inhibitor on Suppressed Neuronal Differentiation in G13C/WT iPSCs(A) qRT-PCR analysis of a neuronal lineage marker (*ASCL1*) in 16-day differentiated cells from WT/WT (clones C1-1 and C2-1) and G13C/WT iPSCs (clones R1-2 and R2-1) treated with a MEK inhibitor (PD184352) at the concentrations of IC~20~, IC~50~, and IC~80~ values as described in [Table S5](#mmc1){ref-type="supplementary-material"}, or 0.1% DMSO (vehicle) (n = 3 independent experiments; mean ± SEM; ^∗^p \< 0.05; ^∗∗^p \< 0.01, ^∗∗∗^p \< 0.001; one-way ANOVA followed by Dunnett\'s test).(B) Immnunocytochemistry of βIII-Tubulin in 16-day differentiated cells from WT/WT and G13C/WT iPSCs treated with PD184352 as described above. Scale bar, 100 μm. PD, PD184352.
Discussion {#sec3}
==========
In the present study, we established *KRAS* mutant and isogenic wild-type iPSC clones from two RALD patients sharing the same somatic heterozygous *KRAS* G13C mutation. In comparison with the isogenic iPSC clones, we showed that activated KRAS with the G13C mutation conferred retained self-renewal of undifferentiated cells and difficulty in differentiation into neurons under permissive conditions. In addition, we generated genome-edited wild-type (WT^ed^/WT) and heterozygous knockout (Δ^ed^/WT) iPSCs from a G13C/WT clone. The phenotypes of parental G13C/WT iPSCs were rescued in WT^ed^/WT iPSCs, and lower self-renewal and higher neuronal differentiation potentials compared with WT^ed^/WT iPSCs was observed in Δ^ed^/WT iPSCs. Accordingly, we concluded that the phenotypes are *KRAS* G13C mutation specific.
First, we found retained high expression of stemness markers in G13C/WT iPSCs, which was observed not only when cells were cultured without bFGF for 5 days, but also in 16-day *in vitro* differentiation. Global gene expression analysis identified differential expression profiles between G13C/WT and WT/WT iPSCs in genes relating to self-renewal of undifferentiated cells, including *POU5F1* and *NANOG*. Their expression was retained at a high level despite the removal of bFGF, suggesting enforced retention of self-renewal of G13C/WT iPSCs. Furthermore, clear differences in the size of OCT4^+^ cell populations between WT/WT and G13C/WT iPSCs could be detected in the 5-day bFGF-depletion assays. The capacity for retaining OCT4 expression was removed by rescuing the G13C allele (WT^ed^/WT), and lowered with the G13C-specific knockout (Δ^ed^/WT). Thus, we concluded that the enforced retention of self-renewal was a *KRAS* mutation-specific phenotype. In general, bFGF is an essential factor for maintaining human ESC and iPSC pluripotency through downstream MEK-ERK and PI3K-AKT pathways ([@bib12], [@bib19], [@bib22]). We postulated that the retained self-renewal of undifferentiated cells observed in G13C/WT iPSCs was caused by the constitutive activation of KRAS and its downstream signals even though iPSCs were cultured without bFGF. We attempted to detect *KRAS* mutation-specific hyper-activation of the KRAS pathways. As expected, RAF1 pull-down assays showed hyper-activated KRAS in G13C/WT iPSCs in the absence of bFGF. Western blot analysis showed that the ERK pathway was largely upregulated, but only slight upregulation was observed in the AKT pathway upon temporal stimulation with bFGF. Moreover, pERK was more upregulated in G13C/WT iPSCs compared with WT/WT iPSCs, but such a trend was not observed in the case of pAKT when cultured without bFGF for a few days. Taken together, the MEK-ERK pathway, but not the PI3K pathway, was considered to be the main pathway involved in the bFGF signal resulting in *KRAS* mutant-specific phenotypes.
To further investigate whether the ERK pathway was responsible for the enforced retention of self-renewal, we performed pharmacological analysis using specific kinase inhibitors. MEK inhibitors (PD184352 and U0126) decreased OCT4^+^ area rates of G13C/WT iPSCs cultured in the absence of bFGF in a concentration-dependent manner. In terms of RAF inhibitors, paradoxical results were obtained between AZ628 and ZM336372; AZ628 lowered OCT4^+^ area rates, whereas ZM336372 elevated them. It has been reported that some RAF inhibitors block MEK-ERK signaling in cells with mutant BRAF, but enhance signaling in cells with wild-type BRAF. These RAF inhibitors include PLX4720 (vemurafenib), GDC-0879 (BRAF inhibitors), and ZM336372 (a CRAF inhibitor) ([@bib17], [@bib18], [@bib32]). Interestingly, among the RAF inhibitor examined, AZ628, a pan-RAF inhibitor, does not induce phosphorylation of MEK or hyper-proliferation of KRAS^WT^/BRAF^WT^ cells ([@bib18]). Our data from G13C/WT iPSCs were in agreement with these previous reports. The PI3K-AKT pathway is also known to be important in maintaining stemness of human ESCs ([@bib12], [@bib22]). The addition of LY294002 in particular led to a decrease in OCT4^+^ area rates, while wortmannin did not affect them, although pAKT was effectively lowered by wortmannin treatment. Recently, LY294002 has been reported to act as not only a PI3K inhibitor, but also as a bromodomain and extraterminal domain (BET) inhibitor ([@bib13]). BRD4, a member of the BET family of proteins, is required for maintenance in human ESCs through the regulation of pluripotency genes, including *POU5F1* and *NANOG* ([@bib11]). Assuming that the effects of LY294002 on OCT4^+^ area rates were caused by BET inhibition, the involvement of PI3K in the retained self-renewal of undifferentiated cells would be moderate under our assay conditions. This is because wortmannin, another chemotype PI3K inhibitor, did not decrease OCT4^+^ area rates at all. Therefore, it could be demonstrated that the enforced retention of self-renewal in G13C/WT iPSCs derived from RALD patients was caused mainly by retained KRAS activation and subsequent prolonged hyper-activation of the RAF-MEK-ERK signals.
In this study, we found suppression of neuronal differentiation from G13C/WT iPSCs as shown in 16-day *in vitro* differentiation experiments. So far, mutations in the RAS pathway have been reported to be implicated in neural diseases. The RASopathies, such as Noonan syndrome, Costello syndrome (CS), cardio-facio-cutaneous syndrome, and neurofibromatosis type 1 (NF-1), are developmental syndromes caused by germline mutations in genes involved in the RAS-MEK pathways. These disorders are all characterized by neurological symptoms, such as mental retardation, cognitive impairment, and increased anxiety. These abnormalities are thought to originate from dysregulated differentiation of neuronal progenitor cells, based on structural CNS characteristic to the RASopathy patients, and the results obtained from animal model studies ([@bib3], [@bib9]). In neural differentiation, bFGF-ERK signaling is known to inhibit *PAX6* induction and act as an inhibitory signal to neural induction in human ESCs ([@bib15]). Moreover, activated RAS signaling caused by mutations of *KRAS* or *NF-1* has been reported to lead to enhanced proliferation of neuronal progenitor cells and aberrant differentiation into neurons in mouse models (e.g., shorter neurites) ([@bib4], [@bib5], [@bib36]). Recently, Rooney et al. reported the generation of CS patient-derived iPSCs harboring a heterozygous constitutively active G12S mutation in HRAS. By demonstrating an increased population of PAX6^+^/MAP2^−^ cells and shorter neurite length in neurons differentiated from HRAS-mutated iPSCs, they suggested that activated HRAS led to an extended neuronal progenitor phase and difficulty in maturation of neurons ([@bib36]). In this study, we manifested the decreased expression of the neuronal markers, particularly *ASCL1*, and reduced βIII-Tubulin^+^/MAP2^+^ cells in G13C/WT iPSC lines cultured for 16 days in a differentiation-permissive condition. Despite taking into consideration the differences in experimental settings between studies, our results would further strengthen the idea that constitutive RAS activation in general likely has the potential to affect fates of stem cells under differentiation-permissive conditions. We speculated that the perturbation of neuronal differentiation in G13C/WT iPSCs occurred mainly due to sustained KRAS activity despite withdrawing bFGF. The inhibited neuronal differentiation was restored completely by the rescue of the G13C allele using genome-editing technology or partially by the treatment of an MEK inhibitor, PD184352. Our findings are in agreement with the previous paper reporting that FGF signaling inhibits neural induction, and inhibition of FGF/ERK upregulates a neuroectodermal fate determinant in human ESCs ([@bib16]).
Overall, our data indicated the significance of KRAS status for the maintenance of iPSC stemness, which would eventually affect aspects such as self-renewal and differentiation propensity. The roles of oncogenic *KRAS* in stemness maintenance have been implicated in some types of stem cells. Oncogenic *KRAS* plays important roles in the imposition of ESC-like transcriptional profiles and upregulation of CSC markers, leading to the tumor initiation and development shown in colorectal and pancreatic cancers ([@bib20], [@bib24], [@bib27]). In a model of retinoic acid (RA)-induced stem cell differentiation to endoderm, oncogenic *KRAS* confers on cells resistance to differentiation associated with the maintenance of stem cell characteristics despite RA treatment ([@bib35]). In addition, oncogenic *KRAS* promotes self-renewal properties and tumor initiation by suppressing non-canonical WNT signaling ([@bib45]). These studies suggest that the involvement of oncogenic *KRAS* in stemness maintenance might be a common feature among various types of stem cells, especially in CSCs. Taken together, the RALD patient-derived isogenic iPSCs would be applicable to research not only on pathology of RALD but also certain aspects of CSCs as a general tool for studying oncogenic *KRAS*-driven stem cell dysfunctions. Using this unique and invaluable tool, the development of pharmacological treatment will likely be achieved for RALD, CSCs, and other oncogenic *KRAS*-related diseases in the future.
Experimental Procedures {#sec4}
=======================
Ethics {#sec4.1}
------
Bone marrow cell samples from patients were used in accordance with the Declaration of Helsinki. Written informed consent for samples to be used for research purposes was obtained from the patient or patient's parents. The study was approved by the Ethics Committee of The Institute of Medical Science, The University of Tokyo (protocol number, 25-3-0701), the Ethics Committee of Tokyo Medical and Dental University (approval numbers, 676), and the Eisai Research Ethics Committee (approval numbers, 2014-0238, 2014-0259, 2015-0238, 2015-0259, 2016-0238, and 2016-0259). The use of viral vectors was approved by the ethics committees of The Institute of Medical Science, The University of Tokyo.
Establishment of iPSCs from RALD Patients {#sec4.2}
-----------------------------------------
Establishment of iPSCs from CD34^+^ hematopoietic stem/progenitor cells was performed as described previously ([@bib23], [@bib41]) at The Institute of Medical Science, The University of Tokyo. Mononuclear cells were extracted by Ficoll density gradient separation from whole bone marrow aspirate obtained from the two individual RALD patients. CD34^+^ cells were isolated from this mononuclear fraction by immunomagnetic separation using a CD34 MicroBead Kit (Miltenyi Biotec) according to the manufacturer\'s protocol. Non-integrating Sendai virus (SeV) vectors harboring human *POU5F1*, *SOX2*, *KLF4*, and *c-MYC* ([@bib30]) were used to transduce isolated CD34^+^ cells for reprogramming into RALD patient-derived iPSCs. Lipofectamine RNAi Max (Thermo Fisher Scientific) was used to transfect established RALD patient-derived iPSCs with small interfering RNA L527 for the removal of SeV vectors. Established clones were screened for mutations in the *KRAS* gene by sequencing genomic DNA. A proportion of CD34^+^ cells obtained from both patients did not carry any *KRAS* mutation. The resulting iPSC clones free of the mutation were retained and used as isogenic controls. Karyotyping was performed at Nihon Gene Research Laboratories (Sendai, Japan).
iPSC Maintenance {#sec4.3}
----------------
RALD patient-derived iPSC clones were introduced into Eisai Tsukuba Laboratories from The Institute of Medical Science. Detailed information of the clones is given in [Table S2](#mmc1){ref-type="supplementary-material"}. In the Eisai Tsukuba Laboratories, iPSCs were maintained on vitronectin (Thermo Fisher Scientific)-coated plates in a culture medium for iPSCs, StemFit (Ajinomoto) added with StemFit\'s supplement package including bFGF and 1× penicillin-streptomycin (P/S; Wako). At passage, iPSCs were dissociated with TrypLE Select (Thermo Fisher Scientific) and cultured with 10 μM Y-27632 (Wako), and the medium was changed to Y-27632-free StemFit the next day. The medium was replaced every 2 days, and iPSCs were passaged every 3 or 4 days. Cell culture was carried out at 37°C in a humidified atmosphere of 20% O~2~. In this study, we used iPSCs at the passage numbers less than 30 after the introduction into Eisai.
*In Vitro* EB-Mediated Differentiation {#sec4.4}
--------------------------------------
*In vitro* differentiation was performed basically according to procedures as described previously ([@bib40]). iPSCs were harvested after cell dissociation with TrypLE Select, and seeded to six-well ultra-low attachment plates (Corning) in bFGF-free StemFit containing 10 μM Y-27632 (Wako) to produce EBs. On the following day, the medium was changed to a fresh one without Y-27632. After the 8-day suspension culture, EBs were collected and transferred to 0.1% gelatin (Sigma)-coated plates to initiate adherent culture. Cells were allowed to differentiate in bFGF-free StemFit containing 10% fetal bovine serum (Thermo Fisher Scientific) for additional 8 days. The media were exchanged basically every 2 days.
Immunocytochemistry {#sec4.5}
-------------------
Cells were fixed by 4% paraformaldehyde (Wako) for 30 min at room temperature. They were blocked and permeabilized in 4% (w/v) Block Ace (AbD Serotec)/cell staining buffer (20 mM Tris-HCl \[pH 7.5\], 150 mM NaCl, and 0.2% Triton X-100) for 30 min at room temperature. Then, cells were incubated with primary antibodies in 0.4% (w/v) Block Ace/cell staining buffer overnight at 4°C, and subsequently incubated with appropriate secondary antibodies with Hoechst 33342 (10 μg/mL; Sigma-Aldrich) in 0.4% (w/v) Block Ace/cell staining buffer for 1.5 hr at room temperature. Detailed conditions of antibodies are given in [Table S6](#mmc1){ref-type="supplementary-material"}. Fluorescence images were acquired on an IX73 (Olympus) or IN-Cell Analyzer (GE Healthcare).
bFGF-Depletion Assays {#sec4.6}
---------------------
Cells were dissociated by TrypLE Select, and 2 × 10^3^ cells were seeded on vitronectin-coated 96-well plates in StemFit containing Y-27632 (10 μM) with or without bFGF and cultured for 5 days. Medium was changed on day 1, to remove Y-27632, and on day 3. Immunocytochemistry of stemness markers was performed as described above. In quantitation, OCT4^+^ and nucleus (Hoechst 33342-stained) areas were measured by an IN-Cell Developer (GE Healthcare). OCT4^+^ rates and expected total cell numbers were calculated as follows:-OCT4^+^ rate (%) = total OCT4^+^ area/Hoechst-positive area × 100-Expected total cell numbers = total Hoechst-positive area/median of individual Hoechst-positive area
Colony-Formation Assays and ALP Staining of Reseeded iPSCs that Had Been Cultured without bFGF {#sec4.7}
----------------------------------------------------------------------------------------------
RALD iPSCs (C2-1 and C2-2 of WT/WT clones; R2-1 and R2-2 of G13C/WT clones) were plated and cultured in StemFit without bFGF for 5 days. After harvesting cells, they were reseeded into six-well plates at 1.0 × 10^4^ cells/well (n = 3), and cultured in StemFit with bFGF. When colonies were formed 7 days after reseeding, cells were stained for ALP using nitrotetrazolium blue chloride (400 μM; Sigma-Aldrich) and 5-bromo-4-chloro-3-indolyl phosphate p-toluidine salt (400 μM; Sigma-Aldrich) in ALP buffer (100 mM NaCl, 100 mM Tris-HCl, and 50 mM MgCl~2~ \[pH 9.5\]). ALP^+^ colony number was measured using BZ-X710 (Keyence).
Pharmacological Examinations {#sec4.8}
----------------------------
We examined effects of specific inhibitors related to RAS pathways ([@bib14], [@bib17], [@bib25], [@bib33], [@bib37], [@bib43]) on bFGF-depletion assays and *in vitro* differentiation. PD184532 and wortmannin were purchased from Sigma. U0126 was obtained from Wako. ZM336372, AZ628, and LY294002 were purchased from Selleck. To exclude the effect of cell toxicity, the concentrations of each compound were selected at the range of lower than 80% cell growth inhibition judged by the percentage of nucleus number relative to the vehicle treatment in bFGF-depletion assays ([Figures S5](#mmc1){ref-type="supplementary-material"}B--S5D).
Author Contributions {#sec5}
====================
K.K., K.Y., M.I., K.T., M.T., and M.O. conceived and designed the study. K.K., K.Y., Y.I., and T.N. performed the experiments and analyzed the data. K.N., M.O., M.N., H.-T.L., M.T., and M.O. contributed to generate the patient-derived iPSCs. K.K., K.Y., H.-T.L., M.T., and M.O. wrote the manuscript. M.I., K.T., T.M., M.T., and M.O. supervised the project.
Accession Numbers {#app1}
=================
Whole exome sequencing data are available in the Japanese Genotype-phenotype Archive (JGA, <http://trace.ddbj.nig.ac.jp/jga>), which is hosted by the DNA Data Bank of Japan (DDBJ), under accession number JGAS00000000136. RNA-seq and microarray data are available in the GEO under accession numbers GEO: [GSE111345](ncbi-geo:GSE111345){#intref0015} and [GSE94141](ncbi-geo:GSE94141){#intref0020}, respectively.
Supplemental Information {#app3}
========================
Document S1. Supplemental Experimental Procedures, Figures S1--S5, and Tables S1, S2, and S4--S7Table S3. WES of iPSCs from Case No. 2, Related to Figure 1Document S2. Article plus Supplemental Information
We thank Drs. Sadakazu Miyashita, Koji Sagane, and Toru Arai for useful advice and comments, and Drs. Haruna Takagi and Kentaro Takahashi for technical assistance. Patient care is provided by Dr. Hiroshi Moritake. This study was partially supported by grants from the Program for Intractable Disease Research Utilizing Disease-specific iPS Cells funded by the Japan Science and Technology Agency (JST)/Japan Agency for Medical Research and Development (A-MED) under grant numbers JP16bm0609006/JP17bm0804004/JP17ek0109233.
Supplemental Information includes Supplemental Experimental Procedures, five figures, and seven tables and can be found with this article online at [https://doi.org/10.1016/j.stemcr.2018.06.008](10.1016/j.stemcr.2018.06.008){#intref0025}.
[^1]: Co-first author
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Spinal cord injury (SCI) is among the issues that affect individual lives and almost can be said all the individuals in society are at equal risk. In 2004, 11000 cases have been diagnosed in United States of America with this condition. SCI is one of the most serious clinical diseases and its prevalence is increasing year by year, although mortality rate from SCI has a decrement about 5%, but disability rate from SCI is almost remained at high level. Among these disabilities, \"paraplegia\" can be mentioned as one of the most important complications. Therefore expedition of locomotor recovery after spinal injury has been a subject of interest to researchers and professionals in neuroscience ([@R1]). The extension of spinal injury can be different in individuals depends on type and area of injury ([@R2]). The most common form of spinal cord injury is contusion that may damages spinal cord because of the physical impact and secondary cell damage by creating glial scarring around the injured spinal cord ([@R3]). Recovery after spinal cord injury will be difficult due to axonal damage ([@R4]), demyelination and scar ([@R5]). Spinal cord injury has two stages: Primary mechanical injuries and secondary injuries that exist through inflammatory responses. Neuropathology demonstrations of SCI contains: Edema, axonal damage, infiltration of inflammatory cells and increased astroglia ([@R6]). Improvement of locomotor recovery after spinal cord injury depends on intensity of tissue damage. Spontaneous recovery after spinal cord injuries is very low ([@R7]). In spinal Restoration process and Further Restoration, macrophages and astrocytes play significant roles through signals activation ([@R8]). The efforts which may be done to improve the performance of injured spinal cord include: reduction of secondary injury development, intervention in neuro-inhibitory environment in injured area, replacing lost tissue cells through cells implantation, renewal of axon myelin loss, increasing potential recovery of local progenitor cells ([@R9]). Different types of cells have been used by researchers to improve locomotor recovery in spinal cord injury ([@R1], [@R10], [@R11]). One of these cells, stem cells derived from human adipose tissue (hADSCs), due to the production of neurotrophic factors such as nerve growth factor, brain-derived neurotrophic factor and glial-derived neurotrophic factor, can be efficient for restoration of function because of supplying appropriate neurological environment at the site of spinal cord injury. In this regard, many studies have shown that stem cells derived from human adipose tissue have the ability to repair traumatic nerve damage ([@R12]). Cells that are extracted from the surface layer of abdominal fat have greater ability to create such an environment rather than cells that are extracted from deep layer ([@R12]). One of the reasons that inhibit axonal regeneration after CNS injury is existence of impermeable cement scar in the site of injury ([@R13], [@R14]). Several family of inhibitory molecules in the extracellular matrix along with reactive astrocytes create dense scar in site of injury that acts as a barrier to axonal regeneration ([@R15]). Chondroitin sulfate proteoglycan (CSPG) produced by astrocytes and oligodendrocytes in site of injury is increased drastically, which results in limiting axonal regeneration ([@R16]). Researchers have shown that CSPG is most abundant extracellular matrix molecules at the site of spinal cord injury and is associated with a Scar ([@R17]). Other therapeutic strategies in the treatment of spinal cord injury that can be used are scar elimination. Chondroitinase ABC has the ability to eliminate these scars and deal with Neuro-inhibitory environment that has been created at the site of spinal cord injury. ChondroitinaseABC (ChABC) is a bacterial enzyme that digests glycosaminoglycan chains (GAG) in CSPG, which is the main component of the extracellular matrix. One of the effects of ChABC which have focused recently by researchers is that the axonal regeneration and functional improvement after lesions of the central nervous system can be done through digestion of GAC and CSPG chains by ChABC ([@R18]-[@R20]).
According to the above description, the aim of present study was to answer to the question: which one is more efficient for Improvement of locomotor recovery after a spinal cord contusion model? Transplantation of hADSCs or injection of ChABC.
Materials and Methods
=====================
Animals
-------
In this study we used adult male wistar rats (N=24) weighting between 250-350g (Pasteur Institute, Tehran). The study approved by medical ethics committee of Iran University of Medical Sciences. Animals kept in animal house standard conditions (temperature 21±3°C, 12 hr light/dark cycle) with free access to food and water.
Isolation of hADSCs
-------------------
After attainment of patients written consents, fatty tissue was prepared from superficial layer of abdomen during liposuction surgery from 25-46 years individuals in Rasul Akram hospital (Iran-Tehran). Isolation of human adipose-derived stem cells was performed according to Dubois *et al* protocol ([@R21]). Fatty tissue was warmed in 37°C water bath before the initiation of Isolation. Then all the Isolation stages were performed under hood sterilized condition. 200 mg of fatty tissue for washing purpose was transferred to the tube containing 1% Penicillin/Streptomycin (Invitrogen) dissolved with warm phosphate-buffered saline (PBS, Invitrogen) and washing was continued until elimination of blood vessels, and connective tissue (commonly 2 times washing). Fatty tissue sample was minced by sterilized scissors and was transferred to the tube containing collagenase type I (Gibco,17100-017, USA) 0.1% and BSA 1% (dissolved with warm PBS)(Invitrogen) for digestion, then kept in water bath for 30 min for total digestion and homogenization of sample. After tissue digestion, the tube containing the sample was centrifuged at room temperature for 5 min at 1200 rpm speed. After discharging supernatant, formed plate was resuspended with BSA 1% solution and was again centrifuged to remove red blood cells using RBC lysis buffer. Ultimately after centrifugation and discharging supernatant, formed plate was resuspended with medium containing DMEM/Ham\'s F-12, FBS 10% and Penicillin/Streptomycin 1% and transferred to the tissue culture flasks. Flasks were maintained in incubator (temperature 37°C, CO~2~ 5%, humidity 98%).
Flowcytometry analysis
----------------------
In order to characterization of hADSCs, isolated cells were fixed in 5^th^ passages (after being harvested by trypsin) in paraformaldehyde 2% for 30 min. After two times washing with PBS, cells were incubated with antibodies against CD90, CD73, CD45, CD44, and CD31 for 30 min. CD44 and CD90 antibodies were directly conjugated with the allophycocyanin (APC). The Rat IgG2b was used for control isotope of CD90 and CD44 as substitute antibody. Goat anti-Rabbit IgG-FITC was used as a secondary antibody for CD31 and CD45 and Goat anti-mouse IgG-FITC was used as a secondary antibody for CD73. Rabbit polyclonal IgG was used as a substitute antibody for control isotope of CD31, CD45 and CD73. Flowcytometry was performed with a BD FACScalibur flow cytometer device (BD Biosciences, USA). Details of used antibodies are summarized in Table [1](#T1){ref-type="table"}.
Tagging Human Adipose- derived Stem cells (hADSCs) with GFP+ Recombinant lentiviral virus
-----------------------------------------------------------------------------------------
Lentivral vector carry Copa-GFP gens under EF1 promoter produce under calcium phosphate standard protocol. lentiviral vector pCDH-311B with EF1-CopGFP (System Bio Inc.) , pMD.2 and p.sPAx.2 (Kindly gift from Dr Trono) was used for transfection HEK293T in 10cm plate with CoPo4 reagents. After 18 h we change medium with fresh DMEM -10% FBS. Recombinant viral collected in 24, 48 and 72 hr after change the medium and any time add 12 ml fresh medium to plate. Collected recombinant viral titer measured with transduction HEK 293T in 6 well plate in different log. Titer was about 1×10^6^-3×10^6^vp/ml. hADSCs transduction achieved with application of polyberen and spinfection enhanced transduction protocol with MOI 4-6. For maximum transduction spinfection repeated with fresh recombinant viruses for 3 times. hADSCs transduction assay with florescent microscope after 72 hr (Figure [2](#F2){ref-type="fig"}).
Spinal cord injury model
------------------------
Animals were anesthetized with IP injection of ketamine (80 mg/kg) and xylazine (10 mg/kg). Animals were placed in the prone position on the covered operating table with warm blankets. After shaving the skin in the thoracic spines area and prepping with Betadine, midline incision was created with a scalpel. After Pushing the subcutaneous fat and muscle to expose the vertebral lamina, laminectomy was performed at T8-T9 levels of spinal cord. Metal cylinder (weighing 10 g and 2 mm in diameter) was released on the exposed spinal cord from distance of 12.5 cm. Then the muscles and skin were sutured with 3/0 Suture. Postoperative care included: Ringer\'s solution administered to prevent dehydration (3 ml IP after surgery), administered gentamicin (0.8 mg/100 g, IP) for 4 days postoperatively and bladder massage twice a day, for all animals was performed.
Contusion model confirmation
----------------------------
To confirm contusion model, seven days after spinal cord injury one animal were selected to evaluation of cavity formation in lesion site. In order to this, mentioned animal were deeply anesthetized with ketamine and xylazine and transcardially perfused with 4% paraformaldehyde in 0.1 mol/l PBS (pH=7.4). A piece of spinal cord (length 1.5 cm), containing contusion location, was maintained in sucrose 30% overnight and was embedded in cryopreservation medium (OCT). Cross-sectional thickness of the 10 µm was made by cryostat from injured area. To observe formed cavity in the lesion site, slides were stained with Nissl staining.
Transplantation procedure
-------------------------
Animals were randomly divided into 4 groups include:Sham operative group (n=6): laminectomy were performed in this groupControl group (n=6): SCI were achieved in this grouphADSCs group (n=6): 1×10^6^ hADSCs in the volume of 10µl were injected through intraspinal injectionChondroitinase ABC group (n=6):10 µl of 100 U/ml Chondroitinase ABC (Sigma, C3667)were injected \[diluted with 0.01% bovine serum albumin (dissolved with PBS)\].
Seven days after SCI, all animals were anesthetized and contusion site were exposed.1×10^6^hADSCs were resuspended with 10µl PBS and aspirated by Hamilton syringe. Cell transplantation and chondroitinase ABC injection were performed by Hamilton syringe with a sterile 30 gauge needle. After connecting the syringe to the micro-injector device (model 780310), needle in the midline were inserted 1-1.5 cm deep in spinal cord and during 2 min 5 µl were injected 1mm rostrally and then 1 mm caudally from the site of injury. For prevention of cells and enzyme leakage from injection site, after 2 min needle was withdrawn from spinal cord at the end of injection.
After each transplantation session, 1 sample of hADSCs from the Hamilton syringe was mounted onto a slide and stained with trypan blue to assess cell viability. Approximately 90% of cells were alive.
Behavioral assessment
---------------------
Locomotor functions were assessed by the open-field walking test during 4 min. Each animal moved freely in a circular field (90 cm in diameter and height of 24 cm) during 4 min. The Animals movement was captured by digital camera during these 4 min. Basso, Beattie, and Bresnahan (BBB) test was performed and recorded for each animal weekly for 8 weeks. Two trained observers who were unaware of the treatment group were saw the films separately and were scored the animals movement on the basis of BBB scale. 21- point open field locomotion score was developed by BBB in order to study the sequence of locomotor recovery patterns and takes into consideration the early (BBB score from 0 to 7), intermediate ([@R8]-[@R13]) and late phases ([@R14]-[@R21]) of recovery ([@R22]). Mean scores of observers were recorded as a BBB score of each animal. BBB scores were recorded one week after SCI. Only the animals that had BBB scores under 2 were entered to study or else they were omitted.
Histology
---------
After the 8^th^ weeks animals were deeply anesthetized with ketamine and xylazine and were perfused transcardially with paraformaldehyde 4%. Segment of Spinal cords which contains lesion site (length 1.5 cm) removed and was maintained in sucrose 30% overnight and was embedded. Serial cross-sections (10 µm in thickness) were produced from samples with cryostat. These sections were stained with cresyl violet for the study of their general histology.
Confirmation of presence hADSCs in tissue
-----------------------------------------
Mentioned sections were observed with flurocense microscope to confirm presence of hADSCs in tissue.
Measurement of the cavity size
------------------------------
For measurement of the cavity size, rats at postoperative 8 weeks were used. Serial cross-sections were stained with cresyl violet (in each animal 10 sections at an interval of 50 µm), and examined under a light microscope equipped with a camera. The area of cavity in the spinal cord was measured with an image processing and analysis program \"Olysia Bio Report Soft Imaging System 3.2\" on consecutive sections.
Statistical analysis
--------------------
The statistical comparisons between groups were carried out using repeated measures analysis of variance (ANOVA) followed with the tukey test for *Post hoc* analysis. Statistical analysis was performed using SPSS version 15. A *P*\<0.05 was accepted to denote statistically significance and all data were presented as mean ± SEM.
Results
=======
Characterization of hADSCs Cell culture
---------------------------------------
Human adipose derived stem cells were cultured in DMEM/F12 with 10% FBS, and the majority of the cells remained in suspension, including cell population. Erythrocytes had been removed during tissue digestion. In order to remove suspended cells, the culture was washed three times with PBS, after which a few attached single cells or cell clumps were observed. DMEM/F12 with 10% FBS added to flasks. Attached cells proliferated and reached approximately 90%confluence in 25-cm^2^ flasks, the primary culture was trypsinized using 0.25% trypsin-EDTA (Invitrogen) and passaged at a culture expansion ratio of 1:4 until passage 5. Spindle-shaped or fibroblast-like morphology of proliferated cells were observed using inverted microscope (Figure [1A](#F1){ref-type="fig"}).
Flow cytometry analysis
-----------------------
hADSCs displayed positive staining for the specific mesenchymal surface markers CD44, CD73 and CD90 (Figuer [2B](#F2){ref-type="fig"}). hADSCs at passage 4 exhibited high levels of CD44, CD73 and CD90 which expressed from 92.35%, 82.65% and 97.67% of the total cell population, respectively. In contrast, only a small proportion of hADSCs expressed hematopoietic stem cell surface markers, including CD31 and CD45 which were expressed at 0.55% and 1.32% of cells, respectively (Figure [1B](#F1){ref-type="fig"}).
Contusion model confirmation
----------------------------
Seven days after the contusion injury, evaluation of sections stained with cresyl violet revealed the formation of several differently sized vacuoles and cystic cavities at the site of injury.
Locomotor function
------------------
At the 7^th^ days after SCI, the contusion site was injected with hADSCs and ChABC. For assessment of locomotor function after injection of hADSCs and ChABC, we used the BBB scale. At 8^th^ weeks after injection (63 days after surgery), injection of hADSCs and ChABC promote locomotor function (*P*\<0.01) such that hADSCs (10.83) and ChABC (10.56) groups show significant increase in BBB Score compared with the control group (5.70) (*P*\<0.01) and this trend indicate that hADSCs and ChABC is necessary for locomotor function recovery. At1^st^week after injection (14 days after surgery) the ChABC group (8.12) show significant increase in BBB Score compared with the control group (1.50) (*P*\<0.001) and this increment was maintained until 8^th^ weeks. In hADSCs group until 3^th^ weeks after injection (28 days after surgery) BBB Score was not shown significant difference compared with control group but at 3^th^ weeks after injection, hADSCs group (8.67) shown significant increase in BBB Score compared with the control group (3.50) (*P*\<0.01) and BBB Score was increased gradually until 8^th^ weeks after injection (63 days after surgery) (Figure [3](#F3){ref-type="fig"}).
Histology and Confirmation presence of hADSCs in tissue
-------------------------------------------------------
For assessment of changes in cavity volume, the host spinal cord tissue was stained with cresyl violet. Non-treated animals with SCI showed the formation of large cavities (Figure [4B](#F4){ref-type="fig"}). The spinal cords of treated animals had cavities much smaller than those of non- treated animals (Figure [4C, D](#F4){ref-type="fig"}). These results showed that hADSCs transplant and Chondroitinase ABC injection reduced the formation of cavities after contusion model of spinal cord injury. Fluorescence microscopy (Olympus AX 70) reveals that GFP-positive hADSCs, transplanted at the site of injury, survived and reorganized around the cavity center (Figure [5](#F5){ref-type="fig"}).
Cavity size
-----------
The spinal cords of hADSCs and ChABC-injected group had cavities much smaller than those of the control group (Figures [4C, D](#F4){ref-type="fig"}). In the hADSCs-injected rats showed a cavity size of 2618461.44 µm^2^ and ChABC-injected rats showed a cavity size of 3338041.09 µm^2^ on average, whereas the control rats showed a value of 6417159.48 µm^2^ on average. These values of cavity volume were significantly different between the hADSCs, ChABC-injected groups and control rats (*P*\<0.001)(Figure [6](#F6){ref-type="fig"}).
Discussion
==========
The most of the human spinal cord injury during the accident, falling down and Sports Injuries occurs due to contusion. Damage to the vertebrae column and eventually inserting bone or inter vertebrae discs into the spinal canal space, is the main cause of the SCI ([@R23]). Therefore, in this study due to the common natural models of the SCI, contusion model is used. In most cases progressive necrosis of spinal cord tissue is caused formation of cavity ([@R24]-[@R26]). Also, in this study cavity was formed at SCI site (Figure [3](#F3){ref-type="fig"}). Flux of inflammatory cells causes a fluid-filled cavity or cyst that this process is known as secondary injury. Secondary injury caused the activation and proliferation of astrocytes that is able to cover holes or cysts with glial scar. The glial scar as an inhibitory factor, degeneration of myelin, prevents proper nerve conduction to correct path ([@R23]). Axonal regeneration with reduction of cavity size as a treatment strategy can improve motor function in SCI ([@R27]-[@R29]).
In this study, the groups treated with hADSCs and ChABC, the Cavity volume was significantly decreased compared to controls. Also in treatment groups, BBB score shows the significant increment rather to control group that Cavity volume in treatment groups is consistent with increased BBB score. This study shows that implantation of hADSCs leads to improvement of locomotor function by reduction in volume of cavity.
Presence of hADSCs in the injury site at least in 8 weeks, causes in volume cavity reduction ([@R30], [@R31]). Fluorescence microscopy confirms the presence of cells in the end of 8^th^ weeks. Other studies are also reported cavity volume loss after SCI with transplantation of cells such as bone marrow stromal cells (BMSCs) ([@R24], [@R32], [@R33]) and neural progenitor cells (NPCs) ([@R34], [@R35]). Like to BMSCs ([@R24], [@R32], [@R33]) and NPCs ([@R34], [@R35]), hADSCs ([@R36]) cannot differentiate totally into neurons after transplanting into the injured spinal cord and remain undifferentiated. According to these explanations, BBB score increase and improve motor function in hADSCs maybe not due to neuronal differentiation of hADSCs but may be due to the production of useful growth factors for neural tissue. Cytokine secretion by hADSCs, such as interleukins, stem cell factor ([@R36], [@R37]), NGF and BDNF ([@R38]) have been reported. Glial cell proliferation and in addition axonal growth at the site of SCI can reduce Cavity volume. Cavity volume reduction maybe not related to a specific factor, but the result is the accumulation of several factors ([@R39]).
One of the therapeutic strategies for SCI is dealing with chondroitin sulfate proteoglycans in order to prevent the formation of glial scar because this inhibitory environment is one of the barriers for axonal growth and recovery of motor function. Attenuation the role of chondroitin sulfate proteoglycans leads to wide range of axonal regeneration cascade and consequently restoring nerve regeneration in injured site. ChABC degrades glial scar ([@R40]), particularly chondroitin sulfate proteoglycans, reduces expression of glial fibrillary acidic protein ([@R41]) in injured site and in this way reduces inappropriate microenvironment in injured site. Field studies showed that this enzyme, ChABC, significantly increases the expression of growth-associated protein-43 and significantly reduces the extent of necrotic area containing Cavity ([@R41]) and improves motor function in SCI ([@R18]). Therefore in this study ChABC was used to counters the Neuro-inhibitory environment. The results of present study also confirm findings of previous studies in such a way that significant cavity volume reduction and motor function improvement was observed in ChABC group. One week after SCI, astrocytes and glial scarring at the site of a lesion creates a Neuro-inhibitory environment where become one of the major challenges for therapeutic interventions in SCI. Currently, the most common intervention in the treatment of SCI is cell therapy in such a way that inappropriate microenvironment does not take into consideration adequately. The aim of our studyis comparing of motor function improvement in adopting of two treatment strategies for SCI, containing counter with inappropriate microenvironment and cell therapy in lesion site.
Conclusion
==========
Our results shows dealing with Neuro-inhibitory environment by ChABC causes reduction in cavity volume and motor function improvement is equal to cell therapy with hADSCs. Innovation of this study regarding its results is that dealing with inappropriate Neuro-inhibitory environment and glial scar by ChABC have equal role compare to cell therapy for improving motor function after SCI and this result in adoption of proper therapeutic strategies for SCI intervention is important.
It is suggested that, Differentiation of transplanted cells into mature neurons also synapse of them with endogenous neurons to be investigated, Locomotor Function tests to be performed at 16 weeks and Scar changes at the cellular and molecular level tobe investigated.
The results described in this paper were part of student thesis. The present study was supported by a grant from Iran University of Medical Sciences and was performed in department of anatomy. The cell culture stage was performed at Cellular and Molecular Research Center of Iran University of Medical Sciences. The Sectioning and staining of tissue sections part of this research was performed at Department of Medical basic Sciences at Iran University of Medical Sciences so we express our deep thanks to Dr Behnam Jameie for her help.
{#F1}
{#F2}
{#F3}
{#F4}
{#F5}
{#F6}
######
Details of used antibodies in flowcytometry
Dilution Cat. number Company Antibody
---------- ------------- ---------------- --------------------------
1/100 559869 BD Biosciences CD90
1/200 17-0441 eBioscience CD44
1/50 ab81720 Abcam CD73
1/100 ab10558 Abcam CD45
1/20 ab28364 Abcam CD31
1/200 17-4031 eBioscience Rat IgG2b
1/100 Ab27478 Abcam Rabbit polyclonal IgG
1/100 AP156F Millipore Goat anti-Rabbit IgG
1/40 Sc-2010 Santa cruz Goat anti-mouse IgG-FITC
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1}
===============
The prevalence of coronary artery disease (CAD) is growing rapidly nowadays, which makes it the leading cause of death in many developed and developing countries. CAD is responsible for 17.7 million deaths, which account for 31% of all global deaths every year. In China, the overall number of estimated prevalent cases of CAD was 93.8 million according to recent data, which resulted in about 4 million deaths annually \[[@B1]\]. Traditionally, CAD is strongly associated with age, obesity, diabetes, and hyperlipidemia. Even sufficient attention was devoted to those traditional risk factors, prevalence rate for CAD still increased remarkably by 14.7% from 1990 to 2016 \[[@B2]--[@B4]\]. Particularly, some individuals with severe artery stenosis (≥90% in coronary arteries) do not exhibit obvious clinical symptom, which may hinder the diagnosis and prevention of CAD. Given this unfulfilled need for deeper understanding and effective preventive for CAD, uncovering new risk factors and biomarkers for CAD is of great importance.
Recently, multiple studies have suggested strong association of gut microbiota with CAD pathogenesis and progression \[[@B4], [@B5]\]. The gut microbiota is influenced by dietary intake and in turn produce metabolites that contribute to affecting the metabolism and immunity of the host. Remarkably, trimethylamine N-oxide (TMAO), a gut microbial metabolite, was indicated to be a proatherogenic factor. Study conducted by Hazen firstly reported elevated TMAO in both atherosclerosis patients and mice model in 2011 \[[@B5]\]. Afterwards, several studies indicated that elevated TMAO level could predict an increased risk of major adverse cardiovascular events \[[@B6]--[@B8]\]. TMAO arises from gut microbiota-mediated metabolism of dietary trimethylamine-containing compounds including choline, betaine, and L-carnitine. Those predecessor products of TMAO were also suggested to be associated with atherosclerosis and/or myocardial ischemia risks. Dietary supplement with choline could enhance atherosclerosis in mice \[[@B9]\]. Plasma level of L-carnitine, which is abundant in red meat, contributed to the prediction of increased risks of cardiovascular disease and major adverse cardiac events \[[@B10]\]. A study which involved 3924 African-American participants indicated higher risk of CAD incident with higher dietary betaine intake \[[@B11]\]. However, the relationship between plasma TMAO, choline, L-carnitine, and betaine levels and extent of arterial stenosis of cardiovascular in CAD patients has not been investigated, and whether the combination of TMAO and its predecessors could help to diagnose CAD and predict the risk of severe stenosis in different gender is still unknown. In this study, we aimed to explore the relationship between plasma TMAO, choline, L-carnitine, and betaine concentrations with CAD incidence and severity of arterial stenosis in male and female CAD patients.
2. Materials and Methods {#sec2}
========================
2.1. Study Population {#sec2.1}
---------------------
We recruited 73 healthy controls (CON group) and 94 CAD patients (CAD group) who were hospitalized for coronary angiography. At least one main coronary artery with luminal stenosis diameter ≥ 50% (determined by quantitative coronary angiogram analysis) was diagnosed as CAD (18 ≤ age ≤ 75 years). Age- and sex-matched CON group was an independently recruited set who underwent coronary artery CT or coronary angiography for chest pain but showed negative results. The CAD group was further divided into severe artery stenosis group (S, *n* = 45) and mild artery stenosis group (M, *n* = 49). Severe artery stenosis was defined as ≥90% stenosis in at least one of the main coronary arteries.
We exclude subjects with an active infection, malignancy, severe liver or heart cerebrovascular diseases, and severe proteinuria (\>3.5 g/day) and those who received probiotics or antibiotic treatment within 1 month of enrollment to minimize potential confusing factors.
Study subjects underwent complete clinical history and physical examination in the study duration. General information including age, sex, weight, and body mass index (BMI) was retrospectively collected from each subject\'s medical records. All subjects were informed before being enrolled in the study.
2.2. Laboratory Test {#sec2.2}
--------------------
Fasting blood samples (5 mL) were collected using EDTA-K treated tubes. Blood samples were centrifuged at 1000 ×*g* for 10 min and stored at −80°C until analysis. Plasma TMAO, choline, L-carnitine, and betaine were measured by high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS) using D9-TMAO, D9-choline, D3-L-carnitine, and D9-betaine internal standards as described previously \[[@B12]\]. Hemoglobin (HGB), fasting blood glucose (Glu), creatinine (Cr), triglyceride (TG), total cholesterol (TCHO), alanine aminotransferase (ALT), glutamic-pyruvic transaminase (AST), low-density lipoprotein (LDL), and high-density lipoprotein cholesterol (HDL-C) were measured using HITACHI7170S automatic biochemical analyzer. Blood pressure (systolic blood pressure (SBP) and diastolic blood pressure (DBP)) and heart rate (HR) were measured after rest for at least 15 min.
2.3. Statistical Analysis {#sec2.3}
-------------------------
Statistical presentation and analysis of the current study were performed using the computer SPSS program (Statistical Package for the Social Science, Chicago). Categorical variables are presented as numbers and percentages. Continuous data are presented as means ± standard deviations (SD) for normal distribution parameters or medians (interquartile ranges, IQR) for nonnormal distribution parameters to investigate the dynamic change of TMAO and its predecessors in study cohort. Student\'s *t-*test or the Mann--Whitney *U* test was used for differences evaluation between two groups. Logistic regression analysis was performed to examine the odds ratio (OR) and 95% confidence interval (95% CI) of TMAO and its predecessor products for CAD and severe artery stenosis in male, female, and all participants; adjustments were made for variables including age, gender, BMI, SBP, DBP, Glu, TG, and Cr. Because the distribution of TMAO, choline, L-carnitine, and betaine was skewed, they were log-transformed in logistic analysis. The area under the receiver-operating characteristic curve (AUC) was calculated to evaluate the value of betaine, choline, L-carnitine, and TMAO in predicting the risk of CAD and severe artery stenosis. A two-tailed *p*-value \< 0.05 was considered statistically significant.
3. Results {#sec3}
==========
3.1. Patients Characteristics {#sec3.1}
-----------------------------
Of the 94 CAD patients and 73 healthy controls, the baseline patient characteristics are displayed in [Table 1](#tab1){ref-type="table"} as categorized by CAD and CON group. 42.47% of CON and 42.55% of CAD group were male. A comparison of the baseline characteristics between groups with or without CAD showed that CAD patients tended to exhibit higher blood pressure (SBP = 138 ± 3; DBP = 81 ± 1 in CAD versus SBP = 128 ± 3; DBP = 75 ± 2 in CON, *p*=0.01) and to have higher Glu (5.46(4.77--6.64) mmol/L in CAD versus 4.96(4.57--5.73) mmol/L in CON, *p*=0.01) and TG (1.85(1.31--2.65) in CAD versus 1.31(0.97--1.85) mmol/L in CON, *p* \< 0.01) concentrations than CON subjects. However, other factors like age, gender, height, weight, BMI, HR, TCHO, HDL, LDL, HGB, Cr, ALT, and AST were similar between two groups (*p* \> 0.05, [Table 1](#tab1){ref-type="table"}).
3.2. Relationship between Plasma TMAO, Choline, L-Carnitine, and Betaine Levels and CAD {#sec3.2}
---------------------------------------------------------------------------------------
We performed a cross-sectional comparison of TMAO, choline, L-carnitine, and betaine concentrations between CAD patients and CON group at first. We observed remarkably higher TMAO concentration in CAD group than CON group (1.46(0.8--2.32) *μ*M versus 1.18(0.67--1.7) *μ*M, *p*=0.03) ([Table 1](#tab1){ref-type="table"} and [Figure 1](#fig1){ref-type="fig"}). No significant difference was found in choline, L-carnitine, and betaine among CAD and CON group. Then we divided the CAD group as severe artery stenosis group (S, *n* = 45) and mild artery stenosis group (M, *n* = 49) to see if the concentrations of those metabolites could further increase in severe artery lesion patients. As showed in [Figure 2](#fig2){ref-type="fig"}, only TMAO showed significantly elevated concentration in S group compared with M group (1.62(0.91--2.81) *μ*M versus 1.27(0.77--1.82) *μ*M, *p*=0.02).
Logistic regression analysis was employed to evaluate the risk of TMAO, choline, L-carnitine, and betaine in CAD and artery lesion. Results showed that TMAO was an independent prognostic risk factor for CAD even adjusted for traditional risk factors including age, gender, BMI, Glu, TG, and Cr in model 1 (OR = 1.81, 95% CI: 1.07--3.09, *p*=0.03). We performed a second model as model 2 which includes choline, L-carnitine, and betaine and all factors in model 1. TMAO remained to be an independent risk factor (OR = 1.12, 95% CI: 1.12--3.57, *p*=0.02) while choline, L-carnitine, and betaine exhibited a very low odds ratio with no statistical significance in CAD (*p* \> 0.05) ([Table 2](#tab2){ref-type="table"}). More than that, we analyzed the risk of TMAO and its predecessors in different gender with CAD. As showed in [Table 2](#tab2){ref-type="table"}, TMAO was independently and significantly associated with the presence of CAD after being adjusted for other traditional risk factors in male subjects (model 3, OR = 2.64, 95% CI: 1.04--6.69, *p*=0.04). When further adjusted for its predecessors, the risk of TMAO remained robust in CAD with an odds ratio of 2.58 (*p*=0.05). No significant association of TMAO with CAD incidence in female subjects was observed (model 5, OR = 1.33, 95% CI: 0.71--2.49, *p*=0.37; model 6, OR = 1.66, 95% CI: 0.76--3.65, *p*=0.2). None of choline, L-carnitine, and betaine showed statistical significance in the relationship of CAD in male or female group (*p* \> 0.05) ([Table 2](#tab2){ref-type="table"}).
Similar results were found in the association of TMAO and its predecessors with extent of artery stenosis. Significant predictive value of plasma TMAO was preserved after adjustment for traditional risk factors in total study cohort (model 7, OR = 1.36, 95% CI: 1.01--1.84, *p*=0.04; model 8, OR = 1.37, 95% CI: 1.01--1.86, *p*=0.05) ([Table 3](#tab3){ref-type="table"}). The odds ratios of TMAO in severe artery stenosis were 1.38 versus 1.33 in male subjects and 1.25 versus 1.14 in female subjects after adjustment with traditional factors and its predecessors, but there is no statistical significance (*p* \> 0.05). The association of choline, L-carnitine, and betaine with the presence of artery stenosis in total, male, and female groups was not statistically significant (*p* \> 0.05).
3.3. Diagnostic Value for TMAO and Its Predecessors in Determining CAD and Artery Stenosis {#sec3.3}
------------------------------------------------------------------------------------------
To explore the role of TMAO and its predecessors in distinguishing CAD from healthy controls and severe artery stenosis from mild artery stenosis, we performed receiver-operating characteristic curve (ROC) analysis. As showed in Figures [3](#fig3){ref-type="fig"} and [4](#fig4){ref-type="fig"}, the AUCs of TMAO in discriminating CAD and severe artery stenosis in total study cohort were 0.61 (*p*=0.03) and 0.62 (*p*=0.02). Combination of TMAO and predecessors showed slightly improved diagnostic accuracy with AUCs of 0.64 (*p* \< 0.01) and 0.66 (*p* \< 0.01) in discriminating CAD and severe artery stenosis. In male group, combination of TMAO and its predecessors or TMAO alone exhibited almost identical AUC for diagnosing CAD and artery lesion (0.64 versus 0.64, *p*=0.04 for CAD; 0.64 versus 0.64, *p*=0.05 for artery stenosis), while in female group, combination of TMAO and its predecessors exhibited better accuracy and higher sensitivity than TMAO alone at diagnosing CAD and artery stenosis extent (AUC of combined metabolites = 0.64, *p*=0.02 versus AUC of TMAO = 0.57, *p*=0.21 in CAD; AUC of combined metabolites = 0.68, *p*=0.01 versus AUC of TMAO = 0.58, *p*=0.26 in severe artery stenosis) (Tables [4](#tab4){ref-type="table"} and [5](#tab5){ref-type="table"}).
4. Discussion {#sec4}
=============
The main finding of the current study is (1) strong association between fasting plasma TMAO levels in patients with CAD and severe artery lesion; (2) elevated fasting plasma TMAO level which could serve as an independent predictor of CAD and the severe artery stenosis; (3) TMAO which is more accurate in diagnosing CAD and severe artery stenosis in men compared with women, while a combination of TMAO, choline, L-carnitine, and betaine showed improved accuracy in determining CAD and artery stenosis risk in both men and women.
Metabolites produced by gut microbiota could directly regulate the pathogenesis and progression of disease, among which TMAO has aroused extensive attention as it has been previously reported to directly promote proatherosclerotic effects and mechanistically involved in atherosclerosis. A case-control study indicated that TMAO was correlated with the SYNTAX score in patients with CAD \[[@B13]\]. Our study extends these observations by showing that plasma TMAO concentration, rather than its predecessors, was significantly associated with high atherosclerotic burden of artery ([Figure 2](#fig2){ref-type="fig"}). More than that, the logistic analysis results suggested TMAO as an independent risk factor in CAD and severe artery stenosis with odds ratios of 1.81 and 1.36 (*p* \< 0.05) after adjustment for traditional factors, and the OR of TMAO slightly increased after adjustment for choline, L-carnitine, and betaine in CAD (OR = 2, 95% CI: 1.12--3.57, *p*=0.02) and in severe artery stenosis (OR = 1.37, 95% CI: 1.01--1.86, *p*=0.05) (Tables [2](#tab2){ref-type="table"} and [3](#tab3){ref-type="table"}).
Main dietary derived predecessors of TMAO such as choline, L-carnitine, and betaine were also indicated to be involved in pathogenesis of CAD \[[@B14]\]. Ueland and his colleagues\' study in middle and elderly men and women suggested that high plasma choline level was positively associated with cardiovascular risk factors \[[@B15]\]. L-Carnitine, a chemical analog of choline, was also reported to be elevated in people with CAD \[[@B16]\]. Koeth investigated plasma L-carnitine levels in patients undergoing cardiac evaluation and revealed risk-predicting ability of L-carnitine in both prevalent cardiovascular disease and incident major adverse cardiac events \[[@B10]\]. Wang\'s study, based on 1879 stable CAD participants, indicated that plasma betaine levels were significantly higher in CVD cases than in controls and there were dose-dependent associations with betaine concentration and the presence of CVD \[[@B5]\]. However, in our study, there are no obvious differences of choline, L-carnitine, or betaine concentration between CAD and CON groups as showed in [Table 1](#tab1){ref-type="table"} and [Figure 1](#fig1){ref-type="fig"}. Subsequently, multivariate analysis indicated that choline, L-carnitine, and betaine were not risk factors related to CAD incidence (choline: OR = 0.89, L-carnitine: OR = 0.98, and betaine: OR = 1.02, *p* \> 0.05) or severity of artery lesion (choline: OR = 1.07, L-carnitine: OR = 1.01, and betaine: OR = 0.98, *p* \> 0.05) (Tables [2](#tab2){ref-type="table"} and [3](#tab3){ref-type="table"}).
Notably, the median concentrations of TMAO in CAD patients in the present study were lower than previously reported (1.46 *μ*M versus 2.2 *μ*M \[[@B17]\]). The median concentrations of choline and betaine in CAD patients enrolled in Ueland and Wang\'s study were higher than patients enrolled in our study (9.9 *μ*M versus 8.37 *μ*M for choline; 41.2 *μ*M versus 32.64 *μ*M for betaine), while the median concentration of the L-carnitine in patients reported in Koeth\'s study was relatively lower than patients in our study (38 *μ*M versus 43.16 *μ*M). These differences may be caused by multiple characteristics between populations such as age structure, criteria of CAD patients, composition of gut microbiota, the Eastern and Western dietary habits, and polymorphism of key metabolic enzyme. Further studies which involve all those factors are needed for more information.
There are significant differences existing between men and women in terms of disease onset, prevalence, and clinical symptoms resulting in the differences of many aspects between men and women like physiology and psychology \[[@B18]\]. However, most of treatment strategies were based on studies conducted on male patients. Many clinical symptoms of female patients with CAD are often atypical, leading in about 65% of women with CAD to miss the diagnosis \[[@B19]\]. Therefore, identifying sensitive biomarkers for different gender to reduce the incidence of CAD is of great scientific and clinical importance. The current study, to our knowledge, revealed a gender-related difference in risk-predicting accuracy of TMAO and its three predecessors in CAD and artery stenosis for the first time. Men might possess several protective physiological mechanisms with respect to TMAO production. There were studies suggested that flavin-containing monooxygenase-3 (FMO3), which was most active in TMAO synthesis pathway, was downregulated by testosterone, and the abundance of FMO3 was relatively lower in male compared to female population \[[@B20]\]. Despite all these, Juergen revealed a significantly lower TMAO level in female subjects than male subjects (5.4 ± 5.6 *μ*M in women versus 7.3 ± 10.0 *μ*M in men) \[[@B21]\]. Our study showed similar results with higher TMAO in men compared with women with CAD (1.33(0.77--2.09) *μ*M versus 1.65(1.04--2.84) *μ*M) (Supplemental Table 1). More than that, significantly elevated TMAO was observed in male CAD patients compared with male controls, but there are no significant differences of TMAO between female CAD and CON groups (Supplemental Table 1).
Results of logistic analysis in our study showed that elevated TMAO concentration tended to have higher odds ratio for both CAD incidence and sever artery stenosis in male (OR = 2.64, 95% CI: 1.04--6.69, *p*=0.04 for CAD; OR = 1.38, 95% CI: 0.93--2.04, *p*=0.11 for severe artery stenosis) than in female subjects (OR = 1.33, 95% CI: 0.71--2.49, *p*=0.37 for CAD; OR = 1.25, 95% CI: 0.74--2.12, *p*=0.4 for severe artery stenosis). None of choline, L-carnitine, or betaine showed significant risk-predicting sensitivity in CAD or severe artery stenosis in both men and women analyzed as covariant either together (Tables [2](#tab2){ref-type="table"} and [3](#tab3){ref-type="table"}) or separately (data not shown). Interestingly, receiver-operating characteristic analysis showed that TMAO (AUC = 0.64, *p*=0.04 for CAD; AUC = 0.64, *p*=0.05 for severe artery stenosis) or combination of TMAO and its predecessors (AUC = 0.64, *p*=0.04 for CAD; AUC = 0.64, *p*=0.05 for severe artery stenosis) exhibited almost identical diagnostic accuracy and sensitivity in discriminating CAD from CON and severe artery stenosis from mild artery stenosis in male subjects. However, TMAO showed poor accuracy at discriminating CAD from CON or severe artery stenosis from mild artery stenosis in female subjects (AUC = 0.57, *p*=0.21 for CAD; AUC = 0.58, *p*=0.26 for severe artery stenosis). Nevertheless, combination of TMAO and its three predecessor products greatly improved the diagnostic accuracy and sensitivity in female subjects with an AUC of 0.64 for CAD (*p*=0.02) and 0.68 for severe artery stenosis (*p*=0.01). These differences between men and women might be related to dietary habits and microbiota composition. Several studies reported that men consume fewer fiber-rich foods and have higher red meat and fat intake compared to women \[[@B22]--[@B24]\], which might lead to the accumulation of trimethylamine-containing compounds and increased CVD risk in men. Evidence from microbiologists suggested that women may harbor a higher ratio of *Firmicutes*/*Bacteroidetes* in comparison to men \[[@B25], [@B26]\]. Gut bacteria in *Firmicutes* phylum are capable of producing short-chain fatty acids, which were demonstrated to contribute to CVD protection and might alleviate the risk of arteriosclerosis which arose from other gut flora metabolites like TMAO \[[@B27]\]. These data could provide us some hints that risk factors like TMAO alone may be less efficient in CAD risk prediction in women due to microbiota background, dietary habits, and metabolic or physiologic characteristics. When TMAO was employed for risk stratification in CAD populations, sex differences should be taken into account and assessed carefully to avoid bias in evaluations of CAD especially in women. Further investigations are needed to clarify these factors. Many cardiovascular events occur in individuals who were regarded as low-risk patients according to the currently used traditional cardiovascular risk factors due to low efficiency \[[@B28]\]. Jacqueline\'s work demonstrated improved cardiovascular risk stratification in women who were supposed at intermediate risk by adding combinations of noninvasive risk markers to traditional factors, while the risk-predicting improvements were limited in men \[[@B29]\]. Similarly, in our study, we did not see significant risk-predicting sensitivity of TMAO in women with respect to CAD incidence or artery lesion, but combination of TMAO, choline, L-carnitine, and betaine exhibited improved performance on CAD and artery lesion diagnosis in women. These results suggested that, regardless of the potent diagnostic role in men, TMAO itself might not be sensitive enough for CAD risk-predicting in women, but overall consideration of TMAO and its predecessors could be more appropriate for general population in CAD risk stratification.
So far, even accumulating data have indicated the link between TMAO and CVD, the causative role of TMAO in cardiovascular pathology is still under discussion. Contradictory observations also indicated that TMAO may not contribute significantly to the progression of early atherosclerotic disease risk \[[@B30]\]. Evidences supporting TMAO as a diagnostic marker for CAD were incoherent, and it remains unclear whether there exist factors other than diet and microbiota that may profoundly influence systemic TMAO levels. Therefore, we discussed the gender-related differences in associations of TMAO and its predecessors in CAD which might provide more information for clinical utility. However, multicenter and larger sample size will be needed to further verify these issues.
This study has several limitations. First, it is a cross-sectional study, so we cannot predict the future development of CAD. Second, the number of subjects in this study was small and selection bias cannot be excluded. And we did not include other potential confounding factors, such as patients\' nutritional status, recent diet, and gut microbiota profile, which might also influence the results. Trimethylamine (TMA) is the direct precursor of TMAO, which was also suggested to be involved in cardiovascular pathology by reducing cardiomyocytes viability according to recent data \[[@B31], [@B32]\]. It would be interesting to know the associations of TMA with the CAD with regard to gender. However, the present study was designed mainly focusing on TMAO which was more stable and less volatile for our HPLC-MS-based detection method other than TMA \[[@B33]\]. Although we have revealed the statistically significant associations of TMAO with CAD which differed by gender, future study which involves sensitive and stable detecting methodologies for metabolites including TMA as well as FMO activity and gut microbiota background is warranted for better understanding of the potential roles of those metabolites in CAD.
5. Conclusions {#sec5}
==============
In conclusion, the current study explored the correlation of TMAO and its three main predecessors, choline, L-carnitine, and betaine, in CAD and severe artery stenosis patients, and suggested a gender-related association of TMAO with CAD and artery stenosis. TMAO alone was powerful in risk stratification of CAD and artery stenosis in men; however, in women, no association of TMAO with risk of CAD as well as extent of artery stenosis was observed. More than that, we found that combination of four metabolites had better performance at disease diagnosis in both men and women compared with TMAO. These findings not only provided new thoughts for gender-related differences referring to onset and progression of cardiovascular disease but also suggested the potential clinical utility of the combination of TMAO, choline, L-carnitine, and betaine in artery burden stratification and CAD diagnosis.
This work was funded by grant from National Development of Key Novel Drugs for Special Projects of China (2017ZX09304014) and the Natural Science Foundation of Hunan Province (2019JJ50966) and also supported by the Hunan Key Laboratory for Bioanalysis of Complex Matrix Samples (2017TP1037).
Data Availability
=================
The data used to support the findings of this study are included within the tables of the paper.
Ethical Approval
================
All subjects were informed before enrollment in the study. The study plan was approved by the Ethical Committee of Xiangya Hospital of Central South University.
Conflicts of Interest
=====================
All authors have no conflicts of interest.
Authors\' Contributions
=======================
Dongsheng Ouyang, Zaixin Yu, and Fei Guo contributed to research idea, study design, and manuscript writing. Jun Zhou contributed to sample and data acquisition. Fei Guo and Zhenyu Li performed sample analysis. Fei Guo contributed to statistical analysis. Each author contributed to important intellectual content during manuscript drafting or revision and accepts accountability for the overall work by ensuring that questions pertaining to the accuracy or integrity of any portion of the work are appropriately investigated and resolved.
Supplementary Materials {#supplementary-material-1}
=======================
######
Supplemental Table 1: distribution of TMAO, choline, L-carnitine, and betaine in different gender with or without CAD.
######
Click here for additional data file.
{#fig1}
{#fig2}
{#fig3}
{#fig4}
######
Baseline characteristics of subjects stratified by CON and CAD.
CON (*n* = 73) CAD (*n* = 94) *p*
-------------------- ---------------------- ---------------------- --------
Male (*n*, %) 31 (42.47%) 40 (42.55%) 1
Age (years) 56 ± 1 57 ± 1 0.59
Height (cm) 160 ± 1 162 ± 1 0.12
Weight (kg) 63 ± 2 66 ± 1 0.29
BMI (kg/m^2^) 25 ± 1 25 ± 0 0.5
SBP (mmHg) 128 ± 3 138 ± 3 0.01
DBP (mmHg) 75 ± 2 81 ± 1 0.01
HR 73 (66--85) 71 (66--80) 0.31
Glu (mmol/L) 4.96 (4.57--5.73) 5.46 (4.77--6.64) 0.01
TCHO (mmol/L) 4.49 ± 0.13 4.61 ± 0.12 0.71
TG (mmol/L) 1.31 (0.97--1.85) 1.85 (1.31--2.65) \<0.01
HDL (mmol/L) 1.03 (0.88--1.25) 0.96 (0.83--1.18) 0.09
LDL (mmol/L) 2.78 ± 0.09 2.83 ± 0.09 0.81
HGB (g/L) 133.5 (122--141.75) 135 (126.5--147) 0.07
Cr (*μ*mol/L) 82.89 ± 2.30 86.54 ± 2.41 0.23
ALT (U/L) 19.1 (15.1--29.48) 21.1 (15.4--32.7) 0.3
AST (U/L) 22.93 (18.35--29.1) 20.8 (18.05--31.5) 0.92
TMAO (*μ*M) 1.18 (0.67--1.7) 1.46 (0.8--2.32) 0.03
Choline (*μ*M) 8.30 (6.90--10.14) 8.37 (6.76--9.56) 0.99
L-Carnitine (*μ*M) 43.08 (33.81--49.51) 43.16 (35.44--48.94) 0.96
Betaine (*μ*M) 37.84 (29.44--42.14) 32.64 (26.11--41.33) 0.08
Continuous data are presented as mean ± SD or median (interquartile range), and categorical variables are presented as counts and percentage (%).
######
Association of TMAO, choline, L-carnitine, and betaine levels with CAD in all, male, and female participants.
------------- --------- ------------ ------ ------ ------------ ------
All
Model 1 Model 2
OR 95% CI *p* OR 95% CI *p*
TMAO 1.81 1.07--3.09 0.03 2.00 1.12--3.57 0.02
Choline 0.89 0.7--1.13 0.33
L-Carnitine 0.98 0.94--1.03 0.47
Betaine 1.02 0.95--1.1 0.55
Male
Model 3 Model 4
OR 95% CI *p* OR 95% CI *p*
TMAO 2.64 1.04--6.69 0.04 2.58 0.99--6.74 0.05
Choline 1.00 0.68--1.48 0.99
L-Carnitine 1.00 0.97--1.04 0.81
Betaine 1.00 0.91--1.1 0.98
Female
Model 5 Model 6
OR 95% CI *p* OR 95% CI *p*
TMAO 1.33 0.71--2.49 0.37 1.66 0.76--3.65 0.20
Choline 0.85 0.64--1.13 0.27
L-Carnitine 1.01 0.96--1.06 0.71
Betaine 0.96 0.89--1.03 0.21
------------- --------- ------------ ------ ------ ------------ ------
OR, odds ratio. 95% CI, 95% confidence interval. OR shown were for plasma TMAO, choline, L-carnitine, and betaine in all, male, and female participants. Models 1, 3, and 5 adjusted for age, gender, BMI, Glu, TG, and Cr. Models 2, 4, and 6 adjusted for all factors in models 1, 3, and 5 plus choline, L-carnitine, and betaine.
######
Association of TMAO, choline, L-carnitine, and betaine levels with severe artery stenosis in all, male, and female participants.
------------- ---------- ------------ ------ ------ ------------ ------
All
Model 7 Model 8
OR 95% CI *p* OR 95% CI *p*
TMAO 1.36 1.01--1.84 0.04 1.37 1.01--1.86 0.05
Choline 1.07 0.89--1.28 0.48
L-Carnitine 1.01 0.99--1.03 0.23
Betaine 0.98 0.93--1.03 0.36
Male
Model 9 Model 10
OR 95% CI *p* OR 95% CI *p*
TMAO 1.38 0.93--2.04 0.11 1.33 0.92--1.93 0.13
Choline 1.08 0.79--1.48 0.64
L-Carnitine 1.02 0.99--1.05 0.20
Betaine 1.00 0.93--1.08 0.93
Female
Model 11 Model 12
OR 95% CI *p* OR 95% CI *p*
TMAO 1.25 0.74--2.12 0.40 1.14 0.65--2.01 0.65
Choline 1.10 0.86--1.41 0.46
L-Carnitine 0.99 0.93--1.05 0.78
Betaine 0.94 0.86--1.02 0.15
------------- ---------- ------------ ------ ------ ------------ ------
OR, odds ratio. 95% CI, 95% confidence interval. OR shown were for plasma TMAO, choline, L-carnitine, and betaine in all, male, and female participants. Models 7, 9, and 11 adjusted for age, gender, BMI, Glu, TG, and Cr. Models 8, 10, and 12 adjusted for all factors in models 7, 9, and 11 plus choline, L-carnitine, and betaine.
######
AUC of TMAO, choline, L-carnitine, and betaine for predicting CAD in all, male, and female participants.
AUC 95% CI *p*
---------------------------------------- ------ ------------ --------
All
TMAO 0.6 0.52--0.69 0.03
TMAO + choline + L-carnitine + betaine 0.64 0.56--0.72 \<0.01
Male
TMAO 0.64 0.51--0.77 0.04
TMAO + choline + L-carnitine + betaine 0.64 0.52--0.77 0.04
Female
TMAO 0.57 0.46--0.69 0.21
TMAO + choline + L-carnitine + betaine 0.64 0.53--0.75 0.02
######
AUC of TMAO, choline, L-carnitine, and betaine for predicting severe artery stenosis in all, male, and female participants.
AUC 95% CI *p*
---------------------------------------- ------ ------------ --------
All
TMAO 0.62 0.52--0.72 0.02
TMAO + choline + L-carnitine + betaine 0.66 0.56--0.75 \<0.01
Male
TMAO 0.64 0.51--0.78 0.05
TMAO + choline + L-carnitine + betaine 0.64 0.5--0.77 0.05
Female
TMAO 0.58 0.42--0.74 0.26
TMAO + choline + L-carnitine + betaine 0.68 0.53--0.83 0.01
[^1]: Academic Editor: Manoel Otavio C Rocha
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#section1-2324709618757259}
============
Takotsubo cardiomyopathy (TTC), also called "broken heart syndrome," "transient apical ballooning," and "stress cardiomyopathy," is an acute cardiac syndrome that mimics myocardial infarction and characterized by transient cardiac wall motion abnormalities. This occurs in the absence of any coronary artery obstruction or acute plaque rupture.^[@bibr1-2324709618757259],[@bibr2-2324709618757259]^ In most cases of TTC, the cardiac wall motion abnormality does not follow a single epicardial coronary artery territory. It is usually characterized by depressed function or akinesis of the mid and apical segments of the left ventricle along with hyperkinesis of the basal walls.^[@bibr1-2324709618757259]^ On the other hand, reverse TTC (r-TTC), or inverted TTC, has been recognized as a variant with a hypocontractile ventricular basal segment along with a hypercontractile apex.^[@bibr2-2324709618757259]^ Cases of r-TTC following surgery seems to be unusual and rare.^[@bibr3-2324709618757259],[@bibr4-2324709618757259]^ In this article, we report a case of r-TTC in a patient who underwent exploratory laparotomy for small bowel obstruction.
Case Presentation {#section2-2324709618757259}
=================
A 44-year-old female patient with a known history of multiple sclerosis (MS) maintained on immunomodulatory agents presented to our emergency department with abdominal pain, nausea, and vomiting. Her past surgical history was only remarkable for appendectomy and partial small bowel resection. The patient was diagnosed with small bowel obstruction after a computed tomography (CT) scan of her abdomen was performed. She was admitted to the surgical service subsequently. However, her clinical status deteriorated and she underwent exploratory laparotomy on the next day. One day after surgery, she started experiencing shortness of breath. Her physical examination was otherwise unremarkable. An electrocardiogram was remarkable only for sinus tachycardia. A CT angiography of the chest for suspected pulmonary embolism was performed and was unremarkable. Serum troponin-T levels done were elevated and peaked at 2.19 ng/mL (reference range: \<0.05 ng/mL). CT angiography of the coronary arteries was normal with a calcium score of zero. Echocardiographic examination done revealed a hyperkinetic apical wall along with a hypokinetic basilar wall of the myocardium suggestive of r-TTC with a left ventricular ejection fraction (LVeF) of 30% ([Video 1](http://journals.sagepub.com/doi/suppl/10.1177/2324709618757259); available in the online version of the article). Therapy with a β-blocker and an angiotensin-converting-enzyme inhibitor was initiated and the patient was discharged on postoperative day 8 with a persistent LVeF of 30%. A follow-up cardiac magnetic resonance imaging was done 14 days later after discharge revealed an improvement in the myocardial function with an LVeF of 54%. A repeat echocardiography also showed normalization of the LVeF.
Discussion {#section3-2324709618757259}
==========
The pathogenesis of both TTC and r-TTC is not well understood. Several mechanisms were postulated. The most widely accepted underlying etiological mechanism behind both types is sympathetic nervous system overactivation. Among the various neurochemical substances associated with cardiac wall motion abnormalities, epinephrine and norepinephrine seem to be the most crucial. This catecholamine surge is believed to mediate a vascular dysfunction leading to coronary artery vasospasm, microvascular dysfunction, hyperdynamic contractility, and direct myocardial toxicity via free radicals formation.^[@bibr5-2324709618757259]^
Furthermore, there seems to be a role in protein signaling within the myocardial cells that mediates a paradoxical negative inotropic effect to protect against the intense activation of β-adrenoceptors. This effect is greatest at the apical myocardium where the β-adrenoceptor density is highest.^[@bibr5-2324709618757259]^ This has been also proven by 123-meta-iodobenzylguanidine myocardial scintigraphy that implied more myocardial sympathetic innervation in the apex.^[@bibr6-2324709618757259]^ This might explain the myocardial stunning affecting the apical wall in TTC. However, it does not explain the hyperkinetic apical wall motion in r-TTC, neither the hypokinesis of the basal wall. It has also been postulated that as catecholamine levels subside after a surge, a quicker apical recovery might happen leading to r-TTC pattern.^[@bibr6-2324709618757259]^ Nevertheless, this again does not explain the hypokinesis observed in the basal wall.
Additionally, it is worth noting that certain clinical features differ between TTC and r-TTC.^[@bibr2-2324709618757259],[@bibr7-2324709618757259]^ Song et al^[@bibr8-2324709618757259]^ and Ramaraj et al^[@bibr9-2324709618757259]^ observed that in r-TTC patients are usually younger, tend to have a lower LVeF, and sustains a quicker myocardial recovery in comparison to TTC. Moreover, since the basilar part of the ventricle is the main involved region in r-TTC, which has more myocardial tissue, cardiac biomarkers are usually more elevated in comparison to TTC.^[@bibr2-2324709618757259],[@bibr9-2324709618757259]^
In the literature, most reported cases of r-TTC occurred after physical or emotional stress. Although surgery is considered a form of stress, it is less described. Nevertheless, it remains unclear whether r-TTC occurring after surgery is secondary to physical or emotional stress solely, or whether there is a role for anesthetic agents in triggering this cardiomyopathy. Furthermore, TTC and r-TTC have been reported in several neurological diseases such as subarachnoid hemorrhage, seizure, and ischemic stroke but rarely in MS patients.^[@bibr10-2324709618757259][@bibr11-2324709618757259][@bibr12-2324709618757259]-[@bibr13-2324709618757259]^ Biesbroek et al^[@bibr11-2324709618757259]^ and Kozu et al^[@bibr12-2324709618757259]^ both reported r-TTC in association with new MS lesion occurrence. They have speculated a possible role for demyelinating brain lesions interfering with sympathetic nervous system regulation. On the other hand, Peller et al^[@bibr10-2324709618757259]^ reported r-TTC in a patient with stable MS, which is similar to our case; however, our patient had the complication after surgery and without any acute electrocardiogram changes, in comparison to the previous cases. It is unclear in our case whether MS had a role in triggering r-TTC or whether surgery itself was the sole stress triggering factor.
Conclusion {#section4-2324709618757259}
==========
Reverse takotsubo cardiomyopathy is a rare type of stress-induced cardiomyopathy that is described mainly following neurological insults. Cases of r-TTC following surgery are less described in the literature. We presented a case of r-TTC in a patient with stable MS who underwent exploratory laparotomy for small bowel obstruction.
Supplementary Material
======================
###### Supplementary material
**Declaration of Conflicting Interests:** The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
**Funding:** The author(s) received no financial support for the research, authorship, and/or publication of this article.
**Ethics Approval:** Our institution does not require ethical approval for reporting individual cases or case series.
**Informed Consent:** Verbal informed consent was obtained from the patient(s) for their anonymized information to be published in this article.
**ORCID iD:** Tamer Akel  <https://orcid.org/0000-0002-5632-5571>
**Supplemental Material:** Supplementary material is available for this article online.
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION
============
Recent studies have demonstrated that the resting membrane conductance (*G~M~*) of skeletal muscle is highly regulated during repetitive firing of short trains of action potentials (APs) that replicate excitation patterns occurring during activity in skeletal muscle ([@bib30],[@bib31]). In fast-twitch rat muscle, this regulation of *G~M~* has two distinct phases. At the onset of AP firing, Phase I involves the inhibition of ClC-1 channels through a PKC-mediated mechanism causing a decline in *G~M~* to ∼40% of its resting value. Then, during prolonged activity, Phase II changes involve the opening of ClC-1 and K~ATP~ channels, increasing *G~M~* to four to five times its value in quiescent fibers. Upon cessation of AP firing, *G~M~* recovers to its level before AP firing in 1--5 min.
In our companion paper (see Pedersen et al. in this issue), we demonstrated the significance of this *G~M~* regulation for subthreshold electrical properties in muscle fibers, and from this predicted the effects of such *G~M~* changes upon their excitability. In particular, our study described the relative sensitivity of the different aspects of muscle excitability, including neuromuscular transmission, sarcolemmal AP propagation, and tubular (t)-system excitation, to such regulation. The study used a linear circuit analysis appropriate for subthreshold electrical membrane phenomena. This approach first allowed for the development of analytical solutions for three cable models of muscle fibers. To determine which of these models gave the best representation of the electrical properties of rat extensor digitorum longus (EDL) muscle fibers in which the *G~M~* regulation has been observed, experimental measurements of membrane impedance properties were compared with the electrical characteristics of the cable models. This demonstrated that circuit models of rat EDL muscle fibers require a substantial luminal t-system resistance to account for experimental observations of their impedance properties and the velocity with which APs propagate in these fibers.
The study went on to predict that such a luminal resistance enhances the propagation velocity of sarcolemmal APs. It also showed that a luminal resistance in series with the t-system membrane divides the voltage gradient between the intracellular and interstitial spaces into voltage drops across the t-system membrane and the luminal resistance. Because this voltage division is highly frequency dependent, the luminal resistance has important consequences for t-system excitation. Thus, high-frequency current will predominantly generate voltage gradients across the luminal resistance, whereas lower frequency components will be important for t-system excitation. The analytic approach of the previous study therefore demonstrated that *G~M~* changes predominantly affect the low-frequency membrane impedance. It was further demonstrated, by convolving experimental APs with the circuit models, that t-system excitation is a low-frequency phenomenon that is highly dependent on *G~M~*. Collectively, our companion study thus demonstrates that models of rat EDL muscle fibers must include a luminal resistance and suggests that neuromuscular transmission and t-system excitation are more sensitive to *G~M~* regulation than is sarcolemmal AP propagation.
The linear circuit analysis used in our companion study was also useful in providing analytic expressions that could distinguish the appropriate equivalent circuit model for rat EDL muscle fibers and quantify the t-system luminal resistance. However, such a linear analysis is necessarily incomplete when exploring the possible physiological roles of the t-system luminal resistance and *G~M~* regulation for the nonlinear membrane phenomena involved in AP propagation. Such nonlinear properties, explored in the present study, include voltage-dependent, time-dependent, and rectifying and Na^+^/K^+^-ATPase-mediated currents, as well as alterations in intracellular and t-system ionic concentrations during repetitive activity.
The aim of the present study was therefore to quantify the influence of t-system luminal resistances and *G~M~* regulation in skeletal muscle excitability and t-system ionic homeostasis. It uses a nonlinear, iterative approach based upon the charge--difference (CD) model of [@bib10] and [@bib12],[@bib13]). The development of a new model was necessary because no existing model simulates the full range of physical and electrophysiological properties that underlie t-system ionic homeostasis and its relationship with the membrane potential. Thus, early models applying circuit theory with voltage- and time-dependent conductances allow for the simulation of individual APs in the absence of ionic concentration or osmotic changes and are therefore unsuitable for simulating trains of APs where ionic concentrations shift substantially ([@bib2]; [@bib1]). Subsequent, more realistic models permitting the simulation of APs within the restricted extracellular space of a whole muscle ([@bib20]), and of t-system K^+^ handling during AP firing ([@bib38]), do not incorporate osmotic water movements. Therefore, they do not reach unique steady-state solutions that are independent of the initial values of key modeled variables, such as intracellular and intra--t-system ion concentrations ([@bib11]).
The modeling approach adopted here encompasses the known determinants of AP propagation and reaches a true history-independent steady state, thereby remaining applicable despite large perturbations in ionic concentration. It incorporates terms describing voltage- and time-dependent ion conductances and Na^+^/K^+^ pump activity; represents muscle fiber surface geometry using 99 linearly connected fiber segments permitting simulation of surface conduction; represents tubular geometry in terms of 20 concentric shells per fiber segment, each separated by a small luminal series resistance as was experimentally verified in our companion study ([@bib32]); and simulates osmotic water fluxes, thereby permitting it to reach a unique history-independent steady state ([@bib11]).
History independence of the modeled variables was demonstrated by initiating the new model from several sets of widely divergent unphysiological values for all ion concentration variables. In each case, the model relaxed to an identical and physiologically reasonable steady state, thereby demonstrating that it was capable of investigating the determinants of intracellular and intra--t-system ionic homeostasis, and additionally allowing for the use of model-derived initial values for variables such as intra--t-system ion concentrations for which there is little available experimental data.
The initial application of this model confirmed the prediction from the preceding analytical study that the t-system luminal resistance enhances the sarcolemmal AP propagation velocity and reduces t-system excitation. It further demonstrated the influence of active AP generation within the t-system on t-system excitation and t-system ionic homeostasis, and the effect of these upon the surface membrane potential. It was then used to explore the role of *G~M~* regulation in determining the threshold stimulus for AP firing, for t-system excitability, and for the homeostasis of K^+^ ions within the t-system during repetitive AP firing. The outputs of simulated trains of APs were shown to be in close quantitative agreement with experimental intracellular recordings of AP trains obtained from rat EDL muscle fibers.
Theory
------
Development of a CD model of rat skeletal muscle
------------------------------------------------
The electrical properties of rat EDL skeletal muscle fibers were modeled on CD principles ([@bib11]), adapting and extending the model of [@bib10] and [@bib12],[@bib13]) to include key nonlinear membrane phenomena of AP excitation and conduction and t-system excitation, as depicted in [Fig. 1](#fig1){ref-type="fig"}. The new model thus simulated the following features: longitudinal subdivision of the model fibers into multiple subsections connected to form a cable model; subdivision of the t-system into a series of concentric shells separated by series resistances representing the t-system luminal resistance; voltage- and time-dependent ion channels within the surface and t-tubular membranes; and ion and osmotic water fluxes across the surface and t-system membranes and the t-system luminal resistances.
{#fig1}
The model parameters were calibrated to the passive electrical properties of rat fast-twitch EDL muscle fibers that were obtained using the analytic model in our companion study ([@bib32]). This allowed its subsequent use in evaluating the physiological significance of the t-system luminal resistance and *G~M~* regulation in active muscle fibers that has been observed previously in AP-firing EDL fibers ([@bib30]). The model was developed to adhere strictly to four key principles: (a) current continuity; (b) conservation of charge and mass; (c) electrodiffusion of ions; and (d) osmotic equilibrium at steady state.
### (a) The principle of current continuity across the resistances and capacitances of the CD model.
The simulation of multiple apposing finite compartments within strict conservation principles requires that any flow of current completes a circuit to ground, and also that such currents, whether capacitative or resistive, correlate precisely with the correct changes in compartmental ion contents. Both current continuity and conservation principles can be achieved straightforwardly for ionic fluxes through ion channels and across resistances by the following equations:$$\Delta\left\lbrack S \right\rbrack_{a} = J_{\text{S}}A_{\text{m}(a)}\Delta t,\ \left( \text{mol\ l}^{- \text{1}}\text{s}^{- \text{1}} \right),$$$$\Delta\left\lbrack S \right\rbrack_{b} = - J_{\text{S}}A_{\text{m}(b)}\Delta t,\ \left( {\text{mol\ l}^{- \text{1}}\text{s}^{- \text{1}}} \right),$$where \[*S*\] represents the concentration of any solute (mol l^−1^), the subscripts *a* and *b* denote the compartments between which the ionic flux, *J* (mol cm^−2^), is flowing, and *A*~m~ is the membrane area per unit volume (cm^2^ l^−1^) for each compartment. Conservation of charge and ion concentration is clear from the relationship Δ\[*S*\]*~a~*/*A*~m(*a*)~ = −Δ\[*S*\]*~b~*/*A*~m(*b*)~.
The CD modeling approach additionally permitted calculation of the less straightforward requirement to characterize the precise ionic concentration changes resulting from capacitive currents, in which association of one ion species with one "plate" of a capacitor might displace or attract different ion species from or to the other plate, such that Δ\[*S*\]*~a~*/*A*~m(*a*)~ ≠ −Δ\[*S*\]*~b~*/*A*~m(*b*)~, in marked contrast to the situation with currents through ion channels. For example, hyperpolarization of the t-tubular membrane could result in the association of Na^+^ and/or Cl^−^ dissociation with the outer membrane surface, but the dissociation of K^+^ from the inner membrane surface. This possibility is made explicit by the following relationships for the surface and t-tubular membrane capacitance (*C*~m~, F cm^−2^) in a muscle fiber with a homogenous t-tubular compartment separated from the bulk extracellular fluid (ECF) by an access resistance, *R*~A~:$$E_{\text{m}} = Q_{\text{sm}}/\left( {C_{\text{m}}A_{\text{m}}} \right),\left( \text{V} \right)$$and:$$E_{\text{t}} = Q_{\text{tm}}/\left( C_{\text{m}}A_{\text{t}} \right),\left( \text{V} \right)$$while:$$E_{\text{A}} = E_{\text{m}} - E_{\text{t}},\left( \text{V} \right)$$where *Q*~sm~ is the net charge associated with the surface membrane capacitance, *Q*~tm~ is the net charge associated with the t-system membrane capacitance, with each expressed per liter of cell volume (C l^−1^), and *A*~m~ and *A*~t~ represent the membrane areas of the sarcolemma and t-system, respectively, referred to total cell volume (cm^2^ l^−1^). *E*~m~ represents the trans-sarcolemmal membrane potential, *E*~t~ is the trans-tubular membrane potential, and *E*~A~ is the potential across the tubular access resistance. Ionic fluxes then directly influence the ion concentrations and hence the net charges within the intracellular (*i*) and t-tubular (*t*) compartments. These relate to the charges associated with membrane capacitances by:$$Q_{\text{i}} = Q_{\text{sm}} + Q_{\text{tm}},\ \left( \text{C\ l}^{- \text{1}} \right)$$$$Q_{\text{t}} = - Q_{\text{tm}}.\ \left( \text{C\ l}^{- \text{1}} \right)$$
These relationships encapsulate the principles that (a) intracellular charge may be associated with either *C*~m~ or *C*~t~; and (b) any deposition of charge on one plate of *C*~t~ must be matched by an equal charge flowing from its opposing plate. These relationships incorporate the assumptions that (a) the outer surface of *C*~m~ is grounded via a pathway of insignificant resistance; and (b) the outer surface of *C*~t~ is grounded via a nonzero access resistance, *R*~A~, thereby allowing *E*~m~ and *E*~t~ to differ. Collectively, these assumptions imply that any alterations in *Q*~t~, for example because of currents across *R*~A~, are matched by equal alterations in charge on *C*~m~, thereby completing a circuit to ground.
### (b) Strict conservation principles in CD modeling of multiple interconnected finite compartments.
By tracking concentration changes and then calculating electrical potentials directly from the concentrations of charge carriers and the associated capacitance terms, the CD approach results in a model that obeys fundamental charge and mass conservation principles. This allows the model to reach true steady states that are independent of the initial values of any modeled variable ([@bib11]). Thus, for a single membrane-bound compartment, the electrical potential with respect to the bulk ECF is given at any time by:$$E_{\text{m}} = Q_{\text{i}}/C_{\text{m}},\ \left( \text{V} \right)$$where *E*~m~ is the membrane potential, and *C*~m~ is the membrane capacitance. *Q* denotes net charge, and the subscript "i" denotes the intracellular space, such that:$$Q_{\text{i}} = F\left( \left\lbrack {Na}^{+} \right\rbrack_{i} + \left\lbrack \text{K}^{+} \right\rbrack_{\text{i}} - \left\lbrack \text{Cl}^{-} \right\rbrack_{\text{i}} + z_{\text{X}}\left\lbrack X \right\rbrack_{\text{i}} \right),\ \left( \text{C\ l}^{- \text{1}} \right)$$where *z*~X~ is the mean charge valency of the various membrane-impermeant solutes, denoted *X*, and *F* is Faraday's constant. The influence of *z*~X~ and \[X\]~i~ on the steady state of the model has been described previously ([@bib10], [@bib11]; [@bib12]).
With the current--continuity assumptions explored above, relationships may be given for *E*~m~ and *E*~t~ in terms of the CDs within each compartment. First, if the t-system is considered as a single homogenous compartment, partially bounded by a membrane with capacitance per unit area that is identical to that of the surface membrane, then:$$E_{\text{m}} = \frac{\left( {\left\lbrack Q \right\rbrack_{\text{i}} + \left\lbrack Q \right\rbrack_{\text{t}}V_{\text{t}}} \right)}{\left( {C_{\text{m}}A_{\text{m}}} \right)},\ \left( \text{V} \right)$$$$E_{\text{t}} = \frac{- \left\lbrack Q \right\rbrack_{\text{t}}V_{\text{t}}}{C_{\text{m}}A_{\text{t}}},\text{(V)}$$$$\begin{matrix}
{\left\lbrack Q \right\rbrack_{\text{i}} = F\left( {\left\lbrack Na \right\rbrack_{\text{i}} + \left\lbrack K \right\rbrack_{\text{i}} - \left\lbrack Cl \right\rbrack_{\text{i}} + z_{\text{X}}\left\lbrack X \right\rbrack_{\text{i}}} \right)\ } \\
{\left\lbrack Q \right\rbrack_{\text{t}} = F\left( {\left\lbrack Na \right\rbrack_{\text{t}} + \left\lbrack K \right\rbrack_{\text{t}} - \left\lbrack Cl \right\rbrack_{\text{t}} + \left\lbrack Z \right\rbrack_{\text{t}}} \right),\mspace{22mu}\text{C\ l}^{- 1}} \\
\end{matrix}$$where *V*~t~ is the volume of the t-system relative to the cell volume, *V*~c~; and *Z*~(t)~ denotes the fixed charge density within the t-tubules (C l^−1^). It is implicit in these equations that the ionic activities of the mobile ions are 1. Sign conventions are as follows: the surface membrane potential (*E*~m~) is expressed as the potential of the intracellular space with respect to the ECF; and the t-system membrane potential (*E*~t(*n*)~) is expressed as the potential of the intracellular space with respect to the potential of the t-system lumen in each shell, *n*.
To calculate transmembrane potentials when the t-system is modeled as multiple concentric shells, the excess charge within the t-system is simply summed, thus:$$E_{\text{m}} = \frac{\left( {\left\lbrack Q \right\rbrack_{\text{i}} + {\sum\limits_{n = 0}^{N}{\left\lbrack Q \right\rbrack_{\text{t}(n)}V_{n}}}} \right)}{\left( {C_{\text{m}}A_{\text{m}}} \right)},\text{(V)}$$$$E_{\text{t(}n\text{)}} = \frac{- \left\lbrack Q \right\rbrack_{\text{t(}n\text{)}}V_{\text{t}}}{C_{\text{m}}A_{\text{t}}},\left( V \right)$$where *V~n~* (dimensionless) is the volume of each t-system shell, *n*, relative to the cell volume, and *V*~t~ is the volume of the t-system relative to cell volume.
### (c) Electrodiffusion of ions.
In all cases, ionic movements were calculated according to both the concentration gradient and the local electrical gradient. This contrasts with previous models of skeletal muscle fibers that consider either diffusional fluxes or electrical currents across t-system luminal resistances, but not both ([@bib9]; [@bib14]; [@bib2]; [@bib1]; [@bib20]). However, preliminary simulations demonstrated that AP activity resulted in alterations in both t-system ion concentrations and t-system electrical potential, thereby necessitating the simulation of both diffusional and electrically driven ion movements.
Transmembrane ionic fluxes were calculated using the Goldman flux equation ([@bib19]). Fluxes across *R~A~* and between t-system shells were calculated as the sum of electrical drift and diffusional fluxes:$$\begin{matrix}
{J_{\text{S((}n\text{-1)}\rightarrow n)} =} \\
{\left( \frac{V_{\text{t}}\sigma_{\text{t}}A_{n}}{\ \Delta} \right)\left( {g_{\text{t(}n\text{)}}\frac{\left\lbrack S \right\rbrack_{n}}{\left\lbrack C \right\rbrack_{n}}\frac{E_{t(n - 1)} - E_{t(n)}}{z_{\text{S}}F} + D\left( {\left\lbrack S \right\rbrack_{(n - 1)} - \left\lbrack S \right\rbrack_{(n)}} \right)} \right)\text{,}\ \left( \text{mol\ s}^{- \text{1}} \right)} \\
\end{matrix}$$where *σ*~t~ is the t-system tortuosity factor ([@bib38]) (dimensionless); *A~n~* is the area of shell *n* (dm^2^); *Δ* is the distance between successive t-system shells (dm); *g*~t~ is the inter-shell conductance (S dm^−1^) for shell *n*; \[*S*\]*~n~*/\[C\]*~n~* is the concentration of *S* relative to the total mobile ion concentration within shell *n* (dimensionless), such that each ionic species, *S*, contributes an appropriate proportion of the total conductance; and *D* is the diffusion coefficient (dm^2^ s^−1^), assumed to be equal for all modeled ions.
Fluxes across *R*~A~ were calculated as follows:$$J_{\text{S(e}\rightarrow\text{0)}} = A_{\text{m}}\left( {\frac{D}{R_{\text{A}}g_{\text{t}}}\left( {\left\lbrack S \right\rbrack_{\text{e}} - \left\lbrack S \right\rbrack_{0}} \right) + \frac{\left\lbrack S \right\rbrack_{0}}{\left\lbrack C \right\rbrack_{0}}\left( \frac{- E_{0}}{R_{\text{A}}z_{\text{S}}F} \right)} \right),\ \left( {\text{mol\ l}^{- \text{1}}\text{s}^{- \text{1}}} \right)$$where the *D*/*R*~A~*g*~t~ term ensures that the ratio between the diffusive and electrical components of the flux across *R*~A~ is equal to that between t-system shells.
### (d) Osmotic equilibrium at steady state.
[@bib10], [@bib11]) showed that when the steady state of a CD model cell is constrained by the requirement for osmotic equilibrium at steady state, then the steady-state values of these variables were uniquely defined by the permeabilities and extracellular concentrations of each ion and by the Na^+^/K^+^ pump density. Furthermore, certain concentrations are tightly constrained at osmotic equilibrium. For example, when Cl^−^ is the only extracellular anion and therefore contributes half of the extracellular osmolality, \[Cl^−^\]~t~ must be equal to \[Cl^−^\]~e~ for t-system osmolality to be equal to extracellular osmolality. Thus, osmotic water movements are fundamental to ensure accurate steady-state behavior that is constrained to what is physically possible. This allows CD modeling to be used to simulate manipulations expected to have a large influence on ion concentrations, in contrast to models that do not reach true steady states ([@bib11]). Thus, transmembrane water fluxes were calculated as:$$J_{\text{H2O}(\text{a}\rightarrow\text{b})} = A_{\text{m}}P_{\text{H2O}}\left( \mathit{\Pi}_{\text{b}} - \mathit{\Pi}_{\text{a}} \right),$$where *Π* is the osmolality of each compartment and *P*~H2O~ is the hydraulic permeability of the membrane, taken as 156 µm s^−1^ ([@bib15]). Water fluxes across *R*~A~ and between t-system shells were assumed to be instantaneous to maintain osmotic equality in all extracellular compartments.
It is clear from the profound changes in t-system volume with manipulations of extracellular ion concentrations or osmolality ([@bib34]; [@bib8]; [@bib6]) that the t-system cannot be treated as a rigid structure. The model therefore allowed the t-system volume to be determined by the balance of osmotic fluxes across the t-system membrane and *R~A~*. This gave stable steady-state values when a small fixed negative charge term was included in the t-system and responded appropriately to changes in extracellular osmolality in preliminary experiments (unpublished data).
Simulation of cable properties and AP firing
--------------------------------------------
A length of muscle fiber was simulated as 99 identical segments. Ion concentrations and membrane potentials were calculated for each segment as described above. Inter-compartment ion fluxes were calculated for each intracellular ionic species according to the prevailing electrical gradient between segments:$$J_{\text{S(}x\rightarrow(x + 1))} = \frac{A_{\text{x}}}{R_{\text{L}}z_{\text{S}}FV_{\text{x}}}\left( \frac{E_{\text{m(}x\text{)}} - E_{\text{m}(x + 1)}}{\Delta L} \right),\left( \text{mol\ dm}^{- \text{3}}\text{s}^{- \text{1}} \right)$$where *A*~x~ is the membrane area of compartment *x* (dm^2^), *V*~x~ is its volume (dm^3^), *R*~L~ is the longitudinal resistance, and Δ*L* is the distance between the centers of adjacent compartments. As each segment was identical, ion concentrations did not differ significantly between adjacent segments; therefore, it was not necessary to include a diffusional flux term.
The simulations were performed using a 100-µm segment length. Initial simulations to determine the optimal segment length compared the behavior of model fibers with 100-µm segment lengths to those with 10-µm segment lengths. Over a range of frequencies from 25 to 3,000 Hz, frequency--velocity relationships for passive sinusoidal currents were not influenced by this change in segment length (difference in velocity \<1% at each frequency; not depicted).
APs were modeled by adding voltage-gated Na^+^ and K^+^ channels to the surface and t-tubular membranes. Voltage-gated ion channels were simulated using Hodgkin--Huxley gating equations ([@bib22]), with conductances and gating parameters calibrated to rat skeletal muscle ([@bib5]; [@bib20]; [@bib38]). Nevertheless, the present model differs significantly from these existing models in several important respects. First, all modeled variables, including t-system ion concentrations and ion channel--gating variables, reach stable steady-state values that are independent of initial conditions. In contrast, these existing models treat some or all ion concentrations as fixed parameters. Second, the present model simulates ion diffusion between the t-system and extracellular space according to the prevailing electrical and concentration differences. In contrast, existing models simulate passive diffusion, yet particularly when *E*~t~ lags *E*~m~ during the early phase of the surface AP, the voltage between the t-system and the ECF was shown to reach almost 100 mV in these earlier studies ([@bib38]).
A stimulus was required to initiate APs. This was applied to the central compartment in the cable model. To maintain strict conservation principles, stimulation currents were simulated as fluxes of K^+^ and Cl^−^ of equal magnitude and opposite direction, thereby approximating stimulation with a KCl-filled microelectrode.
Calibration to rat skeletal muscle
----------------------------------
The physical conductance parameters summarized in [Table I](#tbl1){ref-type="table"} were calculated from experimental measurements of subthreshold sine wave conduction using the analytic model of [@bib32]. Resting membrane permeabilities for Na^+^, K^+^, and Cl^−^ were calculated from membrane conductance measurements, literature values for the ratio *P*~Cl~/*P*~K~, and model fitting of *E*~m~ to the present experimental data for the *P*~Na~/*P*~K~ ratio. Unless otherwise stated, the ionic permeabilities per unit area of the t-system membrane and the sarcolemma were assumed to be identical. The Na^+^/K^+^ pump was simulated using the model of [@bib21], and its membrane density was chosen to produce reasonable resting values of \[Na^+^\]~i~. Finally, voltage-gated channel permeabilities, distribution, and kinetics were as described by [@bib20].
######
Summary of physical and electrophysiological parameters used
Symbol Description Value Reference
--------------------------------- ------------------------------------------------------ --------------------- -----------
Physical parameters
*A*~m~ Surface membrane area 769231 cm^2^ l^−1^ 1
*A*~t~ t-System membrane area 3076924 cm^2^ l^−1^ 1
*V*~t~ Relative t-system volume 3.2e-3 liter.l^−1^ 2, 7
*C*~m~ Membrane capacitance 1e-6 F cm^−2^ 1
*R*~a~ t-System access resistance 72.3 Ω cm^2^ 1
*R*~i~ Longitudinal fiber resistance 180 Ω cm 1
*g*~t~ t-System luminal conductance 3.7e-3 S cm^−1^ 2
*σ*~t~ t-System tortuosity factor 0.21 2
Electrophysiological parameters
*N*~s~ Na^+^/K^+^-ATPase sarcolemmal density 9e-12 cm^−2^ 3
*N*~t~ Na^+^/K^+^-ATPase tubular membrane density 4.5e-12 cm^−2^ 3
*P*~Na~ Background Na^+^ permeability 2.16e-8 cm s^−1^ 4, 5
*P*~K~ Background K^+^ permeability 6.3e-7 cm s^−1^ 4, 6
*P*~Cl~ Background Cl^-^ permeability 2.7e-6 cm s^−1^ 4, 6
*P*~Na(v,s)~ Voltage-gated Na^+^ channel sarcolemmal permeability 0.00265 cm s^−1^ 7
*P*~Na(v,t)~ Voltage-gated Na^+^ channel t-system permeability 0.00034 cm s^−1^ 7
*P*~K(v,s)~ Voltage-gated K^+^ channel sarcolemmal permeability 0.0004 cm s^−1^ 7
*P*~K(v,s)~ Voltage-gated K^+^ channel t-system permeability 0.0001 cm s^−1^ 7
References: 1, [@bib32]; 2, [@bib38]; 3, model derived to maintain \[Na^+^\]~i~ in physiological range; 4, total conductance taken from [@bib32]; 5, *P*~Cl~/*P*~K~ ratio ([@bib7]; [@bib29]); 6, *P*~Na~/P~K~ ratio model derived to obtain membrane potentials in agreement with control experiments; 7, [@bib20]. Hodgkin-Huxley parameters for voltage-gated Na^+^ and K^+^ channels are from [@bib20]. Note that all intracellular ion concentrations and compartmental volumes in the model are variables, the values of which are determined solely by the choice of parameters, as demonstrated in [Fig. 2](#fig2){ref-type="fig"}. Extracellular Na^+^ and K^+^ concentrations matched those of the experimental studies, and extracellular \[Cl^−^\] was chosen to make the solution electroneutral.
######
From legend to [Fig. 2](#fig2){ref-type="fig"}
Variable (units) Initial value 1 Initial value 2 Final value 1 Final value 2
--------------------- ----------------- ----------------- --------------- ---------------
\[*Na^+^\]~i~* (mM) 200 50 5.84 5.84
\[*K^+^\]~i~* (mM) 0 250 180.3 180.3
\[*Cl^-^\]~i~* (mM) 0 50 7.37 7.37
\[*X\]~i~* (mM) 121.4 151.7 108.5 108.5
\[*Na^+^\]~t~* (mM) 0 0.01 157.6 157.6
\[*K^+^\]~t~* (mM) 0.01 1 6.38 6.38
\[*Cl^-^\]~t~* (mM) 0 1.009 138.0 138.0
\[*Z\]~t~* (l^-1^) −0.01 −0.001 −25.4 −25.4
Note that the final value of each modeled variable, shown at time ∞ in [Fig. 2](#fig2){ref-type="fig"}, is both physiologically reasonable and independent of its initial value.
Model initiation and validation
-------------------------------
A critical advantage of CD modeling is that *E*~m~, *E*~t~, and all intracellular and intra--t-system ion concentrations are dependent variables that have unique steady-state values for any given set of physical, permeability, and pump parameters. Therefore, the model may be initiated with arbitrary values of the principal modeled variables (\[*Na*^+^\]~i~, \[*K*^+^\]~i~, \[*Cl*^−^\]~i~, \[*X*\]~i~, \[*Na*^+^\]~t~, \[*K*^+^\]~t~, \[*Cl*^−^\]~t~, and \[*Z*\]~t~). In the present study, the model was run to steady state from two markedly different semi-arbitrary sets of these variables using a single physiological parameter set as listed in [Table I](#tbl1){ref-type="table"}. As shown in [Fig. 2](#fig2){ref-type="fig"}, the eventual steady-state values of all modeled variables were identical in each case. This demonstrates that the new model fully obeys conservation principles. Furthermore, the eventual steady-state values of all variables are physiologically reasonable and in agreement with published data, thereby providing some validation of the model. This procedure was also used to obtain the steady-state values in the multiple-shell t-system model (unpublished data), similarly demonstrating a unique steady-state independent of the starting values of the variables.
{ref-type="table"}), but with two sets of intracellular ion concentrations that were chosen to be markedly different from each other and clearly unphysiological, although fulfilling a constraint of bulk electroneutrality in each compartment. See [Table II](#tbl2){ref-type="table"} below for a summary of the initial and steady-state values of the major modeled variables.](JGP_201110617R_GS_Fig2){#fig2}
MATERIALS AND METHODS
=====================
Animal handling
---------------
Experimental recordings of trains of APs were obtained from surface muscle fibers of intact EDL muscles from 12-wk-old female Wistar rats of our own breed (∼230 g). Animals were kept at 21°C and fed ad libitum living under 12:12-h light/dark conditions. All handling and killing of animals followed Danish welfare regulations. After dissection, the muscles were incubated in standard Krebs-Ringer bicarbonate solution (pH 7.4 at 30°C) containing (in mM): 122 NaCl, 25 NaHCO~3~, 2.8 KCl, 1.2 KH~2~PO~4~, 1.2 MgSO~4~, 1.3 CaCl~2~, and 5.0 [d]{.smallcaps}-glucose. All solutions were equilibrated with a mixture of 95% O~2~ and 5% CO~2~.
Recordings of trains of APs
---------------------------
Trains of APs were recorded using a two-microelectrode setup described in detail previously ([@bib29]). In brief, two sharp electrodes (∼20 MΩ) were inserted into the same muscle fiber, and the train of APs was initiated by injecting an ∼3.5-s train of constant current pulses (3 ms, 300 nA) into the fiber through one electrode while the other electrode recorded the membrane voltage. To allow the electrodes to remain inserted during AP trains, contractile activity of the muscle fibers was reduced by including 50 µM of the specific MHC II inhibitor, *N*-benzyl-*p*-toluene sulphonamide (BTS), to the Ringer's solution as described by [@bib24].
RESULTS
=======
Our companion study ([@bib32]) demonstrated that linear electrical circuit models must incorporate a significant t-system luminal resistance to accurately model the subthreshold electrical properties of rat EDL muscle fibers. It was possible to evaluate the physiological significance of such resistances for sarcolemmal AP propagation velocity and t-system excitation. This led to an evaluation of the role of *G~M~* regulation during repetitive firing of short AP trains ([@bib30]). The present study proceeded to a quantitative analysis of important nonlinear responses not amenable to the preceding analytic methods of [@bib32] using a new nonlinear model based on CD modeling principles ([@bib10]; [@bib12],[@bib13]). First, the frequency dependency of the conduction velocity of passively conducted sine waves was quantified in the new CD model and compared with the previous findings based on linear current--voltage relationships ([@bib32]). Second, the influence of voltage- and time-dependent ion channels within both the surface and t-system membranes was clarified. Third, this permitted an analysis of t-system ion concentration changes and their influence upon surface and t-system membrane potential and excitability during repetitive AP firing. Experimental microelectrode recordings of membrane potential were compared with the results of the CD modeling.
Passive conduction properties of skeletal muscle fibers
-------------------------------------------------------
[Fig. 3 A](#fig3){ref-type="fig"} demonstrates the relationship between frequency and passive conduction velocity in simulations without voltage-gated ion channels under four different degrees of separation of the t-system from the true ECF. Fiber properties were simulated assuming: (a) no *R*~A~ (simple cable model); (b) a physiological *R*~A~ of 72.3 Ω cm^2^ (homogenous t-system model; equivalent to the lumped model of [@bib32]); (c) physiological *R*~A~ and additional smaller series resistances (3.7 mS cm^−1^; [@bib38]) between each of 20 concentric t-system shells (20-shell t-system model, equivalent to the distributed cable model in [@bib32]); and (d) extremely high *R*~A~ (723 Ω cm^2^) simulating near-complete isolation of the t-tubules from the ECF (detubulation model).
{#fig3}
In full agreement with the findings of our companion analysis ([@bib32]), the homogenous t-system model showed faster conduction than the simple cable model, particularly at higher frequencies. This was despite the fundamentally different assumptions and approaches underlying the two models. The 20-shell t-system model ([@bib38]) similarly showed faster conduction velocity \<1,000 Hz, but slightly slowed conduction velocity \>1,000 Hz, relative to the homogenous t-system model. It is interesting to note that both physiological *R*~A~ and intra--t-system series resistances appear to enhance the conduction velocity at frequencies corresponding to those of the AP upstroke ([@bib32]). Further increases in *R*~A~ to yield a detubulation model produced further accelerated conduction at lower frequencies (\<600 Hz) but slowed conduction at all higher frequencies relative to fibers with a physiological *R*~A~, regardless of whether additional intra--t-system resistances were present.
[Fig. 3 (B and C)](#fig3){ref-type="fig"} demonstrates the changes in t-system transmembrane voltage, *E*~t~, and the voltage across the sarcolemma, *E*~m~, during sine wave propagation. In full agreement with the analytical solutions of our companion study, the CD model demonstrates that the voltage division of the intracellular potential between the t-system membrane and *R*~A~ is highly frequency dependent. Thus, at low frequency (50 Hz; [Fig. 3 B](#fig3){ref-type="fig"}), *E*~t~ tracks *E*~m~ relatively closely, indicating that the t-system membrane then dominates the impedance path from the cytosol through the t-system to the ECF. At higher frequency (1,000 Hz; [Fig. 3 C](#fig3){ref-type="fig"}), the amplitude of *E*~t~ is significantly reduced relative to that of *E*~m~, corresponding to *R*~A~ rather than the t-system membrane progressively dominating this impedance path. The larger voltage drop across *R*~A~ when compared with that across the t-system membrane at high frequencies has two consequences. First, the t-system membrane is poorly charged by high-frequency components of the circuit currents during AP propagation. Second, because current and voltage do not lag on a resistive component, *R*~A~ reduces the phase lag between current and voltage at such frequencies. This reduces the time required for high-frequency currents important for AP propagation to charge the sarcolemma. This latter point explains the steeper frequency--velocity relationship observed in models possessing a t-system luminal resistance ([Fig. 3 A](#fig3){ref-type="fig"}) and furthermore, as in the previous linear analysis, suggests that *R*~A~ causes the sarcolemmal AP to propagate faster.
The similarities in frequency--velocity relationships shown by the CD and linear circuit models ([@bib32]) occurred despite the distinctly nonlinear current--voltage relationships for each ionic current and the resultant total currents in the former. These nonlinearities result from the differing intracellular and extracellular concentrations of the major transmembrane charge carriers, K^+^ and Cl^−^, resulting in equivalent membrane resistances varying from 2.0 to 5.3 kΩ in the surface membrane and 0.6 to 1.3 kΩ in the t-system during the 50-Hz sine wave, in each case with the lower resistances occurring with greater depolarization. Thus, the nonlinearity of background current--voltage relationships does not influence passive conduction velocity.
Conduction velocity of actively propagated signals in skeletal muscle fibers
----------------------------------------------------------------------------
The findings above thus demonstrate that CD modeling successfully reproduces experimental measurements and linear modeling of passive frequency--velocity relationships in EDL muscle fibers ([@bib32]). Thus, the simulations of passive conduction demonstrate that increased signal frequency, and increased *R*~A~ at frequencies below ∼800 Hz, each produce a greater conduction velocity, but also decreased t-system excitation. However, AP conduction velocities in skeletal muscle under physiological conditions depend upon a combination of passive signal propagation and active signal regeneration. Furthermore, coordinated excitation--contraction coupling along the entire length of a muscle fiber requires both rapid AP conduction and t-system excitation. Simulations of regenerative AP conduction were therefore performed to: (a) determine the influence of active regeneration upon the conduction velocity of surface APs; (b) investigate the influence of *R*~A~ and the presence or otherwise of voltage-gated channels within the t-system upon AP conduction velocity and t-system excitation; and (c) investigate the influence of the *G~M~* changes observed during repetitive AP firing on surface and t-system excitability.
### The influence of active signal regeneration upon conduction velocity.
[Fig. 4](#fig4){ref-type="fig"} demonstrates the importance of active signal regeneration to sarcolemmal conduction velocity. It displays results from two model fibers with physiological densities of voltage-gated Na^+^ and K^+^ channels after stimulation in the central of 99 compartments into which the fiber long axis was divided. [Fig. 4 A](#fig4){ref-type="fig"} compares voltage responses in six adjacent compartments 3--3.5 mm from the initial stimulus along each model fiber. The solid lines depict results from a model fiber with a physiological density of voltage-gated ion channels. The broken lines depict the passive conduction remaining when the permeabilities of the voltage-gated ion channels in all compartments beyond 3 mm from the stimulus were fixed at their resting values. The latter shows that signal amplitude steeply declines with distance, reflecting an absence of active signal regeneration. Furthermore, the increased spacing of the peaks with distance in the passive conduction model (marked by vertical broken lines) indicates a marked deceleration of the passively propagated signal with distance from its source, relative to the actively propagated signal.
{#fig4}
[Fig. 4 (B--D)](#fig4){ref-type="fig"} goes on to examine the detailed effects of *P*~Na~ upon the conduction characteristics. [Fig. 4 B](#fig4){ref-type="fig"} (inset) shows AP traces resulting from setting the maximum permeability of the voltage-gated Na^+^ channel (*P*~Na(max)~) at 200% (*i*), 100% (*ii*), 50% (*iii*), or 30% (*iv*) of the value expected under control conditions ([@bib20]). The AP magnitude increased and the time to peak decreased with increasing *P*~Na(max)~. The resulting sarcolemmal conduction velocity of actively propagated APs ([Fig. 4 B](#fig4){ref-type="fig"}, squares) then showed a log-linear dependence upon *P*~Na(max)~, similar to the relationship between passive conduction velocity and sine wave frequency ([Fig. 3](#fig3){ref-type="fig"}). This contrasted with the shallower dependence on *P*~Na(max)~ of the velocity of passively conducted AP waveforms over the first 100 µm ahead of the active AP wave front. This is consistent with the highest frequency, fastest conducting components of the AP having the shortest length constants ([@bib32]).
[Fig. 4 C](#fig4){ref-type="fig"} summarizes a power spectral analysis of the AP traces in the inset of [Fig. 4 B](#fig4){ref-type="fig"} obtained by their Fourier transformation to permit a comparison of their frequency content and the conduction velocity of the corresponding APs. It clearly correlated an increased *P*~Na(max)~ with a larger content of high-frequency components in the resulting AP waveforms. This could be compared directly with the frequency dependence shown by the propagation velocity of passive subthreshold sinusoidal voltage changes, through identification of an equivalent frequency for the AP equal to the sine wave frequency, whose conduction velocity matched that of the AP. As passive frequency components with conduction velocities slower than that of the AP cannot contribute to the AP propagation velocity, this equivalent frequency therefore identifies the lowest frequency component of the circuit currents that could possibly contribute to the AP propagation velocity. [Fig. 3 A](#fig3){ref-type="fig"} was used to extract the equivalent frequencies for the different conduction velocities obtained at the different *P*~Na(max)~ values. [Fig. 4 D](#fig4){ref-type="fig"} plots the resulting equivalent frequencies against either AP conduction velocity or *P*~Na(max)~. The relationships were linear for the conduction velocity and log-linear for *P*~Na(max)~ ranging from 0.3 to 10 times the control value. Collectively, these findings clearly illustrate the influence of *P*~Na(max)~ upon the high-frequency components of the AP, and through the latter upon AP conduction velocity.
### The effect of t-system excitability upon surface conduction velocity and t-system excitation.
The simulations depicted in [Fig. 4](#fig4){ref-type="fig"} demonstrate that, over anything but the shortest distances, active signal regeneration by sarcolemmal voltage-gated ion channels enhances conduction velocity. The simulations that follow now analyze the effects of the voltage-gated ion channels additionally present in the t-system membrane upon t-system excitation and surface conduction velocity.
[Fig. 5](#fig5){ref-type="fig"} compares passive (A and C) and active (B and D) t-system responses to a surface AP in a model with a single homogenous t-system compartment (A and B) and in a model with a t-system subdivided into 20 concentric shells of equal thickness (C and D). In each case, changes in *E*~t(*n*)~ lag changes in *E*~m~. In the single-compartment t-system model, the magnitude of the change in *E*~t~ (Δ*E*~t~) is increased from +73.5 mV in the absence of voltage-gated ion channels within the t-system ([Fig. 5 A](#fig5){ref-type="fig"}) to +109.8 mV with active regeneration (B). However, findings from the 20-shell t-system model suggest that the homogenous t-system model underestimates the importance of active regeneration within the t-system. Thus, the 20-shell model shows that the *E*~t~ change in the center of the fiber (Δ*E*~t(19)~) is increased from +40.9 mV in the absence of voltage-gated ion channels within the t-system ([Fig. 5 C](#fig5){ref-type="fig"}) to +118.0 mV with active regeneration (D). This contrast between the two t-system models is of some relevance, as Δ*E*~t~ in the passive single-compartment t-system model would be sufficient for activation of excitation--contraction coupling, whereas Δ*E*~t(19)~ in the passive 20-shell t-system model is well below this activation threshold ([@bib18]).
{#fig5}
[Fig. 5 (E and F)](#fig5){ref-type="fig"} demonstrates the influence of *R*~A~ magnitude upon the peak *E*~t~ change during an AP (E), and upon the sarcolemmal AP conduction velocity (F). In the homogeneous model ([Fig. 5, E and F](#fig5){ref-type="fig"}, broken lines), an increase in *R*~A~ decreases the passive change in *E*~t~ during AP in fibers without voltage-gated ion channels in the t-system (E, triangles), but accelerates the sarcolemmal AP conduction velocity (F, triangles). Similarly, the presence of a series resistance separating successive t-system shells in the 20-shell t-system model ([Fig. 5, E and F](#fig5){ref-type="fig"}, solid lines) produces a further modest acceleration of the sarcolemmal AP (F), while greatly decreasing the passive t-system voltage response (E), relative to that of the homogenous t-system model. However, in the presence of voltage-gated ion channels in the t-system ([Fig. 5, E and F](#fig5){ref-type="fig"}, squares), the change in *E*~t~ during an AP is \>100 mV in both the homogenous and 20-shell t-system models, even for values of *R*~A~ up to three times the physiological value.
Thus, these findings demonstrate that t-system luminal resistance enhances surface AP conduction at the expense of a reduced passive excitation of the t-system. However, even in the presence of a significant access resistance, voltage-gated ion channels within the t-system permit full t-system excitation during an AP. Finally, voltage-gated channels in the t-system do not contribute to the velocity of surface AP conduction.
### The influence of G*~M~* regulation in active muscle on sarcolemmal and t-system excitability.
Despite the advantages of the analytical solution in generating a detailed mechanistic insight into the relationship between *G~M~* and excitability ([@bib32]), the approach was limited to linear analysis. The present study therefore extended the work to determine the influence of *G~M~* changes upon muscle excitability in the presence of voltage-gated ion channels in sarcolemma and in the t-system. The simulations test the prediction that AP excitation at the neuromuscular junction and the t-system, but not along the sarcolemma, should be particularly sensitive to alterations in *G~M~*, owing to sarcolemmal AP propagation being conveyed by high-frequency elements of the circuit currents for which the impedance of a sink membrane region would be relatively insensitive to *G~M~*.
Thus, [Fig. 6](#fig6){ref-type="fig"} shows the influence of Phase I and Phase II *G~M~* changes (see [Fig. 9 C](#fig9){ref-type="fig"}) on the excitation of the sarcolemma AP. During Phase I, it was possible to trigger an AP with a current that was slightly lower than the current that triggered an AP under control conditions, whereas during Phase II, a substantially larger current was required. Thus, sarcolemmal AP excitation is facilitated by Phase I G*~M~* regulation and markedly impaired by Phase II *G~M~* regulation, even when excitation proceeds with currents of short durations comparable to the endplate current.
{#fig6}
[Fig. 6 B](#fig6){ref-type="fig"} shows the subthreshold responses under each condition. *G~M~* strongly influences the downstroke of the membrane potential response, exerting smaller influences on its upstroke and magnitude. The importance of this influence on the downstroke is demonstrated in [Fig. 6 C](#fig6){ref-type="fig"}, in which it is shown that for near-threshold stimuli, the AP upstroke occurs after the cessation of the 1-ms stimulus. [Fig. 6 D](#fig6){ref-type="fig"} additionally demonstrates that APs excited during Phase II were characterized by a reduced amplitude and slower upstroke as compared with APs during Phase I and under control conditions.
Next, sarcolemmal AP propagation and t-system excitation were simulated in the model during control conditions, Phase I, and Phase II. [Fig. 6 E](#fig6){ref-type="fig"} shows the sarcolemmal AP (black lines) and the membrane voltage in the t-system shells (red, blue, and green lines) at 3 mm from the point of excitation. Under control conditions, the simulated APs propagated at 2.03 m s^−1^, whereas during Phase I and Phase II, it propagated at 2.09 and 1.67 m s^−1^, respectively. Thus, Phase I caused only a 3% increase in propagation velocity, whereas Phase II reduced the propagation velocity by 18%. Under conditions when the maximum permeability of the voltage-gated Na^+^ channels was reduced to 50%, the sarcolemma propagation velocity dropped to 1.52 m s^−1^ under control conditions ([Fig. 6 F](#fig6){ref-type="fig"}). The increase in sarcolemmal AP propagation velocity during Phase I became marginally more pronounced (5%), whereas during Phase II, the sarcolemmal AP propagation velocity dropped by 36%. There were also clear-cut effects of *G~M~* regulation on t-system excitation and t-system AP waveform. Thus, the amplitude of the AP in the innermost t-system shell became 12 and 14% larger during Phase I, whereas during Phase II, it was reduced by 19%. With reduced *P*~Na(max)~, the t-system shells did not actually produce an AP, despite the existence of fully propagating sarcolemmal AP during Phase II. The latter finding suggests that *G~M~* regulation during Phase II can cause loss of t-system excitation when voltage-gated Na^+^ channels have been partly inactivated, as can occur through slow inactivation of these channels during repetitive AP firing. This suggests that *G~M~* regulation could contribute to muscle fatigue through reductions in both endplate and t-system excitability.
Changes in t-system ionic homeostasis and membrane potential during AP trains
-----------------------------------------------------------------------------
Skeletal muscle function in vivo generally involves sequential as opposed to lone AP firing. The approach used in this analysis permitted study of ion concentration and membrane potential homeostasis associated with such repetitive firing. It is generally established that repetitive AP firing in skeletal muscle fibers markedly elevates t-system luminal \[K^+^\] ([@bib3]; [@bib38]; [@bib36]). It has also been speculated that these could depolarize the t-system membrane sufficiently to produce significant Na^+^ channel refractoriness and thereby reduce t-system excitability. Thus, tubular K^+^ accumulation has been implicated in muscular fatigue ([@bib37]). Experimental studies report such accompanying depolarizations of the resting membrane potential ([@bib14]).
However, experimental explorations of the possible tubular K^+^ changes, depolarization phenomena, the relationship between them, and the extent to which the voltage changes might provide a means for assessing tubular K^+^ homeostasis during AP firing are precluded by real-time assessments for \[K^+^\]~t~ not being available. Nevertheless, CD modeling provided an approach to such an analysis, as it calculates the concentrations of all ions in all compartments. The present studies determined the extent to which depolarization of the resting membrane potential during an AP train in skeletal muscle fibers might result from t-system K^+^ accumulation, through comparing experimental recordings with simulations of AP trains using the 20-shell t-system model.
[Fig. 7 A](#fig7){ref-type="fig"} shows representative experimental recordings of AP trains with AP-firing frequencies of 6, 15, and 30 Hz in EDL muscle fibers. In each case, the resting membrane potential developed a depolarization over the first ∼0.5 s of AP firing that then remained relatively stable despite continuation of the AP trains for 3.5 s. The magnitude of this depolarization increased with the AP-firing frequency. Thus, on average, the maximum baseline depolarization was 0.97 ± 0.15 mV at 6 Hz (*n* = 13), 2.08 ± 0.12 mV at 15 Hz (*n* = 23), and 4.4 ± 0.33 mV at 30 Hz (*n* = 19). On cessation of AP firing, the resting membrane potentials recovered over ∼1 s to their prestimulation value. Note that the resting membrane potential during AP firing was recorded immediately before each AP was triggered.
{#fig7}
[Fig. 7 B](#fig7){ref-type="fig"} shows sarcolemmal potentials during simulated AP trains of the same duration and frequency as in the experimental records depicted in [Fig. 7 A](#fig7){ref-type="fig"}. [Fig. 7 C](#fig7){ref-type="fig"} then compares the experimental and model records of sarcolemmal depolarization, measured as the difference between the potential immediately before each stimulation and that immediately before the first AP. The CD model successfully replicated the magnitudes of the resting membrane potential depolarizations seen during the experimentally recorded AP trains.
To determine whether the depolarization of the resting membrane potential was associated with alterations in t-system \[K^+^\], [Fig. 7 D](#fig7){ref-type="fig"} shows t-system \[K^+^\] for each of the 20 t-system shells during the AP trains. AP firing led to marked elevations in \[K^+^\]~t~, particularly in the deeper regions of the t-system. At all frequencies, \[K^+^\]~t~ reached a steady level after ∼1 s of AP firing and rapidly recovered after AP firing. The magnitude of change in \[K^+^\]~t~ increased with frequency. Thus, mean t-system \[K^+^\]~t~ increased from 5.3 mM before AP firing to ∼6.8 mM at 6 Hz, ∼8.8 mM at 15 Hz, and ∼11.2 mM at 30 Hz.
The cause of t-system \[K^+^\] accumulation was then further explored. First, simulation of fibers without t-system voltage-gated Na^+^ channels showed a depolarization of only 0.3 mV when stimulated at 15 Hz for 4 s, in comparison with the 2.1-mV depolarization in modeled fibers with normal t-system ion channel densities. This reflected a much reduced accumulation of K^+^ within the t-system (increase of \[K^+^\]~t~ of around 0.5 mM compared with 5.9 mM under normal conditions). Similarly, there was an absence of resting potential depolarization in a model of detubulation (unpublished data). Therefore, the model suggests that t-system AP firing is responsible for the K^+^ accumulation and baseline membrane potential depolarization seen during experimental AP trains, and therefore might be used experimentally to assess whether sarcolemmal APs trigger APs in the t-system.
The simulations demonstrated much smaller sarcolemmal and tubular membrane depolarization than would be expected simply from the alterations in the equilibrium potential of K^+^ that the accumulation of K^+^ in the t-system would induce. This suggested a role for Cl^−^ in stabilization of the resting potential during repetitive AP firing.
The role of Cl^−^ conductance on t-system K^+^ homeostasis and the membrane potential during AP trains
------------------------------------------------------------------------------------------------------
[Fig. 8](#fig8){ref-type="fig"} explores the possible role of Cl^−^ in simulations and experimental measurements of repetitive AP firing using solutions with reduced Cl^−^ concentration. The replacement of extracellular Cl^−^ with a monovalent membrane-impermeant anion produced a marked increase in the magnitude of the depolarization observed during repetitive AP firing in both the experimental recordings ([Fig. 8, A and B](#fig8){ref-type="fig"}) and the simulations ([Fig. 8 C](#fig8){ref-type="fig"}). In previous studies ([@bib30],[@bib31]), *G*~Cl~ was measured in EDL muscle fibers at the different Cl^−^ concentrations also used in the present study.
![The effect of extracellular Cl^−^ on the magnitude of the depolarization during AP trains. Two electrodes were inserted into the same muscle fiber, and 3.5-s-long 15-Hz trains of AP trigger pulses were injected. Experiments were conducted using extracellular solutions without Cl^−^ (*n* = 20) or with different Cl^−^ concentrations of 127 (control; *n* = 11), 80 (*n* = 23), or 50 mM (*n* = 14). (A) Representative recordings from different fibers at the four Cl^−^ concentrations. (B) The average change in resting membrane potential during the AP trains at the four Cl^−^ concentrations. Also presented is the G~Cl~ at the different conditions as reported by [@bib30] and the resulting G~K~/G~Cl~ ratios. (C) The experimental observations of AP trains at the different Cl^−^ concentrations were simulated, and the change in resting membrane potential was determined. (D) The K^+^ concentration in the deepest t-system compartment during the AP trains under the four conditions.](JGP_201110617_RGB_Fig8){#fig8}
Under control conditions G~Cl~ and G~K~ were 1,314 and 144 µS/cm^2^, respectively, giving a G~K~/G~Cl~ of 0.11. Reducing extracellular Cl^−^ to 80 mM increased the depolarization magnitude by around 50% and increased G~K~/G~Cl~ to 0.28. Further reduction in Cl^−^ to 50 mM increased G~K~/G~Cl~ to 0.66 and the depolarization by 143%. Entirely withdrawing Cl^−^ increased the depolarization magnitude by 543%, despite the necessary addition of a small dose of TTX (10^−8^ M) in the solutions to avoid myotonic activity.
These experiments suggest a highly nonlinear relationship of the depolarization during AP firing to G~K~/G~Cl~, in which moderate reductions in G~Cl~, similar to those during Phase I (G~K~/G~Cl~ of around 0.5), only moderately increase the depolarizations, whereas further G~Cl~ reductions markedly increase the depolarizations during AP firing, potentially giving rise to myotonic activity.
Simulations of AP trains with different Cl^−^ concentrations gave similar findings. Thus, [Fig. 8 C](#fig8){ref-type="fig"} shows that for control conditions and for 80 and 50 mM Cl^−^, the magnitude of depolarization during the simulated trains closely agreed with the experimental observations. The model further showed that the increased depolarization with reduced Cl^−^ was associated with minor increases in concentrations of t-system K^+^ during the AP trains ([Fig. 8 D](#fig8){ref-type="fig"}). In accordance with the experiments in Cl^−^-free conditions where TTX had to be included to prevent myotonic AP firing, the simulations in the absence of Cl^−^ resulted in myotonic activity. It was therefore necessary to reduce the maximum permeability of voltage-gated Na^+^ channels in the model to perform simulations without Cl^−^. [Fig. 8 C](#fig8){ref-type="fig"} shows that the depolarization during the simulated AP train in Cl^−^-free conditions caused a marked rise in the depolarization that, however, was not as marked as in the experimental recordings. In Cl^−^-free conditions, the rise in t-system K^+^ during the AP trains was similar to the rise at 50 mM Cl^−^.
*G~M~* regulation during repetitive AP firing has been shown to produce first a decrease in Cl^−^ conductance (Phase I), and then a large increase in K^+^ and Cl^−^ conductances (Phase II) ([@bib30],[@bib31]). The influence of these changes on the magnitude of baseline membrane potential depolarization and t-system ionic homeostasis during repetitive AP firing is demonstrated in [Fig. 9](#fig9){ref-type="fig"}. [Fig. 9 A](#fig9){ref-type="fig"} shows the first, 40th, and 80th trains of 49 APs at 15 Hz and, on a larger scale, the voltage response to a square test current pulse injected in between the AP trains. This demonstrates an increased voltage response immediately before the 40th AP train, illustrating the Phase I *G~M~* decrease, and a decreased voltage response before the 80th AP train, illustrating the Phase II *G~M~* increase. [Fig. 9 B](#fig9){ref-type="fig"} demonstrates that the Phase I *G~M~* decrease results in an approximately twofold increase in baseline depolarization, similar to the increase in baseline depolarization seen in low Cl^−^ solutions ([Fig. 8](#fig8){ref-type="fig"}). Yet, a smaller increase in baseline depolarization compared with initial conditions was also recorded during the Phase II increase in *G~M~*.
![Effects of *G~M~* regulation during Phase I and Phase II on the membrane potential depolarization and t-system K^+^ concentration during trains of APs. (A) Representative experimental recordings of the membrane potential from a muscle fibers that was repeatedly activated to fire 49 APs at 15 Hz every 7 s. Recordings show the first (left), the 40th (middle), and the 80th (right) AP train. To highlight the regulation of *G~M~* during AP firing, the membrane potential responses (underlined above) to the small current injections in between the AP trains have been highlighted below. As previously reported ([@bib30]), the onset of AP firing leads to a reduction in *G~M~*, as shown by an enlarged membrane potential deflection during the current injection (Phase I), whereas prolonged AP firing leads to a marked rise in *G~M~* (Phase II), as revealed by a reduced membrane potential deflection. To determine the effect of such *G~M~* regulation in active muscle fiber on the depolarization of the resting membrane potential during the AP trains, the membrane potential before AP firing was subtracted from the potential as recorded 55 ms after every AP. (B) The average change in resting membrane potential during the AP trains during control conditions (first AP train), during Phase I (40th train), and during Phase II (80th train). (C) *G~M~* during AP firing as originally reported by [@bib30]. (D) Changes in the composite conductances for K^+^ (G~K~) and Cl^−^ (G~Cl~) that underlie the changes in *G~M~* in C. (E) The depolarization of the resting membrane potential during the AP trains was determined in simulated AP trains in a similar way that this depolarization was assessed in the experimental recordings in B. (F) The K^+^ concentration in the deepest t-system compartment during the simulation of AP trains during the three conditions. (G) The magnitude of depolarization of the resting membrane potential during simulated 15-Hz trains (circles) and the corresponding t-system K^+^ in the deepest t-system shell (triangles) for a range of Cl^−^ membrane permeabilities. Typical values for Cl^−^ permeabilities for the three conditions have been indicated.](JGP_201110617_RGB_Fig9){#fig9}
The precise changes in *G~M~* ([Fig. 9 C](#fig9){ref-type="fig"}) and *G*~K~/*G*~Cl~ ratios (D) were obtained from [@bib30] to calibrate the model. Thus, [Fig. 9 E](#fig9){ref-type="fig"} demonstrates simulations showing a similar pattern of a greatly increased baseline depolarization during Phase I and a smaller increase in baseline depolarization during Phase II *G~M~* changes. However, [Fig. 9 F](#fig9){ref-type="fig"} demonstrates that although both Phase I and Phase II increase the baseline depolarization, these *G~M~* changes have opposite influences on t-system K^+^ accumulation. Thus, t-system K^+^ accumulation was increased in Phase I by ∼6 mM compared with 4.5 mM in control conditions, whereas it only was increased by 2 mM in Phase II. In addition, larger excursions in \[K^+^\]~t~ were seen during each AP in Phase II. Finally, [Fig. 9 G](#fig9){ref-type="fig"} shows the influence of membrane Cl^−^ permeability on baseline depolarization and t-system K^+^ accumulation. At low values of *P*~Cl~, K^+^ accumulation and baseline depolarization are large, as shown in [Fig. 8](#fig8){ref-type="fig"}. Between values of *P*~Cl~ of 10^−5^ cm s^−1^ and 10^−6^ cm s^−1^, there is a log-linear relationship between *P*~Cl~ and baseline depolarization, but a smaller influence of *P*~Cl~ on t-system K^+^ accumulation. Thus, in this range that includes the *P*~Cl~ seen in Phase I, the predominant effect of a decrease in *P*~Cl~ is to shift the membrane potential closer to *E*~K~. However, an increase in *P*~Cl~ above normal values, as seen in Phase II, produces a sharp reduction in t-system K^+^ accumulation and yet an increase in the magnitude of the baseline depolarization. This latter effect reflects depolarization of *E*~Cl~ because of an increase in intracellular \[Cl^−^\] as Cl^−^ currents, rather than K^+^ currents, provide the majority of repolarization after each AP. Note that further increases in *P*~Cl~ prevent t-system excitation.
DISCUSSION
==========
The t-system of skeletal muscle fibers both influences and is influenced by the passage of a sarcolemmal AP. It has a capacitance that is considerably greater than that of the sarcolemma itself ([@bib9]; [@bib17]), and therefore its charging could potentially slow the surface conduction velocity. It also possesses voltage-gated ion channels that underlie the t-system AP ([@bib4]; [@bib33]; [@bib28]), which is reflected in recordings of the surface AP as a prominent after-depolarization ([@bib16]). Such t-system regenerative electrical activity is thought to result in an accumulation of K^+^ in the t-system lumen during repetitive firing of APs ([@bib38]; [@bib35]).
It has been proposed that such K^+^ accumulation could induce loss of muscle excitability through depolarization and consequent inactivation of voltage-gated Na^+^ channels, thereby contributing to muscle fatigue ([@bib35]). Key to this theory on the etiology of muscle fatigue are observations of elevated extracellular K^+^ in plasma ([@bib26]) and in the interstitial space of contracting muscle as obtained using microdialysis probes ([@bib25]). However, for two reasons such measurements of extracellular K^+^ confer very little information on muscle excitability and their membrane potential per se. First, the membrane potential depends heavily on the actions of other ions, their permeabilities, and other transport systems. Second, it has been shown that several mechanisms are activated in the active muscle that can either reduce or enhance the force-depressing actions of activity-induced elevation in extracellular K^+^. Thus, muscle activity is associated with elevated Na^+^,K^+^ pump activity ([@bib27]) and changes in *G~M~* ([@bib30],[@bib31]) that might also markedly affect the excitability in working muscle. To deal with the emerging complexity of skeletal muscle excitability, it is desirable to develop mathematical models that will allow multiple parameter changes to be evaluated in combination.
In this study, a model was developed that builds on a realistic t-system geometry as determined using the analytic approach in our companion paper ([@bib32]) to evaluate the role of *G~M~* regulation upon excitability and AP conduction velocity in a system with nonlinear voltage- and time-dependent conductances, and to describe and quantify the interrelationships between *G~M~*, t-system ionic homeostasis and the membrane potential during repetitive AP firing.
The new multiple-compartment model was based on the CD model of [@bib10] and [@bib12],[@bib13]), previously used to investigate ionic and volume homeostasis in amphibian skeletal muscle. The model included background conductances, t-system resistances, and a longitudinal cable structure calibrated to rat skeletal muscle using the analytic model of [@bib32], and voltage- and time-dependent conductances to allow simulation of regenerative AP firing ([@bib38]). The ion concentrations within all compartments of the model were shown to possess unique history-independent steady-state values determined by the physical and ion channel parameters of the model. This allowed for the use of the model to predict ion concentrations within the t-system both at rest and during repetitive activity, as well as sarcolemmal and t-system AP conduction.
The model was first used to simulate the passive conduction properties of rat skeletal muscle in the absence of voltage-gated ion channel activity to extend the work of [@bib32] into a model system with physiological nonlinear background current--voltage relationships. This yielded similar results to those obtained from the model system with linear current--voltage relationships. Thus, the passive conduction velocity of a sinusoidal input current was greater for higher frequency signals, and this frequency--velocity relationship was greatly enhanced by the presence of a significant access resistance (*R*~A~) between the t-system and the extracellular space surrounding the fiber. Furthermore, the presence of additional series resistances within the t-system was shown to steepen this frequency--velocity relationship at higher frequencies (above ∼300 Hz).
The study then modeled the influence of active signal regeneration by voltage-gated ion channels upon surface conduction velocity and t-system excitation. This demonstrated that entrance of a propagating AP into a region lacking voltage-gated ion channels produced a sharp deceleration of conduction velocity over distance. Thus, the presence of AP regeneration by voltage-gated ion channels allowed actively propagated APs to show faster peak-to-peak conduction than did passively propagated AP waveforms, even over very short distances (100 µm or less), despite the clear necessity for passive currents to travel ahead of the AP waveform. Fourier transformation demonstrated that high-frequency components of passively conducted AP waveforms declined steeply over distance, thereby accounting for the reduced conduction velocity. In contrast, active regeneration by voltage-gated ion channels was shown to maintain the high-frequency components and thereby maintained a high conduction velocity. Thus, the analysis predicted a log-linear relationship between the maximum total Na^+^ permeability (*P*~Na(max)~) and AP velocity over a range of values of *P*~Na(max)~ from 0.3 to 10 times the normal physiological value. Fourier analysis then showed that this influence of *P*~Na(max)~ upon conduction velocity could be attributed to a greater content of high-frequency components in the AP and therefore in the local circuit currents that convey AP propagation.
Fast AP conduction is necessary but not sufficient to achieve synchronous contraction along the entire length of a muscle fiber. There is also a requirement for the surface AP to excite the entire t-system sufficiently to activate dihydropyridine receptors and thereby initiate excitation--contraction coupling ([@bib23]). The present study therefore explored the influence upon t-system excitation of the passage of a sarcolemmal AP for several different t-system models. It demonstrated that a model without a t-system luminal series resistance did not require voltage-gated ion channels within the t-system in order for a surface AP to produce sufficient t-system depolarization (to ∼0 mV) for excitation--contraction coupling to occur, despite a physiological *R*~A~. In contrast, a t-system simulated as 20 concentric shells with small physiological resistances between each shell demonstrated a much reduced response (to a peak of only approximately −40 mV in the fiber core) to the passage of a surface AP in the absence of voltage-gated ion channels within the t-system. However, the presence of voltage-gated ion channels within the t-system allowed for the propagation of a full-sized AP through the t-system, producing a peak response of approximately +40 mV in the core of the fiber. The presence or otherwise of voltage-gated Na^+^ channels within the t-system had no influence on the conduction velocity of the sarcolemmal AP. Thus, it appears clear that fast AP propagation requires a significant *R*~A~, but a high *R*~A~ reduces the passive t-system excitation produced by a surface AP, such that regeneration by voltage-gated ion channels within the t-system is required.
Our companion study ([@bib32]) additionally demonstrated that *R*~A~ has an important influence on the relationship between *G~M~* and the response of the sarcolemma and the t-system membrane to subthreshold signals. Thus, it predicted that fiber excitation at the neuromuscular junction and t-system excitation are particularly sensitive to regulation of *G~M~*, compared with sarcolemmal propagation. This prediction arose from the property that sarcolemmal AP propagation is conveyed by high-frequency elements in the circuit currents, and the membrane impedance at such frequencies is largely insensitive to *G~M~*. Despite the advantages of the analytical solution in generating the detailed mechanistic insight into the relationship between *G~M~* and excitability, the approach was limited to a linear analysis of subthreshold responses. It was therefore useful to extend the predictions generated in our companion paper using a model system simulating regenerative AP firing.
During repetitive AP firing, there is first an ∼60% reduction in *G~M~* (Phase I), and then after prolonged activity, there is a four- to fivefold increase in *G~M~* (Phase II), in comparison to control conditions ([@bib30],[@bib31]). The simulations demonstrated that, for 1-ms square current pulses simulating an endplate potential, Phase I produced a minor decrease in the current required to excite an AP, whereas Phase II required a near doubling of this current. This suggests that the sarcolemmal AP excitation is somewhat facilitated by Phase I and markedly reduced during Phase II *G~M~* regulation, even though currents of short duration, comparable to the endplate current, must predominantly flow across the membrane capacitance. Phase II *G~M~* changes were also shown to produce a reduction in the upstroke velocity of the surface AP sufficient to reduce conduction velocity by ∼18%, and a reduction in the magnitude of the t-system AP. Additional reduction of *P*~Na(max)~---simulating activity induced slow inactivation of the voltage-gated Na^+^ channel---produced a failure of t-system AP generation during Phase II *G~M~* changes, despite persistence of the surface AP. This functional disconnection between sarcolemma and t-system AP generation illustrates that EMG recordings might miss a failure of excitability under conditions of high *G~M~* and low *P*~Na(max)~, as might occur after prolonged muscle activity.
Regeneration of the t-system AP by voltage-gated ion channels was shown in the model to take place at the cost of producing a rapid accumulation of K^+^ ions within the t-system during trains of APs. Thus, the K^+^ concentration within the deepest parts of the t-system \[K^+^\]~t(19)~ was shown to more than double within 2 s of 30-Hz stimulation. No such K^+^ accumulation was seen in fibers lacking voltage-gated Na^+^ channels in the t-system, despite the continued presence of voltage-gated K^+^ channels. Direct experimental measurement of ion concentrations within the t-system is difficult or even impossible during periods of dynamic change during electrical activity. However, the increased \[K^+^\]~t~ during repetitive AP firing was shown to produce a stable depolarization of the resting sarcolemmal potential, allowing for comparison of model findings with experimental microelectrode measurements of the membrane potential.
Thus, membrane potential measurements were recorded experimentally from rat EDL muscle fibers exposed to myosin II inhibitor BTS ([@bib24]). The experimental results were in close quantitative agreement with the model simulations, with each showing a baseline resting potential depolarization of ∼1, 2, and 3 mV during 6-, 15-, and 30-Hz AP trains, respectively, and an associated decrease in maximum AP height. The close quantitative agreement between model and experiments on the depolarization during AP trains permitted the use of the model to investigate the underlying alterations in t-system K^+^ homeostasis. This revealed that the dynamics of t-system K^+^ accumulation during AP trains and recovery thereafter are very rapid, completing within 1--2 s. This means that measurements of interstitial K^+^ using microdialysis probes that offer temporal resolutions in minutes might offer rather little information on the K^+^ dynamics in the working muscle ([@bib25]).
Combining experiments and simulations further demonstrated a significant influence of membrane Cl^−^ permeability upon t-system K^+^ accumulation, membrane potential depolarization, and the relationship between these two variables. Reduction of extracellular \[Cl^−^\] by replacement with an impermeable anion greatly increased the magnitude of the depolarization during AP trains in both simulations and parallel experiments. The *G*~K~/*G*~Cl~ ratio has been measured previously under identical experimental conditions ([@bib30]). Thus, reductions in \[Cl^−^\]~e~ increased depolarization in both the model and experimental records by ∼50% for 80 mM (*G*~K~/*G*~Cl~ = 0.28) and \>100% for 50 mM (*G*~K~/*G*~Cl~ = 0.66) when compared with that in solutions of 127 mM (*G*~K~/*G*~Cl~ = 0.11). Both experimental and model muscle fibers showed myotonic behavior in zero-\[Cl^−^\] solutions; therefore, a small dose of TTX (10^−8^ M) was added in the experiments, and a 50% reduction in *P*~Na(max)~ was applied to the model. Despite this manipulation, Cl^−^-free solutions increased the depolarization by 410% in the model and 540% in the experimental studies. This close agreement between the membrane potential records of the model and the experiments prompted detailed analysis of the influence of Cl^−^ reduction upon t-system K^+^ homeostasis using the model. Reduced \[Cl^−^\] was associated with only minor increases in the concentration of t-system K^+^ during the AP trains. Thus, under normal conditions, Cl^−^ influences the resting membrane potential during AP trains by maintaining a high *G*~Cl~/*G*~K~, such that the influence of t-system K^+^ accumulation on the surface membrane potential is small.
Finally, modeling and experimental membrane potential measurements were used to demonstrate and quantify the influence of the physiological Phase I and Phase II changes in *G*~Cl~ and *G*~K~ upon membrane depolarization and t-system K^+^ accumulation that occur during repetitive AP firing. The Phase I reduction in *G*~Cl~ was shown to approximately double the magnitude of the resting potential depolarization during 15-Hz stimulation in both model and experiment, similar to the influence of reduced extracellular \[Cl^−^\]. The model also indicated an increase in t-system K^+^ accumulation from ∼10 to ∼11.5 mM.
The Phase II increase in *G*~K~ and *G*~Cl~ produced a smaller increase in baseline depolarization during 15-Hz AP trains, although the model slightly underestimated the magnitude of the experimentally observed increase. The model showed a large reduction in t-system K^+^ accumulation during Phase II, to a peak of ∼7.5 mM compared with ∼10 mM under control conditions. Thus, during Phase II, depolarization is increased despite a smaller rise in K^+^. Indeed, a biphasic relationship was found for the influence of *G*~Cl~ values on the membrane potential during AP firing, such that both decreases and increases in *G*~Cl~ in the physiological range increase the resting potential depolarization during AP trains. The effect of a reduction in *G*~Cl~ was shown to result from a shift of the membrane potential closer to *E*~K~, which is depolarized during activity because of increased \[K^+^\]~t~. In contrast, an increase in *G*~Cl~ was shown to produce a large contribution of Cl^−^ currents to AP repolarization, thereby causing intracellular Cl^−^ accumulation and hence a depolarization of *E*~Cl~. Phase II also increased the rate of recovery to normal values of \[K^+^\]~t~ after activity. However, the presence of high background ion conductances during AP firing in Phase II produced increases in both inward and outward K^+^ movements. Thus, although Phase II *G~M~* changes appear to accelerate reuptake of K^+^ from the t-system into the cell and thereby enhance recovery of normal ionic concentrations, they also appear to reduce the energy efficiency of APs by increasing the total ionic fluxes required for each AP.
This study therefore provides a quantitative analysis, supported by experimental results, of the relationships between resting conductances, excitability, and t-system ionic homeostasis in skeletal muscle. It demonstrates that active signal regeneration within the t-system produces K^+^ accumulation and quantifies the relationship between this and the depolarization of the resting membrane potential that is observed during trains of APs. It shows that the Phase I ∼60% decrease in *G~M~* during repetitive AP firing enhances endplate and t-system excitability, while also increasing the magnitude of t-system K^+^ accumulation and approximately doubling the membrane potential depolarization during AP trains. In contrast, the Phase II increase in *G~M~* decreases endplate and t-system excitability and also enhances K^+^ reuptake from the t-system, thereby reducing K^+^ accumulation during AP trains. Yet, Phase II also increases the resting potential depolarization during AP trains because of intracellular Cl^−^ accumulation.
The model that is described here therefore reproduces predictions of electrophysiological behavior and ion concentrations, at rest and during activity, within both the sarcoplasm and the t-system, that are in full agreement with experimental and prior theoretical results. In so doing, it reveals the mechanisms of observed changes including baseline resting potential depolarization during AP trains. Furthermore, it provides a platform for further exploration of electrophysiological activity in skeletal muscle that is applicable where steady-state ion concentrations are not known or expected to undergo significant change. This, in contrast to existing models, permits analysis of the influence of activity-related, genetic, and pharmacological manipulations to the normal physiological state.
J.A. Fraser is supported by a David Phillips Fellowship from the Biotechnology and Biological Sciences Research Council (UK). C.L.-H. Huang acknowledges the support of the British Medical Research Council, the Wellcome Trust, and the British Heart Foundation. T.H. Pedersen acknowledges the support of the Danish Medical Research Council.
Christopher Miller served as editor.
Abbreviations used in this paper:APaction potentialCDcharge--differenceECFextracellular fluidEDLextensor digitorum longus*G*~M~resting membrane conductance
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Kidney cancer is a common malignant illness [@B1]. Around 90% pathological type of renal cancers is renal cell carcinoma (RCC), majority of which are subtyped as clear cell renal cell carcinoma (ccRCC) [@B2]. Based on the size and metastasis of tumor, the degree of invasion external of the kidney and the involvement of lymph node, ccRCC is classified into pathological T stages, pathological N stage, metastasis and clinical stage [@B3]. It is well known that the prognosis of this disease is correlated to the pathological stage and the five-year survival rates of these four pathological stages are 95%, 88%, 59% and 20% for the aforementioned stages, respectively [@B4]. Indeed, localized ccRCC can be cured with radical nephrectomy but the prognosis is poor when the disease turned to be metastatic. In case of the late staged ccRCC, traditional chemotherapies are usually tolerant. In recent decades, different oncogenes related to ccRCC had been identified with high-throughput microarray technology [@B5]-[@B7]. Treatments targeted on these discovered genes have been proved more effective than chemotherapies, duo to the target specificity and low adverse effect [@B8]. A number of targeted therapies have been accepted for clinical use, such as anti-vascular endothelial growth factor (VEGF) antibodies, mammalian target of rapamycin (mTOR) and multi-kinase inhibitors [@B9]. Although patients\' survival have been ameliorated with these new treatments, median overall survival and progression-free are virtually 2 years and most cases finally become surrender and resistance [@B8]. Ignorance of interconnection between genes could contribute to the failure of these targeted therapies, as carcinogenesis progression is not only the consequence of deregulation of some tumor suppressors or oncogenes but also the result of complex molecular mechanisms, including the strong interconnection between genes with similar expression patterns. Therefore, to achieve effective individualized treatments for ccRCC, more therapeutic targets should be identified and their interconnection should be determined.
Langfelder et al. firstly used weighted gene co-expression network analysis (WGCNA) to explore a thorough association between different gene sets or between gene sets and clinical characteristics [@B10]. With emerging plenty of microarray or RNA sequencing data, WGCNA has been widely performed to filter modules and hub genes that are correlated to clinical features like grade, metastasis and tumor stages among various tumor types such as hepatocellular carcinoma [@B11] and papillary renal cell carcinoma [@B12]. With regarding to ccRCC, our center analyzed a microarray data with WGCNA and discovered six hub genes (*CCNB2*,*CDC20*,*CEP55*,*TOP2A*,*KIF20A* and*UBE2C*) that were highly correlated with pathologic stage of ccRCC [@B13]. Also in our center, Chen et al. identified a hub gene *FCER1G* through co-expression network analysis of another microarray data and demonstrated this hub gene had connection with progression and prognosis of ccRCC via influencing immune-related pathways [@B14]. In current study, we downloaded a different microarray dataset and tried to build a co-expression network with a systematical biology process of WGCNA. Furthermore, ccRCC and adjacent normal kidney tissues wer harvested to verify the bioinformatic analysis. We aimed to seek and validate other different hub genes which are associated with clinical stages and survival of ccRCC [@B15]-[@B17].
Materials and Methods
=====================
Data collection
---------------
GSE36895 microarray dataset, containing 29 homo ccRCC tissues and 23 homo normal kidney tissues, was downloaded from Gene Expression Omnibus (GEO) database (<http://www.ncbi.nlm.nih.gov/geo/>) for constructing co-expression networks and exploring hub genes. Patient\'s clinical information of ccRCC tissues included age, gender, different grades (I \-- Ⅳ), pathological T stages (I \-- Ⅳ), pathological N stages (I \-- Ⅲ), metastasis (M0 and M1) and clinical stages (I \-- Ⅳ). We also downloaded RNA-sequencing dataset with detailed clinical information from The Cancer Genome Atlas (TCGA) database (<https://genome-cancer.ucsc.edu/>) to validate the gene expression based on the RNA-sequencing technology of IlluminaHiseq.
Data preconditioning
--------------------
The raw data were background corrected, log2 transformed and quantile normalized by Robust Multi-array Averaging (RMA). The \"Affy\" R package was used to summarize median polish probesets which were annotated with the files of Affymetrix annotation. Finally, sample clustering was applied to evaluate the quality of GSE36895 dataset.
Differential expression genes (DEGs) screening
----------------------------------------------
DEGs between ccRCC and normal renal tissues were screened using R software based on \"limma\" R package at a preset threshold with \|log2 fold change (FC)\| \> 1 and p value \< 0.05.
Co-expression network construction
----------------------------------
After verifying the qualification of DEGs\' expression data, a co-expression network was set for the DEGs using R software based on the \"WGCNA\" R package. Pearson\'s correlation matrices were conducted and a weighted adjacency matrix were performed by a formula amn = \|cmn\|^β^ (cmn represents Pearson\'s correlation between genes, amn represents adjacency between genes and the soft-thresholding parameter (β) was able to magnify the correlation between genes through enhancing high correlations and weakening low correlations). In current study, β = 6 was chosen to guarantee a scale-free network. Subsequently, the adjacency was transformed into topological overlap matrix (TOM) and identified modules including similar genes by hierarchically clustering genes [@B18]. To categorize genes with analogous expression into gene modules, an average linkage hierarchical clustering was carried out based on TOM dissimilarity measure with a minimal gene size of 30 for constructing a dendrogram [@B19]. Finally, a cut-line was selected for module dendrogram and merged some modules after dissimilarity of estimated module eigengenes being evaluated.
Discovering the interesting module
----------------------------------
Module eigengenes (MEs) were considered as the most principal component and all genes were summarized into a single characteristic expression profile. The interesting module was identified by calculating the relevance between MEs and clinical feature. The log10 transformation of the p value was defined as gene significance (GS) and the average GS for all genes in the module was defined as the module significance (MS). The module with the highest MS score was chosen as the one related to clinical feature.
In order to investigate the possible mechanism of the association between the interesting module genes and correlated clinical characters, all genes in brown module were uploaded into the DAVID database and analyzed by GO functional enrichment analysis with a cutoff criterion of false discovery rate (FDR) \< 0.01.
Identification and validation of hub genes
------------------------------------------
For interesting module, the hub genes were defined based on module connectivity (Pearson\'s correlation of module membership \> 0.8) and clinical characteristic relation (Pearson\'s correlation of GS \> 0.2). Moreover, protein-protein interaction (PPI) network was built through putting all relevant genes from the module into the Search Tool for Interacting Genes\' Retrieval (STRING). The common hub genes in both co-expression network and PPI network were regarded as "real" hub genes for further analyses.
Efficacy evaluation and survival analysis
-----------------------------------------
TCGA data were utilized to evaluate the association between the expression of the most interesting hub genes and the different pathological stages of ccRCC using Gene Expression Profiling Interactive Analysis (GEPIA) database (<http://www.gepia.cancer-pku.cn>). The survival rate analysis was conducted based on the TCGA database for the assessment of the identified genes\' effects on the prognosis of ccRCC patients. Firstly, patients with mRNA data were classified in two different categories in accordance with each gene\'s median expressions (low vs. high). Patients with methylation data were similarly analyzed. Secondly, analysis was conducted on patients with both mRNA expression and different ccRCC grades data. Finally, we performed Kaplan-Meier survival analysis and the log-rank test by adopting the "survival" R package. One-way analysis of variance (ANOVA) and paired 2-tailed Student\'s t tests were used to analyze the statistical significance of differences of data.
Gene set enrichment analysis (GSEA)
-----------------------------------
Two categories (high vs. low) of the most interesting hub genes in 539 ccRCC patients were classified and the median value of gene expression was applied as the cut-off point. GSEA (<http://software.broadinstitute.org/gsea/index.jsp>) was carried out to investigate potential functions of the most interesting hub genes with a cut-off criteria of \|Enrichment score (ES) \| \> 0.5 and p value \< 0.05.
Human ccRCC and adjacent normal kidney tissues
----------------------------------------------
ccRCC and adjacent normal kidney tissues (n = 15) were obtained from patients undergoing laparoscopic nephrectomy at Zhongnan Hospital of Wuhan University. Two pathologists independently confirmed the histological diagnosis. Half of each specimen was immediately fixed in 4% PFA (paraformaldehyde) and half stored in liquid nitrogen. The use of these ccRCC specimens was approved by the Ethics Committee at Zhongnan Hospital of Wuhan University, and informed consent was obtained from all patients.
Total RNA extraction and real-time RT-PCR
-----------------------------------------
Total RNA was isolated from the frozen tissues using Takara RNAiso Plus (Takara Bio. Inc., Otsu, Shiga, Japan) according to the manufacturer\'s protocol. Genomic DNA (gDNA) was removed and cDNA was reverse-transcribed using Takara PrimeScriptTM RT reagent Kit with gDNA Eraser (Takara Bio. Inc., Otsu, Shiga, Japan) in a T100TM Thermal Cycler System (BioRad, USA). The experimental protocol utilized was first gDNA removal (42 °C, 2 min), followed by reverse transcription (37 °C 15 min, 85 °C 5 s). Subsequently, all samples were amplified by a 25 μl reaction volume in a CFX96TM Real-time PCR Detection System (BioRad, USA), using SYBR® Premix Ex TaqTM Ⅱ (Takara Bio. Inc., Otsu, Shiga, Japan). All samples were run in triplicate. The seven identified genes were investigated. The amplification program was repeated for 40 cycles. For relative quantification, gene expression was normalized to expression of GAPDH housekeeping gene and compared by 2^-ΔΔCT^ method.
Immunohistochemistry
--------------------
For immunohistochemistry, sections were deparaffinized in xylene, followed by graded alcohols. Antigen retrieval was performed in 10 mM sodium citrate buffer (pH 6.0) and heated to boil. Sections were kept in boiled buffer for 2 min. Endogenous peroxidase activity was blocked by using 3% H2O2 solution at room temperature for 10 min. Then sections were incubated with 15% normal goat serum for 15 min at room temperature to block non-specific binding. Primary antibody was applied to the sections on the slides and incubated in a humidified chamber at 4 ◦C overnight. Then the sections were stained by routine immunohistochemistry methods.
Results
=======
Different expression of genes screened
--------------------------------------
After quality evaluation and data preprocessing, the expression matrix was acquired from the 52 samples of GSE36895 dataset (Figure [1](#F1){ref-type="fig"}). With the \|log2FC\| \> 1and p value \< 0.05, a sum of 1624 DEGs (886 down-regulated and 738 up-regulated) were selected for subsequent analyses.
Weighted co-expression network construction and key modules identification
--------------------------------------------------------------------------
Twenty-nine ccRCC samples with clinical information were included for the co-expression analysis with β = 6 used as the soft-thresholding to guarantee a free scale network (scale free R^2^ = 0.85) (Figures [2](#F2){ref-type="fig"}A - D). A sum of 6 different modules were identified (Figure [2](#F2){ref-type="fig"}E). The highest absolute MS score in the Module-feature relationship was found between pathological T stage and brown module (r = - 0.45, p = 0.01; Figure [2](#F2){ref-type="fig"}F), which was chosen for the subsequent analyses. Interestingly, the brown module was also found to be associated with pathological lymph nodes stage (r = - 0.40, p = 0.03) and tumor grade (r = - 0.45, p = 0.01).
A total of 565 genes associated biological relevance in brown module were investigated utilizing DAVID database for GO (Gene Ontology) analyses. As shown in Figure [3](#F3){ref-type="fig"}A, 36 enriched biological procedures were found to be possible mechanisms of how the brown module genes impact on pathological T stage, such as metabolic process (p = 9.63E - 09), oxidation-reduction process (p = 1.05E - 08), oxidoreductase activity (p = 1.72E - 04) and fatty acid beta-oxidation (p = 1.45E - 06).
GSEA of hub genes
-----------------
To explore the underlying roles of these three hub genes, we conducted GSEA to map into KEGG (Kyoto Encyclopedia of Genomes and Genes) pathways database. According to the cut-off criteria with \|ES\| \> 0.5 and p value \< 0.05, a sum of 20 significant gene sets were found and majority of which focused on metabolic relevant pathways. Six representative pathways were "Biosynthesis of unsaturated fatty acids", "Butanoate metabolism", "Peroxisome", "PPAR signaling pathway", "Propanoate metabolism" and "Valine leucine and isoleucine degradation" (Figure [3](#F3){ref-type="fig"}B).
Identification of hub gene
--------------------------
In current study, 18 high connective genes in brown module were selected as hub genes (Table [1](#T1){ref-type="table"}). Moreover, we conducted a PPI network for all genes in brown module using Cytoscape software and the genes associated with more than 7 nodes were regarded as hub node genes (Figure [3](#F3){ref-type="fig"}C). The three common hub genes (*EHHADH*, *ACADM* and *AGXT2*), met both criterions in co-expression and PPI networks, were selected as \"true\" hub genes for further validation (Table [1](#T1){ref-type="table"}).
Hub gene validation
-------------------
The complete data of 539 ccRCC patients in the TCGA dataset were carried out to validate hub genes, demonstrating that all three hub genes exhibited a substantial negative correlation with different ccRCC stages, consistent with above analyses of GSE36895 microarray dataset (Figures [4](#F4){ref-type="fig"}A, C, E). In addition, based on TCGA data, significantly longer overall survival times were shown in patients with higher expression of these three hub genes, indicating that *EHHADH*, *ACADM* and *AGXT2* were prognostic biomarkers for ccRCC (Figures [4](#F4){ref-type="fig"}B, D, F). Furthermore, as shown in Figure [5](#F5){ref-type="fig"}, both protein and mRNA expression of these hub genes were significantly lower in ccRCC tissues compared to normal ones, which were provided and confirmed by The Human Protein Atlas and Oncomine databases. To assess the roles of these three hub genes in ccRCC, gene expression validations were performed, and all of three hub genes were also downregulated in the TCGA database (Figure [6](#F6){ref-type="fig"}A, C, E) and longer overall survival duration was also found in cases of lower expression at each tumor grade. (Figure [6](#F6){ref-type="fig"}B, D, F).
Association of three hub genes methylation level with the prognosis of ccRCC
----------------------------------------------------------------------------
The methylations of the three hub genes identified above were further analyzed with TCGA database. As shown in Figure [7](#F7){ref-type="fig"}(A, C and E), *EHHADH* and *ACADM*) were found hyper-methylated while *AGXT2* hypo-methylated in tumor tissues. The survival curves were drawn to evaluate the association between three hub genes methylation levels and the prognosis of ccRCC, respectively. The two hyper-methylated genes (*EHHADH* and *ACADM*) were associated a shorter OS duration, but the hypomethylated one (*AGXT2*) also showed a shorter OS (Figure [7](#F7){ref-type="fig"}B, D, F).
Expression of the identified hub gene in ccRCC and normal tissues obtained in our institute
-------------------------------------------------------------------------------------------
Expression of *EHHADH*,*ACADM* and *AGXT2* mRNA were determined using quantitative real time RT-PCR and immunohistochemistry between ccRCC samples and normal ones. *EHHADH* expression was most significantly downregulated at the transcription level (p \< 0.0001) in ccRCC. Real time RT-PCR also showed that for the other two genes that *ACADM* (p = 0.0225) and *AGXT2* (p = 0.0019) the mRNA expression was significantly altered in the ccRCC samples (Figure [8](#F8){ref-type="fig"}A, B, C). As shown in Figure [8](#F8){ref-type="fig"} (D - I), in normal tissue, all three genes were mainly present in in renal tubules and partly present in the glomeruli. In ccRCC, the localization is similarly with that in normal ones, the immune positivity is significantly lower than that of normal. That is to say, immunohistochemistry also confirms that the expression of the three genes *EHHADH*,*ACADM* and*AGXT2* in ccRCC is lower than that in normal tissues.
Discussion
==========
Current study identified three novel hub genes (*EHHADH*, *ACADM* and *AGXT2*) through analyzing the co-expression and PPI networks. Our data further demonstrated these hub genes showed a negative relationship with different ccRCC stages, which having impact on overall survival. Also, the methylation of these hub genes was found negatively corelated with survival. In addition, it was found these three hub genes might play important roles in ccRCC prognosis through metabolic related pathways.
GSE36895 data were used. There are 29 cases of ccRCC and 23 normal kidney tissues involved 1624 DEGs (886 down-regulated and 738 up-regulated). These genes were further analyzed with WGCNA through constructing a gene co-expression network based on the expression similarity among samples. WGCNA have been used to discover complex disease related genes, biological pathways and neoplasm treatment targets of Alzheimer\'s disease, osteoporosis and hyperlipidemia. Also, in our center, Yuan et al. analyzed GSE40435 and discovered six hub gene that were highly correlated with pathologic stage of ccRCC [@B13]. Chen et al. identified a hub gene *FCER1G* which might regulate immune-related pathways to tumorigenesis through co-expression network analysis of another microarray data GSE66272 [@B14]. In present study, a larger ccRCC sample size was analyzed with WGCNA and 18 hub genes were screened from brown module. Three common hub genes (*EHHADH*, *ACADM* and *AGXT2*), met both analyses of co-expression and PPI networks, were regarded as "real" hub genes and were further validated. Our data indicated the three hub genes had high connection with clinical prognosis as well as vital biological processes.
In the GEPIA database, we found a trend that the expression of *EHHADH*, *ACADM* and *AGXT2* was negatively correlated with the pathological stages of ccRCC (Figures [4](#F4){ref-type="fig"}A, C, E), which illustrated the critical role of these three hub genses in the progression of ccRCC. In addition, these hub genes were found negatively corelated with survival at tumor grade (Figure [6](#F6){ref-type="fig"}). Indeed, the Oncomine database found a significant lower expression of *EHHADH*, *ACADM* and *AGXT2* in ccRCC tissues than normal kidney tissues (Figures [5](#F5){ref-type="fig"}A, C, E). In addition, immunohistochemistry staining in The Human Protein Atlas database showed that the expression of *EHHADH*, *ACADM* and *AGXT2* proteins were also significantly lower in renal carcinoma compared to normal kidney (Figures [5](#F5){ref-type="fig"}B, D, F). Consistently, the lower expression of the 3 hub genes was determined both at the mRNA and protein levels with using tissues harvested from our institute (Figure [8](#F8){ref-type="fig"}). Thus, our study indicated a negatively role of these three hub genes in clinical stages of ccRCC development. Moreover, we performed survival analysis to validate if these three hub genes were associated with patient prognosis (Figures [4](#F4){ref-type="fig"}B, D, F). According to the GEPIA database, we found that a lower expression of *EHHADH*, *ACADM* and *AGXT2*, a shorter overall survival time. Also, the hyper-methylation of these hub genes was found negatively associated with survival (Figure [7](#F7){ref-type="fig"}). However, the hypo-methylated gene*AGXT2* also showed a shorter OS, which will need further investigation In line with previous report, hyper-methylated and lower expressed genes showed worse prognosis. Therefore, *EHHADH*, *ACADM* and *AGXT2* could be suggested as protective tumor suppressors for ccRCC.
*EHHADH* (3-hydroxyacyl CoA dehydrogenase and enoyl-CoA hydratase) encodes a bifunctional enzyme that is one of the four enzymes of peroxisomal β-oxidation pathway [@B20]. Suto K et al. and Cablé S et al. found that *EHHADH* was lower expressed in hepatocellular carcinoma and colon carcinoma and could be used as a potential prognostic marker [@B21], [@B22]. *ACADM* (medium-chain acyl-CoA dehydrogenase) could catalyze the first dehydrogenation in fatty acyl-CoA beta-oxidation in mitochondria [@B23] and *ACADM* insufficiency might impact on the medium-chain fatty acids which exist abundantly in the beta-oxidation pathway that would indirectly influence the triglycerides metabolism [@B24], [@B25] and play a significant role in cell apoptosis through the function of light chain. *AGXT2* (alanine-glyoxylate aminotransferase 2), a multifunctional mitochondrial aminotransferase, has diverse functions in cellular physiology and its products and substrates are biomarkers of renal, cardiovascular and metabolic diseases [@B26]. Therefore, it was assumed these three hub genes played functional roles in suppressing cancer mainly through metabolic related pathways. Indeed, current study carried out GSEA using KEGG pathways database and found that majority of gene sets involved in metabolic related pathways. Consistently, TCGA Research Network illustrated that epigenetic reprogramming and oncogenic metabolism are the fundamental features of ccRCC. It is known that renal cancer is considered as one of the most deliberated and exemplary of malignancies characterized through metabolic reprogramming [@B27], [@B28]. And genes mutated in renal cancer are complicated in a quantity of disparate pathways regulating various aspects of cellular metabolism, such as iron sensing and/or oxygen, the tricarboxylic acid (TCA) cycle, tumor energetics and glutamine metabolism [@B20], [@B29], [@B30]. When translated to clinical scenarios, these three hub genes would have clinical values for diagnosis and personalized therapy for ccRCC.
Several limitations ought to be noted. In present study, some risk factors like gender, age, tumor grade, metastasis, and pathological stages were analyzed in patients suffered from ccRCC. However, other major established risk factors for ccRCC, such as hypertension and cigarette smoking, were not displayed for analysis during data collection. Additionally, more high quality ccRCC samples are needed to confirm our findings and elucidate the deep possible mechanisms of the effect on pathological stages.
In conclusion, our study identified and validated three novel hub genes including *EHHADH*, *ACADM* and *AGXT2*. Moreover, these three hub genes were found negatively correlated with clinical stages and having impact on patients\' survival. Our novel data suggests these abnormally expressed or methylated genes could be used as therapeutic targets and biomarkers for ccRCC patients to be precisely diagnosed and effectively treated.
The excellent technical assistance of Danni Shan and Shanshan Zhang is gratefully acknowledged. We also would like to acknowledge the KEGG database developed by Kanehisa Laboratories. This study was supported in part by National Natural Science Foundation of China (N.81160086, N.81270843 and N.81770757). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Availability of data and materials
==================================
The present study followed the publication guidelines of Gene Expression Omnibus (GEO) (<https://www.ncbi.nlm.nih.gov/geo/info/disclaimer.html>) and TCGA Research Network (<https://cancergenome.nih.gov/abouttcga/policies/ethicslawpolicy>). Thus, no further ethical approvals were required.
{#F1}
{#F2}
{#F3}
{#F4}
{#F5}
{#F6}
{ref-type="fig"}A, C and E. The Kaplan-Meier survival curve was plotted. It revealed that the overall rates of survival for the patients with 2 hyper-methylated genes (*EHHADH* and *ACADM*) and 1 hypo-methylated gene *AGXT2* were significantly lower, which were shown in Figure [7](#F7){ref-type="fig"}B, D and F.](jcav10p6779g007){#F7}
{#F8}
######
Hub genes in the module related with pathological stage.
Gene Probe Co-expression analysis (cor.geneModuleMembership) Hub gene in PPI network DEG analysis
---------- -------------- --------------------------------------------------- ------------------------- -------------- ----------
EHHADH 205222_at 0.88 YES -1.14 3.06E-04
ACADM 202502_at 0.89 YES -1.10 8.19E-07
AGXT2 229229_at 0.86 YES -2.37 5.67E-07
BBOX1 243018_at 0.93 NO -2.24 1.35E-03
HAO2 220801_s\_at 0.93 NO -3.15 1.82E-10
ANK3 221751_at 0.83 NO -1.26 1.18E-05
GOT1 1553878_at 0.87 NO -1.18 1.24E-06
PHYH 203335_at 0.81 NO -1.28 4.37E-10
SLC27A2 205768_s\_at 0.82 NO -2.49 2.19E-05
SERPINA5 209443_at -0.81 NO -4.00 2.12E-21
DDC 214347_s\_at 0.83 NO -2.45 5.30E-07
HIBCH 203711_s\_at 0.88 NO -1.47 6.50E-07
FBP1 205014_at 0.84 NO -2.80 1.62E-11
HMGCS2 240110_at 0.80 NO -1.87 1.54E-04
DMGDH 231591_at 0.84 NO -1.51 5.66E-04
LRP2 205710_at 0.89 NO -1.66 1.44E-03
CUBN 206775_at 0.80 NO -1.85 9.03E-05
ECHS1 201135_at 0.81 NO -1.45 1.34E-14
[^1]: **^\*^**These authors have contributed equally to this work.
[^2]: Competing Interests: The authors have declared that no competing interest exists.
| {
"pile_set_name": "PubMed Central"
} |
**A**SUM would like to draw your attention to recent research conducted by the University of London and the Katholieke Universiteit Leuven, Belgium, published in the November 2011 issue of *Ultrasound in Obstetrics and Gynaecology*.
The studies suggest that given inter‐observer variability in ultrasound measurements and the significant variation in early embryonic growth, a more conservative approach to the diagnosis of early pregnancy loss is warranted.
The recommendation which has been temporarily endorsed by the RCOG suggests a mean sac diameter (MSD) cut off \> 25 mm and a crown rump length (CRL) cut off\> 7 mm be introduced to minimise the risk of a false positive diagnosis of miscarriage.
While this research awaits confirmation from other centres ASUM suggests interim caution and highlights the importance of transvaginal confirmation of early pregnancy failure.
It should also be noted that many other factors are used when assessing early pregnancy failure, including the presence of a yolk sac, shape of the gestation sac, position within the uterine cavity or cervix, progress from a previous scan and correlation with known gestational age especially in IVF pregnancies.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-nutrients-10-00274}
===============
Humans are holobionts, a complex ecosystem of host-derived cells together with transient and stable microbial symbionts \[[@B1-nutrients-10-00274]\]. A typical human body contains up to 100 trillion microorganisms, equivalent to \~10 times the total number of nucleated cells in the body \[[@B2-nutrients-10-00274],[@B3-nutrients-10-00274],[@B4-nutrients-10-00274],[@B5-nutrients-10-00274]\]. The large intestine is the greatest single human reservoir of microbes, containing at least 30 identified genera and as many as 500 different species \[[@B2-nutrients-10-00274],[@B6-nutrients-10-00274],[@B7-nutrients-10-00274]\]. The inter-relationships that occur within this ecosystem are complex and affect the development and health of the individual \[[@B8-nutrients-10-00274]\].
The association between the development of the gut microbiota and the host's genotype and phenotype has received increasing attention as technological advances in culture-independent techniques (e.g., genomic, transcriptomic, proteomic, and metabolomic) have facilitated the detection of a greater diversity of microbes \[[@B9-nutrients-10-00274]\]. These studies have demonstrated that the composition of the gut microbiota in each infant is idiosyncratic with significant inter-individual variation being evident from the first day after birth \[[@B10-nutrients-10-00274],[@B11-nutrients-10-00274],[@B12-nutrients-10-00274]\]. Individuality, time from birth and mode of feeding were the strongest contributors to variation in the microbiota in 8 infants sampled seventeen times over the first 12 weeks from birth \[[@B13-nutrients-10-00274]\]. Notably, the impact of individuality on microbiota development was more pronounced in breast-fed babies consistent with the impact of fluctuations in environmental effects (e.g., mother-specific fluctuations in milk composition) \[[@B13-nutrients-10-00274]\]. The infant's gut microbial composition increases in number and diversity as they age \[[@B14-nutrients-10-00274],[@B15-nutrients-10-00274]\]. By around three years of age, the infants' gut microbiota will attain a diversity and complexity of composition that resembles the mature adult anaerobic gut microbiota \[[@B4-nutrients-10-00274],[@B14-nutrients-10-00274],[@B16-nutrients-10-00274],[@B17-nutrients-10-00274]\].
It is clear that under "normal" circumstances, the gut microbiota has a symbiotic relationship with the host during which, among other things, it contributes to: the storage and harvesting of energy \[[@B18-nutrients-10-00274]\]; development of the host immune system \[[@B6-nutrients-10-00274],[@B19-nutrients-10-00274],[@B20-nutrients-10-00274],[@B21-nutrients-10-00274]\]; maintenance of intestinal homeostasis \[[@B22-nutrients-10-00274]\]; and, nutrient processing \[[@B12-nutrients-10-00274]\]. Interactions between gut microbes and the host also have a profound effect on an individual's health later in life \[[@B23-nutrients-10-00274]\], while perturbation of the gut microbiota population structure (i.e., dysbiosis) is associated with pathological conditions \[[@B24-nutrients-10-00274]\] that include inflammatory bowel disease (IBD) \[[@B25-nutrients-10-00274]\], obesity, allergic \[[@B26-nutrients-10-00274]\], and autoimmune diseases \[[@B27-nutrients-10-00274]\].
Despite our awareness of the significance of maintaining the mutual relationship between the host and gut microbiota across the life-span, conclusive evidence of the factors that affect the development of the microbiome are not yet available. Various factors have been proposed to affect the early-life development of the microbiota, including: the composition of the maternal microbiome \[[@B28-nutrients-10-00274]\]; mode of birth \[[@B10-nutrients-10-00274],[@B24-nutrients-10-00274],[@B29-nutrients-10-00274]\]; antibiotic usage \[[@B30-nutrients-10-00274]\]; and, length of gestation \[[@B31-nutrients-10-00274],[@B32-nutrients-10-00274]\] ([Figure 1](#nutrients-10-00274-f001){ref-type="fig"}). In this review, we will focus on factors that are known to affect the establishment of the biggest human microbial reservoir---the GI tract and the reciprocal relationship between GI microbiota and GI tract development from in utero to post-natal life.
2. Development of the GI Tract {#sec2-nutrients-10-00274}
==============================
The human gastrointestinal tract or alimentary canal starts from the mouth, extending through well-defined anatomical regions---the oesophagus, stomach, small intestine, colon, rectum---and ending at the anus \[[@B6-nutrients-10-00274]\]. The functional and structural development of the GI tract is a crucial part of human development as the gut must accommodate the diversity of dietary inputs and foreign antigens that are introduced into the human body together with food throughout different stages of life \[[@B33-nutrients-10-00274]\].
The maturation of the human GI tract starts in utero but continues after birth with some functions, such as epithelial barrier mechanisms, accessory structures (e.g., glands), and the intestinal immune system, only becoming fully developed several months or years after birth \[[@B33-nutrients-10-00274]\]. The primitive gut forms from the dorsal section of the yolk sac approximately 22 days after conception, leading to the appearance of the foregut, midgut, and hindgut approximately 25 days after conception \[[@B34-nutrients-10-00274]\]. The stomach appears approximately five weeks post-conception. The midgut rapidly increases in length until it can no longer fit within the developing abdominal cavity and herniates into the vitelline sac before undergoing complex rotations and returning to the abdominal cavity after approximately 10 to 12 weeks of gestation \[[@B34-nutrients-10-00274]\]. GI tract development continues until all the major tissue components of the mature gut are present after approximately 20 weeks' post-conception \[[@B33-nutrients-10-00274]\]. Despite the fact that the GI tract originates from the dorsal section of the yolk sac, there are regional-specific tissue features (e.g., gastric pits, glands villi, and crypts) that differentiate between the different sections of the gut \[[@B33-nutrients-10-00274]\].
Comparative studies have identified abnormalities in gut-associated lymphoid tissue development and decreased antibody production in germ-free mice \[[@B35-nutrients-10-00274]\]. Similarly, germ-free piglets have been shown to have alterations to their intestinal physiology that include reduced turnover of the intestinal epithelial cell and a reduction in mucosal biosynthetic rate when compared to control animals \[[@B36-nutrients-10-00274],[@B37-nutrients-10-00274]\]. Collectively, these findings are consistent with the structural and functional development of the GI tract being affected by the composition and the activity of an individual's microbial flora. Because of the nature of GI tract development, these microbe-specific effects are likely to be due to impacts on developmental processes that occur both in utero and postnatally.
3. Microbial Impacts on In Utero GI Development {#sec3-nutrients-10-00274}
===============================================
We have long assumed that the gut is sterile before birth \[[@B38-nutrients-10-00274],[@B39-nutrients-10-00274],[@B40-nutrients-10-00274],[@B41-nutrients-10-00274],[@B42-nutrients-10-00274]\]. However, this dogma was challenged when several studies identified bacteria, bacterial DNA, or bacterial products in meconium \[[@B10-nutrients-10-00274],[@B43-nutrients-10-00274],[@B44-nutrients-10-00274]\], amniotic fluid \[[@B23-nutrients-10-00274],[@B45-nutrients-10-00274]\], and the placenta \[[@B23-nutrients-10-00274],[@B46-nutrients-10-00274]\]. Yet, evidence of live bacterial culture from placental and amniotic fluid samples remains limited \[[@B23-nutrients-10-00274]\]. Despite the limited evidence, these findings raise the possibility that intrauterine human gut development that prepares the protective barrier necessary for enteral feeding after birth is affected by the development of stage-specific microbiota that begins in utero \[[@B23-nutrients-10-00274]\].
4. The Importance of Fetal Swallowing for GI Development {#sec4-nutrients-10-00274}
========================================================
In order for the microbiome to affect GI development in utero, there must be a mechanism to ensure the selection and exposure of the appropriate microbial population or factors. The obvious medium for such a system is the amniotic fluid that bathes the developing fetus. Notably, the composition of amniotic fluid varies over the course of gestation \[[@B47-nutrients-10-00274]\]. Amniotic fluid is composed primarily of fetal urine, with contributions from secreted lung liquid, buccal secretions, and transmembrane flow \[[@B33-nutrients-10-00274]\]. Amniotic fluid also contains hormones and growth regulators \[[@B48-nutrients-10-00274]\], immune modulating proteins, and microbial components \[[@B23-nutrients-10-00274]\]. It remains unclear how the selection of particular microbes would be achieved within the amniotic fluid; however, interactions between typical environmental factors (e.g., pH, oxygen levels, carbon sources), innate and learned immunity are obvious candidates.
One of the potential route for the exposure of the developing GI to microbes/microbial products within the amniotic fluid can be mediated by swallowing. The human fetus begins swallowing amniotic fluid as early as 10 weeks post-conception \[[@B48-nutrients-10-00274]\]. This coincides with the oesophagus being innervated, which typically occurs by 13 weeks \[[@B33-nutrients-10-00274]\]. In the last trimester of pregnancy, the human fetus swallows between 700 and 1000 mL amniotic fluid per day \[[@B49-nutrients-10-00274],[@B50-nutrients-10-00274]\]. This provides a possible medium by which bacteria and bacterial products (e.g., glycoproteins, RNA and DNA) can be introduced into the developing gut.
Data from human infants born with gut abnormalities, such as gastroschisis and intestinal atresia, suggest that despite nutrition being obtained through the placenta, fetal gut development affects fetal growth in utero \[[@B50-nutrients-10-00274]\]. Babies with a proximal gastrointestinal atresia have reduced birthweight compared to babies born with a distal atresia \[[@B50-nutrients-10-00274],[@B51-nutrients-10-00274]\] and term babies with atresia have significantly reduced birthweight centiles compared with those born prematurely with atresia \[[@B50-nutrients-10-00274]\]. These data are consistent with observations from animal studies, which also have demonstrated the trophic effect of amniotic fluid \[[@B52-nutrients-10-00274]\] and that fetal swallowing is essential for normal development of the gut \[[@B53-nutrients-10-00274]\]. For example, fetal sheep that have undergone oesophageal ligation exhibit reductions in the thickness of the external muscle layers in their stomach, duodenum and proximal small intestine, changes in the length of intestinal villi, and altered rates of epithelial cell migration \[[@B34-nutrients-10-00274]\]. In rabbits that had undergone oesophageal ligation, infusion of amniotic fluid beyond the ligation resulted in relatively normal gut development as compared to that of rabbits with an oesophageal ligation alone \[[@B52-nutrients-10-00274]\]. These findings reveal the importance of swallowing amniotic fluid in utero for the anatomical development of the GI tract.
It remains to be determined if it is the act of swallowing or specific components of the amniotic fluid that are responsible for the observed impacts on development. However, oral administration of labelled *Enterococcus fecium* to pregnant mice resulted in this organism being isolated from the meconium of the newborn pups but not from the control group \[[@B44-nutrients-10-00274]\], as consistent with a mechanism for translocation to the amniotic fluid for fetal ingestion. Consistent with this, additional studies have demonstrated that maternal microbes can be transported to the amniotic fluid \[[@B23-nutrients-10-00274],[@B44-nutrients-10-00274]\] and the placenta \[[@B54-nutrients-10-00274]\]. However, irrevocable evidence for intrauterine transmission of microbes from mother to fetus that results in active colonisation and developmental impacts is still lacking.
5. Factors Affecting GI Microbiome Development after Birth {#sec5-nutrients-10-00274}
==========================================================
The complexity and richness of the microbial community progress through a series of developmental stages that range from the neonatal period before there is an apparent stabilisation after weaning. There are several crucial interrelated factors, together with individuality, which play an important role in shaping the transitioning human GI microbial composition. These factors include age \[[@B55-nutrients-10-00274],[@B56-nutrients-10-00274]\], diet \[[@B24-nutrients-10-00274],[@B29-nutrients-10-00274]\], host genetics \[[@B24-nutrients-10-00274],[@B29-nutrients-10-00274],[@B57-nutrients-10-00274]\], antibiotic usage \[[@B24-nutrients-10-00274],[@B29-nutrients-10-00274],[@B56-nutrients-10-00274]\], the physiology of the colonisation site \[[@B6-nutrients-10-00274]\], mode of birth \[[@B10-nutrients-10-00274],[@B24-nutrients-10-00274],[@B29-nutrients-10-00274]\], type of feeding (i.e., breast milk or formula milk) \[[@B10-nutrients-10-00274],[@B24-nutrients-10-00274]\], and the birth environment of the infants (e.g., NICU) \[[@B10-nutrients-10-00274]\].
The shape of the developing microbial composition is also affected by technical variation ([Figure 2](#nutrients-10-00274-f002){ref-type="fig"}). For example, culture based techniques for the identification of microbes are subject to biases that arise from: (1) oxygen-sensitivity; (2) the recalcitrance of some bacterial species to culture media; and, (3) competition between fast- and slow-growing bacteria. This limits current culture-based methods to the successful isolation of no more than 70% of intestinal microbes in a sample when compared to culture independent approaches \[[@B58-nutrients-10-00274]\].
Culture-independent techniques, typically use high throughput sequencing or array technologies to analyse the extracted nucleic acids from the community. These techniques are subject to variation due to processing, the kits that are used to prepare the DNA for analysis, and the computational analyses themselves. Commercially available kits including the PowerLyzer PowerSoil DNA Isolation Kit (MoBio, Carlsbad, CA, USA) \[[@B43-nutrients-10-00274],[@B59-nutrients-10-00274]\], QIAamp DNA Stool Mini Kit (QIAgen, Hilden, Germany) \[[@B23-nutrients-10-00274]\], MOBIO PowerSoil DNA Isolation Kit (MOBIO) \[[@B27-nutrients-10-00274]\], PowerMag^®^ Soil DNA Isolation Kit (MO-BIO Laboratories, Inc., Carlsberg, CA, USA) \[[@B60-nutrients-10-00274]\], and Fast DNA SPIN Kit (MP BIO, Santa Ana, CA, USA) \[[@B9-nutrients-10-00274],[@B58-nutrients-10-00274]\] have been widely used.
Despite the use of commercial kits for DNA preparation, sample processing methods, including different amounts of starting material (e.g., 100 mg to 200 mg \[[@B43-nutrients-10-00274],[@B59-nutrients-10-00274]\]), also introduce potential biases. Sample storage (i.e., duration and temperature) prior to processing varies (e.g., store at 4 °C and process within 24 h \[[@B27-nutrients-10-00274]\] and 72 h \[[@B61-nutrients-10-00274]\]; snap-freeze in liquid nitrogen; or place on dry ice immediately following collection and store at −80 °C \[[@B43-nutrients-10-00274]\])and contributes to potential biases in the sample composition. Inter-study variation is further compounded by the use of different 16S rRNA gene hypervariable regions (e.g., V5 − V3 \[[@B27-nutrients-10-00274]\]; V1 and V3 \[[@B9-nutrients-10-00274]\]; V1 − V3 \[[@B23-nutrients-10-00274]\]; V3 − V6 \[[@B13-nutrients-10-00274]\]; V2 − V4, V6 + V7 − V8 and V9 \[[@B62-nutrients-10-00274]\] and the most frequent amplified region---the V4 region \[[@B43-nutrients-10-00274],[@B59-nutrients-10-00274],[@B60-nutrients-10-00274]\]) which also contributes to variation between different studies. It is important to keep these systematic biases in mind as we consider the variation between studies.
5.1. Influence of Birth Mode on the Establishment of GI Microbiota {#sec5dot1-nutrients-10-00274}
------------------------------------------------------------------
The mode of birth determines the microbial population that infants are exposed to during birthing. For instance, vaginal birth exposes infants to the microbes that are currently colonising the mother's birth canal. This direct form of inheritance during birth results in infants who are born by vaginal delivery having a similar microbiota to that of their own mother when compared to other mothers \[[@B38-nutrients-10-00274],[@B63-nutrients-10-00274]\]. By contrast, significant overlap between the microbiota of mothers and children that are born by C-section has not been observed \[[@B38-nutrients-10-00274],[@B63-nutrients-10-00274]\]. Rather, environmental factors (e.g., delivery and surgical equipment, air, other infants and healthcare workers) appear to have a greater effect on the microbiome of infants born by C-section \[[@B6-nutrients-10-00274],[@B61-nutrients-10-00274]\]. Recent findings for infants born by C-section indicated that a period of labour prior to the surgery was associated with infants having a microbiota that more closely resembled that of vaginally-born infants, whereas infants born without any labour period had a microbiota that resembled that of the maternal skin \[[@B27-nutrients-10-00274]\]. C-section is suggested to be one of the reasons for early-life microbial disruption and this perturbation in microbial colonisation during infancy affect microbial-host interaction, which can lead to long-term metabolic consequences in the host \[[@B64-nutrients-10-00274],[@B65-nutrients-10-00274],[@B66-nutrients-10-00274]\]. In addition, higher chances to acquire atopic diseases during the first two years after birth also demonstrated by C-section infants as compared to vaginally birth infants based on data from 2500 full-term healthy newborns in LISA-Study \[[@B67-nutrients-10-00274]\].
A good example of the effect of birth mode on the gut microbiota comes from the impact of birth mode on the acquisition of *Lactobacillus* in the infant's GI tract. *Lactobacillus* is highly abundant in, and specific to, the maternal vagina with an IndVal index of 0.922 \[[@B27-nutrients-10-00274]\]. It has been reported that infants born through the mother's birth canal contain *Lactobacillus* as part of their microbiome profile, but that infants born by C-section do not \[[@B65-nutrients-10-00274]\]. These early observations gained further support in a subsequent study that detected significantly fewer *Lactobacillus* genus in the microbiome profile of infants born by C-section (*n* = 17) versus vaginal (*n* = 134) (Detection rate of 6% vs. 37%) \[[@B68-nutrients-10-00274]\]. Notably, the low detection rate of *Lactobacilli* in the intestinal discharge of the C-section infants persisted for the first six months after birth in contrast to vaginally born infants, who had higher and increased detection rates throughout the first six months after birth \[[@B68-nutrients-10-00274]\]. However, this difference in *Lactobacilli* detection rates vanished by three years of age \[[@B68-nutrients-10-00274]\].
The levels of bacteria in the *Bacteroides* and *Clostridium* genera (e.g., *Bacteroides fragilis* and *Clostridium difficile*) within an individual's microbiota are also associated with birth mode \[[@B38-nutrients-10-00274],[@B39-nutrients-10-00274],[@B43-nutrients-10-00274],[@B61-nutrients-10-00274],[@B63-nutrients-10-00274],[@B69-nutrients-10-00274],[@B70-nutrients-10-00274]\]. In the KOALA Birth Cohort study in the Netherlands (*n* = 1032), real-time quantitative PCR assays were used to enumerate different bacterial species from stool samples collected at one month of age \[[@B39-nutrients-10-00274]\]. Infants born by unassisted vaginal birth (*n* = 826) had relatively high numbers of *B. fragilis* and a reduced number of *C. difficile* compared to C-section infants \[[@B39-nutrients-10-00274]\]. By contrast, stool samples from infants born by C-section (*n* = 108) showed the inverse relationship \[[@B39-nutrients-10-00274]\]. The sources of *C. difficile* could be linked to environmental factors rather than the mother, as *C. difficile* was detected on the hands and in the stools from healthy hospital personnel \[[@B39-nutrients-10-00274],[@B71-nutrients-10-00274]\]. Notably, *C. difficile* has been considered as a microorganism that exists exclusively in hospitals \[[@B72-nutrients-10-00274]\] and has been shown to be absent from vaginal swabs from women prior to delivery \[[@B73-nutrients-10-00274],[@B74-nutrients-10-00274]\]. This may explain the high levels of *C. difficile* identified in hospital-born and C-section infants \[[@B39-nutrients-10-00274]\].
At the phylum level, a low abundance of *Bacteroidetes* (*p* = 0.002) was also observed in infants born by C-section (*n* = 9) when compared to vaginally born infants in a study of 24 infants conducted in south-eastern Sweden \[[@B38-nutrients-10-00274]\]. Notably, this reduction in the abundance of *Bacteroidetes* persisted for the first two years after birth \[[@B38-nutrients-10-00274]\]. This is consistent with previous reports that highlight delayed establishment of members of the *Bacteroides* in C-section infants in the first six months \[[@B61-nutrients-10-00274]\] and one year of life \[[@B75-nutrients-10-00274]\]. Notably, members of the *Bacteroides* genus are highly specific to the maternal stool (IndVal index of 0.943) \[[@B27-nutrients-10-00274]\]. Collectively, these findings highlight the central role of exposure to the maternal stool environment during birthing for the early inheritance and establishment of *Bacteroides* in the infant's microbial profile.
Not all studies have found an association between mode of birth (vaginal versus C-section), the inheritance and development of the GI microbiota. For example, a study of 21 infants found that mode of birth did not affect microbial population in preterm babies throughout the first three months after birth \[[@B11-nutrients-10-00274]\]. The incorporation of preterm infants (gestational age of 30 to 35 weeks) in this study could potentially explain this, as another study also found that birth mode was not significantly associated with microbiome composition in preterm infants \[[@B59-nutrients-10-00274]\]. In summary, studies indicate that infants born by C-section tend to have: lower numbers of anaerobes (e.g., *Bacteroidetes*); a less diverse microbiota \[[@B21-nutrients-10-00274],[@B38-nutrients-10-00274],[@B61-nutrients-10-00274]\]; delayed colonisation of microbial population \[[@B43-nutrients-10-00274]\]; and, they acquire atopic diseases \[[@B21-nutrients-10-00274]\] and metabolic disorders \[[@B66-nutrients-10-00274]\] more frequently than infants born by unassisted vaginal birth. However, these studies are complicated by ethnic and geographic diversity and differences in analytical methodologies.
5.2. Impact of Feeding on GI Microbiome Development {#sec5dot2-nutrients-10-00274}
---------------------------------------------------
Milk is the first food that is introduced into the GI tract postpartum and the composition of the milk is believed to directly impact on shaping the early GI microbiota \[[@B3-nutrients-10-00274],[@B4-nutrients-10-00274]\]. This impact can occur through the provision of: essential nutrients for bacterial proliferation \[[@B3-nutrients-10-00274]\]; immunomodulatory molecules \[[@B76-nutrients-10-00274]\]; and, microbes that are capable of colonising the infant \[[@B77-nutrients-10-00274]\]. The possibility that feeding type contributes to the early post-natal development of the GI flora is supported by an observed similarity between microbial composition in colostrum and the meconium from infants who were breast-fed from the first hour after birth \[[@B23-nutrients-10-00274]\]. Shared bacterial DNA (e.g., homologous to *Streptococcus thermophilus*, *Staphylococcus epidermidis*, and *Bifidobacterium longum*) has also been identified in human breast milk and infants' faecal samples \[[@B78-nutrients-10-00274]\]. This relationship is more pronounced between infants, their mother's milk and areolar skin when compared to a random mother (*p* \< 0.001) \[[@B14-nutrients-10-00274]\]. Furthermore, *Bifidobacterium longum* rDNA sequences can be co-detected in infant's faeces, maternal blood, maternal faeces and breast milk collected between one and four weeks post-birth \[[@B78-nutrients-10-00274]\]. Collectively, these results are consistent with the breast milk mediated vertical transfer of microbial communities to the infant's gut \[[@B14-nutrients-10-00274]\].
Culture-based techniques have identified more diverse microbiomes in formula-fed infants when compared to breast-fed infants \[[@B79-nutrients-10-00274]\]. Culture-independent studies have supported this observation \[[@B9-nutrients-10-00274],[@B80-nutrients-10-00274]\]. For example, Lee et al. (2015) characterised the effect of feeding type on the microbiota of 20 Korean infants who were born vaginally at term to healthy mothers \[[@B9-nutrients-10-00274]\]. Faecal samples were collected at four weeks of age from 10 exclusively breast-fed and 10 formula-fed infants. Relatively small amounts of formula supplementation (once every 24 h during the first week after birth) to breast-fed infants shifted the microbial profile towards a pattern that was similar to that observed for exclusively formula-fed infants \[[@B81-nutrients-10-00274]\]. Although some formula-fed infants were exposed to breast milk, these infants were fed with a diet that consisted of 70--100% formula milk \[[@B9-nutrients-10-00274]\]. Five species of bacteria were present in the faecal samples of all the infants in this study (i.e., both breast- and formula-fed group contained *Bifidocbacterium longum*, *Streptococcus salivarius*, *Strepotococcus lactarius*, *Streptococcus pseudopneumoniae*, and *Lactobacillus gasseri*). Lee et al. (2015) argued that the presence of these species in the intestines of these babies must be independent of feeding type, and thus these species constitute common commensal bacteria that are present in four-week-old Korean infants \[[@B9-nutrients-10-00274]\]. However, the relative abundances of these species differed in both groups, with *B. longum*, *S. pseudopneumoniae*, and *L. gasseri* having a greater abundance and *S. salivarius*, and *S. lactarius* having a lesser abundance in breast-fed infants when compared to formula-fed infants. These results are consistent with the hypothesis that the relative abundance of common commensal bacteria is altered by exposure to different feeding patterns, formula or breast milk.
Collectively, formula-fed infants tend to have relatively stable and diverse GI microbial communities that contain higher levels of facultative anaerobes and strict anaerobes when compared to breast-fed infants \[[@B9-nutrients-10-00274],[@B37-nutrients-10-00274],[@B82-nutrients-10-00274],[@B83-nutrients-10-00274]\]. Faecal samples from breast-fed infants are less complex, contain higher numbers of aerobic organisms, and exhibit more dramatic changes in microbial composition over their first year after birth \[[@B9-nutrients-10-00274],[@B82-nutrients-10-00274],[@B83-nutrients-10-00274]\]. Studies suggest that, once weaning (i.e., the introduction of solid foods into the diet) starts, the differences in microbial population between breast and formula-fed infants are lost and the microbial communities converge towards a complex adult microbiome \[[@B3-nutrients-10-00274],[@B6-nutrients-10-00274]\]. However, a recent study (*n* = 107) reported that the continuation of breast milk feeding after the introduction of solid food suppresses the diversification of the microbiota associated with the introduction of solid food \[[@B14-nutrients-10-00274]\]. The effect and mechanism of this suppression is yet to be determined and more studies that allow for other factors, such as ethnicity of subjects, to be controlled are required.
### Nutrients and Microbial Composition of Human Breast Milk
The nutritional composition of human milk varies as lactation progresses. Besides nutrients, breast milk also contains hormones \[[@B37-nutrients-10-00274]\], growth factors \[[@B37-nutrients-10-00274]\], microbiota \[[@B14-nutrients-10-00274],[@B84-nutrients-10-00274]\], immunoglobulin \[[@B37-nutrients-10-00274],[@B85-nutrients-10-00274]\], and enzymes \[[@B37-nutrients-10-00274],[@B85-nutrients-10-00274]\]. The protein content of early human milk from mothers who gave birth to preterm children is higher than that from mothers who gave birth at term \[[@B86-nutrients-10-00274],[@B87-nutrients-10-00274]\]. This is partly explained by a higher demand for protein to support growth in preterm infants compared to infants born at term. However, this protein content declines steadily with lactation \[[@B88-nutrients-10-00274]\] and is negatively associated with milk volume output at six and nine months after birth \[[@B89-nutrients-10-00274]\].
Breast milk that was aseptically-collected from lactating mothers, who birthed at term, contained members of the *Lactobacillus*, *Streptococcus*, *Enterococcus*, *Peptostreptococcus*, *Staphylococcus*, *Corynebacterium*, and *Escherichia* species \[[@B78-nutrients-10-00274]\]. It can be argued that the *Escherichia* species, found specifically (IndVal = 0.95) in the six week old infant gut, may have originated from breast milk. This is consistent with the observed low abundance, and hence low transfer possibility, of this species from other maternal body sites (i.e., skin, nares, oral, vagina and stool) \[[@B27-nutrients-10-00274]\]. The microbial composition of breast milk varies among mothers in terms of beta diversity according to the time post-birth (e.g., ≤6 months after delivery or ≥6 months) \[[@B14-nutrients-10-00274]\]. By contrast, the alpha diversity varies with lifestyles (e.g., rural and urban) \[[@B85-nutrients-10-00274]\]. For example, the alpha diversity of the breast milk associated microbiota between rural and urban women showed significantly higher microbial diversity in the breast milk from rural women \[[@B85-nutrients-10-00274]\]. These differences in the alpha diversity of the breast milk associated microbiota from women with different lifestyles potentially provides for different seeding of the microbial population within the infants GI tract. A longitudinal study from Pannaraj et al. (2017) identified a stable alpha diversity within the breast milk microbiota (within sample diversity) in the first year of their babies' lives. By contrast, the beta diversity (between samples diversity) of the breast milk microbiota increased during the first six months after birth, but slowly reduced when breast milk was no longer a primary source of nutrients \[[@B14-nutrients-10-00274]\].
Human milk oligosaccharides (HMOs) are the third largest component of breast milk. The types and amounts of HMOs produced from mothers who gave birth to preterm and term infants differ \[[@B90-nutrients-10-00274]\]. HMOs are a form of prebiotic and have been reported to be capable of promoting the growth of specific microorganism, including *Bifidobacteria* species \[[@B91-nutrients-10-00274],[@B92-nutrients-10-00274]\] and *Bacteroidetes*, but not pathogenic bacteria, such as *Enterobacteriaceae* \[[@B92-nutrients-10-00274]\]. Although *Bifidobacteria* predominate in both formula- and breast-fed infants \[[@B37-nutrients-10-00274],[@B79-nutrients-10-00274]\], the appearance in formula-fed infants is less frequent than that observed in breast-fed infants of the same age group \[[@B9-nutrients-10-00274],[@B37-nutrients-10-00274],[@B41-nutrients-10-00274],[@B80-nutrients-10-00274],[@B93-nutrients-10-00274],[@B94-nutrients-10-00274]\]. Typically, infants will acquire a broad spectrum of *Bifidobacteria* species from the mother, but not all *Bifidobacteria* species are able to degrade HMOs \[[@B94-nutrients-10-00274]\]. Thus, strains of *Bifidobacteria* that are able to break down the individual specific HMOs that are present in the breast milk are most likely to predominate in the infant's GI tract during early development \[[@B91-nutrients-10-00274],[@B95-nutrients-10-00274]\]. A high lactose content and the presence of sialylated and fucosylated oligosaccharides in human breast milk, when compared to cow milk, can also promote the growth of *Bifidobacteria* over other bacteria \[[@B9-nutrients-10-00274]\].
5.3. The Effect of Antibiotic Usage on Gastrointestinal Microbial Development in Infants {#sec5dot3-nutrients-10-00274}
----------------------------------------------------------------------------------------
The use of antibiotics is more prevalent in infants born via C-section \[[@B39-nutrients-10-00274]\] and in those born preterm when compared to term infants born vaginally \[[@B96-nutrients-10-00274]\]. Maternal and infant exposure to antibiotics during the perinatal period has been linked to increased risks of later onset diseases, such as asthma \[[@B97-nutrients-10-00274]\], obesity \[[@B98-nutrients-10-00274]\], inflammatory bowel disease \[[@B99-nutrients-10-00274]\], and other allergic/inflammatory conditions \[[@B100-nutrients-10-00274]\] in children. Antibiotic exposure during the prenatal, perinatal, and postnatal periods has also been hypothesized to cause a delay in microbial maturation from 6 to 12 months after birth \[[@B101-nutrients-10-00274]\].
Intrapartum antimicrobial prophylaxis (IAP) is believed to be the most frequent source of antibiotic exposure in neonates \[[@B102-nutrients-10-00274]\]. IAP (penicillin, ampicillin, or ampicillin plus erythromycin) \[[@B103-nutrients-10-00274]\] is administered to mothers who are positive for group B *Streptococcus* during labour to reduce the risk of early-onset neonatal infections \[[@B62-nutrients-10-00274]\], such as pneumonia, septicaemia, and meningitis \[[@B104-nutrients-10-00274]\]. Two cohort studies of full term vaginally-born babies have reported reduced alpha diversity in faecal samples from infants that are exposed to maternal IAP when compared to non-IAP exposed infants \[[@B102-nutrients-10-00274],[@B105-nutrients-10-00274]\]. Absolute levels of *Actinobacteria* and *Bacteroidetes* were lower in IAP-exposed infants in two different studies that performed 16s rRNA amplicon sequencing using the Illumina \[[@B102-nutrients-10-00274]\] and Ion Torrent \[[@B62-nutrients-10-00274]\] sequencing platforms. Significantly lower levels of *Bifidobacteriaceae* were also observed in IAP-exposed infants \[[@B62-nutrients-10-00274],[@B102-nutrients-10-00274],[@B105-nutrients-10-00274]\]. By contrast, the *Firmicutes* \[[@B102-nutrients-10-00274]\] and *Proteobacteria* \[[@B105-nutrients-10-00274]\] phyla increased numbers (*p* \< 0.05 and *p* \< 0.062, respectively) in IAP-exposed infants.
Exposure of the maternal system to IAP impacts upon the early GI microbial composition in infants. However, the magnitude of the effect and its relationship to the duration (i.e., short- and long-term) and timing of the exposure is yet to be determined. Despite this, we know that the impact on certain bacterial counts within faecal samples from neonates in early life is affected by the combined effect of IAP exposure, postpartum feeding mode \[[@B105-nutrients-10-00274],[@B106-nutrients-10-00274]\], and birth mode \[[@B107-nutrients-10-00274]\]. Studies conducted to date to understand the effects of these interactions on microbiota maturation have focused on term born babies in western countries \[[@B102-nutrients-10-00274],[@B105-nutrients-10-00274],[@B106-nutrients-10-00274],[@B107-nutrients-10-00274]\]. Therefore, future studies should focus on the combined effects of these factors on GI microbial maturation processes, taking into account gestational age and the ethnicity of the infants.
5.4. Environmental Factors that Affect the Infants' GI Microbiome Development {#sec5dot4-nutrients-10-00274}
-----------------------------------------------------------------------------
Exposure to different extra uterine environments during early gut development contributes to the colonisation and evolution of the infant GI microbiota. Infants born by C-section are speculated to be more susceptible to environmental factors \[[@B20-nutrients-10-00274],[@B91-nutrients-10-00274]\]. This is particularly true for preterm infants who have a higher chance of developing a flora that reflects the Neonatal Intensive Care Unit (NICU), due, in part, to the immaturity of their GIs and prolonged exposure to this environment \[[@B4-nutrients-10-00274]\].
The route of microbial transfer from the immediate environment to infants is hard to verify but studies have shown that microbes from the immediate environment can be isolated from infant faecal samples \[[@B17-nutrients-10-00274],[@B108-nutrients-10-00274]\]. Cross-transmission between patients and dissemination of a multi-drug resistant (MDR) *Acinetobacter baumannii* strain also led to an outbreak in a NICU in Tunis. 31 neonates (gestational ages 26 to 41 weeks) developed MDR *A. baumannii* associated pneumonia and there were 10 deaths due to infection after the MDR *A. baumannii* was transferred from a neonate in the epidemic-associated surgical ward of another hospital \[[@B108-nutrients-10-00274]\]. These results are consistent with observations that infants from different geographical areas or different hospitals harbour different microbial populations \[[@B17-nutrients-10-00274],[@B82-nutrients-10-00274]\]. However, further studies on larger groups of infants are required to verify the reliability and specificity of these results and to extend them to the population level. On the other hand, the PiPS trial, a double-blind randomised placebo-controlled trial of probiotic treatment with *Bifidobacterium breve* for the prevention of sepsis and necrotising enterocolitis in 1310 preterm babies born between 23 and 30 weeks' gestation from 24 hospitals in the southeast England found that the probiotic strain of *Bifidobacterium breve* could be detected in the stool of 37% of babies in the placebo arm, as compared with 85% of the intervention arm, indicating that environmental-related factors resulted in significant cross-colonisation of *B. breve* in babies in this study \[[@B109-nutrients-10-00274]\]. Interestingly, this phase 3 PiPS trial also found no difference in the microbial diversity and richness of the microbiome of babies in the two arms of the study \[[@B110-nutrients-10-00274]\].
It is possible that the hospital environment, handling, feeding, and treatment regimes enhance microbial transmission to neonates \[[@B17-nutrients-10-00274]\]. However, the details of the mechanisms of transmission, dominant microbial populations within the hospital environments and the bacterial strains that have the highest chances to successfully colonise the infants' GI remain elusive and are worth exploring in future studies.
### Oxygen Levels in Term and Preterm Infants and Their Effects on GI Microbiota Development
The microbiota of children born preterm tend to more frequently contain detectable levels of pathogenic microorganisms (e.g., *K. pneumoniae* and *C. difficile*) \[[@B11-nutrients-10-00274],[@B92-nutrients-10-00274],[@B111-nutrients-10-00274]\], reduced microbial diversity \[[@B92-nutrients-10-00274],[@B111-nutrients-10-00274]\], and contain low levels of short chain fatty acids (SCFA) \[[@B11-nutrients-10-00274]\]. Notably, when compared to term infants, preterm infants tend to be dominated by facultative anaerobes, including *Enterobacteriaceae* \[[@B112-nutrients-10-00274]\] and *Enterococcaceae* \[[@B5-nutrients-10-00274],[@B41-nutrients-10-00274]\], and have low levels of anaerobes from the *Bifidobacterium*, *Bacteroides*, and *Atopobium* \[[@B5-nutrients-10-00274],[@B41-nutrients-10-00274]\].
Newborn infants have an aerobic intestine at birth \[[@B113-nutrients-10-00274]\]. The high level of oxygen in the newborn GI tract favours the appearance of facultative anaerobes (e.g., *Enterobacteriaceae*, *Enterococcus*, and *Streptococcus*) \[[@B5-nutrients-10-00274],[@B6-nutrients-10-00274],[@B11-nutrients-10-00274],[@B83-nutrients-10-00274],[@B113-nutrients-10-00274],[@B114-nutrients-10-00274]\]. Other facultative anaerobes are also present in the neonatal GI tract (e.g., *Staphylococci*, *Escherichia coli*, *Enterococcus fecalis*, and *faecium*, *Klebsiella*, *Enterobacter,* and, infrequently *Aeromonas*, *Pseudomonas*, and *Acinetobacter*) \[[@B3-nutrients-10-00274],[@B6-nutrients-10-00274]\]. These early colonizers gradually create a reduced, anaerobic environment within the GI tract by consuming the available oxygen, consequently facilitating the establishment of obligate anaerobes (e.g., *Bifidobacterium*, *Clostridium*, *Bacteroides*, *Veillonella*, *Eubacterium*, and *Ruminococcus* species) \[[@B3-nutrients-10-00274],[@B6-nutrients-10-00274],[@B11-nutrients-10-00274],[@B39-nutrients-10-00274],[@B43-nutrients-10-00274],[@B83-nutrients-10-00274],[@B92-nutrients-10-00274],[@B113-nutrients-10-00274]\]. In addition to reducing oxygen levels and facilitating the colonisation of the GI by strict anaerobes, many of the primary facultative anaerobic colonisers are potentially pathogenic \[[@B83-nutrients-10-00274],[@B113-nutrients-10-00274]\].
Differences in oxygen levels exist between infants born term and preterm \[[@B115-nutrients-10-00274]\]. However, there is currently insufficient evidence to support oxygen levels as a factor contributing to the differences in the establishment of the gut microbiota in preterm infants. It remains likely that the observed differences in oxygen levels in the GI tract of preterm and term infants might be due to medical practices in the NICU. These differences could include the use of continuous positive airway pressure, which can lead to increased air in the preterm GI tract and can cause a delay in the establishment of the stage-specific commensal bacteria. Thus, it remains a possibility that GI oxygen levels contribute to health-related complications in preterm infants through a mechanism that enhances the levels of facultative anaerobes in the preterm gut.
6. Conclusions {#sec6-nutrients-10-00274}
==============
Gastrointestinal and gut microbiota maturation is an intricate, lengthy, and complicated process which starts in utero and continues after birth. Currently, there is no standardised definition of the composition of a "healthy" GI microbiota at different developmental stages and the main factors that contribute to the establishment of the GI microbiota in the early life remain elusive \[[@B63-nutrients-10-00274]\]. Seeding of the GI tract microbiome is suggested to begin in utero \[[@B23-nutrients-10-00274],[@B68-nutrients-10-00274],[@B116-nutrients-10-00274]\] and a representative of the intrauterine environment, meconium \[[@B13-nutrients-10-00274]\], was detected to harbour bacteria that have been detected in amniotic fluid \[[@B117-nutrients-10-00274]\]. GI microbial composition and species abundancy in infants are affected by variables that can directly or indirectly perturb the microbial community throughout the growth periods. Other than the apparent factors discussed in this review, other factors \[[@B10-nutrients-10-00274]\], such as geographical regions, host genetic factors, the effect of mix-feeding (breast milk and formula milk), hygiene level of healthcare providers, brands and content of formula milk consumed, and inter-individual variable are also likely to contribute to GI microbiota development.
Albeit clear that environmental and extrinsic factors affect gut microbiota development, a lack of statistical power limits current studies with preterm infants. This is particularly relevant because nearly 15 million babies are born preterm each year worldwide and the rate of preterm birth is increasing \[[@B118-nutrients-10-00274]\]. Thus, there is a need for a powered longitudinal study that focuses on children born preterm and accounts for ethnicity and other potential confounders.
Crosstalk between host cells (e.g., intestinal brush border cells or immune cells) and the colonising microbiota is likely to be critical for metabolic development and the programming of body immune system in infants \[[@B119-nutrients-10-00274],[@B120-nutrients-10-00274]\]. Although the intrauterine environment has been proven to not be sterile, the influence of the maternal microbial community on the establishment of microbial population in utero is yet to be determined. As such, the studies will require experiments that highlight the pathways of bacteria translocation from the mother to the infants. Notably, the mechanisms of translocation during vaginal birth are still not fully understood, nor are the host characteristics that select the bacteria species that will be inherited from maternal gut. In order to fully understand how the GI microbiota, in its entirety, can be manipulated to enhance health, well-being and performance, it is essential to understand how each of these factors interact with one another to maintain intestinal homeostasis in developing infants, children, and adults.
This work was supported by the DIAMOND trial (Health Research Council ID\#: 16/605), a multicentre factorial design randomised trial with ethics approval from the New Zealand Health and Disability Ethics Committee (16/NTA/90).
C.C.Y.L., F.H.B. and J.M.O.S. were involved in the preparation and editing of the manuscript. All authors approved the final version of the paper.
The authors declare no conflicts of interest.
C-section
Caesarean section
DNA
Deoxyribonucleic acid
GI
Gastrointestinal
HMO
Human milk oligosaccharides
IndVal index
Indicator value index
IAP
Intrapartum antimicrobial prophylaxis
IBD
Inflammatory bowel disease
MDR
Multi-drug resistant
NICU
Neonatal Intensive Care Unit
RNA
Ribonucleic acid
SCFA
Short chain fatty acids
{#nutrients-10-00274-f001}
{#nutrients-10-00274-f002}
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#S0001}
============
The circadian system adjusts physiology and behaviour to the varied demands of the day--night cycle (Czeisler et al. [1999](#CIT0006); Wright et al. [2001](#CIT0040); Roenneberg et al. [2003a](#CIT0026)). To ensure synchrony with the astronomical day, the circadian system entrains to daily environmental signals (zeitgebers = time givers). The light--dark cycle is the most significant zeitgeber for most organisms, including humans (Honma et al. [1987](#CIT0013); Roenneberg et al. [2007](#CIT0029)). Differences in the relationships between an individual's circadian phase and external local time give rise to a distribution of chronotypes across the population, ranging from *early* chronotypes, the proverbial "larks", to *late* chronotypes termed "owls" (Roenneberg et al. [2003a](#CIT0026)).
The timing of light exposure has a differential effect upon circadian phase: early light exposure advances the cycle whilst light late in the internal day delays circadian phase (Czeisler et al. [1989](#CIT0007); Khalsa et al. [2003](#CIT0016)). Thus, exposure to bright artificial light in the evening before bedtime has been associated with a delay in circadian phase, as assessed by measures of subjective chronotype (Martin et al. [2012](#CIT0020); Vollmer et al. [2012](#CIT0037)), subjective sleep timing (Koo et al. [2016](#CIT0018)), salivary melatonin levels (Gordijn et al. [1999](#CIT0011); Benloucif et al. [2008](#CIT0002); Cajochen et al. [2011](#CIT0004)) and core body temperature (Krauchi et al. [1997](#CIT0019)). Furthermore, adolescents living in urban areas and exposed to bright artificial light at night have a later chronotype as assessed by the Munich ChronoType Questionnaire (MCTQ) and Morningness--Eveningness Questionnaire (MEQ), compared to those living in more rural settings (Vollmer et al. [2012](#CIT0037)). By contrast exposure to bright light in the morning results in an advance of the circadian phase of melatonin synthesis and release (Dijk et al. [1989](#CIT0008); Gordijn et al. [1999](#CIT0011); Revell et al. [2005](#CIT0024)). In addition, bright morning light has been used as a therapy for advancing sleep timings in patients with delayed sleep--wake phase disorders (Rosenthal et al. [1990](#CIT0032); Saxvig et al. [2014](#CIT0033)) and, more recently, with social jet lag (SJL) (Geerdink et al. [2016](#CIT0010)).
Despite society's increasing detachment from the natural light--dark cycle, sunlight can still be seen to impact chronotype. Living further east within the same time zone in the Northern Hemisphere is associated with an earlier subjective chronotype in adults assessed using the MCTQ (Roenneberg et al. [2007](#CIT0029)) and in adolescents assessed with the MEQ (Randler [2008](#CIT0022)), most likely as a result of an earlier sunrise time. Seasonal changes are also apparent, such that during the months of increasing day length, subjective chronotype advances with individuals rising earlier (Kantermann et al. [2007](#CIT0014); Allebrandt et al. [2014](#CIT0001)). There is also some evidence that geographical location has an impact upon chronotype. For example, in a study conducted in Brazil, subjective chronotype was assessed using the MCTQ and MEQ in two cities: São Paulo at latitude 23° 32\' S and longitude 46° 38\' W and Natal at 05° 47\' S and 35° 12\' W. Chronotype was found to be earlier in individuals living in Natal, the city closest to the equator (Miguel et al. [2014](#CIT0021)).
Clearly, the pattern of natural light within a particular environment will be critical in defining an individual's phase of entrainment. However, an individual's behaviour within that environment will also play an important role. A recent study compared the same individuals living under their normal urban routines (including artificial light at night) with a period under natural light exposure (camping without artificial light). The findings demonstrated that increased exposure to natural light advanced the circadian phase of all individuals (Wright et al. [2013](#CIT0041); Stothard et al. [2017](#CIT0035)). Increasing photic zeitgeber strength by spending more time outside has also been correlated with self-reported chronotype: the more time spent outside, the earlier the chronotype (Roenneberg and Merrow [2007](#CIT0030); Roenneberg et al. [2015](#CIT0027)).
By studying populations across the Northern Hemisphere and Southern Hemisphere, specifically Oxford, Groningen, Munich, Perth, Melbourne and Auckland, we aimed to investigate the association between geographical location and chronotype and how different aspect(s) of environmental light (timing; length of time spent outside; intensity of light, sleep timings relative to sunrise and sunset) might influence chronotype.
Materials and methods {#S0002}
=====================
Students were recruited from six universities: University of Oxford, UK (51° 45\' N, 1° 15\' W); University of Groningen, The Netherlands (53° 13\' N, 6° 33\' E); LMU, Munich, Germany (48° 8\' N, 11° 34\' E); University of Western Australia, Perth, Australia (31° 57\' S, 115° 51\' E); Monash University, Melbourne, Australia (37° 48\' S, 144° 57\' E) and University of Auckland, New Zealand (36° 50\' S, 174° 44\' E). Students were asked to complete the online version of the MCTQ twice, in May and October of 2010, to control for seasonal influences. Overall, 13 299 individuals completed the MCTQ online. Over half of the participants were excluded from the analysis (see section on 'Data processing'). 6 441 students (mean age 21.5 ± 2.2 years, 67.5% female, see [Table 1](#T0001) for group demographics) were included in the analysis. Daily irradiance, sunrise and sunset times were obtained for May and October 2010. Ethical approval for this study was obtained from the local ethics committee for each university involved in the study.10.1080/07420528.2018.1482556-T0001Table 1.Average demographics, sleep timings and social jet lag for each city. Oxford (*n* = 302)Groningen (*n* = 3050)Munich (*n* = 1919)Perth (*n* = 342)Melbourne (*n* = 368)Auckland (*n* = 460)Age (years)21.1 ± 2.0921.54 ± 2.1422.50 ± 1.9318.96 ± 1.3820.51 ± 1.9020.00 ± 1.98Gender: females (%)161(53.3)1996(65.4)1401(73.0)202(59.1)271(73.6)314(68.3)Workdays (local time) Bedtime00:42 ± 01:0900:19 ± 01:0700:02 ± 01:0623:02 ± 01:3000:12 ± 01:2323:01 ± 01:10Wake-up time07:55 ± 00:5308:02 ± 01:0307:30 ± 01:0107:28 ± 01:1307:20 ± 01:0407:00 ± 01:02Free days (local time) Bedtime01:30 ± 01:2401:14 ± 01:2201:09 ± 01:2300:46 ± 01:3900:59 ± 01:3200:46 ± 01:27Wake-up time09:52 ± 01:2009:53 ± 01:2009:41 ± 01:2309:23 ± 01:3209:45 ± 01:3209:10 ± 01:28Social jet lag (hours)1.44 ± 0.831.39 ± 0.911.66 ± 0.931.46 ± 0.921.65 ± 0.981.62 ± 0.95[^2]
Materials {#S0003}
=========
The Munich ChronoType Questionnaire {#S0003-S2001}
-----------------------------------
The online version of the MCTQ (Roenneberg et al. [2003b](#CIT0031)) was used in the native language of the country of each university. The MCTQ consists of questions concerning sleep timings for both workdays and free days separately, work time and time spent outside. The MCTQ has been validated against actigraphic recordings (Vetter et al. [2015](#CIT0036)) and melatonin rhythms (Kitamura et al. [2014](#CIT0017)). The MCTQ is used to calculate the MSF as the mid-point between sleep onset and sleep end. MSF was corrected for oversleep on free days (MSF~sc~: Mid Sleep point on Free days, Sleep Corrected) that occurs as a result of sleep debt \[MSF~sc~ = MSF -- (SDf -- ((((nWD × SDw) + (7 − nWD)) × SDf)/7)), where SDf is the sleep duration of free days, SDw is the sleep duration of workdays and nWD is the number of workdays (Roenneberg et al. [2004](#CIT0028)). In cases where the numbers of workdays were missing, five workdays were assigned. SJL was also calculated from the MCTQ \[SJL = MSF -- MSD\] where MSF is the mid-sleep point of free days and MSD the mid-sleep point of work days (Wittmann et al. [2006](#CIT0039)).
Light data {#S0003-S2002}
----------
*"Time spent outside"* was self-reported on the MCTQ. The weighted average of the number of hours given for free days and workdays was calculated using the number of workdays also reported on the MCTQ. If no workdays were given, five workdays were assigned. \[time spent outside = ((time spent outside on workdays × number of workdays) + (time spent outside on free days × (7 − number of workdays)))/7\].
*"Light dose"* is a measure of how much light individuals are exposed to over a given period of time. Here, we calculated the average hourly irradiance for the day for participants that completed the MCTQ and normalised this to the "time spent outdoors", averaged for work and free days \[light dose = daily irradiance/day length × time spent outside\].
*"Day length"* for each day of the collection periods, for each city, was calculated using the world clock (<http://www.timeanddate.com/worldclock>).
*"Daily irradiances"* for the collection periods, for both May and October 2010, were obtained from three sources on an hourly basis. The data for Oxford, Groningen and Munich were provided by Dr Lucien Wald (MINES, ParisTech), obtained from Meteosat satellite images and converted to data maps of solar radiation using the Heliosat-2 method (Rigollier et al. [2004](#CIT0025)). The data for Perth and Melbourne were obtained from the Australian Bureau of Meteorology, again derived from satellite images processed by the Australian Bureau of Meteorology. Finally, the data for Auckland were obtained from the New Zealand Meteorological office based on readings from its ground station in Auckland. For all daily irradiance, the data represent light intensity experienced at ground level, taking into account weather conditions, either through processing of satellite data or as data taken at ground level.
Data processing {#S0004}
===============
As only 440 participants completed the questionnaire in both May and October (420 from Northern Hemisphere cities and 20 from Southern Hemisphere cities), longitudinal analysis was not performed and data for these participants were only included in the data analysis from the May collection period. Individuals were excluded if they were outside the age range (17--26 years); did not indicate they were currently living in any of the cities of interest; completed the questionnaire outside May or October 2010; or had reported working shifts during the past 3 months. Individuals were also excluded if they indicated using an alarm clock on free days (an exclusion criterion for chronotyping). For inter-hemispheric comparisons, months were assigned to season (Northern Hemisphere, May, and Southern Hemisphere, October, as spring and *vice versa* as autumn).
Data analysis {#S0005}
=============
Statistical analysis was performed using R version 3.0.1 (2013-05-16, Copyright 2013 The R Foundation for Statistical Computing). Linear mixed-effects models were fitted using the lme package for group and seasonal assessments of MSF~sc~ and light data. For categorical comparisons, linear models were referenced to Oxford for group comparisons, females for sex comparisons and spring for season comparisons. Since age and sex are known to influence chronotype, both of these were included as covariants when modelling MSF~sc~. Spearman's rank correlation analysis was performed to assess associations between MSF~sc~ and light data.
Results {#S0006}
=======
On average, students in this sample reported the following habitual sleep-related times: bed time on workdays: 00:11 ± 01:10 (mean, SD) and nearly an hour later on free days: 01:09 ± 01:24; wake-up time was two hours later on free days (09:45 ± 01:23) compared to workdays (07:43 ± 01:06). A mean of 1.51 ± 0.93 h of SJL was reported. [Table 1](#T0001) details wake-up and bed times for work and free days at each city along with SJL. The sleep midpoint on free days (corrected for over sleep; MSF~sc~) was different between cities but not between seasons, and no city--season interactions were found (see suppl. data model 1). Hence, MSF~sc~ was collapsed across seasons.
When plotting, chronotype and time spent outside against the absolute distance of each city from the equator, MSF~sc~ showed a positive association: with chronotype becoming later with increasing distance ([Figure 1a](#F0001)). Controlling for age and sex, the cities formed three groups for MSF~sc~: Oxford, Groningen and Munich were not statistically different from each other, and had the latest MSF~sc~; Melbourne showed an intermediate MSF~sc~ and was statistically significant from all the other cities; and Perth and Auckland showed the earliest MSF~sc~ and were also not statistically significant from each other (see suppl. data model 2).10.1080/07420528.2018.1482556-F0001Figure 1.Relationship between mid-sleep time points (MSFsc) and ambient light conditions in students living in different cities of the Northern and Southern Hemispheres. Cities are plotted relative their distance to the equator and against (a) Average MSFsc (as time-of-day in hours (decimal time), (b) time spent outside and (c) light dose. Dashed lined boxes represent cities that are not statistically different to each other (see supplementary materials for models). Error bars = standard error of the mean. MSFsc: mid-sleep on free days sleep corrected.
Overall, the students (regardless of city) reported they spend on average (2.20 ± 1.33 h) outside a day resulting in exposure to on average (24.74 ± 20.09 W/m^2^) of light on the day they completed the survey (light dose). Students in Perth reported spending the most amount of time outside (2.76 ± 1.67 h) and experienced the highest intensity of light whilst outside, light dose (62.68 ± 39.11 W/m^2^, see suppl. data model 3 and 4). Whereas students in Melbourne reported spending the least amount of time outside (1.89 ± 1.49 h), students in Oxford received the lowest light dose (19.43 ± 14.31 W/m^2^). In relation to geographical location, the amount of time spent outside was not associated with the distance of each city from the equator ([Figure 1b](#F0001)), but light dose did show an association, with the cities nearest to the equator experiencing a higher light dose, except for Auckland ([Figure 1c](#F0001)).
MSF~sc~ was positively, although weakly, correlated with time spent outside (rho = 0.036, *p* = 0.005) indicating that the longer students spent outside the later their sleep midpoint on free days. Time spent outside binned for MSF~sc~ in 30 min intervals, showed a stronger positive correlation (rho = 0.86, *p* = 0.011, [Figure 2a](#F0002)). However, this was only statistically significant with the removal of the outlier of MSF~sc~ binned from 06:30 to 07:00. Light dose was not correlated with MSF~sc~ for raw (rho = − 0.0005, *p* = 0.97) or binned data (rho = − 0.15, *p* = 0.71, [Figure 2b](#F0002)). Using a linear mixed effect model, taking age and sex into account, time spent outside but not light dose was found to have a significant effect on MSF~sc~ (see suppl. data model 5). However, the addition of time spent outside into the model did not remove the effect of city; moreover, no city-time spent outside interaction was found (see suppl. data model 6). This suggests that although the amount of time spent outside does have an influence on sleep midpoint on free days, other differences between the cities are also important.10.1080/07420528.2018.1482556-F0002Figure 2.MSFsc and ambient light conditions. Average time spent outside (a) and light dose (b) for 30 minute bins of MSFsc across all students regardless of city. Error bars represent the standard error of the mean. (c) The shorter the proportion of the light window (time between sunrise and sunset) students are awake for, the later MSFsc for work and free days. (d) The later students wake up after sunrise (represented by line) the later MSFsc for work and free days. (e) The later student go to bed after sunset (represented by line) the later MSFsc for work and free days. MSFsc: mid-sleep on free days sleep corrected.
To define the time of day, the students in this population were most likely to receive natural light, the timing of sunrise and sunset the day each student completed the survey was compared to their self-reported sleep timings for work and free days. The proportion of daylight (i.e. between sunrise and sunset) during which time students were awake was negatively correlated to MSF~sc~ for both work (rho = − 0.4, *p* \< 0.001) and free days (rho = −0.71, *p* \< 0.001, [Figure 2c](#F0002)), indicating that students with the latest MSF~sc~ were only likely to be awake for around 40% of the daylight period. This is because students wake up after sunrise rather than going to bed before sunset ([Figure 2d](#F0002) and [2e](#F0002), respectively). About 75.5% of students woke up after sunrise on workdays and 98.1% on free days, with 15.7% waking up 5 h after sunrise on free days (1.2% on workdays).
The impact of geographical location on MSF~sc~ was found to be most influenced by the timing of sunset. The average MSF~sc~ per city was plotted against the timing of sunrise and sunset for the day the survey was completed, as well as time spent outside and light dose ([Figure 3](#F0003)). The timing of sunset showed a positive association with MSF~sc~: the later sunset, the later MSF~sc~. Interestingly, no association was seen for sunrise. Collectively, these data suggest that the timing of sunset and therefore the amount of light in the evenings may be more influential on sleep midpoint on free days than the amount of light in the morning in this population.10.1080/07420528.2018.1482556-F0003Figure 3.Geographical location, MSFsc and ambient light conditions. Average MSFsc for each city plotted against light dose (a), time spent outside (b), time of sunrise (c) and time of sunset (d). Sunset times given as hours from midnight.
Discussion {#S0007}
==========
The findings from this study suggest that in a university student population, a later chronotype is associated with (i) living further from the equator, (ii) a later sunset, (iii) spending more time outside and (iv) waking up after sunrise. Significantly, we did not find that light intensity was associated with chronotype. Initially, these findings appear to contradict earlier studies where increased light exposure is associated with an earlier chronotype in the general population (Roenneberg and Merrow [2007](#CIT0030); Wright et al. [2013](#CIT0041); Roenneberg et al. [2015](#CIT0027); Stothard et al. [2017](#CIT0035)). However, our findings are derived exclusively from a university student population aged between 17 and 26 years. Thus, we suggest that the age and occupation of our population increase the likelihood that these individuals will experience relatively little light exposure in the morning whilst encountering more light exposure later in the day, when light has a delaying effect upon the circadian system.
In a sample of approximately 200,000 individuals from primarily Central Europe and North America, spending more time outside was associated with an earlier chronotype (Roenneberg and Merrow [2007](#CIT0030); Roenneberg et al. [2015](#CIT0027)). However, when age was taken into consideration, 15--20 year olds did not show a significant correlation between time spent outside and chronotype, and 20--25 year olds had only a weak correlation (Roenneberg et al. [2015](#CIT0027)). Our sample of over 6000 students (17--26 years) falls across these age ranges and also differs from the Roenneberg sample (MCTQ database) in several important aspects. A key difference is the work status of the populations studied: all individuals within our sample are university students. However the general population sample from the MCTQ database would have included individuals who were studying at school or university, or working. It is possible, therefore, that imposed work schedules could result in more morning vs. evening light exposure in the general MCTQ population. In addition, the data for the current study was collected exclusively in May and October, whilst the general population sample was collected all year round, which might also have an impact upon the timing of light exposure.
The timing of light exposure has a differential effect upon circadian phase: early light exposure advances the cycle whilst light late in the internal day delays circadian phase (Czeisler et al. [1989](#CIT0007); Khalsa et al. [2003](#CIT0016)). In our student population, we found that longer time spent outside the later chronotype, which would suggest that our population was exposed to more phase delaying evening light than phase advancing morning light. Although it was not possible to determine the timing of light exposure definitively from our study, we provide several lines of evidence that support the importance of evening light in this population. In the present study, we demonstrated that the later students wake up after sunrise the later MSF~sc~. As a result, individuals are likely to be exposed to a photoperiod with a greater proportion of evening phase delaying vs. morning phase advancing light. Clearly, future studies will need to define the phase relationship between the internal circadian and external environmental light cycle. Moreover the timing of sunset rather than sunrise was found to be most associated with MSF~sc~ in our population. Previously, a longitudinal study of around 55 000 individuals has reported that MSF~sc~ tracks sunrise and not sunset (Kantermann et al. [2007](#CIT0014)). However, again, the broad demographics of this population make direct comparisons to our population difficult, but it is possible that the association of young adult university students (comparable to our population) is masked by other individuals in the sample. Finally, the sensitivity of young adults to evening light has recently been demonstrated in two studies. In 20 healthy young adults (mean age of 23), later light onset and offset has been associated with later melatonin onset as assessed using dim light melatonin onset (Wams et al. [2017](#CIT0038)). A mathematical model of sleep timing based on the experimentally derived effects of light on the human circadian clock, and interaction of the circadian clock and sleep homeostat, predicts a similar finding. Individuals with a longer intrinsic clock and hence later chronotype are predicted to be more susceptible to evening light, causing even more of a delay in the circadian cycle (Skeldon et al. [2017](#CIT0034)).
Geographical location was found to be associated with chronotype: the closer to the equator the earlier chronotype, and in this regard, our findings are consistent with previous findings (Miguel et al. [2014](#CIT0021)). However, this association was assumed to be driven by higher environmental light intensities closer to the equator. Interestingly, it was only the duration of time spent outside -- not the intensity of light -- that was found to influence MSF~sc~ in our study except for subjects in Auckland. Although the intensity of light has been shown to impact on the entraining properties of light pulses under experimental conditions (Boivin et al. [1996](#CIT0003); Zeitzer et al. [2005](#CIT0044); Duffy and Czeisler [2009](#CIT0009)), a saturation effect on shifting the phase of the melatonin rhythm has been reported above approximately 1000 lux (Zeitzer et al. [2000](#CIT0043)), equivalent to approximately 7.9 W/m^2^ (based on the approximation that 1 lux = 0.0079 W/m^2^ for solar irradiance). Considering the lowest average light intensity reported in this study was 19.43 W/m^2^, and therefore well above saturation intensities, it is perhaps unsurprising that no effect of light intensity emerged. Instead, in our population, it appears that the association between chronotype and geographical location is due to the timing of sunset.
This study reported on a large sample of over 6 000 university students collected in term time during spring and autumn. Such numbers help mitigate the limitation of a cross-sectional assessment of chronotype. However, longitudinal studies are needed to determine precisely how an individual's chronotype changes with environmental light levels and age. Although based on self-reported sleep timings, the MCTQ is a validated measure of chronotype (Kantermann et al. [2015](#CIT0015)). The reliability of self-reported time spent outside as a proxy for light dose is less certain. The amount and type of environmental light exposure will be influenced by various factors including photoperiod, weather conditions and the level of urbanisation. Although these have been taken into consideration in this study as much as possible (weather conditions accounted for in measures of daily irradiance and photoperiod in proportional assessment of daylight students awake for), objective assessment of time spent outside and light monitoring need to be undertaken to define when individuals go outside and the nature of their light exposure (inside vs. outside). Furthermore, exposure to artificial light was not addressed in this study, which of course will have an added impact on circadian physiology. Of particular interest in this young student population is the impact of light-emitting devices on sleep and the circadian clock. Although such devices have been found to impact sleep and circadian timing (Cajochen et al. [2011](#CIT0004); Chang et al. [2015](#CIT0005)), the findings are mixed (Heath et al. [2014](#CIT0012); Rangtell et al. [2016](#CIT0023)) and these changes are often small; thus, the real-world significance of these findings remains unclear (Zeitzer [2015](#CIT0042)). The findings from this study, however, emphasise that environmental evening light exposure may need to be tailored for different populations. With the rapid growth in the diversity of energy-efficient light-emitting devices, robust, evidence-based advice is needed to ensure that individuals get the right kind of light at the right time of day to reinforce robust entrainment of the sleep--wake cycle.
In conclusion, we report that in this young adult university student population, time spent outside is associated with a later chronotype. This seems to be linked to the fact that this population spends more time outside in the evening and that dusk light exposure will have a phase delaying effect upon their circadian biology. Moreover, we found that the closer students lived to the equator the earlier their chronotype. Significantly, this also appears to be associated with the timing of sunset rather than sunrise. Collectively, our results emphasise the fact that the age and occupation of individuals will likely impact profoundly upon the timing of their light exposure and hence their phase of entrainment. Moreover, this work highlights the need for future longitudinal studies that will define these relationships with greater precision.
Declaration of interest {#S0008}
=======================
KP, TR, LW, MM, LF, MG, DR, GW and KW declare no conflicts of interest. RGF is in receipt of funding from Circadian Therapeutics. MG is working as a consultant for Philips Sleep & Respiratory care. SMWR reports that he has served as a consultant through his institution to Vanda Pharmaceuticals, Philips Respironics, EdanSafe, The Australian Workers' Union, National Transport Commission, and Transport Accident Commission, and has through his institution received research grants and/or unrestricted educational grants from Vanda Pharmaceuticals, Takeda Pharmaceuticals North America, Philips Lighting, Philips Respironics, Cephalon, and ResMed Foundation, and reimbursements for conference travel expenses from Vanda Pharmaceuticals. His institution has received equipment donations or other support from Optalert™, Compumedics, and Tyco Healthcare. He has also served as an expert witness and/or consultant to shift work organizations. SMWR also serves as a Program Leader in the Cooperative Research Centre for Alertness, Safety and Productivity.
TLS reports her institution has received equipment donations or other support from Philips Lighting, Philips Respironics, Optalert™ and Compumedics. TLS serves as a Project Leader in the Cooperative Research Centre for Alertness, Safety and Productivity. The study was partly supported by the EU 6th Framework Integrated Project 0187241 (EUCLOCK), the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre based at Oxford University Hospitals NHS Trust, Oxford University (A90305 and A92181 to KW and RGF), and the Wellcome Trust (Investigator award, 106174/Z/14/Z to RGF and Strategic award for the SCNi, 098461/Z/12/Z). LW is MINES ParisTech personnel. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, or the Department of Health.
Supplemental material {#S0009}
=====================
Supplemental data for this article can be accessed [here](https://doi.org/10.1080/07420528.2018.1482556).
###### Supplemental Material
[^1]: Color versions of one or more of the figures in the article can be found online at [www.tandfonline.com/icbi](http://www.tandfonline.com/icbi).
[^2]: Mean and standard deviation shown.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
Chargaff\'s first parity rule based on the nucleotide composition of double-stranded DNA states that the complementary nucleotides have the same abundance values.^[@DSP021C1],[@DSP021C2]^ This is explained by the DNA double-helix model in which A pairs only with T and G pairs only with C.^[@DSP021C3]^ Chargaff and his colleagues^[@DSP021C4],[@DSP021C5]^ came with a similar observation of compositional relationship between the complementary nucleotides even within individual DNA strands of bacterial chromosomes. In the post-genomic era, this intra-strand relationship between the complementary nucleotides is observed in double-stranded genomes of viruses, bacteria, archaea and eukaryotes, which is known as Chargaff\'s second parity rule or intra-strand parity (ISP).^[@DSP021C2]^ There is no such defined rule to describe ISP in chromosomes like the base-pairing rule in Chargaff\'s first parity. ISP is also observed between the complementary oligonucleotides in chromosomes,^[@DSP021C6]--[@DSP021C9]^ which has been attributed to genome-wide large-scale inversion, inversion transposition^[@DSP021C10]^ and coding sequence compositional symmetry between the strands.^[@DSP021C9]^ Violation of ISP is observed with respect to organellar (mitochondria and plastids) genomes of some organisms, single-stranded viral genomes or any RNA genome.^[@DSP021C11]--[@DSP021C13]^
Theoretically, under no strand bias in terms of mutation and selection, the base complementary relationship easily explains the presence of ISP in chromosomes.^[@DSP021C14],[@DSP021C15]^ However, several evidences now prove that both the strands are not identical in terms of mutation/selection.^[@DSP021C16]^ This results into violation of ISP in sub-chromosomal regions. Longer the sub-chromosomal region, smaller is the violation of ISP observed.^[@DSP021C17]^ The mechanisms that are responsible to cause violation are defined under three categories.^[@DSP021C18]^ First, DNA replication: leading strand (LeS) is found to be composed of more K nucleotides (G and T) than the complementary M (A and C) nucleotides and the reverse holds true for the lagging strand (LaS).^[@DSP021C19]^ This is due to the fact that the LeS which functions as the template for Okazaki fragment synthesis (functions as template for LaS) remains exposed more as single-stranded than the LaS (functions as template for LeS) during replication that results into higher deamination of the cytosine residues^[@DSP021C20],[@DSP021C21]^ in LeS (cytosine gets deaminated 140 times faster in ssDNA than in dsDNA^[@DSP021C22]^). In addition, the influence of Okazaki fragments and the sliding DNA clamp proteins associated with the synthesis of LaS create functional asymmetry of the mismatch repairing system on DNA.^[@DSP021C23]^ Second, transcription: genes are preferentially located in the LeS than in the LaS to avoid head on collision between the machineries of replication and transcription.^[@DSP021C24]^ During transcription, the non-template strand remains more exposed as single-stranded than the template strand, which causes asymmetry in cytosine deamination between the strands.^[@DSP021C22]^ The transcription-coupled repair system also acts only upon the template strand and thereby contributes to the strand asymmetry.^[@DSP021C25]^ Third, translation: uses of synonymous codons are influenced by differential abundance of tRNA molecules which results into the differential abundance of complementary nucleotides at the third position of family box codons. This causes parity violation.^[@DSP021C14]^ In spite of these factors favoring violations of the parity in chromosomes, ISP is observed in an entire chromosome due to the cancellation effect of the local violations in opposite directions.^[@DSP021C14]^
Evolutionary biologists are more interested to understand the role of mutation and/or selection in the violation of ISP by analyzing the weakly selected or selectively neutral regions (third position of family box codons and non-coding regions) in chromosomes.^[@DSP021C14],[@DSP021C26]^ Whether any specific feature(s) is/are associated with chromosomes exhibiting ISP is yet to be understood. Shioiri and Takahata^[@DSP021C27]^ studied ISP by finding out the total AT skew (ATS) and GC skew (GCS) in the chromosomes of several bacteria. In their study, out of 36 bacterial chromosomes, *Xylella fastidiosa* exhibited maximum ATS and GCS. They observed variable ATS/GCS among chromosomes of different strains of a species as well as chromosomes within a bacterial cell. They also observed ATS and GCS may be different from each other within a chromosome. Since, they did not do any statistical analysis of the skew, the significance of the variability observed among chromosomes was not discussed by them. The usual statistical tool used to find out ISP in chromosomes is a correlation analysis of oligonucleotides abundance described by Prabhu.^[@DSP021C6]^ The ISP study between the complimentary mononucleotides is important because it has been proven that oligonucleotide parity and mononucleotide parity are independent.^[@DSP021C8]^ Baisnée *et al*.^[@DSP021C8]^ studied parity in chromosomes by measuring the S^1^ index which is defined as the sum of the absolute values of the differences between complementary oligonucleotides (*n* mer) frequencies (*n* varies from 1 to 9 mer). Both these methods do not measure the statistical significance of differences between the abundance values of a mono/oligonucleotide and its reverse complement. For example, if a chromosome carries significant similarity between the abundance values of A and T but carries significant difference between the abundance values of G and C, this will not be identified separately. Similarly, the above methods are unable to find out parity violations in chromosomes with respect to the abundance values of an oligonucleotide and its reverse complement. We have developed a methodology here that can independently study ISP between S nucleotides (any oligonucleotide and its reverse complement) as well as between W nucleotides using the abundance values of mononucleotides. We use the well-known Kolmogorov--Smirnov (KS) test to study the frequency distribution of the compositional abundance values of the mononucleotides in a chromosome sequence, which gives the statistical significance of the similarity between the distributions of complementary nucleotides. This we called as intra-strand frequency distribution parity (ISFDP), which has been used here to study the chromosomes of bacteria, archaea and eukaryotes.
Materials and methods {#s2}
=====================
Frequency distribution calculation {#s2a}
----------------------------------
Chromosome sequences of different bacteria, archaea and eukaryotes (Tables [1](#DSP021TB1){ref-type="table"}[](#DSP021TB2){ref-type="table"}--[3](#DSP021TB3){ref-type="table"}) were obtained from the genome information broker, DDBJ site ([www.gib.genes.nig.ac.jp](www.gib.genes.nig.ac.jp)). Bacterial chromosomes were chosen randomly from the database starting the genus name from A to Z. Chromosome sequences of different strains belonging to the same species in the case of bacteria were taken in several cases to do the intra-species comparison. Each chromosome sequence was divided into smaller-size sequences of 1000 nucleotides each starting from the beginning, and the abundance value of the four nucleotides was determined using the computer program (developed for this study). The distribution of the abundance values of complementary nucleotides in different fragments were analyzed by the KS non-parametric test using XLSTAT program^[@DSP021C28]--[@DSP021C30]^ (Kovach Computing Services, Anglesey, Wales). *H*~0~: distribution patterns of any two nucleotides/oligonucleotides in a chromosome are similar; *H*~A~: there is a difference between the two distributions. Owing to the large sample size, similarity was considered at the *P*-value of \>0.01, weak similarity was considered at the *P*-value between 0.01 and 10^−4^, and the value of \<10^−4^ was considered as strong violation similarity. Group-frequency distributions of the abundance values were plotted to observe the frequency-distribution parity. In the case of the di- and trinucleotides, the abundance values were determined using a different computer program (developed here for this study) in the segments for the 16 dinucleotides and 64 trinucleotides. The analysis was done as described for the mononucleotides earlier.
Angular replication asymmetry of the chromosomes was calculated with the help of the information on *ori* (origin) and *ter* (termination) cited in the websites (<http://www.cbs.dtu.dk/services/GenomeAtlas/suppl/origin/> and <http://pbil.univ-lyon1.fr/software/Oriloc/oriloc.html>). The chromosomal region starting from *ori* to *ter* was considered as the leading region in the Watson strand (Ws) and the remaining portion of the chromosome as the lagging region. For a circular chromosome, the angular replication asymmetry was calculated as the amount of angular distance of leading region deviating from 180°.
Proportionate distribution of forward- and reverse-encoded sequences in a DNA strand {#s2b}
------------------------------------------------------------------------------------
From the DDBJ site, only coding sequences were downloaded. A continuous stretch of the nucleotide sequence was made from all the sequences by removing the gene names. This resembled a DNA strand only composed of forward-encoded sequences. Frequency distribution analysis was done on this. In another approach, 50% of the above strand was made reverse complement by *in silico* followed by joining with the rest. This resembled a DNA strand composed of 50% forward-encoded and 50% reverse-encoded sequences. Frequency-distribution study was carried out as described above.
Identification of leading and LaS region {#s2c}
----------------------------------------
ATS and GCS analyses of the chromosome sequences were done as described earlier.^[@DSP021C21]^ This was used to find out the tentative leading and lagging portions in a DNA strand.
Relative proportion of coding sequence distribution {#s2d}
---------------------------------------------------
This was found out by deducting ORF numbers between Ws (top strand) and Crick strand (Cs: bottom strand) followed by dividing that with the total number of ORFs. Gene orientation information was obtained from the website (<http://cmr.jcvi.org/tigr-scripts/CMR/CmrHomePage.cgi>).
Results {#s3}
=======
ISFDP in chromosomes of bacteria {#s3a}
--------------------------------
In this study, a total of 112 bacterial chromosomes were considered, which includes different lineages of bacteria such as protobacteria, cyanobacteria, firmicutes, actinobacteria etc. Samples from each group were taken randomly. The bacteria included in the sample comprised a GC% variation from a minimum of 28% to a maximum of 75% and chromosome size variation from 580 kb to a maximum of 9105 kb. We have studied the frequency distributions of the abundance values of mononucleotides in the uniform sub-chromosomal length of 1000 nucleotides. A collective analysis of the nucleotide abundance values from all the segments of a chromosome was done by frequency distribution smooth curves using Microsoft Excel, and the similarity of the distributions of two complementary nucleotides was tested using the KS test (XL-Stat; <http://www.xlstat.com/en/download>). Figure [1](#DSP021F1){ref-type="fig"}A(i), B(i), C(i), D(i) and E(i) represents the smooth curves of frequency distributions of nucleotides in chromosomes *Campylobacter jejuni* RM1221 (30.31%), *Escherichia coli* K12 MG1655 (50.79%), *Xanthomonas campestris* pv. *campestris* (*Xcc*; 65.07%), *X. fastidiosa* 9a5c (52.68%) and *X. fastidiosa* Temecula (51.78%). Smooth curves of complementary nucleotides overlap with each other in the first three chromosomes, whereas those of non-complementary ones do not. In the fourth chromosome, none of the curves overlap with each other. In *E. coli* chromosome \[Fig. [1](#DSP021F1){ref-type="fig"}B(i)\], all the four smooth frequency curves are close to each other due to the closeness of the abundance values of the nucleotides, whereas in the graphs of *C. jejuni* and *Xcc*, the smooth frequency curves of W (A and T) and S (G and C) nucleotides are distinctly separated as GC% the chromosome are toward both extremes. The distribution was studied by the KS test and the results of the four chromosomes are shown in Fig. [1](#DSP021F1){ref-type="fig"}A(ii, iii), B(ii, iii), C(ii, iii), D(ii, iii) and E(ii, iii). The graphs generated by the KS test suggest the complete overlapping between the complementary nucleotides in the chromosomes except the one of *X. fastidiosa* strain, which is in concordant with the smooth frequency curve. The distributional similarity between the complementary nucleotides is called as ISFDP. A total of 112 bacterial, 49 archaea and 18 eukaryotic chromosomes (Tables [1](#DSP021TB1){ref-type="table"}[](#DSP021TB2){ref-type="table"}--[3](#DSP021TB3){ref-type="table"}) were analyzed by the KS test to study ISFDP. The *P*-values between the A and T distributions as well as between the G and C distributions are given in Tables [1](#DSP021TB1){ref-type="table"}[](#DSP021TB2){ref-type="table"}--[3](#DSP021TB3){ref-type="table"}.
![(**A**--**E**) Frequency distribution of nucleotides in chromosomes. Smooth curves present the group-frequency distribution of the four nucleotides a (square), t (asterisk), g (triangle) and c (rhombus). The *X*-axis represents the abundance values of the nucleotide spanning a range, whereas the *Y*-axis represents the frequency of the abundance values. In (A), the chromosome is AT rich; in (B), the chromosome is composed of similar AT and GC and in (C), the chromosome is GC rich. This is also evident from the group-frequency distribution curve. The smooth frequency curves of complementary nucleotides in these chromosomes are overlapping with each other. The KS test is shown for S and W nucleotides separately adjacent to the figures, respectively \[a(ii, iii)--e(ii, iii)\]. The KS test is in concordance with the curve obtained by smoothing group-frequency distribution. In (D) and (E), the group-frequency distribution for the chromosomes of two strains of *X. fastidiosa* is shown. In 9a5c strain chromosome, the smooth frequency curve between the complementary nucleotides does not overlap which is also suggested by the KS test. However, in Temecula 1 strain chromosome, the parity is maintained.](dsp02101){#DSP021F1}
######
ISFDP analysis in bacterial chromosomes
Serial number Strain name Size (kb) GC% KS (W) KS (S) \|(∑A − ∑T)\|/(∑A + ∑T) \|(∑G − ∑C)\|/(∑G + ∑C) Bacterial group TB (°)
--------------- -------------------------------------------------------------- ----------- ------- -------------- -------------- ------------------------- ------------------------- ------------------ --------
1 *Acinetobacter* sp. ADP1 3598 40.43 0.745 *0.006* 0.00068 0.00484 G-Proteobacteria 7.07
2 *Actinobacillus pleuropneumoniae* L20 serotype 5b 2274 41.3 0.436 0.819 0.00187 0.00109 NA
3 *Actinobacillus succinogenes* 130Z 2319 44.91 0.312 0.291 0.00232 0.00291
4 *Aeromonas hydrophila* subsp. *hydrophila* ATCC 7966 4744 61.55 0.88 0.19 0.00141 0.00139
5 *Aeromonas salmonicida* subsp. *salmonicida* A449 4702 58.51 0.04 0.959 0.00215 0.00073
6 *Agrobacterium tumefaciens* C58 (circular chromosome) 2841 59.38 **\<0.0001** **\<0.0001** 0.00694 0.00967 A-Proteobacteria 7.37
7 *Alkaliphilus oremlandii* OhILAs 3123 36.26 **\<0.0001** **\<0.0001** 0.00615 0.01324 Firmicutes NA
8 *Anaeromyxobacter dehalogenans* 2CP-C 5013 74.9 0.077 *0.001* 0.00476 0.00249 D-Proteobacteria 70.57
9 *Anaeromyxobacter* sp. Fw109-5 5277 73.53 0.712 *0.008* 0.00073 0.00216 7.48
10 *Bacillus anthracis* Ames 5227 35.38 *0.004* **\<0.0001** 0.00215 0.00581 Firmicutes NA
11 *Bacillus anthracis* \'Ames Ancestor\' 5227 35.38 *0.003* **\<0.0001** 0.00215 0.00582 7.48
12 *Bacillus anthracis* Sterne 5228 35.38 *0.008* **\<0.0001** 0.00221 0.00588 7.46
13 *Bacillus subtilis* 4214 43.52 0.219 0.234 0.00212 0.00224 13.69
14 *Bacillus thuringiensis* Al Hakam 5257 35.43 0.123 *0.002* 0.00042 0.00081 NA
15 *Bacillus thuringiensis* serovar konkukian 97-27 5237 35.41 0.015 **\<0.0001** 0.00194 0.00438 3.98
16 *Bordetella parapertussis* 12822 4773 68.1 0.433 **\<0.0001** 0.00247 0.00776 B-Proteobacteria 37.01
17 *Bordetella pertussis* Tohama 1 4086 67.72 0.861 **\<0.0001** 0.00022 0.00390 71.28
18 *Bradyrhizobium japonicum* USDA 110 9105 64.06 0.512 0.31 0.00070 0.00038 A-Proteobacteria 7.07
19 *Bradyrhizobium* sp. BTAi1 8264 64.92 0.381 0.01 0.00100 0.00163 NA
20 *Brucella melitensis* 16M 1177 57.35 0.472 *0.008* 0.00227 0.00312
21 *Campylobacter concisus* 13826 2052 39.43 0.033 0.048 0.00038 0.00599 E-Proteobacteria
22 *Campylobacter curvus* 525.92 1971 44.54 0.028 0.752 0.00745 0.00282
23 *Campylobacter jejuni* RM1221 1777 30.31 0.574 0.23 0.00330 0.00436 8.69
24 *Campylobacter jejuni* subsp. *jejuni* 81116 1628 30.54 0.491 0.029 0.00250 0.00613 NA
25 *Campylobacter jejuni* subsp. *jejuni* NCTC 11168 1641 30.55 0.067 0.132 0.00296 0.00457 10.25
26 Candidatus *Desulfococcus oleovorans* Hxd3 3944 56.17 0.258 0.133 0.00199 0.00157 Firmicutes NA
27 *Caulobacter crescentus* CB15 4016 67.22 0.042 0.171 0.00396 0.00188 A-Proteobacteria 8.56
28 *Chlamydia muridarum* Nigg 1072 40.34 0.221 0.853 0.00107 0.00337 Chlamydiae 1.17
29 *Chlamydia trachomatis* AHAR-13 1044 41.31 0.228 0.284 0.00230 0.00059 1.30
30 *Chlamydophila abortus* S263 1144 39.87 0.534 *0.002* 0.00065 0.00361 0.57
31 *Coxiella burnetii* Dugway 7E9-12 2158 42.44 *0.004* *0.001* 0.00592 0.00573 G-Proteobacteria NA
32 *Coxiella burnetii* RSA 493 1995 42.66 0.014 0.467 0.00198 0.00029 31.15
33 *Desulfovibrio desulfuricans* G20 3730 57.84 0.59 *0.001* 0.00189 0.00322 Firmicutes 10.70
34 *Desulfovibrio vulgaris* subsp. *vulgaris* DP4 3462 63.01 0.3 0.159 0.00152 0.00106 D-Proteobacteria NA
35 *Desulfovibrio vulgaris* subsp. *vulgaris* Hildenborough 3570 63.14 0.557 0.082 0.00143 0.00024 4.78
36 *Enterobacter sakazakii* ATCC BAA-894 4368 56.77 0.167 0.388 0.00359 0.00044 G-Proteobacteria NA
37 *Enterobacter* sp. 638 4518 52.98 0.645 0.39 0.00169 0.00163 NA
38 *Escherichia coli* 536 4938 50.52 0.714 0.084 0.00062 0.00328 7.40
39 *Escherichia coli* APEC O1 5082 50.55 0.779 0.576 0.00032 0.00070 NA
40 *Escherichia coli* CFT073 5231 50.48 0.112 0.92 0.00173 0.00080 5.66
41 *Escherichia coli* E24377A 4979 50.62 0.736 0.128 0.00205 0.00212 NA
42 *Escherichia coli* HS 4643 50.82 0.328 0.469 0.00151 0.00207
43 *Escherichia coli* K12 MG1655 4639 50.79 0.732 0.587 0.00054 0.00113 4.28
44 *Escherichia coli* UTI89 5065 50.6 0.51 0.237 0.00076 0.00203 3.70
45 *Escherichia coli* W3110 4646 50.8 0.873 0.729 0.00073 0.00091 12.64
46 *Frankia alni* ACN14A chromosome 7497 72.82 0.463 0.036 0.00141 0.00139 Actinobacteria NA
47 *Frankia* sp. CcI3 5433 70.08 0.808 0.662 0.00129 0.00017
48 *Haemophilus influenzae* 86-028NP 1914 38.16 0.886 0.654 0.00089 0.00044 G-Proteobacteria
49 *Haemophilus influenzae* PittEE 1813 38.04 0.544 0.038 0.00054 0.00317
50 *Haemophilus influenzae* PittGG 1887 38.01 0.125 **\<0.0001** 0.00005 0.01016
51 *Haemophilus influenzae* Rd KW20 1830 38.15 0.154 *0.004* 0.00298 0.00472 46.61
52 *Helicobacter acinonychis* Sheeba 1553 38.18 *0* 0.596 0.00869 0.00164 E-Proteobacteria NA
53 *Helicobacter hepaticus* ATCC 51449 1799 35.93 0.161 **\<0.0001** 0.00499 0.01518 46.54
54 *Helicobacter pylori* J99 1643 39.19 0.246 0.256 0.00259 0.00510 10.97
55 *Lactobacillus acidophilus* NCFM 1993 34.72 0.382 **\<0.0001** 0.00066 0.01644 Firmicutes 19.54
56 *Lactobacillus brevis* ATCC 367 2291 46.22 0.023 **\<0.0001** 0.00271 0.02882 NA
57 *Lactobacillus delbrueckii* subsp. *bulgaricus* ATCC BAA-365 1856 49.69 0.491 0.264 0.00201 0.00087
58 *Lactobacillus reuteri* F275 1999 38.87 *0.001* **\<0.0001** 0.00122 0.01040
59 *Lactococcus lactis* subsp. *cremoris* MG1363 2529 35.75 0.233 0.056 0.00352 0.00524
60 *Lactococcus lactis* subsp. *cremoris* SK11 2438 35.86 0.399 0.521 0.00147 0.00136
61 *Magnetococcus* sp. MC-1 4719 54.17 *0.001* **\<0.0001** 0.00490 0.01198 Magnetococcus
62 *Magnetospirillum magneticum* AMB-1 4967 65.09 0.031 **\<0.0001** 0.00339 0.00288 A-Proteobacteria 2.14
63 *Methylobacillus flagellatus* KT 2971 55.72 0.03 0.916 0.00226 0.00135 B-Proteobacteria 10.57
64 *Methylococcus capsulatus* Bath 3304 63.59 0.145 *0.004* 0.00150 0.00287 G-Proteobacteria NA
65 *Mycobacterium leprae* TN 3268 57.8 *0.003* **\<0.0001** 0.00378 0.00609 Actinobacteria 7.04
66 *Mycobacterium* sp. KMS 5737 68.44 0.389 0.478 0.00030 0.00060 NA
67 *Mycobacterium tuberculosis* F11 4424 65.62 0.366 *0.007* 0.00006 0.00198
68 *Mycobacterium ulcerans* Agy99 5631 65.47 **\<0.0001** **\<0.0001** 0.00433 0.00374
69 *Mycoplasma gallisepticum* R 996 31.45 0.18 0.615 0.00626 0.00021 Tenericutes 9.32
70 *Mycoplasma genitalium* G37 580 31.69 *0* 0.148 0.01219 0.00433 3.75
71 *Mycoplasma hyopneumoniae* J 897 28.52 0.033 0.599 0.01020 0.00067 NA
72 *Mycoplasma pneumoniae* M129 816 40.01 *0.001* 0.115 0.01767 0.00243 16.23
73 *Neisseria gonorrhoeae* FA 1090 2153 52.69 0.07 0.033 0.00601 0.00144 B-Proteobacteria 9.20
74 *Neisseria meningitidis* MC58 2273 51.52 0.695 *0.004* 0.00135 0.00806 NA
75 *Nitrobacter hamburgensis* X14 4406 61.72 0.332 0.53 0.00112 0.00041 A-Proteobacteria
76 *Nitrobacter winogradskyi* Nb-255 3402 62.05 0.011 **\<0.0001** 0.00323 0.00294 37.15
77 *Nitrosococcus oceani* ATCC 19707 3481 50.32 0.02 0.056 0.00530 0.00243 G-Proteobacteria 8.39
78 *Nitrosomonas eutropha* C91 2661 48.49 0.992 0.318 0.00043 0.00162 B-Proteobacteria NA
79 *Nostoc* sp. PCC 7120 6413 41.35 0.134 0.857 0.00129 0.00162 Cyanobacteria
80 *Pseudomonas entomophila* L48 chromosome 5888 64.16 0.657 0.251 0.00078 0.00173 G-Proteobacteria 1.99
81 *Pseudomonas fluorescens* PfO-1 6438 60.52 *0.003* 0.028 0.00443 0.00222 3.18
82 *Pseudomonas putida* F1 5959 61.86 0.602 0.013 0.00113 0.00187 36.81
83 *Ralstonia eutropha* H16 2912 66.78 0.238 0.47 0.00483 0.00023 B-Proteobacteria NA
84 *Ralstonia solanacearum* GMI1000 chromosome 3716 67.04 0.056 **\<0.0001** 0.00636 0.00581 22.40
85 *Rhizobium etli* CFN 42 4381 61.27 0.107 **\<0.0001** 0.00175 0.01177 A-Proteobacteria 17.65
86 *Rhizobium leguminosarum* bv. *viciae* 3841 5057 61.09 *0.001* **\<0.0001** 0.00363 0.01196 NA
87 *Rickettsia bellii* RML369-C 1522 31.65 *0* **\<0.0001** 0.00859 0.01514 26.08
88 *Rickettsia conorii* Malish 7 1268 32.44 0.584 0.052 0.00294 0.00634 16.28
89 *Rickettsia rickettsii* \'Sheila Smith\' 1257 32.47 0.575 *0.002* 0.00182 0.00767 NA
90 *Rickettsia typhi* Wilmington 1111 28.92 0.919 *0.007* 0.00020 0.01395 26.15
91 *Salmonella enterica* subsp. *enterica* serovar Typhi CT18 4809 52.09 0.267 0.043 0.00151 0.00152 G-Proteobacteria 9.85
92 *Salmonella typhimurium* LT2 4857 52.22 0.89 0.585 0.00043 0.00008 3.58
93 *Shigella boydii* Sb227 4519 51.21 0.571 *0.001* 0.00022 0.00249 11.05
94 *Shigella flexneri* 58401 4574 50.92 0.48 0.268 0.00147 0.00214 NA
95 *Staphylococcus aureus* RF122 2742 32.78 0.788 0.427 0.00130 0.00247 Firmicutes 0.10
96 *Staphylococcus epidermidis* ATCC 12228 2499 32.1 **\<0.0001** **\<0.0001** 0.01246 0.01087 21.12
97 *Staphylococcus haemolyticus* JCSC1435 2685 32.79 *0.001* *0* 0.00584 0.00643 NA
98 *Streptococcus mutans* UA159 2030 36.83 0.111 0.046 0.00403 0.00679
99 *Streptococcus pyogenes* MGAS2096 1860 38.73 0.619 0.15 0.00133 0.00154 3.71
100 *Streptococcus thermophilus* CNRZ1066 1796 39.08 0.05 0.863 0.00537 0.00459 2.63
101 *Streptomyces coelicolor* A3(2) 8667 72.12 *0.001* 0.037 0.00394 0.00134 Actinobacteria NA
102 *Thermotoga maritima* MSB8 1860 46.25 0.171 **\<0.0001** 0.00344 0.01548 Thermotogae 39.15
103 *Thermotoga petrophila* RKU-1 1824 46.09 0.733 **\<0.0001** 0.00013 0.01687 NA
104 *Thiobacillus denitrificans* ATCC 25259 2909 66.07 0.962 0.086 0.00027 0.00059 B-Proteobacteria 5.70
105 *Vibrio cholerae* O395 3024 47.78 **\<0.0001** 0.069 0.00514 0.00105 G-Proteobacteria NA
106 *Vibrio fischeri* ES114 1332 37.03 *0.001* 0.037 0.00994 0.00491
107 *Xanthomonas campestris* pv. *campestris* ATCC 33913 5076 65.07 0.196 0.719 0.00302 0.00038
108 *Xanthomonas oryzae* pv. *oryzae* KACC 10331 4941 63.69 0.87 0.499 0.00104 0.00065
109 *Xylella fastidiosa* 9a5c 2679 52.68 **\<0.0001** **\<0.0001** 0.04727 0.05291 62.97
110 *Xylella fastidiosa* Temecula 1 2519 51.78 0.044 *0* 0.00379 0.01093 6.44
111 *Yersinia pestis* CO92 4653 47.64 0.649 *0.001* 0.00090 0.00520 NA
112 *Yersinia pseudotuberculosis* IP32953 4744 47.61 0.969 *0.001* 0.00124 0.00496
TB, termination bias. Chromosomes of bacteria analyzed in this study. The KS test for significance between the frequency distribution of complementary nucleotide values are given as: KS (W) between A and T and KS (S) between G and C. In bacteria, archaea and eukaryotes, *P*-values of \<10^−4^ (strong violation of ISFDP) are shown in bold and *P*-values of \<0.01 but≥10^−4^ (weak violation of ISFDP) are shown in italics. The *P*-value between 10^−4^ and 10^−3^ is shown as 0.000. Relative absolute abundance value difference between the complementary nucleotides is given by \|(∑A − ∑T)\|/(∑A + ∑T) and \|(∑G − ∑C)\|/(∑G + ∑C) for ATS and GCS, respectively. In chromosome of *X. fastidiosa* 9a5c, the GCS/ATS value is highest suggesting that the difference between the abundance values of complementary nucleotides is high. The *P*-value by the KS test is in concordant with the ATS/GCS suggesting that the abundance difference can be represented by the frequency distribution study of the nucleotides. Similar relation is also observed in other chromosomes.
######
ISFDP analysis in archaea chromosomes
Serial number Strain name Size (kb) GC% KS (W) KS (S) \|(∑A − ∑T)\|/(∑A + ∑T) \|(∑G − ∑C)\|/(∑G + ∑C) Archaea group
--------------- -------------------------------------------------- ----------- ------- -------------- -------------- ------------------------- ------------------------- ---------------
1 *Aeropyrum pernix* K1 1670 56.3 *0.001* 0.025 0.01292 0.00695 Crenarchaeota
2 *Archaeoglobus fulgidus* DSM 4304 2179 48.5 0.037 0.093 0.00365 0.00350 Euryarchaeota
3 *Caldivirga maquilingensis* IC-167 2078 43.08 0.586 0.643 0.00146 0.00104 Crenarchaeota
4 Candidatus *Methanoregula boonei* 6A8 2543 54.51 0.058 0.191 0.00311 0.00108 Euryarchaeota
5 *Cenarchaeum symbiosum* 2046 57.34 0.101 *0.006* 0.00574 0.00161 Crenarchaeota
6 *Haloarcula marismortui* ATCC 43049 chromosome 1 3132 62.35 0.252 0.905 0.01075 0.00024 Euryarchaeota
7 *Halobacterium* sp. NRC-1 2015 67.88 0.862 0.313 0.00056 0.00151
8 *Haloquadratum walsbyi* DSM 16790 3133 47.85 0.578 0.027 0.00160 0.00523
9 *Hyperthermus butylicus* DSM 5456 1668 53.7 0.019 0.908 0.00531 0.00100 Crenarchaeota
10 *Ignicoccus hospitalis* KIN4/I 1298 56.5 0.118 0.901 0.00199 0.00014
11 *Metallosphaera sedula* DSM 5348 2192 46.21 *0* **\<0.0001** 0.00668 0.01423 Crenarchaeota
12 *Methanobrevibacter smithii* ATCC 35061 1854 31.02 **\<0.0001** **\<0.0001** 0.02048 0.03768 Euryarchaeota
13 *Methanocaldococcus jannaschii* DSM 2661 1666 31.4 0.132 0.031 0.00450 0.01128
14 *Methanococcoides burtonii* DSM 6242 2576 40.74 0.078 *0.002* 0.00266 0.00845
15 *Methanococcus aeolicus* Nankai-3 1570 30.02 0.218 0.52 0.00399 0.00063
16 *Methanococcus maripaludis* C5 1781 32.99 *0.001* 0.065 0.00846 0.00454
17 *Methanococcus maripaludis* C6 1745 33.4 0.045 0.045 0.00553 0.00224
18 *Methanococcus maripaludis* C7 1773 33.27 0.256 0.784 0.00430 0.00088
19 *Methanococcus maripaludis* S2 1662 33.08 0.021 0.08 0.00619 0.00259
20 *Methanococcus vannielii* SB 1721 31.31 0.505 0.519 0.00364 0.00400
21 *Methanocorpusculum labreanum* Z 1806 49.97 0.606 0.05 0.00097 0.00404
22 *Methanoculleus marisnigri* JR1 2479 62.04 0.816 0.745 0.00234 0.00000
23 *Methanopyrus kandleri* AV19 1696 61.22 0.556 0.032 0.00230 0.00471
24 *Methanosaeta thermophila* PT 1880 53.53 0.673 *0.004* 0.00018 0.00595
25 *Methanosarcina acetivorans* C2A 5752 42.67 **\<0.0001** 0.839 0.00628 0.00083
26 *Methanosarcina barkeri* Fusaro 4838 39.27 **\<0.0001** *0.003* 0.00475 0.00391
27 *Methanosarcina mazei* Goe1 4097 41.47 0.252 0.812 0.00212 0.00079
28 *Methanosphaera stadtmanae* DSM 3091 1768 27.62 *0.002* 0.275 0.00897 0.00652
29 *Methanospirillum hungatei* JF-1 3545 45.14 **\<0.0001** 0.015 0.00951 0.00411
30 *Methanothermobacter thermautotrophicus* Delta H 1752 49.52 0.022 0.114 0.00566 0.00166
31 *Nanoarchaeum equitans* Kin4-M 491 31.55 0.549 0.177 0.00000 0.00127 Nanoarchaeota
32 *Natronomonas pharaonis* DSM 2160 2596 63.42 0.473 0.228 0.00174 0.00091 Euryarchaeota
33 *Nitrosopumilus maritimus* SCM1 1646 31.15 **\<0.0001** *0.002* 0.00921 0.00855 Crenarchaeota
34 *Picrophilus torridus* DSM 9790 1546 35.96 0.296 *0.008* 0.00096 0.00793 Euryarchaeota
35 *Pyrobaculum aerophilum* IM2 2223 51.34 *0.001* **\<0.0001** 0.00727 0.01022 Crenarchaeota
36 *Pyrobaculum arsenaticum* DSM 13514 2122 54.98 0.795 0.431 0.00138 0.00316
37 *Pyrobaculum calidifontis* JCM 11548 2010 57.13 0.148 0.337 0.00294 0.00008
38 *Pyrobaculum islandicum* DSM 4184 1827 49.58 0.305 0.436 0.00085 0.00183
39 *Pyrococcus abyssi* 1766 44.69 0.652 0.574 0.00219 0.00342 Euryarchaeota
40 *Pyrococcus furiosus* DSM 3638 1909 40.75 0.754 0.757 0.00004 0.00094
41 *Pyrococcus horikoshii* OT3 1739 41.86 0.133 *0.002* 0.00229 0.01262
42 *Staphylothermus marinus* F1 1571 35.71 **\<0.0001** **\<0.0001** 0.02078 0.02726 Crenarchaeota
43 *Sulfolobus acidocaldarius* DSM 639 2227 36.69 0.413 0.526 0.00309 0.00124
44 *Sulfolobus solfataricus* P2 2993 35.77 *0.007* 0.747 0.00533 0.00241
45 *Sulfolobus tokodaii* 7 2695 32.78 *0.005* 0.029 0.00521 0.00659
46 *Thermococcus kodakarensis* KOD1 2089 51.98 0.062 0.328 0.00418 0.00160 Euryarchaeota
47 *Thermofilum pendens* Hrk 5 1782 57.66 0.014 *0.005* 0.00346 0.00665 Crenarchaeota
48 *Thermoplasma acidophilum* DSM 1728 1565 45.99 0.016 0.016 0.00680 0.00383 Euryarchaeota
49 *Thermoplasma volcanium* GSS1 1585 39.91 0.055 0.361 0.00404 0.00263
Chromosomes of archaea analyzed in this study. The KS test for significance between the frequency distribution of complementary nucleotide values are given as KS (W) between A and T and KS (S) between G and C. In bacteria, archaea and eukaryotes, *P*-values of \<10^−4^ (strong violation of ISFDP) are shown in bold and *P*-values of \<0.01 but ≥10^−4^ (weak violation of ISFDP) are shown in italics. The *P*-value between 10^−4^ and 10^−3^ is shown as 0.000. Relative absolute abundance value difference between the complementary nucleotides is given by \|(∑A − ∑T)\|/(∑A + ∑T)and \|(∑G − ∑C)\|/(∑G + ∑C) for ATS and GCS, respectively. In chromosome of *X. fastidiosa* 9a5c, the GCS/ATS value is highest suggesting the difference between the abundance values of complementary nucleotides is high. The *P*-value by the KS test is in concordant with the ATS/GCS suggesting that the abundance difference can be represented by the frequency distribution study of the nucleotides. Similar relation is also observed in other chromosomes.
######
ISFDP analysis in eukaryotes chromosomes
Serial number Strain name Size (kb) GC% KS (W) KS (S) \|(∑A − ∑T)\|/(∑A + ∑T) \|(∑G − ∑C)\|/(∑G + ∑C) Eukaryotes group
--------------- ------------------------------------------------ ----------- ------- --------- -------------- ------------------------- ------------------------- ------------------
1 *Guillardia theta* nucleomorph chromosome 01 197 25.64 0.411 0.468 0.00080 0.00517 Cryptophyta
2 *Guillardia theta* nucleomorph chromosome 02 181 26.7 0.435 0.35 0.00451 0.00356
3 *Guillardia theta* nucleomorph chromosome 03 175 26.81 0.671 0.403 0.00051 0.00622
4 *Leishmania major* Friedlin chromosome 01 270 62.84 0.055 **\<0.0001** 0.01290 0.02500 Euglenozoa
5 *Plasmodium falciparum* 3D7 chromosome 01 644 20.52 *0.001* 0.69 0.02184 0.01210 Alveolata
6 *Plasmodium falciparum* 3D7 chromosome 05 1344 19.32 *0.006* *0.005* 0.01288 0.01482
7 *Plasmodium falciparum* 3D7 chromosome 11 2036 18.95 0.043 0.027 0.00339 0.00994
8 *Plasmodium falciparum* 3D7 chromosome 12 2272 19.31 0.05 0.677 0.00597 0.00376
9 *Plasmodium falciparum* 3D7 chromosome 13 2733 19.11 0.105 0.266 0.00422 0.00914
10 *Plasmodium falciparum* 3D7 chromosome 14 3292 18.43 0.258 0.062 0.00275 0.00730
11 *Saccharomyces cerevisiae* S288C chromosome 01 231 39.14 0.731 0.088 0.00100 0.01231 Fungi
12 *Saccharomyces cerevisiae* S288C chromosome 04 1532 37.9 0.807 0.379 0.00240 0.00345
13 *Saccharomyces cerevisiae* S288C chromosome 07 1091 38.05 0.285 0.85 0.00136 0.00080
14 *Saccharomyces cerevisiae* S288C chromosome 12 1079 38.44 0.055 0.461 0.00325 0.00173
15 *Saccharomyces cerevisiae* S288C chromosome 15 1092 38.13 0.181 0.64 0.00584 0.00387
16 *Schizosaccharomyces pombe* 972h chromosome 01 5574 36.09 0.4 0.076 0.00073 0.00086
17 *Schizosaccharomyces pombe* 972h chromosome 02 4510 35.92 0.461 0.825 0.00207 0.00039
18 *Schizosaccharomyces pombe* 972h chromosome 03 2453 36.23 0.152 0.012 0.00217 0.00369
Chromosomes of eukaryotes analyzed in this study. The KS test for significance between the frequency distribution of complementary nucleotide values are given as KS (W) between A and T and KS (S) between G and C. In bacteria, archaea and eukaryotes, *P*-values of \<10^−4^ (strong violation of ISFDP) are shown in bold and *P*-values of \<0.01 but ≥10^−4^ (weak violation of ISFDP) are shown in italics. The *P*-value between 10^−4^ and 10^−3^ is shown as 0.000. Relative absolute abundance value difference between the complementary nucleotides is given by \|(∑A − ∑T)\|/(∑A + ∑T) and \|(∑G − ∑C)\|/(∑G + ∑C) for ATS and GCS, respectively. In chromosome of *X. fastidiosa* 9a5c, the GCS/ATS value is highest suggesting that the difference between the abundance values of complementary nucleotides is high. The *P*-value by the KS test is in concordant with the ATS/GCS suggesting that the abundance difference can be represented by the frequency distribution study of the nucleotides. Similar relation is also observed in other chromosomes.
Out of 112 bacterial chromosomes, 60 chromosomes exhibited ISFDP, 16 chromosomes exhibited violation between S nucleotides as well as between W nucleotides, 30 chromosomes exhibited violation only between S nucleotides and 7 chromosomes exhibited violation only between W nucleotides (Table [4](#DSP021TB4){ref-type="table"}). Chromosomes of *Alkaliphilus oremlandii* OhILAs (36.26%), *Agrobacterium tumefaciens* C58 (circular; 59.38%), *Mycobacterium ulcerans* Agy99 (65.47%), *Staphylococcus epidermidis* ATCC 12228 (32.1%) and *X. fastidiosa* 9a5c (52.68%) exhibited strong violations between S nucleotides as well as between W nucleotides. Chromosomes of the three *Bacillus anthracis* (35.35%) strains, *Lactobacillus reuteri* F275 (38.87%), *Magnetococcus* sp. MC-1 (54.17%), *Mycobacterium leprae* TN (57.8%), *Rhizobium leguminosarum* bv. *viciae* 3841 (61.09%) and *Rickettsia bellii* RML369-C (31.65%) exhibited strong violation between S nucleotides as well as weak violation between W nucleotides. Chromosomes of *Coxiella burnetii* Dugway 7E9-12 (42.44%) and *Staphylococcus haemolyticus* JCSC1435 (32.79%) exhibited weak violation between S nucleotides as well as between W nucleotides. Chromosome of *Vibrio cholerae* O395 (47.78%) exhibited strong violation of ISFDP only between W nucleotides. Similarly, there are six chromosomes where weak violations only between W nucleotides were observed. Chromosomes of *Bacillus thuringiensis* serovar konkukian 97-27 (34.41%), *Bordetella parapertussis* 12822 (68.1%), *Bordetella pertussis* Tohama 1 (67.72%), *Haemophilus influenzae* PittGG (38.01%), *Helicobacter hepaticus* ATCC 51449 (35.93%), *Lactobacillus acidophilus* NCFM (34.72%), *Lactobacillus brevis* ATCC 367 (46.22%), *Nitrobacter winogradskyi* Nb-255 (62.05%), *Ralstonia solanacearum* GMI1000 chromosome (67.04%), *Rhizobium etli* CFN 42 (61.27%), *Thermotoga maritima* MSB8 (46.25%) and *Thermotoga petrophila* RKU-1 (46.09%) exhibited strong violation only between S nucleotides. Similarly there are 17 chromosomes exhibited weak violation only between S nucleotides. An interesting finding that came from this study is that violations of ISFDP within a chromosome with respect to S and W nucleotides may not be of similar magnitudes. This study suggests that although ISFDP is commonly observed among chromosomes, its violation is not as rare as described earlier.^[@DSP021C13]^ ISFDP violation found in bacteria belongs to different groups, possessing different GC% and with different genome sizes.
######
Summary of ISFDP violations in chromosomes of Bacteria, Archaea and Eukaryotes
Organism Number of chromosomes Number of chromosomes exhibiting ISFDP for both W and S Number of chromosomes violating\* ISFDP for both W and S Number of chromosomes violating ISFDP only between S nucleotides Number of chromosomes violating ISFDP only between W nucleotides
------------ ----------------------- --------------------------------------------------------- ---------------------------------------------------------- ------------------------------------------------------------------ ------------------------------------------------------------------
Bacteria 112 60 15 (5^a^+8^b^+0^c^+2^d^) 30 (13^e^+17^f^) 07 (1^g^+6^h^)
Archaea 49 30 06 (2^a^+2^b^+2^c^+0^d^) 06 (0^e^+6^f^) 07 (2^g^+5^h^)
Eukaryotes 18 15 01 (0^a^+0^b^+0^c^+1^d^) 01 (1^e^+0^f^) 01 (0^g^+1^h^)
\*Violation of ISFDP includes both weak (10^−2^ \> *P* ≥ 10^−4^) and strong (*P* \< 10^−4^).
^a^Strong violation between S nucleotides as well as between W nucleotides.
^b^Strong violation between S nucleotides but weak violation between W nucleotides.
^c^Weak violation between S nucleotides but strong violation between W nucleotides.
^d^Weak violation between S nucleotides as well as between W nucleotides.
^e^Strong violation only between S nucleotides.
^f^Weak violation only between S nucleotides.
^g^Strong violation only between W nucleotides.
^h^Weak violation only between W nucleotides.
Usually, different strains within a species are found to be similar with respect to ISFDP such as the eight *E. coli* strains were observed to exhibit ISFDP between S nucleotides as well as between W nucleotides, the three *B. anthracis* strains are found to be similar in terms of their ISFDP violation (strong violation of ISFDP between S nucleotides as well as weak violations of ISFDP between W nucleotides). However, variation among the strains of a bacterial species with respect to ISFDP was observed as follows: out of the two strains of *C. burnetii*, Dugway 7E9-12 strain violated ISFDP, whereas RSA 493 strain exhibited ISFDP. Out of the four *H. influenza* strains, 86-028NP and PittEE exhibited violation of ISFDP, whereas PittGG and Rd KW20 exhibited strong and weak violations only between S nucleotides, respectively. *Xylella fastidiosa* 9a5c exhibited strong violation of ISFDP, whereas *X. fastidiosa* Temecula 1 exhibited weak violation of ISFDP only between S nucleotides. These are called as intra-species ISFDP violations. Chromosomes of four species of *Mycobacterium* genus exhibited a large difference among each other with respect to ISFDP. Chromosome of *Mycobacterium* sp. KMS (68.44%) exhibited parity between S nucleotides as well as between W nucleotides, whereas chromosome of *M. ulcerans* Agy99 (65.47%) exhibited strong violation of the parity between S nucleotides as well as between W nucleotides.
ISFDP in chromosomes of archaea and eukaryotes {#s3b}
----------------------------------------------
Out of the 49 archaea chromosomes, 30 exhibited ISFDP, 6 exhibited violations of it between S nucleotides as well as between W nucleotides, 6 exhibited violations only between S nucleotides and 7 exhibited violations only between W nucleotides (Table [4](#DSP021TB4){ref-type="table"}). Chromosomes of *Methanobrevibacter smithii* ATCC 35061 (31.02%) and *Staphylothermus marinus* F1 (35.71%) exhibited strong violation of ISFDP between S nucleotides as well as between W nucleotides. Chromosomes of *Metallosphaera sedula* DSM 5348 (46.21%) and *Pyrobaculum aerophilum* IM2 (51.34%) exhibited strong violations between S nucleotides but weak violations between W nucleotides. Strong violation between W nucleotides and weak violation between S nucleotides were observed in chromosomes of *Methanosarcina barkeri* Fusaro (39.27%) and *Nitrosopumilus maritimus* SCM1 (31.15%). This suggests that within a chromosome, the magnitude of parity violation between S nucleotides may be different from that between W nucleotides in archaea also like that of bacteria. Intra-species parity violation was also observed in archaea in the case of *Methanococcus maripaludis*. The C5 strain exhibited ISFDP violation between W nucleotides but exhibited parity between S nucleotides. The C6, C7 and S2 strains exhibited ISFDP between S nucleotides as well as between W nucleotides.
Out of the 18 eukaryotic chromosomes belonging to five species, 15 chromosomes exhibited ISFDP (Table [4](#DSP021TB4){ref-type="table"}). Strong violation of ISFDP only between S nucleotides is observed in *Leishmania major* Friedlin chromosome 01 (62.84%). *Plasmodium falciparum* 3D7chromosome 05 exhibited weak violation of parity between S nucleotides as well as between W nucleotides, whereas chromosome 01 exhibited violation of parity only between W nucleotides. The other four chromosomes of *P. falciparum* exhibited parity between S nucleotides as well as between W nucleotides. Similarly, the eight chromosomes of *Saccharomyces cerevisiae* even though exhibited parity between S nucleotides as well as between W nucleotides, the *P*-values either for S nucleotides or for W nucleotides is of more than 10-fold difference among the chromosomes. This differential ISFDP violation observed among chromosomes of an organism suggests that there may not be any strict rule inside a cell to maintain ISFDP.
ISFDP between complementary oligonucleotides in chromosomes {#s3c}
-----------------------------------------------------------
ISP between compositional abundance values of complimentary oligonucleotides is well reported. We studied here the frequency distribution of complementary di- and trinucleotides in chromosomes as described for mononucleotides. The smooth curves of oligonucleotide frequencies have been shown in [Supplementary data](http://dnaresearch.oxfordjournals.org/cgi/content/full/dsp021/DC1). In [Supplementary Fig. S1](http://dnaresearch.oxfordjournals.org/cgi/content/full/dsp021/DC1)a and b, the frequency distributions of dinucleotides have been shown for *E. coli* K12 MG1655 and *Pseudomonas entomophila* L48 chromosome (64.16%). Out of the 12 smooth frequency curves (four palindromic dinucleotides were excluded), overlapping of the curves between complementary dinucleotides is observed. In Fig. [2](#DSP021F2){ref-type="fig"}, though the abundance values of aa, tt, tg and ca dinucleotides in *E. coli* chromosome are close, the distributions between the complementary dinucleotides are found only overlapping and that of the non-complementary ones are different. The distributions for aa and tt follow a higher standard deviation (values not shown) than that of tg and ca. Similarly, gg and cc dinucleotides distributions exhibit a higher standard deviation (values not shown) than that of the dinucleotides tc and ga, although the abundance values of the four dinucleotides are close to each other. The significance of the similarity was studied by the KS test which suggested that the frequency distributions between complementary dinucleotides are statistically similar. Apart from this, dinucleotides distribution parity has been studied in three more bacterial chromosomes, two archaea chromosomes and one eukaryotic chromosome (data not shown) and similar result has been observed. In [Supplementary Fig. S2](http://dnaresearch.oxfordjournals.org/cgi/content/full/dsp021/DC1)i and ii, the distribution of 22 trinucleotides of *E. coli* K12 MG1655 chromosome is shown. Like dinucleotides, overlapping between the distributions of complementary trinucleotides is also observed. Distribution similarity between complementary trinucleotides was studied by the KS test for the 64 trinucleotides which suggested that the distributions of complementary trinucleotides within a strand are similar. The same study was done in one more bacterial chromosome (data not shown) and similar results were obtained. Although we did not analyzed the chromosomes of archaea and eukaryotes for trinucleotide distribution parity, it is expected to be there because these chromosomes had exhibited ISFDP for mononucleotides as well as dinucleotides.
{#DSP021F2}
ISFDP weakly correlates with Chargaff\'s second parity {#s3d}
------------------------------------------------------
Comparison of ISFDP was done with the ATS/GCS in chromosomes to find out whether one can define the other. GCS was compared with ISFDP violation between S nucleotides and ATS was compared with ISFDP violation between W nucleotides. Among the bacterial chromosomes, maximum GCS was found in *X. fastidiosa* 9a5c with the value 0.0529. All of the 16 chromosomes with GCS ≥0.01 were found to violate ISFDP (14 strongly violated and 2 weakly violated). Out of the 18 chromosomes with GCS ≥0.005 but \<0.01, 6 exhibited insignificant violation, 7 exhibited strong violation and 5 exhibited weak violation of ISFDP. Similarly, out of 56 chromosomes with GCS ≥0.001 but \<0.005, 5 exhibited strong violation, 11 exhibited weak violation and 40 exhibited insignificant violation. Out of the 22 chromosomes with GCS \<0.001, except *B. thuringiensis* Al Hakam chromosome (with GCS value 0.00081 exhibited weak violation of ISFDP) all other exhibited insignificant violation. Maximum ATS was found in *X. fastidiosa* 9a5c with the value 0.04727. Out of the five chromosomes with ATS ≥0.01, four were found to violate ISFDP (two strongly violated and two weakly violated), whereas *Mycoplasma hyopneumoniae* J exhibited insignificant violation (with ATS 0.0102). Out of the 14 chromosomes with ATS ≥0.005 but \<0.01, 6 exhibited insignificant violation, 3 exhibited strong violation and 5 exhibited weak violation of ISFDP. Out of the 67 chromosomes with ATS ≥0.001 but \<0.005, 57 exhibited parity, 1 strongly violated and 9 violated weakly between the W nucleotides. All the 26 chromosomes with ATS ≤0.001 exhibited insignificant violation of ISFDP. These results suggest that chromosomes with high ATS/GCS (≥0.01) have a stronger propensity to violate ISFDP and chromosomes with low ATS/GCS (≤0.001) have a stronger propensity to exhibit ISFDP. However, chromosomes with intermediate ATS/GCS (≥0.001 and ≤0.01) have the possibility of either exhibiting parity or violating the parity.
Correlation analysis was done between the *P*-values (from the KS test between) of W nucleotides and ATS as well as between the *P*-values (from the KS test between) of S nucleotides and GCS. The *r*-values are −0.5572 and −0.4526 for W and S nucleotides, respectively. This suggests that the correlation between the two ISP features is weak. The correlation between ATS and GCS is 0.629, which suggests that parity violation between S nucleotides weakly correlates with parity violation between W nucleotides within a chromosome. Unlike ATS and GCS correlation, no correlation was found between the *P*-values (the KS test) of W nucleotides and that of S nucleotides, which supports that ISFDP and Chargaff\'s second parity are not the same.
In the case of the archaea chromosomes, the ISFDP analysis revealed similar results to that of bacterial chromosomes. Maximum GCS with the value 0.03768 was found in the chromosome of *M. smithii* ATCC 35061 (31.02%) followed by the value 0.02726 in *S. marinus* F1 (35.71%), in which significant ISFDP violation was also observed. In the GCS interval 0.005 \< GCS ≤ 0.01, there were eight chromosomes out of which five exhibited weak violation and three exhibited insignificant violation of ISFDP. Out of the 24 chromosomes in the interval 0.001 \< GCS ≤ 0.005, 2 exhibited weak violation and 22 exhibited insignificant violation of ISFDP. These results suggest that chromosomes with high ATS/GCS (≥0.01) are most likely going to violate ISFDP and chromosomes with low ATS/GCS (≤0.001) are most likely going to exhibit ISFDP. However, chromosomes with intermediate ATS/GCS ((≥0.001 and≤0.01) have the possibility of either exhibiting parity or violating the parity. Pearson\'s correlation coefficient between ATS and GCS was found to be 0.707847, which is similar to that of the bacterial analysis. The *r*-values between ATS and the *P*-values of KS (W) as well as GCS and the *P*-values of KS (S) were found to be −0.57495 and −0.47557, respectively, suggesting a weak correlation.
The chromosomes with asymmetric replication topography are more prone to ISFDP violation in bacteria {#s3e}
----------------------------------------------------------------------------------------------------
Bacterial chromosome is a single replicon. Owing to the bidirectional mode of replication, one part of a strand is synthesized as LeS whereas the other part is synthesized as LaS. In most of the chromosomes, the mutational strand asymmetry causes K nucleotides \> M nucleotides in LeS and the reverse in (K nucleotides \< M nucleotides) in LaS. In an ideal case where the termination site is located symmetrically with respect to the origin of replication in a chromosome, the excess of K nucleotides in LeS will be similar to the excess of M nucleotides in LaS and therefore will cancel each other to exhibit Chargaff\'s second parity in chromosomes. Potential replication origin and termination sites for different chromosomes based on ATS, GCS, coding sequence skew, nucleotide skew at the third position of codons and oligonucleotides skew in chromosomes have been reported,^[@DSP021C31],[@DSP021C32]^ which has been reviewed in detail.^[@DSP021C33]^ Out of the 112 bacterial chromosomes analyzed in this study, information regarding the potential site for the origin and termination of 56 chromosomes is available. ISFDP violation between S nucleotides was compared with the angular deviation of termination site because G \> C in LeS is a more universal feature of chromosomes than T \> A in LeS. Of the 112 chromosomes, maximum angular deviation of 71.28° is reported in *B. pertussis* Tohama 1. Out of the 14 chromosomes where ≥20° angular deviation was observed, 12 exhibited violation of ISFDP between S nucleotides. *Pseudomonas putida* F1 (61.86%) with 36.8° and *C. burnetii* RSA 493 (42.66%) with 31.14° angular deviations exhibited insignificant parity violation. Out of the 11 chromosomes with deviation ≥10° but \<20°, 4 chromosomes exhibited ISFDP violation between S nucleotides. Out of the 30 strains with deviation ≥1.0° and ≤10°, 9 chromosomes exhibited parity violation between S nucleotides. *Chlamydophila abortus* S263 with angular deviation only 0.569°, parity violation was observed only between S nucleotides. This study indicates that chromosomes with higher asymmetric topography are more prone to violate the parity. However, chromosomes with symmetric replication topography were also observed to violate the parity.
The correlation coefficient between angular deviations and GCS as well as ATS values are 0.474 and 0.357, respectively, suggesting a weak correlation. The correlation between angular deviations and *P*-value of S (the KS test between S nucleotides) as well as that of W (the KS test between W nucleotides) are −0.259 and −0.048, respectively. The angular deviation in *X. fastidiosa* 9a5c is 62.96°, whereas the same in Temecula 1 is 6.44°. The difference in the magnitude of ISFDP violation between the strains might be attributed to the chromosome topography. Comparison for the four *H. influenzae* strains could not be done due to the unavailability of information for all the strains. The Rd KW20 chromosome (that violated ISFDP) has the angular deviation 46°, which might be an important factor to violate ISFDP. Archaea chromosomes have been reported to have more replication origin like eukaryotic chromosomes. Therefore, replication topography will not be applicable to study ISFDP violations in these cases.
Composition of forward- and reverse-encoded sequences within DNA strands might influence the parity {#s3f}
---------------------------------------------------------------------------------------------------
Most of the regions in prokaryotic chromosomes are composed of coding sequences. Presence of both forward- and reverse-encoded sequences in bacterial chromosomes has been proposed for the observation of Chargaff\'s second parity in chromosomes.^[@DSP021C8],[@DSP021C9]^ So we analyzed only coding sequences in chromosomes of bacteria and archaea to study ISFDP as follows (Fig. [2](#DSP021F2){ref-type="fig"}): in one way (Case I), a DNA strand is only composed of only forward-encoded sequences, and in the other way (Case II), a DNA strand is composed of 50% forward-encoded and 50% reverse-encoded sequences. The result is shown for *E. coli* chromosome (Fig. [3](#DSP021F3){ref-type="fig"}A and B). The smooth frequency curves of complementary nucleotides overlap in Fig. [3](#DSP021F3){ref-type="fig"}B, whereas in Fig. [3](#DSP021F3){ref-type="fig"}A, they do not overlap. The significance of these overlaps were studied by the KS test which suggests that the similarity between the distribution of complementary nucleotides in Case II. Similar results were obtained by the analysis of several bacterial (10) and archaea (15) chromosomes.
{#DSP021F3}
A comparative analysis between the Ws and Cs in a chromosome with respect to their composition of forward-encoded sequences was done in *X. fastidiosa* species as well as in *H. influenza* species. The relative differences in the compositional abundance values of forward sequences in Ws and Cs of *X. fastidiosa* 9a5c and *X. fastidiosa* Temecula 1 chromosomes are 0.078 and 0.015, respectively, which indicate that the proportion of forward- and reverse-encoded sequence in 9a5c strain is more disproportionate than that of Temecula 1 strain, which might be the reason for a stronger parity violation in the former. The relative differences of the compositional abundance values of forward-encoded sequences in Ws and Cs of *H. influenzae* 86-028NP (exhibits parity) and *H. influenzae* Rd KW20 (violates parity) chromosomes are 0.030 and 0.005, respectively, which suggest that the proportion of forward- and reverse-encoded sequences in 86-028NP strain is more disproportionate than that of Rd KW20 strain. This is in contrast to the result of *X. fastidiosa*, i.e. parity violation is observed in the strain (Rd KW20) with more proportionate gene distribution between Ws and Cs, whereas insignificant parity violation is observed in chromosome with disproportionate gene distribution between the strands. A quantitative estimation of the coding sequences in both the strands of the chromosomes was done in few other bacteria and archaea such as *A. tumefaciens*, *B. subtilis*, *E. coli*, *M. smithii* and *S. marinus* (Fig. [4](#DSP021F4){ref-type="fig"}). Maximum difference of ORF numbers between Ws and Cs was found in *S. marinus*, in which the parity violation was also observed. However, the relative difference of ORFs between the strands is found more in *B. subtilis* than in *M. smithii*. The former exhibited the parity whereas the latter violated it. *Agrobacterium tumefaciens* was shown to possess minimum relative difference of ORF numbers between the strands but violates parity. The results from this indicate that a higher disproportionate composition of forward- and reverse-encoded sequences within a strand has greater propensity to parity violation. However, proportionate composition of the sequences not necessarily implies the exhibition of parity.
{#DSP021F4}
Discussion {#s4}
==========
We have described in this study a new ISP feature in chromosomes, which is found in bacteria, archaea and eukaryotes. The methodology used to study this parity gives the statistical significance of similarity between the two distributions of complementary nucleotides/oligonucleotides. The basic qualitative feature of ISFDP is not changing for a chromosome even the segmentation is done at randomly taking any point out of the first 1000 nucleotides as the starting point. In other words, the sampling fluctuation is not affecting the feature. The correlation between the ISFDP and ISP is not strong, which is in accordance with the view that similarity in the total abundance values of two complementary nucleotides will not always yield similarity in their frequency distribution pattern. However, violation of ISP will definitely exhibit violation of ISFDP. Around 50% of the chromosomes in bacteria are found to exhibit ISFDP violations. Chromosomes of *H. influenzae* Rd KW20, *M. tuberculosis* F11, etc., which have been reported to exhibit ISP, are found to violate ISFDP.^[@DSP021C27]^
ISFDP violation observed in all possible combinations in chromosomes: (i) violation of parity between S nucleotides as well as between W nucleotides; (ii) only between S nucleotides and only between W nucleotides. The correlation between ATS and GCS is found to be not strong suggesting that parity violation between S nucleotides not necessarily always associate with parity violations between W nucleotides and *vice versa*. This can be called as intra-chromosomal parity violations. ISFDP violations of different magnitudes were found among chromosomes of different strains belonging to a species which can be referred as intra-species parity violations. Examples are *C. burnetii*, *H. influenzae* and *X. fastidiosa*. These intra-chromosomal and intra-species violations suggest that there may not be any strict rule existing in cells to maintain ISFDP in chromosomes. Differential ISP among chromosomes within a species and between chromosomes within a bacterium has already been reported in *Chlamydophila pneumoniae* strains and *Deinococcus radiodurans* R1 chromosomes,^[@DSP021C27]^ respectively. However, these were not considered significant in their study due to the lack of statistical proof. Oligonucleotide skew patterns also have been found to be variable among strains of *Yersinia pestis*. These intra-species variations in the chromosomal features are interesting and need in-depth analysis of the genome sequences to find out the reason that might reveal the reason for ISP/ISFDP violation in chromosomes and between the two ISP features.
Enrichment of LeS with K nucleotides over M nucleotides and the *vice versa* in LaS due to the mutational strand asymmetry is a general observation in chromosomes. Owing to the bidirectional replication, GCS/ATS in LeS is cancelled with GCS/ATS in LaS which results in the establishment of parity in chromosomes. The cancellation effect indirectly suggests that the compositional abundance values between the two complementary nucleotides even though they differ within a sub-chromosomal region. This is in support of the observation here that chromosomes with higher GCS/ATS values are violating ISFDP and chromosomes with lower GCS/ATS are exhibiting the parity. However, the chromosomes with intermediate range GCS/ATS are found to exhibit parity as well as violate parity and this violation is independent of genome GC%. For example, *Streptococcus mutans* UA159, *Rickettsia conorii* Malish 7, *C. jejuni* subsp. *jejuni* 81116, *Campylobacter concisus* 13826 and *Lactococcus lactis* subsp. *cremoris* MG1363, *Helicobacter pylori* J99 are (all AT-rich organisms) chromosomes with GCS≥0.005 that exhibit ISFDP between S nucleotides, whereas chromosomes of *B. anthracis* strains (AT rich) with similar GCS (\>0.005) violate ISFDP between S nucleotides. So ISFDP in these chromosomes is an interesting aspect of future research.
In concordant with the view of the bidirectional replication and establishment of parity in chromosomes, several chromosomes with higher asymmetric replication topography were found to violate ISFDP. The exceptions are *P. putida* F1 and *C. burnetii* RSA 493 chromosomes with 36° and 31° angular deviations, respectively. Chromosomes of *C. abortus* S263 and *Magnetospirillum magneticum* AMB-1, with very less angular deviations 0.57° and 2.14°, respectively, are found violating ISFDP. This indicates that features apart from the replication topography might contribute to the parity establishment in chromosomes. Proportionate composition of forward-encoded sequences between the two strands though thought to be responsible to establish the parity after the analysis of artificially constructed chromosomes, several observations went against it. The extreme case is *A. tumefaciens* where the composition is very much proportionate but violations of ISFDP are strong. So the two factors such as asymmetric replication topography and disproportionate composition of forward-encoded sequences between the strands in chromosomes that were assumed to play important roles in determining ISFDP violations were found to be insufficient.
In spite of different selection/mutation pressures on chromosomes as exemplified by codon usage,^[@DSP021C34]^ replication topography,^[@DSP021C31]^ isochores^[@DSP021C35]^ and GCS/ATS,^[@DSP021C21]^ the tendency of the chromosomes of all types toward maintaining the ISFDP is interesting. Since ISFDP and ISP are the outcomes of compositional abundance of nucleotides (mono/oligo), theories proposed for ISP might hold true for ISFDP. The Nussinov--Forsdyke hypothesis is that stem--loop potential has an adaptive advantage, and therefore an important factor driving the compositional symmetry (ISP) between the complementary oligonucleotides^[@DSP021C36],[@DSP021C37]^ has been challenged recently by Chen and Zhao^[@DSP021C38]^ for human chromosomes. This indicates that the stem--loop (recombination) hypothesis might not be the only explanation for ISP in chromosomes. Baisnée *et al*.^[@DSP021C8]^ have argued that the reverse complement symmetry does not result only from point mutation or from recombination, but from a combination effect of different mechanisms at different orders.^[@DSP021C8]^ Two independent reports have theoretically shown that multiple inversion events in chromosomes can establish ISP.^[@DSP021C10],[@DSP021C39]^ Though this hypothesis looks fine theoretically, frequent inversion unable to explain the universal observation of opposite GCS/ATS in LeS and LaS,^[@DSP021C40]^ gene distribution asymmetry between the strands^[@DSP021C41]^ and the maintenance of gene orders among different bacterial chromosomes.^[@DSP021C42]^ This hypothesis also does not describe any functional significance/advantage of the ISP/ISFDP feature, which is so wide spread in chromosomes. Theoretically, it has also been argued that the mismatch error repairing system is responsible to establish Chargaff\'s second parity rule in chromosomes.^[@DSP021C13]^ However, the intra-chromosomal parity violation observed in eukaryotes (this study) goes against this hypothesis.
We think the important factor that determines ISP/ISFDP in chromosomes is the bidirectional replication. This causes one part of a strand Ws/Cs as LeS and the other part as LaS. The strand mutational asymmetry and gene distribution asymmetry between LeS and LaS therefore cancel out each other within the strand to exhibit the parity. In the case of ssDNA/ssRNA viruses, gene distribution is restricted to one strand only depending on which these are called as either plus or -- strand viruses. The genome size is also not large (\<10 kb) in these phages^[@DSP021C43],[@DSP021C44]^ and during replication, one strand only acts as the template on which the other strand is made. Most likely these features are responsible for violating the parity in these genomes. The advantages of bidirectional replication in bacteria and archaea where the nucleus is absent are as follows: (i) quicker completion of replication than the unidirectional mode of replication and (ii) the meeting of the two replication forks might be sending some signal to the cell for the completion of chromosome replication where the nucleus is absent. Symmetric replication topography will help to terminate the replication from the origin in a lesser time in comparison with an asymmetric topography. So the selection pressure to maintain the symmetric replication topography in fast-growing bacteria is likely to be more than that in slow-growing bacteria. This proposition has similarity with the Selection Mutation Drift theory proposed for codon usage^[@DSP021C45]^ in bacteria. Our study of ISFDP of *Vibrio* species (the generation time is 0.2--0.3 h; fast-growing) in this context seems to be also not holding true here because its chromosomes violate ISFDP between W nucleotides. Moreover, comparison of generation time^[@DSP021C40]^ with asymmetry in replication topography of chromosomes^[@DSP021C32]^ exhibits no correlation (data not shown). More research on this aspect will give a conclusive result if the growth rate has any relation with parity establishment in chromosomes. In conclusion, our study has revealed an interesting aspect of ISP. Future research will reveal the reason for the presence of this parity in chromosomes.
Supplementary data {#s5}
==================
[Supplementary data are available at www.dnaresearch.oxfordjournals.org](http://dnaresearch.oxfordjournals.org/cgi/content/full/dsp021/DC1).
Supplementary Material
======================
###### \[Supplementary Data\]
A part of this work has been presented as a poster by S.K.R. in the International Conference ISMB2008 at Toronoto, Canada. Support to S.K.R. from DST (India), INSA (India) and Tezpur University for attending this conference is thankfully acknowledged. We thank the department of Biotechnology, Govt. of India for awarding MSc student fellowships to A.K. and P.K.J. The authors thank to J. R. Lobry for his critical comments on the work and are very much grateful to D. R. Forsdyke (the reviewer of the manuscript) for his comments on the manuscript.
[^1]: Edited by Hiroyuki Toh
[^2]: Present address: Molecular Biophysics Unit, Indian Institute of Science, Bangalore 560012, India.
[^3]: Present address: MS-04/603, Kendriya Vihar, Sector 56, Gurgaon, Haryana 122011, India.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction
===============
Cholelithiasis is a major health problem easily encountered in clinics and 120,000 Koreans visited the hospital for gallstone diseases in 2012 with an annual increase rate of 7.3% from 2007 to 2012. Health burden related to cholelithiasis in Korea has been great with costs of nearly \$1.6 billion in 2012 and an annual increase rate of 8.6% from 2007 to 2012.^\[[@R1]\]^ The increasing prevalence and health burden of cholelithiasis should be addressed.
The prevalence and risk factors of asymptomatic cholelithiasis vary with race, diet, culture, and geographic differences. Prevalence of 13.3% to 50.5% has been reported for asymptomatic cholelithiasis in Western countries, including United States and Europe,^\[[@R2],[@R3]\]^ whereas the prevalence of asymptomatic cholelithiasis in Eastern countries, including Korea, has been reported to be lower than that of Western countries with 2.0% to 10.7%.^\[[@R4]--[@R6]\]^ Age, female sex, obesity, dyslipidemia, diabetes mellitus, metabolic syndrome, rapid weight loss, total parenteral nutrition, chronic disease including Crohn disease, cystic fibrosis, and chronic liver disease, and drugs including octreotide, ceftriaxone, and statin were previously identified as risk factors for asymptomatic cholelithiasis.^\[[@R7]\]^
However, previous studies reported conflicting results for some risk factors of asymptomatic cholelithiasis. In some reports from Eastern countries, female sex was not found to be a risk factor for asymptomatic cholelithiasis and the prevalence of asymptomatic cholelithiasis was not significantly different between sexes or even higher in males than females.^\[[@R4],[@R6]\]^ In a study on sex differences in risk factors for asymptomatic cholelithiasis including 3573 subjects from China, a high level of fasting plasma glucose was a risk factor in males and hypertriglyceridemia or obesity in females.^\[[@R5]\]^ Identifying differences in the prevalence and risk factors for asymptomatic cholelithiasis might be important in management of asymptomatic cholelithiasis and planning preventive strategies for asymptomatic cholelithiasis.
The aim of this study was to evaluate sex differences in the prevalence and risk factors for asymptomatic cholelithiasis in Korean health screening examinees.
2. Methods
==========
2.1. Subjects and measurements
------------------------------
Examinees who underwent examinations through health promotion center at 5 hospitals in Daegu-Gyeongbuk province from January 2014 to December 2014 were included and analyzed retrospectively. Examinees younger than 18 years were excluded. All examinees were checked for height, weight, waist circumference (WC), and blood pressure (BP), and underwent laboratory tests after overnight fasting, including fasting plasma glucose, triglyceride, total cholesterol, high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol concentration, hepatitis B surface antigen (HBsAg), and hepatitis C virus antibody (anti-HCV). Ultrasonography of the abdomen was performed in all examinees.
The prevalence and risk factors for asymptomatic cholelithiasis were compared between male and female. Institutional review board approval was obtained for this study (2015--07--046).
2.2. Definition of variables
----------------------------
Diagnosis of cholelithiasis was made by abdominal ultrasound if there was the presence of echogenic densities casting distal acoustic shadow or mobile upon postural change with or without distal acoustic shadowing.^\[[@R8]\]^ Overweight was defined if body mass index (BMI) score was between 23 and 25 kg/m^2^ and obesity as BMI score \>25 kg/m^2^ in both sexes according to definition suggested by World Health Organization criteria for the Asia Pacific region.^\[[@R9]\]^ Central obesity was defined as WC \>90 cm in male and \>80 cm in female^\[[@R10]\]^ or waist-to-height ratio (WHtR) \>0.5.^\[[@R11]\]^ High BP was defined as systolic BP \>140 mmHg or diastolic BP \>90 mmHg. Hypertriglyceridemia was defined as a triglyceride level ≥150 mg/dL, hypercholesterolemia as a total cholesterol level ≥200 mg/dL, low HDL-cholesterol level as a HDL-cholesterol level \<40 mg/dL in male and \<60 mg/dL in female and high LDL-cholesterol as LDL-cholesterol level ≥130 mg/dL. High fasting glucose was defined as fasting plasma glucose level ≥110 mg/dL. Chronic hepatitis B infection was defined if examinee showed positive result for HBsAg and chronic hepatitis C if positive for anti-HCV.
2.3. Statistical analysis
-------------------------
Categorical data were presented as the number of cases and percentages. Continuous variables were shown as mean ± standard deviation. Differences were tested for statistical significance using the Student *t* test and the Pearson *χ*^2^ test. A Cox regression analysis was performed for identification of risk factors for asymptomatic cholelithiasis. Regardless of the statistical tests, the level of significance was defined as *P* \< 0.05. Statistical analyses of the data were performed using SPSS 20 (IBM SPSS, Chicago, IL).
3. Results
==========
3.1. Baseline characteristics and prevalence of asymptomatic cholelithiasis
---------------------------------------------------------------------------
Among a total of 30,556 examinees who underwent health screening examination, 12 examinees under age 18 years were excluded. Mean age of the 30,544 examinees included in this study was 47.3 ± 10.9 years and male to female ratio was 1.4:1 (17,966:12,578). Asymptomatic cholelithiasis was diagnosed in 1268 examinees with overall prevalence of 4.2%. No significant difference in overall prevalence of asymptomatic cholelithiasis was observed between male and female (4.3% vs. 4.0%, *P* = 0.159).
In both sexes, the prevalence of asymptomatic cholelithiasis increased with age (*P* \< 0.05). The prevalence of asymptomatic cholelithiasis in females and males in their 20s, 30s, 40s, 50s, 60s, and over 70 s was 2.2% vs. 1.4%, 2.9% vs. 2.0%, 3.8% vs. 3.5%, 4.0% vs. 5.5%, 6.1% vs. 7.0%, and 8.8% vs. 9.1%, respectively (Fig. [1](#F1){ref-type="fig"}). The average increase in the prevalence of asymptomatic cholelithiasis every decade of age was 1.46-fold in males and 1.32-fold in females. In age \<40 years, significantly higher prevalence of asymptomatic cholelithiasis was observed in females compared with males (2.7% vs. 1.9%, *P* = 0.020), whereas males showed significantly higher prevalence than females at age over 50 years (6.2% vs. 5.1%, *P* = 0.012) (Fig. [2](#F2){ref-type="fig"}).
{#F1}
{#F2}
The mean age of males and females was 47.7 ± 10.5 years and 46.7 ± 11.4 years, respectively (*P* \< 0.001). Proportion of obesity was 39.0% (7014/17,966) in male examinees and 18.7% (2357/12,578) in female examinees (*P* \< 0.001). Central obesity was observed in 8522 (47.4%) male examinees and 4225 (33.8%) female examinees (*P* \< 0.001). Hypertriglyceridemia was observed in 6370 (35.5%) male examinees and 1600 (12.7%) female examinees (*P* \< 0.001). High fasting glucose was observed in 2235 (12.4%) males and 710 (5.6%) females (*P* \< 0.001). Chronic hepatitis B infection was observed in 815 (4.6%) male examinees and 472 (3.8%) in female examinees (*P* = 0.001). Chronic hepatitis C infection was observed in 90 (0.5%) males and 42 (0.3%) females (*P* = 0.032) (Table [1](#T1){ref-type="table"}).
######
Baseline characteristics of examinees.
Age, serum triglyceride and HDL-cholesterol, fasting plasma glucose, and proportion of obesity, central obesity, chronic hepatitis B infection, and high BP were significantly higher in examinees with asymptomatic cholelithiasis compared to examinees without asymptomatic cholelithiasis (*P* \< 0.05) (Table [2](#T2){ref-type="table"}).
######
Comparison of baseline characteristics of examinees according to presence of asymptomatic cholelithiasis.
In examinees older than 50 years, higher fasting plasma glucose and proportion of obesity and central obesity were observed in males than females, whereas higher age, total cholesterol, HDL-cholesterol, and LDL-cholesterol were observed in females than males (*P* \< 0.05) (Table [3](#T3){ref-type="table"}). In examinees younger than 40 years, males had higher age, total cholesterol, triglyceride, LDL-cholesterol, fasting plasma glucose, and proportion of obesity, central obesity, chronic hepatitis C infection, and high BP than females and higher HDL-cholesterol was observed in females than in males (*P* \< 0.05) (Table [4](#T4){ref-type="table"}).
######
Baseline characteristics of examinees older than 50 years.
######
Baseline characteristics of examinees younger than 40 years.
3.2. Risk factors for asymptomatic cholelithiasis
-------------------------------------------------
In univariate analysis, age (≥50 years), obesity, central obesity, high fasting glucose, high BP, and high LDL-cholesterol were risk factors for asymptomatic cholelithiasis in male examinees, and age, obesity, central obesity, high fasting glucose, high BP, hypertriglyceridemia, low HDL-cholesterol, and chronic hepatitis B infection were risk factors for females.
Multiple logistic regression analysis indicated that age (odds ratio \[OR\] 2.139, 95% confidence interval \[CI\] 1.839--2.488, *P* \< 0.001), obesity (OR 1.171, 95% CI 1.009--1.359, *P* = 0.038) and high BP (OR 1.361, 95% CI 1.146--1.617, *P* \< 0.001) were risk factors for asymptomatic cholelithiasis in males, whereas age (OR 1.301, 95% CI 1.082--1.565, *P* = 0.005), obesity (OR 1.817, 95% CI 1.484--2.224, *P* \< 0.001), hypertriglyceridemia (OR 1.554, 95% CI 1.233--1.958, *P* \< 0.001), and chronic hepatitis B infection (OR 1.525, 95% CI 1.027--2.265, *P* = 0.036) in females (Tables [5](#T5){ref-type="table"} and [6](#T6){ref-type="table"}).
######
Univariate and multivariate analysis for risk factors of asymptomatic cholelithiasis in male examinees.
######
Univariate and multivariate analysis for risk factors of asymptomatic cholelithiasis in female examinees.
In stratification of examinees as normal, overweight, and obese according to the level of BMI, positive correlation was observed between the degree of BMI and prevalence of asymptomatic cholelithiasis with statistical significance in males and females (3.9% vs. 4.2% vs. 4.7%, *P* = 0.025 vs. 3.3% vs. 3.5% vs. 6.7%, *P* \< 0.001).
4. Discussion
=============
The overall prevalence of asymptomatic cholelithiasis in Korean health screening examinees was 4.2% in this study, comparable with previous reports of Eastern countries. Lower prevalence of asymptomatic cholelithiasis compared to Western countries despite recent westernized dietary changes in Korea suggests that other factors such as genetic or environmental factors might be important in the formation of asymptomatic cholelithiasis. Reasons for this difference in prevalence of asymptomatic cholelithiasis might be diverse. First, the proportion of pigment cholelithiasis is higher in Eastern countries than that of cholesterol stones, which is more predominant in Western countries. A study conducted in Korea reported a higher proportion of pigment stones compared to cholesterol stones in patients who underwent cholecystectomy owing to cholelithiasis.^\[[@R12]\]^ Second, obesity is less prevalent in Korean females compared with Western countries, but similar in males. A study conducted in the United States using data from the National Health and Nutrition Examination Survey (2007--2012) reported obesity of 36.8% in females and 35.0% in males, whereas obesity was reported in 30.5% of females and 34.9% of males in Korea.^\[[@R13],[@R14]\]^ Similarity in proportion of obesity in males and difference in females can explain in part the higher prevalence of asymptomatic cholelithiasis in males compared with females in Korea.
Age is a well-known risk factor for asymptomatic cholelithiasis and the prevalence of asymptomatic cholelithiasis increased with age in many studies. This study showed that the risk of asymptomatic cholelithiasis increased with age in both sexes showing consistent results with previous reports.^\[[@R15]\]^ Increased formation of cholelithiasis with age is suggested to be related to a longer period of exposure to various risk factors for cholelithiasis and gallbladder dysmotility secondary to sedentary activity in old age.^\[[@R16],[@R17]\]^
Female sex is an important risk factor for asymptomatic cholelithiasis, and most previous studies reported a higher prevalence of asymptomatic cholelithiasis in female than male.^\[[@R18]\]^ In the present study, females showed significantly higher prevalence of asymptomatic cholelithiasis than males only in age below 40 years. This increased risk of cholelithiasis in females is related to the estrogen effect, pregnancy, use of oral contraceptives, or hormonal replacement therapy.^\[[@R18],[@R19]\]^
Interestingly, in this study, average increase in the prevalence of asymptomatic cholelithiasis every decade of age was 1.46-fold in males and 1.32-fold in females. As a result, the prevalence of asymptomatic cholelithiasis showed a greater increase in males than females older than 50 years, and at age above 50 years, significantly higher prevalence of asymptomatic cholelithiasis was observed in males than females. Higher prevalence of asymptomatic cholelithiasis in females younger than 40 years was likely to be related to estrogen effect or pregnancy, whereas higher prevalence of asymptomatic cholelithiasis in males than females older than 50 years suggests that decreased estrogen effect or pregnancy diminished the formation of cholelithiasis in females, whereas increase in other lithogenic factors led to the formation of asymptomatic cholelithiasis in males. In this study, males older than 50 years had higher BMI, triglyceride level, and fasting plasma glucose than females, and these differences presumably influenced higher formation of asymptomatic cholelithiasis in males than females older than 50 years.
Obesity, another well-known risk factor for asymptomatic cholelithiasis in the general population, is mainly related to cholesterol stone formation. Mechanisms of cholelithiasis formation in obesity include loss of gallbladder contractility, increased level of cholesterol secretion from liver, and bile supersaturation by increase in biliary secretion of cholesterol. A large prospective study of obese women reported a strong linear association between BMI and the incidence of cholelithiasis.^\[[@R20]\]^ This study also showed statistically significant positive correlation between the level of BMI and prevalence of asymptomatic cholelithiasis when examinees were stratified as normal, overweight, and obese according to the level of BMI.
WC, WHtR, and waist-to-hip ratio indicate the degree of central obesity, whereas BMI represents the parameter of obesity. BMI is known to be related to risk of cholelithiasis in females but not in males because BMI poorly reflects central obesity in males.^\[[@R3]\]^ However, in this study, BMI was a risk factor for asymptomatic cholelithiasis in both sexes. Although waist-to-hip ratio is known to reflect central obesity more than BMI, in this study hip circumference was not measured due to retrospective design. In this study, central obesity was not found to be a risk factor for asymptomatic cholelithiasis in both sexes. Owing to the retrospective nature of the present study, measurement of WC might have been slightly different between hospitals and this might have influenced the association between central obesity and asymptomatic cholelithiasis. Central obesity is associated with formation of cholesterol cholelithiasis and these differences in association between parameters of obesity and asymptomatic cholelithiasis in the present study might be because of higher proportion of pigment cholelithiasis than cholesterol stones in Eastern countries.
Among lipid profiles, hypertriglyceridemia, a known risk factor for cholesterol stones,^\[[@R5]\]^ was a risk factor for asymptomatic cholelithiasis in females only. Proportion of examinees with hypertriglyceridemia was not significantly different between examinees with and without asymptomatic cholelithiasis in males (34.9% vs. 35.5%, *P* = 0.757). The mechanisms of increased risk of cholelithiasis formation related to hypertriglyceridemia are increase in the activity of HMG CoA reductase and subsequent increase in biliary cholesterol saturation and decrease in gallbladder sensitivity to cholecystokinin and contractility. Conflicting result of hypertriglyceridemia as risk factor for asymptomatic cholelithiasis between sexes in this study might be related to more frequent alcohol consumption in Korean male than female resulting in higher incidence of alcohol-induced hypertriglyceridemia in males compared to females.
In this study, high fasting glucose was a risk factor for asymptomatic cholelithiasis in univariate analysis, although it was not statistically significant in multivariate analysis. Insulin resistance is also associated with reduced gallbladder motility, overactivation of the rate-limiting enzyme for cholesterol synthesis, and cholesterol supersaturation in the bile, which leads to cholelithiasis formation.
An interesting point of this study was that chronic hepatitis B infection was found as a risk factor for asymptomatic cholelithiasis only in female sex, and chronic hepatitis C infection was not significantly associated with the formation of asymptomatic cholelithiasis. Previous studies have reported conflicting results for chronic hepatitis B infection as a risk factor for cholelithiasis,^\[[@R21],[@R22]\]^ and previous reports suggested chronic hepatitis C as a risk factor for cholelithiasis.^\[[@R23]\]^ Chronic liver inflammation increases the risk of cholelithiasis^\[[@R24]\]^ and chronic inflammation caused by chronic hepatitis B or C might increase the risk of cholelithiasis. Risk of cholelithiasis was reported to show correlation with the degree of liver cirrhosis.^\[[@R25]\]^ In addition, proportion of examinees with chronic hepatitis C was small and the use of anti-HCV as a surrogate for chronic hepatitis C infection may overestimate the prevalence of chronic hepatitis C infection in this study. This discrepancy between the present study and previous reports is presumably related to predominantly higher prevalence in chronic hepatitis B compared to chronic hepatitis C in Korea.
In this study, high BP was found as a risk factor of asymptomatic cholelithiasis in male sex. Some previous studies have reported on the relation between high BP and cholelithiasis.^\[[@R26],[@R27]\]^ Risk factors for high BP were age, obesity, diabetes, and dyslipidemia, and cholelithiasis shares these as risk factors.
There were several limitations in this study. First, because data were analyzed retrospectively, information about other risk factors for asymptomatic cholelithiasis such as history of pregnancy, parity, rapid weight reduction, or the use of lithogenic drugs were not investigated. Further studies including these risk factors are required to identify reasons for the differences in the prevalence of asymptomatic cholelithiasis, particularly in females younger than 40 years. Second, males were significantly older than females; thus, comparison of overall incidence of asymptomatic cholelithiasis was difficult because risk of asymptomatic cholelithiasis increased with age. However, the influence of age might be small in this study because the difference in age between males and females was only 1 year and the prevalence of asymptomatic cholelithiasis was also analyzed according to age group. Third, type of asymptomatic cholelithiasis was not evaluated and as this is important in the evaluation of prevalence and risk factors for asymptomatic cholelithiasis, studies concerning composition of asymptomatic cholelithiasis are needed. Nevertheless, this study was one of the few studies to evaluate sex differences in the prevalence and risk factors for asymptomatic cholelithiasis, particularly in a non-Western area.
In conclusion, overall prevalence of asymptomatic cholelithiasis in Korean health screening examinees was 4.2% with similar prevalence in both sexes. Females showed higher prevalence of asymptomatic cholelithiasis compared to males younger than 40 years, whereas it was higher in males older than 50 years. Age and obesity were risk factors for asymptomatic cholelithiasis in both sexes. Males had additional risk factors of high BP and females had hypertriglyceridemia and chronic hepatitis B infection.
Abbreviations: anti-HCV = hepatitis C virus antibody, BMI = body mass index, BP = blood pressure, CI = confidence interval, HBsAg = hepatitis B surface antigen, HDL = high-density lipoprotein, LDL = low-density lipoprotein, OR = odds ratio, WC = waist circumference, WHtR = waist-to-height ratio.
The authors have no potential conflict of interest to declare.
| {
"pile_set_name": "PubMed Central"
} |
Index
=====
1. Introduction to Cholesterol-Based Compounds
1
2. Drug Delivery Applications..............................................................................
3
3. Anticancer, Antimicrobial, and Antioxidant Compounds.......................................
15
4. Cholesterol-Based Liquid Crystals.....................................................................
27
5. Cholesterol-Based Gelators...............................................................................
35
6. Bioimaging Applications..................................................................................
41
7. Synthetic Applications.....................................................................................
48
8. Miscellaneous................................................................................................
56
9. Conclusions...................................................................................................
59
Funding..............................................................................................................
60
Conflicts of Interest..............................................................................................
60
Abbreviations List.................................................................................................
60
References............................................................................................................
62
1. Introduction to Cholesterol-Based Compounds {#sec1-molecules-24-00116}
==============================================
Cholesterol (cholest-5-en-3β-ol) is considered to be a lipid-type molecule, being one of the most important structural components of cell membranes. Chemically, cholesterol is a rigid and almost planar molecule with a steroid skeleton of four fused rings, three six-membered and one five-membered, conventionally lettered from A to D (1,2-cyclopentanoperhydrophenanthrene ring system) ([Figure 1](#molecules-24-00116-f001){ref-type="fig"}A). Therefore, the cholesterol molecule comprises four essential domains ([Figure 1](#molecules-24-00116-f001){ref-type="fig"}B). In domain I, the polarity of the 3-hydroxy group constitutes an active site for hydrogen bond interactions with a myriad of biological molecules (e.g., phospholipids in membranes) \[[@B1-molecules-24-00116]\]. In domain II, the absence of methyl groups at C-4 and C-14 influences directly the planarity of the molecule, while in domain III, the natural (*R*) configuration at C-20 determines the "right-handed" conformation of the side chain. Finally, in domain IV, the conformation and length of the side chain is of prime relevance to intermolecular contacts \[[@B2-molecules-24-00116]\]. The presence of a hydrophilic 3-hydroxy headgroup on the A-ring, together with a hydrophobic hydrocarbon body, give the molecule an amphiphilic nature, which makes cholesterol the most recognized sterol.
Cholesterol plays a vital role in life, particularly in cell membranes and as a precursor to the biosynthesis of several steroid hormones. In cell membranes, which are essentially constituted by a double layer of phospholipids, cholesterol has great influence on membrane fluidity, microdomain structure (lipid rafts), and permeability by interacting with both the hydrophilic headgroups and the hydrophobic tails of phospholipids. In addition, modifications of the stereochemistry and oxidation states of the fused rings, the side chain, as well as the functional groups of cholesterol, lead to a wide variety of biologically important molecules, such as bile acids, vitamin D, and several steroid hormones \[[@B1-molecules-24-00116],[@B2-molecules-24-00116]\]. Interestingly, 13 Nobel Prizes have been awarded to scientists who studied the structure of cholesterol, its biosynthetic pathway, and metabolic regulation. Unfortunately, cholesterol has gained a bad reputation because it is increasingly associated with several cardiovascular and neurodegenerative diseases, among others \[[@B1-molecules-24-00116],[@B3-molecules-24-00116]\].
Over the years, cholesterol has risen as an attractive starting material or a model system for organic synthesis due to its easily derivatized functional groups, availability, and low cost. Many useful chemical and enzymatic reactions are now widely used for multistep steroid transformations, leading to products of practical importance. The chemical transformations range from simple ones, such as manipulations of functional groups, to more complex ones, such as C-H activation or C-C bond formation with organometallic reagents. In 2014, a purely synthetic chemistry review was published, dealing only with the advances in cholesterol chemistry since 2000, focusing on cholesterol oxidation reactions, substitution of the 3β-hydroxy group, addition to the C5=C6 double bond, C-H functionalization, and C-C bond forming reactions. However, this review paper excluded simple derivatization reactions of cholesterol such as the preparation of carboxylic and inorganic acid esters, aliphatic and aromatic ethers, simple acetals, or glycosides \[[@B4-molecules-24-00116]\]. From our perspective, the simpler chemical transformations very often lead to the preparation of new cholesterol-based molecules with potential applications in several important research fields. Therefore, in this review, we focused our attention on publications from 2014 to date and described not only the synthesis of cholesterol-based new molecules, but also the application of these molecules in different fields, such as drug delivery; bioimaging; liquid crystals; gelators; anticancer, antimicrobial, and antioxidant applications; as well as purely synthetic applications. However, some interesting papers published before 2014 were included to fill some of the lacking papers from the 2014 review paper. Throughout the text, several reaction schemes will be depicted to describe the chemical reaction involved in the preparation of the cholesterol-based compounds. For simplification purposes, the structures of cholesterol will consistently be represented using the abbreviations depicted in [Figure 2](#molecules-24-00116-f002){ref-type="fig"}.
2. Drug Delivery Applications {#sec2-molecules-24-00116}
=============================
Drug delivery is a method or process of administering a pharmaceutical compound to achieve a therapeutic effect in humans or animals. Drug delivery systems can in principle provide enhanced efficacy, reduced toxicity, or both for various types of drugs. Liposomes are the most common and well-investigated nanocarriers for targeted drug delivery because they have demonstrated efficiency in several biomedical applications by stabilizing therapeutic compounds, overcoming obstacles to cellular and tissue uptake, and improving the biodistribution of compounds to target sites in vivo \[[@B5-molecules-24-00116]\].
In 2014, Vabbilisetty and Sun reported a study of terminal triphenylphosphine carrying anchor lipid effects on a liposome surface by postchemically selective functionalization via Staudinger ligation, using lactosyl azide as a model ligand. They synthesized two different anchor lipids, one of them based on the cholesterol molecule (Chol-PEG~2000~-thiphenylphosphine **3**), which was synthesized through an amidation reaction of synthetic Chol-PEG~2000~-NH~2~ **1** with 3-diphenylphosphino-4-methoxycarbonylbenzoic acid *N*-hydroxysuccinimide (NHS) active ester **2** ([Scheme 1](#molecules-24-00116-sch001){ref-type="scheme"}) \[[@B6-molecules-24-00116]\].
The authors verified that the Staudinger ligation could be carried out under mild reaction conditions in aqueous buffers without a catalyst and in high yields. The encapsulation and releasing capacity of the glycosylated liposome based on cholesterol were evaluated, respectively, by entrapping 5,6-carboxyfluorescein (CF) dye and monitoring the fluorescence leakage. It was concluded that Chol-PEG~2000~-thiphenylphosphine **3** is particularly suitable for the ligation of water-soluble molecules and can accommodate many chemical functions, being potentially useful in the coupling of many other ligands onto liposomes for drug delivery purposes \[[@B6-molecules-24-00116]\].
In 2015, a new method was reported for the deposition of a single lipid bilayer onto a hard polymer bead starting from discoidal bicelles and using chemoselective chemistry to hydrophobically anchor the lipid assemblies, using cholesterol bearing an oxyamine linker. The synthesis of oxyamine-terminated cholesterol **6** involved two steps, starting with a Mitsunobu reaction of compound **4**, followed by a reaction of **5** with hydrazine hydrate ([Scheme 2](#molecules-24-00116-sch002){ref-type="scheme"}) \[[@B7-molecules-24-00116]\].
The discoidal bicelles were prepared in water media upon mixing dimyristoylphosphatidylcholine (DMPC), dihexanoylphosphatidylcholine (DHPC), dimyristoyltrimethylammonium propane (DMTAP), and the oxyamine-terminated cholesterol derivative **6**, in a specific molar ratio. These bicelles were exposed to aldehyde-bearing polystyrene (PS) beads and readily underwent a change to a stable single lipid bilayer coating at the bead surface. This approach may be advantageous in depositing membrane proteins at such surfaces for analytical, diagnostic, or therapeutic applications (namely drug delivery) \[[@B7-molecules-24-00116]\].
Cholesterol chloroformate **7** was used as a lipid anchor for hydrophobization of arabinogalactan (AG), a liver-specific high galactose containing a branched polysaccharide, through a two-step reaction sequence that yielded a novel polysaccharide lipid, conjugated ligand **9** (**Chol**-**AL**-**AG**), with a bifunctional spacer β-alanine (AL) ([Scheme 3](#molecules-24-00116-sch003){ref-type="scheme"}) \[[@B8-molecules-24-00116]\].
Ligand **9** was used to prepare conventional liposomes (CLs) and surface-modified liposomes (SMLs) through the reverse phase evaporation technique. These new liposomes were characterized by different techniques exhibiting the required particle size for targeting tumor and infectious cells. In vitro biological studies showed an enhanced binding affinity and cellular uptake of SMLs compared to CLs by HepG2 cells, making SMLs an interesting new approach for targeted drug delivery in liver cancer therapeutics \[[@B8-molecules-24-00116]\].
Crucianelli et al. reported in 2014 a new delivery system based on liposomes containing dioleoylphosphatidilcholine (DOPC) and mannose 6-phosphate (M6P)-functionalized cholesterol **14**. For this purpose, M6P cholesteryl conjugate **14** (**Chol**-**M6P**) was synthesized following a three-step route starting from cholesterol derivative **10**, as depicted in [Scheme 4](#molecules-24-00116-sch004){ref-type="scheme"} \[[@B9-molecules-24-00116]\].
This novel vector system, designed to target lysosomes, was loaded with a model compound calcein to investigate intracellular trafficking in a 3T3-NIH cell line using a confocal and fluorescence microscopy technique. The affinity of the M6P group for the **CI-M6PR** receptor enabled these liposomes to carry calcein along the route leading to lysosomes, in opposition to calcein itself, which did not internalize into cells. These results suggest that liposomes containing **Chol-M6P 14** appear to be promising vectors in the selective targeting of lysosomes for enzyme replacement therapy or anticancer therapy \[[@B9-molecules-24-00116]\].
The importance of liposomes in drug delivery applications is well recognized. In this context, Silva and coworkers reported the synthesis of cholesterol-based neoglycoconjugates of **19** (galactose-Gal and *N*-acetylglucosamine-GlcNAc) for further incorporation into liposomes. The glycoconjugates were synthesized through a copper-catalyzed 1,3-dipolar cycloaddition (CuAAC) reaction of glycosyl azides (**18**) with cholesterol derivative **17** ([Scheme 5](#molecules-24-00116-sch005){ref-type="scheme"}) \[[@B10-molecules-24-00116]\]. The authors carried out biodistribution in vivo studies to evaluate the targeting of these carbohydrate-coated liposomes, concluding that they showed high uptake by the liver, spleen, and kidneys and no significant accumulation into other organs. Furthermore, it was demonstrated that liposomes with galactose in the surface preferentially target the liver cells. The results suggest that this kind of liposome might be a promising delivery system for therapeutic agents in hepatic diseases \[[@B10-molecules-24-00116]\].
To develop a drug delivery system for potential theranostic applications, Škorpilová et al. synthesized the fluorescent macrostructure **23** containing sesquiterpene lactone trilobolide (Tb), cholesterol, and a green-emitting boron dipyrromethene (BODIPY) dye. The synthesis of compound **23** involved a three-step sequence, starting from the CuAAC reaction of propargyl cholesterol **20** with BODIPY dye **21**, followed by functionalization with sesquiterpene lactone trilobolide ([Scheme 6](#molecules-24-00116-sch006){ref-type="scheme"}) \[[@B11-molecules-24-00116]\]. This fluorescent cholesterol conjugate **23** was successfully incorporated into liposome formulations, which showed promising immunomodulatory properties in primary rat macrophages and improved drug distribution in U-2 OS and HeLa cancer cells. The study of the intracellular trafficking pattern of liposomes revealed two populations: One localized on the cell membrane and the other inside the cell, this last one closely related to cell death. This new liposomal cholesterol conjugate **23** not only retains the biological properties of pure trilobolide, but also enhances bioavailability, and thus has potential for use in theranostic applications \[[@B11-molecules-24-00116]\].
Recently, Lin et al. reported the synthesis of a fluorescent triple-responsive block-graft copolymer **27**, bearing cholesteryl- and pyrenyl-side groups, with a disulfide (S-S) bridging point joining the hydrophilic and hydrophobic chains. The synthesis of such polymers relied on a typical click reaction between PNiPAAm~10~-S-S-P(αN~3~CL)~10~ **26**, pyrenylmethyl 4-pentynoate **25**, and cholesteryl 4-pentynoate **24**, affording PNiPAAm~10~-S-S-P(αN~3~CL~10~-g-PyrePA~3~/-CholPA~7~) **27** ([Scheme 7](#molecules-24-00116-sch007){ref-type="scheme"}) \[[@B12-molecules-24-00116]\]. Experimental results indicated that copolymer **27** could undergo self-assembly into polymeric micelles with excellent fluorescence performance in aqueous solution. The drug-loading capacity of cholesteryl-grafted copolymer **27** was evaluated using doxorubicin (DOX) as a template drug, and the results showed reasonable DOX-loading capacity. The authors also demonstrated that DOX-loaded micelles enter the cells at a substantially faster rate than their free-form counterparts, effectively inhibiting HeLa cell proliferation \[[@B12-molecules-24-00116]\].
In 2014, the synthesis of a new dual-imaging and therapeutic agent for improved efficacy in Boron Neutron Capture Therapy (BNCT) in cancer treatment was reported \[[@B13-molecules-24-00116]\]. The compound consists of a carborane unit (ten boron atoms) bearing a cholesterol unit on one side (to pursue incorporation into the liposome bilayer) and a Gd(III)/1,4,7,10-tetraazacyclododecane monoamide complex on the other side (as an magnetic resonance imaging (MRI) reporter to attain the quantification of the B/Gd concentration). The synthesis of the target compound Gd-B-AC01 (**37**) relied on an eight-step synthetic strategy, which ended with the complexation of **36** with Gd(III) in aqueous solution at pH 6.5 ([Scheme 8](#molecules-24-00116-sch008){ref-type="scheme"}). This dual probe **37** was functionalized with a polyethylene glycol (PEG)ylated phospholipid containing a folic acid residue at the end of the PEG chain. These liposomes presented interesting features such as the ability to selectively concentrate high amounts of boron in human ovarian cancer cells (IGROV-1), enough to perform efficient BNCT treatment with significantly reduced uptake by healthy cells in the surrounding regions. Furthermore, these liposomes, which can be used as nanoplatforms to deliver both Gd and B agents, can, in principle, be used for the simultaneous delivery of antitumor drugs such as DOX \[[@B13-molecules-24-00116]\].
Zhang and coworkers studied the behavior of nanoparticles (NPs) formed by self-assembly of amphiphilic poly\[*N*-(2-hydroxypropyl)methacrylamide\] (*p*HPMA) copolymers bearing cholesterol side groups (**39**) as potential drug carriers for solid tumor treatment ([Figure 3](#molecules-24-00116-f003){ref-type="fig"}) \[[@B14-molecules-24-00116]\].
The behavior of such NPs in human serum albumin (HSA) protein environment was evaluated using mixed solutions of NPs from polymer conjugates with or without the anticancer drug doxorubicin bounded to them, **39** and **40**, respectively. The authors found that in the absence of DOX, a small amount of HSA molecules bind to the cholesterol groups of the NPs by diffusing through the loose *p*HPMA shell or get caught in meshes formed by the ***p*HPMA** chains. On the other hand, the presence of DOX strongly hinders these interactions, and for that reason the delivery of DOX by these NPs in the human body is not affected by the presence of HSA \[[@B14-molecules-24-00116]\].
Recently, Singh and coworkers reported the biofunctionalization of the surface of β-cyclodextrin nanosponge **41** (**β-CD-NSP**) with cholesterol, expecting to improve its cellular binding ability. The β-CD-NSP was functionalized by grafting cholesterol hydrogen succinate (CHS) through a coupling reaction, affording **β-CD-NSP-CHS 42** ([Scheme 9](#molecules-24-00116-sch009){ref-type="scheme"}) \[[@B15-molecules-24-00116]\].
The cytotoxicity assays showed that **β-CD-NSP 41** was nontoxic and that the surface biofunctionalized with CHS **42** improved both the therapeutic and drug delivery efficacy of DOX. The experimental results also demonstrated that CHS grafting may enhance DOX adsorption due to the hydrophobic charge on the surface. Therefore, the surface-engineered CD-NSP could be used as a carrier for low water-soluble small drug molecules to improve solubility and bioavailability in site-specific drug delivery systems \[[@B15-molecules-24-00116]\].
In attempting to develop an intelligent drug delivery for cancer chemotherapy, Li et al. synthesized dual redox/pH-sensitive amphiphilic copolymer **44** and cholesterol-modified poly(β-amino esters)-grafted disulfide poly (ethylene glycol) methyl ether \[**PAE(-SS-mPEG)-*g*-Chol**\]. The precursor PAE-SS-mPEG **43** was successfully synthesized via Michael-type step polymerization using disulfide linkage-containing PEG segment. Finally, cholesterol was incorporated into the hydroxy-pendant group trough an esterification reaction, affording the copolymer **PAE(-SS-mPEG)-*g*-Chol 44** ([Scheme 10](#molecules-24-00116-sch010){ref-type="scheme"}) \[[@B16-molecules-24-00116]\].
The authors verified the interesting physicochemical properties of copolymer **44**, namely redox and pH sensitivity. Doxorubicin-loaded hybrid polymer-lipid NPs (DOX-HDPLNPs) were prepared, and drug-loading capacity, delivery efficacy, and redox- and pH-triggered drug release behavior in vitro were studied. The results showed that DOX-HDPLNPs enhanced loading capacity and improved cellular uptake ability, as well as serum stability. The anticancer potential in tumor-bearing mice was addressed, indicating that the DOX-HDPLNPs prepared with redox- and pH-sensitive copolymer with disulfides and PEGylated lipid could efficiently enhance therapeutic efficacy with low cytotoxicity and side effects. Both in vitro and in vivo experiments indicated that DOX-HDPLNPs enhanced therapeutic efficacy with high cellular uptake and negligible cytotoxicity compared to the free drug DOX. Therefore, HDPLNPs can be considered to be smart delivery systems for hydrophobic anticancer drug delivery \[[@B16-molecules-24-00116]\].
Tran et al. developed a copolymer in 2014, constituted of polynorbonene-cholesterol/ poly(ethylene glycol) \[**P(NBCh9-*b*-NBPEG**)\] **45**, that undergoes self-assembly to form a long circulating nanostructure capable of encapsulating the anticancer drug DOX with high drug loading ([Figure 4](#molecules-24-00116-f004){ref-type="fig"}) \[[@B17-molecules-24-00116]\].
The authors found that the doxorubicin-loaded nanoparticles (DOX-NPs) were effectively internalized by human cervical cancer cells (HeLa) and that they showed dose-dependent cytotoxicity. Moreover, the DOX-NPs showed good in vivo circulation time and preferential accumulation in tumor tissue with reduced accumulation in the heart and other vital organs, and significantly inhibited tumor growth in tumor-bearing severe combined immunodeficient (SCID) mice. Based on these results, DOX-NPs can become useful carriers in improving tumor delivery of hydrophobic anticancer drugs \[[@B17-molecules-24-00116]\].
A new series of amphiphilic diblock terpolymer poly(6-*O*-methacryloyl-[d]{.smallcaps}-galactopyranose)- -*b*-poly(methacrylic acid-*co*-6-cholesteryloxyhexyl methacrylate) bearing attached galactose and cholesterol grafts \[**PMAgala-*b*-P(MAA-*co*-MAChol)s**\] **49** were prepared via Reversible Addition Fragmentation chain Transfer (RAFT) copolymerization followed by deprotection of galactose in the presence of trifluoroacetic acid (TFA) ([Scheme 11](#molecules-24-00116-sch011){ref-type="scheme"}) \[[@B18-molecules-24-00116]\].
The new terpolymers (**49**) were studied for in vitro DOX release, and the results revealed high stability of the DOX-loaded terpolymer micelles under neutral conditions and significantly fast responsive DOX release. In addition, the results of fluorescence microscopy revealed that the DOX encapsulated in the synthesized diblock terpolymer PMAgala~18~-*b*-P(MAA~26~-*co*-MAChol~9~)/DOX micelles could be uptaken and delivered into cell nuclei in an efficient way, and their intracellular trafficking pathway could be altered compared to the free DOX control. The new terpolymers (**49**) could therefore be strongly considered for future smart nanoplatforms toward efficient antitumor drug delivery \[[@B18-molecules-24-00116]\].
In 2014, a reduction-responsive polymersome based on the amphiphilic block copolymer **PEG-SS-PAChol 52** was developed. The synthesis of **52** was achieved using **PEG-SS-Br 50**, a versatile atom transfer radical polymerization (ATRP) macroinitiator, and a cholesterol-containing acrylate **51**, using CuBr as a catalyst and *N*,*N*,*N*′,*N*″,*N*″-pentamethyldiethylenetriamine (PMDETA) as a ligand ([Scheme 12](#molecules-24-00116-sch012){ref-type="scheme"}) \[[@B19-molecules-24-00116]\].
The polymersome **52** was studied to come up with robust nanocarriers able to release their content inside the cells upon contact with the intracellular reducing environment. The physical crosslinking by a smectic phase of **52** in the hydrophobic sublayer, as well as the introduction of a disulfide bridge that links the hydrophilic PEG and hydrophobic blocks present in **52**, were key features that gave stability, robustness, and reduction sensitivity to the polymersome. The results showed sensitivity of the block copolymer **52** to reduction, and the fluorescence dequenching of calcein both in glutathione (GSH) solution and in vitro with the mouse macrophage cells pretreated with GSH-OEt demonstrated the breakdown of polymersome under reduction conditions. To achieve significant calcein release, high concentrations of GSH and long incubation times were necessary. These reduction-responsive polymersomes (**52**) could be used as drug carriers with very long circulation profiles and slow release kinetics \[[@B19-molecules-24-00116]\].
Recently, two new sterol-anchored polyethylene glycols, **55** and **58**, were reported as potential alternatives to conventional phosphatidylethanolamine-PEGs. Their synthesis relied on the esterification reaction of cholesterol derivatives **53** and **56** with PEGs **54** and **57**, as depicted in [Scheme 13](#molecules-24-00116-sch013){ref-type="scheme"} \[[@B20-molecules-24-00116]\].
The authors studied the biophysical properties of liposomes containing these two sterol-anchored PEGs, **55** and **58**, which exhibited an array of canonical PEGgylated-liposome behaviors including retention of encapsulated small molecules, low serum protein adsorption, and reduced cellular uptake, yet they did not exhibit long circulation \[[@B20-molecules-24-00116]\].
Polymeric micelles are known for their variety of therapeutic applications. In this field, two amphiphilic polymers were successfully synthesized using hyaluronic acid (HA), cholesterol, and octadecanoic acid as hydrophobic groups. Only the synthesis of cholesterol containing polymer **HA-SA-CYS-Chol 60** is depicted in [Scheme 14](#molecules-24-00116-sch014){ref-type="scheme"}, since the other hydrophobic groups do not fit in the scope of this paper. Nevertheless, the authors concluded that different properties of hydrophobic groups of the amphiphilic carrier are closely implicated in the stability and drug-loading capacity of the amphiphilic carrier and micelles. **HA-SA-CYS-Chol 60** presented a lower critical micellar concentration, producing docetaxel (DTX)-loaded micelles of a smaller particle size, higher encapsulation efficiency, and drug loading, when compared to the other hydrophobic tails \[[@B21-molecules-24-00116]\]. Furthermore, in vivo animal studies revealed very good tumor-targeting properties and efficient antitumor effects at very low concentrations, with low systemic toxicity of **HA--SA--CYS--Chol 60** micelles \[[@B21-molecules-24-00116]\].
A new liposomal formulation for drug delivery purposes was recently developed, based on the *N*-terminal cholesterol conjugation with a mitochondria-penetrating peptide (MPP) sequence, consisting of four amino acids \[phenylalanine-arginine-phenylalanine-lysine (FRFK)\]. More specifically, the synthesis of cholesterol-phenylalanine-arginine-phenylalanine-lysine (**Chol-FRFK**) **64** was achieved by coupling cholesteryl chloroformate **7** with amino acid-bound resins (**62**), followed by resin cleavage using TFA and the removal of protecting groups ([Scheme 15](#molecules-24-00116-sch015){ref-type="scheme"}) \[[@B22-molecules-24-00116]\]. The authors developed the liposomes using dioleoyl-*sn*-glycero-3-phosphoethanolamine (DOPE) and **Chol-FRFK 64** for delivery of the hydrophobic drug antimycin A specifically targeted toward mitochondria and lung cancer A549 cells. The results indicated that this formulation can effectively deliver the encapsulated drug to the mitochondria because of the small size and moderately cationic charge of the liposomes, enabling cellular uptake with low toxicity. The liposomes were found to be stable for long periods at room temperature, and they acted synergistically with antimycin A, leading to the complete disruption of inner membrane potential \[[@B22-molecules-24-00116]\].
In 2016, six new cholesterol-derived cationic lipids, **68**--**73**, were synthesized via ether or ester linkages with different head groups ([Scheme 16](#molecules-24-00116-sch016){ref-type="scheme"}), which were used to create cationic liposomes for nonviral gene delivery vectors \[[@B23-molecules-24-00116]\]. The authors studied the relationship between the structure of the synthesized lipids and the transfection efficiency and optimized gene transfection conditions of the liposomes. They found that the chemical structure of head groups and the linkage between cholesterol and head groups play important roles in gene delivery efficiency. Furthermore, lipids **69** and **73** exhibited higher transfection efficiency and lower toxicity than those of the tested commercial liposomes DC-Chol and lipofectamine 2000, even in the presence of serum \[[@B23-molecules-24-00116]\].
In 2015, Vulugundam and coworkers reported the design and synthesis of new redox-active monomeric **76** and **77**, and dimeric (gemini) **79** and **80**, cationic lipids based on ferrocenylated cholesterol derivatives for the development of gene delivery systems ([Scheme 17](#molecules-24-00116-sch017){ref-type="scheme"}). The cationic cholesterols **76** and **77**, as well as **79** and **80**, were incorporated into co-liposomes and shown to be transfection-efficient. The authors found that redox activities of co-liposomes and their lipoplexes could be regulated using the alkyl ferrocene moiety. The vesicles possessing ferrocene in the reduced state induced an efficient gene transfection capability using pEGFP-C3 plasmid DNA in three cell lines, even better than the commercial lipofectamine 2000 (Lipo 2000). This evidence suggests that these redox-driven systems could be used in gene delivery applications where transfection needs to be performed spatially or temporally \[[@B24-molecules-24-00116]\].
A series of macrocycle polyamine (cyclen and 1,4,7-triazacyclononane (TACN))-based cationic lipids **85** and **88**, bearing cholesterol as a hydrophobic tail, were synthesized through ring-opening reactions ([Scheme 18](#molecules-24-00116-sch018){ref-type="scheme"}). These cationic lipids, **85** and **88**, were used in combination with 1,2-dioleoyl-*sn*-glycero- -3-phosphoethanolamine (DOPE) to prepare lipoplexes, which efficiently condense DNA into nanoparticles with a proper size and zeta potential \[[@B25-molecules-24-00116]\].
Lipid **85**, containing cyclen as a headgroup, demonstrated lower toxicity and better transfection efficiency (TE) in vitro, when compared to the commercial reference lipofectamine 2000 in both 7402 and A549 cancer cells. Furthermore, the authors rationalized the good serum tolerance of **85** due to the presence of a hydroxy group in its structure. These promising results indicated that cationic-lipid **85** should be considered for nonviral gene vectors in in vivo applications \[[@B25-molecules-24-00116]\].
Aiming to extend the existent library of polycationic amphiphiles, Puchkov et al. designed and synthesized a new molecule, **92**, based on triethylenetetramine and cholesterol (a spermine analogue containing the same number of amino groups but differing in the number of methylene units). The synthesis of the polycationic amphiphile **92** was based on the selective transformation of primary amines into secondary ones via nitrobenzenesulfonamides, and the molecule of cholesterol was incorporated through alkylation of bis(sulfonamide) **89** with bromo derivative of cholesterol **90** ([Scheme 19](#molecules-24-00116-sch019){ref-type="scheme"}) \[[@B26-molecules-24-00116]\]. The authors used the triethylenetetramine-based amphiphile **92** to prepare cationic liposomes and concluded that the transfection properties of delivery nucleic acids in eukaryotic cells were inferior to those with amphiphiles based on spermine. Despite the polyamines (triethylenetetramine and spermine) having the same number of amino groups, their distribution was significantly different, which may have resulted in the difference in their transfection activity \[[@B26-molecules-24-00116]\].
A newly designed arginine-conjugated cholesterol derivative, **94**, was recently reported for the preparation of cationic liposomes and their interaction with paclitaxel (PTX), a widely used anticancer drug. The synthesis of cholesterol-arginine ester (CAE) conjugate **94** was carried out through *N*-amidation of [l]{.smallcaps}-arginine ethyl ester **93** with cholesteryl chloroformate **7** ([Scheme 20](#molecules-24-00116-sch020){ref-type="scheme"}) \[[@B27-molecules-24-00116]\].
The authors conducted molecular dynamic simulations as well as in vitro studies with the PTX-loaded liposomes. The results showed that these cationic liposomes enhanced loading efficiency and stability over the conventional liposomes, which can be rationalized based on the hydrogen bonding between CAE and PTX and the deeper penetration of PTX in the bilayer. Moreover, these novel liposomes demonstrated improved cytotoxicity in three different cell lines (MDA MB 231, H5V, and HDMEC) and enhanced endothelial cell migration inhibition compared to conventional liposomes. The absence of genotoxicity makes cholesterol-arginine ester **94** an interesting biocompatible cationic ligand in drug delivery applications \[[@B27-molecules-24-00116]\].
The design and synthesis of thermosensitive polymers of *N*-(2-hydroxypropyl)methacrylamide mono/dilactate of different molecular weights and composition bearing a cholesterol anchor **98** (**Chol-*p*HPMAlac**) was reported in 2014 ([Scheme 21](#molecules-24-00116-sch021){ref-type="scheme"}). These new cholesterol-based polymers **98** were incorporated onto liposome formulations loaded with DOX. The authors concluded that the release of DOX from such liposome formulations was effective at low temperatures and could be adjusted according to the grafting density of **Chol-*p*HPMAlac 98**. **Chol-*p*HPMAlac 98** with a cloud point of 19.0 °C and a *M*~n~ of 10.0 kDa showed interesting releasing features because it was stable at body temperature, releasing its content only under hyperthermia conditions. These releasing features make these liposomes interesting for local drug delivery using hyperthermia \[[@B28-molecules-24-00116]\].
Recently, Asayama and coworkers reported a byproduct-free PEGylation method for the modification of insulin. The strategy involves the reaction of cholesterol chloroformate **7** with aminopropyl mPEG in the presence of triethylamine to afford the conjugate **Chol-U-Pr-mPEG 99** ([Scheme 22](#molecules-24-00116-sch022){ref-type="scheme"}), complexation with insulin in aqueous solution, and subsequent freeze-drying \[[@B29-molecules-24-00116]\].
The **Chol-U-Pr-mPEG**/insulin complex not only preserved the insulin conformation, but also was shown to be effective in its protection from hydrolysis by protease and in the suppression of blood glucose levels in mice \[[@B29-molecules-24-00116]\].
3. Anticancer, Antimicrobial, and Antioxidant Compounds {#sec3-molecules-24-00116}
=======================================================
Many new cholesterol derivatives bearing a wide range of bioactive scaffolds have been developed in the search for new anticancer, antimicrobial, or antioxidant agents with improved efficacy. In this context, Rodríguez et al. described an efficient synthesis of (6*E*)-hydroximinosteroid homodimers (**105**) linking two steroidal monomers at position 3 of the steroid scaffold via ruthenium-catalyzed cross-metathesis reaction ([Scheme 23](#molecules-24-00116-sch023){ref-type="scheme"}). The synthesis of the precursor monomers was carried out through a five-step reaction sequence starting from cholesterol **28** ([Scheme 23](#molecules-24-00116-sch023){ref-type="scheme"}) \[[@B30-molecules-24-00116]\].
The cytotoxic activity of (6*E*)-hydroximinosteroid homodimers (**105**) was evaluated in vitro using human lung carcinoma A549, colon adenocarcinoma HCT-116, human Caucasian glioblastoma multiform T98G, and human pancreatic adenocarcinoma PSN1 cells. Only homodimer **105** (*n* = 2) showed selective cytotoxicity against HCT-116 cells: However, it presented no activity against the remaining cell lines. Nevertheless, the monomer counterparts **106** and **107** showed better cytotoxic activity against all cell lines when compared to homodimer **105** \[[@B30-molecules-24-00116]\].
Richmond et al. reported the synthesis of four new (6*E*)-hydroximinosteroids (**109**), starting from the corresponding ketones (**108**) derived from cholesterol. The authors evaluated the cytotoxicity of all the prepared compounds (**109**) and compared the results to those of five polyhydroxylated sulfated analogs (**110**) ([Scheme 24](#molecules-24-00116-sch024){ref-type="scheme"}) \[[@B31-molecules-24-00116]\].
Upon evaluation of the cytotoxic activity of the steroidal oxime **109** against two prostate carcinoma cell lines (PC-3 and LNCaP), the authors concluded that oxime **109** (R^1^ = R^4^ = OH, R^2^ = R^3^ = H) was the most active compound for PC-3, while for LNCaP the trisulfated analog **110** (R^5^ = H, R^6^ = OSO~3~Na) was the most active one \[[@B31-molecules-24-00116]\].
A new greener methodology involving steroidal epoxides as intermediates for the synthesis of steroidal β-aminoalcohols was recently reported. The synthesis of β-aminoalcohol **112** involved two steps: i) Epoxidation of cholesterol **28** conducted by *m*-chloroperoxybenzoic acid (*m*-CPBA); and ii) solvent-free aminolysis of epoxide **111** mediated by sulfated zirconia ([Scheme 25](#molecules-24-00116-sch025){ref-type="scheme"}) \[[@B32-molecules-24-00116]\].
The antiproliferative activity of the cholesterol-based β-aminoalcohol **112** was evaluated using MCF-7 cells, and the results showed better cytotoxic effects than cholesterol **28** itself, either by crystal violet staining (CVS) or 3-(4,5-dimethylthiazo-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assays. Furthermore, cell images obtained by Harris' hematoxylin and eosin staining protocol evidenced formation of apoptotic bodies because of the presence of cholesterol β-aminoalcohol **112** in a dose-dependent fashion \[[@B32-molecules-24-00116]\].
The synthesis of new steroidal 5α,8α-endoperoxides starting from different steroids, including cholesterol, was reported, involving a four-step synthetic protocol. It involved the introduction of a diene in the cholesterol **28** structure through allylic bromination followed by elimination, and finally a photoinduced formation of the cholesterol-based 5α,8α-endoperoxide **115** ([Scheme 26](#molecules-24-00116-sch026){ref-type="scheme"}) \[[@B33-molecules-24-00116]\].
The authors evaluated the in vitro antiproliferative activities of the 5α,8α-endoperoxides against human cancer cell lines derived from various human cancer types, such as human hepatocellular cancer cell lines (HepG2, SK-Hep1) and human breast cancer cell lines (MDA-MB-231, MCF-7). It was found that some compounds exhibited potent anticancer activities through inducing cancer cell apoptosis against the four tested cancer cell lines, particularly the cholesterol-based 5α,8α-endoperoxide **115**, which was the most promising derivative, presenting IC~50~ values ranging from 8.07 to 12.25 μM \[[@B33-molecules-24-00116]\].
A six-step synthetic route based on cholesterol **28** as a starting material was designed to prepare two new steroidal thiadiazole derivatives, **121** (R = H, Me), with an A-homo lactam and a B-norsteroidal skeleton ([Scheme 27](#molecules-24-00116-sch027){ref-type="scheme"}) \[[@B34-molecules-24-00116]\]. The antiproliferative activity of compounds **118**--**121** against various cancer cell lines was evaluated, and the results showed that compounds **120** (R = Ph) and **121** (R = Me) displayed excellent selective inhibition to the A-549 (human lung carcinoma) cell line, with IC~50~ values of 7.8 and 8.0 µM, respectively \[[@B34-molecules-24-00116]\].
In 2017, Martínez-Pascual et al. reported a new three-step method for the synthesis of 6a-aza-[b]{.smallcaps}-homo steroidal lactam **124** using cholesterol **28** as a starting material. This new methodology involved the formation of a hydroximino intermediate **123** obtained in a two-step sequence from the starting cholesterol **28** ([Scheme 28](#molecules-24-00116-sch028){ref-type="scheme"}) \[[@B35-molecules-24-00116]\].
The new compound **124** was evaluated as an antiproliferative agent against six human solid tumor cell lines, displaying only moderate activity against the screened cell lines \[[@B35-molecules-24-00116]\].
D'yakonov et al. synthesized two new hybrid compounds based on cholesterol and 1,14-tetradeca-(5*Z*,9*Z*)-dienedicarboxylic acid, **127** and **129**, which were synthetic analogues of natural (5*Z*,9*Z*)-dienoic acids. The synthetic methodology relied on the preparation of cholesterol-based oximes **126** and **128** and their further esterification using 1,14-tetradeca-(5*Z*,9*Z*)-dienedicarboxylic acid ([Scheme 29](#molecules-24-00116-sch029){ref-type="scheme"}) \[[@B36-molecules-24-00116]\]. The authors evaluated the in vitro cytotoxic activities of the synthesized compounds **126**--**129** against Jurkat (leukemia), K562 (myelogenous leukemia), U937 (lung), HeLa (cervical), and Hek293 (kidney) human cell lines. The results showed that the hybrid molecules **127** and **129** efficiently induced apoptosis of the studied cell lines and were substantially more cytotoxic than their cholesterol oxime precursors **126** and **128** \[[@B36-molecules-24-00116]\].
Cholesterol **28** was used as a template for the synthesis of a series of 2-methoxybenzoate analogs, bearing function groups such as carbonyl **131**, hydroxyl **132**, and thiosemicarbazones **133**, which were evaluated as potential new anticancer agents. The synthetic route involved the reaction of cholesterol **28** with 2-methoxybenzoyl chloride and the subsequent functionalization of the 7 position of the steroid core with several functional groups ([Scheme 30](#molecules-24-00116-sch030){ref-type="scheme"}) \[[@B37-molecules-24-00116]\].
All of the synthesized cholesterol derivatives were evaluated for their in vitro antiproliferative activities against CNE-2 (nasopharyngeal), BEL-7402 (liver), HepG2 (liver), and Skov3 (ovarian) human cancer cells, as well as HEK-293T human kidney epithelial cells. The results demonstrated that the presence of the 7-hydroxy group (compound **132**) doubled the antiproliferative activity over the nonhydroxylated compound **130**. Furthermore, none of the evaluated compounds showed inhibitory activity on HEK-293T normal cells, making them good candidates for cancer treatment \[[@B37-molecules-24-00116]\].
The synthesis of a bis(cyclam)capped cholesterol lipid (**139**) was recently reported by Peters and coworkers, who also evaluated its bioactivity using primary chronic lymphocytic leukemia (CLL) cells. The synthesis of the bis(cyclam)capped cholesterol lipid relied on a four-step methodology, as depicted in [Scheme 31](#molecules-24-00116-sch031){ref-type="scheme"} \[[@B38-molecules-24-00116]\]. It was found that the bis(cyclam)capped cholesterol lipid **139** was water-soluble and self-assembled into micellar and nonmicellar aggregates in water. The authors also found that the bis(cyclam)capped cholesterol lipid **139** was as effective as the commercial drug AMD3100 in reducing chemotaxis along CXCL12 gradients, showing that **139** may be effective in disrupting the migration of CLL cells into protective niches such as the bone marrow and lymphoid organs \[[@B38-molecules-24-00116]\].
In 2015, a paper describing the synthesis, as well as the antimicrobial and cytotoxic activities, of ten pharmacophoric motifs through CuAAC of chloroquinoline and glucose azide substrates with propargyl compounds such as chalcones, theophylline, and cholesterol was published. Within the scope of this review, only the synthesis of cholesterol-based derivatives **141** and **143** is presented ([Scheme 32](#molecules-24-00116-sch032){ref-type="scheme"}). Interestingly, the results from the antimicrobial evaluation showed that among the ten synthesized conjugates, triazole **143** exhibited the highest antibacterial activity against *E. coli* and *S. aureus*, and moderate antifungal activity against *A. flavus* and *C. albicans*. Furthermore, the sugar-cholesterol conjugate **143** displayed the best in vitro cytotoxic activity against the prostate cancer PC3 cell line \[[@B39-molecules-24-00116]\].
Two cholesterol derivatives (3β-azidocholest-5-ene (**144**) and (3β)-3-(prop-2-yn-1-yloxy)- -cholest-5-ene (**20**)) were used as starting materials for the preparation of three-motif pharmacophoric conjugates including cholesterol, 1,2,3-triazole, and either a chalcone, a lipophilic residue, or a carbohydrate tag \[[@B40-molecules-24-00116]\]. The first set of cholesterol conjugates was prepared through the reaction of 3β-azidocholest-5-ene **144** with propargylated chalcones or lactose derivatives under CuAAC conditions, affording chalcone conjugates **145** and **146** and lactose conjugates **147** and **148** ([Scheme 33](#molecules-24-00116-sch033){ref-type="scheme"}) \[[@B40-molecules-24-00116]\].
A second set of cholesterol conjugates was prepared once again through CuAAC of (3β)-3-(prop-2-yn-1-yloxy)cholest-5-ene (**20**) with azido alkanols (**149**) and 3β-azidocholest-5-ene (**144**), affording cholesterol-triazole alkanols (**150**) and a triazole-linked cholesterol dimer (**152**), respectively ([Scheme 34](#molecules-24-00116-sch034){ref-type="scheme"}) \[[@B40-molecules-24-00116]\]. Furthermore, compound **150** was converted in the respective bromo alkane **151** through a substitution reaction in the presence of CBr~4~ ([Scheme 34](#molecules-24-00116-sch034){ref-type="scheme"}) \[[@B40-molecules-24-00116]\]. A carbohydrate-tagged set of cholesterol compounds was prepared by the CuAAC reaction of (3β)-3-(prop-2-yn-1-yloxy)- -cholest-5-ene (**20**) with the appropriate glycosyl azides **153** and **155**, affording compounds **154** and **156**, respectively, upon cleavage of the acetyl protecting groups ([Scheme 34](#molecules-24-00116-sch034){ref-type="scheme"}) \[[@B40-molecules-24-00116]\].
Another carbohydrate-tagged compound, **159**, was synthesized through the reaction of cholesterol **28** with an appropriate glycosyl donor **157** in a three-step protocol, as depicted in [Scheme 35](#molecules-24-00116-sch035){ref-type="scheme"} \[[@B40-molecules-24-00116]\]. The authors screened all the cholesterol conjugates for their in vitro antimicrobial and anticancer activities. Among all compounds, the chalcone-triazole-cholesterol derivative **145** (R = NMe~2~) was the one with the most promising antimicrobial activity, being as active as the controls against *E. coli*, *S. aureus* and *C. albicans*. Concerning the cytotoxic potential of the cholesterol conjugates, the cholesterol-triazole-lactoside congener **147** displayed the best in vitro cytotoxic effect against the prostate cancer PC3 cell line, with similar cytotoxicity to that of DOX, used as a control \[[@B40-molecules-24-00116]\].
A new methodology for the synthesis of steroidal pyrazolines (**162**) through the reaction of cholest-5-en-7-ones (**160**) with 2,4-dinitrophenylhydrazine (**161**) was reported by Shamsuzzaman and coworkers in 2016 ([Scheme 36](#molecules-24-00116-sch036){ref-type="scheme"}) \[[@B41-molecules-24-00116]\]. The reaction proceeded by a well-known 1,4-/1,2-addition/dehydration mechanism to an α,β-unsaturated carbonyl compound. The new steroid-based pyrazolines (**162**) were evaluated for their in vitro antibacterial activity against three different strains (*E. coli*, *Corynebacterium xerosis*, and *S. epidermidis*), in which compound **162** (R = H) was the most active against *C. xerosis* and *S. epidermidis*, with minimum inhibitory concentrations similar to the positive control gentamicin. Compound **162** (R = H) also demonstrated moderate activity against fungal strains *Mucor azygosporus*, *Claviceps purpurea*, and *A. niger*, being the most effective compound tested. The in vitro anticancer activity against five human cancer cell lines (SW480 (colon), HeLa (cervical), A549 (lung), HepG2 (hepatic), HL-60 (leukemia)) of pyrazolines (**162**) was also screened, with the chlorinated compound **162** (R = Cl) the most active \[[@B41-molecules-24-00116]\].
The same research group reported a green simple synthesis of steroidal 2*H*-pyran-2-ones (**163**), starting from 3-substituted cholest-5-en-7-ones (**160**) and ethyl acetoacetate in the presence of chitosan as an ecofriendly heterogeneous catalyst ([Scheme 37](#molecules-24-00116-sch037){ref-type="scheme"}) \[[@B42-molecules-24-00116]\]. The synthesized steroidal 2*H*-pyran-2-ones (**163**) were tested in vitro against two cancer cell lines (HeLa (cervical) and Jurkat (leukemia)) and one normal cell line (PBMC: Peripheral blood mononuclear cell). All the tested compounds (**163**) exhibited moderate-to-good activity against the two human cancer cell lines and were less toxic against the noncancer cell line. Furthermore, the antioxidant potential of these new compounds (**163**) was also evaluated, exhibiting lower 2,2-diphenyl-1-picrylhydrazyl radical (DPPH) radical scavenging activity than the positive control, ascorbic acid \[[@B42-molecules-24-00116]\].
A series of new steroidal pyrimidine derivatives (**167**) was prepared through the multicomponent reaction of cholestan-6-ones (**164**) with urea (**166**) and benzaldehyde (**165**) in the presence of trimethylsilyl chloride (TMSCl) as catalyst ([Scheme 38](#molecules-24-00116-sch038){ref-type="scheme"}) \[[@B43-molecules-24-00116]\]. The antitumor activity of these steroidal pyrimidine-functionalized scaffolds (**167**) was screened against three human cancer cell lines, MDA-MB231 (breast), HeLa (cervical), and HepG2 (hepatic), and one noncancer normal cell line, PBMC, by MTT assay. All tested compounds showed cytotoxicities against the three cancer cell lines. Particularly, compound **167** (R = H) exhibited the highest cytotoxicity against the three cancer cell lines. However, all cases were lower than DOX, used as a positive control \[[@B43-molecules-24-00116]\]. The authors also addressed the antioxidant activity of the pyrimidine compounds (**167**), concluding that these new compounds presented reduced DPPH radical, hydroxyl radical, nitric oxide radical, and H~2~O~2~ scavenging potential than [l]{.smallcaps}-ascorbic acid, used as a control. Moreover, the IC~50~ values pointed out that the scavenging activity of the tested compounds were in the order of nitric oxide radical \< hydrogen peroxide \< DPPH radical \< hydroxyl radical \[[@B43-molecules-24-00116]\].
The cases in which the attachment of a heterocycle in the steroid backbone changes the biological properties of the steroid molecule are not so rare, and often are an interesting platform for the development of new pharmacophores. In this context, Saikia et al. reported the synthesis of steroidal heterocyclic compounds (**170**) through the solvent-free microwave-assisted epoxide ring opening with nitrogen nucleophiles \[[@B44-molecules-24-00116]\]. The first series of *N*-heterocycles was synthesized by the reaction of nitrogen nucleophiles with the epoxide **169**, which was prepared in a three-step synthetic route starting from cholesterol acetate **125** ([Scheme 39](#molecules-24-00116-sch039){ref-type="scheme"}) \[[@B44-molecules-24-00116]\].
The synthesis of another set of *N*-heterocycles, **173**, was accomplished using a mixture of epoxides (**171** (α) and **172** (β) (4:1)) as starting materials, which were obtained through the direct epoxidation of cholesterol acetate **125** ([Scheme 40](#molecules-24-00116-sch040){ref-type="scheme"}) \[[@B44-molecules-24-00116]\]. It is worth noticing that compound **173** was obtained as a diastereomeric mixture, which upon recrystallization in ethanol provided pure alcohol **173**.
The authors also considered the dehydration of the obtained cholesterol-based *N*-heterocycles (**170** and **173**), which was successfully accomplished using a catalytic amount of sulfuric acid in acetic acid, affording compounds **174** and **175**, respectively ([Scheme 41](#molecules-24-00116-sch041){ref-type="scheme"}) \[[@B44-molecules-24-00116]\].
Finally, the in vitro antibacterial activity of all compounds was evaluated, and the *N*-heterocycles **170** (Het = 4-nitroimidazole, piperidine, morpholine, thiomorpholine, tetrahydroisoquinoline) and dehydrated *N*-heterocycles **174** (Het = 4-nitroimidazole, morpholine) demonstrated moderate effects against the tested microorganisms (*E. coli*, *P. syringae*, *B. subtilis*, *P. vulgaris* and *S. aureus*). Specifically, compound **170** (Het = piperidine, morpholine, and thiomorpholine) inhibited all the tested strains, and the **170** (Het = tetrahydroisoquinoline) derivative showed inhibition against three gram-negative bacterial strains, *E. coli*, *P. syringae*, and *P. vulgaris*. The authors also concluded that the removal of the hydroxyl group decreased the antimicrobial activity of the tested compounds \[[@B44-molecules-24-00116]\].
Recently, Morake and coworkers synthesized a series of artemisinin-cholesterol conjugates, **177**, **179**, **180**, **182**, **184**, **186**, and **188**, expecting that the putative cholesterol transporters may enhance the activity of the parent drug (artemisinin) against malaria and tuberculosis \[[@B45-molecules-24-00116]\]. The conjugates were designed to have different *O*- or *N*-linkers, such as ether, ester, and carbamate, varying the length of each linker as well. The first set of conjugates, **177**, **179**--**180**, was synthesized from cholesterol **28** or cholesteryl chloroformate **7** with dihydroartemisinin **178** or artesunate **176** ([Scheme 42](#molecules-24-00116-sch042){ref-type="scheme"}) \[[@B45-molecules-24-00116]\].
A second set of conjugates, **182**, **184**, and **186**, was synthesized starting from a specific artemisinin derivative, **181** or **183**, appropriately substituted with a piperazine group at C-10, through reaction with the appropriate cholesterol derivative, **7** or **185** ([Scheme 43](#molecules-24-00116-sch043){ref-type="scheme"}) \[[@B45-molecules-24-00116]\].
Furthermore, the authors designed a final set of compounds bearing a carbamate linker, **188**. The synthesis of this set of compounds was carried out through an amidation reaction of cholesteryl chloroformate **7** with the appropriate amine derivative, **187**, bearing different lengths of alkyl chains ([Scheme 44](#molecules-24-00116-sch044){ref-type="scheme"}) \[[@B45-molecules-24-00116]\].
The antimalarial activity of the novel artemisinin-cholesterol conjugates **177**, **179**, **182**, **184**, **186**, and **188** were evaluated against *Plasmodium falciparum* (*Pf*) NF54, K1, and W2 strains, in which the conjugates of **186** (*N*-linked artemisinin-cholesterol conjugates) were the most active derivatives. However, the potency of these compounds was lower than the precursors artemether and artesunate. The authors rationalized these results based on the low solubility in the culture medium given by cholesterol moiety, which may have affected the efficacies of the artemisinin-cholesterol conjugates. On the other hand, concerning the activities against *Mycobacterium tuberculosis* (*Mtb*) H37Rv, the conjugates displayed enhanced efficacy over the parent drug artemisinin \[[@B45-molecules-24-00116]\].
The synthesis of three new cholesterol conjugates, **190**, **193**, and **194**, via CuAAC reaction was recently reported \[[@B46-molecules-24-00116]\]. These conjugates were prepared either to have a ferrocene-chalcone moiety **190** or sugar moieties **193** and **194** as well, both linked by a triazole group ([Scheme 45](#molecules-24-00116-sch045){ref-type="scheme"}) \[[@B46-molecules-24-00116]\].
The antimicrobial activities of these cholesterol conjugates were evaluated in vitro against *E. coli*, *S. aureus*, *A. flavus*, and *C. albicans*. Surprisingly, the authors found that the cholesterol conjugate bearing ferrocene-chalcone moiety **190** was completely inactive against all the tested bacteria. On the other hand, sugar conjugates **193** and **194** showed moderate inhibitory activity against *E. coli*, *A. flavus*, and *C. albicans*, being even less potent than control compounds ampicillin and amphotericin B \[[@B46-molecules-24-00116]\].
Employing a one-pot multicomponent reaction procedure using (thio)semicarbazide hydrochloride **196** and ethyl 2-chloroacetoacetate **195** allowed the preparation of a series of steroidal oxazole and thiazole derivatives (**197**) ([Scheme 46](#molecules-24-00116-sch046){ref-type="scheme"}) \[[@B47-molecules-24-00116]\].
The antimicrobial activity of the new steroidal compound **197** was evaluated against two gram-negative (*E. coli* and *P. aeruginosa*) and two gram-positive bacterial strains (*S. aureus* and *L. monocytogenes*). Additionally, the bioactivity against pathogenic fungi (*C. albicans* and *C. neoformans*) was also addressed. The authors found that most of the compounds exhibited good antibacterial and antifungal activity against the tested strains. In addition, the compounds also showed interesting antibiofilm activity against *S. aureus* biofilm. Molecular docking studies showed effective binding of the steroidal compound **197** with amino acid residues of DNA gyrase and glucosamine-6-phosphate synthase through hydrogen bonding interactions \[[@B47-molecules-24-00116]\].
Given the increasing importance of steryl ferulates \[3-*O*-(*trans*-4-feruloyl)sterols\] in pharmaceutical applications, Begum and coworkers reported the microwave-assisted synthesis of steryl ferulates from several steroids \[[@B48-molecules-24-00116]\]. The synthesis of cholesterol-based steryl ferulate **199** is exemplified in [Scheme 47](#molecules-24-00116-sch047){ref-type="scheme"}, in which microwave (MW) irradiation played a crucial role in the esterification step with *trans*-4-*O*-acetylferulic acid **198** \[[@B48-molecules-24-00116]\].
The authors evaluated the antioxidant capacity (DPPH radical scavenging, total antioxidant capacity, and reducing power) of all synthesized steryl ferulates in comparison to equimolar mixtures of steryl ferulates and γ-oryzanol (a natural mixture of steryl ferulates, abundant in cereal bran layers). The results showed that the mixture of steryl ferulates and γ-oryzanol was a better radical scavenger than most individual ferulates, including the cholesterol-based one, **199** \[[@B48-molecules-24-00116]\].
4. Cholesterol-Based Liquid Crystals {#sec4-molecules-24-00116}
====================================
A liquid crystal is basically a state of matter that has properties between those of conventional liquids and those of solid crystals. The classification of liquid crystals was proposed in the 19th century and is based on molecular arrangement. Since then, liquid crystals have been divided into smectic (from the Greek word "smegma", meaning soap) and nematic (from the Greek word "nema", meaning thread) crystals. In smectic liquid crystals, molecules are arranged so that their major axes are parallel, and their centers of mass lie in one plane. There are many different smectic phases characterized by different types and degrees of positional and orientational order. The most common ones are the smectic A phase, in which the molecules are oriented along the layer normal, and the smectic C phase, in which the molecules are tilted away from it. Nematic phases are the simplest liquid crystalline phases formed, since they only have long-range orientational order (of, e.g., molecules, columns) and no degree of long-range translational order \[[@B49-molecules-24-00116]\]. There is also a chiral variant of nematic or smectic phases, when the molecules of the liquid crystalline substance are chiral, with these phases denoted N\* or Sm(A/B)\* (an asterisk denotes a chiral phase), respectively. These phases are often called the cholesteric phases, because they were first observed for cholesterol derivatives \[[@B49-molecules-24-00116]\].
In 2014, Hiremath reported the synthesis of two new series of cholesterol-biphen-4-yl 4-(*n*-alkoxy)benzoate conjugates (**203**), linked through either odd-parity or even-parity spacers ([Scheme 48](#molecules-24-00116-sch048){ref-type="scheme"}) \[[@B50-molecules-24-00116]\]. The compounds in **203** are optically active, and both series of conjugates show a frustrated liquid crystalline state, with a thermodynamically stable twist grain boundary phase with a chiral smectic C structure (TGBC\*) over an exceedingly wide thermal range \[[@B50-molecules-24-00116]\].
The author explained such behavior based on the combined effect of extended geometry (conformation), strong chirality, and the enantiomeric excess of the molecules. Furthermore, the conjugates of **203** with an odd-parity spacer show an additional phase, the blue one. The clearing transition temperatures and the associated enthalpies alternate where the odd members exhibit lower values compared to those of even members. These results clearly demonstrate that the geometry (rod-like and bent conformation) and the thermal behavior of the conjugates of **203** are greatly influenced by the spacer parity \[[@B50-molecules-24-00116]\].
A series of similar conjugates of **206**, containing cholesterol, triazole, and biphenylene units, were synthesized via CuAAC chemistry ([Scheme 49](#molecules-24-00116-sch049){ref-type="scheme"}). Different flexible spacers were introduced in the system to evaluate the effect on the mesophase formation as well as the influence of the presence of a triazole linker \[[@B51-molecules-24-00116]\].
The authors concluded that short (*n* = 5 and 6) and medium (*n* = 7, 8, and 9) alkyl spacers exhibit enantiotropic SmA\* and monotropic SmC\* phases, whereas the conjugate possessing the longest spacer (*n* = 10) favors the formation of enantiotropic SmA and N\* phases. A close correlation between the transition temperatures and the increase in the length of the methylene spacer was also observed, and a higher clearing point was observed for the even spacers. Further comparison studies with **(*S*)-2MBbip-*n*-Chol 207** ([Scheme 49](#molecules-24-00116-sch049){ref-type="scheme"}) demonstrated that the triazole ring plays a crucial role in the mesophase formation, wherein apart from the molecular dipole, the subtle electrostatic interaction and van der Waals forces enhance the SmC\* phase \[[@B51-molecules-24-00116]\].
A study involving the design, synthesis, and mesomorphic properties of the first examples of cholesterol-based calixarene liquid crystals was reported in 2015 by Guo and coworkers \[[@B52-molecules-24-00116]\]. Novel cholesterol-1,3-bis-substituted calix\[4\]arene **209** and cholesterol-tetra-substituted calix\[4\]arene **210** derivatives were synthesized by reacting cholesterol-chlorinated derivatives (**208**) with calix\[4\]arene, as depicted in [Scheme 50](#molecules-24-00116-sch050){ref-type="scheme"}.
The liquid crystalline behaviors of cholesterol-calix\[4\]arene compounds **209** and **210** were studied, and both showed excellent mesomorphic properties of the columnar molecular arrangement of the calix\[4\]arene bowlic column, with cholesterol units as ancillary lateral columns. Furthermore, the authors demonstrated that compounds with longer spacers and more cholesterol units, such as **210**, are better for good mesomorphic properties \[[@B52-molecules-24-00116]\].
Following this study, similar calix\[4\]arene-cholesterol derivatives with Schiff-base bridges (**213**) were synthesized ([Scheme 51](#molecules-24-00116-sch051){ref-type="scheme"}), and the influence of complexation behaviors on their mesomorphic properties was investigated \[[@B53-molecules-24-00116]\]. Like the previous cholesterol-calix\[4\]arene compounds (**210**), these Schiff-base bridged compounds (**213**) presented mesomorphic properties with a molecular arrangement of the calixarene bowlic column and Schiff-base cholesterol units as ancillary lateral columns as well. However, upon complexation with AgClO~4~, no mesophase was observed, suggesting that the mesomorphic properties of compound **213** could be tuned by ion-complexation behavior \[[@B53-molecules-24-00116]\].
Recently, novel columnar liquid crystals (LCs) based on symmetric hairpin-shaped cholesterol tetramers with Schiff-base spacers were prepared, and their mesomorphic behaviors were investigated by different techniques. The new molecules were synthesized through the reaction between a cholesterol dimer, **214**, and phenylenediamines or bis-hydrazides working as spacers containing hydrogen bonds, affording compounds **215** and **216** ([Scheme 52](#molecules-24-00116-sch052){ref-type="scheme"}) \[[@B54-molecules-24-00116]\].
The results indicated good hexagonal columnar liquid crystalline behaviors, with three molecules arranged as a disc of the columnar hexagonal state. In addition, the symmetric cholesterol tetramers with rigid cores or hydrogen-bonding cores strongly favored the formation of a columnar mesophase \[[@B54-molecules-24-00116]\].
The preparation of a series of tetramers (**218**), based on azobenzene decorated with cholesterol units, was also recently reported. These oligomeric compounds bearing different alkyl spacers were synthesized by reacting azobenzene tetracarboxylic acid (**217**) with cholesteryl derivatives (**200**) ([Scheme 53](#molecules-24-00116-sch053){ref-type="scheme"}) \[[@B55-molecules-24-00116]\].
Among the synthesized compounds, it was found that oligomers with *n* = 1, 5, and 8 exhibited an enantiotropic N\* phase, while the other oligomers showed a monotropic N\* phase, upon cooling from an isotropic state. Interestingly, oligomers with *n* = 1 and 8 formed spherulites in their crystalline state, dispersed for hundreds of micrometers in the case of the oligomer with *n* = 1. Moreover, both oligomers (*n* = 1 and 8) had photoisomerization in dilute solutions and Langmuir monolayers, in opposition to the liquid crystalline state, in which no photoisomerization was observed \[[@B55-molecules-24-00116]\].
Cholesterol-based nonconventional liquid crystals have been studied by Gupta and coworkers. They reported the synthesis of novel functional discotic oligomeric materials based on 3,4,9,10-tetrasubstituted perylene, one of which bore the cholesterol units of **220** ([Scheme 54](#molecules-24-00116-sch054){ref-type="scheme"}) \[[@B56-molecules-24-00116]\].
The cholesterol derivative **220** was found to be a nonconventional LC at room temperature: However, a monotropic nematic (N\*) phase on cooling was achieved. The authors also demonstrated that the combination of rod and disc-like moieties sufficiently perturbed the molecular shape to yield calamitic mesophases. Additionally, this hybrid material showed interesting fluorescence emission properties, making it suitable for a range of optoelectronic applications \[[@B56-molecules-24-00116]\].
Recently, the synthesis of perylene derivatives with two (**223**) or four cholesterol units (**225**) at bay-position or both in bay-position and imide position, respectively, was reported ([Scheme 55](#molecules-24-00116-sch055){ref-type="scheme"}). The authors addressed the influence of the number as well as the position of the cholesterol units on the mesomorphic and photophysical properties of these new liquid crystals \[[@B57-molecules-24-00116]\]. The authors concluded that more cholesterol units significantly lowered the mesophase temperature, created wider scopes of phase transfer temperatures, and increased the fluorescence. Furthermore, it was found that a longer spacer between perylene and cholesterol units was ideal for mesomorphic properties as well as to enhance the fluorescence of the compounds \[[@B57-molecules-24-00116]\].
A year later, Chen et al. reported the synthesis of three different perylene-based liquid crystals bearing different bay-rigid spacers (**228**). These new liquid crystals were synthesized starting from a perylene derivative (**227**) with six alkyl chains on the imides positions by coupling two phenyl (biphenyl or naphthyl)-bridging cholesterol units (**226**) at bay positions ([Scheme 56](#molecules-24-00116-sch056){ref-type="scheme"}) \[[@B58-molecules-24-00116]\]. Investigations addressing the mesomorphic properties of these perylene-based compounds (**228**) demonstrated that all derivatives ordered hexagonal columnar liquid crystalline behaviors, despite the functionalization of the bay positions with aromatic spacers. Derivatives with larger and rigid aromatic spacers presented higher phase transition temperatures as well as smaller scopes of mesophase temperatures. The authors also concluded that rigid and larger aromatic groups showed stronger emission and higher fluorescence quantum yield. These results suggested that by adjusting the structures of spacers on the bay position, both mesomorphic and photophysical properties are likely to be tuned depending on the purpose of the liquid crystal \[[@B58-molecules-24-00116]\].
Aiming to explore the potentially interesting mesomorphic properties of liquid crystals, Champagne and coworkers reported the synthesis of a synthetic liquid crystal dimer (**233**) and two of its monomer analogues (**231**) based on cholesterol mesogens \[[@B59-molecules-24-00116]\]. The synthesis relied on the CuAAC reaction of a cholesteryl azide (**229**) with α,ω-di-*O*-propargyl-TEG (**232**) and *O*-monopropargylated-TEG (**230**) linkers, as depicted in [Scheme 57](#molecules-24-00116-sch057){ref-type="scheme"}. Several experimental studies were carried out, showing that both monomers (**231**) as well as the dimer (**233**) formed a smectic A liquid crystalline phase with comparable layer spacing. The authors explained this feature by the formation of a bilayer structure in the case of the monomers (**231**) and a monolayer structure for the dimer (**233**). Concerning the thermal stability of the self-assembled phases, the clearing temperature increased around 10 °C from **231** (R = Ac) to **231** (R = H). Molecular modeling studies rationalized the features of the liquid crystalline phases based on the different chemical functional groups present in each class of materials, allowing different kinds of intermolecular interactions, such as dipole-dipole interaction, hydrogen-bonding, as well as London dispersion forces, which greatly affected the self-assembly behavior of the three cholesterol derivatives \[[@B59-molecules-24-00116]\].
The synthesis of four new aliphatic polycarbonate copolymers (**mPEG~43~-*b*-P(MCC-C~n~**)~51~ (**236**) (*n* = 1--4) containing cholesteryl groups as side chain mesogenic units was achieved through the coupling reaction between **mPEG~43~-*b*-PMCC~51~** (**235**) with a side carboxyl group and chiral cholesteryl derivatives (**234**) with different numbers of methylene groups, bearing a terminal hydroxyl group ([Scheme 58](#molecules-24-00116-sch058){ref-type="scheme"}) \[[@B60-molecules-24-00116]\].
The authors studied the liquid crystal behavior of both chiral cholesteryl compounds (**234**) and the block copolymers based on cholesterol (**236**). The results demonstrated that the chiral compounds (**234**) exhibited an enantiotropic mesophase of an SmA phase and cholesteric phase except for **234** (*n* = 1), which only showed an SmA phase. The block copolymers showed an enantiotropic mesophase of an SmA phase except for **mPEG~43~-*b*-P(MCC-C~1~)~51~** (**236**) (*n* = 1), with the mesophase temperature range of the copolymers (**236**) being greater than those of the corresponding chiral compounds (**234**). It was also concluded that a longer spacer tended to stabilize the mesophase more than a shorter one and showed a wide mesophase range. These new polycarbonate copolymers with longer spacers based on cholesterol exhibited mesophase states below body temperature, which makes them good candidates for drug delivery applications \[[@B60-molecules-24-00116]\].
The synthesis of the cholesterol-triazine-BODIPY trimers **239** and **240** with one or two cholesterol units involved the reaction of cyanuric chloride-substituted BODIPY derivative **238** with an esterified cholesterol derivative (**237**), using different reaction conditions ([Scheme 59](#molecules-24-00116-sch059){ref-type="scheme"}) \[[@B61-molecules-24-00116]\].
The cholesterol-triazine-BODIPY trimers **239** and **240** exhibited distinct mesomorphic properties, dependent on the number of cholesterol units. The one-cholesterol unit derivative **239** showed nematic liquid crystal behavior, while the two-cholesterol unit **240** was a hexagonal columnar liquid crystal. The photophysical properties of both compounds were also addressed, and the authors concluded that both derivatives presented good fluorescence intensities with higher quantum yields and larger Stokes shifts when compared to their precursors. The authors claimed that this study reported the first examples of cholesterol-BODIPY liquid crystals, in which the introduction of a cholesterol unit was favorable for both liquid crystalline behavior and improved fluorescence \[[@B61-molecules-24-00116]\].
The synthesis of two series of λ-shaped dicholesteryl-based conjugates, **242** and **245**, containing a Schiff base core linking two cholesteryl ester units was reported. The first series of compounds was prepared based on a Williamson etherification between the Schiff-base (**241**) and cholesteryl bromo-alkanoates (**200**) to afford **XSB-*n*-Chol** (*n* = 4--10) derivatives (**242**) ([Scheme 60](#molecules-24-00116-sch060){ref-type="scheme"}) \[[@B62-molecules-24-00116]\]. The synthesis of **SB-10-Chol** (**244**) was slightly different and involved the alkylation of 2,4-dihydroxybenzadehyde (**243**) by cholesteryl bromo-decanoate (**200**) followed by condensation with 4-aminophenol to afford **OHSB-10-Chol** (**245**) ([Scheme 60](#molecules-24-00116-sch060){ref-type="scheme"}) \[[@B62-molecules-24-00116]\]. The study of the liquid crystal properties of the conjugates **242** and **245** indicated that the compounds had enantiotropic chiral nematic behavior, with an exception for short conjugates, which formed an additional SmA phase along with the narrow intermediary TGB phase. All compounds showed mesogenic properties, as they could form oily streaks, fan-shaped filaments, and Grandjean textures in the liquid crystalline state. The authors also found that long spacer compounds vitrified to form stable cholesteric glassy states instead of crystallization. Furthermore, the mesomorphic temperature range increased alongside the length of the spacer (from *n* = 4 to *n* = 10), showing an odd-even alternation on the clearing and transition temperatures \[[@B62-molecules-24-00116]\].
In 2015, Frizon and coworkers described the synthesis of and preliminary studies on the thermal and photophysical properties of selenium liquid crystals containing cholesterols **247**, **249**, **251**, and **253**. The synthesis of three new series of selenide **247**/**251** and diselenide compounds **249**/**253** was accomplished via esterification of cholesterol **28** with the appropriate selenide **246**/**250** or diselenide acid **248**/**252** ([Scheme 61](#molecules-24-00116-sch061){ref-type="scheme"}) \[[@B63-molecules-24-00116]\].
All synthesized compounds presented good thermal stability. Six of them showed liquid crystal properties, in which selenide **251** and alkyl diselenides **249** (*n* = 2) and **249** (*n* = 3) exhibited an SmC\* mesophase, whereas aryl diselenide **253**, with higher structural rigidity, showed a chiral enantiotropic smectic A (SmA\*) mesophase. Furthermore, all these new selenide-cholesterol compounds showed higher glutathione peroxidase-like activity than the standard ebselen, with selenide **249** (*n* = 2) the most active \[[@B63-molecules-24-00116]\].
A series of glycosteroids (**256**) constituted by cholesterol and distinct glycosidic moieties were synthesized by coupling propargyl 1-*S*-propargyl [d]{.smallcaps}-glucose, [d]{.smallcaps}-galactose, or [l]{.smallcaps}-rhamnose (**255**) to cholesterol scaffold **254** through a CuAAC reaction ([Scheme 62](#molecules-24-00116-sch062){ref-type="scheme"}) \[[@B64-molecules-24-00116]\]. This study aimed to analyze if the sugar structure as well as the heteroatom linked to the anomeric position had an impact on the liquid-crystalline properties of the glycosteroids (**256**). The mesomorphic temperature range found for the glycosteroids (**256**) was higher than that generally reported in the literature, but similar to that reported for other glycosteroids. All the studied glycosteroids (**256**) showed great phase stability compared to those already studied, and interestingly, glycosteroids (**256**) (sugar = [d]{.smallcaps}-glucose; X = S) showed no decomposition even at 200 °C. These results offer new possibilities in the development of new high-temperature captors or detectors \[[@B64-molecules-24-00116]\].
5. Cholesterol-Based Gelators {#sec5-molecules-24-00116}
=============================
Low molecular weight organic gelators (LMOGs) are small organic molecules that self-assemble in water or organic solvents, forming a 3D network that entraps the liquid phase, resulting in gel formation. In recent years, these classes of compounds have attracted much attention because of their range of applications, for example as alternative biomaterials for drug delivery or tissue engineering \[[@B65-molecules-24-00116],[@B66-molecules-24-00116]\]. New generations of steroidal low molecular mass gelators (LMGs) are usually designed through the assembly of various building units such as a steroid derivative (S), a linker unit (L), and often an aromatic platform (A) around which the steroid units can be positioned through linkers. The good gelation ability of the steroidal LMGs led to the development of a series of steroid-based gelators commonly classified as ALS, arranged in A(LS)~2~, A(LS)~3~, LS, or LS~2~ molecular types \[[@B65-molecules-24-00116]\].
In 2014, an interesting study was reported involving the design of an uncommon class of cholesteryl-based triangular A(LS)~3~-type low molecular mass gelators and the exploration of their gelation and anion-sensing applications. The design strategy was based on placing three cholesteryl derivatives using linker units around melamine or benzene-1,3,5-tricarbonyl chloride as aromatic platform precursors. The synthesis of compounds **257** and **259** involved the reaction of cholesteryl chloroformate **7** with different amines in one- or two-step procedures ([Scheme 63](#molecules-24-00116-sch063){ref-type="scheme"}) \[[@B67-molecules-24-00116]\].
This study also involved the evaluation of gelation and self-assembly properties of this new class of compounds by comparing them to the existing cholesteryl-based LMGs. The results indicated that the gelation and self-assembly properties of compounds **257** and **259** could be controlled by modification of the structural features of the A(LS)~3~-type molecule. Increasing the length of the linker units, the fibrous xerogel networks assembled into more porous fiber networks. Moreover, the authors found that the compounds **257** and **259** could be used as selective sensors for F^−^, and their selectivity could be enhanced by increasing the chain length of their linker units \[[@B67-molecules-24-00116]\].
Two new cholesterol-based compounds (**261**) were also reported as fluoride-responsive organogels. Their design was based on the coupling of compounds in **260**, bearing azo units as the chromophore and a pyrazole group as the anion acceptor, with the cholesteryl chloroformate **7** ([Scheme 64](#molecules-24-00116-sch064){ref-type="scheme"}) \[[@B68-molecules-24-00116]\].
The authors observed that structural modifications on the benzyl core of compound **261** (R = H or NO~2~), hydrogen bonding, hydrophobic interactions, as well as π-π stacking interactions, had considerable influence on the gel-sol transition properties. Moreover, they also found that the gel was selectively fluoride-responsive among the tested anions, expressing gel-sol transition and red-purple color changes easily detected by the naked eye \[[@B68-molecules-24-00116]\].
Following the purposes of the selective detection of F^−^, a new coumarin-based supramolecular gelator (**267**) was designed \[[@B69-molecules-24-00116]\]. The reported compound **267** follows a simple architecture that bears a coumarin-appended 1,2,3-triazole coupled with cholesterol, synthesized in a six-step route as depicted in [Scheme 65](#molecules-24-00116-sch065){ref-type="scheme"}. The coumarin moiety acts as a fluorescence signaling unit, the 1,2,3-triazole as a linker and as an anion binding site, and cholesterol as a hydrophobic surface.
The authors concluded that cooperative hydrogen bonding between phenolic OH and a 1,2,3-triazole ring as well as hydrophobic-hydrophobic interactions of the cholesteryl groups in compound **267** played a crucial role in the formation of an organogel. Furthermore, it was demonstrated that compound **267** organogel was sensitive for F^−^ and HP~2~O~7~^3−^ detection by means of gel phase transformation as well as fluorimetrically, showing considerable changes in emission properties \[[@B69-molecules-24-00116]\].
A novel cholesterol-based organogelator containing D-A (donor-acceptor) pairs (salicylaldehyde and naphthalimide units) (**272**) was synthesized \[[@B70-molecules-24-00116]\]. The synthetic strategy relied on the introduction of the electron-rich salicylaldehyde group into a naphthalimide-based organogelator through a Schiff-base reaction ([Scheme 66](#molecules-24-00116-sch066){ref-type="scheme"}). This cholesterol-based organogelator (**272**) was found to form stable and chiral gels with different optical properties and morphologies in several organic solvents. An interesting feature of compound **272** was the changing of the color and emission color of the organogel in benzene, which varied from yellow-green to red during the thermoreversible sol-gel transformation, demonstrating for the first time solvent-controlled multiple color emission achieved in a monocomponent gel system. This feature makes the organogel **272** quite suitable for applications in optical switches, sensors, and smart materials \[[@B70-molecules-24-00116]\].
To develop new supramolecular gelators, Panja and coworkers synthesized pyrrole and furan-based pyridine/pyridinium bisamides containing cholesteryl units in their architecture \[[@B71-molecules-24-00116]\]. The synthesis of cholesterol-based bisamides (**274**) was achieved through the coupling reaction of cholesteryl chloroacetate derivate **262** with the pyridine ring nitrogens in bisamide **273** ([Scheme 67](#molecules-24-00116-sch067){ref-type="scheme"}).
The gelation properties of both bisamide **273** and bisamides with a cholesteryl unit attached (**274**) were evaluated. In aqueous DMSO, compound **274** (X = O) exhibited nongelation properties, while compound **274** (X = NH) produced a light yellow colored gel. This suggests that the heteroatom of the aromatic linker played a crucial role in gelation. The organogel formed by compound **274** (X = NH) revealed itself to be a good anion sensor, since the gel state was selectively ruptured into solution in the presence of F^−^ and AcO^−^ anions. Interestingly, the gel rupture induced by F^−^ was recovered upon the addition of Fe^3+^. This feature is very useful in the visual distinction of F^−^ from AcO^−^ anions \[[@B71-molecules-24-00116]\].
A different kind of fluorescent organogelator based on cholesterol containing benzothiadiazole fluorophores **276** and **278** was designed and synthesized by Sun and coworkers ([Scheme 68](#molecules-24-00116-sch068){ref-type="scheme"}). The authors aimed to understand the role of hydrogen bonding and π--π interactions and to study the changes of fluorescent properties in the process of gelation of cholesterol-based π-conjugated organogels \[[@B72-molecules-24-00116]\].
The authors studied three methods of gel preparation (heating-cooling process, ultrasonic treatment, and mixed solvents, at room temperature) and found that π--π and H-bonding interactions should be the key contributors in forming gels of **276**, while in gel formations of **278**, only π--π interactions seemed to matter. The obtained results suggest that these two multiple-stimuli responsive luminescent gels, **276** and **278**, can be used as smart soft materials sensitive to temperature, solvent, ultrasound, and Hg^2+^ \[[@B72-molecules-24-00116]\].
Recently, Panja and Ghosh reported three related works involving cholesterol conjugates bearing three different moieties (dithioacetal **280**, diaminomalononitrile **281**, and diazine **282** functional groups) for sensing a series of cations such as Hg^2+^, Cu^2+^, Ag^2+^, and Fe^2+^ \[[@B73-molecules-24-00116],[@B74-molecules-24-00116],[@B75-molecules-24-00116]\]. The three cholesterol conjugates were synthesized using the same three-step methodology, except for the final step, which involved the reaction of the intermediate benzaldehyde **279** with 1-dodecanethiol, diaminomalononitrile, and hydrazine to afford cholesterol conjugates **280**, **281**, and **282**, respectively ([Scheme 69](#molecules-24-00116-sch069){ref-type="scheme"}). The cholesterol-dithioacetal conjugate **280** was used for the detection of Hg^2+^ and incorporated two distinct components: i) A cholesterol motif to assist the self-assembly of the molecules through hydrophobic interaction; and ii) a thiol part that was used as the reaction-based recognition unit of the molecule \[[@B73-molecules-24-00116]\].
The authors studied the sensing mechanism for Hg^2+^ of the cholesterol-dithioacetal conjugate, realizing that the specific Hg^2+^-induced deprotection of the thioacetal functionality of **280** resulted in sol-to-gel transition in DMF/H~2~O (1:1, *v*/*v*) through the formation of precursor aldehyde **279**. The authors also claimed that this was the first chemodosimeter that functions as a selective "naked-eye" Hg^2+^-detector by showing in situ sol-to-gel conversion \[[@B73-molecules-24-00116]\].
The cholesterol-diaminomalononitrile conjugate **281** was found to form supramolecular gels in dimethylformamide (DMF)/H~2~O and 1,2-dichlorobenzene, as confirmed by rheological studies. In addition, the authors verified that the gel formed in DMF/H~2~O was more stable and robust than the one obtained from 1,2-dichlorobenzene, due to strong intermolecular forces among the gelators in DMF/H~2~O. Furthermore, it was also established that cholesterol-diaminomalononitrile **281** gel was selective for visual recognition of Hg^2+^ and Cu^2+^ ions, and for sensing hydrazine based on the dosimetric interaction of the malononitrile motif with hydrazine \[[@B74-molecules-24-00116]\].
Concerning the cholesterol-diazine conjugate **282**, the authors demonstrated that it could form nice gels with Ag^+^ and Fe^3+^ ions in a CHCl~3~/CH~3~OH mixture solvent, using the diazine moiety as a metal ion binding site. The gelator **282** was able to distinguish Ag^+^ and Fe^3+^ with the aid of tetrabutylammonium chloride, tetrabutylammonium bromide or fluoride, and ammonium thiocyanate. Furthermore, the authors proved that there was no interference of Fe^2+^ ions in the detection of Fe^3+^ ions, as in the case of most chemosensors and gelators \[[@B75-molecules-24-00116]\].
The effect of different spacer lengths containing two, three, five, six, ten, or twelve carbon atoms on cholesterol-based azobenzene organogels **285** and **286** was investigated \[[@B76-molecules-24-00116]\]. For this purpose, a series of seven azobenzene-cholesterol compounds was synthesized through esterification reactions of cholesterol derivatives of **283** (bearing different spacer lengths) with 4′-carboxy-4-methoxyazobenzene **284** carried out in the presence of *N*,*N*′-dicyclohexylcarbodiimide (DCC) and dimethylaminopyridine (DMAP) in dichloromethane, as depicted in [Scheme 70](#molecules-24-00116-sch070){ref-type="scheme"}.
Typical reversible *trans*-*cis* and *cis*-*trans* isomerization of the azobenzene units was observed upon UV-Vis irradiation, giving the compounds **285** and **286** recoverable photoresponsive properties. Differential scanning calorimetry studies revealed that the spacer length plays a crucial role in the gelation phenomenon. Interestingly, among the tested compounds, only **285** (*n* = 6) could form a gel, and in specific solvents such as ethanol, isopropanol, and butan-1-ol. Furthermore, the authors concluded that the solvents, intermolecular H-bonding, and van der Waals interactions affected the aggregation mode and morphology of the gels \[[@B76-molecules-24-00116]\].
In 2016, a study was reported on liquid crystal (LC) and gelation-based self-assembly, as well as the photoresponsive behavior of a new unsymmetrical azobenzene-cholesterol based dimesogen, **288** \[[@B77-molecules-24-00116]\]. This molecule assembles a CN group at one end and a cholesterol carbonate, fixed through an oxyethylene spacer, to the opposite end of the azobenzene unit ([Scheme 71](#molecules-24-00116-sch071){ref-type="scheme"}).
Compound **288** presented the capacity of acting as a chiral mesogenic dye dopant to induce a high helical-twisting chiral phase in the common nematic phase of 5CB. In addition, the gels of **288** formed in organic solvents exhibited multiple stimuli-responsive behaviors upon exposure to environmental stimuli such as temperature, light, and shear forces. The photoresponsive character was also proven in solution, in LC and gel states. These properties give to compound **288** potential applications in displays, as chiral mesogenic dye dopants, photochemical molecular switches, and new versatile LMGs \[[@B77-molecules-24-00116]\].
A new series of liquid crystal gelators (**290**) with photoresponsive and aggregation-induced emission (AIE) properties was synthesized by connecting cholesterol derivatives **200** and tetraphenylethylene (an important AIEgen) to a central azobenzene moiety through esterification reaction ([Scheme 72](#molecules-24-00116-sch072){ref-type="scheme"}) \[[@B78-molecules-24-00116]\]. The authors included variations in the alkyl chain spacer (*n* = 0, 1, 3, 5) to adjust the distance between cholesterol and azobenzene, while a fixed alkyl chain was placed between azobenzene and tetraphenylethylene ([Scheme 72](#molecules-24-00116-sch072){ref-type="scheme"}). The liquid crystal properties of compounds in **290** were assessed, and the results showed that all compounds exhibited, in pure state, smectic A LC phases, enantiotropic for **290** (*n* = 0) and (*n* = 3), but monotropic for **290** (*n* = 1) and (*n* = 5). The gelation properties of compound **290** demonstrated that **290** (*n* = 3) and (*n* = 5) form stable gels in appropriate solvents or solvent mixtures, while **290** (*n* = 0) and (*n* = 1) cannot form gels in a range of solvents. An interesting feature of both **290** (*n* = 3) and (*n* = 5) LMOGs is that they have significantly enhanced emissions induced by molecular self-assembly into fibril or ribbon-like nanostructures \[[@B78-molecules-24-00116]\].
Three new cholesteryl-based A(LS)~2~- and A(LS)~3~-type LMGs, **292**, **294**, and **296**, without hydrogen bond linkers, were reported in the literature, synthesized through esterification reactions of acid chlorides **291**, **293**, and **295** with cholesterol **28** in the presence of DMAP ([Scheme 73](#molecules-24-00116-sch073){ref-type="scheme"}) \[[@B79-molecules-24-00116]\]. The study of the gelation properties in various organic solvents indicated that the number and position of the substituents in the cholesteryl moieties attached to a benzene ring had a great influence on the gelation as well as in the aggregation behaviors of the A(LS)~2~- and A(LS)~3~-type LMOGs. Among these three gelators, **294** and **296** showed efficient gelation abilities even without hydrogen bond linkers, in contrast with the meta-substituted **292**, which did not gelate in any tested solvent \[[@B79-molecules-24-00116]\].
Recently, the synthesis of a new pillar\[6\]arene-functionalized cholesterol derivative (**298**), acting as an LMG, was reported in the literature \[[@B80-molecules-24-00116]\]. In this new compound, the host--guest pillar\[6\]arene **300** was linked to a cholesterol unit by the long alkyl chain, as well as amide groups ([Scheme 74](#molecules-24-00116-sch074){ref-type="scheme"}). This new pillar\[6\]arene-cholesterol **298** was found to form an organogel in cyclohexane/hexan-1-ol (10:1, *v*/*v*), which was reversibly responsive to temperature, share stress, and partially host--guest interaction introduced by ferrocenyl iminium derivative **299**. In the case of the addition of ferrocenyl iminium derivative **299**, the organogel could be tuned into a solution and tuned back into the organogel upon addition of per-butylated pillar\[6\]arene **300**. This interesting feature could be explained on the basis of host--guest interactions of individual **300** with cationic guest **299** that bound with pillar\[6\]arene-cholesterol gelator **298** \[[@B80-molecules-24-00116]\].
In 2015, the development of a new kind of self-healing, degradable, and biocompatible polypeptide hydrogel based on self-assembly between cholesterol-modified triblock poly([l]{.smallcaps}-glutamic acid)-*block*-PEG-*block*-poly([l]{.smallcaps}-glutamic acid) \[(**PLGA-*b*-PEG-*b*-PLGA)-*g*-Chol**\] **302** and β-cyclodextrin (β-CD)-modified poly([l]{.smallcaps}-glutamic acid) (**PLGA-*g*-β-CD**) **303** ([Figure 5](#molecules-24-00116-f005){ref-type="fig"}) was reported in the literature \[[@B81-molecules-24-00116]\].
The authors observed that the hydrogel formation was based on the host and guest linkage between β-cyclodextrin (β-CD) and cholesterol, and that their viscoelastic behavior depended on polymer concentration as well as the β-CD/Chol molar ratio. Those hydrogels showed very interesting self-healing capabilities, good cytocompatibility, excellent flexibility, and quick colorant diffusion. With all these features, it is anticipated that these self-healable hydrogels may have important applications in tissue engineering \[[@B81-molecules-24-00116]\].
6. Bioimaging Applications {#sec6-molecules-24-00116}
==========================
Imaging techniques, particularly fluorescence imaging techniques, have become powerful tools for noninvasive visualization of biological processes in real time with high spatial resolution. Methods to "see into the body" or "see into cells" are essential for the diagnosis and treatment of a disease, as well as for research into the basic processes of life. Therefore, bioimaging techniques to visualize physiological or pathophysiological changes in the body and cells have become increasingly important in biomedical sciences \[[@B82-molecules-24-00116]\].
The synthesis of a series of BODIPY-based fluorogenic dyes was reported, involving the CuAAC reaction of a nonfluorescent BODIPY-azide, **304**, with a series of nonfluorescent alkyne molecules, including *O*-propargylated cholesterol **20** ([Scheme 75](#molecules-24-00116-sch075){ref-type="scheme"}) \[[@B83-molecules-24-00116]\]. The most interesting molecule was the cholesterol-linked dye **305**, which presented red-shifted absorption and emission wavelengths and displayed its preferential accumulation at the intracellular membranes over the plasma membrane of HeLa cells. This result offers potential applications of cholesterol-BODIPY conjugate **305** in the bioimaging of cholesterol trafficking in living cells and organisms \[[@B83-molecules-24-00116]\].
Byrd and coworkers reported the synthesis of a crosslinker containing two independent cholesterol units, with or without a photoaffinity label, guided by computational methods based on a model for the transfer of a cholesterol molecule between two proteins, NPC1 and NPC2 \[[@B84-molecules-24-00116]\]. The synthesis of crosslinker **314** (without a photoaffinity label) involved several steps, especially because of the demanding six-step synthetic route of one of the portions that constitutes the crosslinker **314** ([Scheme 76](#molecules-24-00116-sch076){ref-type="scheme"}) \[[@B84-molecules-24-00116]\].
Another cholesterol-based crosslinker (**322**) with a photoaffinity label was also synthesized ([Scheme 77](#molecules-24-00116-sch077){ref-type="scheme"}) \[[@B84-molecules-24-00116]\]. The synthesis of such a compound involved two stages: i) The preparation of an appropriate carboxylic acid cholesterol moiety (**318**) ([Scheme 78](#molecules-24-00116-sch078){ref-type="scheme"}) \[[@B84-molecules-24-00116]\]; and ii) the linkage between compounds **318** and **312** (previously synthesized) ([Scheme 76](#molecules-24-00116-sch076){ref-type="scheme"}) \[[@B84-molecules-24-00116]\].
The authors claimed that with the appropriate connection of the two cholesterol molecules **314** and **322**, both proteins (NPC1 and NPC2) are simultaneously occupied in a manner that stabilizes the protein--protein interaction, allowing detailed structural analysis of the resulting complex. Furthermore, the introduction of a photoaffinity label in one of the cholesterol moieties, **322**, should allow the covalent attachment of one of the units into its respective protein-binding pocket. The compounds synthesized in this work may be interesting tools for studying the transfer of cholesterol between cholesterol-binding proteins \[[@B84-molecules-24-00116]\].
Two cholesterol-based fluorescent lipids, **326** and **329**, were synthesized using nitrobenzoxadiazole (NBD) or rhodamine B, respectively, linked by an ether alkyl chain ([Scheme 79](#molecules-24-00116-sch079){ref-type="scheme"}). Compounds **326** and **329** were incorporated into liposome formulations, aiming to create and validate their use as fluorescent probes for lipoplex tracking, without interfering with green fluorescent protein (GFP) \[[@B85-molecules-24-00116]\]. The authors concluded that both compounds **326** and **329** did not interfere with the expression of GFP plasmid, obtaining live cell images without any interference. Furthermore, microscopic observations clearly showed that these fluorescent lipids had minimal self-quenching and photobleaching effects. The results indicated that the synthesized compounds **326** and **329** may be considered for the development of fluorescent probes to trace the intracellular trafficking of cholesterol-derived cationic liposomes \[[@B85-molecules-24-00116]\].
Reibel et al. prepared radiolabeled-^18^F polymer compounds based on linear PEG **332** and novel linear-hyperbranched amphiphilic polyglycerol (*hb*PG) **334**, using cholesterol **28** as a lipid anchor via CuAAC chemistry of propargylated compounds **330** and **333** with radiolabeled-^18^F azide **331** ([Scheme 80](#molecules-24-00116-sch080){ref-type="scheme"}) \[[@B86-molecules-24-00116]\].
The authors also carried out direct labeling of cholesterol **28** with ^18^F ([Scheme 80](#molecules-24-00116-sch080){ref-type="scheme"}) and performed in vivo positron emission tomography (PET) studies as well as ex vivo biodistribution studies in mice with both polymers (**Ch-PEG~27~-CH~2~-triazole-TEG-^18^F 332** and **Ch-PEG~30~-*hb*PG~24~-CH~2~-triazole-TEG-^18^F 334**) and ^18^F-cholesteryl fluoride **336**. These three new derivatives were incorporated into liposome formulations. The results showed that both polymers **332** and **334** were quickly excreted by renal function, whereas ^18^F-cholesteryl fluoride **336** showed some retention in the lung, liver, and spleen. Liposome formulations with the new polymers showed different physical properties from those of the conventional liposomes with ^18^F-cholesteryl fluoride **336**, as well as fast uptake by the liver, spleen, and lung. Furthermore, the novel *hb*PG-polymer liposomes of **334** showed similar behavior to the PEG-shielded vesicles, enhancing multifunctionality without the loss of pharmacokinetic properties. This approach opens new possibilities in the field of polymer tracking in vivo and liposome tracing in mice via PET \[[@B86-molecules-24-00116]\].
In 2015, Palakollu and Kanvah designed and synthesized cholesterol-conjugated chromophores of α-cyanostilbene/diene **338** and **340** exhibiting intramolecular charge transfer (ICT) and aggregation-induced enhanced emission (AIEE). Compounds **338** and **340** were easily prepared from the reaction of cholesterol chloroformate **7** with either a stilbene **337** or diene derivative **339** ([Scheme 81](#molecules-24-00116-sch081){ref-type="scheme"}) \[[@B87-molecules-24-00116]\].
The authors carefully studied the absorption and emission properties of both cholesterol conjugates **338** and **340** and their parent chromophores **337** and **339**. An ICT behavior was observed for diene compounds **339** and **340**, whereas for stilbene compounds **337** and **338** a remarkable AIEE behavior was detected. The lack of AIEE characteristics in dienes may be explained by the competing nonradiative losses due to double bond flexibility. Nevertheless, the most interesting conclusion of the optical properties study was that the random aggregates formed by stilbene **337** in aqueous media became highly ordered upon cholesterol conjugation **338**. Furthermore, the interaction with sodium cholate stimulated the formation of self-assembled structures in nanoscale dimensions, making these conjugates the starting point for the development of several bioimaging probes \[[@B87-molecules-24-00116]\].
In 2016, Wercholuk and coworkers synthesized a fluorescent-labeled cholesterol molecule (**342**) by treating cholesteryl chloroformate **7** with 4-amino-1,8-naphthalimides (**341**) ([Scheme 82](#molecules-24-00116-sch082){ref-type="scheme"}) \[[@B88-molecules-24-00116]\]. The authors expected that such conjugates might serve one of two roles, depending on whether the toxicity of the fluorophore was retained in the conjugates: As reporters for following in vivo uptake or catabolism of cholesterol, or as "Trojan horse" antibiotics. The results pointed out that the new compounds (**342**) emitted blue light in nonpolar solvents, and its lipid portion incorporated into liposomal membrane bilayers quickly, leaving the fluorophore exposed to the external aqueous environment. Compounds in **342** were incubated with *Mycobacterium smegmatis* mc2 155, which displayed stable integration of the fluorescent-labeled cholesterols into bacterial membranes in vivo. Although fluorophores are toxic to prokaryotic cells, the new cholesterol conjugates (**342**) are not, and therefore they could be considered for the evaluation of cholesterol uptake in prokaryotic organisms \[[@B88-molecules-24-00116]\].
In the same year, Bernhard et al. reported an interesting paper in which they studied two strategies for the bioconjugation of bombesin (BBN), a well-known peptide, the receptor of which is overexpressed at the surface of tumor cells and which has been conjugated in several probes \[[@B89-molecules-24-00116]\]. They used subphthalocyanines (SubPcs), which are interesting probes for optical imaging. One of these strategies involved the entrapping of SubPc into a liposome and subsequently grafting BBN to the SubPc-containing liposome to afford a biovectorized liposome. The synthesis of cholesterol derivatives **346** and **347** used in their work was achieved by the reaction of dimethylaminopropyne **344** or 3-azidodimethylpropylamine **345** with cholesterol bromo ester **343** to afford cholesteryl-ammonium species **346** (alkynyl) and **347** (azide), respectively ([Scheme 83](#molecules-24-00116-sch083){ref-type="scheme"}) \[[@B89-molecules-24-00116]\].
Once the cholesteryl-ammonium species **346** and **347** were prepared, the pre-bioconjugation strategy started from grafting the biomolecule to one liposome's component (i.e., cholesterol additive) prior to the preparation of the liposome, to afford BBN-cholesterol conjugates **348** and **349**. The conjugation of BBN-azide with cholesteryl-alkyne **346** (i.e., pre-functionalization by copper-catalyzed click chemistry) was carried out in the presence of copper sulfate and sodium ascorbate as the reducing agents ([Scheme 84](#molecules-24-00116-sch084){ref-type="scheme"}) \[[@B89-molecules-24-00116]\]. Alternatively, BBN-bicyclononyne and cholesteryl-azide **347** were reacted without the Cu catalyst to afford conjugate **349** ([Scheme 84](#molecules-24-00116-sch084){ref-type="scheme"}) \[[@B89-molecules-24-00116]\]. This strategy was employed using liposomes containing graftable cholesterol derivatives, revealed itself as a more suitable approach in addressing the stability of SubPcs, and was achieved by copper-free click-chemistry on the outer face of the liposome. This study demonstrated that both azido- and alkynyl-liposomes are good entry points for a bioconjugation or biovectorization approach (on the outer face of the liposome), which offers a second chance for fluorophores with no reactive functional group available on their backbone, a way of imitating bioconjugation with a biomolecule (i.e., an indirect approach offered to achieve future site-specific targeting of tumors) \[[@B89-molecules-24-00116]\].
A series of new hybrid compounds (Ch-DAINs), **355**, **356**, and **360**, bearing a green fluorescent protein-chromophore analogue, 4-(diarylmethylene)imidazolinone (DAIN), and a cholesten or cholestane, was recently reported as a candidate for viscosity-dependent and cholesterol-responsive fluorescent molecules \[[@B90-molecules-24-00116]\]. The synthesis of Ch(en)-DAINs **355** and **356** was carried out through a condensation reaction of methyl imidates **352** or **358** (obtained from cholestenone through Beckmann rearrangement followed by methylation) with *N*-(diarylmethylene) glycinates **353** or **354** ([Scheme 85](#molecules-24-00116-sch085){ref-type="scheme"}) \[[@B90-molecules-24-00116]\]. Likewise, Ch(an)-DAINs **359** and **360** were obtained following the same synthetic strategy with an additional double-bond hydrogenation step ([Scheme 85](#molecules-24-00116-sch085){ref-type="scheme"}) \[[@B90-molecules-24-00116]\].
Among the tested compounds, cholesten DAINs **355** and **356** increased their fluorescence intensity in viscous solvents such as triglycerides. Besides, compound **355** showed good cholesterol-responsive emission, which increased linearly with the amount of cholesterol in the lipid bilayer. The responsiveness displayed by cholesten DAIN **355** to cholesterol was improved relatively to the known viscosity probes, 9-(2,2-dicyanovinyl)julolidine (DCVJ) and Laurdan \[[@B90-molecules-24-00116]\].
7. Synthetic Applications {#sec7-molecules-24-00116}
=========================
The regio- and stereoselective formation of *O*-glycosidic bonds between carbohydrates and steroids is still a demanding process, despite the considerable progress in carbohydrate chemistry in the last years. The direct electrochemical glycosylation of steroids is an alternative: However, it has several drawbacks. In attempting to solve the problem, Tomkiel et al. screened several derivatives of cholesterol as sterol donors in electrochemical reactions with sugar alcohols \[[@B91-molecules-24-00116]\]. The authors tested sixteen cholesterol substrates in the presence of 1,2:3,4-di-*O*-isopropylidene-α-[d]{.smallcaps}-galactopyranose (**362**), concluding that cholesteryl diphenylphosphate **361** was the best compound for the purpose, affording 3β-*O*-(1′,2′:3′,4′-di-*O*-isopropylidene-α-[d]{.smallcaps}-galactopyranos-6′-yl)-cholest-5-ene (**363**) in 54% yield ([Scheme 86](#molecules-24-00116-sch086){ref-type="scheme"}) \[[@B91-molecules-24-00116]\].
Following this work, the same authors reported in 2015 the use of 3α,5α-cyclocholestan-6β-yl alkyl and aryl ethers (**364**) as a cholesteryl donor in the electrochemical synthesis of glycoconjugates (**363**) ([Scheme 87](#molecules-24-00116-sch087){ref-type="scheme"}) \[[@B92-molecules-24-00116]\]. The reaction worked well for all the tested compounds, but the best yields were achieved for ethyl, benzyl, phenyl, and *tert*-butyldimethylsilyl (TBDMS) ethers (51%, 50%, 58%, and 52%, respectively). Unfortunately, an isomerization side reaction was observed for the less reactive cholesteryl esters, affording the compounds in **365** ([Scheme 87](#molecules-24-00116-sch087){ref-type="scheme"}) \[[@B92-molecules-24-00116]\].
To develop step-economy syntheses of cholesteryl glycosides, Davis and coworkers reported a methodology for the synthesis of α-[d]{.smallcaps}-cholesteryl glycosides **369** and **372**, using a one-pot per-*O*-trimethylsilyl glycosyl iodide glycosylation ([Scheme 88](#molecules-24-00116-sch088){ref-type="scheme"}) \[[@B93-molecules-24-00116]\]. The methodology relied first on the generation of glucosyl or galactosyl iodide through the reaction of per-*O*-TMS glucoside **366** or **370** with iodotrimethylsilane (TMSI), which was directly cannulated into a solution of cholesterol, tetrabutylammonium iodide (TBAI), and *N*,*N*-diisopropylethylamine (DIPEA), and the mixture was stirred for 2 days at room temperature. After that, the product was treated with methanol and Dowex-50WX8-200 acidic resin to remove the silyl protecting groups, affording compounds **367** and **371** ([Scheme 88](#molecules-24-00116-sch088){ref-type="scheme"}) \[[@B93-molecules-24-00116]\]. These glycosides were subsequently esterified using regioselective enzymatic acylation of the 6-hydroxy group with tetradecanoyl vinyl ester **368** ([Scheme 88](#molecules-24-00116-sch088){ref-type="scheme"}) \[[@B93-molecules-24-00116]\].
This methodology involving the glycosylation of cholesterol followed by enzymatic regioselective acylation allowed expansion of the acylated α-cholesteryl glycoside inventory to include galactose analogues. The glycosylation of per-*O*-silylated glucose provided better α-selectivity (39:1) than past syntheses (8:1 α-selectivity) and higher glycosylation yields due to the armed nature of per-*O*-silyl donors \[[@B93-molecules-24-00116]\].
Mao and coworkers developed a novel glycosyl coupling reaction, involving a photoinduced direct activation mechanism of thioglycosides (**373**) and subsequent *O*-glycosylation in the absence of photosensitizer \[[@B94-molecules-24-00116]\]. In their studies, the authors used several sugars, amino acids, and cholesterol **28** (75%) as substrates ([Scheme 89](#molecules-24-00116-sch089){ref-type="scheme"}). The authors showed that the activation of thioglycosides upon UV irradiation followed by the oxidation of Cu(OTf)~2~ led to the in situ formation of species that could undergo glycosylation to afford glycosides without the need for a photosensitizer. The proposed mechanism involved i) homolytic cleavage of a C-S bond to generate a glycosyl radical and ii) oxidation to an oxacarbenium ion promoted by Cu(OTf)~2~, and sequential *O*-glycosylation \[[@B94-molecules-24-00116]\].
In 2015, Davis and coworkers reported the synthesis of cholesteryl-α-[d]{.smallcaps}-lactoside **378** via generation and trapping of stable β-lactosyl iodide **376**. The iodide derivative **376** was prepared quantitatively under non-in situ anomerization and metal-free conditions by reacting commercially available β-per-*O*-acetylated lactose **375** with trimethylsilyl iodide \[[@B95-molecules-24-00116]\]. The introduction of cholesterol occurred under microwave conditions to afford the corresponding glycoconjugate **377** in 59% yield ([Scheme 90](#molecules-24-00116-sch090){ref-type="scheme"}). Cholesterol glycoconjugate **377** was further deacetylated using sodium methoxide to afford cholesteryl α-[d]{.smallcaps}-lactoside **378** in 88% yield ([Scheme 90](#molecules-24-00116-sch090){ref-type="scheme"}). This glycosylation method can be employed on sterically demanding nucleophiles such as cholesterol and has potential applications in accessing structurally diverse cholesteryl glycoside analogs \[[@B95-molecules-24-00116]\].
A new efficient method for the synthesis of cholesteryl glucosides starting from sucrose **379** was recently developed \[[@B96-molecules-24-00116]\]. This method lays down a five-step synthetic route that involves the initial protection of disaccharide **379** hydroxy groups, and upon acidic hydrolysis at its anomeric center, the pyranosyl moiety **381** is converted into trichloroacetimidate derivative **383** ([Scheme 91](#molecules-24-00116-sch091){ref-type="scheme"}).
The final two steps rely on the formation of the glycosidic bond to cholesterol **28** followed by the removal of the protecting groups, affording the desired cholesteryl glucoside **384** ([Scheme 91](#molecules-24-00116-sch091){ref-type="scheme"}). The authors claimed that the major advantage of this strategy was the use of the readily available and cheap sucrose **379** as starting material. In addition, the methodology proved to be fast, cost-saving, and high-yielding, representing a competitive preparation method for these natural compounds \[[@B96-molecules-24-00116]\].
In 2014, Algay and coworkers extensively explored the versatility of nitrile oxide alkyne cycloaddition (NOAC) chemistry for the formation of cholesterol conjugates anchored by way of a polar, aromatic, metabolically stable isoxazole nucleus \[[@B97-molecules-24-00116]\]. The first series of compounds produced in this paper involved i) the microwave-assisted formation of propargyl ethers (**386**) in 62%--70% yield ([Scheme 92](#molecules-24-00116-sch092){ref-type="scheme"}a), and ii) the reaction of cholesterol propargyl ethers (**386**) with phenyl nitrile oxide (generated in situ from benzaldehyde oxime upon exposure to an ethanolic solution of chloramine-T) ([Scheme 92](#molecules-24-00116-sch092){ref-type="scheme"}b). This last reaction was carried out at room temperature or under microwave heating depending on the length of the spacing between the bulky lipid and the reacting alkyne, affording isoxazoles (**387**) in fair to excellent yields (35%--91%) \[[@B97-molecules-24-00116]\]. The authors extended a bit further this reaction to prepare biologically relevant cholesterol fluorescent probes such as steroid--coumarin (**391**) (75%) and steroid--azobenzene conjugates (**389**) (56%) ([Scheme 92](#molecules-24-00116-sch092){ref-type="scheme"}). It is known that long-chain hydrophilic linkers are very attractive for bioconjugation and therefore, in this paper, the authors also synthesized three new ether-linked isoxazole-cholesterol conjugates (**396**) in 29%--58% yield ([Scheme 92](#molecules-24-00116-sch092){ref-type="scheme"}) \[[@B97-molecules-24-00116]\].
Another series of isoxazole-cholesterol conjugates (**401**) was also prepared, starting from cholesterol chloroformate **7** and bearing an amidocarbamate linker following the four-step synthetic route depicted in [Scheme 93](#molecules-24-00116-sch093){ref-type="scheme"} \[[@B97-molecules-24-00116]\].
The nontrivial synthesis of aryl ethers of natural alcohols drove the authors to test the NOAC chemistry in the assembly of aryl ether cholesterol conjugates \[[@B97-molecules-24-00116]\]. Therefore, isoxazole-linked aryl cholesterol ether **404** was prepared from the aldehyde-functionalized aryl ether **402** through subsequent oximation and cycloaddition reactions, as depicted in [Scheme 94](#molecules-24-00116-sch094){ref-type="scheme"}.
Finally, the authors used the potential of NOAC chemistry to prepare a steroidal glycoconjugate, **407**, and the selective tethering of one or two cholesterol units, **409** and **410**, respectively, to a thymidine skeleton was demonstrated by trapping of the same dipole by 5′-protected mono- or bis-propargylated thymidines ([Scheme 95](#molecules-24-00116-sch095){ref-type="scheme"}) \[[@B97-molecules-24-00116]\].
In 2016, Alarcón-Manjarrez and coworkers reported the synthesis of two dimeric steroidal terephthalates, **415** and **416**, from epimeric 4,5-seco-cholest-3-yn-5-ols **413** and **414**, using a five-step synthetic route with cholesterol **28** as a starting material \[[@B98-molecules-24-00116]\]. The synthetic route first involved the Oppenauer oxidation of cholesterol **28**, followed by epoxidation, to afford a mixture of epoxides (**411**) (α:β = 1:4) ([Scheme 96](#molecules-24-00116-sch096){ref-type="scheme"}). Then, an Eschenmoser-Tanabe fragmentation followed by carbonyl group reduction provided the epimeric alkynols **413** and **414** in a 1:2 ratio ([Scheme 96](#molecules-24-00116-sch096){ref-type="scheme"}). Finally, the treatment of each one of the epimeric alkynols **413** and **414** with terephthaloyl chloride led to the symmetrical axial and equatorial dimers **415** and **416**, respectively ([Scheme 96](#molecules-24-00116-sch096){ref-type="scheme"}) \[[@B98-molecules-24-00116]\].
The authors proceeded to crystallographic analysis of the compounds and concluded that the facial hydrophobicity of the steroidal skeletons had crucial influence on the crystal packing in which the dimeric molecules were forced to accommodate these fragments only with a few hydrogen-bonding interactions. This feature originated a cisoid conformation for **415** and a linear conformation for **416** \[[@B98-molecules-24-00116]\].
Shibuya et al. reported in 2016 the synthesis of (24*S*)-hydroxycholesterol (24*S*-OHChol) esters, which are involved in neuronal cell death, through catalysis with acyl-CoA:cholesterol acyltransferase-1 (ACAT-1) \[[@B99-molecules-24-00116]\]. The authors studied the esterification of (24*S*)-OHChol **417** with *cis*-oleoyl chloride under basic conditions and obtained mono-oleates **418** and **419** and bis-oleate **420** in 39%, 9%, and 20% yields, respectively ([Scheme 97](#molecules-24-00116-sch097){ref-type="scheme"}). The protection of (24*S*)-OH with a trifluoroacetyl group was also attempted, affording mono-trifluoroacetates **421** and **422** in 33% and 14% yields, respectively, and the bis-trifluoroacetate **423** in 21% yield ([Scheme 97](#molecules-24-00116-sch097){ref-type="scheme"}) \[[@B99-molecules-24-00116]\]. The authors took advantage of the mono-trifluoroacetate **421** to prepare the stearoyl and palmitoyl esters **427** and **428** in 68% and 75% yields, respectively, as depicted in [Scheme 98](#molecules-24-00116-sch098){ref-type="scheme"} \[[@B99-molecules-24-00116]\]. Finally, the authors also reported the use of esters of unsaturated long-chain fatty acids, such as linoleic (LA), arachidonic (AA), and docosahexaenoic (DHA), to react with cholesterol derivative **422** in order to prepare linoleate **430**, arachidonoate **431,** and docosahexaenoate **432** esters, in 52%, 74%, and 66% yields, respectively, in a two-step synthetic route, as depicted in [Scheme 99](#molecules-24-00116-sch099){ref-type="scheme"} \[[@B99-molecules-24-00116]\].
Recently, Sarkar et al. reported interesting work dealing with the preparation of diverse ring-A or ring-B oxo-functionalized steroids in a green fashion involving solvent-free solid supports \[[@B100-molecules-24-00116]\]. The authors used cholesterol derivatives such as 4β-hydroxycholesterol **433**, which was functionalized into three different keto-steroids, **434**, **435**, and **436**, in 55%, 10%, and 10% yields, respectively, employing *p*-toluenesulfonic acid and SiO~2~ (silica 60--120 mesh) as solid support ([Scheme 100](#molecules-24-00116-sch100){ref-type="scheme"}) \[[@B100-molecules-24-00116]\]. Interestingly, if the reaction was attempted in the solution phase at room temperature using either dichloromethane or ethanol as solvents, cholest-4-en-3-one **434** was obtained exclusively in 64% and 60% from dichloromethane and ethanol, respectively. The procedure on solid silica was applied to the other cholesterol derivative, namely 4β,7α-dihydroxycholesterol **437**, which was converted into four distinct keto-steroids: (i) cholest-5-en-7-one (**438**, 8%), (ii) cholesta-3,5-dien-7-one (**439**, 13%), (iii) cholesta-4,6-dien-3-one (**440**, 17%), and (iv) 5a-cholestane-4,7-dione (**441**, 7%) ([Scheme 100](#molecules-24-00116-sch100){ref-type="scheme"}) \[[@B100-molecules-24-00116]\]. It is worth noticing that if the reaction of 4β,7α-dihydroxycholesterol **437** was carried out in dichloromethane as solvent, cholesta-4,6-dien-3-one (**440**, 54%) was found to be the only product formed. This was found to be a facile procedure for the synthesis of dienone **440** from cholesterol via triol **437** \[[@B100-molecules-24-00116]\].
Cholesterol derivatives can also be used as starting materials for the synthesis of fused nitrogen heterocycles. This was the case for 4-cholesten-3-one **350**, which was involved in the preparation of A-ring dehydropiperazine **443** (90% yield) through a microwave-assisted annulation reaction with ethylenediamine **442** in the presence of basic alumina ([Scheme 101](#molecules-24-00116-sch101){ref-type="scheme"}) \[[@B101-molecules-24-00116]\].
The proposed mechanism should encompass the initial oxidation of the allylic protons of the conjugated ketone via enolate intermediate to afford a diketo intermediate. Then, the condensation with ethylenediamine followed by a Michael addition and autoxidation reactions afforded the dehydropiperazine derivatives \[[@B101-molecules-24-00116]\].
Recently, Ansari and coworkers reported an efficient and green synthetic method for the preparation of steroidal pyridines \[[@B102-molecules-24-00116]\]. Such methodology relied on the utilization of MgO NPs as a heterogeneous, mild, and reusable catalyst, in a multicomponent one-pot protocol, taking advantage of the usefulness of the microwave irradiation as an alternative heating source. The series of substituted fused pyridines (**444**) were obtained in 80%--89% yield from the reaction of steroidal ketones (**164**) with malononitrile/methylcyanoacetate, benzaldehyde, and ammonium acetate in ethanol using MgO NPs as a catalyst ([Scheme 102](#molecules-24-00116-sch102){ref-type="scheme"}) \[[@B102-molecules-24-00116]\].
One of the key mechanistic steps in this kind of multicomponent reaction is the standard Knoevenagel condensation of benzaldehyde and malononitrile/methyl cyanoacetate. The effect of MgO NPs can be rationalized on this basis since they are known as a highly effective heterogeneous base catalyst for Michael addition and Knoevenagel condensation reactions with Mg^2+^ (Lewis acid) and O^2−^ (Lewis base) sites along with various cationic and anionic vacancies in the lattice \[[@B102-molecules-24-00116]\].
The A-ring of cholesterol was also functionalized by fusing pyrimidines at the steroidal 2,3-position. These new steroidal compounds were synthesized through a microwave-assisted three-component reaction of 2-hydroxymethylene-3-ketosteroid (**445**), benzaldehydes (**446**), and ammonium acetate, affording cholesterol-fused pyrimidines (**447**) in good yields (78%--88%) ([Scheme 103](#molecules-24-00116-sch103){ref-type="scheme"}) \[[@B103-molecules-24-00116]\].
The authors' mechanism was based on: (i) microwave-assisted reaction of ammonia (released from decomposition of ammonium acetate) with 2-hydroxymethylene-3-ketosteroid to afford a β-aminoketoimine intermediate; (ii) their condensation reaction with benzaldehydes to afford a diamine intermediate; and (iii) cyclization and subsequent auto-oxidation to give the cholesterol-fused pyrimidines \[[@B103-molecules-24-00116]\].
A two-step method for the preparation of steroid-fused 4,6-diaryl substituted pyridines has been reported \[[@B104-molecules-24-00116]\]. The synthetic protocol relied on the Michael addition of 5α-cholestan-3-one **448** with chalcones (generated in situ by the base-catalyzed reaction of acetophenones (**449**) and benzaldehydes (**446**)), affording 3,5-diaryl-1,5-dicarbonyl 5α-cholestan-3-one derivatives (**450**) (88%--94%) ([Scheme 104](#molecules-24-00116-sch104){ref-type="scheme"}). Then, the intermediates (**450**) were used as substrates in a microwave-assisted solid phase reaction with urea in the presence of BF~3~·OEt~2~ to give 4,6-diaryl substituted pyridines (**451**) in good yields (81%--93%) ([Scheme 104](#molecules-24-00116-sch104){ref-type="scheme"}) \[[@B104-molecules-24-00116]\].
The authors proposed a mechanism for the formation of a pyridine ring, which may start with the release of ammonia by urea under microwave heating, which forms an imine by reaction with one carbonyl group. Next, the BF~3~·OEt~2~-promoted nucleophilic attack of the imine NH-group on the activated carbonyl functionality facilitated an aza-cyclization reaction, affording a 1,4-dihydropyrinine intermediate upon which aromatization gave the desired 4,6-diarylpyridines \[[@B104-molecules-24-00116]\].
In 2015, Schulze and coworkers developed a new method for the synthesis of model asphaltene compounds. The reported methodology was based on a multicomponent cyclocondensation reaction of 2-aminoanthracene **452** with aromatic aldehydes and 5-α-cholestan-3-one **448** ([Scheme 105](#molecules-24-00116-sch105){ref-type="scheme"}) \[[@B105-molecules-24-00116]\].
The authors found that the actual catalyst for this reaction was indeed hydriodic acid, which is formed in situ from the reaction of iodine with water. Carrying the reaction under anhydrous conditions, it was proven that iodine itself did not promote the reaction, as generally assumed. Using this methodology, the authors prepared a library of optically active steroidal naphthoquinolines (**453**) in acceptable yields (40%--53%) \[[@B105-molecules-24-00116]\].
8. Miscellaneous {#sec8-molecules-24-00116}
================
The design of (supra)molecular switches and machines has a key feature that relates to the control of mechanical motions at the molecular level. In this field, rotaxanes have attracted much attention because they offer the possibility of restricting the freedom of motion to some well-defined pathways, such as the translational motion of a rotaxane's ring along its axis in a shuttling manner. The synthesis of a novel nonsymmetrical bistable pH-sensitive rotaxane with a cholesterol stopper at one end and a tetraphenylmethane group at the other end (**457**), has been reported \[[@B106-molecules-24-00116]\]. The synthesis of both terminal ends was challenging, and therefore we only describe here the final step, which consisted of joining both axes of the nonsymmetrical rotaxane, the alkyne **454**, and the azide **455** through CuAAC chemistry, affording compound **456** ([Scheme 106](#molecules-24-00116-sch106){ref-type="scheme"}). The formation of a pH-sensitive bistable rotaxane **457** was achieved by methylation of the triazole ring using methyl iodide ([Scheme 106](#molecules-24-00116-sch106){ref-type="scheme"}). The authors verified that the crown ether part changed its preferred position on the axis because of the protonation state of a secondary amine. More specifically, the crown ether was located around the secondary ammonium ion as the best binding site in the protonated form. On the other hand, NMR analysis showed that upon deprotonation of the ammonium ion, the triazolium ion became the better binding site, which caused the ring to shuttle along the axis toward this position ([Scheme 106](#molecules-24-00116-sch106){ref-type="scheme"}) \[[@B106-molecules-24-00116]\].
Venkataraman and coworkers reported the two-step synthesis of cholesterol-functionalized aliphatic *N*-substituted 8-membered cyclic carbonate monomer **459** ([Scheme 107](#molecules-24-00116-sch107){ref-type="scheme"}) \[[@B107-molecules-24-00116]\]. Cholesterol-based monomer **459** was employed in organocatalytic ring-opening polymerization to produce PEGylated amphiphilic diblock copolymers (using a commercially available macroinitiator), polyethylene glycol monomethyl ether (mPEG-OH) **460** ([Scheme 107](#molecules-24-00116-sch107){ref-type="scheme"}). The authors evaluated the behavior of these copolymers in aqueous media, concluding that they self-assembled to form unique nanostructures, including disk-like micelles. The experimental results also suggested that the prepared copolymers can be used as inexpensive steric stabilizers for liposomes, making them suitable for several biomedical applications \[[@B107-molecules-24-00116]\].
Recently, a cholesterol-modified poly([l]{.smallcaps}-cysteine) copolymer, **466**, that can undergo unusual micelle-to-vesicle transformation of polypeptides triggered by oxidation, was synthesized following a three-step protocol starting from cholesteryl 3-bromopropylcarbamate **462** ([Scheme 108](#molecules-24-00116-sch108){ref-type="scheme"}) \[[@B108-molecules-24-00116]\]. The thioether groups in the side chains of **466** were further oxidized to the corresponding sulfone derivative **467** ([Scheme 108](#molecules-24-00116-sch108){ref-type="scheme"}). The authors demonstrated that oxidation of the thioether groups in the side chains could change the packing characteristics of cholesterol groups and the peptide backbone, resulting in the transformation of a β-sheet to an α-helix conformation, combined with an interesting morphological transition from micelle-like structures to vesicles. Moreover, changing the secondary structure as well as the morphology endowed the polymer assemblies with excellent specificity for controlled payload release and improved cell interaction in response to ROS. These interesting formulations had excellent anticancer properties both in vitro and in vivo \[[@B108-molecules-24-00116]\].
Gramine \[*N*-(1*H*-indol-3-ylmethyl)-*N*,*N*-dimethylamine\] is a well-known indole derivative and is often used as synthon for the preparation of a large variety of substituted indoles with important biological activities. In this context, Kozanecka and coworkers reported the use of gramine (**470**) to synthesize cholesterol (**471**) and cholestanol dimers (**472**) consisting of two molecules of sterols connected by an N(CH~3~)~2~ group ([Scheme 109](#molecules-24-00116-sch109){ref-type="scheme"}) \[[@B109-molecules-24-00116]\].
These new steroid dimers (**471** and **472**) were shown to interact in vitro with the human erythrocyte membrane, changing the discoid erythrocyte shape, which resulted in induced stomatocytosis or echinocytosis. The authors also demonstrated that these new dimers were capable of interfering with membrane phospholipid asymmetry and loosening the molecular packing of phospholipids in the bilayer at sublytic concentrations. Moreover, the dimers **471** and **472** possessed a higher capacity for changing the erythrocyte membrane structure and its permeability than steroids alone did \[[@B109-molecules-24-00116]\].
A new multifunctional pyridine derivative was synthesized and studied as an efficient initiator for the polymerization of diethylvinylphosphonate (DEVP). The authors used a new pyridine compound (**473**) in the thiol-ene click reaction (a well-established coupling method) to link together poly-DEVP and thiocholesterol **95** ([Scheme 110](#molecules-24-00116-sch110){ref-type="scheme"}) \[[@B110-molecules-24-00116]\].
Compound **474** exhibited good thermal response and low cytotoxicity against human embryonic renal cell lines (HEK-293) and immortalized human microvascular endothelial cells (HMEC-1). It was concluded that the introduction of the thiocholesterol anchor unit was advantageous regarding toxicity when compared to polymers without functionalization. The thiocholesterol conjugate **474** is interesting for many applications, since it is water-soluble, thermo-responsive, and biocompatible \[[@B110-molecules-24-00116]\].
To take advantage of the important biological properties of cholesterol and glutathione for the cells, a cholesterol-glutathione (**Chol-GSH**) bioconjugate (**478**) was designed and used as a model amphiphilic biomolecule to make a co-assembly with lysozyme using a dialysis-assisted approach \[[@B111-molecules-24-00116]\]. The synthetic route toward the **Chol-GSH** bioconjugate **478** involved a five-step reaction sequence, including esterification, 1,3-dipolar cycloaddition, and thiol‒disulfide exchange reactions ([Scheme 111](#molecules-24-00116-sch111){ref-type="scheme"}).
The authors applied a dialysis-assisted method of Ch-GSH and lysozyme to prepare bioactive self-assembled structures, which showed that hydrophobic cholesterol located in the walls, and hydrophilic GSH and lysozyme on the inner and outer surfaces. This result was explained based on the electrostatic interaction between GSH and lysozyme, which provided a driving force for the self-assembly, maintaining the bioactivity of lysozyme in the self-assembly process \[[@B111-molecules-24-00116]\].
9. Conclusions {#sec9-molecules-24-00116}
==============
In this review, the role of cholesterol-based compounds in different research areas such as drug delivery, biological activities, liquid crystals, gelators, bioimaging, and purely synthetic applications was highlighted. In the drug delivery field, several examples of cholesterol derivatives were highlighted due to their applications in preclinical and clinical liposomal drug formulations to decrease membrane fluidity and provide favorable drug retention properties. Furthermore, in the last few years, some series of new cholesterol derivatives have also been developed for pharmacological applications as anticancer, antimicrobial, or antioxidant agents. In the bioimaging field, cholesterol has been used as a lipid anchor attached to fluorophores to study cellular membrane trafficking, imaging of cholesterol density, and liposome tracing, among many other bioimaging applications. The fact that cholesterol conjugates have much scientific interest in the field of materials science due to their liquid crystal phase behavior, as well as the ability to promote self-organization and hydrophobic interactions in aqueous media (gelation properties), was also demonstrated in this review. In this review, a general perspective was given of the main applications of cholesterol derivatives in several research fields, but also a concise perspective of the advances in their synthetic chemistry. Therefore, we described the synthetic pathway for different cholesterol derivatives alongside the corresponding application of the new compounds to furnish a general view from the synthetic and biological aspects of the most recently reported cholesterol-based compounds.
Thanks are due to the University of Aveiro, Instituto Politécnico de Bragança, FCT/MEC for financial support of the QOPNA (FCT UID/QUI/00062/2013) and CIMO (UID/AGR/00690/2013) research units, through national funds, and where applicable cofinanced by the FEDER, within the PT2020 Partnership Agreement; and also to the Portuguese NMR Network. This work was also supported by the Integrated Programme of SR&TD "pAGE--Protein aggregation Across the Lifespan" (reference CENTRO-01-0145-FEDER-000003), co-funded by the Centro 2020 program, Portugal 2020, European Union, through the European Regional Development Fund. H. M. T. Albuquerque thanks the pAGE project for his Post-Doc grant (BPD/UI98/4861/2017).
The authors declare no conflict of interest.
Ac
acetyl
Ac
2
O
acetic anhydride
AcOH
acetic acid
AG
arabinogalactan
AIBN
2,2′-azobis(2-methylpropionitrile)
AIEE
aggregation induced enhanced emission
AL
β-alanine
AscONa
sodium ascorbate
ATRP
atom transfer radical polymerization
BBN
bombesin
Bn
benzyl
Boc
2
O
di-
tert
-butyl decarbonate
BODIPY
boron dipyrromethene
BtOH
N
-hydroxybenzotriazole
Bz
benzoyl
CAE
cholesterol-arginine ester
β-CD
β-cyclodextrin
β-CD-NSP
β-cyclodextrin nanosponge
CDI
carbonyldiimidazole
CF
5,6-carboxyfluorescein
Chol
cholesterol
Chol-OA
oxyamine-terminated cholesterol
CHS
cholesterol hydrogen succinate
Ch-T
chloramine-T
CL
conventional liposomes
CuAAC
copper(I)-catalyzed azide-alkyne cycloaddition
CVS
crystal violet staining
Cyclen
1,4,7,10-tetraazacyclododecane
CYS
cystamine
DAIN
4-(diarylmethylene)imidazolinone
DBU
1,8-diazabicyclo\[5.4.0\]undec-7-ene
DCC
N
,
N
′-dicyclohexylcarbodiimide
DCVJ
9-(2,2-dicyanovinyl)julolidine
DHPC
dihexanoylphosphatidylcholine
DIAD
diisopropyl azodicarboxylate
DIPEA
N
,
N
-diisopropylethylamine
DMAP
dimethylaminopyridine
DMF
dimethylformamide
DMPC
dimyristoylphosphatidylcholine
DMSO
dimethyl sulfoxide
DMTAP
dimyristoyltrimethylammonium propane
DNA
deoxyribonucleic acid
DOPC
dioleoylphosphatidilcholine
DOPE
1,2-dioleoyl-
sn
-glycero-3-phosphoethanolamine
DOX
doxorubicin
DOX-NPs
doxorubicin loaded nanoparticles
DPPA
diphenylphosphoryl azide
DPPH
2,2-diphenyl-1-picrylhydrazyl radical
DTX
docetaxel
EDAC
1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride
EDCl
N
-(3-dimethylaminopropyl)-
N
′-ethylcarbodiimide hydrochloride
Et
2
N
diethylamine
Et
3
N
triethylamine
Et
2
O
diethyl ether
EtOAc
ethyl acetate
EtOH
ethanol
GFP
green fluorescent protein
GSH
glutathione
HA
hyaluronic acid
hb
PG
linear-hyperbranched amphiphilic polyglycerol
HOBt
N
-hydroxybenzotriazole
HSA
human serum albumin
IC
50
half-maximal inhibitory concentration
ICT
intramolecular charge transfer
l
-AA
l
-ascorbic acid
LC
liquid crystal
LMGs
low-molecular-weight gelators
MeCN
acetonitrile
MeOTf
methyl triflate
m
-CPBA
m
-chloroperoxybenzoic acid
min
minutes
MPP
mitochondria-penetrating peptide
MTT
3-(4,5-dimethylthiazo-2-yl)-2,5-diphenyltetrazolium bromide
MW
microwave irradiation
NaOAc
sodium acetate
NaOMe
sodium methoxide
NBD
nitrobenzoxadiazole
NBS
N
-bromosuccinimide
NHS
N
-hydroxysuccinimide
NOAC
nitrile oxide alkyne cycloaddition
NPs
nanoparticles
PBS
phosphate-buffered saline
PDC
pyridinium dichromate
PEG
polyethylene glycol
PET
positron emission tomography
p
HPMA
poly\[
N
-(2-hydroxypropyl)-methacrylamide\]
PMDETA
N
,
N
,
N
′,
N
″,
N
″-pentamethyldiethylenetriamine
PPA
phenylpropanolamine
PPh
3
triphenylphosphine
p
-TsCl
p
-toluenesulfonyl chloride
PTX
paclitaxel
PyBOP
benzotriazole-1-yl-oxytripyrrolidinophosphonium hexafluorophosphate
RAFT
reversible addition fragmentation chain transfer
rt
room temperature
SA
succinic anhydride
SCID
severe combined immunodeficient
SML
surface-modified liposomes
SubPcs
subphthalocyanines
TACN
1,4,7-triazacyclononane
Tb
trilobolide
TBAB
tetrabutylammonium bromide
TBAF
tetrabutylammonium fluoride
TBAH
tetrabutylammonium hydroxide
TBAITBDMS
tetrabutylammonium iodide
tert
-butyldimethylsilyl
TBDMSCl
tert
\-
butyldimethylsilyl chloride
Tb-N
3
VA
trilobolide 8-
O
-azidovalerate
TBTA
tris\[(1-benzyl-1
H
-1,2,3-triazol-4-yl)methyl\]amine
t
-BuOH
tert
-butanol
t
-BuOK
potassium
tert
-butoxide
t
-BuOOH
tert
-butyl hydroperoxide
TE
transfection efficiency
TEG
tetraethylene glycol
TFA
trifluoroacetic acid
TFAA
trifluoroacetic anhydride
THF
tetrahydrofuran
TIS
triisopropylsilane
TMSCl
trimethylsilyl chloride
TMSITMSOTf
iodotrimethylsilanetrimethylsilyl trifluoromethanesulfonate
TOAB
tetraoctylammonium bromide
TsCl
p
-toluenesulfonyl chloride
Figures and Schemes
===================
{#molecules-24-00116-f001}
{#molecules-24-00116-f002}
{#molecules-24-00116-sch001}
{#molecules-24-00116-sch002}
{#molecules-24-00116-sch003}
{#molecules-24-00116-sch004}
![Synthesis of cholesterol-based neoglycoconjugates derived from [d]{.smallcaps}-galactose and *N*-acetylglucosamine. Reagents and conditions: **a**) 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDAC), DMAP, acetone/THF, rt, 4 h; **b**) glycosyl azide, CuSO~4~·5H~2~O, AscONa, THF/H~2~O, rt, 4 h; **c**) NaOMe, MeOH, 0 °C, 1 h.](molecules-24-00116-sch005){#molecules-24-00116-sch005}
![Synthesis of cholesterol-trilobolide conjugate. Reaction conditions: **a**) CuSO~4~·5H~2~O, AscONa, tris\[(1-benzyl-1*H*-1,2,3-triazol-4-yl)methyl\]amine (TBTA), dimethylformamide (DMF), microwave irradiation (MW), 60 °C, 90 min; **b**) amino-PEG~4~-acetylene, *N*-(3-dimethylaminopropyl)-*N*′-ethylcarbodiimide hydrochloride (EDCl), DMAP, *N*-hydroxybenzotriazole (HOBt), DMF, rt, 24 h; **c**) Tb-N~3~VA, CuSO~4~·5H~2~O, AscONa, TBTA, DMF, MW, 60 °C, 5 h.](molecules-24-00116-sch006){#molecules-24-00116-sch006}
{#molecules-24-00116-sch007}
{#molecules-24-00116-sch008}
{#molecules-24-00116-f003}
{#molecules-24-00116-sch009}
{#molecules-24-00116-sch010}
{#molecules-24-00116-f004}
{#molecules-24-00116-sch011}
{#molecules-24-00116-sch012}
{#molecules-24-00116-sch013}
{#molecules-24-00116-sch014}
{#molecules-24-00116-sch015}
{#molecules-24-00116-sch016}
{#molecules-24-00116-sch017}
{#molecules-24-00116-sch018}
{#molecules-24-00116-sch019}
{#molecules-24-00116-sch020}
{#molecules-24-00116-sch021}
{#molecules-24-00116-sch022}
{#molecules-24-00116-sch023}
{#molecules-24-00116-sch024}
{#molecules-24-00116-sch025}
{#molecules-24-00116-sch026}
{#molecules-24-00116-sch027}
![Synthesis of 6a-aza-[b]{.smallcaps}-homo lactams. Reagents and conditions: **a**) NaNO~2~, Ac~2~O, BF~3~OEt~2~, AcOH, rt, 1 h; **b**) Na~2~CO~3~, CH~2~Cl~2~/MeOH (1:1), reflux, 2.5 h; **c**) SOCl~2~, dioxane, rt, 20 min.](molecules-24-00116-sch028){#molecules-24-00116-sch028}
{#molecules-24-00116-sch029}
{#molecules-24-00116-sch030}
{#molecules-24-00116-sch031}
![Synthesis of pharmacophoric motifs by copper-catalyzed 1,3-dipolar cycloaddition (CuAAC). Reaction conditions: **a**) CuSO~4~·5H~2~O, [l]{.smallcaps}-AA, THF/H~2~O (4:1), reflux.](molecules-24-00116-sch032){#molecules-24-00116-sch032}
{#molecules-24-00116-sch033}
![Synthesis of cholesterol-triazole dimer, cholesterol-triazole alkanes, and cholesterol-triazole carbohydrates. Reagents and conditions: **a**) CuSO~4~·5H~2~O, [l]{.smallcaps}-AA, THF/H~2~O (5:1), reflux, 3 h; **b**) CBr~4~, PPh~3~, CH~2~Cl~2~, rt, overnight; **c**) NaOMe, MeOH, rt, 2 h.](molecules-24-00116-sch034){#molecules-24-00116-sch034}
{#molecules-24-00116-sch035}
{#molecules-24-00116-sch036}
{#molecules-24-00116-sch037}
{#molecules-24-00116-sch038}
{#molecules-24-00116-sch039}
{#molecules-24-00116-sch040}
{#molecules-24-00116-sch041}
{#molecules-24-00116-sch042}
{#molecules-24-00116-sch043}
{#molecules-24-00116-sch044}
![Synthesis of cholesterol-triazole conjugates bearing ferrocene-chalcone and sugar moieties. Reagents and conditions: **a**) [l]{.smallcaps}-AA, CuSO~4~·5H~2~O, THF/H~2~O (6:1), 60--80 °C, 4 h.](molecules-24-00116-sch045){#molecules-24-00116-sch045}
{#molecules-24-00116-sch046}
{#molecules-24-00116-sch047}
{#molecules-24-00116-sch048}
{#molecules-24-00116-sch049}
![Synthesis of calix\[4\]arene-cholesterol derivatives. Reagents and conditions: **a**) calix\[4\]arene, K~2~CO~3~, KI, MeCN, reflux, 24 h; **b**) 208, K~2~CO~3~, KI, MeCN, reflux, 36 h.](molecules-24-00116-sch050){#molecules-24-00116-sch050}
![Synthesis of calix\[4\]arene-cholesterol derivatives with Schiff-base bridges. Reagents and conditions: **a**) AcOH, MeOH/CHCl~3~ (1:1), reflux, 8 h.](molecules-24-00116-sch051){#molecules-24-00116-sch051}
{#molecules-24-00116-sch052}
{#molecules-24-00116-sch053}
{#molecules-24-00116-sch054}
{#molecules-24-00116-sch055}
{#molecules-24-00116-sch056}
{#molecules-24-00116-sch057}
{#molecules-24-00116-sch058}
{#molecules-24-00116-sch059}
{#molecules-24-00116-sch060}
{#molecules-24-00116-sch061}
{#molecules-24-00116-sch062}
{#molecules-24-00116-sch063}
{#molecules-24-00116-sch064}
{#molecules-24-00116-sch065}
{#molecules-24-00116-sch066}
{#molecules-24-00116-sch067}
{#molecules-24-00116-sch068}
{#molecules-24-00116-sch069}
{#molecules-24-00116-sch070}
{#molecules-24-00116-sch071}
{#molecules-24-00116-sch072}
{#molecules-24-00116-sch073}
![Synthesis of pillar\[6\]arene-functionalized cholesterol. Reagents and conditions: **a**) 11-bromoundecanoyl chloride, DIPEA, CHCl~3~, rt, 12 h; **b**) **Mono-OH-P\[[@B6-molecules-24-00116]\]**, Cs~2~CO~3~, DMF, rt, 24 h.](molecules-24-00116-sch074){#molecules-24-00116-sch074}
{#molecules-24-00116-f005}
{#molecules-24-00116-sch075}
{#molecules-24-00116-sch076}
{#molecules-24-00116-sch077}
{#molecules-24-00116-sch078}
{#molecules-24-00116-sch079}
![Synthesis of radioactive polymers **Ch-PEG~27~-CH~2~-triazole-TEG-^18^F**, **Ch-PEG~30~-*hb*PG~24~-CH~2~-triazole-TEG-^18^F**, and 3-\[^18^F\]Fluoro-cholest-5-ene. Reagents and conditions: **a**) CuSO~4~, AscONa, PBS, DMSO, 70 °C, 15 min; **b**) CuSO~4~, AscONa, PBS, EtOH, 70 °C, 15 min; **c**) methanesulfonyl chloride, Et~3~N, CH~2~Cl~2~, rt, 16 h; **d**) tetrabutylammonium hydroxide (TBAH), \[^18^F\]Fluoride, MeCN, 120 °C, 20 min.](molecules-24-00116-sch080){#molecules-24-00116-sch080}
{#molecules-24-00116-sch081}
{#molecules-24-00116-sch082}
{#molecules-24-00116-sch083}
![Synthesis of bombesin (BBN)-cholesterol conjugates. Reagents and conditions: **a**) *N*-2-2-\[2-(*N*-azidoethoxy)ethoxy\]acetyl-bombesin, CuSO~4~·5H~2~O, AscONa, DMF, rt, 8 h; **b**) *N*-2-\[2-(2-(*N*-4-(((bicyclo\[6.1.0\]non-4-yn-9-ylmethoxy)carbonyl)amino)ethoxy)ethoxy\]acetyl-bombesin, DMF, rt, 15 h.](molecules-24-00116-sch084){#molecules-24-00116-sch084}
{#molecules-24-00116-sch085}
![Electrochemical oxidation of cholesteryl diphenylphosphate in the presence of 1,2:3,4-di-*O*-isopropylidene-α-[d]{.smallcaps}-galactopyranose.](molecules-24-00116-sch086){#molecules-24-00116-sch086}
{#molecules-24-00116-sch087}
![Synthesis of cholesteryl α-[d]{.smallcaps}-glucopyranoside and its enzymatic regioselective acylation. Reagents and conditions: **a**) iodotrimethylsilane (TMSI), OH-Chol, tetrabutylammonium iodide (TBAI), DIPEA, 4 Å molecular sieves, CH~2~Cl~2~, rt, 48 h; **b**) Dowex-50WX8-200, MeOH, 2 h, rt; **c**) Novozym 435, acetone, 40 °C, 24 h; **d**) Novozym 435, THF/pyridine (4:1), 40 °C, 96 h.](molecules-24-00116-sch088){#molecules-24-00116-sch088}
{#molecules-24-00116-sch089}
{#molecules-24-00116-sch090}
{#molecules-24-00116-sch091}
{#molecules-24-00116-sch092}
{#molecules-24-00116-sch093}
{#molecules-24-00116-sch094}
{#molecules-24-00116-sch095}
{#molecules-24-00116-sch096}
{#molecules-24-00116-sch097}
{#molecules-24-00116-sch098}
{#molecules-24-00116-sch099}
{#molecules-24-00116-sch100}
![Synthesis of 4′-dehydrocholest-4-eno\[3,4-*e*\]piperazin-6-one. Reagents and conditions: **a**) basic Al~2~O~3~, MW, 120 °C, 5 min.](molecules-24-00116-sch101){#molecules-24-00116-sch101}
{#molecules-24-00116-sch102}
{#molecules-24-00116-sch103}
{#molecules-24-00116-sch104}
{#molecules-24-00116-sch105}
![Synthesis of cholesterol-based rotaxane. Reagents and conditions: **a**) Cu(CH~3~CN)~4~BF~4~, tris\[(1-benzyl-1*H*-1,2,3-triazol-4-yl)methyl\]amine, CH~2~Cl~2~, rt, 48 h; **b**) MeI, rt, 72 h; **c**) NaBF~4~, CH~2~Cl~2~/acetone/H~2~O (1:1:2), rt, 18 h.](molecules-24-00116-sch106){#molecules-24-00116-sch106}
![Synthesis of PEGylated amphiphilic diblock copolymers. Reagents and conditions: **a**) diethanolamine, Na~2~CO~3~, THF/H~2~O (2:1), 0 °C to rt, 16 h; **b**) ethyl chloroformate, Et~3~N, THF, 0 °C to rt, 16 h; **c**) 1,8-diazabicyclo\[5.4.0\]undec-7-ene (DBU), CH~2~Cl~2~, rt, 2 h.](molecules-24-00116-sch107){#molecules-24-00116-sch107}
{#molecules-24-00116-sch108}
{#molecules-24-00116-sch109}
{#molecules-24-00116-sch110}
{#molecules-24-00116-sch111}
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#sec1}
============
Several recent papers have highlighted the importance of the gut microbiome and its potential role in the human body in functions such as immune response, physiology, and metabolism.^[@ref1]−[@ref6]^ Imbalance in the gut microbiota, also known as dysbiosis, has been linked to various diseases including inflammatory bowel disease, atopy, arthritis, certain cancers, and obesity.^[@ref7]−[@ref11]^ In recent years, the analysis of metabolites in fecal matter has become an important aspect of the study of functional aspects of the gut microbiome and its relationship with human health. Metabolomics has emerged as an important approach for the measurement of metabolites. Using metabolomics, one can identify and quantify many small molecules or metabolites in biological samples such as urine, plasma, or feces using analytical techniques such as nuclear magnetic resonance (NMR) and mass spectrometry coupled to chromatography.^[@ref12],[@ref13]^ Examination of these metabolites allows detailed metabolic phenotyping and examination of altered pathways under certain conditions.^[@ref14],[@ref15]^ To this end, application of metabolomics to fecal samples has enhanced our understanding of certain conditions and provided evidence for the link between diet and gut microbiota activity.^[@ref16]−[@ref18]^
However, despite the increasing application of metabolomics to fecal samples, there is a lack of studies in examining the stability of fecal samples. Metabolites and microbial DNA can be degraded in a fecal sample through oxidation, enzymatic degradation, and hydrolysis which can occur during fecal collection and storage before the sample is prepared for analysis.^[@ref19],[@ref20]^ Owing to this degradation, it is essential to have an optimized method for fecal collection and storage in order to reduce degradation of metabolites and DNA which would allow for more accurate and reproducible results in the area of gut microbiome investigation.^[@ref20]^ There are few studies available which investigate the different collection and storage methods and their impact on metabolite and bacteria levels.
Gratton et al. demonstrated that storing a fecal sample over time at different temperatures before processing for analysis has a large impact on the metabolic profiles of human feces. An increase in branched-chain amino acids (BCAAs) and aromatic amino acids was reported after a fecal sample underwent a freeze--thaw cycle before extracting fecal water.^[@ref21]^ This information has supported a previous paper by Saric et al. which has also demonstrated an increase in BCAAs after freeze--thaw cycles of the fecal sample prior to fecal water extraction.^[@ref22]^ Furthermore, a number of studies have illustrated the effects of different sample-processing procedures on the metabolite levels. An overview of the impact of steps such as homogenization, filtration, centrifugation, and solvent extraction has been previously presented.^[@ref23]−[@ref25]^
The impact of storage conditions on bacterial community levels has been examined. Roesch et al. demonstrated a 10% change in bacterial community levels in a fecal sample when stored at −80 °C at different time points post receiving the sample.^[@ref26]^ However, other studies have reported no significant difference in results because of differing storage conditions on bacterial community levels in fecal samples, and^[@ref27],[@ref28]^ others have recommended a collection protocol to minimize the impact.^[@ref29]^ With respect to the fecal metabolome, there is no such consensus, and further studies are warranted.
In summary, analyzing the gut microbiome has become an integral part of many human studies, and assessment of the fecal water metabolome can yield valuable information. However, more work is needed to examine the impact of storage and processing procedures on the metabolite levels in fecal water. The objective of this study was to examine the impact of different storage conditions prior to metabolite extraction on the fecal water metabolome. The work highlights the need for standardized procedures.
Results {#sec2}
=======
A total of nine healthy participants were included in this study including four males and five females. Two of the individuals supplied samples on two separate occasions resulting in a total of 11 sample sets. The mean age was 34 years and participants had a mean body mass index (BMI) of 24.1 ± 2.78 kg/m^2^. A summary of the baseline characteristics of the participants is presented in [Table [1](#tbl1){ref-type="other"}](#tbl1){ref-type="other"}.
###### Demographic Data of Participants Included in the Study[a](#t1fn1){ref-type="table-fn"}
characteristics male (*n* = 4) female (*n* = 5)
-------------------- ---------------- ------------------
age (years) 34 ± 9 35 ± 14
weight (kg) 77.55 ± 7.91 66.00 ± 12.24
height (m) 1.76 ± 0.08 1.68 ± 0.08
BMI (kg/m^2^) 25.1 ± 2.6 23.2 ± 2.9
waist to hip ratio 0.87 ± 0.05 0.82 ± 0.09
All values shown are mean ± SD.
Examination of the principal component analysis (PCA) revealed that the interindividual variation was the dominant source of variation on the dataset. The samples of the individuals were grouped together in the PCA scores plot ([Figure [1](#fig1){ref-type="fig"}](#fig1){ref-type="fig"}A). To explore the impact of storage, we employed a row-wise centering of the data and this resulted in separation of the samples according to the storage type ([Figure [1](#fig1){ref-type="fig"}](#fig1){ref-type="fig"} B).
{#fig1}
To examine the impact of the storage further, a univariate analysis approach was employed. A total of 14 compounds from the fecal water analysis were significantly different across the three storage conditions. Significant metabolites from repeated measures ANOVA corrected by the false discovery rate (FDR) are presented in [Table [2](#tbl2){ref-type="other"}](#tbl2){ref-type="other"}. Interestingly, the metabolites were predominantly increased following freezing at −80 °C prior to preparation of fecal water ([Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"}). The sample stored on ice for 24 h had the highest fidelity to the freshly prepared samples. Closer examination of the significantly different metabolites revealed that the BCAAs, aromatic amino acids, Krebs cycle intermediates, and monosaccharides were found at higher levels in fecal water prepared from frozen samples. However, the short chain fatty acid (SCFA), butyrate, was lower in fecal water from frozen/defrosted, and on ice samples. [Figure [3](#fig3){ref-type="fig"}](#fig3){ref-type="fig"} shows the typical spectra of fecal water analyzed from fresh, frozen, and on ice samples.
{#fig2}
{#fig3}
###### Differential Fecal Metabolites between the Different Conditions of Sample Storage[a](#t2fn1){ref-type="table-fn"}
metabolite *P*-value FDR description
--------------- ----------- ------- ----------------------------------------------------------
aspartate \<0.001 0.008 Krebs cycle: oxaloacetate transamination
butyrate \<0.001 0.017 short chain fatty acid
fructose \<0.001 0.008 monosaccharide
fumarate \<0.001 0.004 dicarboxylic acid related to the Krebs cycle
glucose \<0.001 0.008 monosaccharide
glutamate \<0.001 0.008 proteinogenic amino acid related to cellular metabolism.
isobutyrate 0.001 0.017 BCAA-derivative
isoleucine \<0.001 0.004 BCAA
leucine \<0.001 0.004 BCAA
nicotinate \<0.001 0.004 energy metabolism in the living cell and DNA repair
phenylalanine \<0.001 0.009 aromatic amino acid
threonine \<0.001 0.008 alpha amino acid
tyrosine \<0.001 0.008 aromatic amino acid
valine \<0.001 0.004 BCAA
BCAA, branched-chain amino acid. *P*-values of repeated measures ANOVA were corrected for multiple comparisons using the Benjamini--Hochberg (FDR) procedure.
Discussion {#sec3}
==========
The present findings point to differences in levels of several metabolites following different sample storage conditions prior to metabolite extraction. Overall, BCAAs and derivatives, aromatic amino acids, SCFAs, Krebs cycle intermediates, and monosaccharides were sensitive to the storage procedure. Our study highlights the urgent need to standardize processing parameters for the preparation of fecal water in order to obtain meaningful data.
This current lack of standardization is aggravated by the fact that fecal analysis entails a higher sensitivity to differences in study design than most other biofluid analyses.^[@ref25]^ Previous work has demonstrated that insufficient sample collection can introduce bias because of the heterogeneity of fecal samples. Typically, despite fecal samples being highly heterogeneous, the fecal sampling method is to take a small portion and extract the metabolites.^[@ref25]^ Gratton and colleagues observed that fecal water extracted from 15 g of feces contained different concentrations of certain amino acids and carboxylic acids compared with fecal water extracted from a lower amount of 350 mg of fecal samples.^[@ref21]^ Furthermore, proper sample storage is essential to avoid additional bias post sampling. Both microbial and enzymatic activities affect the fecal metabolome. Hence, both storage temperature of the fecal samples and time before preparation of fecal water can have an impact on the metabolome and on the microbial composition.^[@ref21],[@ref30]^ Based on the findings of the present study, sample storage prior to the extraction of the metabolites is critical for the final levels of fecal metabolites. With the exception of butyrate, in which the highest levels were found in fresh fecal samples, metabolite levels increased in samples prepared from frozen fecal matter.
Similar results were observed in studies by Gratton et al. and Saric et al.^[@ref21],[@ref22]^ In these previous studies, the relative concentrations of BCAAs and aromatic amino acids were elevated in the fecal samples following a freezing cycle.^[@ref21],[@ref22]^ In our study, the BCAAs and aromatic amino acids phenylalanine and tyrosine were increased in samples prepared from frozen fecal matter. The impact of freezing was not limited to BCAAs and aromatic amino acids, and indeed significant changes were observed for other metabolites such as fumarate, isobutyrate, nicotinate, and aspartate. Butyrate levels were decreased in samples prepared from fecal samples kept on ice and frozen samples. However, there is no simple interpretation for the unique decrease of this SCFA. The same pattern consisting of an increase of BCAAs and a decrease of SCFAs was observed in lyophilized feces compared with wet feces.^[@ref22]^ In this study, the authors suggested that the decreased relative concentrations of SCFAs after the lyophilization process were connected to the instability and variability of these compounds over time.^[@ref22]^ This would support the decrease observed in the present study.
Our findings support the call for care in interpretation of metabolites following freeze--thaw cycles of faeces and support the concept that these alterations are not because of improved recovery of fecal metabolites.^[@ref25]^ Previous work suggested that these alterations in the fecal metabolome may be related to the lysis of cells causing the release of microbial intracellular contents.^[@ref21],[@ref25]^ When a fecal sample is frozen, ice crystals form within the cells and these crystals damage the cell walls. Once the sample is thawed, the damaged cell walls allow the release of intracellular contents.^[@ref31],[@ref32]^ This hypothesis is supported by the fact that several metabolites such as glutamate, BCAAs, Krebs cycle intermediates, and glucose increased only following freezing of the fecal sample. Taken together, it is evident that the procedure used for sample preparation can have an impact on the metabolic profile and care is needed when interpreting results from the fecal water metabolome. Interestingly, our results highlight that avoidance of freezing fecal samples by storage on ice for 24 h increased the fidelity to the freshly processed samples. For many large studies, preparation of fecal water from fresh samples will not be possible, and our results support the potential of storage on ice for 24 h until extraction of fecal water is possible.
The present study has a number of strengths and limitations worth mentioning. The same fecal sample was used for each of the three storage conditions allowing us to disentangle the effects from the larger interindividual variation. Furthermore, the study design enabled comparison to a fresh sample processed immediately following collection. A limitation of the current study is the small sample size; however, it is worth noting that the sample size is larger than previous work in the area.^[@ref21]^
Conclusions {#sec4}
===========
Analysis of the fecal water metabolome has the potential to enhance our understanding of the importance of the gut microbiome in a number of health and disease conditions.^[@ref33],[@ref34]^ However, the lack of standardization in the processing of fecal samples may lead to misinterpretation of the data. The present findings support the need for a standardized and robust sample preparation for fecal water. Importantly, freezing of solid fecal matter prior to fecal water extraction should be avoided.
Materials and Methods {#sec5}
=====================
Participants and Sample Collection {#sec5.1}
----------------------------------
Nine participants donated a total of eleven fecal samples. Participants were healthy males or females aged between 18 and 60 years of age with a BMI range of \>18.5 and \<29 kg/m^2^. Exclusion criteria included smokers, anyone taking medication or nutritional supplements, pregnant or lactating, or diagnosed with a medical disease. The study was approved by the Human Research Ethics Committee at University College Dublin (LS_16_91_Gibbons_Brennan) and written informed consent was obtained from each participant prior to participation in the study.
Each participant received a sample collection pack consisting of a plastic container, two zip-lock freezer bags, gloves, and two blue plastic bags. Subjects were asked to save the full bowel movement, record the time when the sample was produced, not contaminate the sample with urine or toilet paper, and store the sample in a cool environment prior to attending the visit center. Participants were asked to collect the sample on the day of the study visit to ensure that it was as fresh as possible. The samples were immediately placed on ice using different conditions as described below.
Sample Processing {#sec5.2}
-----------------
Each sample was weighed for total weight before being processed. Subsamples from each fecal sample were taken. From these subsamples, the following three conditions were examined. The first sample was processed immediately to obtain a fresh fecal water sample. The second sample was kept for 24 h on ice, and the third sample was placed in the −80 °C freezer for 24 h. The following day, the frozen sample was thawed and the fecal water extraction was performed.
Each subsample was processed in the same way to obtain a fecal water sample regardless of storage conditions. In short, 5 g of the sample was weighed into a 50 mL conical tube and 2× w/v of sterile phosphate-buffered saline was added. The sample was homogenized manually using a tissue grinder and centrifuged at 2654×*g* for 1 h at 4 °C. Following this, the samples were aliquoted and centrifuged a further two times at 16 000×*g* for 30 min at 4 °C before filtering (a 0.7 μm pore Minisart Filter followed by a 0.2 μm pore Whatmann filter) and freezing at −80 °C until NMR analysis.
NMR Spectroscopy {#sec5.3}
----------------
To prepare the samples for NMR spectroscopy, 60 μL of D~2~O and 10 μL of sodium trimethylsilyl propionate-\[2,2,3,3-2*H*4\] (TSP) (0.05 g/mL) were added to 540 μL of fecal water. NMR spectra were acquired on a 600 MHz Varian NMR spectrometer by using a CPMG pulse sequence at 25 °C. Spectra were acquired with 16 384 data points and 128 scans. Water suppression was achieved during the relaxation delay of 3 s. All ^1^H NMR spectra were referenced to TSP at 0.0 parts per million (ppm) and processed manually with Chenomx NMR Suite (version 7.5) by using a line broadening of 0.2 Hz, followed by phase correction and baseline correction. Metabolites were identified by Chenomx NMR Suite. Spectra were then bucketed in domains of 0.005 ppm excluding both TSP and water regions. Data were normalized to the sum of the spectral integral.
Metabolite identification was performed using Chenomx NMR Suite 8.3 Profiler (Chenomx Inc., Edmonton, Canada). Further confirmations to the assignment of proton peaks were provided by comparing the chemical shifts with those available in the Human Metabolome Database (<http://www.hmdb.ca>).
Statistical Analyses {#sec6}
====================
The dataset containing 1780 variables was log transformed prior statistical analysis. PCA on samples was carried out on all samples to obtain an overview of the data. Because we analyzed the same sample per individual using different storage conditions, the interindividual variability may obscure the actual issue of interest in a multivariate approach.^[@ref35]^ Therefore, for visualization purposes, the standardized projection of the individual effects was used by a row-wise mean centering^[@ref36]^ of samples, subtracting the within-subject variation for each individual.^[@ref37]^ Then, a new "multilevel" PCA score plot was performed exhibiting the variation because of the sample class.^[@ref36]^ To assess differences between the three conditions of sample storage, repeated measures ANOVA followed by Bonferroni post hoc comparisons were carried out on the dataset. Benjamini--Hochberg FDR was used to correct analysis for multiple comparisons.^[@ref38]^ Statistical significance was considered at a *P*-value of \<0.05. All statistical analyses were performed in the open source statistical programming environment R v3.3.1.
The authors declare no competing financial interest.
The authors acknowledge funding from the European Research Council (647783).
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#sec1_1}
============
Diarrhoeal disease causes an estimated 1.8 million deaths per year ([@B1]). Despite evidence of reduction in mortality over the last 50 years ([@B2],[@B3]), diarrhoeal disease continues to be a major killer of children aged less than five years and a principal cause of morbidity for most impoverished children of the world. It is well-known that most of these deaths are preventable with existing disease- control strategies ([@B4]).
Although there are numerous causes for the lack of greater progress in the control of diarrhoeal diseases, it is clear that our investments in related research over the last 20 years have not had the greatest attainable impact. It is now increasingly recognized that research priorities do not optimally address the needs of children in developing countries ([@B5],[@B6]). Setting research priorities is clearly a challenging and imperfect process that relies on the best data available and the knowledge of experts in the field to fill in knowledge gaps. Clearly, data regarding the number and cause of deaths and the coverage of interventions are limited and imperfect. How to integrate available data with expert opinion is an evolving area. Classically, they were derived from expert group meetings and, more recently, Delphi exercises have been employed as an improved strategy to incorporate expert opinion in decision-making processes. Cost-effectiveness analyses have been used for prioritizing among health interventions but have not been systematically used for determining the most promising research. In any case, cost-effectiveness analysis would have limited usefulness without a critical assessment of the likelihood that investments in research would result in reduction in the burden of disease. The Child Health and Nutrition Research Initiative (CHNRI) was founded to encourage and support research on the important child health problems in low- and middle-income countries. The CHNRI developed a structured process that was designed to measure the likelihood that funding-specific research questions would be successful in reducing child morbidity and mortality. This novel methodology was used here for assessing the priority for funding particular avenues of research to address the burden of disease caused by childhood diarrhoea ([@B7],[@B8]).
MATERIALS AND METHODS {#sec1_2}
=====================
Selection of group members and research options {#sec1_2_1}
-----------------------------------------------
As part of a larger exercise of assessing research priorities for major child health conditions, the CHNRI decided that the exercise should consider research for burden of reduction of diarrhoeal disease by 2015 among children aged less than five years. Since the currently-accepted disease-burden measure is the disability-adjusted life-year (DALY), which incorporated both mortality and morbidity components of disease, our scoring exercise included both mortality and morbidity components. However, due to the current thinking on disease weighting, most burden of diarrhoeal disease is related to mortality rather than morbidity, and scorers were asked to respect this current thinking. The CHNRI Secretariat selected two group members (CL and MK). These two members defined the list of research options in communication with IR, based on a systematic framework for listing research options relating to a single disease developed by the CHNRI ([@B7],[@B8]). This systematic approach enables comprehensive listing and equal treatment of research options in different broad research domains: epidemiologic research, health policy and systems research (HSPR), research intended to improve existing interventions, and research to develop new interventions. The list of research questions was intentionally limited to less than 50 to allow individuals to be able to complete the scoring process in a single day. These research options included different strategies in diarrhoeal disease control encompassing those aimed at improving water and sanitation infrastructure, those targeting healthcare-delivery strategies, those addressing nutritional deficiencies, and research to evaluate novel diagnostics and vaccines. The research options were then categorized as either: (a) Health policy and systems research (HSPR) that aimed to improve the efficiency and coverage of known interventions; (b) Research that improved existing interventions by making them more affordable or deliverable; or (c) Research options to develop entirely new interventions. Although a research option could encompass multiple different research questions, it was made sufficiently narrow in scope to be able to anticipate specific research project derivatives that could be evaluated by the scoring process.
A further 17 experts were invited to participate, of whom eight completed and returned priority scores for a total group of scorers comprising 10 individuals \[Group of scorers: Shinjini Bhatnager, Zulfiqar A. Bhutta, Olivier Fontaine, Margaret Kosek, Claudio F. Lanata, Dilip Mahalanabis, Mohammed Abdus Salam, John D. Snyder, Cesar Victora, and Damian G. Walker\].
These individuals were categorized as physicians with expertise in infectious diseases, gastroenterologists, public-health researchers specialized in programmatic issues, a health economist, and public-health researchers in areas other than programme development and evaluation. Each one scored the individual research options independently using a five-component structured model developed by the CHNRI to evaluate health research. The five components consist of the following: (a) likelihood that the research option can yield new knowledge in an ethical manner; (b) likelihood that the research findings will lead to efficacious and effective interventions; (c) likelihood that the intervention derived from the research would be affordable and deliverable to the population of interest; (d) most likely maximum burden of disease reduction that could be derived from interventions resulting from research within the option; and (e) likely impact that the derivates of the research will have on equity.
Scores were computed as percentage of maximal obtainable points for each of the major five components being evaluated and then combined for an overall score. The scores outline both limitations and strengths of each research option. Each of the five intermediate scores reflect the likelihood that the research option will be answerable, that it will result in an effective intervention, that the resulting intervention will be deliverable, that the resulting intervention will increase equity, and an estimation of the maximal disease-burden impact an intervention resulting from the research is foreseen to have. When added together, this overall score becomes a quantitative measure of the collective optimism that research in that area can have substantial impact prior to 2015. Although this system easily accommodates weighting of these five options by donors or regional agencies or other stakeholders, we have presented the unweighted results. The process of scoring is presented in greater detail elsewhere ([@B7]-[@B10]).
RESULTS {#sec1_3}
=======
The listing of research options yielded 46 options. Twenty-one research options were designed as health policy and systems research to increase the efficiency of interventions already in place, 10 options addressed research to improve the affordability and deliverability of known interventions, and 15 options were primary research to develop new interventions. The complete list of the 46 research options is presented in Table [1](#T1){ref-type="table"}. The questions guiding the scoring of the research options by each criteria are shown in Table [2](#T2){ref-type="table"}. An excel file that facilitates the scoring exercise by providing a spreadsheet for the input of scores is available online (<http://www.icddrb.org/jhpn>).
######
List of 46 research options scored by diarrhoeal disease experts
Research option
----------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
RO1: Health policy and systems research (HPSR) to increase access to ORS packets at all times in all sites for all children who may need it
RO2: Research to generate new knowledge (mostly effectiveness studies) to increase the use of low-osmolarity ORS
RO3: Health policy, systems, and education/behaviour modification research to increase the percentage of infants with exclusive breastfeeding at \<6 month of age
RO4: Health policy, systems, and education/behaviour modification research to increase the percentage of infants and children, aged less than 2 years, who are breastfed
RO5: Health system research to increase the coverage of measles vaccine
RO6: HPSR to improve the coverage of rotavirus vaccine in countries with the greatest needs
RO7: Systems and education/behaviour modification research to increase water consumed per person per day
RO8: System research to measure the effectiveness of piped water systems on diarrhoea if they are installed at the community vs in the home
RO9: System research to measure the effectiveness of piped water systems on diarrhoea if they are installed so as to provide intermittent vs 24-hour availability
RO10: Systems and education/behaviour modification research to increase the coverage of sewage systems
RO11: Systems and education/behaviour modification research to increase the prevalence of effective latrines
RO12: Health policy, systems, and education/behaviour modification research to increase the proportion of women and children washing their hands effectively to improve hand-washing promotion
RO13: Education/behaviour modification research to increase the energy density of weaning foods at the household level (in areas with food availability)
RO14: HPSR to allow that all mothers with a child with diarrhoea will know how to recognize danger-signs for timely referral/self-referral of severe cases
RO15: HPSR to improve the quality of care of moderate/severe diarrhoea cases through standardized case management
RO16: HPSR to improve prescription of appropriate antibiotics for dysentery
RO17: Efficacy/effectiveness studies of interventions of behaviour modification to reduce baby bottle-use
RO18: Education/behaviour modification research to increase the use of refrigerators for storage of weaning foods
RO19: Efficacy/effectiveness studies and education/behaviour modification research to increase consumption of *Lactobacillus* GG probiotic
RO20: Efficacy/effectiveness studies of interventions of behaviour modification to increase potties-use/improved faece-disposal practices
RO21: HPSR to generate new knowledge to increase the coverage of vitamin A supplementation
RO22: System and community research to reduce costs/improve deliverability and increase the coverage of piped water systems
RO23: Effectiveness, costs, sustainability, system and behavioural modification/cultural research to increase the use of point-of-use water disinfection: implementation of point-of-use treatment and water-storage practices
RO24: Research to develop new ways of sewage-treatment systems that will make them affordable to developing countries
RO25: Research to improve the deliverability, measure effectiveness, and determine the sustainability of fly-control interventions
RO26: Effectiveness studies and studies that will reduce the cost/improve the deliverability of cholera vaccines in high-burden countries
RO27: Cash-transfer programmes to improve diet quality and nutrition in poor areas
RO28: Policy, systems, and education/behaviour modification research to improve current strategies aiming at improving the quality of diet of family in areas with low access to good diets
RO29: Effectiveness, HPSR, and educational/behaviour modification studies to improve the deliverability/cost of zinc treatment in diarrhoea-control programmes in several regions of the world with different epidemiological profiles
RO30: Efficacy, effectiveness and cost studies that will increase the use of zinc food-fortification programmes in developing countries
RO31: Cost-effectiveness studies of rotavirus vaccine in different epidemiologic contexts
RO32: Develop norovirus vaccines
RO33: Develop *Shigella* vaccines
RO34: Develop ETEC vaccines
RO35: Develop *Campylobacter* vaccines
RO36: Develop EPEC vaccines
RO37: Develop *Helicobacter pylori* vaccines
RO38: Develop vaccines for *Entamoeba histolytica*
RO39: Develop new measles vaccines that will be heat-stable and able to immunize newborns
RO40: Solar ovens to keep weaning foods above \>50 °C for a day
RO41: Low cost, no electrical/no fuel consuming refrigerators to storage food at the household level
RO42: New antibiotics for drug-resistant *Shigella*
RO43: New antibiotics for drug-resistant cholera
RO44: Develop interventions that will reduce bacterial contamination of crops irrigated with contaminated water in developing countries
RO45: Further development of antisecretory agents in the management of paediatric diarrhoea
RO46: Develop the technology to deliver zinc to children using prolong dosing intervals
EPEC=Enteropathogenic *Escherichia coli* ETEC=Enterotoxigenic *Escherichia coli* ORS=Oral rehydration solution
RO=Research option
######
Questions with which to assess the 5 criteria for each selected research option. For an interactive spreadsheet for scoring, see website (<http://www.icddrb.org/jhpn>)
Scoring criteria Question 1 Question 2 Question 3
---------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Criteria 1: Likelihood that the research will lead to new knowledge in an ethical way 1.1Would you say the research question is well-framed and endpoints are well-defined? 1.2Based on: (a) the level of existing research capacity in proposed research and (b) the size of the gap from current level of knowledge to proposed endpoints, would you say that a study can be designed to answer the research question and to reach the proposed endpoints of the research? 1.3Do you think that a study needed to answer the proposed research question would obtain ethical approval without major concerns?
Criteria 2: Assessment of the likelihood that the intervention which would be developed through the research would be efficacious 2.1Based on the best existing evidence and knowledge, would the intervention which would be developed/improved through proposed research be efficacious? 2.2Based on the best existing evidence and knowledge, would the intervention which would be developed/improved through the proposed research be effective? 2.3Would you say that the evidence upon which these opinions are based (answers to prior 2 questions) is of high quality?
Criteria 3: Likelihood that the intervention based on the research would be affordable, deliverable, and sustainable in the population of interest 3.1Taking into account the level of difficulty with intervention delivery from the perspective of the intervention itself (e.g. design, standardizability, safety), the infrastructure required (e.g. human resources, health facilities, communication and transport infrastructure) and users of the intervention (e.g. need for change of attitudes or beliefs, supervision, existing demand), would you say the endpoints of the research would be deliverable within the context of interest? 3.2Taking into account the resources available to implement the intervention, would you say that the endpoints of the research would be affordable within the context of interest? 3.3Taking into account government capacity and partnership requirements (e.g. adequacy of government regulation, monitoring, and enforcement; governmental intersectoral coordination, partnership with civil society and external donor agencies; favourable political climate to achieve high coverage), would you say that the endpoints of the research would be sustainable within the context of interest?
Criteria 4: Assesment of the maximal potential for disease-burden reduction 4.1Taking into account the results of conducted intervention trials, or for the new interventions the proportion of avertable burden under an ideal scenario, would you say that the sucessful reaching of research endpoints would have the capacity to remove 5% of the burden or more? 4.2To remove 10% or more? 4.3To remove 15% or more?
Criteria 5: Assessment of the impact of proposed research on equity 5.1Would you say that the present distribution of the disease burden affects mainly the underpriveleged in the population? 5.2Would you say that either (a) mainly the underpriveleged, or (b) all segments of the society equally would be the most likely to benefit from the results of the proposed research after its implementation? 5.3Would you say that the proposed research has the overall potential to improve equity in disease-burden distribution in the long term (e.g. 10 years)?
Criterion 1: Generating new knowledge {#sec1_3_1}
-------------------------------------
Priority scores for research options evaluated solely on the criterion of their ability to generate new knowledge ranged from 57.0% to 95.8%. The top five research options that were predicted to encounter minimal obstacles in their realization are listed in Table [3](#T3){ref-type="table"}. The research option that received the highest score for this criterion was the conduction of cost-effectiveness studies of rotavirus vaccines in different epidemiologic contexts. The second ranked research option by this criterion was effectiveness studies to evaluate the expanded use of low-osmolarity oral rehydration solution (ORS). Two research alternatives received equal scores to rank third. These options included health policy and systems research to improve the deliverability and cost of zinc treatment in diarrhoeal disease programmes and health policy and systems research to improve coverage of rotavirus vaccine in countries with the greatest burden of disease. Health policy and systems research to improve case management of moderate and severe cases of diarrhoeal illness by using standardized case management and the development of new antibiotics for the treatment of drug-resistant shigellosis tied for fourth place in the ranking. The fifth rank was occupied by two evenly-ranked options of health policy and services research to increase access to ORS envelops at all times to all children who need it and health policy and services research to increase the percentage of infants exclusively breastfed up to the age of six months.
######
Top 5 options for generating new knowledge (Criterion 1)
Rank Category Research option Score (%)
------ ---------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- -----------
1 3 RO31: Cost-effectiveness studies of rotavirus vaccine in different epidemiologic contexts 95.8
2 2 RO2: Research to generate new knowledge (mostly effectiveness studies) to increase the use of low-osmolarity ORS 88.3
3 3 RO29: Effectiveness, health policy, and systems research (HPSR) and educational/behaviour modification studies to improve the deliverability/cost of zinc treatment in diarrhoea-control programmes in several regions of the world with different epidemiological profiles 86.7
3 2 RO6: HPSR to improve coverage of rotavirus vaccine in countries with the greatest needs (impact on morbidity and mortality) 86.7
4 2 RO15: HPSR to improve the quality of care for moderate and severe diarrhoea cases through standardized case management (impact on mortality only) 85.2
4 4 RO42: New antibiotics for drug-resistant enteropathogens: *Shigella* 85.2
5 2 RO1: HPSR to increase access to ORS packets at all times in all sites for all children who may need it 85.0
5 2 RO3: Health policy, systems, and education/behaviour modification research to increase percentage of infants with exclusive breastfeeding \<6 months of age 85.0
ORS=Oral rehydration solution
RO=Research option
Low scoring research options in this category were predominated by enteric vaccines that are currently in early stages of development as these are unlikely to yield clinical trials demonstrating efficacy prior to 2015. Scores for *Campylobacter*, enteropathogenic *Escherichia coli, Entamoeba histolytica,* and norovirus all obtained scores for answerability ranging between 57% and 68%.
Criterion 2: Efficacy and effectiveness {#sec1_3_2}
---------------------------------------
The priority scoring of research options based solely on their predicted potential to lead to (or improve) efficacious and effective interventions yielded scores ranging from 32.0% to 97.9% (Table [4](#T4){ref-type="table"}). The top research option when judged by this criterion was the study of cost-effectiveness of rotavirus vaccine in different epidemiologic contexts. The second-ranking research option was health policy and services research to increase the access to ORS for all children who may need it. Effectiveness, health policy and services research, and educational/behavioural modification studies to improve the deliverability and cost of zinc treatment in diarrhoea-control programmes in several regions of the world with different epidemiologic profiles occupied the third rank. Tied for the fourth ranking was research to generate new knowledge (mostly effectiveness studies) to increase use of low-osmolarity ORS and health systems research to increase the coverage of measles vaccine. The fifth-ranking research option in terms of efficacy and effectiveness was health policy and systems research to improve the quality of care of moderate/severe diarrhoea cases through standardized case management.
######
Top 5 options for efficacy/efficaciousness (Criterion 2)
Rank Category Research options Score (%)
------ ---------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- -----------
1 3 RO31: Cost-effectiveness studies of rotavirus vaccine in different epidemiologic contexts 97.9
2 2 RO1: HPSR to increase access to ORS packets at all times in all sites for all children who may need it 93.3
3 3 RO29: Effectiveness, HPSR, and educational/behaviour modification studies to improve the deliverability/cost of zinc treatment in diarrhoea-control programmes in several regions of the world with different epidemiological profiles 90.7
4 2 RO2: Research to generate new knowledge (mostly effectiveness studies) to increase use of low-osmolarity ORS 86.7
4 2 RO5: Health-systems research to increase the coverage of measles vaccine 86.7
5 2 RO15: HPSR to improve the quality of care for moderate and severe diarrhoea cases through standardized case management 85.4
HPSR=Health policy and systems research; ORS=Oral rehydration solution; RO=Research option
Research options that received low scores in this area were diverse and included research relating to fly control, the improved storage of weaning foods, and the development of interventions meant to curb bacterial contamination of crops.
Criterion 3: Sustainability and deliverability {#sec1_3_3}
----------------------------------------------
Priority scores for options evaluated based on sustainability and deliverability alone had the widest variation in obtained scores of the five criteria with scores ranging from 9% to 90%. The highest score in terms of sustainability and deliverability was given to effectiveness studies to increase the use of low-osmolarity ORS which obtained a priority score of 90% (Table [5](#T5){ref-type="table"}). The second ranking was to studies relating to the uptake and evaluation of zinc in diarrhea-control programmes in the different epidemiologic contexts. The third ranking was shared by research options for health policy and systems research to improve the prescription of appropriate antibiotics for dysentery and to increase the coverage of vitamin A supplementation. The evaluation of the cost-effectiveness of rotavirus vaccine in the different epidemiologic contexts and health policy and systems research to improve the quality of care of moderate and severe diarrhoea cases through standardized case management were ranked fourth. The fifth-ranking option was health policy, systems and education and behavioural modification research to increase the proportion of women and children washing their hands effectively.
######
Top 5 options for sustainability and deliverability (Criterion 3)
Rank Category Research option Score (%)
------ ---------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- -----------
1 2 RO2: Research to generate new knowledge (mostly effectiveness studies) to increase the use of low-osmolarity ORS 90.0
2 3 RO29: Effectiveness, HPSR, and educational/behaviour modification studies to improve the deliverability/cost of zinc treatment in diarrhoea-control programmes in several regions of the world with different epidemiological profiles 88.9
3 2 RO16: HPSR to improve prescription of appropriate antibiotics for dysentery 85.2
3 2 RO21: HPSR to generate new knowledge to increase the coverage of vitamin A supplementation (to reduce severity of diarrhoea and improve mortality) 85.2
4 3 RO31: Cost-effectiveness of rotavirus vaccine in different epidemiologic contexts 83.3
4 2 RO15: HPSR to improve the quality of care for moderate and severe diarrhoea cases through standardized case management 83.3
5 2 RO12: Health policy, systems, and education/behaviour modification research to increase the proportion of women and children washing their hands effectively 77.8
HPSR=Health policy and systems research; ORS=Oral rehydration solution; RO=Research option
Research options scoring poorly in deliverability and sustainability included options to increase the use of refrigerators for the storage of weaning foods, cash-transfer programmes to improve the quality of diet and nutrition in poor areas, and vaccines for *E. histolytica* and *Camplylobacter*.
Criterion 4: Maximal potential for diseaseburden reduction {#sec1_3_4}
----------------------------------------------------------
Priority scores to judge the maximum potential for disease-burden reduction ranged from 8% to 79% (Table [6](#T6){ref-type="table"}). The top two scoring research options related to use of rotavirus vaccine. The top scoring research option achieved a priority score of 79.2% and was aimed at evaluating the cost-effectiveness of rotavirus vaccination in the different epidemiologic contexts while health policy and systems research to improve the coverage of rotavirus vaccination in countries with the greatest need had a 76.7% priority score. Health policy and services research to increase access to ORS envelops received the third-ranking position and was followed in the ranking by health policy and services research to improve the deliverability and cost of zinc treatment in diarrhoea-control programmes in several regions of the world with different epidemiologic profiles. The fifth-ranking option was health policy and systems research to train mothers in the recognition of danger signs for the timely self-referral of severe cases of diarrhoea.
######
Top 5 option for maximal potential to decrease disease burden (Criterion 4)
Rank Category Research option Score (%)
------ ---------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- -----------
1 3 RO31: Cost-effectiveness of rotavirus vaccine in different epidemiologic contexts 79.2
2 2 R06 : HPSR to improve the coverage of rotavirus vaccine in countries with the greatest needs 76.7
3 2 RO1: Health policy and systems research (HPSR) to increase access to ORS packets at all times in all sites for all children who may need it 75.0
4 3 RO29: Effectiveness, HPSR, and educational/behaviour modification studies to improve the deliverability/cost of zinc treatment in diarrhoea-control programmes in several regions of the world with different epidemiological profiles 70.0
5 2 RO14: HPSR to allow that all mothers with a child with diarrhoea will know how to recognize danger-signs for timely referral/self-referral of severe cases 64.8
ORS=Oral rehydration solution; RO=Research option
Low scores in maximal potential for disease-burden reduction were obtained by research options relating to fly-control interventions, effectiveness, and deliverability of cholera vaccine, and efficacy and effectiveness studies relating to use of probiotic *Lactobacillus* GG.
Criterion 5: Equity {#sec1_3_5}
-------------------
The scores for equity were the highest of any of the criteria and ranged from 43.0% to 98.3%. All five top options received priority scores greater than 90% (Table [7](#T7){ref-type="table"}). The top score was given to health systems research to increase the coverage of measles vaccine. The second-ranking research option was policy, systems, and education and behaviour modification research to improve current strategies aiming at improving the quality of diet of families in areas with low access to good diets. Systems and education and behavioural modification research to increase the prevalence of latrines was ranked third. Educational and behavioural modification research to increase the energy density of weaning foods at the household level in areas with food availability and systems and behavioural research to improve water consumed per person per day obtained equal scores to rank fourth among the selected research options. The fifth-ranking option was health policy and systems research to improve the quality of care of moderate and severe cases of diarrhoea through standardized case management.
######
Top 5 options for equity (Criterion 5)
Rank Category Research option Score (%)
------ ---------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- -----------
1 2 R05 : Health systems research to increase the coverage of measles vaccine 98.3
2 3 RO28: Policy, systems, and education/behaviour modification research to improve current strategies aiming at improving the quality of diet of family in areas with low access to good diets 96.7
3 2 RO11: Systems and education/behaviour modification research to increase the prevalence of effective latrines 93.3
4 2 RO13: Education/behaviour modification research to increase the energy density of weaning foods at the household level (in areas with food availability) 92.6
4 2 RO7: Systems and education/behaviour modification research to improve water consumed per person per day 92.6
5 2 RO15: HPSR to the improve the quality of care for moderate and severe diarrhoea cases through standardized case management 90.7
HPSR=Health policy and systems research; RO=Research option
Low priority scores were obtained by research options to evaluate probiotic *Lactobacillus* GG, new antisecretory agents, and vaccines for *E. histolytica*, norovirus, and *H. pylori*.
### Combined results {#sec1_3_6_1}
The overall priority scores that were assigned to research options by computing unweighted means of the five intermediate scores ranged from 35.0% to 85.2% (Table [8](#T8){ref-type="table"}). The top scoring research items, overall, were predominantly (7/10) options that aimed at increasing the efficiency and coverage of interventions with known effectiveness. The research option receiving the highest priority score addressed effectiveness, health policy and services research to improve the deliverability and cost of zinc treatment in diarrhoea-control programmes in areas with different epidemiologic profiles. The second-ranking research option, which proposed conducting cost-effectiveness studies of rotavirus vaccine in different epidemiologic contexts received a priority score less than 1% lower than the top-ranking option. Health policy and services research to increase access to ORS envelops and to improve the quality of care of moderate to severe cases of diarrhoea through standardized case management were ranked third and fourth respectively. Effectiveness studies to increase the use of low-osmolarity ORS received the fifth-ranking priority score and was followed closely by health policy, systems, and education and behavioural modification research to increase the percentage of infants exclusively breastfed for the first six months of life. Health systems research to increase the coverage of measles and rotavirus vaccines were ranked seventh and eighth respectively. Health policy and systems research to aid mothers in the recognition of danger-signs in cases of severe illness to improve self-referral ranked ninth, and the tenth priority ranking was assigned to efficacy, effectiveness and costing studies to evaluate the increased use of zinc-fortification programmes in developing countries.
######
Top 10 research options overall by five criteria
Rank Category Research option Score (%)
------ ---------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- -----------
1 3 RO29: Effectiveness, HPSR, and educational/behaviour modification studies to improve the deliverability/cost of zinc treatment in diarrhoea-control programmes in several regions of the world with different epidemiological profiles 85.2
2 3 RO31: Cost-effectiveness studies of rotavirus vaccine in different epidemiologic contexts 85.0
3 2 RO1: Health policy and systems research (HPSR) to increase access to ORS packets at all times in all sites for all children who may need it 81.6
4 2 RO15: HPSR to improve the quality of care for moderate/severe diarrhoea cases through standardized case management 80.0
5 2 RO2: Research to generate new knowledge (mostly effectiveness studies) to increase the use of low-osmolarity ORS 78.7
6 2 RO3: Health policy, systems, and education/behaviour modification research to increase the percentage of infants with exclusive breastfeeding \<6 months of age 77.4
7 2 RO5: Health systems research to increase the coverage of measles vaccine 77.2
8 2 RO6: HPSR to improve the coverage of rotavirus vaccine in countries with the greatest needs 75.0
9 2 RO14: HPSR to allow that all mothers with a child with diarrhoea will know how to recognize danger-signs for time referral/self-referral of severe cases 74.4
10 3 RO30: Efficacy, effectiveness and cost studies that will increase the use of zinc food-fortification programmes in developing countries 73.0
HPSR=Health policy and systems research; ORS=Oral rehydration solution; RO=Research option
### Vaccines {#sec1_3_6_2}
It is worth noting that the development of new vaccines for the improved control of the burden of diarrhoeal disease obtained low scores (Table [9](#T9){ref-type="table"}). The highest-scoring vaccine was a measles vaccine that would be both heat-stable and effective when administered in the neonatal period. Research directed towards obtaining this goal received a priority score of 65.1%. The highest-scoring vaccine directly targeting an enteric organism was a vaccine against *Shigella*, which obtained a priority score of 63.0%. Determining the cost-effectiveness of the currently-available cholera vaccine received a score of 61.7%.
######
Unweighted priority scores (%) for development of novel vaccines to diminish the burden of diarrhoeal disease
Vaccine Criterion 1(%) Criterion 2(%) Criterion 3(%) Criterion 4(%) Criterion 5(%) Overall priority score (%)
-------------------------------------------------------------------- ---------------- ---------------- ---------------- ---------------- ---------------- ----------------------------
New measles vaccines that will be heat-stable and able to immunize
newborns 76.7 77.8 59.3 50.0 61.7 65.1
*Shigella* 83.3 83.3 44.4 35.2 68.5 63.0
ETEC 66.7 63.0 37.0 42.6 68.5 55.6
EPEC 63.3 43.8 25.0 37.0 68.5 47.5
Norovirus 68.3 54.2 31.3 29.2 44.4 45.5
*Helicobacter pylori* 63.3 41.7 24.1 14.8 42.6 37.3
*Entameoba histolytica* 63.3 43.8 14.6 8.3 55.6 37.1
*Campylobacter* 56.7 41.7 12.5 13.0 51.9 35.1
EPEC=Enteropathogenic *Escherichia coli*; ETEC=Enterotoxigenic *Escherichia coli*
There is little question that the group of scorers was not biased against vaccines as a whole or enteric vaccines. Two of the overall top 10 research options reflected research regarding the appropriate usage of rotavirus vaccine in different settings. In addition to this, health policy and service research to increase the coverage of the current measles vaccine was ranked seventh overall. In general, the existing vaccines with proven efficacy and potential for expanded access were scored higher than theoretical possibilities in vaccine research. Furthermore, as the time set to expect the benefits of the research in terms of disease-burden reduction is 2015, this timetable challenges vaccine research in areas in which knowledge on basic science regarding the basis of acquired immunity or the global burden of disease is lacking. However, even with an efficacy of 95% in preventing mortality and 60% in preventing morbidity, few enteric pathogens cause an aetiologic fraction of morbidity and mortality of diarrhoeal disease large enough to compare with interventions that are independent of aetiology, and scores for criterion 4 (maximum disease-reduction fraction) were, therefore, correspondingly low. While the aetiologic fractions may vary somewhat by region, few pathogens cause more than 10% of incident episodes globally, putting a practical ceiling on the overall diarrhoeal disease-burden reduction available by their control even with high efficacy and coverage. Further stifling the scoring were issues regarding to deliverability, and to a lesser extent, equity.
### Water and sanitation {#sec1_3_6_3}
It is somewhat surprising to see the predominantly intermediate ranks received by the research options that could broadly be categorized as dealing with water and sanitation, the theoretical mainstay to control the transmission of enteric diseases. Research options covered a broad range of issues, including those that deal with quality and quantity of water (research option 7, 8, 9, 22, and 23; research options addressing the improved disposal of human excreta (research option 10, 11, 24, and 44); and a single research option addressing handwashing (research option 12). As a group, these obtained high scores in terms of equity and answerability and scores predominantly in the second quartile for maximum disease-burden reduction but lower scores in other categories, including deliverability and sustainability.
### Nutrition {#sec1_3_6_4}
Improving nutritional status has a key role in decreasing the burden of diarrhoeal disease. Two of the top 10 priority scores, overall, went to research options that included zinc, a micronutrient that has shown effects in decreasing the duration of diarrhoeal episodes and the risk of developing persistent diarrhoea. Increasing the coverage of vitamin A supplementation and increasing the energy-density of weaning foods in areas with food availability received priority scores of 71.2% and 67.5% respectively and were in the top quartile of rankings. Research to improve the quality of diet in areas with limited food availability obtained a more modest priority score of 61.0%. Cash-transfer programmes designed to improve diet in areas with restricted access were given a priority score of only 47.1% hampered in large part by the perceived problems in delivery and maintenance of the programmes.
DISCUSSION {#sec1_4}
==========
The research options that appear on the list are not comprehensive. These are rather meant to be a selection of the top research options representative of various strategies of diarrhoeal disease control. These include a balance between the development of new interventions with research that addresses improved implementation of interventions that have already been shown to be efficacious. We feel that the structured nature of the process of scoring research options leads to a significant improvement from previously-used methods involving expert consensus. The output is a list of ranking options developed by technical experts scoring independently to avoid problems associated with group dynamics in decision-making. The structure of the scoring process minimizes bias through its transparency and resulting accountability.
The prioritization of research options has important implications for the assignment of available funds that are intended for the control of diarrhoeal disease and the improvement of child health. This approach to improve the decision-making process has steered away from informal meetings of experts that were classically used because of concerns of decision-making flaws resulting from both group dynamics (i.e. groupthink) ([@B11]) and vested interest in the development of programmes in the particular field of interest of the expert. The Delphi process was meant to overcome these shortcomings through anonymity of scorers, structured feedback, and progressive feedback to distill or focus-group opinion. However, the Delphi output lacks the transparency of the CHNRI process, and the reasons behind the scorers\' decisions are unclear. While the Delphi thereby works to obtain a consensus, the CHNRI process generates a quantitative output that allows for the calculation of uncertainty in the evaluation of research options through analysis of variance of the individual expert\'s scores. Furthermore, the CHNRI process elucidates the specific contexts within which the priorities are set (preferably with the investors in health research). It offers an approach to systematic listing of large numbers of research investment options and comparison between options from different research domains using the same set of criteria, thus balancing between high-risk and low-risk options. Its systematic nature in scoring research options decreases individual bias while independent scoring by many experts removes the possibility of a few individuals dominating the decisions on priorities.
One of the shortcomings of this scoring exercise done at this level and applied to a local level is related to problems of the context. When scoring is done, as was the case here, to be broadly applied to the majority of population of the developing world, assumptions are made that may not be truly representative of certain areas. A key example of this relates to issues of deliverability, cost, and sustainability. This criterion is the one most likely to vary between areas with different levels of political stability, public infrastructure, and economic resources. For these reasons, at the regional level, it is worth having a group of experts re-score research options for this particular criterion as this may optimize the overall evaluation of different research options in local contexts. An excel file that facilitates this process is available at <http://www.icddrb.org/jhpn>.
A second shortcoming of this exercise was the limited availability of selected experts for the exercise. Although 19 experts were invited to participate, only 10 individuals contributed their priority scores. In the future, we would recommend linking the scoring activity to a meeting to facilitate higher participation rates among appropriate experts.
Despite limitations of the CHNRI process, it has several advantages over existing priority-setting techniques. First, the transparency of the process in and of itself can be a roadmap to researchers and potential funding agencies in the area of interest. For example, if experts give a low scores for certain research option relating to issues surrounding deliverability and equity, researchers in the area will be stimulated to develop novel delivery strategies to deliver the intervention to those with the greatest need, and granting agencies can direct programme announcements to fund these proposals. Like the Delphi exercise, it controls the influence of group dynamics, although it does so to a greater extent because scorers are not 'refocused' or redirected by serial scoring exercises.
It is clear that the domination of health policy and systems research among those obtaining the higher ranks is a consequence of results (effects on the burden of disease) being expected by 2015. A longer timeframe (e.g. 50 years) might allow more long-term strategic events to obtain higher scores. Furthermore, although the framework for scoring is transparent and systematic and it follows that rational answers to narrowly-formatted questions minimize personal biases, there is a possibility that a different group, composed of more policy-makers and programme officers rather than the group that preformed this exercise, may yield somewhat different results. Despite these limitations, we believe that this priority-setting exercise is a useful guide to investors in health research targeting diarrhoeal disease, who hope to observe measurable results prior to the end-date of the millennium development goals.
ACKNOWLEDGEMENTS {#sec1_5}
================
This study was supported by the Child Health and Nutrition Research Initiative (CHNRI) of the Global Forum for Health Research, under support from the World Bank.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Botulinum toxins have been used for over 20 years for aesthetic procedures to improve the appearance of the face.[@b1-ccid-5-053],[@b2-ccid-5-053] Botulinum toxin type A injections are the most common nonsurgical procedures performed in the US with almost 2.5 million procedures carried out in 2010.[@b3-ccid-5-053]
There are several commercially available botulinum toxin type A products. OnabotulinumtoxinA (Botox^®^/Vistabel^®^; Allergan Inc, Irvine, CA) is indicated for the treatment of glabellar frown lines and is commonly used for the treatment of facial wrinkles.[@b4-ccid-5-053] The terms "incobotulinumtoxinA", "NT 201", "Xeomin^®^", and "Bocouture^®^" (Merz Pharmaceuticals GmbH, Frankfurt, Germany) all refer to the same botulinum toxin type A (150 kDa) that, unlike onabotulinumtoxinA, is free from complexing proteins. IncobotulinumtoxinA is currently licensed widely across Europe, the US, and parts of South America and Asia for aesthetic indications. IncobotulinumtoxinA has demonstrated clinical efficacy in aesthetic indications in a number of clinical trials.[@b5-ccid-5-053]--[@b7-ccid-5-053]
In comparative, head-to-head trials in healthy volunteers, and in the therapeutic indications of blepharospasm and cervical dystonia, incobotulinumtoxinA had an identical time course of action (eg, time to onset, duration of effect, and time to waning of effect) as onabotulinumtoxinA.[@b8-ccid-5-053] Furthermore, in a large head-to-head study comparing incobotulinumtoxinA with onabotulinumtoxinA for the treatment of glabellar frown lines, the percentage of responders 4 and 12 weeks after injection of the same number of units (U) of either preparation were similar, and demonstrated that both treatments were highly effective according to the assessment of independent raters, investigators, and subjects.[@b7-ccid-5-053] No statistically significant difference in efficacy was observed in a proof-of-concept study in the treatment of crow's feet[@b6-ccid-5-053] and in the treatment of forehead lines[@b9-ccid-5-053] using a 1:1 dose ratio of the two products.
The aim of this retrospective analysis was to investigate the clinical efficacy of incobotulinumtoxinA compared with onabotulinumtoxinA or abobotulinumtoxinA (Dysport^®^/Azzalure^®^; Ipsen Ltd, Berkshire, UK) when used in daily practice by physicians to treat wrinkles of the upper face. IncobotulinumtoxinA was launched in Germany in 2005, making it the best suited location for this study by ensuring sufficient data could be obtained for this analysis including in off-label indications.
Parameters relating to subject and physician satisfaction, the time interval between doses administered (duration between treatment cycles), dosages used, and adverse effects experienced were investigated. Any differences in these parameters may indicate that the product with lower subject and physician satisfaction, shorter interval between treatments, or higher dosages was less clinically effective.
Materials and methods
=====================
The parameters used as indicators of clinical efficacy in daily practice were subject and physician satisfaction, the time interval between injections, dosage, and adverse effects. In order to collect the relevant data to address these questions, physicians known to use incobotulinumtoxinA at 41 sites in Germany were contacted and asked to complete questionnaires ([Table 1](#t1-ccid-5-053){ref-type="table"}) based on an inspection of files for those subjects who had received at least two consecutive botulinum toxin type A injections, but not more than three, in the upper face within 12 months during the last 2 years (ie, to treat glabellar frown lines, lateral periorbital wrinkles, and/or horizontal forehead lines, which includes common on- and off-label usage in clinical practice). A different questionnaire was filled in for each indication, giving a maximum of three completed questionnaires per subject. The selected subjects had therefore been treated according to the treatment flows shown in [Figure 1](#f1-ccid-5-053){ref-type="fig"}. For the analysis, subjects were divided into two groups: subjects who did not change product and subjects who changed product (irrespective of whether the change occurred at visit 2 or visit 3 or both).
The documentation period was from March 2011 to June 2011 and data that would allow a subject to be identified were not collected. Male or female subjects aged 18 years and over who had received treatment with botulinum toxin type A were eligible for inclusion in this study. Initial treatments and touch-up treatments were reasons to exclude a subject from the study. No ethics committee approval was required for this retrospective study without invasive measures.
In daily practice, it is very common to treat "aesthetic units", for example the "upper face", rather than single isolated areas, such as forehead lines, the glabella or crow's feet. Consequently, analysis was performed on all treatments of the upper face rather than single isolated areas. Differences between treatment groups were assessed using appropriate statistical analyses. Any differences in subject and physician satisfaction and adverse effects were analyzed using Fisher's Exact test, while analysis of doses at each visit and treatment intervals used the Wilcoxon--Mann--Whitney test. The Student's *t*-test was used to analyze the total mean dose across all visits. Data were collected from a sufficient number of subjects (n = 1256) to enable statistical analyses to be carried out.
Results
=======
Of the 41 sites contacted, 21 sites returned completed questionnaires. In total, 2316 questionnaires were returned and 46 were excluded. Forty-five questionnaires were excluded because they only recorded one visit (rather than the required minimum of two) and one was excluded because it was a duplicate, leaving 2270 evaluable questionnaires. In total, data from 1256 subjects were included. Demographic and baseline characteristics are shown in [Table 2](#t2-ccid-5-053){ref-type="table"}. Subject numbers in the incobotulinumtoxinA and onabotulinumtoxinA treatment groups were sufficient to ensure that robust statistical data could be obtained, but the number of abobotulinumtoxinA injections was so low (1.6%, which might reflect daily practice in Germany), that no statistical evidence could be conveyed. Hence, these were removed from the analysis. Most subjects received incobotulinumtoxinA injections and the majority of subjects did not change product within the time limits of this retrospective analysis ([Table 2](#t2-ccid-5-053){ref-type="table"}). The most common reason for product change given was that the usual product was unavailable at the time of treatment.
Subject and physician satisfaction
----------------------------------
The vast majority of subjects were satisfied with their treatment (96.4% for incobotulinumtoxinA and 95.8% for onabotulinumtoxinA). There was no statistically significant difference in subject satisfaction between the two products (*P* = 1.000). Similarly, the rates of physician satisfaction were also very high for both products: 96.3% and 95.3% were satisfied with incobotulinumtoxinA and onabotulinumtoxinA, respectively (*P* = 0.825).
Interval between two treatments
-------------------------------
Any difference in the mean treatment interval (ie, the time between the first and second injections \[interval 1\] or the second and third injections \[interval 2\]; [Figure 1](#f1-ccid-5-053){ref-type="fig"}) was assessed separately in subjects who did not change product in order to give an indication of the duration of the effect. There was no statistically significant difference between the treatment intervals in subjects who did not change product ([Table 3](#t3-ccid-5-053){ref-type="table"}). The mean length of interval 1 was 25.25 weeks and 24.90 weeks for subjects treated with incobotulinumtoxinA and onabotulinumtoxinA, respectively (*P* = 0.9646). For interval 2, the mean length was 22.43 weeks and 21.95 weeks, respectively (*P* = 0.8696).
Because the group of subjects who did change product included subjects who changed product at visit 2 or visit 3 or both, no conclusions could be drawn from any differences between the lengths of the treatment intervals in this group of subjects.
Dosage
------
In order to analyze the dosages administered, subjects were again divided into two groups: those who did not change product and those who did. Within these two groups, the mean dosage of each product was calculated at each visit. For the subjects who did not change product, the mean dosages for incobotulinumtoxinA and onabotulinumtoxinA were 18.92 U and 18.79 U at visit 1 (*P* = 0.4335), 18.12 U and 18.44 U at visit 2 (*P* = 0.4262), and 18.20 U and 18.94 U at visit 3 (*P* = 0.6900), respectively. In the group of subjects who did change product, the mean dosages for incobotulinumtoxinA and onabotulinumtoxinA at each visit were 17.48 U and 16.99 U (visit 1; *P* = 0.9138), 17.13 U and 18.63 U (visit 2; *P* = 0.3168), and 17.97 U and 18.11 U (visit 3; *P* = 0.7007), respectively. The mean total treatment dose for the upper face at each treatment visit did not differ significantly between incobotulinumtoxinA and onabotulinumtoxinA for subjects who did and those who did not change product ([Table 4](#t4-ccid-5-053){ref-type="table"}). Furthermore, there was no statistically significant difference between the mean total dose of incobotulinumtoxinA and onabotulinumtoxinA across all visits (*P* = 0.35).
Safety
------
Adverse effects were analyzed in subjects who did not change product in order to assess any differences in adverse effects experienced by subjects receiving either product. In these subjects, all the adverse effects were mild or moderate in intensity. There were no severe side effects reported for either product. Of the subjects treated only with incobotulinumtoxinA, 6.7% experienced adverse effects compared with 9.5% of those treated with onabotulinumtoxinA, though this was not a statistically significant difference (*P* = 0.178). Localized pain was reported in 2.2% of subjects receiving incobotulinumtoxinA and 3.4% of subjects receiving onabotulinumtoxinA while local hematoma was reported in 3.0% and 2.7% of subjects, respectively. Headache was reported by 1.7% of incobotulinumtoxinA-treated subjects and 2.0% of onabotulinumtoxinA-treated subjects.
Discussion
==========
This retrospective study, comparing use of botulinum toxin type A preparations in daily clinical practice, analyzed subject and physician satisfaction, the time interval between doses administered (duration between treatment cycles), dosages used, and adverse effects experienced, and found no differences between incobotulinumtoxinA and onabotulinumtoxinA. The low number of abobotulinumtoxinA injections meant that they were excluded from the analysis. This low number presumably reflects the habits of the physicians who responded to the questionnaire. Amongst these physicians, abobotulinumtoxinA is the least commonly used of the three products included in this study. More subjects were treated with incobotulinumtoxinA compared with onabotulinumtoxinA due to the fact that mainly incobotulinumtoxinA injectors were approached. However, all physicians were equally proficient and experienced in administering all products and both the subject and physician satisfaction were similar for incobotulinumtoxinA and onabotulinumtoxinA.
Subject and physician satisfaction levels were similarly high for both products, and product change was rare and most often simply because the usual product was unavailable at the time, reflecting physician and subject confidence and satisfaction with both products.
The treatment intervals in subjects who did not change product were of a similar length. However, it should be noted that the length of time between two treatments is influenced by several factors, including financial and time constraints on the subject.
In this study no questions were asked relating to the reasons why subjects returned for subsequent treatments when they did, so the impact of the factors that may affect treatment interval could not be assessed here. In addition, the intention had been to analyze the different treatment areas separately, but this was not possible due to the confounding factor that allocation of individual injection points to a single isolated area was not necessarily possible (for example, some glabellar injection points also affect forehead lines and vice versa). Thus, the data were pooled and analyzed as "upper face".
Similar clinical efficacy between incobotulinumtoxinA and onabotulinumtoxinA with regard to the other parameters tested suggests that these two products have similar clinical efficacy when used at the same dosage. In addition, the total dosages for both products were similar and low, as expected for cosmetic procedures in the face. Therefore, the results from this large retrospective study (n = 1256) are in line with published results of smaller prospective clinical studies which have shown similar efficacy of incobotulinumtoxinA compared with onabotulinumtoxinA in the treatment of glabellar frown lines in 381 subjects,[@b7-ccid-5-053] crow's feet in 21 subjects,[@b6-ccid-5-053] and forehead lines in 12 subjects[@b9-ccid-5-053] in aesthetic indications using identical dosages, as well as in therapeutic indications with similar dosages.[@b10-ccid-5-053],[@b11-ccid-5-053]
The results from the large number of subjects presented in this retrospective analysis of daily clinical practice support those observed in prospective, randomized, controlled trials. Therefore, evidence from both routine clinical practice and the clinical trial setting suggests that incobotulinumtoxinA and onabotulinumtoxinA have similar clinical efficacy in therapeutic and aesthetic indications. In addition, the results of this retrospective analysis demonstrated that the identical dosage (20 U) of both products for the treatment of glabellar frown lines is appropriate.
Conclusion
==========
In daily aesthetic practice, similar clinical efficacy between incobotulinumtoxinA and onabotulinumtoxinA in terms of subject and physician satisfaction, dosage given, and safety were observed. These data support comparable and similar therapeutic efficacy of these two products and clinicians may alternate between incobotulinumtoxinA and onabotulinumtoxinA as product availabilities dictate.
This retrospective analysis and publication were supported by Merz Pharmaceuticals GmbH.
**Disclosure**
WP has served as a consultant and lecturer for Allergan Inc, Merz Pharmaceuticals GmbH and Galderma Pharma SA. MI has served as a consultant, speaker, and investigator in clinical trials for Merz Pharmaceuticals GmbH. UK received honoraria as an investigator in clinical trials and speaker for Merz Pharmaceuticals GmbH. L-PK has spoken at workshops and training sessions for Merz Pharmaceuticals GmbH. WGP-D has acted as a consultant and lecturer for Allergan Inc, Merz Pharmaceuticals GmbH, and Galderma Pharma SA. TP has served as a consultant and speaker at workshops and meetings for Merz Pharmaceuticals GmbH and Galderma Pharma SA. She has received honoraria as an investigator in clinical trials for Merz Pharmaceuticals GmbH, Galderma Pharma SA, and Ipsen Pharma. Editorial assistance was provided by Ogilvy 4D and funded by Merz Pharmaceuticals GmbH.
{#f1-ccid-5-053}
######
Components of questionnaire completed by physicians
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
***Variables***
**Demographics and baseline characteristics**
The following variables were considered as demographics and baseline characteristics: Gender (male/female)Age (years)Used botulinum toxin (Bocouture^®^/Xeomin^®^; Vistabel^®^/Botox^®^; Azzalure^®^/Dysport^®^)Treated indication (glabellar frown lines, horizontal forehead lines, crow's feet)
**Side effects**
Investigators were asked if any side effects occurred during the treatment period. If so, the nature of the side effects could be specified by choosing one or more of the following possible answers: Local painLocal hematomaHeadachePtosis of the upper eyelidPtosis of the eyebrowsOtherIn addition, the severity of the most relevant side effect could be assessed (mild, moderate, severe)
**Satisfaction with product**
The investigator was asked to assess his/her satisfaction with the product (physician's satisfaction) as well as to transcribe the subject's satisfaction from subject documentation. The following answers were possible: SatisfiedNot satisfiedUnknown
**Change of product**
Participants were asked if the neurotoxin product was changed during the treatment period. If so, the following reasons could be assigned: Dissatisfaction with cosmetic resultDissatisfaction due to side effectsOther
**Treatment details**
The treatment details were recorded for at least two but not more than three injections. The following data were collected: Treatment dateTime interval until the next treatment in the same indicationOverall dose for the treatment in this indicationIndividual doses for each injection point, where it was possible to choose between 24 different injection points
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
######
Demographic and baseline characteristics
n (%)
---------------------------- -------------
Total number of subjects 1256 (100)
Gender
Male 105 (8.4)
Female 1151 (91.6)
Indication
Glabellar frown lines 1096 (48.3)
Horizontal forehead lines 621 (27.4)
Crow's feet 553 (24.4)
Product
IncobotulinumtoxinA 1998 (88.0)
OnabotulinumtoxinA 236 (10.4)
AbobotulinumtoxinA 36 (1.6)
Product change
Missing 3 (0.1)
No 2018 (88.9)
Yes 249 (11.0)
Age, years
Mean 47.4
STD 10.91
Maximum 94.0
Median 46.0
Minimum 20.0
**Abbreviation:** STD, standard deviation.
######
Average interval between treatments in the upper face for subjects without product change
Interval 1 Interval 2
------------- -------------- ------------ -------------- -------
Mean, weeks 25.25 24.90 22.43 21.95
STD 15.21 12.65 10.55 7.87
Median 22.43 23.29 20.57 20.86
*P* = 0.9646 *P* = 0.8696
**Note:** *P* value calculated using the Wilcoxon--Mann--Whitney test.
**Abbreviation:** STD, standard deviation.
######
Mean total treatment doses in the upper face at each treatment visit for subjects with and without product change
Treatment visit Subjects without product change Subjects with product change
----------------- --------------------------------- ------------------------------ -------------- -------
1
Mean 18.92 18.79 17.48 16.99
STD 9.2 8.77 9.62 6.38
*P* = 0.4335 *P* = 0.9138
2
Mean 18.12 18.44 17.13 18.63
STD 8.85 6.80 6.62 9.34
*P* = 0.4262 *P* = 0.3168
3
Mean 18.20 18.94 17.97 18.11
STD 9.39 4.68 5.8 8.01
*P* = 0.6900 *P* = 0.7007
**Note:** *P* value calculated using the Wilcoxon--Mann--Whitney test.
**Abbreviation:** STD, standard deviation.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
A large number of studies have focused on subjective well-being[1](#Fn1){ref-type="fn"} (see Diener [@CR14]; Helliwell [@CR25]; Kahneman et al. [@CR30]), and many have found that married people have higher subjective well-being (Mikucka [@CR49]; Stutzer and Frey [@CR71]; Waite and Gallagher [@CR79]). Yet the increase in cohabitation---not just as a prelude to marriage but also as an alternative partnership type and an accepted setting for parenthood (Perelli-Harris et al. [@CR59])---raises questions as to whether only marriage has beneficial effects. Given that cohabitation shares many of the same characteristics as marriage---for example, intimacy, emotional and social support, and joint residence---cohabitors may have similar well-being to those who are married (Soons et al. [@CR70]; Zimmerman and Easterlin [@CR83]).
One of the key issues when analyzing the relationship between partnership status and subjective well-being (SWB) is selection. Cross-sectional studies have often recognized the inability to disentangle selection and causality (e.g., Lee and Ono [@CR41]; Soons and Kalmijn [@CR69]). Longitudinal studies tend to use fixed-effects (FE) models, which focus on within-individual transitions in partnership status, a process usually occurring in younger adulthood and covering a relatively short period of the life course (Kalmijn [@CR32]; Musick and Bumpass [@CR53]; Soons et al. [@CR70]). These studies did not examine the long-term effects of partnership in midlife, after the majority of individuals have made decisions about marriage and childbearing. Midlife[2](#Fn2){ref-type="fn"}---after women's prime reproductive period, and when being married may influence one's identity and well-being---is an understudied part of the life course (Lachman [@CR37]). At this point in life, the initial boost in happiness may have declined (Soons et al. [@CR70]; Zimmermann and Easterlin [@CR83]), and raising children may confound SWB (Balbo and Arpino [@CR5]; Margolis and Myrskylä [@CR47]).
FE models also do not directly consider individuals who do not experience a change in partnership status: that is, those who cohabit and never marry. Thus, we cannot tell whether certain groups---for example, those living in a coresidential partnership who have a low propensity to marry---would benefit if they married rather than cohabited; or alternatively, whether the benefits to marriage may be more pronounced for those who have a higher propensity to marry. Given the interest in marriage promotion policies, targeting low-income individuals in countries such as the United Kingdom, it is important to examine whether those unlikely to marry would be happier if they did marry. To address these selection processes, we use propensity score--weighted regression analysis to pinpoint how early life and/or current conditions influence conditions in midlife. This method allows us to address baseline bias and differential treatment bias (Morgan and Winship [@CR52]), which occur when the link between marriage and well-being varies across subgroups.
Cross-sectional research has found a "happiness gap" between cohabitation and marriage in most countries, but the size of the gap appears to vary and may be linked to the acceptance and prevalence of cohabitation in a society (Soons and Kalmijn [@CR69]) or gender context and religious norms (Lee and Ono [@CR41]). However, it is unclear whether the long-term effects of selection in different countries operate similarly, especially because union duration, childbearing experience, and meanings of cohabitation differ across countries (Hiekel et al. [@CR27]; Perelli-Harris et al. [@CR60]). In addition, the heterogeneity of treatment effects may vary, indicating that marriage has different benefits for different groups, depending on the country. Here we compare the association between partnership type and SWB in the United Kingdom, Australia, Germany, and Norway, which have experienced substantial increases in cohabitation over the past few decades but have different family policies (Perelli-Harris and Sánchez Gassen [@CR61]) and cultural orientations toward marriage (Perelli-Harris et al. [@CR60]). Each of these contexts leads us to predict a certain relationship between partnership status and SWB.
This study addresses a number of gaps in the literature. First, we provide new insights into selection due to childhood background and current characteristics for the association between union type and SWB. Second, we examine whether marriage may be especially advantageous for partnered individuals who have a lower or higher propensity to marry. Third, we analyze the extent to which the relationship between partnership type and SWB varies by country and gender. More broadly, analyzing how cohabitation differs from marriage for individuals' well-being will contribute to our understanding of the meaning and consequences of cohabitation as well as the extent to which these meanings differ across contexts.
Theoretical Background {#Sec2}
======================
A large body of research has investigated the beneficial aspects of marriage for well-being (for reviews, see Nelson-Coffey [@CR56]; Waite and Gallagher [@CR79]). These studies posited that married partners benefit from sexual and emotional intimacy, companionship, and daily interaction (Kamp Dush and Amato [@CR33]; Umberson et al. [@CR75]). Spouses help each other cope with stress by providing social and emotional support. Recognition from a spouse may provide symbolic meaning in life (Umberson et al. [@CR75]). Additionally, sharing a household can lead to economies of scale, and married spouses could profit from a larger friendship and kin network (Ross and Mirowsky [@CR66]; Umberson and Montez [@CR76]). All these mechanisms could enhance SWB.
Nonetheless, given dramatic social change over the past decades, the benefits to marriage may be declining (Liu and Umberson [@CR43]). A recent study comparing 87 countries found that the life satisfaction advantage of married men compared with unmarried men has waned over the last three decades, suggesting that marriage has become less advantageous (Mikucka [@CR49]). This decline may be partially due to the increase in cohabitation, especially in high-income countries. Cohabitation may be taking on much of the form and function of marriage (Cherlin [@CR10]), especially as cohabiting unions become longer and involve children. Similar to married couples, cohabiting couples share a household and may benefit from similar intimacy, support, care, and family networks. Normative expectations to marry have become weaker, and the tolerance for nonmarital arrangements has increased in most countries (Treas et al. [@CR74]).
A large body of research, however, has found that cohabitors often differ from married couples. Across countries, cohabitors have lower second birth rates (Perelli-Harris [@CR57]), are less likely to pool incomes (Gray and Evans [@CR22]; Lyngstad et al. [@CR45]), have lower relationship quality (Wiik et al. [@CR81]), and are more likely to dissolve their relationships (Galezewska et al. 2017), even if they have children (Musick and Michelmore [@CR54]). Qualitative research from Europe and Australia has suggested that many still think of cohabitation as a less-committed type of union than marriage and instead oriented toward freedom and independence (Perelli-Harris et al. [@CR60]). Marriage may thus still be desired by most people but more as a cultural ideal or status symbol.
Recently, most scholars have used FE models to examine whether cohabitation is similar to marriage in increasing SWB. This approach allows the testing of set-point theory, which posits that individuals have a baseline level of happiness that cannot be permanently modified by life events, such as union formation. This theory has been tested in a range of settings, and the findings support a positive effect of marriage and cohabitation on SWB (Musick and Bumpass [@CR53]; Soons et al. [@CR70]; Zimmermann and Easterlin [@CR83]), with cohabitation having a weaker effect (Kalmijin 2017). Some studies, however, have questioned set-point theory and found that different model specifications can result in long-term improvements for marriage (Anusic et al. [@CR3]). Overall, however, most studies indicated that, on average, marriage provides a boost to well-being, with cohabitation providing a weaker boost, and individuals return to original happiness levels in the long term.
Selection Processes {#Sec3}
-------------------
Selection processes---also referred to as *baseline bias* (Morgan and Winship [@CR52])---select people into marriage and may be responsible for higher SWB in midlife. These processes can begin early in childhood and continue into adulthood (Elo [@CR17]; Kuh et al. [@CR36]; Umberson et al. [@CR75]). For example, parents' education and socioeconomic status (SES) are strongly associated with adult life satisfaction (Frijters et al. [@CR20]; Layard et al. [@CR40]), but can also influence decisions about cohabitation and marriage, especially around the time of a first birth (Koops et al. [@CR35]; Wiik [@CR80]). Parental divorce in childhood may have long-term effects on future SWB, both emotionally and financially (Amato [@CR2]), and lead the children of divorced parents to choose cohabitation over marriage for their first relationship (Perelli-Harris et al. [@CR58]). Thus, any positive association between marriage and SWB may not be due to the benefits of marriage but instead childhood conditions and experiences, which may both select people into marriage and lead to higher SWB.
Prior research has consistently found that current conditions, such as income and education, are also important for understanding SWB (e.g., Diener et al. [@CR15]; Layard [@CR39]). Unemployment has a persistent negative effect on SWB, especially for men (Clark et al. [@CR13]). Household income and partner's education and employment status may be particularly important for women if they are not working. In addition, people with higher education and better economic prospects are more likely to marry (Kalmijn [@CR31]; Sweeney [@CR73]) and stay married (Matysiak et al. [@CR48]). Another key indicator of SWB in adulthood is poor health (Binder and Coad [@CR9]), which can lead to poor relationship quality and potentially influence marriage decisions. Unlike selection in childhood, selection in adulthood could be confounded with partnership formation and happiness; for example, a woman could drop out of the labor market after marriage, or unhappy people could be more likely to lose their job.
Differential Treatment Bias {#Sec4}
---------------------------
Selection processes result in individuals having a different propensity to cohabit or marry, which can produce *differential treatment bias* (Morgan and Winship [@CR52]). The effect of the treatment, in this case marriage, can vary based on selection characteristics. As we discuss earlier, most prior research has shown that marriage is selective of those who grew up in stable married-parent families with higher SES, and cohabitation is selective of those raised by single mothers or low-income parents (Wiik [@CR80]). Thus, the benefits of marriage over cohabitation may be differentially distributed according to these background characteristics (Su et al. [@CR72]). The differential treatment effect may result in marriage affecting SWB through different potential mechanisms.
### High Propensity to Marry, Positive SWB {#Sec5}
Individuals with a high propensity to marry may be more likely to have higher SWB if they marry, because they may be more affected by social norms and expectations to marry, especially because marriage can be considered an expression of status (Cherlin [@CR11]). They may benefit more from the legal security and access to courts that official marriage provides when property or resources are combined, especially for couples who have children. Thus, marriage may protect their standard of living and provide a sense of security in case of divorce (Perelli-Harris et al. [@CR60]).
### High Propensity to Marry, Negative/Neutral SWB {#Sec6}
Those with a higher propensity to marry may not benefit more from marriage than cohabitation, because they have the emotional, social, and financial resources to provide a positive outlook regardless of marital status.
### Low Propensity Positive SWB {#Sec7}
People with a lower propensity to marry may benefit more from marriage, because it is recognized not only legally but also by family, friends, and the community, who may then provide greater social support (Marcussen [@CR46]; Umberson and Montez [@CR76]). Rather than symbolizing social status, marriage may signify having achieved stability and protection. Married people may also have greater trust in the long-term prospects of their relationship given that marriage is usually intended for life.
### Low Propensity to Marry, Neutral/Negative SWB {#Sec8}
Marriage may not be advantageous to those with a low propensity to marry, if individuals have no desire to marry their partners. Those from a disadvantaged background are more likely to have partners less suitable for marriage---for example, because they are unemployed or have fewer resources and greater debt. They may be happier cohabiting because of the high perceived risk of divorce and its associated costs.
Gender Differences {#Sec9}
==================
Although some prior studies have found few gender differences in the effects of marriage on psychological well-being (e.g., Williams [@CR82]), men and women may receive different benefits from being in marriage compared with cohabitation (Liu and Umberson [@CR43]), and these benefits may differ by the propensity to marry. Previous studies have argued that marriage provides men with greater social recognition and support, thereby positively influencing their well-being (Ross et al. [@CR67]). Men may benefit from the status of marriage more than women, particularly in the workplace (Killewald and Gough [@CR34]). On the other hand, if men benefit more from emotional support and regular sexual intimacy, cohabitation may provide advantages similar to those of marriage. Women could benefit more from marriage because of the higher economic resources and legal protection that provides them with a sense of safety (Waite [@CR78]), which are especially important when bearing and raising children. Women may also feel a sense of satisfaction from the wedding and achieving normative social aspirations (Berrington et al. [@CR8]). Nonetheless, women may prefer cohabitation if they oppose the traditional, patriarchal constraints of marriage. Disadvantaged women may not want to marry unsuitable partners who cannot achieve a certain economic bar (Edin and Reed [@CR16]; Lichter et al. [@CR42]; Smock et al. [@CR68]). In such cases, marriage for women may be more detrimental to SWB than cohabitation.
Cross-National Differences {#Sec10}
--------------------------
Welfare state context, family policies, and cultural attitudes shape partnership formation and may influence the association between partnership formation and SWB. Selection processes may also differ across countries, which may in turn produce heterogeneous treatment effects, with some groups benefiting more from marriage than others. Table [1](#Tab1){ref-type="table"} outlines how welfare state, legal system, social norms, and strength of selection would predict differences in SWB by partnership on average and by those with a high or low propensity to marry in each country.Table 1Brief description of welfare state, legal status of cohabitation, social norms, and social selection in each country (see the text for references): Summary of average and differential expectations for men and womenWelfare StateLegal Status of CohabitationSocial NormsSocial Selection Into Midlife CohabitationAverage Expectations for MenDifferential Expectations for Men ^a^Average Expectations for WomenDifferential Expectations for Women ^a^United KingdomLiberal welfare stateInferior/ignored in most policy areasCohabitation is accepted, but marriage is often preferred.Strong based on disadvantageSignificant difference between cohabitation and marriage, but eliminated after selection taken into accountHP, neutral SWBLP, positive SWBSignificant difference between cohabitation and marriage, but eliminated after selection taken into accountHP, neutral SWBLP, positive SWBAustraliaLiberal welfare state, with higher benefitsEquivalent to marriage (after 0.5 year or with children)Cohabitation is accepted, but marriage is often preferred.Strong based on disadvantageSignificant difference between cohabitation and marriage, but eliminated after selection taken into accountHP, neutral SWBLP, positive SWBSignificant difference between cohabitation and marriage, but eliminated after selection taken into accountHP, neutral SWBLP, neutral SWBGermanyConservative welfare stateInferior/tax advantages to marriage, especially if one partner earns moreMarriage is usually ideal in West; cohabitation is much more accepted in East.Weak based on disadvantageSignificant difference between cohabitation and marriageHP, neutral SWBLP, positive SWBSignificant difference between cohabitation and marriageHP, neutral SWBLP, positive SWBNorwaySocial-democratic welfare regimeMostly equivalent to marriage (after two years or with children)Cohabitation and marriage are equal, but marriage is sometimes preferred.Weak based on disadvantageNo difference between cohabitation and marriageHP, neutral SWBLP, neutral SWBNo difference between cohabitation and marriageHP, neutral SWBLP, positive SWB^a^HP = high propensity to marry. LP = low propensity to marry.
The United Kingdom has been classified as a liberal welfare state (Esping-Andersen [@CR18]) with a minimal safety net and a prioritization of means-tested benefits. In England, the Conservative party promoted marriage as an institution, and the current UK government legislated tax breaks for married couples in 2013 (BBC [@CR7]). Lawyers have described the legal situation for cohabitors as chaotic, with cohabiting couples receiving little access to family courts upon separation and receiving no inheritance rights upon death of a partner (Barlow [@CR6]). Cohabitation is strongly selective of disadvantage (Perelli-Harris et al. [@CR62]), and the means-tested welfare system may provide further disincentives to marry. Thus, the lack of rights for cohabitors and the social norms favoring marriage should on average make marriage more advantageous than cohabitation. However, differences between cohabitation and marriage should disappear after selection is taken into account. Nonetheless, both men and women in partnered unions with a low propensity to marry should have higher SWB if married (low propensity, positive SWB), because marriage would provide them with a greater sense of security and stability, especially important when the welfare state does not provide sufficient support.
Although Australia is often classified as a liberal welfare state, it has higher levels of benefits (Arts and Gelissen [@CR4]), which may result in women being less dependent on marriage. The state also allows cohabiting couples the same access to family courts as married couples upon union dissolution and similar rights to inheritance (Evans [@CR19]). Nonetheless, highly educated individuals are more likely to be married than lower-educated individuals (Evans [@CR19]; Heard [@CR24]). Thus, we expect stark differences in SWB between cohabitation and marriage on average but also that these differences will be eliminated after taking selection into account. Given the strong state support for cohabiting couples, women with a low propensity to marry would not be better off if married (low propensity, neutral/negative), but men with a low propensity to marry would still benefit from the perceived social status and stability achieved through marriage (low propensity, positive SWB).
Germany has a traditional welfare state with a generally conservative view on families (Esping-Andersen [@CR18]). Marriage was enshrined in the Constitution, and the German state directly promotes marriage with tax incentives, especially relevant when couples have children, and the mother drops out of the labor force (Perelli-Harris and Sánchez Gassen [@CR61]). Most laws related to health insurance, inheritance, and property regulation favor married couples. Given the normative and legal privileging of marriage, we expect on average that SWB will be higher among those who are married. On the other hand, tax benefits to marriage apply only to couples in which the man earns substantially more than the wife; dual-earner couples would be just as well off if they cohabit. Thus, we expect that marriage will provide fewer benefits for both men and women with a high propensity to marry (high propensity, neutral/negative SWB) because they would be happy regardless of marriage. Men and women with a lower propensity to marry, however, would benefit more from marriage, because it would provide them with a greater sense of security and stability (low propensity, positive SWB).
Norway's social-democratic welfare state, with its emphasis on gender equality and individual autonomy (Esping-Andersen [@CR18]), may have facilitated the increase in cohabitation (Lappegård and Noack [@CR38]). Norwegian policies tend to support dual-earner families by providing parental leave and public childcare; but the tax and transfer system, which focuses on individuals, may lower the incentives for couples to marry (Perelli-Harris and Sánchez Gassen [@CR61]). The legal system has also harmonized many rights and responsibilities between cohabiting and married people, especially for those with children and in long-term unions; however, inheritance rights are still reserved for married spouses. Overall, the general tolerance of cohabitation in Norway (Treas et al. [@CR74]) and high levels of nonmarital childbearing suggest that on average, cohabiting and married individuals should have similar levels of SWB, regardless of their likelihood to marry. On the other hand, prior research has indicated that low-educated women are more likely to give birth within cohabitation than highly educated women (Perelli-Harris et al. [@CR62]), indicating that cohabitation is selective of disadvantage. Thus, we expect that women with a low propensity to marry would benefit from marriage because of the increased social support and family networks that marriage would provide to disadvantaged women (low propensity, positive SWB).
Data, Methods, and Measures {#Sec11}
===========================
Data {#Sec12}
----
We employ four harmonized data sets, which are the best suited in each country for studying the effects of partnership status in midlife. For the United Kingdom, we use the UK Household Longitudinal Study (UKHLS), a nationally representative, household-based longitudinal survey (University of Essex [@CR77]). The survey started in 2009 with approximately 51,000 individuals and is conducted annually. Our sample comes from the fourth wave conducted in 2012/2013, with a total of 47,157 individuals surveyed. Our sample aged 38--50 includes 8,941 individuals who answered the SWB and partnership questions.
The Australian data come from the Household, Income and Labour Dynamics in Australia (HILDA), a nationally representative, household-based longitudinal survey. The survey started in 2001 and annually interviews all adults over age 15 (13,969 individuals) in the selected households. The sample expanded with a general top-up at Wave 11 in 2012; in 2013 (our analysis year), 13,536 individuals were interviewed. After excluding those not aged 38--50 and who did not answer the SWB and partnership questions, our sample comprises 3,787 individuals.
For Germany, we use the Socio-Economic Panel (SOEP), a representative longitudinal study of households, with all household members interviewed annually (from age 15). SOEP began in 1984 with 12,290 individuals. Apart from the inclusion of participants from the former East-German state after German reunification, this study had several refreshment samples over its 30-year duration in order to assure national representation. Our sample comes from the 2013 wave, which surveyed 24,113 individuals, 8,830 of whom were aged 38--50.
We also use the Norwegian Generations and Gender Survey (GGS) because no equivalent panel surveys study partnerships in midlife in Norway, and cohabitation among couples without children has been only recently recorded in the registers. The GGS is a nationally representative cross-sectional survey of respondents aged 18--79 and has information that can approximate our longitudinal design. It includes 15,114 individuals interviewed by telephone and a self-administered questionnaire (SAQ) in 2007, combined with data from administrative records. After excluding those who did not answer the SWB or partnership questions, we include 2,785 individuals aged 38--50 in our sample.
Methods {#Sec13}
-------
In this study, we use propensity score--weighted regression analysis to address both types of selection: baseline bias and differential treatment bias (e.g., Su et al. [@CR72]). This approach has the advantage of addressing selection while incorporating contemporaneous factors that may influence SWB. Figure [1](#Fig1){ref-type="fig"} presents our analytical approach, depicting the variables from childhood (0--16) through to midlife (38--50). Childhood conditions are used in the propensity score and included as controls in the weighted regression models. Prior family experiences, captured for the age range 16--50, and current factors measured at the time of the interview are included as controls in the weighted regression models.Fig. 1Analytic approach
In the first step of the analysis, we estimate propensity scores using logistic regression for the probability of selection into the treatment group---marriage---based on a set of observed characteristics (Rosenbaum and Rubin [@CR65]). See Table [A1](#MOESM1){ref-type="media"} in the online appendix for the logistic regression results for each country. In the second step, the propensity scores are used as weights in regression analyses (Morgan and Todd [@CR51]). Through weighting, the treatment and control group become comparable. For example, cohabiting individuals who are similar to married individuals with respect to background characteristics are up-weighted, and cohabitors with features dissimilar to those of the married are down-weighted. We obtain three weights that function as conditional predicted probabilities of being married at midlife. The first weight is the average treatment effect (ATE), which estimates the effect of marriage on SWB for the entire sample, controlling for selection and other characteristics (Eq. ([1a](#Equ1){ref-type=""})). The second weight is the average treatment effect on the treated (ATT), which estimates the effect of marriage for those with a high propensity to marry (Eq. ([1b](#Equ2){ref-type=""})). The third weight, the average treatment effect of the control (ATC), shows the effect that marriage would have on SWB for those with a low propensity to marry relative to cohabiting (Eq. ([1c](#Equ3){ref-type=""})).$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$ {\displaystyle \begin{array}{l}\mathrm{For}\kern0.5em {d}_i=1:{w}_{i, ATE}=\frac{1}{{\hat{p}}_i}\\ {}\mathrm{For}\kern0.5em {d}_i=0:{w}_{i, ATE}=\frac{1}{1-{\hat{p}}_i}\end{array}} $$\end{document}$$$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$ {\displaystyle \begin{array}{l}\mathrm{For}\kern0.5em {d}_i=1:{w}_{i, ATT}=1\\ {}\mathrm{For}\kern0.5em {d}_i=0:{w}_{i, ATT}=\frac{{\hat{p}}_i}{1-{\hat{p}}_i}\end{array}} $$\end{document}$$$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$ {\displaystyle \begin{array}{l}\mathrm{For}\kern0.5em {d}_i=1:{w}_{i, ATC}=\frac{1-{\hat{p}}_i}{{\hat{p}}_i}\\ {}\mathrm{For}\kern0.5em {d}_i=0:{w}_{i, ATC}=1.\end{array}} $$\end{document}$$
Weighting seeks to make the treatment and control groups comparable, or balanced, in terms of certain characteristics. We test how well the data are balanced by estimating the average standardized mean difference between treatment and control groups for all categorical covariates, and the standardized difference in standard deviations for continuous covariates (see Morgan and Todd [@CR51]). Table [A2](#MOESM1){ref-type="media"} in the online appendix shows that the weights successfully balance the data, as shown in the improved average standardized mean and standard deviation. Estimating the same regression models (available on request) for sample members of the treatment group who have a counterpart in the control group---*cases of common support* (Morgan and Todd [@CR51])---supports our initial interpretation. We also use a combination of propensity score weighting with covariate adjustment, which can correct a small imbalance when the propensity score weighting may not have sufficiently balanced the sample (Morgan and Todd [@CR51]). Although propensity score analyses are effective at removing bias caused by observed variables, omitted variables cannot be considered in this adjustment.
In a third step, we estimate ordinary least squares (OLS) regressions using the three propensity weights for men and women in each country (Eq. ([2](#Equ4){ref-type=""})). Although the estimation of the propensity scores requires complex modeling, the final analysis is generally straightforward. *Y*is level of SWB; *D* is the treatment variable (being married); δ̂ is the estimated effect of *D* on *Y*, adjusted for **X**; and **X** is a vector of observed variables that are thought to determine *D* and *Y*. We first estimate the average effect of being married on SWB with regressions that applied the ATE weights. We next assess causal effect heterogeneity by comparing the results from ATT- and ATC-weighted regressions. In other words, we compare the effect of marriage for those with a high propensity to marry (ATT) with the effect of marriage on those with a low propensity to marry (ATC).$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$ Y=\upalpha \hat{\mkern6mu} +\updelta \hat{\mkern6mu} \mathrm{D}+\mathbf{X}\upbeta \hat{\mkern6mu} +\upvarepsilon . $$\end{document}$$
Measures {#Sec14}
--------
We harmonize the variables in each survey, although region remains context-specific. Missing observations for independent variables are imputed using the mi impute command in Stata 13.0.
### Dependent Variable {#Sec15}
In all countries, SWB is measured with a single item in the latest wave analyzed.[3](#Fn3){ref-type="fn"} The responses are recorded on an 11-point scale ranging from 0 = completely dissatisfied to 10 = completely satisfied with life, except in the United Kingdom, where the scale ranges from 1 = completely dissatisfied to 7 = completely satisfied. The single-item life satisfaction measure is widely used in demographic research (e.g., Balbo and Arpino [@CR5]) and has similar psychometric characteristics to multiple-item scales (Cheung and Lucas [@CR12]).
### Age {#Sec16}
All models include a yearly control for age, given that the age range is 38--50. Other specifications of the age range do not alter the results.
### Childhood Characteristics {#Sec17}
We distinguish three types of childhood background characteristics important for union formation and well-being: (1) region, and parent's nativity; (2) family structure in childhood; and (3) parental SES. We choose the most relevant measure of region for each country.
Large regional differences exist across the United Kingdom, and we control for four major regions: (1) Scotland, Ireland, and North England; (2) Midlands and Wales; (3) South West England; and (4) South East England. In Australia, we include a control only for differences between rural and urban areas: prior studies have indicated these are the most salient for understanding geographic disparities (Monnat and Beeler Pickett [@CR50]). In Germany, the East-West divide is particularly important for family formation. In eastern Germany, nearly two-thirds of children were born in cohabitation; in western Germany, only one-third were born in cohabitation (Perelli-Harris et al. [@CR59]). Because of these regional distinctions and migration from the East to the West after reunification, we control for the following four categories: (1) West Germany; (2) East Germany stayed; (3) East moved to West Germany; and (4) born outside Germany. In Norway, we control for four regions: (1) Oslo area; (2) East; (3) South and West; and (4) Mid and North.
In all countries, we include indicators for whether the respondent and the respondent's parents were born in the country. Family structure in childhood includes a dummy variable for whether the respondent lived with both parents at age 15 (Norway) or up to age 16 (United Kingdom, Australia, and Germany). SES of parents includes mother's and father's education (low, medium, and high), whether mother worked, and father's occupation (low, medium, high, and not employed) when respondent was aged 14 or 15.
### Partnership and Childbearing Experiences {#Sec18}
As we discuss earlier, cohabiting partnerships could be very similar to marital partnerships, especially with regard to union duration and childbearing behavior. These experiences could also be endogenous to choice of partnership type. Longer union duration often signals a deeper investment in the relationship (Lyngstad et al. [@CR45]), which could reduce differences between cohabitation and marriage. Current union duration is included as a linear variable.[4](#Fn4){ref-type="fn"} Parenthood also directly affects SWB, but results from previous studies are mixed, indicating that children can have a positive or negative impact (Aassve et al. [@CR1]; Balbo and Arpino [@CR5]; Myrskalä and Margolis [@CR55]; Pollmann-Schult [@CR64]). Partnership and childbearing histories were collected in different ways in each survey. In the UKHLS, HILDA, and GSOEP, they were collected retrospectively in the first wave of the survey and updated during each additional panel. In the Norwegian GGS, the histories were collected retrospectively at the main wave of the survey (2007). Here, childbearing history is represented with three categories: (1) respondent has no child; (2) respondent has at least one child, but not with current partner; and (3) respondent has at least one child with current partner.[5](#Fn5){ref-type="fn"} Other specifications of childbearing history, including total number of children, do not change the main effects of partnership status on SWB.
The experience of union dissolution is another important control because it can have long-term effects on health and mental well-being, even if individuals repartner (Amato [@CR2]; Hughes and Waite [@CR29]). Separation can also influence decisions about repartnering: people who have separated are more likely to choose cohabitation for subsequent partnerships (Galęzewska et al. [@CR21]). A dummy variable indicates whether the respondent ever experienced separation and/or divorce. We also test models restricted to repartnered individuals with children, and the results do not change substantially.
Finally, relationship satisfaction is another key factor that may be very important for mediating the effect of marriage on well-being. Any association between marriage and SWB may not be due to marriage itself but instead may be due to married people, on average, having higher relationship satisfaction, which is highly correlated with SWB. Thus, some individuals in cohabiting relationships may have similar levels of well-being to those in marital relationships if their relationship provides them with similar levels of satisfaction. On the other hand, relationship satisfaction is most likely endogenous to both marriage and SWB. People who are more satisfied with their relationship are more likely to have higher levels of SWB (Kamp Dush and Amato [@CR33]), and happier couples are more likely to marry (Gustavson et al. [@CR23]). In addition, cohabitors are usually less satisfied with their relationships than married individuals (Wiik et al. [@CR81]). We include relationship satisfaction in our models to control for the potential similarities between cohabitation and marriage, but relationship satisfaction could also be interpreted as a mediator between marriage and SWB, and it may be a proxy for the decision to marry. Relationship satisfaction is measured with a scale from 0 (very unhappy with relationship) to 10 (very happy with relationship) for Australia and Norway. For the United Kingdom, the scale ranges from 0 to 7, and we rescale it to be similar to the other countries. For Germany, because relationship satisfaction was not asked, we control for satisfaction with family life, measured on a scale of 0--10.
### Current Situation {#Sec19}
We control for contemporaneous factors measured at the time of the most recent survey, which could be considered endogenous but have been found to be very important for SWB (Kamp Dush and Amato [@CR33]). We include self-rated health measured on a five-level scale (from 1 = poor to 5 = excellent). The socioeconomic background of the person is represented by education (low, medium, or high), employment status (employed, out of the labor force, or unemployed), and household income (quintiles). Partner's education is also measured in three categories (low, medium, or high); however, partner's employment status is a dummy variable (employed or out of the labor force).
Results {#Sec20}
=======
Descriptive Statistics {#Sec21}
----------------------
Table [2](#Tab2){ref-type="table"} presents (1) the percentage of individuals living with and without a partner for the entire sample and (2) the percentage married and cohabiting among those who are partnered. It also shows mean SWB (with confidence intervals) by partnership type. Immediately, we see large significant differences in SWB between the partnered and unpartnered in all countries. Differences between cohabitation and marriage, however, are significant only in the United Kingdom and Australia. Table [3](#Tab3){ref-type="table"} presents the key independent variables that may explain differences between the two relationship types. Because Table [3](#Tab3){ref-type="table"} indicates gender difference in levels of SWB, we perform separate analyses by gender.Table 2Percentage and number of those partnered or unpartnered, and married or cohabiting, mean subjective well-being, and 95 % confidence intervals (CI), men and women aged 38--50United KingdomAustraliaGermanyNorway%\
(*n*)Mean (95 % CI)%\
(*n*)Mean (95 % CI)%\
(*n*)Mean (95 % CI)%\
(*n*)Mean (95 % CI)Partnered677.44697.80807.45748.35(6,006)(7.39,7.49)(2,629)(7.75,7.85)(7,085)(7.41,7.49)(2,051)(8.30,8.42)Unpartnered336.55317.23206.49267.84(2,935)(6.45,6.65)(1,158)(7.13,7.33)(1,745)(6.40,6.58 )(734)(7.70,7.97)Total *N* of Sample8,9417.213,7877.648,8307.212,7858.22(7.16,7.26)(7.59,7.69)(7.17,7.24)(8.14,8.30)Married837.48877.84887.46848.43(4,988)(7.42,7.53)(2,288)(7.78,7.89)(6,269)(7.42,7.50)(1,727)(8.37,8.49)Cohabiting177.25137.48127.35168.24(1,018)(7.12,7.37)(341)(7.32,7.64)(816)(7.24,7.46)(324)(8.09,8.39)Total *N* of Partnered6,0067.432,6297.807,0857.442,0518.40(7.39,7.49)(7.45,7.85)(7.40,7.47)(8.33,8.47)*Source:* Own calculations using UKHLS (United Kingdom), HILDA (Australia), SOEP (Germany), and GGS (Norway).Table 3Descriptive statistics for cohabiting (COH) and married (MAR) men and women in midlifeUnited KingdomAustraliaGermanyNorwayMenWomenMenWomenMenWomenMenWomenCOHMARCOHMARCOHMARCOHMARCOHMARCOHMARCOHMARCOHMARSubjective Well-being7.37.47.27.57.47.87.67.97.17.57.57.48.28.38.18.5 Mean/SD2.12.22.42.31.41.31.71.31.51.51.51.51.21.31.71.3Family Behavior Union duration in years11.216.511.818.59.814.912.616.78.521.28.721.811.717.613.920.2 Mean/SD7.76.78.26.97.47.28.27.66.211.55.710.16.56.36.76.3 Ever separated (%) No previous cohabiting union36752976448847866762786253785377 Separated/divorced64257124561253143338223847224723 Children with partner (%) No children50324638377325449307236184 Child with previous partner1551842142162920432024111811 Child with current partner36633658428947892771277353846485 Relationship satisfaction^a^6.67.06.56.87.58.17.57.97.98.37.88.18.58.78.28.7 Mean/SD2.12.12.22.22.21.92.32.01.71.51.71.51.51.42.01.4Childhood Background Parental separation^b^ Yes292028202415231627132115118107 No71807280768577847387798589929093 Both parents native (%) Yes79667766385748567985909093909490 At least one foreign-born213423346243524421151010710610 Mother's education (%) Low35394340605357562032233044444839 Medium58555153313327277155705848444449 High76679141817913712812812 Father's education (%) Low3535413733363840816101433324328 Medium55535052534344417065776656484948 High1012101214211819221913201120824 Mother's employment status (%) Not employed32393032414342462128233233372931 Employed68617068595758547972766867637169 Father's employment status (%) Not employed7687889964841323 Employed93949293929291919496929699979897 Father's occupation (%) Low59546357262528232028272374657463 Medium1112910293336323539524423302231 High3034283345423645453321333646Current Situation Education (%) Low1716201332213335565825192923 Medium48384040464229286764696352533639 High35464047223738372830262923283538 Household income quintiles (%) First159128188158125851282931 Second21182515262023199119131483530 Third24222122212120222021241823192018 Fourth2025232724252724283837362929913 Fifth202619281126152730262228223678 Employment status (%) Out of labor force741618124202135419551212 Unemployed8473432194841111 Employed85927779849378788891887794948787 Self-rated health3.43.53.33.53.43.53.53.53.53.63.43.53.73.73.73.7 Mean/SD1.21.21.21.21.01.01.00.90.90.90.80.91.11.11.21.1 Partner's education (%) Low14121916333329218105626139 Medium42404742362946437266646463536554 High44483442313925362024313035412237 Partner's employment status (%) Out of labor force252317925201672525138121285 Employed75778391758084937575879288889295 Region^c^ 133353834676861686258616115201520 222192121333239322114251235273528 3119119----------------346324382735 434373036----------------142482326152417Total *N*4912,2535272,7351781,0841631,2044202,9213963,348150774174953 %18821684148612881386118916841585*Source:* Own calculations with UKHLS, HILDA, SOEP, and GGS; data are weighted.^a^For Germany, we include satisfaction with family life because relationship satisfaction was not asked.^b^Parents separated during childhood.^c^Region in United Kingdom: 1 = Scotland, Ireland, and North England; 2 = Midlands and Wales; 3 = South West England; and 4 = South East England. In Germany: 1 = West Germany, 2 = East Germany stayed, 3 = East moved to West Germany, and 4 = born outside Germany. In Australia: 1 = urban, and 2 = rural. In Norway: 1 = Oslo area, 2 = East, 3 = South and West, and 4 = Mid and North.
OLS Regressions {#Sec22}
---------------
Table [4](#Tab4){ref-type="table"} presents coefficients for marriage relative to cohabitation for OLS regression models of SWB in midlife (full tables available on request). The unweighted column shows the unconditional association between marriage and mean SWB. The ATE column presents the average treatment effect after we apply weights, which can indicate the extent to which selection processes are biasing the results. The next two columns address our research questions about differential treatment bias: ATT refers to the average treatment effect on the treated (those in a partnership with a high propensity to marry), and ATC refers to the average treatment effect on the controls (those in a partnership with a low propensity to marry). The first row includes only controls for partnerships status and age. The second row includes all our control variables.[6](#Fn6){ref-type="fn"} The third row adds relationship satisfaction (satisfaction with family life in Germany), which is important to examine separately because of endogeneity.Table 4OLS weighted regression coefficients for the association between marriage and subjective well-being relative to cohabitation at midlife (ages 38--50)MenWomenUnweightedATEATT: High Propensity to MarryATC: Low Propensity to MarryUnweightedATEATT: High Propensity to MarryATC: Low Propensity to MarryUnited Kingdom (1) Married vs. cohabiting + age0.0860.0210.0030.1010.327\*\*0.289\*0.2350.283\*\*(0.105)(0.104)(0.105)(0.105)(0.120)(0.126)(0.128)(0.125) (2) + Childhood characteristics + partnership behavior + person's and partner's SES in current year^a^--0.055(0.114)--0.079(0.118)--0.093(0.121)--0.013(0.116)0.225(0.134)0.334\*(0.150)0.300(--0.173)0.405\*(0.145) (3) + Satisfaction with relationship--0.1360.069--0.182--0.1120.1320.2580.2240.279(0.112)(0.114)(0.117)(0.113)(0.129)(0.147)(0.149)(0.140) Number of observations3,5563,262Australia (1) Married vs. cohabiting + age0.351\*\*0.314\*0.310\*0.335\*\*0.245\*0.1810.1700.261\*(0.109)(0.131)(0.134)(0.120)(0.116)(0.127)(0.127)(0.132) (2) + Childhood characteristics + partnership behavior + person's and partner's SES in current year^a^0.211(0.110)0.193(0.126)0.193(0.129)0.199(0.120)0.214(0.116)0.156(0.120)0.147(0.121)0.232(0.123) (3) + Satisfaction with relationship0.0280.0180.0160.0360.056--0.001--0.0110.082(0.105)(0.123)(0.126)(0.114)(0.109)(0.108)(0.108)(0.114) Number of observations1,2621,367Germany (1) Married vs. cohabiting + age0.166\*0.1270.1110.234\*\*0.024-0.035--0.0440.037(0.085)(0.095)(0.100)(0.078)(0.090)(0.101)(0.103)(0.096) (2) + Childhood characteristics + partnership behavior + person's and partner's SES in current year^a^0.134(0.082)0.119(0.099)0.118(0.103)0.130(0.081)0.016(0.089)--0.032(0.091)0.027(0.091)--0.094(0.092) (3) + Satisfaction with relationship0.0860.1190.1240.0890.047--0.034--0.029--0.098(0.076)(0.093)(0.097)(0.074)(0.082)(0.085)(0.086)(0.082) Number of observations3,3413,744Norway (1) Married vs. cohabiting + age0.0660.0750.0730.0870.274\*0.336\*\*0.341\*\*0.314\*(0.114)(0.103)(0.104)(0.105)(0.110)(0.119)(0.121)(0.125) (2) + Childhood characteristics + partnership behavior + person's and partner's SES in current year^a^0.126(0.115)0.137(0.103)0.116(0.106)0.111(0.109)0.289\*(0.112)0.361\*\*(0.113)0.393\*\*\*(0.117)0.279\*(0.113) (3) + Satisfaction with relationship--0.054--0.068--0.074--0.0440.0540.1450.1670.026(0.103)(0.093)(0.092)(0.099)(0.100)(0.107)(0.111)(0.097) Number of observations9241,127*Source:* Own calculations with UKHLS, HILDA, SOEP, and GGP.^a^Childhood characteristics: region of origin, parent's nativity, parental separation during childhood, mother's and father's education, mother's and father's employment status, and father's occupational level. Partnership behavior: union duration, ever separated, and children within partnership. Respondent's socioeconomic background in current year: educational level, employment status, household income, and self-rated health. Partner's characteristics in current year: partner's education and partner's employment.\**p* \< .05; \*\**p* \< .01; \*\*\**p* \< .001
Table [4](#Tab4){ref-type="table"} shows substantial differences by country, gender, and propensity to marry. In the United Kingdom, marriage is not associated with higher SWB for men relative to cohabitation. However, marriage seems to be, on average, more beneficial for women (ATE) (*p* \< .05) and for partnered individuals with a low propensity to marry (ATC) (*p* \< .01). These findings suggest that women with a high propensity to marry are just as happy regardless of whether they marry. Women with a low propensity to marry, however, seem to be happier if they marry. Contrary to expectations, including our large battery of controls increases the magnitude of the coefficients, thus implying that marriage becomes even more important, especially for those with a low propensity to marry. After relationship satisfaction is included, however, statistical differences between marriage and cohabitation are eliminated. This result may imply that the quality of the relationship matters more than whether it is legally recognized, or that only women with high-quality relationships and suitable marriage partners marry.
For Australian men, marriage is associated with higher SWB when only partnership status and age are included in the models. The level of significance is slightly higher for the unweighted and ATC (*p* \< .01). Although all married people have higher levels of SWB than cohabitors, those with a lower propensity to marry would receive a slightly higher benefit if they married. However, when control variables are introduced, the coefficients in the weighted and unweighted models are no longer statistically significant, implying that selection processes matter in Australia. For Australian women, however, marriage provides benefits only in the unweighted and the ATC weighted regression models (*p* \< .05); those with a low propensity to marry would be happier if they were to marry. However, significant differences between married and cohabiting women again disappear after controls are included.
For German women, marriage is not significantly different from cohabitation, suggesting that marriage does not provide additional SWB benefits over living with a partner. However, we find that before weighting, German married men have higher levels of SWB; after weighting, those living in a partnership with a low propensity to marry have higher levels of SWB (*p* \< .01). After controls are introduced, though, the significant differences for men disappear, and the magnitude decreases. Even though the results show heterogeneity of treatment effects, they are surprising given our expectation of stronger differences between marriage and cohabitation in Germany.
Finally, for Norwegian men, as expected, marriage is not significantly different from cohabitation. For Norwegian women, however, all married women have higher levels of SWB compared with cohabiting women, regardless of the propensity to marry. Both those with a low propensity to marry and those with a high propensity to marry would experience benefits from marriage if they were to marry. Although control variables reduce the magnitude of the coefficients for those with a low propensity to marry (*p* \< .05), they increase the magnitude and level of significance for those with a high propensity to marry (*p* \< .001) and for the ATE (*p* \< .001). This result is quite surprising because the Norwegian legal and social context suggests that marriage provides few advantages to SWB, and we would not expect that those with a high propensity to marry need to marry to be happy. However, we find that Norwegian women would have higher SWB if they married. Nonetheless, our analyses control for only socioeconomic characteristics and self-rated health; unobserved factors, such as personality or other psychological factors closely related to SWB, may be more important for decisions to cohabit or marry. Thus, although we can say that a large range of background characteristics do not eliminate differences by partnership type, we cannot truly ascertain a causal relationship for Norwegian women. In addition, an individual's SWB may be dependent on the quality of the relationship with the partner. Including relationship satisfaction in the models eliminates differences between marriage and cohabitation, potentially implying that happy cohabiting relationships contribute just as much to SWB as high-quality marital relationships. On the other hand, Norwegian women may not be marrying because they have not found the right partner, which could be making them unhappy. Because relationship quality is measured only at the time of the survey, we cannot adjudicate between these explanations.
Discussion {#Sec23}
==========
Prior studies examining the effects of SWB on partnership status have found that individuals receive a minor boost in happiness after moving in with a partner and a larger boost after marriage, although these effects generally wear off as married partners return to their set-point happiness (Kalmijn [@CR32]; Musick and Bumpass [@CR53]; Soons et al. [@CR70]). What these studies cannot show is whether marriage is beneficial to those who are unlikely to marry, and the extent to which any marriage benefits are due to characteristics that select people into marriage. Prior studies have also not specifically focused on the effects of marriage relative to cohabitation in midlife, after most people have married, and the initial honeymoon period of marriage is over. Our study produced some surprising findings that indicate not only differences by country and gender but also differences by the propensity to marry.
First, contrary to prior studies (e.g., Ono and Lee 2012; Soons and Kalmijn [@CR69]), our results indicate that relative to cohabitation, marriage does not automatically provide a boost to SWB in all countries. On average, cohabiting men in the United Kingdom and Norway and women in Germany have levels of SWB that are similar to those of married men in midlife, even without controls. These findings suggest that in some countries, cohabitation may provide benefits similar to those of marriage, such as shared intimacy, pooled resources, and emotional support (Musick and Bumpass [@CR53]; Perelli-Harris and Styrc [@CR63]).
Second, our results show that, on average, marriage does differ from cohabitation for Australian men and Norwegian women. Without any controls for selection, married individuals in these countries have higher levels of SWB than those in cohabiting partnerships. In Australia, these differences disappear after we include controls, indicating that cohabitation is selective of disadvantage, in accordance with prior studies (Evans [@CR19]; Heard [@CR24]). For Norwegian women, however, our entire battery of controls cannot eliminate average differences between cohabitation and marriage. This finding is quite surprising because some have argued that cohabitation and marriage in Norway, and Scandinavian countries in general, are indistinguishable (Heuveline and Timberlake [@CR26]). Yet here, we see persistent differences between the two partnership types. Our models, however, take into account important sociodemographic status and childhood background variables but may be missing key psychological characteristics or other attributes that are associated with marriage. We cannot rule out the possibility that people with certain personality traits or preferences are more likely to marry. Thus, we are reluctant to interpret our results as having a causal effect.
Nonetheless, the findings suggest that marriage is more important in Norway than often assumed. Focus group research found that marriage is associated with romance and love, even if it occurs sometime after childbearing (Lappegård and Noack [@CR38]), or as a capstone later in life (Holland [@CR28]). Although cohabitation may seem to be identical to marriage superficially, marriage may be indicative of a closer partnership on a deeper level. Marriage for women may be symbolic of a more committed loving relationship, and if marriage does not happen by midlife, the lack of marriage might have detrimental effects on SWB. The elimination of marriage effects when we include relationship satisfaction suggests this may be the case. On the one hand, the results may indicate that relationship quality is more important than type of partnership, and cohabiting and married women have similar SWB. On the other hand, relationship satisfaction may instead be a proxy for marriage given that happier couples are generally more likely to marry (Wiik et al. [@CR80]). Then again, the marital contract may in fact improve relationship quality for women. Thus, although we urge caution in interpreting this result one way or the other, it seems to be plausible that marriage, on average, has positive effects for women in Norway.
Third, our results demonstrate the heterogeneity of treatment effects for German men and British and Australian women. Partnered individuals who have a lower propensity to marry based on childhood selection mechanisms (ATC) have lower SWB if they cohabit rather than marry. Those who have a high propensity to marry (ATT), on the other hand, would not receive any benefits from marriage. For German men and Australian women, controlling for selection mechanisms and partnership experiences eliminates differences between cohabitation and marriage, again demonstrating that selection is more consequential for SWB than partnership status.
For British women, however, we find the persistent effect of marriage on SWB for those who are less likely to marry, despite the large number of controls. These results suggest that marriage may provide some benefits for disadvantaged women, as was also found in a study on mental well-being (Perelli-Harris and Styrc [@CR63]). Our findings here, however, indicate that women from disadvantaged backgrounds would be better off if they did marry, but not because they or their partners have low education, poor employment conditions, or low income, which would make them unhappy. Instead, marriage seems to be associated with happiness for other reasons. The findings could be due to unobserved selection mechanisms related to personality, appearance, or other psychological factors that make them less-attractive marriage partners. On the other hand, they may be unhappy because they do not want to marry partners who do not live up to their expectations, or they may disagree about whether to get married, which could have a greater effect on women than men. The mediating effect of relationship quality suggests this may be the case; those who have higher-quality relationships marry and have higher levels of SWB. Qualitative research provides a deeper explanation for this finding: focus group participants from all educational levels in Britain generally agreed that marriage signaled a more committed relationship than cohabitation, but low-educated women stated that although they would like to marry, cohabitation was more common among their peers. For these women, marriage was a low priority compared with other responsibilities such as housing and children, but they nonetheless aspired to have a wedding (Berrington et al. [@CR8]), and perhaps they were unhappier because they could not achieve their goals.
Our study is not without limitations. First, as mentioned, life satisfaction may be endogenous to partnership decisions: happier people may be more likely to marry than cohabit (Luhmann et al. [@CR44]). Although we focus on controlling for selection mechanisms in childhood, before individuals enter into partnerships, our data do not include a direct measure of happiness in childhood or around the time of entrance into partnership. Propensity score--weighted regression is also unable to control for unobserved factors not available in our surveys. Marriage may be selective of other individual characteristics such as personality, emotional control, or attractiveness. This is particularly important for Norwegian and British women, where differences between cohabitation and marriage persist until relationship satisfaction is included in the models. Second, despite our concerted effort to harmonize the variables across our surveys, differences in survey design and variable construction may limit comparability across surveys. Finally, our definition of midlife (ages 38--50) is relatively narrow. But because cohabitation has increased only within the past few decades, the sample size for cohabiting individuals in the older cohorts is too small. Therefore, future research must continue to evaluate these relationships as cohabitation increases throughout midlife.
Despite these limitations, this study provides evidence that the relationship between partnership and SWB is not straightforward and is context-specific. The context, especially the policy context, does not always operate in predictable ways. For example, given that the German government privileges the marital breadwinner model, we would have predicted that married women in Germany would be happier than cohabiting women; however, we find no differences between marriage and cohabitation. This finding suggests that despite policies to encourage marriage, cohabitation in Germany may not be stigmatized; indeed, Treas et al. ([@CR74]) found that Germans have a more positive view of living in cohabitation without marriage intentions than those in Great Britain and Australia. Thus, the policy climate may not shape the association between partnership status and SWB as much as changing values and the specific meaning of cohabitation. This also holds true for Norwegian women, who (as discussed earlier) appear happier if they are married, despite a gender-friendly policy regime that stipulates few legal distinctions between cohabitation and marriage (Lappegård and Noack [@CR38]). Regardless of an increasing number of children born in cohabitation and a lack of stigmatization toward cohabitors, Norwegian women still seem to value marriage and the symbolic implication of the wedding. Research on individual countries needs to recognize that context may be shaping the relationship between factors.
Finally, this study not only demonstrates the role of selection in accounting for the association between marriage and SWB but also illustrates how selection is heterogeneous and differs according to the propensity to marry. Such findings can have important policy implications given that those with a low propensity to marry---namely, the disadvantaged---are likely to be targeted by pro-marriage policymakers. At first glance, our findings seem to suggest that in some countries, those least likely to marry would benefit from marriage-promotion policies: if they were to marry, they would be happier. However, for German men and Australian women, the effect of marriage disappears after the inclusion of more controls. For disadvantaged women in the United Kingdom, marriage may matter, especially if it provides legal protection and a sense of security (Barlow [@CR6]; Berrington et al. [@CR8]). The effect in the United Kingdom disappears after a control for relationship satisfaction is included, implying that policymakers could focus on improving relationship quality, possibly through relationship support organizations that provide counseling, although it is also important to recognize that these women may be unhappy because they are unable to find a suitable partner. On the whole, however, our study indicates that especially after selection and relationship satisfaction are taken into account, differences between marriage and cohabitation disappear in all countries. Marriage does not cause higher SWB; instead, cohabitation is a symptom of economic and emotional strain. Thus, our findings imply that in order to increase SWB, policymakers should aim to reduce disadvantages---both in childhood and adulthood---instead of creating incentives to marry.
Electronic supplementary material
=================================
{#Sec24}
ESM 1(PDF 363 kb)
"Happiness" and "life satisfaction" are often used interchangeably. *Life satisfaction* is typically defined as a process of assessing individuals' perceived quality of life based on current circumstances (Diener et al. [@CR15]). Here we use the term *subjective well-being* because it best reflects satisfaction with overall life.
We define *midlife* as aged 38--50. Older cohorts are less likely to be in a cohabiting partnership, and our sample becomes too small to analyze.
United Kingdom: "Please choose the number which you feel best describes how dissatisfied or satisfied you are with the following aspects of your current situation: Your overall life." Australia and Germany: "All things considered, how satisfied are you with your life? Again, pick a number between 0 and 10 to indicate how satisfied you are." Norway: "On a scale from 0 to 10, where 0 means 'Not satisfied at all' and 10 means 'completely satisfied,' how satisfied are you, on the whole, with your current life?"
A quadratic specification of union duration does not change the main results.
A respondent who has a child with a previous partner and the current partner are classified as having a child with the current partner.
We also ran models (available on request) that follow our theoretical framework and include each set of control variables separately: (1) childhood characteristics; (2) the respondent's partnership and fertility behavior; and (3) socioeconomic background of the respondent and partner in the current year, which could be endogenous. These models did provide some nuances: for example, men's own SES was particularly important for eliminating differences between cohabitation and marriage in Germany and Australia. Overall, however, the more detailed models do not change our main story. Therefore, we show tables with all control variables included at once.
**Publisher's Note**
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This research has received funding from European Research Council Grant 263794 CHILDCOHAB, the UK Economic and Social Research Council Centre for Population Change (Grant RES-625-28-0001), Australian Research Council Discovery Project funding (DP110104439), and Research Council of Norway grant (236926/H20). Understanding Society is an initiative funded by the Economic and Social Research Council and various government departments, with scientific leadership by the Institute for Social and Economic Research, University of Essex, and survey delivery by NatCen Social Research and Kantar Public. The research data are distributed by the UK Data Service. The HILDA Project was initiated and is funded by the Australian Government Department of Social Services (DSS) and is managed by the Melbourne Institute of Applied Economic and Social Research (Melbourne Institute). The German Socio-Economic Panel Study (SOEP) was distributed by the German Institute for Economic Research (DIW), Berlin.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-molecules-23-00890}
===============
As an important predecessor of submicron particles (PM2.5) and ozone pollution in the atmosphere, volatile organic compounds (VOCs) emitted from industrial manufacturing have received increasing attention in recent years \[[@B1-molecules-23-00890],[@B2-molecules-23-00890]\]. Among the VOCs, the benzene series (BTEX) and acetone have always been simultaneously utilized in paints, solvents, and raw materials in the chemical and printing industry \[[@B3-molecules-23-00890],[@B4-molecules-23-00890],[@B5-molecules-23-00890]\]. Exposure usually causes a number of environment-related health problems, including dizziness, nausea, organ damage, and even cancer \[[@B6-molecules-23-00890]\]. Thus, developing suitable abatement methods for BTEX/acetone emission control is urgent and significant. Conventional technologies for VOC removal include adsorption, thermal combustion, chemical absorption, and catalytic oxidation. However, such methods are not sufficiently cost-effective or suitable for removal of dilute concentrations (\<1000 ppm) of contaminants under high space velocity because of low efficiencies and high energy consumption \[[@B7-molecules-23-00890]\]
In recent years, non-thermal plasma (NTP) has been regarded as an energy-saving, efficient, and promising method for low-concentration VOC abatement due to its environmentally-friendly nature, fast ignition response, and strong oxidative degradation ability \[[@B8-molecules-23-00890],[@B9-molecules-23-00890]\]. At room temperature, quantities of highly energetic electrons and reactive species generated in the discharge area trigger a cascade of plasma chemistry reactions, resulting in the removal of pollutants \[[@B10-molecules-23-00890],[@B11-molecules-23-00890],[@B12-molecules-23-00890]\]. Several studies on BTEX removal, side product analysis, and degradation mechanisms by NTP have been reported over the past few years. Satoh et al. explored the effect of O~2~ proportion in carrier gas on the removal of benzene by different manners of discharge. The results shows that at low oxygen concentration, the byproducts are primarily C~2~H~2~, HCN, NO, and HCOOH, while only HCOOH is found at high oxygen concentrations \[[@B13-molecules-23-00890]\]. Stefan et al. investigated the degradation of cyclohexene and BTEX in an NTP air purifying system, and the degradation efficiency order of benzene (\<1%) \< xylene (3%) ≈ ethylbenzene \< toluene (11%) ≈ cyclohexene was found \[[@B14-molecules-23-00890]\]. Our previous research indicates that the conversion of low-concentration benzene, toluene, and *p*-xylene increases from 2%, 19%, and 49%, respectively, at an energy density (ED) of 10 J·L^−1^ under positive corona discharge for BTEX \[[@B15-molecules-23-00890]\]. Unlike BTEX, there are few reports focused on acetone degradation by NTP. In research, monolithic ceramic catalysts and CuO/γ-Al~2~O~3~ are often added into the plasma reactor or after the reaction to intensify the decomposition of acetone plasma \[[@B16-molecules-23-00890],[@B17-molecules-23-00890]\]. With respect to byproduct investigation, Narengerile et al. evaluated the acetone decomposition efficiency in DC water plasmas at atmospheric pressure. It was found that aqueous acetone can be successfully decomposed into H~2~, CO~2~, CO, and CH~4~, but unwanted byproducts, such as HCOOH and HCHO, also form \[[@B18-molecules-23-00890]\].
However, some limitations on the study of acetone and BTEX removal by NTP still exist as follows. Firstly, acetone, as a representative of oxygen-containing VOCs, is hardly decomposed, but has the highest emission limit (100 mg·m^−3^) among VOCs from petrochemical industry emissions \[[@B19-molecules-23-00890]\]. However, few studies have focused on its removal effectiveness, especially at low concentrations. Secondly, in general typical industrial emissions contain a blend of VOCs \[[@B20-molecules-23-00890],[@B21-molecules-23-00890]\]. However, research often focuses on single-component VOCs rather than mixtures of VOCs, which is not in accordance with real emission conditions. Thirdly, though BTEX concentrations are far below those of acetone in real life, their compounds are characteristic constituents of the gaseous effluents of wastewater treatments in petrochemical plants \[[@B19-molecules-23-00890]\]. Whether there is an interaction between acetone and BTEX when treated together remains unknown.
Accordingly, this study focuses on a mixture of acetone and BTEX degradation using corona discharge that aims at investigating possible influencing mechanisms in removal efficiency and COx selectivity under NTP treatment. In addition, the impact of BTEX types on NO~x~, O~3~, organic byproduct formation, and the acetone degradation pathway are also studied by experimental and theory calculations in this paper.
2. Experimental {#sec2-molecules-23-00890}
===============
2.1. Experimental System {#sec2dot1-molecules-23-00890}
------------------------
The experimental system is shown in [Figure 1](#molecules-23-00890-f001){ref-type="fig"}. It consists of a coaxial link tooth wheel-cylinder plasma reactor with a 25 kV/5 mA negative direct current (DC) high voltage power supply, reaction gas supply, and analytical instrumentation. It is a stainless steel cylinder with an inner diameter of 42 mm and a length of 300 mm that serves as the ground electrode of the plasma reactor. The high voltage electrode is a stainless steel rod (o.d. 6 mm) through which 10 discharge teeth wheels are linked with a space interval of 10 mm, while each wheel has six discharge cusps. The effective discharge length and discharge intervals are 100 and 16 mm, respectively. The visual appearance of the discharge is that of a gleamy plasma column completely filling the inter-electrode space, representing a streamer-like corona discharge.
2.2. Experimental Methods {#sec2dot2-molecules-23-00890}
-------------------------
Experiments with both the single acetone and acetone with BTEX (benzene, toluene, or *p*-xylene) were conducted in this study. Gaseous VOCs and water vapor were introduced by passing air through a temperature-controlled bubble tower and they were then mixed with dilution air in a mixing chamber to reach the desired concentration. Relative humidity (RH) of the reaction gas was controlled at 50% at room temperature (298 K). The concentrations of benzene, toluene and *p*-xylene were all 50 ppm, while that of acetone was 250 ppm. The total flow rate was 2.0 L·min^−1^. All the reagents used in this study were analytically pure, obtained from the Beijing Chemical Corporation (Beijing, China).
The outlet concentrations of the VOCs, CO~x~ (CO and CO~2~), O~3~, NO~x~ (NO and NO~2~) were respectively detected. The VOCs in the gas stream were analyzed by an online gas chromatograph (Agilent, model 6890N, Santa Clara, CA, USA), equipped with a flame ionization detector (FID) and a 30.0 m × 320 μm HP-5 capillary column. The column temperature was 373 K and that of the detector was 423 K. The conversion of each VOC by non-thermal plasma (NTP) decomposition was defined by η, calculated according to Equation (1). The energy density (ED, J·L^−1^) was used to evaluate the validity of NTP technology, which was calculated according to Equation (2). In the present work, all the decomposition results were compared and discussed based on the ED. The outlet concentrations of CO and CO~2~ were analyzed by a gas chromatograph (Techcomp, model GC7890II, Beijing, China) equipped with an FID detector, a TDX-01 packed column, and a methane conversion oven prior to the detector. The CO~x~ selectivity was adopted to characterize the mineralization degrees of the VOCs in the present work, defined by SCO~x~ according to Equation (3). O~3~ was monitored by the Ozone Monitor (2B Technologies, 106-L, Boulder, CO, USA) according to the ultraviolet absorption method, while NO~x~ was monitored according to the *N*-(1-naphthyl) ethylene diamine dihydrochloride spectrophotometric method. $$\eta = \frac{C_{inlet} - C_{outlet}}{C_{inlet}} \times 100\text{\%},$$ $$ED = \frac{U \times I}{Q} \times 60$$ $${~S}_{{CO}_{x}} = \frac{\left( {2 - x} \right)C_{CO} + \left( {x - 1} \right)C_{{CO}_{2}}}{\sum n_{inlet}C_{inlet}^{\prime}\eta} \times 100\text{\%}$$ where *C~inlet~* and *C~outlet~* are inlet and outlet concentrations of the pollutant (ppm), respectively; *U* and *I* are the applied voltage (kV) and discharge current (mA), respectively, both of which can be automatically detected by the power supply equipment, checked before experiments; and *Q* is the flow rate of the reaction gas (L·min^−1^). C′~inlet~ is the inlet concentration of the pollutant (mg·m^−3^); CCO~2~ and CCO are the outlet concentrations of CO~2~ and CO (mg·m^−3^); and ninlet indicates the number of carbon atoms of the VOC.
2.3. Characterization {#sec2dot3-molecules-23-00890}
---------------------
In order to clarify the mechanism of VOC degradation by NTP, the variations in the intermediate products were detected by a mass spectrometer (MS OmnistarTM, model GSD 301 O2, Preddvor, Slovenia) and Fourier transform infrared spectroscopy (FTIR, Nicolet 6700, Santa Monica, CA, USA). The mass spectrometer was used to explore the change of intermediate species during the reaction process. An electron multiplier and a tungsten filament were used to detect the specific charge, while the SEM voltage was controlled at 1.4 kV. The first mass was 44.5 u and the last mass was 100 u, except for *p*-xylene (110 u). The scan speed was 5 s. FTIR was used to investigate the change of the gas molecule functional group during the reaction process; all spectra were collected in the 3700--600 cm^−1^ frequency range at a resolution of 4 cm^−1^. In total, 16 scans were averaged for each spectrum, while the gathering time was 20 min.
The organic side products were collected using 5 mL methyl alcohol for 15 min, and then identified using gas chromatography--mass spectrometry (GC--MS, Shimadzu GC-2020, Kyoto, Japan) in electron impact mode (70 eV), equipped with a 60.0 m × 0.25 mm × 0.25 μm column (Agilent, HP5MS, Santa Clara, CA, USA), A 200 °C ion source was chosen. Argon was chose as the carrier gas and the flow rate was 1 mL·min^−1^.
The theory calculation was based on the Gaussian 09 package program combined with density functional theory (DFT). The molecular geometry of the VOCs was optimized using the DFT (B3LYP) method with a 6-311G++(d,p) basis set. The atomic partial charge, bond length, and electrostatic potential of HOMO orbits of the VOCs were calculated, and the corresponding results are listed in [Table S1](#app1-molecules-23-00890){ref-type="app"}.
3. Results and Discussion {#sec3-molecules-23-00890}
=========================
3.1. Effect of BTEX on the Acetone Removal Efficiency {#sec3dot1-molecules-23-00890}
-----------------------------------------------------
Acetone and BTEX conversions are studied as a function of energy density under the conditions of individual contaminants and two VOC mixtures. As shown in [Figure 2](#molecules-23-00890-f002){ref-type="fig"}a, when acetone is treated individually, its conversion increases from 0% to 71% with increasing ED. This result is similar to that of a previous study by Zheng et al. \[[@B22-molecules-23-00890]\]. However, when acetone is degraded together with three kinds of BTEX under plasma treatment, its degradation efficiency reduces to less than 20% with an energy density of 1200 J·L^−1^, implying that BTEX has a significant negative impact on acetone conversion. The degree of influence of three different kinds of BTEX on acetone removal efficiency can be ordered as follows: toluene ≈ *p*-xylene \> benzene.
However, it can be seen from [Figure 2](#molecules-23-00890-f002){ref-type="fig"}b that benzene conversion is hardly affected when it coexists with acetone, while the removal efficiencies of toluene and *p*-xylene decrease with the introduction of acetone. In addition, it is found that whether treated together with acetone or not, *p*-xylene conversion is much higher than that of toluene under the same ED. It is known that high-energy electrons ranging from 1011 to 1014 cm^−3^ contribute to VOC decomposition during the NTP process by two means: (1) directly breaking VOC molecules into organic fragments via high-speed collision; and (2) interaction with carrier gas molecules to generate large quantities of chemical reactive species (e.g., •O, •OH, O~3~, and metastable N~2~) for VOC removal \[[@B23-molecules-23-00890]\]. On one hand, the original concentration of BTEX (50 ppm) in this study is far lower than that of acetone (250 ppm), and thus the probability of effective electron collision with acetone molecules may not change significantly after BTEX introduction. In other words, the types and concentration levels of active radicals in the NTP reactor should play a dominant role in contributing to BTEX inhibiting acetone degradation. On the other hand, the structural difference between the three kinds of VOCs is shown in the methyl group and its quantity in the benzene. Considering the efficiency difference shown in [Figure 2](#molecules-23-00890-f002){ref-type="fig"}, besides the benzene ring effect, it can be deduced that the hydrogen abstraction reactions on the methyl group may occur during BTEX degradation, also influencing the acetone plasma--chemical reactions, removal efficiency, and the degradation product.
3.2. Effect of the BTEX on CO~x~ Selectivity {#sec3dot2-molecules-23-00890}
--------------------------------------------
[Figure 3](#molecules-23-00890-f003){ref-type="fig"} shows the selectivity of CO~x~ as function of ED for the NTP process. It can be seen that in the case of BTEX and acetone together, the CO~x~ yield shows two stark divergences of trends: addition of *p*-xylene significantly promotes CO~x~ selectivity, thereby improving acetone mineralization; while CO~x~ production decreases after adding benzene or toluene into the NTP reaction system, meaning that more organic intermediates form. [Figures S1 and S2](#app1-molecules-23-00890){ref-type="app"} also provide the CO and CO~2~ selectivity under the binary and single-component VOC degradation processes. It is shown that more CO is generated when methylic BTEX and acetone are treated together, but the tendency is reversed for benzene introduction. By contrast, the selectivity of CO~2~ in the binary VOC degradation process is obviously lower than in single-component VOCs. Because more active free radicals are consumed when BTEX is treated together with acetone, the rate of CO~x~ mineralization and further oxidation of CO to CO~2~ is restricted. Equally important is that the hydrogen abstraction reaction on the methyl group occurred more easily with toluene or *p*-xylene degradation, which can lead to CO formation despite an adversity to the total oxidation of pollutants \[[@B24-molecules-23-00890],[@B25-molecules-23-00890]\].
3.3. NO~2~ and O~3~ Formation Analysis {#sec3dot3-molecules-23-00890}
--------------------------------------
[Figure 4](#molecules-23-00890-f004){ref-type="fig"} and [Figure 5](#molecules-23-00890-f005){ref-type="fig"} provide the relationships between NO~2~, O~3~ concentrations and ED. As shown in the two figures, NO~2~ and O~3~ production monotonically increases with the increase of ED under both single and binary VOC mixture sets. Notably, the maximum NO~2~ concentration is less than 1 ppm, and NO is hardly detected in this study, so the effect of BTEX adding on NO~x~ formation may be negligible. It believed that O~3~ concentration is greater in the absence of VOCs at a relatively high ED (\>200 J·L^−1^) considering ozone consumption by the oxidation of VOCs \[[@B26-molecules-23-00890],[@B27-molecules-23-00890]\]. Therefore, even though VOC mixture degradation under the NTP process follows complex mechanisms, there may be some correlations between the ozone concentrations and pollutant treatment efficiency. [Figure 6](#molecules-23-00890-f006){ref-type="fig"} gives the effect of O~3~ concentration on VOC concentrations. It can be seen that each BTEX conversion enhances significantly with increasing ED when degraded simultaneously with acetone, while acetone follows an inverse pattern. This phenomenon can be explained by the fact that BTEX oxidizes faster than acetone, since the former has a higher reaction rate constant with O~3~ and •OH \[[@B28-molecules-23-00890]\]. Furthermore, there is significant positive linear correlation between O~3~ concentration and *p*-xylene conversion, while it is poor for other VOCs. Combined with the lowest yield of O~3~ for xylene-containing atmosphere degradation ([Figure 5](#molecules-23-00890-f005){ref-type="fig"}), it is reasonable to infer that O~3~ or its precursor from the plasma--chemical reactions can interact with the methyl group from the *p*-xylene molecule.
3.4. Transient Study during NTP Treatment {#sec3dot4-molecules-23-00890}
-----------------------------------------
[Figure 7](#molecules-23-00890-f007){ref-type="fig"} and [Table S1](#app1-molecules-23-00890){ref-type="app"} show the detailed in situ FT-IR spectra of the acetone and acetone--BTEX mixture effluents under plasma treatment at an ED of 1600 J·L^−1^. When acetone is degraded alone, as the reaction time extends the bands at 1027 and 1054 cm^−1^, ascribed to the O-O stretching vibrations of the ozone, gradually increase. Besides, bands appear at 2126, 2096, 1731, 1441, 1373, 1343, 1302, 1210, 1161, 1121, 1093, and 885 cm^−1^ during the NTP process. Among them, the acetone species are characterized by the ν (C=O) at 1731 cm^−1^, δ (CH~3~) at 1441--1371 cm^−1^ and δ (C-C) at 1161 cm^−1^ \[[@B29-molecules-23-00890],[@B30-molecules-23-00890]\]. As the reaction time continues, the intensity of the bands attributed to olefins (885 cm^−1^), formic acid (1093 and 1121 cm^−1^), oxalic acid (1302 cm^−1^), alcohol (1343 cm^−1^), and CO (2126 and 2096 cm^−1^) continuously increases on the spectra \[[@B16-molecules-23-00890],[@B31-molecules-23-00890],[@B32-molecules-23-00890]\]. Narengerile's research results indicate that HCOOH, HCHO, and soot are the main organic byproducts when acetone is decomposed by direct current (DC) plasma \[[@B18-molecules-23-00890]\]. Therefore, the bands belong to the ν (C=O) stretching vibration of the aldehydes, which is usually located at around 1700 cm^−1^, and may be covered by the characteristic peak of acetone at 1730 cm^−1^.
For the two composite VOC atmospheres, the FTIR spectra in the range of 1800--1000 cm^−1^ are significantly different from those of the acetone system. The bands at 821 cm^−1^, 1475 cm^−1^, 1516--1526 cm^−1^, and 1660--1684 cm^−1^ represent the characteristic vibrational peaks of the benzene ring. The byproducts related to BTEX degradation consist of the phenolic hydroxyl (OH, 1252--1258 cm^−1^), aldehyde (CHO, 1759 cm^−1^), and carboxyl (COOH, 1787 cm^−1^) group, indicating that phenol, benzaldehyde, and benzoic acid may form under NTP treatment \[[@B33-molecules-23-00890]\]. In the case of the acetone--xylene composite, the signal-to-noise ratio is significantly lower than for the acetone--benzene and acetone--toluene systems, meaning *p*-xylene becomes extremely unstable under the discharge condition. Additionally, the characteristic peaks of *p*-xylene at 1516 cm^−1^ disappear rapidly after treatment with NTP, suggesting its reaction rate is faster than for benzene and toluene. As shown in the [Figure 7](#molecules-23-00890-f007){ref-type="fig"}, production of O~3~ is greater as the reaction progresses, while the reverse result is found when benzene or toluene is treated alone ([Figure S3](#app1-molecules-23-00890){ref-type="app"}). In general, ozone generated by NTP can be partly consumed by the oxidation of VOCs, but this is suppressed when acetone and BTEX coexist \[[@B27-molecules-23-00890]\]. One possible explanation is that a harder degradation of acetone inhibits the reaction between BTEX and ozone, thereby augmenting O~3~ concentrations.
[Figure 8](#molecules-23-00890-f008){ref-type="fig"} shows the in situ MS spectrum of VOCs degradation at various discharge times. It is obvious that the production of *m*/*z* 45, 46, 58, 78, 91, and 106 occurs during NTP treatment. Among these, *m*/*z* 58, 78, 91, and 106 can be mostly attributed to acetone, benzene, toluene, and *p*-xylene, while *m*/*z* 45 and 46 can be attributed to the oxalic acid or fragment of carboxyl and formic acid, respectively. From this finding, together with the FTIR results, it can be speculated that a hydrogen abstraction from the benzene ring leads to a phenyl radical formation, which would react with •OH to form benzaldehyde and phenol or be further oxidized to benzoic acid. According to previous research, the hydrogen abstraction reaction from the methyl group will inevitably occur when toluene and acetone are decomposed by NTP and sequentially a reaction occurs with the excited •OH to form HCOOH, HCHO, and Cox \[[@B16-molecules-23-00890],[@B25-molecules-23-00890]\]. It is reasonable to infer the proportion of formic acid in the carboxylic acid byproduct by the peak area ratio of *m*/*z* 46 to 45. When BTEX and acetone are treated together, the proportion of formic acid is clearly reduced compared with acetone alone, indicating that the acetone decomposition pathway is influenced by BTEX. Moreover, [Figure 8](#molecules-23-00890-f008){ref-type="fig"} also gives the acetone degradation rate under different atmospheric conditions. It is clear that the benzene and toluene slightly accelerate the acetone removal but *p*-xylene lessens it. As is well known, benzene can be directly oxidized to various ring-opening byproducts under NTP, while for toluene or xylene, substituent groups on the benzene are more susceptible to decomposition, and benzene series byproduct generation occurs easily \[[@B15-molecules-23-00890],[@B34-molecules-23-00890]\]. Therefore, more methyl group fragments derived from *p*-xylene or toluene decomposition may inhibit the acetone removal, which can also be decomposed to the methyl group and other products under NTP.
3.5. Organic Product Analysis by GC-MS {#sec3dot5-molecules-23-00890}
--------------------------------------
The organic byproducts of the effluent gas at ED of 1600 J·L^−1^ for single and binary VOCs systems are collected by GC-MS, as shown in [Figure 9](#molecules-23-00890-f009){ref-type="fig"}. The main byproducts during acetone degradation are annular and long-chain oxy-organics containing aldehyde, ethers, esters, etc. The number of organic byproducts, including both aromatic and ring-opening byproducts generated for benzene removal, is significantly greater than for toluene and *p*-xylene. Furthermore, it is seen that ring-opening byproducts show a significant decrease with the amount of methyl groups on the benzene ring. The major byproducts in the effluent gas of *p-*xylene treatment are benzene, toluene, 4-methyl benzaldehyde, 4-methylbenzoic acid, and benzyl methyl ether. When acetone and BTEX are treated together, the number of light ring-opening byproducts like dimethyl oxalate, methyl dimethoxyacetate, and dimethyl maleate is reduced, while the number of byproducts with more substituent groups on the benzene ring such as o-tolualdehyde, o-nitrophenol, and *p*-toluic acid rises. This phenomenon implies that instead of BTEX ring-opening, methylation and electrophilic substitution reactions in the benzene ring are favored under *co*-treatment of acetone and BTEX by NTP. Therefore, it can be concluded that the methyl group on acetone, toluene, and *p*-xylene decomposing into methyl fragments or radicals may have an immediate impact on the byproduct formation through a series of plasma chemical reactions.
3.6. Theoretical Analysis of Intermediates {#sec3dot6-molecules-23-00890}
------------------------------------------
In order to explore the mechanisms of the influence of BTEX on acetone decomposition, acetone and BTEX molecules underwent geometry optimization using the Gaussian 09 package program. Since VOCs are attacked by highly active oxygen species like •OH and •O through the electrophilic reactions, the attack position of active radicals on pollutants should have the greatest electron density in the HOMO orbital of the molecule, in accordance with the "frontier orbital" theory \[[@B35-molecules-23-00890],[@B36-molecules-23-00890]\]. Therefore, the bond length, atomic charge, and electrostatic surface potential of total density and HOMO orbitals of the VOCs molecules are also calculated by the density functional theory, as shown in [Figure S4 and Table S1](#app1-molecules-23-00890){ref-type="app"}. As seen in [Figure S3](#app1-molecules-23-00890){ref-type="app"}, the positive electrostatic potential of acetone molecule is mainly concentrated in the region around the C1 and C3 atoms, whose atomic charge is −0.763 e on has the strongest electronegativity. Meanwhile, the C1--C2 bonds have the longest bond length (1.517 Å), so the cleavage of the methyl group would be the easiest step during the acetone decomposition process.
According to the ESP of HOMO orbitals for BTEX, it is found that introducing electron-donating methyl groups enhances the electron density of the benzene ring, facilitating attack by free radicals. Therefore, the substituent group position on the benzene ring would be easily occupied by electrophilic •OH radicals, resulting in further aromatic byproduct formation. On the other hand, as listed in [Table S1](#app1-molecules-23-00890){ref-type="app"}, the carbon--carbon bond length on the benzene ring is shorter than that between the benzene ring and the substituent group. Thus, *p*-xylene with more methyl groups is more easily substituted through electrophilic reactions than other types of BTEX. Considering the higher electron density on the benzene ring, greater energy should be provided for ring-opening, thereby inhibiting the non-aromatic hydrocarbon byproducts, which may be a reasonable explanation for the GC-MS result. Consequently, after the --CH~3~ on toluene and *p*-xylene is removed from the aromatic ring, intermediates such as phenol, benzaldehyde, and benzoic acid would be generated. At the same time, the falling methyl fragment would have a negative impact on acetone decomposition because it may lead to broken acetone regeneration.
3.7. BTEX Effect on Acetone Degradation Pathway {#sec3dot7-molecules-23-00890}
-----------------------------------------------
The decomposition mechanism of acetone under various plasma treatment conditions has been reported and discussed by previous researchers \[[@B16-molecules-23-00890],[@B17-molecules-23-00890],[@B18-molecules-23-00890],[@B30-molecules-23-00890]\]. The relative reaction rate constants involved are summarized in [Table S2](#app1-molecules-23-00890){ref-type="app"}. When humidity is introduced into the reactor, the main reactive species are oxygen and hydroxyl radicals because O~3~ can act as a source of their formation, in accordance with Reactions (4)--(7) \[[@B37-molecules-23-00890]\]:$$\left. N_{2}~ + ~O_{2}~ + hv\rightarrow~N_{2} + ~2O, \right.$$ $$\left. H_{2}O~ + ~hv\rightarrow~H + ~OH \right.$$ $$\left. O_{3}~ + ~hv\rightarrow~O_{2} + ~O \right.$$ $$\left. H_{2}O~ + ~O~\rightarrow~2OH \right.$$
High-energy electrons •OH and N~2~\* directly dissociate acetone molecules into the fragment of •CH~3~ and CH~3~CO• via Reaction (8). Subsequently, the recombination of •CH~3~ and the oxygen species via several steps (Reactions (9)--(13)) can form CH~3~O~2~•, CH~3~O• and HCHO. The latter is further oxidized to HCO• by •OH or divided into CO and H (Reactions (14) and (15)) \[[@B38-molecules-23-00890]\]. Additionally, HCO• is also the precursor of the CO and CO~2~, according to Reactions (16--19). Besides, a trace of NO~2~ can be detected, which may follow Reactions (20)--(23) \[[@B39-molecules-23-00890]\].
C
H
3
C
O
C
H
3
→
C
H
3
\+
C
H
3
C
O
,
C
H
3
C
O
\+
O
→
C
H
3
\+
C
O
2
C
H
3
\+
O
2
→
C
H
3
O
2
C
H
3
\+
O
3
→
C
H
3
O
\+
O
2
C
H
3
\+
O
→
H
C
H
O
\+
H
C
H
3
O
\+
O
2
→
H
C
H
O
\+
O
H
H
C
H
O
\+
O
H
→
H
C
O
\+
H
2
O
H
C
H
O
→
2
H
\+
C
O
H
C
O
→
C
O
\+
2
H
H
C
O
\+
O
→
C
O
\+
O
H
C
O
\+
O
→
C
O
2
C
O
\+
O
H
→
C
O
2
\+
H
O
2
N
2
→
2
N
N
\+
O
3
→
N
O
\+
O
C
H
3
O
2
\+
N
O
→
C
H
3
O
\+
N
O
2
\+
H
O
2
N
O
\+
O
→
N
O
2
On the basis of previous results, hydrogen abstraction, ring-opening, and isomerization are the major possible chemical pathways for BTEX oxidation by non-thermal plasma \[[@B11-molecules-23-00890],[@B15-molecules-23-00890],[@B40-molecules-23-00890]\]. Nevertheless, each degradation pathway depends on the oxygen or hydroxyl radicals, and hence the acetone degradation is unavoidably influenced by BTEX because of radical consumption. On the one hand, in the light of a semi-empirical first-order kinetic model provided by previous studies, the reaction rate constant with O for the removal of *VOCs* in the NTP reactors can be predicted using Reactions (24) and (25) when the *VOC* removal is below 95% \[[@B26-molecules-23-00890],[@B41-molecules-23-00890]\]. $$O + \left\lbrack {VOCs} \right\rbrack\overset{k}{\rightarrow}~products,$$ $$\frac{\left\lbrack {VOCs} \right\rbrack}{\left\lbrack {VOCs} \right\rbrack_{0}}~ = e^{- ED/\beta}$$ where *k*, \[*VOC*\]~0~, and *β* are represent the reaction rate constant, original concentration of the VOC, and the regression parameter that is equal to −1/*k*, respectively. The *k* values calculated by different VOC atmospheres as a function of the hydrogen weight fraction are shown in [Figure 10](#molecules-23-00890-f010){ref-type="fig"} and [Table S3](#app1-molecules-23-00890){ref-type="app"}, because dehydrogenation reaction by O tends to be carried out on the methyl groups of VOCs undergoing treatment. The result shows that the order of BTEX reactivity is as follows: *p*-xylene \> toluene \> benzene, in agreement with the descending order of the methyl group amount. Furthermore, its impact trend on acetone reactivity implies that a greater hydrogen fraction in BTEX generally results in a greater loss of the acetone rate constant. On the other hand, since other possible pathways of VOC degradation may include ion--molecule, electrophilic and hydroxyl radical, and ozone reactions, ionization energy, proton affinity, and reaction constants (298 K) with the •OH and O~3~ of pollutants are referred to from the NIST and listed in [Table 1](#molecules-23-00890-t001){ref-type="table"} \[[@B42-molecules-23-00890]\]. It is found that ozone reactivity is negligible, since reaction rate constants for ozone are far lower than for the hydroxyl radical. Both the sequence of •OH reaction constants and ease of ionization follow the (decreasing) order *p*-xylene \> toluene \> benzene \> acetone, in consensus with the O reaction rate mentioned above. Thus, it is suggested that BTEX may be decomposed firstly rather than acetone. However, the order of proton affinities is as follows: acetone \> *p*-xylene \> toluene \> benzene, implying acetone is chemically more reactive with electrophilic reagent than BTEX. It thus stands to reason that acetone can react with electrophilic intermediate species generated from hydrogen abstraction reaction on the methyl group under BTEX decomposition, thereby changing the original degradation pathway.
3.8. Roles of Methyl Species in the Acetone Degradation {#sec3dot8-molecules-23-00890}
-------------------------------------------------------
In the plasma environment, acetone can be firstly decomposed into CH~3~• and CH~3~CO• by active radicals and high-energy electrons, as mentioned above (Equation (9)). Then, these two intermediates further go through a series of oxidation reactions to generate CO, CO~2~ and other byproducts by collisions with O or •OH radicals. As is well known, the methyl group (−CH~3~) usually plays a key role in the C1 chemistry, because it is the main precursor of HCHO, HCOOH, and other organic micro-molecules \[[@B43-molecules-23-00890]\]. Therefore, it is believed that the fate of the hydrogen abstraction reaction on the methyl group largely determines the oxidation regime of acetone, as pointed by Alzueta's research \[[@B24-molecules-23-00890]\]. When acetone is treated together with methylic BTEX, amounts of intermediate products from methyl group reaction are inevitably generated and further participate in acetone degradation. For Reaction (9), there is an unfavorable shift in the equilibrium through the increase in the products, resulting in the decreased degradation efficiency of acetone, as well as that of toluene and *p*-xylene. This may be also a reasonable explanation for the immunity effect of benzene shown in [Figure 2](#molecules-23-00890-f002){ref-type="fig"}. Furthermore, considering a limited number of active radicals (mainly •OH), amounts of CH~3~• from acetone degradation would react with CO rather than being completely oxidized to CO~2~, leading to CO~2~ selectivity decreasing. Finally, the methyl group with electron-donating effect can promote the electron density of the benzene ring, which makes electrophilic •OH more easily consumed on the benzene ring than acetone, thereby intensifying aromatic byproduct formation. Therefore, a summary schematic diagram of the influencing mechanism of BTEX on the acetone degradation pathway is proposed, as shown in [Figure 11](#molecules-23-00890-f011){ref-type="fig"}.
4. Conclusions {#sec4-molecules-23-00890}
==============
In this research, the effect of BTEX on acetone degradation by non-thermal plasma was investigated in depth under negative DC corona discharge in humid air. The primary findings are as follows:(1)BTEX has a significant negative impact on acetone conversion when they are treated together. The degree of influence of the three different kinds of BTEX on acetone removal efficiency can be ordered as follows: toluene ≈ *p*-xylene \> benzene.(2)*p*-xylene significantly promotes CO generation, thereby improving CO~x~ selectivity; while benzene or toluene decreases CO~x~ selectivity after coexistence with acetone.(3)Based on the results of in situ experiments and GC-MS, it is found that the quantity of ring-opening byproducts is reduced, while the number of aromatic byproducts rises under abatement of mixture VOCs, indicating that instead of BTEX ring-opening, methylation, and electrophilic substitution reactions in the benzene ring are favored during decomposition.(4)It is deduced that methyl radicals play an important inhibiting role in degradation efficiency and CO~x~ selectivity, which is mainly shown through: (1) unfavorable shifting of the equilibrium of acetone degradation reaction through the increase in the products; (2) reaction with limited active radicals to generate CO rather than highly oxygenated CO~2~; and (3) intensified side reactions on the benzene ring, thereby promoting aromatic byproduct formation.
This work was financially supported by the Beijing Natural Science Foundation (No. 8182033), the National key research and development program (2017TFC0211800), and the American Energy Foundation (No. 6326012).
**Sample Availability:** Not available.
The following are available online, Figure S1: Effects of energy density on the Sco of binary (a) and single (b) component VOCs degradation process. Reaction conditions: Acetone: 250 ppm, BTEX: 50 ppm, 50% RH and total flow: 2 L/min, Figure S2: Effects of energy density on the Sco2 of binary (a) and single (b) component VOCs degradation process. Reaction conditions: Acetone: 250 ppm, BTEX: 50 ppm, 50% RH and total flow: 2L/min, Figure S3: In situ FTIR results of the single component VOCs degradation process (ED = 1600 J/L), Figure S4: The optimized acetone and BTEX structures, corresponding atomic charge and ESP of total density and HOMO orbital, Table S1: Main bond length of acetone and BTEX from theory calculations, Table S2: Relative reactions and rate constants, Table S3: Reaction rate constant (k) and β parameter.
######
Click here for additional data file.
L.H. and X.L. conceived and designed the experiments; L.H. and D.X. performed the experiments; L.H. and X.L. analyzed the data; L.H., D.X. and H.W. contributed reagents/materials/analysis tools; L.H. and X.L. wrote the paper.
The authors declare no conflict of interest.
{#molecules-23-00890-f001}
{#molecules-23-00890-f002}
{#molecules-23-00890-f003}
{#molecules-23-00890-f004}
{#molecules-23-00890-f005}
{#molecules-23-00890-f006}
{#molecules-23-00890-f007}
{#molecules-23-00890-f008}
{#molecules-23-00890-f009}
{#molecules-23-00890-f010}
{#molecules-23-00890-f011}
molecules-23-00890-t001_Table 1
######
The values for ionization energies, proton affinities, and reaction constants for reactions with the •OH and O~3~ of VOCs.
VOCs IE (eV) PA (kJ/mol) *k*~OH~ (cm^3^ molecule^−1^s^−1^) *k*~O3~ (cm^3^ molecule^−1^s^−1^)
------------ --------- ------------- ----------------------------------- -----------------------------------
Acetone 9.70 812.0 3.90 × 10^−14^ nf
Benzene 9.24 750.4 1.28 × 10^−12^ 1.72 × 10^−22^
Toluene 8.83 784.0 6.16 × 10^−12^ 3.9 × 10^−22^
*p*-xylene 8.44 794.4 1.52 × 10^−11^ nf
nf---not found in the NIST.
| {
"pile_set_name": "PubMed Central"
} |
All relevant data are within the paper.
Introduction {#sec005}
============
Acute airway infections are common causes of early childhood hospitalization. Among these infections, acute bronchiolitis is a viral infection commonly encountered in infants and toddlers with acute symptoms of wheezy cough and dyspnea \[[@pone.0121906.ref001], [@pone.0121906.ref002]\]. In addition, an increasing trend in medical visits and total hospitalizations along with the diagnosis of bronchiolitis has been reported in young children in recent decades \[[@pone.0121906.ref003], [@pone.0121906.ref004]\]. Asthma is a severe form of hyperactive airway disease which may present as an acute attack or chronic persistent pattern, and the influence on daily activities can be a serious problem. The clinical symptoms of wheezing and dyspnea in acute asthmatic attacks are similar to acute bronchiolitis. Therefore, whether there is a relationship between bronchiolitis and later childhood asthma, and even adulthood asthma has been investigated for many years \[[@pone.0121906.ref005]--[@pone.0121906.ref009]\]. It has been reported that recurrent wheezing, reduced pulmonary function, and the development of later asthma may occur in young children with bronchiolitis \[[@pone.0121906.ref010]--[@pone.0121906.ref023]\], however there is currently no conclusive evidence \[[@pone.0121906.ref024], [@pone.0121906.ref025]\].
In addition to bronchiolitis, other acute airway infections are not unusual in hospitalized infants and toddlers. Airway inflammation caused by infections may also be related to later childhood asthma. Elucidating the potential relationship between childhood asthma and various kinds of acute airway infections occurring in infants and toddlers is therefore warranted.
Taiwan′s National Health Insurance Research Database (NHIRD) includes comprehensive claims data from the National Health Insurance (NHI) program. This program covers more than 99.5% of residents in Taiwan, and children younger than 18 years of age account for approximately 23% of the whole population in Taiwan \[[@pone.0121906.ref026]\]. The NHIRD provides reliable data for population-based disease research \[[@pone.0121906.ref027]--[@pone.0121906.ref030]\]. These datasets contain aggregated secondary data without personal identification, including patient′s age, gender, admission date, discharge date, diagnosis, expenses, laboratory examination items, detail drug prescription codes and operational codes \[[@pone.0121906.ref031]\]. They have been used in extensive medical research fields from neonates to adults \[[@pone.0121906.ref026]--[@pone.0121906.ref028], [@pone.0121906.ref032]--[@pone.0121906.ref038]\]. It is trustworthy to analyze the population-based relationship between early acute airway infections and later childhood asthma development using these datasets.
We hypothesized that children with a history of early hospitalization due to acute airway infections may have a higher risk of childhood asthma. The purpose of this study was to analyze the relationship between early hospitalization due to acute airway infections, including acute bronchiolitis and other airway infections, in children younger than 3 years of age and subsequent childhood asthma when they are 3 to 10 years of age.
Materials and Methods {#sec006}
=====================
Study population {#sec007}
----------------
Claims data from the Longitudinal Health Insurance Database (LHID) 2010 of the National Health Insurance Research Database (NHIRD) of Taiwan between 1997 and 2010 were retrieved for analysis and comparisons. The LHID 2010 includes claims data of 1,000,000 randomly sampled beneficiaries from the 2010 registry of beneficiaries from the NHIRD. There are no significant differences in age distribution, gender distribution, or average insured payroll-related amount between the patients in the LHID 2010 and the original NHIRD according to the Bureau of National Health Insurance (NHI) in Taiwan \[[@pone.0121906.ref039], [@pone.0121906.ref040]\]. This study was approved by the Institutional Review Board of Taipei Veterans General Hospital, Taipei, Taiwan (VGHIRB No.:2013-06-011BC). There was no consent given because the data were analyzed anonymously and no personal information could be connected in this study.
Information of children younger than 36 months who were hospitalized from 1997 to 2000 were retrieved, and those diagnosed with acute bronchiolitis and other acute airway infections on discharge were enrolled into our study group. The diagnostic codes were based on the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) codes, including acute bronchiolitis (466.1), and other acute airway infections (464.1--464.5, 466.0, and 480--486) ([Table 1](#pone.0121906.t001){ref-type="table"}) \[[@pone.0121906.ref041]\]. The children in the control group were selected from the remaining patients during the same enrollment period who did not have any admission records or outpatient records of the above diagnoses. We randomly enrolled four times the number of age- and gender-matched children for the control group for comparisons. The exclusion criteria were a diagnostic record of asthma (ICD-9-CM: 493) at age younger than 3 years. The remaining children were grouped into hospitalized airway infection (HAI) and control groups, and the HAI group was further sub-grouped into bronchiolitis and other HAI subgroups ([Fig 1](#pone.0121906.g001){ref-type="fig"}). The basic data of the enrolled children, including age and gender were recorded and analyzed. Data on potential comorbidities including preterm (ICD-9-CM: 765), congenital heart disease (CHD) (ICD-9-CM: 745, 746, 747), congenital respiratory disease (ICD-9-CM: 748), and chronic lung disease (ICD-9-CM: 770.7) were also retrieved.
10.1371/journal.pone.0121906.t001
###### Diagnostic Codes Defined as Acute Bronchiolitis, Other Acute Airway Infections and Asthma of the Enrolled Children.
[^a^](#t001fn001){ref-type="table-fn"}
{#pone.0121906.t001g}
ICD-9-CM Code Diagnosis
----------------------------------- ------------------------------------------------------
**Acute bronchiolitis**
**466.1** Acute bronchiolitis
**Other acute airway infections**
**464.1** Acute tracheitis
**464.2** Acute laryngotracheitis
**464.3** Acute epiglottitis
**464.4** Croup
**464.5** Supraglottitis, unspecified
**466.0** Acute bronchitis
**480** Viral pneumonia
**481** Pneumococcal pneumonia
**482** Other bacteria pneumonia
**483** Pneumonia-other specified organism
**484** Pneumonia in infectious disease classified elsewhere
**485** Bronchopneumonia, organism unspecified
**486** Pneumonia, organism unspecified
**Asthma**
**493** Asthma
^a^Based on International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) code.
{#pone.0121906.g001}
Each child was then individually tracked by medical care records from 3 to 10 years of age. The record of a diagnosis of asthma (ICD-9-CM: 493) from the admission datasets (≥ 1 admission with discharge diagnosis of asthma) or outpatient datasets were recorded and analyzed. The diagnosis of asthma in the outpatient dataset was included only if they had 4 or more records of visits due to asthma, and having used one of the following anti-asthmatic medicines at each visit: adrenergics (Anatomical Therapeutic Chemical (ATC) codes: R03A, R03C), xanthines (R03DA), or steroids/β2 agonists (R03BA, R03AK06, R03AK07). The age at first diagnosis of asthma between 3 to 10 years was defined as the age at onset. Medical care conditions including ambulatory visit frequency, admission frequency, admission diagnosis, and medical expenses between 3 to 10 years of age were also retrieved and compared among groups.
Data analysis {#sec008}
-------------
The dataset from the NHIRD was retrieved using Microsoft SQL Server 2008 R2 for database decoding, and SPSS (version 19, SPSS Inc., Chicago, IL, USA) was used for data analysis. SigmaPlot 12.0 (Systat Software Inc. San Jose, CA, USA) and SPSS were used to create graphical representations. One way ANOVA followed by post hoc Student Newman Keul or *t* tests were used to compare means of continuous variables as appropriate, and the chi-square test was used to compare categorical data among different groups. Logistic regression modeling was used to analyze odds ratios (ORs) of the study group compared to the control group. Kaplan-Meier survival analysis with the log rank test, Breslow test, and Tarone-Ware test was used to compare the cumulative asthma event-free ratios of the enrolled children among different groups. For time-to-event analysis of the longitudinal follow-up, Cox regression analysis was performed to analyze hazard ratios between any two groups. A two-sided *p* value of less than 0.05 was considered to determine statistical significance.
Results {#sec009}
=======
A total of 4,967 young children hospitalized with acute airway infections were identified, and we identified another 19,868 age- and gender-matched children for comparison. After excluding those with an early (\< 36 months of age) diagnosis of asthma, there were 3,264 children in the HAI group, including 1,981 children with bronchiolitis and 1,283 children with other HAIs, and 18,527 children in the control group ([Fig 1](#pone.0121906.g001){ref-type="fig"}). There were more boys than girls in each group. The mean age of the children at first diagnosis of bronchiolitis was significant younger than that in the other HAI subgroup (*p* \< 0.001) and the mean index age of control group at enrollment (*p* \< 0.001) ([Table 2](#pone.0121906.t002){ref-type="table"}).
10.1371/journal.pone.0121906.t002
###### Characteristics of the children at enrollment and onset of childhood asthma at age 3 to 10 years.
{#pone.0121906.t002g}
HAI group Control group
------------------------------------------------------------------- --------------------------------------------------- ------------------------------------------------------------------------------------------ --------------------------------------------------- --------------
**At enrollment (0--2 years old)**
Case no. 3264 1981 1283 18527
Age (months, mean ± SD) 12.9 ± 9.6 9.5 ± 7.7[^a^](#t002fn003){ref-type="table-fn"} [^*c*^](#t002fn005){ref-type="table-fn"} 18.2 ± 9.9[^a^](#t002fn003){ref-type="table-fn"} 12.5 ± 9.4
Male (% of total) 1891 (57.9) 1184 (59.8) 707 (55.1)[^b^](#t002fn004){ref-type="table-fn"} 11026 (59.5)
Female (% of total) 1373 (42.1) 797 (40.2) 576 (44.9)[^b^](#t002fn004){ref-type="table-fn"} 7501 (40.5)
**Onset of childhood asthma (3--10 years old)**
Case number
3--10 y (% of enrolled case)[\*](#t002fn002){ref-type="table-fn"} 530 (16.2)[^a^](#t002fn003){ref-type="table-fn"} 351 (17.7)[^a^](#t002fn003){ref-type="table-fn"} 179 (14.0)[^b^](#t002fn004){ref-type="table-fn"} 2159 (11.7)
3--5 y (% of total) 411 (77.5) 272 (77.5) 139 (77.7) 1487 (68.9)
6--10 y (% of total) 119 (22.5) 79 (22.5) 40 (22.3) 672 (31.1)
Age
(months, mean ± SD) 59.1 ± 19.2[^a^](#t002fn003){ref-type="table-fn"} 59.6 ±19.5[^a^](#t002fn003){ref-type="table-fn"} 58.2 ± 18.7[^a^](#t002fn003){ref-type="table-fn"} 64.8 ± 20.1
(years, mean ± SD) 4.9 ± 1.6[^a^](#t002fn003){ref-type="table-fn"} 5.0 ± 1.6[^a^](#t002fn003){ref-type="table-fn"} 4.8 ± 1.6[^a^](#t002fn003){ref-type="table-fn"} 5.4 ± 1.7
Gender
Male (% of total) 322 (60.8) 219 (62.4) 103 (57.5)[^b^](#t002fn004){ref-type="table-fn"} 1404 (65.0)
Female (% of total) 208 (39.2) 132 (37.6) 76 (42.5)[^b^](#t002fn004){ref-type="table-fn"} 755 (35.0)
Abbreviations: HAI, hospitalized airway infection; y, years old.
\*% of enrolled children with early childhood airway disease (0--2 years old).
^a^ *p* \< 0.001 vs. control group;
^b^ *p* \< 0.05 vs. control group,
^*c*^ *p* \< 0.001 vs. other HAI subgroup.
With regards to the occurrence of asthma at 3 to 10 years old, there were 530 (16.2%) children diagnosed with asthma in the overall HAI group, which were significantly higher than the proportion of the control group (11.7%) (*p* \< 0.05). The mean ages of the children diagnosed with asthma in the bronchiolitis and other HAI subgroups were both significantly younger than in the control group (*p* \< 0.001). Although there were still more boys than girls in each group, the proportion of male children was significantly lower in other HAI subgroup (57.5%) than control group (65.0%) ([Table 2](#pone.0121906.t002){ref-type="table"}).
Most cases of asthma were diagnosed when they were 3 to 5 years old (approximately 78% in both HAI subgroups, and 69% in the control group) ([Table 2](#pone.0121906.t002){ref-type="table"}), and the proportion of cases was markedly higher in both HAI subgroups than the control group at 3 years of age ([Fig 2](#pone.0121906.g002){ref-type="fig"}). The control group had the highest asthma event-free rate and the bronchiolitis subgroup had the lowest rate over time. There were significant differences in the asthma event-free rates during the 8 years of follow-up among the 3 groups ([Fig 3](#pone.0121906.g003){ref-type="fig"}) (*p* \< 0.05). The time-to-event analysis showed that the hazard ratios were 1.583 (95% CI: 1.414--1.772), and 1.226 (95% CI: 1.053--1.428) in the bronchiolitis and other HAI subgroups compared to the control group, respectively. In addition, the hazard ratio of the bronchiolitis subgroup compared to the other HAI subgroup was 1.228 (95% CI: 1.075--1.542). Therefore, the risk of developing asthma at an age of 3--10 years was significantly higher in the bronchiolitis subgroup and the lowest in the control group, and the other HAI subgroup had a medium risk.
{#pone.0121906.g002}
{#pone.0121906.g003}
The ORs, adjusted for age and gender, of the children with early HAI for the occurrence of childhood asthma at an age of 3--10 years (*p* \< 0.05) and of 3--5 years (*p* \< 0.001) were significantly higher than in the control group for both bronchiolitis and other HAI subgroups ([Table 3](#pone.0121906.t003){ref-type="table"}). Between the ages of 6--10 years, only the children with bronchiolitis still had a significantly higher OR compared to the control group (*p* = 0.018). For each potential comorbidity, the ORs of the study groups with regards to preterm, congenital respiratory disease, or chronic lung disease did not significantly differ from the control group (all *p* \> 0.05). However, a significantly higher OR for the children with CHD in the bronchiolitis subgroup was noted at an age of 3--5 years compared to the control group (*p* = 0.008) ([Table 3](#pone.0121906.t003){ref-type="table"}).
10.1371/journal.pone.0121906.t003
###### Adjusted odds ratios of children with early hospitalized airway infections (HAIs) (\< 3 years) and childhood asthma (3--10 years), and those with combined comorbidities of preterm birth, congenital heart disease, or congenital respiratory diseases compared to the control group.
{#pone.0121906.t003g}
3--10 y 3--5 y 6--10 y
----------------------------- ------------- ---------------------- ----------------------- ----------------------
**All children**
HAI 3264 (100) 1.470 (1.328--1.630) 1.849 (1.624--2.106) 1.044 (0.894--1.218)
Bronchiolitis 1981 (100) 1.633 (1.443--1.848) 1.928 (1.648--2.256) 1.243 (1.038--1.487)
Other HAI 1283 (100) 1.229 (1.043--1.449) 1.730 (1.420--2.107) 0.745 (0.569--0.976)
Control 18527 (100) 1 1 1
**Preterm (ICD = 765)**
HAI 110 (3.4) 1.164 (0.629--2.154) 1.557 (0.693--3.499) 0.824 (0.351--1.938)
Bronchiolitis 86 (4.3) 1.232 (0.639--2.376) 1.528 (0.641--3.645) 0.949 (0.389--2.315)
Other HAI 24 (1.9) 0.931 (0.296--2.929) 1.659 (0.436--6.311) 0.402 (0.051--3.179)
Control 164 (0.9) 1 1 1
**CHD (ICD = 745,746,747)**
HAI 260 (8.0) 1.244 (0.856--1.808) 1.838 (1.160--2.911) 0.689 (0.382--1.243)
Bronchiolitis 187 (9.4) 1.458 (0.971--2.189) 1.973 (1.193--3.263) 0.911 (0.495--1.677)
Other HAI 73 (5.7) 0.754 (0.364--1.560) 1.503 (0.684--3.306) 0.156 (0.021--1.147)
Control 674 (3.6) 1 1 1
**CRD (ICD = 748)**
HAI 47 (1.4) 0.656 (0.213--2.019) 1.310 (0.235--7.360) 0.452 (0.117--1.750)
Bronchiolitis 39 (2.0) 0.627 (0.194--2.028) 0.917 (0.142--5.925) 0.559 (0.143--2.179)
Other HAI 8 (0.6) 0.810 (0.130--5.028) 3.667 (0.424--31.726) 0
Control 1 1 1
**CLD (ICD = 770.7)**
HAI 9 (0.0) 0.667 (0.059--7.475) 2.250 (0.125--40.656) 0
Bronchiolitis 8 (0.0) 0.762 (0.067--8.665) 2.571 (0.141--47.017) 0
Other HAI 1 (0.0) 0.0 0 0
Control 19 (0.0) 1 1 1
Abbreviations: y, years old; CHD, congenital heart disease; CLD, chronic lung disease arising in the perinatal period; CRD, congenital respiratory disease; HAI, hospitalized airway infections; aOR, adjusted odds ratio, adjusted by age and gender; ICD, International Classification of Diseases, 9th Revision, Clinical Modification code.
The analysis of subsequent medical care requirement of the enrolled children at an age of 3--10 years also showed significantly higher rates of total hospitalizations, ambulatory visits, and medical expenses for children with HAI, including both subgroups, compared to the control group (all *p* \< 0.01) ([Table 4](#pone.0121906.t004){ref-type="table"}). In each patient with asthma, the number of admissions with a diagnosis of asthma was significantly higher in the bronchiolitis subgroup (*p* \< 0.001) than the control group, but not in the other HAI subgroup (*p* \> 0.05) ([Table 4](#pone.0121906.t004){ref-type="table"}). Comparing both HAI subgroups, the children in the bronchiolitis subgroup had a higher number of hospitalizations due to asthma, higher number of ambulatory visits, and overall medical expenses than the other HAI subgroup at age 3 to 10 years (all *p* \< 0.01) ([Table 4](#pone.0121906.t004){ref-type="table"}).
10.1371/journal.pone.0121906.t004
###### Subsequent hospitalization and ambulatory visit frequencies and expenses in the enrolled children at age 3 to 10 years.
{#pone.0121906.t004g}
HAI group[\*](#t004fn002){ref-type="table-fn"} Control group
------------------------------------------ ----------------------------------------------------- -------------------------------------------------------------------------------------------- ----------------------------------------------------- ---------------
**All hospitalizations (any diagnosis)**
All enrolled case no. 3264 1981 1283 18527
Frequencies 2272 1525 747 6318
Hospitalization/case 0.7 ± 1.4[^a^](#t004fn003){ref-type="table-fn"} 0.8 ± 1.6 [^a^](#t004fn003){ref-type="table-fn"} 0.6 ± 1.1[^a^](#t004fn003){ref-type="table-fn"} 0.3 ± 0.9
**Hospitalizations due to asthma**
Asthma case no. 530 351 179 2159
Frequencies 726 559 167 1729
Hospitalizations/asthma case 1.4 ± 2.1[^a^](#t004fn003){ref-type="table-fn"} 1.6 ± 2.4[^a^](#t004fn003){ref-type="table-fn"} [^b^](#t004fn004){ref-type="table-fn"} 0.9 ± 1.3 0.8 ± 1.8
**Frequency of ambulatory visits**
Frequencies 420,226 264,764 155,462 2,119,346
Visits/child 129 ± 76[^a^](#t004fn003){ref-type="table-fn"} 134 ± 77[^a^](#t004fn003){ref-type="table-fn"} [^b^](#t004fn004){ref-type="table-fn"} 121 ± 74[^a^](#t004fn003){ref-type="table-fn"} 114 ± 67
**Medical expense**
All expense (USD\$) 5,734,192 3,211,446 37,261,008 37,261,008
Expense/case (USD\$) 2,741 ± 7,066[^a^](#t004fn003){ref-type="table-fn"} 2,895 ± 8,854[^a^](#t004fn003){ref-type="table-fn"} [^b^](#t004fn004){ref-type="table-fn"} 2,503 ± 4,843[^a^](#t004fn003){ref-type="table-fn"} 2,011 ± 2,538
Abbreviations: HAI, hospitalized airway infections; USD, United States dollar.
\*children who were admitted due to lower airway infections when they were younger than 3 years old.
^a^ *p* \< 0.01 vs. control group.
^b^ *p* \< 0.01 vs. other HAI group.
Discussion {#sec010}
==========
The results of this study showed a significantly higher occurrence of childhood asthma in children who were hospitalized at an early age (\< 3 years old) due to acute airway infections, including both bronchiolitis and other acute HAIs. Furthermore, we also demonstrated that most cases of asthma developed in children aged 3--5 years, the presence of CHD increased the risk in children with bronchiolitis, and the risk of early bronchiolitis was higher than other HAIs.
Viruses are the leading causes of acute lower respiratory tract infections in young children, and respiratory syncytial virus (RSV) has been reported to be the most common pathogen \[[@pone.0121906.ref001], [@pone.0121906.ref042]\]. Previous investigations have mostly focused on the role of RSV infection-associated bronchiolitis and the development of later childhood asthma \[[@pone.0121906.ref010], [@pone.0121906.ref012]--[@pone.0121906.ref020], [@pone.0121906.ref025]\], however some studies have also reported the influence of non-RSV viruses \[[@pone.0121906.ref023], [@pone.0121906.ref043]--[@pone.0121906.ref046]\]. The current study investigated the potential influence of relatively severe acute airway infections in children requiring hospitalization regardless of the causal organisms, and we found that acute airway infections other than bronchiolitis also increased the risk of subsequent childhood asthma. This suggests that attention should be paid to young children hospitalized for different types of acute airway infections due to the possibility of later childhood asthma.
Previous investigations on the epidemiology of bronchiolitis in Taiwan have demonstrated that bronchiolitis in young children occurs all year round with two slight peaks in spring and autumn \[[@pone.0121906.ref011], [@pone.0121906.ref047]--[@pone.0121906.ref049]\], and this is different from North America and Europe where a predominant peak is noted in winter \[[@pone.0121906.ref001], [@pone.0121906.ref002], [@pone.0121906.ref023]\]. Ethnic and environmental differences may exist, however published reports tend to show a higher likelihood of developing asthma or recurrent wheezing in children who suffer from early bronchiolitis in many countries \[[@pone.0121906.ref011], [@pone.0121906.ref012], [@pone.0121906.ref014], [@pone.0121906.ref021], [@pone.0121906.ref023], [@pone.0121906.ref045], [@pone.0121906.ref050], [@pone.0121906.ref051]\]. Therefore, the risk of developing asthma in children with early airway infections seems a general phenomenon, and our results also support this phenomenon in children living in Taiwan.
With regards to the age at onset, our results showed that most cases of childhood asthma were initially diagnosed when the children were 3--5 years old, regardless of whether their initial problems were bronchiolitis, other HAIs, or neither of these ([Table 3](#pone.0121906.t003){ref-type="table"}). In particular, there were significantly more cases of asthma in both HAI groups than in the control group at 3 years of age ([Fig 2](#pone.0121906.g002){ref-type="fig"}). This means that children with early HAIs are more likely to suffer from an earlier onset of childhood asthma.
Sigurs et al. reported that the development of asthma, recurrent wheezing or impaired pulmonary function is more common in children with RSV-infected bronchiolitis \[[@pone.0121906.ref013]--[@pone.0121906.ref016]\]. However, some meta-analyses have reported no significant differences compared to control groups in children older than 5 years \[[@pone.0121906.ref052]--[@pone.0121906.ref054]\]. In the current study, we only analyzed the development of childhood asthma. Although the ORs of the bronchiolitis subgroup at an age of 6--10 years were still significant (*p* = 0.018), they were not as high as in those between 3--5 years ([Table 3](#pone.0121906.t003){ref-type="table"}). In addition, in the other HAI group, the risk of asthma was not higher than the control group at an age of 6--10 years. These findings suggest that infection-associated airway hyperactivity may gradually resolve as the children grow older and their airways become larger.
Prematurity, CHD, and chronic lung disease have been reported to be potential risk factors for severe bronchiolitis in early childhood \[[@pone.0121906.ref055]\]. In children born prematurely, the risk of developing childhood asthma was reported to be higher than full-term children \[[@pone.0121906.ref056], [@pone.0121906.ref057]\]. Our study demonstrated that early HAI would not further increase the risk of developing asthma in preterm-born children, nor would the children with CRD or CLD. In children with CHD, the relationship to childhood asthma was rarely reported \[[@pone.0121906.ref058]\]. Our results revealed that the occurrence of bronchiolitis in children with CHD significantly increased the risk of developing asthma at age 3--5 years. Therefore, children with CHD and who suffer from bronchiolitis should be followed carefully for the potential risk of developing later childhood asthma, and their parents should be informed about this.
Genetic factors have been reported to be associated with early bronchiolitis and later asthma in children \[[@pone.0121906.ref059], [@pone.0121906.ref060]\]. In the current study, we focused on the association between severe airway infections and later asthma in children living in Taiwan. However, we lacked data to compare genetic differences, and further investigations are warranted to elucidate this issue.
Comparing the bronchiolitis and other HAI subgroups at an age of 6--10 years, we demonstrated higher odds ratios, more medical care requirement, and a higher number of hospitalizations due to asthma in the children with bronchiolitis (Tables [3](#pone.0121906.t003){ref-type="table"} and [4](#pone.0121906.t004){ref-type="table"}). These findings suggest that bronchiolitis may have a more severe influence on airways than other HAIs, although both conditions can increase the risk of later asthma development at an age of 3--5 years. Thus, we suggest pediatricians to pay more attention to children suffering from bronchiolitis for the development of later childhood asthma up to 10 years of age.
Considering the cause and effect relationship between early HAIs and later childhood asthma, there may be some possibilities. For one thing, children with asthmatic characteristics may have susceptible airways and be easier to get HAIs during early childhood. For another thing, early HAIs in little kids may influence their airway development, and further increase the risk of developing asthma during later childhood. These two mechanisms may be synergistic. The present study is only an observation using claims data, so it is not easy to elucidate the cause-effect relationship. A further study on this issue will be necessary in the future.
There are some limitations to the present study. The main limitation is that NHIRD is derived from claims data of medical care providers in the NHI program, but the laboratory data and patients′ history were not included. Therefore, accurate viral or bacterial diagnoses could not be obtained, and we could not further evaluate the influence of microbial patterns in childhood asthma. We also could not evaluate the role of other potential confounding factors, such as patients′ birth weight, presence of older siblings, maternal smoking, family circumstances, etc. In addition, the diagnosis of asthma was based on the diagnostic codes of the in- and out-patients claims data without supporting evidence of pulmonary function tests.
Conclusions {#sec011}
===========
Young children hospitalized with bronchiolitis or other HAIs during the first 3 years of life may have a higher risk of developing childhood asthma at age 3 years or older, and require more medical care and expense in the following years. The parents of children with HAIs at age 0--2 years should be informed that there is a higher incidence of childhood asthma and that they should see a physician upon signs of obstructive airway disease, especially in children with bronchiolitis and underlying CHD, and at an age of 3 to 5 years.
This work was based on the datasets of the National Health Insurance Research Database provided by the Bureau of National Health Insurance (BNHI), Department of Health, Executive Yuan, Taiwan. The data interpretation and conclusions do not represent those of the respective institutions or agencies. All authors thank Mr. Jian-Ping Lin for his help in the data mining and statistical analysis.
[^1]: **Competing Interests:**The authors have declared that no competing interests exist.
[^2]: Conceived and designed the experiments: MJJ YSL PCT. Performed the experiments: MJJ. Analyzed the data: MJJ. Contributed reagents/materials/analysis tools: MJJ YSL CFY WJS. Wrote the paper: MJJ.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1}
===============
With the increase in life expectancy, there is, proportionately, a dramatic increase in age-related health conditions including, in particular, eye diseases. Among all the eye pathologies, diabetic retinopathy (DR) is currently one of the leading causes of blindness in the working-age population in developed countries \[[@B1]--[@B3]\]. Epidemiological data show that the prevalence of DR is around 30% in the diabetic population and that the annual incidence varies from 2% to 6% \[[@B4]\]. Every year, up to 1% of diabetic patients develop serious ocular complications leading to both poor quality of life and a socioeconomic burden \[[@B5]\].
With regard to the Italian setting, a study conducted in the Veneto Region, using clinical fundus examination, confirmed literature data with a prevalence of DR of 26.6% \[[@B6]\]. A significant worldwide increase in the number of persons affected by diabetes has been estimated for the next 10 years, reaching approximately 380 million by 2025, mostly in developing countries. In Europe, compared with 2007, there is expected to be an increase of 10 million persons with a prevalence of diabetes below 10% \[[@B7]\].
As the prevalence of diabetes is expected to rise in the future, an associated increase in DR cases should also be expected. However, recently, data suggest that the prevalence of DR may decrease, due to the intensification of the screening programmes and better control of risk-factors, thus underlying their effectiveness and the significance \[[@B8]\]. In this view, scholars and practitioners widely agreed that DR, at any stage of progression, requires specific management: from the screening programmes (useful for an early diagnosis) to the definition of pharmacological treatments needed. In recent years, the strategies for DR prevention have moved from the traditional ophthalmological examination to a faster digital retinal imaging acquisition and grading of DR. The application of these new screening programmes, followed by prompt diagnosis and a better timely management, is well known to prevent significantly the risk of diabetic blindness \[[@B9], [@B10]\], as well as to decrease all the costs related to the investigated pathology. In fact, from an economic point of view, the annual cost per patient affected by DR is approximately twice as high as those patients with diabetes only \[[@B11]\]: thus, the implementation of effective and recognized screening programmes could represent cost-effective strategies \[[@B12], [@B13]\], one useful to narrow both the economic and the social burden of DR.
Despite the implementation of successful national screening programmes worldwide, the lack of diffuse screening activities for DR in Italy is delaying the diagnosis and prompt treatment of DR.
Public screening programmes and prompt treatments have been shown to reduce lifetime costs related to visual disability \[[@B14], [@B15]\]. For these reasons, vision impairments as a result of DR should be considered easy to prevent, and the development of systematic programmes of screening should represent an urgent and primary healthcare need.
Moving on from these premises, since DR should be one of the major priorities for the healthcare services, the present study aimed at investigating the feasibility related to the introduction of a specific and accurate screening programme in the catchment area of Treviso (defined as ULSS 9), in comparison with the "no prevention" strategy, including also the costs averted to blindness, in terms of the validity of the intervention and the direct costs absorbed (efficiency) by the regional healthcare service. The analysis was designed assuming the regional healthcare service (Veneto Region) point of view.
2. Materials and Methods {#sec2}
========================
2.1. Study Design {#sec2.1}
-----------------
In order to achieve the previously mentioned objective, a perspective study for the implementation of a screening programme devoted to DR was planned to be performed between September and December 2012, after receiving the approval from Ethic Committee of the Local Health Authority. As DR is a multiprofessional and a multidimensional eye disease, a multidisciplinary group was involved in the project, including general practitioners, diabetes experts, administrative staff, nurses, epidemiologists, and ophthalmologists. The involvement of different healthcare professionals in the screening programme could influence the resources absorption and the optimisation of the screening pathway \[[@B12], [@B13], [@B16]--[@B18]\].
The screening involved all types (type 1 and type 2) of diabetic patients living in the area of Ponzano (Treviso) that has 5.000 inhabitants and is part of one of the Local Health Authorities of the Veneto Region: the ULSS 9 of Treviso. Diabetic subjects were accurately identified by crossing different databases: primary and secondary diagnosis for hospitalisation, drugs and delivery systems, prescriptions, health care procedures covered by the diabetic code, and the diabetic code assigned to patients. In particular, the following patients were invited to attend the screening programme: (i) outpatients going to a hospital specialist visit to the diabetologists, or first visit to ophthalmologist; (ii) patients going to the territorial pharmacist, in order to retrieve the specific diabetes drugs; and (iii) patients referring to the general practitioners, with a diagnosis of diabetes. Individuals who were already diagnosed with DR were not included in the study.
Initially, a personal letter of invitation was sent to all the diabetic target population (*N* = 498) to be enrolled in the screening programme. All the patients attending the screening procedure have signed a specific informed consent form, consistent with ethical aspects. Screening for diabetic retinopathy was made using a nonmydriatic fundus camera. A semiobscured visiting room was used to optimise physiological mydriasis before each exam. Three nonmydriatic, 45° field, digital retinal images were captured, in accordance with a previously validated technique. The three fundus images encompassed the following retinal fields: field 1 centred on the macula, field 2 centred on optic disc, and field 3 midperipheral superior-temporal field \[[@B19]\]. The images were obtained by trained paramedical staff (in particular, nurses). All images were electronically transmitted to the reading centre and stored in an online secured database called "Eye Knowledge Network" for the second step, online examination. Retinal images were graded for DR at the Ca\' Foncello Treviso Hospital, where the Ophthalmology Unit is located, by two experts and certified readers who were members of the Reading Centre staff of the University of Padova. DR and diabetic macular edema (DME) were graded in accordance with the International Classification proposed by the American Academy of Ophthalmologists \[[@B20]\]. When the quality of the images was "inadequate" for the clinical evaluation and when fundus images were graded as "positive," the patients were referred for further ophthalmologic examination. "Positive" findings included retinal changes that required specialist management: moderate and severe nonproliferative DR, proliferative DR, DME, and/or any other retinal abnormality. Further ocular and diagnostic examinations or treatments, if necessary, were then planned. Fundus images were graded as "negative" if no DR or nonsight threatening DR was detected. A report, with the results of the screening and the correct follow-up timetable for the "negative" screened population, was sent to the patient\'s general practitioner within 1 month after the screening.
2.2. Economic Evaluation {#sec2.2}
------------------------
From a methodological point of view, in order to quantify the impact of the introduction of screening programme in the clinical practice, both an activity based costing (ABC) analysis and a budget impact analysis (BIA) were conducted.
The ABC \[[@B21], [@B22]\] is useful for the enhancement of the average costs related to each phase of the screening pathway. In particular, the main objective of the ABC is the measurement of the costs and the performance activities, taking into account also the related human and materials resources for the proper development of the procedure. In this view, the economic impact of each patient was determined utilising the following components: (i) human resources (i.e., individuals involved in the different phases of the screening programme, such as administrative staff, nurses, and ophthalmologists); (ii) materials and equipment; (iii) pharmacological and/or laser treatment. The ABC did not take into consideration the costs related to the delivery of the invitation letter: different technologies could be implemented in order to carry out this task (telephonic invitation, e-mail contact, personal contact, or letter), differing for the related economic value.
After the implementation of the ABC, the BIA was applied. A BIA allows the prediction of the potential financial impact of a new technology adoption, into a healthcare system \[[@B23]\]; influencing, in a positive or in a negative manner, the healthcare expenditure and considering both a specific point of view and a determined time horizon. In this view, two scenarios were simulated, thus comparing the so-called "do nothing strategy" with the implementation of a proper screening programme. In particular, in both scenarios, the occurrence and the related cost of blindness were taken into consideration as direct healthcare costs. Since the analysis assumed the Healthcare Regional Services point of view, the intangible and indirect items of expenditure were excluded from the study.
The economic analysis used the 2015 Italian Outpatients and Hospital Admissions Reimbursement Tariffs. Drug costs were derived from the officially published NHS price list. If necessary, economic values were reported in "euros," considering the 2015 inflation rate, using the Consumer Price Index for healthcare expenditure, thus making economic measures comparable, and being based on the same year of reference.
In order to ensure the robustness of the result, a sensitivity analysis was carried out. In particular, both the attendance rate to the screening programme and the percentage of patients undergoing the complete eye examination were modified, thus understanding if significant changes in the feasibility of the programme occur.
Literature reported that patients\' compliance with DR screening is not optimal worldwide, hence reporting attendance rates ranging from a minimum of 32% and a maximum of 92% \[[@B24]--[@B27]\]. Further analysis reported a variation in the eye examination rate. The Local Health Authority involved declared that it could reach about the 80%--85% if collaboration with community-based organisations, as well as greater information activities of DR risks among citizens, is implemented.
Moving on from these premises, 4 different analyses were performed.
3. Results {#sec3}
==========
3.1. The Sample under Assessment {#sec3.1}
--------------------------------
498 individuals were identified as being diabetic patients within the area of Ponzano (that has 5,000 inhabitants); thus, it emerged that the prevalence of the diabetes in the investigated town was around 10%.
Out of the 498 diabetic patients originally invited to enroll in the screening programme, 340 accepted to be evaluated (68%), although it was possible to confirm a response rate equal to 80% since 57 patients did not attend the screening programme because they had already undergone a complete eye examination that showed no signs of DR development. No patients with a previous DR diagnosis were enrolled in the present study.
The study population was composed predominantly of males (55%); the average age of the sample was 68 years (range: 26--93). On average, patients have been suffering from diabetes for 20 years (range: 2--41). The most common comorbidities developed by patients were hypertension (40%) and dyslipidaemia (38%). 324 patients (95%) successfully completed the procedure, though in 16 cases (5%) images were noncaptured due to either systemic conditions of the patients, insufficient mydriasis, or other technical reasons related to the fundus camera. The quality of digital images was adequate for the interpretation in 260 patients (80%), although it was ungradable in 64 cases (20%) due to insufficient dilation, media opacity, poor fixation, or the absence of one of the captured fields.
As a result, taking into account the gradable 260 patients, 225 were classified as "negative" (87%) and 35 as "positive" (13%). Based on the entire screened population, 115 patients (34%) were referred to an ophthalmologist. Of these, 16 (14%) were patients from whom it was impossible to obtain images, 64 (56%) had no evaluable images, and 35 (30%) were "positive" cases.
Of the previously mentioned 115 patients, 92 (91%) underwent a full ophthalmological examination, giving the following results: in 67 cases (73%), DR was not detected, though 25 patients (27%) presented signs of DR. In particular, 24% (*N* = 6) had mild nonproliferative DR, 52% (*N* = 13) had moderate nonproliferative DR, 12% (*N* = 3) had preproliferative DR, and 12% (*N* = 3) had proliferative DR. Concomitant DME was present in 36% (*N* = 9) of these patients.
In 9 cases (36%), prompt treatment with intravitreal injections and/or laser photo-coagulation was required for proliferative and severe preproliferative retinopathy or macular edema.
3.2. Results from the Study Conducted in Ponzano {#sec3.2}
------------------------------------------------
After these general remarks, regarding the screened population, an economic evaluation was required in order to investigate the amount of resources dedicated for this specific innovative programme.
It should be noted that the Veneto Region 2015 tariffs were taken into account in order to investigate the value of the complete screening programme, thus also including the entire cycle of intravitreal injections.
With reference to the abovementioned distribution of the patients, the economic resources absorption, with regard to the whole population screened pathway, was divided into four distinct and logical phases:340 patients attended the screening programme, involving 1 healthcare professional (nurse) who spent 12 minutes per patient and considering the equipment amortization.324 patients completed the procedure, after which 1 ophthalmologist interpreted their digital images, spending about 5 minutes per patient, considering also the workstation and the administrative staff costs.out of the 115 patients referred for a complete examination, only 92 underwent an ophthalmology examination, in order to obtain an in-depth analysis of the disease (in the third phase, it is important to note that the costs of all the materials and the drugs utilised by the clinicians for the complete eye examination are included in the "first visit" reimbursement tariff).Nine patients, who were suffering from a severe stage of pathology, received pharmacological and/or laser treatment, thus considering that both the human resources and the materials/equipment costs are included in the procedure costs, as detailed in [Table 2](#tab2){ref-type="table"}.It was first necessary to evaluate the costs related to the human resources involved in the screening programme. Thus, the gross monthly salary, related to each specific professional title, was taken into consideration. In particular, its time value per minute was multiplied by the time dedicated to each procedure.
[Table 1](#tab1){ref-type="table"} shows the categories of cost impacting on each phase and the total cost per phase.
[Table 1](#tab1){ref-type="table"} shows that the fourth phase absorbed the most part of the economic resources, presenting the economic evaluation of the treatment for 9 patients. In particular, three groups of patients suffering from RD or DME were considered: (i) 4 patients (45.5%) received a therapeutic cycle of intravitreal injections with Ranibizumab alone; (ii) 3 individuals (30%) received laser treatment as a support of Ranibizumab injection; and (iii) 24.5% (2 patients) received Dexamethasone, thus being consistent with literature and real-life data \[[@B28]\]. Treatment costs were related to a period of 12 months, in which clinicians administered to the patients a cycle composed of, on average, 3.61 (Ranibizumab) or 1.3 (Dexamethasone) intravitreal injections, based on the observation of the clinical pathway of the nine patients under investigation in the health authority of reference and consistent with other national and international literature evidence \[[@B28], [@B29]\]. No surgical interventions were performed in the observed population.
Costs for an intravitreal injection were distinguished as follows: (a) the cost of the specific drug (€644.73 for Ranibizumab and €951.75 for Dexamethasone) and (b) the cost of the procedure carried out by the health authority of reference (€290.00, independently of the administered drug). Patients performing laser therapy absorbed an additional cost equal to €81.28 for every single procedure.
The costs related to these two treatment phases are detailed in [Table 2](#tab2){ref-type="table"}.
With reference to Tables [1](#tab1){ref-type="table"} and [2](#tab2){ref-type="table"}, the amount of all the screening costs for patients within the area of Ponzano was equal to €35,899.81, treating the 340 patients who attended the programme.
3.3. Results from the Economic Evaluation Testing the Feasibility of the Screening Programme in the Grater Treviso Catchment Area {#sec3.3}
---------------------------------------------------------------------------------------------------------------------------------
After the costs of the screening programme phases had been calculated, the prevention activity was then extended to the Treviso catchment area (considering the whole ULSS 9), in order to estimate the feasibility of the screening procedure in a larger population.
This area presents 22,000 estimated diabetes cases. If the response rate was the same as that in the pilot study (80%), 17.600 patients would be expected to attend the screening programme within the first year. Considering an average work shift of 7 hours per day, a single nurse, completely devoted to the screening activities, could be able to perform about 7,000 procedures per year. With regard to the ophthalmologist and the administrative staff, assuming an average work shift of 8 hours, it emerges that the clinician could produce on average 9,600 medical reports per year, and the administrative staff could generate a maximum amount of 19,200 documents per year. As a result, 3 members of the paramedical staff (nurses), 2 ophthalmologists, and 1 administrative would also be required to treat the abovementioned attending patients.
Investment in the required equipment would be made to cover the need of the expected diabetic citizens; in particular, it should be allocated to 5 pieces of equipment, thus requiring an overall investment of €72,000.
Considering the incidence of blindness as a result of DR (0.002% per year, taking into account the whole ULSS 9 that has 419,728 inhabitants), 8 new cases every year would be expected if the screening programme was not applied. The screening programme conducted reported an effectiveness equal to 74.59%, calculated as the difference between individuals attending the screening programme and individuals who did not participate in the prevention activities, or patients who denied to go to the ophthalmologist for the complete eye examination. Considering the Treviso catchment area, 4,400 individuals did not attend the screening programme whereas 1,190 did not perform the in-depth visit, thus reaching a total amount of 5,590 individuals (25.41% of the overall invited population). In this view, in the innovative scenario, only 2 patients would develop blindness.
Literature evidence \[[@B30]\] shows that the cost of blindness is approximately \$18,670 (i.e., €16,803), including only medical and direct costs.
In addition, therapeutic treatment should be administered to 8% of DR patients, considering both the screening pathway and the "no prevention" strategy. In particular, 381 patients referring to the screening pathway scenario and 609 patients (thus considering a DR occurrence rate equal to 34.60% \[[@B31]\] within the diabetic population) for whom "no prevention" strategy was implemented received drugs therapy. The distribution of the treated population, considering the administered drug (74.5% for Ranibizumab alone or with the support of laser therapy and 25.5% for Dexamethasone \[[@B28], [@B29]\]), as well as the treatment frequency within a 12-month time horizon, was assumed to be the same as the conducted study previously described.
With reference to these data, the total costs of the two different pathways are presented in [Table 3](#tab3){ref-type="table"}. In this view, it emerged that in both scenarios the more significant economic resources absorption was related to the treatment phase: at 12-month time horizon, the administration of Ranibizumab with the support of laser therapy, Ranibizumab alone, or Dexamethasone required, on average, €3,667.80, €3,374.38, and €1,614.28, respectively, per patient.
[Table 4](#tab4){ref-type="table"} reports the results of the sensitivity analyses, demonstrating an overall economic advantage in all the cases in which the prevention activity is implemented, thus ensuring the robustness of the BIA result. In particular, it emerged that the economic benefits of the screening programme implementation are more sensitive to the higher number of screened individuals, in comparison with the decrease of the blindness events that occurred.
More investments are required from the healthcare service, if the screening programme would cover a larger number of the target population: in particular, achieving a maximum attendance rate equal to 92%, the healthcare service would equally benefit of an economic saving of −2.38%, always resulting in the preferable solution.
4. Discussion {#sec4}
=============
Screening for diabetic retinopathy is important because the majority of patients who develop DR show no symptoms until diabetic macular edema and/or proliferative diabetic retinopathy are present, thus confirming that, in the early stages of the investigated disease, any notable symptoms affect the patients.
Although the beginning of the DR has an asymptomatic nature, the attendance rate reported in the proposed study was equal to 80%, thus demonstrating the effectiveness and the validity of the screening programme. An increase in the attendance rate could be achieved with an improvement and a diversification of the communication tools to inform patients concerning the importance of the eyes examinations and controls, thus enhancing their awareness on this field. In particular, literature \[[@B32]\] reported that telehealth or telemedicine programmes may facilitate early DR diagnosis and timely treatment, preserving vision.
The results of the study show the importance of a screening programme, from an economic point of view, one leading to a substantial saving of €271,543.32 (−13.71%) in comparison with the "no prevention" strategy. In addition, three-field colour, 45-degree, nonmydriatic images have demonstrated a sensitivity and specificity of 82% and 92%, respectively, in the diagnosis of DR, representing an effective tool in a screening setting \[[@B19]\].
Traditionally, ophthalmologists evaluate patients for DR by mydriatic indirect ophthalmoscopy. The references for the correct follow-up and management are the AAO and national guidelines for DR \[[@B33], [@B34]\]. However, despite the relevance of implementing a screening programme, in the real world, these recommendations are seldom adhered to. The increasing number of diabetic patients, in particular due to the population ageing, delays access to the next ophthalmologic examination. It has been estimated that only 50% of the known diabetic patients receive the recommended regular eye examinations \[[@B35]\], something that may be considered a real concern. Eye screening offers the possibility of identifying the early signs of DR, thus preventing visual loss due to DME. As previously described, new technological screening procedures based on digital mydriatic and nonmydriatic fundus images present multiple advantages. Images can be taken by trained nonspecialist operators and can be viewed "online" by specifically trained ophthalmologists in a deferred time, thus sparing and optimising resources. This could lead to a significant benefit with the decreasing of waiting lists, a phenomenon perceived by citizens as a serious problem of modern healthcare systems that compromises the coverage of their health needs. In addition, because of their easy and safe use, without the administration of drugs, nonmydriatic cameras can be placed in primary care settings in order to improve the access to care.
A large-scale application of this screening strategy could spare unnecessary examinations for "negative" patients, thus preventing irreversible loss of visual acuity for persons affected by retinopathy, due to long waiting lists for eye examination.
In particular, it emerged that the screening programme presented in the study has the potential to reduce the prevalence of blindness due to DR in the Veneto Region.
With reference to this, the nonmydriatic fundus camera is not only effective but also cost-effective in the investigated greater area and leads to significant benefits for both the regional healthcare service and for the patients. Vision loss is associated not only with a large increase of costs due to the management of this condition but also with the compromise of a patient\'s quality of life, thus representing a significant social burden and requiring a future in-depth analysis. In fact, with the rapid ageing of the population, and considering that older adults could stay active and also productive in this specific part of their lives, blindness should be a public healthcare priority.
Healthcare regulators and policy makers will benefit from the implementation of adequate screening programmes, optimising effectiveness and resources allocation within this specific target population. In this view, the present study would represent the first attempt to extent the current theories and models into the practical context of the healthcare sector, extending the results of the study, from the Italian setting to other European and international contexts. The results would contribute in the advancement and establishment of organisational and management models to be applied in the prevention sector, freeing up hospital resources for more severe cases of patients and reducing the economic and social burden of waiting lists for ophthalmic procedures. The early diagnosis and the process of taking charge of the patients could be considered an effective way to offer a better healthcare delivery to the patients, a clinical pathways optimisation but, in particular, a possible economic saving for the healthcare services.
A future interesting contribution for policy makers could be also the definition of the best screening programme organisational setting, maximizing both effectiveness and efficiency, involving different healthcare professionals, such as general practitioners and pharmacists. New approaches to the screening development could reduce healthcare expenditure and increase the attendance rate for patients.
In this view, further researches could be addressed to the proposal of a reimbursement tariff for the diabetic screening activities, with reference to different clinical setting in which the procedure could be performed, as well as the implementation of a multidimensional assessment (through the health technology assessment tool) in order to understand all the possible implications (organisational, social, and equity aspects) of the adoption of screening programmes, measured into the clinical practice, generating significant advancement to these findings and their robustness.
Competing Interests
===================
The authors declare that there is no conflict of interests regarding the publication of this paper.
######
Cost of the screening programme, distinguished by phases.
Phases Human resources Materials and equipment Drugs Total
----------- --------------------- ------------------------- ------------ ------------
Phase I €1,056.23 €2,112.46 --- €3,168.69
Phase II €1,170.98 €1,170.39 --- €2,341.37
Phase III €3,109.10 --- --- €3,109.10
Phase IV --- --- €27,280.65 €27,280.65
*Total*€*35,899.81*
######
Details of the treatment\'s costs.
Treatment option Cost of the drug administered Procedure Total cost for a single injection Total cost for a therapeutic cycle Total cost for the treated population (9 patients)
----------------------------- ------------------------------- ----------- ----------------------------------- ------------------------------------ ----------------------------------------------------
Ranibizumab + laser therapy €644.73 €371.28 €1,016.01 €3,667.80 €9,903.07
Ranibizumab €644.73 €290.00 €934.73 €3,374.38 €22,928.93
Dexamethasone €951.75 €290.00 €1,241.75 €1,614.28 €3,559.48
######
Economic resources related to the investigated procedures.
Screening programme pathway
----------------------------- ------------------
Phase I €164,026.11
Phase II €121,230.90
Phase III €160,924.94
Phase IV (treatment) €1,154,724.70
Investment in equipment €72,000.00
Blindness €35,843.46
*Total* €*1,708,750,11*
"Do nothing" strategy
Blindness €134,424.00
Treatment €1,845,869.43
*Total* €*1,980,293.43*
*Economic savings* *−*€*271,543.32*
*−13.71%*
######
Sensitivity analyses.
Sensitivity analysis Screening attendance rate Eye examination rate "To do nothing" strategy Screening programme pathway Difference (€) Difference (%)
-------------------------- --------------------------- ---------------------- -------------------------- ----------------------------- ---------------- ----------------
Sensitivity analysis I 32% 75% €1,980,293.43 €811,332.32 −€1,168,961.11 −59.03%
Sensitivity analysis II 92% 75% €1,980,293.43 €1,933,104.55 −€47,188.87 −2.38%
Sensitivity analysis III 80% 85% €1,980,293.43 €1,789,070.03 −€191,223.39 −9.66%
Sensitivity analysis IV 80% 90% €1,980,293.43 €1,869,389.96 −€110,903.47 −5.60%
[^1]: Academic Editor: Tamer A. Macky
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction
===============
The ATP-binding cassette transporter A1 (ABCA1) acts as a vehicle for cellular cholesterol which after crossing cell membrane bounds to acceptor molecule such as apolipoprotein (apo) A \[[@B1]-[@B3]\]. Thus, ABCA1 influences the initial steps in high density lipoprotein (HDL) formation and in reverse cholesterol transport. The ABCA1 protein belongs to ABC proteins family, which are ingredients of biological membranes and use ATP to transfer various particles such as lipids \[[@B1]\]. The ABCA1 protein gene is located in the chromosome 9 in the area 9q31.1. This gene encodes a protein which is expressed in many tissues such as liver, macrophages, intestines, lungs etc. Several ABCA1 gene polymorphisms were identified, including rs2230806 (R219K) and rs2230808 (R1587K), which are mainly associated with the HDL cholesterol (HDL-C) concentration. The R219K results in a single amino acid change in codon 219 from arginine to lysine. The K allele of the R219K polymorphism has been related to low coronary artery disease (CAD) risk \[[@B4]\] and to lower triglycerides (TGs) concentration \[[@B5]\]. As far as concern the levels of HDL-C the reports are still confusing \[[@B4],[@B6]\]. The R1587K which is located in the extracellular loop of the ABCA1 protein, results in a single amino acid change in codon 1587. This polymorphism has been consistently associated with low HDL-C concentration \[[@B7],[@B8]\].
This study was undergone to evaluate the influence of these two ABCA1 gene polymorphisms on lipid profile \[total cholesterol, TGs, HDL-C and low density lipoprotein cholesterol (LDL-C)\] in young nurses. We also tested if there are any differences in frequency of ABCA1 gene polymorphisms between individuals with low and high HDL-C concentration.
2. Materials and methods
========================
Subjects
--------
The genotyping of 308 Greek female students aged 22.5 (±2.3) years who were attended to the University of Nursing of Technological and Educational Institution was performed. All students had no personal history of CAD and were not taking any drugs. Also, exclusion criteria were diabetes mellitus, thyroid and liver disease, high alcohol consumption, professional athleticism and any chronic disease.
All women were attended to the University every day and were staying for 8-10 hours. Women were eating at the school canteen which served typical Mediterranean food. Only one (evening) meal daily was most likely to be different in each student.
Additionally, subject were divided to those with high (HDL-C \>70 mg/dl) and low (HDL-C \<40 mg/dl) HDL-C concentration.
The University of Nursing of Technological and Educational Institution ethics committee approved the protocol of this study. All subjects signed an informed consent form.
Blood Chemistry
---------------
Plasma total cholesterol, TGs, HDL-C and apo A1 were measured using enzymatic colorimetric methods on Roche Integra Biochemical analyzer with commercially available kits (Roche). The serum LDL-C concentration was calculated using the Friedewald formula only in patients with TGs concentration \< 400 mg/dl.
DNA analysis and determination of blood lipids
----------------------------------------------
The ABCA1 gene polymorphisms (R219K and R1587K) were detected using polymerase chain reaction (PCR) and restricted fragment length polymorphism analysis (RFLP\'s). The PCR was performed using Taq polymerase KAPATaq. For R219K polymorphism the oligonucleotide primers which were used are AAAGACTTCAAGGACCCAGCTT and CCTCACATTCCGAAAGCATTA \[[@B9]\]. PCR was subjected to 95°C for 5 min, thirty cycles of 95°C for 30 s, 55°C for 30 s and 72°C for 30s and final extension to 72°C for 7 min, producing a fragment of 309 bp. This fragment was subsequently cleaved by EcoNI, creating fragments for R allele 309 bp and for K allele 184 bp and 125 bp, which were subjected to electrophoresis on an agarose gel 4% and visualized with ethidium bromide.
For R1587K polymorphism the oligonucleotide primers which were used are AAGATTTATGACAGGACTGGACACGA and TGAATGCCCCTGCCAACTTTAC \[[@B8]\]. PCR was subjected to 95°C for 5 min, thirty cycles of 95°C for 30 s, 60°C for 30 s and 72°C for 30s and final extension to 72°C for 7 min, producing a fragment of 139 bp. This fragment was subsequently cleaved by BssSI, creating fragments for R allele 117 bp and 22 bp and for K allele 139 bp, which were subjected to electrophoresis on an agarose gel 4% and visualized with ethidium bromide (Figure [1](#F1){ref-type="fig"}, Figure [2](#F2){ref-type="fig"}).
{#F1}
{#F2}
Statistical analysis
--------------------
The results are given as mean ± standard deviation (SD) or as median and interquartile range (IQR) according to normality of continuous variables. All qualitative variables are presented as absolute or relative frequencies. All biochemical variables were assessed for normality of distribution employing the Shapiro-Wilk test and non parametric statistical tests were used if appropriate. However, non parametric variables were initially normalized; TGLs by the log10 transformation - apoA1 and TC by square root transformation and parametric criteria were employed, all providing the same results as non parametric tests. However, TGL variable was extremely skewed and we decided to use the median \[IQR\] presentation and keep the result of the non parametric test that was employed.
Differences in lipid levels for the various genotypes were evaluated with one - way analysis of variance (ANOVA) or its non-parametric analogue Kruskal - Wallis H statistic. The Pearson\'s chi-square test was employed for the categorical variables.
All tests were two-tailed and statistical significance was established at 5% (p \< 0.05). Data were analyzed using Stata™ (Version 10.1 MP, Stata Corporation, College Station, TX 77845, USA).
3. Results
==========
Clinical and laboratory parameters
----------------------------------
Demographic data, clinical characteristics and lipid profile of the study cohort are shown in Table [1](#T1){ref-type="table"}. R and K allele frequencies appear to have equal distributions in both ABCA1 polymorphisms (p \> 0.05) (Table [2](#T2){ref-type="table"}). The frequencies of R219K genotypes were 50.97% for RR, 40.91% for RK and 8.12% for KK, whereas the frequencies of R1587K genotypes were 47.08% for RR, 41.56% for RK and 11.36% for KK. Both frequencies were found in Hardy-Weinberg equilibrium.
######
Characteristics of the study population.
*Demographic data* *Lipid profile(in mg/dl)*
-------------------------------- --------------------------- ------------------- -------------------
Number of subjects 308 Total Cholesterol 196.6 (59.7)
Age (ys) 22.5 (2.3) TGs 87 \[60.5 - 149\]
BMI (Kg/m^2^) 21.5 \[19.8 - 24.2\] HDL-C 69.2 (25.9)
Waist (cm) 87.0 (12.5) LDL-C 103.5 (38.8)
Apo A 152.8 (53.5)
***Clinical characteristics***
Smoking (yes/no) 110/176 (38.5%/61.5%)
HDL-C: high density lipoprotein cholesterol, LDL-C: low density lipoprotein cholesterol, ApoA1: Apolipoprotein A1, TGs: triglycerides. Data are expressed as mean ± standard deviation (SD) or as median and interquartile range (IQR) according to normality of continuous variables. Qualitative variables are presented as absolute and relative frequencies.
######
R and K allele frequencies according to R219K and R1587K polymorphisms.
ABCA1 R allele frequency K allele frequency p value\*
------------ -------------------- -------------------- -----------
**R219K** 0.72 0.28 0.11
**R1587K** 0.68 0.32
Fisher\'s exact test
R219K and R1587K polymorphisms
------------------------------
The distribution of R219K and R1587K polymorphisms was investigated according to Low HDL-C (n = 46) and High HDL-C concentration (n = 104). No statistical difference was observed in both polymorphisms when compared to Low vs High HDL-C concentration (p = 0.44 and p = 0.48, respectively). Moreover, no difference in the distribution of the R219K genotypes was detected according to lipid profile (Table [2](#T2){ref-type="table"}). Also, no differences in the distribution of K and R carriers of R219K polymorphism was detected according to lipid profile (p = 0.87).
The R1587K genotypes differed significantly according to total cholesterol, LDL-C and TGs concentration (p = 0.023, p = 0.014 and p = 0.047, respectively) (Table [3](#T3){ref-type="table"}). Significant difference in LDL-C concentration was detected between RK and RR genotypes of the same polymorphism (110.6 mg/dl vs 96.9 mg/dl, respectively, p = 0.004), Figure [3](#F3){ref-type="fig"}. However, the LDL-C concentration did not differ between women with the KK and RK genotype of the R1587K polymorphism (104.52 mg/dl vs 110.64 mg/dl, p = 0.4). Total cholesterol levels was higher in women with the RK compared to women with the RR genotype of the R1587K polymorphism (207.41 mg/dl vs 187.69 mg/dl, p = 0.006), whereas total cholesterol levels did not differ between women with RK and KK genotypes (207.41 mg/dl vs 193.66 mg/dl, p = 0.22). Finally, a significant difference was observed in the levels of TGs according to the R1587K polymorphism, with the RK genotype women having higher TGs concentration in comparison to the RR genotype (134.25 mg/dl vs 108.89 mg/dl, p= 0.014). However, TGs levels did not seem to differ between RK and KK genotypes of the same polymorphism (134.25 mg/dl vs 108.06 mg/dl, p= 0.11). Also, no differences in the distribution of K and R carriers of R1587K polymorphism was detected according to lipid profile (p = 0.5).
######
Blood lipid levels according to ABCA1 R1587K polymorphism in all genotypes.
*Lipid Profile (in mg/dl)* Genotype Mean SD P\*
---------------------------- -------------- ------------ ---------------- ------------
RR 187.69 59.19
***Total cholesterol*** RK 207.41 59.55 0.023
KK 193.65 57.49
RR 68.94 26.35
***HDL-C*** RK 69.89 25.33 0.88
KK 67.48 26.57
RR 96.97 38.48
***LDL-C*** RK 110.64 37.26 0.014
KK 104.52 41.51
RR 147.8 53.23
***ApoA1 (mg/dl)*** RK 158.46 54.33 0.26
KK 152.54 50.65
**Genotype** **Median** **IQR** **P**^*†*^
RR 83 \[56 - 123\]
***TGs*** RK 98 \[63.5 - 180\] 0.047
KK 79 \[61 - 134\]
HDL-C: high density lipoprotein cholesterol, LDL-C: low density lipoprotein cholesterol, ApoA1: Apolipoprotein A1, TGs: triglycerides. \*P values among genotypes from Anova test performance - ^†^P values among genotypes from Kruskal Wallis test performance.
{#F3}
4. Discussion
=============
We examined the probable impact of the ABCA1 polymorphisms as a genetic influence on lipid profile in Greek young nurses.
R219K gene polymorphism
-----------------------
The frequency of R allele of R219K polymorphism in our study was 72% similar to Pasdar et al \[[@B10]\] who reported 73.2% in controls (68% in patients with ischemic stroke) and Porchay et al \[[@B11]\] \[D.E.S.I.R participants (Data from an Epidemiological Study on the Insulin Resistance)\] who reported 71.7%. Conversely, Clee et al \[[@B5]\] reported frequency of 46% in Dutch men with proven CAD who participated in the Regression Growth Evaluation Statin Study.
Concerning lipid profile, the possibly influence of the R219K polymorphism is still evaluated. For example, Hodoğlugil et al in Turks individuals with low HDL-C concentration found the association of R219K polymorphism with HDL-C concentration \[[@B12]\]. Frikke- Schmidt et al in Danish population did not found any association between R219K polymorphism and individuals with Low or High HDL-C concentration \[[@B7]\]. On the other hand, Kakko et al in Finnish women \[[@B13]\] found the association of 219K allele with higher HDL-C concentration. Opposite, Clee et al \[[@B5]\] reported no differences according to HDL-C concentration between K or R allele, although carriers of K allele had significant lower TGs concentration in relation to carriers of R allele \[[@B5]\]. Furthermore, younger homozygotes of K allele had higher cholesterol efflux and HDL-C concentration compared to homozygotes of R allele. Sandhofer et al \[[@B14]\] studied male population and did not find any association of R219K polymorphism and lipid profile. Also, Cenarro et al \[[@B15]\] evaluated the R219K polymorphism in patients with familial hypercholesterolemia with and without premature CAD and reported that K allele was more frequent in subjects without premature CAD compared to individuals with premature CAD \[[@B15]\]. Li et al \[[@B16]\] investigated the relation of R219K polymorphism with the manifestation of CAD in Chinese patients and did not report any significant correlation. However, the TGs concentration were significant higher and HDL-C concentration significant lower in patients with RR genotype than those with KK genotype. Similarly, Delgado-Lista J et al \[[@B17]\] did not find association of R219K polymorphism and lipid profile.
According to our results from the young Greek female nurses (living and working in the similar conditions) we did not find any association with HDL-C concentration or other lipid parameters. Also, no association between R219K polymorphism and Low or High HDL-C subgroups was found although small number involved in these groups may be a limitation.
R1587K gene polymorphism
------------------------
The frequency of R allele of R1587K polymorphism in our study was 68%. Pasdar et al \[[@B10]\] reported 75.9% and Frikke-Schmidt reported 76% \[[@B18]\].
Also, Frikke-Schmidt et al \[[@B7]\] found that this polymorphism is overexpresed in individuals (men and women) with low HDL-C concentration. Furthermore, there was a gradual decrease in HDL-C levels about 0.07 mmol/l (2.7 mg/dl) for RK genotype and 0.11 mmol/l (4.2 mg/dl) for the RR genotype. Tregouet et al \[[@B19]\] stated that R1587K has impact on the apo A1 concentration. Wang et al \[[@B20]\] did not found any relation of R1587K with lipids levels in patients with type 2 diabetes mellitus who were treated with rozigliatoze. The study of Clee et al \[[@B5]\] in Danish population has shown that carriers of K allele had lower HDL-C concentration in comparison with the KK genotype. In a multi-analysis including age, BMI, smoking and TGs as independent variables, the R1587K polymorphism remained significant factor for the prediction of HDL-C concentration. Pasdar et al \[[@B10]\] found the association of R1587K polymorphism and apo A concentration and this was not related to CAD. Cohen et al \[[@B21]\] supported that rare alleles with major phenotypic effects contribute significantly to low HDL-C levels in the general population. However, Tupitsina et al \[[@B9]\] observed that in patients with CAD the R1587K polymorphism did not affect lipids levels. Slatter et al \[[@B22]\] investigated the prevalence of mutations and common SNPs in ABCA1 in 154 low HDL-C individuals and 102 high HDL-C individuals. The R1587K SNP was over represented in low HDL-C individuals. Ksiazek et al in a small study of 50 individuals reported a trend (p = 0.07) in terms of association between TGs concentration and R1587K genotype \[[@B23]\]. In our study no association between R1587K polymorphism and HDL-C levels was found. Also, no association between R1587K polymorphism and Low or High HDL-C subgroups was found. However, individuals with RK genotype had significantly higher TGs, total cholesterol and LDL-C concentration compared to RR genotype.
Reduced circulating HDL-C can be caused by either genetic and/or environmental factors (sedentary lifestyle, diabetes mellitus, smoking, obesity or a diet enriched in carbohydrates). The potential mechanisms of interaction between genetic variations and phenotypes contribution are not fully understood. This happens because each study involved different studying population and the environmental interactions could not be ruled out. Thus, in our study the environmental influence was partially diminished. The advantage of our study was that the study cohort was almost homogenous, since the nurses most of the time were following the same day to day program and were eating in the same school canteen. Thus the influence of diet or physical activities were unlikely, which may partially explain the lack of association of R219K or R1587K polymorphisms and HDL-C concentration. However, the influence of smoking cannot be ruled out since 38.5% of students were smokers.
It is well known from clinical trials, that 1% change in LDL-C is associated with a 1% change of cardiovascular events, which means that individuals born with favorable genotype (in this case RR genotype of R1587K) have already some advantage compare to less favorable genotypes, since they have lower LDL-C. At that time, this was only a clinical observation. Hopefully in near future we will be able to identify high risk patients according to genetic testing, which the assessment in these days has same cost, time implications and replication problems in independent studies, which are disadvantages in routine clinical practice. Nevertheless, these limitations may become less relevant as technology develops.
In summary, the R1587K polymorphism of ABCA1 gene was associated with altered lipid levels in Greek young nurses. Women with RK genotype had higher TGs, total and LDL-C concentration compared to RR genotype. These observations may be significant in assessing the risk of CAD since a 1% change in LDL-C is associated with a 1% change of cardiovascular events and TGs concentration were documented to play a significant role in women.
Competing interests
===================
The authors declare that they have no competing interests.
Authors\' contributions
=======================
VK participated in the development of hypothesis, drafting of the manuscript and carried out the genetic analysis, GK conceived the study and participated in the development of the hypothesis, the study design and drafting of the manuscript, AM participated in the molecular genetic studies, AK performed the statistical analysis and drafting of the manuscript, GV and AK collected the blood samples, DD participated in revising the manuscript critically for important intellectual content, CM and CD participated in the study design and its coordination. All authors read and approved the final manuscript.
| {
"pile_set_name": "PubMed Central"
} |
To the Editor,
I read with great interest the paper by Temiz et al. ([@ref1]) entitled "Effects of cinacalcet treatment on QT interval in hemodialysis patients" published as Epub ahead of print for The Anatolian Journal of Cardiology 2015. They aimed to evaluate the effects of a calcimimetic drug (cinacalcet) on corrected QT values (QTc) in patients with end-stage renal disease (ESRD). They found a prolongation of QTc value compared with baseline QTc value after cinacalcet treatment. I have a few comments.
QTc interval is the time from beginning of QRS to the end of T wave. In other words, it consists of depolarization and repolarization phases of cardiac tissue. Prolongation of QTc represents delayed cardiac repolarization and can be related to ventricular arrhythmias and sudden cardiac death ([@ref2], [@ref3]). Although it is well-known that QTc prolongation is common in patients with ESRD, the exact mechanism of this cardiac repolarization abnormality has not been established ([@ref2]--[@ref5]). Hemodialysis is one reason QTc interval may be affected. Researchers have demonstrated increased QTc intervals in patients with ESRD, especially after end of hemodialysis. This is largely attributed to rapid changes in plasma electrolyte levels during hemodialysis ([@ref4]). Therefore, it is crucial exactly when electrocardiography (ECG) is performed. In the study by Temiz et al. ([@ref1]), it is possible that the timing of ECG may have influenced measurement of QTc.
Electrolyte disturbances are common in patients with ESRD and can cause changes in cardiac ionic polarization, resulting in altered QTc interval ([@ref2]--[@ref5]). Foglia et al. ([@ref5]) determined QT prolongation in patients with primary renal hypokalemia-hypomagnesemia, and demonstrated that decreased levels of potassium and magnesium could alter duration of action potential in cardiac cell membrane. In the study by Temiz et al. ([@ref1]), it may be helpful to present electrolyte levels of the patients at the time of ECG whether electrolytes have influential effects on QTc measurements or not.
Finally, measurement of QTc interval has some technical difficulties such as presence of U waves or inverted T waves and intraobserver variability, which are not mentioned clearly in the study by Temiz et al. ([@ref1]) and can affect the precise measurement of QTc value ([@ref2]--[@ref5]).
In conclusion, I think that this study would be stronger with these additional data mentioned above and we can easily understand the role of cinacalcet on QTc interval in patients with ESRD.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-micromachines-11-00622}
===============
The limitations of traditional computer architectures have become explicit as the industry reaches to the end of Dennard scaling \[[@B1-micromachines-11-00622]\] and Moore's law \[[@B2-micromachines-11-00622]\] where the data movement is dominated over both overall system energy and performance. More than 90% of the energy consumed by an instruction is spent on memory access \[[@B3-micromachines-11-00622]\]. Considering the current station in which 90% of the overall data has been produced in the last two years, which corresponds to a 9× increase in the total amount \[[@B4-micromachines-11-00622]\], the computer architectures responsible in the processing of all these data must be optimized in terms of data handling methodology. Certainly, the most important domain that needs such a massive amount of data is signal processing. The emergence of artificial intelligence (AI) and big data has dramatically increased the importance of signal processing since the raw data must be processed to obtain better accuracy achievement. On the other hand, there is no corresponding development in computer architectures to handle such an enormous amount of data at the same rate. Excessive increase in the amount of data to be processed and increasing complexity of computational tasks force the researchers towards more data-centric architectures rather than today's processor-centric ones. One such application that highly requires data-centric computational platforms is stencil codes that are used in many computational domains \[[@B5-micromachines-11-00622],[@B6-micromachines-11-00622]\].
The bottleneck of the current systems is generally caused by the communication between processor and memory. The memory systems cannot supply the data to the processor at the required processing rate. Moreover, the energy consumption spent on data access is an order of magnitude higher than the computation cost due to the out of chip access \[[@B3-micromachines-11-00622],[@B7-micromachines-11-00622]\]. The ideal solution is combining processor and memory at the same location to alleviate the limited connection link between them. For this reason, there are recently many research attempts aiming either bringing the processor near the memory (i.e., near-memory computing) \[[@B8-micromachines-11-00622]\] or integrating them (i.e., in-memory computing) \[[@B9-micromachines-11-00622],[@B10-micromachines-11-00622]\]. In-memory computation architectures are very diverse, ranging from analog computation by using the non-volatile memories \[[@B11-micromachines-11-00622],[@B12-micromachines-11-00622],[@B13-micromachines-11-00622],[@B14-micromachines-11-00622]\] through the in-DRAM processing between the DRAM rows \[[@B15-micromachines-11-00622]\]. Among them, associative processors (APs) propose an applicable solution that performs the noise-free digital computation through the binary memory devices (e.g., memristor, SRAM, STT-RAM) \[[@B16-micromachines-11-00622],[@B17-micromachines-11-00622]\]. Associative processors can be considered as a type of single instruction multiple data (SIMD) processor that combines the functionalities of processor and memory in the same location \[[@B16-micromachines-11-00622]\]. In AP, the operations are performed directly on the data residing in memory without moving them. Each memory row behaves as an individual processor together with its own special set of registers. Since an operation can be performed on all memory words in parallel, the execution time of operations does not depend on the vector size. This feature solves the memory-wall problem of traditional von Neumann architectures since there is no inter-dependence between memory and processor \[[@B18-micromachines-11-00622]\]. Even though the inherent latency of associative processors is much higher than the traditional architectures, it can result in better throughput and energy efficiency if the required degree of parallelism is demonstrated by the application \[[@B19-micromachines-11-00622]\]. In applications characterized by data parallelism, associative processors (APs) accomplish a remarkable acceleration \[[@B20-micromachines-11-00622]\], and can be employed as an accelerator near the main processor \[[@B21-micromachines-11-00622]\].
Stencil codes are a class of iterative kernels which update a given array (generally 2D or 3D) with respect to a specific pattern \[[@B22-micromachines-11-00622]\]. This pattern is called as a stencil. The code performs a sequence of iterations through a given array. In each iteration, all the elements of the arrays (i.e., cells) are updated. Stencil computations are highly used in the scientific computation domain for many purposes, including image processing, solving differential equations, computational fluid dynamics simulations (e.g., weather prediction), etc. Due to its importance, there are many studies in the literature that aims to propose an efficient architecture implementation for stencil codes \[[@B23-micromachines-11-00622],[@B24-micromachines-11-00622],[@B25-micromachines-11-00622]\]. Most of the studies are headed towards to field-programmable gate arrays (FPGAs) or graphical processing units (GPUs) based implementations since traditional central processing unit (CPU)-based solutions cannot fulfill the parallel processing requirements. As an example, the study in \[[@B26-micromachines-11-00622]\] proposes a GPU-based 2D stencil implementation using CUDA. The implementation exploits the multi-threading and optimizes the shared memory usage in GPUs. In \[[@B27-micromachines-11-00622]\], OpenCL implementation of four 3D stencil computations is proposed for GPU architectures, which exhibits superior performance than CUDA-based alternatives. In \[[@B28-micromachines-11-00622]\], a multi-core CPU based implementation is proposed together with the corresponding software optimization. In \[[@B29-micromachines-11-00622]\], an OpenCL-based FPGA implementation of some stencil codes is proposed in which a high-level synthesis language is used to generate the stencil codes. Similarly, in \[[@B30-micromachines-11-00622]\], a custom optimized high-level synthesis flow is presented for both area and throughput optimization. The FPGA-based approaches can be considered as near-memory architecture where the memory bottleneck problem is mitigated through the distributed internal memory inside the FPGA fabric. The study in \[[@B31-micromachines-11-00622]\] proposes a multi FPGA-based stencil implementation. The study in \[[@B32-micromachines-11-00622]\] proposes a parameterizable, generic VHDL template for parallel 2D stencil code applications on FPGAs instead of high-level synthesis solutions. In FPGA-based solutions, the performance is limited by both memory bandwidth and the amount of internal memory and logical resources inside the FPGA. After reaching their limits, increasing the parallelism does not increase the performance. The same rule also applies to GPU and CPU based implementations as well. Therefore these architectures limit the degree of parallelism to the number of cores that can be fit in a given chip area and available energy budgets. Considering the case that size and quality of the data are increasing rapidly, it is obvious that there is a need for more efficient domain-specific processor architectures to manage an enormous amount of data for stencil codes as pointed by the computational trends for beyond the Moore's Law and Dennard Scaling \[[@B33-micromachines-11-00622]\].
The stencil computation generally requires basic operational complexity (i.e., a sum of weighted products), but large external memory bandwidth \[[@B26-micromachines-11-00622],[@B29-micromachines-11-00622]\]. This is due to that it requires a number of accesses to the memory while updating each point. Therefore, most implementations of stencil code on traditional architectures suffer from bandwidth limitations \[[@B26-micromachines-11-00622],[@B34-micromachines-11-00622]\]. As a promising solution, associative in-memory processors take advantage of content addressable memories, which provides an area-efficient, in-memory processing solution by integrating the computation and storage. In in-memory solutions, memory bandwidth can be considered as the amount of whole memory. For this reason, this study proposes a 2D stencil kernel architecture based on associative in-memory processing to eliminate the memory bottleneck. The study shows the two implementations by using both SRAMs and memristors. Since stencil codes are memory bound (i.e., the ratio of memory access to computation is high), APs provide a good processing environment for them. Furthermore, the implementation provides a considerable amount of energy savings and speedups in the system through approximate computing at some reasonable level. The rest of the study is organized as follows: In the following section, the background knowledge of both associative processors and stencil codes is presented. [Section 3](#sec3-micromachines-11-00622){ref-type="sec"} introduces the proposed accelerator architecture in detail. Experimentation and evaluation results are discussed in [Section 4](#sec4-micromachines-11-00622){ref-type="sec"}. The final section concludes the work.
2. Background {#sec2-micromachines-11-00622}
=============
2.1. Associate Processor {#sec2dot1-micromachines-11-00622}
------------------------
Almost all computer architectures use traditional Boolean logic to perform logical and arithmetic operations. On the other hand, there are many other techniques as well to perform the operations non traditionally. Associative computing is one of them that exploits the associativity principles of memories for logical and arithmetic computations. The architecture of an associative processor (AP) is presented in [Figure 1](#micromachines-11-00622-f001){ref-type="fig"}, which consists of a content addressable memory (CAM), controller, interconnection circuit, and some specific registers (key, mask, and tag). The CAM stores the data on which operations are performed. The $key$ and $mask$ registers are used to search a given data inside the specified columns of the CAM. The key register keeps the data to be searched inside the CAM. The $mask$ register points the specified column locations. The $tag$ registers are used to keep track of row locations that have the searched data. Therefore, each row has its own single bit tag register even though mask and key registers are common for all CAM rows. The controller generates the instructions (key and mask pairs) for the corresponding operation (e.g., addition, subtraction, etc.) and checks the tag bits to carry on the operations. The rows tagged with logic-1 means that the corresponding CAM row has been matched with the given key and mask value. For example, if the key is set as 101 and mask as 011, the tag bits of the corresponding rows whose first and seconds bits are logic-1 and logic-0 respectively become logic-1. The third bit is not searched for logic-0 since its corresponding mask bit is logic-0 (i.e., not activated). The interconnection matrix is a basic circuit-switched matrix which used to communicate with other APs as column parallel fashion. The architecture can also incorporate low-power mechanisms such as selective compare, where within a lookup table (LUT) pass, the matched rows are not precharged again since it is not possible to get another match in this row \[[@B35-micromachines-11-00622]\]. As the most important part of the CAM array, the cells can either be implemented by traditional SRAM memory or alternatively by emerging non-volatile memories such as memristor (ReRAM) or STT-RAM. This study shows the two implementation candidates for APs, which are SRAM-based and ReRAM-based. [Figure 1](#micromachines-11-00622-f001){ref-type="fig"} shows the corresponding cell implementations. The traditional NOR-type CAM cell is used for SRAM-based implementation \[[@B36-micromachines-11-00622]\]. ReRAM-based implementations exploit the two-transistor, two-memristor ternary CAM cell structure, as studied in \[[@B37-micromachines-11-00622]\]. In both of the implementations, the functionally is performed exactly as same, but there are some trade-offs between them. For example, ReRAM-based implementation minimizes the static power consumption since the cells are non-volatile \[[@B38-micromachines-11-00622]\], but requires higher energy consumption during the write operation. On the other hand, SRAM-based implementation suffers from static energy consumption which becomes more severe as process technology improves, but provides low-cost write and less delay.
In traditional processor architectures, the data are sent over the functionality. In other words, the data are read from the main memory and sent to the processor to perform the operations on them. On the contrary, in associative processing, the operands stay inside the processor (i.e., in memory), and the functionality is sent over the data. Therefore, operations are performed inside the memory as in-place without moving them. An operation on AP is carried out by consecutive $compare$ and $write$ phases. During the compare phase, the content is selected inside the memory, and in the write phase, the corresponding functionality is applied to the selected rows which hold the corresponding data. Depending on the desired arithmetic operation, the controller sets the mask and key values by referencing a lookup table (LUT) for compare and write operations. The following example clarifies the in-place addition operation on AP.
[Figure 2](#micromachines-11-00622-f002){ref-type="fig"} illustrates the complete flow for in-place addition of two $4 \times 1$ 2-bit signed vectors, A and B, i.e., $\left. B\left\lbrack i \right\rbrack\leftarrow B\left\lbrack i \right\rbrack + A\left\lbrack i \right\rbrack,i = 0\ldots 3 \right.$, where the tables in the first row correspond to in-place addition LUT, and the others show the progress in the CAM content together with the key/mask values and the tag status. Initially, A contains (i.e., columns 1-0) the values of \[1; −1; 1; −2\] and B (i.e., columns 3-2) contains the values of \[0; −2; 1; −2\] in binary 2's complement. Cr (carry) column (i.e., column 4) is initially all 0 s. In LUT, the highlighted entry shows the applied key on the masked columns. Each entry corresponds to a combination of different Cr, B, A. Even though there is a total of eight ($2^{3}$) combinations, only four of them are used since others have no effect on the operation \[[@B35-micromachines-11-00622]\]. In each CAM of the figure, the key value from the compare column of the LUT is searched in the masked columns of the CAM. The arrows specify the flow of the operation. In the first row, partial addition operation is performed on the first bits of A and B while Cr holds the carry. Therefore, the mask register is set as logic-1 for Cr and the first columns of A and B. The second row similarly corresponds to addition on second bits. After each comparison, the matching rows are tagged with logic-1, as indicated in its vertical tag register. Then, corresponding LUT entry (shown in the write column of the LUT) is written only to the masked cells in the rows whose tag register is logic-1. For example, in the first table, "011" is searched in Cr, B, and A columns, respectively. The third-row matches by indicating a logic-1 in its tag register. As a result, logic-1 is written to the Cr column, and logic-0 is written for the B column. Normally, this operation represents a full adder for the combination of 0 + 1 + 1, where the result is logic-0 and carry is logic-1 ("10" as together). By applying all combinations of inputs on each bit locations, column-wise full addition is performed. The process is repeated for all the passes in the prescribed order shown in [Figure 2](#micromachines-11-00622-f002){ref-type="fig"}. Finally, the value stored in Cr and B becomes \[1; −3; 2; −4\] which is equal to B+A (i.e., \[0 + 1; −2 + −1; 1 + 1; −2 + −2\]). In general, adding two vectors that are *m*-bit wide takes $8m$ cycles ($4m$ compares and $4m$ writes), independently of the vector size. Considering the case of huge vector operations, in-memory associative processing eliminates the memory access costs and provides great performance advantage by its SIMD-like processing on each memory row.
2.2. Stencil Codes {#sec2dot2-micromachines-11-00622}
------------------
As introduced in the introduction, stencil codes are basic computational kernels that update an input array by following a specific pattern and mathematical equation. This update is performed over the whole array iteratively until a degree of convergence is obtained (e.g., a dependable weather prediction). The most common stencil type is the Laplace equation in which a cell is updated with respect to the average of its four neighboring cells. If the cell itself also included in averaging, the stencil code is named 5-point Jacobi iteration. The other stencil types are named as 7-point, 9-point, and 25-point, both on 2D or 3D data, which provides a weight for each cell by including more cells to the computation, respectively. Even though the figure presents three different types, there are many different types of kernels which perform different operations by following a different pattern like finite-difference time-domain (FDTD) stencil \[[@B39-micromachines-11-00622]\]. Depending on the shape of neighborhood cells, a different data processing application is obtained. In this way, the stencil codes can also be used for signal processing, especially on 2D image data. [Figure 3](#micromachines-11-00622-f003){ref-type="fig"} shows the three different stencil types with the visualization of their computation patterns and equations.
Even though computation seems trivial for stencil applications, memory bottleneck becomes a big problem since the computation is tightly coupled to the memory. Most stencil codes are categorized as memory-bound \[[@B29-micromachines-11-00622]\]; therefore, they suffer from memory bottleneck. Therefore, an efficient parallel implementation becomes very crucial. For this reason, GPUs are employed as the best processing environment until now rather than CPUs \[[@B5-micromachines-11-00622],[@B26-micromachines-11-00622]\]. The reason is that even though GPUs have simpler and slower processing cores than CPUs, they can provide a high throughput to process such a huge amount of data since GPU cores have more memory bandwidth. When compared to GPUs, APs have much simpler cores, and its core density is huge (i.e., one memory row is like a processing core that can handle basic stencil computation) as presented in [Section 2.1](#sec2dot1-micromachines-11-00622){ref-type="sec"}. Furthermore, AP performs the operations on the data directly, which virtually boosts memory bandwidth to memory size. At that point, a truly in-memory implementation of these applications on APs can provide more benefits than GPU-based implementations. Many studies in the literature prove that AP-based implementations of data-intensive applications have superior performance than the traditional correspondences \[[@B20-micromachines-11-00622],[@B40-micromachines-11-00622],[@B41-micromachines-11-00622],[@B42-micromachines-11-00622],[@B43-micromachines-11-00622]\], including the applications that has processing flow similar to stencil codes like fast Fourier Transform (FFT) \[[@B44-micromachines-11-00622]\]. Therefore, it is obvious that another memory-bound application of the stencil code can get a benefit, which is the main idea of this study. The following section presents the implementation in the AP in detail.
3. Accelerator Architecture for 2D Stencils {#sec3-micromachines-11-00622}
===========================================
[Figure 4](#micromachines-11-00622-f004){ref-type="fig"} shows the proposed pipelined implementation of a 2D stencil (Laplace) in three AP stages where the data is transferred through the fixed interconnections between the APs. Each pipeline stage in the architecture performs the multiplication and addition operations with the corresponding neighboring cells and steers the data to the next stage. Since the communication pattern between the stages is known before, a fixed pattern can be defined in the circuit instead of having a configurable communication switch for which the area and energy costs are higher than the CAM array itself \[[@B19-micromachines-11-00622]\]. The pattern is the same for all three stencil types evaluated in this study ([Figure 3](#micromachines-11-00622-f003){ref-type="fig"}).
In order to perform a stencil kernel on 2D data, data are sent to the accelerator as column-wise. This communication between the external memory (generally a DRAM) and AP can be handled by high speed dedicated buses so that CPU cycles are not wasted during the transmission. Each time, one column of the 2D data is placed to the AP sequentially starting from the first row. On the AP accelerator, the first stage keeps a three-column window inside to perform the stencil operation. The second and third stages also perform the addition operation between upper and lower neighboring cells together with the averaging operation to compute the final results as column-wise. For 5-point and 9-point stencil, these stages can also perform the multiplication operations with weights. On the other hand, the weight is generally set to get the average of these neighbor pixels. In that case, the operation can be converted to the sum of products, and multiplication operations can be performed in the last stage to get the average. In this case, a constant multiplication operation can be performed on all the rows for the faster and energy-efficient alternative. Compared to traditional CPU or GPU architectures, APs minimizes the memory access since data is moved once to the accelerator. Then the results are written back to the memory after an iteration. On the other hand, traditional architectures need to access memory whenever a cache miss occurs, so this leads to a huge number of circulation between the memory and the cache.
4. Evaluation {#sec4-micromachines-11-00622}
=============
In the evaluation of proposed accelerator architectures, a cycle-accurate AP simulator is used, which can realistically perform the circuit simulations on Synopsys HSPICE in an iterative manner. For the transistors, 65nm predictive technology models are used \[[@B45-micromachines-11-00622]\]. For the memristor, a fabricated nano-second switching time device is referenced in the ReRAM-based AP architecture which has a size of 50 nm. Its corresponding SPICE model in \[[@B46-micromachines-11-00622]\] is used in the simulations. Since the AP supports fixed-point computation, data bitwidth is set as 32-bit. All data moving costs to the accelerator are taken into account as well as computation. For the stencil types, the multiplication constants are selected as equal (i.e., to perform the averaging operation). On the other hand, the architecture can support any type of numerical weights. In the circuit implementation, a CAM buffer is added in the first stage to increase the throughput so that during data movement, the computation can also be performed concurrently. Therefore, the total architecture consists of three computational stages and one buffer to receive the DRAM data. The following three subsections provide the details of the evaluation.
4.1. Fixed-Point Computation {#sec4dot1-micromachines-11-00622}
----------------------------
Due to the energy and performance issues of the traditional computer architectures performing on floating point, there is a trend towards using fixed-point architectures for the sake of performance and energy in the applications that can tailor some degree of inaccuracy, especially in the field of artificial intelligence and signal processing. Even some recent GPUs proposes a configurable precision architecture that can both perform operations on floating-point as well as fixed-point by delivering higher operations/second \[[@B47-micromachines-11-00622]\]. The stencil codes can also be evaluated under this class of applications, which can get benefit from the fixed-point computation. To evaluate this opportunity, a set of simulations were carried out in both floating-point (64-bit) and fixed point (32-bit). For the sake of simulation time, which takes more than one week for $256 \times 256$ matrix sizes, only $64 \times 64$ matrices were evaluated for three different stencil codes. Even though most FPGA/GPU-based stencil applications in the literature use floating-point arithmetic, our simulation results reported that 32-bit fixed-point calculation gave almost identical results to the 32-bit floating-point since the data were kept within a limited range during the stencil iteration (i.e., a kernel update includes averaging at the end). [Figure 5](#micromachines-11-00622-f005){ref-type="fig"} shows the peak signal-to-noise ratios (PSNRs) of three stencil codes over the iterations where the PSNR was computed with respect to double-precision (64-bit) floating-point. The value is computed as ${PSNR} = 10\,\cdot\,\log_{10}\left( {peakval}^{2}/{MSE} \right)$, where the $peakval$ is 1 for the normalized 64-bit floating-point numbers, and $MSE$ corresponds to the mean-squared error between the fixed-point and floating-point results. Overall, the computation yields a high SNR rate of more than 100 dB. According to the results, the difference between the two computations was slightly increased over iterations (i.e., PSNR value decreased). On the other hand, the PSNR was settled down to its minimal value after some number of iterations. The results were very reasonable because the data kept in a limited range during the stencil. As an example, during weather prediction, the temperature range of the weather is generally limited in a range.
4.2. Comparison of Performance {#sec4dot2-micromachines-11-00622}
------------------------------
The performance of the accelerator depends on some factors. If the array size is assumed as nxm where n is the number of rows and m is the number of columns, the run-time can be formalized as $\max\left( {n \times t_{write},m \times t_{comp}} \right)$ where $t_{write}$ is the write speed to the CAM while reading data from DRAM and $t_{comp}$ is the total time that the slowest stage can finish its computation. As long as the array fits into the CAM as row-wise, $t_{comp}$ does not depend on the number of rows. It only depends on the bitwidth of the operands, and the next subsection on approximate computing presents the results on the effect of bitwidth. As stated in the previous section in which 32-bit fixed-point representation is enough for accurate computation, it is used to represent the numbers. [Figure 6](#micromachines-11-00622-f006){ref-type="fig"} shows the run time results of three stencil codes with variable matrix sizes on the SRAM-based architecture. Compared to SRAM-based architecture, ReRAM-based architecture had the same compare time, on the other hand, ReRAM requires two cycles for a write operation, and each write to ReRAM takes around one ns for the used memristor model. Therefore, ReRAM write operation was 4× slower than SRAM write. For this reason, ReRAM-based architecture was 50% slower. On the other hand, ReRAM-based implementation provided a 66% better area utilization in the memory area compared to SRAM-based cells since it was very compact and consists of two transistors and two memristors, where the memristor had a size of 50 nm. Furthermore, memristor-based implementation can facilitate probabilistic computing through its inherent stochasticity, which is a potential advantage over the traditional technology \[[@B48-micromachines-11-00622]\]. It is highly possible that it will come to prominence in the near future as the dark silicon area becomes more obvious \[[@B1-micromachines-11-00622]\]. In the results, the matrix size was selected as 4096 × m, where m changed between $2^{10}$ and $2^{16}$. As seen in the figure, Laplace transform took the least time since there was no required multiplication operation since the multiplication by 0.25 could be easily handled by shifting the point location in the number representation. For this reason, the data movement time dominated over the computation time. The 5-point and 9-point stencils gave almost the same results; however, the weight representation of 9-point stencil allowed faster multiplication since its binary representation had one less logic-1 compared to 5-point stencil. For them, the computation time dominated over communication time. On the other hand, if the total number of rows of the matrix exceeded 8 K, the communication cost dominated over computation, as presented in [Figure 7](#micromachines-11-00622-f007){ref-type="fig"}.
4.3. Approximate Stencil Computing {#sec4dot3-micromachines-11-00622}
----------------------------------
Approximate computing is another promising approach for energy-efficient digital system designs, especially for error-tolerant applications like signal processing in the multimedia domain or neural networks \[[@B49-micromachines-11-00622]\]. In this approach, the accuracy requirement of the system is sacrificed at an acceptable level for the sake of performance and energy gains \[[@B50-micromachines-11-00622]\].
As stated in [Section 2.1](#sec2dot1-micromachines-11-00622){ref-type="sec"}, an arithmetic operation can be started with any of the bits by disregarding their remaining right bits and go through the most significant bits since all operations are performed as bit-wise in the AP. For this reason, the associative computing provides a natural way of bit-wise dynamic approximate computing. Approximate computing is highly demanded, especially for signal processing applications to trade-off the accuracy for the sake of energy consumption and performance. In order to witness the effect of approximate in-memory computing on 2D stencil codes, the proposed accelerator was simulated under variable bit widths. [Figure 8](#micromachines-11-00622-f008){ref-type="fig"} shows the accuracy (i.e., similarity index) vs. speedup results with changing the number of bits. According to the results, 2.56× speedup was possible with an accuracy degradation of less than 1% when the bit width of operands was set to 20-bit instead of 32-bit. The speedup, in turn, provided more than 50% reduction in total energy consumption. This situation provided a perfect opportunity for edge devices at which power consumption was crucial. Compared to the traditional implementation of stencil codes on GPUs and FPGAs, the APs provided finer-grain reconfigurability for approximate computing.
5. Conclusions {#sec5-micromachines-11-00622}
==============
This study shows a step towards solving the bottleneck problem in stencil applications through in-memory associative processing. The methodology mainly proposes combining the memory and CPU in the same place and exploiting each memory row as an individual CPU. To demonstrate this, a 2D stencil kernel is implemented in associate processors, and a comparison is made between the different stencil implementations. The results show that AP can provide an advantage for huge data amounts. Furthermore, the proposed methodology allows for bit-wise dynamic approximate computing, which is useful for signal processing applications. According to the results, the approximation at some reasonable level provides a considerable amount of energy savings and speedup in the system. Although the study focuses on stencil applications, it can be generalized to other signal and image processing applications on a massive amount of data such as convolution, filtering (edge detection, finite impulse response, etc.), and Fourier transform.
Conceptualization, H.E.Y.; investigation, H.E.Y.; methodology, H.E.Y. and K.N.S.; project administration, K.N.S.; software, H.E.Y.; supervision, K.N.S. and A.M.E.; validation, H.E.Y.; writing---original draft, H.E.Y. and A.M.E. All authors have read and agreed to the published version of the manuscript.
This research was supported by King Abdullah University of Science and Technology (KAUST) AI Initiative.
The authors declare no conflict of interest.
The following abbreviations are used in this manuscript: AIArtificial IntelligenceCPUCentral Processing UnitGPUGraphical Processing UnitAPAssciative ProcessorCAMContent Addressable MemoryFFTFast Fourier TransformFPGAField Programmable Gate ArraysPSNRPeak Signal-to-Noise RatioMSEMean-squared Error
{#micromachines-11-00622-f001}
{#micromachines-11-00622-f002}
{#micromachines-11-00622-f003}
{#micromachines-11-00622-f004}
{#micromachines-11-00622-f005}
{#micromachines-11-00622-f006}
{#micromachines-11-00622-f007}
{#micromachines-11-00622-f008}
| {
"pile_set_name": "PubMed Central"
} |
Research in contextEvidence before this studyEar and mastoid disease is a common disease, which demands early and appropriate diagnosis with otoscopy or otoendoscopy, but is not trivial in local clinics and the diagnosis rate even by otolaryngologists using ear images show an unsatisfactory accuracy, as low as 73%. So far, the best known study for automatic diagnosis of ear disease using images has been done with tympanic membrane using a shallow neural network of relatively small data size (n \~ 390) with an accuracy of 86.84%, however, the previous method is only capable of partially diagnosing middle ear disease.Added value of this studyThis is the first study to utilize a deep learning scheme to classify tympanic membrane otoendoscopic images into six diagnostic categories, especially including attic retractions and tumors, using a large database (*n* = 10,544), and the deep learning model covers most of the ear diseases in the clinic, not only on the middle ear but also on the external ear. It also deals with an unstandardized clinical image set as-is without image quality control, which makes the current system adaptable to the real-world clinical setting. The ensemble classifier, which we propose, shows better performance than using a single transferred deep learning model with an accuracy of 93·67%.Implications of all the available evidenceAccording to our evaluation on the relationship between database size and the performance of the transfer deep learning models, current study suggests the need for a sufficient size of the database for a reliable classification performance in the medical image domain.Due to the high accuracy and the diagnostic coverage in the proposed model, clinicians with less experience in otoendoscopy, or other specialty physicians such as pediatricians, emergency, or family medicine doctors could be benefitted from the model and thus it may result in alleviating the burden of the growing number of patients with hearing impairment.Alt-text: Unlabelled Box
1. Introduction {#s0020}
===============
Ear and mastoid disease (International Statistical Classification of Diseases and Related Health Problems (ICD) code H.60-H.95) is a common disease that can easily be treated with early medical care. Nevertheless, if one does not receive timely detection and appropriate treatment, it may leave sequelae, such as hearing impairment. In the evaluation of ear and mastoid disease in the clinic, physical examination using conventional otoscopy or otoendoscopy as well as history taking is the first step. However, diagnosis by non-otolaryngologists using otoscopy or otoendoscopy is highly susceptible to misdiagnosis \[[@bb0005]\]. In a study by Pichichero, Poole \[[@bb0010]\], the correct diagnosis rate of otitis media diagnosed by 514 pediatricians using pneumatic otoscope was an average rate of 50%. The study also shows a higher (compared to pediatricians) but not a satisfactory accuracy of 73% when diagnosed by 188 otolaryngologists. This low diagnostic accuracy implies that diagnosis of ear disease without the help of additional resources such as imaging or acoustic testing is difficult even for specialists. The short of specialists in the local clinic and their relatively low diagnostic accuracy calls for a new way of diagnostic strategy, in which machine learning may play a significant role.
As far as we know, relatively few machine learning studies have been conducted for automated diagnosis of ear disease using otoscopic images. Myburgh and colleagues reported auto-diagnosis of otitis media, with an accuracy of 81·58% by decision tree and 86·84% by neural network method \[[@bb0015]\], which conducted a classification of tympanic membrane into five groups between normal eardrum, otitis media with perforation, acute otitis media, otitis media with effusion and cerumen impaction. However, the classification categories lack important and critical diagnosis such as attic retraction.
For clinical use, the current study is conducted to provide a reliable diagnosis of otitis media, attic retraction, atelectasis, tumors, and otitis externa, using deep learning for otoscopy photos of the eardrum and external auditory canal (EAC). These categories cover most of the domain of ear diseases that could be diagnosed using otoendoscopy in the clinics. For this, we proposed an ensemble classifier of two best-performing deep neural networks evaluated for ear images.
Deep learning or deep neural network has been introduced to various fields of medicine successfully. For example, in the field of ophthalmology, the machine learning result is comparable to a level of specialist \[[@bb0020], [@bb0025], [@bb0030]\]. Most of these studies utilize convolutional neural network (CNN), a supervised deep learning method. However, building CNN from scratch requires a large amount of dataset and computational power, which is not practical in many application areas. Instead, public CNN models pretrained for natural images could be reused and fine-tuned to a specific application, which is called transfer learning. In transfer learning, most network layers in a public network model are transferred to a new model, followed by a new fully-connected layer that classifies those features into a new set of classes. Studies with transfer learning for medical imaging showed high classification accuracy comparable to, or even better to building CNN from scratch \[[@bb0035],[@bb0040]\].
This study is composed of the following three main parts. First, we evaluated the performance of nine public models to choose the best models in terms of accuracy and training time for the current application. Based on this evaluation, ensemble classifier to combine multiple models\' classification results was proposed, which is expected to increase the overall classification performance than using a single classifier. Second, although transfer learning is known to be efficient in a relatively small dataset (as in labelled medical images), the dependency of the classification accuracy and model type on the size of the dataset is not exampled yet. Thus, we tested the performance of the classifier depending on the data size. We also conducted optimization of the model configuration, by assigning a hidden layer in the fully connected network layer, and changing colour channels in the image database. Finally, we showed and discussed the characteristics of the proposed model for diagnosing ear diseases in the clinical setting.
2. Materials and methods {#s0025}
========================
2.1. Patient selection and data acquisition {#s0030}
-------------------------------------------
Data from patients who visited the outpatient clinic in Severance Hospital otorhinolaryngology department from the year 2013 to 2017 were used. As a routine, patients had their otoendoscopic photo taken upon visit. Drum photos were taken with either 4 mm or 2.7 mm OTOLUX 0-degree telescope (MGB Endoskopische Geräte GmbH Berlin, Germany) tethered to Olympus OTV-SP1 video imaging system (Olympus Corporation, Japan), by otolaryngology residents, faculty or experienced nurses. The image resolution was 640 by 480 pixels. A total of 19,496 endoscope photos were reviewed for labelling. Since otoendoscopic findings of post-surgery status are mostly subjective and rely on the surgeon, 7602 photos were excluded. Additionally, 1350 photos were excluded since the photos were not appropriate for examination, for example, sites not related to eardrum or EAC, duplicates, the picture was significantly blurred due to handshakes or focus problems, or the author could not agree despite attending physician\'s medical records, acoustic and radiologic test results. Since photos were taken by several clinicians, and the external auditory canal is subject to individual variation, the composition of photography was not standardized; colour arrangements, white balance, eardrum size, location, rotation, angle, and light reflection in images were variable, but the photo was analyzed as-is to reflect real-life clinical setting. In addition, partially visible eardrums due to the image\'s field of view not containing the whole eardrum were included in the analysis. Finally, a total of 10,544 otoendoscopic images of eardrum and EAC from patients were analyzed. This retrospective study was approved by the Severance Hospital Institutional Review Boards.
2.2. Labelling of images {#s0035}
------------------------
Photos of eardrums and its surrounding EAC were taken with otoendoscope and were labelled into six categories. The classification was done according to *Colour Atlas of Endo-Otoscopy* \[[@bb0045]\]. A normal eardrum and EAC included: 1) completely normal eardrum, 2) normal but showing healed perforation, 3) normal with some tympanosclerosis. Abnormal findings included: 1) tumorous condition which includes middle ear tumors, EAC tumors, and cerumen impaction, 2) otitis media with effusion, 3) eardrum erosions, otitis externa, 4) perforation of the eardrum, 5) attic retraction/atelectasis ([Fig. 1](#f0005){ref-type="fig"}). Some classes have relatively small numbers of samples for training. In order to balance the sample size for each class, we merged several sub-classes into a class according to their similarity in diagnosis and treatment. Three normal diagnoses are trained as one big "Normal" class. "Tumor" class include cerumen impaction, EAC tumors, and middle ear tumors since they share a common property that the eardrum is not well-visible, and since they often require surgical procedures. Attic retraction (or destruction) and eardrum atelectasis has been merged into "Aradom" class, since it shares common pathogenesis and physical findings, and often requires surgical intervention. Otitis externa and myringitis have been merged into "Myriaom-otex" class since otorrhea is the main symptom, the physical finding is similar, and first-line treatment is antibiotics.Fig. 1Decision tree for labelling of otoendoscopy image and six diagnostic classes. Classes that were used for training are marked with an asterisk. EAC: external auditory canal.Fig. 1
The number of images used and samples representing each classification is shown in [Fig. 2](#f0010){ref-type="fig"}A. If there are more than one features of the ear disease, for example, tympanic perforation with attic retraction, it was labelled as tympanic perforation, according to our labelling priority. The priority is based on the certainty of the diagnosis and clinical importance, for example, requiring surgical intervention. Of note, image acquisition was not standardized in any fashion and was labelled and trained as-is. Examples of diversity in image acquisition include differences in white balance, image composition, presence of cerumen, position of the eardrum in the image. The exemplary images for diversity in the normal class are presented in [Fig. 2](#f0010){ref-type="fig"}B.Fig. 2A) Examples of six classifications of ear disease, sorted by labelling priority (total *n* = 10,544). B) Example of image diversity labelled "Normal".Fig. 2
An in-house graphic user interface software implemented on MATLAB2019a® (MathWorks, Inc., Natick, Massachusetts, United States) was used for manual labelling. As mentioned above, study by Pichichero, Poole \[[@bb0010]\], confirms the limited accuracy by a single physician is below 75% at best, numerous methods were used for labelling the ground truth of otoendoscopic image. The images of eardrum and EAC were labelled by the first author, and all the images were double checked by reviewing electronic medical record written by attending physician at the time, who had at least 10 years of experience in a tertiary referral center. Since this study is retrospective, additional clinical demographic and symptomatic data was used. In addition, acoustic test results (pure tone audiometry, impedance audiometry) were often available along with computed tomography, magnetic resonance imaging results were used for more accurate labelling. For example, if pure tone audiometry and/or impedance audiometry data was available, it was used for labelling otitis media with effusion or tympanic perforation. Temporal bone computed tomography or magnetic resonance imaging results were also used if available. If the classification of otoendoscopic image could not come to an agreement even after reviewing all of the available information, it was discarded.
2.3. Training transfer learning network models {#s0040}
----------------------------------------------
Public deep learning models pretrained with ImageNet database (<http://www.image-net.org>), capable of classifying 1000 natural objects, were used for training the model for otoendoscopic images. Among many deep learning models publicly available, Alexnet \[[@bb0050]\], GoogLeNet \[[@bb0055]\], ResNet \[[@bb0060]\] (ResNet18, ResNet50, ResNet101), Inception-V3 \[[@bb0065]\], Inception-ResNet-V2 \[[@bb0070]\], SqueezeNet \[[@bb0075]\], and MobileNet-V2 \[[@bb0080]\] were used and compared since these network models are known to show higher performance in the accuracy when compared with any other networks with similar prediction time. Smaller size networks models were also included to see the performance for online processing. When transferring layers in public models to new models, we replaced the last fully connected layer of each model with a new fully-connected layer with six output nodes, followed by a softmax activation function. The training was conducted using an Adaptive moment estimation (ADAM) \[[@bb0085]\] with a batch size of 50, the maximum epoch of 20 and an initial learning rate of 0·0001. The initial learning rate of 0·0001, which may seem low, was selected according to our experience that using conventional learning rate of 0·01 to 0·001 did not converge in the current application. For the fully connected layer, we assigned weight and bias learning rate factors of 10 to render faster learning in the new layer than in the transferred layers. This study was conducted using Deep Learning Toolbox in MATLAB 2019a over four graphics processing unit (GPU) in the DGX station (NVIDIA, inc., USA). To augment data, we conducted random X and Y translation of input images from −45 to 45 pixels, random rotations from −30 to 30 degrees, random scales between 0.8 and 1.2 and random left/right flips to render translation, rotation, scale and left/right invariance.1.***Selection of the best two models:*** the best two among nine models were selected by evaluating the performance of each model in terms of accuracy and calculation time. From a total of 10,544 otoendoscopic images, 80% of the images were used for training; 20% were left out for validation of the model. This training-validation step was done twice with different sets of training and validation data. According to the mean accuracy and calculation time, we chose two models.2.***Performance according to data size, and hidden layer in the fully connected layer and colour channels:*** We evaluated the performance for all nine models for the half and a quarter of the full dataset, to compare those with the performance trained with all the data. We also evaluated the performance for a model with an additional hidden layer (node size =25) between the input layer and the output layer in the fully connected neural network. We also evaluated the performance for changing colour (RGB) orders by changing R and G channels in the image data since public network models were trained with natural images different from the current ear images.3.***Ensemble classifier:*** We generated an ensemble classifier that combines classifiers\' outputs from the best two models. Each classifier scores the probability of an input image to be one of six classes and the maximal score among all classes is chosen as a predicted label. The ensemble classifier adds the two scores from the two models for an input image and the class having a maximal score is chosen to be the image\'s label.4.***Cross-validation:*** For the best two models and the ensemble classifier, we conducted five-fold cross-validation for each classifier and evaluated the classification performance in terms of accuracy.
3. Results {#s0045}
==========
Accuracy, training time (GPU time), the number of parameters of each transferred model is presented in [Table 1](#t0005){ref-type="table"}. The number of parameters was referred from the MATLAB official web site (<https://www.mathworks.com/help/deeplearning/ug/pretrained-convolutional-neural-networks.html>). There was no significant improvement in the models with a hidden layer (number of nodes = 25, H25 in [Table 1](#t0005){ref-type="table"}). The average accuracy between different sets of data size showed significant improvement according to data size - 78·88% for data set of *n* = 2000, 85·62% for data set of *n* = 5000, and 90·21% for the entire data set of *n* = 10,544 ([Fig. 3](#f0015){ref-type="fig"}). The performance of nine models was evaluated without a hidden layer in the fully connected neural network. The best accuracy was yielded by the Inception-ResNet-V2 (92·1%), Inception-V3 (92%) and ResNet101 (91·55%) in order. Despite its accuracy, the Inception-ResNet-V2 (33,283 s) had three times longer training time than those of Inception-V3 (11,938 s) or ResNet101 (12,215 s). Therefore, we finally chose Inception-V3 and ResNet101 as best transferred network models for the subsequent analysis.Table 1Performance table of training models.Table 1Transferred modelsAccuracyGPU time (seconds)Parameters (millions)Number of layersFullFull-H25QuarterHalfSqueezeNet85.5585.573.582.841371.2468Alexnet87.283.673.782.638056125ResNet1890.6590.283.486425611.772MobileNet-v290.7589.879.984.970323.5155GoogLeNet90.988.768.285.551047144Resnet5091.291.481.386.3730225.6177Resnet10191.5591.783.686.112,21544.6347Inception-v39292.184.189.511,93823.9316InceptionResnet-v292.191.982.286.933,28355.9825[^2][^3][^4][^5][^6]Fig. 3Accuracy grouped by sample size.The bar represents 95% Confidence interval.Quarter: Data set of *n* = 2000.Half: Data set of *n* = 5000.Full: Data set of *n* = 10,544.\*: statistically significant \[Mann-Whitney test\].Fig. 3
From these two models, we generated an ensemble classifier, which decides the image label according to the sum of the two models\' scores for the given image ([Fig. 4](#f0020){ref-type="fig"}). [Fig. 5](#f0025){ref-type="fig"} shows examples of improvement using the ensemble classifier by evaluating the sum of classification scores of the two network models. Repeated measures one-way ANOVA for the 5-fold cross-validation tests ([Fig. 6](#f0030){ref-type="fig"}) showed that the ensemble model was significantly better than the other two models in accuracy \[*p* = 0·0005, Repeated measures one-way ANOVA\].Fig. 4Schematic example of how ensemble method classifies given otoendoscopic image.The cells shaded in gray background is the final prediction for each model.Fig. 4Fig. 5Examples of inconsistencies between prediction models, and obtaining better accuracy with ensemble model.Incept: Prediction using Inception-V3 model, ResN: Prediction using ResNet101 model,Ensemble: Prediction using ensemble of both models, Target: Ground truth.The second row shows classification scores of Inception-V3 model.The third row shows classification scores of ResNet101 model.Classification scores are represented in the following order: Aradom-Myriaom-Normal-Ome-Tp-Tumor.Fig. 5Fig. 6Accuracy comparison between three methods.The bar represents 1 Standard deviation.Mean accuracy is 0·9154, 0·9251, 0·9373, respectively.\*: statistically significant (*P* = 0·0001) \[Repeated measures one-way ANOVA\].Fig. 6
Overall, the system was able to achieve an average of 93·73% diagnostic accuracy. [Fig. 7](#f0035){ref-type="fig"} displays the confusion matrices for Inception-V3, ResNet101, and the ensemble classifier at the fold (among 5-folds) having a maximal accuracy. [Fig. 8](#f0040){ref-type="fig"} shows a representative figure of classification result from "InceptionV3 + ResNet101 ensemble" model.Fig. 7Confusion matrices for Inception-V3, ResNet101, and ensemble classifier at the fold (among 5-folds) having a maximal accuracy.Target class in x axis refers to ground truth label.Output class in y axis refers to classification by InceptionV3 based model.No: Normal eardrum and external auditory canal (including some tympanosclerosis, healed perforation).Tp: tympanic perforation, Ar: Attic retraction or adhesive otitis media.My: myringitis and/or otitis externa.Om: Otitis media with effusion.Tu: middle ear or external auditory canal tumor or cerumen impaction.Fig. 7Fig. 8Representative figure of classification results from "InceptionV3 + ResNet101 ensemble" model.Abbreviations are identical to [Fig. 7](#f0035){ref-type="fig"}. Labeling is ordered as Ground truth--Classification.Tu-No: ground truth is tumor, but the system classified as normal.Om-Tu: ground truth is otitis media with effusion, but the system classified as tumor.Fig. 8
4. Discussion {#s0050}
=============
Despite many efforts to improve diagnostic accuracy, diagnosis of otitis media mainly relies on otoscopy and often relies on physician\'s experience \[[@bb0090]\]. Diagnosis by otoendoscopy requires expertise in image diagnosis; in a study with video-presented examination for diagnosis, otolaryngologists performed significantly better than pediatricians and general practitioners \[[@bb0095]\]. Even for otolaryngologists, diagnosis of otitis media by otoendoscopy is not trivial. In a study, a series of surveys for diagnosis of eardrum images with twelve fellowship-trained neurotologists in the United States was conducted with overall correctness of diagnosis for ear pathologies ranging from 48·6 to 100%. Along with the diagnosis, the reviewer was also asked to rate the confidence of diagnosis, which revealed overall mean 8·1 out of 10, which means even for a specialist, they are only about 80% certain about their diagnosis on average. In these situations, the current deep network model could help physicians by suggesting possible diagnosis based on otoendoscopic image, and they could achieve better diagnostic accuracy by combining clinical information along with suggestion.
The current image classification model, based on transfer learning with deep convolutional neural network, classified middle ear and EAC pathologies into six categories with a mean accuracy of 93·73%, which is unprecedented in terms of both accuracy and diagnostic diversity. This high accuracy for multiple classes is partly due to the size of the current database. In the current model, 10,544 labelled otoendoscopic images were used, which is significantly bigger than any other studies to our knowledge. Previous studies from other groups utilized 391 and 389 images and yielded 80·6% and 86·84% of five classes focused only on otitis media \[[@bb0015],[@bb0100]\]. Compared to other previous studies, the advantage of the current model is that this study included almost all eardrum and EAC pathologies, especially tumors, attic retraction, and eardrum atelectasis, which is a crucial part of diagnosis in the real-world clinical setting. Retraction of the eardrum may indicate chronic otitis media, especially if retraction pocket or destruction is present in the attic area and should not be missed in the clinic. Middle ear tumors such as glomus tumor and congenital cholesteatoma are rare, and due to its rare prevalence, there is a considerable chance of missing the diagnosis unless examined by an experienced physician with clinical suspicion. The current image classification model is the first to diagnose these pathologies.
It should be noted that we intentionally included all the clinical ear images (except for no ear images), without any selection bias for training. Unlike X-ray images or histology slide images, there is no standardization for image acquisition or quality controls in the ear images. White balance is not always equal, which in turn leads to inconsistent skin or eardrum colour. Camera exposure may not optimally focus on eardrum in case of a tortuous external auditory canal or mass blocking the eardrum. Blurry or out of focus images happen quite often. The eardrum is not always in the centre of the image. Rotation, tilting of the image is inconsistent. Such example is illustrated in [Fig. 2](#f0010){ref-type="fig"}B. We included most of the images as long as a clinician could get an impression for diagnosis upon given image. We speculate that this practical image database (including uncleaned data) makes the performance of the model to be dependent on the database size.
Reducing the number of images for training to 2000 images, and 5000 images, the average accuracy was declined to a level of around 80% and 86%. Mann Whitney tests for accuracy among the three conditions showed statistical significance ([Fig. 3](#f0015){ref-type="fig"}). These results indicate that it is hard to get a satisfactory result with a small amount of data for training even in the transfer learning, at least in the current ear diagnosis. If the otoendoscopic images were acquired in a standardized setting, similar accuracy could have been achievable with fewer images. A big amount of data is advantageous for a deep network model to find features that explain various disorders regardless of clinical conditions.
Another thing to note is that as the data set gets bigger, the performance gap between training models gets reduced ([Table 1](#t0005){ref-type="table"}). With this in mind, the efficiency of training (training time versus accuracy) for each model should be considered in choosing the best model. As for InceptionResNet-V2 model, the accuracy is only 0·5% better than Inception-V3 model, yet requiring almost 3 times as much training time, in other words, processing power. Therefore, Inception-V3 model seems to be the best choice considering the efficiency of training. When it comes to execution time, the number of parameters are related to calculation time. The MobileNet-V2 model, which has only 3·5 million parameters, achieved 90·75% accuracy, which is not best in terms of accuracy. However, given that it has fewer parameters than other competitors, this model could be more useful in devices with less processing powers, such as mobile phones. In this case, CNN models with fewer parameters may be optimal with acceptable diagnostic accuracy, given that the model has been trained with a sufficiently large number of images.
Adding an additional fully connected layer in front of the final classifiers did not help in this study. Changing colour (RGB) orders by switching R and G channels in the image data show similar or lower accuracy than utilizing natural RGB channels. Although all these variations were not beneficial in the current study, we think it is too early to conclude the generality for these schemes in other applications.
Transfer learning method is popularly used in the medical image analysis as it makes it possible to apply deep learning techniques to a relatively small dataset without significantly sacrificing accuracy. It shows a highly reliable accuracy in various medical image diagnosis \[[@bb0035],[@bb0040]\]. This study is in line with the previous studies of transfer learning with fine-tuning to make it applicable to the specific domain of medical image diagnosis. In the field of ophthalmology, transfer learning was applied to diagnose retinal optical coherence tomography (OCT) images, allowing similar accuracy to a model with the full training data with less training data \[[@bb0105],[@bb0110]\]. Also, there are studies focused on microscopic histological images utilizing transfer learning for classification \[[@bb0115]\]. Since transfer learning is efficient in training time, several transfer models can be built practically with a given data set. Instead of using one model, several studies have combined different models to improve classification accuracy \[[@bb0120],[@bb0125]\]. In those studies, transfer learning has been used as feature extractors. Features from each model are concatenated to train a new network. Utilizing trained models as feature extractors are limited in the fine-tuning of the transfer layers since the classifier is independently trained with feature extraction networks. Furthermore, feature sets from multiple models may contain redundant information as the number of parameters increases. In contrast to combining features to retrain a new classifier, we simply combined classification scores of each model and determined image labels according to the maximal scores (the probability of being the class). The classification using an ensemble of Inception-V3 and ResNet101 model ([Fig. 4](#f0020){ref-type="fig"}) increased diagnostic accuracy significantly ([Fig. 6](#f0030){ref-type="fig"}). Usually, transferred Inception-V3 model is a better performer, but in some cases, transferred ResNet101 is more accurate, and the ensemble method was able to take advantage of combining inconsistencies of the two models. [Fig. 5](#f0025){ref-type="fig"} shows an example for this ensemble approach, where shows the score of each model, which is softmax value representing the probability of each classification. Upon inspection, the classifier that has a sum of classification scores close to 2 (maximum 1 for each model), which means the model is almost certain about the diagnosis, tends to be chosen by ensemble model. It resembles a case conference between two physicians, in this case, Inception-V3 and ResNet101, arguing over the right diagnosis and the one with more certainty winning the argument.
Based on the confusion matrix of the ensemble classifier, the classification system is a good performer in the diagnosis of normal, otitis media with effusion, tympanic perforation, and tumors which exceeded over 90% accuracy. Additional representative figure ([Fig. 8](#f0040){ref-type="fig"}) illustrates examples of otoendoscopic images. As for otitis externa or myringitis, accuracy is 77·91% with 89·33% sensitivity and 99·02% specificity. It often misdiagnosed as tympanic perforation or tumorous condition; in some cases of myringitis, the EAC may be whitish, wet circular fashion, with the centre being dark, mimicking large perforation. Also, it may be confused with tumors, which makes sense since it hinders the proper view of the eardrum and external auditory canal. Mostly, these images often contain discharges, crusts in the EAC, which should have been removed prior to taking images for better accuracy. Label "ARADOM" refers to attic retraction or adhesive otitis media, accuracy is 85·78% with 90·19% sensitivity and 98·25% specificity. It was commonly misdiagnosed as tympanic perforation or normal. In non-severe cases, retraction could be subtle and clinicians may find it hard to decide whether it is normal or grade I retraction by Tos or Sade classification \[[@bb0130], [@bb0135], [@bb0140]\]. On the other hand, severe cases of attic retraction or middle ear atelectasis often reveal the ossicles inside the tympanic membrane, sometimes making it hard to distinguish between total perforation and severe atelectasis. Detecting attic retraction is very important since it implies underlying chronic otitis media, and it often requires surgical treatment to prevent progression. This model\'s capability of predicting attic retraction or adhesive otitis media, with an accuracy of 85·78% is the most important and practical technologic advancement to be of use in clinics.
In terms of predicting normal and abnormal otoendoscopic findings, overall sensitivity and specificity is 93·69% and 96·82%, respectively. It shows the possibility to be used for screening of ear disease in regular routine health checkup.
[Fig. 8](#f0040){ref-type="fig"} illustrates examples of otoendoscopic image classified in the ensemble model. Trivial cases tend to be appropriately classified, and most misclassified items have some ambiguities; the classification system tends to be not entirely wrong about the diagnosis. For example, in [Fig. 8](#f0040){ref-type="fig"}, image labelled TUM-TP, meaning ground truth is tumor, but there is also tympanic perforation present, it does have tympanic perforation, which the classification system has labelled accordingly and is also a correct diagnosis.
Acquisition of the otoendoscopic image could be easily done by non-doctors with a little degree of training, and remote diagnosis based on the otoendoscopic photo may not significantly differ whether the photo was taken by otolaryngologist or telehealth facilitator \[[@bb0145]\]. In areas short of otolaryngologists, some other speciality doctors (pediatricians, family medicine, or general practitioners), or even non-doctors could take otoendoscopic photos, and analyze images for ear disease based on our system and decide the next step. If the diagnosis is normal or otitis media with effusion, observation is recommended. Otherwise, if otitis externa or myringitis is suggested, physicians of other specialities could try antibiotics before referring the patient to an otolaryngologist for further intervention. For attic retraction, tympanic perforation, or tumors, referring the patient to otolaryngologist would be appropriate for the next step. A system based on the current study could aid early diagnosis of one of the most common childhood illness, otitis media \[[@bb0150]\], which may alleviate the burden of growing number of patients with hearing impairment.
For otolaryngology specialists, this model could be useful for generating second opinion and be used for double checking the diagnosis, especially tumors and attic retractions which could have been missed due to insufficient experience or low clinical suspicion. This model did not take external patient factors such as age, presence of fever and otologic symptoms such as (pulsatile) tinnitus, ear fullness, otalgia, hearing loss, otorrhea, etc. In real-life clinical settings, physicians may take otoendoscopic image and correlate the current classifier\'s results with clinical information for diagnosis. In turn, the current deep learning classifier may be trained with these non-image information for better diagnosis rate.
Considering many previous studies regarding the diagnosis rate of ear disease, which were \<80% on average \[[@bb0010],[@bb0095],[@bb0155]\], we carefully claim that this automated diagnosis image classification system can perform better than an average otolaryngologist specialist, and since this classification system covers most of ear disease domains including attic retraction, tumors, which is unprecedented, it is ready for use in real-world clinical settings. Ultimately, it may help the world ease the burden of hearing impairment by contributing to early diagnosis of ear disease.
Data sharing statement {#s0055}
======================
The data are not available for public access because of patient privacy concerns but are available from the corresponding author on reasonable request approved by the institutional review boards of Yonsei university college of medicine.
Funding sources {#s0060}
===============
This research was supported by Brain Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT(NRF-2017M3C7A1049051) and by a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant number HI18C0160).
Author contributions {#s0065}
====================
D.C designed the study, collected & analyzed the data. C·P and H.P. designed deep transfer learning system and performed training with S.S. D.C drafted the manuscript with J.C and H.P. and all authors reviewed and revised the manuscript.
Declaration of Competing Interest
=================================
None.
None.
[^1]: Equally contributed first authors.
[^2]: "Quarter" set used about 2000 images for training and validation.
[^3]: "Half" set used about 5000 images for training and validation.
[^4]: "Full" represents an average accuracy of twice evaluation of full dataset (80% training and 20% validation).
[^5]: "Full-H25" represents adding additional 25 fully connected hidden layer to "Full" model.
[^6]: GPU time represents the processing power needed for training the model.
| {
"pile_set_name": "PubMed Central"
} |
Background {#Sec1}
==========
Fasciolosis is a zoonotic parasitic disease caused by infection with the digenetic trematode flukes of the genus *Fasciola*. While *Fasciola hepatica* is prevalent in temperate regions, *F. gigantica* is more widespread in Africa and Asia \[[@CR1], [@CR2]\]. Migration of these flukes inside the body of the host causes severe damage to the liver parenchyma and gall-bladder \[[@CR3]--[@CR5]\]. Buffaloes are economically important animals for the farming communities in developing countries. Infection of buffaloes with *F. gigantica* is common in southern China and other geographic regions of the world \[[@CR6]\]. Infection can cause poor animal health and significant loss of meat and milk production, with considerable financial implications \[[@CR3], [@CR7]\]. *Fasciola gigantica* flukes specifically target the liver of their definitive host. Effective and balanced local immunity is therefore essential for detecting and controlling these hepatotropic parasites, and for limiting hepatic damage.
Liver flukes are, however, efficient immune-modulators and produce many effectors in order to exploit the host immune response to ensure their survival. A recent study in experimentally infected buffaloes reported a modest increase in the level of Th2-type immune cytokines during early *F. gigantica* colonization and immunosuppression during chronic *F. gigantica* infection \[[@CR8]\]. Other studies reported a pro-inflammatory or a mixed Th1/Th2 immune response during early infection, and heightened Th2 and Treg responses during chronic infection. This heightened response was assumed to play roles in restoring the host tissue integrity by damping excessive inflammatory response \[[@CR9], [@CR10]\]. How inflammation contributes to the pathogenesis of *F. gigantica* is a complex and multi-faceted story that is still unfolding.
Hepatic immune-inflammatory mechanisms are essential to maintain liver homeostasis and, if dysregulated (e.g. due to parasite infection), can lead to liver pathology and dysfunction. The abnormal production of cytokines and/or transcription factors can lead to inadequate control of *Fasciola* infection \[[@CR11], [@CR12]\]. CD4^+^ T-cells are subdivided into Th1, Th2, Th17 and regulatory T-cells (Treg) subsets, based on their pattern of cytokine production \[[@CR13]\]. Transcription factors T-bet, GATA-3, Foxp3 and ROR-γτ play important roles in the differentiation of Th1, Th2, Treg and Th17 cells respectively, and mediate the production of cytokines in these cells \[[@CR14]--[@CR16]\]. Although immunological impairment and polarization of the Th1/Th2 balance are major consequences of *F. gigantica*-induced liver pathology, the expression profile and dynamic changes of Th1/Th2 cytokines during *F. gigantica* infection has not been completely elucidated. Also, the role of Th17 and Treg cells in the pathogenesis of *F. gigantica* infection is still not well-defined.
In the present study, we hypothesized that *F. gigantica* infection impairs the balance of Th subsets (Th1/Th2/Th17) and Treg, thus contributing to the immune-pathogenesis of fasciolosis. A temporal study of gene expression of nine cytokines (IFN-γ, IL-1β, IL-12B, IL-4, IL-6, IL-10, IL-13, IL-17A and TGF-β) and four transcription factors (T-bet, GATA-3, Foxp3 and ROR-γτ) in the livers from buffaloes infected with *F. gigantica* was conducted using quantitative real-time PCR (qRT-PCR). Data analyses revealed a large number of differentially regulated genes, which exhibited temporal profiles of expression across the time course study. Our results showed evidence of a step-change in gene expression from an 'early' TGF-β-associated immune-suppersive response (3--10 dpi), to a mixed Th1/Th2 immune response (28--70 dpi) and a 'late' predominantly Th1/Treg-driven response (98 dpi). These results provide new insights into the dynamic immune response of buffaloes to *F. gigantica* over the course of experimental infection.
Methods {#Sec2}
=======
Parasite strain {#Sec3}
---------------
Eggs of *F. gigantica* were obtained from the gall bladder and faeces of naturally infected buffaloes slaughtered for human consumption at local abattoirs (Nanning, Guangxi, P.R. China). Protocols used for the preparation of *F. gigantica* eggs, snail infection with miracidia and harvesting of encysted metacercariae (EM), were performed as previously described \[[@CR16]\]. EM were stored in sterile phosphate buffered solution (PBS) at 4 °C. We employed PCR amplification and sequencing of the second internal transcribed spacer (ITS-2) of ribosomal DNA (rDNA) to genotype EM, as described previously \[[@CR17]\]. Species identity was confirmed as *F. gigantica* based on absolute homology to the known ITS-2 sequence of *F. gigantica* from Guangxi province (GenBank: AJ557569). The viability of EM was examined microscopically and only those with viability greater than 90% were used.
Animals {#Sec4}
=======
Eight to ten month-old (80--100 kg body weight) water buffaloes (*n* = 35) were obtained from a local breeder and were identified as swamp type by karyotypic analysis. Buffaloes were kept in separate concrete floor pens. Commercial feed and clean water were provided ad libitum for all animals during the entire study period. None of the buffaloes had been used previously for any experimental procedure. Animals were confirmed as negative in terms of prior infection with liver flukes, by negative fecal examination and negative serum *F. gigantica*-specific IgG-antibody-based ELISA prior to the start of the study. All animals were treated with a single dose of triclabendazole (5% *w*/*v*) in order to eliminate any potential existing fluke infection that may have been missed on laboratory examination. Following triclabendazole treatment, buffaloes were allowed to acclimatize for 30 days to avoid any residual efficacy of the treatment on the establishment of experimental *F. gigantica* infection.
Animal inoculation and tissue collection {#Sec5}
----------------------------------------
Thirty-five buffaloes were assigned randomly to seven different groups (5 buffaloes/group). Group I was composed of 5 buffaloes that were mock-incoulated with PBS only. Buffaloes in Groups II-VII were each infected with 500 viable metacercariae by oral gavage. Control buffaloes were euthanized at the start of the experiment to obtain baseline values for hepatic tissue pathology and gene expression. Animals from each of the six infected groups were sacrificed and their livers were harvested at 3, 10, 28, 42, 70 and 98 days post infection (dpi), for histopathological and molecular studies. Group III-VII buffaloes were examined clinically on a weekly basis for the development of clinical signs of fasciolosis. The control group served as a baseline point of reference for monitoring the progressive changes in gene expression over the course of infection. However, the inclusion of matched control groups receiving placebo (treated with vehicle only) and euthanized at the same time points as infected groups would have strengthened the power of the study. Alternatively, control buffaloes could have been kept alive (rather than killing them at the start) and liver punch biopsy samples obtained from them at each of the above time points. Despite the early sacrifice of the control animals for reasons of economy and resources, we believe the effects we have observed to be due to the experimental manipulation.
Gross examination and histopathological evaluation {#Sec6}
--------------------------------------------------
At six time points after infection (indicated above), animals in each infected group were sacrificed, their livers were harvested and examined for pathological lesions and the presence of the flukes. Parasite eggs were recovered by filtering bile fluid through a 0.15 mm pore size mesh. *Fasciola gigantica* infection was confirmed by observing gross pathological lesions, associated with flukes in the livers and/or by the presence of flukes and eggs in the bile ducts. Samples of liver tissue (\~8 g) showing pathological lesions were collected from each animal. Tissue samples were resuspended in 10% PBS-buffered formalin solution overnight, then dehydrated in alcohol, rinsed in xylene, and embedded in paraffin. 3 μm sections of paraffin-embedded tissue were mounted onto glass slides, and stained with hematoxylin and eosin (H&E). Stained tissue sections were examined microscopically at 400× magnification and imaged using a Zeiss Axio Imager manual upright research microscope. Additional liver tissue samples for RNA extraction were collected and kept in RNA store buffer (Tiangen Biotech, Beijing, China), snap frozen in liquid nitrogen and stored at -80 °C.
RNA isolation {#Sec7}
-------------
Total RNA was extracted from frozen liver tissue samples by RNAprep Pure Tissue Kit (Tiangen Biotech, Beijing, China) following the manufacturer's instructions. RNA integrity was examined by 2% agarose gel electrophoresis and quantified by NanoDrop 2000/2000c Spectrophotometer analysis (Thermo Scientific, Waltham, US).
Quantification of cytokine and transcription factor gene expression {#Sec8}
-------------------------------------------------------------------
Quantitative gene expression analysis was performed on liver samples obtained from uninfected control animals, and from infected animals on 3, 10, 28, 42, 70 and 98 dpi. Levels of mRNA expression of nine cytokines (IFN-γ, IL-1β, IL-12B, IL-4, IL-6, IL-10, IL-13, IL-17A and TGF-β) and four transcription factors (T-bet, GATA-3, Foxp3 and ROR-γτ) were determined using quantitative real-time PCR (qRT-PCR). All qRT-PCR primers used in the study are described in Table [1](#Tab1){ref-type="table"}. Complementary DNA (cDNA) was synthesized from 500 ng RNA samples using a PrimerScript™ RT reagent kit (TaKaRa Bio, Dalian, China). qRT-PCR was performed using SYBR®Premix Ex Taq™ II (Tli RNaseH Plus, TaKaRa Bio) and a CFX96 real-time PCR instrument (Bio-Rad, Hercules, US). To determine the specificity of amplification, melting curve analysis was applied to all final PCR products. The efficiency of qRT-PCR, and relative quantification (*RQ*) of gene expression, were analyzed using the comparative 2^--ΔΔ*Cq*^ method \[[@CR18]\]. The level of expression of each gene was normalized using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as the reference housekeeping gene.Table 1List of primers used in the SYBR green-based qRT-PCR analysisGene targetPrimer sequence (5′--3′)Product length (bp)ReferenceGAPDHFCCTGCACCACCAACTGCTTG222\[[@CR62]\]RTTGAGCTCAGGGATGACCTTGIFN-γFGTCTCCTTCTACTTCAAACT253\[[@CR63]\]RATTCTGACTTCTCTTCCGCTTGF-βFCGTGCTAATGGTGGAATAC208Present studyRGCCAGGAATTGTTGCTATAIL-1βFCTAGCCCATGTGTGCTGAAG59\[[@CR62]\]RCCTTTACTTGGCTCTTCACCIL-4FCAGCATGGAGCTGCCT177\[[@CR64]\]RACAGAACAGGTCTTGCTTGCIL-6FCTGCAATGAGAAAGGAGATA191\[[@CR63]\]RGGTAGTCCAGGTATATCTGAIL-10FCTGTGCCTCTCCCCTAGAGT236\[[@CR62]\]RGCAGCTAGCTCCACAAGGAAIL-12BFCAGGGACATCATCAAACCAG213\[[@CR63]\]RCTTGTGGCATGT GACTTTGGIL-13FAGAACCAGAAGGTGCCGCT50\[[@CR65]\]RGGTTGAGGCTCCACACCATGIL-17AFCTACAGTGAACTGGAAGGAAC554Present studyRAAAAGGGGCTGGGTCTT-betFCCTGGACCCAACTGTCAACT171\[[@CR66]\]RGAAACTCGGCCTCATAGCTGGATA-3FGATCAAGCCCAAGCGAAGG124Present studyRCCGCAGGCATTGCAGACAFoxp3FGACAGCACCCTTTCGACTGT191\[[@CR66]\]RCTCCAGAGATTGCACCACCTROR-γτFCTACAGTGAACTGGAAGGAAC554Present studyRAAAAGGGGCTGGGTCT*Abbreviations*: *F* forward primer, *R* reverse primer
Data analysis {#Sec9}
-------------
Statistical analysis and graph production were performed using GraphPad Prism (GraphPad Software Inc., La Jolla, CA, USA, version 6.02.). Levels of cytokine and transcription factor mRNA expression between uninfected and infected groups were compared at different time points after infection using one-way analysis of variance (ANOVA) with *post-hoc* LSD multiple comparison tests. Results were presented as *F* ~(DFn,\ DFd)~ and *P*-value. Pearson's correlation coefficients (*r*-value) were used to detect any correlation between the measured level of Th1, Th2, Treg and Th17 immune cytokines and transcription factors (T-bet, GATA-3, Foxp3 and ROR-γτ) gene expression between infected groups, followed by a two-tailed *post-hoc* test and presented as a *P-*value. Data shown represent the mean ± SEM of results from five buffaloes. The level of significance for all analyses was evaluated with a confidence interval \> 95% (*P* \< 0.05).
Results {#Sec10}
=======
Gross and histopathological attributes {#Sec11}
--------------------------------------
Even though buffaloes did not exhibit clear clinical signs, *F. gigantica* induced a wide range of pathological lesions over the course of infection (Table [2](#Tab2){ref-type="table"}). During the early stage of infection (3--70 dpi), immature flukes migrated through the intestinal wall, abdominal cavity, liver capsule and liver parenchyma, to reach the bile ducts where they attain sexual maturity. Migration of the juvenile flukes through the host's tissues was characterized by classical inflammatory signs and accumulation of fibroblasts towards the end of this stage, but without obvious fibrosis. The total number (and average length) of flukes recovered from the livers of infected groups at 28, 42, 70 and 98 dpi were: 3 flukes (1.5 mm), 22 flukes (2.5 mm), 65 flukes (8 mm), and 36 flukes (14 mm), respectively. No fluke was observed in the liver or bile duct before 28 dpi, probably because they were too small to be visible to the naked-eye. Hepatic hemorrhage, swelling, necrosis and viscous bile were first detected at 10 dpi. Tissue fibrosis was observed at 70 dpi, along with disappearance of hemorrhage. Seven parameters were used to describe the histopathological features in buffalo livers during infection (Table [3](#Tab3){ref-type="table"}). Histopathological characteristics of the infected liver tissues (Fig. [1](#Fig1){ref-type="fig"}) included an infiltration of inflammatory cells, such as neutrophils and lymphocytes at 3 dpi. As infection progressed, an accumulation of eosinophils, monocytes and red blood cells (RBCs) was detected (10--28 dpi), while fibroblast formation and bile duct hyperplasia were observed later (42 dpi). During the chronic stage (70--98 dpi), significant fibrosis and necrosis, without calcification, were observed. Fluke eggs were detected in the bile and feces of infected animals at 98 dpi. Granulomas with necrotic centers, heavy infiltrates of lymphocytes, RBCs and eosinophils were present at 98 dpi.Table 2The presence (+) and absence (−) of gross pathological lesions, flukes and fluke eggs in the liver of buffaloes experimentally infected with *F. gigantica*GroupDpiNo. of animalsHemorrhageSwellingParasite tunnelFibrosisNecrosisBile duct hyperplasiaViscous bileVisible flukesEggsI05----------------II35------------------III105++----+--+----IV285++----+--++--V425--++--++++--VI705--+++++++--VII984^a^--++++++++^a^RNA from one animal in this group VII did not pass the quality control check and was therefore excluded from the analysis Table 3The presence (+) and absence (−) of the histopathological changes observed in the liver of buffaloes experimentally infected with *F. gigantica*GroupDpiNo. of animalsInflammatory infiltrationRBCsEosinophilsNeutrophilsFibroblastsLymphocytesMonocytesI05--------------II35+----+--+--III105++++--+--IV285++++--++V425+++++++VI705+++++++VII984+++++++ Fig. 1Histopathological characteristics of the livers of buffaloes infected with *Fasciola gigantica*. **a** At 3 dpi, there was local hyperemia associated with mild filtration of lymphocytes and neutrophils. **b** At 10 dpi, there was scattered vacuolation of hepatocytes, consistent with fat, along with mild to moderate focal necrosis. **c** At 28 dpi, diffuse intravascular coagulation, severe infiltration of inflammatory cells mainly neutrophils and lymphocytes, and granular degeneration of the cytoplasm were observed. **d** At 42 dpi, moderate to severe multifocal hemorrhages and necrosis, infiltration of eosinophils, RBCs and monocytes, and accumulation of fibroblasts were detected. **e** At 70 dpi, there were mild to moderate multifocal bile duct hyperplasia and focal coagulative necrosis associated with collagen deposition. **f** At 98 dpi, severe periportal fibrosis associated with multifocal inflammatory infiltrate, cellular debris, moderate multifocal hemorrhages, and granulomas with necrotic centres were detected. **g** Liver tissue from uninfected animal showed normal histological architecture of the hepatic tissue. **h** Adult flukes in the intrahepatic bile duct, along with epithelial hyperplasia of the duct. In all figures, tissue sections were stained with H&E and arrows point at the corresponding morphological features described above. *Scale-bars*: **a-g**, 50 μm; **h**, 100 μm
Gene expression of cytokines {#Sec12}
----------------------------
To determine how immune response changed over the course of infection, gene expression of nine cytokines was assessed in the liver tissue of 29 infected buffaloes and five uninfected buffaloes using qRT-PCR. RNA from one animal from group VII did not pass the quality control check and was therefore excluded from the analysis. Results showed that infection had a significant impact on gene expression of Th1, Th2, Th17, and Treg cytokines (Fig. [2a](#Fig2){ref-type="fig"} and Table [4](#Tab4){ref-type="table"}). The transcriptional profile of cytokine genes showed an immunosuppressive state during early infection, a mixed Th1/Th2 response as the infection progressed, and a shifting to Th1/Treg response, associated with greater histopathological changes and fibrosis during late stage infection.Fig. 2Temporal changes of the mRNA expression of cytokines and transcription factors in the liver of buffaloes experimentally infected with *Fasciola gigantica*. The X axis represents days post infection (dpi), control animals are represented by empty bars and Y axis represents the mRNA relative expression of target genes relative to *GAPDH* based on 2^--ΔΔ*Cq*^ calculation. Columns show the means and error bars show SEMs. The log number of mRNA relative expression of **a** IFN-γ, IL-1β, IL-4, IL-6, IL-10, IL-12B, IL-13, IL-17A, TGF-β and **b** T-bet, GATA-3, Foxp3, and ROR-γτ, are shown. **c** Trends in the ratios between each pair of transcription factors during the course of infection, which was used as indicators of the balance between Th1/Th2 and Treg/Th17. Significant differences of each time-point compared with control non-infected (NC) group: \**P* \< 0.05; \*\**P* \< 0.01; \*\*\**P* \< 0.001 or \*\*\*\**P* \< 0.0001 (analyzed by one-way ANOVA, *post-hoc* LSD test) Table 4Relative gene expression in the liver of buffaloes experimentally infected with *F. gigantica* compared to the control group. The relative quantities were obtained according to the comparative *Cq* method (2^--ΔΔ*Cq*^). *P* values indicate statistical significance (*P* \< 0.05)Gene TargetANOVAMultiple comparisons3 dpi10 dpi28 dpi42 dpi70 dpi98 dpi*F* ~(DFn,\ DFd)~*PRQPRQPRQPRQPRQPRQP*IFN-γ4.693~(6,\ 21)~0.00352.080.43265.510.12822.060.73583.210.45909.270.007913.910.0003TGF-β4.738~(6,\ 19)~0.004113.930.00033.070.48841.670.80363.850.34081.630.83712.660.6062IL-1β11.05~(6,\ 26)~\< 0.00011.310.97671.270.981114.250.00070.260.81372.310.802528.45\< 0.0001IL-42.877~(6,\ 24)~0.02917.010.584719.40.489297.210.002827.90.31219.430.774967.140.0630IL-67.636~(6,\ 24)~0.00015.500.85292.810.936678.10.00021.940.972823.140.280594.560.0008IL-102.514~(6,\ 26)~0.04715.990.70039.920.318421.180.06942.30.95796.240.490232.670.0036IL-12B5.595~(6,\ 26)~0.00085.920.47863.990.474613.680.07590.860.98255.370.531224.84\< 0.0001IL-136.56~(6,\ 20)~0.00062.090.84597.800.152211.870.02595.450.349910.80.042627.33\< 0.0001IL-17A3.532~(6,\ 21)~0.01420.330.46120.070.36945.950.01140.260.43270.280.44002.990.6354T-bet2.173~(6,\ 28)~0.0762.160.79230.250.78304.730.10183.930.41583.990.460611.380.0113GATA-33.319~(6,\ 28)~0.01350.840.81011.420.95795.180.00354.800.08891.340.93971.80.6952Foxp31.455~(6,\ 28)~0.22960.700.90541.070.97493.030.04243.970.37053.400.44583.050.1372ROR-γτ4.387~(6,\ 28)~0.0032.250.75674.570.378010.280.00023.240.56372.680.70762.550.7148*Abbreviations*: *RQ*, relative quantity (mean); *F* ~(DFn,\ DFd)~, degrees of freedom with numerator and degrees of freedom denominator inside the subscript parentheses
The gene expression of Th1 cytokines (IFN-γ and IL-12B) and IL-10 showed no significant change during early infection. IFN-γ showed significantly higher expression at 70 dpi (*P* = 0.0079) and 98 dpi (*P* = 0.0003), when compared with the control. Relative gene expression of both of IL-12B (*P* \< 0.0001) and IL-10 (*P* = 0.0036) peaked at 98 dpi. Fluke infection also upregulated IL-4 expression at 28 dpi (*P* = 0.0028), whereas IL-13 significantly increased at 28 dpi (*P* = 0.0259) and 70 dpi (*P* = 0.0426), peaking at 98 dpi (*P* \< 0.0001). TGF-β mRNA was rapidly elevated at 3 dpi (*P* = 0.0003), followed by a decrease to the basal level and remained low thereafter. Upregulation of IL-17A was noted at 28 dpi (*P* = 0.0114), however expression was relatively low based on a high *Cq* value, suggesting it to be of little diagnostic value as a marker for Th17 cytokine in this study. The kinetics of the cytokines IL-1β and IL-6 mRNA were similar, both peaking at 28 dpi (*P* = 0.0007 for IL-1β; *P* = 0.0002 for IL-6) and 98 dpi (*P* \< 0.0001 for IL-1β; *P* = 0.0008 for IL-6).
Spearman's correlation analysis between each pair of cytokines was performed based on the *RQ* values from infected groups (Fig. [3a](#Fig3){ref-type="fig"}). This analysis revealed a significant positive correlation between the expression of IL-12B mRNA and Th2 cytokines (IL-4: *r* ~(29)~ = 0.44, *P* = 0.0336; IL-13: *r* ~(29)~ = 0.41, *P* = 0.0542), and between IFN-γ and IL-12B (*r* ~(29)~ = 0.62, *P* = 0.0017) or IL-13 (*r* ~(29)~ = 0.51, *P* = 0.0124). The gene expression of IL-10 was significantly correlated with both Th1 and Th2 cytokines (IFN-γ: *r* ~(29)~ = 0.49, *P* = 0.0171; IL-12B: *r* ~(29)~ = 0.50, *P* = 0.014; IL-4: *r* ~(29)~ = 0.80, *P* \< 0.0001; IL-13: *r* ~(29)~ = 0.51, *P* = 0.014) and with IL-17A (*r* ~(29)~ = 0.73, *P* \< 0.0001). Positive correlation was also found between IL-4 and IL-13 (*r* ~(29)~ = 0.44, *P* = 0.0336) gene expression. There was only a statistically significant relationship between IL-17A expression and IL-4 *r* ~(29)~ = 0.87, *P* \< 0.0001) and IL-10 (*r* ~(29)~ = 0.73, *P* \< 0.0001), but not with any of the other cytokines. Although not significant, the relative mRNA expression of TGF-β showed an inverse relationship to any other cytokine.Fig. 3Pearson's correlation analysis between mRNA expressions of Th1/Th2/Th17/Treg type **a** cytokines, **b** transcription factors, and **c** cytokines vs transcription factors in the liver of buffaloes infected with *Fasciola gigantica* (tested samples, *n* = 29) by pairwise comparison. Connecting lines illustrate positive (*r* \> 0) and negative (*r* \< 0) correlation indices
Expression profiles of transcription factors {#Sec13}
--------------------------------------------
*Fasciola gigantica* infection altered the expression of T-bet, GATA-3 and ROR-γτ, with the expression of Foxp3 being least affected (Fig. [2b](#Fig2){ref-type="fig"} and Table [4](#Tab4){ref-type="table"}). Compared with the control, mRNA of T-bet was highly expressed at 98 dpi (*P* = 0.0113), while the expression of the GATA-3, Foxp3 and ROR-γτ genes was upregulated at 28 dpi (*P* = 0.0035, *P* = 0.0424 and *P* = 0.0002, respectively). Spearman's correlation analysis indicated that the expression of T-bet and Foxp3 was correlated (*r* ~(29)~ = 0.63, *P* = 0.0011). Positive correlation was also found between the expression of ROR-γτ and GATA-3 (*r* ~(29)~ = 0.43, *P* = 0.039) or Foxp3 (*r* ~(29)~ = 0.42, *P* = 0.0439) (Fig. [3b](#Fig3){ref-type="fig"}). As shown in Fig. [3c](#Fig3){ref-type="fig"}, the mRNA expression of GATA-3 correlated significantly with that of IL-4 (*r* ~(29)~ = 0.54, *P* = 0.0072), but was less correlated with another Th2 cytokine IL-13 (*r* ~(29)~ = 0.017, *P* = 0.9387), whereas, the expression of ROR-γτ was correlated with that of IL-17A (*r* ~(29)~ = 0.54, *P* = 0.0072). There was a tendency of correlation between the expression of T-bet and Th1 cytokine IFN-γ (*r* ~(29)~ = 0.22, *P* = 0.305) or IL-12B (*r* ~(29)~ = 0.33, *P* = 0.1265) and between the expression of Foxp3 and IL-10 (*r* ~(29)~ = 0.38, *P* = 0.0718).
Th1/Th2/Th17/Treg balance {#Sec14}
-------------------------
To better understand the temporal trends in gene expression of Th1, Th2, Th17 and Treg in the liver post-infection, the ratios between the expression of transcription factors were evaluated using the pairwise comparison method (Fig. [2c](#Fig2){ref-type="fig"}). The *RQ* ratio of T-bet/GATA-3 was employed as an indicator of Th1/Th2 cytokine gene expression pattern. This ratio was higher in the liver at 98 dpi when compared to the uninfected group (*P* = 0.0049). This analysis also revealed that Th1/Th17 (*P* = 0.0246) and Treg/Th17 (*P* = 0.0126) were both upregulated at 98 dpi, when compared with the samples tested at other time points (3--70 dpi). However, no significant change was found in either Th1/Treg, Th2/Treg, or Th2/Th17 expression ratios at any of the examined time points. These data suggest a predominance of the Th1 and Treg cellular immune response at 98 dpi.
Discussion {#Sec15}
==========
Our findings provide an increased understanding of the immuno-inflammatory response of buffaloes to *F. gigantica* infection through examining the correlation between gene expression of nine cytokines and four transcription factors in Th1, Th2, Th17 and Treg cells, together with the liver pathology during early, mid and late stages of infection. We initially characterized gross and histopathological changes in the liver at six time points following infection compared to uninfected control. Although infected buffaloes did not exhibit clinical signs, they developed hepatic gross pathologies similar to what have been observed in other studies \[[@CR15], [@CR19]\]. Histopathological features also agreed with previous studies on *F. gigantica* \[[@CR9], [@CR20], [@CR21]\] and *F. hepatica* \[[@CR22]\]. An unexpected finding was that none of the examined liver from infected animals showed calcium deposits. The difference in liver calcification between our study and previous reports may relate to variations in the experimental conditions, such as using different infectious doses, different parasite strains, the permissibility of the host species, the duration of the experimental infection, and using single infection vs repeated infections \[[@CR21]--[@CR23]\].
Next, we have shown that buffaloes showed a substantial liver immune response subsequent to oral challenge with 500 *F. gigantica* encysted metacercariae. Certain genes exhibited shifting temporal expression patterns as the infection proceeded. Expression of cytokine and transcription factor genes were decreased at 3 and 10 dpi, suggesting a general immunosuppression state to allow parasite colonization. Cytokine reduction during the early stage of *F. gigantica* infection has been reported previously \[[@CR24], [@CR25]\]. A previous study has documented the downregulation of MHC-II related genes, as well as the suppression of the host pro-inflammatory (Th1) immune response during early *F. gigantica* infection \[[@CR16]\]. This immunosuppression might be attributed to an increased expression of the immunosuppressive cytokine TGF-β. While TGF-β was upregulated at 3 dpi, Foxp3 (Treg transcription factor) was upregulated at a later stage of infection (28 dpi), indicating that Tregs (also known as Foxp3-expressing cells) play a more diverse role than causing early immunosuppression in regulating immune responses. Although increased TGF-β at 3 dpi may have promoted the generation of Foxp3-expressing Tregs, increased IL-6 at 28 dpi may have abrogated the suppressive function of Tregs \[[@CR26]\] by antagonizing TGF-β-induced Treg generation and stimulating Treg differentiation to Th17, as reported previously \[[@CR27]\].
The inflammatory cascades triggered by *F. gigantica* must be tightly coordinated in order to avoid severe liver pathology. Th2 cells mediate humoral responses via induction of IL-4, IL-5, IL-9, IL-10 and IL-13 cytokines, which are important for controlling extracellular parasites \[[@CR28]\]. In agreement with this, our results at 28 dpi showed a significant increase in the expression of the anti-inflammatory Th2-associated cytokines (IL-4, IL-13) and transcription factor (GATA-3), indicating the induction of Th2 response in liver tissue, probably to limit the fluke development and to balance the increased levels of inflammatory cytokines IL-1β, IL-6 and IL-17A. This response was expected because IL-4 is known to have the greatest effect in inducing Th2 differentiation \[[@CR29]\] and the GATA-3 transcription factor is a key regulator of Th2 differentiation \[[@CR30]\]. IL-4 stimulation may have activated STAT6, which is the major signal transducer in IL-4-mediated Th2 differentiation \[[@CR31]\]. STAT6 is known to promote Th2 differentiation by stimulating GATA-3 \[[@CR32]\]. Therefore, this coordinated response may be mediated by the parasite and/or the host in order to promote parasite survival while minimizing host hepatic damage. Th2-predominant response together with the suppression of Th1/Th17 response has been reported previously \[[@CR33]--[@CR36]\]. Interestingly, some molecules secreted by *Fasciola* species can suppress the differentiation of Th17 cells independently of Th2 cells differentiation, by altering the function of dendritic cells \[[@CR37]\].
Although CD4^+^ Th17 cells play a role in the control of a variety of parasitic infections, the role of IL-17 produced by CD4^+^ Th17 cells in immunity to *F. gignatica* has not been clearly defined \[[@CR38], [@CR39]\]. Our current and previous studies \[[@CR8]\], demostrated that IL-17 may play a role in the inflammatory process during early *F. gigantica* infection. Th17 cells are a newly-identified class of effector T cell, which produces IL-17A and IL-17F. Th17 cells can contribute to resistance to many intracellular \[[@CR40]\] and extracellular parasites \[[@CR8]\]. In addition to controlling infection, IL-17 expression has been associated with inflammatory and allergic responses \[[@CR41]\]. TGF-β and IL-6 act cooperatively and non-redundantly to promote Th17 activity \[[@CR27]\]. In agreement with this, TGF-β was upregulated at 3 dpi, and both IL-6 and IL-17A were upregulated at 28 dpi. IL-6 was found to be important in suppressing Treg generation in order to promote Th17 differentiation as discussed above. The expression of retinoic acid-related orphan receptors ROR-γt, a key transcription factor in Th17 differentiation \[[@CR42], [@CR43]\], was positively correlated with the *RQ* level of IL-17A at 28 dpi.
At 70 and 98 dpi a mixed Th1/Th2 type profile, supported by co-dominance of IFN-γ, IL-1β, IL-12B, IL-6, IL-4 and IL-13 cytokines, was observed as consistent with the result obtained by Kumar \[[@CR44]\]. The high expression of Th1 cytokines (IL-1β and IL-6), suggesting acute inflammatory activity, correlated with hemorrhage, necrosis and severe infiltration of inflammatory cells in the liver. Th2 cytokines, on the other hand, often lead to chronic inflammatory response by promoting fibrosis and tissue remodeling \[[@CR45], [@CR46]\]. The accumulation of fibroblasts during chronic infection suggests that Th2 cytokine-mediated tissue repair was taking place. This state of immune homeostasis is probably required for parasite persistence and was correlated with an increased number of flukes. The T-bet/GATA-3 ratio, a marker of Th1/Th2 \[[@CR47]\], was not altered during acute infection, but was significantly increased when the infection became chronic, suggesting more bias towards a Th1-mediated inflammatory response during late infection. This observation contradicts previous studies in mice, cattle and buffaloes infected with *F. gigantica* that reported a mixed Th1/Th2 response with a predominance of a Th2-biased pattern \[[@CR48], [@CR49]\]. Other studies have reported a Th2 response in early infection and an increased Th0-type response during chronic *F. gigantica* infection in cattle \[[@CR35]\] and immunosuppression during chronic *F. gigantica* infection in buffaloes \[[@CR8]\]. These differing findings between studies may result from different experimental conditions.
A synergism between IL-1β and IL-17 has previously been reported \[[@CR50]\]. In our study, a strong correlation was observed between IL-17A and IL-4 and IL-10, as well as between ROR-γτ and GATA-3 and Foxp3, implying that induction of Th17 was paralleled by induction of both Th2 and Treg. This refutes the antagonistic relationship previously reported between a polarized Th2/Treg immune response and suppression of Th1/Th17 cytokines \[[@CR11], [@CR51]--[@CR53]\]. A gradual upregulation of IL-10 mRNA expression was observed during early infection, in agreement with our previous finding in buffalo's serum \[[@CR8]\]. IL-10 levels significantly increased at 98 dpi, which was also reported in a study on *F. hepatica*-infected sheep \[[@CR54]\]. IL-10 is considered to be a regulatory cytokine rather than a Th2 type cytokine \[[@CR55], [@CR56]\]. High levels of IL-10 can suppress excessive inflammation induced by the parasite \[[@CR57], [@CR58]\]. In our study, elevated levels of IL-10 and the concurrent upregulation of Th1 and Th2 cytokine genes in the liver suggest that IL-10 was required for the Th2 response to *F. gigantica* infection, instead of serving an anti-inflammatory function. Our data also showed that the *RQ* level of Foxp3 increased at 28 dpi. This level remained relatively high over the course of infection, which may have contributed to the Th1/Th2 balance in order to enhance parasite survival in the liver and to protect the host from over-inflammation \[[@CR59]\]. Foxp3-expressing Treg cells are critical to maintaining immune homeostasis by minimizing tissue pathology, while modulating host immune response against helminth infections \[[@CR60], [@CR61]\]. The monitoring of Th17 and Treg cell offers a promising new perspective on the pathogenesis of *F. gigantica* infection and deserves further exploration.
Conclusions {#Sec16}
===========
This is the first study to characterize the correlation between the expression of Th1, Th2, Th17, and Treg cytokines and transcription factors with liver pathology in buffaloes, during the course of experimental *F. gigantica* infection. The expression patterns for the examined genes indicated that there were periods of differential regulation during *F. gigantica* infection, which may suggest either a mechanism of immune evasion based on modulation of transcription or a mechanism used by host tissue to limit the infection and tissue damage. Gene expression profiles showed a significant T-cell adaptive immune suppression between 3 and 10 dpi to facilitate parasite colonization and a more substantial transcript differential expression change (mixed Th1/Th2 immune response) from 28 to 70 dpi, which may contribute to the parasite survival while minimizing host tissue damage. During late infection (98 dpi), the response was biased towards Th1/Treg in order to limit the host's Th2 protective response and promote chronic infection, which conincided with an increase in the number of recovered flukes and greater histopathological changes and fibrosis. Clearly, regulation of immune cytokine and transcription factor gene expression during *F. gigantica* infection is a complex mechanism involving a variety of parasite-specific and host-specific factors, perhaps depending on the parasite's needs for survival during various phases of infection. Research into the mechanisms governing differential expression of these immune-related genes, may shed light on the actual role of these immune regulatory factors in the pathogenesis of *F. gigantica* infection in buffaloes. Much remains to be elucidated about the role of various types of inflammatory responses, and the cross-talk between types of T-helper responses, in fasciolosis.
ANOVA
: Analysis of variance
cDNA
: Complementary deoxyribonucleic acid
*Cq*
: Quantification cycle
DFd
: Degrees of freedom denominator
DFn
: Degrees of freedom numerator
DNA
: Deoxyribonucleic acid
DPI
: Days post-infection
ELISA
: Enzyme linked immunosorbent assay
EM
: Encysted metacercariae
Foxp3
: Forkhead box P3
GAPDH
: Glyceraldethde-3-phosphate dehydrogenase
GATA
: GATA binding protein
H&E
: Hematoxylin and eosin
IFN
: Interferon
IL
: Interleukin
ITS
: Internal transcribed spacer
LSD
: Least significant difference
mRNA
: Messenger ribonucleic acid
qRT-PCR
: Quantitative reverse transcription-polymerase chain reaction
RBC
: Red blood cells
rDNA
: Ribosomal DNA
RNA
: Ribonucleic acid
ROR
: Retinoid-related orphan nuclear receptor
RQ
: Relative quantitation
SEM
: Standard error of the mean
T-bet
: T-box expressed in T cells
TGF
: Transforming growth factor
Th
: T helper
Treg
: Regulatory T cell
WPI
: Weeks post-infection
We are grateful to Xiao-Xuan Zhang, Wen-Bin Zheng and Jian-Gang Ma from Lanzhou Veterinary Research Institute, Chinese Academy of Agricultural Sciences, and Yang-Qun Kang, Xue-Fang Mei, Chang-Hong Hu, Yu Zhang, Yao-Yao Zhang, and Yi-Ying Liang from Guangxi University for their technical assistance.
Funding {#FPar1}
=======
Project financial support was provided by National Key Basic Research Program (973 Program) of China (Grant No. 2015CB150300) and by National Natural Science Foundation of China (Grant No. 31260605).
Availability of data and materials {#FPar2}
==================================
The data supporting the findings of this article are included within the article.
WS, WYH and XQZ conceived and designed the study, and critically revised the paper. WS, ZYW, FKZ and KJL prepared the experimental samples. WS, ZYW and ZAS performed the experiments. WS, ZYW, FKZ and DYW analyzed the data. WS drafted the manuscript. HME helped in the data analysis and critical revision of the manuscript. All authors read and approved the final manuscript.
Ethics approval {#FPar3}
===============
The study design was reviewed and approved by the Animal Ethics Committee of Guangxi University. Animals used in the study were handled in accordance with good animal practices as required by the Animal Ethics Procedures and Guidelines of the People's Republic of China.
Consent for publication {#FPar4}
=======================
Not applicable.
Competing interests {#FPar5}
===================
The authors declare that they have no competing interests.
Publisher's Note {#FPar6}
================
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
| {
"pile_set_name": "PubMed Central"
} |
All sequences were deposited in GenBank Sequence Read Archives ([www.ncbi.nlm.nih.gov](http://www.ncbi.nlm.nih.gov)) under accession number SRP001226. All environmental metadata are available in S1 Table.
Introduction {#sec001}
============
Globally, over a quarter-million square kilometers of marine ecosystems are threatened by low dissolved oxygen (DO) levels, or hypoxia, which can result in the exclusion or death of resident macroorganisms creating so-called 'dead-zones' \[[@pone.0135731.ref001]\]. The number of hypoxic coastal areas has grown at an exponential rate of 5.54% year^-1^ \[[@pone.0135731.ref002]\], largely due to increased eutrophication \[[@pone.0135731.ref003],[@pone.0135731.ref004]\]. Seasonal hypoxia is detrimental to coastal ecosystems and is known to have negative effects on fish \[[@pone.0135731.ref005]\], crustacea \[[@pone.0135731.ref006]\], benthic invertebrate communities \[[@pone.0135731.ref007]\], and trophic interactions and energy flow \[[@pone.0135731.ref008]\]. However, Bacteria and Archaea remain active in hypoxic waters and perform important functions in mineralization of organic matter and other biogeochemical cycling \[[@pone.0135731.ref001],[@pone.0135731.ref009]\].
Extensive research has gone into understanding the thresholds of hypoxia for management strategies aimed at avoiding catastrophic collapses of ecosystems. In a review of over 800 published experiments, Vaquer-Sunyer and Duarte \[[@pone.0135731.ref002]\] report that published thresholds of hypoxia span a broad range from 0.29 mg O~2~ L^-1^ to 4 mg O~2~ L^-1^, but on average, sublethal effects on marine benthic macroorganisms occur when DO concentration drops below 2.61 ± 0.17 mg O~2~ L^-1^ and lethal effects occur at 2.05 ± 0.09 mg O~2~ L^-1^, with many taxa of Crustacea and Fishes demonstrating negative effects of hypoxia near 4 mg O~2~ L^-1^. It is worth noting, however, that thresholds for hypoxia likely differ between coastal open ocean hypoxic zones, as the organisms in each may be adapted to different durations of hypoxia \[[@pone.0135731.ref010]\]. Lacking from our current knowledge is whether a threshold of hypoxia for bacterial communities exists, and, if so, at what DO concentration do bacterial communities experience taxonomic and functional shifts.
Understanding the effect of hypoxia on lower trophic levels including plankton assemblages is crucial for elucidating whole ecosystem effects and interactions among trophic levels. Many microbial studies have focused on truly oxygen deficient zones where anaerobic metabolisms dominate including the Eastern Tropical South Pacific \[[@pone.0135731.ref011]\] and North Pacific \[[@pone.0135731.ref012]\], Arabian Sea \[[@pone.0135731.ref013],[@pone.0135731.ref014]\], Cariaco Basin \[[@pone.0135731.ref015],[@pone.0135731.ref016]\], Baltic Sea \[[@pone.0135731.ref017]\], Black Sea \[[@pone.0135731.ref018]--[@pone.0135731.ref020]\], and Saanich Inlet \[[@pone.0135731.ref021]\]. A recent meta-analysis examining global patterns of microbial community composition in oxygen-minimum zones found that particular bacterial and archaeal taxa, specifically Proteobacteria, Bacteroidetes, marine group A, Actinobacteria, and Planctomycetes, are common in suboxic waters and that these taxa tend to co-occur in a non-random pattern along the oxycline \[[@pone.0135731.ref022]\]. Only some of the members of these suboxic-associated phyla are known to use electron acceptors alternative to oxygen, which might suggest that mechanisms other than oxygen-depletion drive changes in the bacterial community. Though these previous studies have demonstrated that particular bacterial taxa persist in oxygen-minimum zones, the threshold concentration of DO at which bacterial communities change remains unclear. The bacterial community that exists at dysoxic (0.66--2.96 mg O~2~ L^-1^) and suboxic (0.03--0.66 mg O~2~ L^-1^) levels has the potential to become increasingly important in global biogeochemical cycles as oxygen-minimum zones in both the open and coastal ocean continue to expand with climate change \[[@pone.0135731.ref023]\], thus emphasizing the importance of defining a threshold DO concentration for Bacteria.
We examined spatial and temporal variation in bacterial communities in Hood Canal, WA, USA between April and October of 2007. Hood Canal is a long and narrow glacial fjord located 80 miles west of Seattle, Washington, USA ([Fig 1A](#pone.0135731.g001){ref-type="fig"}), that seasonally experiences periods of low and even completely depleted DO concentrations as a result of naturally occurring physical and hydrographical conditions \[[@pone.0135731.ref024]\]. This system has relatively small-scale, enclosed circulation patterns and rapidly changing abiotic conditions. Throughout the year, Hood Canal remains highly stratified, which limits the movement of oxygenated waters \[[@pone.0135731.ref025]\]. Furthermore, human activity over the past few decades has exacerbated nutrient loading particularly in the southern reaches of Hood Canal \[[@pone.0135731.ref026]\], in turn increasing the frequency and duration of recent hypoxia events \[[@pone.0135731.ref027]\]. Fish and macroinvertebrates in Hood Canal are directly affected by hypoxia through mortality or distributional shifts \[[@pone.0135731.ref028],[@pone.0135731.ref029]\]. Particularly significant fish kill events near the southern reaches of the canal have occurred in the Falls of 2002, 2003, 2004, and 2006 as well as in the Spring of 2006 when the oxygen-deprived deep-water mass shoaled to the surface as a result of a combination of southerly winds and fresh-water intrusion from the Puget Sound at the North end of the canal \[[@pone.0135731.ref030]\]. Additionally, mesozooplankton assemblages in Hood Canal demonstrated altered composition and behavior when DO dropped below \~1.5 mg O~2~ L^-1^ \[[@pone.0135731.ref031]\]. These shifts have implications for predator-prey dynamics and trophic energy transfer and suggest that hypoxia may affect the entire biologic community in Hood Canal. We examined the bacterial communities associated with the seasonal variability in abiotic factors in Hood Canal, including depleted DO concentration. We hypothesized that bacterial communities in Hood Canal demonstrate non-random changes in composition and taxa richness related to variation in abiotic factors, and that these community patterns are associated particularly with changing levels of DO.
). Note changes in scale on both the x- and y-axes.](pone.0135731.g001){#pone.0135731.g001}
Methods {#sec002}
=======
Sample collection {#sec003}
-----------------
High resolution depth profiles of DO and chlorophyll *a* for two mid Hood Canal, WA stations, Hama Hama and Sister's Point, were obtained via Oceanic Remote Chemical Analyzer (ORCA) buoys maintained by the Northwest Association of Networked Ocean Observing Systems ([http://www.nanoos.org](http://www.nanoos.org/)) \[[@pone.0135731.ref032]\]. Water samples were collected during April, June, and October of 2007 into Niskin bottles mounted on a rosette, equipped with a CTD (conductivity, temperature, and depth) sampler (Sea-Bird Electronics, Bellevue, WA, USA) that measured in-situ temperature and salinity. April and October samples were collected from two stations near the middle of Hood Canal: Hama Hama and Sister's Point. During the June sample collection, two additional stations (Bangor and Lynch Cove) were added in order to survey a full north to south transect of the Canal. At each sampling station and time, samples were collected both at five meters depth (herein referred to as surface samples) as well as ten meters above the bottom substrate (herein referred to as deep samples). These so-called deep samples varied in depth from 140−145 m at Hama Hama, 125 m at Bangor, 40−50 m at Sister's Point, and 12 m at Lynch Cove, as the depth of the water column decreases near the southern reaches of the canal. The timing of the collection points were intended to coincide with the late spring algal bloom (April), subsequent die-off (June), and a smaller autumn algal bloom (October), which affect DO concentration and nutrient availability ([S1 Table](#pone.0135731.s003){ref-type="supplementary-material"}). A field permit was not required for the sampling in this study, as there are no protections in place for the waters or organisms that would be impacted by our sampling and sampling did not affect any endangered or protected species.
Immediately upon retrieval of the Niskin bottles, DO samples were collected using surgical tubing into calibrated glass bottles with ground glass stoppers and were analyzed using a modified Winkler titration \[[@pone.0135731.ref033]\]. The remaining water was used for additional measurements of biotic and abiotic parameters. For inorganic nutrients, 50 mL of water was passed through a 0.45-μm syringe filter and immediately frozen until being analyzed at the University of Washington's Oceanography Marine Chemistry Lab for NO~3~ ^-^ (detection limit 0.15 μM), NO~2~ ^-^ (detection limit 0.02 μM), NH~4~ ^+^ (detection limit 0.12 μM), PO~4~ ^3-^ (detection limit 0.03 μM), and SiOH~4~ (detection limit 0.59 μM) following the Protocols for the Joint Global Ocean Flux Survey (JGOFS) Core Measurements (1994) \[[@pone.0135731.ref034]\]. To measure bacterial abundance, 5 mL of water was preserved in formaldehyde with a final concentration of 1% and stored at -80°C until 0.5--1.0 mL aliquots were filtered onto 0.22-μm polycarbonate filters and stained with 4',6-diamidino-2-phenylindole (DAPI). The DAPI filters were viewed under a Nikon Eclipse 80i with UV light from an X-cite Series EXFO to count the number of cells present in fifteen randomly selected fields using NIS-Elements BR 3.0 software then averaged to quantify the total bacterial abundance. Principal components analyses and all other statistical analyses were conducted in *R* using the *vegan* package (<http://cran.r-project.org/web/packages/vegan/index.html>), unless noted otherwise. Metadata are provided in [S1 Table](#pone.0135731.s003){ref-type="supplementary-material"}.
DNA extraction and processing {#sec004}
-----------------------------
For bacterial community analyses, 300 mL of water was filtered onto 0.22-μm Supor filters and preserved with 3 mL of solution of 0.5 M NaCl, 10 mM Tris, and 100 mM EDTA, then immediately frozen at -80°C. A Qiagen DNeasy Tissure Kit was used to extract DNA from the filters following the manufacturer's instructions for extraction from gram-negative Bacteria (Qiagen Valencia, CA). The DNA extracts were sent to MBL's Keck Facility for PCR amplification and sequencing on the GS-FLX Titanium 454 platform (see Sogin, et al., 2006 for previously described methods \[[@pone.0135731.ref035]\]). Briefly, the V6 region of the 16S rRNA gene was amplified via polymerase chain reaction (PCR) using a pool of bacterial specific primers, excluding amplification of Archaea (see Huse, et al., 2010 for description of V6 primer pool \[[@pone.0135731.ref036]\]). Sequencing was performed in conjunction with the International Census of Marine Microbes (ICoMM) project ([http://icomm.mbl.edu](http://icomm.mbl.edu/)), which uses a standardized 454 pyrosequencing pipeline to sequence marine microbial samples from across the globe \[[@pone.0135731.ref037]\]. All sequences were deposited in GenBank Sequence Read Archives ([www.ncbi.nlm.nih.gov](http://www.ncbi.nlm.nih.gov/)) under accession number SRP001226.
As part of the ICoMM project, the sequence data were initially analyzed using a standardized pipeline \[[@pone.0135731.ref035],[@pone.0135731.ref036],[@pone.0135731.ref038]\]. Sequences were pre-clustered using a single-linkage algorithm to smooth sequencing errors and reduce noise. Sequences were then clustered using average-linkage into operational taxonomic units (OTUs) based on 97% sequence identity. Prior to community analyses, we removed sequences that were highly similar to chloroplast sequences.
Statistical analyses {#sec005}
--------------------
Abiotic factors, including NO~3~ ^-^, NO~2~ ^-^, NH~4~ ^+^, PO~4~ ^3-^, SiO~4~ ^2-^, salinity, temperature, and DO concentration, were log-transformed prior to analyses to meet assumptions of normality and homogenization of variance. Euclidean distance between samples was calculated for environmental variables \[[@pone.0135731.ref039]\]. Principal components analysis (PCA) was performed to assess dominant trends in environmental variables between samples across depths, sites, and seasons. A Monte Carlo randomization test was used to test the significance of each principal component at an alpha level of 0.05.
To minimize the effect of sequencing effort on richness estimates, all samples were randomly subsampled down to 3663 sequences, which was the number of sequences of the sample with the fewest number of sequences ([Table 1](#pone.0135731.t001){ref-type="table"}), the richness measures were calculated, then the process was repeated and averaged over 1000 iterations. Observed taxa richness, or the number of OTUs identified at the pre-defined sequence identity, as well as the Chao richness estimator were calculated using the subsampling method and averaged over 1000 iterations ([Table 1](#pone.0135731.t001){ref-type="table"}). We used multiple regression modeling with a requirement of α \< 0.05 for a variable to enter the model to determine which abiotic factors best described the changes in taxa richness, evenness (Pielou's J), and proportion of functional groups across all samples.
10.1371/journal.pone.0135731.t001
###### Number of sequences obtained in each sample, as well as the observed and Chao estimates of taxa richness and Pielou's J index of taxa evenness for all Hood Canal samples based on an operational taxonomic unit cutoff of 97% sequence identity.
{#pone.0135731.t001g}
Sample Location Season Depth Number of sequences Observed richness Chao Pielou\'s J
---------- ---------------- --------- --------- --------------------- ------------------- ------ -------------
HH_APR_S Hama Hama April Surface 14743 155 376 0.58
HH_APR_D Hama Hama April Deep 14897 512 980 0.75
SP_APR_S Sister's Point April Surface 10482 173 353 0.68
SP_APR_D Sister's Point April Deep 14739 554 1303 0.75
BA_JUN_S Bangor June Surface 14019 239 430 0.71
BA_JUN_D Bangor June Deep 11108 302 595 0.72
HH_JUN_S Hama Hama June Surface 6729 262 537 0.73
HH_JUN_D Hama Hama June Deep 11540 467 840 0.74
SP_JUN_S Sister's Point June Surface 7077 254 516 0.70
SP_JUN_D Sister's Point June Deep 10368 546 1219 0.74
LC_JUN_S Lynch Cove June Surface 11595 193 342 0.72
LC_JUN_D Lynch Cove June Deep 8379 339 600 0.70
HH_OCT_S Hama Hama October Surface 3663 447 1043 0.71
HH_OCT_D Hama Hama October Deep 11445 498 966 0.75
SP_OCT_S Sister's Point October Surface 4446 238 576 0.62
SP_OCT_D Sister's Point October Deep 9795 590 1193 0.76
BA = Bangor; SP = Sister's Point; HH = Hama Hama; LC = Lynch Cove; APR = April; JUN = June; OCT = October; S = Surface; D = Deep.
We used multivariate statistical analyses to identify important drivers of the relationship between bacterial community composition and environmental factors. We selected a Sorensen abundance-based dissimilarity measure because it incorporates probability of detection of taxa shared between two samples to account for under-sampling, which is common when studying microbial communities \[[@pone.0135731.ref040]\]. Similarly to the alpha diversity calculations, the bacterial community matrix was randomly subsampled to the number of sequences of the smallest sample then the Sorensen abundance-based similarity was calculated between samples to account for sequencing effort bias \[[@pone.0135731.ref041],[@pone.0135731.ref042]\] and averaged over 1000 iterations.
Analysis of similarity (ANOSIM) tests examined the effects of factors including depth, site, season, or DO (high or low) on bacterial community composition of each sample and significance was tested using 999 random permutations of the community composition dataset. A non-parametric iterative *BIO-ENV* analysis was employed to select the set of abiotic variables most correlated with changes in bacterial community composition \[[@pone.0135731.ref043]\].
Threshold Indicator Taxa Analysis (TITAN) detects changes in taxa distributions along an environmental gradient that exists over space and/or time \[[@pone.0135731.ref044]\]. The analysis assesses the synchrony among taxa change points as evidence for community thresholds. Prior to TITAN, the sixteen Hood Canal samples were randomly subsampled to the smallest number of sequences present in a single sample (3663 sequences). In order to accurately calculate an indicator value (IndVal) for each taxon, the taxon must be present in three or more samples. Therefore, taxa with narrow distributions (present in less than three samples) were eliminated from each subsampled dataset. Finally, all bacterial abundance data were log-transformed to down-weight highly abundant types. TITAN was run in *R* using a minimum split of three samples, 100 random permutations to calculate IndVal p-values, and 100 random permutations for the bootstrapping calculation. Subsampling and TITAN analysis was repeated ten times to confirm that stochastic changes in sampling do not affect the results of TITAN.
To validate the DO threshold identified by TITAN, we used a non-parametric ANOSIM test grouping samples either by depth or DO level split at either 4 mg O~2~ L^-1^ or 6 mg O~2~ L^-1^. The R-values from ANOSIM, which describes how well samples group by factor based on community composition, were compared to determine the best factor for grouping.
Finally, hierarchical clustering using average linkage was used to cluster samples based on dissimilarity of community composition quantified by the Sorensen abundance-based estimator to identify the DO concentration at which the community composition changed significantly. All three analyses, TITAN, ANOSIM, and clustering were performed in *R* using scripts from Baker and King \[[@pone.0135731.ref044]\]; *vegan*, *cluster*, and *pvclust* packages and scripts written for this study, which are available upon request.
A similarity percentage (SIMPER) analysis was used to identify bacterial taxa contributing most to the similarity within groups of samples defined by DO and thus identify taxa strongly associated with either high or low DO \[[@pone.0135731.ref045]\]. SIMPER analysis identifies taxa that contribute the most to the similarity in community composition between samples of the same group, or similarly, taxa that contribute the most to dissimilarity in community composition across groups of samples. Thus, the taxa identified by SIMPER are significant indicators of a particular factor, in this case low DO. Dissolved oxygen is usually higher in surface waters than in deep waters, so to separate a depth effect on taxa distribution from an oxygen effect, the SIMPER analysis was run a second time using sampling depth as a factor rather than oxygen level. To test the significance of an oxygen effect over a depth effect, a generalized linear model was run using an intercept of 0 and a slope of 1, which would imply a null hypothesis that there is no difference between a depth or DO effect. Studentized residuals were calculated for each taxon, and a Bonferroni correction was applied to calculate a 95% confidence interval because of the high number of taxa or data points. Taxa that demonstrated a significant effect of DO level rather than sampling depth were those that fell above the 95% confidence interval, and thus had a studentized residual greater than 4.2169. For taxa identified to be significant indicators of high or low DO, their taxonomy was further resolved by using the V6-16S rRNA gene sequences to perform both a blastn search against the NCBI nr/nt database in (<http://blast.ncbi.nlm.nih.gov/Blast>) and also by aligning the V6-16S rRNA gene sequences against the arb-silva database using the online SINA v1.2.11 aligner (<http://www.arb-silva.de/aligner/>) \[[@pone.0135731.ref046]\]. All alignments were performed in June and July of 2014.
Results {#sec006}
=======
Environmental context {#sec007}
---------------------
Using high-resolution measurements of DO and chlorophyll *a* ([Fig 1B](#pone.0135731.g001){ref-type="fig"}), it was apparent that patterns of hypoxia in 2007 were not as extreme as in past years when low DO had caused fish kills. Yet, our samples captured a wide range of abiotic conditions and likely represent typical seasonal variability in Hood Canal. Water samples were collected from Hama Hama and Sister's Point ([Fig 1A](#pone.0135731.g001){ref-type="fig"}) in April, directly following the major spring phytoplankton bloom as indicated by the heightened level of chlorophyll *a* in the surface water layer at both stations, marked by high DO concentrations (12.32−13.18 mg O~2~ L^-1^) and low nutrient concentrations in the surface waters (0.03−0.73 μM NO~3~ ^-^) and low DO concentrations (3.59−3.94 mg O~2~ L^-1^) and high nutrient concentrations in deep waters ([Fig 1B](#pone.0135731.g001){ref-type="fig"}). June samples were collected at the end of the phytoplankton bloom, and thus the surface waters were characterized by slightly lower DO (8.66−10.25 mg O~2~ L^-1^) and nutrient concentrations than in April. At the same time, the deep waters contained low DO (2.95−7.12 mg O~2~ L^-1^) and high nutrient concentrations. October samples were collected after a lesser fall phytoplankton bloom. Dissolved oxygen and nutrient concentrations in October were in the mid to high range, relative to previous months, in both water depths, which were less stratified than earlier in the year ([S2 Fig](#pone.0135731.s002){ref-type="supplementary-material"} and [S1 Table](#pone.0135731.s003){ref-type="supplementary-material"}). Bacterial abundances across all stations and time points ranged from 2.16 × 10^5^ cells mL^-1^ to 2.70 × 10^6^ cells mL^-1^, with an average abundance of 8.46 × 10^5^ cells mL^-1^ across all samples.
Using the environmental variables measured at the time of water sample collection, a principal components analysis (PCA) was conducted to visualize abiotic factors driving the differences among samples ([Fig 2](#pone.0135731.g002){ref-type="fig"}). Nitrate, DO, salinity, and phosphate were most correlated with PCA axis one, which explained 57% of the total measured abiotic variation among samples. Samples appear to fall into two groups along axis one, mostly described by depth. An additional 19% of the total variation in abiotic parameters was explained by PCA axis two, which was most correlated with ammonium concentration.
{#pone.0135731.g002}
Bacterial alpha diversity in Hood Canal, WA {#sec008}
-------------------------------------------
Observed bacterial taxa richness, or the number of OTUs (defined by 97% sequence similarity clusters) recovered, in Hood Canal ranged from 155 OTUs to 590 OTUs per sample and was significantly higher in the deep-water samples than in the surface water samples by an average factor of two (F~1,14~ = 22.55; p \< 0.01) ([Table 1](#pone.0135731.t001){ref-type="table"}, [Table 2](#pone.0135731.t002){ref-type="table"}, [Fig 3](#pone.0135731.g003){ref-type="fig"}). Neither station (horizontal space) nor season had a significant effect on bacterial richness (results not shown). By applying a forward-selection multiple regression model using all abiotic variables measured in this study, DO concentration alone best explained the variation in observed bacterial richness across samples (adjusted R^2^ = 0.738, p \< 0.01, [Fig 3](#pone.0135731.g003){ref-type="fig"}). Taxonomic richness peaked at a DO concentration of 4 mg O~2~ L^-1^ ([Fig 3](#pone.0135731.g003){ref-type="fig"}).
{#pone.0135731.g003}
10.1371/journal.pone.0135731.t002
###### Continuous abiotic variables affect the diversity and composition of bacterial communities in Hood Canal, WA, USA (α = 0.05 for all tests).
The strength of richness and evenness, measured by Pielou's J, models was assessed by the adjusted R^2^ value from multiple linear regressions, and the strength of the composition models was assessed by the Spearman's rho correlation coefficient from BIO-ENV analyses.
{#pone.0135731.t002g}
Response Classification Model Adjusted R^2^ Spearman\'s rho
------------- --------------------- ---------------------------- --------------- -----------------
Richness Bacterial domain DO(−) 0.738
Alphaproteobacteria Sal(+) 0.585
Flavobacteria DO(−) 0.705
Gammaproteobacteria DO(−) 0.759
Pielou\'s J Bacterial domain PO~4~(+) 0.524
Alphaproteobacteria NO~3~(−), Si(−) 0.837
Flavobacteria PO~4~(+), Sal(+), NO~3~(−) 0.854
Gammaproteobacteria DO(−) 0.429
Composition Bacterial domain DO 0.732
Alphaproteobacteria DO, PO~4~ 0.692
Flavobacteria DO 0.765
Gammaproteobacteria DO 0.792
Across all samples, over 57% of the sequences clustered into only 42 unique OTUs (defined by 97% sequence identity), while less than 2% of the sequences clustered into over 2300 unique OTUs. The abundant OTUs were typically widespread across all the samples, whereas each of the less abundant OTUs was detected in one or two samples. Uneven distribution of individuals into taxa is not unusual in bacterial communities \[[@pone.0135731.ref035]\], which we further explored by calculating an index of taxon evenness, Pielou's J ([Table 1](#pone.0135731.t001){ref-type="table"}) that provided an additional window into bacterial community structure. Evenness of taxa was higher in deep waters than in the surface water (F~1,14~ = 9.517, p \< 0.01), while neither season nor sampling site had an effect on the evenness of the bacterial community ([Table 2](#pone.0135731.t002){ref-type="table"}). Using multiple regression modeling, taxa evenness was best explained by PO~4~ ^3-^ (adjusted R^2^ = 0.524, p \< 0.01, [Table 2](#pone.0135731.t002){ref-type="table"}).
Although an uneven distribution of individuals into taxa is common in bacterial communities \[[@pone.0135731.ref035]\], changes in the relative proportion of the dominant classes provide insight into factors impacting bacterial community structure ([Fig 4](#pone.0135731.g004){ref-type="fig"}). Sampling depth explained a significant amount of the variation in relative proportion of the fifteen most abundant classes across samples (G-test; G = 37.585, p \< 0.01), whereas season or horizontal space had no significant effect (alpha \< 0.05). Of the most abundant bacterial classes, the majority of sequences recovered from all samples belonged to three classes: *Flavobacteria*, *Alphaproteobacteria*, and *Gammaproteobacteria*. In surface waters these classes comprised 88.5% of sequences, whereas in deep waters they comprised only 68.7% of sequences, which demonstrates higher taxa evenness in deep-water samples. The proportions of *Alphaproteobacteria* and *Gammaproteobacteria* were not significantly different between surface and deep waters; however, the proportion of Flavobacterial sequences was significantly higher in the surface water samples compared to deep-water (F~1,14~ = 21.463; p \< 0.01). The majority of *Flavobacteria* detected in the Hood Canal surface samples are members of the Flavobacteriaceae family, which are known to be highly associated with algal particles and are important players in organic matter degradation as part of the microbial loop \[[@pone.0135731.ref047],[@pone.0135731.ref048]\]. The high relative abundance of Flavobacteriaceae in the surface water samples of Hood Canal reflect the considerable algal bloom dynamics in the estuary and a recent pulse of organic matter particles into the euphotic zone. The five most abundant taxa in each sample included common, ubiquitous marine bacterial taxa such as those belonging to SAR11, SAR86, Flavobacteriaceae, Thalassobacter, SAR406, SAR324, and Rhodobacteriacaea. However, as DO concentration decreased across samples, the relative abundances of the dominant types changed as well ([Fig 4](#pone.0135731.g004){ref-type="fig"}).
{#pone.0135731.g004}
Patterns of functionally relevant taxa {#sec009}
--------------------------------------
In order to evaluate potential functional changes in the bacterial communities, we examined changes in the relative abundances of several groups of bacteria in our dataset known to be important in specific biogeochemical processes ([Table 3](#pone.0135731.t003){ref-type="table"}). Two groups of ammonia oxidizing bacteria (AOB), *Nitrosomonas* and *Nitrosococcus*, were detected in Hood Canal. The relative abundance of AOB was higher in deep water than surface water (F~1,14~ = 18.98, p \< 0.01) and also demonstrated a significant negative relationship with DO concentration (adjusted R^2^ = 0.5765, p \< 0.01). Given that we used bacterial-specific 16S rRNA primers, we did not detect ammonia-oxidizing Archaea, which are also known to inhabit Hood Canal \[[@pone.0135731.ref009]\]. The relative abundance of two groups of nitrite-oxidizing bacteria (NOB) in Hood Canal, *Nitrospira* and *Nitrospina*, was also significantly higher in the deep water than in the surface water (F~1,14~ = 24.9, p \< 0.01). Additionally, the relative abundance of NOB was not only strongly associated with a decrease in DO but also with decreased NH~4~ ^+^ concentration and temperature (adjusted R^2^ = 0.9315, p \< 0.01). There were no significant patterns of increased *Cyanobacteria* abundance by water depth, sampling season, or sample site; however, the relative abundance of this photosynthetic phylum was strongly associated with higher temperatures (adjusted R^2^ = 0.8812, p \< 0.01). The relative abundance of methylotrophic bacteria in Hood Canal, including Methylobacteriaceae, Methylococcaceae, Methylocystaceae, and Methylophilaceae, also demonstrated no significant patterns across water depths, season, or sampling site but their abundances did increase with decreasing DO and higher temperature (adjusted R^2^ = 0.5984, p \< 0.01).
10.1371/journal.pone.0135731.t003
###### Abiotic factors affect the average proportion of various bacterial functional groups.
A forward selection stepwise multiple linear regression modeling approach was employed with a requirement of p \< 0.05 to enter the model.
{#pone.0135731.t003g}
Functional group Total no. of reads Average proportion[\*](#t003fn003){ref-type="table-fn"} Model Adjusted R^2^
------------------- -------------------- --------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------------- ---------------
Nitrite oxidizers 2484 0.0136 DO([−](#t003fn002){ref-type="table-fn"}), NH~4~([−](#t003fn002){ref-type="table-fn"}), Temp([−](#t003fn002){ref-type="table-fn"}) 0.9315
Ammonia oxidizers 300 0.0012 DO([−](#t003fn002){ref-type="table-fn"}) 0.5765
Methylotrophs 701 0.0044 DO([−](#t003fn002){ref-type="table-fn"}), Temp([+](#t003fn001){ref-type="table-fn"}) 0.5984
Cyanobacteria 2241 0.0136 Temp([+](#t003fn001){ref-type="table-fn"}) 0.8812
(+) indicates a significant positive association (p \< 0.01)
(−) indicates a significant negative association (p \< 0.01)
\*Average proportion of the functional group within each sample.
A bacterial community threshold for DO {#sec010}
--------------------------------------
Change in bacterial community composition (BCC) was examined across all possible combinations of abiotic variables measured within this study, and DO concentration alone was most correlated with dissimilarities in the composition of bacterial communities among samples (Spearman's rho = 0.7320, [Table 2](#pone.0135731.t002){ref-type="table"}). Further, we examined the relevance of a threshold of DO concentration for bacterial communities in Hood Canal by applying three separate statistical tests: 1) a parametric assessment of the community data using hierarchical clustering, 2) a non-parametric assessment of the community data using an ANOSIM, and 3) an analysis of community thresholds determined by patterns of individual taxa using TITAN. Using a parametric clustering approach the BCC samples were plotted on a dendrogram based on the dissimilarity in BCC between pairs of samples. The clustering dendrogram reveals two distinct clusters of samples based solely on community composition data without any *a priori* knowledge of environmental factors ([Fig 5A](#pone.0135731.g005){ref-type="fig"}). Most of the surface water samples fell in the high DO cluster, except for the surface sample collected in October, which clustered with the lower DO samples as expected given the DO level at that sample was below 5.18 mg O~2~ L^-1^, the level that defined the low DO cluster. Similarly, all deep samples clustered with each other except for the deep sample collected in June from Bangor when DO concentrations were higher than 7.12 mg L^-1^. The two clusters suggest that DO is a strong factor in shaping BCC.
{#pone.0135731.g005}
In the second analysis, the samples were again plotted based on the pairwise dissimilarities in BCC using non-parametric multidimensional scaling (NDMS). The NMDS further supports that two groups of samples exist and that these two groups are delineated by DO, rather than depth alone ([Fig 5B](#pone.0135731.g005){ref-type="fig"}). An analysis of similarity (ANOSIM) confirmed that depth and DO are significant descriptors of the two groups of samples, but the association of DO with community composition is stronger than the association with depth (ANOSIM: R~DO~ = 0.827, p~DO~ \< 0.01; R~depth~ = 0.6144, p~depth~ \< 0.01).
Lastly, we examined taxon-specific responses to the DO gradient to identify a threshold concentration of DO at which both positively and negatively-associated taxa change, thus giving a community-wide change point. TITAN identified a threshold value of 6.15 mg O~2~ L^-1^ for both positively and negatively associated taxa ([Fig 6](#pone.0135731.g006){ref-type="fig"}). This implies that a shift in BCC occurs at 6.15 mg O~2~ L^-1^ (95% CI \[3.62, 8.78\]) from a community dominated by taxa that are positively-associated with DO to a community dominated by negatively-associated taxa. Though the confidence interval surrounding the community change-point predicted by TITAN is broad, the other two independent statistical tests (hierarchical clustering and NMDS) also suggest a community change point existing between 5.18 mg O~2~ L^-1^ and 7.12 mg O~2~ L^-1^.
{#pone.0135731.g006}
Dominant taxa in Hood Canal, WA and indicator taxa of low DO {#sec011}
------------------------------------------------------------
The relative proportions of the fifteen most abundant bacterial classes changed significantly as DO dropped below 6 mg O~2~ L^-1^ ([Fig 4](#pone.0135731.g004){ref-type="fig"}, G-test; G = 44.564, p \< 0.01). Shifts in the most abundant bacterial classes prompted further investigation into the individual taxa that may be representative of changing DO conditions. By comparing the abundances of taxa across our samples using pyrosequencing data, we identified taxa that may be indicators of low DO in Hood Canal. We identified 23 taxa that had significantly higher DO SIMPER values than depth values at an alpha level of 0.05 ([Fig 7A and 7B](#pone.0135731.g007){ref-type="fig"}), suggesting that these taxa drive dissimilarity between high and low DO communities and can thus be considered indicators of different DO conditions. Only two taxa identified from SIMPER analysis were indicators of high DO waters in Hood Canal. One of these taxa, Alphaproteobacteria_03_61, was most similar to members of the family Rhodobacteriacaea that are commonly detected in marine surface waters, particularly during phytoplankton blooms ([S2 Table](#pone.0135731.s004){ref-type="supplementary-material"}). The remaining 21 taxa were indicators of low DO waters in Hood Canal. The majority of the low DO indicator taxa were common marine bacterial groups including Rhodobacteracaea, SAR11, SAR406, *Nitrospina*, *Methylobacter*, and SAR86, which are often found across a wide range of marine habitats ([S2 Table](#pone.0135731.s004){ref-type="supplementary-material"}). These indicator taxa were abundant OTUs throughout the entire dataset but have significant shifts with decreasing DO, suggesting that these OTUs may be specialized members of common, ubiquitous groups that have undergone niche differentiation.
{#pone.0135731.g007}
Discussion {#sec012}
==========
On a global scale, DO content in coastal waters has decreased in recent decades and is arguably becoming one of the most important indicators of aquatic ecosystem health \[[@pone.0135731.ref001]\]. We investigated changes in bacterial community diversity and composition along a DO gradient in Hood Canal, WA, and, for multiple indices of bacterial community structure, we found a role for DO that was stronger than that of the other abiotic factors assessed here. Further, we identified a threshold concentration of DO at which the bacterial community composition shifted significantly, suggesting that the structure of bacterial communities is intimately linked to DO.
A strong negative relationship between bacterial richness and DO was detected across all samples ([Fig 3](#pone.0135731.g003){ref-type="fig"}), which highlights a pattern consistent with other systems including the seasonally anoxic Canadian fjord, Saanich Inlet \[[@pone.0135731.ref021]\] and in the open ocean water column in the Eastern Tropical North Pacific \[[@pone.0135731.ref012]\]. However, when DO concentration falls below hypoxic levels in Saanich Inlet and the Eastern Tropical North Pacific, bacterial richness has been shown to decrease sharply. Expanding our sampling gradient to extremely low DO conditions periodically present in Hood Canal could elucidate such a potential unimodal relationship between DO and richness ([Fig 3](#pone.0135731.g003){ref-type="fig"}). Indirect effects of oxygen-depletion on the bacterial community at low DO could explain the increasing richness at low DO. For instance, energy is transferred from macrofauna to microbes as oxygen is depleted as oxygen-dependent organisms either avoid hypoxic areas or die \[[@pone.0135731.ref001],[@pone.0135731.ref008],[@pone.0135731.ref049]\]. This transfer of energy to the bacterial community as DO decreases could also allow for higher taxa richness exemplifying a common theme in ecology where increasing bioavailable energy allows for coexistence of multiple taxa resulting in greater biodiversity \[[@pone.0135731.ref050]--[@pone.0135731.ref052]\].
Dissolved oxygen played a strong role in shaping bacterial community composition, as well. Our data suggest a significant shift in bacterial community composition at a threshold DO concentration between 5.18 and 7.12 mg O~2~ L^-1^, which is not only higher than the published thresholds of 2--4 mg O~2~ L^-1^ for higher trophic-level organisms \[[@pone.0135731.ref002]\], but also much higher than the minimum concentration for aerobic metabolisms (\< 96 ng O~2~ L^-1^) \[[@pone.0135731.ref053],[@pone.0135731.ref054]\] as well as the maximum DO concentration for denitrification (160 μg O~2~ L^-1^) \[[@pone.0135731.ref055]\]. Therefore, the direct effects of oxygen on bacteria may not be the cause of the compositional shift we detected. Alternatively, indirect effects through trophic interactions including altered bacterial grazing or viral infection may favor a different bacterial community, as has been shown under other climate change scenarios such as increased temperature and UV radiation \[[@pone.0135731.ref056]\]. Additionally, abiotic factors not measured in this study that may co-vary with DO may be causing a compositional shift. For example, the composition and quality of dissolved organic matter can strongly influence bacterial community composition \[[@pone.0135731.ref057]\]. Nonetheless, evidence for significant changes to the bacterial community at threshold DO concentrations above the conventional definition of hypoxia suggests that impacts of decreasing DO have implications for ecosystem processes and health before sublethal and lethal effects on macrofauna are detected.
We identified 21 indicator taxa of low DO that were present across the majority of samples spanning a full range of DO, but their relative abundances change significantly at the threshold DO concentration. These low-DO indicator taxa were identified as members of common and ubiquitous marine bacteria including *Roseobacter*, SAR11, SAR202, SAR406, SAR86, OM190, and *Nitrospina*, which are often detected in coastal regions, estuaries, upwelling zones, and other pelagic oxygen minimum zones ([S2 Table](#pone.0135731.s004){ref-type="supplementary-material"}). Other OTUs from these common marine groups were also detected throughout our dataset, but were not identified as indicators of low DO. This may suggest that particular low-DO adapted ecotypes of common marine bacteria exist, highlighting the taxonomic and functional redundancy of bacterial communities \[[@pone.0135731.ref058]\]. Ecotypes of SAR11 are known to be distributed across chemical and physical gradients in marine surface waters suggesting that niche separation has led to the diversification within this group \[[@pone.0135731.ref059],[@pone.0135731.ref060]\]. It is possible that similar mechanisms are responsible for shaping the distribution of indicator taxa we detected. Further research into the indicator taxa we detected here will elucidate the adaptations of these potential ecotypes that are responsible for providing a competitive advantage at low DO.
Our initial survey across a gradient of spatial and temporal conditions in Hood Canal has provided a snapshot into the complex relationship between bacterial community structure and DO. An obvious limitation of this study is that a few discrete samples do not capture fine-scale patterns or more extreme hypoxia conditions; yet with even limited sampling we were able to detect a significant shift in community composition at a threshold level of DO. Expanded sampling in both space and time will aid in refining this threshold concentration and the mechanisms for compositional shifts. The long residence time of water in Hood Canal \[[@pone.0135731.ref025]\] combined with the reduced grazing pressure on bacteria at depth or lower DO as a result of a shift in the distribution of higher trophic levels where DO is depleted \[[@pone.0135731.ref031]\] could also allow for the detection of more bacterial taxa, many of which may be in a dormant or inactive state. Dormancy can contribute to high taxonomic diversity and also act as a reservoir of functional diversity that can be resuscitated under favorable conditions \[[@pone.0135731.ref061]\].
As bacterial communities shift in diversity and composition, it is likely that the bacterially mediated nutrient and energy cycles also change, which has implications for changes in global cycles as oxygen minimum zones continue to expand. The hydrography of Hood Canal generates a wide gradient of abiotic factors including DO that change on an annual scale, which establishes a unique natural laboratory in which these questions can be explored. We are only beginning to understand how microbial communities are responding to decreasing DO, but further questions into the effects on individual functional groups of Bacteria will be crucial in predicting the global consequences of expanding oxygen minimum zones.
Supporting Information {#sec013}
======================
###### Rarefaction curves of (A) All 16S sequences, (B) Alphaproteobacterial sequences only, (C) Gammaproteobacterial sequences only, and (D) Flavobacterial sequences only.
Each solid line represents the sequence accumulation curve for each sample in the dataset. The dotted line represents a 1:1 line, which is the maximum taxa accumulation per sequence. The dashed, vertical line in each plot represents the number of sequences to which each set was subsampled. Note that plot A is on a different scale from plots B-D.
(TIF)
######
Click here for additional data file.
###### Depth profiles of physical and chemical data recorded during sampling.
\(A\) April sampling at Hama Hama and Sister's Point, (B) during June sampling at all stations, and (C) during October sampling at Hama Hama and Sister's Point.
(PDF)
######
Click here for additional data file.
###### Abiotic data collected to accompany each bacterial pyrosequencing sample and used in PCA and regression analyses.
(PDF)
######
Click here for additional data file.
###### Indicator taxa of high or low DO conditions in Hood Canal, WA, as identified by SIMPER analysis, and their associated top BLAST matches on NCBI.
"Contrib.depth" and "Contrib.DO" describe the relative contribution of a particular taxon to the dissimilarity between sample groups defined by water depth or DO content, respectively. "Hi" or "Lo DO ave abund" describes the average abundance of a particular taxon in the samples grouped as either high or low DO, respectively.
(XLSX)
######
Click here for additional data file.
The authors wish to thank Allan Devol, Jan Newton, and Wendi Ruef for supplying oxygen and chlorophyll *a* data from ORCA buoys in Hood Canal, which are operated by the University of Washington and maintained as part of the Northwest Association of Networked Ocean Observing Systems (NANOOS), part of the U.S. Integrated Ocean Observing System (IOOS). We thank Michael Dyen for assistance in sample processing and data collection and KathiJo Jankowski for helpful and insightful comments on the manuscript.
[^1]: **Competing Interests:**The authors have declared that no competing interests exist.
[^2]: Conceived and designed the experiments: CW GR MCHD. Performed the experiments: CW GR MCHD. Analyzed the data: RLS GR MCHD. Contributed reagents/materials/analysis tools: RLS CW GR MCHD. Wrote the paper: RLS GR MCHD.
[^3]: Current address: School of Oceanography, University of Washington, Seattle, Washington, United States of America
| {
"pile_set_name": "PubMed Central"
} |
1. Background {#sec1}
=============
A woman\'s probability of death due to pregnancy is unacceptably high in low and middle‐income countries than in high‐income countries \[[@B1]\]. It is estimated that nearly 30 million women in Africa become pregnant each year and about 250,000 of them die from pregnancy and childbirth‐related causes \[[@B2]\] even though most pregnancy and childbirth complications can be prevented or treated by skilled care during pregnancy, childbirth, and the immediate postnatal period \[[@B1]\]. Furthermore, antenatal care (ANC) reduces maternal and perinatal morbidity and mortality both directly, through early detection and treatment of pregnancy‐related complications, and indirectly, through the identification of women and girls at increased risk of developing complications during labour and delivery \[[@B3]\]. For instance, antenatal care decreases the likelihood of maternal anaemia during pregnancy and the delivery of a premature or low birth weight baby \[[@B3], [@B4]\]. Notwithstanding the crucial role of antenatal care in maternal and new‐born health, only 40% of all pregnant women in low‐income countries had the recommended antenatal care visits in 2015 \[[@B1]\].
Globally, several studies and reports have highlighted the positive influence male involvement has on the successful implementation of maternal and child health programmes and interventions. For instance, available evidence suggests that in patriarchal societies such as Ghana, access to and survival of maternal and child healthcare services requires the active involvement of men \[[@B5], [@B6]\]. Men in patriarchal settings have tremendous control over their spouses and a woman must receive permission and money from her husband to seek healthcare \[[@B7]--[@B9]\]. As decision‐makers for families, men in some parts of Northern Ghana consult with soothsayers to decide treatment for pregnant women and serve as the final authority on where and when pregnant women should seek medical care \[[@B10]\]. This is regardless of the fact that for some obstetric complications such as haemorrhage, the window of opportunity to respond and save the life of the mother may be measured in hours \[[@B2]\].
Despite the crucial role of men in maternal and child health, it is extremely unlikely to find a husband at an antenatal clinic or in a delivery room in many low and middle‐income countries. In a study conducted in Johannesburg, only 14% of women reported that a partner had attended ANC with them during their current pregnancy \[[@B11]\]. Similarly, a study among married men in Nepal reported 39.3% of male involvement in ANC \[[@B8]\]. Men understand antenatal care as an affair for women and thus inappropriate for them. In some communities, men who rise above these strict gender roles to support their wives during pregnancy and accompany them to the antenatal clinic are ridiculed and stigmatised by other members of their community \[[@B10]\].
Active male participation in antenatal care and childbirth can play an important role in addressing the first and second delays in seeking care: thus delay in recognising problems and deciding to seek care and delay in reaching care \[[@B8], [@B12]\]. The presence of a male partner in a delivery room will provide emotional support for the mother, decrease pain perception, establish an early relationship between a father and the infant and perhaps encourage the practice of family planning \[[@B13]\]. Notwithstanding the benefits of involving men in maternal and child healthcare, it is unclear how women view such new developments in Sub‐Saharan Africa \[[@B6]\]. In addition, many studies in the area focus on self‐reported behaviours of men on the barriers and determinants of male involvement without seeking the perspectives of women on the subject \[[@B8], [@B14], [@B15]\]. Interventions to promote men\'s involvement in maternal and child health care are less likely to succeed if the views and concerns of women are not considered \[[@B6]\]. This study provides an understanding of women\'s perspective on men\'s involvement in antenatal care, labour, and childbirth in the Northern Region of Ghana.
2. Materials and Methods {#sec2}
========================
This cross‐sectional study was conducted at the antenatal clinic of Tamale Teaching Hospital (TTH), a tertiary health facility in the Northern Region of Ghana. The hospital serves as a referral centre for all the primary and secondary health facilities in the three regions of the North and the northern part of Brong‐Ahafo Region. It provides advanced clinical health services, collaborates with the University for Development Studies, Tamale, for the training of undergraduate and postgraduate students, and undertakes research that influences treatment and policy in the health sector \[[@B16]\]. The institutional review committee of the Tamale Teaching Hospital reviewed the study protocol and granted approval for the study. The purpose of the study and the rights of participants during the study were clearly explained to each participant and written consent obtained for their participation. The participants were informed they could refuse to answer any questionnaire they are not comfortable with or withdraw from the study at any point without any condition or threat. Only the authors had access to the information collected from participants and the confidentiality of the information was ensured in accordance with the data protection act.
2.1. Study Population, Sample Size, and Sampling Method {#sec2.1}
-------------------------------------------------------
Three hundred pregnant women were recruited for the study. In line with the inclusion criteria of the study, only pregnant women who were attending ANC at the Tamale Teaching Hospital were enrolled into the study. Pregnant women who were unwilling to give consent and those who did not attend ANC at TTH were excluded from the study. Cochran\'s formula was used to estimate the sample size for the study using 14% prevalence of male partner attendance at the antenatal clinic \[[@B11]\], an assumption of 95% confidence interval, and 5% degree of error. A minimum sample size of 185 was estimated. However, the final sample size was increased to 300 to boost the power of the study. Systematic random sampling method was used to recruit pregnant women for the study. In this type of probability sampling method, study participants are chosen at regular intervals (estimated by dividing the projected population size by the sample size) from a sample frame after randomly selecting the first participant \[[@B17]\]. A review of the ANC attendance register for the month prior to the period of data collection showed that an average of 80 pregnant women visits the clinic for ANC services daily. Data collection for the study was scheduled to take place within a month during which nearly 1800 pregnant women were expected to visit the antenatal clinic. The estimated monthly attendance was divided by the sample size (300) to give a sampling interval of 6. The first participant for each day of data collection was randomly selected. After that, every sixth (6^th^) eligible pregnant woman was selected to participate in the study until the sample size was obtained.
2.2. Data Collection Procedure {#sec2.2}
------------------------------
A structured questionnaire was used to collect data from the participants on socio‐demographic characteristics such as age, marital status, religion, ethnicity, educational level, and occupation; male involvement in antenatal care; and male partner companionship during labour and childbirth. For the purposes of this study, "male involvement" was defined as the attendance and participation of men in antenatal care services during the visits of their spouse and their presence and support during childbirth. The items for the questionnaire were designed after a thorough literature review of similar studies in peer‐reviewed journals \[[@B11], [@B13], [@B18]\]. A senior midwife and a public health specialist reviewed the questionnaire and deemed the items appropriate and content valid. Three final year nursing students administered the questionnaire to participants. They were trained on how to obtain consent and administer the study questionnaire to the participants. Women visiting the antenatal clinic were informed about the study during health talk at the clinic and assessed for eligibility afterwards. Voluntary consent for participation was sought from eligible participants and the study questionnaire was only administered to women who agreed to participate in the study after the health talk. The research assistants and the principal investigator explained to the participants how to complete the questionnaire and addressed their concerns. Participants who had no formal education and could not read or write were assisted by the research assistants to complete the questionnaire, this was reported in less than 20% of the participants. Prior to data collection, a native speaker proficient in both English language and the main local language (Dagbani) translated the questionnaire into Dagbani. A second native speaker reviewed the translated instrument to ensure clarity and to eliminate ambiguities. The data collectors, who were native speakers of the main local language, then used the translated questionnaire for all participants with no formal education. The principal investigator supervised the data collection exercise and reviewed all completed questionnaires at the end of each day.
2.3. Data Analysis {#sec2.3}
------------------
Data were cross‐checked for completeness, coded in Microsoft Excel spreadsheet, and analysed using STATA Version 14.0 (College Station, Texas 77845, USA). Socio‐demographic characteristics and women\'s perspectives on men\'s involvement in antenatal care, labour, and childbirth were described in tables using frequencies and percentages. Univariate and multivariable logistic regression models were used to determine socio‐demographic factors associated with women\'s perspectives on men\'s involvement in antenatal care, labour, and childbirth estimating Odds ratio with 95% confidence intervals and *p* values. The univariate logistic regression was applied in the initial analysis and factors with *p*‐value \< .05 were selected for inclusion in the multivariable logistic regression analysis to determine independent predictors of women\'s perspectives on men\'s involvement in antenatal care, labour, and childbirth. In both the Univariate and multivariable regression models, the significance level was set at \<.05.
3. Results {#sec3}
==========
[Table 1](#tab1){ref-type="table"} shows the socio‐demographic characteristics of the study sample. More than half (55.3%) of the 300 pregnant women were within the age group of 20 to 29 years with an overall mean age of 28 (SD = 5.21) years. Majority of the women were married (98.3%), Muslims (68.3%), and Mole‐Dagombas (61.0%). A little over two‐fifths (40.3%) of the participants attended or completed secondary education and 50.3% of them were employed in the formal sector. Most (31.7%) of the women were carrying their second pregnancy and only 16.3% of them had had four or more pregnancies. Vaginal delivery (75.2%) was the major mode of delivery among the pregnant women who had given birth in the past.
3.1. Women\'s Perspective on Male Involvement in Antenatal Care {#sec3.1}
---------------------------------------------------------------
The perspectives of the pregnant women on men\'s involvement in antenatal care are presented in [Table 2](#tab2){ref-type="table"}. More than four‐fifths (86.0%) of the participants indicated their desire for male involvement in antenatal care services. Most of them said men would learn about pregnancy and childbirth (33.9%); and how to support women during pregnancy (32.5%) when they participate in antenatal care. Among the women who did not support male participation in antenatal care, 45.2% believed men do not have time to attend ANC while 22.6% believed pregnancy is women\'s affair. More than half (56.3%) of the study sample indicated their partners ever accompanied them to the hospital for antenatal care. However, only 33.7% of them said their partners participated in antenatal care services. Among the women that said their partners had never accompanied them to the antenatal clinic, 87.0% indicated they had never asked their partners to accompany them to the clinic and 43.9% said they had never asked their partners because they believed they do not have time to attend ANC services.
3.2. Association of Women\'s Socio‐Demographic Factors and Male Involvement in ANC {#sec3.2}
----------------------------------------------------------------------------------
[Table 3](#tab3){ref-type="table"} presents the association of women\'s socio‐demographic characteristics and men\'s involvement in ANC. The majority of women in the age range of 20--29 reported their husbands ever accompanied them to ANC (62.7%). However, only 38.5% of the male partners actually participated in at least one ANC service. In the case of pregnant women who were less than 20 years, none of their partners participated in any ANC service whereas for women aged 40--48 years only one participant indicated her partner ever participated in one of the ANC services. More than half of the women who were married (*n* = 167, 56.6%) and those who were Christians (*n* = 70, 73.7%) indicated their husbands accompanied them to ANC; however, most of these male partners did not participate in ANC services. The results in [Table 3](#tab3){ref-type="table"} clearly revealed that most of the male partners who accompanied their partners to the ANC did not participate in any of the ANC services. There was no significant difference in women\'s previous mode of delivery and their report of male attendance and participation in ANC.
3.3. Women\'s Perspective about Male Partner Companionship during Labour and Childbirth {#sec3.3}
---------------------------------------------------------------------------------------
[Table 4](#tab4){ref-type="table"} shows the perspectives of women about male partner presence during childbirth in the delivery room. More than four‐fifths (84.7%) of the women expressed their desire for male partner companionship during labour and childbirth. Of these women, 37.5% believed the presence of a male partner will provide emotional support, 21.5% said it will provide the woman with an opportunity to express her problems to a familiar person and 18.9% indicated men will treat women better if allowed in the delivery room during the delivery of their wives. Most of the forty‐six women who did not support the presence of male partners in the delivery room stated that delivery is a women\'s affair (38.02%) and that men do not have any role to play in the delivery room (28.10%).
3.4. Socio‐Demographic Determinants of Women\'s Perspective about Male Partner Involvement in ANC {#sec3.4}
-------------------------------------------------------------------------------------------------
[Table 5](#tab5){ref-type="table"} presents the odds ratio and 95% confidence intervals of the socio‐demographic determinants of women\'s perspective about male involvement in ANC. As shown in the unadjusted model, women aged 20--29 years (OR 2.34, 95%CI 0.25, 22.26) and 30--39 years (OR 1.06 95%CI 0.11, 9.93) were more likely to encourage male involvement in ANC when compared with women less than 20 years. However, those aged 40--48 years (OR 0.92, 95%CI 0.07, 11.58) were less likely to encourage male involvement in ANC. Married women were (OR 9.85, 95%CI 1.59, 60.81) more likely to encourage male involvement in ANC compared to women who were unmarried.
We observed that the probability of encouraging male involvement in ANC decreased with increased level of education of women: primary/JHS (OR 3.60, 95%CI 1.08, 12.02) secondary (OR 3.19, 95%CI 1.34, 7.61), and tertiary (OR 1.92, 95%CI 0.81, 4.56). A similar pattern was observed in the adjusted analysis: primary/JHS (OR 3.66, 95%CI 1.01, 13.27), secondary (OR 2.16, 95%CI 0.81, 5.75), and tertiary (OR 1.50, 95%CI 0.58, 3.89).
A significant relationship was observed between the number of pregnancies \[gravida 2, *p* = 0.021; gravida 3, *p* = 0.010; gravida 4 or more, *p* = 0.046\] and a woman\'s likelihood to encourage male involvement in ANC. After accounting for the effect of other significant covariates, there was good evidence to suggest that married women (*p* = 0.002), women with only primary/Junior High School education (*p* = 0.048) and those with two (*p* = 0.010), three (*p* = 0.008), or ≥4 (*p* = 0.044) previous pregnancies had a desire for male partner involvement in ANC.
3.5. Socio‐Demographic Determinants of Women\'s Perspective about Male Partner Companionship during Labour and Childbirth {#sec3.5}
-------------------------------------------------------------------------------------------------------------------------
[Table 6](#tab6){ref-type="table"} presents the odds ratios, confidence intervals, and *p* values of the socio‐demographic predictors of women\'s perspective about male partner companionship in labour and childbirth. There was no significant relationship between the age of the women, number of pregnancies, marital status of the women, and their perspective about male partner companionship in labour and childbirth. We observed that the odds of a woman encouraging the attendance of men at childbirth increased significantly with increased level of education in both the unadjusted and adjusted models: primary/JHS (OR 2.25, 95%CI 0.89, 5.71); AOR 1.94, 95%CI 0.68, 5.57), secondary (OR 5.11, 95%CI 2.23, 11.72; AOR 4.91, 95%CI 1.69, 14.27), and tertiary (OR 6.19, 95%CI 2.36, 16.25; AOR 9.75, 95%CI 2.51, 37.91). The results revealed that women employed in the formal sector (OR 1.12, 95%CI 0.38, 3.33) were more likely to support the presence of a male companion at delivery. As shown in [Table 6](#tab6){ref-type="table"}, women who were carrying their second pregnancy had 1.76 (95%CI 0.72, 4.29) increased odds of encouraging the presence of a male partner at birth compared to women who were pregnant for the firsts time. However, women who were pregnant for the third time (OR 0.63, 95%CI 0.28, 1.41) and those who were carrying their fourth or more (OR 0.95, 95%CI 0.37, 2.44) pregnancy were less likely to encourage the presence of a male partner at birth. Previous caesarean delivery was significantly associated with a 3.67 (95%CI 1.07, 12.60) increased odds of a woman encouraging male partner presence at delivery. This association was no longer apparent after adjustment with other covariates.
4. Discussion {#sec4}
=============
The women in this study expressed favourable support for male partner involvement in antenatal care services and as companions during labour and childbirth, which is in line with earlier studies in several low‐income countries \[[@B11], [@B13], [@B18], [@B20], [@B21]\]. This was probably because most of the women we surveyed believed that men who attend ANC with their partners acquire useful knowledge on how to support their wife\'s during pregnancy and childbirth. In line with the expectation of the women, a meta‐analysis of 14 studies on the impact of male involvement on maternal health outcome in low and middle‐income countries found that male involvement was significantly associated with reduced odds of postpartum depression and improved utilisation of skilled birth attendance and postnatal care \[[@B22]\]. Likewise, in Ethiopia, Mohammed et al., found that the probability of attending at least one ANC visit, first ANC visit within first trimester, and utilising skilled facility‐based delivery were higher in women with greater male partner involvement \[[@B23]\]. However, contrary to the current findings, an earlier study in Ghana reported that many Ghanaian women do not want their male partners involved beyond the provision of money and transport for maternal and child healthcare services \[[@B6]\]. In this study, women who did not support male participation in antenatal care said men do not have time to attend ANC and that pregnancy is a woman\'s affair. Both men and women alike in several studies have expressed similar views across Africa \[[@B6], [@B15], [@B24], [@B25]\].
Despite the desire of the current study participants for their partners to attend ANC, only a little over half of them reported that their partners ever accompanied them to ANC, which is consistent with similar studies in Uganda \[[@B14],[@B15]\], and Kenya \[[@B5]\]. Moreover, approximately two‐thirds of the male partners who accompanied their spouses did not participate in any ANC services. This lack of enthusiasm by most men to accompany their partners and participate in ANC services could be cultural because ANC attendance is considered a woman\'s affair as expressed in several studies \[[@B6], [@B24], [@B25]\]. Men may feel discomforted attending a female dominated program and discussing sexual and reproductive health issues with third parties. Fear of societal stigma and unfriendly clinic environment have also been reported as obstacles to male participation in maternal and child health services \[[@B10], [@B26]\]. In Ghana, Aborigo et al. found that in the Kassena‐Nankana Districts in the Upper East Region, men who accompany their wives to the antenatal clinic are called "kana‐kadona" or "bakana" which means "women\'s rivals" or "man‐woman" respectively, suggesting that such men exhibit female tendencies \[[@B10]\]. Men are conscious of these cultural barriers in the community and this may explain why most of the women reported that their husbands have never participated in any ANC service and the lack of interest on their part to ask their partners to accompany them for ANC services.
We found that younger women were more likely to encourage male partner involvement in ANC compared to older women. The reason could be that the younger women were likely to experience anxiety and fear during pregnancy than older women who might have experienced pregnancy several times. There was a significant relationship between a woman\'s marital status and the likelihood to encourage male involvement in ANC. Married women were 9.8 times more likely to support male partner involvement compared to unmarried women. This is probably because out‐of‐wedlock pregnancies are frowned upon and considered a shameful act in many African societies including Ghana, and this may discourage unmarried women from visiting the clinic with the father of their unborn child. Interestingly, the probability of encouraging male involvement among the women in this study decreased with an increased level in education, which agrees well with the finding of a study conducted in Nigeria \[[@B25]\]. However, further evidence, through large studies, is needed to adequately understand why educated women were not in support of male involvement in ANC.
Regarding the presence of men during childbirth, the majority of the women encouraged it because they believed the presence of a companion would provide emotional support, give them an opportunity to express their problems to a familiar person, and inspire men to treat their wives better after observing the delivery process. Several studies in low‐income countries corroborate these findings \[[@B13], [@B21]\]. However, the current arrangements of most labour wards in Ghana do not make provision for male partners to be with their wives during childbirth. In most clinical settings, pregnant women are admitted to a labour ward without much privacy to prevent male partners from witnessing the delivery of other women. An earlier study in Ghana reported that men are usually sent away from the delivery room during labour and are only called back after delivery and newborn care was completed \[[@B27]\]. Women who did not support male involvement in childbirth indicated that childbirth was a woman\'s responsibility and that men do not have a role to play in the delivery room.
We found that the odds of a woman encouraging the attendance of male partner at childbirth increased significantly with increased level of education. This result is consistent with those of other studies \[[@B19], [@B25], [@B28]\] and suggests that maternal years of formal education have an influence on healthcare utilisation. In line with the findings of Morhason‐Bello et al. \[[@B28]\], the current study revealed that women who were employed in the formal sector were more likely to ask for male companionship during delivery.
5. Limitations {#sec5}
==============
Our study has several limitations that must be considered in the application of our findings. Firstly, the study was conducted in an urban city and the findings may not be applicable to women in rural communities. Secondly, we interviewed only women who visited the antenatal clinic of a single hospital and this may limit the generalizability of the findings to pregnant women who visited other facilities and women who did not utilise antenatal care services during pregnancy. Thirdly, the study employed a cross‐sectional study design, making it difficult to establish a causal relationship between the outcome variables and the explanatory variables.
6. Conclusion {#sec6}
=============
Male involvement in antenatal care and childbirth received overwhelming support from the women in this study. Their perspectives were that male involvement would expose men to information about pregnancy and childbirth to help support their wives during pregnancy. Furthermore, the women indicated that the presence of a companion during childbirth would provide emotional support, give them an opportunity to express their problems to a familiar person, and inspire men to treat their wives better. Women\'s marital status, educational level, and number of pregnancies were identified as significant determinants of a woman\'s perception of male participation in ANC. Likewise, the educational level of women was found to be a significant predictor of a woman\'s perception of male involvement in childbirth. Interventions to encourage male involvement in pregnancy, labour and childbirth should consider the provision of male‐friendly environment at antenatal clinic and facilities to accommodate male companions in labour rooms as desired by the women in this study.
The authors would like to thank the management and staff of Tamale Teaching Hospital and all the participants for their cooperation.
ANC:
: Antenatal care
TTH:
: Tamale teaching hospital.
Data Availability
=================
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Ethical Approval {#sec8}
================
The Department of Research and Development of the Tamale Teaching Hospital reviewed the study protocol and granted approval for the study. The purpose of the study and the rights of participants during the study were clearly explained to each participant and written consent obtained for their participation. The participants were informed they could refuse to answer any questionnaire they were not comfortable with or withdraw from the study at any point without any condition or threat. Only the authors had access to the information collected from participants and the confidentiality of the information was ensured in accordance with the data protection act.
Conflicts of Interest
=====================
The authors declare that they have no conflicts of interest.
Authors\' Contributions {#sec10}
=======================
SM designed the study, wrote the protocol and performed the statistical analysis. IB and IA participated in data collection, managed the literature search, and discussion of the findings. All authors read and approved the final manuscript.
######
Background characteristics of participants.
Characteristic Number Percent
------------------------------------------ ----------- ---------
*Age (in years)*
\<20 5 1.67
20--29 166 55.33
30--39 115 38.33
40--48 14 4.67
Mean (SD) 28 (5.21)
*Marital status*
Unmarried^a^ 5 1.67
Married 295 98.33
*Religion*
Christianity 95 31.67
Islam 205 68.33
*Ethnicity*
Mole‐Dagombas 183 61.00
Akan 19 6.33
Gonja 21 7.00
Frafra 18 6.00
Chokosi 11 3.67
Dagati 14 4.67
Others 34 11.33
*Education level*
No formal education 50 16.67
Primary/JHS 45 15.00
Secondary 121 40.33
Tertiary 84 28.00
*Employment status*
Unemployed 93 31.00
Formal employment^b^ 151 50.33
Informal employment^c^ 56 18.67
*Number of pregnancies*
One 90 30.00
Two 95 31.67
Three 66 22.00
Four or more 49 16.33
*Previous mode of delivery ( n* = 210*)*
Vaginal birth 158 75.24
Caesarean delivery 52 24.76
^a^Unmarried: single, divorced, widowed, cohabiting. ^b^Formal employment: public and private sector employees. ^c^Informal employment: farmer, artisan, trader.
######
Women\'s perspective on male involvement in antenatal care.
Characteristic Number Percent
------------------------------------------------------------------------- -------- ---------
*Should men participate in ANC services*
Yes 258 86.00
No 42 14.00
*Why should men participate in ANC services (n = 258)^a^*
In case of infection, we can both be treated 86 14.1
He will learn about pregnancy and childbirth 207 33.93
He will learn how to support a pregnant woman 198 32.46
It is a demonstration of his love 48 7.87
It will make women feel supported 71 11.64
*Why should men stay away from ANC (n = 42)^a^*
Many men do not have time 21 45.65
Men need to concentrate on their work 5 10.87
ANC is not a place for a man 5 10.87
Pregnancy is women\'s affair 15 32.61
*Has your partner ever accompanied you to ANC*
Yes 169 56.33
No 131 43.67
*Did he participate in any ANC service (n = 169)*
Yes 57 33.73
No 112 66.27
*Did you ever asked your partner to accompany you to the ANC (n = 131)*
Yes 17 12.98
No 114 87.02
*Why have you never asked him to accompany you to the ANC (n = 114)^a^*
I know he will not accept to accompany me 40 16.88
He does not have time to attend ANC with me 104 43.88
He is not always around 57 24.05
He has nothing to do at the clinic 3 1.27
His presence will make me uncomfortable 4 1.69
ANC is for only women 11 4.64
His other wives will be unhappy 18 7.59
^a^Multiple responses were allowed.
######
Association of women\'s socio‐demographic factors and male partner involvement in ANC.
Characteristics Has your partner ever accompanied you to ANC *p* value Did he participate in any ANC service (*n* = 169) *p* value
----------------------------- ---------------------------------------------- ------------ --------------------------------------------------- ------------ ------------ ----------
*Age (in years)*
\<20 3(60.00) 2(40.00) 0.009^a^ 0(0.00) 3(100.00) 0.275^a^
20--29 104(62.65) 62(37.35) 40(38.46) 64(61.54)
30--39 59(51.30) 56(48.70) 16(27.12) 43(72.88)
40--48 3(21.43) 11(78.57) 1(33.33) 2(66.67)
*Marital status*
Unmarried 2(40.00) 3(60.00) 0.656^a^ 0(0.00) 2(100.00) 0.550^a^
Married 167(56.61) 128(43.39) 57(34.13) 110(65.87)
*Religion*
Christianity 70(73.68) 25(26.32) \<0.001 27(38.57) 43(61.43) 0.263
Islam 99(48.29) 106(51.71) 30(30.30) 69(69.70)
*Education level*
No formal education 15(30.00) 35(70.00) \<0.001 0(0.00) 15(100.00) 0.005^a^
Primary/JHS 10(22.22) 35(77.78) 5(50.00) 5(50.00)
Secondary 81(66.94) 40(33.06) 26(32.10) 55(67.90)
Tertiary 63(75.00) 21(25.00) 26(41.27) 37(58.73)
*Employment status*
Unemployed 56(37.09) 95(62.91) \<0.001 30(44.12) 38(55.88) 0.044
Formal employment 68(73.12) 25(26.88) 17(30.36) 39(69.64)
Informal employment 45(80.36) 11(19.64) 10(22.22) 35(77.78)
*Number of pregnancies*
One 54(60.00) 36(40.00) \<0.001 21(38.89) 33(61.11) 0.510^a^
Two 66(69.47) 29(30.53) 23(34.85) 43(65.15)
Three 33(50.00) 33(50.00) 10(30.30) 23(69.70)
Four or more 16(32.65) 33(67.35) 3(18.75) 13(81.25)
*Previous mode of delivery*
Vaginal birth 83(52.53) 75(47.47) 0.258 26(31.33) 57(68.67) 0.994
Caesarean delivery 32(61.54) 20(38.46) 10(31.25) 22(68.75)
^a^Fisher\'s exact test.
######
Women\'s perspective on male partner companionship during labour and childbirth.
Characteristic Number Percent
-------------------------------------------------------------------------------- -------- ---------
*Should a man be allowed in the delivery room during the delivery of his wife*
Yes 254 84.67
No 46 15.33
*Why should men be allowed in the delivery (*n* = 42)^a^*
To treat women better afterwards 81 18.88
To provide emotional support 161 37.53
It may encourage men to support and practice family planning 33 7.68
Helps the mother bear labour pain 32 7.46
An opportunity for the woman to express her problem to a familiar person 92 21.45
Strengthens the couple\'s relationship 30 6.99
*Why should men stay away from the delivery room (n = 46)^a^*
Delivery is women\'s affair 46 38.02
Men do not have role to play in the delivery room 34 28.1
Men may not love their wives the same way after observing the delivery 18 14.88
He may disturb the health professionals 23 19.01
^a^Multiple responses were allowed.
######
Socio‐demographic determinants of women\'s perspective about male partner involvement in ANC.
Characteristics Unadjusted model Adjusted model
----------------------------- ------------------- ---------------- --------------------- -------
*Age (in years)*
\<20 1.00
20--29 2.34(0.25, 22.26) 0.458
30--39 1.06(0.11, 9.93) 0.961
40--48 0.92(0.07, 11.58) 0.946
*Marital status*
Unmarried 1.00 1.00
Married 9.85(1.59, 60.81) 0.014 37.58(3.74, 377.75) 0.002
*Religion*
Christianity 1.00
Islam 1.31(0.66, 2.60) 0.446
*Education level*
No formal education 1.00 1.00
Primary/JHS 3.60(1.08, 12.02) 0.037 3.66(1.01, 13.27) 0.048
Secondary 3.19(1.34, 7.61) 0.009 2.16(0.81, 5.75) 0.123
Tertiary 1.92(0.81, 4.56) 0.140 1.50(0.58, 3.89) 0.400
*Employment status*
Informal employment 1.00
Formal employment 0.74(0.26, 2.07) 0.564
Unemployed 0.67(0.26, 1.74) 0.408
*Number of pregnancies*
One 1.00 1.00
Two 0.29(0.10, 0.83) 0.021 0.19(0.52, 0.67) 0.010
Three 0.24(0.81, 0.71) 0.010 0.17(0.04, 0.63) 0.008
Four or more 0.30(0.09, 0.98) 0.046 0.22(0.05, 0.96) 0.044
*Previous mode of delivery*
Vaginal birth 1.00
Caesarean delivery 1.87(0.73, 4.78) 0.190
######
Socio‐demographic determinants of women\'s perspective about male partner companionship during labour and childbirth.
Characteristics Unadjusted model Adjusted model
----------------------------- -------------------- ---------------- ------------------- -------
*Age (in years)*
\<20 1.00
20--29 1.93(0.21, 18.22) 0.564
30--39 1.00(0.11, 9.38) 1.00
40--48 0.92(0.073, 11.58) 0.946
*Marital status*
Unmarried 1.00
Married 3.80(0.62, 23.42) 0.150
*Religion*
Christianity 1.00
Islam 0.42(0.19, 0.95) 0.036
*Education Level*
No formal Education 1.00 1.00
Primary/JHS 2.25(0.89, 5.71) 0.088 1.94(0.68, 5.57) 0.217
Secondary 5.11(2.23, 11.72) \<0.001 4.91(1.69, 14.27) 0.004
Tertiary 6.19(2.36, 16.25) \<0.001 9.75(2.51, 37.91) 0.001
*Employment status*
Informal employment 1.00
Formal employment 1.12(0.38, 3.33) 0.839
Unemployed 0.46(0.18, 1.18) 0.108
*Number of pregnancies*
One 1.00
Two 1.76(0.72, 4.29) 0.214
Three 0.63(0.28, 1.41) 0.258
Four or more 0.95(0.37, 2.44) 0.905
*Previous mode of delivery*
Vaginal birth 1.00
Caesarean delivery 3.67(1.07, 12.60) 0.039
[^1]: Academic Editor: Luca Marozio
| {
"pile_set_name": "PubMed Central"
} |
1.. Introduction
================
Polyaniline (PANI) has been used in the fabrication of various types of enzyme-based biosensors because of its porous structure, as well as its adequate conductivity and thermal stability \[[@b1-sensors-09-04635]-[@b6-sensors-09-04635]\]. To stabilize the immobilized enzyme in the matrix of PANI film, glutaraldehyde (GA) is usually employed as a bifunctional agent to crosslink enzyme molecules \[[@b7-sensors-09-04635]-[@b10-sensors-09-04635]\], but the crosslinking efficiency under standard conditions is not always satisfactory \[[@b10-sensors-09-04635],[@b11-sensors-09-04635]\], which results in the lower sensitivity and poor stability of the resulting biosensor. Previously, we have electrochemically synthesized the PANI film on a Pt electrode in the presence of bovine serum albumin (BSA), a lysine-rich enzyme, which provides extra free *ε*-amino groups for the further crosslinking of HRP with glutaraldehyde and has significantly improved the effectiveness of a PANI modified biosensor \[[@b1-sensors-09-04635]\]. Nevertheless, to enhance the efficiency and stability of enzyme immobilization on an electrode is still the major issue of fabricating enzyme-based biosensors.
Over the past few years, immobilizations of enzymes in well-defined mesoporous silica materials have been proven to be promising for enhancing the thermal stabilities and maintaining the catalytic activities of enzymes \[[@b12-sensors-09-04635]-[@b14-sensors-09-04635]\]. Enzymes entrapped inside the silica mesopores are less susceptible to pH and temperature alternations and organic solvents as well \[[@b12-sensors-09-04635],[@b15-sensors-09-04635],[@b16-sensors-09-04635]\]. Among them, SBA-15, which is synthesized in the presence of nonionic triblock copolymer P123 as a template under acidic conditions, exhibits well-ordered hexagonal pore arrays of uniform pore size \[[@b17-sensors-09-04635],[@b18-sensors-09-04635]\]. Meanwhile, SBA-15 possesses a large surface area and internal silanol hydroxyls that have affinities for physical adsorption of enzyme molecules \[[@b19-sensors-09-04635]-[@b23-sensors-09-04635]\]. Recently, SBA-15 has been successfully employed to entrap glucose oxidase (GOD) to construct a glucose biosensor, achieving with enhanced sensitivity, long-term stability and reproducibility \[[@b24-sensors-09-04635],[@b25-sensors-09-04635]\]. In addition, monoclonal antibodies were immobilized in SBA-15 for the detection of antigen (cTnI) in the serum of patients, which is more convenient and superior to the conventional enzyme-linked immunoadsorbent assay (ELISA) \[[@b26-sensors-09-04635]\]. For the detection of hydrogen peroxide (H~2~O~2~), SBA-15 loaded with hemoglobin (Hb) has shown a fast amperometric response, a low detection limit, and good stability \[[@b27-sensors-09-04635]\].
In this study, we further exploited the application of mesoporous SBA-15 in entrapping HRP and constructed an amperometric GA/SBA-15(HRP)/PANI/Pt biosensor by immobilizing the SBA-15(HRP) on the electrochemically synthesized PANI film on a Pt electrode. The composite biosensor was then characterized and evaluated for the detection of H~2~O~2~ with cyclic voltammetry. In addition, its linear correlation, sensitivity and stability were investigated.
2.. Results and Discussion
==========================
2.1.. Characterizations of SBA-15 Mesoporous Silica
---------------------------------------------------
The X-ray diffraction (XRD) pattern of calcined SBA-15 \[line (a) in [Figure 1](#f1-sensors-09-04635){ref-type="fig"}\] revealed well-resolved peaks of 2θ at 0.812, 1.392 and 1.588, which represented the characteristic (100), (110), and (200) reflections of hexagonal mesoporous materials with *p6mm* symmetry \[[@b17-sensors-09-04635]\]. The total specific surface area, total pore volume, and BJH pore diameter of the SBA-15 were estimated by N~2~ adsorption-desorption isotherm ([Figure 2](#f2-sensors-09-04635){ref-type="fig"}) and Barrett-Joyner-Halenda (BJH) calculation ([Table 1](#t1-sensors-09-04635){ref-type="table"}).
[Figure 3](#f3-sensors-09-04635){ref-type="fig"} shows the TEM image of calcined SBA-15 with well-ordered hexagonal array mesopores. The pore diameter was approximately 100 Å, which was close to the center of the pore size distribution (*ca.* 92 Å) shown in the inset of [Figure 2](#f2-sensors-09-04635){ref-type="fig"}. The pore size distribution indicated the major pore diameter of mesopores ranged between 80 and 110 Å, which permitted the easier access of HRP molecules because the dimension of the native HRP (MW: ∼ 44 kDa) in a neutral buffer solution was predicted to be 62 × 43 × 12 (Å)^3^ by a scanning tunneling microscopy (STE) study \[[@b28-sensors-09-04635]\]. Meanwhile, the average pore diameter of SBA-15 estimated by the BJH method was *ca.* 76 Å ([Table 1](#t1-sensors-09-04635){ref-type="table"}), which was below the major pore size distribution (80∼110 Å), suggesting the presence of micro-channels in the interior of SBA-15.
2.2.. Immobilization of HRP in SBA-15 Mesopores
-----------------------------------------------
When compared with SBA-15, the total specific surface area and the total pore volume of SBA-15(HRP), decreased modestly by about 11% and 8%, respectively, indicating the successful entrapment of HRP within the pores of SBA-15 ([Figure 2](#f2-sensors-09-04635){ref-type="fig"} and [Table 1](#t1-sensors-09-04635){ref-type="table"}). The loading of HRP in SBA-15 was also confirmed with ABTS enzymatic assay, which nearly 395 units of HRP was stably retained by one gram of SBA-15 with the procedure described in the Experimental section. On the other hand, the XRD pattern of SBA-15(HRP) \[line (b) in [Figure 1](#f1-sensors-09-04635){ref-type="fig"}\] matched with that of unloaded SBA-15 \[line (a) in [Figure 1](#f1-sensors-09-04635){ref-type="fig"}\] although the intensity was decreased, and exhibited the similar mesoporous parameters listed in [Table 1](#t1-sensors-09-04635){ref-type="table"}, suggesting SBA-15(HRP) had an analogous mesoporous structure to that of SBA-15. Furthermore, the similar pore distribution of SBA-15 and SBA-15(HRP), shown in the inset of [Figure 2](#f2-sensors-09-04635){ref-type="fig"}, implied that the retained HRP did not block the entrances of mesopores, but rather resided in the inner space of mesopores. This conclusion was further supported by the identical BJH pore diameters before and after loading of HRP ([Table 1](#t1-sensors-09-04635){ref-type="table"}).
Furthermore, we utilized the Coomasie Brilliant Blue R-250, which is commonly employed to stain proteins in SDS-PAGE gel analysis, to stain SBA-15 and SBA-15(HRP), respectively. Both were then washed with de-staining solution several times. Our result showed that the blue dye was retained by SBA-15(HRP), but not by SBA-15, suggesting HRP was stably immobilized in SBA-15(HRP). In order to investigate whether the immobilized HRP in SBA-15(HRP) was entrapped inside the pore of SBA-15 or was adsorbed on the surface of SBA-15, we examined another mesoporous silica, MCM-41, which was synthesized by a similar approach as that used for SBA-15, but its pore diameter was estimated as 25 Å, therefore, HRP was supposed to be completely excluded by MCM-41. We found that the adsorption of HRP on the surface of silica MCM-41 was relatively unstable and most of the blue stain was washed away. Therefore, the stably immobilized HRP in SBA-15(HRP) was most likely entrapped inside the pores of SBA-15.
2.3.. The Surface Morphology of Electrodes
------------------------------------------
After depositing the SBA-15 on the electrochemically synthesized PANI/Pt electrode, SEM was employed to illustrate the surface morphology of the constructed electrodes. As shown in [Figure 4a](#f4-sensors-09-04635){ref-type="fig"}, the SBA-15 formed aggregations with the size of several micrometers, in which stably attached to the surface or filled in the inner matrix of fibrous PANI film. [Figure 4(c)](#f4-sensors-09-04635){ref-type="fig"} presents the surface morphology of the constructed GA/SBA-15(HRP)/PANI/Pt electrode, where SBA-15 particles appeared to be covered by the thin layer formed by glutaraldehyde. If glutaraldehyde was not applied (data not shown), the electrode exhibited the similar surface morphology as that of SBA-15/PANI/Pt electrode in [Figure 4(b)](#f4-sensors-09-04635){ref-type="fig"}, however, the stability was relatively poorer than that of GA/SBA-15(HRP)/PANI/Pt electrode and the comparison was further discussed later.
2.4.. Electrochemical Response of Electrodes towards H2O2
---------------------------------------------------------
[Figure 5](#f5-sensors-09-04635){ref-type="fig"} shows the cyclic voltammograms of electrodes in response to 18.5 mM of H~2~O~2~ in 0.1 M phosphate buffer (pH 6.2). The PANI modified Pt electrode displayed considerable cathodic response to H~2~O~2~, as represented by line (a) in [Figure 5](#f5-sensors-09-04635){ref-type="fig"}, and was similar to that reported in our previous publication \[[@b1-sensors-09-04635]\]. The deposition of unloaded SBA-15 on the PANI film decreased the peak current \[line (b) in [Figure 5](#f5-sensors-09-04635){ref-type="fig"}\], implicating the reduction of conductivity due to the electronically insulating SBA-15 particles on the PANI film. However, the immobilization of SBA-15(HRP) on the PANI film dramatically enhanced the cathodic response as expected \[line (c) in [Figure 5](#f5-sensors-09-04635){ref-type="fig"}\]. According to the results shown in [Figures 1](#f1-sensors-09-04635){ref-type="fig"} and [2](#f2-sensors-09-04635){ref-type="fig"} as well as [Table 1](#t1-sensors-09-04635){ref-type="table"}, the mesopores of SBA-15 provided the electro-active proteins with a unique environment that might prevent protein aggregation during immobilization. A recent publication has demonstrated that the SBA-15 mesoporous materials accelerated the electron transfer between the entrapped enzyme and electrode \[[@b29-sensors-09-04635]\]. The similar conclusions were also made for other mesoporous silica \[[@b30-sensors-09-04635]-[@b32-sensors-09-04635]\] and the electron hopping mechanism was proposed to facilitate the electron transfer inside silica mesopores \[[@b29-sensors-09-04635],[@b33-sensors-09-04635]-[@b35-sensors-09-04635]\].
Meanwhile, the novel constructed electrode exhibited a nice linear correlation with H~2~O~2~ in the range of 0.02 to 18.5 mM \[*R*^2^ = 0.997, line (c) in [Figure 6](#f6-sensors-09-04635){ref-type="fig"}\]. The sensitivity of GA/SBA-15(HRP)/PANI/Pt electrode towards H~2~O~2~ was obtained as 89.46 μA·mM^-1^·cm^-2^, which was better than those of 66.21 μA·mM^-1^·cm^-2^ for PANI/Pt and 53.71 μA·mM^-1^·cm^-2^ for SBA-15/PANI/Pt electrodes, as shown by lines (a) and (b) in [Figure 6](#f6-sensors-09-04635){ref-type="fig"}, respectively. Although the amount of HRP actually entrapped in the GA/SBA-15(HRP)/PANI/Pt electrode was much less than that employed for direct immobilization of HRP on the PANI film (GA/HRP/PANI/Pt electrode) as we previously reported \[[@b1-sensors-09-04635]\], their cathodic response and sensitivity were all comparable. The inset in [Figure 6](#f6-sensors-09-04635){ref-type="fig"}, also displayed their linear correlations with H~2~O~2~ in the range of 0.02 to 4 mM, and the limit detection of GA/SBA-15(HRP)/PANI/Pt sensor was about 10 μM.
2.5.. The Stability of Constructed Electrode
--------------------------------------------
It has been reported that the electrochemical synthesized PANI film on the Pt electrode may be unstable during the sequential cyclic voltammetric measurements, in which can be improved by the implanted bovine serum albumin \[[@b1-sensors-09-04635]\]. The same results are shown in [Figure 7](#f7-sensors-09-04635){ref-type="fig"}, where the cathodic current towards 1.96 mM H~2~O~2~ plunged down near 50% for PANI/Pt (line a) and 43% for GA/HRP/PANII/Pt (line b), respectively, with the second measurement. By the fifth measurement of a 16-day period, both electrodes lost near 50% of their initial responses. However, the cathodic current dropped about 30% for SBA-15(HRP)/PANI/Pt (line c) and 20% for GA/SBA-15(HRP)/PANI/Pt (line d) with the second measurement, but the decay observed for GA/SBA-15(HRP)/PANI/Pt during the subsequent measurements was insignificant, therefore we concluded that glutaraldehyde might be able to stabilize the entrapped HRP through crosslinking or forming the thin film, as indicated by [Figure 4c](#f4-sensors-09-04635){ref-type="fig"}. Meanwhile, the improvement of PANI film stability was probably ascribed to the filling of SBA-15 in the matrix of PANI network as shown in [Figures 4(b) and (c)](#f4-sensors-09-04635){ref-type="fig"}.
The entrapment of enzyme in the mesopores of SBA-15 has been proposed to occur by means of electrostatic interaction, simple adsorption, and entrapment \[[@b36-sensors-09-04635]\], therefore the interactions between enzyme and the inorganic inner pore surface of SBA-15 were not tight enough. In this study, the loaded SBA-15 was rinsed with phosphate buffer for several times until no measurable leaching of HRP was detectable, based on the Bradford assay as well as ABTS enzymatic assay. However, significant reductions of the cathodic responses for both SBA-15(HRP)/PANI/Pt and GA/SBA-15(HRP)/PANI/Pt were still observed after the initial cyclic voltammetric measurement, in which were partially resulted from the exclusion of enzyme molecules by the electrostatic repulsion. Nevertheless, the SBA-15 loaded with HRP provided the PANI/Pt electrode with not only an enhanced sensitivity but also an improved stability, in which could be further improved by employing glutaraldehyde.
3.. Experimental Section
========================
3.1.. Chemicals
---------------
Horseradish peroxidase (HRP), copolymer poly(ethylene glycol)-block-poly(propylene glycol)- block-poly(ethylene glycol) (EO~20~PO~70~EO~20~, Pluronic P123 with the molecular weight of 5,800), 1,3,5-Trimethylbenzene (TMB), and tetraethyl orthosilicate (TEOS, 98%) were commercially available from Sigma-Aldrich Corp. (St. Louis, MI, USA). Glutaraldehyde (GA, 25%, v/v), hydrogen peroxide (35%, v/v), aniline monomer were obtained from Merck KGaA (Darmstadt, Germany). All other reagents used for buffer and standard solution preparation were were of analytical grade and purchased from various commercial sources.
3.2.. Preparation of SBA-15 Mesoporous Silica
---------------------------------------------
SBA-15 was prepared according to the procedure described by Lettow *et al.* \[[@b17-sensors-09-04635]\] with minor modifications. In a routine preparation, 2 g of Pluronic P123 was completely dissolved in 75 mL of 1.6 M HCl at 35 °C with stirring, 0.2 g of TMB as a swelling agent was then added with stirring for 1 h. Later on, 4.25 g of TEOS was added to serve as the silica source. The mixture was further stirred for 24 h at 35 °C and aged without stirring for 48 h at 100 °C, while the solids were recovered by filtration and air-dried at room temperature. Finally, the product was calcined under 600 °C for 2 h to remove remaining triblock copolymer.
3.3.. Characterizations of SBA-15 Mesoporous Silica
---------------------------------------------------
The X-ray diffraction (XRD) measurements of calcined SBA-15 were performed on a X\'Pert MPD pro diffractometer (PANalytical, ALMELO, The Netherlands) using CuK~α~ radiation (λ = 1.5418 Å) in the range of 0.3 ∼ 3° 2θ with step of 0.03° per second. The specific surface area, total pore volume, and pore size distribution of the SBA-15 were estimated by nitrogen adsorption and desorption under 77K with a Micromeritics ASAP 2020 instrument (Norcross, GA, USA), and BJH pore diameter was obtained based on the Barrett-Joyner-Halenda (BJH) calculation. The mesoporous structure was imaged by a JEOL transmission electron microscope (TEM, JEM-2010 at 200 kv, Tokyo, Japan) and by a JEOL field emission scanning electron microscope (SEM, JSM-7000F at 15 kV, Tokyo, Japan).
3.4.. Assay of Protein Activity
-------------------------------
The HRP activity assay was performed by the 1-Step™ ABTS protocol (PIERCE Chemical Co., Rockford, IL, USA) according to manufacturer\'s procedure. In brief, one milliliter of assay mixture contained 150 μL of 1-Step™ ABTS reagent and 1,850 μL of diluted HRP solution in 0.1 M phosphate buffer (pH 6.2). The mixture was incubated at room temperature for 15 minutes and the reaction was stopped by adding 100 μL of stop solution containing 1% SDS. The absorbance was measured at 410 nm with a Genesys 2 UV-vis spectrophotometer (Rochester, NY, USA). For HRP immobilized in SBA-15, the reaction was carried out in a 1.5 mL Eppendorf tube and the final product was centrifuged with 7,000 rpm for 5 minutes. The supernatant was subjected to the measurement of absorbance at 410 nm.
3.5.. Immobilization of HRP
---------------------------
Prior to immobilization, enzyme stock solution was prepared by dissolving 16.8 mg of HRP (5,000 units) in 1 mL of 0.1 M phosphate buffer (pH 6.2), and then was aliquoted and stored at -80 °C. To perform enzyme immobilization, 20 μL of HRP stock solution (100 units) was first diluted to 200 μL by 0.1 M phosphate buffer (pH 6.2) and subsequently 10 mg of SBA-15 was suspended in the enzyme solution for 1 h at 4 °C on a rotator. The loaded SBA-15 (SBA-15(HRP)) was recovered by centrifugation with 7,000 rpm for 5 min and the supernatant was also collected for protein analysis. The SBA-15(HRP) was washed by 0.1 M phosphate buffer for at least 5 times, and was finally resuspended in 10 mL of 0.1 M phosphate buffer (pH 6.2). The amount of HRP immobilized on SBA-15 was determined according to Bradford assay with a Genesys 2 UV-vis spectrophotometer (Rochester, NY, USA) and HRP activity was assessed by 1-Step™ ABTS method (PIERCE Chemical Co., Rockford, IL, USA) according to the manufacturer\'s procedure.
3.6.. Fabrication of the Biosensor
----------------------------------
The PANI/Pt electrode was prepared following a procedure similar to that described our previous publication \[[@b1-sensors-09-04635]\]. The Pt/ceramic electrode with a desired pattern (area: 0.28 cm^2^) was constructed by sputtering platinum to a ceramic plate (area: 2 cm^2^) with a shadow mask for 600 sec on a sputter instrument (JFC-1200, JEOL, Tokyo, Japan), then washed with 3M NaOH and 3M HCl, rinsed with water, and finally dried under 50 °C for one hour. A certain amount of aniline was then electropolymerized onto the Pt/ceramic base by immersing the working electrode into a solution containing 1 M HCl, 0.1 M aniline, whilst the potential was swept from 0.0 to 1.0 V for certain cycles under an ambient condition. The fashioned PANI/Pt electrodes were then immersed in 0.1 M phosphate buffer (pH 4.0) and reduced at -0.5 V for 20 min to remove the remaining chloride ions that were possibly embedded in the polymer matrix. It was then oxidized in the same phosphate buffer at 0.6 V for 10 min. To construct SBA-15/PANI/Pt and SBA-15(HRP)/PANI/Pt electrodes, 20 μL SBA-15 or SBA-15(HRP) solution was carefully dropped onto the surface of PANI/Pt electrode and air-dried at room temperature. The electrodes were then stabilized by dropping a 2.5% (v/v) glutaraldehyde solution, and incubated at 4 °C for overnight to form the covalent linkages. Afterward, the constructed electrode was rinsed with PBS buffer (pH 5.6) thoroughly and stored in a 4 °C refrigerator. The measurements were preferred to be performed within 48 hours.
3.7.. Electrochemical Measurement
---------------------------------
A PC-controlled CHI621B electrochemical analyzer (CH Instruments, Austin, TX, USA) was employed to run cyclic voltammetric experiments for electrode preparation and hydrogen peroxide measurement. All experiments were performed in a miniature electrochemical cell using a modified PANI/Pt electrode (area: 0.28 cm^2^) as the working electrode, a platinum wire as the auxiliary electrode, and an Ag/AgCl (3M NaCl) electrode as the reference electrode. The reductions of H~2~O~2~ on electrodes were quantified with cyclic voltammetry in 0.1 M phosphate buffer (pH 6.2). The buffer had undergone deoxygenation with highly pure nitrogen for 20 min before a certain amount of H~2~O~2~ was added. During the calibration, pure nitrogen gas was gentle purged on the surface of the sample solution to create an anaerobic atmosphere.
4.. Conclusions
===============
We have presented a new strategy for the fabrication of hydrogen peroxide biosensor based on entrapping HRP in mesoporous SBA-15 and depositing on a PANI modified Pt electrode. Our results further indicated that the synthetic SBA-15 particle possessed well-defined pore geometry and high internal surface area that was able to enhance the physical adsorption of enzyme molecules. The proper pore size distribution of SBA-15 was suitable for the entrapment of HRP and maintaining its bioactivity. Meanwhile, the novel GA/SBA-15(HRP)/PANI/Pt biosensor exhibited enhanced sensitivity and a fine linear correlation between the cathodic response and the concentration of H~2~O~2~ in the range of 0.02 to 18.5 mM (*R*^2^ = 0.997). In particular, the current approach by utilizing SBA-15 to entrap HRP provided the biosensor with improved stability for multiple measurements. As shown in [Figure 5](#f5-sensors-09-04635){ref-type="fig"}, the lost of current response was mainly occurred during the initial cyclic voltammetric measurement, suggesting a few of HRP molecules were excluded by the applied potential.
Furthermore, the pore size of SBA-15 can be easily adjusted by controlling the synthesis conditions in the presence of pore expanding reagents, such as 1,3,5-trimethylbenzene (TMB) \[[@b21-sensors-09-04635],[@b37-sensors-09-04635]-[@b39-sensors-09-04635]\], and biomolecules with different molecular mass can be entrapped in SBA-15 of proper pore size accordingly. Meanwhile, the internal surface of SBA-15 has been successfully modified by various organic functional groups, such as amine, thiol, and carboxylic acid, thereby providing additional improvement to minimize the leaching of biomolecules \[[@b20-sensors-09-04635]\]. Nevertheless, the entrapment of biomolecules in SAB-15 may provide new aspects of enhancing the performance of enzyme-based biosensors, in particular offering better stability and multiple usages.
This work was supported by grants from National Science Council of ROC (NSC 93-2214-E-029- 006) to Y. Gu.
{#f1-sensors-09-04635}
{#f2-sensors-09-04635}
{#f3-sensors-09-04635}
{#f4-sensors-09-04635}
{#f5-sensors-09-04635}
{#f6-sensors-09-04635}
{#f7-sensors-09-04635}
######
Pore characterizations of SBA-15 and SBA-15(HRP).
**Sample** *A***~BET~ (m^2^/g)** *V***~total~ (cm^3^/g)** *a***~0~ (Å)** *D***(Å)**
------------- ----------------------- -------------------------- ---------------- ------------
SBA-15 708.5 0.92 12.56 76
SBA-15(HRP) 632.3 0.85 12.16 76
***A*~BET~**: total specific surface area; ***V*~total~**: total pore volume; ***a*~0~**: lattice parameter; **D**: BJH pore diameter.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-molecules-23-01525}
===============
*Hedyotis diffuse* Willd. (HD) is a well-known Chinese folk-medicine with a spectrum of pharmacological activities, including anti-cancer, antioxidant, anti-inflammatory, anti-fibroblast, immunomodulatory and neuroprotective effects, especially the anti-cancer effect in practice \[[@B1-molecules-23-01525]\]. Almost 200 compounds have been identified in HD, including iridoids, flavonoids, anthraquinones, phenylpropanoids, phenolics and their derivatives, sphingolipids, volatile oils and miscellaneous compounds \[[@B1-molecules-23-01525],[@B2-molecules-23-01525],[@B3-molecules-23-01525]\].
*Hedyotis corymbosa* (L.) Lam. (HC), another species of the same genus, is also used interchangeably in China as a health supplement and for disease prevention. It is reported to possess antioxidant \[[@B4-molecules-23-01525],[@B5-molecules-23-01525]\], anti-inflammatory \[[@B6-molecules-23-01525]\], hepatoprotective \[[@B7-molecules-23-01525],[@B8-molecules-23-01525]\], antitumor \[[@B9-molecules-23-01525],[@B10-molecules-23-01525]\], antimalarial \[[@B11-molecules-23-01525]\] and anti-nociceptive \[[@B12-molecules-23-01525]\] activities. Iridoids, carboxylic acids, flavonoids, phenolics and their derivatives, triterpenes, anthranquinones and coumarins were isolated from HC \[[@B13-molecules-23-01525],[@B14-molecules-23-01525],[@B15-molecules-23-01525]\]. Iridoid glycosides were reported as the main constituents \[[@B16-molecules-23-01525]\]. Oleanolic acid and ursolic acid were also considered as biologically active ingredients \[[@B17-molecules-23-01525],[@B18-molecules-23-01525]\].
HD and HC are closely related species of the Rubiaceae family. Due to their similar morphology, they are often mixed up. Recently, a systematic survey on confusable Chinese herbal medicines has revealed that HC is indiscriminately sold as HD in wholesale markets or food markets \[[@B19-molecules-23-01525]\]. This confusion in the market has led to a growing concern about the identification and quality evaluation of HD and HC.
Several methods using various techniques have been established to distinguish between these two species, such as loop-mediated isothermal amplification technique (LAMP) \[[@B20-molecules-23-01525]\], fluorescence microscopy \[[@B21-molecules-23-01525]\], thin layer chromatography (TLC) \[[@B22-molecules-23-01525]\], DNA sequencing of the complete internal transcribed spacer region and chemical analysis \[[@B23-molecules-23-01525]\], phylogenetic utility of nuclear ribosomal DNA (nrDNA) internal transcribed spacers (ITS) \[[@B24-molecules-23-01525]\], high-performance liquid chromatography (HPLC) \[[@B25-molecules-23-01525]\], etc. As a result, markers such as hedyotiscone A \[[@B22-molecules-23-01525]\], scandoside methyl ester \[[@B25-molecules-23-01525]\], (9*R*,10*S*,7*E*)-6,9,10-trihydroxyoctadec-7-enoic acid \[[@B26-molecules-23-01525]\] for HC, 6-*O*-(*E*)-*p*-coumaroyl scandoside methyl ester \[[@B23-molecules-23-01525],[@B25-molecules-23-01525]\], (10*S*)-hydroxypheophytin a \[[@B23-molecules-23-01525]\], 6-*O*-(*E*)-*p*-coumaroyl scandoside methyl ester-10-methyl ether and 6-*O*-*p*-feruloyl scandoside methyl ester \[[@B25-molecules-23-01525]\] for HD have been found. The UPLC-UV (detection wavelength at 254 nm) fingerprint of HC was also established to distinguish it from HD \[[@B27-molecules-23-01525]\]. The contents of oleanolic acid and ursolic acid were significantly different \[[@B28-molecules-23-01525]\].
Untargeted metabolomics, with the ability to profile diverse classes of metabolites, is primarily used to compare the overall small-molecule metabolites of different samples \[[@B29-molecules-23-01525]\]. It is mainly applied in metabolites identification through mass-based search strategy followed by manual or automated verification. The combination of ultra-high performance liquid chromatography (UPLC) separation, quadrupole time-of-flight tandem mass spectrometry (Q/TOF-MS) detection and the automated data processing software UNIFI with a scientific library is frequently applied in the characterization of chemical constituents of herbal medicines \[[@B30-molecules-23-01525],[@B31-molecules-23-01525],[@B32-molecules-23-01525],[@B33-molecules-23-01525]\] and traditional Chinese medicine injection recently \[[@B34-molecules-23-01525]\]. High-resolution tandem mass spectrum can provide an accurate and specific mass when the coeluting components possess different *m*/*z* values. UNIFI, a high throughput, comprehensive, simple and efficient platform, offers the approach to integrate data acquisition, data mining, library searching and report generation. The Traditional Medicine Library within the platform contains more than 6000 compounds from 600 herbs.
The aim of the study was search for potential biomarkers in order to systematically screen chemical components and the non-targeted metabolomic analysis of the two species, and in turn providing the basis for establishment of HC and HD quality criterion in the future. UPLC-QTOF-MS^E^, UNIFI platform and multivariate statistical analyses, such as principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) were used to profile these two herbs. The established method could enable us to find the similarities and differences between them, and provide data for the establishment of HC and HD quality criterion in the future. This comprehensive and unique phytochemical profile study revealed the structural diversity of secondary metabolites and the different patterns in HC and HD. The method developed in this study can be used as a standard protocol for identifying and discriminating species of HC and HD.
2. Experimental {#sec2-molecules-23-01525}
===============
2.1. Materials and Reagents {#sec2dot1-molecules-23-01525}
---------------------------
HC and HD were purchased from herbal markets or collected from their respective cultivation areas in China ([Table 1](#molecules-23-01525-t001){ref-type="table"}). The corresponding voucher specimens had been deposited in the Research Center of Natural Drug, School of Pharmaceutical Sciences, Jilin University, China. All the HC and HD samples were identified with the macroscopic and microscopic characters according to the *Standard of Chinese Medicinal Materials in Guangdong Province* (2004 Edition) and the *Standard of Chinese Medicinal Materials in Shaanxi Province* (2015 Edition). In these Standards, the identified methods only focus on the different macroscopic and microscopic characters. As the chemical constitutes are concerned, both oleanolic acid and ursolic acid are used to quality control. That is to say, there are no biomarkers to distinguish HC from HD.
Acetonitrile and methanol were UPLC-MS pure grade (Fisher Chemical Company, Geel, Belgium). Formic acid for UPLC was purchased from Sigma-Aldrich Company (St. Louis, MO, USA). Deionized water was purified using a Millipore water purification system (Millipore, Billerica, MA, USA). All other chemicals were of analytical grade. For reference substance, ursolic acid (110742-201622), citric acid (111679-201602), chlorogenic-acid (110753-201716), geniposide (110749-201718), luteolin 7-*O*-*β*-[d]{.smallcaps}-glucopyranoside (111968-201602), rutin (100080-201409), quercetin (100081-201610), kaempferol (110861-201611), hesperidin (110721-201617) were purchased from the National Institutes for Food and Drug Control (Beijing, China). Scandoside (20170503), alizarin 1-methyl ether (20170608) were purchased from Nanjing DASF Biotechnology Co., Ltd. (Nanjing, China). Scandoside methyl ester (20171001), 5,6,7,4′-tetramethoxyflavone (20171011), geniposidic acid (20171024) were purchased from Sichuan Weikeqi Biotechnology Co., Ltd. (Chengdu, China). 6-Methoxy-8-methylcoumarin (16018), sanlengdiphenyllactone (15025) were provided by the Research Center of Natural Drugs, School of Pharmaceutical Sciences, Jilin University, China.
2.2. Sample Preparation and Extraction {#sec2dot2-molecules-23-01525}
--------------------------------------
All the whole plants, including HC (HC1\~HC10) and HD (HD1\~HD10), were air-dried, grinded and sieved (40 mesh) to get the homogeneous powder respectively. Then, the powder of 20 samples (200 mg per sample) were extracted respectively with 80% methanol (2L × 3) at 80 °C for three times (3 h each time) with the reflux method. The extraction procedure is repeated until the extracted solution is colorless. After filteration, the extracts of each sample were combined, concentrated and evaporate to dryness. As a result, 20 desiccated extract powders were obtained. Each powder was dissolved in 1.0 mL of 80% methanol. Subsequently, each methanolic solution was filtered and injected directly into the UPLC system. The volume injected of each sample was 2 μL for each run. Furthermore, the methanol blank were run with the same gradient program between two samples during the whole sample list. The wash volume between injections was enough for avoiding carry over. Meanwhile, 20-μL aliquots of each HD and HC sample were mixed to obtain a quality control (QC) sample, which contained all of the components in the analysis. The QC sample was run every five samples to monitor the stability of the system.
2.3. Ultra-High Performance Liquid Chromatography with Quadrupole Time-of-Flight Tandem Mass Spectrometry (UPLC-QTOF-MS) {#sec2dot3-molecules-23-01525}
------------------------------------------------------------------------------------------------------------------------
The separation and MS detection of components were performed on a Waters Xevo G2-XS QTOF mass spectrometer (Waters Co., Milford, MA, USA) connected to the UPLC system through an electrospray ionization (ESI) interface. UV wavelength did not trigger the MS detection of components. The column used was an ACQUITY UPLC BEH C~18~ (100 mm × 2.1 mm, 1.7 μm) from Waters Corporation (Milford, MA, USA). The mobile phases consisted of eluent A (0.1% formic acid in water, *v*/*v*) and eluent B (0.1% formic acid in acetonitrile, *v*/*v*) with a flow rate of 0.4 mL/min following a liner gradient program: 10% B from 0 to 2 min, 10--90% B from 2 to 25 min, 90% B from 25 to 26 min and 90--10% B from 26 to 26.1 min. The temperature of the UPLC column and sample was set at 30 °C and 15 °C. Mixtures of 10/90 and 90/10 water/acetonitrile were used as the strong wash and the weak wash solvent respectively. The optimized instrumental parameters were as follows: capillary voltage floating at 2.6 kV (ESI^+^) or 2.2 kV (ESI^−^), cone voltage at 40 V, source temperature at 150 °C, desolvation temperature at 400 °C, cone gas flow at 50 L/h and desolvation gas flow at 800 L/h. In MS^E^ mode, collision energy of low energy function was set to 6 V, while ramp collision energy of high energy function was set to 20--40 V. Each sample was analyzed by UPLC-QTOF-MS^E^ mode; data acquisition was performed via the mass spectrometer by rapidly switching from a low-collision energy (CE) scan to a high-CE scan during a single LC run. The low-CE experiment provides information about the intact molecular ion, e.g., \[M+H\]^+^, while the high-CE scan generates fragment ion information. Alignment of the low-CE and high-CE data is automatically performed by the software. To ensure mass accuracy and reproducibility, the mass spectrometer was calibrated over a range of 100--1200 Da with sodium formate. Leucine enkephalin was used as external reference of Lock Spray™ infused at a constant flow of 10 μL/min. In addition, MassLynx data were recorded in continuous mode during acquisition.
2.4. Chemical Information Database for the Components of HC and HD {#sec2dot4-molecules-23-01525}
------------------------------------------------------------------
In addition to the Waters Traditional Medicine Library in the UNIFI software, a systematic investigation of chemical constituents was conducted. A self-built database of compounds isolated from HC and HD was established by searching online databases such as China Journals of Full-Text Database (CNKI), PubMed, Medline, Web of Science and ChemSpider. The name, molecular formula and structure of components from HC and HD were obtained in the database.
2.5. Data Analysis by UNIFI Platform {#sec2dot5-molecules-23-01525}
------------------------------------
Data analysis was performed on UNIFI 1.7.0 software (Waters, Manchester, UK). Emphasis was put on analyzing structural characteristics and MS fragmentation behaviors, especially for characteristic fragments. Minimum peak area of 200 was set for 2D peak detection. The peak intensity of high energy over 200 counts and the peak intensity of low energy over 1000 counts were the selected parameters in 3D peak detection. A margin of error up to 5 ppm for identified compounds was allowed. We selected positive adducts containing +H and +Na and negative adducts including +COOH and −H. For exact mass accuracy, with leucine enkaplin as the reference compound, \[M+H\]^+^ 556.2766 was used for positive ion and \[M−H\]^−^ 554.2620 was used for negative ion in the UNIFI platform.
The MS raw data were processed using the streamlined workflow of UNIFI software to quickly identify the chemical components that met the match criteria with the Traditional Medicine Library. Firstly, an in-house scientific library was created including the information of chemical components from the target herbs based on the literature, saved as Mol file format, and then, the newly built library was imported into the analysis method, in virtue of some compounds being missing in the Traditional Medicine Library. Secondly, the raw data was compressed by Waters Compression and Archival Tool v1.10 and imported into the software. Thirdly, automated screening and identification were performed by the UNIFI platform instead of manually extracting each individual chromatographic peak, calculating the elementary composition and then analyzing MS fragmentation behaviors. Fourthly, we set up a filter to refine results, being mass error between −5 and 5 ppm, and additionally, response value greater than 6000. Finally, further verification of compounds by comparison with retention time of reference substances and characteristic MS fragmentation patterns reported in literature was carried out. After processing and filtering of the data by UNIFI, all selected components were listed for further verification, including information such as compound name, chemical structure, mass error, adducts, response, extracting ion chromatograms and spectra of low energy and high energy. The components were listed by descending response order and confirmed by reference substances or comparison with literatures.
2.6. Metabonomics Analysis {#sec2dot6-molecules-23-01525}
--------------------------
MarkerLynx XS V4.1 software (Waters, Manchester, UK) was used to process the raw data for alignment, deconvolution, data reduction, etc. As a result, the list of mass and retention time pairs with corresponding intensities for all the detected peaks from each data file. The main parameters were as follows: retention time range 0--26 min, mass range 100--1200 Da, mass tolerance 0.10, minimum intensity 5%, marker intensity threshold 2000 counts, mass window 0.10, retention time window 0.20, and noise elimination level 6. Furthermore, also with the MarkerLynx XS V4.1 software, principle component analysis (PCA) and orthogonal projections to latent structures discriminant analysis (OPLS-DA) were applied to analyze the above resulting data. Whether these two species are different would depend on the separation between HD and HC groups. The obvious separation in PCA score plots means they are differentiated. The supervised pattern recognition approach OPLS-DA can visualize and depict general metabolic variation between two groups. To identify the metabolites contributing to the discrimination, S-plots and VIP-plots were obtained via OPLS-DA analysis to find potential biomarkers that significantly contributed to the difference among HC and HD. Each spot in S-plots represents a variance. The importance of each variance to classification is determined by the value of variable importance in the projection (VIP) and metabolites with VIP value above 2.0 were considered as potential markers.
3. Results {#sec3-molecules-23-01525}
==========
3.1. Identification of Components from HC and HD {#sec3dot1-molecules-23-01525}
------------------------------------------------
A total of 113 compounds were identified or tentatively characterized in both positive and negative mode from HC and HD ([Table 2](#molecules-23-01525-t002){ref-type="table"}), the base peak intensity (BPI) chromatograms are shown in [Figure 1](#molecules-23-01525-f001){ref-type="fig"}, and their chemical structures are shown in [Figure 2](#molecules-23-01525-f002){ref-type="fig"}. In HC and HD 109 and 104 compounds were characterized, respectively. Both herbs are rich in natural components with various structural patterns, including iridoids, flavonoids, organic acids and organic acid esters, tannins, alcohols, ketones, coumarins, anthraquinones, monoterpenes, triterpenoids, etc. Some of these compounds have isomers may be distinguished based on characteristic MS fragmentation patterns reported in literature, or comparison of retention times to reference substances.
80 common constituents were identified from HC and HD. Among them, there were eleven iridoids (compounds **6**, **8**, **11**, **14**, **18**, **20**, **29**, **51**, **53**, **58** and **59**), thirteen flavonoids (compounds **7**, **17**, **25**, **26**, **27**, **31**, **36**, **37**, **38**, **39**, **43**, **56** and **61**), one monoterpene (compound **10**), one anthraquinone (compound **68**), two ketones (compounds **34** and **67**), three tannins (compounds **4**, **73** and **60**), five alcohols (compounds **13**, **80**, **82**, **98** and **99**), and the rest are organic acids and organic acid esters, triterpenoids, coumarins, alkaloid, phenol, amide and glycoside. The contents of above components were similar in these two herbs.
3.2. Biomarker Discovery for HD and HC {#sec3dot2-molecules-23-01525}
--------------------------------------
PCA, a classic unsupervised lowering-dimension pattern recognition model, can be used to select distinct variables and to find potential biomarkers. It was firstly established based on the spectra of HD and HC samples to discern the presence of inherent similarities in mass spectral profiles as displayed in [Figure 3](#molecules-23-01525-f003){ref-type="fig"}. Two parameters, R^2^ (cum) and Q^2^ (cum), are commonly used to assess the quality of the PCA model, with values close to 1.0 indicative of good fitness and predictive ability. In the present study, R^2^X (cum) and Q^2^ (cum) were 0.6909 and 0.6257, respectively, indicating good fitness and prediction of the constructed PCA model.
Based on the obtained PCA score plots ([Figure 3](#molecules-23-01525-f003){ref-type="fig"}), the 20 samples were obviously divided into two main groups according to different species (HD and HC). The HD samples were noticeably overlapping, which indicates good similarity among them, and this result was also observed for HC samples. Meanwhile, the HD group and the HC group were completely separated, indicating that these two species herbs could be differentiated. The QC samples were between the two species, which came from the fact that they were mixed volumetrically in 50%.
In order to distinguish HD from HC, OPLS-DA models were built in both positive and negative modes. OPLS-DA score plot, S-plot, variable trend and VIP (variable importance in the projection) values were obtained to understand which variables are responsible for separation \[[@B109-molecules-23-01525]\].
As shown in [Figure 4](#molecules-23-01525-f004){ref-type="fig"}, OPLS-DA models were constructed to discriminate the difference under the already established separation between different groups based on the PCA results. Each model has 2 score components (HD and HC). These scores are weighted averages of the original ones, hence providing a good summary. In addition, these scores display the separation of the groups in both ESI^+^ and ESI^−^ modes. The scores t\[1\] (*x*-axis) and to\[1\] (*y*-axis) are the two most important new variables in summarizing and separating the data. Each point in the plot corresponds to an observation. The groups are shown in different shapes and the separation of the groups is easily visible in t\[1\]. The to\[1\] score values show the variation within each class. This variation can either be caused by biological variation or by systematic changes in the experimental setup.
[Figure 5](#molecules-23-01525-f005){ref-type="fig"} displays the variable importance (VIP) versus the PLS-regression coefficients. Important X-variables have large positive VIP values and large positive or negative coefficient values. The covariance p\[1\] and correlation p(corr)\[1\] loadings from a two class OPLS-DA model were shown here in S-Plot format ([Figure 6](#molecules-23-01525-f006){ref-type="fig"}). The points are Exact Mass/Retention Time pairs (EMRTs). The upper right quadrant of the S-plot shows those components which are elevated in HC, the control group, while the lower left quadrant shows EMRTs elevated in HD, the treated group. The farther along the *x*-axis the greater the contribution to the variance between the groups, while the farther the *y*-axis the higher the reliability of the analytical result. Based on VIP values (VIP \> 4) ([Figure 5](#molecules-23-01525-f005){ref-type="fig"}) and *p* values (*p* \< 0.05) \[[@B110-molecules-23-01525]\] from univariate analysis, and the identification of components from HC and HD ([Table 2](#molecules-23-01525-t002){ref-type="table"}), 33 robust known biomarkers enabling the differentiation between HD and HC were discovered and marked in S-plots ([Figure 6](#molecules-23-01525-f006){ref-type="fig"}). In order to systematically evaluate the biomarkers, a heatmap was generated from these biomarkers (shown in [Figure 7](#molecules-23-01525-f007){ref-type="fig"}), which shows distinct segregation between two species.
4. Discussion {#sec4-molecules-23-01525}
=============
There are 109 and 104 compounds characterized from HC and HD respectively. Sixty compounds were identified in ESI^−^ mode and 53 compounds were identified in ESI^+^ mode. According to the BPI chromatograms of HC and HD, it seems that ESI^−^ ionization mode is better than ESI^+^ based on the quantity and the responses of the identified compounds, but it is still necessary to run the ESI^+^ mode because some compounds showed better respond than in ESI^−^ mode.
It was revealed that HD and HC differed in their chemical composition according to the HPLC analysis \[[@B19-molecules-23-01525]\]. It was also indicated that 6-*O*-(*E*)-*p*-coumaroyl scandoside methyl ester and 6-*O*-(*E*)-*p*-coumaroyl scandoside methyl ester-10-*O*-methyl ether were the main components of HD. In 2007, Liang et al. reported that HD and its substitutes could be identified based on HPLC chemical fingerprints and mass spectrometric analysis \[[@B25-molecules-23-01525]\]. MS combined with UV spectra and literature values was used to obtain the chemical information. As a result, four compounds, asperuloside, 6-*O*-(*E*)-*p*-coumaroyl scandoside methyl ester, 6-*O*-(*E*)-*p*-coumaroyl scandoside methyl ester-10-methyl ester and 6-*O*-*p*-feruloyl scandoside methyl ester were recommended to be used as chemical markers for quality evaluation and chemical authentication of HD and its substitutes. In addition, scandoside methyl ester detected in the chromatograms of HC can be used as the characteristic peaks \[[@B25-molecules-23-01525]\]. Furthermore, a previous report found that hedyotiscone A could be used to differentiate HC from HD using TLC method \[[@B22-molecules-23-01525]\]. In our study, asperuloside, 6-*O*-(*E*)-*p*-coumaroyl scandoside methyl ester-10-methyl ester, scandoside methyl ester, 6-*O*-*p*-feruloyl scandoside methyl ester and hedyotiscone A were shared in HC and HD, but the reported result concerning 6-*O*-(*E*)-*p*-coumaroyl scandoside methyl ester was consistent with our findings.
In the other record, another marker compound, 10(*S*)-hydroxypheophytin a, isolated with a yield of 22 mg from 600 g of HC, was identified exclusively in HD \[[@B23-molecules-23-01525]\]. It is a pity that it was not be detected under our experimental conditions. Similarly, (9*R*,10*S*,7*E*)-6,9,10-trihydroxyoctadec-7-enoic acid, isolated with a yield of 47.9 mg from 20 kg of HC, was reported to be used to differentiate HC from HD \[[@B26-molecules-23-01525]\]. It was not be detected under our experimental conditions either.
In this study, 33 known compounds enabling the robust differentiation between HC and HD were detected. For HC, there were 18 potential biomarkers, including three iridoids (**23**, **55**, **66**), eight flavonoids (**30**, **35**, **40**, **42**, **47**, **71**, **75**, **81**), two tannins (**19**, **45**), two ketones (**22**, **91**), one alcohol (**92**), two monoterpenes (**89**, **90**). Among these potential biomarkers, the contents of nine components (**19**, **22**, **23**, **30**, **35**, **40**, **45**, **66**, **92**) in HC were much greater than in HD. Compounds **42**, **47**, **55**, **71**, **75**, **81**, **89**, **90** and **91** could be detected only in HC. It's worth mentioning that two iridoids, compounds **55** (hedycoryside B) and **66** (hedycoryside A), with high responses in UPLC-MS might be used for rapid identification of HC. For HD, there were 15 potential biomarkers including two iridoids (**52**, **50**), eight flavonoids (**41**, **44**, **49**, **54**, **57**, **62**, **79**, **84**), one tannin (**46**), one ketone (**70**), and three anthraquinones (**69**, **77**, **78**). Among them, the contents of eleven components (**41**, **44**, **46**, **49**, **52**, **57**, **62**, **70**, **77**, **78**, **79**) in HD were much higher than those in HC. Compounds **50**, **54**, **69** and **84** were detected only in HD. In addition, two anthraquinones, compounds **69** (1,3-dihydroxy-2-methylanthraquinone) and **78** (2-hydroxy-3-methylanthraquinone) with high responses in UPLC-MS might be used for rapid identification of HD.
However, there are still some unresolved issues. Firstly, the pharmaceutical effects associated with these identified compounds should be screened in the future. Secondly, as shown in BPI chromatograms, though 113 compounds were identified, there are still some unidentified components. Further research should be carried out based on the formula of these unknown compounds. Thirdly, source material is not seasonable as it was collected during summer time. Fourthly, collecting HC and HD in the same area may be the better way for comparison. But in this study, Haikou City for HC and Fuzhou City for HD were visited. To some extent, the collection of these samples might be used as negative controls for another species because it could eliminate the influence of the region on the analysis of the sample. But unfortunately, the regional factor should not be considered as there should be more samples per region.
5. Conclusions {#sec5-molecules-23-01525}
==============
Under the optimized conditions, a total of 109 chemical compounds with different structural types were identified from HC and 104 from HD. The similarities and differences between these two herbs were also highlighted in the paper. Various structural patterns including iridoids, flavonoids, organic acids and organic acid esters, tannins, alcohols, ketones, coumarins, anthraquinones, monoterpenes, triterpenoids were presenting in these two herbs, of which there were 80 shared compounds in HC and HD. There is quite a difference in the parent structures types between HC and HD. A total of 33 robust biomarkers enabling the differentiation between HC and HD were discovered. For HC and HD, 18 and 15 potential biomarkers, respectively, were identified in this paper. Two iridoids, hedycoryside B (compound **55**) and hedycoryside A (**66**) might be used for rapid identification of HC, and two anthraquinones, 1,3-Dihydroxy-2-methylanthraquinone (compound **69**) and 2-Hydroxy-3-methylanthraquinone (**78**) might be used for rapid identification of HD based on their presence and content. Actually, these solid biomarkers are recommended for further use in the recognition and distinction between HC and HD. The results provided reliable characterization profiles to identify these two herbs and to clarify the fundamental pharmacological substances. Different chemical compositions will inevitably lead to different biological effects of HC and HD in clinical application. HC should not be used as substitute of HD. The results provided data on the chemical constituents of HC and provide a reference for the quality control of HD in the aspect of quantitative determination.
**Sample Availability:** Samples of the compounds 6-Methoxy-8-methyl coumarin and Sanlengdiphenyllactone are available from the authors.
J.L. conceived and designed the experiments; Y.W., C.W. and H.L. performed the experiments; Y.W., Y.L. (Yunhe Liu), Y.L. (Yameng Li) and Y.Z. were responsible for data analysis. J.L. wrote the paper. J.L. and P.L. assisted paper revision.
This research was supported by the Biomedicine Special Foundation for Government-Univeristy Cooperation Project of Jilin Province \[No. SXGJSF2017-1-1-(02)\].
The authors declare that they have no conflicts of interest concerning this article.
{#molecules-23-01525-f001}
######
Chemical structures of compounds identified in HD and HC.
{#molecules-23-01525-f003}
{#molecules-23-01525-f004}
{#molecules-23-01525-f005}
{#molecules-23-01525-f006}
{#molecules-23-01525-f007}
molecules-23-01525-t001_Table 1
######
The list of the tested samples from China.
Sample No. Source Collection Time
------------ ------------------------------------------------------ -------------------
HC 1 Guangzhou City, Guangdong Province, China; market 15 September 2016
HC 2 26 August 2017
HC 3 Haikou City, Hainan Province, China; market 11 August 2017
HC 4 Nanning City, Guangxi Province, China; field 8 July 2016
HC 5 5 July 2017
HC 6 Kunming City, Yunnan Province, China; market 13 August 2016
HC 7 1 September 2017
HC 8 Shenzhen City, Guangdong Province, China; cultivated 28 September 2017
HC 9 Luoding County, Guangdong Province, China; market 24 August 2016
HC 10 12 July 2017
HD 1 Nanning City, Guangxi Province, China; field 12 July 2016
HD 2 20 July 2016
HD 3 Luoding County, Guangdong Province, China; market 13 August 2016
HD 4 15 July 2017
HD 5 Guangzhou City, Guangdong Province, China; market 13 July 2017
HD 6 Shenzhen City, Guangdong Province, China; cultivated 21 September 2016
HD 7 21 August 2017
HD 8 Kunming City, Yunnan Province, China; market 8 August 2016
HD 9 Fuzhou City, Fujian Province, China; field 22 August 2017
HD 10 13 September 2017
molecules-23-01525-t002_Table 2
######
Compounds identified from HD and HC by UPLC-QTOF-MS^E^.
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
No. t~R~ (min) Formula Calculated Mass (Da) Theoretical Mass (Da) Mass Error (ppm) MS^E^ Fragmentation Identification Source Ref.
----------- ------------ -------------------- ---------------------- ----------------------- ------------------ ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------- ------------ --------------------------------
**1** 0.61 C~7~H~12~O~6~ 192.0633 192.0634 −0.4 191.0560 \[M−H\]^−^, 129.0190 \[M−C~2~H~2~O~2~\]^−^, 127.0407 \[M−OH−COOH\]^−^ Quinic acid HC, HD s
**2** 0.69 C~4~H~6~O~5~ 134.0216 134.0215 0.9 133.0144 \[M−H\]^−^, 115.0037 \[M−OH\]^−^, 71.0150 \[M-OH-COOH\]^−^ 2-hydroxy-succinic acid HC, HD \[[@B35-molecules-23-01525]\]
**3** 0.75 C~6~H~8~O~7~ 192.0270 192.0270 0.0 191.0197 \[M−H\]^−^, 173.0085 \[M−OH\]^−^, 117.0193 \[M−OH−CH~2~COOH\]^−^, 111.0089 \[M−2 × OH−COOH\]^−^, 101.0247 \[M−2 × COOH\]^−^, 89.0250 \[M−CH~2~COOH−COOH\]^−^ Citric acid HC, HD s
**4** 0.80 C~9~H~10~O~4~ 182.0582 182.0579 1.2 227.0564 \[M+HCOO\]^−^, 165.0558 \[M-OH\]^−^, 153.0555 \[M-CHO\]^−^, 137.0611 \[M−OH−CHO\]^−^, 125.0244 \[M−2CH~3~−CHO\]^−^ Syringaldehyde HC, HD a
**5** 0.93 C~19~H~19~N~3~O 305.1505 305.1528 7 328.1397 \[M+Na\]^+^, 132.0812 \[M−C~10~H~9~N~2~O\]^+^, 117.0626 \[M−C~11~H~12~N~2~O\]^+^, 107.0503 \[M−C~11~H~10~N~2~-NCH~3~\]^+^ Wuchuyuamide I HC, HD \[[@B36-molecules-23-01525]\]
**6** 0.94 C~16~H~20~O~10~ 372.1051 372.1057 −1.4 371.0978 \[M−H\]^−^, 315.0723 \[M−C~3~HO\]^−^, 167.0712 \[M−Glu−COOH\]^−^, 153.0192 \[M−Glu−C~3~HO\]^−^, 123.0451 \[M−Glu−C~3~HO~3~\]^−^ Deacetyl asperuloside HC, HD \[[@B37-molecules-23-01525]\]
**7** 1.11 C~22~H~18~O~10~ 442.0866 442.0900 −7.8 443.0938 \[M+H\]^+^, 319.0772 \[M−C~6~H~5~O~3~\]^+^, 145.0255 \[M−C~13~H~10~O~7~\]^+^ (+)-Epicatechol 3-gallate HC, HD \[[@B38-molecules-23-01525]\]
**8** 1.20 C~17~H~24~O~11~ 404.1318 404.1319 −0.2 449.1300 \[M+HCOO\]^−^, 353.0872 \[M−OH−OCH~3~\]^−^, 247.1184 \[M−OH−C~6~H~5~O~3~\]^−^, 241.0720 \[M−Glu\]^−^, 211.0610 \[M−OCH~3~−Glu\]^−^ Scandoside methyl ester HC, HD s
**9** 1.34 C~11~H~10~O~5~ 222.0519 222.0528 −4.0 223.0592 \[M+H\]^+^, 209.0418 \[M−CH~3~\]^+^, 191.0318 \[M−OH−CH~3~\]^+^, 181.0501 \[M−C~2~H~3~O\]^+^, 179.0680 \[M−COOH\]^+^, 163.0364 \[M−OH−C~2~H~3~O\]^+^ 4-*O*-acetyl-caffeic acid HC, HD \[[@B39-molecules-23-01525]\]
**10** 1.34 C~10~H~10~O~4~ 194.0568 194.0579 −5.8 195.0641 \[M+H\]^+^, 181.0501 \[M−CH~3~\]^+^, 179.0680 \[M−OH\]^+^, 163.0364 \[M-CH~3~−OH\]^+^, 149.0581 \[M−COOH\]^+^, 145.0256 \[M−OH−OCH~3~\]^+^ 3-Hydroxy-4-methoxycinnamic acid HC, HD \[[@B40-molecules-23-01525]\]
**11** 1.44 C~16~H~22~O~10~ 374.1207 374.1213 −0.6 419.1189 \[M+HCOO\]^−^, 357.1190 \[M−OH\]^−^, 343.0975 \[M−CH~2~OH\]^−^, 313.0909 \[M−2 × CH~2~OH\]^−^, 257.0671 \[M−C~5~H~4~O~3~\]^−^ Geniposidic acid HC, HD s
**12** 1.45 C~8~H~8~O~3~ 152.0490 152.0473 9.2 175.0382 \[M+Na\]^+^, 136.0598 \[M−OH\]^+^, 119.0494 \[M−2 × OH\]^+^, 91.0561 \[M−OH−COOH\]^+^ 4-Hydroxybenzeneacetic acid HC, HD \[[@B41-molecules-23-01525]\]
**13** 1.46 C~10~H~14~O~5~ 214.0840 214.0841 −0.4 213.0768 \[M−H\]^−^, 195.0657 \[M−OH\]^−^, 181.0498 \[M−CH~2~OH\]^−^, 177.0554 \[M−2 × OH\]^−^, 163.0395 \[M−OH−CH~2~OH\]^−^, 151.0397 \[M−C~2~H~5~O~2~\]^−^, 149.0593 \[M−3 × OH−CH~3~\]^−^ Guaiacyl glycerol HC, HD \[[@B42-molecules-23-01525]\]
**14** 1.56 C~18~H~24~O~12~ 432.1260 432.1268 −1.8 431.1187 \[M−H\]^−^, 269.0663 \[M−Glu\]^−^, 165.0552 \[M−Glu−OH−C~3~H~5~O~2~\]^−^ Asperulosidic acid HC, HD \[[@B43-molecules-23-01525]\]
**15** 1.83 C~16~H~18~O~9~ 354.0939 354.0951 −3.4 355.1011 \[M+H\]^+^, 163.0383 \[M−quinic acid\]^+^, 145.0264 \[M−quinic acid−OH\]^+^ Chlorogenic acid HC, HD s
**16** 1.90 C~7~H~12~O~6~ 192.0632 192.0634 −0.7 191.0632 \[M−H\]^−^, 173.0445 \[M−OH\]^−^, 137.0239 \[M−2 × OH\]^−^, 121.0291 \[M−4 × OH\]^−^ 1,3,4,5-Tetrahydroxycyclohexanecarboxylic acid HC, HD a
**17** 2.02 C~33~H~40~O~21~ 772.2067 772.2062 0.6 817.2049 \[M+HCOO\]^−^, 609.1443 \[M−Glu\]^−^ Kaempferol-3-*O*-sophoroside-7-*O*-*β*-[d]{.smallcaps}-glucopyranoside HC, HD \[[@B44-molecules-23-01525]\]
**18** 2.09 C~16~H~22~O~11~ 390.1152 390.1162 −2.7 389.1079 \[M−H\]^−^, 209.0454 \[M−Glu\]^−^, 165.0549 \[M−Glu−OH−CH~2~OH\]^−^, 121.0658 \[M−Glu−OH−CH~2~OH−COOH\]^−^ Scandoside HC, HD s
**19 \#** 2.39 C~20~H~30~O~13~ 478.1677 478.1686 −1.8 523.1659 \[M+HCOO\]^−^, 293.0873 \[M−C~9~H~11~O~4~\]^−^, 151.0395 \[M−furanosyl−Glu\]^−^ 3,4,5-Trimethoxyphenyl−6-*O*-[d]{.smallcaps}-apio-*β*-[d]{.smallcaps}-furanosyl-*β*-[d]{.smallcaps}-glucopyranoside HC, HD\ \[[@B45-molecules-23-01525]\]
(HC\>\>HD)
**20** 2.50 C~18~H~22~O~11~ 414.1157 414.1162 −1.0 459.1139 \[M+HCOO\]^−^, 367.1029 \[M−OH−CH~2~OH\]^−^, 251.0555 \[M−Glu\]^−^, 191.0352 \[M−Glu−CH~3~COOH\]^−^, 177.0190 \[M−Glu−CH~2~COOCH~3~\]^−^ Asperuloside HC, HD \[[@B46-molecules-23-01525]\]
**21** 2.65 C~9~H~8~O~4~ 180.0423 180.0423 0.2 179.0350 \[M−H\]^−^, 165.0192 \[M−CH~3~\]^−^, 135.0451 \[M−COOH\]^−^ (4-Methoxyphenyl)-oxoacetic acid HC, HD \[[@B47-molecules-23-01525]\]
**22 \#** 3.07 C~15~H~20~O~8~ 282.1108 282.1103 1.3 327.1090 \[M−H\]^−^, 165.0556, 147.0452 \[M−Glu\]^−^, 121.0294 \[M−Glu−CH~3~−CO\]^−^ Androsin HC, HD\ \[[@B48-molecules-23-01525]\]
(HC\>\>HD)
**23 \#** 3.27 C~17~H~24~O~10~ 388.1370 388.1370 0.1 433.1352 \[M+HCOO\]^−^, 355 \[M−OCH~3~\]^−^, 353.0876 \[M−OH−CH~3~\]^−^, 337.0932 \[M−OH−OCH~3~\]^−^, 225.0770 \[M−Glu\]^−^, 193.0506 \[M−Glu−OCH~3~\]^−^ Geniposide HC, HD\ s
(HC\>\>HD)
**24** 3.82 C~14~H~17~NO~6~ 295.1052 295.1056 −1.2 340.1034 \[M+HCOO\]^−^, 167.0346 \[M−N−C~6~H~5~−2OH\]^−^, 166.0508 \[M−3 × OH−C~6~H~5~\]^−^ Prunasin HC, HD \[[@B49-molecules-23-01525]\]
**25** 4.01 C~15~H~10~O~7~ 302.0414 302.0426 −4.3 303.0486 \[M+H\]^+^, 153.0171 \[M−C~8~H~6~O~3~\]^+^, 127.0389 \[M−C~9~H~6~O~4~\]^+^ Moric acid HC, HD a
**26** 4.12 C~27~H~30~O~17~ 626.1493 626.1483 1.6 625.1420 \[M−H\]^−^, 609.1424 \[M−OH\]^−^, 595.1373 \[M−CH~2~OH\]^−^, 400.0883 \[M−OH−CH~2~OH−Glu\]^−^, 300.0282 \[M−Glu\]^−^ Quercetin−3-sophoroside HC, HD \[[@B50-molecules-23-01525]\]
**27** 4.14 C~43~H~48~O~25~ 964.2508 964.2485 2.4 963.2435 \[M−H\]^−^, 903.2227 \[M−CH~2~OH\]^−^, 757.1849 \[M−C~11~H~11~O~4~\]^−^, 625.1419 \[M−C~11~H~11~O~4~−Glu\]^−^ Quercetin-3-*O*-(6-*O*-feruloyl-*β*-[d]{.smallcaps}-glucopyranosyl)-(1→2)-*β*-[d]{.smallcaps}-galactopyranosyl-(1→2)-*β*-[d]{.smallcaps}-glucopyranoside HC, HD a
**28** 4.15 C~15~H~18~O~8~ 326.1002 326.1002 0.0 371.0984 \[M+HCOO\]^−^, 163.0403 \[M−Glu\]^−^, 119.0504 \[M−Glu−COOH\]^−^ trans-*p*-Coumaric acid-4-*O*-glucoside HC, HD \[[@B51-molecules-23-01525]\]
**29** 4.19 C~19~H~26~O~12~ 446.1421 446.1424 −0.7 491.1403 \[M+HCOO\]^−^, 371.0986 \[M−OCH~3~−C~2~H~3~O\]^−^, 283.0824 \[M−Glu\]^−^, 163.0403 \[M−Glu−OCH~3~−C~3~H~5~O~2~\]^−^, 119.0504 \[M−Glu−OH−C~5~H~8~O~4~\]^−^ Daphylloside HC, HD \[[@B52-molecules-23-01525]\]
**30 \#** 4.24 C~21~H~20~O~11~ 448.1000 448.1006 −1.3 449.1072 \[M+H\]^+^,415.1006 \[M−2 × OH\]^+^, 397.0920 \[M−3 × OH\]^+^, 287.0490 \[M−Glu\]^+^, 137.0587 \[M−Glu−C~7~H~3~O~3~\]^+^ Luteolin 7-*O*-*β*-[d]{.smallcaps}-glucopyranoside HC, HD\ s
(HC\>\>HD)
**31** 4.39 C~26~H~28~O~14~ 564.1486 564.1479 1.3 563.1414 \[M−H\]^−^, 403.1260 \[M−OH−C~9~H~6~O~2~\]^−^, 275.0578 \[M−OH−CH~2~OH−C~6~H~5~O−apiofuranosyl\]^−^ Apiin HC, HD \[[@B53-molecules-23-01525]\]
**32** 4.48 C~9~H~10~O~3~ 166.0631 166.0630 0.7 165.0558 \[M−H\]^−^, 147.0451 \[M-OH\]^−^, 119.0501 \[M−COOH\]^−^, 103.0556 \[M−OH−COOH\]^−^ Phloretic acid HC, HD \[[@B54-molecules-23-01525]\]
**33** 4.59 C~10~H~8~O~4~ 192.0413 192.0423 −4.8 193.0486 \[M+H\]^+^, 178.0247 \[M−CH~3~\]^+^, 122.0350 \[M−C~3~H~2~O~2~\]^+^ Scopoletin HC, HD \[[@B55-molecules-23-01525]\]
**34** 4.75 C~11~H~16~O~3~ 196.1097 196.1099 −0.2 197.117 \[M+H\]^+^, 179.1057 \[M−OH\]^+^, 167.0688 \[M−2× CH~3~\]^+^, 147.0436 \[M−2 × CH~3~-OH\]^+^ Loliolide HC, HD \[[@B56-molecules-23-01525]\]
**35 \#** 4.96 C~26~H~28~O~16~ 596.1375 596.1377 −0.4 595.1302 \[M−H\]^−^, 300.0280 \[M−Glu−Xyl\]^−^ Isoetin-7-*O*-*β*-[d]{.smallcaps}-glucopyranosyl-2′-*O*-*β*-[d]{.smallcaps}-xyloypyranoside HC, HD\ \[[@B57-molecules-23-01525]\]
(HC\>\>HD)
**36** 4.98 C~15~H~12~O~7~ 304.0573 304.0583 −2.8 349.0555 \[M+HCOO\]^−^, 195.0294 \[M−C~6~H~5~O~2~\]^−^, 179.0323 \[M−OH−C~6~H~5~O~2~\]^−^, 151.0036 \[M−C~8~H~7~O~3~\]^−^ Dihydroquercetin HC, HD \[[@B58-molecules-23-01525]\]
**37** 5.04 C~15~H~10~O~7~ 302.0424 302.0427 −0.3 303.0496 \[M+H\]^+^, 287.0541, 127.0395 \[M−C~9~H~6~O~4~\]^+^ 5,7,8,3′,4′-pentamethoxy Flavonoids HC, HD \[[@B59-molecules-23-01525]\]
**38** 5.04 C~27~H~30~O~16~ 610.1538 610.1534 0.7 611.1611 \[M+H\]^+^, 465.1016 \[M−Rha\]^+^, 303.0493 \[M−Glu−Rha\]^+^ Rutin HC, HD s
**39** 5.34 C~21~H~20~O~12~ 464.0948 464.0955 −1.4 463.0876 \[M−H\]^−^, 301.0353 \[M−Glu\]^−^ Quercetin-3-*O*-glucopyranoside HC, HD \[[@B60-molecules-23-01525]\]
**40 \#** 5.34 C~22~H~22~O~10~ 446.1212 446.1213 −0.2 447.1285 \[M+H\]^+^, 429.1118 \[M−OH\]^+^, 175.0383 \[M−Glu−C~6~H~3~O\]^+^, 163.0388 \[M−Glu−C~6~H~5~−OCH~3~\]^+^, 131.0489 \[M−Glu−C~7~H~3~O~3~\]^+^ Acacetin 7-*O*-*β*-[d]{.smallcaps}-glucopyranoside HC, HD\ \[[@B61-molecules-23-01525]\]
(HC\>\>HD)
**41 \*** 5.71 C~37~H~38~O~19~ 786.2011 786.2007 0.5 831.1993 \[M+HCOO\]^−^, 565.1556 \[M−CH~2~OH−C~10~H~9~O~3~\]^−^, 379.0657 \[M−Glu−CH~2~OH−C~6~H~5~O−C~7~H~7~O~2~\]^−^ Allivictoside F HC, HD\ \[[@B62-molecules-23-01525]\]
(HD\>\>HC)
**42 \#** 5.75 C~20~H~18~O~11~ 434.0848 434.0849 −0.2 433.0775 \[M−H\]^−^, 300.0280 \[M−Ara\]^−^, 163.0401 \[M−Ara−C~6~H~4~O~3~\]^−^, 147.0450 \[M−H−Ara−C~6~H~4~O~3~−OH\]^−^ Quercetin-3-*O*-*β*-Arabinopyranose HC \[[@B63-molecules-23-01525]\]
**43** 5.79 C~27~H~30~O~15~ 594.1588 594.1585 0.6 593.1515 \[M−H\]^−^, 285.0403 \[M−Glu−Rha\]^−^ Kaempferol 3-glucoside-7-rhamnoside HC, HD \[[@B64-molecules-23-01525]\]
**44 \*** 5.89 C~28~H~32~O~16~ 624.1682 624.1690 −1.3 625.1755 \[M+H\]^+^, 501.1583 \[M−C~6~H~4~O~2~\]^+^, 479.1155 \[M−Rha\]^+^, 465.0997 \[M−Rha−CH~3~\]^+^, 317.0637 \[M−Rha−Glu\]^+^ Isorhamnetin-3-rutinoside HC, HD\ \[[@B65-molecules-23-01525]\]
(HD\>\>HC)
**45 \#** 5.93 C~20~H~12~O~8~ 380.0560 380.0532 6.5 425.0542 \[M+HCOO\]^−^,163.0399 \[M−CO−C~11~H~5~O~5~\]^−^ Phelligrindins [d]{.smallcaps}-9 HC, HD\ \[[@B66-molecules-23-01525]\]
(HC\>\>HD)
**46 \*** 6.10 C~26~H~32~O~11~ 520.1941 520.1945 −0.6 565.1923 \[M+HCOO\]^−^, 501.1766 \[M−OH\]^−^, 489.1748 \[M−CH~2~OH\]^−^, 339.1233 \[M−Glu\]^−^ Matairesinol monoglucoside HC, HD\ \[[@B67-molecules-23-01525]\]
(HD\>\>HC)
**47 \#** 6.17 C~36~H~36~O~19~ 772.1864 772.1851 1.7 771.1791 \[M−H\]^−^, 565.1548 \[M−*p*-Hydroxy-cinnamic acid−CH~2~OH\]^−^ Allivictoside G HC \[[@B62-molecules-23-01525]\]
**48** 6.38 C~8~H~14~O~2~ 187.1049 187.0977 0.4 187.0977 \[M−H\]^−^, 169.0871 \[M−OH\]^−^, 125.0973 \[M−OH−COOH\]^−^, 97.0663 \[M−OH−C~3~H~5~O~2~\]^−^ Azelaic acid HC, HD \[[@B68-molecules-23-01525]\]
**49 \*** 6.52 C~27~H~32~O~15~ 596.1749 596.1741 1.3 595.1676 \[M−H\]^−^, 549.1621 \[M−OH−CH~2~OH\]^−^, 387.1073 \[M−OH−CH~2~OH−Rha\]^−^, 369.0977 \[M−2 × OH−CH~2~OH−Rha\]^−^, 163.0400 \[M−C~18~H~24~O~12~\]^−^ Neoeriocitrin HC, HD\ \[[@B69-molecules-23-01525]\]
(HD\>\>HC)
**50 \*** 6.82 C~27~H~32~O~14~ 580.1808 580.1792 2.5 625.1790 \[M+HCOO\]^−^, 529.1359 \[M−OH−OCH~3~\]^−^, 517.1356 \[M−OCH~3~−CH~2~OH\]^−^, 417.1204 \[M−Glu\]^−^, 193.0510 \[M−C~17~H~23~O~10~\]^−^, 147.0449 \[M−OCH~3~−Glu−C~10~H~9~O~4~\]^−^ 6-*O*-*Z*-*p*-feruloyl scandoside methyl ester HD \[[@B70-molecules-23-01525]\]
**51** 6.89 C~26~H~30~O~13~ 550.1683 550.1686 −0.6 595.1665 \[M+HCOO\]^−^, 433.14811 \[M−OH−C~4~H~4~O~3~\]^−^, 403.13121 \[M−C~9~H~7~O~2~\]^−^, 387.1093 \[M−Glu\]^−^, 355.0823 \[M−Glu−OCH~3~\]^−^ 10-*O*-*E*-*p*-courmaroyl scandoside methyl ester HC, HD \[[@B71-molecules-23-01525]\]
**52 \*** 7.07 C~26~H~30~O~13~ 550.1683 550.1686 −0.6 549.1610 \[M−H\]^−^, 595.1663 \[M+HCOO\]^−^, 387.1086 \[M−Glu\]^−^, 370.0789 \[M−Glu−CH~3~\]^−^, 193.0503 \[M−Glu−OCH~3~−C~9~H~7~O~3~\]^−^ 6-*O*-*p*-coumaroyl scandoside methyl ester HC, HD\ \[[@B16-molecules-23-01525]\]
(HD\>\>HC)
**53** 7.13 C~27~H~32~O~13~ 564.1843 564.1843 0 609.1825 \[M+HCOO\]^−^, 549.1613 \[M−CH~3~\]^−^, 387.1086 \[M−CH~3~−Glu\]^−^, 387.1086 \[M−CH~3~−C~10~H~9~O~2~\]^−^, 370.0789 \[M−2 × CH~3~−C~10~H~9~O~2~\]^−^, 337.1070 \[M−OCH~3~−OH−Glu\]^−^ 6-*O*-(*E*)-*p*-coumaroyl scandoside methyl ester-10-methyl ester HC, HD \[[@B72-molecules-23-01525]\]
**54 \*** 7.18 C~27~H~32~O~14~ 580.1800 580.1792 1.3 579.1727 \[M−H\]^−^, 399.1051 \[M−Glu\]^−^, 223.0604 \[M−Glu−C~10~H~12~O~4~\]^−^ Nobiletin-3-*O*-*β*-[d]{.smallcaps}-glucoside HD \[[@B73-molecules-23-01525]\]
**55 \#** 7.58 C~24~H~28~O~12~ 508.1581 508.1581 0.1 553.1563 \[M+HCOO\]^−^, 345.0977 \[M−Glu\]^−^, 223.0602 \[M−Glu−C~7~H~4~O~2~\]^+^ Hedycoryside B HC \[[@B74-molecules-23-01525]\]
**56** 7.88 C~15~H~10~O~7~ 302.0438 302.0426 3.9 303.0511 \[M+H\]^+^, 287.0549 \[M−OH\]^+^, 153.0181 \[M−C~8~H~6~O~3~\]^+^, 152.0565 \[M−C~7~H~4~O~4~\]^+^ Quercetin HC, HD s
**57 \*** 7.91 C~15~H~10~O~6~ 286.0489 286.0477 −3.7 287.0540 \[M+H\]^+^,163.0361 \[M−C~6~H~4~O~3~\]^+^,149.0589 \[M−C~6~H~4~O~3~−OH\]^+^, 131.0487 \[M−C~6~H~4~O~3~−2 × OH\]^+^ Kaempferol HC, HD\ s
(HD\>\>HC)
**58** 7.94 C~23~H~26~O~11~ 478.1471 478.1475 −0.9 477.1398 \[M−H\]^−^, 355.1035 \[M−benzoic acid\]^−^, 315.0879 \[M−Glu\]^−^, 285.0406 \[M−Glu−C~2~H~3~\]^−^, 241.1076 \[M−OH−benzoic acid−C~3~H~2~O~3~\]^−^ Hedycoryside C HC, HD \[[@B13-molecules-23-01525]\]
**59** 8.12 C~24~H~28~O~12~ 508.1585 508.1581 0.7 553.1567 \[M+HCOO\]^−^, 345.0976 \[M−Glu\]^−^, 207.0655 \[M−Glu−benzoic acid\]^−^, 137.0245 \[M−Glu−benzoic acid−C~4~H~4~O~2~\]^−^ 10-*O*-benzoyl scandoside methyl ester HC, HD \[[@B43-molecules-23-01525]\]
**60** 8.88 C~14~H~12~O~4~ 244.0738 244.0736 0.9 245.0811 \[M+H\]^+^, 227.0693 \[M−OH\]^+^, 135.0429 \[M−C~6~H~3~O~2~\]^+^, 119.0493 \[M−C~6~H~3~O~2~−OH\]^+^, 95.0512 \[M−C~8~H~5~O~2~\]^+^ Piceatannol HC, HD \[[@B75-molecules-23-01525]\]
**61** 9.26 C~17~H~14~O~7~ 330.0750 330.0740 3.2 331.0823 \[M+H\]^+^, 315.0485 \[M−CH~3~\]^+^, 301.0679 \[M−OCH~3~\]^+^, 207.0647 \[M−OH−C~6~H~5~O~2~\]^+^ 5,3′,4′-Trihydroxy-6,7-dimethoxy flavonoids HC, HD \[[@B76-molecules-23-01525]\]
**62 \*** 9.23 C~28~H~34~O~15~ 610.1909 610.1898 1.8 609.1836 \[M−H\]^−^, 401.1232 \[M−OH−OCH~3~−Rha\]^−^, 193.0513 \[M−2 × Rha−C~6~H~3~O\]^−^, 177.0557 \[M−2 × Rha−C~6~H~3~O~2~\]^−^ Hesperidin HC, HD\ s
(HD\>\>HC)
**63** 9.62 C~20~H~30~O~5~ 350.2078 350.2093 −4.2 351.2151 \[M+H\]^+^, 293.2123 \[M−C~2~H~4~O~2~\]^+^, 275.1999 \[M−C~2~H~4~O~2~−OH\]^+^, 257.1917 \[M−C~2~H~4~O~2~−OH−OH\]^+^, 105.0713 \[M−C~6~H~6~O~3~−C~6~H~12~O~2~\]^+^ 14-Andrographolide HC, HD \[[@B77-molecules-23-01525]\]
**64** 9.62 C~16~H~28~O~2~ 252.2113 252.2089 8.7 275.2006 \[M+Na\]^+^, 195.1389 \[M−C~4~H~8~\]^+^, 155.1050 \[M−C~7~H~14~\]^+^, 151.1110 \[M−C~6~H~12~O\]^+^ 7-Hexadecenoic acid-16-hydroxy-*O*-lactone HC, HD a
**65** 9.98 C~11~H~10~O~3~ 190.0625 190.0630 −2.8 191.0697 \[M+H\]^+^, 177.0533 \[M−CH~3~\]^+^, 159.0427 \[M−CH~3~−OH\]^+^, 105.0348 \[M−C~6~H~5~O\]^+^ 6-Methoxy-8-methyl coumarin HC, HD s
**66 \#** 10.00 C~24~H~28~O~11~ 492.1634 492.1632 −0.4 537.1616 \[M+HCOO\]^−^, 329.1028 \[M−Glu\]^−^, 207.0622 \[M−Glu−C~7~H~5~O\]^−^, 195.0664 \[M−Glu−OCH~3~−C~7~H~5~O\]^−^, 163.0397 \[M−Glu−OCH~3~−C~8~H~7~O~2~\]^−^ Hedycoryside A HC, HD\ \[[@B13-molecules-23-01525]\]
(HC\>\>HD)
**67** 10.08 C~11~H~16~O~2~ 180.1145 180.1150 −2.9 181.1218 \[M+H\]^+^, 163.1114 \[M−O\]^+^, 121.1022 \[M−C~2~HO~2~\]^+^ 5,6,7,7*α*-Tetrahydro-4,4,7*α*-trimethyl-2(4*H*)-benzofuranone HC, HD \[[@B77-molecules-23-01525]\]
**68** 10.31 C~15~H~10~O~4~ 254.0589 254.0579 3.7 255.0661 \[M+H\]^+^, 240.0411 \[M−CH~3~\]^+^, 224.0466 \[M−OCH~3~\]^+^ Alizarin 1-methyl ether HC, HD s
**69 \*** 10.64 C~15~H~10~O~4~ 254.0579 254.0579 0 253.0506 \[M−H\]^−^, 224.0477 \[M−CH~2~OH\]^−^ 1,3-Dihydroxy-2-methylanthraquinone HD \[[@B78-molecules-23-01525]\]
**70 \*** 11.03 C~15~H~8~O~4~ 252.0424 252.0423 0.7 251.0352 \[M−H\]^−^, 223.0399 \[M−O−CH\]^−^, 207.0449 \[M−COO\]^−^ Sanlengdiphenyllactone HC, HD\ s
(HD\>\>HC)
**71 \#** 11.65 C~21~H~22~O~8~ 402.1311 402.1315 −0.8 403.1384 \[M+H\]^+^, 387.1084 \[M−CH~3~\]^+^, 373.0905 \[M−2 × CH~3~\]^+^, 359.1092 \[M−CH~3~−OCH~3~\]^+^ Chuan Nectein HC \[[@B79-molecules-23-01525]\]
**72** 11.82 C~15~H~16~O~4~ 260.1065 260.1049 6.3 261.1138 \[M+H\]^+^, 205.0499 \[M-C~4~H~7~\]^+^, 190.0262 \[M-C~5~H~9~\]^+^, 177.0543 \[M-C~5~H~9~O\]^+^, 162.0316 \[M-OCH~3~-C~5~H~9~\]^+^ 5-Prenyloxy-7-methoxycoumarin HC, HD a
**73** 11.85 C~20~H~26~O~4~ 330.1805 330.1831 −7.8 331.1878 \[M+H\]^+^, 149.0953 \[M-OH-C~10~H~13~O~2~\]^+^, 131.0489 \[M-OH-CH~3~-C~10~H~13~O~2~\]^+^, 135.0803 \[M-OCH~3~-C~10~H~12~O~2~\]^+^, 121.0646 \[M-OCH~3~-C~12~H~17~O~2~\]^+^ Dihydroguaiac acid HC, HD \[[@B80-molecules-23-01525]\]
**74** 11.87 C~15~H~14~O~4~ 258.0889 258.0892 −1.3 259.0962 \[M+H\]^+^, 244.0707 \[M-CH~3~\]^+^, 229.0480 \[M-2 × CH~3~\]^+^, 227.0684 \[M-OCH~3~\]^+^, 217.0474 \[M-C~3~H~5~\]^+^, 212.0444 \[M-CH~3~-OCH~3~\]^+^ Hedyotiscone A HC, HD \[[@B81-molecules-23-01525]\]
**75 \#** 12.11 C~19~H~18~O~6~ 342.1103 342.1103 −0.2 343.1175 \[M+H\]^+^, 327.08434 \[M−CH~3~\]^+^, 313.06864 \[M−2 × CH~3~\]^+^, 299.08954 \[M−CH~3~−OCH~3~\]^+^, 285.07454 \[M−2 × CH~3~−OCH~3~\]^+^ 5,6,7,4′-Tetramethoxyflavone HC s
**76** 12.11 C~16~H~28~O~3~ 268.2056 268.2038 6.2 291.1949 \[M+Na\]^+^, 217.1566 \[M−CH~3~−OH\]^+^, 132.0863 \[M−OH−C~2~H~5~−C~4~H~7~O~2~\]^+^ 13-Hydroxy-9,11-Hexadecandienoic acid HC, HD b
**77 \*** 12.41 C~16~H~12~O~4~ 286.0731 268.0736 −0.4 269.0804 \[M+H\]^+^, 254.0557 \[M−CH~3~\]^+^, 251.06537 \[M−OH\]^+^, 239.0689 \[M−OCH~3~\]^+^, 225.0540 \[M−OCH~3~−CH~3~\]^+^ Methylisotropine-1-methylether HC, HD\ a
(HD\>\>HC)
**78 \*** 12.44 C~15~H~10~O~3~ 238.0630 238.0630 0.2 237.0558 \[M−H\]^−^, 224.0471 \[M−CH~3~\]^−^, 208.0518 \[M−OH−CH~3~\]^−^ 2-Hydroxy-3-methylanthraquinone HC, HD\ \[[@B82-molecules-23-01525]\]
(HD\>\>HC)
**79 \*** 12.49 C~15~H~10~O~5~ 270.0524 270.0528 −1.6 269.0451 \[M−H\]^−^, 237.0555 \[M−2 × OH\]^−^ 5-Dehydroxykaempferol HC, HD\ \[[@B83-molecules-23-01525]\]
(HD\>\>HC)
**80** 12.51 C~17~H~24~O~3~ 276.1730 276.1725 1.8 277.1803 \[M+H\]^+^, 259.1608 \[M−OH\]^+^, 231.1774 \[M−CH~3~−2 × OH\]^+^, 213.1633 \[M−CH~3~−3 × OH\]^+^, 203.1776 \[M−3 × OH−C~2~H~3~\]^+^, 201.1612 \[M−3 × OH−C~2~H~5~\]^+^ (10*E*)1,10-Heptadeca-diene-4,6-diyne-3,8,9-triol HC, HD \[[@B84-molecules-23-01525]\]
**81 \#** 12.74 C~22~H~24~O~9~ 432.1411 432.1420 −2.0 433.1484 \[M+H\]^+^, 418.1231 \[M−CH~3~\]^+^, 403.0998 \[M−2 × CH~3~\]^+^, 388.0763 \[M−3 × CH~3~\]^+^, 385.0857 \[M−CH~3~−OCH~3~\]^+^, 372.1131 \[M−2 × OCH~3~\]^+^, 357.0934 \[M−CH~3~−2 × OCH~3~\]^+^ 3′,4′,5′,5,6,7,8-Seven-methoxyflavone HC \[[@B85-molecules-23-01525]\]
**82** 12.80 C~17~H~24~O~2~ 260.1774 260.1776 −0.8 305.1756 \[M+HCOO\]^−^, 135.0813 \[M−C~3~H~7~−C~5~H~5~O\]^−^, 125.0969 \[M−C~2~H~5~−C~7~H~5~O\]^−^, 121.0656 \[M−C~4~H~9~−C~5~H~5~O\]^−^ Fakalinediol HC, HD \[[@B86-molecules-23-01525]\]
**83** 13.35 C~30~H~48~O~5~ 488.3497 488.3502 −0.9 533.3479 \[M+HCOO\]^−^, 291.1956 \[M−C~12~H~20~O~2~\]^−^, 195.1029 \[M−C~19~H~29~O~2~\]^−^, 171.1025 \[M−C~21~H~33~O~2~\]^−^ 3*β*,19*α*,23-Trihydroxyurs-12-en-28-oic acid HC, HD \[[@B87-molecules-23-01525]\]
**84 \*** 13.36 C~17~H~14~O~6~ 314.0793 314.0790 0.8 315.0866 \[M+H\]^+^, 300.0618 \[M−CH~3~\]^+^, 282.04958 \[M−OCH~3~\]^+^, 111.04458 \[M−CH~3~−C~10~H~6~O~4~\]^+^ 5,3′-Dihydroxy-7,4′-dimethoxyflavone HD \[[@B88-molecules-23-01525]\]
**85** 13.96 C~27~H~28~N~2~O~4~ 444.2060 444.2049 2.5 445.2133 \[M+H\]^+^, 385.1887 \[M−C~2~H~3~O~2~\]^+^, 224.1062 \[M−C~12~H~13~NO~3~\]^+^, 194.1172 \[M−C~16~H~13~NO~2~\]^+^, 134.0970 \[M−C~2~H~3~O~2~−C~16~H~13~NO~2~\]^+^ Gold Amide Alcohol Ester HC, HD \[[@B89-molecules-23-01525]\]
**86** 14.44 C~17~H~32~O~2~ 268.2398 268.2402 −1.4 313.2380 \[M+HCOO\]^−^, 251.2019 \[M−CH~3~\]^−^, 183.1388 \[M−C~6~H~13~\]^−^, 129.0918 \[M−C~10~H~17~\]^−^ Methyl cis-9-hexadecenoate HC, HD a
**87** 14.81 C~18~H~34~O~4~ 314.2460 314.2457 0.9 337.2352 \[M+Na\]^+^, 139.1118 \[M−C~9~H~18~O~3~\]^+^, 125.09614 \[M−C~10~H~20~O~3~\]^+^ Dibutyl sebacate HC, HD a
**88** 14.81 C~18~H~30~O~2~ 278.2244 278.2246 −0.8 279.2316 \[M+H\]^+^, 249.1834 \[M−C~2~H~5~\]^+^, 217.1935 \[M−CH~3~−COO\]^+^, 191.1801 \[M−C~4~H~6~O~2~\]^+^, 163.1483 \[M−C~6~H~10~O~2~\]^+^ 9,12,15-Octadecatrienoic acid HC, HD \[[@B90-molecules-23-01525]\]
**89 \#** 15.25 C~26~H~32~O~6~ 440.2193 440.2199 −1.4 441.2266 \[M+H\]^+^, 389.2315 \[M−C~3~H~2~O\]^+^, 340.1657 \[M−C~3~H~2~O−C~2~H~3~O\]^+^, 147.0437 \[M−C~17~H~25~O~4~\]^+^ Isofeterin HC \[[@B91-molecules-23-01525]\]
**90 \#** 15.25 C~20~H~28~O~4~ 332.2016 332.1988 7.9 355.1908 \[M+Na\]^+^, 241.1946 \[M−OH−CH~2~OH−COO\]^+^, 217.1189 \[M−OH−CH~2~OH−CH~3~−C~4~H~8~\]^+^, 161.1320 \[M−OH−CH~2~OH−CH~3~−C~6~H~5~O~2~\]^+^ 14-Deoxy-11,12-dihydroandrographolide HC \[[@B92-molecules-23-01525]\]
**91 \#** 15.69 C~15~H~22~O 218.1659 218.1671 −5.5 219.1731 \[M+H\]^+^, 163.1106 \[M−C~4~H~7~\]^+^, 161.0935 \[M−CH~3~−C~3~H~6~\]^+^ *α*-Turmerone HC a
**92 \#** 15.87 C~15~H~28~O~2~ 240.2090 240.2089 285.2072 \[M+HCOO\]^−^, 223.2068 \[M−OH\]^−^ Isodonsesquitin A HC, HD\ \[[@B93-molecules-23-01525]\]
(HC\>\>HD)
**93** 16.02 C~16~H~30~O~2~ 254.2251 254.2246 1.8 277.2143 \[M+Na\]^+^, 137.1316 \[M−C~4~H~9~−CH~2~COOH\]^+^, 109.1012 \[M−C~4~H~9~−C~3~H~6~COOH\]^+^ *Z*-11-Hexadecenoic acid HC, HD \[[@B94-molecules-23-01525]\]
**94** 16.02 C~20~H~28~O~3~ 316.2025 316.2038 −4.1 317.2098 \[M+H\]^+^, 289.1787 \[M−C~2~H~4~\]^+^, 277.2151 \[M−C~2~H~2~O\]^+^, 251.1930 \[M−C~4~H~4~O\]^+^, 235.1667 \[M−C~5~H~7~O\]^+^, 221.1503 \[M−CH~3~−C~5~H~7~O\]^+^ 7*β*-Senecioyloxyoplopa-3(14)*Z*,8(10)-dien-2-one HC, HD a
**95** 16.03 C~34~H~58~O~4~ 530.4316 530.4335 −3.4 553.4208 \[M+Na\]^+^, 483.3400 \[M−3CH~3~\]^+^, 317.2060 \[M−OCH~3~−C~13~H~27~\]^+^, 315.1595 \[M−2 × CH~3~−C~13~H~27~\]^+^, 313.1703 \[M−OH−CH~3~−C~13~H~27~\]^+^ Ferulic acid esters lignoceric HC, HD a
**96** 16.23 C~16~H~30~O~2~ 254.2258 254.2246 4.6 277.2151 \[M+Na\]^+^, 137.1329 \[M−C~2~H~5~−C~4~H~6~O~2~\]^+^, 123.1168 \[M−C~2~H~5~−C~5~H~8~O~2~\]^+^, 111.1171 \[M−C~8~H~14~O~2~\]^+^ Palmitoleic acid HC, HD \[[@B95-molecules-23-01525]\]
**97** 16.23 C~20~H~28~O~3~ 316.2021 316.2038 −5.5 317.2094 \[M+H\]^+^, 301.2068 \[M−OH\]^+^, 277.2147 \[M−C~2~H~2~O\]^+^, 259.2029 \[M−CH~3~−COOH\]^+^, 215.1763 \[M−C~2~H~2~O−COOH\]^+^, 141.0911 \[M−C~11~H~15~\]^+^ Terminalic acid HC, HD \[[@B96-molecules-23-01525]\]
**98** 16.61 C~20~H~26~O~3~ 314.1859 314.1882 −7.2 315.1932 \[M+H\]^+^, 159.1158 \[M−OH−C~8~H~9~O~2~\]^+^, 133.1005 \[M−C~10~H~13~O~3~\]^+^, 147.1165 \[M−OH−C~9~H~11~O~2~\]^+^ Oxyphyllacinol HC, HD \[[@B97-molecules-23-01525]\]
**99** 16.91 C~20~H~26~O~3~ 314.1854 314.1882 −8.9 315.1927 \[M+H\]^+^, 191.1040 \[M−OH−C~8~H~9~\]^+^, 173.1307 \[M−OH−C~7~H~7~O~2~\]^+^, 135.0799 \[M−OH−OCH~3~−C~10~H~13~\]^+^ Neonootkatol HC, HD \[[@B98-molecules-23-01525]\]
**100** 17.08 C~17~H~30~O~2~ 266.2646 266.2646 0.1 311.2228 \[M+HCOO\]^−^, 183.1387 \[M−C~6~H~12~\]^−^, 249.2224 \[M−OH\]^−^ 7,10-Dienylhexadecanoic acid methyl ester HC, HD a
**101** 17.37 C~18~H~32~O~3~ 296.2355 296.2351 1.3 295.2283 \[M−H\]^−^, 277.2176 \[M−OH\]^−^, 233.2262 \[M−O−COOH\]^−^, 183.1024 \[M−CH~3~−5×CH~2~−2×CH\]^−^,125.0968 \[M−OH−C~10~H~17~O\]^−^, 123.1180 \[M−O−CH~2~COOH−C~7~H~13~\]^−^ Coronaric acid HC, HD \[[@B99-molecules-23-01525]\]
**102** 17.37 C~18~H~32~O~3~ 296.2355 296.2351 1.3 295.2283 \[M−H\]^−^, 277.2176 \[M−OH\]^−^, 233.2262 \[M−O−COOH\]^−^, 125.0968 \[M−C~10~H~17~O~2~\]^−^, 123.1180 \[M−COOH−C~8~H~15~O\]^−^ Vernonia acid HC, HD \[[@B100-molecules-23-01525]\]
**103** 17.39 C~30~H~46~O~4~ 470.3398 470.3396 0.3 471.347 \[M+H\]^+^, 455.3448 \[M−OH\]^+^, 437.3382 \[M−2 × OH\]^+^, 425.3421 \[M−COO\]^+^, 420.2712 \[M−2 × CH~3~−OH\]^+^, 409.3449 \[M−OH−COO\]^+^, 383.3309 \[M−CH~3~−CO−COO\]^+^ Caryophylloside HC, HD \[[@B101-molecules-23-01525]\]
**104** 17.87 C~20~H~28~O~3~ 316.2019 316.2038 −6.2 317.2092 \[M+H\]^+^, 235.1672 \[M−C~5~H~6~O\]^+^, 189.1622 \[M−C~5~H~6~O−COOH\]^+^, 179.1418 \[M−OH−CH~3~−C~7~H~8~O\]^+^ Saurufuran B HC, HD \[[@B102-molecules-23-01525]\]
**105** 17.87 C~18~H~28~O~2~ 276.2088 276.2089 −0.3 277.2161 \[M+H\]^+^, 235.1672 \[M−C~3~H~6~\]^+^, 217.1967 \[M−CH~2~COOH\]^+^, 207.1729 \[M−OH−C~3~H~6~\]^+^, 189.1623 \[M−C~3~H~6~COOH\]^+^ Stearidonic acid HC, HD \[[@B103-molecules-23-01525]\]
**106** 15.99 C~18~H~30~O~3~ 294.2197 294.2195 0.7 293.2124 \[M−H\]^−^, 275.2016 \[M−OH\]^−^, 211.1340 \[M−C~6~H~12~\]^−^, 185.1180 \[M−C~8~H~14~\]^−^, 182.1305 \[M−OH−C~7~H~13~\]^−^ (*E*,*E*)-9-Oxooctadeca-10,12-dienoic acid HC, HD \[[@B104-molecules-23-01525]\]
**107** 18.61 C~20~H~28~O~3~ 316.2021 316.2038 −7.0 317.2089 \[M+H\]^+^, 283.1680 \[M−OH−CH~3~\]^+^, 259.2034 \[M−CH~3~−COOH\]^+^, 235.1680 \[M−C~5~H~5~O\]^+^ Saurufuran A HC, HD \[[@B103-molecules-23-01525]\]
**108** 18.61 C~16~H~30~O~2~ 254.2270 254.2246 8.7 277.2162 \[M+Na\]^+^, 179.1405 \[M−OH−C~4~H~9~\]^+^, 165.1260 \[M−OH−C~5~H~11~\]^+^, 151.1111 \[M−OH−C~6~H~13~\]^+^, 125.0963 \[M−OH−C~8~H~15~\]^+^ Hexadecenoic acid HC, HD \[[@B105-molecules-23-01525]\]
**109** 18.68 C~18~H~34~O~3~ 298.2511 298.2508 1.1 297.2438 \[M−H\]^−^, 279.2332 \[M−OH\]^−^, 155.1076 \[M−C~9~H~18~O\]^−^ Ricinolic acid HC, HD \[[@B106-molecules-23-01525]\]
**110** 19.51 C~20~H~30~O~3~ 318.2174 318.2195 −6.6 319.2247 \[M+H\]^+^, 239.1776 \[M−COOH−CH~2~OH\]^+^, 233.193 \[M−C~4~H~3~O~2~\]^+^, 189.1630 \[M−OH−C~6~H~7~O~2~\]^+^ Andrograpanin HC, HD \[[@B107-molecules-23-01525]\]
**111** 21.64 C~30~H~48~O~3~ 456.3579 456.3604 −4.8 501.3561 \[M+HCOO\]^−^, 340.2808 \[M−2 × OH−C~6~H~12~\]^−^, 277.2159 \[M−C~12~H~20~O\]^−^, 223.2062 \[M−COOH−C~14~H~19~\]^−^ Ursolic acid HC, HD s
**112** 21.72 C~28~H~48~O~2~ 416.3678 416.3654 5.3 439.357 \[M+Na\]^+^, 342.3004 \[M−OH−C~4~H~9~\]^+^, 327.2377 \[M−2CH~3~−C~4~H~9~\]^+^, 277.2119 \[M−C~10~H~21~\]^+^, 249.1820 \[M−CH~3~−C~11~H~23~\]^+^ *γ*-Tocopherol HC, HD \[[@B108-molecules-23-01525]\]
**113** 23.33 C~19~H~38~O~4~ 330.2776 330.2770 1.6 353.2668 \[M + Na\]^+^, 313.2733 \[M−OH\]^+^, 283.2593 \[M−2 × OH−CH~3~\]^+^, 269.2161 \[M−OH−C~3~H~7~\]^+^, 239.2376 \[M−C~3~H~5~O~3~\]^+^ Palmitin HC, HD a
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
\* Characteritic component in HD; \# Characteritic component in HC; s: Identified with reference substance. a: Compared with spectral data obtained from Wiley Subscription Services, Inc. (USA); b: Compared with NIST Chemistry WebBook; HD: *Hedyotis diffuse* Willd.; HC: *Hedyotis corymbosa* (L.) Lam.
| {
"pile_set_name": "PubMed Central"
} |
{
"pile_set_name": "PubMed Central"
} |
|
Tunçalp Ӧ, Were WM, MacLennan C, Oladapo OT, Gülmezoglu AM, Bahl R, Daelmans B, Mathai M, Say L, Kristensen F, Temmerman M, Bustreo F. Quality of care for pregnant women and newborns---the WHO vision. BJOG 2015;122:1045--1049.25929823
In any reproduction of this article there should not be any suggestion that WHO or the article endorse any specific organization or products. The use of the WHO logo is not permitted. This notice should be preserved along with the article\'s URL.
In 2015, as we review progress towards Millennium Development Goals (MDGs), despite significant progress in reduction of mortality, we still have unacceptably high numbers of maternal and newborn deaths globally. Efforts over the past decade to reduce adverse outcomes for pregnant women and newborns have been directed at increasing skilled birth attendance.[1](#bjo13451-bib-0001){ref-type="ref"}, [2](#bjo13451-bib-0002){ref-type="ref"} This has resulted in higher rates of births in health facilities in all regions.[3](#bjo13451-bib-0003){ref-type="ref"} The proportion of deliveries reportedly attended by skilled health personnel in developing countries rose from 56% in 1990 to 68% in 2012.[4](#bjo13451-bib-0004){ref-type="ref"} With increasing utilisation of health services, a higher proportion of avoidable maternal and perinatal mortality and morbidity have moved to health facilities. In this context, poor quality of care (QoC) in many facilities becomes a paramount roadblock in our quest to end preventable mortality and morbidity.
QoC during childbirth in health facilities reflects the available physical infrastructure, supplies, management, and human resources with the knowledge, skills and capacity to deal with pregnancy and childbirth---normal physiological, social and cultural processes, but prone to complications that may require prompt life‐saving interventions. Research shows that it is necessary to go beyond maximising coverage of essential interventions to accelerate reductions in maternal and perinatal mortality and severe morbidity.[5](#bjo13451-bib-0005){ref-type="ref"} Moreover, there is a complex interplay of experiences of mistreatment and lack of support that impact women\'s childbirth experiences and outcomes.[6](#bjo13451-bib-0006){ref-type="ref"}
Moving beyond 2015, the World Health Organization (WHO) envisions a world where 'every pregnant woman and newborn receives quality care throughout pregnancy, childbirth and the postnatal period.' This vision is in alignment with two complementary global action agendas conceptualised by WHO and partners in 2013--2014---\'Strategies toward Ending Preventable Maternal Mortality (EPMM)\' and 'Every Newborn Action Plan (ENAP)'.[7](#bjo13451-bib-0007){ref-type="ref"}, [8](#bjo13451-bib-0008){ref-type="ref"} It is articulated at a critical time when the global community is developing the new Global Strategy for Women\'s, Children\'s and Adolescents\' Health (2016--2030) for the post‐2015 Sustainable Development Goal era.[9](#bjo13451-bib-0009){ref-type="ref"}
Although indirect causes of maternal death are increasing (27.5% of maternal deaths), globally, over 70% of maternal deaths occur as a result of complications of pregnancy and childbirth such as haemorrhage, hypertensive disorders, sepsis and abortion.[10](#bjo13451-bib-0010){ref-type="ref"} Complications of preterm birth, birth asphyxia, intrapartum‐related neonatal death and neonatal infections together account for more than 85% of newborn mortality.[11](#bjo13451-bib-0011){ref-type="ref"} Therefore, the time of childbirth and the period immediately after birth are particularly critical for maternal, fetal and neonatal survival and well‐being. Effective care to prevent and manage complications during this critical period is likely to have a significant impact on reducing maternal deaths, stillbirths and early neonatal deaths---a triple return on investment.[12](#bjo13451-bib-0012){ref-type="ref"} Within this critical period, quality of care improvement efforts would target essential maternal and newborn care and additional care for management of complications that could achieve the highest impact on maternal, fetal and newborn survival and well‐being. Based on the current evidence on burden and impact, the following specific thematic areas have been identified as high priority for this vision:[10](#bjo13451-bib-0010){ref-type="ref"}, [11](#bjo13451-bib-0011){ref-type="ref"}, [12](#bjo13451-bib-0012){ref-type="ref"}
Essential childbirth care including labour monitoring and action and essential newborn care at birth and during the first week;Management of pre‐eclampsia, eclampsia and its complications;Management of postpartum haemorrhage;Management of difficult labour by enabling safe and appropriate use of medical technologies during childbirth;Newborn resuscitation;Management of preterm labour, birth and appropriate care for preterm and small babies;Management of maternal and newborn infections.
To end preventable maternal and newborn morbidity and mortality, every pregnant woman and newborn need skilled care at birth with evidence‐based practices delivered in a humane, supportive environment. Good quality of care requires appropriate use of effective clinical and non‐clinical interventions, strengthened health infrastructure and optimum skills and attitude of health providers, resulting in improved health outcomes and positive experience of women and providers. Moreover, quality of care is considered a key component of the right to health, and the route to equity and dignity for women and children.[13](#bjo13451-bib-0013){ref-type="ref"}
So, what is quality of care? To underpin this vision, we need a common understanding of what it means. This WHO vision defines quality of care as 'the extent to which health care services provided to individuals and patient populations improve desired health outcomes. In order to achieve this, health care needs to be safe, effective, timely, efficient, equitable, and people‐centred.'[14](#bjo13451-bib-0014){ref-type="ref"}, [15](#bjo13451-bib-0015){ref-type="ref"} Operational definitions for the characteristics of quality of care are defined in Box [1](#bjo13451-fea-0001){ref-type="boxed-text"}.
###### Operational definitions for the characteristics of QoC definition[14](#bjo13451-bib-0014){ref-type="ref"}, [15](#bjo13451-bib-0015){ref-type="ref"}
{#bjo13451-sec-1001}
*Safe---*delivering health care which minimises risks and harm to service users, including avoiding preventable injuries and reducing medical errors*Effective---*providing services based on scientific knowledge and evidence‐based guidelines*Timely---*reducing delays in providing/receiving health care*Efficient---*delivering health care in a manner which maximises resource use and avoids wastage*Equitable---*delivering health care which does not vary in quality because of personal characteristics such as gender, race, ethnicity, geographical location or socioeconomic status*People‐centred---*providing care which takes into account the preferences and aspirations of individual service users and the cultures of their communities
Quality of care is a multi‐dimensional concept. Therefore, a framework with important domains of measurement and pathways to achieve the desired health outcomes is required to identify the action points to improve the quality of care. Since the Donabedian model of quality of care for health facilities was proposed in 1988, WHO and others have developed strategic thinking to operationalise key characteristics of QoC, using different elements from the provision of care as well as the experience of care, integral to maternal and newborn care provided in the facilities.[15](#bjo13451-bib-0015){ref-type="ref"}, [16](#bjo13451-bib-0016){ref-type="ref"}, [17](#bjo13451-bib-0017){ref-type="ref"}, [18](#bjo13451-bib-0018){ref-type="ref"}, [19](#bjo13451-bib-0019){ref-type="ref"} WHO has also advanced health systems thinking by identifying six building blocks---service delivery; health workforce; information, medical products, vaccines and technologies; financing, and leadership/governance---creating a structure from where health systems analysis and intervention points can be established.[20](#bjo13451-bib-0020){ref-type="ref"}
Building on these developments, the framework (Figure [1](#bjo13451-fig-0001){ref-type="fig"}) conceptualises QoC for maternal and newborn health by identifying domains of QoC which should be targeted to assess, improve and monitor care within the context of the health system as the foundation. Health systems create the structure which enables access to quality care and allows for the process of care to occur along two important and inter‐linked dimensions of provision and experience of care.
{#bjo13451-fig-0001}
Based on this framework, QoC for pregnant women and newborns in facilities requires competent and motivated human resources and the availability of essential physical resources. Also, evidence‐based practices for routine and emergency care, actionable information systems where record keeping enables review and audit mechanisms, and functional referral systems between levels of care should be in place. Experience of care includes firstly effective communication---a woman (or her family if required) should feel that she understands what is happening, what to expect and knows her rights. Secondly, she should receive care with respect and dignity. Thirdly, she should have access to the social and emotional support of her choice.
Improved QoC increases the likelihood of desired individual and facility‐level outcomes---health outcomes, coverage of key practices and people‐centred outcomes---with a focus on the identified high priority thematic areas described above. Although our framework focuses on the care provided in the facilities, it should be noted that communities and service users have a critical role in identifying their own needs and preferences, and in managing their own health. Perspectives of women, their families and communities, on the quality of maternity care services influence decisions to seek care and are essential components for creating a demand for and access to quality maternal and newborn services.[6](#bjo13451-bib-0006){ref-type="ref"} Community engagement, therefore, is an important aspect to be considered.
A number of strategies that guide implementation efforts to improve QoC have been proposed. Many of these primarily focus on adapting interventions and working to overcome barriers to adaptation and implementation.[21](#bjo13451-bib-0021){ref-type="ref"} However, these strategies do not always address the fundamental issue of achieving a balance between conformity to the evidence‐based practices and accommodating contextual differences, which underlies successful implementation. Moreover, without the appropriate tools and materials available in a user‐friendly format, health systems are less likely to implement an evidence‐based intervention, and, even if implemented, they may be suboptimal.
In this vision, WHO will use a QoC improvement strategy, an adaptation of the 'Plan‐Do‐Study Act' (PDSA) cycle model[22](#bjo13451-bib-0022){ref-type="ref"} based on evidence synthesis, best practice and experience. This strategy provides a roadmap for continuous quality improvement. It starts by setting aims and building teams to achieve desired outcomes through implementation of evidence‐based change packages (individual, multi‐faceted and/or complex interventions depending on the context and the needs). It also incorporates capacity strengthening and other strategies to maximise the chances for sustaining the implementation.[23](#bjo13451-bib-0023){ref-type="ref"}, [24](#bjo13451-bib-0024){ref-type="ref"} In this context, quality improvement should achieve the standards set for both provision and experience of care.
Consolidating the framework and the improvement strategy described above, WHO will develop a comprehensive approach to provide guidance to global and national stakeholders to realise this vision. Figure [2](#bjo13451-fig-0002){ref-type="fig"} depicts how the WHO approach consolidates the QoC framework and improvement strategy, and highlights the identified strategic areas.
{#bjo13451-fig-0002}
In line with its organisational mandate (research, norms and standards, support for implementation, monitoring and evaluation),[25](#bjo13451-bib-0025){ref-type="ref"} six strategic areas have been identified for WHO to contribute to ending preventable mortality and morbidity among mothers and newborns. The QoC definition and framework will inform this evidence‐based and systematic approach to (1) research, (2) guideline development, (3) standards of care, (4) identification of effective intervention strategies for quality improvement, (5) development of monitoring indicators at global, national and facility levels, and (6) capacity strengthening for quality improvement research, measurement and programming. Work in these strategic areas will support the maternal and newborn QoC improvement strategy and ensure implementation based on robust data, while including targeted country‐level capacity strengthening and technical support.
Given the progress made in MDG‐4 and MDG‐5 in the past 15 years, with increases in coverage of skilled attendance and essential intervention, the next phase should, in addition, target multiple domains of quality of care to reduce further the burden of preventable mortality and morbidity, integrated as part of the Global Strategy for Women\'s, Children\'s and Adolescents\' Health.
Disclosure of interests {#bjo13451-sec-0003}
=======================
None declared. Completed disclosure of interests form available to view online as supporting information.
Contribution of authorship {#bjo13451-sec-0004}
==========================
The idea of this commentary was conceived by ÖT, AMG, MT and RB. ÖT, AMG, RB, WW, CM, OO, BD, MM, LS, FK, MT and FB all contributed the content and development of the article. All authors reviewed and agreed to the final version of this manuscript. All of the co‐authors are staff members at the World Health Organization.
Details of ethics approval {#bjo13451-sec-0005}
==========================
No ethics approval required.
Funding {#bjo13451-sec-0006}
=======
None.
Supporting information
======================
######
######
Click here for additional data file.
######
######
Click here for additional data file.
######
######
Click here for additional data file.
######
######
Click here for additional data file.
######
######
Click here for additional data file.
######
######
Click here for additional data file.
######
######
Click here for additional data file.
######
######
Click here for additional data file.
######
######
Click here for additional data file.
######
######
Click here for additional data file.
######
######
Click here for additional data file.
######
######
Click here for additional data file.
We acknowledge the valuable contributions of Every Newborn Action Plan (ENAP) and Ending Preventable Maternal Mortality (EPMM) Steering Teams and Advisory Groups to our work on quality of maternal and newborn care and on the revitalisation of the global attention. The Bill and Melinda Gates Foundation made funds available to WHO for advancing work on quality of maternal and newborn care.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Background
----------
Hypertension, diabetes, and hypercholesterolemia are the global leading risks of cardiovascular mortality \[[@ref1]-[@ref3]\]. Health behaviors such as engaging in exercise, balanced diet, and weight control reduce one's risk of cardiovascular mortality. Therefore, in addition to medication, the management of health behavior is crucial in patients with multiple risks of cardiovascular mortality \[[@ref4]\]. Clinical guidelines recommend a combined self-management strategy of health behaviors and appropriate medication use \[[@ref3],[@ref5]\] A recent self-management approach in line with the Chronic Care Model (CCM) specifies that health behavior management should be used to manage coexisting illnesses \[[@ref6]-[@ref8]\]. Owing to the importance of self-management in patient-centered health care in combination with the increased use of mobile devices (including smartphones and tablets), there is a need to develop an efficient, affordable, and sustainable self-management strategy-based elecronig program that targets high-risk individuals \[[@ref9]-[@ref11]\].
Research concerning mobile health (mHealth) innovations to support populations with chronic illnesses and improve their health behaviors is growing \[[@ref4],[@ref9],[@ref11]\]. A systematic review showed that the use of apps in mHealth has the potential to improve health outcomes among patients with chronic diseases through enhanced self-management \[[@ref12]\]. A number of randomized controlled trials (RCTs) have assessed the effectiveness of mobile phone- or tablet-assisted self-management programs in addressing cardiovascular disease \[[@ref13]\] or chronic hepatitis \[[@ref14]\]. Although there is a need to organize intervention programs to improve the health outcomes of patients with chronic illnesses \[[@ref6]\], few mHealth trials have addressed this \[[@ref3],[@ref5]\].
Objectives
----------
We therefore aimed to determine the efficacy of a self-management strategy-based electronic program for patients who had been treated for hypertension, diabetes, or hypercholesterolemia and who had at least one indicator of poor disease control. To do so, we provided patients with an intervention program via a Web-based health management program (mobile app or PC-Web-based) \[[@ref15]\].
Methods
=======
Study Design
------------
We conducted this study with 106 patients within 2 months of treatment termination, and the patients were randomly assigned to either the control group or the intervention group (ie, *Smart Healthing*; [Multimedia Appendices 1](#app1){ref-type="supplementary-material"} and [2](#app2){ref-type="supplementary-material"}). Each physician from 2 study hospitals screened patients for the eligibility criteria by reviewing their medical records and blood test results at outpatient clinics. A clinical research coordinator at each hospital explained the study details to the participants who met the eligibility criteria ([Figure 1](#figure1){ref-type="fig"}). The patients who were eligible to participate were recruited by the physician in-charge and were asked to provide written informed consent to the researchers. The institutional review boards at the 2 hospitals approved the study protocol (numbers 1707-084-870 and B-1802/453-401). The trial was performed in accordance with the Good Clinical Practice guidelines and the Declaration of Helsinki. All staff who were involved in screening and recruiting participants were certified by their institutions for ethical conduct of research (Collaborative Institutional Training Initiatives).
{#figure1}
Participants
------------
From November 2017 to March 2018, we identified patients with at least one indicator of poor disease control among patients who had been treated for hypertension, diabetes, or hypercholesterolemia. We recruited patients who met the following criteria: (1) aged ≥19 years; (2) diagnosed with hypertension, diabetes, or hypercholesterolemia; (3) failed to meet 1 or more of the following clinical goals: (i) glycated hemoglobin (HbA~1c~) \<7.0%, (ii) systolic blood pressure (SBP) \<140 mmHg, or (iii) low-density lipoprotein (LDL) cholesterol \<130 mg/dL; (4) had a smartphone and personal computer (for the electronic program-based health care program); and (5) understood the study's purpose.
Patients were excluded from the study if they met any of the following criteria: (1) had medical conditions that would limit participation adherence (as confirmed by their referring physician \[eg, dyspnea and severe depression\]); (2) could not speak, understand, or write Korean; or (3) could not understand the content of the provided materials owing to poor eyesight and/or hearing.
Randomization
-------------
We used an internet-based Clinical Research and Trial management system by the Centers for Disease Control and Prevention for participant randomization. The patients were randomly assigned (1:1) to the intervention or control group based on a random computer-generated number. To minimize the effects of potential confounding variables, we randomized participants stratified by disease type with the clinical indicators (hypertension, diabetes, or hypercholesterolemia). The research assistants executed face-to-face procedures and therefore could not be blinded when assigning participants to groups.
Control
-------
The attention control group was encouraged to continue their usual care and routine medications and to study a health educational booklet about chronic diseases. The booklet noted 12 healthy life habits: positive thinking, regular exercise, balanced diet, proactive living, regular checks-ups, helping others, regular religious life, quitting smoking, drinking cessation, work-life balance, living with loved ones, and taking medication.
Intervention
------------
The intervention group received the self-management strategy-based electronic program, whereas the control group received basic educational material about disease content. We developed the Smart Management Strategy for Health--based electronic program and utilized the comprehensive and multifaceted Smart Management Strategy for Health strategies. The self-management strategy-based electronic program used in this study was a 3-month Smart Management Strategy for Health Intervention, and it is comprised of an app and a Web-based program. On the basis of our literature review and interviews, we developed a conceptual framework for the Smart Management Strategy for Health intervention that incorporates management strategies for overcoming crises and developing healthy management strategies. The Smart Management Strategy for Health intervention includes the following 9 strategies: (1) assessment, (2) reality acceptance, (3) preparation for change, (4) decision making, (5) planning, (6) environment creation, (7) action, (8) feedback and maintenance, and (9) core strategies. All of these strategies can help patients overcome a disease crisis and develop healthy self-management skills \[[@ref16],[@ref17]\].
The program covered 4 areas: self-assessment, self-planning, self-learning, and self-monitoring by automatic feedback. We targeted 4 priority areas for intervention---positive thinking, balanced diet, physical activity, and medication. The 20 learning sessions included *12 Rules for Highly Effective Health Behavior* and health management strategies.
The self-management strategy-based electronic program was used for 12 weeks. The patients were provided with a manual with detailed instructions on how to use the program to both increase its usage rate and decrease the dropout rate. Self-evaluations were conducted with regard to the participants' self-management competence and health practices before and after the program (excellent, moderate, and poor). In addition, the patients wrote health mission statements that included their life goals, health practice goals, obstacles, and methods to overcome them and detailed promises in relation to the self-management strategy-based electronic program. Self-learning was structured with the health management strategy and health information on 12 health behavior rules. The patients received daily health educational content from the self-management strategy-based electronic program. Every week, the patients learned 1 health behavior among the 4 essential rules, and they could selectively study the other 8 health behavior rules. By graphically displaying the participants' blood glucose levels, blood pressure, and weight to them, it was possible for the participants to track any changes.
The patients could create their own health management weekly plan for the 4 essential health behavior rules and monitor their progress and health. The weekly plan addressed dieting, vegetable and fruit consumption, physical activity, and daily medication schedule. More specifically, the weekly physical activity plan included the activity's type, length of time, intensity, and schedule. The self-management strategy-based electronic program included an automatic push function and alarms for the scheduled physical activities, medications, and assessments to remind participants of their plans. After 1 week, the patients were provided with feedback to motivate and help them plan for the following week. Through periodic monthly assessments, the program identified changes in their essential health behaviors and provided feedback on monthly changes through a comparison of their prior month's results to help patients change their behavior.
Measures
--------
The primary outcome was the percentage of subjects that met the target clinical indicators (HbA~1c~ \<7.0%, SB*P*\<140 mmHg in clinic, or LDL cholesterol \<130 mg/dL).
The secondary outcomes included the originally proposed clinical indicator outcomes---physical activity, depression, self-management strategies, and health behaviors after 12 weeks in the program. The patients' self-management strategies were assessed with a short form of the Smart Management Strategy for Health, which is a 3-set, 16-factor, 30-item tool (ie, core strategies, 10 items; preparation strategies, 10 items; and implementation strategies, 10 items) that assesses patients' abilities to overcome health-related crises \[[@ref17]\]. Physical activity was measured with the modified version of the Godin Leisure-time Exercise Questionnaire, which is widely used, reliable, and valid \[[@ref16]\]. The modified version adds average duration to the original questions of average frequency of light, moderate, and strenuous exercise per week. We evaluated depression with the Patient Heath Questionnaire-9 (PHQ-9). The participants were asked to measure their 12 health behaviors with 5 scales: (1) precontemplation, (2) contemplation, (3) preparation, (4) action, and (5) maintenance, which are all based on the transtheoretical model \[[@ref18],[@ref19]\].
We assessed the proportion of patients with a ≥1.0% decrease in their HbA~1c~ level from the baseline, the proportion of patients with a ≥10-mmHg decrease in SBP from the baseline, and a ≥15% decrease in LDL cholesterol level. We also assessed the proportion of patients with either a decrease or no change in PHQ-9 score from the baseline, a ≥5- metabolic equivalent of task increase in physical activity level, a ≥10% increase in self-management strategy, and a ≥3 habits increase in the maintenance of the 12 health habits.
The participants completed baseline questionnaires before randomization at the clinics. After 12 weeks, we conducted follow-up assessments with the participants with regard to the primary and secondary outcomes. When patients did not complete a questionnaire item, the clinical research coordinator documented the reason.
Statistical Analysis
--------------------
Providing 80% power to detect a 30% proportion difference in patients achieving disease control with a 2-tailed alpha value of less than .05, we calculated that it was necessary to have 42 patients per group. We predicted a 20% dropout rate and aimed at recruiting 53 patients in each group. A multiple imputation approach was used to impute scores for missing values for the intent-to-treat analysis. The imputed values were used for the covariates analyses but not for the descriptive statistics.
We used a Student *t* test or Pearson chi-square test to determine significant differences in the baseline characteristics between the intervention and control groups. We used an analysis of covariance to estimate between-group changes in the clinical outcome numbers with general linear modeling, adjusting for the baseline score and age. We compared the participants' changes from their baseline values with their values after 12 weeks in the program. Pearson chi-square test was used to assess between-group differences in the proportion of patients with improvement (overall, depression, and physical activity). Enhanced self-management strategies and health habits were also estimated by using a Pearson chi-square test.
We used STATA version 14.2 (STATA) for all statistical analyses. A two-sided *P* value \<.05 was considered significant.
Results
=======
Study Participants
------------------
The study team contacted 281 patients between October 27, 2017, and March 26, 2018. Of these, 124 patients were eligible, and 18 were excluded because of screening failures or refusal to participate. Finally, 53 were randomized to the intervention group and 53 to the control group ([Figure 1](#figure1){ref-type="fig"}). Except for age and residence, all baseline characteristics were similar between the 2 groups ([Table 1](#table1){ref-type="table"}). More specifically, compared with the control group, the intervention group was older, and they were more likely to reside in metropolitan areas (*P*=.001 and .04, respectively).
######
Baseline characteristics of participants.
-------------------------------------------------------------------------------------------------------------------
Characteristics Intervention group (n=53), n (%) Control group (n=53), n (%)
---------------------------------------- ---------------------------------- ----------------------------- ---------
**Age (years)** \ \
\ 20-49 8 (13) 18 (34)
\ 50-59 17 (33) 21 (40)
\ 60-69 26 (50) 8 (15)
\ ≥70 2 (4) 6 (11)
**Sex** \ \
\ Male 31 (58) 29 (55)
\ Female 22 (42) 24 (45)
**Marital status** \ \
\ Married 47 (89) 45 (85)
\ Unmarried 4 (8) 6 (11)
\ Separated/bereaved 2 (4) 2 (4)
**Educational status** \ \
\ High school or less 18 (34) 18 (34)
\ ≥College or university 35 (66) 35 (66)
**Presence of religion** \ \
\ Yes 36 (68) 27 (51)
\ No 17 (32) 26 (49)
**Residence** \ \
\ Metropolitan 39 (74) 28 (52)
\ Urban or rural 14 (26) 25 (47)
**Monthly income (1000 KRW^a^/month)** \ \
\ ≤3999 14 (26) 22 (42)
\ 4000-4999 10 (19) 9 (17)
\ ≥5000 29 (55) 22 (42)
**Employment status** \ \
\ Employed 36 (68) 33 (62)
\ Unemployed/retired 17 (32) 20 (38)
**Disease^b^** \ \
\ Diabetes mellitus 26 (49) 21 (40)
\ Dyslipidemia 23 (43) 23 (43)
\ Hypertension 11 (21) 14 (26)
-------------------------------------------------------------------------------------------------------------------
^a^KRW: Korean Won.
^b^Some participants have been diagnosed with more than one disease.
Success Rate for Achieving Goals
--------------------------------
[Table 2](#table2){ref-type="table"} describes the percentage of patients' achieved goals. The intervention group showed a significantly higher success rate for achieving the targeted levels for each of the 3 clinical indicators after 12 weeks, and this higher success rate remained significant when stratified by starting medication with the Mantel-Haenszel method (*P*\<.05). With regard to disease, the patients with hypertension in the intervention group showed significant improvement compared with the control group (72.7% vs 35.7%, *P*=.035; Mantel-Haenszel chi-square test). These results for the patients with diabetes and hypercholesterolemia were nonsignificant.
We found a significant reduction of HbA~1c~ in the intervention group compared with the control group (0.71 vs 0.22, respectively; between-group difference=--0.54, 95% CI --0.98 to --0.11; *P*=.014). The patients with hypertension exhibited a greater reduction in SBP in the intervention group compared with the control group; however, this result was nonsignificant (17.5 mmHg vs 11.6 mmHg; *P*=.41). For the patients with hypercholesterolemia, both the intervention and control groups showed a reduction in LDL cholesterol, and the between-group difference was nonsignificant (23.7 mg/dL vs 25.3 mg/dL; *P*=.72).
######
Differences in clinical outcomes controlling for the primary disease.
----------------------------------------------------------------------------------------------------------------------
Time point Intervention group (n=60) Control group (n=57) *P* value^a^ *P* value^b^
--------------------- --------------------------- ---------------------- -------------- -------------- ----- ----- ---
**All diseases** \ 60 \ 37 .01 .02
\ Baseline 0 \ 0 \ \ \
\ 12 weeks 36 \ 21 \ \ \
**Diabetes^c^** \ 54 \ 38 .35 .43
\ Baseline 0 \ 0 \ \ \
\ 12 weeks 14 \ 8 \ \ \
**Hypertension^d^** \ 73 \ 36 .07 .04
\ Baseline 0 \ 0 \ \ \
\ 12 weeks 8 \ 5 \ \ \
**Dyslipidemia^e^** \ 61 \ 35 .08 .1
\ Baseline 0 \ 0 \ \ \
\ 12 weeks 14 \ 8 \ \ \
----------------------------------------------------------------------------------------------------------------------
^a^All reported *P* values are 2-sided, with *P*\<.05 considered as statistically significant.
^b^Stratified analysis by starting medication (Mantel-Haenszel method).
^c^Intervention: n=26; control: n=21.
^d^Intervention: n=11; control: n=14.
^e^For both intervention and control: n=23.
Differences in Other Clinical Outcomes, Health Outcomes, and Self-Management Measures
-------------------------------------------------------------------------------------
Among the intervention group, 20% of patients with diabetes exhibited a ≥1% decrease in HbA~1c~ (compared with 0% in the control group; [Table 3](#table3){ref-type="table"}).
In the intervention group, 73% of the participants showed a decrease or no change in depressive symptoms (vs 51% in the control group). The Smart Healthing program strengthened the implementation strategy of the modified Smart Management Strategy for Health greater in the intervention group (57.5%) than in the control group (33.3%). However, the differences in the core and preparation strategies for both the intervention and control groups were nonsignificant (*P*=.53 and .30, respectively; [Table 4](#table4){ref-type="table"}). There were no important harms or unintended effects observed in either group.
######
Differences in clinical measures.
Differences (clinical outcomes) Intervention group (n=53), n (%) Control group (n=53), n (%) *P* value
------------------------------------------------------------------------------------------------ ---------------------------------- ----------------------------- -----------
≥1.0 percentage point decrease in glycated hemoglobin level (intervention: n=25; control n=19) 5 (20) 0 (0) .04
≥10 mmHg decrease in systolic blood pressure (intervention: n=10; control: n=9) 8 (80) 5 (56) .25
≥15% low-density lipoprotein decrease (intervention: n=15; control: n=17) 7 (47) 7 (41) .76
######
Differences in health outcomes and self-management measures.
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Differences Intervention group (n=53), n (%) Control group (n=53), n (%) *P* value
-------------------------------- ------------------------------------------------------------------------------------------- ----------------------------- ----------- -----
**Health outcomes** \ \ \
\ Decrease or no change in PHQ-9^a^ score (intervention: n=41; control: n=39) 30 (73) 20 (51) .04
\ ≥5 metabolic equivalent of task physical activity (intervention: n=41; control: n=39) 29 (71) 32 (82) .23
\ ≥Increase in 3 of the 12 health habits (intervention: n=41; control: n=39) 12 (29) 11 (28) .92
**Self-management strategies** \ \ \
\ ≥10% increase in the *Core Strategy of SAT*^b^ (intervention: n=41; control: n=39) 13 (32) 15 (38) .53
\ ≥10% increase in the *Preparation Strategy of SAT* (intervention: n=38; control: n=39) 15 (39) 20 (51) .30
\ ≥10% increase in the *Implementation Strategy of SAT* (intervention: n=40; control: n=33) 23 (58) 13 (33) .03
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
^a^PHQ-9: Patient Health Questionnaire-9.
^b^SAT: Smart Management Strategy for Health Assessment Tool.
Discussion
==========
Principal Findings
------------------
This RCT indicated that this study's self-management strategy-based electronic program effectively encouraged patients with at least one indicator of poor disease control for diabetes, hypertension, or hypercholesterolemia to meet key guideline criteria (HbA~1c~, SBP, and LDL cholesterol). The patients with hypertension showed a significant improvement in SBP from their baseline values in comparison with the control groups. There was also a significant reduction in HbA~1c~ in the intervention group compared with the control group. We are particularly encouraged by these findings, and we posit that this study's self-management strategy-based electronic program can more effectively support disease control in comparison with the usual care strategies.
The proportion of patients with controlled hypertension increased significantly more in the intervention group than in the control group. The proportion of patients with controlled diabetes and hypercholesterolemia also increased more in the intervention group than in the control group; however, these findings were nonsignificant, which could be a result of the study's small sample size. These improvements in the primary outcomes in our trial support the findings from earlier trials with same clinical indicators. Concerning the secondary outcomes, the mean change in HbA~1c~ and the proportion of patients with a significant decrease in HbA~1c~ level from their baseline values were both higher in the intervention than the control (ie, usual care) group \[[@ref3],[@ref6],[@ref20]-[@ref24]\].
There are several possible explanations for our findings. First, the intervention strategies were based on the Smart Management Strategy for Health program. The intervention significantly increased the participants' Implementation Strategy scores for self-management. It is possible that the CCM self-management program thus helps individuals to develop preferences for how to manage their own care \[[@ref7],[@ref8]\] It is assumed that most patients want to remain independent; however, these preferences and patients' daily behaviors may change over time because of their symptoms, the treatments they undergo and their goals \[[@ref8]\]. Patients with chronic illnesses must manage the medical and emotional strain of their health condition(s) \[[@ref8],[@ref25]\]. The Smart Management Strategy for Health supports patients to help them overcome a disease crisis and develop health-related self-management skills \[[@ref26],[@ref27]\] The fact that this intervention integrates self-management strategies with electronic program in the CCM highlights how mHealth can address cardiovascular risks \[[@ref28]\].
Second, a user‐centered electronic program has the potential to improve clinical indicators among those living with chronic diseases by allowing users to obtain information from the mobile- and Web-based pages at their own pace, to flexibly review material as needed \[[@ref29],[@ref30]\] and by facilitating the management of multiple health behaviors \[[@ref11]\]. Third, the noted self-management program can help patients by providing immediate, easy, and continual access to the intervention \[[@ref4],[@ref8],[@ref15]\]. This electronic program intervention may thus provide a critical route to successful chronic care. From a clinical perspective, it could be valuable to link this electronic program-based program with face-to-face or telephone counseling \[[@ref31]-[@ref33]\].
Furthermore, we did not observe any significant changes concerning the examined health habits. There are 2 possible explanations for this finding. First, the intervention might not have been intensive enough to modify patients' long-held health habits. Second, the 12-week intervention period might have been too short to observe any meaningful changes in health habits \[[@ref5]\].
Despite well-established data on chronic disease management, its uptake into routine clinical practice remains limited. Further innovation, optimization, and rigorous research in customized mobile technology might improve health care delivery and outcomes \[[@ref5],[@ref12]\].
Limitations
-----------
Several limitations of our study should be noted. First, our sample included 3 types of cardiovascular risks of varying severity and a small number of patients; thus, we lacked the power to determine meaningful differences. Further studies with larger sample sizes and distinct cardiovascular risks are needed to confirm the efficacy of the intervention. Second, approximately one-quarter of the patients in the intervention group did not complete the follow-up at the 12-week mark. Missing data for these patients may have resulted in an underestimation of the efficacy of the intervention program. Third, as the patients were aware of their group, the self-reported changes in depression, physical activity, self-management strategy, and health behaviors could have been influenced by that awareness and not just by the intervention itself. Fourth, there might be attrition bias. A quarter of the intervention group was lost to follow up, and many patients of this group were older, which may be associated with less use of newer technology. This loss may lead to an overestimate of effect. Fifth, although the attention control group was encouraged to continue their usual care and routine medications and to study a health educational booklet about chronic diseases, the Hawthorne effect may be still relevant. Finally, our trial was relatively short, and we do not know whether the changes associated with the program would be maintained over a longer period. Additional research on the long-term efficacy of this intervention, including a full-scale RCT, is warranted to confirm the efficacy of this program.
Conclusions
-----------
A short-term self-management strategy-based electronic program intervention may improve clinical outcomes among patients with cardiovascular risks. More research with context-specific trials is needed to enhance these findings, to ensure the long-term generalizability and sustainability of the program, and to indicate the cost-effectiveness of this intervention.
This study was supported by the Korea Health Technology R&D Project through the Korea Health Industry Development Institute and by the Ministry of Health and Welfare, Republic of Korea (grant number: HI16C0455). The authors would like to thank Editage for English language editing.
Conflicts of Interest: None declared.
Screenshot of the app.
Screenshot of the intervention (Smart Healthing).
CONSORT-EHEALTH checklist (V 1.6.1).
CCM
: chronic care model
HbA~1c~
: glycated hemoglobin
LDL
: low-density lipoprotein
mHealth
: mobile health
PHQ-9
: Patient Health Questionnaire-9
RCT
: randomized controlled trial
SBP
: systolic blood pressure
| {
"pile_set_name": "PubMed Central"
} |
We read with interest the recent paper by Wang et al. \[[@B1]\]. Authors analyzed the results of the published literature where Chinese herbs were used as osteoporosis therapy (as measured by BMD). And authors concluded that Chinese herbs have merits in improving lumbar spine BMD as compared to the placebo or other standard antiosteoporotic drugs.
However, there was a serious issue with this papaer. Due to significant differences existed among participants, study design, intervention, and outcome measurement, statistical heterogeneity was noted in the analysis of this study. According to the Cochrane Handbook for Systematic Reviews \[[@B2]\], *I* ^2^ ranges between 0% and 100%; *I* ^2^ values of 25%, 50%, and 75% are referred to as low, moderate, and high estimates. *I* ^2^ statistic greater than 50% suggested moderate heterogeneity, and a random effects model should be used. Instead, a fixed effects model was used for *I* ^2^ statistic less than 50%, which showed that heterogeneity could be neglected \[[@B2]\]. In the present study by Wang et al. \[[@B1]\], *I* ^2^ value was 94% in the analysis of Chinese herbs versus placebo on spine BMD; 96% in the analysis of Chinese herbs versus placebo on femoral neck BMD; 84% in the analysis of Chinese herbs versus standard antiosteoporotic drugs on lumber spine BMD; and 0% in the analysis of Chinese herbs versus standard antiosteoporotic drugs on the femoral neck BMD. But all the authors used fixed effects model regardless of the heterogeneity which was 0% or 96%. Is this reasonable?
In my opinion, the present study by Wang et al. \[[@B1]\] gives us an important message: Chinese herb is effective in treatment of bone loss among patients with osteoporosis. Authors have done excellent job, and credit should be given to this work. But analysis method used in this study was irrational and should not be neglected, for this may influence the final results.
The author declare that there is no conflict of interests.
[^1]: Academic Editor: Alexandra Deters
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
The Chilean population is rapidly ageing due to a process of fast demographic transition in the second half of the twentieth century ([@R19]). Further understanding of possible influences on health in later life, including mental health, is therefore important particularly as the prevalence of depression among older people in Chile is high in comparison with other Latin American countries ([@R1]; [@R31]). In this study, we investigate Chilean grandparents' help to grandchildren and associations between providing such help and grandparents' subsequent mental well-being.
Numerous studies have documented the contributions that grandparents make to the support of younger family members, including providing assistance with the care of grandchildren. The benefits of this support for the recipients, and for society as a whole, are well recognised ([@R7]; [@R21]) but evidence on implications for grandparents is mixed. Altruistic behaviours and balanced intergenerational exchanges are hypothesised to have mental health benefits ([@R6]; [@R8]) but caring for children can be stressful and may limit opportunities for other forms of activity, social engagement and self-care with consequent negative health implications ([@R11]; [@R14]; [@R27]).
Much previous research on grandparenting and mental health has been based on studies from the US and has focussed on grandparents providing custodial or intensive care for grandchildren. In general, these studies have reported poorer mental (and physical) health for grandparents in 'skipped generation' households with primary responsibility for grandchildren, and for grandparents living with grandchildren in three generation households ([@R8]; [@R14]; [@R15]; [@R17]) although some longitudinal studies suggest that negative effects reduce over time ([@R2]; [@R27]).
However, it is unclear to what extent these findings may reflect prior characteristics and experiences of the grandparents involved ([@R26]), especially as custodial grandparenting in the US is often precipitated by mental health or addiction problems of the child\'s mother ([@R14]), or whether similar associations apply to grandparents providing lesser amounts of help. [@R10] found that grandparent carers had poorer mental health than other grandparents even before assuming care for a grandchild and some indications that grandmothers providing less intensive help to grandchildren had reduced risks of depression. Similarly, [@R6] in analyses of another US longitudinal study in which baseline mental health characteristics were controlled, found that providing moderate amounts of help to grandchildren was protective against depression two-three years later for grandfathers, although not for grandmothers. Some other studies have also reported gender differences in associations between grandparenting and mental health. Analyses of the US National Survey of Families and Households undertaken by [@R16], for example, found more depression among caregiving grandmothers than equivalent grandfathers.
The applicability of results from studies of grand-parenting in the US or other high-ncome Western countries to low- or mid-income societies with different patterns of family and household organisation is also questionable. In Latin America, three generation households are much more prevalent than in North America or Europe ([@R5]; [@R28]) and involvement in extended family life may be beneficial for older people\'s mental health. Previous longitudinal studies of grandparenthood and mental health are lacking but results from cross sectional analyses suggest that for older Cubans, for example, social networks centred on children and the extended family are associated with a low frequency of depressive symptoms ([@R23]). Results from China also show a different pattern of associations from those reported in the US. One cross sectional study of a rural Chinese population found that older parents living in three-generation households or with grandchildren in skipped-generation households had better psychological well-being than those living in single-generation households, a finding attributed in part to the cultural value attached to multi-generational co-residence ([@R24]).
In this article, we use longitudinal data on a representative sample of older people resident in the Greater Santiago area of Chile to investigate the relationship between providing help to grandchildren and mental health. To our knowledge, this is the first longitudinal investigation of the implications of providing grandchild care for the mental health of grandparents in a Latin American population.
Aims and hypotheses {#s2}
===================
The first aim of the research reported here was to analyse factors associated with the provision of help to grandchildren in a sample of Chilean grandparents. On the basis of previous studies, we expected that women would be more involved in providing care for grandchildren than men but that grandfathers might provide more material assistance. We also expected that factors related either to potential demand for grandparent help (number of grandchildren, age of youngest grandchild and grandchild in the household) or ability of grandparents to provide it (health status and competing activities) would be associated with differentials in provision.
The second aim of the study was to investigate associations between provision of this help and mental health outcomes two years later. As summarised in the introduction, previous research and theory provide conflicting evidence on the possible direction of any such association in the older Chilean population. Studies from the US suggest that coresidential or custodial grandparenting, but possibly not provision of lesser amounts of help, has negative effects on mental health. In Latin American societies, three generational family connections, including co-residence, may be more 'normative' and possibly beneficial for older people\'s mental health. On the other hand, the poorer economic and physical health status ([@R31]) of older Chileans compared with North Americans may increase vulnerability to stresses attendant on providing care for grandchildren.
Data {#s3}
====
We used data collected in 2005 from a sample of people aged 66--68 resident in low or middle income areas of the Santiago Metropolitan area of Chile. The sample comprised participants in a cluster randomised controlled trial primarily designed to investigate the cost effectiveness of a nutritional supplement and/or exercise programme on pneumonia incidence, walking capacity and body mass index. The restricted age range was chosen in order to include respondents just below the threshold age for automatically receiving the nutritional supplement the effectiveness of which the trial aimed to evaluate. Full details of the trial methodology and primary findings have been reported elsewhere ([@R3], [@R4]). In brief, the sampling strategy involved recruiting from age-eligible people registered with 20 health centres in low- or middle-income areas of Santiago. Exclusion criteria were inability to walk unaided; having sought medical advice for unplanned weight loss in the past three months; already consuming the nutritional supplements the trial was designed to evaluate; planning to move house within the next three months or poor cognitive function (Mini Mental State Examination short form test score of \< 13 and Pfeffer score of 6 or more). This sampling strategy resulted in identification of 2649 eligible participants of whom 2002 were recruited to the trial, a response rate of 76%. All participants gave signed informed consent and the trial was approved by the Institutional Review Board at INTA, University of Chile and the London School of Hygiene & Tropical Medicine Ethics Committee.
Participants were interviewed at baseline in their local community centre or at home if unable to visit the community centre. Information was collected on physical and mental health status; on socio-economic and demographic characteristics, including level of education, family and household structure and number of grandchildren, and on participation in various activities. These included provision of help to grandchildren, participation in community organisations, and paid and unpaid work. At the end of the trial, 24 months after enrolment, respondents were re-interviewed when the baseline questionnaire (with a few minor amendments) was re-administered. By the time of this follow up, 28 respondents had died; 1669 of the remainder (85%) were successfully interviewed. Most loss to follow-up was due to inability to locate respondents at their previous address, despite several attempts. Additionally, small proportions were known to have moved out of the study area or refused reinterview.
Measures {#s4}
========
Grandchildren, help to grandchildren and family and household characteristics {#s5}
-----------------------------------------------------------------------------
Respondents were asked how many grandchildren they had and for the ages of the youngest and oldest. A social definition of grandchildren was chosen as being most appropriate to the study population with interpretation of who constituted a grandchild left to respondents. An additional question asked about 'other children you consider to be like grandchildren' who were included with grandchildren. Respondents were asked whether they 'regularly helped' grandchildren (including children considered to be like grandchildren) and approximate hours per week spent helping (none; less than 2; 2--4 or 4 or more). They were also asked if they regularly provided help with money or material goods that grandchildren needed. Information collected on all household members and their relationship to the respondent was used to derive a three-category household type variable distinguishing those living alone or just with a spouse/ partner; those living with other relatives (whether or not they also lived with a partner) not including a grandchild, and those living with other relatives (with or without a partner) including one or more grandchildren. Information on marital or partnership status was dichotomised into married or living with a partner versus unmarried/unpartnered.
Mental health and well-being outcomes {#s6}
-------------------------------------
Life satisfaction was measured using an indicator derived from responses to four items from Neugarten\'s life satisfaction scale ([@R18]): 'As I grow older things seem better than I thought they would be'; 'These are the best years of my life'; 'I feel my age but it doesn\'t worry me' and 'These are the worst years of my life'. Respondents were asked whether they agreed, disagreed or neither agreed nor disagreed with these statements and a score was derived distinguishing those with positive attitudes (yes to the first three items and no to the fourth); those with negative attitudes (no to the first three items and yes to the fourth) and an intermediate group, termed neutral, with other mixtures of responses. The second indicator of mental health was score on the Mental Component Summary (MCS) of the SF-36 scored using an algorithm derived for Chilean older people ([@R12]). This was included as a continuous variable. Depression was measured using the 15 item Geriatric Depression Scale (GDS) ([@R22]) with those with scores of 5 or more considered as having depressive symptoms.
Other co-variates {#s7}
-----------------
Other covariates were selected on the basis of the research questions and the previous literature on factors associated either with provision of help to grandchildren and/or with mental health outcomes. Indicators of socio-economic status and resources included years of education (grouped 0--5 or not known, 6--8, 9+) and household income per household member which was used in analyses of variations in transfers of money or material goods to grandchildren. Physical health status was measured using an indicator of functional health based on responses to 15 questions about limitations in specific areas of function or mobility including reaching; lifting; bending; stair climbing; walking; running and also bathing and dressing. On the basis of the observed distribution, we distinguished three groups: those with 0--2 limitations (low); those with 3--4 limitations (mid) and those with five or more limitations (high).
Variables based on questions about paid and unpaid work and participation in community organisations were included as participation in these activities might compete with the provision of help to grandchildren and are known to be associated with mental health ([@R13]; [@R25]). In Chile competing demands from work are likely to be particularly relevant, and to vary by level of education, even though those included in this study were older than the legal retirement ages of 65 for men and 60 for women. This is because the part privatised system introduced in the early 1980s worked reasonably well for workers with regular jobs, but not for the many Chileans, particularly the less well educated and women, who rely on part-time, seasonal or informal work and so have limited opportunities for acquiring the 20 years of contributions required for even the minimum pension ([@R30]). For those lacking contributory pensions the only other state provision available at the time this study was conducted was rationed access to low-level 'assistance' pensions ([@R30]). We defined paid work as being employed, working in a family business (whether paid or unpaid), or undertaking various self-employed activities. Community work was defined as voluntary work for a community, charitable, educational or similar body or participation in a community group, club or other organisation. For most variables considered, item non-response and consequent missing data was trivial. The exception was information on household income which was missing for 6.7% of the baseline sample of grandparents.
Analysis {#s8}
========
The analysis was restricted to the 608 men (94%) and 1352 women (also 94%) who reported having at least one grandchild. We first investigated factors associated with the provision of help to grandchildren at baseline. Hours of help provided (time help) was analysed using ordinal logistic regression with three outcome categories; no help given (reference); less than four hours per week, and four or more hours per week. Models were specified in accordance with the previous literature and study research questions. Model 1 includes variables related to demand or need for help, namely grandchild characteristics and family and household situation (number of grandchildren; age of youngest grandchild; household type and marital status). Number of grandchildren was included as a continuous variable; age of youngest grandchild was included as a categorical variable distinguishing four groups (0--4, 5--10, 11+, age not known). Model 2 additionally includes potential confounding variables and variables related to ability to provide help and possible competing demands (educational status, functional health limitations, paid work and involvement in voluntary or community activities). We also used logistic regression to analyse the dichotomous outcome provision of help with money or material goods, including equivalised household income as an additional co-variate, but report only briefly on results of this here due to limitations of space.
In the second stage of the analysis, we analysed differences in the three mental health outcomes at follow-up (life satisfaction; SF-36 MCS score and depression score). Life satisfaction and depression were dichotomised into outcomes representing positive life satisfaction versus neutral or negative and GDS score of 0--4 (reference) versus 5 or more (indicative of depressive symptoms). These were analysed using logistic regression. Linear regression was used to analyse the continuous SF-36 MCS score after preliminary checks that the distribution met assumptions required. Three models were fitted in these analyses. In addition to the main variable of interest (help to grandchildren), Models 1 and 2 were as specified in the first part of the analysis and included grandchild and family and household variables (Model 1) and additionally variables relating to education, physical health and work and community activities (Model 2). Model 3 also included the baseline value of the mental health outcome being investigated. In the interests of brevity, we show here only results from Models 1 and 3 and comment briefly on differences between Model 2 and the final Model 3; full results are available from the authors on request. We dichotomised the follow-up outcome measures of life satisfaction and depression in the interests of model parsimony but retained more information in the equivalent baseline co-variate and used a trichotomous indicator distinguishing positive, neutral and negative life satisfaction scores and GDS scores of 0--4, 5--9 and 10+ respectively.
Random-effect models were used to account for clustering (by health centre) in all regression analysis. All analyses were undertaken using STATA version 11 or 12 and results presented are for those with available data for all variables included in the relevant series of models.
Results {#s9}
=======
Descriptive results {#s10}
-------------------
[Table 1](#T1){ref-type="table"} presents descriptive information about the sample at baseline. Grandfathers had an average of 6.9 grandchildren and 55% had one or more grandchildren aged under 5. Grandmothers had on average slightly more grandchildren (8.2) and slightly older grandchildren, reflecting women\'s generally earlier age at parenthood and so grandparenthood. Co-residence with one or more grandchildren was slightly more usual among grandmothers (43%) than grandfathers (38%). Three quarters of the sample provided some time help to grandchildren; the proportion providing four or more hours per week was higher among grandmothers (55%) than grandfathers (48%). However, grandmothers were slightly less likely than grandfathers to provide help to grandchildren with money or material goods (54% versus 59%). Few grandparents provided only help with material goods and overall 70% of those providing time help, also provided help with money. There were large gender differences in other characteristics of sample members. Compared to women, men had higher levels of education, were much more likely to be married or partnered and to have paid work, although a higher proportion of women than men were involved in community organisations. Women\'s functional health was much worse than that of men; 54% had five or more functional limitations compared with 29% of men.
######
Characteristics of grandparents at baseline, 2005.
Grandfathers Grandmothers
----------------------------------------------------------------------------- -------------- -------------- ------------- -------
Provision of help to grandchildren 608 1264
None 21.55 22.63
Money/things they need only 3.78 3.88
Time only: 0 \> 4 hours per week 6.91 6.80
Time only: 4+ hours per week 12.34 16.53
Money + 0 \> 4 hours help per week 19.41 11.63
Money + 4 + hours help per week 36.02 38.53
All providing help with money[∗](#T1-FN1){ref-type="table-fn"} 59.21 54.04
All providing 0 \> 4 hours help per week[∗∗∗](#T1-FN1){ref-type="table-fn"} 26.33 18.43
All providing 4 + hours help per week[∗∗](#T1-FN1){ref-type="table-fn"} 48.36 55.06
Grandchild characteristics 608 1264
Number of grandchildren[∗∗∗](#T1-FN1){ref-type="table-fn"} 6.92 (5.18) 605 8.15 (6.33) 1249
Age group of youngest[∗∗∗](#T1-FN1){ref-type="table-fn"} 608 1264
0−4 54.77 45.81
5--10 26.48 30.62
11+ 7.57 11.87
Not known 11.18 11.71
Household type[∗](#T1-FN1){ref-type="table-fn"} 608 1264
Alone or with spouse only 29.77 24.84
With other relatives (+/− partner), not including grandchild 32.73 32.52
With other relatives (+/− partner), including grandchild 37.50 42.64
Married/partner[∗∗∗](#T1-FN1){ref-type="table-fn"} 85.86 56.96
Education (years)[∗∗∗](#T1-FN1){ref-type="table-fn"} 608 1264
\<6 (incl. not known) (Low) 25.66 36.31
6--8 (Mid) 41.45 39.72
9+ (High) 32.89 23.97
Other activities 608 1263
Has paid work[∗∗∗](#T1-FN1){ref-type="table-fn"} 51.81 27.16
Has community work[∗∗](#T1-FN1){ref-type="table-fn"} 32.40 40.66
Functional limitations[∗∗∗](#T1-FN1){ref-type="table-fn"} 608 1264
5 or more (High) 29.11 54.11
2--4 (Mid) 22.70 24.53
0--2 (Low) 48.19 21.36
Notes: Difference between grandfathers and grandmothers.
∗∗∗*p* \< 0.001; ∗∗*p* \< 0.01; ∗*p* \> 0.05.
[Table 2](#T2){ref-type="table"} shows the life satisfaction, SF-36 MCS and depression scores at baseline and follow-up for those included in the follow-up; comparable information for the whole baseline sample is also presented as a guide to possible bias resulting from differential inclusion in the follow-up by baseline mental health status. 60% of women and 61% of men had had positive life satisfaction scores at baseline; by follow up these proportions had decreased slightly to 53% and 57%, respectively. Mean SF36-MCS for the sample present at both rounds of the survey were also slightly lower (worse) at follow-up and lower for women than men. 21% of men and 36% of women had GDS scores of five or more (indicating depression) at follow-up compared with 20% and 32%, respectively at baseline. Slight differences in the baseline values of these variables for those included in the follow-up and the whole baseline sample are suggestive of some association between non-participation in the follow-up and worse initial mental health, but differences were not statistically significant.
######
Indicators of mental well-being at baseline (2005) and follow-up (2007).
Grandfathers Grandmothers
------------------------------- -------------- -------------- -------------- -------------- -------------- --------------
Life satisfaction score (*N* = 607) (*N* = 465) (*N* = 465) (*N* = 1263) (*N* = 1081) (*N* = 1081)
Negative (%) 10.54 9.89 12.69 12.98 13.04 16.74
Neutral(%) 30.31 28.17 30.54 27.08 27.10 29.88
Positive(%) 59.14 61.94 56.77 59.94 59.85 53.38
SF-36: mental component score (*N* = 603) (*N* = 464) (*N* = 464) (*N* = 1255) (*N* = 1072) (*N* = 1072)
Mean (SD) 52.84 (7.48) 53.29 (7.01) 53.21 (7.28) 48.62 (8.41) 48.83 (8.34) 48.67 (8.57)
GDS score (*N* = 600) (*N* = 461) (*N* = 461) 1255 (*N* = 1073) (*N* = 1073)
0--4(%) 78.00 80.26 78.74 64.46 65.80 63.56
5--9(%) 15.67 14.53 14.10 22.63 21.71 20.97
10 + (%) 6.33 5.21 7.16 12.91 12.49 15.47
Notes: Difference between grandfathers and grandmothers in follow-up sample.
Life satisfaction: *p* \> 0.05; SF-MCS: *p* \< 0.001; GDS score: *p* \< 0.001.
Provision of help to grandchildren at baseline {#s11}
==============================================
[Table 3](#T3){ref-type="table"} shows results from ordinal logistic regression models of provision of time help to grandchildren at baseline. Results from all models show an inverse association between provision of more help to grandchildren and age of youngest grandchild and a positive association between provision of help and living in a household including at least one grandchild. Among grandfathers, there was a negative association between provision of help and number of grandchildren and among grandmothers provision of help was positively associated with being married or partnered. Educational level, functional health and engagement in paid or community work were not significantly associated with provision of help among either grandfathers or grandmothers.
######
Results from mixed effects ordinal regression models of hours of help provided to grandchildren at baseline.
Grandfathers Grandmothers
------------------------- ----------------------------------- ------------------------------------------ -------------- ------- ------------------------------------------ ------- ------- ------------------------------------------ ------- ------- ------------------------------------------ ------- -------
Number of grandchildren −0.05[∗∗](#T3-FN1){ref-type="table-fn"} −0.08 −0.01 −0.04[∗∗](#T3-FN1){ref-type="table-fn"} −0.07 −0.01 −0.01 −0.03 0.01 −0.01 −0.03 0.01
Age youngest grandchild 5--10 −0.48[∗](#T3-FN1){ref-type="table-fn"} −0.87 −0.08 −0.52[∗∗](#T3-FN1){ref-type="table-fn"} −0.89 −0.14 −0.43[∗∗](#T3-FN1){ref-type="table-fn"} −0.75 −0.11 −0.43[∗∗](#T3-FN1){ref-type="table-fn"} −0.74 −0.12
11 + −1 21[∗∗∗](#T3-FN1){ref-type="table-fn"} −1.83 −0.59 −1.17[∗∗∗](#T3-FN1){ref-type="table-fn"} −1.74 −0.60 −1.24[∗∗∗](#T3-FN1){ref-type="table-fn"} −1.62 −0.85 −1.24[∗∗∗](#T3-FN1){ref-type="table-fn"} −1.64 −0.85
Not known 1.09[∗∗∗](#T3-FN1){ref-type="table-fn"} −1.50 −0.68 −1.14[∗∗∗](#T3-FN1){ref-type="table-fn"} −1.54 −0.73 −0.64[∗∗](#T3-FN1){ref-type="table-fn"} −1.08 −0.21 −0.66[∗∗](#T3-FN1){ref-type="table-fn"} −1.09 −0.24
Household type With others, not with grandchild −0.16 −0.45 0.13 −0.14 −0.45 0.17 −0.09 −0.36 0.18 −0.10 −0.37 0.17
With others, including grandchild 1.67[∗∗∗](#T3-FN1){ref-type="table-fn"} 1.31 2.04 1.70[∗∗∗](#T3-FN1){ref-type="table-fn"} 1.33 2.08 2.07[∗∗∗](#T3-FN1){ref-type="table-fn"} 1.74 2.41 2.07[∗∗∗](#T3-FN1){ref-type="table-fn"} 1.75 2.40
Marital status Married 0.44 0.00 0.87 0.42 −0.03 0.87 0.38[∗∗](#T3-FN1){ref-type="table-fn"} 0.10 0.62 0.35[∗](#T3-FN1){ref-type="table-fn"} 0.08 0.63
Education Mid 0.05 −0.27 0.38 0.10 −0.28 0.48
High 0.27 −0.07 0.61 0.09 −0.21 0.39
Paid work Yes −0.37 −0.85 0.12 0.14 −0.03 0.31
Community work Yes 0.11 −0.31 0.52 0.04 −0.30 0.38
Functional limitations Mid −0.05 −0.41 0.31 0.06 −0.19 0.30
Low 0.17 −0.22 0.56 −0.26 −0.50 −0.02
*N* 605 605 1248 1248
Notes: Reference categories: Youngest grandchild aged 0--4; lives alone or with spouse only; not married/partnered; low education (\< 6 years); no paid work; no community work; high functional limitation.
∗∗∗*p* \< 0.001; ∗∗*p* \< 0.01; ∗*p* \> 0.05.
Results from logistic regression models of helping grandchildren with money or material goods (not shown) showed that provision of such help was positively associated with living with a grandchild for both grandfathers and grandmothers and that grandfathers who did not know the age of their youngest grandchild were less likely to provide material help. Among men, being married, having a higher equivalised household income and a high (worse) level of functional limitation were also positively associated with provision of material help. There was a negative association between involvement in community work and grandmothers' help to grandchildren with money or material goods.
Provision of help to grandchildren and psychosocial health at follow-up {#s12}
=======================================================================
Life satisfaction {#s13}
-----------------
As shown in [Table 4](#T4){ref-type="table"}, providing help of four or more hours per week was positively associated with grandfathers' life satisfaction at follow-up in both Model 1 and, although slightly attenuated, also in the final fully adjusted Model. The effect of providing smaller amounts of help was not statistically significant although odds ratios were positive. As would be expected, there was a strong association between baseline and follow-up life satisfaction but no other baseline characteristics were associated with life satisfaction at follow-up in the models shown, although in Model 2 (which did not include baseline life satisfaction) there was a significant association between a low level of functional limitation and positive life satisfaction at follow-up (OR 1.97, 95% CI 1.25--3.08). There was no indication of a positive effect of helping grandchildren on grandmothers' life satisfaction and in fact, odds ratios were below 1, although not significantly so, in all models. In the fully adjusted model, living with relatives rather than alone or just with a spouse was positively associated with good life satisfaction, particularly for those living with relatives not including grandchildren. Participation in community activities and a low or mid, rather than high, level of functional limitations were also associated with good life satisfaction.
######
Results from mixed effects logistic regression models of time help to grandchildren and life satisfaction at follow up.
Grandfathers Grandmothers
-------------------------------- ----------------------------------- -------------------------------------------- -------------- --------------------------------------- ---------------------------------------- ------ ------- --------------------------------------- ------ ------ ------------------------------------------ ------ -------
Hours of help to grandchildren 0 \> 4 1.36 0.81 2.29 1.21 0.69 2.13 0.70 0.48 1.02 0.73 0.48 1.09
4+ 1.99[∗∗](#T4-FN1){ref-type="table-fn"}1.19 3.32 1.78[∗](#T4-FN1){ref-type="table-fn"} 1.02 3.13 0.78 0.56 1.08 0.81 0.56 1.17
Number of grandchildren 1.01 0.97 1.05 1.00 0.96 1.04 0.99 0.97 1.01 0.99 0.97 1.01
Age youngest grandchild 5--10 1.10 0.71 1.71 1.09 0.68 1.75 1.07 0.81 1.42 1.23 0.90 1.68
11 + 1.64 0.76 3.57 1.72 0.73 4.04 1.16 0.77 1.74 1.27 0.81 1.99
Not known 1.67 0.83 3.33 1.63 0.77 3.46 1.05 0.67 1.63 1.15 0.70 1.88
Household composition With others, not with grandchild 1.02 0.63 1.65 1.08 0.64 1.83 1.44[∗](#T4-FN1){ref-type="table-fn"} 1.04 2.01 1.71[∗∗](#T4-FN1){ref-type="table-fn"} 1.19 2.46
With others, including grandchild 0.94 0.57 1.53 0.89 0.52 1.53 1.25 0.90 1.75 1.45[∗](#T4-FN1){ref-type="table-fn"} 1.01 2.10
Marital status Married 1.17 0.65 2.08 1.23 0.71 2.12 10.7 0.84 1.38 1.00 0.76 1.32
Education Mid 1.07 0.57 2.01 0.93 0.68 1.26
High 1.08 0.64 1.80 0.75 0.52 1.07
Paid work Has paid work 0.94 0.62 1.43 0.87 0.64 1.18
Community work Has Community work 1.19 0.77 1.84 1.34[∗](#T4-FN1){ref-type="table-fn"} 1.02 1.76
Functional limitations Mid 1.49 0.83 2.66 1.60[∗∗](#T4-FN1){ref-type="table-fn"} 1.16 2.21
Low 1.46 0.90 2.38 2.38[∗∗∗](#T4-FN1){ref-type="table-fn"} 1.67 3.40
Life satisfaction Neutral 2.58[∗](#T4-FN1){ref-type="table-fn"} 1.12 5.96 3.29[∗∗∗](#T4-FN1){ref-type="table-fn"} 1.95 5.54
Positive 9.62[∗∗](#T4-FN1){ref-type="table-fn"} 4.32 21.43 10.40[∗∗∗](#T4-FN1){ref-type="table-fn"} 6.33 17.08
N 464 1068 1068
Notes: Reference categories: Youngest grandchild aged 0--4; lives alone or with spouse only; not married/partnered; low education (\<6 years); no paid work; no community work; high functional limitation; life satisfaction negative.
∗∗∗*p* \< 0.001; ∗∗*p* \< 0.01; ∗*p* \> 0.05.
SF36-MCS {#s14}
--------
Grandfathers providing some help, but less than four hours per week, to grandchildren had better follow-up MCS scores than other grandfathers when only family and household characteristics were taken into account (Model 1), but this association was attenuated and ceased to be significant when baseline health and socioeconomic status was controlled (Model 2, not shown); in this model a low level of functional health limitation was positively associated with higher MCS score, but in the final model (Model 3) only the association between baseline and follow-up MCS was significant ([Table 5](#T5){ref-type="table"}). For grandmothers there was an inverse association between number of grandchildren and MCS in Model 1 (and Model 2) but in the final model the only co-variates significantly associated with higher follow-up MCS were baseline MCS and low or mid (rather than high) levels of functional limitation.
######
Results from mixed effects linear regression models of time help to grandchildren and SF36-MCS at follow up.
Grandfathers Grandmothers
-------------------------------- ----------------------------------- --------------------------------------- -------------- ------- ----------------------------------------- ------- ------ ----------------------------------------- ------- ------- ----------------------------------------- ------- ------
Hours of help to grandchildren 0 \< 4 2.40[∗](#T5-FN1){ref-type="table-fn"} 0.33 4.47 0.92 −0.56 2.41 −0.27 −2.08 1.54 0.59 −1.11 2.30
4+ 2.03 −0.51 4.57 1.07 −0.46 2.60 −0.22 −2.17 1.72 0.34 −1.03 1.71
Number of grandchildren −0.06 −0.11 −0.29 −0.07 −0.22 0.10 −0.13[∗∗](#T5-FN1){ref-type="table-fn"} −0.21 −0.05 −0.06 −0.13 0.01
Age youngest grandchild 5--10 −0.58 −2.86 −1.69 −0.20 −1.67 1.28 0.37 −1.20 1.93 0.85 −0.18 1.88
11+ −0.32 −2.63 1.98 0.79 −0.75 2.33 −0.88 −2.46 0.69 −0.33 −1.50 0.84
Not known 1.28 −1.79 4.53 1.56 −0.82 3.93 0.29 −2.01 2.58 0.70 −1.42 2.83
Household composition With others, not with grandchild 0.60 −1.42 2.61 0.15 −1.36 1.66 1.03 −0.30 2.36 1.08 −0.07 2.23
With others, including grandchild 0.78 −1.28 2.84 0.32 −1.45 2.10 −0.02 −2.05 2.02 0.02 −1.19 1.24
Marital status Married 1.32 −0.91 3.54 1.15 −0.66 2.95 0.14 −1.00 1.29 −0.07 1.00 0.85
Education Mid 0.28 −1.42 1.98 0.55 −0.61 1.71
High 1.05 −0.20 2.29 −0.11 −1.55 1.34
Paid work −0.55 −1.62 0.52 −0.28 −1.33 0.76
Community work 0.74 −0.38 1.85 0.09 −0.92 1.10
Functional limitations Mid 0.20 −1.60 2.00 1.26[∗∗](#T5-FN1){ref-type="table-fn"} 0.53 2.00
Low 0.90 −0.35 2.14 3.06[∗∗∗](#T5-FN1){ref-type="table-fn"} 0.51 0.60
MCS 0.59[∗∗∗](#T5-FN1){ref-type="table-fn"} 0.47 0.70 0.56[∗∗∗](#T5-FN1){ref-type="table-fn"} 0.51 0.60
N 463 463 1059
Notes: Reference categories: Youngest grandchild aged 0--4; lives alone or with spouse only; not married/partnered; low education (\< 6 years); no paid work; no community work; high functional limitation.
∗∗∗*p* \< 0.001; ∗∗*p* \< 0.01; ∗*p* \> 0.05.
Depression {#s15}
----------
For men, helping grandchildren was associated with lower odds of depression in Model 1, but in the final Model only baseline depression and better functional health were associated with lower odds of depressive symptoms at follow-up. For grandmothers helping grandchildren for four or more hours per week appeared protective against depression in the fully adjusted model (OR 0.65, 95% CI 0.44, 0.98); the odds ratio for those providing less help (rather than none) was similar but not significant (OR 0.66, 95% CI 0.42, 1.04). Similarly, to results for grandfathers, lower levels of functional limitation and fewer depressive symptoms at baseline were negatively associated with depression at follow-up ([Table 6](#T6){ref-type="table"}).
######
Results from mixed effects logistic regression models of time help to grandchildren and depression at follow up.
Grandfathers Grandmothers
-------------------------------------------------------- ----------------------------------- --------------------------------------- -------------- ------ ----------------------------------------- ------ ------- --------------------------------------- ------ ------ ------------------------------------------ ------ -------
Hours of help to grandchildren (ref. None) 0 \> 4 0.52[∗](#T6-FN1){ref-type="table-fn"} 0.27 0.98 0.59 0.28 1.24 0.91 0.62 1.35 0.66 0.42 1.04
4+ 0.53[∗](#T6-FN1){ref-type="table-fn"} 0.29 0.98 0.57 0.28 1.16 0.79 0.56 1.13 0.65[∗](#T6-FN1){ref-type="table-fn"} 0.44 0.98
Number of grandchildren 1.00 0.96 1.05 1.00 0.95 1.05 1.02[∗](#T6-FN1){ref-type="table-fn"} 1.01 1.05 1.01 0.98 1.03
Age youngest grandchild (ref. \>5) 5--10 1.20 0.71 2.03 1.08 0.58 2.01 0.84 0.62 1.13 0.66[∗](#T6-FN1){ref-type="table-fn"} 0.47 0.94
11+ 1.29 0.54 3.09 1.24 0.44 3.50 0.81 0.52 1.24 0.61 0.37 1.01
Not known 0.54 0.21 1.41 0.41 0.14 1.22 0.77 0.48 1.25 0.58 0.33 1.01
Household composition (ref. alone or just with spouse) With others, not with grandchild 1.06 0.58 1.94 1.28 0.63 2.61 0.83 0.59 1.18 0.78 0.52 1.17
With others, including grandchild 1.32 0.71 2.43 1.41 0.69 2.87 1.14 0.80 1.62 1.01 0.67 1.52
Marital status Married 0.68 0.35 1.33 0.68 0.31 1.50 0.83 0.64 1.08 0.81 0.60 1.10
Education (ref. Low) Mid 1.04 0.54 1.99 0.78 0.56 1.09
High 0.58 0.28 1.22 0.73 0.49 1.10
Paid work Has paid work 0.83 0.48 1.43 1.23 0.88 1.72
Community work Has community work 0.99 0.55 1.76 0.88 0.65 1.19
Functional limitations Mid 0.51 0.25 1.03 0.61[∗∗](#T6-FN1){ref-type="table-fn"} 0.43 0.88
Low 0.37[∗∗](#T6-FN1){ref-type="table-fn"} 0.20 0.71 0.33[∗∗∗](#T6-FN1){ref-type="table-fn"} 0.22 0.51
GDS score 5--9 9.08[∗∗∗](#T6-FN1){ref-type="table-fn"} 4.77 17.28 4.09[∗∗∗](#T6-FN1){ref-type="table-fn"} 2.92 5.73
10+ 7.91[∗∗∗](#T6-FN1){ref-type="table-fn"} 2.85 22.00 15.99[∗∗∗](#T6-FN1){ref-type="table-fn"} 9.50 26.91
*N* 460 460
Notes: Reference categories: Youngest grandchild aged 0--4; lives alone or with spouse only; not married/partnered; low education (\<6 years); no paid work; no community work; high functional limitation; GDS score 0--4.
∗∗∗*p* \< 0.001; ∗∗*p* \< 0.01; ∗*p* \> 0.05.
Provision of material help and mental health {#s16}
============================================
Results from similar analyses of associations between the provision of money or other material goods and indicators of mental health at follow-up are summarised in [Table 7](#T7){ref-type="table"}. Grandfathers who helped their grandchildren in this way had slightly better MCS scores at follow up. Provision of this type of help was also positively associated with good life satisfaction at follow-up in Model 2 but not significantly so when baseline life satisfaction was taken into account (Model 3). Among grandmothers helping grandchildren with money or material goods was negatively associated with good life satisfaction at follow up (OR 0.73, 95% C.I. 0.54, 0.97).
######
Associations between provision of help with money or material goods and mental health outcomes at follow-up: Results from mixed effects regression models.
Grandfathers Grandmothers
------------------- --------- --------------------------------------- ------- -------------- ----- --------------------------------------- ------- ------ -----
Life satisfaction Model 1 1.45 0.98 2.16 457 0.81 0.62 1.05 982
Model 2 1.59[∗](#T7-FN1){ref-type="table-fn"} 1.05 2.39 0.80 0.61 1.05
Model 3 1.36 0.88 2.12 0.73[∗](#T7-FN1){ref-type="table-fn"} 0.54 0.97
β β
MCS Model 1 1.12 −0.26 2.57 456 0.42 −0.52 1.36 973
Model 2 1.40[∗](#T7-FN1){ref-type="table-fn"} 0.01 2.80 0.33 −0.60 1.25
Model 3 1.03[∗](#T7-FN1){ref-type="table-fn"} 0.20 2.05 0.52 −0.34 1.39
OR OR
GDS Model 1 0.86 0.52 1.37 453 0.92 0.70 1.20 975
Model 2 0.70 0.41 1.17 0.96 0.72 1.28
Model 3 0.69 0.39 1.23 0.98 0.71 1.36
Notes: Model 1: Controlling for number of grandchildren; age of youngest grandchild; household composition, marital status. Model 2: Additionally controlling for educational status; functional health status, equivalised household income, paid work; community participation.
Model 3: Additionally controlling for baseline value of relevant mental health outcome.
∗*p* \> 0.05.
Discussion {#s17}
==========
This analysis of data from a prospective study of 2000 older Chileans aged 66--68 firstly highlights the importance of the grandparental role and the contribution made by grandparents to their families. Fertility in Chile has declined very rapidly since the 1970s but those included in this study had relatively large families (six children on average) and 94% of sample members had one or more grandchildren, grandfathers having on average seven grandchildren and grandmothers eight. Forty percent of grandparents lived with at least one grandchild, 75% provided some time help every week and 50% four or more hours of help per week. Our data were restricted to a specific narrow age range which makes it harder to make comparisons with other populations but the level of provision of help seems similar to results reported from reported from middle income Asian countries ([@R9]) but higher than in North America or Europe. [@R10] analysis of data from the 1998 US HRS, for example, found that 7% of 50--80 year old grandparents lived with a grandchild and only 13% of grandmothers and 7% of grandfathers provided 200 hours or more of care to grandchildren a year. The 2004 data on 10 countries included in the Surveys of Health, Ageing and Retirement in Europe analysed by [@R7] showed that 28% of grandparents provided help to grandchildren at least weekly, although these proportions were higher in Southern European countries with, for example, 35% of Spanish grandmothers providing at least weekly help. Results from our study are consistent with other sources on the extent of three generational living arrangements in Latin America ([@R28]).
Despite the importance of the grandparent role in Latin American societies, very few previous studies have investigated either the provision of help by grandparents or implications of such provision for the psychosocial health of grandparents. We found that the provision of help was associated with demographic factors indicative of need or demand for such help, such as living with a grandchild, having a grandchild aged under five and, for grandfathers, having a larger number of grandchildren. Little variation by grandparent characteristics, such as educational level, involvement in other activities or health status, was found, apart from an association between higher incomes and provision of material assistance by grandfathers. Overall, these findings are consistent with research on the salience of family links in Latin American societies and suggest that providing help to grandchildren is viewed as a normative response to family need rather than a matter of personal choice. Gender differences in the associations between providing help to grandchildren and aspects of mental health suggest that this may be particularly true for grandmothers. Thus, we found that for grandfathers providing help to grandchildren was associated with the positive indicators of mental health - better life satisfaction for those providing four or more hours per week of time help and higher scores on the mental health component of the quality of life measure for those providing material help. Among grandmothers, however, helping with money was *negatively* associated with good life satisfaction and the direction of association between providing time help also tended to be negative, although not significantly so. However, grandmothers who helped grandchildren for four or more hours per week had a lower risk of depression at follow-up. Other studies have also found gender differences in mental health effects of helping or looking after grandchildren and attributed these to men\'s more detached role in childrearing and care ([@R2]; [@R27]). This may be particularly important in Latin American societies given strongly differentiated gender roles and responsibilities ([@R20]; [@R23]). For women helping grandchildren may represent an addition to other family responsibilities whereas for men helping may more often be matter of choice. Further investigation, including qualitative studies, might help to see whether the effects of helping grandchildren on different aspects of mental health depend on the extent to which this activity is viewed as 'fun' or 'duty' and how this varies by gender.
Our findings contrast with the results from some US and European studies, in general, indicating some positive effects of active grandparenting on mental health. This difference may reflect the fact that much previous research has focussed only on effects of intensive grandparenting but also the different context and different pathways to co-residence with grandchildren. If, as our results suggest, active grandparent-hood has benefits for older Chileans' mental health it is important to assess whether possible future changes in family and household patterns may have adverse effects on older people, especially as co-residence with grandchildren is currently both common and a strong influence on grandparental help. It would also seem important to examine whether associations are similar among grandparents older than those we consider as in Chile, as elsewhere, mental health problems are more prevalent among the older than the younger old ([@R1]).
This study is, to our knowledge, the first longitudinal investigation of grandparenting and mental health in Chile, a rapidly ageing population with a very high prevalence of late-life depression. The study does however have some limitations. The data relate to a narrow age range and our sample was drawn from low- and middle-income areas of the Santiago Metropolitan area. Chile is a predominantly urbanised society and 34% of the total population live in Santiago ([@R29]), however, our results may not be applicable to those in rural areas, high income Chileans or to grandparents older or younger than those we consider. Response rates at baseline and follow-up were relatively good (76% and 85%, respectively) and there were no significant differences between the mental health of those retained in the study and those lost to follow-up but bias due to non response remains a possibility and confirmation and further investigation of our findings in other studies is warranted.
The Cost-effectiveness Evaluation of a Nutrition supplementation and EXercise programme (CENEX) study was funded by The Wellcome Trust with support from the Ministry of Health, Chile. We are very grateful to all members of the CENEX study team involved in data collection and to all participants in the study.
| {
"pile_set_name": "PubMed Central"
} |
{.601}
| {
"pile_set_name": "PubMed Central"
} |
Arthur Amman, President of Global Strategies for HIV Prevention ( <http://www.globalstrategies.org>), tells this story: "I recently met a physician from southern Africa, engaged in perinatal HIV prevention, whose primary access to information was abstracts posted on the Internet. Based on a single abstract, they had altered their perinatal HIV prevention program from an effective therapy to one with lesser efficacy. Had they read the full text article they would have undoubtedly realized that the study results were based on short-term follow-up, a small pivotal group, incomplete data, and unlikely to be applicable to their country situation. Their decision to alter treatment based solely on the abstract\'s conclusions may have resulted in increased perinatal HIV transmission."
Amman\'s story shows the potentially deadly gap between the information-rich and the information-poor. This gap is not the result of lack of technology or of money, but of a failure of imagination. We live in the most information-rich era of history, when the Internet allows immediate global dissemination of crucial health information, and the interlinking of online information creates an integrated, living body of information---the ultimate vision of which is the semantic Web ( <http://www.w3.org>).
What is preventing such a living Web? For scientific and medical information, two obstacles are vested interests and traditions. The role of copyright, which was developed when the dissemination of work was on paper, is crucial. Initially, applying copyright to medical articles protected both the intellectual investment of authors and the commercial investment of publishers. Authors of scientific articles handed over their copyright to publishers to prevent unauthorized print copying. Thus, the prevention of unauthorized copying helped to disseminate information by providing a valid business model for publishers. But the proliferation of subscription-based medical and scientific journals led to readers having to pay more and more to publishers in order to keep up with current knowledge, and also led to an increasing fragmentation of knowledge between different publishers.
The Internet provides the means to revolutionize publishing in two crucial ways. First, it makes it possible to disseminate health information at no charge to anyone in the world with online access. Although it costs money to peer review, edit, produce, and host an online article, this is a onetime, fixed cost. If research funders are willing to pay this cost, then the published work can be made freely available to all readers worldwide, and there would be no need for journal subscriptions. This is one way of financing an open-access model of publishing ( <http://www.earlham.edu/~peters/fos/overview.htm>).
Second, because the internet allows not just ease of access but ease of reuse, an article\'s usefulness is limited only by a user\'s imagination. To allow this, the traditional role of copyright has to change. Instead of publishers using copyright to restrict use, authors can retain copyright and grant the public the right to creatively reuse their work. Licenses such as those developed by Creative Commons ( <http://creativecommons.org>), which facilitate rather than prohibit reuse, are used by the open-access publishers PLoS (DOI: [10.1371/journal.pbio.0020228](10.1371/journal.pbio.0020228)) and BioMed Central (BMC). The result as Jan Velterop, Director of Open Access at Springer, says is that "copyright can be used for what it is meant to in science, not to make the articles artificially scarce and in the process restrict their distribution, but instead, to ensure that their potential for maximum possible dissemination can be realised" ( <http://www.soros.org/openaccess/scholarly_guide.shtml>).
The potential benefits of such a change are vast. No longer will physicians have to base their practice on half truths. Instead, everyone from patients to policymakers can read for themselves the evidence on which crucial science and health policy decisions are made. One example of a paper with potentially profound public health implications is the first randomized trial of male circumcision to prevent HIV infection (DOI: [10.1371/journal.pmed.0020298](10.1371/journal.pmed.0020298))---having this paper and all related discussions freely available has allowed a lively, informed debate to flourish.
Will poorly funded researchers be excluded from publishing in open-access journals? This concern is addressed by publishers such as PLoS and BMC, who waive fees for authors who cannot pay, and who strictly separate decisions on publication from ability to pay. This is not a radical departure into subsidies, but an accepted part of distributing publishing costs across the scientific community.
Increasingly, funders of research also realize the benefit of an open-access model of publishing. The United Kingdom\'s Wellcome Trust ( <http://www.wellcome.ac.uk/doc_WTD002766.html>) mandates its funded authors to make their work publicly available; the United States National Institutes of Health are encouraging it ( <http://publicaccess.nih.gov>), and governments and funding bodies are signing up to declarations on open access ( <http://www.zim.mpg.de/openaccess-berlin/berlindeclaration.html>).
By regaining control of copyright, the medical and scientific communities could ensure that publishing is no longer driven by the interests of publishers, but rather by the needs of society.
**Citation:** The *PLoS Medicine* Editors (2006) The impact of open access upon public health. PLoS Med 3(5): e252.
This Editorial is co-published in *The Bulletin of the World Health Organization* as part of a theme issue on intellectual property and public health. DOI: 10.2471/BLT.06.032409
[^1]: E-mail: <[email protected]>
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#sec1_1}
============
Parsonage-Turner syndrome (PTS), also known as neuralgic amyotrophy or acute brachial plexus neuritis, is a disorder characterized by the sudden appearance of severe shoulder pain, usually unilateral, followed a few days to a week later by progressive motor weakness. At times, some dysesthesiae and numbness may coexist. The key to establishing the diagnosis is the sequence of symptoms and confirmation by electroneuromyography.
We report the case of an 86-year-old man referred to our internal medicine department for severe left shoulder pain. In this patient, diagnosis was complicated by the occurrence of herpetic lesions on the upper left arm and thoracic region.
Case Presentation {#sec1_2}
=================
An 86-year-old right-handed man presented to his general practitioner with complaints of acute left shoulder pain that awoke him one night. There was no history of previous shoulder trauma. The patient had suffered from a low-grade stage IV non-Hodgkin B-cell lymphoma 4 years earlier that was treated with rituximab chemotherapy alone without radiotherapy, which led to remission. An ultrasound of his shoulder revealed a scapulohumeral periarthritis. An intra-articular corticoid infiltration was performed on an outpatient basis. He was referred to our department 7 days after the onset of the symptoms because of persistent shoulder pain.
At admission, the patient presented a hypertensive crisis that was attributed to the severe shoulder pain and was managed with nitroglycerin and calcium-channel blockers. The patient denied any neck pain, headache, fever, chills or weight loss. Neurologic examination was normal at the time, with preserved sensory and motor functions of the left arm. The patient was oriented and showed no signs of meningeal irritation. Routine laboratory tests were normal.
Two days following hospital admission, i.e. 9 days after the beginning of the left shoulder pain, he presented a vesicular rash over the left D1 and D2 dermatomes. An immunofluorescence test of a skin sample was positive for herpes varicella-zoster virus. A 10-day treatment of intravenous acyclovir in conjunction with prednisone was started. Pain was managed with paracetamol, morphine, pregabalin and physiotherapy. Two days after the vesicular eruption, i.e. 11 days after the beginning of the shoulder pain, he developed a marked weakness of the left arm. This weakness was severe on elbow flexion (1/5), shoulder abduction (1/5) and arm external rotation (1/5), whereas elbow extension and handgrip remained normal (5/5). There was no sensory loss. Deep left bicipital tendon reflex was absent.
Needle electromyography (table [1](#T1){ref-type="table"}) showed evidence of partial denervation of the left biceps, brachioradialis and deltoid muscles, consistent with peripheral motor nerve involvement of the C5 and C6 myotomes. As the rhomboid and serratus anterior muscles were not affected, it is possible that the denervation was related to motor axonal lesions within the upper trunk of the brachial plexus. Neurography showed responses of small but symmetric amplitudes, in particular for the sensory nerve conductions (table [2](#T2){ref-type="table"}). A CT scan of the cervical spine showed mild cervical osteoarthritis without spinal compromise. A thoracic CT scan showed no adenopathy or mass associated with the lymphoma and ruled out a local root or plexus compression. MRI of the brain showed a global cerebral atrophy that was attributed to the age of the patient. MRI of the brachial plexus was not performed.
The diagnosis of troncular, motor neurological lesions due to herpes zoster infection was initially suspected. However, the symptoms, their timing, as well as the neurological territories affected -- upper brachial plexus for the motor deficit, dermatomes of the lower brachial plexus D1 and upper thoracic region D2 for the vesicular rash -- casted doubts on this hypothesis. The symptoms seemed more probably related to PTS rather than to herpes zoster infection of the upper left arm. The clinical presentation, the results of the CT scan of the cervical and thoracic spine, and the motor-evoked potentials made cervical myelopathy unlikely.
Over the course of the following days, the weakness persisted. The patient was transferred to a rehabilitation unit to continue physiotherapy. A second electroneuromyography was performed 24 days after the first, i.e. 6 weeks after the onset of the shoulder pain (tables [1](#T1){ref-type="table"}, [2](#T2){ref-type="table"}). It showed signs of severe denervation of the left biceps, deltoid, supraspinatus and brachioradialis muscles. The neurological findings and normal transcranial motor-evoked potentials were not suggestive of cervical myelopathy.
Subsequently, the patient experienced a general deterioration and sepsis caused by *Staphylococcus aureus* infection. Despite antibiotic therapy, he died 2 months after admission to the rehabilitation unit, i.e. 3 months after the onset of the symptoms.
Discussion {#sec1_3}
==========
The clinical presentation of a herpetic eruption and a motor deficit in the same limb may lead to diagnostic pitfalls. The first diagnosis that may come to mind in this situation is a motor monoparesis due to herpes zoster infection. However, this early impression may be challenged in this situation. Our hypothesis is that the patient presented two different successive conditions. First, left shoulder pain followed 11 days later by left arm weakness, possibly related to PTS, then a vesicular eruption of the left arm caused by varicella-zoster virus reactivation, as confirmed by immunofluorescence. The time elapsed between the eruption and the weakness, only 2 days, is rather short to be explained by a herpes zoster motor monoparesis. Moreover, the distance between the lesions of the sensory ganglions causing the skin vesicular rash (dermatomes D1 and D2) and the motor axonal lesions causing denervation within the upper myotomes of the brachial plexus (C5-C6) also makes the diagnosis of herpes zoster monoparesis unlikely. Sensory findings, if present, are usually much less prominent than motor deficits in PTS. The contrary is true in zoster neuropathies, in which sensory lesions are usual and motor deficits rare (with the notable exception of zoster facial palsy).
Although all weak muscles observed to exhibit denervation belonged to the C5 and C6 myotomes, a radiculopathy appears unlikely in this patient since (1) it should have very seriously affected both C5 and C6 roots simultaneously in order to explain the complete or subcomplete denervation of the muscles of the anterior arm, and, in such a case, it would then probably have affected the spinatus and brachioradialis muscles with a similar severity; (2) a mechanical lesion of the roots would have caused sensory symptoms in the C5 and C6 dermatomes; (3) denervation would have concerned the rhomboid and serratus muscles. The severe, incomplete axonal lesion was thus more likely localized within the upper trunk. Symmetrical sensory neurography (from the C6 dermatome; digit I) without clinical sensory loss (in particular in the C5-C6 dermatomes), points to a lesion that essentially affected the motor axons. This is in line with the possibility of a selective targeted immunologic lesion, and thus in good agreement with the suspected causal mechanism of PTS. We hypothesize, therefore, that the development of herpes zoster was a consecutive phenomenon, possibly favored by the PTS itself, the lymphoma and the intra-articular corticoid infiltration.
PTS is a neuritis of unknown cause affecting the brachial plexus with an overall incidence of 1.64 cases per 100,000 individuals \[[@B1], [@B2], [@B3], [@B4]\]. The classic description of PTS is a condition in which the patient first develops an abrupt, severe and constant unilateral shoulder pain that can extend proximally to the neck but also distally to the upper arm, forearm and hand. Pain lasts from a few hours to several weeks, with an average duration of 4 weeks. The weakness appears within 24 h of the onset of pain in approximately one-third of patients but can take up to 4 weeks to occur \[[@B5]\]. A sensory deficit may also occur but its prevalence varies, depending on studies, from 66% to only a minority of patients \[[@B6], [@B7], [@B8]\]. The upper trunk of the brachial plexus is the most frequent site of the lesion \[[@B6]\]. Clinical observation and electromyogram studies suggest that the lesions are often multifocal within the plexus or in the individual branches \[[@B9], [@B10]\]. Electroneuromyography is helpful to confirm the diagnosis, to rule out other possible causes of painful weakness of the upper limb (radiculopathies, thoracic outlet syndrome or mononeuropathies), and to define the prognosis. Frequently, the syndrome is initially mistaken for an arthropathy of the shoulder, as was the case in this patient. Blood and cerebrospinal fluid analyses are generally unhelpful. A chest radiograph or CT scan can be performed if there is any suspicion of malignancy.
The treatment is based on pain management, NSAIDs, prednisolone and neuroleptics \[[@B11], [@B12]\]. This condition generally carries a good prognosis, as about 75% of all patients recover completely within 2 years, and 89% by the end of the third year. Patients with predominantly lower plexus involvement have a slower recovery. The rate of recurrence varies among studies, from 5 to 26%, and the second episode typically turns out less severe \[[@B4]\].
Herpes zoster can cause segmental paresis with a reported incidence of 3-5% of limb weakness \[[@B13]\]. The paresis affecting the same zone typically occurs 2-3 weeks after the herpetic rash. The prognosis of functional recovery with zoster paresis is good, with a complete or partial recovery reported in 75% of patients within 1 or 2 years \[[@B14]\]. The administration of antiviral therapy may have a positive effect on the course of the paresis. In both PTS and zoster paresis, prevention of muscle contracture and atrophy by physiotherapy is an important part of the treatment \[[@B15]\].
The etiology of PTS is unclear but many possible promoting factors have been proposed including trauma, recent surgery, infection, heavy exercise, immunization and autoimmune conditions. Although the literature reports a direct relationship between herpes zoster and plexus neuritis \[[@B6], [@B15]\], the causal relationship between this viral infection and PTS is not yet established. The hypothesis that a number of neuralgic amyotrophy cases may be caused by unrecognized herpes neuritis is tenable, in particular in case of zoster sine herpete.
Conclusion {#sec1_4}
==========
PTS is a diagnostic challenge because it may mimic several conditions causing pain and weakness around the shoulder, such as rotator cuff tears, acute calcifying tendinitis, impingement syndromes, cervical radiculopathy, tumors of the brachial plexus and spinal cord and compressive nerve injuries. In the reported case, the presence of concomitant shingles was a pitfall, with the potential to mask the underlying PTS and to lead to a misdiagnosis of herpetic brachial plexopathy. A detailed clinical history and physical examination, as well as the help of electroneuromyography were essential to distinguish the neurological structures involved and to ascertain the correct diagnosis.
######
Needle electromyography
Fib. PSW Fase. Description and voluntary activity
------------------------------ ------ ----- ------- ----------------------------------------------------------------
Days 12 and 14 (left side)
Supraspinatus +/− +/− 0 Simple recording-1 MUP firing at 25 Hz
Infraspinatus \+ +/− 0 Increased insertional activity; poor intermediate pattern
Deltoid ++ +/− 0 Intermediate pattern
Biceps brachii \+ \+ 0 No voluntary activity
Brachioradialis \+ \+ 0 Intermediate pattern
Rhomboid major // // // Clinically normal; no needle EMG
Day 38 (left side)
Rhomboid major 0 0 0 Interference pattern; normal size MUPs
Serratus anterior (2 sites) 0 0 0 Interference pattern; normal size MUPs
Supra spinatus ++ ++ 0 Intermediate pattern; high frequencies (25 Hz or more audible)
Deltoid +++ +++ 0 Simple recording; 1 MUP firing at \>25Hz
Biceps brachii ++++ +++ 0 Increased insertional activity; no voluntary activity
Brachialis ++++ +++ 0 No voluntary activity
Brachioradialis ++ ++ 0 Intermediate pattern with frequencies \>25 Hz
Fib. = Fibrillation; PSW = positive sharp wave; Fasc. = fasciculation; MUP = motor unit potential.
0 = No spontaneous activity; +/− = 1 Fib. (or PSW) in 1 region of the muscle; + = 2 Fibs, (or PSWs) in 2 different regions; ++ = \>2 Fibs, (or PSWs) in several regions; +++ = abundant Fibs, (or PSWs) in most regions; ++++ = abundant Fibs, (or PSWs) in all regions with no voluntary activity and no response to electrical nerve stimulation; // = not studied.
######
Nerve conduction studies
Sensory Latency ms Amplitude μV Duration ms Area μV × ms Distance mm Conduction velocity m/s Temperature °C
-------------------- ------------ -------------- ------------- -------------- ------------- ------------------------- ----------------
*Days 12 and 14*
Median
Wrist--digit II L 3.0 8 1.9 5 142 48 34.1
Wrist--digit II R 2.7 6 1.8 6 131 49 33.0
Palm-digit II R 1.5 9 1.5 4 76 51 33.0
Wrist--digit I L 2.3 10 2.2 5 105 46 34.2
Wrist--digit I R 1.8 11 1.6 6 83 45 33.0
Radial
Wrist--digit I L 2.0 2 1.2 2 100 50 33.9
Wrist--digit I R 1.8 2 1.1 2 85 46 33.0
Ulnar
Wrist--digit V R 2.3 10 1.7 6 117 52 33.0
*Day 38*
Median
Wrist--digit II L 2.7 8 2.1 4 135 50 31.9
Ulnar
Wrist--digit V L 2.4 9 2.1 6 125 52 33.0
Motor Latency ms Amplitude mV Duration ms Area mV × ms Distance mm Conduction velocity m/s CMCT (ms)
----------------------- ------------ -------------- ------------- -------------- ------------- ------------------------- -----------
Ulnar
ADM L-wrist 2.5 6.8 7.2 21.4 55
Wrist--elbow below 6.6 6.8 7.5 20.7 220 54
Elbow below--above 8.2 6.4 7.9 21.8 70 44
Elbow above--axilla 12.0 6.0 7.2 19.6
Axilla--elbow 15.1 5.8 7.2 19.6
F waves ADM L--Wrist 29.8
Median
ADM--elbow 9.4 (+)0.1 11.1 (+)0.6
MEP (TMS)
ADM L-cortex R 22.0 4.5 9.5 18.9 6.3
Sensory neurography is antidromic; for amplitude, peak-to-peak amplitude was measured; for duration, negative take-off to positive peak duration was measured. Motor neurography: for the compound muscle action potential, amplitude, duration and area of the negative peak were measured. ADM = Abductor digiti minimi; MEP = motor-evoked potential; TMS = transcranial magnetic stimulation; CMCT = central motor conduction time; (+) = positive peak.
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#s1}
============
Biological responses to radiation are induced in irradiated cells mainly as a result of DNA damage. However, many studies indicate that biological responses to radiation are not always limited to the irradiated cells but can be induced in neighboring unirradiated 'bystander' cells. This phenomenon, often called radiation-induced bystander response, was first described by Nagasawa and Little in 1992 \[[@RRT068C1]\]. So far the range of biological effects demonstrated to be induced in bystander cells via signals from irradiated cells includes sister chromatid exchange \[[@RRT068C1], [@RRT068C2]\], cell death \[[@RRT068C3]--[@RRT068C8]\], chromosomal instability \[[@RRT068C9]\], and mutations \[[@RRT068C10], [@RRT068C11]\]. These findings have had a large impact on radiobiology because they may have important implications for the estimation of risk to human health associated with exposure to low-dose radiation. The risks associated with low-dose ionizing radiation are estimated by extrapolating data obtained after exposure to intermediate doses using a linear non-threshold (LNT) model. The discovery of radiation-induced bystander responses and other non-targeted effects has triggered a dispute over the validity of the LNT model because a non-linear response at a low dose is a characteristic of these phenomena. The controversy over this issue has not been resolved, in part, because the number and location of the radiation-track traversals cannot be monitored or controlled for each cell when a broad radiation field is used, which is often true in low-dose radiation experiments. The microbeam cell irradiation system, which enables observation of cellular responses of individual irradiated and non-irradiated cells equally, is a powerful tool to elucidate the mechanisms underlying the biological responses to low-dose radiation, including bystander responses.
We used a synchrotron X-ray microbeam irradiation system developed at the Photon Factory, High Energy Accelerator Research Organization, KEK \[[@RRT068C12]--[@RRT068C15]\], and found that cell death is more prevalent in cells irradiated with X-ray microbeams when only nuclei, rather than the whole cells, are irradiated \[[@RRT068C16], [@RRT068C17]\]. Furthermore, we recently showed that the biphasic increase in bystander cell death was dose-dependent when nuclei of targeted cells were exposed to X-ray microbeams \[[@RRT068C7], [@RRT068C18]\]. Our findings indicated that cell death, both in irradiated and bystander cells, was modified by the site of energy deposition within the cells.
As a next step, we measured the mutation frequency in bystander cells neighboring those with irradiated nuclei. Because mutations are a prerequisite of carcinogenesis, our results may provide fundamental information for evaluating the carcinogenic risk posed by exposure to low doses of ionizing radiation.
MATERIALS AND METHODS {#s2}
=====================
Cell culture and sample preparation {#s2a}
-----------------------------------
V79 Chinese hamster lung cells were cultured in minimum essential medium-alpha (MEMα; Nacalai Tesque Inc., Nakagyo-ku, Kyoto, Japan) containing 10% fetal bovine serum (FBS; Nichirei Biosciences Inc., Chuo-ku, Tokyo, Japan), 100 U/ml penicillin (Invitrogen Corp, Carlsbad, California, USA), 100 µg/ml streptomycin (Invitrogen), and 15 mM 4-(2-hydroxyethyl)piperazine-1-ethanesulfonic acid (HEPES; Nacalai Tesque) and then incubated in a humidified incubator maintained at 37°C in an atmosphere containing 5% CO~2~. To irradiate cells with microbeams, 1.0 × 10^5^ V79 cells were seeded on custom-designed polypropylene-based dishes (34 mm in diameter), the bottoms of which were composed of a 3-µm polypropylene film (Toray Industries Inc., Chuo-ku, Tokyo, Japan), and incubated overnight. Before X-ray irradiation, the cell nuclei were stained with a 2 µM solution of Hoechst 33258 (Dojindo Molecular Technologies Inc., Kamimashiki-gun, Kumamoto, Japan) for 1 h. At the time of irradiation, the Hoechst solution was replaced with 5 ml of fresh medium.
Microbeam irradiation {#s2b}
---------------------
Monochromatic X-ray microbeam irradiation was performed using the synchrotron X-ray microbeam irradiation system installed at the BL-27B station in the Photon Factory \[[@RRT068C12]--[@RRT068C15]\]. The procedures for microbeam irradiation and dosimetry have been previously described \[[@RRT068C7], [@RRT068C17]\]. To ensure that the irradiation for initial stimulation was similar to that used in previous studies \[[@RRT068C7], [@RRT068C18]\], nuclei of five isolated single cells located at the center of each dish were selected as targets. The positions of these nuclei, defined as the center of mass of Hoechst33258-stained nuclei images, were stored in the controlling computer of the system. We irradiated the five targeted cell nuclei with 10 × 10 µm^2^ 5.35 keV X-ray beams, and the exposure rates were 8.5 × 10^−3^ ± 3.4 × 10^−5^ C/kg/s (1.0 × 10^4^ ± 4.2 × 10^1^ photons/s in 10 × 10 µm^2^ beam). In our study, the 'nuclear-averaged dose', at which the absorbed energy is divided by the mass of nucleus as described in Maeda *et al.* \[[@RRT068C17]\], was used as a measure of radiation doses and the dose rate was 1.8 × 10^−1^ ± 7.3 × 10^−4^ Gy/s.
Determination of bystander cell survival {#s2c}
----------------------------------------
The fraction of bystander cells that survived was measured using a colony-formation assay. After irradiation, the culture medium was removed, and the cells were washed twice with phosphate buffered saline (PBS). Immediately thereafter, 2 ml of fresh medium was added to the dishes, and the cells were cultured for 3 h. Next the cells were harvested using a trypsin-ethylene diamine tetraacetic acid (EDTA) solution (0.05% trypsin, 0.53 mM EDTA•4Na; Invitrogen); the harvested cells were diluted and plated in a 100-mm cell culture dish at approximately 150 viable cells per dish. After incubation for 6 days, the cells were fixed with HC Tissue Fixative MB (Amresco Inc., Solon, Ohio, USA) for 25 min at room temperature and rinsed twice with PBS. Following that the cells were stained with 1% methylene blue (Wako Pure Chemical Industries Ltd, Chuo-ku, Osaka, Japan) solution. Colonies containing more than 50 cells were counted as survivors.
*HPRT* mutation assay {#s2d}
---------------------
The *HPRT* mutation assay is a method commonly used to study the genetic changes and genomic instability \[[@RRT068C19], [@RRT068C20]\]. After irradiation, the culture medium was removed, and the cells were washed twice with PBS. Immediately thereafter, 2 ml of fresh medium was added to the dishes, and the cells were cultured for 3 h. Cells were harvested by trypsinization and transferred to a T-75 cell culture flask containing fresh medium. Cells were maintained for 8 days and were subcultivated every 2 days to allow for phenotypic expression. Then, 1 × 10^6^ cells were harvested and seeded onto 100-mm cell culture dishes with fresh medium containing 10 µg/ml 6-thioguanine (Wako) and incubated for 6 days. Cells in dishes were fixed and stained using the same method described above for the colony-formation assay, and the colonies (i.e. *HPRT* mutants) were scored. The mutation frequency was expressed as the number of resistant colonies divided by the total number of viable cells at the time of selection.
Cell culture with NO scavenger after irradiation {#s2e}
------------------------------------------------
2-(4-Carboxyphenyl)-4,4,5,5-tetramethylimidazoline-1-oxyl-3-oxide, sodium salt (carboxy-PTIO; Dojindo Molecular Technologies Inc.) is a specific scavenger of NO \[21, 22\]. Directly after irradiation, the cells were incubated with medium containing carboxy-PTIO instead of normal fresh medium. During the clonogenic assays for the measurement of surviving fractions and of mutation frequencies, cells were also incubated with medium containing carboxy-PTIO instead of normal fresh medium. The concentration of carboxy-PTIO in the medium was set to 20 µM, because the concentration of NO in the medium was not expected to exceed 20 µM after irradiation \[[@RRT068C7]\], as indicated by studies in which the concentration of NO~2~^−^, an oxidization product of NO, in the medium after irradiation was measured with Griess reagent \[[@RRT068C23]\].
Statistical analysis {#s2f}
--------------------
Statistical analysis was performed on the data obtained from at least three independent experiments. All the results are expressed as means ± standard error (SE). Significant levels were assessed using Student\'s *t* test. Analysis of the correlations between bystander cell death and mutation frequency in the bystander cells was assessed using analysis of variance (ANOVA). A probability (*P*) value \< 0.05 was considered to indicate statistical significance.
RESULTS {#s3}
=======
Determination of the incubation period for the assays of bystander responses {#s3a}
----------------------------------------------------------------------------
We first determined the incubation period for trypsinization to harvest the cells for the colony formation assay. Cell nuclei (*n* = 5) located in the center of a dish were irradiated with 1 Gy of 10 × 10 µm^2^ square 5.35 keV X-ray beams and incubated for 0--6 h. Then, the surviving fractions of the cell populations on the dishes were determined using the colony-formation assay. We set the incubation period as 3 h, because the decrease in surviving fractions reached a plateau 3 h after irradiation (Fig. [1](#RRT068F1){ref-type="fig"}). Fig. 1The surviving fraction of bystander V79 cells is plotted as a function of the time after irradiation. Data were taken from more than three independent experiments. The error bars represent standard errors (SEs).
Bystander cell death induced by microbeam-irradiated V79 cells {#s3b}
--------------------------------------------------------------
We measured the dose--survival relationship in bystander V79 cells (Fig. [2](#RRT068F2){ref-type="fig"}). We observed a dose-dependent biphasic increase in the death of bystander cells. The surviving fraction decreased to 0.87 ± 0.015 when nuclei were irradiated with a nuclear-averaged dose of 1 Gy; however, at higher doses the surviving fraction was stable at approximately 0.94. Because the fraction of irradiated cells in the dish was \< 0.5 × 10^−4^, it is likely that the decrease in the surviving fraction was the result of bystander responses. Our previous studies showed the same type of biphasic dose--response effect as that observed in this study during the death of bystander cells \[7, 8, 18\]. Fig. 2.The surviving fractions of bystander V79 cells are plotted as a function of the nuclear-averaged dose in the irradiated cells. Cells were incubated with (open circles) or without (closed circles) the nitric oxide (NO)-specific scavenger, 2-(4-carboxyphenyl)-4,4,5,5-tetramethylimidazoline-1-oxyl-3-oxide (carboxy-PTIO), commencing immediately after irradiation. Data were taken from at least three independent experiments. The error bars represent standard errors (SEs). An asterisk indicates *P* \< 0.05, and double asterisks indicate *P* \< 0.01.
In the previous study, 2.0 × 10^3^ individual V79 cells were seeded on dishes, and five cells at the center of the dish were irradiated \[[@RRT068C7], [@RRT068C18]\]. On the other hand, in the present study, 1.0 × 10^5^ cells were seeded on dishes. Although, the shape of the dose--survival curves and the magnitude of bystander cell death were almost equal in both studies \[[@RRT068C7], [@RRT068C18]\], the fraction of irradiated cells among the cell population was 0.25% and 0.005%, respectively. These results indicate that initial small stimulation (i.e. irradiation to only five cell nuclei in the cell population) is sufficient to saturate the bystander cell death.
*HPRT* mutation in the bystander cell population {#s3c}
------------------------------------------------
The dose--response relationship between radiation dose and mutation frequency in the bystander cell population was determined using the *HPRT* mutation assay (Fig. [3](#RRT068F3){ref-type="fig"}). The background mutation frequency in the control, non-irradiated cells, was 2.6 × 10^−5^ ± 1.3 × 10^−6^. The mutation frequency in bystander cells decreased significantly (*P* \< 0.01) to 5.3 × 10^−6^ ± 1.3 × 10^−6^ when five target nuclei were irradiated with a nuclear-averaged dose of 1 Gy; however, at higher doses the mutation frequency returned to background levels. The biphasic decrease in mutation frequency was similar to the biphasic decrease in bystander cell death. Fig. 3*HPRT* mutation frequencies in bystander cells are plotted as a function of the nuclear-averaged dose in the irradiated cells.Cells were incubated with (open squares) or without (closed squares) the nitric oxide (NO)-specific scavenger, 2-(4-carboxyphenyl)-4,4,5,5-tetramethylimidazoline-1-oxyl-3-oxide (carboxy-PTIO), commencing immediately after irradiation. Data were taken from at least three independent experiments. The error bars represent standard errors (SEs). An asterisk indicates *P* \< 0.05, and double asterisks indicate *P* \< 0.01.
The effects of NO on bystander cell death and mutagenesis in the population of bystander cells {#s3d}
----------------------------------------------------------------------------------------------
Recently, we showed that NO is a principal mediator of bystander cell death \[[@RRT068C7]\]. Therefore, we investigated the role of NO in bystander cell death and mutations. Treatment with 20 µM carboxy-PTIO, a specific scavenger of NO \[[@RRT068C21], [@RRT068C22]\], was not cytotoxic to V79 cells (Table [1](#RRT068TB1){ref-type="table"}). The effects of carboxy-PTIO in post-irradiation incubation are shown in Figs [2](#RRT068F2){ref-type="fig"} and [3](#RRT068F3){ref-type="fig"}. The dose-dependent biphasic increase in the death of bystander cells was not observed when the cells were incubated with carboxy-PTIO (Fig. [2](#RRT068F2){ref-type="fig"}). Furthermore, the dose-dependent biphasic decrease in mutation frequency was not observed when the cells were incubated with carboxy-PTIO (Fig. [3](#RRT068F3){ref-type="fig"}). These results clearly show that NO plays an important role, not only in the induction of death of bystander cells, but also in the suppression of spontaneous mutagenesis in bystander cells. Table 1.Cytotoxicity of incubating V79 cells with 20 µM 2-(4-carboxyphenyl)-4,4,5,5-tetramethylimidazoline-1-oxyl-3-oxide (carboxy-PTIO)carboxy-PTIO (20 µM)--+Colony-forming efficiency0.92 (±0.070)0.91 (±0.050)Mutation frequency2.6 × 10^−5^ (±7.8 × 10^−6^)2.6 × 10^−5^ (±6.0 × 10^−6^)
DISCUSSION {#s4}
==========
Our findings showed a biphasic dose--response relationship between the mutation frequency in bystander cells and irradiation dose when the nuclei of targeted cells were exposed to X-ray microbeams. Previous studies on bystander responses, especially those investigating mutations in bystander cells, mostly focused on the effects caused by irradiation with low fluences of α-particles. For example, Nagasawa and Little found an unexpectedly high frequency of *HPRT* mutations in CHO cells exposed to very low fluences of α-particles; they concluded that the mutations occurred in unirradiated cells via bystander responses \[[@RRT068C10]\]. Similarly, Zhou *et al.* reported that when 20% of human-hamster hybrid (A~L~) cells were irradiated with 20 α-particles using charged particle microbeams, the frequency of mutations in the human chromosome 11 in the other 80% of the cells (i.e. bystander cells) quadrupled relative to background levels \[[@RRT068C11]\]. Our findings, unlike those from previous studies, showed that the *HPRT* mutation frequency in bystander V79 cells decreased at radiation doses around 1 Gy (Fig. [3](#RRT068F3){ref-type="fig"}). The mutation frequency significantly decreased (*P* \< 0.01) from 2.6 × 10^−5^ (background level) to 5.3 × 10^−6^ at a nuclear-averaged dose of approximately 1 Gy; however, at higher doses the mutation frequency returned to the background level. Interestingly, this dose--response effect is quite similar to that observed in the death of bystander cells (Fig. [2](#RRT068F2){ref-type="fig"}). ANOVA indicated a significant correlation (*P* \< 0.05) between bystander cell death and *HPRT* mutation frequency. The similarity of these behaviors indicates the possibility that bystander cell death and mutations in bystander cells are not independent phenomena.
Some groups reported a similar dose--response in neoplastic transformation after exposure to low-dose radiation. Although the phenomena described in those studies are not bystander responses, the discussions in those reports might be useful to help consider the mechanisms underlying our results. Azzam *et al.* reported that low-dose γ-ray irradiation reduces the risk of neoplastic transformation from the spontaneous level to one-third or one-fourth of that level \[[@RRT068C24]\]. They proposed that exposure to low doses of radiation resulted in an increased capacity for the error-free repair of DNA double-strand breaks \[[@RRT068C24]\]. In addition, Redpath and his colleagues reported that low-dose γ-ray or X-ray irradiation protected a human hybrid cell line against spontaneous neoplastic transformation. Interestingly, the dose--response in their work was biphasic, similar to that observed in our *HPRT* mutation study. They suggested that a reduction in the oxidative stress, possibly as a consequence of the upregulation of antioxidants by low doses of irradiation, might be a mechanism mediating those phenomena \[[@RRT068C25], [@RRT068C26]\]. Because in our study the viability of bystander cells decreased in the same dose range as that of the reduction of *HPRT* mutation frequency, the activation of antioxidant functions, rather than the upregulation of DNA repair capability, may be related to the suppression of mutagenesis. The oxidative damage of nucleotides within DNA or precursor pools that is caused by reactive oxygen species (ROS) is thought to play an important role in spontaneous mutations. Intracellular levels of ROS are persistently high in genetically unstable cells \[[@RRT068C27]--[@RRT068C29]\], and clones of those unstable cells secrete factors such as cytokines and persistent free radicals that contribute to the perpetuation of the unstable phenotype \[[@RRT068C30]\]. Bystander responses might suppress the secretion of those factors and/or elevate the antioxidant functions in the bystander cells population.
To date, two major classes of intercellular signaling events, direct cell-to-cell contact via gap junctions \[[@RRT068C11], [@RRT068C31]--[@RRT068C33]\] and indirect communication via secreted factors \[[@RRT068C34], [@RRT068C35]\], are known to be involved in the induction of bystander responses. Hou *et al.* reported that most mutations induced in bystander cells were point mutations and suggested that these mutations may have been caused by ROS \[[@RRT068C36]\]. However, Zhou *et al.* reported that oxidative stress or hydroxyl radicals, which have a very short half-life, were not chief mediators of mutations in bystander cells \[[@RRT068C11]\]; they showed that the inhibition of gap-junction communication suppressed the mutagenesis in bystander cells \[[@RRT068C11], [@RRT068C37]\]. V79 cells, the cells used in this study, exhibit some degree of gap-junction intercellular communication \[[@RRT068C38], [@RRT068C39]\]. However, gap junction-mediated bystander responses may have been rare in our study because we seeded the cells on the irradiation dish at low densities, and most cells did not contact other cells during the short incubation periods after irradiation. Therefore, it is likely that the observed bystander responses were due to indirect communication mediated by secreted factors. Certain bystander signaling factors secreted from irradiated cells, such as NO \[[@RRT068C7], [@RRT068C8], [@RRT068C40]--[@RRT068C45]\] and transforming growth factor-β1 (TGF-β1) \[[@RRT068C46]--[@RRT068C48]\], apparently act as mediators between irradiated and bystander cells. Recently, we showed that NO is a principal mediator of bystander cell death; the addition of carboxy-PTIO, a specific scavenger of NO, to the culture medium suppresses the biphasic bystander cell death \[[@RRT068C7]\]. The dose-dependent biphasic increase in the death of bystander cells was suppressed by scavenging of the NO in a similar manner (Fig. [2](#RRT068F2){ref-type="fig"}). Interestingly, scavenging of the NO also suppresses the dose-dependent biphasic decrease of the mutation frequency in bystander cells (Fig. [3](#RRT068F3){ref-type="fig"}). These results clearly indicate an increase in NO-mediated bystander cell death participated in mechanisms that suppressed mutagenesis in bystander cells. In many cases \[[@RRT068C4]--[@RRT068C8], [@RRT068C49]\], approximately 5--20% of the bystander cells were killed by radiation-induced bystander responses. However, it is not clear why bystander responses induce cell death at those frequencies. Recently, Egashira *et al.* reported that exposure to NO causes mitochondrial degeneration and subsequent cell killing in cells that have low antioxidative functions \[[@RRT068C50]\]. Because NO is a major mediator of bystander cell death, the cells that are genetically unstable because of defects in their antioxidative activity might be selectively killed by bystander responses. Thus, the secretion of factors that contributed to the perpetuation of the unstable phenotype may have been suppressed, and the antioxidant activity in surviving cells may have been increased; therefore, mutagenesis may have been suppressed in the bystander cells (Fig. [3](#RRT068F3){ref-type="fig"}). Our group has reported that the biphasic NO-mediated bystander cell death was induced by X-ray microbeams also in normal human fibroblast WI-38 cells \[8\]. A reduction in the mutation frequency may occur in normal human bystander cells if NO-mediated bystander cell death selectively kills genetically unstable cells.
Radiation-induced bystander responses are generally thought to increase the risk of low-dose radiation to human health because many cytotoxic phenomena are observed in bystander cells. However, previous studies, as well as this study, have reported the presence of a radiation-protective bystander phenomena \[[@RRT068C48], [@RRT068C51], [@RRT068C52]\]. These phenomena are part of the cellular homeostatic response and it is difficult to categorize these effects as beneficial or harmful. To assess the effects of radiation-induced bystander responses on human health precisely, it is necessary not only to elucidate the mechanisms mediating individual endpoints, but also to understand the biological significances for the one living system. Our results indicated that radiation-induced bystander responses can enhance selective cell killing of genetically unstable cells in the bystander cell population, and this selective cell death might act as a protective mechanism that competes with increases in non-lethal and potentially carcinogenic damage (e.g. mutations). These hypothetical protective effects may provide an adaptive advantage to the organism.
FUNDING {#s5}
=======
This work was supported in part by a Grant-in-Aid for Young Scientists (A) (21681006) from the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan, and a Grant-in-Aid for Young Scientists (B) (23710076) from Japan Society for the Promotion of Science (JSPS).
The authors would like to express their sincere gratitude to Drs Mika Maeda, Hiroshi Maezawa and Kensuke Otsuka for their valuable discussions concerning this study. We thank Ms Masako Mizuno for her technical assistance. The study was approved by the Photon Factory Program Advisory Committee (proposal numbers 2008G624 and 2010G040).
| {
"pile_set_name": "PubMed Central"
} |
{#sp1 .438}
{#sp2 .439}
{#sp3 .440}
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1}
===============
In 2010 there were estimated 70,530 new cases of bladder cancer in the United States, and over 500,000 current survivors \[[@B1], [@B2]\]. The standard of care in the United States for muscle-invasive bladder cancer is radical cystoprostatectomy for men and anterior exenteration for women. There is also evidence that earlier cystectomy in high-risk superficial bladder cancer improves long-term survival \[[@B3]\]. Options for urinary diversion after cystectomy include noncontinent conduits, continent cutaneous diversions, and orthotopic bladder substitutes. With the improvement of surgical technique in recent years, continent diversions and orthotopic bladder substitutes have been shown to have similar perioperative complication rates, cancer control, and morbidity \[[@B4], [@B5]\].
With the proliferation of urinary diversion options for bladder cancer that have comparable cancer control and complication rates, quality of life becomes an important factor to consider. Health-related quality of life (HRQOL) refers to the physical, psychological, and social domains of health that are influenced by a person\'s experiences, beliefs, expectations, and perceptions \[[@B6]\]. The task of translating the subjective components of health into a quantitative value is a complex one, drawing from a variety of fields in the social sciences. HRQOL is measured with questions, or items, whose answers can be converted to numerical scores. Many questionnaires, or instruments, have been developed to assess the various aspects of HRQOL. An ideal instrument should be valid (measures what it reports to measure), reliable (able to give the same result on several occasions given stable disease), and responsive (able to detect true but clinically meaningful changes). The process of psychometric and clinical testing is beyond the scope of this article, but an instrument can only be correctly considered "validated" if it has undergone the established rigors of developmental testing so that its validity, reliability, and responsiveness can be described quantitatively \[[@B7]\].
This article reviews the methods for defining HRQOL, the challenges in measuring HRQOL in bladder cancer and the existing literature comparing HRQOL after various methods of urinary diversion.
2. Measuring HRQOL {#sec2}
==================
HRQOL instruments can be either generic or disease-specific. Generic instruments are applicable to all patients regardless of their illness. They may address issues like bodily pain, energy or fatigue, and limitations in physical activity that are common to many disease processes. Examples of generic instruments are the Medical Outcomes Study 36-Item Short Form (SF-36) \[[@B8]\] and the Sickness Impact Profile (SIP) \[[@B9]\].
There are multiple instruments that assess cancer-specific HRQOL, including the European Organization for Research and Treatment of Cancer-QOL (EORTC-QLQ-C30) \[[@B10]\] and the Functional Assessment of Cancer Therapy general form (FACT-G) \[[@B11]\]. The current version of the EORTC-QLQ-C30 contains 30 items that are grouped into five functional scales (physical, role, emotional, cognitive, and social), three symptoms scales (fatigue, nausea and vomiting, and pain), and an overall HRQOL scale. The broadness of this instrument makes it generally applicable to all cancer states, but it lacks the specificity to address issues that may be unique to a particular type of cancer \[[@B12]\].
Disease-specific instruments focus on the issues that are relevant to patients with a particular disease state. Some disease-specific instruments are developed from validated generic instruments, with additional items added to make them more specific to the disease of interest. For example, 12 items focusing on urinary tract symptoms, intestinal symptoms, sexual symptoms, and stoma concerns were added to the FACT-G to create the FACT-BL, designed to address HRQOL in bladder cancer. Likewise, 17 items were added to the FACT-G questionnaire to create the Vanderbilt Cystectomy Index (FACT-VCI), which has been separately validated to measure HRQOL following radical cystectomy and urinary diversion for bladder cancer \[[@B13]\]. The European Organization for Research and Treatment of Cancer (EORTC) has also added items to the EORTC-QLQ-C30 core questionnaire to make it specific for superficial and invasive bladder cancer, and is in the process of validating these instruments \[[@B14]\]. More recently, the Bladder Cancer Index (BCI) has been developed and validated to assess health outcomes specific to localized bladder cancer \[[@B15]\]. Refer to [Table 1](#tab1){ref-type="table"} for a comparison of bladder cancer specific HRQOL instruments.
3. Challenges of Measuring HRQOL in Bladder Cancer {#sec3}
==================================================
Measuring HRQOL in bladder cancer has its unique difficulties. All of the currently available bladder cancer-specific instruments contain items evaluating the urinary domain, but the items mostly address general problems with urinary control (e.g., "I have trouble controlling my urine" or "I urinate more frequently than usual"). Leaking from a stoma, however, may be a very different experience than leaking per urethra. Different diversions also come with a different set of potential side effects (e.g., metabolic derangements) and impact on body image. It is difficult to account for all these factors in a single questionnaire that can be applied to all patients with bladder cancer.
Consequently, many of the bladder cancer-specific instruments target an even narrower population of bladder cancer patients. As shown in [Table 1](#tab1){ref-type="table"}, the EORTC QLQ-BLS24 is targeted to superficial bladder cancer and will include items related to the bother of repeated cystoscopies, whereas the FACT-VCI was validated to target patients who undergo radical cystectomy for bladder cancer.
In addition, sexual function is an important domain when discussing the impact of bladder cancer and its treatment. Since men and women experience different issues related to sexual function, it is challenging to create a questionnaire that is applicable to both sexes. The BCI sexual domain asks gender-neutral questions addressing sexual desire and arousal, ability to climax, and genital sensation. But men may experience issues with erection and ejaculation, while women may have problems with vaginal lubrication and dyspareunia, which even a disease-specific instrument like the BCI would not be specific enough to distinguish \[[@B16]\].
On the other hand, as instruments are tailored more specifically to the disease of interest, they may be less likely to detect unanticipated effects of the disease. For example, the BCI contains 36 items distributed among the 3 primary domains that are expected to be relevant in bladder cancer: urinary, bowel, and sexual. If a treatment for localized bladder cancer were to unexpectedly cause a neurologic side effect, this disease-specific instrument would not be sensitive enough to detect this change.
4. Current Literature: HRQOL after Cystectomy {#sec4}
=============================================
In 2005, Porter and Penson published a systematic review examining HRQOL outcomes among different types of urinary diversion after radical cystectomy \[[@B17]\]. After excluding studies that did not specifically address bladder cancer patients, studies that included radiotherapy as treatment, and studies that did not compare at least 2 types of urinary diversion (ileal conduit, orthotopic neobladder, or continent cutaneous urinary reservoir), 15 studies were appropriate for analysis. The review found no randomized, controlled studies. Only 1 study was performed prospectively and included preoperative baseline measurements \[[@B18]\]. The other studies were cross-sectional, consisting of a single-mailed or clinic-administered instrument. Of 15 studies, 10 (67%) used a previously validated health-related quality of life instrument, while 10 (67%) used an instrument developed by the investigators without previous validation, either in conjunction with a validated instrument or as the sole measure of quality of life. The authors concluded that the available data did not conclusively show that any form of urinary diversion was superior to another in terms of HRQOL.
They also noted that limitations to the literature included the lack of baseline assessment before cystectomy, the lack of longitudinal data to evaluate the impact over time, and the lack of a bladder cancer-specific and validated instrument to measure quality of life. Despite the inability to draw conclusions regarding the superiority of any method of diversion after cystectomy, some common patterns were identified across the studies that may help future research. Three studies indicated that urinary leakage was more of a problem with conduit diversion than with continent diversion \[[@B19]--[@B21]\]. Three studies indicated that patients with neobladder or continent reservoir were more likely to travel than patients with conduit diversion \[[@B22]--[@B24]\]. Two studies indicated that patients with continent diversions scored better on social function domains than those with conduit diversions \[[@B25], [@B26]\].
In 2005, Gerharz et al. published a review that rated studies by levels of evidence and grades of recommendations set out by the International Consultation on Urological Diseases modification of the Oxford Centre for Evidence-Based Medicine \[[@B27]\]. They identified no studies with level I evidence and also concluded that there was no evidence to claim that continent reconstruction provides better quality of life than conduit diversion.
Since the publication of these reviews, there have been a handful of studies published in the last 5 years that evaluate HRQOL after cystectomy in bladder cancer. In 2007 Gilbert et al. used the recently validated bladder cancer-specific instrument, the Bladder Cancer Index (BCI), to assess HRQOL in patients treated with a variety of interventions, including cystectomy and endoscopic procedures \[[@B28]\]. They identified all bladder cancer patients in an institutional bladder cancer database from a high-volume tertiary referral center. From this population, 315 (45%) completed the BCI and were included in the study. There was no pretreatment HRQOL assessment. The cases were stratified into 4 groups according to treatment type: native bladder without intravesical treatment, native bladder with intravesical treatment, cystectomy with ileal conduit, and cystectomy with orthotopic neobladder. They found that cystectomy groups scored relatively lower than native bladder groups in all the function and bother domains (urinary, bowel, sexual). Interestingly, the cystectomy with orthotopic neobladder group scored significantly lower in urinary function scores than the other 3 groups, challenging the commonly held belief that continent urinary diversion offers improved HRQOL outcomes compared to incontinent diversion. However, despite the difference in functional scores, the urinary bother scores did not differ between the cystectomy groups, leading the authors to speculate that perhaps neobladder patients are willing to compromise physiologic urinary function in exchange for anatomic urinary function.
Two recent prospective studies looked at body image and its impact of HRQOL. In 2009 Somani et al. published the results of a study that prospectively evaluated 32 patients undergoing radical cystectomy for bladder cancer; 29 ileal conduit diversions and 3 neobladder replacements were performed \[[@B29]\]. The study evaluated HRQOL preoperatively and 9--12 months after cystectomy and urinary diversion using the SEIQoL-DW, a validated generic instrument administered by an independent researcher, and the EORTC QLQ-C30. Patients identified several nonhealth-related determinants of quality of life (family, relationships, finance) that were most important. None of the patients mentioned body image as an important determinant of quality of life. When asked specifically about appearance, 76% in the ileal conduit diversion group said that it was quite or very important but only 14% thought that it would change significantly after the operation.
In a separate prospective study, Hedgepeth et al. evaluated changes in body image and quality of life in 139 patients undergoing neobladder replacement after cystectomy and 85 patients undergoing ileal conduit diversion, compared to a reference group of patients who had cystoscopy \[[@B30]\]. HRQOL outcomes were measured using the Bladder Cancer Index (BCI), which consists of 36 questions covering urinary, bowel, and sexual health domains. Function is measured by items describing frequency and control of bodily or stoma function, and bother is measured by items describing how much a patient is troubled by their symptoms. Participants were assessed preoperatively, as well as at several time points in the 8 years after surgery. The authors concluded that patients with continent urinary diversions experience greater functional urinary impairment with little difference in overall bother. Body image for both neobladder and ileal conduit diversion patients worsened immediately after surgery but improved in a linear fashion over time. Ileal conduit diversion patients eventually returned to baseline body image levels, with no significant difference from the cystoscopy group in the long-term. Neobladder patients never returned to baseline body image levels.
Acknowledging that the sociocultural milieu in which patients live can influence their perception of health and illness, studies have also begun to explore the role of cultural differences in HRQOL. In a 2007 prospective study comparing male patients from Sweden and Egypt undergoing radical cystectomy and orthotopic neobladder substitution for locally advanced bladder cancer, Månsson et al. reported that Egyptian men were more likely to express depression and anxiety on the Hospital Anxiety and Depression Scale (HADS), a validated generic instrument to detect mood disorders, as well as score lower on the FACT-BL than their Swedish counterparts \[[@B31]\]. The authors did warn against pitfalls in comparing patients from different cultural groups, such as the inherent demographic differences that may bias the results, as well as the difficulty of preserving conceptual and linguistic equivalence when translating HRQOL questions into a different language. However, this represents an important step in acknowledging that patient-assessed outcomes differ in patients from different sociocultural backgrounds.
Investigators are also beginning to evaluate HRQOL in patients undergoing bladder preservation therapy in the setting of invasive bladder cancer. Lagrange et al. in a prospective, multicenter study published this year evaluated 53 patients with muscle-invasive disease in a Phase II trial who underwent pelvic radiation and concurrent chemotherapy with cisplatin and 5-fluorouracil \[[@B32]\]. Subjects completed the EORTC QLQ-C30 pretreatment, as well as at several time points post treatment. The results indicated that 70% maintained good scores for bladder function for 24 months, at which point deterioration in symptoms, such as frequency, pain, and urinary control, was seen. This study is limited by the lack of a disease-specific instrument, and there are not yet any studies comparing HRQOL in bladder preservation therapy with the standard of care, cystectomy with urinary diversion.
5. Conclusion {#sec5}
=============
Improvements have been made in recent years in the field of HRQOL research in bladder cancer. More prospective studies are being published with HRQOL instruments administered prior to treatment as well as at multiple points in time after treatment. However, the vast majority of available studies are still retrospective and cross-sectional. This study design is less able to distinguish differences in health status that may simply reflect underlying preoperative differences.
Another major limitation to existing literature is the lack of an available bladder cancer-specific, validated instrument to measure health-related quality of life. With the development of the Bladder Cancer Index, as well as the growing popularity of the FACT-BL and the bladder cancer module of the EORTC QLQ-C30, researchers should be better equipped to assess HRQOL in bladder cancer patients.
Even with the improvements of the last 5 years, there is still not conclusive evidence that any type of urinary diversion offers superior HRQOL outcomes. In fact, though postcystectomy patients commonly experience urinary and sexual problems, HRQOL remains good irrespective of the method of urinary diversion. It has been speculated that the reason for this finding is that patients who receive thorough preoperative counseling end up choosing the method that is most suitable for them and are well prepared for the adjustment after surgery \[[@B27]\]. It seems that patient education, careful evaluation of each patient\'s unique clinical and psychosocial situation, and active patient participation in treatment decisions remain crucial to good postoperative quality of life.
Although progress has been made in evaluating HRQOL in postcystectomy bladder cancer patients, there is still a need for well-designed, prospective studies. In light of recent evidence demonstrating that despite marked differences in functional outcomes, body image and ultimately overall bother are similar between methods of urinary diversion, it would be interesting to conduct studies looking at the way preoperative counseling and patient education impact HRQOL post treatment.
This material is the result of work supported in part by resources from the VA Puget Sound Health Care System, Seattle, Washington.
######
Comparison of bladder cancer specific HRQOL instruments.
*Target population* *No of items* *Domains* *Validation*
---------------------------- ------------------------------------------ --------------- ----------- -------------- ---------------------------------------------------------------------------------------------------------- ---------- ---------------------------------------------------------
Bladder Cancer Index (BCI) Localized bladder cancer 36 No Yes (10) Yes (14) Yes (12) Test-retest 0.73--0.95, internal consistency 0.77--0.91
EORTC QLQ-BLS24 Superficial bladder cancer (Ta, T1, CIS) 24 No Yes Yes (including items related to intravesical treatment: fever, malaise, bother of repeated cystoscopies) Yes Pending
EORTC QLQ-BLM30 Invasive bladder cancer (T2, T3, T4a/b) 30 No Yes Yes (including items related to urostomy probems, catheterization) Yes Pending
FACT-BL Bladder cancer 45 FACT-G Yes (4) Yes (3) + Stoma (2) Yes (2) FACT-G: test-retest 0.82--0.92
FACT-VCI Bladder cancer, after radical cystectomy 45 FACT-G Yes (4) Yes (4) Yes (4) Test-retest 0.79, internal consistency 0.83--0.85
[^1]: Academic Editor: Maxwell V. Meng
| {
"pile_set_name": "PubMed Central"
} |
The content published in Cureus is the result of clinical experience and/or research by independent individuals or organizations. Cureus is not responsible for the scientific accuracy or reliability of data or conclusions published herein. All content published within Cureus is intended only for educational, research and reference purposes. Additionally, articles published within Cureus should not be deemed a suitable substitute for the advice of a qualified health care professional. Do not disregard or avoid professional medical advice due to content published within Cureus.
Introduction
============
Percutaneous coronary intervention with stent implantation is a successful treatment for patients with obstructive coronary artery disease and has been shown to improve symptoms and reduce mortality \[[@REF1],[@REF2]\]. Previously, bare metal stents (BMS) were used, and now the majority of recently used stents are drug-eluting stents (DES). First and second-generation DES and BMS present the risk of in-stent thrombosis, with newer second-generations demonstrating the lowest risk of late stent thrombosis \[[@REF3],[@REF4]\]. This serious complication can occur acutely (within the first 24 hours of stent deployment), subacutely (within 30 days), late (within the first year), or very late (more than a year post deployment). Most patients with stent thrombosis present with ST-segment elevation on electrocardiogram (STEMI). It is caused by a total or near total thrombotic occlusion of the intracoronary stent. Few cases of very long stent thrombosis are reported in the literature with the longest reported period being 11 years \[[@REF5],[@REF6]\]. We report a patient who presented with a very late stent thrombosis (VLST) that occurred 12 years after implantation. It is unique being the longest duration reported so far in the literature. We will cover in our discussion the possible underlying mechanisms and risk factors, as well as the implications of stent thrombosis on our practice.
Case presentation
=================
A 78-year-old man presented to our emergency department because of acute onset chest pain that started two hours prior to presentation. The pain was retrosternal, pressure-like, moderate in intensity and started upon awakening from sleep. His past medical history is significant for type II diabetes mellitus and pancreatic cancer that was treated with the Whipple procedure 27 years ago. He had coronary artery disease status post percutaneous angioplasty with stenting of the mid right coronary artery (RCA) 12 years ago, and stenting of the proximal circumflex and proximal RCA 17 years ago. A paclitaxel drug-eluting stent (PES) 3.0 x 24 mm was used to stent the mid RCA 12 years ago (Figure [1](#FIG1){ref-type="fig"}).
{#FIG1}
The patient is a former cigarette smoker, and does not consume alcohol, caffeine, or illicit drugs. At presentation, he was in mild distress, complaining of typical chest pain persistent despite aspirin administration. On physical examination, the patient was noted to be diaphoretic. His heart rate was 60 bpm, and blood pressure was 110/75 mmHg. Electrocardiogram (ECG) showed ST-segment elevation in the inferior leads II, II, and avF (Figure [2](#FIG2){ref-type="fig"}).
{#FIG2}
The patient was loaded with aspirin and clopidogrel and emergently taken to the cardiac catheterization laboratory. Coronary angiography showed thrombotic occlusion of mid RCA DES placed 12 years ago. Immediate percutaneous coronary balloon angioplasty was performed followed by a 3.5 x 16 mm everolimus drug-eluting stent (EES) deployment at a maximum inflation pressure of 14 atm. Following the intervention, excellent angiographic appearance of the artery was obtained with a 0% residual stenosis (Figure [3](#FIG3){ref-type="fig"}).
{#FIG3}
Due to the acuity of the situation and the patient's unstable hemodynamic status, it was difficult to obtain intravascular images of the lesion and tell with certainty what mechanism led to this event. The patient was stabilized and monitored for the following 24 hours with no further complications, and was then successfully discharged home.
Discussion
==========
Stent thrombosis occurring at any time is a serious complication carrying a significant risk of death. Although VLST is infrequent, it is being reported more in the literature with DES. According to the academic research consortium, our patient fits the criteria of having a "definite stent thrombosis" even with the absence of intravascular imaging. Angiographic diagnosis is made by identifying a thrombus originating in or within 5 mm of the stent along with the presence of one of the following criteria: acute onset of ischemic symptoms, new ischemic ECG changes, typical rise and fall in cardiac biomarkers, or pathologic confirmation following thrombectomy or by autopsy \[[@REF7]\]. Etiology, pathogenesis, and predictive factors have not yet been established due to relatively low prevalence of this condition and its multifactorial nature. The possible mechanisms of thrombosis that have been evaluated through real-time imaging studies using intravascular ultrasound, angioscopy, or optical coherence tomography, and tissue histology are still not fully understood. However, potential mechanisms and explanations of this late in-stent thrombosis include (1) delayed neointimal coverage (uncovered stent struts), (2) stent underexpension, (3) ongoing vessel inflammation, (4) neoatherosclerosis rupture, and (5) late stent malapposition, the latter being the most common \[[@REF8]-[@REF10]\]. Risk factors for stent thrombosis in general are well known, but some of these factors have been specifically associated with VLST. These include smoking history at the time of the stent implantation, presence of thrombus, multivessel disease, type C lesions, longer total stented length, and overlapping stents \[[@REF11]\]. Another major factor affecting the risk of stent thrombosis is the type of stent used. Several concerns have emerged about the higher risk of VLST with first-generation DES: paclitaxel drug-eluting stent (PES) and sirolimus drug-eluting stent (SES) \[[@REF12],[@REF13]\]. Studies comparing PES to newer generation DES showed that EES is associated with a significant reduction of ST at long-term follow-up, and PES having a higher risk of VLST \[[@REF4],[@REF14],[@REF15]\].
Conclusions
===========
As we report the longest case of stent thrombosis so far in the literature, we advocate the need of heightened awareness of these risks especially with first-generation PES and the other risk factors mentioned above. Further studies are needed to better understand the exact pathology behind very late in-stent thrombosis and help preventing them. Until then, we advise physicians to address as much as possible the underlying modifiable risk factors mentioned above. And when available, the use of intravascular imaging pre- and post-stenting is highly encouraged to guide selecting the appropriate stent size and adequate deployment of the stent on a healthy endothelium, and ensuring well apposition and expansion of the stent, particularly when dealing with proximal lesions.
The authors have declared that no competing interests exist.
Consent was obtained by all participants in this study
| {
"pile_set_name": "PubMed Central"
} |
Despite the significant progress in structure determination of membrane transporters, elucidating the dynamics of substrate movement through a transporter or a channel has remained a challenge ([@bib11]; [@bib23]). Crystal structures provide a static snapshot of the membrane transporter, but the substantial conformational changes that occur during substrate movement are far more difficult to capture. This latter aspect requires additional biochemical, genetic, or spectroscopic inputs to provide insights into the nature of the conformational changes ([@bib11]).
Over the years, a number of genetic and biochemical techniques have been exploited to obtain insights into the mechanistic workings of membrane transporters. These approaches have provided key support for our understanding, such as in the \"alternating access model\" of substrate transport seen in LacY permease and other transporters ([@bib1]; [@bib15]; [@bib23]). The primary focus of these approaches, however, has been to target the residues involved in substrate binding and substrate access. An aspect that has lagged behind in these studies is the nature of the conformational changes in regions not directly involved in substrate access, and how these other global changes in conformation in the protein contribute toward the transport process.
Here, we have attempted to partially address this lacuna and describe an approach we used to perform "charged/polar residue scanning" of the hydrophobic face of a transmembrane helix of a membrane transporter; we followed this with detailed biochemical and genetic analysis. [Hgt1p](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748) is a high-affinity glutathione transporter of the yeast *Saccharomyces cerevisiae* ([@bib5]) that belongs to the relatively uncharacterized Oligopeptide Transporter family ([@bib14]; [@bib17]; [@bib24]; [@bib25]). Homologs in other yeasts have also been shown to function in glutathione transport ([@bib4]; [@bib9]; [@bib34]; [@bib35]; [@bib36]) and in some cases oligopeptide transport ([@bib27]; [@bib29]), whereas more remote members are transporters of metal--amino acid conjugates ([@bib6]; [@bib7]; [@bib18]; [@bib37]). The *S. cerevisiae* paralogue, [OPT2](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000006398), also appears to have a role in glutathione homeostasis while functioning at the yeast peroxisomes ([@bib10]).Very little structure--function information is available for members of this family, and the lack of functionality of a cysteine-free mutant has made it difficult to apply methods, such as the Substituted Cysteine Accessibility Method (SCAM) ([@bib13]; [@bib21]). The TMD9 of [Hgt1p](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748) has been identified as being important in substrate binding, and two key residues, F523 and Q526, are thought to line the channel on the hydrophilic face of the helix ([@bib19]; [@bib34]).On the basis of the helical wheel arrangement of TMD9, it appears that the hydrophobic side of the helix might interact with the lipid matrix and/or other transmembrane segments of the protein. Residues of this face of the TMD9 helix were systematically replaced with lysine, glutamine, and glutamic acid. These replacements in the hydrophobic patch are expected to be deleterious to the protein and would then be subjected to mutagenesis to identify functional suppressors in other parts of the protein, with the premise that critical interaction disrupted by the primary mutation would be compensated by mutations at second sites.
Among the different charged mutants created, only six were nonfunctional, revealing a surprising tolerance of charged residues in the hydrophobic part of TM helices. I524, proximal to the substrate binding residues, was the only position that did not tolerate any charged residues. Suppressor analysis of all the nonfunctional mutants yielded second-site suppressors only in the case of I524K and I524Q, both of which involved a G202Q, G202K, or G202I substitution in the hydrophilic loop of [Hgt1p](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748) between TMD3 and TMD4, but G202Q/K/I alone was not deficient in transport activity. Charged and polar residue mutagenesis of P525, another residue close to the substrate binding residues, revealed that mere proximity to these residues was not responsible for the observations with I524. The results suggest that I524 in the hydrophobic face is in a conformationally critical region for substrate translocation and requires the involvement and possible interaction with region G202, near TMD3.
Materials and Methods {#s1}
=====================
Chemicals and reagents {#s2}
----------------------
All the chemicals used in this study were analytical grade and obtained from commercial sources. Media components were purchased from Difco (Detroit, MI) Sigma Aldrich, (St. Louis, MO), HiMedia, (Mumbai, India), Merck India Ltd (Mumbai, India), and USB Corporation (Cleveland, OH). Oligonucleotides were purchased from Sigma India. Restriction enzymes, Vent DNA polymerase, and other DNA-modifying enzymes were obtained from New England Biolabs (Beverly, MA). DNA sequencing kit (ABI PRISM 310 XL with dye termination cycle sequencing ready reaction kit) was obtained from Perkin Elmer, (Norwalk, CT). Gel-extraction kits and plasmid miniprep columns were obtained from QIAGEN (Valencia, CA) or Sigma (St. Louis, MO). \[^35^S\] GSH (specific activity 1000 Ci mmol^-1^) was purchased from Bhabha Atomic Research Centre, Mumbai, India. HA-Tag (6E2) mouse monoclonal antibody and horse anti-mouse HRP-linked antibody were bought from Cell Signaling (Danvers, MA). Alexa Flour 488 conjugated goat anti-mouse antibody was obtained from Molecular Probes (Eugene, OR). Hybond ECL (nitrocellulose) membrane and ECL plus Western blotting detection reagents were purchased from Amersham Biosciences (UK).
Strains, media, and growth conditions {#s3}
-------------------------------------
The *Escherichia coli* strain DH5α was used as a cloning host. The *Saccharomyces cerevisiae* strain used in this study was ABC 817 *(MATa [his3](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000005728)Δ1 [leu2](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000000523)Δ0 [met15](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000004294)Δ*-0 *[ura3](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000000747)Δ0 [hgt1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)Δ*::*[LEU2](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000000523))* ([@bib5]). *S. cerevisiae* was regularly maintained on yeast extract, peptone, and dextrose (YPD) medium. *S. cerevisiae* synthetic defined minimal medium (SD) contained yeast nitrogen base, ammonium sulfate, and dextrose supplemented with histidine, leucine, and methionine (when required) at 50 mg/liter ([@bib16]). Glutathione was added as required. Growth, handling of bacteria and yeast, and all the molecular techniques used in the study were according to standard protocols ([@bib31]).
Site-directed mutagenesis {#s4}
-------------------------
*[HGT1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)*, tagged with a Hemagglutinin (HA) tag at the C-terminus, was subcloned downstream of the TEF promoter at the *Bam* HI and *Eco* RI sites of a modified p416TEF vector ([@bib21]). This construct was used as a template for site-directed mutagenesis for creation of different site-directed mutants of [Hgt1p](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748) by splice overlap extension strategy. The mutations, K,Q,E, for each residue were generated using a single mutagenic oligonucleotide exploiting degenerate base pairs at the desired position in the different mutagenic oligonucleotides ([Supporting Information](http://www.g3journal.org/content/suppl/2015/03/16/g3.115.017079.DC1/017079SI.pdf), [Table S1](http://www.g3journal.org/content/suppl/2015/03/16/g3.115.017079.DC1/TableS1.pdf)). The PCR products generated with these oligonucleotides were subcloned back into the TEF vector background using appropriate restriction sites for subsequent analyses. The resulting mutants were sequenced to confirm the presence of the desired nucleotides changes and to rule out any undesired mutations introduced during the mutagenic procedure.
The dual complementation-cum-toxicity plate assay for assessing *HGT1* functionality {#s5}
------------------------------------------------------------------------------------
The complementation-cum-toxicity assay has been described previously ([@bib21]). The *S. cerevisiae [met15](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000004294)*Δ *[hgt1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)*Δ (ABC 817) was transformed with a single-copy centromeric vector expressing wild-type or different mutants of TMD9 of *[HGT1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)* expressed downstream of the TEF promoter. Transformants were grown in minimal media containing methionine and other supplements, without uracil, overnight. These cultures were reinoculated in the same media and allowed to grow until they reached the exponential phase. An equal number of cells were harvested, washed with water, and resuspended in sterile water to an OD600 of 0.2. These were serially diluted 1:10, 1:100, and 1:1000; 10 μl of these cell resuspensions were spotted on minimal medium containing different concentrations of glutathione (15, 30, 50, 100, 150, 200 μM) or methionine (200 μM) as sole organic sulfur source. Because *[HGT1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)* is under the strong TEF promoter, it is able to confer the ability to grow on glutathione at low concentrations of glutathione. However, at higher concentrations of glutathione, excessive uptake leads to toxicity in growth. Using this growth at low concentrations, and lack of it at higher concentrations, we have been able to grade the transporter defects, which is more informative than a mere growth assay at lower concentrations. The plates were incubated at 30° for 2--3 d and photographs were taken.
### Hydroxylamine mutagenesis of the pTEF-HGT1:
The random *in vitro* mutagenesis of the charged/polar residues of *[HGT1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)* was done using hydroxylamine and a previously described protocol ([@bib30]; [@bib35]). The 10 μg plasmid DNA was incubated in 0.5 ml of hydroxylamine solution (90 mg NaOH, 350 mg hydroxylamine HCl in 5 ml water, pH ∼6.5, freshly made before use). This mixture was incubated at 37° for 24 hr and mutagenized DNA was directly purified using Qiagen column. The purified plasmid DNA concentration was measured by running on agrose gel along with DNA ladder. The randomly mutagenized plasmid was directly transformed in *S. cerevisiae [met15](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000004294)Δ [hgt1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)Δ* strain and selected on 10 µg glutathione containing plates, and functional transformants were used to prepare plasmid DNA following a retransformation.
Glutathione transport assay {#s7}
---------------------------
The *S. cerevisiae* ABC 817 strain (*[met15](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000004294)Δ[hgt1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)Δ*) was transformed with different plasmid constructs bearing wild-type or *[HGT1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)* mutants under TEF promoter and was grown in minimal media containing methionine and other supplements, without uracil, overnight. These cultures were reinoculated in the same media and allowed to grow until they reached the exponential phase. Cells were harvested, washed, and put on ice in a MES-buffered medium until the transport was initiated. Transport experiments were carried out with \[35^S^\]-GSH as described earlier ([@bib21]). The results were expressed as nmol of glutathione-mg.protein-1 min-1. For the measurements of total protein, the 100 μl of the above cell suspension (cell suspension volume used for the transport assay) was boiled with 15% sodium hydroxide for 10 min, followed by neutralization of total cell lysate by addition of hydrochloric acid; 100 μl of this crude cell lysate was incubated with 0.1% Triton X-100 for 10 min and total protein was estimated by using the Bradford reagent (Sigma) using bovine serum albumin as a standard. For saturation kinetics, the initial rate of glutathione uptake was measured at a range of glutathione concentrations from 12.5 μM to 800 μM, with specific activity being kept constant at each concentration. The initial rate of glutathione uptake was determined by measuring the radioactive glutathione accumulated in the cells at 30-sec and 180-sec time points in the ABC 817 strain transformed with different plasmids constructs bearing wild-type and *[HGT1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)* mutants under TEF promoter or only vector. After subtracting the initial rate of glutathione uptake in vector from the initial rate of glutathione uptake in the different test constructs, the extent of glutathione uptake corresponding to the test construct was determined.
Preparation of cell extract and Western blot analysis {#s8}
-----------------------------------------------------
Crude cell extracts were prepared and Western blot analysis was done as described previously ([@bib21]) using a modified Western blot method ([@bib20]). Densitometry analysis of the unsaturated band signals was performed using the Scion Image software to quantify the protein expression levels in the different mutants. The resulting signal intensity was normalized with respect to the band surface area (in square pixels) and expressed in arbitrary units. The relative protein expression levels in the mutant [Hgt1p](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748) were represented as percentage expression relative to wild-type [Hgt1p](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748).
Cellular localization of the mutants by confocal microscopy {#s9}
-----------------------------------------------------------
To localize the different charged residue mutants of TMD9 of [Hgt1p](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748) that were created, indirect immunofluorescence was performed using a published protocol modified as described earlier ([@bib21]). For staining of ER (endoplasmic reticulum), live yeast cells were incubated with the ER-Tracker TM Red dye (BODIPY TR glibenclamide; Invitrogen, USA) according to the manufacturer's instructions and by other published literatures ([@bib12]; [@bib26]; [@bib28]). Images were obtained with an inverted LSM510 META laser scanning confocal microscope (Carl Zeiss) fitted with a Plan-Apochromat ×100 (numerical aperture, 1.4) oil immersion objective. The 488-nm line of an argon ion laser was directed over an HFT UV/488 beam splitter, and fluorescence was detected using an NFT 490 beam splitter in combination with a BP 505 to 530 band pass filter. ER-Tracker fluorescence was detected at Ex/Em 587/615. Images obtained were processed using Adobe Photoshop version 5.5
Results {#s10}
=======
Analysis of the residues of the hydrophobic face of the TMD9 helix by mutation to charged residues {#s11}
--------------------------------------------------------------------------------------------------
The helical wheel diagram of TMD9 shows a topological sidedness. The residues F523 and Q526, shown to be important in substrate binding, are located on the more polar face of the channel and are thought to directly interact with the substrate ([@bib19]; [@bib34]). The hydrophobic face of the helix, in contrast, is expected to interact with the lipid bilayer or with other TMDs. These residues in TMD9 are A509, V513, L517, L520, I524, and I528. To investigate this face and these residues more systematically, we considered a novel strategy. In this approach, each of the residues on this face was mutated to multiple charged residues, lysine, glutamine, or glutamic acid (using a single degenerate oligonucleotide), and nonfunctional mutants would then be subjected to suppressor analysis to obtain further insights ([Figure 1, A and B](#fig1){ref-type="fig"}). The residues described above were mutated as indicated and the mutants were subjected to an initial functional characterization using the previously designed sensitive plate assay, termed a complementation-cum-toxicity assay ([@bib19]; [@bib21]). The assay is based on the ability of *[HGT1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)* when expressed from the TEF promoter to permit growth on glutathione as a sole sulfur source in an *S. cerevisiae* strain (ABC 817). This strain is an organic sulfur auxotroph defective in glutathione uptake. At low glutathione concentrations (15 µM), *[HGT1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)* expressed under the TEF promoter complements the growth defect in the ABC 817 strain ([@bib5]) but results in toxicity when cells are grown in medium containing 50-µM or higher glutathione concentrations ([@bib21]; [@bib33]). The reasons for this toxicity have been examined elsewhere ([@bib22]), but the dual assay permits one to discern a gradation in functionality. Plasmids bearing either the WT *[HGT1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)* or the different mutants as described above were individually transformed into the *S. cerevisiae* ABC 817 strain, and the transformants were analyzed for their ability to confer complementation and/or toxicity to the cells over a range of glutathione concentrations.
{#fig1}
Based on their inability to complement the glutathione transporter defect, the mutants L517Q, L520Q, L520E, I524E, I524K, and I524Q were nonfunctional. V513K was severely defective but still mildly functional. The remaining mutants (A509Q, A509K, A509E, V513Q, V513E, L517K, L517E, L520K, and I528E) surprisingly had no functional defect and were comparable to WT ([Figure 2A](#fig2){ref-type="fig"} and [Table 1](#t1){ref-type="table"}). In case of I528, only I528E was constructed, because it was the only mutant plasmid recovered among the 20 different mutants of I528 that we analyzed.
{#fig2}
###### Summary of the charged residues mutants of *HGT1*
Residue Mutants Functional Activity
--------- --------- ---------------------
A509 A509Q +++
A509K +++
A509E +++
V513 V513Q +++
V513K \+
V513E +++
L517 L517Q ---
L517K +++
L517E +++
L520 L520Q ---
L520K +++
L520E ---
I524 I524Q ---
I524K ---
I524E ---
I528 I528E +++
Summary of the charged residue mutants of *HGT1* on functional activity of the transporter using dual complementation-cum-toxicity assay.
To corroborate these growth assays with biochemical assays, the mutants that were nonfunctional and severely defective were evaluated by measuring the rate of ^35^S-GSH uptake in the ABC 817 strain. The uptake data correlated well with the plate assay. Among the defective mutants, V513Q displayed some transport activity (∼18%) in agreement with the plate assay. However, the others mutants L517Q, L520Q, L520E, I524E, I524Q, and I524K had an almost complete loss in functional activity, showing very minimal uptake ability in agreement with the plate assay ([Figure 2B](#fig2){ref-type="fig"})
Expression and cell surface targeting of nonfunctional mutants {#s12}
--------------------------------------------------------------
To determine whether the nonfunctional mutant proteins are expressed and targeted properly, we first examined their steady-state expression levels by immunoblotting. Equal amounts of the crude protein extracts prepared from the *[met15](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000004294)∆[hgt1](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748)∆* strain transformed with the different mutants were loaded onto the gel, electroblotted to the membrane, and probed with anti-HA monoclonal antibody. L520Q and L520E showed almost complete loss in protein expression level. V513K and L517Q mutant showed detectable, but significantly lower, expression, whereas the other mutants I524K, I524Q, and I524E showed slightly lower, but otherwise comparable, levels to WT, and their expression ranged between 65% and 92% of wild-type ([Figure 3A](#fig3){ref-type="fig"}).
{#fig3}
The nonfunctional mutants were also studied for subcellular localization by indirect immunofluorescence using anti-HA monoclonal antibody. No signal was observed either in L520Q or in L520E, which probably reflects the almost complete loss of protein expression. In V513K, L517Q, I524Q, I524E, and I524K, a signal was observed at the cellular periphery of the cells, although they also showed a small amount of intracellular signal in addition to the cell surface signal ([Figure 3B](#fig3){ref-type="fig"}, [Figure S1](http://www.g3journal.org/content/suppl/2015/03/16/g3.115.017079.DC1/FigureS1.pdf)). This suggested that these mutants did not carry a significant trafficking defect.
G202K/I and G202K/Q mutations in the loop region between TMD3 and TMD4 are able to restore function to the nonfunctional I524K and I524Q mutants {#s13}
------------------------------------------------------------------------------------------------------------------------------------------------
Among all the hydrophobic residues that were evaluated by this charged residue scanning, it was only the charged residues at position I524 (I524K, I524K, and I524Q) that could not be tolerated as seen in functional assays. At all other positions, at least one of the charged residues was tolerated ([Table 1](#t1){ref-type="table"}).
The loss of function mutants I524K/Q/E were further subjected to suppressor analysis to examine if these mutants that were otherwise properly expressed and localized might be restored to functionality by compensatory mutations. Mutants were mutagenized *in vitro* with hydroxylamine treatment and transformed into the yeast ABC 817 strain. The resulting transformants were screened for a functional phenotype by direct selection on minimal media containing 10 μM GSH. For those clones bearing the original mutation, the full gene was sequenced to identify any additional mutations. The I524Q mutation yielded eight suppressors, of which two were revertants, whereas the remaining were either G202K or G202Q ([Table 2](#t2){ref-type="table"}). The I524K mutations yielded 10 suppressors, of which five were revertants, whereas the remaining were either G202K or G202I ([Table 2](#t2){ref-type="table"}). It is interesting that in all cases it was the G202 that was mutated to different amino acids (G202Q, G202K, and G202I) ([Table 2](#t2){ref-type="table"}). G202 is located in the intracellular loop between TMD3 and TMD4, but at the junction of TMD3.
###### Suppressor isolated from hydroxylamine mutagenesis--based intragenic suppressor analysis of mutant*s*
Primary Mutation No. of Suppressors Obtained at 10 µM of GSH Concentration Suppressor Distribution (No.)
------------------ ----------------------------------------------------------- -------------------------------
V513K 6 513V (4)
513N (1)
513Q (1)
L517Q 4 517L (3)
517A (1)
L520Q 6 520L (6)
L520E 8 520L (8)
I524Q 8 524I (2)
I524Q G202Q (2)
I524Q G202K (4)
I524E 8 524I (8)
I524K 10 524I (5)
I524K G202K (3)
I524K G202I (2)
We also subjected the remaining nonfunctional mutants to this suppressor analysis. The V513K suppressor involved same-site mutations with conversion to 513N and 513Q. Similarly, the L517Q suppressors yielded a mutation at the same site, 517A. In the case of L520Q and L520E, several suppressors were isolated but they all reverted back to the wild-type sequence. Thus, second-site suppressors were obtained only for I524Q and I524K mutations and were found to involve the G202 residue.
To investigate the I524-G202 interaction in greater detail, we created several additional charged mutants of I524 (I524D, I524R, I524N) as well as G202 (G202D, G202R, G202N, G202E) and evaluated the functional interaction of the different combinations ([Table 3](#t3){ref-type="table"}). Interestingly, the fresh charged residues at the I524 position that we introduced were all nonfunctional, although some partial functionality was observed with the I524N mutant.
###### Genetic analysis of I524 and G202 mutants and interaction involving I524-G202 in functional activity of the Hgt1p using dual complementation-cum-toxicity assay
Residue at position 524 Residue at Position 202 Functional Activity
------------------------- ------------------------- ---------------------
I G +++ (WT)
Q G ---
E G ---
K G ---
D G ---
R G ---
N G −/+
I K +++
I I +++
I Q +++
Q K +++
Q Q +++
Q R +++
Q D ---
Q N ---
Q E ---
K I +++
K K +++
K N ---
K D ---
K E ---
R R ---
R N ---
R K ---
D K ---
D N ---
N R ---
Detailed genetic analysis of I524 and G202 mutants and interaction involving I524-G202 in functional activity of the Hgt1p using dual complementation-cum-toxicity assay.
The interaction of these various combinations as measured by the growth assay is summarized in [Table 3](#t3){ref-type="table"}. Among the various G202 mutations that were made, only G202R and G202Q were able to suppress the I524Q mutation in a manner similar to G202K and suggested that the charge or polarity at this position was critical for suppression. However, in contrast, the I524R mutant (unlike the I524K mutant) could not be suppressed by G202K. Disruption of the activity by this mutation was therefore causing other defects in the functionality. None of the other mutations that included negatively charged residues could restore functionality in these interactions.
Functional and kinetic analysis of the I524K and I524Q suppressors {#s14}
------------------------------------------------------------------
To evaluate the suppressors in greater detail, the mutants I524K G202K, I524K G202I, I524K G202Q, I524Q G202K, and I524Q G202Q were spotted on different concentrations of glutathione and evaluated by the dual complementation-cum-toxicity assay. The suppressors G202K, G202Q, and G202I with their primary mutation I524K and I524Q were able to suppress the phenotypic defect caused by I524Q and I524K in terms of their ability to complement growth on glutathione. All these suppressors exhibited a functional activity almost similar to wild-type in being able to complement at lower concentrations of glutathione. However, the suppressors displayed a slight loss in functional activity in terms of their ability to confer toxicity to the cells at high concentrations of glutathione ([Figure 4A](#fig4){ref-type="fig"}). To corroborate the growth assays with actual transport data, we measured the initial rate of ^35^S-glutathione uptake in the ABC 817 strain transformed with the suppressor mutants of [Hgt1p](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748). The uptake was 45--62% as compared to wild-type ([Figure 4B](#fig4){ref-type="fig"}). These results indicated that the suppressors had regained significant activity.
{#fig4}
To determine the effect of suppressor mutations in the absence of the primary mutation, the G202Q, G202K, and G202I mutations were subcloned into the wild-type plasmid backbone. The mutants were functionally evaluated using the plate based dual complementation-cum-toxicity assay ([Figure S2](http://www.g3journal.org/content/suppl/2015/03/16/g3.115.017079.DC1/FigureS2.pdf)). The individual mutants G202Q G202K and G202I appeared to have an activity similar to wild-type. The protein expression levels of these suppressors with primary mutation were also similar to wild-type and properly localized to the cell surface ([Figure 4, C and D](#fig4){ref-type="fig"}).
Because of the low activity of the I524K/E/Q mutations, their kinetics could not be evaluated. However, the restoration of activity by the G202K/I/Q suppressors in these backgrounds provided the opportunity to get some insight into these mutants. We determined the kinetic parameter by measuring the initial rate of glutathione uptake over a range of glutathione concentrations. I524K and I524Q had almost null activity and could not be included for the kinetic study. We determined the kinetic parameters for I524Q G202K. Compared to the *Km* of the WT 48.1 ± 10.5 μM ([@bib5]), the *Km* for the suppressor mutant was significantly higher at 282.8± 44.5 μM ([Figure S3](http://www.g3journal.org/content/suppl/2015/03/16/g3.115.017079.DC1/FigureS3.pdf)).
Charged/polar residue scanning of P525 of TMD9 of Hgt1p {#s15}
-------------------------------------------------------
As I524 was located between the previously identified substrate binding residues F523 and Q526, examination was needed regarding the possibility that the inability of I524 to tolerate charged/polar residues might be due to this proximity to the substrate binding site rather than as a consequence of it being on the hydrophobic face and being involved in interactions critical for the channel functioning. To evaluate this possibility, we decided to examine if the residue, P525 adjacent to I524 and also located between F523 and Q526, could tolerate any charged/polar residues. P525 was mutated to lysine, glutamine, and glutamic acid by site-directed mutagenesis. These mutants were also subjected to an initial functional characterization by growth assays. P525K and P525E showed a severe effect on functional activity of [Hgt1p](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748), although they still retained partial activity, but P525Q had no discernible loss in function and was comparable to WT ([Figure 5](#fig5){ref-type="fig"}). Thus, polar residues at this position could be tolerated despite the proximity to F523 and Q526.
{#fig5}
The severely affected mutants P525K and P525E were also subjected to suppressor analysis. Suppressors were screened for a functional phenotype by direct selection on a minimal media containing 10 μM GSH. Plasmids were retrieved, re-examined for their functional activity using the growth-based plate assay, and sequenced. In one clone, P525K had converted to 525Q. All other suppressors for both P525K and P525E had reverted back to the wild-type sequence P525.
Discussion {#s16}
==========
The work described in this manuscript has explored a novel strategy for gaining insight into the mechanisms of membrane transporter functioning. In this approach, the six residues that formed the hydrophobic face of a TMD helix whose polar face was presumed to line the substrate translocation channel was subjected to "charged/polar-residue scanning" mutagenesis followed by genetic suppressor and biochemical analysis.
One of the most striking observations revealed from this study was that membrane transporters are surprisingly tolerant of charged residues, even in the most hydrophobic part of the TMDs. Previous efforts whereby charges were introduced into TM domains of membrane proteins were done either to examine the role in translocation/trafficking ([@bib8]) or to determine if the membrane domain spanned the membrane lipid bilayer with a helical ([@bib32]). In both cases, introduction of charged residues led to significant loss in function of the proteins. In contrast, out of the 16 mutations made to charged/polar residues (to lysine, glutamine, or glutamic acid), only six of these showed a complete loss in activity. In those cases, when an effect was seen, the effect of the replacement depended on both the position and the residue being introduced. Thus, for example, V513Q was functional, but V513K showed very low functionality; in contrast, L517Q was nonfunctional but L517K was functional. The loss of functionality in these cases was shown to be due to loss of protein expression, but it is clear that in the absence of any structural information it is difficult to predict the effects of introducing different charge/polarity at a particular position. Only at one of the six positions, position I524, was there a complete lack of tolerance to any charged/polar residues. I524E, I524K, or I524Q were all completely nonfunctional. However, all other positions on this face were capable of accepting at least one of the charged/polar residues. The I524 residue mutated to charged residues was also the only residue that allowed isolation of intragenic suppressors. This might suggest that other residues are not directly involved in interaction with other domains, and might also be an explanation for why charged/polar residues at these positions are more easily tolerated
Among the six substituted changes that led to complete loss in activity, three of them, L520Q, L520E, and L517Q, appeared to have destabilized the protein because no protein could be detected in L520Q and L520E and a very reduced protein expression was seen in L517Q. It was interesting that charged/polar residue changes at I524 affected the function of the protein but did not affect the protein expression to any significant extent. These mutants were also properly localized to the cell surface. There are thus two possible explanations for the complete loss in activity seen in I524K, I524Q, and I524E. The first one is that considering the proximity to the substrate binding residues; the changes were somehow drastically interfering with the substrate-binding. This could have been resolved by kinetic analysis; however, the activity was too low to allow us to subject these mutants to any kind of kinetic analysis. We thus examined how P525, another residue located between F523 and Q526, could tolerate charged residues. We subjected P525 to charged/polar residue scanning mutagenesis. P525K and P525E were nonfunctional, but P525Q was functional. Thus, charged/polar residues were functional at position P525, suggesting that mere proximity to the substrate binding residues could not account for the behavior of the I524K/Q/E mutants.
That brings us to the second plausible explanation for the complete loss of function of the I524K, I524E and I524Q mutants, namely, that the residue was on the region of the hydrophobic face that was part of the dynamics of channel movement and function. If this was the case, then it should involve interactions with other domains and, if so, we should be able to isolate suppressor mutations at other parts of the protein that might be able to suppress this loss of function. It is interesting in this context that although all the nonfunctional mutants were subjected to suppressor analysis, only second-site suppressor mutations were identified for I524Q and I524K. The position at which the second-site mutations had occurred was at G202, and it is predicted to be located in the loop between TMD3 and TMD4, close to TMD3. Interestingly, neither of these residues (neither I524 nor G202) was conserved in the oligopeptide transporter (OPT) super family. Interestingly, second-site suppressors were not observed even in the search for suppressors of P525K and P525E.
G202Q by itself is not an inhibitory mutation, but in the background of I524Q and I524K it functioned as a genetic suppressor that reverses the incapacitating constraints introduced by the I524Q and I524K mutation. Interestingly, the second-site suppressor was always found at the G202 position for both nonfunctional mutations I524Q and I524K. These studies suggest that TMD9 and TMD3 may be adjacent to, or in close contact with, each other. Sequence analysis of OPT homologs for any evolutionary coupling of G202 and I524 was carried out; however, no such coupling seemed to exist (data not shown). The higher *K~m~* in the suppressor suggests that the interaction has an impact on the substrate binding, but the major role of this interaction seemed to go beyond the effects on substrate binding. An explanation for the G202 with Q/K/I substitutions is that it could shift the local secondary structure probability from loop to helix and may produce helical turn in TMD3. Irrespective of this or other effects of G202 mutations, it is clear that the region plays a critical role in substrate translocation in conjunction with the hydrophobic face of TMD9. More biophysical, biochemical, or crystal structure studies would be required, however, to confirm these events.
In conclusion, we have used a novel strategy that we refer to as charged/polar-residue scanning mutagenesis of the hydrophobic face of TM helices as an approach to investigating membrane transporters and that could be added to the list of approaches currently being used to investigate this class of proteins. Genetic suppressor analysis for the determination of interacting domains is not a new approach, even in membrane transporters, and has been successfully applied previously ([@bib2]; [@bib3]), but the systematic application of this strategy to the hydrophobic face of the helix reveals a novel way of gaining insight into this class of proteins. The application of this approach to TMD9 of the yeast glutathione transporter, [Hgt1p](http://www.yeastgenome.org/cgi-bin/locus.fpl?dbid=S000003748), has yielded several important insights, and we believe that this approach can be successfully applied to other transporters as well.
This work was supported in part by grant-in-aid projects to A.K.B. from the Departments of Science and Technology and Department of Biotechnology, Government of India, and the JC Bose National Fellowship. A.T. was the recipient of a Research Fellowship from Council of Scientific and Industrial Research, Government of India.
Supporting information is available online at <http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.115.017079/-/DC1>
Communicating editor: R. A. Sclafani
[^1]: Present address: Laboratory of Gene Regulation and Development, *Eunice Kennedy Shriver* National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD 20892.
| {
"pile_set_name": "PubMed Central"
} |
The left atrium (LA) serves three major roles in maintaining left ventricular (LV) filling and overall cardiovascular performance: a reservoir that stores pulmonary venous return during LV contraction \[[@r11]\], a conduit that continues to passively transfer pulmonary venous flow during LV diastole, and a booster pump function that actively augments LV filling during atrial systole \[[@r31]\]. LA function helps to preserve cardiac output and maintain an effective LV stroke volume \[[@r21], [@r22], [@r29]\]. In patients with LV dysfunction, the importance of the atrial contribution to ventricular filling is emphasized by the development of clinical signs and symptoms of heart failure when LA contraction is impaired \[[@r24], [@r30]\].
In humans, the method that accurately determines the values of LA area and volume are two-dimensional (2D) echocardiography, multislice computed tomography, cardiac magnetic resonance imaging, and real time three-dimensional echocardiography \[[@r12], [@r32]\]. In veterinary patients, the phasic size of the LA (area and volume) \[[@r1], [@r2], [@r6], [@r13], [@r14], [@r19], [@r26], [@r28], [@r34]\] and the percentage fractional area change using 2D echocardiography have been an area of focus because of their utility in measuring LA function \[[@r14], [@r19], [@r26], [@r28]\].
A novel echocardiographic technique on the basis of the 2D speckle tracking method enabled automatic analysis of the time-LA area or volume curve representing LA phasic function in humans \[[@r20], [@r23]\] and dogs \[[@r14], [@r28]\]. The LA fractional area change during booster pump function (LA-FAC~act~) obtained via two-dimensional speckle tracking echocardiography (2D-STE) was lower in dogs with progressively more severe myxomatous mitral valvular heart disease \[[@r1], [@r26]\].
Moreover, strain imaging using 2D-STE is currently being developed for quantification of LA myocardial deformation by tracking the LA wall from frame to frame throughout the cardiac cycle and focuses on calculating the deformation parameter (strain) and the rate of deformation change (strain rate \[SR\]) automatically \[[@r1], [@r5], [@r25]\]. Notably, LA booster pump dysfunction indicated by strain imaging using 2D-STE was shown to be the best predictor of heart failure complications in dogs and had a higher predictive power for evaluating congestive heart failure over the LA-FAC~act~ \[[@r25]\].
Volume load dependency of echocardiographic indices is of clinical concern when using the indices in heart diseases associated with volume overload: LA dysfunction can be masked by the enhancing effect of the volume loading on LA function indices \[[@r31]\]. In a previous experimental study using healthy beagles, we have shown that the volumetric LA function indices (i.e., LA fractional area changes) determined with 2D-STE are volume load-dependent and enhanced by cardiac volume loading \[[@r27]\]. On the other hand, the degree of volume load dependency on LA function indices derived from strain imaging using 2D-STE remains unclear in dogs.
Therefore, the aim of this study is to elucidate the effect of clinically relevant changes of acute volume loading on strain and SR parameters derived using the 2D-STE method in dog models. The results of the present study could describe the degree of volume load dependency on LA myocardial deformation for further therapeutic strategy and prognostic information of the LA.
MATERIALS AND METHODS {#s1}
=====================
Animals
-------
Six laboratory beagles (aged 1--3 years, with body weight of 8.8 to 11.4 kg), which were part of an experimental unit at Hokkaido University, were enrolled in this study. All dogs were healthy and had no abnormalities of cardiac function on the basis of routine physical examination, including blood examination, electrocardiogram (ECG), and standard echocardiography (including M-mode, pulsed-wave Doppler, and color flow Doppler--based imaging). All procedures were reviewed and approved by the laboratory animal experimentation committee of the Graduate School of Veterinary Medicine, Hokkaido University (approval No. 15-0087).
Procedure
---------
The protocol used in this study was the same as in a previous report \[[@r27]\]. An intravenous infusion route was established in each dog on the left and right cephalic veins with a 20-gauge over-the-needle catheter, and a 24-gauge over-the-needle catheter was placed in the left or right dorsal pedal artery to directly monitor arterial blood pressure. Each dog was administered atropine sulfate (Mitsubishi Tanabe Pharma Corp., Osaka, Japan) 0.05 mg/kg, subcutaneously, cefazolin sodium hydrate (Astellas Pharma Inc., Tokyo, Japan) 20 mg/kg intravenously (IV), and heparin sodium (Ajinomoto Pharmaceuticals Co., Ltd., Tokyo, Japan) 100 units/kg IV, and sedated with butorphanol tartate (Meiji Seika Pharma Co., Ltd., Tokyo, Japan) 0.2 mg/kg IV and midazolam hydrochloride (Astellas Pharma Inc., Tokyo, Japan) 0.1 mg/kg IV. Then, anesthesia was induced with administration of propofol (Mylan Inc., Canonsburg, PA, U.S.A.) 6 mg/kg IV. Thereafter, each dog was endotracheally intubated, and anesthesia was maintained with isoflurane (DS Pharma Animal Health Co., Ltd., Osaka, Japan) 1.75 to 2.0% in 100% oxygen. End-tidal partial pressure of carbon dioxide was continuously monitored and maintained between 35 and 45 mmHg with mechanical ventilation, with a tidal volume of 10 to 15 m*l*/kg and a respiratory rate of 10 to 12 breaths/min. Heart rate and arterial pressure measured with arterial catheterization were continuously recorded with a commercial polygraph instrument (Nihon Kohden Co., Ltd., Tokyo, Japan).
A 6F, 12-cm introducer sheath (St. Jude Medical Inc., Minnetonka, MN, U.S.A.) was percutaneously inserted into the right external jugular vein using the Seldinger technique in each dog which was positioned in the position of left lateral recumbency. A 5F, 75-cm Swan-Ganz catheter (Edwards Lifesciences Corp., Irvine, CA, U.S.A.) was advanced into the pulmonary artery with fluoroscopy guidance. The catheter was connected to polygraph equipment for acquisition of hemodynamic data.
Following a stabilization period of about 10 min, baseline recordings of hemodynamic and echocardiographic indices were performed. Thereafter, cardiac preload was increased by IV infusion of warmed lactated Ringer solution (Terumo Corp., Tokyo, Japan) at 150 m*l*/kg/hr for 90 min \[[@r27]\]. This dose was modified from the dose used in previous studies \[[@r16], [@r17]\].
After the fluid infusion began, hemodynamic and echocardiographic evaluations were performed every 15 min. The hemodynamic data were obtained before echocardiography at each time point assessment. Following the final echocardiographic examination, each dog was administered furosemide (Sanofi K K, Tokyo, Japan) 4 to 6 mg/kg IV and allowed to recovery from anesthesia.
Hemodynamic assessment
----------------------
All hemodynamic data including heart rate, mean arterial blood pressure, mean pulmonary arterial pressure, pulmonary capillary wedge pressure (PCWP), mean right atrial pressure, and cardiac output were recorded by a polygraph instrument and digitally stored. Mechanical ventilation was briefly stopped during the recordings of hemodynamic indices. The distal and proximal ports of a Swan-Ganz catheter were used to measure pulmonary arterial and right atrial pressures, respectively. The PCWP was determined when the balloon at the end of the Swan-Ganz catheter was inflated to be wedged in a small pulmonary artery. After pressure recordings, cardiac output was determined using the thermodilution method with the injection of a 5 m*l* bolus of cold saline (0.9% NaCl) into the right atrium through the proximal port of a Swan-Ganz catheter. Stroke volume was calculated by dividing cardiac output by heart rate. For pressure measurements, the mean of five consecutive cardiac cycles was calculated, and the average of four measurements was calculated for cardiac output.
Standard echocardiographic methods
----------------------------------
Echocardiography was performed by the same experienced investigator (KN) using a Toshiba Artida^TM^ echocardiographic system (Toshiba Medical System Corp., Tochigi, Japan) with a 3- to 7-MHz sector probe transducer array. All echocardiographic indices were recorded when dogs were in an expiratory phase. An ECG trace (lead II) was recorded simultaneously with echocardiographic imaging by the ECG equipment on the ultrasonographic device, in addition to that on the polygraph instrument. The mean of 3 consecutive cardiac cycles was calculated for all echocardiographic indices, including those determined by 2D-STE.
Pulsed-wave Doppler echocardiography was performed to measure the transmitral flow velocity from the left apical four-chamber view. The sample gate for transmitral flow was placed at the tip of the mitral valve leaflets when they were opened \[[@r3]\]. The following indices were measured: peak velocity of the early diastolic wave (E wave), peak velocity of the late diastolic wave (A wave), and the ratio of the peak velocity of the E wave to the peak velocity of the A wave. These indices were not determined when the E and A waves were completely or partially fused.
The aortic Doppler flow profile was obtained with the sample gate positioned immediately below the aortic valve from the left apical five-chamber view. Left ventricular ejection time (ET) was measured as the interval from the onset to the end of the aortic flow. Left ventricular pre-ejection period (PEP) was measured as the interval from the start of the QRS complex to the beginning of aortic flow. The ratio of the PEP to ET was calculated.
Myocardial motion velocities derived from tissue Doppler imaging were recorded with the sample gate placed at the septal mitral annulus from the left apical four-chamber view \[[@r3]\]. The peak velocity of the systolic wave (S′ wave), peak velocity of the early diastolic wave (E′ wave), and peak velocity of the late diastolic wave (A′ wave) were measured, and the ratio of the peak velocity of the E′ wave to the peak velocity of the A′ wave and the ratio of the peak velocity of the E wave to the peak velocity of the E′ wave were calculated. These indices other than the peak velocity of the S′ wave were not determined when the E′ and A′ waves were completely or partially fused.
Additionally, from the tissue Doppler imaging velocities of myocardial motion at the septal mitral annulus, the isovolumic relaxation time (IVRT) and the isovolumic contraction time (IVCT) was measured: the IVRT was correspondence to the interval from the end of the S′ wave to the beginning of the E′ wave, while the IVCT was correspondence to the interval from the end of the A′ wave to the beginning of the S′ wave.
2D-STE of the LA
----------------
From the left apical four-chamber view, the image used for strain imaging using 2D-STE of LA was acquired with the frequency, depth, and sector width adjusted for optimization of frame rate (between 151 and 229 frames per rate). The image of three consecutive cardiac cycles was digitally stored at each assessment point for later offline analysis.
The obtained echocardiographic images were analyzed with 2D wall motion-tracking software (Toshiba Medical Systems Corp., Tochigi, Japan) by one investigator (AD) \[[@r5], [@r25]\]. The LA strain and SR were analyzed by 2D-STE using the QRS complex on the ECG trace as the initiation of the calculation. The LA endocardial surface was manually traced along the clearly visualized internal edge of the LA wall in that frame using the point-and-click method, and the epicardial surface of the LA wall was automatically generated by offline software. After tracking, the software generates an optimal region of interest with the LA myocardial wall thickness divided into six segments with adjustable width to fit the entire LA myocardial wall throughout the cardiac cycle, thus creating the longitudinal strain and SR curve for each atrial segment and a mean curve of all six segments (global strain and SR) for each dog. A cine loop preview was used to confirm an adequate speckle pattern generation following movement of the LA myocardium. All images included in the study were visually inspected for image quality (an adequate image without dropout speckle pattern). The LA strain and SR for each of the three phasic functions were measured from the global curve at three different time points ([Fig. 1](#fig_001){ref-type="fig"}Fig. 1.Two-dimensional speckle tracking echocardiographic images illustrating measurement of strain and strain rate (SR) curves during each phasic function of the left atrium (LA). The LA myocardium was automatically divided into six segments, as shown in the apical four-chamber view. White lines represent the average global strain and SR. From the onset of ventricular systole, the mean LA strain curve presented the first positive peak and decrease to a plateau during diastasis, followed by the second positive peak at the atrial contraction phase, and finally the negative peak after atrial contraction. For the SR profile, the first positive peak during ventricular systole (SR~s~) and two negative peaks at early (SR~e~) and late (SR~a~) ventricular diastole were obtained.) \[[@r5], [@r25]\]: minimum strain (S~min~) at negative peak during the ventricular end-diastole, maximum strain (S~max~) at peak during the ventricular systole, and strain before atrial contraction (S~a~). The LA strain corresponding reservoir function (ɛS), that for conduit function (ɛE), and that for booster pump function (ɛA) were calculated at the period of ventricular systole, early ventricular diastole, and late ventricular diastole using the following equations, respectively.
ɛS=S~max~−S~min,~ ɛE=S~max~−S~a,~ ɛA=S~a~−S~min~
Similarly, the SR for reservoir function (SR~S~), SR for conduit function (SR~E~), and SR for booster pump function (SR~A~) were calculated at the time of ventricular systole at positive peak, early ventricular diastole at first negative peak, and late ventricular diastole at second negative peak, respectively, as follows:
SR~S~=max1, SR~E~=min1, SR~A~=min2
As well as LA strain and SR curves, the software automatically generated a LA volume curve which was calculated by the monoplane area-length method of LA. LA volumes for each of the three phasic functions were measured: maximal LA volume (V~max~), the volume at the frame before the opening of the mitral valve starts; preatrial contraction LA volume (V~preA~), the volume at the frame before the P wave on the ECG; and minimal LA volume (V~min~), the volume at the frame at the mitral valve closure. The total, passive, and active LA emptying FVCs indicating reservoir, conduit, and booster pump function, respectively, were calculated based on the following formulae:
LA-FVC~total~=100 × (V~max~−V~min~)/V~max,~
LA-FVC~passive~=100 × (V~max~−V~p~)/V~max,~
LA-FVC~active~=100 × (V~p~−V~min~)/V~p~
Statistical analysis
--------------------
Statistical analyses were performed on JMP Pro 12.2.0 software (SAS Institute, Cary, NC, U.S.A.). Normal distribution of the data was confirmed by a Shapiro-Wilk test. A linear mixed model was developed with time (baseline, 15, 30, 45, 60, 75, and 90 min) as a categorical fixed effect and dog identity as a random effect. The *F* test was performed to assess the effect of time on the values of the measured variables. Pairwise comparisons between the baseline and each time point were performed by obtaining the least-squares means and using the Bonferroni correction to account for multiple comparisons. The relationship between PCWP and each of the indices of LA functional strain/SR were investigated using multiple regression analysis. In model 1, the PCWP and dummy coding of the enrolled dogs were included as covariates (linear regression model). In model 2, the quadratic terms of PCWP and dummy coding of the enrolled dogs were entered as covariates (quadratic regression model). For each LA function parameter, model 2 was accepted if the effect of the quadratic term of the PCWP was significant, and a log-likelihood ratio χ^2^ test revealed that model 2 had a fit superior to that of model 1. After constructing each model, assumptions of linearity, normality, homoscedasticity, and independence of the residuals were evaluated by inspection of the standardized residual plots and quantile plots.
RESULTS {#s2}
=======
Change in hemodynamic variables
-------------------------------
The changes in hemodynamic variables before (baseline) and at each assessment point after IV infusion of fluid are summarized in [Table 1](#tbl_001){ref-type="table"}Table 1.Least square mean (95% CI) obtained from linear mixed model for hemodynamic data before (baseline) and each times point during acute volume loading in 6 healthy beaglesVariablesBaseline15 min30 min45 min60 min75 min90 minMBP55 (48--63)55 (48--62)57 (49--64)58 (51--65)60 (53--67)60 (53--67)59 (52--67)Heart rate (beats/min)110 (103--117)101 (94--108)^a)^109 (102--116)110 (103--117)113 (106--120)114 (107--121)109 (102--116)PAPm (mmHg)9.5 (7.1--11.9)12.7 (10.2--15.1)^a)^15 (12.6--17.4)^a)^16.3 (13.9--18.8)^a)^17.3 (14.9--19.8)^a)^17.8 (15.4--20.3)^a)^17.5 (15.1--19.9)^a)^PCWPm (mmHg)3.3 (1.1--5.6)8 (5.8--10.2)^a)^10.3 (8.1--12.6)^a)^12 (9.8--14.2)^a)^12.7 (10.4--14.9)^a)^13.3 (11.1--15.6)^a)^14 (11.8--16.2)^a)^RAPm (mmHg)0.3 (−1.1--1.7)5.5 (4.1--6.9)^a)^6.5 (5.1--7.9)^a)^7.2 (5.8--8.6)^a)^7.5 (6.1--8.9)^a)^8.2 (6.8--9.6)^a)^8.3 (6.9--9.7)^a)^Cardiac output (*l*/min)2.1 (1.8--2.4)2.6 (2.3--2.8)^a)^2.8 (2.5--3.1)^a)^2.9 (2.7--3.2)^a)^3 (2.7--3.3)^a)^3.1 (2.8--3.4)^a)^3.1 (2.8--3.4)^a)^Stroke volume (m*l*)19 (17--22)25 (23--28)^a)^26 (24--28)^a)^27 (24--29)^a)^27 (24--29)^a)^28 (25--30)^a)^28 (26--31)^a)^a) Value differs significantly (*P*\<0.05) from corresponding baseline value. PAP=Pulmonary arterial blood pressure. RAP=Right atrial blood pressure. PCWP=Pulmonary capillary wedge pressure. MAP=Mean arterial blood pressure as measured via arterial catheterization.. Mean PCWP and cardiac output were significantly greater than at baseline from 15 to 90 min after acute volume loading began. Heart rate and mean arterial blood pressure were not significantly changed from baseline at 15 to 90 min for all 6 dogs, whereas stroke volume was significantly changed from baseline at 15 to 90 min.
Change in echocardiographic parameters
--------------------------------------
The changes in echocardiographic parameters before (baseline) and at each assessment point after acute volume loading began are summarized in [Table 2](#tbl_002){ref-type="table"}Table 2.Least square mean (95% CI) obtained from linear mixed model for conventional echocardiographic parameters before (baseline) and each times point during acute volume loading in 6 healthy beaglesVariablesBaseline15 min30 min45 min60 min75 min90 minE wave (m/sec)0.71 (0.65--0.77)0.84 (0.78--0.89)^a)^0.85 (0.79--0.91)^a)^0.83 (0.77--0.89)0.91 (0.85--0.97)^a)^0.91 (0.85--0.97)^a)^0.85 (0.79--0.91)^a)^A wave (m/sec)0.43 (0.33--0.53)0.48 (0.38--0.58)0.55 (0.45--0.66)0.55 (0.45--0.65)0.53 (0.42--0.63)0.56 (0.45--0.66)0.49 (0.38--0.59)Ratio of E and A1.72 (1.33--2.10)1.81 (1.42--2.19)1.58 (1.19--1.97)1.66 (1.27--2.05)1.69 (1.29--2.07)1.77 (1.39--2.16)1.86 (1.47--2.25)PEP (msec)74 (66--81)66 (59--74)^a)^66 (58--73)^a)^62 (55--70)^a)^60 (52--67)^a)^61 (53--68)^a)^62 (54--69)^a)^ET (msec)185 (170--200)235 (220--250)^a)^239 (223--254)^a)^243 (228--258)^a)^255 (239--270)^a)^253 (238--269)^a)^261 (246--276)^a)^Ratio of PEP and ET0.40 (0.36--0.44)0.28 (0.25--0.32)^a)^0.27 (0.24--0.31)^a)^0.26 (0.22--0.29)^a)^0.23 (0.19--0.27)^a)^0.24 (0.20--0.28)^a)^0.24 (0.20--0.27)^a)^E′ wave (cm/sec)6.47 (4.95--7.98)9.04 (7.52--10.55)^a)^9.75 (8.23--11.27)^a)^9.17 (7.66--10.69)^a)^9.88 (8.36--11.39)^a)^9.29 (7.78--10.82)^a)^8.72 (7.20--10.24)^a)^A′ wave (cm/sec)3.54 (2.35--4.73)4.80 (3.61--5.99)5.18 (3.99--6.36)^a)^5.56 (4.37--6.75)^a)^6.53 (5.34--7.71)^a)^5.17 (3.98--6.36)^a)^5.29 (4.09--6.48)^a)^S′ wave (cm/sec)5.44 (4.59--6.29)5.42 (4.57--6.27)5.56 (4.71--6.41)5.57 (4.72--6.42)5.96 (5.11--6.81)5.75 (4.90--6.59)5.41 (4.56--6.26)Ratio of E and E'11.38 (9.68--13.09)9.39 (7.68--11.09)8.98 (7.28--10.69)9.16 (7.45--10.86)9.26 (7.55--10.96)10.17 (8.46--11.88)9.89 (8.18--11.59)IVCT55 (45--64)49 (40--59)53 (43--62)47 (37--57)46 (36--56)46 (36--56)48 (39--58)IVRT54 (47--62)57 (49--64)59 (51--66)56 (48--63)60 (52--67)59 (52--67)56 (48--63)a) Value differs significantly (*P*\<0.05) from corresponding baseline value. A wave=Peak velocity of the A wave. A′ wave=Peak velocity of the A′ wave. ET=Left ventricular ejection time. E wave=Peak velocity of the E wave. E′ wave=Peak velocity of E′ wave. IVCT=Isovolumic contraction time. IVRT=Isovolumic relaxation time. PEP=Left ventricular pre-ejection period. S′ wave=Peak velocity of S′ wave.. Peak velocities of E, A, E′, and A′ wave were determined without fusion of the E and A wave and the fusion of the E′ wave and A′ wave for all dogs. Acute volume loading induced a significant increase from baseline at 15 to 90 min in peak velocity of the E wave, E′ wave, and at 30 min in the A′ wave. LV ET was significantly increased and the ratio of LV PEP to ET was significantly decreased from baseline at 15 to 90 min. The IVCT and IVRT did not show a significant change after volume infusion. For LA strain and SR variables ([Table 3](#tbl_003){ref-type="table"}Table 3.Least square mean (95% CI) obtained from linear mixed model for LA strain and strain rate before (baseline) and each times point during acute volume loading in 6 healthy beaglesVariablesBaseline15 min30 min45 min60 min75 min90 minɛS19.4 (15.5--23.3)31.1 (27--35.2)^a)^32.2 (28.3--36)^a)^32.5 (28.6--36.4)^a)^33.9 (30--37.8)^a)^33.1 (29.3--37)^a)^30.3 (26.4--34.1)^a)^ɛE13.6 (10.3--16.9)22.3 (18.9--25.7)^a)^23.9 (20.7--27.3)^a)^21.8 (18.5--25.1)^a)^21.5 (18.2--24.7)^a)^22.6 (19.3--25.8)^a)^20.7 (17.4--24)^a)^ɛA5.8 (2.6--8.9)8.8 (5.5--12)8.3 (5.1--11.4)10.4 (7.2--13.5)^a)^12.1 (8.9--15.3)^a)^10.4 (7.3--13.6)^a)^9.7 (6.6--12.9)^a)^SR~S~1.4 (0.8--2)2.3 (1.7--2.9)^a)^2.3 (1.7--2.9)^a)^2.5 (1.9--3.1)^a)^3.1 (2.5--3.7)^a)^2.6 (1.9--3.2)^a)^2.2 (1.5--2.8)^a)^SR~E~2.5 (1.9--3.2)3.5 (2.8--4.1)^a)^3.3 (2.7--3.9)^a)^3.2 (2.6--3.8)^a)^3.4 (2.8--4)^a)^3.3 (2.7--3.9)^a)^3.1 (2.5--3.7)^a)^SR~A~1.3 (0.9--1.7)1.9 (1.5--2.2)^a)^1.7 (1.4--2.1)^a)^1.9 (1.5--2.3)^a)^1.8 (1.4--2.2)^a)^1.7 (1.3--2.1)^a)^1.5 (1--1.9)a) Value differs significantly (*P*\<0.05) from corresponding baseline value. ɛS/ SR~S~, ɛE/ SR~E,~ ɛA/ SR~A~ represent strain/strain rate during reservoir, conduit and booster pump function, respectively.), acute volume loading caused a significantly increased LA strain from baseline, corresponding to reservoir function and conduit function from baseline at 15 to 90 min and at 45 to 90 min in ɛA ([Fig. 2](#fig_002){ref-type="fig"}Fig. 2.Representative software generated a strain curve for a single cardiac cycle of a healthy beagle at baseline (A) and at 90 min after cardiac acute volume loading (B).). Acute volume loading caused a significant increase from baseline in all phasic functions of SR at 15 min. The LA volumes and LA-FVCs at each assessment point after acute volume loading are shown in [Table 4](#tbl_004){ref-type="table"}Table 4.Least square mean (95% CI) obtained from linear mixed model for LA phasic function variables derived from 2D-STE before (baseline) and each times point during acute volume loading in 6 healthy beaglesVariablesBaseline15 min30 min45 min60 min75 min90 minVmin6.12 (4.47--7.77)6.91 (5.24--8.58)7.93 (6.29--9.58)6.66 (5.01--8.30)7.44 (5.79--9.09)8.52 (6.87--10.17)^a)^8.76 (7.11--10.41)^a)^VpreA7.31 (4.58--10.04)8.94 (6.17--11.71)10.33 (7.60--13.06)9.33 (6.59--12.06)10.89 (8.16--13.62)^a)^11.79 (9.07--14.53)^a)^11.75 (9.02--14.48)^a)^Vmax11.05 (7.03--15.08)16.63 (12.54--20.71)^a)^19.19 (15.17--23.22)^a)^16.59 (12.57--20.62)^a)^18.83 (14.81--22.86)^a)^21.19 (17.17--25.22)^a)^20.51 (16.48--24.53)^a)^Fractional volume change (%)Total43.6 (38.7--48.4)58.1 (52.9--63.2)^a)^58.8 (53.9--63.6)^a)^59.9 (55--64.8)^a)^60.1 (55.3--65)^a)^59.2 (54.4--64.1)^a)^57.4 (52.5--62.3)^a)^Passive32.3 (26.1--38.5)45.8 (39.4--52.2)^a)^46.7 (40.5--52.8)^a)^43.9 (37.7--50.1)^a)^42.7 (36.5--48.9)^a)^44.9 (38.7--51.1)^a)^43.3 (37.1--49.5)^a)^Active16.5 (10.2--22.7)22.9 (16.4--29.4)22.6 (16.4--28.9)28.3 (22.1--34.6)^a)^29.7 (23.5--35.9)^a)^25.5 (19.3--31.8)^a)^24.2 (17.9--30.5)^a)^a) Value differs significantly (*P*\<0.05) from corresponding baseline value. LA-FVC~total~, LA-FVC~passive~, LA-FVC~active~ represent left atrial fractional volume change during reservoir, conduit and booster pump function, respectively.. The V~min~, V~preA~, and V~max~ were obtained from speckle tracking analysis. The V~preA~ was significantly increased from baseline at 60 to 90 min, at 15 to 90 min in V~max~, and at 75 to 90 min in V~min~.
A quadratic multiple regression model (model 2) provided to be a better fit model when compared with the linear regression model (model 1) for relationships between PCWP and all phasic functions of LA function indices, including LA strains, SRs, and LA-FVCs ([Table 5](#tbl_005){ref-type="table"}Table 5.Maximum likelihood estimates (95% CIs), adjusted coefficients of determination (*R*^2^), and results of log-likelihood ratio χ^2^ tests for multiple linear (1) and quadratic (2) regression models of the association between variables of left atrial phasic function and PCWP in 6 healthy BeaglesVariablesModel 1Model 2Log-likelihood ratio χ^2^ testPCWP*R^2^*PCWPPCWP^2^*R^2^*Accepted*P*-valueɛS0.8(0.4--1.2)0.550.7(0.5--0.9)−0.13(−0.2 to −0.1)0.79Q\<0.0001ɛE0.5(0.3--0.8)0.500.4(0.2--0.7)--0.08(−0.1 to −0.04)0.68Q\<0.0001ɛA0.4(0.2--0.6)0.640.3(0.1--0.5)--0.05(−0.1 to −0.02)0.72Q0.0017SR~S~0.8(0.03--0.1)0.440.07(0.02--0.1)--0.01(−0.02 to −0.002)0.53Q0.0067SR~E~0.05(0.004--0.1)0.550.03(−0.01--0.1)--0.01(−0.02 to −0.005)0.68Q\<0.0001SR~A~0.03(−0.003--0.1)0.420.001(−0.04--0.04)--0.01(−0.02 to −0.003)0.58Q0.0067LA-FVC~Total~1.2(0.7--1.6)0.530.9(0.6--1.3)--0.15(−0.2 to −0.1)0.77Q\<0.0001LA-FVC~Passive~0.9(0.4--1.4)0.510.7(0.3--1.1)--0.1(−0.2 to −0.04)0.62Q0.0017LA-FVC~Active~0.8(0.4--1.3)0.590.7(0.3--1.1)--0.1(−0.2 to −0.03)0.67Q0.0067L=Linear regression model (model 1). PCWP^2^ =Quadratic term of PCWP. Q=Quadratic regression model (model 2).; [Fig. 3](#fig_003){ref-type="fig"}Fig. 3.Relationships between pulmonary capillary wedge pressure (PCWP) in 6 healthy beagles and the following indices of left atrial (LA) phasic function. Quadratic regression lines better fit the relationships between PCWP and LA deformation indices, indicating three phases of LA function.).
DISCUSSION {#s3}
==========
The major finding of our study was that the LA phasic function assessed by strain imaging with 2D-STE was enhanced during experimental cardiac volume overload as the volumetric LA function indices (i.e., LA-FVCs) did. This study was the first to report the relationship between acute volume loading and the three phasic functions of LA obtained via strain imaging with 2D-STE in dogs and to evaluate the response of the strain indices to acute hemodynamic change.
A technical limitation of the volumetric LA function indices including LA-FACs and LA-FVCs is volume load dependency \[[@r31]\]. In heart diseases with volume overload, LA dysfunction evaluated on the basis of these indices can be masked by the enhancing effect of the volume loading on them. In the present study, LA phasic function assessed by LA-FVCs were enhanced with acute volume loading as LA-FACs did in our previous study \[[@r27]\].
Regarding the change in LA phasic function, it is known that there are many significant determinants of cardiovascular factors. The reservoir function is modulated by LA intrinsic relaxation, LA chamber stiffness, and the property of LV contraction \[[@r11], [@r31]\]. LA conduit function is related to the early diastolic pressure between LA and LV, and relaxation of the LV \[[@r31]\]. The booster pump of LA function is determined by intrinsic LA contractility, LA preload (i.e., LA volume before atrial contraction), and LV end-diastolic filling pressure and compliance \[[@r31]\].
During the early atrial contraction phase, in the same way as for LV, the LA function follows Starling's law, and the booster pump function was enhanced in response to an increasing cardiac preload, as supported by increased volume before atrial contraction (V~preA~) identified in this study \[[@r31]\].
The increase in volume load enhanced the reservoir function by stimulating the LA booster pump function and LV systolic function as determined by the reduction in the ratio of the LV PEP and ET and by increasing the cardiac output according to the Frank starling mechanism in this study \[[@r15], [@r27], [@r31], [@r33]\].
In addition, the parameter indicating conduit function was higher after infusion of volume. In the study reported here, the peak velocity of the E and the E′ wave was increased with acute volume loading, whereas the ratio of peak velocities of the E to velocity of the E′ wave and the ratio of the velocity of the E to the velocity of the A wave were not changed significantly. The effect of preload on that ratio was suggested as a minimal preload dependency \[[@r17]\]. The enhancement of conduit function was mainly caused by increased LA pressure during the early diastolic period, as suggested by the increase in the indicator of LA pressure (increased peak velocities of E wave) in this study. Mitral valve E velocity is a variable that is not only primarily determined by the filling pressure gradient between LA and LV but also influenced by relaxation \[[@r8]\]. Furthermore, the increase in conduit function might be surpassed by the effect of increased reservoir phase. The LA reservoir and conduit function is correlated with maintaining LV performance, whether the LA is initially filled with blood or volume and consequent relaxation of LV, resulting in increased blood entering the LV. However, the relationship between the LA reservoir and conduit is likely to vary depending on atrial pressure, volume, and neural control, as described in a previous investigation \[[@r29]\].
In the present study, there were quadratic relationships between PCWP and LA function indices derived from strain imaging with 2D-STE: the LA function indices were initially enhanced and then started to be impaired during cardiac volume overload. This finding is in line with a previous study where the volumetric LA function indices (changes in LA diameter) representing reservoir and booster pump functions were initially enhanced and then impaired in healthy dogs given cardiac volume loading with IV infusion of dextran \[[@r15]\]. During later phases of volume load, results of previous studies suggested that LA afterload, as suggested by an increase in PCWP in this study, could suppress LA booster pump (afterload mismatch of the LA might be induced) and reservoir functions \[[@r15], [@r27]\]. Indeed, those LA functional parameters during reservoir and booster pump function in this study were not decreased from baseline in the later phase. This observation might imply that the enhancing effect of the volume load of those functions could have been offset by its suppressive effect \[[@r27]\].
There have been conflicting reports between LA function indices determined by strain imaging with 2D-STE in humans. A previous study enrolling infants with patent ductus arteriosus demonstrated that cardiac volume overload secondary to this disease was associated with the impairment of reservoir and booster pump functions evaluated with LA strains and SRs \[[@r4]\]. On the contrary, another previous study including healthy humans showed that the acute decrease in cardiac volume load caused by a tilt maneuver was associated with the impairment of reservoir, conduit, and booster pump functions assessed on the basis of LA strains \[[@r10]\]. The discrepancies between the present study and above-mentioned previous studies might have resulted from the difference in the duration of the change in cardiac volume load, the difference in the degree of the change in cardiac volume load, or the difference in the direction of the change in cardiac volume load.
Several limitations should be noted in this study. First, our study lacked the use of a gold standard for measurement of the mechanical function of the LA. We used a Swan-Ganz catheter and right heart catheterization for measuring hemodynamic variables. LA pressure-volume loop analysis would have strengthened the results. Although the LA pressure-volume loop analysis may be needed to measure LA intrinsic properties \[[@r31]\], the invasive nature of this method and the expertise required to obtain appropriate data limit the application of this approach. In addition, the measurement of LV properties was made only by echocardiographic parameters. Second, we did not measure the LA function indices after examination of diuretic administration. Therefore, it remains unknown whether the LA deformation indices would be decreased after the volume unloading effect. Third, the number of dogs enrolled in study was relatively small, such that we could not determine a significant change in some echocardiographic data. The smaller sample could explain the low power of the statistics. Moreover, we could not clearly determine the indices of conduit and booster pump function in some images, which might possibly be related to the observed increase in the heart rate caused by acute volume loading \[[@r18]\]. Fourth, a possible effect of general anesthesia on cardiac function could not be excluded. As described in a previous study \[[@r9]\], isoflurane alters the active and passive mechanical properties of the LA. This agent can depress LA myocardial contractility, delay relaxation, and also enhance reservoir function \[[@r9]\]. Therefore, the enhancing effect of acute volume loading on LA phasic function in this study might have been affected by the use of isoflurane. Fifth, it is possible that our results could not be extrapolated to dogs with higher PCWP. In the present study, the degree of the elevation of PCWP was relatively mild (mean PCWP of about 15 mmHg). In general, a mean PCWP greater than 15 to 25 mmHg can be associated with left-sided heart failure \[[@r7]\]. Furthermore, this was a clinically normal animal study of the acute volume load effect intervention on the LA strain imaging that we could not exclude the possibility of chronic adaptations in awake clinical dogs. Such chronic adaptations may lead to different response to changes in loading condition.
In conclusion, the LA phasic functions assessed by strain imaging with 2D-STE are affected by changes in acute volume loading condition and correlated with invasive measurement of PCWP in clinically normal dogs. Therefore, the diagnosis on the basis of LA phasic function obtained by strain and SR analysis obtained from 2D-STE should be considered with caution. Strain variables obtained from 2D-STE may serve as a sensitive indicator and provide additional information of the dogs with acute volume loading-stated heart diseases.
The authors did not receive any specific grant for this research from funding agencies in the public, commercial, or not-for-profit sectors. This research was presented as an oral presentation at the 161st meeting of the Japanese Society of Veterinary Science (JSVS) and 2018 ACVR/IVRA Joint Scientific Conference, Omni, Fort Worth, Texas, U.S.A.
[^1]: These authors contributed equally to this work.
| {
"pile_set_name": "PubMed Central"
} |
Symptomatic patients {#s1}
====================
Background {#s1a}
----------
Patients are traditionally considered 'recently symptomatic' if they have suffered a carotid territory transient ischaemic attack or stroke within the preceding 6 months. In the 1980s, there was controversy as to whether carotid endarterectomy (CEA) conferred any benefit over best medical therapy (BMT) in patients with an ipsilateral carotid stenosis. Two landmark randomised controlled trials (RCTs), the European Carotid Surgery Trial (ECST) and the North American Symptomatic Carotid Endarterectomy Trial (NASCET), determined that CEA conferred significant benefit over BMT in patients with an ipsilateral 50%--99% internal carotid artery (ICA) stenosis,[@R1] using the NASCET method for measuring carotid stenosis severity.[@R2] Subgroup analyses suggested that it was possible to identify certain imaging/clinical features that were associated with a higher risk of stroke on BMT.[@R3] Clinical features of increased benefit conferred by CEA include: increasing age (especially patients aged \>75 years), recency of symptoms, male sex, hemispheric versus ocular symptoms, cortical versus lacunar stroke and increasing medical comorbidities.[@R3] Imaging features associated with an increased risk of stroke on medical therapy include: irregular versus smooth plaques, increasing stenosis severity (but not subocclusion), contralateral occlusion, tandem intracranial disease and a failure to recruit the intracranial collateral circulation.[@R3]
CEA versus CAS in recently symptomatic patients {#s1b}
-----------------------------------------------
### 30-day outcomes {#s1b1}
Nine RCTs recruited symptomatic patients only,[@R4] while five randomised both symptomatic and asymptomatic patients between CEA and carotid artery stenting (CAS).[@R13] The most influential national/international RCTs comparing CEA with CAS in symptomatic patients include: the Endarterectomy Versus Angioplasty in Patients with Symptomatic Severe Carotid Stenosis (EVA-3S) trial, the Stent-Protected Angioplasty versus Carotid Endarterectomy (SPACE) study, the International Carotid Stenting Study and the Carotid Revascularisation versus Stenting Trial (CREST).[@R8] The principle 30-day endpoints for these four RCTs are detailed in [table 1](#T1){ref-type="table"}.
######
30-day risks following CEA and CAS in trials that randomised \>500 recently symptomatic patients into EVA-3S, SPACE, International Carotid Stenting Study (ICSS) and CREST[@R8]
30-day risks EVA-3S[@R8] SPACE[@R9] ICSS[@R11] CREST\*[@R18]
------------------------ ------------- ------------ ------------ --------------- ------ ------ ------ ------
Death 1.2% 0.8% 0.9% 1.0% 0.8% 2.3%
Any stroke 3.5% 9.2% 6.2% 7.2% 4.1% 7.7% 3.2% 5.5%
Death/any stroke 3.9% 9.6% 6.5% 7.4% 4.7% 8.5% 3.2% 6.0%
Death/disabling stroke 1.5% 3.4% 3.8% 5.1% 3.2% 4%
Death/stroke/MI 5.2% 8.5% 5.4% 6.7%
Cranial nerve injury 7.7% 1.1% 5.3% 0.1% 5.1% 0.5%
\*Only includes symptomatic patients from CREST.
CAS, carotid artery stenting; CEA, carotid endarterectomy; CREST, Carotid Revascularisation versus Stenting Trial; EVA-3S, Endarterectomy Versus Angioplasty in Patients with Symptomatic Severe Carotid Stenosis; MI, myocardial infarction; SPACE, Stent-Protected Angioplasty versus Carotid Endarterectomy.
[Table 2](#T2){ref-type="table"} details ORs (95% CIs) for 30-day death/stroke in the four main RCTs, where only the symptomatic patients randomised within CREST were included within the meta-analysis.
######
ORs (95% CIs) for 30-day death/stroke for CEA versus CAS in EVA-3S, SPACE, ICSS and CREST\*
Trial OR (95% CI)
--------------- ---------------------
EVA-3S[@R8] 0.38 (0.16 to 0.84)
SPACE[@R9] 0.89 (0.55 to 1.42)
ICSS[@R11] 0.53 (0.35 to 0.80)
CREST\*[@R18] 0.52 (0.29 to 0.92)
Meta-analysis 0.59 (0.42 to 0.81)
\*Only symptomatic patients from CREST were included.
CAS, carotid artery stenting; CEA, carotid endarterectomy; CREST, Carotid Revascularisation versus Stenting Trial; EVA-3S, Endarterectomy Versus Angioplasty in Patients with Symptomatic Severe Carotid Stenosis; SPACE, Stent-Protected Angioplasty versus Carotid Endarterectomy.
The Carotid Stent Trialists Collaboration (CSTC) have undertaken a number of subgroup analyses to determine factors associated with poorer outcomes after CAS and CEA, which may influence how individual symptomatic patients are treated.
#### CAS operator experience {#s1b1a}
In EVA-3S, SPACE and ICSS, the 30-day rate of death/stroke was not influenced by lifetime CAS practitioner stenting experience (P=0.8). However, the 30-day rate of death/stroke was significantly higher in symptomatic patients who were treated by CAS practitioners with a low annual CAS volume (≤3 procedures per annum; 30-day death/stroke=10.1%; adjusted risk ratio=2.30 (95% CI 1.36 to 3.87)), versus intermediate in-trial CAS volumes (3--6 procedures per annum; 30-day death/stroke=8.4%; adjusted risk ratio=1.93 (95% CI 1.14 to 3.27)), compared with patients treated by higher annual in-trial volume practitioners (\>6 procedures per year; 30-day death stroke=5.1%).[@R19]
#### Effect of age in recently symptomatic patients {#s1b1b}
The CSTC pooled data from EVA-3S, SPACE, ICSS and CREST, regarding the effect of increasing age on 30-day death/stroke after CEA and CAS.[@R20] There was no evidence of any association between increasing patient age and an increased risk of death/stroke after CEA. However, increasing age was associated with increasing procedural risks in symptomatic patients undergoing CAS. Compared with CAS patients aged \<60 years, performing CAS in patients aged 70--74 years was associated with a significant increase in 30-day death/stroke (OR 4.01 (95% CI 2.19 to 7.32)). In CAS patients aged \>80 years (compared with CAS patients\<60 years), the 30-day risk of death/stroke was increased by 4.15 (95% CI 2.20 to 7.84).[@R20]
Compared with CEA, 30-day rates of death/stroke were no different after CAS in recently symptomatic patients aged \<70 years of age. However, there was a progressive increase in the risk of death/stroke after CAS (compared with CEA) which became significant at age 70--74 (OR 2.09 (95% CI 1.32 to 2.32)), increasing to an OR of 2.43 (95% CI 1.35 to 4.38) for CAS patients aged \>80 years.[@R20]
#### Recency of symptoms {#s1b1c}
There is now a worldwide drive towards performing carotid interventions as soon as possible after onset of symptoms. This is because evidence suggests that the risk of stroke in the first 7--14 days after onset of symptoms is significantly higher than previously thought, while delays to CEA are associated with significant reductions in the benefit conferred by CEA.[@R3] The CSTC undertook an individual patient meta-analysis of outcomes, stratified for the time delay between symptom onset and undergoing CEA/CAS.[@R21] Patients undergoing CAS within 0--7 days after symptom onset were significantly more likely to suffer a perioperative stroke (9.4%), compared with CEA (2.8%) (OR 3.4 (95% CI 1.01 to 11.8)). Patients undergoing CAS within 8--14 days after symptom onset were also significantly more likely to suffer a perioperative stroke (8.1%) compared with CEA (3.4%) (OR 2.4 (95% CI 1.0 to 5.7)).[@R21]
### Late outcomes after CEA/CAS in symptomatic patients {#s1b2}
#### Late ipsilateral stroke {#s1b2a}
Each of the four largest RCTs have shown that once the perioperative period has elapsed, late rates of ipsilateral stroke were no different to CEA, indicating that CAS was as durable as CEA.[@R9]
#### Late survival {#s1b2b}
In CREST, CEA was associated with a 2.3% risk of perioperative myocardial infarction (MI), which was significantly higher than the 1.1% observed after CAS (OR 2.0 (95% CI 1.06 to 3.8), P=0.03)).[@R14] In a CREST subgroup analysis, patients suffering a perioperative MI faced a threefold increase in late mortality (HR 3.4 (95% CI 1.7 to 6.0), P=0.001).[@R25] This was interpreted at the time as meaning that anyone with a history of history of cardiovascular disease should preferentially undergo CAS, rather than CEA.[@R25]
However, reduced survival after a perioperative MI needs to be balanced against a similar effect of a perioperative stroke on late survival. In CREST, CAS was associated with a 4.1% risk of perioperative stroke, which was significantly higher than the 2.3% observed after CEA (OR 1.79 (95% CI 1.14 to 2.82), P=0.01).[@R14] In a further CREST subgroup analysis, patients suffering a perioperative stroke also faced a significant increase in late mortality (HR 2.78 (95% CI 1.63 to 4.76)).[@R26] In a separate meta-analysis, Vincent *et al* reported that CAS was associated with a 0.3% absolute reduction in perioperative MI, which was offset by a 1.8% increase in perioperative stroke.[@R27]
Translating evidence into clinical practice in symptomatic patients {#s1c}
-------------------------------------------------------------------
[Table 3](#T3){ref-type="table"} summarises the 2018 European Society for Vascular Surgery (ESVS) recommendations for the management of symptomatic carotid disease.[@R28] As can be seen, the guidelines advise that there is a role for both CEA and CAS, but the levels of evidence are slightly lower for CAS than for CEA. This is because 30-day risks of death/stroke in the RCTs were significantly higher after CAS than after CEA, and there remain concerns that results obtained in the RCTs may not be generalisable into routine clinical practice. In a systematic review, Paraskevas *et al* observed that 13/18 administrative dataset registries (72%) reported 30-day death/stroke rates in excess of the recommended 6% risk threshold following CAS in symptomatic patients, while 5/18 (28%) reported stroke rates in excess of 10%. This compares with 1/18 registries, which reported 30-day death/stroke rates exceeding 6% in patients undergoing CEA.[@R29]
######
2018 ESVS recommendations for managing patients with symptomatic carotid artery disease[@R28]
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ----------- ---------
CEA is recommended in patients reporting carotid territory symptoms \<6 months and who have a 70%--99% carotid stenosis, provided the documented procedural death/stroke rate is \<6%. Class I Level A
CEA should be considered in patients reporting carotid territory symptoms \<6 months and who have a 50%--69% carotid stenosis, provided the documented procedural death/stroke rate is \<6%. Class IIa Level A
It is recommended that most patients who have suffered carotid territory symptoms \<6 months and who are aged \>70 years and who have 50%--99% stenoses should be treated by CEA, rather than by CAS. Class I Level A
When revascularisation is indicated in patients who with carotid territory symptoms \<6 months and who are aged \<70 years, CAS may be considered an alternative to CEA, provided procedural death/stroke rates are \<6%. Class IIb Level A
When revascularisation is considered appropriate in symptomatic patients with 50%--99% stenoses, it is recommended that this be performed as soon as possible, preferably within 14 days of symptom onset. Class I Level A
Patients who are to undergo revascularisation within the first 14 days after onset of symptoms should undergo CEA, rather than CAS. Class I Level A
In recently symptomatic patients with 50%--99% stenoses and anatomical and/or medical comorbidities that are considered by the multidisciplinary team to make them 'higher-risk for CEA, CAS should be considered as an alternative to endarterectomy, provided the documented procedural death/stroke rate is \<6%. Class IIa Level B
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ----------- ---------
The colour of the text boxes identifies the class and level of evidence.
CAS, carotid artery stenting; CEA, carotid endarterectomy; CREST, Carotid Revascularisation versus Stenting Trial; ESVS, European Society for Vascular Surgery.
Asymptomatic patients {#s2}
=====================
Background {#s2a}
----------
Patients considered to be asymptomatic have either reported no carotid territory symptoms at any time in the past, or at least 6 months have elapsed since the most recent symptom. Two landmark RCTs, the Asymptomatic Carotid Atherosclerosis Study (ACAS) and the Asymptomatic Carotid Surgery Trial (ACST), determined that CEA conferred a small but significant benefit over BMT in patients with an ipsilateral 60%--99% ICA stenosis.[@R30] Unlike in NASCET and ECST, it was more difficult to identify subgroups of patients who were at higher (or lower) risk of stroke if treated medically. The available data suggested that males gained greater benefit than females and that patients aged \>75 years gained no benefit from CEA. Interestingly, the presence of a contralateral occlusion and increasing stenosis severity was not associated with an increased risk of late stroke on medical therapy in the RCTs.[@R32]
CEA versus CAS in asymptomatic patients {#s2b}
---------------------------------------
### 30-day outcomes {#s2b1}
Four RCTs exclusively randomised asymptomatic patients,[@R33] while five included asymptomatic patients within the trial as well as symptomatic patients.[@R13] In the latter studies, outcomes were not always stratified for symptom status. [Table 4](#T4){ref-type="table"} details the main 30-day outcomes from five RCTs where data were provided for asymptomatic patients.
######
30-day morbidity and mortality in randomised trials comparing CEA and CAS in asymptomatic patients
30-day outcomes Lexington[@R34] CREST-1\*[@R18] ACT-1[@R35] SPACE-2[@R33] Mannheim[@R36]
------------------------ ----------------- ----------------- ------------- --------------- ---------------- ------ ------ ------ ------ ------ ------
Death/stroke 0% 0% 1.4% 2.5% 1.7% 2.9% 2.0% 2.5% 0.0% 1.5% 2.9%
Death/disabling stroke 0% 0% 0.3% 0.5% 0.6% 0.6%
\*Only asymptomatic patients in CREST-1 were included.
ACT-1, Asymptomatic Carotid Trial 1; CAS, carotid artery stenting; CEA, carotid endarterectomy; SPACE, Stent-Protected Angioplasty versus Carotid Endarterectomy.
### Late outcomes {#s2b2}
The Lexington study, CREST and ACT-1 observed that once the perioperative period had elapsed, there was no difference in rates of late ipsilateral stroke, suggesting that CAS was as durable as CEA.[@R22]
Translating evidence into clinical practice {#s2c}
-------------------------------------------
Unlike the symptomatic RCTs, which continue to retain the same relevance in the modern era, there are concerns that the ACAS and ACST trials (which recruited patients up to 25 years ago) may not be as relevant as when published in 1995 and 2004, respectively.[@R30] This is mainly because of increasing evidence that the risk of stroke on 'modern BMT' may not be as high as previously thought and there is evidence that the annual risk of stroke on BMT may have declined by about 70% since ACAS first reported in 1995.[@R32] These concerns were recognised in the 2018 ESVS carotid guidelines where it was recommended that only patients with one or more clinical and/or imaging features that might make them higher risk for stroke on BMT should be considered for CEA or CAS.[@R28] These imaging and clinical criteria are summarised in [table 5](#T5){ref-type="table"} and readers are referred to the 2018 ESVS carotid guidelines, where greater detail has been provided regarding the magnitude of benefit (in terms of stroke reduction) associated with each of these clinical/imaging parameters.[@R28]
######
2018 ESVS Guidelines: clinical/Imaging features associated with an increased risk of stroke in patients with asymptomatic carotid stenosis treated medically[@R28]
------------ -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Clinical History of contralateral TIA or stroke
CT/MRI ipsilateral 'silent' infarction
Ultrasound Stenosis progression\>20%; spontaneous embolisation on TCD; impaired cerebral vascular reserve; large volume plaques (\>80 mm^2^); predominantly echolucent plaques; large juxta-luminal black area (\>8 mm^2^)
MRI Intraplaque haemorrhage
------------ -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
ESVS, European Society for Vascular Surgery; TCD, transcranial Doppler ultrasound; TIA, transient ischaemic attack.
While this decision to target CEA/CAS into a smaller cohort of asymptomatic patients has not always met with universal approval,[@R38] it was necessary as (currently) 95% of all asymptomatic patients undergoing a carotid intervention ultimately undergo an unnecessary intervention.[@R32] Interestingly, the American Heart Association guidelines advise that only 'highly selected' asymptomatic patients should be considered for CEA (or CAS), but they have never defined exactly what 'highly selected' means.[@R39]
[Table 6](#T6){ref-type="table"} summarises the 2018 ESVS recommendations for the management of asymptomatic carotid disease. As with symptomatic patients, the ESVS guidelines advise that there is a potential role for both CEA and CAS, but the levels of evidence are slightly less for CAS than for CEA. This is because 30-day risks of death/stroke in the largest RCTs, which used credentialed (experienced CAS practitioners),[@R18] were only just within the accepted 3% risk threshold and there remain concerns that the results obtained in the RCTs may not be generalisable into routine clinical practice. In a systematic review, Paraskevas *et al* observed that 9/21 administrative dataset registries (43%) reported 30-day death/stroke rates in excess of the recommended 3% risk threshold after CAS in asymptomatic patients, while 7/21 (33%) reported stroke rates in excess of 4%. This compares with 1/21 registries which reported 30-day death/stroke rates exceeding 3% in patients undergoing CEA.[@R29]
######
2018 ESVS recommendations for managing patients with asymptomatic carotid artery disease[@R28]
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ----------- ---------
In 'average surgical risk' patients with an asymptomatic 60%--99% stenosis, CEA should be considered in the presence of 1+ imaging characteristics that may be associated with an increased risk of late ipsilateral stroke\*, provided perioperative stroke/death rates are \<3% and the patient's life expectancy exceeds 5 years. Class IIa Level B
In 'average surgical risk' patients with an asymptomatic 60%--99% stenosis in the presence of 1+ imaging characteristics that may be associated with an increased risk of late ipsilateral stroke\*, CAS may be an alternative to CEA, provided perioperative stroke/death rates are \<3% and the patient's life expectancy exceeds 5 years. Class IIb Level B
CAS may be considered in selected asymptomatic patients who have been deemed by the multidisciplinary team to be 'high-risk for CEA' and who have an asymptomatic 60%--99% stenosis in the presence of 1+ imaging characteristics that may be associated with an increased risk of late ipsilateral stroke\*, provided procedural risks are \<3% and the patient's life expectancy exceeds 5 years. Class IIb Level B
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ----------- ---------
\*See [table 5](#T5){ref-type="table"} for clinical/imaging features.
The colour of the text boxes identifies the class and level of evidence.
CAS, carotid artery stenting; CEA, carotid endarterectomy; ESVS, European Society for Vascular Surgery.
**Funding:** This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
**Competing interests:** None declared.
**Provenance and peer review:** Commissioned; internally peer reviewed.
**Guest chief editor:** J David Spence
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#sec1-1}
============
Thrombosis of the cerebral venous sinuses (CVT) has been described in adults and children with nephrotic syndrome.\[[@ref1][@ref2]\] Acute post-infectious glomerulonephritis (APIGN), the prototype of acute nephritic syndrome is not reported to increase the risk for venous thrombosis. Here we report a case of CVT in a patient with APIGN.
Case Report {#sec1-2}
===========
A 13-year-old girl presented with recent onset generalized edema, decreased urine output, high-colored urine, and persistent vomiting. Two weeks ago she had fever and cough, which was successfully treated with a short course of antibiotics. On admission her blood pressure was 160/100 mm of Hg, urine showed 2+ protein and dysmorphic red blood cells. The other relevant investigations are shown in [Table 1](#T1){ref-type="table"}. Ultrasound abdomen revealed normal sized kidneys. She was treated with anti-hypertensives and diuretics. Over next 3 days, her blood pressure came down and urine output improved, but serum creatinine increased to 5.1 mg/dl. Hence, she was started on intravenous methylprednisolone injections at a dose of 750 mg/day for 3 days, followed by oral prednisolone at 1 mg/kg.
######
Laboratory parameters at admission
On the 5^th^ day after initiation of steroid therapy, she developed recurrent episodes of generalized tonic-clonic seizures associated with altered sensorium. Her blood pressure at the time of seizures was 130/80 mm of Hg. On examination, she was found to have left sided hemiparesis. Non-enhanced computed tomography (CT) brain showed an infarct in the left temporo-parietal region, mild midline shift, and cerebral edema. Hyperdensities were observed in the sagittal sinus, right sigmoid, and transverse sinuses \[[Figure 1a](#F1){ref-type="fig"}\]. A CT venogram showed an empty delta sign with filling defects in right transverse and sigmoid sinus extending to the right internal jugular vein \[Figures [1b](#F1){ref-type="fig"}, [2a](#F2){ref-type="fig"} and [b](#F2){ref-type="fig"}\]. She was started on anticoagulation with continuous infusion of unfractionated heparin (UFH). Hemodialysis was initiated through right femoral catheter in view of persistent renal failure. The seizures were controlled and sensorium improved over the next 1-week. Heparin was switched over to warfarin at the end of 7 days. She was supported with hemodialysis for 1-week, subsequently her renal function started to improve. After the initial decline, the serum creatinine remained static at 4.8 mg/dl. Her ANA, lupus anticoagulant (LA), anticardiolipin antibody, and ANCA were negative. We could not proceed with thrombophilia work up due to financial constraints.
{#F1}
{#F2}
Renal biopsy revealed enlarged glomeruli showing endocapillary proliferation, with neutrophils and occasional eosinophils in the capillary lumina. Glomerular basement membrane (GBM) thickness was normal. A segmental cellular crescent was present in one glomerulus. Tubules, interstitium, and vessels were normal. Immunofluorescence microscopy (IF) showed diffuse granular deposits of IgG and C3 (3+ intensity) along the capillary loops. Tubules showed simplification of the lining epithelium. Interstitium and vessels were unremarkable. The renal biopsy was consistent with post-infectious glomerulonephritis \[[Figure 3a](#F3){ref-type="fig"}--[d](#F3){ref-type="fig"}\].
{#F3}
Even though the biopsy was suggestive of post-infectious glomerulonephritis, we decided to continue corticosteroids in view of incomplete recovery of renal function. The patient was discharged on prednisolone 40 mg/day, warfarin and antiepileptics. Over the next 6 weeks, her serum creatinine decreased to 1 mg/dl. Her erythrocyte sedimentation rate decreased to 20 mm/1^st^ h. Prednisolone was given for a total duration of 3 months. Anti-epileptics and anticoagulants were stopped after 6 months. Currently, the patient is off anticoagulation for the last 8 months; without any recurrent episodes of thrombosis. On last follow-up, her blood pressure was 120/80 mm of Hg, serum creatinine 0.8 mg/dl, and 24 h urine protein was \<150 mg/dl with normal urine sediment.
Discussion {#sec1-3}
==========
CVT is considered to be less common when compared to thrombosis of other vascular beds. Apart from a few case series and single case reports, there are no reliable estimates on the prevalence of CVT in glomerular diseases. CVT has been described in patients with idiopathic nephrotic syndrome, minimal change disease, and membranous nephropathy.\[[@ref1][@ref2]\] Nephrotic syndrome induces a hypercoagulable state due to increased blood viscosity, elevated levels of procoagulant proteins like fibrinogen, factor V, VIII, and X, and deficiency of coagulation inhibitors like antithrombin III, protein S and C. CVT is reported to occur at the onset of nephrotic syndrome as well as during relapses. A poorly controlled nephrotic state, as well as steroid resistance confer a high risk for the development of CVT.
Apart from disease related factors, drugs can also trigger venous thrombosis. The volume depletion induced by aggressive diuretic therapy can trigger thrombosis in individuals with significant proteinuria. Exogenous corticosteroid administration is associated with increased thromboembolic tendencies in the initial days of exposure.\[[@ref3]\] There have been reports of high dose corticosteroid therapy predisposing to CVT in patients with multiple sclerosis with no additional risk factors for thrombosis.\[[@ref4]\] The proposed mechanisms include increased release of von Willebrand factor, shortened prothrombin time (PT) and impaired fibrinolytic activity.
The association between PIGN and CVT does not appear to be direct. The patient had hypoalbuminemia and elevated lipid levels suggestive of a nephrotic state. The coexisting acute tubular necrosis would have been responsible for the lower rates of urinary protein excretion. The volume depletion induced by loop diuretics in conjunction with low serum albumin levels, elevated lipids, and high dose corticosteroids might have precipitated CVT in this patient. There was no evidence of SLE or antiphospholipid antibodies. We did not test for deficiencies of protein C, protein S, antithrombin III, factor V, and prothrombin gene mutations due to financial constraints. The current evidence does not recommend routine thrombophilia testing in all patients with first episode of CVT as it has a low value for predicting recurrences.\[[@ref5][@ref6][@ref7]\] It is possible that the patient might be having a preexisting mild thrombophilia, which in the presence of appropriate triggers might have precipitated CVT.
Seizure is not an uncommon complication of PIGN. The most common cause of seizures in PIGN is hypertensive encephalopathy. There have been few reports of posterior reversible encephalopathy syndrome causing recurrent seizures in PIGN. Since CVT can also present as seizures, a high index of suspicion and appropriate early investigations are required to identify this entity.
Conclusions {#sec1-4}
===========
We wanted to highlight the occurrence of CVT in a patient with APIGN, a disease which is not usually associated with a hypercoagulable state. The nephrotic state resulting from APIGN combined with the volume depletion would have precipitated CVT in this patient. It is possible that CVT might be underdiagnosed because of the wide variability in clinical presentation. Recognizing this rare complication is important because of the potentially devastating outcome, if left untreated.
Financial support and sponsorship {#sec2-1}
---------------------------------
Nil.
Conflicts of interest {#sec2-2}
---------------------
There are no conflicts of interest.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-sensors-20-02144}
===============
In the 4th industrial revolution, Structural Health Monitoring(SHM) is becoming a hot issue in the construction industry \[[@B1-sensors-20-02144],[@B2-sensors-20-02144]\] in which nondestructive evaluation and serviceability monitoring are essential features. Nondestructive evaluation plays a critical role in assuring that structural components and systems perform in a reliable and cost-effective fashion. This mechanism does not affect the future usefulness of the object or material. On the other hand, serviceability monitoring is performed in real time throughout its service life span \[[@B3-sensors-20-02144],[@B4-sensors-20-02144]\]; this is directly related to controlling the structural responses caused by either deflection, cracks, vibration, creep, or a combination of them. Various techniques have been utilized for the assessment of structure performance \[[@B5-sensors-20-02144]\]. Sensors are essential components and have different purposes based on the nature of the techniques. In general, we can classify sensors into two types based on the connection. Contact sensors which are commonly known for having physical interactions with the target structure. A Linear Variable Differential Transformer(LVDT), piezoelectric transducer, fiber optic sensor and acoustic emission sensor are common examples of this type \[[@B6-sensors-20-02144],[@B7-sensors-20-02144],[@B8-sensors-20-02144]\]. On the other hand, non-contact sensors are known for acquiring responses from the target material without making direct or indirect contact. Laser sensor systems, drones with vision-based sensors using cameras, wireless rechargeable sensor networks, radar sensor networks and lidar sensor systems are grouped as this type \[[@B9-sensors-20-02144],[@B10-sensors-20-02144],[@B11-sensors-20-02144]\]. Recently, non-contact sensors are being commonly utilized due to their portability, easy use in harsh surroundings, and so on. Knowing about the deflection of a beam or any structure using TLS has a big advantage from the perspective of site conditions. For the sake of describing the benefits of this study in the real world, aged structures, considering their spatial positions and site conditions, are not safe due to perilous structural conditions, inconvenience, insecurities and slippery site conditions. Consequently, the Light Detection And Ranging (LiDAR) system has become prominent in structural health monitoring. Despite such sensors being used for detection, measurement and characterization of hidden and/or apparent defects using advanced techniques, there are still many questions regarding the dimensions of the sensors for structural health monitoring.
Terrestrial Laser Scanning (TLS), or lidar, is a crucial non-contact optical sensor that analyzes the structural Three Dimensional(3D) shape in terms of a very dense 3D point cloud. This technology is a recent innovation in the spatial information of data acquisition, allowing for a scanned area to be digitally captured with unprecedented resolution and precision. However, the acquired data are influenced by several factors. Among them, random errors due to inherent physical properties are often difficult to eliminate while error due to environmental and scanning geometry issues can be removed through analysis \[[@B12-sensors-20-02144]\]. Consequently, attention should be given to data acquisition by considering the position, incidence angle, and color of a specimen \[[@B13-sensors-20-02144]\]. As a result, it is possible to dwindle noises tremendously.
In recent years, various approaches have been used to estimate the elements for serviceability limit states such as accumulated deflection, crack and dynamic displacement \[[@B14-sensors-20-02144],[@B15-sensors-20-02144],[@B16-sensors-20-02144]\]. Several researchers have published papers concerning the serviceability assessment in support of structural health monitoring \[[@B17-sensors-20-02144],[@B18-sensors-20-02144],[@B19-sensors-20-02144]\]. Park et al. conducted an experiment to define the deflection of a steel structure element via TLS using the geometrical shape of the specimen \[[@B20-sensors-20-02144]\]. Even though the model of Cabaleiro et al. is not valid due to torsion, they tried to model a deformed beam caused by concentric loads and torsional forces on the specimen after scrutinizing the max deflection with respect to the allowable building codes \[[@B21-sensors-20-02144]\]. Some researchers proposed how to estimate the deflection of a structural element, i.e., a beam, by integrating photogrammetry with TLS. For instance, Gordon measured the beam deflection with respect to benchmark photogrammetric data \[[@B22-sensors-20-02144]\]. Zogg and Ingensand tried to monitor a deformation of real structure called Felsenau Viaduct(CH) bridge which is a part of Swiss highway via with TLS. The deformation on their study is mainly caused by settlement and tilting of a structure. Eventually, they showed that TLS could replace the area wide precise levelling in monitoring of a structure deformation since the maximum difference in between these two system is less than 1 mm \[[@B23-sensors-20-02144]\]. Olsen et al. conducted a damage assessment for full scale structural test specimen by identifying volumetric change and its deformation \[[@B24-sensors-20-02144]\]. Cabaleiro et al. utilized TLS data for checking of a deflection and stresses caused by torsion especially for open cross section which have a very low torsional strength. They compared the result obtained from the proposed methodology with the measurement taken during an experiment and Finite Element Modeling(FEM). As a result, they could show that their proposed algorithm is much closer to the measurement taken directly which is considered as a ground truth data \[[@B25-sensors-20-02144]\].
Naturally, Lidar data are highly vulnerable and easily affected by noise \[[@B26-sensors-20-02144]\]. Different scholars have been studying the factors which affect the Lidar data regarding structural health monitoring \[[@B13-sensors-20-02144],[@B27-sensors-20-02144]\]. On the other hand, some researchers have used different types of denoising methods and optimization techniques in order to diminish their effect \[[@B28-sensors-20-02144],[@B29-sensors-20-02144]\]. Notice that, outliers and noise are differing conceptually. Indeed, these two terms have really ambiguous meaning to differentiate by the researchers. However, Sagado et al. defined outliers and noise individually. According to their definition, an outlier is a data point which is different from the remaining data whereas noise can be defined as mislabeled examples (class noise) or errors in the values of attributes (attribute noise) \[[@B30-sensors-20-02144]\]. Genetic Algorithm(GA) is also one of the crucial optimization methods used in order to attain the target effectively without the noise affection. GA has been conducted in different topics of structural health monitoring by different scholars. Optimization of sensor placement which provides the best possible performance is one of many critical topics \[[@B31-sensors-20-02144]\]. However, estimating a structural deflection for the purpose of assessing the performance via a genetic algorithm is not prominently conducted so far. Rather, many researchers have performed this optimization method to determine a structure deflection both analytically and numerically \[[@B32-sensors-20-02144]\]. GA can be also incorporated with a regression methodology to increase the robustness of curve fitting. Indeed, there are many techniques that are capable to increase the robustness of a curve fitting technique when we have numerous numbers of outliers in the data; Least Absolute Deviation (LAD), M-estimation, and S estimation are popular schemes which can considerably improve estimation precision \[[@B33-sensors-20-02144]\].
This study presents an effective algorithm for measuring the structural deflection by enriching previous studies. Improvements are made according to the following five main steps of the procedure listed; the acquisition of TLS data for the loading and unloading scenarios, fitting of a plane for the point cloud acquired during the unloading scenario using a robust genetic algorithm, transformation of the scanner coordinates into local structural coordinates, curve fitting of transformed data for the loading case, and eventually, estimation and comparison of deflection between contact sensors. Furthermore, our research illustrates the performance of the proposed procedure with a validating experiment in which deflection measurements are simulated based on the loading scenario.
2. Basic Principles {#sec2-sensors-20-02144}
===================
2.1. Least Square Regression {#sec2dot1-sensors-20-02144}
----------------------------
A plane can be described by a normal vector $n = \left\lbrack A,B,D \right\rbrack^{T}$ perpendicular to the plane, and a vector on the plane connected to a known point $p_{1}$ = ($x_{1}$, $y_{1}$, $z_{1}$) and an arbitrary point $p_{2}$ = (*x*, *y*, *z*) is described by $p_{2} - p_{1}$ since the normal vector and vector in the plane are perpendicular to each other, the dot product of these two vectors should be null \[[@B34-sensors-20-02144]\]. $${\overset{\rightarrow}{n} \cdot \left( {\overset{\rightarrow}{P_{2}} - \overset{\rightarrow}{P_{1}}} \right) = 0}\,\,\,\because\,\,\,\left( \overset{\rightarrow}{n}\bot\left( \overset{\rightarrow}{P_{2}} - \overset{\rightarrow}{P_{1}} \right) \right)$$ For the sake of simplifying the over-determined problem, begin by removing one component by constraining the solution space. Thus we assume the coefficient D is the unit:$${Ax_{i} + By_{i} + z_{i}} + C = 0$$
Keeping the assumption that the *z*-component of the data is functionally dependent on the *x* and *y*-components and given a set of samples $\left( x_{i},y_{i},z_{i} \right)$, determine *A*, *B*, and *C* so that the plane *z* = *Ax* + *By* + *C* best fits the samples in the sense that the sum of the squared errors between the $z_{i}$ and the plane values $Ax_{i} + By_{i} + C$ is minimized \[[@B35-sensors-20-02144]\]. This can be written via Equation ([3](#FD3-sensors-20-02144){ref-type="disp-formula"}). $$E\left( A,B,C \right) = {\sum\left\lbrack {z_{i} - \left( {Ax_{i} + By_{i} + C} \right)} \right\rbrack}^{2}$$
2.2. Genetic Algorithm {#sec2dot2-sensors-20-02144}
----------------------
Stochastic optimization is a process of seeking a maximum or minimum value of a mathematical or statistical function in the presence of randomness. GA is one of the common method which include in this kind of optimization techniques. It is used to find the optimal solution(s) to a given computational problem maximizing or minimizing a function. A genetic algorithm is a random-based classical evolutionary algorithm \[[@B36-sensors-20-02144]\]. This principle of continual improvement over generations is utilized by evolutionary algorithms to optimize solutions to a problem. In the initial generation, a population composed of different individuals is generated randomly or by other methods. An individual is a solution to the problem, which can vary in quality: the quality of an individual with regard to the problem is called the fitness, which reflects the adequacy of the solution to the problem to be solved. The higher the fitness of an individual, the more likely it is to pass some or all of its genotype to individuals of the next generation. Increasing the robustness of an algorithm improves the output. The primary distinguishing features of such algorithms are encoding, a selection mechanism, a crossover mechanism, a mutation mechanism, and a culling mechanism. Such algorithms can optimize multiple objectives simultaneously and can be used as black boxes since they do not assume any properties of the mathematical model to be optimized \[[@B37-sensors-20-02144]\]; their only real limitation is the computational complexity.
3. Proposed Approach for Computation of Deflection {#sec3-sensors-20-02144}
==================================================
The proposed method is applied to measure artificial deflection of a structural element from TLS data and compare this with the contact sensor mounted during the experiment. The deflection is estimated by taking the difference in shape change between the loading and unloading scenarios. In the unloading case, first the resulting point cloud data are represented by plane and an extracted line, which help to set the spatial position of a local specimen coordinate system. Once the coordinate system is transferred to the new system, we can define and visualize a deflection curve for each loading case. The flowchart in [Figure 1](#sensors-20-02144-f001){ref-type="fig"} shows the steps needed to estimate the deflection of a structure. The subsequent subsections explain the proposed method in detail.
3.1. Acquisition of the Point Cloud and Pre-Processing {#sec3dot1-sensors-20-02144}
------------------------------------------------------
Representing or modeling real-world scenarios with virtual worlds makes it easy to apply analytical theories and visualize the effects. Among them, capturing an object using a 3D digital device, i.e., TLS in our case, enables us to measure accurately any changes in shape with magnificent resolution. Basically, The TLS point cloud tells us the spatial position (*X*,*Y*,*Z*) of a point with respect to its own scanner coordinate system. Additionally, TLS provides us with information about the scalar field that represents intensity, RGB color information, and time in some cases. Indeed, not every sensors can produce RGB color information for the users. Some brands like, RIEGL LMS-Z420i, has a camera system that detached from the main system which is used to acquired the color data of an objects. whereas in our case the system has in-built camera system which helps to produce RGB color data with the corresponding *X*, *Y*, and *Z* coordinates. In this study, only the spatial information of a point cloud is utilized for the proposed analysis to achieve the required solution. Pre-processing is a technique that improves the accuracy and resolution of data, along with filtering techniques that serve to enhance or highlight the spatial characteristics of an image data set. Manual trimming and segmentation are employed in this study since they are considered enough for analyzing the data. After removing the outliers and segmentation, the number of points on the flange became 39,730 while those on the web entity become 163,024 points. [Figure 2](#sensors-20-02144-f002){ref-type="fig"} depicts the resulting scanned point cloud representation of a specimen with the corresponding solid 3D representation of an object.
3.2. Characterizing the Point Cloud in Terms of Definable Mathematical Elements {#sec3dot2-sensors-20-02144}
-------------------------------------------------------------------------------
Our data form a point cloud set which has information about its spatial position. We only have a collection of point cloud data which describe the specimen. Therefore, representing this data by a plane and/or a surface, which can be defined by some mathematical equation, is significant for further analysis. Extracting the line which helps us to locate the local structural coordinate system, transforming the coordinate of a point from scanner coordinate system to local structural coordinate system and computing the deflection of a specimen anywhere along with the span are some tasks that can be computed whenever the acquired point cloud data is defined by mathematical plane and/or surface.
### 3.2.1. Least Squares Genetic Algorithm(LS-GA) Based Plane Fitting {#sec3dot2dot1-sensors-20-02144}
The principle of least squares sum in the curve fitting problem, which minimizes the difference between the data and model of the output, is useful. However, in this study, the basic principle of least squares is incorporated with a notable stochastic optimization known as a genetic algorithm for the purpose of decreasing the difference between the data and model \[[@B38-sensors-20-02144]\]. Consequently, this is particularly useful if we utilize an error metric in terms of the functional form and optimize it using a GA. Holland and his student stated that GA works with an encoding of the individuals throughout the algorithm \[[@B36-sensors-20-02144]\]. Originally, they constructed a GA based on binary numbers. Once we have fixed the number of chromosomes in the population, the basic algorithm operators (Selection, Crossover, and Mutation) start to conduct iteratively until it converges to optimal solutions. Currently, all these complication and processes are simplified using different platforms. Among them, the global optimization toolbox in MATLAB software is utilized for our studies. The essential parameters for running the GA in this study are \[[@B37-sensors-20-02144]\]: Fitness function: The fitness function is a mathematical formulation of the desired optimization problem. It determines how suitable a solution is. The magnitude of the residual, which defines the difference between the model and data examines the individuals formed throughout all generation. This can be described in Equation ([4](#FD4-sensors-20-02144){ref-type="disp-formula"}): $$Fitness\mspace{720mu} value = \sum\limits_{i = 0}^{N}{d_{i}}^{2} = \sum\limits_{i = 0}^{N}\left( {P_{i} - F\left( {A,B,C} \right)} \right)^{2}$$ where $F\left( A,B,C \right)$ is the point model, normally defined as the nearest data point *P*Population: The population is a set of individuals that have a chance to be the fittest among them. An individual is characterized by a set of variables called its genes. It is a basic building block of for the algorithm. In our case, this refers to the parameter of the regression model. These genes are joined into a string to form a chromosome (individuals). It is simply a series of binary numbers (0 and 1), which encode all values of parameter of the regression model. $I_{1},I_{2}$ up to $I_{200}$ that are shown in [Figure 3](#sensors-20-02144-f003){ref-type="fig"} are list of chromosomes in the population which have a binary representation of all the parameters *A*, *B*, and *C* of the plane. Our GA is one of the unique behaviors of this optimization techniques; it works only on the number of chromosomes inside the population. The number of chromosomes in the population defined by the user is labeled as the population size. The building block and their corresponding chromosomes in the population is shown in the [Figure 3](#sensors-20-02144-f003){ref-type="fig"}.Fitness scaling function: If we consider only multiplying the fitness by some factor, then as the name implies, this does not change the relationships among the population. Rather, a much more sensible fitness scaling is an affine transformation, where scaled(f) = a ∗ f + b, of the fitness value, is seen via formulae, the values are multiplied by some number and then offset by another either up or down. This parameter plays a crucial role in GA by limiting the tendency of the strongest solution to overwhelm the weaker ones thus avoiding premature convergence.Selection: The selection is a parameter for choosing two parents from the population for the purpose of mating. This process continues in every single generation until one of the stopping criteria is attained. Parents, i.e., a pair of individuals, are chosen based on their fitness score.Crossover/Reproduction: This parameter mainly depicts how the selected individuals mate with each other and create new offspring for the next generation. In this process, part of the individuals is exchanged with its mate based on the user-defined crossover probability.Elite Count: This is the number of individuals with the best fitness values in the current generation that have been retained for the next generation. Because of their robust fitness score, more copies of these individuals are achieved in the subsequent generation.Crossover fraction: The fraction of individuals in the next generation, other than the elite children, that are created via crossover.Mutation: A process of changing of a bit (gen) within a bit string (chromosome). This is done to maintain diversity within the population and prevent premature convergence.Migration: An exchange of information (exchange of individuals) between the sub-populations. [Figure 4](#sensors-20-02144-f004){ref-type="fig"} illustrates how the parameter of genetic algorithm is functioning in one generation. This feature of a genetic algorithm allows us to find the optimal parameter of a plane with respect to an error metric.
### 3.2.2. Taguchi Experimental Design {#sec3dot2dot2-sensors-20-02144}
Even though the prevalent method of modeling a plane fitting is based on deterministic ordinary differential equations, a stochastic method such as a genetic algorithm is more appropriate and well established whenever the amount of noise is significant \[[@B38-sensors-20-02144],[@B39-sensors-20-02144],[@B40-sensors-20-02144]\]. However, GAs have limitations regarding the optimality of individuals which satisfy the given fitness function unless an appropriate input parameter is selected. Consequently, to overcome this drawback, a different scheme has been utilized for tuning a suitable parameter for a genetic algorithm \[[@B41-sensors-20-02144],[@B42-sensors-20-02144],[@B43-sensors-20-02144]\]. Among various techniques, the Taguchi experimental design approach is used for tuning the parameters of the GA solver. Dr. Taguchi developed the design of an experiment based on well-defined guidelines. Taguchi's design uses orthogonal arrays that estimate the effects of factors on the response mean and variation. An orthogonal array means the design is balanced such that the factor levels are weighted equally. Taguchi uses the following convention for naming the orthogonal arrays: $L_{a}\left( b^{c} \right)$, where *L* stands for Latin square, *a* is the number of experimental runs, *b* is the number of levels of each factor, and c is the number of variables \[[@B44-sensors-20-02144]\]. We can easily assess each factor individually because of utilizing an Orthogonal Array on the design experiment. As a result, the effect of one factor does not affect the estimation of a different factor. This can reduce the time and cost associated with the experiment when fractionated designs are used. Therefore, Taguchi's design of experiments plays a crucial role in the reduction of number of trial experiments \[[@B45-sensors-20-02144]\]. The parameters in Taguchi's design are selected by considering the effects of individual value on the algorithm output. Here, nine parameters and their levels are selected and summarized in [Table 1](#sensors-20-02144-t001){ref-type="table"}.
Among the various preferences for experiment design, a robust design is identified using the signal-to-noise ratio (SNR). This is because the higher values SNR identify the control settings that minimize the effects of the noise factors. Basically, the Taguchi design of experiment is used for two optimization processes: (a) use the signal-to-noise ratio to identify those control factors that reduce variability. (b) identify control factors that shift the mean to the target and have a small or no effect on the signal-to-noise ratio \[[@B45-sensors-20-02144]\]. A SNR is suggested by Taguchi for cases in which the response standard deviation is related to the mean. The purpose of SNR in Taguchi's approach to robust parameter design is to provide an easy-to-use performance criterion that takes the process mean and variance into account. Among the numerous SNR values, "the smaller-the-better" philosophy used in this study, as shown in Equation ([5](#FD5-sensors-20-02144){ref-type="disp-formula"}). $${SNR} = - 10{log}\left\lbrack {\sum\limits_{i = 1}^{n}\left( \frac{{y_{i}}^{2}}{n} \right)} \right\rbrack$$ where; SNR is signal to noise ratio, $y_{i}$ is the response variables, and *n* is number of response values. $\sum_{i = 1}^{n}$ implies the summation over n response values at the outer array points.
### 3.2.3. Extraction of the 3D Intersection Line {#sec3dot2dot3-sensors-20-02144}
Defining the coordinates of points using the new basis is important because it allows for visualization and quantification of the real shape changes of the specimen. However, it is not easy to find out the appropriate position of a new basis (local structural coordinate system) since we have a collection of points. Therefore, considering one axis (in this case, the X-axis) coincides with intersection 3D line between the flange fitted plane $P_{f}$ and the web fitted plane $P_{w}$ is an efficient way of determining the position with the new basis as shown in the [Figure 5](#sensors-20-02144-f005){ref-type="fig"}. Suppose the direction vectors for the flange and web planes in the [Figure 5](#sensors-20-02144-f005){ref-type="fig"} are $N_{f}\left( i_{f},j_{f},k_{f} \right)$ and $N_{w}\left( i_{w},j_{w},k_{w} \right)$, respectively. Since these two planes cross each other perpendicularly, the vector product of their normal vectors is equivalent to the direction vector *S* of their line of intersection, $L_{i}$ ([6](#FD6-sensors-20-02144){ref-type="disp-formula"}). $$N_{f}\left( i_{f},j_{f},k_{f} \right) \times N_{w}\left( i_{w},j_{w},k_{w} \right) = S$$
However, this information is not enough to extract the equation of the line. Thus, picking one arbitrary known point for our case results in the intersection of the line of intersection of the two planes $L_{i}$ with the left edge of the plane $L_{l}$, and this point is inserted in to the following symmetric form in Equation ([7](#FD7-sensors-20-02144){ref-type="disp-formula"}) for 3D lines. $$\frac{x - x_{0}}{a} = \mspace{720mu}\frac{y - y_{0}}{b} = \frac{z - z_{0}}{c} = t$$ where; *a*, *b* and *c* are direction vectors of the line $L_{i}$ i.e., $S = \left( a,b,c \right)$; $x_{0}$, $y_{0}$ and $z_{0}$ are the coordinates of a point on a line, taken here as (0,−7.495,−0.402).
Accordingly, the real spatial position of the new basis (*x*′,*y*′,*z*′) can be used to determine the coordinate transformation, followed by setting the x′-axis of the local structural coordinates on the extracted intersection line of the planes, as shown in [Figure 5](#sensors-20-02144-f005){ref-type="fig"}.
### 3.2.4. Transformation of Coordinates {#sec3dot2dot4-sensors-20-02144}
To easily evaluate the real changes in shape of the specimen by visualizing the effect of loading, the reference frame is transformed from the scanner coordinate system to the local structural coordinate system. This implies that the acquired data set coordinates should be defined using the local structural coordinate system. A rigid transformation (also called an isometry) is a transformation which preserves the shape and size of the object. Reflections, translations, rotations, and/or combinations of these three transformations can be categorized as rigid transformations \[[@B46-sensors-20-02144]\]. A translation $T_{O^{\prime}O}$ is a transformation which slides an object by a fixed range from point *O* to point *O*′, as in [Figure 5](#sensors-20-02144-f005){ref-type="fig"}. All the other points move the same distance in the same direction whenever the coordinates of the point cloud are multiplied by the translation matrix, as shown by Equation ([8](#FD8-sensors-20-02144){ref-type="disp-formula"}). After a translation, which is defined as moving the origin of the scanner coordinates to the hypothetical spatial position of the local structural coordinates, a rotation is applied. A rotation $R_{\theta\_ i,O\prime}$ is a transformation that twirls an object about a fixed point *O*′, called the center of rotation, with an anlge $\theta$ in either the *x*, *y*, *z*, or a combination of them; the rotation directions are called the Euler angles \[[@B47-sensors-20-02144]\]. $$\left\lbrack T \right\rbrack = \begin{bmatrix}
1 & 0 & {\mspace{720mu}\mspace{720mu}\mspace{720mu}\begin{matrix}
0 & {\mspace{720mu}\mspace{720mu} 0} \\
\end{matrix}} \\
0 & 1 & {\mspace{720mu}\begin{matrix}
{\mspace{720mu}\mspace{720mu} 0} & {\mspace{720mu}\mspace{720mu} 0} \\
\end{matrix}} \\
\begin{matrix}
0 \\
X_{O}^{\prime} \\
\end{matrix} & \begin{matrix}
0 \\
Y_{O}^{\prime} \\
\end{matrix} & \begin{matrix}
\begin{matrix}
{\mspace{720mu}\mspace{720mu} 1} \\
{\mspace{720mu}\mspace{720mu} Z}_{O}^{\prime} \\
\end{matrix} & \begin{matrix}
{\mspace{720mu} 0} \\
1 \\
\end{matrix} \\
\end{matrix} \\
\end{bmatrix}$$
The rotation matrices for rotating a vector about the x-axis by an angle $\alpha$, about y-axis by an angle $\beta$ and about the z-axis by an angle $\gamma$ are given by Equations (9a)--(9c), respectively. $$R_{({\alpha,\mspace{720mu}\mspace{720mu} x^{\prime}})}\mspace{720mu} = \mspace{720mu}\begin{bmatrix}
1 & 0 & {\mspace{720mu}\mspace{720mu}\mspace{720mu}\begin{matrix}
0 & {\mspace{720mu}\mspace{720mu}\mspace{720mu}\mspace{720mu}\mspace{720mu}\mspace{720mu}\mspace{720mu}\mspace{720mu} 0} \\
\end{matrix}} \\
0 & {\cos\alpha} & {\mspace{720mu}\begin{matrix}
{- \sin\alpha} & {\mspace{720mu}\mspace{720mu} 0} \\
\end{matrix}} \\
\begin{matrix}
0 \\
0 \\
\end{matrix} & \begin{matrix}
{\sin\alpha} \\
0 \\
\end{matrix} & \begin{matrix}
\begin{matrix}
{\mspace{720mu}\mspace{720mu}\cos\alpha} \\
0 \\
\end{matrix} & {\mspace{720mu}\begin{matrix}
0 \\
1 \\
\end{matrix}} \\
\end{matrix} \\
\end{bmatrix}$$ $$R_{({\alpha,\mspace{720mu}\mspace{720mu} y^{\prime}})}\mspace{720mu} = \mspace{720mu}\begin{bmatrix}
{\cos\beta} & 0 & \begin{matrix}
{\sin\beta} & 0 \\
\end{matrix} \\
0 & 1 & {\mspace{720mu}\begin{matrix}
{\mspace{720mu}\mspace{720mu} 0} & {\mspace{720mu}\mspace{720mu}\mspace{720mu}\mspace{720mu} 0} \\
\end{matrix}} \\
\begin{matrix}
{- \sin\beta} \\
0 \\
\end{matrix} & \begin{matrix}
0 \\
0 \\
\end{matrix} & \begin{matrix}
\begin{matrix}
{\cos\beta} \\
0 \\
\end{matrix} & \begin{matrix}
0 \\
1 \\
\end{matrix} \\
\end{matrix} \\
\end{bmatrix}$$ $$R_{({\alpha,\mspace{720mu}\mspace{720mu} z^{\prime}})}\mspace{720mu} = \mspace{720mu}\begin{bmatrix}
{\cos\gamma} & {- \sin\gamma} & \begin{matrix}
0 & 0 \\
\end{matrix} \\
{\sin\gamma} & {\cos\gamma} & \begin{matrix}
0 & 0 \\
\end{matrix} \\
\begin{matrix}
0 \\
0 \\
\end{matrix} & \begin{matrix}
0 \\
0 \\
\end{matrix} & \begin{matrix}
\begin{matrix}
1 \\
0 \\
\end{matrix} & {\mspace{720mu}\begin{matrix}
0 \\
1 \\
\end{matrix}} \\
\end{matrix} \\
\end{bmatrix}$$
Once the transformation of a rotation angle into a rotation matrix is successfully completed, the transformation matrix in Equation ([10](#FD10-sensors-20-02144){ref-type="disp-formula"}) is obtained by combining the above-described rotation matrices with the translation matrix as follows:$$\left\lbrack M \right\rbrack = \mspace{720mu}\left\lbrack T \right\rbrack \cdot \left\lbrack R_{({\alpha,{\mspace{720mu} x}^{\prime}})} \right\rbrack \cdot \left\lbrack R_{({\beta,{\mspace{720mu} y}^{\prime}})} \right\rbrack \cdot \left\lbrack R_{({\gamma,{\mspace{720mu} z}^{\prime}})} \right\rbrack$$ where, $\left\lbrack M \right\rbrack$ is 4 × 4 transformation matrix; $\left\lbrack R_{({\alpha,{\mspace{720mu} x}^{\prime}})} \right\rbrack$ is a rotation matrix for rotating of a vector by an angle $\alpha$ with respect to the *x*′ axis; $\left\lbrack R_{({\beta,{\mspace{720mu} y}^{\prime}})} \right\rbrack$ is Rotation matrix for rotation of a vector by angle $\beta$ with respect to *y*′ axis; and $\left\lbrack R_{({\gamma,{\mspace{720mu} z}^{\prime}})} \right\rbrack$ is Rotation matrix for rotation of a vector by angle $\gamma$ with respect to *z*′ axis.
In the three-dimensional world, four coordinates are necessary when considering the perspective of a scene. In projective space, two parallel lines appear to meet at the horizon, which is not the case in Euclidean space. Therefore, mathematicians use homogeneous coordinates which can represent the N-dimensional coordinates using N + 1 numbers \[[@B48-sensors-20-02144]\]. Equation ([11](#FD11-sensors-20-02144){ref-type="disp-formula"}) shows, how Cartesian coordinates can be described in terms of homogeneous coordinates. Consequently, the transformation matrix \[*M*\] is defined as a 4 × 4 matrix by redefining the rotation and translation matrices in terms of adding an additional dimension to the coordinates. $$\left( {X,Y,Z} \right) \equiv \left( {x,y,z,w} \right)\mspace{720mu}\mspace{720mu}\mspace{720mu}\mspace{720mu}\forall\mspace{720mu}\mspace{720mu}\mspace{720mu}\mspace{720mu}\left\{ \begin{array}{l}
{X = {x/w}} \\
{Y = {y/w}} \\
{Z = {z/w}} \\
\end{array} \right.$$ where, $\left( X,Y,Z \right)$ are the Cartesian coordinates and $\left( x,y,z,w \right)$ are the homogeneous coordinates.
Now, the coordinates of a point cloud acquired in a different loading are multiplied by the transformation matrix described in Equation ([10](#FD10-sensors-20-02144){ref-type="disp-formula"}) to obtain the new coordinates so as to identify and quantify the actual shape change of a specimen. This can be described with the following Equation ([12](#FD12-sensors-20-02144){ref-type="disp-formula"}):$$\left( {x^{\prime},y^{\prime},z^{\prime},1} \right) = \mspace{720mu}\left\lbrack M \right\rbrack \ast \left( {x,y,z,1} \right),$$ where, $\left( {x^{\prime},y^{\prime},z^{\prime},1} \right)$ is the point coordinates w.r.t. the new basis; $\mspace{720mu}\mspace{720mu}\left( {x,y,z,1} \right)$ is the point coordinates w.r.t. the old basis; $\left\lbrack M \right\rbrack$ is the transformation matrix.
3.3. Estimation of the Deflection Curve from the Loading Scenario {#sec3dot3-sensors-20-02144}
-----------------------------------------------------------------
After obtaining the coordinate transformation, it is possible to consider changes in the shape of an object formed by the point cloud that are equivalent to the actual object. In the last phase of the proposed algorithm, we fit the curved surface as a 2nd degree curve on the longitudinal axis and a linear curve on the transverse axis for each individual loading scenario using the transformed point clouds. The GA fitness function appears more complicated than the previous one since the error metric formulation is now considered as a two degree curve caused by loading. This can be described by the following Equation ([13](#FD13-sensors-20-02144){ref-type="disp-formula"}):$$Fitness\mspace{720mu} value = \mspace{720mu}\sum\limits_{i = 0}^{N}{d_{i}}^{2} = \mspace{720mu}\sum\limits_{i = 0}^{N}\left( {P_{i} - F\left( {A,B,C,D,E} \right)} \right)^{2}$$ where $d_{i}^{2}$ is the squared error which depicts the difference between the actual data and the model; $P_{i}$ is the measured data of the *z*-coordinate of a point cloud; $F\left( {A,B,C,D,E} \right) = \left\lbrack {A{x_{i}}^{2} + Bx_{i} + Cy_{i} + Dx_{i}y_{i} + E} \right\rbrack$; $x_{i}$ is the measured *x*-coordinate of a point cloud; $y_{i}$ is the measured *y*-coordinate of a point cloud; *A*, *B*, *C*, *D* & *E* are the coefficient parameter of the model; and *N* is the number of points in the point cloud data.
Like the unloading scenario, the resulting point cloud from loading is fitted as a curved surface according to the parameters obtained from genetic optimization. As discussed earlier, the method of finding suitable parameters for the optimization function, i.e., Taguchi's Design, is also utilized here. Because we assumed that lateral torsional buckling of a specimen is negligible due to the stiffeners, the longitudinal profile deflected shape of the specimen is our main focus. As a result, taking the longitudinal line which lies on the curved surface as a deflection curve makes it easy to estimate the deflection of the required position. Even though we set the local structural coordinates according to the intersection between the upper flange and the front side of the web, the center line of the deflected shape of a curved surface (flange) accurately describes the actual effect of the loading on the specimen. According to our experimental setup, which is a fixed-fixed support of beam, it is known that there is zero deflection at the end point of the specimen. Finally, the results obtained from the actual LVDT measurements and the analysis of what is proposed are compared and contrasted.
4. Experimental Study {#sec4-sensors-20-02144}
=====================
The proposed method has been validated through experiment carried out at Sungkyunkwan university. The experiment involved a steel box girder section specimen under different loadings. The experiment took place at the Concrete Material Lab (Suwon), where the University is located. [Figure 6](#sensors-20-02144-f006){ref-type="fig"} shows how all the entities are synchronized during the experiment. As shown in [Figure 6](#sensors-20-02144-f006){ref-type="fig"}, the main entities that play essential roles during the experiment are:Universal Testing Machine (UTM): This device is related mainly to the loading. The load is applied perpendicularly midway from the top flange face with the help of a hydraulic system. The loading system is controlled in real time via the UTM, i.e., the hydraulic power unit, load measuring unit and control devices, which are linked with the loading unit. ACE-USS200 model of Servo-Hydraulic Universal Testing Machines, which having 200 ton loading capacity, was utilized for this study. Servo-Hydraulic UTM can be controlled via a multi-functional remote control handset that is located on the frame, a digital control unit or Material Testing Program (MTP) software was installed on the PC connected to the Control Unit. It can carry out tensile and yield, compression, flexure tests with load and displacement controls.Linear variable displacement transducer (LVDT): LVDT is a sensor that converts the linear movement of the object the LVDT is coupled to into a variable corresponding to the electrical signal proportional to that movement. This contact sensor measures the real time displacement of a specimen by attaching the rod element, which is a combination of the core, core extension, and probe tip, lightly to the bottom flange face during unloading. CDP-50 type of LVDT was utilized for our experiment.Terrestrial laser scanning(TLS): It measures a scanned object by emitting laser pulses and recording the subsequent intensity of their return after reflection. Leica scan station C5 scanner which is operated based on the time of flight principle was utilized for our experimentation. We have used the highest resolution mode of resolution. According to Leica specifications, this kind of mode has 0.02 m × 0.02 m resolution. Furthermore, there are 2530 × 2181 points in the horizontal and vertical directions, respectively.Specimen: The beam utilized for this study is a steel box girder, SS 400-6T, which has dimensions of 0.4 m × 0.8 m × 2 m. the specimen has a transverse stiffener, which stiffens the flange and web against out of plane deformation. The specimen is welded every 45 cm throughout the entire span in both the right and left webs inside the box. As a result, transverse deformation of the specimen is trivial in this study.
4.1. Design of the Experiment {#sec4dot1-sensors-20-02144}
-----------------------------
The indoor experiment is setup as follows:Setting the position of a specimen through the UTM machine keeps all the necessary alignments both horizontally and vertically. The scanner device stands 2.5 m away from the front web face while considering which factors affect the accuracy of the data. Target color, incidence angle, range and intensity are the main factors that affect the point cloud noise and parametric model fitting \[[@B13-sensors-20-02144]\]. Consequently, this setup considers all the results from Bolkas et al. in accordance with the available space in the laboratory.The contact sensor (LVDT) is mounted below the specimen at three different positions. One is at the center and the other two LVDTs are fixed 55 cm from the left and right edges individually, as shown in [Figure 7](#sensors-20-02144-f007){ref-type="fig"}. These LVDTs and the UTM machine are connected to a computer.The specimen is scanned without any loading by setting the required field of view. The field of view for scanning an object should reduce the outliers caused by objects out side of the target.Once we are done scanning the specimen scene without loading, we apply loading via the UTM until the center LVDT reading reaches 1mm. The 1 mm sag is attained at a 57.33 KN loading. Again, the scanning process starts over by pausing the applied load and keeping the 1 mm sag.In this fashion, specimen scanning is carried out for different deflection sizes for the corresponding loadings. [Table 2](#sensors-20-02144-t002){ref-type="table"} summarizes the induced load for each case along with their corresponding LVDT sensor readings.After capturing all the necessary data with the USB, which was plugged in to the scanner during the scanning process, we changed the file format from .PLY to .PTS using cyclone which is a software module of Leica, for the purpose of using the cloud compare software. Once we have the data file format which is capable to utilize via with cloud compare, it is easy to apply manual segmentation of an object entity, removing the outliers and preparing the data for further analysis. one of the advantage of this software regarding with removal of an outliers, it provides segmentation command in different shape using polylines. [Figure 2](#sensors-20-02144-f002){ref-type="fig"}a,b, which depict the data with and with out the outliers respectively, are obtained from this software.
4.2. Validation Results {#sec4dot2-sensors-20-02144}
-----------------------
As described in previous sections, the specimen is scanned exactly after excitation, which leads to changes in shape followed by scanning without any loading. Once all specimen shape information is obtained for each load case, removal of the outliers and segmentation are carried out. The point cloud obtained from the experiments was processed using the proposed algorithm. [Figure 2](#sensors-20-02144-f002){ref-type="fig"}b depicts the resulting cloud data after preprocessing.
### 4.2.1. Selection of Optimal Parameters for GA {#sec4dot2dot1-sensors-20-02144}
As shown in [Table 1](#sensors-20-02144-t001){ref-type="table"}, eight control parameters at four values and one with two are identified in this study. Due to the number of parameters and their values considered in here, $L32^{\prime}\left( 2^{1}x4^{8} \right)$ orthogonal array is suitable for the experiment design. Orthogonal Arrays (OAs) provide a set of well balanced and minimal experiments. This array assumes that there is no interaction between any two factors. The experiments are repeated five times to increase the consistency of the experiment response. The analysis for the proposed Taguchi design is analyzed using a statistical software MINITAB19. Signal-to-noise ratios (SNR), which are log functions of the desired output as described in Equation ([5](#FD5-sensors-20-02144){ref-type="disp-formula"}), serve as objective functions for optimization, as well as help in data analysis and the prediction of optimum results. In the Taguchi method, the word "optimization" means determining the best values for the control factors. In turn, the best values for the control factors are those that maximize the signal-to-noise ratios. This can also be described as the resulting best values for the control factors so that the fitness functions are negligible, since the fitness functions are directly proportional to the error metric during plane fitting. Based on Taguchi two-step method rules, performance of a design is checked by maximizing the SN ratio and adjusting the mean to the target values. The Higher values of the Signal-to-noise ratio identify control factor settings that minimize the effects of the noise factor. In this study, the fitness value which is a response for the design experiment has a smaller the better characteristic. [Figure 8](#sensors-20-02144-f008){ref-type="fig"} depicts the graphical representation of a robust level of parameter for the Genetic Algorithm. As a result, the optimal combination of control factors for the genetic algorithm to minimize the error is depicted in [Table 3](#sensors-20-02144-t003){ref-type="table"}.
### 4.2.2. Computation of Deflection Based on the Genetic Algorithm {#sec4dot2dot2-sensors-20-02144}
Because of the nature of a lidar system, which is a collection of points, it is difficult to point out the edge of a specimen that would be considered as the "x-axis" of the local structural coordinate system. Therefore, representing the point cloud data by definable mathematical elements, planes, and lying one axis of a coordinate on a line which is obtained by correlating these planes are the easiest and optimal approach. A stochastic optimization, the genetic algorithm in this case, is utlized in this study to find the optimal planes that are most compatible with the data acquired. Even though this kind of optimization is prominently used for nonlinear forms and forms for which no derivative information exists \[[@B39-sensors-20-02144]\], it also plays an enormous role in finding the optimal solution even in heavily noisy spaces \[[@B40-sensors-20-02144]\]. Accordingly, the fitness function for the genetic algorithm is directly derived from the sum of the squares of the residuals, as shown in Equation ([3](#FD3-sensors-20-02144){ref-type="disp-formula"}).
Using the optimal control parameters shown in [Table 3](#sensors-20-02144-t003){ref-type="table"}, the individual plane entity was fitted using the fitness function for the GA from the least square error (Equation ([3](#FD3-sensors-20-02144){ref-type="disp-formula"})), as shown in [Figure 5](#sensors-20-02144-f005){ref-type="fig"}. The coefficients of the plane equation, i.e., the best individuals in genetic algorithm language, are obtained through a built-in MATLAB code, the Global Optimization toolbox, after determining the optimal parameter. By utilizing the Taguchi experimental design and obtaining the optimal parameter, the fitness values started to converge in the early generations (around the 3rd) and stopped at the 68th generation by satisfying the 1st stopping criteria (i.e., the average change in fitness value is less than the function tolerance). [Figure 9](#sensors-20-02144-f009){ref-type="fig"} depicts the mean and best fitness values with respect to each generation and the best individual values for the flange entity plane.These best individuals are 0.0083, −0.0257 and −0.5946. Therefore, the resulting planes which represent the acquired data for each entity of the specimen looks as follows: $G\left( {I,J} \right) = \mspace{720mu} - 0.5946 + 0.0083 \cdot I - 0.0257 \cdot J$... Equation of a plane that represents the flange part;$F\left( {C,D} \right) = \mspace{720mu} - 7.491 - 2.0996 \cdot C + 0.0123 \cdot D$ ... Equation of a plane that represents the web part.
After obtaining the plane equation that represents the web part in a similar way to the flange part, the 3D intersection line $L_{i}$, which is defined as the longitudinal axis of the local beam structural coordinates, is developed by considering Equation ([7](#FD7-sensors-20-02144){ref-type="disp-formula"}) parametrically. This gives us: $$\begin{matrix}
{x = 1.0003 \cdot t;} \\
{y = - 7.493 - 2.0987 \cdot t;} \\
{z = - 0.4022 + 0.0621 \cdot t.} \\
\end{matrix}$$
Now, it is possible to position at least one axis of the new coordinate system, the x-axis in our case, along the intersection line described in the above figure. From [Figure 5](#sensors-20-02144-f005){ref-type="fig"}, the difference between points $O^{\prime}$ and Q gives us the vector which lies on the new x-axis, called *X*${}^{\prime}$. Before performing the rotation, we should determine the geometric correlation between the vector $O^{\prime}$Q and the old basis (*X*,*Y*,*Z*) after translating of point *O* to point $O^{\prime}$. [Figure 10](#sensors-20-02144-f010){ref-type="fig"} shows the resulting point cloud data after performing the required transformation. This information is helpful in quantifying the deflection of a structure by applying the transformation matrix to the acquired point cloud data during the loading scenario.
Similar to the loading scenario, the fitness function for the loading case is obtained using Equation ([13](#FD13-sensors-20-02144){ref-type="disp-formula"}). A 2nd degree of polynomial curve is constructed for the fitness function considering the complexity degree of the structure based on given experimental study. This leads us to compute the best individuals similar as before, but here, we have five individual parameters for the curved surface, as illustrated in [Figure 11](#sensors-20-02144-f011){ref-type="fig"}. This provides the deflection shape of the top flange specimen as depicted in [Figure 12](#sensors-20-02144-f012){ref-type="fig"}. Even though our x-axis lies on the intersection between the flange and web as shown in [Figure 5](#sensors-20-02144-f005){ref-type="fig"}, the center line in [Figure 12](#sensors-20-02144-f012){ref-type="fig"} indicates a larger deflection shape of the structure induced by the point load. Therefore, squeezing the plane towards the center transversely from each side gives us the required hypothetical deflection curve, which enables us to calculate the deflection of a specimen throughout its entire length. Here, only the deflection curve, which appears 4 mm at the center LVDT sensor, is shown diagrammatically. The rest of the deflections constructed using the proposed model along with their corresponding load contact sensor measurements are depicted in [Figure 13](#sensors-20-02144-f013){ref-type="fig"}.
The resulting equation of center line ℄ from [Figure 12](#sensors-20-02144-f012){ref-type="fig"} was calibrated by considering its 2nd degree curve behaviour. Some boundary conditions appear because of the support condition. As shown in [Figure 6](#sensors-20-02144-f006){ref-type="fig"}a, the support conditions used in this experiment lead us to determine the displacement and the moments at the support (i.e., $\delta_{({x = 0,x = L})} = 0\mspace{720mu}\mspace{720mu}\&\mspace{720mu}\mspace{720mu} M_{({x = 0,x = L})} = 0$), such that the moment at the midpoint is a maximum.This calibration is needed because of the inherent random error of the instrument and the natural difference between the theoretical and experimental solutions. In the following, we illustrate how a signal equation can be calibrated by employing the boundary equation for the 4 mm loading scenario. The deflection curve obtained from the proposed analysis is as follows: $Z = - 2.138 \times 10^{- 3} - 7.092 \times 10^{- 3} \cdot X + 3.925 \times 10^{- 3} \cdot X^{2}$
According to basic beam theory, the max deflection for a simply supported beam must attained at the middle of the span. This implies that *dZ*/*dx* = 0,$dZ/dx = 7.85 \times 10^{- 3} \cdot x - 7.092 \times 10^{- 3\mspace{720mu}} = 0$
$x = 0.90344$$\ldots\ldots$ while, *x* must be 1. This tells us the resulting signal deviates by $\left( 1 - 0.90344 \right) = 0.09656$ from the real value. Therefore, the signal needs to be adjusted by +0.09656 along the x-direction and −0.002138 along the z-direction. $\left. dZ/dx = 7.85 \times 10^{- 3} \cdot - K = 0\Leftrightarrow x = 1\mspace{720mu}\mspace{720mu}\mspace{720mu}\therefore\mspace{720mu}\mspace{720mu}\mspace{720mu} K = 7.85 \times 10^{- 3} \right.$$dZ/dx = 7.85 \times 10^{- 3} \cdot x - 7.85 \times 10^{- 3}$$\int{dZ/dx} = \mspace{720mu}\int\left( 7.85 \times 10^{- 3} \cdot x - \mspace{720mu} 7.85 \times 10^{- 3} \right)$$Z = 3.925 \times 10^{- 3} \cdot X^{2} - 7.85 \times 10^{- 3} \cdot X$
Thus, the deflection curve after calibration of the measured data is obtained for the loading scenario *z* = 4 mm.
Finally, the deflection of a loaded beam was estimated from the deflection curve and validated via the corresponding contact sensor (LVDT). The difference between the LVDT and proposed methodology results are illustrated in [Figure 13](#sensors-20-02144-f013){ref-type="fig"}.The model for the center LVDT, which is at 1 m, almost coincides with the actual sensor measurements, while the models for the edge LVDTs are not. This is happened because;
- Since the specimen has transverse stiffener at 45 cm apart along both sides, it is very stiff along with the loading. However, the stiffness of a specimen is not uniform throughout the span with increasing load. A part of the specimen in between two stiffeners has not been equally disturbed with the part where exactly stiffener is welded. Consequently, the global deflection curve may not be expected exactly as a 2nd degree parabolic.
- Secondly, the nature of a specimen has also its part in affecting the result for the edge LVDTs. Hence, the beam is labelled as a deep beam because of its span-to-depth ratio and the concentrated load with it. Therefore, the shear effect is predominant than flexural in our specimen. This implies that the deflection curve which is expected from the flexural effect is affected to some extent. Even the data shown for edge LVDTs are biased by this effect.
As a result, the proposed approach of representing the deflection shape in terms of a parabolic curve may only accurately capture the deflection formed throughout the whole span length except for around the mid-span. For emphasis, in the case of edge LVDTs in [Figure 13](#sensors-20-02144-f013){ref-type="fig"}a,c, the resulting deflection from the LVDT in each load scenario is almost the same as with the nominal one. However, this situation is incompatible with the proposed model since we are dealing with a deflection curve considered as a 2nd degree polynomial curve. It is obvious that if we consider a deflection curve as explained above while keeping the max deflection (at the center) equal to the nominal deflection, the deflection at the edge LVDT is less than the nominal one. Consequently, the proposed methodology is validated only for the middle LVDT sensor. As shown in [Figure 13](#sensors-20-02144-f013){ref-type="fig"}d, the proposed algorithm is accurate and effective for nominal deflections of 2 and 3 mm since the error is between $\pm 1$. For the case of nominal deflections of 1 and 4 mm, the error increases to $\pm 4$ and $\pm 2.6$, respectively.
5. Conclusions {#sec5-sensors-20-02144}
==============
In this work, we proposed a method for estimating the deflection of a beam based on a non-contact sensor, TLS. We simultaneously applied regression curve fitting with a genetic algorithm. The accustomed method, least squares regression fitting, is used to determine the fitness function which measures the suitability of the solution generated by the GA. The GA minimizes the fitness function obtained from the error equation. In addition, transformation of the coordinates is a crucial step in this study. This is because in order to define any geometrical changes in a structure with respect to its own axes, the scan data obtained from the device must also visualize the target structure with all its parts and elements so that periodic maintenance is predetermined. The equation for the curve is extracted from the surface fitting after transforming the coordinates of the data. This equation represents the deflection curve and, therefore, allows us to estimate the deflection value of a beam at any position across its span. The deflection value at various points was validated using the corresponding direct contact sensor, LVDT.
The proposed method is practical in applying the concept of structural health monitoring in different structures using non-contact sensors. Besides computation of a deflection, the geometric shape of a structure at the time of scanning can be visualized throughout this proposed methodology. This leads to regulating a structure health condition all over the area. As shown in the result, this algorithm is suitable for long-spanned beam. However, it is also effective in determining the deflection around the mid-span.
M.B.M. developed the concept for monitoring structural deflection, performed the analysis, designed the methodology and drafted the manuscript. D.L. and G.C. configured the hardware and software setup of the experiment and provided a guidance. The equipment utilized in the presented work was provided by S.P. All authors have read and agreed to the published version of the manuscript.
This work is financially supported by the Korea Ministry of land Infrastructure and Transport (MOLIT) through the "Smart City Master and Doctor Course Grant Program" and a National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIP) (No. NRF-2017R1A2B3007607, No.NRF-2018R1D1A1B07047218).
The authors declare no conflict of interest.
The following abbreviations are used in this manuscript: LIDARLight Detection And RangingTLSTerrestrial Laser ScanningGAGenetic Algorithm3DThree-DimesionLVDTLinear Variable Differential TransformerLS-GALeast Square Based Genetic AlgorithmDOEDesign of ExperiemntRGBRed Green BlueSNRSignal to Noise RatioUTMUniversal Testing MachineOAOrthogonal Arrays
{#sensors-20-02144-f001}
{#sensors-20-02144-f002}
{#sensors-20-02144-f003}
{#sensors-20-02144-f004}
{#sensors-20-02144-f005}
{#sensors-20-02144-f006}
{#sensors-20-02144-f007}
{#sensors-20-02144-f008}
{#sensors-20-02144-f009}
{#sensors-20-02144-f010}
{#sensors-20-02144-f011}
{#sensors-20-02144-f012}
{#sensors-20-02144-f013}
sensors-20-02144-t001_Table 1
######
Genetic Algorithm Parameters and their levels.
Factors Levels
--------- -------------------------- -------------------- -------------- ------------------- ----------------------
A Migration Direction Forward Both \- \-
B Population Size 50 100 150 200
C Fitness Scaling Function Rank Proportional Top Shift Linear
D Selection Function Stochastic Uniform Remainder Roulette Tournament
E Elite Count 2 5 10 20
F Crossover Function Scattered Two Point Heristic Arthmetic
G Crossover Fraction 0.8 0.6 0.4 0.2
H Mutation Function Gaussian Uniform Adaptive Feasible Constraint Dependent
I Hybrid Function None FminSearch Patternsearch Fminunc
sensors-20-02144-t002_Table 2
######
Load cases and LVDT readings for the corresponding nominal deflection.
Loading Nominal Load LVDT Reading (mm)
--------- --------- -------- ------------------- ------- -------
Case 1 1mm 57.33 1.204 1.005 1.055
Case 2 2mm 200.47 2.086 2.014 1.99
Case 3 3mm 380.85 3.046 3.022 2.949
Case 4 4mm 480.84 4.109 4.029 4.188
sensors-20-02144-t003_Table 3
######
Optimal combination parameters from the response table.
A B C D E F G H I
--------------- ---------- ---- ---- ---- ---- ---- ---- ---- ---- ----
Rank SN Ratio 9 4 6 5 3 2 7 8 1
Mean 9 4 5 6 8 2 7 3 1
Optimum Level SN Ratio 2 4 3 1 2 4 2 4 4
Mean 1 1 4 3 4 3 2 2 1
Combination A2 B4 C3 D1 E2 F4 G2 H4 I4
[^1]: Current address: Sungkyunkwan University, Seobu-ro 2066, Korea.
| {
"pile_set_name": "PubMed Central"
} |
WAN files are available from the <https://figshare.com/s/5297ddc238766def6afc> database. Wireless data are within the paper and its Supporting Information files.
Introduction {#sec001}
============
Vehicular cloud computing is a promising paradigm that aims at merging mobile cloud computing and vehicular networking, in order to give arise to integrated communication-computing platforms \[[@pone.0191577.ref001]\]. In vehicular cloud computing, vehicles can be either the service providers to enrich existing cloud services by providing various on-road information (e.g., traffic condition like Pics-on-Wheels proposed by \[[@pone.0191577.ref002]\]) or be the service consumers to enjoy existing centralized Internet cloud services \[[@pone.0191577.ref003]--[@pone.0191577.ref005]\]. One of the key features of vehicular cloud computing is high mobility \[[@pone.0191577.ref005]--[@pone.0191577.ref009]\]. The running job may be interrupted by the random arrival and departure of vehicles \[[@pone.0191577.ref003],[@pone.0191577.ref008]\]. We must be able to address the mobility and provide an effective control scheme to guide service conditions and cloud resources \[[@pone.0191577.ref010]--[@pone.0191577.ref011]\]. Thus fault-tolerant schemes are designed to provide reliable and continuous services in vehicular cloud computing despite the unavailable of some vehicles \[[@pone.0191577.ref012]--[@pone.0191577.ref017]\]. As an essential building block of the fault-tolerant scheme, a failure detector (FD) plays a critical role in the engineering of such dependable systems \[[@pone.0191577.ref018]--[@pone.0191577.ref019]\]. An optimized FD should find the vehicles' failure in a timely and accurate manner \[[@pone.0191577.ref020]\].
To improve connectivity of vehicular cloud computing, the Roadside Unit (RSU) is deployed along the road \[[@pone.0191577.ref021]--[@pone.0191577.ref023]\]. For example, many future Internet applications \[[@pone.0191577.ref024]\] which are delay and delay-jitter sensitive (such as Netflix and VTube) are benefit from RSU. Normally, they use the solar as power input due to the unavailability or excess expense of wired electrical power \[[@pone.0191577.ref025]--[@pone.0191577.ref028]\]. The solar power are easily affected by the natural environment, e.g., solar power cannot be acquired at night or cloudy day. The energy supply of RSU is unstable, so the energy capacity of RSU is limited. According to the U.S. Department of Transportation \[[@pone.0191577.ref029]\], it is estimated that 40% of all initial rural freeway roadside infrastructure would have to be solar powered by 2050. A breakdown of the deployment costs also found that over 63% of these roadside infrastructure costs would be consumed by solar energy provisioning, e.g., solar panels, batteries, and their associated electronics. Thus, it is important to reduce the energy consumption of RSUs \[[@pone.0191577.ref028],[@pone.0191577.ref030]\].
Existing main adaptive FDs (such as Chen FD \[[@pone.0191577.ref031]\], *φ*-FD \[[@pone.0191577.ref032]\], ED FD \[[@pone.0191577.ref033]\] etc.) keep a sliding window (*WS*) that contains information about received messages to make an estimate of the state (trusted or suspected of having failed) of a monitored node. These FDs need a certain memory space to save a large history message window. At each detection cycle, a large amount of calculation is needed to compute the probability distribution parameters and detector parameters \[[@pone.0191577.ref034]\]. For most RSUs with battery, these overhead of FDs can exacerbate the battery consumption.
In this paper, aiming at the RSUs with battery, we have presented the Energy-Efficient Failure Detector (2E-FD). It does not rely on the probability distribution of message transmission delay, or on the maintenance of history message windows. We use the arrival time of last message to estimate the arrival time of next message. In addition, the dynamic safety margin, which is computed by the single exponential smoothing method, is used to improve the accuracy of failure detection. To evaluate the performance of 2E-FD, some best-known existing FDs are selected in terms of detection time, mistake rate and query accuracy probability. Besides, we also measure the battery consumption of RSU with different FDs. The experimental results show that 2E-FD is able to reduce the battery consumption of RSU and provide an adaptive failure detection service with high accuracy.
The rest of this paper is organized as follows. In section 2, the related work of vehicular cloud computing and failure detectors is introduced. Section 3 introduces the system model and presents the implementation of 2E-FD. Section 4 carries out the experiments on real traces and tests the battery consumption of RSU with different FDs. Finally, the work is concluded in section 5.
Related work {#sec002}
============
In this section, vehicular cloud computing is firstly introduced. Second, the quality of service (QoS) metrics of FD is introduced. Finally, several existing main adaptive FDs are presented.
Vehicular cloud computing {#sec003}
-------------------------
Olariu et al. \[[@pone.0191577.ref035]--[@pone.0191577.ref037]\] advocate the concept of vehicular cloud, which coordinates the computing, sensing, communication and storage resources to provide services to authorized users. Different from conventional Internet cloud with dedicatedly installed hardware, vehicular cloud leverages the already available resources on vehicles. In vehicular cloud computing, vehicles communicate with the data centers where the wanted cloud services locate using vehicles-to-infrastructure (V2I) communications, e.g., LTE, WiMax. In addition, vehicles can also work in an autonomous way solely relying on the vehicle-to-vehicle (V2V) communication capabilities. Accordingly, we think that vehicular cloud can be described by a loose two-tier architecture. The first-tier is the Internet cloud computing platform (e.g., data centers) while the second-tier consists of many vehicular cloudlet \[[@pone.0191577.ref003]\]. The vehicular cloudlet is made up of RSU and vehicles. A user can acquire cloud services from either the first-tier data center or the second-tier vehicular cloudlet. An architecture example is illustrated in [Fig 1](#pone.0191577.g001){ref-type="fig"}.
{#pone.0191577.g001}
QoS metrics of failure detector {#sec004}
-------------------------------
Many distributed applications have some timing constraint on the behaviors of FDs \[[@pone.0191577.ref019],[@pone.0191577.ref038]\]. It is not acceptable that a node is suspected hours after than it has crashed or the FD outputs several false positives. To solve this problem, Chen \[[@pone.0191577.ref031]\] proposed a series of metrics to specify the QoS of FD: how fast it detects actual failures and how well it avoids false detections. These metrics can quantitatively represent the detection speed and accuracy. We use *T* or *S* to represent whether a node is trusted or suspected. *T*-transition means that the output of the detector changes from *S* to *T*, while *S*-transition means that the output of the detector changes from *T* to *S*. The following three primary metrics are used to describe the QoS of a FD.
Detection time (*T*~*D*~) is the time that elapses from the moment when a node crashes to the time when it starts being suspected, i.e., when the final *S*-transition occurs.
Mistake rate (*λ*~*M*~) is the number of mistakes that a FD makes per unit time, i.e., it represents how frequently a FD makes mistakes.
Query accuracy probability (*P*~*A*~) is the probability that the output of a FD is correct at a random time.
The first metric is related to a failure detector's speed, while the remaining relate to its accuracy. In many cases, the mistake rate is not sufficient to describe the accuracy of a FD; simultaneously, the query accuracy probability is also needed. For example, [Fig 2](#pone.0191577.g002){ref-type="fig"} shows that both FD~1~ and FD~2~ are detecting the node *p*. The two FDs have the same mistake rate (0.125) but different query accuracy probabilities (0.75 and 0.5).
{#pone.0191577.g002}
Adaptive failure detector {#sec005}
-------------------------
The adaptive FDs are designed to adapt to changing network conditions and application requirements \[[@pone.0191577.ref039]\]. In most adaptive FDs, their implementations are based on a heartbeat strategy. Existing main adaptive FDs (Chen FD, *φ* FD and ED FD) work as follows:
Chen et al. \[[@pone.0191577.ref031]\] proposed the QoS-based adaptive failure detection algorithm based on a probability network model. This algorithm assumes that node *p* sends heartbeat message *m* to node *q* periodically. The recent *n* heartbeat messages *m*~1~, *m*~2~, ⋯, *m*~*n*~ stored in a sliding window at *q*. *A*~1~, *A*~2~, ⋯, *A*~*n*~ are their receipt times according to *q*'s local clock. Then, the expected arrival time of next heartbeat message is estimated by: $$EA_{(k + 1)} = \frac{1}{n}{\sum\limits_{i = k - n - 1}^{k}{(A_{i} - \eta*i) + (k + 1)\eta}}$$ where *η* is the sending interval, decided by the QoS requirement of user. The freshpoint *τ*~*k*+1~ of next heartbeat message is consist of *EA*~(*k*+1)~ and constant safety margin *α*. One has $$\tau_{k + 1} = EA_{(k + 1)} + \alpha$$
This FD estimates the arrival time of next heartbeat message based on a constant safety margin.
In *φ*-FD, the output is a continuous value *φ* to represent the suspicion level of the monitored node, rather than a binary value simply representing true or suspect. It keeps a sliding window to save the most recent arrival time of heartbeat messages, and assumes that the heartbeat inter-arrival time follows a normal distribution. Then the value of *φ* can be calculated as follows: $$\varphi(T_{now}) = - log_{10}(P_{later}(T_{now} - T_{last}))$$ where the *T*~*last*~ is the time when the most recent heartbeat message is received, *T*~*now*~ is the current time, and *P*~*later*~(*t*) is the probability of a heartbeat message will arrive more than t time units later than the previous one. According to the assumption of heartbeat inter-arrival time, *P*~*later*~(*t*) is given by the following equation: $$P_{later}(t) = \frac{1}{\sigma\sqrt{2\pi}}{\int_{t}^{\infty}e^{- \frac{{(x - \mu)}^{2}}{2\sigma^{2}}}}dx = 1 - F(t)$$ where *F*(*t*) is the cumulative distribution function of a normal distribution with mean *μ* and variance *σ*^2^. When the applications query the *φ*-FD at time *T*~*now*~, *φ* FD will return a value of *φ* to them. Then each application compares the value *φ* with its threshold Φ, which is given by different QoS requirements. If *φ* \> Φ, a certain action is trigged. Thus, *φ*-FD can meet different QoS requirements of multiple applications simultaneously.
ED FD is similar to the *φ*-FD. The difference is in the assumption of distribution of heartbeat inter-arrival time. In ED FD, it assumes that the heartbeat inter-arrival time follows an exponential distribution. Then, the suspicion level is given by a value, called *e*~*d*~, which is calculated as follows: $$e_{d} = F(T_{now} - T_{last})$$ $$F(t) = 1 - e^{- \frac{t}{\mu}}$$ where *T*~*now*~, *T*~*last*~ and *μ* have the same meaning as for the *φ* FD. For ED FD, the threshold is called *E*~*d*~.
For a FD, it will belong to the class *◇P* if it satisfies the following properties \[[@pone.0191577.ref040]\]:
Strong completeness: eventually every node that crashes is permanently suspected by every correct node.
Eventually strong accuracy: there is a time after which a correct node is no longer suspected by any correct node.
Implementation {#sec006}
==============
System model {#sec007}
------------
We consider a partially synchronous system consisting of a finite set of nodes ∏ = {*p*~1~, *p*~2~, ⋯, *p*~*n*~}. Each node behaves correctly until it crashes and is unable to recover. Any two nodes can be connected by an unreliable communication channel. Because most FDs are implemented using the UDP protocol, we assume that the communication channel between nodes is a fair-lossy channel \[[@pone.0191577.ref041]\], i.e., no message can be copied or modified and no new message can be created, and if a node *p* continues sending a message *m* to *q*, *q* will eventually receive *m*.
We assume the existence of some global time (unbeknownst to nodes), denoted as global stabilized time (GST), and that nodes always make progress; furthermore, at least *δ* \> 0 time units elapse between consecutive steps (the purpose of the latter is to exclude the case where nodes require an infinite number of steps in finite time).
To simplify the description, consider a system that consists of only two nodes *p* and *q*, where *q* is monitoring *p*. Node *p* sends a message to *q* every *η* time (sending interval) or is subject to crashing. Node *q* suspects node *p* if it does not receive any heartbeat message from *p* for a period of time determined by the freshpoint.
The implementation of 2E-FD {#sec008}
---------------------------
To improve the battery consumption of RSU with failure detector, we propose an energy-efficient failure detector 2E-FD. When a heartbeat message is sent to the receiver, the message delay *d*~*i*~ can be calculated by $$d_{i} = T_{now} - (i - 1)*\eta$$ where *η* is the sending interval, *T*~*now*~ is the arrival time of new heartbeat message. We assume that the message delay of next heartbeat message *d*~*i*+1~ is equal to *d*~*i*~, so the expected arrival time of next heartbeat message is $$EA_{i + 1} = j*\eta + d_{i}$$
According to single exponential smoothing method \[[@pone.0191577.ref042]\], for the each *d*~*i*~, the new predictive delay ${\hat{d}}_{i + 1}$ is computed from the formula: $${\hat{d}}_{i + 1} = \alpha*{\hat{d}}_{i} + (1 - \alpha)*d_{i}$$ where *α* (0 ≤ *α* ≤ 1) is a constant between 0 and 1, which controls how rapidly the ${\hat{d}}_{i + 1}$ adapts to the delay change. So the safety margin (*SM*) can be estimated by: $$SM_{i + 1} = \beta\left| {{\hat{d}}_{i + 1} - d_{i}} \right|$$ where *β* is a variable, chose such that there is an acceptably small probability that the delay for the heartbeat message will exceed timeout. At last, the freshpoint of heartbeat message (*i* + 1) can be computed by $$\tau_{i + 1} = EA_{i + 1} + SM_{i + 1}$$
2E-FD is unable to get the communication delay from the sender to the receiver when it is lost. To ensure the effectiveness of the proposed approach, we fill the gaps with a value computed by $d_{k} = (\frac{k - l}{2}) \cdot \eta + d_{l}$, where *d*~*k*~ is the estimation of message delay of unreceived heartbeat message *m*~*k*~ from *p*, *d*~*l*~ is the message delay of receiving last heartbeat message *m*~*l*~ from *p*.
2E-FD employs the heartbeat approach as the basic failure detection strategy. To simply the description, suppose the system consists of only two nodes *p* and *q*, where *q* is monitoring *p*. The detection algorithm is shown in [Fig 3](#pone.0191577.g003){ref-type="fig"}.
{#pone.0191577.g003}
For the node *p*, it sends heartbeat message to node *q* at interval *i* \* *η* (*i* ≥ 0). For the node *q*, two tasks will be executed. One task is to suspect the node *p* when *q* didn't receive any heartbeat message during the last freshpoint *τ* of *q*'s clock. Another task is to compute the freshpoint *τ* according to the arrival time of new heartbeat message. Every time after receiving the new arrival heartbeat message, node *q* needs to compute the delay from sending to receiving and safety margin of next heartbeat message.
Correctness proof {#sec009}
-----------------
From the theory point-of-view, the 2E-FD can satisfy the strong completeness and eventually strong accuracy. Therefore, our FD belongs to class *◇P*, which is sufficient to solve the consensus problem. The 2E-FD implements a FD of class *◇P* under the condition of the system model defined in section 3. The evidence is as follows.
According to the system model, there is an upper bound on node speeds and on message transmission times after GST. We assume that Δ~*msg*~ is the upper bound and node *p* will not send any heartbeat message after it crashes. The heartbeat message *m*~*k*−1~ is the last heartbeat message from *p* before *p* crashes. So node *q* can calculate the freshpoint *τ*~*k*~ of next heartbeat message based on [Eq (11)](#pone.0191577.e013){ref-type="disp-formula"}.
{#pone.0191577.e015g}
τ
k
=
E
A
k
\+
S
M
k
Because *EA*~*k*~ = (*k* − 1) \* *η* + *d*~*k*−1~ and *d*~*k*−1~ \< Δ~*msg*~, one gets $$EA_{k} < (k - 1)*\eta + \Delta_{msg}$$
For the safety margin $SM_{k} = \beta\left| {{\hat{d}}_{k} - d_{k - 1}} \right|$, we have $${\hat{d}}_{k + 1} = \alpha \cdot {\hat{d}}_{k} + (1 - \alpha) \cdot d_{k}$$
By recursion property of equation, one gets $$\begin{array}{ll}
{\hat{d}}_{k + 1} & {= \alpha(\alpha{\hat{d}}_{k - 1} + (1 - \alpha)d_{k - 1}) + (1 - \alpha)d_{k}} \\
& {= \alpha^{2}{\hat{d}}_{k - 1} + \alpha(1 - \alpha)d_{k - 1} + (1 - \alpha)d_{k}} \\
& \cdots \\
& {= \alpha^{k + 1}{\hat{d}}_{0} + \alpha^{k}(1 - \alpha)d_{0} + \alpha^{k - 1}(1 - \alpha)d_{1} + \ldots + (1 - \alpha)d_{k}} \\
\end{array}$$
Because $d_{0} = {\hat{d}}_{0} = 0$, $$\begin{array}{ll}
{\hat{d}}_{k + 1} & {= \alpha^{k - 1}(1 - \alpha)d_{1} + \alpha^{k - 2}(1 - \alpha)d_{2} + \ldots + (1 - \alpha)d_{k}} \\
& {= (1 - \alpha){\sum\limits_{i = 1}^{k}{\alpha^{k - i}d_{i}}}} \\
\end{array}$$
Therefore, using the same methods, we get $${\hat{d}}_{k} = (1 - \alpha){\sum\limits_{i = 1}^{k - 1}{\alpha^{k - i}d_{{}_{i}}}}$$
Because 0 ≤ *d*~*i*~ \< Δ~*msg*~, therefore $$0 \leq {\hat{d}}_{k} < \alpha(1 - \alpha^{k - 1})\Delta_{msg}$$
From inequalities [(18)](#pone.0191577.e023){ref-type="disp-formula"}, we get $$SM_{k} < \beta\left| {\alpha(1 - \alpha^{k - 1})\Delta_{msg}} \right| = \beta \cdot \alpha(1 - \alpha^{k - 1})\Delta_{msg}$$
Thus, the freshpoint *τ*~*k*~ of next heartbeat message is $$\tau_{k} = EA_{k} + SM_{k} < (k - 1) \cdot \eta + \Delta_{msg} + \beta \cdot \alpha(1 - \alpha)\Delta_{msg}$$
From inequation [(20)](#pone.0191577.e025){ref-type="disp-formula"}, all components of *τ*~*k*~ are bounded, so we can deduce *τ*~*k*~ is bounded. If for each message *m*~*k*−1~ received from node *p*, node *q* activates a bounded timeout, then there is a time after which *q* suspects *p*, if it receives no new message from *p*. Thus the 2E-FD satisfies the strong completeness property.
Every time *q* time out and *p* is correct, then *β* is increased. There is a time *t*~*bound*~ where safety margin *SM* is larger than Δ~*msg*~. The heartbeat message from node *p* must be received by node *q*. As a result, node *q* can avoid false detection, and *SM* stops increasing. Thus, when *SM* becomes higher than Δ~*msg*~, a correct node is no longer suspected by any correct node. This indicates the 2E-FD satisfies the eventually strong accuracy property.
Evaluation and performance {#sec010}
==========================
In this section, we will evaluate and analyze the performance of 2E-FD through comparative experiments. First, the 2E-FD was compared with the failure detection algorithms in terms of detection time, mistake rate and query accuracy probability. Then, we tested the battery consumption of RSU with different failure detection algorithms in a simulation environment.
2E-FD performance evaluation and comparison {#sec011}
-------------------------------------------
To enhance the authenticity of the experiment, we referred to the method in paper \[[@pone.0191577.ref043]\] that used the same trace files to replay different schemes of FDs and calculated the QoS metrics. This method ensures that all schemes of FDs are compared in the same network condition. The trace files are obtained from two network conditions (Wireless and WAN). For the accrual FDs (*φ*-FD and ED FD), it is necessary to compute the parameters of FD based on multiple history messages. In these experiments, *φ*-FD and ED FD shared the same fixed window size (1000), while the Chen FD used two different sliding windows (*WS* = 1 and *WS* = 1000). Furthermore, we did not analyze the sampled data until the sliding window was full, since the behavior of the FDs is stable only after that moment. The parameters of FDs are configured as follows: the basic parameters of 2E-FD are set as *α* = 0.85, and in order to find the best QoS and compare with the others, here *β* ∈ \[10^−6^, 10^6^\]. For *φ*-FD, the parameters are set the same as in \[[@pone.0191577.ref032]\]: Φ ∈ \[0.5, 16\]. For Chen FD, the parameters are set the same as in \[[@pone.0191577.ref031]\]: *α* ∈ \[0, 10000\]. For ED FD, the parameters are set the same as in \[[@pone.0191577.ref033]\]: *E*~*d*~ ∈ \[10^−4^, 10\].
For the Wireless case, two nodes are selected to represent the detecting node and monitored node respectively. The two nodes communicated through a WiFi (802.11g) network. The detecting node was equipped with a 900MHz ARM Cortex A7 processor, 1G RAM and Cent OS 6.5 operating system. While the monitored node was equipped with a 2GHz Intel Xeon processor, 1G RAM and Cent OS 6.5 operating system. The heartbeats were sent with a target of one heartbeat every 100.5ms (standard deviation: 7.87ms; min.: 0.002ms; max.: 948.96ms). During the experiment, the round-trip time (RTT) was measured to be at a low rate. The average RTT was 1.83ms, with a minimum of 1.175ms, and a maximum of 21.953ms. More than 1 million heartbeats were sent.
[Fig 4](#pone.0191577.g004){ref-type="fig"} shows the results of the mistake *λ*~*m*~ vs. detection time *T*~*D*~ in the Wireless scenario. The x-coordinate represents the detection time, and y-coordinate represents the mistake rate. [Fig 5](#pone.0191577.g005){ref-type="fig"} shows the results of query accuracy probability *P*~*A*~ vs. detection time *T*~*D*~ in the same scenario.
{#pone.0191577.g004}
{#pone.0191577.g005}
From the Figures, we found that all the FDs follow the same general tendency. However, our FD outperforms the others in the Wireless scenario. This improvement is because most late heartbeats were caught by the freshpoint under the same network conditions. In [Fig 4](#pone.0191577.g004){ref-type="fig"}, when *T*~*D*~ \< 0.2*s*, the mistake rate of 2E-FD has obvious decrease than other FDs. This is because that our FD can quickly adapt to the network conditions than other FDs. From [Fig 5](#pone.0191577.g005){ref-type="fig"}, our FD has higher query accuracy probability than others. When 0.15*s* \< *T*~*D*~ \< 0.4*s*, our FD and ED-FD seem to have similar query accuracy probability. At the same time, it is clear that the performance of Chen FD with *WS* = 1 is better than Chen FD with *WS* = 1000 in terms of mistake rate and query accuracy probability. This may be because that many burst data and too old data affect the calculation of freshpoint when sliding window size increases.
For the WAN case, the experiment was carried out on two nodes respectively located in Japan and Switzerland \[[@pone.0191577.ref032]\]. In the experiment, all heartbeat messages were transmitted using UDP protocol, and the monitored node *p* sent heartbeat messages to node *q* at the interval of 100ms, and *q* recorded all the arriving time of the received messages into the trace file according to its local clock. The experiment lasted a week, and more than 5 million heartbeat messages has been collected, where the average RTT is 283.3ms and message loss rate is approximately 0.4%.
Figs [6](#pone.0191577.g006){ref-type="fig"} and [7](#pone.0191577.g007){ref-type="fig"} show the results of the mistake rate *λ*~*m*~ and query accuracy probability *P*~*A*~ vs. detection time *T*~*D*~ in the WAN scenario. Similar to the result in the Wireless scenario, the mistake rate and query accuracy probability of all of the FDs have an identical tendency with increasing detection time. In [Fig 6](#pone.0191577.g006){ref-type="fig"}, the 2E-FD presents the lowest mistake rate (an improvement of up to 50%). It shows that our predictive method can improve the detection accuracy effectively. In [Fig 7](#pone.0191577.g007){ref-type="fig"}, our FD has an obvious improvement compared with ED FD when *T*~*D*~ \< 0.35*s*. Furthermore, when the mistake rate or query accuracy probability is the same, our FD performs shorter detection time. In the WAN, 2E-FD behaves better than the others FDs in terms of quick speed and high accuracy.
{#pone.0191577.g006}
{#pone.0191577.g007}
Comparison of battery consumption {#sec012}
---------------------------------
RSUs are static devices that disseminate the data stored in the infrastructure to the passing-by vehicles periodically \[[@pone.0191577.ref028]\]. In addition, RSU may exploit Infrastructure-to-Vehicular single-hop IEEE 802.11-like wireless links for data dissemination \[[@pone.0191577.ref011]\]. Based on these, we simulated the working environment of RSU by organizing several nodes. These nodes were connected by wireless link. One node was selected to represent the RSU, and the other nodes represented the vehicles. The RSU, which was equipped with 500mAh battery, was responsible for disseminating data to vehicles \[[@pone.0191577.ref025],[@pone.0191577.ref028]\]. The vehicles may be failure via the fault injection method. At each experiment, we deployed different FD to the RSU and measured the running time of RSU.
For the accrual FDs (*φ*-FD and ED FD) and Chen FD, they need to calculate the detector parameters and maintain the history window of detection messages at every detection period. We have selected different sliding window size setting (from *WS* = 100 to *WS* = 10000) and have made a detailed comparison of battery consumption of RSU. To make the experiment general, we re-did the experiments 5 times with the same environment, and the same parameters for each failure detection algorithms. Finally, we recorded the running time of RSU with different failure detection algorithms (as shown in [Fig 8](#pone.0191577.g008){ref-type="fig"}).
{#pone.0191577.g008}
From the [Fig 8](#pone.0191577.g008){ref-type="fig"}, it shows that the running time of RSU without any FD is the longest and improved by 6% compared to the RSU with 2E-FD. For the RSU with FD, the running time of RSU with *φ*-FD is the shortest, and it quickly decreases as the sliding window size changes. This is because that the battery consumption of RSU is exacerbated when the parameters of normal distribution model are calculated. The calculation of parameters needs the statistical data from the entire window. While the running time of RSU with 2E-FD is the longest (an improvement of up to 12% than *φ*-FD when the sliding window size is 10000), and it is not affected by sliding window size. For the RSU with Chen FD (*WS* = 1), its running time is not affect by sliding window size too. However, its running time is shorter than the RSU with 2E-FD. This might be because that the accuracy of 2E-FD is higher than Chen FD (*WS* = 1) due to the dynamic safety margin. 2E-FD avoids sending more data to the failure nodes so as to save the battery consumption of RSU. The RSU in the experiment shown in [Fig 8](#pone.0191577.g008){ref-type="fig"} at most connects 10 nodes. In vehicular cloud computing, in order to maintain the connectivity of systems, each RSU is required to connect many vehicles. Therefore, the fact that 2E-FD can reduce battery consumption of RSU is more significant in real systems.
Conclusion {#sec013}
==========
Failure detection plays a very important role in vehicular cloud computing. In this paper, we introduced the energy-efficient failure detector (2E-FD). It has been proven that 2E-FD is a failure detector of class *◇P*. By using the prediction of last message and the dynamic safety margin, the 2E-FD can quickly adapt to the network conditions and provide acceptable QoS failure detection service. Moreover, 2E-FD is able to save the battery consumption of RSU because it does not need to compute the distribution parameters and maintain the sliding window. Through the comparative experiments, the results showed that 2E-FD demonstrates a better performance in terms of detection speed, accuracy and battery consumption. Therefore, 2E-FD is suitable to be layout in the vehicular cloud computing for providing failure detection service.
Supporting information {#sec014}
======================
###### Wireless.
(ZIP)
######
Click here for additional data file.
This work was supported by the National Natural Science Foundation of China (No. 61100029 and 61370085).
[^1]: **Competing Interests:**The authors have declared that no competing interests exist.
| {
"pile_set_name": "PubMed Central"
} |
###### Article summary
Strengths and limitations of this study
=======================================
- This study protocol is based on well-established methodological standards.
- While great efforts were made to ensure inclusion of the most relevant predictors of panic-like anxiety, it is possible that some unidentified but relevant variables were missed.
Introduction {#s1}
============
Chest pain in Emergency Medicine {#s1a}
--------------------------------
Chest pain accounts for approximately 5% of all emergency department (ED) consultations,[@R1] and over 50% of cases remain unexplained at discharge.[@R2; @R3; @R4; @R5; @R6] In Canada, approximately 400 000 patients/year present in the ED with unexplained chest pain (UCP).[@R2] [@R4]
Burden of UCP {#s1b}
-------------
Despite a generally favourable prognosis, 80% of cases of UCP persist for up to 12 years after initial medical evaluation.[@R7; @R8; @R9; @R10; @R11; @R12; @R13] Many patients with UCP (41--60%) report limitations in daily functioning (eg, housework, walking and exercising) and work absenteeism or disability (17--35%).[@R7] [@R8] [@R10; @R11; @R12] [@R14; @R15; @R16; @R17; @R18; @R19; @R20; @R21; @R22] Moreover, the occupational impairments associated with UCP are comparable or more severe than those associated with cardiac chest pain.[@R7] [@R22] The negative impact of UCP on quality of life and day-to-day functioning is considerable and may be observed for up to 10 years after symptom onset.[@R7]^--^[@R9] [@R11] [@R12] [@R14; @R15; @R16; @R17; @R18; @R19; @R20; @R21; @R22]
Despite the benign origin of their pain, patients with UCP report persistent fear of serious health conditions.[@R11] [@R20] [@R22] [@R23] They are frequent users of healthcare services, including emergency care, and often undergo multiple invasive tests (eg, coronary angiograms).[@R9] [@R14] [@R22] [@R24]^--^[@R27] In Canada, the average duration of an ED consultation for a UCP patient is 11 h and one in four patients arrives in the ED by ambulance.[@R3] The direct annual cost associated with UCP in the USA is estimated to amount to eight billion US dollars.[@R28] [@R29]
UCP is also associated with significant psychological distress that can become chronic in the absence of targeted interventions.[@R30; @R31; @R32] In fact, 20--40% of patients present a psychiatric disorder at the time of ED consultation [@R3] [@R6] [@R33; @R34; @R35] and 15% report suicidal ideation.[@R3] [@R33] [@R34] [@R36] Unfortunately, fewer than 5% of patients are referred to a mental health professional for psychiatric or psychological treatment.[@R3]
While the cause of UCP may be unclear, the literature clearly demonstrates that UCP is highly prevalent and often chronic, and that it constitutes a significant burden for patients and society alike.
Aetiology of UCP {#s1c}
----------------
Although pathologies such as microvascular angina and gastroesophageal reflux may be at the origin of some cases of UCP,[@R29] panic attacks are the most prevalent condition associated with UCP in ED.[@R3] [@R34] [@R35] [@R37; @R38; @R39] As many as 44% of patients with UCP experience panic attacks in the month prior to ED consultation.[@R3] [@R34] [@R35] [@R37; @R38; @R39]
A panic attack is defined as a discrete period of intense fear or discomfort that peaks in a few minutes.[@R40] [@R41] Fear or discomfort is accompanied by at least four of the following symptoms: chest pain, palpitations, dyspnoea, a feeling of suffocation, hot or cold flashes, sweating, nausea, feeling faint, paraesthesia, trembling, fear of death, depersonalisation and fear of losing control or going crazy.[@R40] [@R41] Panic attacks may be an isolated phenomenon or may occur in the context of a psychiatric disorder; the most common psychiatric disorder in which panic attacks occur is panic disorder.[@R42] The 1-year prevalence of panic attacks in the general adult population is 8--11%;[@R42] [@R43] the prevalence is four to six times higher among patients presenting with UCP.[@R3] [@R42] [@R43]
The literature clearly demonstrates that panic attacks with and without panic disorder constitute a significant mental health problem with serious consequences.[@R3] [@R36] [@R42; @R43; @R44; @R45; @R46; @R47] For simplicity, the term *panic-like anxiety* (PLA) will be used to refer to panic attacks with or without panic disorder.
Consequences of PLA in patients with UCP {#s1d}
----------------------------------------
PLA may be responsible for a significant portion of the negative consequences of UCP.[@R3] [@R6] [@R15] [@R30; @R31; @R32] [@R34] [@R35] [@R37] PLA is associated with a greater frequency of UCP episodes and increased risk of chronicity.[@R15] [@R30; @R31; @R32] Quality of life is lower and functional limitations levels are higher in patients with UCP and PLA.[@R15] [@R30] [@R31] [@R37] Moreover, in patients with UCP, PLA is associated with at least a threefold increase in psychiatric morbidity and suicidal ideation.[@R3] [@R34] Similarly, use of medical resources nearly doubles when PLA is present.[@R31] [@R32]
In patients with UCP, PLA is associated with elevated morbidity, excessive health services use and a negative prognosis. Unfortunately, more than 92% of cases of PLA remain undiagnosed at the time of discharge from ED.[@R3] [@R34] [@R35]
Identifying PLA in patients with UCP {#s1e}
------------------------------------
Several factors may contribute to the current low rate of PLA identification in patients in ED. First, PLA patients and physicians alike tend to focus on physical symptoms and on potential organic causes.[@R48] Second, the identification of PLA is complicated by the similarity between PLA symptoms and symptoms of medical conditions such as coronary artery disease. Third, the limited time available for clinical evaluation in ED settings may be insufficient to identify psychological causes of symptoms.[@R49] Finally, some ED physicians are unfamiliar with PLA or believe that it is not their role to identify psychiatric problems.[@R50] However, other physicians recognise the importance of improving identification and treatment of PLA in the ED settings.[@R51; @R52; @R53]
Researchers and clinicians seeking methods for increasing PLA identification rates must take into consideration certain constraints related to the clinical practice of emergency medicine, notably the brief period of time available to assess patients.
Importance of screening for PLA in ED patients with UCP {#s1f}
-------------------------------------------------------
Increasing the rate of identification of a problem is not in itself sufficient to improve clinical outcomes for patients.[@R54] [@R55] Gates [@R54] and Stiell and Wells [@R55] propose five criteria for determining the importance of a detection procedure and its potential impact on patients' clinical outcomes: (1) the problem has an impact on public health; (2) the problem is sufficiently prevalent; (3) effective treatments are available to reduce morbidity; (4) early diagnosis improves patient prognosis and (5) additional investigations or treatments are acceptable to patients.
The current data demonstrate that more accurate identification of PLA in ED patients with UCP could improve clinical outcomes. First, PLA in patients with UCP is a prevalent health problem with serious consequences for patients and society. Second, research demonstrates that morbidity associated with PLA in patients suffering from UCP can be greatly reduced via evidence-based treatments.[@R56; @R57; @R58] For example, 80--95% of patients with PLA show significant improvement and attain an adequate level of functioning following cognitive-behavioural therapy.[@R57] [@R59] [@R60] Several evidence-based treatment methods for PLA have proven to be effective in patients with UCP.[@R38] [@R61; @R62; @R63] Third, given that PLA tends to worsen over time,[@R32] [@R42] [@R64; @R65; @R66; @R67] negatively influencing treatment response, early diagnosis improves prognosis.[@R25] [@R42] [@R64] [@R66] [@R68; @R69; @R70; @R71] Finally, the criterion of acceptability to patients appears to have been met. Participation rates for patients with PLA and UCP approached for inclusion in a study are generally over 70%.[@R38] [@R61; @R62; @R63] In addition, 80% of primary care patients with PLA agreed to receive psychiatric care.[@R72]
The development of interventions designed to improve identification of PLA associated with UCP in ED appears to be indicated. A central factor for such an intervention is the availability of a suitable screening instrument, that is, an instrument that is efficient and acceptable to emergency physicians.[@R54] [@R55] The use of decision aids such as screening instruments is recognised as an effective method for improving clinical decision-making.[@R55]
Panic Screening Score {#s1g}
---------------------
To our knowledge, our team has developed the only two screening instruments for PLA in ED patients with UCP.[@R73] [@R74] One instrument, the *Panic Screening Score* (PSS; [figure 1](#BMJOPEN2013003877F1){ref-type="fig"}), was designed for use with patients with UCP and identifies panic attacks with and without panic disorder.[@R73] In addition, the PSS has the advantage of being brief (four items) and easy to use. We have shown that the PSS is eight times more sensitive in detecting PLA associated with UCP than is clinical evaluation by an emergency physician.[@R3] [@R73] In addition, the PSS offers a good combination of sensitivity and specificity ([table 1](#BMJOPEN2013003877TB1){ref-type="table"}) and these properties have been shown to be stable in a retrospective validation and preliminary prospective validation.[@R73] [@R75] As of now, the PSS is the briefest and most effective screening instrument for PLA associated with UCP in emergency settings. Although the PSS has good specificity, its sensitivity needs to be improved ([table 1](#BMJOPEN2013003877TB1){ref-type="table"}).
######
Predictive validity of the PSS[@R73]
Derivation (n=201) Retrospective validation (n=305)
--------------------------- ---------------------------- ----------------------------------
Sensitivity 63% (95% CI 52 to 73%) 53% (95% CI 44 to 62%)
Specificity 84% (95% CI 76 to 90%) 85% (95% CI 78 to 89%)
Positive predictive value 74% (95% CI 62 to 83%) 72% (95% CI 62 to 80%)
Negative predictive value 76% (95% CI 68 to 89%) 71% (95% CI 65 to 77%)
Positive likelihood ratio 3.89 (95% CI 2.5 to 6.05) 3.45 (95% CI 2.35 to 5.04)
Negative likelihood ratio 0.44 (95% CI 0.33 to 0.59) 0.55 (95% CI 0.46 to 0.67)
PSS, Panic Screening Score.
{#BMJOPEN2013003877F1}
Summary {#s1h}
-------
Early identification of PLA in ED patients with UCP appears to be the strategy of choice for reducing morbidity, chronicity and overuse of healthcare services. The PSS is a concise and effective instrument that represents the most promising method for achieving this objective. The present study will represent a major step towards the clinical application of the PSS and the early diagnosis and treatment of PLA in ED patients with UCP.
Methods and analysis {#s2}
====================
Objectives {#s2a}
----------
The objectives of this prospective cohort study are to (1) refine the PSS; (2) validate the revised version of PSS in an independent sample; (3) estimate the reliability of the revised PSS and (4) assess the acceptability of the instrument among ED physicians.
Methodological framework {#s2b}
------------------------
The research methods and statistical analyses used in this study are based on clinical decision rule standards[@R55]: The outcome must be clearly defined and assessed blindly;The predictors must be clearly defined, standardised and evaluated without knowledge of patient status;The reliability of the variables studied must be demonstrated;Participants must be selected without bias and must represent a large range of clinical and demographic characteristics. Ideally, the study should be conducted in several centres in order to increase external validity;Appropriate statistical analyses must be used;The sample size must be sufficient to permit valid statistical analyses;The sensibility of the instrument must be adequate. It must have a clear objective, good content validity and clinical relevance, and must be easy to use in the context of targeted practice;The capacity of the instrument to identify patients with (sensitivity) and without (specificity) the condition must be demonstrated;Measures must be taken to guarantee the appropriate application of the instrument.
The procedures for refining the PSS (objective 1) are based on Stiell and Wells'^[@R55]\ recommendations.^ First, the PLA predictors that were valuable but not essential in previous studies will be reassessed.[@R73] Next, the potential predictors of PLA that were not assessed at the time of the initial study will be evaluated.
Participants {#s2c}
------------
This study will consecutively recruit 3000 patients at the two emergency services of the Centre de santé et de services sociaux Alphonse-Desjardins (University-Affiliated Hospital of Lévis and Paul-Gilbert Hospital). To be eligible, patients will have to present UCP as defined by (1) the absence of an identifiable cause (eg, pneumothorax, pneumonia); (2) absence of chest trauma; (3) absence of new malignant cardiac arrhythmia and (4) a score of 2 or less on the modified version of the *thrombolysis in myocardial infarction score*.[@R76] [@R77] This simple instrument stratifies patients presenting chest pain according to the probability of mortality or myocardial infarction in the 30 days following the ED visit.[@R76; @R77; @R78] A result ≤2 is associated with low incidence (3%) of mortality or cardiac events; scores \>2 are associated with a higher incidence (28%). Scores are obtained by summing values that correspond to the following characteristics: (1) age ≥65 years (1 point); (2) known coronary stenosis ≥50% or history of revascularisation (1 point); (3) deviation of the ST segment ≥0.5 mm (5 points) and (4) elevated rate of cardiac enzymes defined as troponins I ≥99th centile.[@R77] A value of zero is assigned for criteria 3 and 4 if the physician does not order the tests necessary to obtain the information.
Patients will be excluded if they (1) present a terminal illness; (2) present a severe communication problem that could interfere with the administration of the questionnaire; (3) present a psychotic state, major cognitive deficit or other condition that could invalidate the interview and (4) are legally incompetent or younger than 18 years of age.
Procedure {#s2d}
---------
With the assistance of a research nurse, the emergency physicians will assess the eligibility of all patients presenting with UCP. Physicians will complete the PSS for every eligible and consenting patient. To assess the reliability of the PSS items, a second physician will independently complete the PSS for at least 10% of patients. As in other similar studies and due to constraints such as the availability of a second physician, the reliability assessment will be conducted using a convenience sample.[@R79; @R80; @R81]
To assess the primary outcome and potential predictors of PLA, all participants will complete a telephone interview within 72 h following recruitment. Interviewers will be blind to the patient\'s PSS scores. Since the evaluation of the criterion standard will occur after the patient\'s discharge, physicians will be blind to the results of the PLA evaluations.
The research nurse will review the ED computerised database every day to ensure that all potentially eligible patients were assessed. A registered nurse will contact patients who were missed and request consent for the telephone interview. For non-consenting patients, only age, gender and time of ED visit will be recorded. This step will enable us to identify potential selection biases and ensure quality control.
The acceptability of the PSS to emergency physicians will be assessed at the end of the study. The time taken to administer the PSS will also be evaluated in 10% of cases randomly selected. The initiation of the questionnaire will be defined as the moment the physician asks the first question; the end of the questionnaire will be defined as the moment the patient finishes answering the final question. This measure will be used to assess whether or not the PSS is sufficiently brief for the ED.
Measures {#s2e}
--------
- *Eligibility evaluation form*: This form contains the inclusion and exclusion criteria. It also records the patient's contact information, age, gender and time of ED visit.
- *PSS and PSS---revised version*: The PSS score is calculated by summing the points assigned to the answer for each of the four questions. A score ≥6 is considered to be a positive result: the patient presents an elevated probability of PLA. A revised version of the PSS will be administered to patients during the validation phase (objective 2). The revised version will include only the items selected during the optimisation phase.
- *Anxiety disorders interview schedule for Diagnostic and Statistical Manual of Mental Disorders, fourth edition (ADIS-IV)*:[@R82] [@R83] The Panic disorder module of the French (or English) version of the interview will be administered by telephone. As recommended by experts in the field, ADIS-IV will serve as the criterion standard for the identification of PLA.[@R84] The Panic disorder module has shown excellent reliability for the identification of PLA (k≥0.80).[@R34] PLA is defined as either the presence of a panic attack during the previous month or the presence of panic disorder.[@R73]
- *Additional predictors for the refinement phase (objective 1):* Additional predictors include the four potential predictors of PLA identified but not included in the final version of the PSS,[@R73] as well as the items selected following a pilot study [@R75] and a review of the literature: Fear of dying associated with chest pain;[@R85] [@R86]Item number 4 (fear of fainting) of the Anxiety Sensitivity Index[@R87] [@R88];Item number 28 (feeling overwhelmed by one\'s problems) on the State-Trait Anxiety Inventory;[@R89] [@R90]The *Autonomic Nervous System Questionnaire*[@R91];The four panic disorder items from the Patient Health Questionnaire-15;[@R92] these items were selected because they demonstrated sensitivity and specificity between 75% and 96% for PLA in patients in primary care and psychosomatic settings[@R93];The modified version of the *Life Events Stress Scale*.[@R94] [@R95] This scale has an excellent internal consistency (0.87). The questionnaire includes 10 items, each of which corresponds to a category of stressful events. Patients are asked to specify whether or not each event has occurred and, if so, whether or not it occurred within the previous six months. For the purpose of this study, patients will be asked to specify whether or not the event occurred in the last month. The intensity of the stress associated with each event will be evaluated on a five-point Likert scale. This questionnaire was selected because the occurrence or aggravation of PLA is preceded by stressful life events in 80% of cases.[@R96]
- Modified version of the *Ottawa Acceptability of Decision Rules Instrument* (OADRI)[@R97]: This 12-item questionnaire assesses the acceptability of clinical decision rules by physicians. The questionnaire has good internal consistency (0.80--0.86) and good construct validity.[@R97] Ten items will be used in the present study as some information will not be available to ED physicians at the time of assessment. The first excluded item concerns the instrument\'s validation data; the second excluded item concerns the impact of the instrument on the use of clinical resources.
- The PSS administration time assessment record sheet: This document includes instructions for assessing the PSS administration time and recording the result.
Quality control {#s2f}
---------------
Emergency physicians will receive a 30 min training session on how to use the PSS. The session will include an overview of the study, inclusion and exclusion criteria and guidelines for the administration of the PSS and for scoring patients' answers. The training session will be developed jointly with the Centre de liaison sur l\'intervention et la prévention psychosociale. This non-profit organisation is specialised in knowledge transfer and dissemination of research results.
Over a 1-month period, a research nurse will periodically observe each physician as he or she administers the PSS, in order to obtain feedback and to identify problems. This step will be repeated during the implementation of the refined version of the PSS.
Telephone interviews will be conducted by graduate students in psychology. Each student will receive 1 day of training on the administration of ADIS-IV, followed by weekly supervision with a clinical psychologist. Telephone interviews will be recorded to facilitate supervision. A random sample of 25% of recorded interviews will be used to evaluate inter-rater agreement on the diagnosis of PLA. During the optimisation phase (objective 1), the recordings will also be used to evaluate inter-rater agreement on each of the additional interview items. These supervision and inter-rater agreement procedures have been proven effective in our previous studies and generated excellent diagnostic reliability for PLA.[@R3] [@R34]
Data analysis {#s2g}
-------------
Participants' sociodemographic data will be presented in descriptive form. To evaluate the representativeness of the sample, participant data will be compared with data of the eligible patients who declined to participate. Continuous variables that meet the assumptions of normality will be analysed with Student\'s t test. Otherwise, Wilcoxon Mann-Whitney\'s non-parametric test will be used. The χ^2^ test will be used for categorical variables. Inter-rater agreement for PLA diagnosis on ADIS-IV[@R82] [@R83] will be assessed with Cohen\'s κ coefficient.
As the predictive performance of a clinical rule is usually overestimated in the sample used in its development, it is important to evaluate the rule in an independent sample.[@R55] [@R98] In this study, a temporal validation procedure will be used. The refinement analyses (objective 1) will be conducted using data from the first 1500 patients; the validation of the refined version of the PSS (objective 2) will be performed using data from the subsequent 1500 patients.
Refinement of the PSS (objective 1) {#s2h}
-----------------------------------
The reliability of the four PSS items and the 15 potential predictors will be evaluated using Cohen\'s κ tests or weighted κ. Only the items with a good κ coefficient (k≥0.6) will be included in further analysis.[@R55] We will use the two types of analysis recommended by experts in the field, recursive partitioning and regression,[@R55] to refine the PSS with data collected from the first 1500 patients. Recursive partitioning generally results in a more sensitive instrument, whereas regression yields models with a higher global predictive value.[@R55] [@R99]
This study will use the recursive partitioning technique known as Classification and Regression Trees.[@R100] The construction of Classification and Regression Trees will be automated, but manual intervention will be used if some of the concurrent predictors are more useful than others (eg, more reliable, more representative or easier to use). We will dichotomise continuous variables by selecting the most effective cut-off point for identifying patients with PLA.
Log-binomial regression analysis will also be performed. This type of analysis is preferred to logistic regression because it provides exact relative risks rather than ORs. Furthermore, when the prevalence of the dependent variable is greater than 10%, the estimate of the relative risk by logistic regression is imprecise.[@R101] In this study, the expected prevalence of PLA is 44%.[@R3] Variables associated with PLA in univariate log-binomial regression (p≤0.15) will be considered in the multivariate analysis. The multivariate log-binomial regression will be performed using the ascending stepwise method. Multicollinearity between variables will be verified. If correlations ≥0.80 are obtained, the analysis will be repeated using only one of the intercorrelated items, in order to obtain the most effective model. The regression equation will be converted into a score by assigning points to each answer; point assignment will be based on the magnitude of corresponding regression coefficients according to the *Framingham study risk score function*.[@R102] The result is a simple score that provides probability estimates that correspond to the score generated by the multivariate regression model. The cut-off score that yields the best predictive validity will be selected based on the area underneath the receiver-operating characteristic (ROC) curve and the measures of predictive validity.
The calibration (Hosmer-Lemeshow test) and discriminating validity (area under the ROC curve, sensitivity, specificity, likelihood ratios, predictive values) of the two optimised versions of the PSS will be evaluated. CIs of 95% will be calculated for each of the discrimination measures. The version of the PSS that is simplest (ie, the version with the fewest items) and offers the best discrimination will be evaluated (objective 2).
Validation of the refined PSS (objective 2) {#s2i}
-------------------------------------------
The refined version of the PSS will be prospectively validated in a validation sample (n=1500). Sensitivity, specificity, predictive values and likelihood ratios will be calculated with 95% CIs.
Evaluation of the PSS reliability {#s2j}
---------------------------------
Cohen\'s κ coefficient will be used to assess the level of inter-rater agreement (reliability) for the refined PSS (presence or absence of PLA) and for each of its components. In the case of variables including three or more categories, a weighted κ coefficient will be calculated.[@R73] [@R103]
Acceptability of the PSS {#s2k}
------------------------
Descriptive data on the PSS administration time will be reported, including the mean, median and range. The total score on the OADRI and the level of endorsement for each item will also be reported in descriptive form.
Justification of the sample size and feasibility {#s2l}
------------------------------------------------
On the basis of our previous study, we estimate the minimum prevalence of PLA in these settings to be 40%.[@R3] The sample composed of the first 1500 patients will be used for analyses related to the refinement of the PSS. It will include approximately 600 patients with PLA. This number exceeds the minimum ratio of 10 cases for each variable in the regression analysis.[@R104] A subsample of 1500 patients will allow us to obtain a CI of 95%±3.9%, for a sensitivity equivalent to that reported in the original PSS study (63%).
The EDs at the Centre de santé et de services sociaux Alphonse-Desjardins receive approximately 110 000 patients each year, and approximately 2% of patients visiting the two EDs present UCP. Of 2200 eligible patients per year, we expect that 10% will be missed and 20% will decline to participate. Our final recruitment estimate is therefore 3000 participants in 24 months (1500/year).
Ethics and dissemination {#s3}
========================
The research ethics committee at the Centre de santé et de services sociaux Alphonse-Desjardins approved this protocol. The study will not affect usual care and the ethical considerations are minimal. Patients' verbal consent to complete the PSS and to be contacted by telephone will be solicited. Verbal consent will be obtained again at the time of the telephone interview. All data will be treated according to standard guidelines for ensuring patient confidentiality.
The results of the study will be presented in scientific conferences and published in peer-reviewed scientific journals. Further dissemination through workshops aimed at emergency physicians in clinical settings and a dedicated website is planned.
Discussion and conclusions {#s4}
==========================
This study is designed to validate an effective and efficient screening instrument for PLA in ED patients with UCP. The PSS will help emergency physicians determine the likelihood of PLA, in turn facilitating appropriate referrals to mental health professionals or family physicians for confirmation of the diagnosis and treatment. Treatment for PLA significantly reduces associated morbidity and excessive use of healthcare services, and has an overall favourable cost/benefit ratio.[@R26] [@R38] [@R61; @R62; @R63] [@R105; @R106; @R107] This study will result in a screening tool with the potential to have a tangible clinical impact for ED patients with UCP and PLA. Further research will focus on assessing the impact of use of the PSS and on validating the instrument in other settings, such as cardiology and primary care clinics.[@R25] [@R31]
Supplementary Material
======================
###### Author\'s manuscript
The authors wish to thank Stéphanie Hamel for her help in preparing this manuscript and Sarah Roberts for her help in editing this manuscript.
**Collaborators :** Julie Carrier.
**Contributors:** GF-B was responsible for the original idea, literature review and study design. ID helped draft the manuscript, assisted with the methodology and revised the manuscript for important intellectual content. PA and JP revised the manuscript for important intellectual content. RPF assisted with the methodology and revised the manuscript for important intellectual content. CED assisted with the statistical design and methodology and revised the manuscript critically for important intellectual content. All authors read and approved the final manuscript.
**Funding:** This work is supported by the Canadian Institutes of Health Research (CIHR), grant number: 126125.
**Competing interests:** None.
**Ethics approval:** Centre de santé et de services sociaux Alphonse-Desjardins.
**Provenance and peer review:** Not commissioned; peer reviewed for ethical and funding approval prior to submission.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
There were an estimated 65,500 new cases of ovarian cancer in 2012 in Europe with 42,700 deaths^[@CR1]^. It ranks fifth as the cause of cancer death in women and is the most deadly gynecological cancer due to late stage diagnosis^[@CR2]^. Ovarian cancer is not a single disease, but a group of tumors classified depending on the cells it involves. Thereby, there are ovarian epithelial tumors, sex cord-stromal tumors (implicating granulosa and theca cells) and germ cell tumors^[@CR3]^.
Advanced ovarian epithelial cancer patients undergo surgery, in order to reduce all macroscopic visible disease. Early and advanced stage epithelial cancers are treated with a combination therapy of platinum and taxane. Unfortunately, approximately 70% of the patients present a relapse during the first 3 years^[@CR4]^. The most common therapy regimen for sex cord- stromal ovarian tumors is the combination of bleomycin, etoposide and cisplatin (BEP)^[@CR5]^. Even though these tumors show a good response rate after BEP treatment, a high relapse rate is observed several months after the completion of the treatment^[@CR6]^.
Overexpression of Fibroblast Growth Factor 1 (FGF1) has been linked to high grade serous ovarian tumors and poor survival^[@CR7],[@CR8]^. Furthermore, FGF1 has been associated with tumor growth in nude mice injected with ovarian cells overexpressing FGF1^[@CR9]^. In ovarian epithelial cisplatin-resistant cell lines overexpressing FGF1, its knock-down by shRNA, restores sensitivity to cisplatin^[@CR8]^.
FGF1 belongs to the FGF family that counts 22 members^[@CR10],[@CR11]^. FGF1 regulates cell proliferation, differentiation and survival^[@CR12]--[@CR19]^. FGF1 acts through FGFR--dependent or FGFR--independent pathways^[@CR16],[@CR19]--[@CR21]^. Indeed, FGF1 is mainly intracellular under physiological conditions and secreted only under specific stress conditions^[@CR22]--[@CR24]^. Whereas FGF1 has been shown to interact with intracellular proteins such as CK2, FIBP, p34, nucleolin, and p53^[@CR18],[@CR21],[@CR25]--[@CR28]^, its intracellular activities are not fully understood. Nevertheless, FGF1 intracellular activities are crucial for cell survival since FGF1 represses the pro-apoptotic activity of p53. We previously showed in rat embryonic fibroblasts and pheochromocytoma PC12 cell line that FGF1 promotes p53 degradation and inhibits both p53 phosphorylation on serine 15 and p53 transcriptional activities^[@CR16],[@CR17]^. We also showed that FGF1 interacts with p53 in PC12 cells^[@CR18]^.
p53 is a key regulator of apoptosis^[@CR29]^. Its ability to induce apoptosis is mediated by the transactivation of pro-apoptotic genes such as *Bax*^[@CR30]^, *NOXA* and *PUMA*^[@CR31]^. p53 also triggers apoptosis by relocating at the mitochondrion. Indeed, p53 mitochondrial translocation allows its interaction with both the anti-apoptotic proteins BCL-X~L~ and Bcl-2, leading to their inhibition^[@CR32]^, and the pro-apoptotic proteins Bax and BAK, provoking their activation^[@CR33],[@CR34]^.
As FGF1 can inhibit p53-dependent apoptosis, we hypothesized that FGF1 could affect the apoptotic response to etoposide (an activator of p53-dependent apoptosis) in ovarian tumor cells. We tested our hypothesis in the ovarian granulosa cell line COV434 that expresses wild-type p53 protein^[@CR35]^. In the present study, we showed that FGF1 is able to attenuate etoposide and cisplatin-induced apoptosis in COV434 cells. Under etoposide treatment, FGF1 only shows a moderate impact on p53 stability or activation and p53 transcriptional activities do not appear to be involved in COV434 cells apoptosis. However, p53 mitochondrial activities are important for COV434 cells apoptosis and FGF1 regulates p53 mitochondrial translocation.
Results {#Sec2}
=======
FGF1 overexpression protects COV434 ovarian granulosa cells from etoposide-induced apoptosis {#Sec3}
--------------------------------------------------------------------------------------------
We first determined whether FGF1 overexpression could be sufficient to induce resistance to chemotherapeutic agents such as cisplatin or etoposide in ovarian cells. We thus established stable cell lines constitutively overexpressing FGF1 (COV434-FGF1). To induce p53-dependent cell death, we used the well-known genotoxic stress inducer etoposide. Among the hallmarks of apoptosis, we monitored (i) the decrease of the inner mitochondrial membrane potential (ΔΨm) that reflects mitochondrial depolarization during apoptosis, (ii) cytochrome c release arising from mitochondrial outer membrane permeabilization (MOMP) and (iii) caspases activation.
Cell condensation and loss of ΔΨm reflected by low DiOC~6~(3) staining were first examined by flow cytometry analysis. Following a 16 h-long etoposide treatment, the percentage of COV434-FGF1 cells with small size and low ΔΨm (apoptotic cells) was significantly lower than for parental or mock-transfected COV434 cells (Fig. [1a](#Fig1){ref-type="fig"}). Similar results were obtained using cisplatin instead of etoposide (Supplementary Fig. [S1A](#MOESM1){ref-type="media"}). Therefore, FGF1 overexpression partially inhibits etoposide- and cisplatin-induced apoptosis in COV434 cells.Fig. 1FGF1 overexpression protects COV434 cells from etoposide-induced apoptosis.**a** Upper panel: Average flow cytometry quantification of apoptotic cells characterized by their low DIOC staining and cell condensation (DIOC^-^, Size^-^) ± SEM for 3 experiments done in triplicate. Non-transfected COV434 (NT), two COV434-Mock clonal cell lines and three COV434-FGF1 clonal cell lines were treated with etoposide (25 µg/mL) for 16 h, or not treated (Ctl), The t-tests compare to NT Eto. Lower panel: FGF1 levels in non-transfected, mock and FGF1 overexpressing clones using western blot analysis. Total proteins are visualized with the Biorad stain free system. **b** Immunofluorescence study for cytochrome c release. COV434-Mock C1 and -FGF1 C1 cells were treated with 25 µg/mL etoposide for 4 h. Cells were stained with an anti-cytochrome c antibody (green) and TO-PRO-3 (blue) to visualize nuclei (left panels). Scale bar represents 40 µm. The histogram presents the average percentages ± SEM for 3 independent experiments of cells exhibiting cytochrome c release (right panel). The t-test compares to Mock cells similarly treated. **c** Upper panel: Western blot analysis of total proteins for procaspase-9, cleaved caspase-9 and −3 and PARP levels. COV434-Mock and -FGF1 cells were treated or not (0 h) with 25 µg/mL etoposide for 2, 4, 6, or 16 h. Lower panel: histograms present the average fold-change decrease of cleaved caspase-9, cleaved caspase-3 and cleaved PARP in COV434-FGF1 cells ± SEM from pooled results of three FGF1 overexpressing clones (*n* = 8). Two-tailed unpaired t-tests results are shown as \* for *P* ≤ 0.05, \*\* for *P* ≤ 0.01 and \*\*\* for *P* ≤ 0.001
Release of cytochrome c from intermembrane mitochondrial space was then studied using immunofluorescence. Following 4 h of etoposide treatment, FGF1 significantly decreased the release of cytochrome c (Fig. [1b](#Fig1){ref-type="fig"}). This suggests that FGF1 inhibits MOMP during etoposide-induced cell death.
We further examined the effect of FGF1 overexpression on the cleavage of procaspase-9 and procaspase−3 and the caspase target PARP following an etoposide treatment. In COV434-FGF1 cells, the cleavage of these apoptotic markers is delayed and less pronounced than in COV434-Mock cells (Fig. [1c](#Fig1){ref-type="fig"}). Similar results were obtained with cisplatin instead of etoposide (Supplementary Figs. [S1B](#MOESM1){ref-type="media"}, [S1C](#MOESM1){ref-type="media"}).
Therefore, FGF1 overexpression in COV434 cells is able to prevent a decrease in ΔΨm, cytochrome c release and the subsequent activation of caspases arising during etoposide and cisplatin-induced apoptosis. These results suggest that FGF1 acts either upstream or at the mitochondrial level to promote survival.
FGF1 overexpression attenuates the etoposide-induced G2/M cell cycle arrest {#Sec4}
---------------------------------------------------------------------------
As etoposide provokes an S phase delay and a G2 phase accumulation of treated cells^[@CR36]^, we explored the effect of FGF1 overexpression on cell cycle. This was done by cytometry analysis after Hoechst 33342 staining.
Distribution of cells in the different cell cycle phases was similar for COV434 (COV434-NT), COV434-Mock and COV434-FGF1 cells in the absence of etoposide (Fig. [2a, b](#Fig2){ref-type="fig"}). After 16 h of etoposide treatment, we observed a decrease of the percentage of cells in G1 phase and an increase of the percentage of cells in S and G2/M phases in all cell populations as expected. Interestingly, a significantly lower proportion of COV434-FGF1 cells accumulated in G2/M phases compared to COV434-NT and -Mock cells. In conclusion, FGF1 partly inhibits the etoposide-induced cell cycle arrest at G2/M phases in COV434 cells.Fig. 2FGF1 overexpression attenuates the etoposide-induced G2/M cell cycle arrest.Non-transfected (NT), mock and FGF1-overexpressing COV434 cells were treated with etoposide for 16 h, or untreated (Ctl). DNA was stained with Hoechst 33342 and cellular content was analyzed by flow cytometry. **a** Cytograms showing cell cycle phase distribution (G1, S, G2/M). **b** The histograms show the average percentages of cells in each cell cycle phase ± SEM. Results for 4 independent experiments in replicate (*n* = 10). T-tests compare to NT Eto. \**P* ≤ 0.05, \*\**P* ≤ 0.01, \*\*\**P* ≤ 0.001
The FGF1 anti-apoptotic activity involves mainly a FGFR-independent pathway {#Sec5}
---------------------------------------------------------------------------
FGF1 lacks a secretion signal peptide and is not secreted under basal conditions. As expected, affinity chromatography of conditioned media using heparin sepharose showed that FGF1 is not detected in the culture medium of COV434-FGF1 cells in the absence of etoposide (Fig. [3a](#Fig3){ref-type="fig"}). Under stress conditions such as hypoxia or heat shock, FGF1 can be secreted through non-canonical pathways and activate FGFRs on neighboring cells^[@CR37]^. Therefore, we wanted to evaluate the contribution of the FGFR-dependent and FGFR-independent pathways in the anti-apoptotic activity of FGF1. For this purpose, the FGFR-dependent pathway was inhibited using the FGFR1/3 inhibitor PD173074. The efficiency of this inhibitor towards rFGF1 activity was confirmed. Indeed, recombinant FGF1 (rFGF1) added in the culture medium partly protected COV434 cells from etoposide-induced cell death and the addition of PD173074 abrogated this anti-apoptotic activity (Fig. [3b](#Fig3){ref-type="fig"}). Interestingly, PD173074 only partially reversed FGF1-induced resistance to etoposide-induced apoptosis in COV434-FGF1 cells (Fig. [3c](#Fig3){ref-type="fig"}). These data suggest that overexpressed FGF1 acts mainly through FGFR-independent pathways to exert its anti-apoptotic activity in COV434 cells.Fig. 3Both FGFR-dependent and FGFR-independent pathways are involved in FGF1 anti-apoptotic activity.**a** Western blot analysis for FGF1 levels in total extracts and conditioned media of non-transfected COV434 (NT), COV434-Mock and COV434-FGF1 cells. Endogenous FGF1 is detected in all total extracts whereas exogenous FGF1-V5-His is seen only in FGF1-overexpressing cells as expected. **b** Average apoptosis rates ± SEM for 3 experiments done in triplicate measured by flow cytometry of COV434-Mock cells. Cells were pretreated or not with the FGFR1/3 inhibitor PD173074 (25 nM for one hour), followed or not by a treatment with 50 ng/mL of recombinant FGF1 (rFGF1) supplemented with 10 µg/mL heparin for 24 h. On the next day, these treatments were renewed adding or not etoposide for 16 h. **c** Average apoptosis rates ± SEM for 3 experiments done in triplicate measured by flow cytometry of COV434-Mock and COV434-FGF1 cells pretreated or not with 25 nM PD173074 for 24 h, and treated or not with etoposide (25 µg/mL for 16 h). Two-tailed *t*-tests are indicated by \* for *P* ≤ 0.05, \*\* for *P* ≤ 0.01, and \*\*\* for *P* ≤ 0.001
FGF1 does not affect p53 transcriptional activities in the COV434 ovarian cell line {#Sec6}
-----------------------------------------------------------------------------------
Etoposide is known to activate the pro-apoptotic p53 protein. In our previous studies, we have shown that FGF1 is able to regulate p53-dependent apoptosis by decreasing p53 stability, phosphorylation and transcriptional activities in fibroblast and neuronal cells^[@CR16]--[@CR19]^. We thus wondered whether FGF1 could exert its anti-apoptotic activities in a similar way in COV434 cells. First, we examined p53 protein levels and phosphorylation at serine 15, a post-translational modification crucial for its transcriptional activity^[@CR38]^.
Without etoposide, there is no significant difference in p53 protein levels between COV434-Mock and -FGF1 cells. After 2 or 3 h of etoposide treatment, we observed significantly lower p53 levels in COV434-FGF1 compared to Mock cells but no difference after 16 h of treatment (Fig. [4a](#Fig4){ref-type="fig"} and Supplementary Fig. [S2A](#MOESM1){ref-type="media"}). Even though p53 levels are lower in COV434-FGF1 at 2 and 3 h, levels of p53 phosphorylated at serine 15 are not lower and even significantly higher after 16 h of etoposide treatment (Fig. [4a](#Fig4){ref-type="fig"} and Supplementary Fig. [S2B](#MOESM1){ref-type="media"}).Fig. 4p53 transcriptional-dependent activities are dispensable for the apoptosis of the COV434 ovarian cell line.**a** COV434-Mock and COV434-FGF1 cells were treated or not with 25 µg/mL etoposide for 1, 2, 3 or 16 h. Total protein extracts were analyzed for p53 and Ser15-phosphorylated p53 by western blotting. **b** COV434 Mock and COV434-FGF1 cells were treated or not with etoposide for 1, 2, 3 or 16 h. Total proteins were analyzed for PUMA, Bax and p21 protein levels by western blotting. **c** Average apoptosis rates ± SEM for 2 experiments done in triplicate measured by flow cytometry of non-transfected COV434 (NT), Mock and FGF1 cells treated with etoposide for 16 h. Cells were pretreated or not for 90 min with the p53 transcriptional inhibitor pifithrin-alpha (PFT-α, 30 µM). **d** Western blot analysis of p21 and Bax protein levels in COV434 Mock and FGF1 cells pretreated with PFT-α (30 µM for 90 min), followed by an etoposide (25 µg/mL) treatment for 6 or 16 h. Two-tailed *t*-tests results are shown by \* for *P* ≤ 0.05, \*\* for *P* ≤ 0.01, and \*\*\* for *P* ≤ 0.001
We next examined the protein levels of p53 transcriptional targets such as *PUMA*, *Bax*, *p21,* and *TIGAR*. We have previously shown that *Bax* and *PUMA* transactivation by p53 is attenuated in the presence of FGF1 in rat embryonic fibroblasts and pheochromocytoma PC12 cell line^[@CR16],[@CR17]^. Unexpectedly, no decrease in the protein levels of PUMA, Bax, p21 and TIGAR was seen in etoposide treated COV434-FGF1 cells. They are even significantly more elevated in COV434-FGF1 compared to COV434-Mock cells after 3 or 16 h of etoposide treatment (Fig. [4b](#Fig4){ref-type="fig"}, Supplementary Figs. [S3](#MOESM1){ref-type="media"}, [S4](#MOESM1){ref-type="media"}).
Finally, we investigated the involvement of p53 transcriptional-dependent activities in etoposide-induced apoptosis in COV434 cells using pifithrin-α (PFT-α), a p53 transcriptional activity inhibitor^[@CR39]^. A decrease in apoptosis is expected with PFT-α treatment if transcriptional activities of p53 are important for etoposide-induced cell death in COV434 cells. On the opposite, no decrease in the percentage of apoptotic cells (after 16 h of etoposide treatment) was observed upon treatment with PFT-α. PFT-α even slightly increased apoptosis following a 16 h etoposide treatment in COV434-FGF1 cells (Fig. [4c](#Fig4){ref-type="fig"}).
The blockage of the transcriptional activity of p53 was confirmed by testing the protein levels of its transcriptional targets (*p21* and *Bax*) in the presence of PFT-α. A decrease of p21 and Bax protein levels was seen 6 or 16 h after etoposide treatment in the presence of PFT-α, confirming its efficiency (Fig. [4d](#Fig4){ref-type="fig"}).
In conclusion, p53-transcriptional activity remains unaffected after FGF1 overexpression in COV434 cells and seems dispensable for etoposide-induced apoptosis in these cells.
The p21 anti-apoptotic activity is not involved in FGF1-induced resistance to etoposide {#Sec7}
---------------------------------------------------------------------------------------
As p21 displays anti-apoptotic activities, among others^[@CR40]^, we examined the significance of the p21 accumulation observed in COV434-FGF1 cells after an etoposide treatment. Our goal was to determine if p21 anti-apoptotic activity was responsible for FGF1 anti-apoptotic activity. Following a transient transfection using a p21 transcript-targeting siRNA or a control siRNA, COV434-Mock, and COV434-FGF1 cells were treated with etoposide for 6 h. In mock cells, we observed a significant increase of the apoptosis rates following p21 knockdown, whereas p21 knockdown showed no significant increase of COV434-FGF1 cells apoptosis (Fig. [5a](#Fig5){ref-type="fig"}). Furthermore, we observed a higher rate of cleavage of the pro-caspases-9, pro-caspases-3, and PARP upon p21 knockdown in COV434-Mock cells but not in COV434-FGF1 cells (Fig. [5b](#Fig5){ref-type="fig"}).Fig. 5p21 anti-apoptotic activities are not necessary for FGF1-induced resistance to etoposide.**a** COV434 Mock and COV434 FGF1 cells were transfected or not with scramble (scr) or p21 siRNA. Upper panel: average apoptosis rates ± SEM for 2 experiments done in duplicate were measured by flow cytometry in cells treated with etoposide for 6 h, or not treated (Ctl). Lower panel: Western blot analysis for p21 protein levels in COV434 Mock and COV434 FGF1 cells transfected with scr siRNA or p21 siRNA. **b** Western blot analysis of total proteins for p21 and for procaspase-9, cleaved caspase-9, and cleaved caspase-3, and PARP levels. COV434 Mock and FGF1 cells transfected with scr or p21 siRNA were treated or not with etoposide for 16 h. One experiment representative of 3 independent experiments is shown. \**P* ≤ 0.05, \*\**P* ≤ 0.01, \*\*\**P* ≤ 0.001
Altogether these results indicate that p21 exerts anti-apoptotic activities in COV434 cells, but the accumulation of p21 in COV434-FGF1 cells after an etoposide treatment does not seem responsible for FGF1 anti-apoptotic activities.
FGF1 regulates p53 pro-apoptotic mitochondrial activities {#Sec8}
---------------------------------------------------------
Decreasing p53 protein levels with p53-targeting siRNA lowered the apoptotic rate of COV434 cells. This demonstrates the involvement of p53 in the etoposide-induced apoptosis in this cell line (Fig. [6a](#Fig6){ref-type="fig"}). Apoptosis rate was slightly decreased in p53 knockdown cells probably due to the remaining high levels of p53 (Fig. [6a](#Fig6){ref-type="fig"}). To confirm that p53 is able to kill COV434 cells, clonogenic survival assays were done. COV434 cells were transfected with an empty vector or a vector encoding p53^WT^. After 12 days of selection with geneticin, no clones were obtained for COV434 cells overexpressing p53^WT^. Thus p53 can induce cell death in COV434 cells (Fig. [6b](#Fig6){ref-type="fig"}). Given that p53 can exert its pro-apoptotic activities directly at the mitochondrion we tried to determine if mitochondrial p53 is able to kill COV434 cells^[@CR41]^. Clonogenic survival assays were also done for COV434 cells transfected with a vector encoding p53 fused to the transmembrane domain of BCL-X~L~ (p53^CTB^). p53^CTB^ is predominantly localized at the mitochondria as we confirmed by immunofluorescence (data not shown). As with p53^WT^, no clones were obtained for COV434 cells transfected with p53^CTB^ (Fig. [6b](#Fig6){ref-type="fig"}). Furthermore, transient transfections of p53^WT^ or p53^CTB^ showed that mitochondrial p53 is able to induce apoptosis in COV434 cells under basal and etoposide conditions (Supplementary Fig. [S5](#MOESM1){ref-type="media"}).Fig. 6p53 is involved in the induction of apoptosis in COV434 cells.**a** COV434 cells were transfected or not (NT) with scramble (scr) or p53 siRNA. Upper panel: histogram represents average apoptosis rates ± SEM for 2 experiments done in duplicate measured by flow cytometry in cells treated with etoposide for 17 h, or not treated (Ctl). Lower panel: Western blot analysis for p53 protein levels in COV434 cells transfected with scr siRNA or p53 siRNA. **b** COV434 cells were transfected with an empty vector (Mock), a vector encoding p53 wild-type (p53^WT^) and a vector encoding p53 fused to the mitochondrial transmembrane domain of BCL-X~L~ (p53^CTB^). Cells were treated with of G418 for one week and then stained with ethidium bromide and visualized using Chemidoc (Biorad)
The involvement of p53 transcription-independent mitochondrial activities in the apoptosis of COV434 cells was also examined by using the p53 mitochondrial localization inhibitor pifithrin-µ (PFT-μ, 2-phenylethynesulfonamide). PFT-µ blocks p53-mitochondrial translocation by interfering with the HSP70/p53 complex formation^[@CR42]^. While PFT-α was inefficient to inhibit apoptosis upon 16-hour-long etoposide treatment, PFT-µ significantly inhibited the etoposide-induced apoptosis of COV434-NT, COV434-Mock, and COV434-FGF1 cells (Fig. [7a](#Fig7){ref-type="fig"}). A decrease of mitochondrial p53 level was observed in the presence of PFT-μ and etoposide compared to etoposide alone. Because of data disparity, this decrease is not statistically significant (Fig. [7b](#Fig7){ref-type="fig"}).Fig. 7FGF1 regulates p53-mitochondrial localization.**a** Average apoptosis rates ± SEM for 3 experiments done in triplicate were measured by flow cytometry in non-transfected (NT), mock and FGF1-overexpressing COV434 cells treated with etoposide for 16 h. These cells were pretreated or not with the p53 mitochondrial localization inhibitor pifithrin-mu (PFT-µ, 10 µM for 90 min). **b** Non-transfected COV434 cells were pretreated or not with PFT-µ (10 µM for 90 min) prior to etoposide treatment (25 µg/mL for 2h30). Mitochondrial localization of p53 was determined by western blot analysis of enriched mitochondrial fractions (upper panel). Quantification of mitochondrial p53 normalized to TOM40 for four experiments (lower panel). *\**Molecular weight lane. **c** Cytosolic, mitochondrial and total proteins of COV434-Mock, and COV434-FGF1 cells, treated or not with etoposide for 4 h, were analyzed for p53 and FGF1 localization by western blot (left panel). Quantification of mitochondrial p53 normalized to TOM40 (right panel). Results are from 6 independent experiments, means ± SEM, two-tailed *t*-test results are shown by \* for *P* ≤ 0.05, \*\* for *P* ≤ 0.01, and \*\*\* for *P* ≤ 0.001
To determine whether FGF1 could affect the mitochondrial localization of p53 as does PFT-µ, we next investigated the amount of p53 in the mitochondrial fraction. Without etoposide, mitochondrial p53 level is higher in COV434-Mock than in COV434-FGF1 cells. After a 4-hour-long etoposide treatment, mitochondrial accumulation of p53 was seen in both cell lines. Nevertheless, COV434-Mock still presented higher amounts of mitochondrial p53 than COV434-FGF1 (Fig. [7](#Fig7){ref-type="fig"}c). Recombinant FGF1 added to COV434 parental cells also decreased mitochondrial localization of p53 under etoposide treatment (Supplementary Fig. [S6](#MOESM1){ref-type="media"}). Interestingly, FGF1 was detected in the mitochondrial fraction of FGF1 overexpressing cells (Fig. [7c](#Fig7){ref-type="fig"}).
These results suggest that p53 mitochondrial activities are important for etoposide-induced apoptosis in COV434 cells. Furthermore, FGF1 overexpression decreases mitochondrial levels of p53 which could explain the FGF1 anti-apoptotic activity.
Discussion {#Sec9}
==========
Our purpose was to understand the role of FGF1 in chemoresistance in ovarian tumors. Of note, FGF1 is not the sole member of the FGF family that is associated with chemoresistance. FGF2 levels are frequently elevated in solids and hematological cancers. Interestingly expression of FGF2 correlates with resistance of ovarian tumors (among others) to paclitaxel^[@CR43]^. In addition, FGF2 can even serve as prognostic factor for some cancer types^[@CR44]^. As FGF1, we observed that FGF2 decreased etoposide-induced apoptosis of COV434 cells (supplementary Fig. [S7](#MOESM1){ref-type="media"}). Further work will help to determine if FGF2 also affects mitochondrial p53.
In this study, in agreement with FGF1 putative role in chemoresistance, we show that overexpression of FGF1 in the COV434 cell line renders cells less prone to etoposide- and cisplatin-induced apoptosis. Furthermore, we provide new data on the mechanisms involved in apoptotic cell death of these cells. Previous studies had shown that they undergo apoptosis when treated with etoposide or cisplatin and that cisplatin-induced apoptosis partially depends on p53^[@CR45],[@CR46]^. In agreement with these data, etoposide and cisplatin induced the accumulation and activation of p53 in our study. p53 is wild-type in the COV434 cell line^[@CR47]^ and accumulates at the nucleus (our unpublished data) and mitochondria in the presence of a chemotherapeutic agent. Therefore, p53 could display transcriptional and non-transcriptional activities. Our results using chemical inhibitors targeting these activities suggest that p53 transcriptional activities are rather anti-apoptotic since PFT-α increased cell death induced by etoposide. p21 could be the major actor of this anti-apoptotic activity as we observed that PFT-α strongly inhibits its accumulation. Moreover, p21 knockdown using RNA interference markedly increases etoposide-induced apoptosis.
The p21 anti-apoptotic activity is associated with its cytoplasmic localization which relies on its phosphorylation by Akt^[@CR48]^. In the cytoplasm, p21 can interact with various proteins involved in apoptosis, including procaspase-3 (for review see^[@CR49],[@CR50]^). Thereby, p21 inhibits procaspase-3 cleavage and subsequent activation^[@CR51]^. In testicular cancer, cytoplasmic p21 has been associated to cisplatin resistance^[@CR52]^. Furthermore, recent studies described high levels of cytoplasmic p21 in cisplatin-resistant ovarian cells (C13\* cells) and the restoration of sensitivity to cisplatin when this cytoplasmic localization is impaired. The Hsp27 chaperone that is also overexpressed in C13\* resistant cells could be responsible for the accumulation of cytoplasmic p21^[@CR53],[@CR54]^. Here, we confirmed the anti-apoptotic activity of p21 upon stress-induced cell death of COV434 ovarian cells. Since p21 can be cleaved by caspases during apoptosis^[@CR55]^, the accumulation of p21 observed in FGF1 overexpressing cells could be a consequence of the diminished activity of caspases. This could allow for the maintenance of the anti-apoptotic activity of p21. Nevertheless, markedly decreasing overall p21 levels upon RNA interference in FGF1-overexpressing cells leads to a non-significant increase in apoptosis. The p21 anti-apoptotic activity may thus not be the sole or main factor involved in FGF1 anti-apoptotic activities. Another explanation could be that p21 localization is modified by FGF1 and that RNA interference was not sufficiently efficient to decrease cytoplasmic p21 levels in our experiments. Interestingly, we observed a more pronounced accumulation of cytosolic p21 in COV434-FGF1 cells vs. COV434 mock cells upon etoposide treatment (Supplementary Fig. [S8](#MOESM1){ref-type="media"}).
Only rare data are found in the literature concerning the role of p53 in the apoptosis of granulosa tumor cells. Woods and colleagues showed p53 involvement in cisplatin-induced cell death of COV434 and KGN cells, two cell lines commonly used as models of juvenile and adult GCT respectively^[@CR46]^. To our knowledge, we are the first to investigate the role of p53 mitochondrial activities during cell death of COV434 cells. We show here that p53 accumulates at the mitochondrion upon etoposide treatment of these cells and that inhibition of this mitochondrial localization using pifithrin-µ decreases cell death. Moreover, as pifithrin-µ, FGF1 diminished p53 accumulation at the mitochondrion. Therefore in COV434 cells, FGF1 affects p53 activities by a novel mechanism, which does not rely on the inhibition of transcriptional activities of p53 such as in neuronal and fibroblast cell lines^[@CR16]--[@CR19]^. This is the first report highlighting the regulation of mitochondrial p53 by FGF1.
Besides its regulation of p53 mitochondrial localization, FGF1 could also promote survival through other mechanisms such as regulation of the subcellular location of the oncoprotein MUC1-C. MUC1-C is known to attenuate genotoxic stress-induced cell death, conferring resistance to chemotherapeutic reagents such as cisplatin and etoposide^[@CR56]^. MUC1-C interacts with activated BAX and inhibits its oligomerization at the mitochondrion^[@CR57]^. Interestingly, FGF1 can induce mitochondrial localization of MUC1-C^[@CR58]^. Therefore, one could think that, in COV434-FGF1 cells, MUC1-C could compete with p53 for binding to pro-apoptotic members of Bcl-2 family, thereby inhibiting MOMP.
How FGF1 affects p53 mitochondrial localization must be further investigated. The molecular mechanisms involved in mitochondrial localization of p53 are not fully understood but candidate proteins are TRAF6, Tid1, and CHCHD4. The TRAF6 E3 ubiquitin ligase restricts the mitochondrial localization of p53 in basal conditions^[@CR59]^. The TRAF6 E3 ligase mediates K63-linked ubiquitination of p53 in the cytosol. Following cisplatin or etoposide treatment, cytosolic p53 is less ubiquitinated by TRAF6 and accumulates at the mitochondrion. Interestingly, *Traf6*^*−**/**−*^ mice cells undergo spontaneous apoptosis in the thymus, spleen and lungs, and the mitochondrial localization of p53 is largely involved in this cell death. It will be interesting to investigate the consequence of FGF1 overexpression on TRAF6 activity in our COV434-FGF1 cells but also in A2780-CIS cells that resist cisplatin and overexpress FGF1 in comparison to A2780 cisplatin sensitive cells^[@CR8]^. Further work must also be done to determine if FGF1 regulates the mitochondrial import protein Mia40/CHCHD4 and the matrix protein Tid1 that interact with p53 and promote its mitochondrial localization^[@CR60]--[@CR62]^.
The precise localization of FGF1 at the mitochondrion must be further investigated as FGF1 has been proposed to interact with the inner membrane protein SFXN1^[@CR21]^. Mitochondrial subfractionation must thus be done to determine whether FGF1 associates with the outer mitochondrial membrane or localizes inside the mitochondrion (in the inner membrane, the intermembrane space or the matrix). Considering that we and others have shown that p53 and FGF1 can interact^[@CR18],[@CR21]^ and that p53 can localize in the different compartments of the mitochondria^[@CR63],[@CR64]^, it will be interesting to determine whether p53 and FGF1 are found at the same place and interact at the mitochondrion.
The fact that FGF1 was found to interact with the mitochondrial proteins SFXN1 and mthsp70/GRP75/mortalin^[@CR21],[@CR65]^ led us to propose that FGF1 could mediate its survival activities by a direct action at the mitochondrion. Mthsp70/GRP75/mortalin is a chaperone protein predominantly found at the mitochondrion where it participates in the import of proteins in association with TIM proteins. Mortalin is overexpressed in various cancers and can interact with p53 (see ref. ^[@CR66]^ for a review). This interaction is seen as a mechanism of cytoplasmic retention of p53 to prevent the nuclear activities of this transcription factor. For example, in hepatocellular carcinomas cell lines, mortalin interacts with p53 and its downregulation induces nuclear localization of p53^[@CR67]^. Furthermore, apoptosis induced by mortalin knock-down is reversed by PFT-µ, suggesting that mitochondrial activities of p53 are inhibited by mortalin^[@CR68]^. As FGF1 and p53 interact with mortalin and FGF1 regulates p53 mitochondrial localization, one could think that FGF1 may inhibit mitochondrial localization of p53 by promoting mortalin/p53 interaction.
In conclusion, our study provides insights into the mechanisms of cell death in COV434 juvenile granulosa tumor cells and into chemoresistance induced by FGF1. We show for the first time that mitochondrial p53 induces cell death of COV434. Since p53 is not mutated in GCTs, contrarily to most epithelial ovarian tumors, progress in understanding the molecular mechanisms of cell death in these cells is of crucial importance. As GCTs are less frequent than epithelial ovarian cancers (EOC), the study of GCTs has been relatively neglected and therapeutic strategies for EOC have been applied to GCTs^[@CR47]^. However these tumors greatly differ on both molecular and morphological levels. Investigating the molecular mechanisms involved in response to chemotherapy in GCTs is thus of major interest. Further work must be done to determine if the same mechanisms (p53-induced cell death and FGF1 anti-apoptotic activity) are found in adult (KGN) and juvenile tumor cells. More generally, this study provides the molecular basis for the comprehension of the role of FGF1 in the resistance to chemotherapy in ovarian cancers and could potentially be extended to other cancers overexpressing FGF1.
Materials and methods {#Sec10}
=====================
Cell culture, transfections, and chemicals {#Sec11}
------------------------------------------
The COV434 cell line was a kind gift from Sandrine Caburet. COV434 were cultured in DMEM/F12 medium (ThermoFisher, Waltham, MA, USA) supplemented with 10% FBS (ThermoFisher), 1% GlutaMAX (ThermoFisher), 100 μg/ml penicillin and 100 U/ml streptomycin (ThermoFisher) at 37 °C, 5% CO~2~. COV434-Mock and COV434-FGF1 were obtained following stable transfection with 2.5 µg of pcDNA3.1D V5-His plasmid containing or not the *FGF1* coding sequence, using 8 µl of Lipofectamine LTX and 2.5 µL Reagent plus (Life Technologies, Carlsbad, CA, USA). Transfected cells were selected in 500 µg/mL of Geneticin G418 (Life Technologies) for two weeks. The isolated clones were amplified and used for experiments. For p21 and p53 knockdown experiments, 3 × 10^5^ cells were plated on 6-well plates. At 30% confluence, siRNA transfection was done using 80 pmol of human p21 siRNA, human p53 siRNA or siRNA-A control (Santa Cruz, Dallas, TX, USA) and 4 µl lipofectamine RNAimax (Invitrogen). Etoposide (Sigma-Aldrich, St Louis, MO, USA) was used at 25 µg/mL. Pifithrin-α and pifithrin-µ (2-phenylethynesulfonamide, PFTμ) are from Enzo Lifesciences Farmingdale, NY, USA, and FGFR1/3 inhibitor from Tocris Bioscience, Bristol, UK. Recombinant FGF1 (R&D Systems, Minneapolis, MN, USA) was used in combination with 10 µg/mL heparin (Sigma-Aldrich).
Clonogenic survival assays {#Sec12}
--------------------------
COV434 cells were transfected with an empty vector or vectors encoding p53^WT^ or p53^CTB^ (kindly given by Pr Ute M Moll) as described above for stable transfections. After 12 days of selection in geneticin containing medium, clones were stained with ethidium bromide following the procedure described by Guda et al^[@CR69]^. Images were acquired using Chemidoc system and the ImageLab software (BIORAD).
Flow cytometry {#Sec13}
--------------
To monitor apoptosis, 5 × 10^5^ cells were seeded on 6-well plates. Following the appropriate drug treatment, they were trypsinized, harvested and centrifuged. The cellular pellet was suspended in 0.2 µM of DiOC~6~(3) (Molecular Probes, Eugene, OR, USA) containing culture medium and submitted to flow cytometry after an incubation of 30 min at 37 °C. For cell cycle analysis, the collection procedure was as for apoptosis except that the cellular pellet was suspended in culture medium containing 1 µg/mL of Hoescht 33342 (82261, Sigma-Aldrich) for 20 min at 37 °C. Cytometry experiments were performed on *BD LSRFortessa*™ cell analyzer (BD Biosciences Franklin Lakes, NJ, USA). Analysis was done using the BD FACSDiva software (BD Biosciences).
Total protein extraction and mitochondrial fractionation {#Sec14}
--------------------------------------------------------
For total protein extracts, 2 × 10^6^ cells were plated on 60 mm dishes. Following an appropriate drug treatment, they were trypsinized, harvested and centrifuged. The cellular pellet was suspended in ELB buffer (250 mM NaCl, 50 mM HEPES, 5 mM EDTA, 0.1% NP40, 0.1 M DTT) supplemented with protease 1/100 inhibitors cocktail (Cat.No.04693116001, Roche, Mannheim, Germany) and 0.2 mM of sodium orthovanadate/phosphatase inhibitors (cat.567540, Calbiochem, San Diego, CA, USA). For mitochondrial isolation, 2 × 10^7^ cells cultured on 100 mm dishes were harvested, centrifuged and washed with PBS. Fractions enriched in mitochondria were obtained either using the Mitochondrial isolation kit for mammalian cells (Thermofisher) according to the manufacturer's guidelines or following procedure described elsewere^[@CR64]^. The purified mitochondria were lysed in TBS-CHAPS 2% (EUROMEDEX, Strasbourg, France) supplemented with protease inhibitors.
Conditioned media and affinity chromatography {#Sec15}
---------------------------------------------
Cell extracts and conditioned media from COV434-mock or COV434-FGF1 cells were collected and heparin-sepharose purified as previously described^[@CR19]^. FGF1 content was determined by western-blot analysis.
Immunoblotting {#Sec16}
--------------
Equal quantities of proteins (10 to 30 µg) were run on Mini-PROTEAN TGX Stain Free precast polyacrylamide 4--20% gels (BIORAD) in Tris glycine-SDS buffer (BIORAD). Proteins were transferred on Immobilon-P PVDF membranes (Millipore) in Tris glycine-Ethanol 20% buffer (BIORAD). Stain Free technology allows visualizing total proteins without use of any dye. The following antibodies were used at dilutions from 1/200 to 1/1000: anti-caspase 9 (32539, Abcam, Cambridge, UK), anti-cleaved caspase 3 (5A1E, Cell Signaling, Danvers, MA, USA), anti-PARP (Cell signaling 92845), anti-FGF1 (R&D AB-32-NA, Minneapolis, MN, USA) and anti-FGF1 (Cat.No.010-24161, WAKO, Osaka, Japan), anti-p53 DO1 (Santa Cruz Sc-126), anti-p53pSer15 (Cell signaling 92845), anti-PUMA N-19 (Santa Cruz Sc-19187), anti-Bax I-19 (Santa Cruz Sc-930), anti-p21 C-19 (Santa Cruz Sc-397-G) and anti-TOM40 H-300 (Santa Cruz Sc-11414). Secondary antibodies were HRP coupled (Jackson Immunoresearch, West Grove, PA, USA) and the revelation was performed using Clarity Western ECL Blotting Substrate (BIORAD). Secondary antibodies were HRP coupled (Jackson Immunoresearch) and the detection was performed using Clarity Western ECL Blotting Substrate (BIORAD). The chemiluminescent signal was captured by Chemidoc (BIORAD) and quantification was performed with the ImageLab software (BIORAD).
Immunofluorescence {#Sec17}
------------------
1.2 × 10^6^ cells were plated on 60 mm dishes containing coverslips. Following the appropriate drug treatment, cells were fixed with 3.7% paraformaldehyde at 60% confluence then washed with PBS and permeabilized for 30 min in 0.5% Tween20. Nonspecific sites were blocked with PBS-BSA 1% for 30 min. Primary (anti-cytochrome c, BD Pharmingen, Cat.556432) and secondary antibodies (anti-mouse AlexaFluor488, ThemoFischer scientific) diluted in PBS-BSA 3% were incubated for 60 min each. Nuclei were stained with 1 µM TO-PRO-3 reagent (Invitrogen). Mounting was done using ProLong Gold Antifade Mountant (Invitrogen). Image acquisition was performed at the CYMAGES imaging facility (on Leica TCS SPE and Leica TCS SP8 confocal microscopes) and images were analyzed using the ImageJ software.
Statistical analysis {#Sec18}
--------------------
Two-tailed unpaired Student's *t*-tests were performed.
Electronic supplementary material
=================================
{#Sec19}
Supplementary figures
**Electronic supplementary material**
**Supplementary Information** accompanies this paper at (10.1038/s41389-018-0033-y).
**Publisher\'s note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Financial supports were obtained from the University of Versailles Saint-Quentin-en-Yvelines, the IUT de Vélizy/Rambouillet, the Ecole Pratique des Hautes Etudes, and the Ligue Nationale contre le Cancer. The authors wish to thank Dr Sandrine Caburet for providing COV434 cells and Dr Sébastien Gaumer for critical reading of this manuscript. The authors also greatly acknowledge Benoît Maury and Aude Jobart-Malfait of the CYMAGES facility.
Conflict of interest {#FPar1}
====================
The authors declare that they have no conflict of interest.
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#sec1-1}
============
In an era of increased complexity and escalating costs of clinical research, a focus on personalized medicine and patient empowerment, drug development is undergoing a metamorphosis. Typically, when a clinical trial is conducted, clinicians and patients collaborate with the sponsor for determining the safety and effectiveness of a molecule under experimental treatment. Trials are usually designed keeping in mind the feasibility and ease with which the sponsor can conduct the study. This can lead to a large number of costly and complex trials being conducted, without addressing the patient\'s convenience or needs. As per an IMS Health Study forecast, global spending on medicines is targeted to increase 30% to \$1.3 trillion by 2018.\[[@ref1]\] The spend has grown, but patient\'s needs are not being met. Thus, patient-centric drug development is now becoming the model that the industry is following. Today, patients are aware, technology-driven, and informed-driving the change in mindset and way clinical trials are being approached and conducted.
DIGITAL PATIENT -- PATIENT AT THE CENTER OF THE NETWORK ENVIRONMENT {#sec1-2}
===================================================================
A number of factors can influence a patient\'s decision to participate in a clinical trial including financial, social, philanthropic or altruistic.\[[@ref2]\] Prior to enrolling in trials, patients today invest time in learning about the disease of interest, the drug mechanism for cure, locations where trials are being conducted and results from similar trials to name a few, through the internet. This has enhanced the quality of discussions occurring between the physician and patient. Benefits of engaging these "digital patients" have been realized. Contract research organizations (CROs) are offering a ready repository of patients who have enrolled in different trials to pharma companies keen on conducting trials-driving cost and effort saving. For example, quintiles have more than three million patients under their "Digital Patient Unit" program which also helps to use real world patients for the faster testing of inclusion/exclusion criteria and allows sponsors to prescreen the subjects and refer them, if needed, to clinical trial sites.\[[@ref3]\]
INVOLVING PATIENTS FROM THE INITIAL STAGES IN DRUG DEVELOPMENT {#sec1-3}
==============================================================
Pharma are using technology to run contests and competitions to gather ideas and feedback on trial designs, informed consent forms (ICFs) and protocols. Open discussions on draft protocols specially focused on endpoints and visited structure are encouraged, and updates are made based on the responses received. Some of the areas where patients are being actively engaged are listed below.
Seeking patient inputs in informed consent forms {#sec2-1}
------------------------------------------------
The draft guidelines on the informed consent released by the Food and Drug Administration (FDA) in July 2014 could be used as a ready reference for building effective ICFs. While building ICFs, an approach, which is behavioral but science-based helps uncover patient insights. The usage of simple language easily deciphered by the participating subject will establish expectations regarding foreseen and unforeseen risks. ICFs that are short and precise will hold the attention. Clearly defining study objectives, end points and providing a detailed summary would retain the interest of the patient. It could also help in understanding the enrollment risks and also highlight factors that could be adjusted in the protocol for minimizing those risks.\[[@ref3]\] Having the ICF reviewed by a layman could also be considered as it would shed light on content that is too scientific in nature. Understanding the section of the protocol that led to distress while review (e.g., surgical procedure related requirement) could help one make amendments, if needed, in earlier stages of protocol development. Seeking clarification from patients on their understanding of the protocol could provide insights on building ICFs that would be better accepted by a larger patient population.\[[@ref4]\]
Building patient friendly protocols and grooming study staff {#sec2-2}
------------------------------------------------------------
While designing the study protocols, time should be invested in understanding the lifestyle of the patient population. Real world patients providing their inputs to building study protocols and study designs, which are more real and closer to life experiences, cost-effective and patient-centric is needed. Designing patient friendly visit schedules that are more flexible is important. In the case, patients are dependent on caregivers, considering the schedule of a caregiver would be beneficial. For better retention, care should be taken to provide facilities and an environment conducive to patients, especially during prolonged on-site visits, or visits that require patients to be in a fasted state. By being sensitive to a patient\'s comfort and needs, a better retention rate could be achieved.\[[@ref5]\]
Companies should provide appropriate trainings to the study staff and groom them to handle sensitive situations, especially when invasive procedures are followed.
PROMOTING OPEN COMMUNICATION CHANNELS BETWEEN BIOPHARMA, PHYSICIANS, PATIENTS, AND MEDIA {#sec1-4}
========================================================================================
The traditional communication mechanisms where Biopharma companies were at the center and messages were directed toward physicians, who in turn shared the messages with patients, have changed. There is a need to adopt newer tools and technologies that drive two-way communication. Promoting and accelerating direct to consumer advertising where companies still control the delivery of the messages, but the loop in media for a greater effect. Companies are actively using social listening techniques to review what patients are discussing online regarding disease state and issues to drive better patient participation and relationships.\[[@ref6]\] Researchers are skeptical regarding the type and extent of clinical information that is being shared and discussed by patients via online chats, such as protocol details, their experience of participating in trials, adverse event reactions etc., that can introduce bias for future trials.\[[@ref7]\]
Understanding the importance social media and its role in the dissemination of information, FDA has released draft guidance for industry on social media usage in June 2014. It provides recommendations for presenting benefit and risk information for FDA regulated prescription drugs or devices using social media such as Twitter, Yahoo, and Google.\[[@ref8]\] Medicine\'s New Zealand also released its updated code of practice in June 2014 where it has included its guidance on social media usage by the pharmaceuticals. This was done to separate social media from other type of advertising and clearly outline pathway for pharma companies keen to use social media channels -- an indication of increasing global acceptance of social media usage.\[[@ref9]\]
ENHANCING THE FOCUS ON PATIENT EDUCATION, ENGAGEMENT AND RETENTION USING TECHNOLOGY {#sec1-5}
===================================================================================
To improve clinical outcomes, increase patient satisfaction and incur profit revenue, engaging patients in their own healthcare is critical. The US government is encouraging the use of Electronic Health Records (EHRs) via HealthIT.gov and promoting incentives to doctors who use EHRs meaningfully to reduce medical errors and improve the quality of care. CROs and independent service providers are designing educative websites that can be accessed via mobiles or the internet for educating patients on diseases.
A few examples of available websites and tools below
Agency for Health Research and Quality maintained by US Department of Health and Human ServicesClinicalResearch.comWELVU -- Mobile First, an iPad- and iPhone-based educative tool providing medical illustration, quality scores, and health outcomes to engage patientsKrames patient education from StayWellExitCare OnScreen™ video solutions for patient education.\[[@ref10][@ref11][@ref12][@ref13]\]
The retention of patients in a trial is the key to the success of the overall project. Acurian, a service provider for recruitment and retention services uses platforms such as Facebook and Myspace for patient referrals and retention strategies.\[[@ref14]\] The easier it is to be compliant to study schedule, the better is the retention till the end. Dose compliance tracking tools like MediGuard™ enable reminders to be set up for dose intake.
USE OF CROWDSOURCING TECHNIQUES FOR ASSESSING THE PULSE OF THE PUBLIC {#sec1-6}
=====================================================================
Crowdsourcing has been used since long as a powerful tool to engage the masses in other industries. Wikipedia is one classic example. To maintain 23 million articles rich in content, the company uses crowd participation where the site is maintained by a community of passionate 80,000 users, who in turn are incentivized via a gamified award mechanism.\[[@ref15]\]
Companies are holding contests where relevant stakeholders, including medical communities, patient communities, and researchers are looped in to provide responses to survey questions via tools like "protocol builder." The FDAs approval of the first completely crowd-sourced protocol for multiple sclerosis by Transparency Life Science\'s reconfirms the potential that regulatory agencies see in the application of crowdsourcing. Online patient communities such as Mediguard.org and Clinical research.com are instrumental in changing face of healthcare, clinical trials and outcomes. They are beneficial to patients as connections can be developed between people with similar conditions, sharing clinical trials information and advice on management of diseases. The communities also aid in patient recruitment by prescreening potential participants online. Digital observational research where data are directly collected from patients and compared to data results in healthcare records (with reduced physician involvement) can be especially helpful for postmarketing surveillance of products consumed for extended periods.\[[@ref16]\]
Researchers, however, still feel concerned while opening complex clinical problems to a large number of strangers. Issues related to the misuse of intellectual property, the lack of surety in receiving solutions-raises questions on the effectiveness of crowdsourcing. However, companies can leverage on the potential of crowdsourcing and actively engage patients via several methods. Survey questions could be targeted to seek public views on inclusion-exclusion criteria, visit schema, alternative endpoints, etc. Inputs on the protocol design could also be requested from the physicians specialized in areas other than the study indication, e.g., Seeking inputs from an endocrinologist for the study indication diabetes could shed some light on the addition of some biochemical tests for specific analyte tracking crucial to diabetes. Also asking direct questions in the survey related to what kind of results patients expect to see at the end can help derive the primary and secondary endpoint for the trials.\[[@ref17]\]
One of the best-known examples where the application of crowdsourcing in research yielded great results was "Foldit." Here, a group of online gamers decoded the molecular structure of a monomeric retroviral protease, a long-standing scientific problem, by a protein folding game.\[[@ref18]\]
Pharma giants such as Eli Lily, AstraZeneca and Cleveland Clinic, have partnered with companies like innocentive. The company hosts challenges and supports the crowdsourcing of innovation solutions across the globe.
DRIVING THE DATA TRANSPARENCY PRINCIPLE FOR BUILDING TRUST AND CONFIDENCE {#sec1-7}
=========================================================================
The need for data transparency in clinical trials has been long discussed since it helps tackle multiple issues of concern. Having data transparency can help building trust and confidence in patients participating in trials. It also helps researchers prevent unnecessary duplication of trials, primarily when results of the similar trials have indicated that the product may be unsafe. The US Department of Health and Human services rolled out the notice of proposed rulemaking by detailing the process to be followed to ensure requirements established by FDA Amendments Act to strengthen public access to clinical trial data\[[@ref19]\] clearly demarcating cases where some companies would be apprehensive to share patient data citing patient confidentiality. To ensure data transparency, pharma companies can take simple, but effective measures like publishing the copy of the clinical trial protocol on safe sites like [www.clinicaltrials.gov](www.clinicaltrials.gov) and [www.clinicaltrialsregister.eu](www.clinicaltrialsregister.eu), making the results of the trials available irrespective of the outcome (positive or negative), publishing the results of the trials in peer-reviewed journals or company websites, ensuring that the results are available in languages well understood by the participants, etc.\[[@ref20]\]
ROLE OF TECHNOLOGY IN DRIVING PATIENT-CENTRICITY {#sec1-8}
================================================
The paradigm shift that we see in the way clinical research is evolving is attributed to a very large extent to technology and its widespread access to public.
HealthPatch MD, a wearable biosensor partnered between Vital Connect and Medidata helps in efficient remote patient monitoring. Its sensors and advanced algorithm provide continuous measurement of electrocardiogram grading, respiratory rate, skin temperature, heart rate, physical activity, etc. The combined technology (Medidata Clinical cloud and biosensor device) enables near real-time review of patient\'s health metrics\[[@ref21]\]Iodine, a health information website developed a new web-based application for cold and flu season. The app helps consumers to review \>300 options related to medication for cold and flu and compare medications that can help cure their own symptom\[[@ref22]\]Reg4all or Registries for All developed by Genetic Alliance with support from Sanofi works on the principle of matchmaking between patients and clinical trials. The tool provides privacy and flexibility to patients to decide, which groups can have a view/access to their data thus empowering the patientsTreato, a data mining company, monitors conversations of patients on Facebook, Twitter, and patient forums and helps pharma companies make sense of data that is being published by millions of patients across the globe. The company captures near real time comments from social media using a combination of natural language processing algorithms, patient language dictionaries, and big data analytics. These conversations help pharma gain insights into patients' lives and focus their efforts in the correct areas of drug development\[[@ref23]\]Patients like me have strong inbuilt mechanism where data are pulled from clinicaltrials.gov each night and it is matched with the list of 2,50,000 registered patients across globe helping the patients get better and updated results pertaining to the trials that might be conducted in their vicinity. The site also provides a forum when open communication and information exchange occurs between patients and feedback to improve the trial design can also be shared \[[Figure 1](#F1){ref-type="fig"}\].\[[@ref24]\]
{#F1}
CONCLUSION {#sec1-9}
==========
The above graph depicts a six-fold increase in the last 12 years related to a number of searches related to the term "patient-centered" in PubMed.\[[@ref25]\] It is indicative of the curiosity and awareness that exists in the industry for this concept.
Thus, we see that as the focus on healthcare and its access to patient\'s increases, it calls for the clinical research industry to adapt. It needs to consider technology advancement such as technology driving patient education, application of crowdsourcing to drive patient\'s contributions towards innovation, activities like ICF and protocol build based on patient\'s inputs, the usage of social media for data transparency as an integral part of the patient-centric move. The future seems to be bright where pharma and patients have the potential to work symbiotically that could result in better research results, improved health care and profits.
I would like to acknowledge insights and guidance offered by Dr. Nimita Limaye that has helped me while drafting this article.
**Source of Support:** Nil.
**Conflict of Interest:** None declared.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#sec1}
============
The minute pirate bugs (Hemiptera: Heteroptera: Anthocoridae) comprise 500 to 600 species of Heteroptera worldwide. These insects are important as natural members of the predatory fauna ([@bibr17]) and as biological control agents in many agroecosystems ([@bibr31]; [@bibr34]; [@bibr36] ). Two species of pirate bugs, *Orius insidiosus* (Say) and *Orius pumilio* (Champion), were found coexisting on an organic farm in north central Florida ([@bibr26]). Reports of both species have been made from Central America, Mexico, Jamaica, Cuba, and the United States ([@bibr08]; [@bibr16]; [@bibr09]; [@bibr07]; [@bibr26]). In the continental U.S., the range of *O. insidiosus* encompasses the areas east of the Rocky Mountains, north to Canada, and south to Florida. It is the most widespread species of *Orius* in the western hemisphere. The coexistence of *O. insidiosus* and *O. pumilio* in the United States has only been observed in Florida, where the northerly limit of *O. pumilio* appears to be Alachua County, Florida. The two species were found together during the spring on the flowers of two umbelliferous plants, false Queen Anne\'s lace (*Ammi majus* L.) and Queen Anne\'s lace (*Daucus carota* L.) (Apiales: Umbelliferae \[Apiaceae\]), where both pirate bugs apparently preyed on Florida flower thrips, *Frankliniella bispinosa* (Morgan) (Thysanoptera: Thripidae). Throughout the four weeks of study, the demographics for the two species differed markedly. The *O. insidiosus* population was heavily weighted toward males, with a sex ratio of 2.7 males: females ([@bibr26]), and the *O. insidiosus* outnumbered the *O. pumilio* population by 3.6-fold. These observations have led to questions about ecological niches, interspecific competition, and the potential for interspecific mating competition and sexual conflict between the two species.
Fundamental to understanding these interactions is a clear grasp of the taxonomic relationships between *O. insidiosus* and *O. pumilio*. Genital morphologies have conventionally served as key taxonomic characters for the Heteroptera in general and for the Anthocoridae in particular. The size, shape, and orientation of the copulatory tube in the female ([@bibr05]; [@bibr22]; [@bibr35]; [@bibr03]; [@bibr30]) and the shape of the paramere in the male ([@bibr23]; [@bibr09]; [@bibr22]) are used as diagnostic characters in identifying *Orius* species. In addition to genitalic morphologies, the appearance, coloration, and morphometrics of antennal segments, head, thorax, scent glands, and wings have all contributed to the taxonomic identity of genera and species (e.g. [@bibr09], [@bibr10]).
However, these conventional identifying characters do not assure certainty in discriminating interacting field populations of other anthocorids. For example, *Anthocoris antevolens* White (Heteroptera: Anthocoridae) has a broad range in North America. Comparison of specimens from various regions results in a high degree of morphological variability and uncertainty in differentiating them from the closely related species *Anthocoris musculus* (Say) ([@bibr12]; [@bibr13], [@bibr14], [@bibr15]). Because the morphologically different populations of *A. antevolens* are overlapping and sympatric with *A. musculus* populations, potential inter- and intraspecific matings among all these groups could occur. Indeed, laboratory trials showed that insemination can occur in certain heterospecific pairings ([@bibr13], [@bibr15]). Thus, the identity of these species was re-examined based on morphologies of various body parts and genitalia, and included examining the relatedness of mitochondrial DNA sequences. These investigators demonstrated that there was significant uncertainty in identifying any of the populations as a species especially when utilizing the current taxonomic keys.
With their widespread distributions, *Orius* species may also prove to be adapted and reproductively isolated as local and regional populations along with closely related species. In Japan, *Orius* species have been shown to be adapted latitudinally in their diapause characteristics perhaps reflecting their genetic relatedness and adaptations that have led to speciation ([@bibr18]; [@bibr19]; [@bibr29]). Whether other aspects of physiological and behavioral adaptation in anthocorids, both to changes in the physical environment and to pressures exerted by closely related species, result in divergence of populations and ultimately to speciation remains to be discovered.
In order to more reliably assess the relationship between these two species of *Orius*, the morphologies of male and female genitalia, the abilities to cross-mate and produce subsequent egg development, and comparisons of genomic sequence have been examined here. While both species had previously been reported to occur in Florida ([@bibr02]; [@bibr16]; [@bibr09]), only limited observations have been made for the two species in areas where they coexist; these observations included limited monitoring of populations over a period of weeks during the daylight hours in the spring of 2008 ([@bibr26]; unpublished data, 2009) in the flower heads of two species of Umbelliferae (*A. majus* and *D. carota*). A more detailed morphological and genetic description of the relationship between these two species in Florida may contribute to a discussion of interspecific convergence (or divergence) in genitalic morphology and gene sequences, and the potential significance of inter- and conspecific mating competition, sexual conflict, and competition for resources such as prey and pollen.
Materials and Methods {#sec2}
=====================
Insect colonies {#sec2a}
---------------
A colony of *O. insidiosus* was established from a field population collected in Alachua County, Florida in 2008 ([@bibr26]). A colony of *O. pumilio* was established in 2002 from insects collected in Bronson, Levy County, Florida, 18 miles southwest of the site from where the *O. insidiosus* population was collected ([@bibr26]). Subcolonies of 24 h egg collections from each species were set up from primary colonies, and insects were allowed to develop for 13--14 d before use.
Interspecific matings {#sec2b}
---------------------
From sub-colonies of each species, unsexed 5^th^ instar nymphs were individually isolated by species and allowed to molt to adults in microtiter plate wells. Adults were anaesthetized in place with CO2 0--2 d following the adult molt, sexed, and mating was initiated at day 0 in conspecific and heterospecific crosses. Each replicate consisted of 10 males and 10 females in a Petri dish (50 ×× 9-mm) covered with a nylonscreened (23-mm diameter, 0.2-mm sieve size) tight-fit lid, containing 0.2 g shredded parchment paper and 0.75 g of 5% sucrose Hydrocapsules (ARS Inc.). Each group was moved into a 0.6-L Mason jar for oviposition at day 7. Jars contained 1 green bean, 3.5 g buckwheat hulls, 1.0 g Hydrocapsules®®, and 0.1 g *Ephestia kuehniella* eggs. Every 2--3 days until day 21, eggs were counted, beans were replaced, and 0.15 ml *E. kuehniella* eggs were added. At day 18, surviving adults were counted and their species identities were noted. Oviposition was terminated at day 21. For yolk protein ELISA analysis, adult mating groups were set up and mated as above at day 0 and collected at day 6, survivors were counted, and 3 females were collected for each replicate and stored at -80°°C in a 2-ml microcentrifuge tube.
Scanning electron microscopy {#sec2c}
----------------------------
Terminal segments or isolated parameres were dissected from males anaesthetized with CO2 or preserved in 100% ethanol and then transferred into 20% KOH for 2 d at room temperature for clearing. Segments and parameres were transferred from KOH through H~2~O, 25%, 50%, and 100% ethanol in succession over \<1 hr, and stored in 100% ethanol. Segments or parameres were dried from ethanol and mounted with carbon adhesive tabs on aluminum stubs. Specimens were Au/Pd sputter-coated (Denton DeskII sputter coater, Denton Vacuum, [www.dentonvacuum.com](www.dentonvacuum.com)) and then examined with an S-4000 FE-SEM microscope (Hitachi, [www.hitachi-hta.com](www.hitachi-hta.com)). Digital images were acquired and analyzed with Quartz PCI v8 software (Quartz Imaging Corp., [www.qrtz.com](www.qrtz.com)). Ten specimens of each species were examined from multiple perspectives and magnifications.
Light microscopy {#sec2d}
----------------
The copulatory tubes of adult females were examined in seven specimens of *O. pumilio* and six specimens of *O. insidiosus* from the Gainesville laboratory cultures. In addition, the copulatory tubes of field collected adult females from 13 specimens of *O. insidiosus* collected from eight states in the eastern, central, and western U.S., and of two specimens of *O. pumilio* from southern Florida were examined. Examination of insects from outside of the Gainesville geographic area was done to confirm that copulatory tubes in the Gainesville specimens were representative. The insects from the Gainesville laboratory colonies were preserved in alcohol while field-collected specimens were either dried or freshly killed. The distal half of the abdomen was dissected from each female, soaked in a 10% KOH solution for approximately 4 hr (preserved specimens only) and then transferred to a drop of water on a microscope slide. Abdominal segment VII, bearing the copulatory tube, was removed with the use of microtools. The segment was positioned with the interior side up and covered with a glass slip. Water was drawn out with a paper tissue as needed to flatten the segment. The copulatory tube was photographed at 200x using a digital camera attached to the microscope, and measured using digital software.
DNA/PCR analyses {#sec2e}
----------------
The 18S ribosomal gene internal transcribed spacer 1 (ITS-1) has been used previously to assess the phylogenetic relationship among various *Onus* species including *O. insidiosus* and *Orius tristicolor* (White) from North America ([@bibr11]). Genomic DNA sequence data for ITS-1 were derived from specimens of *O. insidiosus* and *O. pumilio* collected from our laboratory cultures and from *O. tristicolor* collected from butterfly bush, *Buddleja* sp., growing in Wapato, Washington for comparison with previously published sequences from anthocorid species. Genomic DNAs of 50 pooled adult *O. insidiosus, O. pumilio*, or *O. tristicolor* were isolated using the Wizard Genomic DNA Isolation System (Promega, [www.promega.com](www.promega.com)) according to the manufacturer\'s protocol for animal tissues. Direct PCR for the ITS-1 and flanking regions were amplified from the genomic DNAs as template using *Taq* PCR kit (New England Biolabs, [www.neb.com](www.neb.com)) with the forward primer, 5′?-ACCGCCCGCGCTACTACCGAT -3′?, and reverse primer, 5′?-TGTTCATGTGTCCTGCAGTTCACA-3′? (Integrated DNA Technologies, [www.idtdna.com](www.idtdna.com)), as identified by Muraji et al. ([@bibr18]). The amplification program was 94°° C for 3 min with 40 cycles of 92°° C for 40 s, 58°° C for 40 s, and 72°° C for 40 followed by 72°° C for 4 min. The PCR products were cloned into pGEM-T Easy (Promega) and sequenced on contract using ABI 3130 DNA sequencing at Interdisciplinary Center for Biotechnology Research (University of Florida). A total of six sequences were examined for each species to establish nucleotide identities.
The nucleotide sequences for ITS-1 reported from anthocorid species were aligned using ClustalW as a component of Mac Vector 7.2.3 software (MacVector, Inc., [www.macvector.com](www.macvector.com)). The phylogenetic and molecular evolutionary analyses were conducted with PAUP\*\* v 4.0b10 ([@bibr32]) and Mega version 4 ([@bibr33]). Alignments were subjected to analysis with PAUP\*\* Maximum Parsimony and trees were established as rooted using *Cimex lectularius* Linnaeus as the outgroup with a 50% majority rule consensus and a bootstrap of 1000 replicates. The same sequence alignment was subjected to analysis with Mega4 Neighbor-Joining, Minimum Evolution and Maximum Parsimony-Close-Neighbor Interchange, and all phylograms were established as rooted using *C. lectularius* as the outgroup with a 50% majority rule consensus and a bootstrap of 1000 replicates.
######
Egg production and mortality in conspecific and heterospecific matings.
Yolk protein quantification {#sec2f}
---------------------------
A monoclonal antibody-based ELISA was used as described ([@bibr25]), with minor modification (Shapiro and Shirk in press), to quantify the amount of yolk protein in adult female *O. insidiosus* and *O. pumilio*.
Results {#sec3}
=======
Interspecific matings {#sec3a}
---------------------
The outcome of interspecies matings were examined to determine whether hybrid progeny could be produced. Conspecific crosses between *O. insidiosus* or *O. pumilio* adults resulted in large numbers of eggs (857 and 424 eggs, respectively; [Table 1](#t01){ref-type="table"}) and successful production of progeny as would be expected. When placed together in heterospecific crosses, the males of both species attempted mating with the females of the other species. In contrast to the conspecific crosses, heterospecific crosses between the two species resulted in few eggs (37 eggs for *O. pumilio* ♂? ×× *O. insidiosus* ♀?) or no eggs (*O. insidiosus* ♂? ×× *O. pumilio* ♀?). From the 37 eggs laid by the *O. pumilio* ♂? ♀? *O. insidiosus* ♀? cross only 3 F1 offspring were recovered, but they did not produce offspring themselves. During the course of the experiment, male mortality averaged 22% while female mortality was generally lower (18%).
Morphology of the male paramere {#sec3b}
-------------------------------
As in other Anthocoridae, the left parameres of *O. insidiosus* and *O. pumilio* are dorsally situated on the ninth abdominal segment ([Figure 1A, B](#f01){ref-type="fig"}). During copulation, the abdomen of the male of either species is extended around and beneath (ventral to) the female. Subsequently the paramere rotates, pivoting counterclockwise around the axis of its embedded root ([Figure 1C, E](#f01){ref-type="fig"}) and comes into contact ventrally with the copulatory tube of the female. Under lower magnifications, differences in the morphology of the excised parameres from *O. insidiosus* and *O. pumilio* are not discernable although parameres from *O. pumilio* appear more robust than those from *O. insidiosus* when viewed laterally (edgewise).
{#f01}
{#f02}
When viewed with the SEM, the images of the parameres ([Figures 1](#f01){ref-type="fig"}--[2](#f02){ref-type="fig"}) revealed structural features not readily visible at the lower magnification. In both species, the flagellum of the paramere was relatively short and somewhat blade- or spear-shaped ([Figures 1 C--F](#f01){ref-type="fig"}, [2 A--B](#f02){ref-type="fig"}). Morphological differences between the species were noted that were consistent with drawings of Herring ([@bibr09], Figs. 14, 15). However, caution must be exercised when comparing the parameres because differences in the angular view of the parameres can cause visual distortions of the structures. The cone of the *O. insidiosus* paramere ([Figure 2A](#f02){ref-type="fig"}) was shorter than that in *O. pumilio* ([Figure 2B](#f02){ref-type="fig"}), did not spiral in as extensive an arc, and was more spear-shaped compared to that of *O. pumilio*. The cone of *O. pumilio* had a more spatulate shape and spiraled further out of the plane ([Figure 2B](#f02){ref-type="fig"}). A groove (""canal"" in [@bibr23] or ""furrow"" in Pééricart 1972) was noticeable on the cone of the paramere, extending from at least the junction of the cone and flagellum towards the apical tip of the cone ([Figure 2A--D](#f02){ref-type="fig"}). According to Pééricart ([@bibr22]), the furrow extends to the ventral side of the cone (not shown). The flagellum in the paramere of *O. pumilio* ([Figure 1F](#f01){ref-type="fig"}) was also relatively longer than that of *O. insidiosus* ([Figure 1E](#f01){ref-type="fig"}).
In both species, there was a noticeable pliability in the dissected paramere at the junction of the cone and flagellum; i.e. the flagellum appears to be flexibly hinged at this junction (arrowheads in [Figures 1 C, D](#f01){ref-type="fig"}; [2 A, B](#f02){ref-type="fig"}). Both species also exhibited a pronounced furrow or groove in the flagellum ([Figure 2B, C](#f02){ref-type="fig"}), shown clearly extending along the length of the flagellum for *O. pumilio* ([Figure 2D](#f02){ref-type="fig"}). This flagellar groove has not previously been reported in any species of *Orius*. Finally, a single pit sensillum (arrows, [Figure 2E, F](#f02){ref-type="fig"}) was apparent on the outer surface close to the tip of the cone in both species. At the highest magnifications, the structure of the sensillum of both species revealed a peg within the pit ([Figure 2G, H](#f02){ref-type="fig"}). The function of this structure is not known.
Morphology of the female copulatory tube {#sec3c}
----------------------------------------
The copulatory tube in both species was observed as a short, sclerotized, and thickwalled cylinder slightly offset from the longitudinal midline of the sternite ([Figure 3A--C](#f03){ref-type="fig"}). For both species, the sperm pouch located at the terminal end of the copulatory tube was destroyed during the dissections and is not shown. A thin-walled apical section usually observed in the copulatory tube of females in this genus ([@bibr04], [@bibr22]) was not visible in either species. The copulatory tubes of the two species differed in length, orientation and presence of a basal sclerotized mound. The copulatory tube in *O. pumilio* was composed of a cylindrical tube about 25 µµm long (range in Gainesville specimens: 15--37 µµm; range in two south Florida specimens: 15--26 µµm) that was embedded in or continuous with a sclerotized basal mound ([Figure 3B](#f03){ref-type="fig"}). The width of the cylindrical tube is generally 1.3 to 2.0 times its length (range in all dissected specimens: 30--48 µµm). The tube was tilted toward the longitudinal midline of the abdomen, although in some specimens or dissections it could be nearly erect. The basal mound and cylindrical portion together had the appearance of a truncated cone ([Figure 3B](#f03){ref-type="fig"}). In *O. insidiosus*, the copulatory tube was about 35 µµm in length (range in Gainesville specimens: 31--48 µµm), and was approximately as wide as it was long ([Figure 3C](#f03){ref-type="fig"}). The copulatory tube was generally parallel to the longitudinal midline of the abdomen, but in some specimens or dissections was oriented slightly away from the midline. In this species, there was no sclerotized mound surrounding the base of the cylindrical tube. In both species, the general appearance of the copulatory tube in the Gainesville specimens was similar to appearance in specimens from outside of the Gainesville area.
{#f03}
ITS-1 conservation {#sec3d}
------------------
The ITS-1 PCR primers produced a product of approximately 600bp that included the ITS-1 and flanking rDNA sequences from genomic DNA of each of the 3 species tested, *O. insidiosus, O. pumilio*, and *O. tristicolor*. Alignment with previously published sequences for *O. insidiosus* and *O. tristicolor* ([@bibr11]) showed only 3 base pairs were different over the length of the ITS-1 between the *O. insidiosus* published sequence and the one produced here ([Figure 4](#f04){ref-type="fig"}). The sequence for *O. tristicolor* showed 14 bp different between the two sequences. Because the published sequences were derived from samples collected in Arizona ([@bibr11]), these differences may indicate regional divergence in the ITS-1 sequence for these two species. When compared with *O. insidiosus*, the ITS-1 sequence from *O. pumilio* had 472 of 521 (91%) shared base identities, while there were only 433 of 516 (84%) shared base identities with *O. tristicolor*. Similarly, *O. insidiosus* and *O. tristicolor* had 432 of 521 (83%) shared base identities.
To assess the relative phylogenetic relationships among the various *Orius* species from which ITS-1 sequences have been reported, molecular evolutionary analyses were conducted to resolve the alignment of the ITS-1 sequences for anthocorids ([@bibr11]) with those identified here using multiple approaches. Four different phylograms were generated utilizing different selective criteria, and consistently the *Orius* species found in Japan segregated as a clade with *O. minutus, O. sauteri*, and *O. strigicollis* grouped closely together as previously observed ([@bibr11]). The other major clade observed in all four phylograms included the three species from North and Central America, i.e. *O. insidiosus, O. pumilio*, and *O. tristicolor*, as a related group with *O. insidiosus* and *O. pumilio* the more closely related ([Figure 5](#f05){ref-type="fig"}).
Yolk protein quantification in interspecific matings {#sec3e}
----------------------------------------------------
Conspecific and heterospecific matings between *O. insidiosus* and *O. pumilio* were done to assess the physiological impact of potential interspecific interactions. When conspecific matings were maintained for 27 d prior to ELISA and dissections, no significant differences in yolk protein (22 µµg/female in *O. insidiosus vs*. 17 µµg/female in *O. pumilio*) or egg contents (10.2 eggs/female in *O. insidiosus* vs. 10.7 eggs/female in *O. pumilio*) were found ([Table 2](#t02){ref-type="table"}).
) between the sequences for a species designates a mismatched base with the published sequence, and the (\*\*) below the sequences designates a conserved base present in all aligned sequences. High quality figures are available online.](f04_01){#f04}
When heterospecific crosses were conducted for 6 d, the conspecific control matings resulted in approximately 25% more yolk protein per female in *O. insidiosus* than in *O. pumilio* (24.9 *vs*. 19.3 µµg/female, respectively) ([Table 3](#t03){ref-type="table"}). However, when the heterospecifically mated females were examined, the yolk protein content of the females of both species was at least 10-fold less than that of the conspecifically mated females (1.3 µµg/female for *O. pumilio* ×× *O. insidiosus* and 1.8 µµg/female for *O. insidiosus* ×× *O. pumilio*) ([Table 3](#t03){ref-type="table"}). These yolk protein contents were similar to those observed in fed, unmated *O. pumilio* females (Shapiro and Shirk, in press).
{#f05}
Discussion {#sec4}
==========
The closely-related minute pirate bugs, *O. insidiosus* and *O. pumilio*, both occur in Central America, Mexico, Jamaica, Cuba, and Florida ([@bibr08]; [@bibr16]; [@bibr09]; [@bibr26]) apparently in sympatry. Recently these two species have been found coexisting naturally in the flowers of Umbelliferae (Apiaceae) where they fed on an abundance of Florida flower thrips, *F. bispinosa* ([@bibr26]). These floral feeding stations evidently provided an environmental oasis where both species found plentiful food and an increased opportunity for mating. However, the skewed abundance of male *O. insidiosus* and the dominant numbers of *O. insidiosus* over *O. pumilio* at this site ([@bibr26]) suggest that the population dynamics of the two species are perturbed in this setting. Because these two species are similar in external appearance ([@bibr09]), we were interested in assessing whether these are true, reproductively-isolated species that have been shown to coexist in sympatry, or whether their sympatry could potentially lead to hybrid progeny or disrupted mating patterns as both species apparently compete for one ecological niche. However, the findings presented here confirm *O. insidiosus* and *O. pumilio* are clearly distinct species. This conclusion was based on detailed comparisons of male and female genital morphology, by examination of 18s rDNA ITS-1 DNA sequences, by the lack of significant egg production following interspecific mating, and by the inability to effect a change in female reproductive physiology (egg maturation) following interspecific mating ([Table 3](#t03){ref-type="table"}).
Genitalia typically serve as key taxonomic characters in the Heteroptera and are used extensively in the Anthocoridae to identify species as well as for developing higher level groupings of species ([@bibr06]; [@bibr22]; [@bibr35]; [@bibr03]). The size, shape, and orientation of the copulatory tube in the female ([@bibr05]; [@bibr22]; [@bibr35]; [@bibr03]; [@bibr30]) and the shape of the paramere in the male ([@bibr23]; [@bibr09]; [@bibr22]) are used as diagnostic characters in identifying *Orius* species. An illustration of the copulatory tube in *O. insidiosus* can be found in Silveira et al. ([@bibr30]), but one in Carayon ([@bibr06]) is incorrectly labeled as a copulatory tube from *O. insidiosus*. The micrographs of the copulatory tube from *O. pumilio* presented here are the first descriptions published for this species. The copulatory tube was somewhat longer in *O. insidiosus* than in *O. pumilio*, and the copulatory tube of *O. insidiosus* was oriented parallel to the abdominal midline while that of *O. pumilio* was slightly tilted toward the midline. The copulatory tube of *O. pumilio* also had a broad, sclerotized basal mound that was not present in *O. insidiosus*.
######
Yolk protein content of fed and mated females.
######
Yolk protein contents of females from conspecific and heterospecific matings of *Orius insidiosus* and *O. pumilio*.
Males are often identified to species by the morphology of the left paramere ([@bibr22], [@bibr35]). The paramere functions during copulation by apparently guiding the soft tissue of the phallus through the groove between flagellum and cone into the copulatory tube of the female ([@bibr22]). The right paramere is vestigial or nonexistent in the Anthocoridae ([@bibr24]). The parameres of *O. pumilio* and *O. insidiosus* are illustrated in Kelton ([@bibr16]) and Herring ([@bibr09]). However, these drawings fail to show substantial morphological detail. To acquire greater detail for comparing the parameres in *O. insidiosus* and *O. pumilio*, specimens were examined with scanning electron microscopy. Differences in the shapes of the parameres were especially distinct at these higher magnifications. The less prominent, spear-shaped cone in *O. insidiosus* contrasted with the spatulate, elongated cone in *O. pumilio*. In both of these species, the flagellum was reduced in length and curvature from those described for other *Orius* species from North America ([@bibr09]), although that of *O. pumilio* was longer than the flagellum of *O. insidiosus*. A novel unanticipated structural feature of the parameres in these two species was the presence of a single sensillum distinctly visible on the outer curvature of the cone of the parameres. The function of the sensillum was not examined here.
Several molecular markers have previously been used to establish relatedness among species or geographic variants of anthocorids, among them sequences of the 18s rDNA internal transcribed spacer 1 (ITS-1). Previously, ITS-1 was used to examine the relationships between *Orius* species found in Japan ([@bibr11]; [@bibr18]). The ITS-1 sequences of *O. insidiosus* and *O. pumilio* shared only 91% homology while there was only 1bp difference between the ITS-1 sequence of *O. insidiosus* from Florida (present results) and those previously reported for specimens originating in Arizona ([@bibr11]). Neither *O. insidiosus* nor *O. pumilio* ITS-1 sequences shared more than 84% homology with the sequence of *O. tristicolor*, the other major North American species. Regardless of the method used to establish the phylograms, the phylogenetic analyses based on the ITS-1 sequences for the anthocorids consistently associated the three *Orius* species from North America in one clade and the major group of *Orius* species from Japan in another clade. The grouping of the *Orius* species from Japan is consistent with previous reports ([@bibr11]). These comparisons further support the taxonomic groupings based on morphological features and phylogenetically place *O. insidiosus* and *O. pumilio* as closely related species.
The interspecific differences in shapes of parameres and copulatory tubes described above may not provide mechanical fits adequate for interspecific matings. To test whether successful interspecific mating is possible and can result in hybrid progeny, *O. insidiosus* and *O. pumilio* were cross-mated. Although no eggs were produced when *O. pumilio* females were paired with *O. insidiosus* males, a few eggs were laid by *O. insidiosus* females paired with *O. pumilio* males. However, only three of these eggs hatched and produced nymphs that reached the adult stage. These studies substantiated the reproductive isolation between these two species, but do raise the possibility that some interspecific mating and egg production may occur in nature, although the potential for viable hybrid progeny seems minimal. Furthermore, even if copulation and sperm transfer were successful in interspecific matings, those matings did not result in increased yolk protein synthesis and accumulation much greater than the levels observed in fed-but-unmated females (Shapiro and Shirk, in press). Although a few oviposited eggs were obtained from the *O. pumilio* ×× *O. insidiosus* matings, the small number of eggs could have been oviposited by very few females. Adult female *O. pumilio* require both feeding and mating to stimulate the female to achieve maximum yolk protein production and develop eggs; if the female is only fed, but left unmated, a low level of yolk protein is produced yet no eggs develop (Shapiro and Shirk, in press). The level of yolk protein observed in the females of both species from the interspecific matings was consistent with fed, but unmated females in the interspecific matings. This suggests that either the interspecific males were unsuccessful in copulation and sperm transfer and thus ineffective at mating, or that the act of copulation and the transfer of sperm were not sufficient to stimulate a switch in reproductive physiology.
Horton et al. ([@bibr13]) examined mating behavior and success between three populations of *Anthocoris antevolens* White (Hemiptera: Heteroptera: Anthocoridae), two of which were sympatric. As with the present species of *Orius*, males of *A. antevolens* had variable success in mating with females from other populations. While one cross between individuals of the two sympatric populations showed a low frequency of success in insemination, the reciprocal cross showed no success. In the case of *A. antevolens*, the reproductive isolation was interpreted as evidence that the species actually comprises a complex of an unknown number of cryptic species ([@bibr13], [@bibr15]). These findings punctuated the historical uncertainty concerning taxonomic relationships between *A. musculus* (Say) and *A. antevolens*, particularly highlighting their status as separate species ([@bibr15]). Perhaps this quandary could be extended to *Orius* as well, for example in questioning whether the broad distribution of *O. insidiosus* might in fact result in broad diversification of cryptic species. Diapause is one example of a character that exhibits such diversity within the genus *Orius* and within species, and latitudinal changes commonly result in corresponding change in the inducibility of diapause ([@bibr18]; [@bibr19]; [@bibr29]). Whether other aspects of physiological and behavioral adaptation in anthocorids, both to changes in the physical environment and to pressures exerted by closely related species, result in divergence of populations and speciation remains to be discovered.
These morphological, molecular, genetic, and ecological observations support a close phylogenetic relationship between the two species, and leave room for speculation regarding species divergence. It was concluded that these two species are more closely related to each other than either one is to *O. tristicolor*, or to Old World species. As often noted, accelerated divergence and speciation may be expected in sympatric species when sexual conflict occurs, especially between males of one species and females of the other species ([@bibr01], [@bibr21]). Further behavioral studies may clarify such interactions in these *Orius* species, and a study on the effects of interspecific sex ratios on reproduction is ongoing.
The authors wish to thank Dr. Jean M. G. Thomas and Richard B. Furlong for excellent technical assistance as well as Gary Parsons (A. J. Cook Arthropod Collection, Michigan State University, East Lansing) for loaning specimens of *O. pumilio*. Nucleotide sequences have been submitted to GenBank, and the accession numbers are GU214726 (*Orius insidiosus*), GU214727 (*Orius tristicolor*), and GU214728 (*Orius pumilio*). The use of trade, firm, or corporation names in this publication is for the information and convenience of the reader. Such use does not constitute an official endorsement or approval by the United States Department of Agriculture or the Agricultural Research Service of any product or service to the exclusion of others that may be suitable.
**ELISA**,
: Enzyme-linked immunoserological assay;
**ITS-I**,
: Internal transcribed spacer I;
**PCR**,
: polymerase chain reaction;
**rDNA**,
: ribosomal DNA;
**SEM**,
: scanning electron microscopy
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#sec1}
============
Invasive cytomegalovirus (CMV) disease is a well-recognized complication following solid organ transplant; however CMV is infrequently reported to cause significant small bowel pathology. Most commonly CMV manifests as a viral syndrome with fever and neutropenia, with common gastrointestinal manifestations including esophagitis, gastritis, colitis, or hepatitis \[[@ref1]\]. Further, the reported cases and series of CMV causing gastrointestinal stricturing are often reported in pediatric patient groups, generally with birth prematurity \[[@ref2], [@ref3]\]. Adult cases of CMV-associated stricture are unusual and have heretofore been reported only in those with acquired immune deficiency syndrome and severe immunodeficiency, but these reports highlight esophageal strictures due to CMV \[[@ref4], [@ref5]\]. To our knowledge, there are no reported instances of CMV-associated small intestinal stricture in a relatively immunocompetent adult.
CASE REPORT {#sec2}
===========
Our patient is a 69-year-old gentleman who underwent orthotopic liver transplant in April of 2010 for cryptogenic cirrhosis. His maintenance therapy consisted of rapamune due to chronic kidney disease but was transitioned to mycophenolate mofetil (MMF) monotherapy in 2014 and continued to have normal allograft function. In the spring of 2018, he developed CMV viremia; immunosuppression was held, and CMV therapy was initiated with valganciclovir. Within a month he had cleared the CMV viremia and was restarted on MMF for immunosuppression; however he proceeded to be readmitted to the hospital service approximately six times over the following 2 months for intolerance of solid foods. He was ultimately taken to surgery for lysis of adhesions, where a strictured segment of the ileum was identified 20 cm from the ileocecal value. Six centimeters of the small intestine were resected, and primary bowel anastomosis was performed. The mucosal surface was remarkable for a centrally located area of stricture with 60% luminal narrowing. Pathology of the resected segment showed no ischemic changes but ulcerations of the mucosa causing the stricture, with immunostaining positive for invasive CMV disease. The patient has subsequently had no recurrence of or readmittance for small bowel obstruction or food intolerance.
DISCUSSION {#sec3}
==========
CMV disease is a well-reported posttransplant infection, most frequently causing allograft hepatitis and infrequently causing biliary strictures \[[@ref6], [@ref7]\]. Stricturing manifestations of intestinal CMV disease more commonly present in term or preterm neonates \[[@ref2]\]. Series reporting episodes of CMV enteritis demonstrate about a 40% incidence of small intestinal disease, with the esophagus and colon compromising an equal percent of cases \[[@ref8]\]. To our knowledge, there is one report of CMV enteritis an adolescent liver transplant recipient, who presented with duodenal bleeding requiring pancreas preserving duodenectomy but no stricture \[[@ref9]\]. The reports of intestinal CMV infections in children emphasize that all patients had significant causes for immunodeficiency, such as HIV, prematurity, or age less than 6 months. There appears to be no other report in the literature of *adult* posttransplant patients suffering from CMV stricturing small bowel enteritis.
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#s1}
============
Breast cancer is the most frequently diagnosed cancer and the leading cause of cancer death among females worldwide \[[@R1]\]. Accumulating publications have reported that genetic factors play important roles in the pathogenesis of both sporadic and familial breast cancer \[[@R2]--[@R6]\]. Recent genome-wide association studies (GWASs) focusing on evaluating common single nucleotide polymorphisms (SNPs) have identified more than 70 genetic susceptibility loci for breast cancer \[[@R6]--[@R12]\]. However, these newly identified genetic factors, along with known high-penetrance breast cancer susceptibility genes, only explain a small portion of the heritability for this cancer \[[@R9]\].
The discovery of submicroscopic copy number variations (CNVs) present in our genomes has changed dramatically our perspective on DNA structural variation and disease. It is now thought that CNVs encompass more total nucleotides and arise more frequently than SNPs. CNVs may account for 13% of the human genome and have been supposed to explain some of the missing heritability for complex diseases after the findings from GWASs \[[@R13]--[@R16]\].
Recently a deletion in the APOBEC3 gene cluster was identified \[[@R17], [@R18]\]. This CNV is a deletion located between exon 5 of APOBEC3A and exon 8 of APOBEC3B, resulting in a fusion gene with a protein sequence identical to APOBEC3A, but with a 3'-UTR of APOBEC3B. It has been found that the deletion frequency was highly variable, rare in African and European populations (frequency of 0.9% and 6%, respectively), more common in East Asian and American populations (36.9% and 57.7%), and almost fixed in Oceanic populations (92.9%) \[[@R17]\]. Komatsu et al. first discovered an increased but statistically non-significant risk of breast cancer associated with APOBEC3 deletion in Japanese \[[@R19]\]. Then Long\'s study strongly indicated a positive correlation of APOBEC3 deletion with an elevated breast cancer risk in Chinese \[[@R20]\]. Later, Xuan\'s study in European and Rezaei\'s study in Iranian both showed the similar positive correlation with statistical power \[[@R21], [@R22]\]. However, the latest study by Göhler only revealed statistically non-significant results in Sweden and Marouf\'s study in Moroccans even indicated contrary results \[[@R23], [@R24]\]. As mentioned above, the results derived from current publications are inconclusive and even conflicting to each other, therefore we believe that it\'s necessary to perform a meta-analysis of available patient data to gain greater statistical power on this issue, with an expectation to obtain a pooled estimate much closer to the unknown truth and consequently to provide useful evidences and suggestions for early breast cancer screening and future investigation.
RESULTS {#s2}
=======
Study identification {#s2_1}
--------------------
For the process of eligible studies' identification and selection, 16 publications were initially retrieved from PubMed and Embase databases. After further screening according to the inclusion and exclusion criteria, 6 relevant reviews and 4 molecular mechanism research studies were excluded. Finally, a total of 6 case-control studies with 18241 subjects were identified to be eligible for this meta-analysis \[[@R19]--[@R24]\]. The main characteristics of included studies were summarized in Table [1](#T1){ref-type="table"}.
###### Characteristics of studies on association between APOBEC3 gene deletion and breast cancer
Author Year Area Ethnicity Sample size Source of Controls Cases Controls Genotyping Method *P* for HWE in controls Quality Score
--------- ------ --------- ----------- ------------- -------------------- ------- ---------- ------------------- ------------------------- --------------- ------ ----- ------------------------------------ ------ ---
Marouf 2016 Morocco Caucasian 226 200 PB 207 19 0 175 25 0 Real-time qualitative PCR (Taqman) 0.35 8
Göhler 2016 Sweden Caucasian 782 1559 PB 633 142 5 1295 251 13 KASPar or Life Technologies assays 0.83 8
Rezaei 2015 Iran Caucasian 262 217 PB 154 103 5 148 63 6 Real-time qualitative PCR 0.82 9
Xuan 2013 Europe Caucasian 1671 1602 PB 1275 376 20 1279 314 9 Real-time qualitative PCR 0.03 8
Long 2013 China Asian 5792 5830 PB 2045 2805 942 2530 2638 662 Real-time qualitative PCR 0.52 9
Komatsu 2008 Japan Asian 50 50 PB 22 21 7 21 23 2 Real-time qualitative PCR 0.16 8
PB, population-based; PCR, polymerase chain reaction; HWE, Hardy-Weinberg equilibrium.
Quantitative data analyses {#s2_2}
--------------------------
Finally, 6 epidemiological individual studies including 8783 cases and 9458 controls were enrolled in this meta-analysis. All studies reported detailed number of three genotypes Insertion/Insertion (I/I), Insertion/Deletion (I/D), Deletion/Deletion (D/D). To be comprehensive, the correlation of APOBEC3 CNVs with breast cancer risk was analyzed by five different comparison models: allele contrast, dominant, recessive, homozygous, and heterozygous.
Firstly, in the allele contrast model, the APOBEC3 deletion variation was found to be significantly correlated with a higher breast cancer risk compared with the no-deletion allele (D *vs* I: OR = 1.29, 95% CI = 1.23-1.36) (Figure [1A](#F1){ref-type="fig"}). For the ethnicity-specific subgroup analysis, the results indicated that both Asians and Caucasians with the APOBEC3 deletion allele possessed an increased breast cancer susceptibility. Summary of the ORs and 95% CIs for breast cancer risk and APOBEC3 deletion under different genetic models was shown in Table [2](#T2){ref-type="table"}.
{#F1}
###### ORs and 95% CI for breast cancer risk and APOBEC3 deletion under different genetic models
Genetic models n OR \[95% CI\] *P*~(OR)~ Model (method) *I*-square (%) *P* ~(H)~ *P* ~(Begg)~ *P* ~(Egger)~
------------------------------------ --- ---------------------- ----------- ---------------- ---------------- ----------- -------------- ---------------
Allele contrast (D *vs* I)
All 6 1.29 \[1.23 - 1.36\] \< 0.001 F (M-H) 36.4 0.164 0.260 0.115
Caucasian 4 1.18 \[1.05 - 1.32\] 0.005 F (M-H) 34.3 0.207 \- \-
Asian 2 1.32 \[1.25 - 1.39\] \< 0.001 F (M-H) 0.0 0.951 \- \-
Dominant model (D/D+I/D *vs* I/I)
All 6 1.34 \[1.26 - 1.43\] \< 0.001 F (M-H) 52.3 0.063 0.452 0.119
Caucasian 4 1.20 \[1.06 - 1.35\] 0.004 F (M-H) 45.0 0.141 \- \-
Asian 2 1.40 \[1.30 - 1.51\] \< 0.001 F (M-H) 0.0 0.508 \- \-
Recessive model (D/D *vs* I/D+I/I)
All 5 1.51 \[1.36 - 1.68\] \< 0.001 F (M-H) 22.5 0.271 1.000 0.809
Caucasian 3 1.26 \[0.74 - 2.15\] 0.389 F (M-H) 44.7 0.164 \- \-
Asian 2 1.52 \[1.37 - 1.69\] \< 0.001 F (M-H) 6.4 0.301 \- \-
Homozygous model (D/D *vs* I/I)
All 5 1.75 \[1.56 - 1.95\] \< 0.001 F (M-H) 17.2 0.305 1.000 0.591
Caucasian 3 1.34 \[0.79 - 2.29\] 0.280 F (M-H) 39.5 0.192 \- \-
Asian 2 1.77 \[1.57 - 1.98\] \< 0.001 F (M-H) 0.0 0.456 \- \-
Heterozygous model (I/D *vs* I/I)
All 6 1.28 \[1.19 - 1.36\] \< 0.001 F (M-H) 39.4 0.143 0.133 0.194
Caucasian 4 1.19 \[1.05 - 1.35\] 0.006 F (M-H) 47.7 0.125 \- \-
Asian 2 1.31 \[1.21 - 1.42\] \< 0.001 F (M-H) 0.0 0.340 \- \-
CI, confidence intervals; I, insertion; D, deletion; *P* ~(OR)~, *P* for odds ratios; *P* ~(H)~, *P* for heterogeneity; n, number of included studies; F, fixed-effect model; M-H, Mantel-Haenszel method.
The positive association between APOBEC3 deletion and breast cancer susceptibility was also found by the dominant model in both Asians and Caucasians (D/D+I/D *vs* I/I: for total, OR = 1.34, 95% CI = 1.26-1.43; for Asians, OR = 1.40, 95% CI = 1.30-1.51; for Caucasians, OR = 1.20, 95% CI = 1.06-1.35) (Figure [1B](#F1){ref-type="fig"}). Similarly, evidences from the heterozygous analysis also support a prominent association between APOBEC3 deletion and breast cancer risk, and no statistically significant ethnic variation was observed (I/D *vs* I/I: for total, OR = 1.28, 95% CI = 1.19-1.36; for Asians, OR = 1.31, 95% CI = 1.21-1.42; for Caucasians, OR = 1.19, 95% CI = 1.05-1.35) (Figure [1E](#F1){ref-type="fig"}).
As Marouf\'s study \[[@R24]\] included no individuals with D/D genotype, it was excluded in the comparison by recessive model (D/D *vs* I/D+I/I) and homozygous model (D/D *vs* I/I). After the exclusion, positive results were only found in Asians but not in Caucasians (D/D *vs* I/D+I/I: for total, OR = 1.51, 95% CI = 1.36-1.68,; for Asians, OR = 1.52, 95% CI = 1.37-1.69; for Caucasians: OR = 1.26, 95% CI = 0.74-2.15) (D/D *vs* I/I: OR = 1.75, 95% CI = 1.56-1.95; for Asians, OR = 1.77, 95% CI = 1.57-1.98; for Caucasians: OR = 1.34, 95% CI = 0.79-2.29) (Figure [1C and 1D](#F1){ref-type="fig"}).
Sensitivity analysis {#s2_3}
--------------------
Sensitivity analysis was conducted by deleting one study at a time to examine the influence of individual study to the pooled ORs. No single study materially altered the pooled ORs when they were sequentially deleted, suggesting that our results were stable and robust (data not shown).
Publication bias {#s2_4}
----------------
The funnel plot revealed no obvious publication bias with a symmetrical distribution of study results around the pooled measurement of effect, indicating that publication bias was generally not a factor influencing the results (Figure [2](#F2){ref-type="fig"}). Egger\'s test and the Begg\'s test were employed to detect the potential publication bias, and also confirmed no statistically significant publication bias in this meta-analysis. All *P*-values from the Egger\'s test and the Begg\'s test were listed in Table [2](#T2){ref-type="table"}.
{#F2}
DISCUSSION {#s3}
==========
The APOBEC3 gene family, including APOBEC3A, APOBEC3B, APOBEC3C, APOBEC3D, APOBEC3E, APOBEC3F, APOBEC3G, and APOBEC3H, plays pivotal roles in intracellular defense against viral infections \[[@R25], [@R26]\]. The APOBEC3 genes family encodes cytosine deaminases that have been demonstrated to play key roles in innate immunity through their ability to mutagenize viral DNA and restrict viral replication \[[@R27], [@R28]\]. Recent advances in cancer genomics, together with biochemical characterization of the APOBEC3 enzymes, have implicated that APOBEC3 germline variations are specifically involved in many human cancers, including breast, bladder, lung (adenocarcinoma and squamous cell carcinoma), head/neck, and cervix \[[@R29], [@R30]\]. Moreover, the APOBEC3 gene may play a role in carcinogenesis by triggering DNA mutation \[[@R31]\]. Stephen Henderson and Tim Fenton have reviewed the evidences linking these enzymes to carcinogenesis including the potential mechanisms that misdirect APOBEC3 activity to the host genome, the links to viral infection, and the association between a common APOBEC3 polymorphism and cancer risk \[[@R27]\].
Up to now, emerging studies support a positive association between APOBEC3 CNVs and cancer risk \[[@R32], [@R33]\], particularly for breast cancer \[[@R19]--[@R24]\]. What\'s more, it\'s reported that the expression of APOBEC3 genes is regulated by estrogen (ER) \[[@R19]\], a hormone that plays a central role in the etiology of breast cancer. And Burns et al. provided evidences that APOBEC3B is overexpressed in breast tumors and cell lines and that the APOBEC3B mutation signature is statistically more prevalent in the breast tumor database of The Cancer Genome Atlas (TCGA) than is expected \[[@R34]\]. Also, APOBEC3B has recently been verified to be a marker of pure prognosis and poor outcomes for ER^+^ breast cancer, which strongly suggests that genetic aberrations induced by APOBEC3B contribute to breast cancer progression \[[@R35]\]. Additionally, a recent re-analysis of public breast cancer mutation data reported that breast cancers in carriers of the deletion show more mutations of the putative APOBEC-dependent genome-wide signatures than cancers in non-carriers. Their results suggested that the APOBEC3A/3B germline deletion allele confers cancer susceptibility through increased activity of APOBEC-dependent mutational processes, although the mechanism by which this occurs remains unknown \[[@R36]\].
Herein we performed a comprehensive databases search for all the eligible studies that reported the association between APOBEC3 CNVs and breast cancer risk. After pooling all the available data, we got a conclusion that the APOBEC3 deletion mutation significantly increases the risk of breast cancer, and the results are especially stable in Asians. OR (95% CI) of 1.29 (1.23-1.36) associated with deletion allele compared with no deletion allele strongly suggested a positive relationship between APOBEC3 deletion and breast cancer risk. And generally, consistent results were observed under the other four genetic models. When assessing by the homozygous and hetarozygous model, APOBEC3 deletion was also significantly associated with breast cancer risk, with OR (95% CI) of 1.28 (1.19-1.36) associated with one-copy deletion and 1.75 (1.56-1.95) associated with two-copy deletion compared with subjects with no deletion. The results present a possibility that the deletion copy number may be directly proportional to the breast cancer susceptibility.
In particular consideration of the influence of ethnicity, we conducted stratified analysis by ethnicity. Notably, the results showed that the relationship is weaker in Caucasians than in Asians. The ORs in Caucasian subgroup were always lower than those in Asian subgroup, and results in recessive model and homozygous model analysis revealed no statistical significance. Although analysis by other three genetic models showed positive results, we cannot convincingly conclude that APOBEC3 deletion would confer risk for breast cancer in Caucasian populations from the current assessment. More studies with large sample size in Caucasian populations are warranted to verify our results.
Our work has several strengths. Above all, this is by far the first meta-analysis evaluating the association between APOBEC3 CNVs and breast cancer risk. And it was comprehensively analyzed by five different comparison models. Besides, all the included studies were of high quality with NOS scores ranged from 8-9 score, which is critical for the reliability of our pooled results. What\'s more, no obvious heterogeneity was observed in all the analysis by five models and the ethnicity subgroup analysis further reduced the heterogeneity. Also, sensitivity analysis indicated no study to be deleted and no evident publication bias was detected. All these points significantly increased the statistical power of our analysis.
Despite its strengths, some limitations should be taken into consideration. Firstly, three of the included studies were based on a relatively small sample of less than 500 subjects and Long\'s study with 11622 subjects weights much higher than others. Secondly, only literatures in English were included in our study. Finally, many other factors could influence our analysis, such as distinct covariant factors, various genotyping methods among studies and non-coincident baseline characteristics of different samples. All of these potential discrepancies interfere with the standardization of our pooled data, but we were unable to conduct further stratified analyses for lack of detailed information. Thus, further studies investigating this issue should consider the factors mentioned above.
In summary, our current work indicates that a high copy number of APOBEC3 deletion confers risk for breast cancer, suggesting a possibility that APOBEC3 CNVs has a good screening accuracy for breast cancer.
MATERIALS AND METHODS {#s4}
=====================
Search strategy {#s4_1}
---------------
Potentially eligible articles were obtained through searching PubMed and Embase databases up to April 2016. The literature search was performed using free-text words combined with Medical Subject Headings (MeSH). Gene-specific terms (APOBEC3 or apolipoprotein B mRNA editing enzyme, catalytic polypeptide 3) were combined with polymorphism-specific terms (polymorphism or polymorphisms or variation or variations or variant or mutation or mutations or genotype or genotypes) and disease-specific terms (breast cancer or breast cancers) to retrieve eligible studies. References from retrieved articles were further screened manually for other potentially available reports.
Inclusion and exclusion criteria {#s4_2}
--------------------------------
Inclusion criteria for studies were the following: (1) case-control or cohort study design; (2) evaluating associations between APOBEC3 gene deletion and breast cancer risk; (3) providing OR estimates with 95% CIs or sufficient data for calculation; (4) published in English; and (5) performed on humans. Exclusion criteria were the following: (1) reviews and comments; (2) insuffient data for calculation; (3) performed on animals; and (4) duplication of a previous publication.
Study selection and data extraction {#s4_3}
-----------------------------------
Data extraction was performed independently by two investigators (Y. Han and Q. Qi) and disagreements were adjudicated by a third reviewer (Q. He). For each study, general characteristics such as the first author, publication year, country and ethnicity of patients, sample size, and genotyping method were collected.
Quality assessment {#s4_4}
------------------
The Newcastle-Ottawa Scale and Agency for Healthcare Research and Quality (<http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp>; maximum score = 9 points) was used to evaluate the methodological quality, which scored studies based on the selection of patients, the comparability of the groups, and the quality of the sampling process. A study awarded a score of 0-3, 4-6, or 7-9 was considered as a low-, moderate-, or high-quality study, respectively. Two authors (Y. Han and Q. Qi) independently assessed the study quality, and inconsistency was discussed with another reviewer-author (Q. He), who acted as an arbiter.
Statistical analysis {#s4_5}
--------------------
To obtain a more comprehensive assessment of associations between APOBEC3 deletion and breast cancer risk, five different comparison models were used: allele contrast, dominant, recessive, homozygous and heterozygous. Risk estimates were expressed as ORs and 95% CIs. Heterogeneity arising from pooled individual studies was examined by the *I*^2^ test and Q test. Values of *P* \> 0.10 for the Q test or *I*^2^ \< 50% was considered lack of heterogeneity and a fixed-effect model was used; otherwise, a random-effect model was used. Subgroup analysis by ethnicity was used to detect and reduce potential source of heterogeneity among studies. Sensitivity analysis was conducted by excluding one study at a time. We depicted the Begg\'s funnel plot and computed the Egger regression asymmetry test to assess the probability of publication bias. Values of *P* \< 0.05 were indicative of statistically significant publication bias. Data analyses were carried out using Stata software, version 11.0 (Stata Corporation; College Station, TX, USA). All *P* values were two sided and *P*-values \< 0.05 were considered statistically significant.
This work was supported by the National Natural Science Foundation of China (81372334).
**CONFLICTS OF INTEREST**
The authors declare no conflicts of interest.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1}
===============
Amyloid aggregates formed by misfolded intrinsically disordered proteins are implicated in many diseases \[[@B1]\]. Here, we focus on human islet amyloid polypeptides (hIAPPs) that aggregate into parallel *β*-sheets upon the interaction with membrane surfaces \[[@B2]--[@B4]\]. The resulting aggregates are detrimental to pancreas *β*-cells, leading to the onset of type II diabetes \[[@B5]--[@B7]\], since they disrupt the membrane integrity \[[@B8]--[@B13]\], increasing the membrane permeability to water and ions \[[@B14]\]. Understanding surface-specific biomolecular interactions between hIAPP aggregates and lipid membranes at the molecular level is therefore crucial for revealing the molecular factors controlling amyloidogenesis.
Previous studies have relied mainly on bulk detection techniques, including circular dichroism (CD) \[[@B15]\], NMR \[[@B16]--[@B18]\], EPR \[[@B19], [@B20]\], 2D-IR \[[@B21]--[@B23]\], AIR-FTIR \[[@B24]--[@B26]\], fluorescence spectroscopy \[[@B27], [@B28]\], Raman spectroscopy \[[@B29]\], and infrared reflection absorption spectroscopy (IRRAS) \[[@B30], [@B31]\]. However, many specific questions concerning the interfacial region of interactions between membranes and proteins remain unclear. Some of the outstanding questions are as follows: Do the amyloid aggregates start forming at membrane surfaces, or in the bulk solution? How do the aggregates orient on membrane surfaces? Are the aggregation products parallel or antiparallel *β*-sheets? Does the orientation of the aggregate affect the integrity of cell membranes? What is the kinetics of misfolding at membrane surfaces? Is it different from misfolding in the bulk solution? Do molecular inhibitors affect aggregation on membrane surfaces? Recent developments in nonlinear sum frequency generation (SFG) vibrational spectroscopy have demonstrated SFG as an intrinsically surface-selective technique with submonolayer sensitivity and label-free detection capability, thus showing great promise to address the above questions, shedding light on the role of the membrane during the aggregation of hIAPP and other amyloid proteins \[[@B32], [@B33]\].
During the last three decades, we have witnessed the emergence and fast development of nonlinear optical spectroscopic techniques, among which SFG vibrational spectroscopy has gained tremendous attention. Unique advantages of SFG include surface selectivity, submonolayer sensitivity, chiral-selectivity, phase-sensitivity, and label-free detection capabilities \[[@B34]--[@B36]\], which make SFG a promising tool for structural characterization of interfaces, including a wide range of applications in material science \[[@B37]\], characterization of polymers \[[@B38], [@B39]\], and catalytic systems \[[@B40], [@B41]\]. Recently, applications have been extended to environmental \[[@B42]--[@B44]\] and biological systems \[[@B32], [@B33], [@B45]--[@B52]\] with unprecedented discoveries beyond the capabilities of conventional tools. For example, SFG has been used to study a variety of atmospherically significant systems at the vapor/aqueous interface to elucidate the organization and reactions in aerosols that contain inorganic/organic compounds \[[@B42]--[@B44]\], small molecules \[[@B42]\], and fatty acids \[[@B42]\] as solutes. Furthermore, SFG has been applied to probe the interfacial structures, orientation, and kinetics of biologically relevant molecules at interfaces, such as proteins \[[@B32], [@B33], [@B45]--[@B47]\], DNAs \[[@B48], [@B49]\], and lipids \[[@B45], [@B50]--[@B52]\], providing insights into the functions of these molecules at biological interfaces and facilitating further research into biomedical science and engineering. Nowadays, SFG is established as a valuable technique to understand physical, chemical, and biological processes at the molecular level \[[@B53]\]. In particular, applications to interfacial biological systems span across important research topics in membrane biophysics \[[@B54]\], surface self-assembly \[[@B55], [@B56]\], peptides at cell surfaces \[[@B57], [@B58]\], DNA hybridization \[[@B48], [@B59]\] and adsorption \[[@B60]\], and amyloid interactions with membrane surfaces \[[@B61]--[@B63]\].
This review focuses on the recent development and application of SFG for probing hIAPP interacting with lipid membranes and discusses the implications of hIAPP/membrane interactions in the studies of type II diabetes \[[@B14], [@B61], [@B62], [@B64], [@B65]\]. The review also summarizes the basic theoretical background and experimental methods of SFG, supplemented with a brief discussion about the application of SFG to other amyloidogenic proteins, and concludes with an outlook of SFG in applications to systems of interest in biological and medical sciences.
2. The SFG Method {#sec2}
=================
2.1. Basic Principles of SFG {#sec2.1}
----------------------------
Sum frequency generation (SFG) vibrational spectroscopy applies two laser beams that interact with the system of interest simultaneously \[[@B34]--[@B36], [@B66]--[@B69]\], one with visible (e.g., 532 nm) or near-infrared (e.g., 800 nm) frequency *ω* ~vis~ and the other with infrared (IR) frequency *ω* ~IR~ ([Scheme 1](#sch1){ref-type="fig"}). When the visible and infrared pulses overlap in time and space, light with the sum of the two frequencies *ω* ~SFG~ = *ω* ~vis~ + *ω* ~IR~ is generated by the interaction of the incident beams with the system at the interface. Fixing *ω* ~vis~ and scanning or dispersing IR frequencies *ω* ~IR~ over a range, the SFG spectra are recorded by monitoring the SFG intensity as a function of *ω* ~SFG~ with a monochromator and a CCD. The SFG signal is dramatically enhanced when *ω* ~IR~ is in resonance with a vibrational frequency of a molecule at an interface, thus showing a peak in the spectrum. The SFG peaks typically exhibit homogeneous and inhomogeneous broadening, leading to various line-shapes due to the constructive or destructive interference with neighboring bands modulated by changes in the molecular orientation as induced by interactions with the surface and the surrounding environment \[[@B70]\]. Thus, the SFG spectra contain detailed structural information reported in terms of line-shape, peak position, and polarization dependence. Extracting that structural information from the SFG spectra, however, requires rigorous theoretical modeling and computer simulations. Hence, the combination of SFG and computational modeling can be used as a label-free and*in situ* analytical methodology for effective characterization of systems at interfaces. In the following sections, we will illustrate the applications of SFG to the studies of IAPP at membrane surfaces \[[@B14], [@B61], [@B62], [@B64], [@B65]\].
2.2. Surface-Specificity, Monolayer Sensitivity, and Polarization Dependence of SFG Spectroscopy {#sec2.2}
------------------------------------------------------------------------------------------------
As a nonlinear optical technique, SFG measures the second-order susceptibility, *χ* ^(2)^, that gives intrinsic surface selectivity \[[@B33], [@B34], [@B71]--[@B74]\]. The second-order susceptibility, *χ* ^(2)^, is the direct product of the complex conjugate of the Raman polarizability derivative matrix (the same as the original matrix when derivative values are real numbers) and the IR dipole derivative vector and thus is a second-order tensor. The generated electric field of SFG signal is related to the electric field of the two incident laser beams, *E* ~vis~ and *E* ~IR~, via the tensor elements *χ* ~*ijk*~ ^(2)^:$$\begin{matrix}
{E_{\text{SFG}}^{i} \propto {\sum\limits_{j,k}{\chi_{ijk}^{(2)}E_{\text{vis}}^{j}E_{\text{IR}}^{k}}},} \\
\end{matrix}$$where *i*, *j*, and *k* specify the direction of the Cartesian component of the optical fields and can be denoted by *x*, *y*, and *z*, in the laboratory coordinates.
The tensor elements of bulk phases with inversion symmetry (e.g., gas, solution, and amorphous solid) are isotropically averaged to zero, so long as the frequency of the visible beam is not in resonance with an electronic excitation \[[@B33], [@B73], [@B74]\]. This is because molecules rotate freely and diffuse, adopting random orientations. In contrast, molecules at surfaces give prominent SFG signals since they have ordered alignment across the surface region, and thereby their tensor elements are nonzero. Furthermore, the second-order susceptibility is proportional to the square of the molecular density at surfaces and thus is sensitive to the change of molecular coverage, enabling SFG to detect a monolayer of molecules at an interface. The monolayer sensitivity is critical for the characterization of biological samples that are difficult to purify in large quantities. As a result of such unique surface-specificity and monolayer sensitivity, the SFG method is free from contributions of the bulk medium and is thus an ideal optical method to probe membrane surfaces and their interactions with other biomolecules.
SFG spectroscopy has the intrinsic sensitivity to chiral structures because the measured second-order susceptibility tensor elements *χ* ~*ijk*~ ^(2)^ are three-dimensional. When the three indices *i*, *j*, and *k* are distinct from one another (i.e., *i* ≠ *j* ≠ *k*), the susceptibility tensor captures the features of the chiral Cartesian coordinate system with three distinct axes and thus comprises information about the chirality at interfaces \[[@B33]\]. Therefore, surface chirality can be directly measured through *χ* ~*ijk* (*i*≠*j*≠*k*)~ ^(2)^ making chiral SFG spectroscopy highly sensitive and easier for interpretation, compared to more conventional chiroptical methods, such as circular dichroism, Raman optical activity, and optical rotation dispersion, which require higher-order couplings between electronic and magnetic dipoles. Due to its high chiral sensitivity, SFG could provide previously unattainable molecular information about protein and biomolecules in chiral supramolecular and hierarchic structures at surfaces \[[@B74], [@B75]\].
SFG measurements can be modulated by various polarization settings. For a particular experiment, one can modulate the incident visible and infrared beams and detect the SFG beam in either*s*- or*p*-polarization \[[@B66], [@B69]\], or even mixed polarizations \[[@B48], [@B72], [@B76]\]. With the use of only*s*- or*p*-polarization, there are in total 2 × 2 × 2 = 8 polarization settings for experimental geometries, including*ssp*,*ppp*,*sps*,*pss*,*spp*,*psp*,*pps*, and*sss*, where the first, second, and third indices indicate the polarization of the SFG, visible, and infrared beams, respectively ([Scheme 2](#sch2){ref-type="fig"}). The*psp*,*spp*, and*pps* polarization settings are chiral-selective and thus can be used to probe the chiral SFG spectra as discussed previously \[[@B33]\]. The others are achiral polarization settings that are sensitive to different vibrational modes. Altogether, chiral and achiral SFG spectroscopy can provide a comprehensive analysis of vibrational modes of chiral or nonchiral molecules at interfaces. In more advanced measurements, one can even determine the absolute orientation of molecules at interfaces by performing a global analysis of various polarization-modulated spectra \[[@B77]\]. With these capabilities, SFG can report on structures and orientations of molecules and proteins at surfaces, offering a methodology to address mechanistic questions on amyloid aggregation that would otherwise be difficult to tackle by using more conventional methods.
2.3. SFG Experiments {#sec2.3}
--------------------
The setup of the SFG spectrometer has been extensively described \[[@B77]--[@B79]\]. Currently, scanning and broad-bandwidth SFG spectrometers are most commonly used. Scanning SFG spectrometers use picosecond pulsed IR and visible beams and scan the IR frequency stepwise at a fixed visible frequency \[[@B80]\]. A typical broad-bandwidth SFG spectrometer consists of a femtosecond (\~100 fs) pulsed IR beam and a picosecond (2--100 ps) pulsed visible beam \[[@B81]\]. Because the femtosecond IR pulse includes a wide frequency range of more than 200 cm^−1^ in the mid-IR region, broad-bandwidth SFG spectrometers can acquire spectra by one shot without scanning the IR frequency. The one-shot scheme allows for monitoring the kinetics of protein conformational changes at interfaces \[[@B55], [@B82]\]. In this review, we focus on recent studies of hIAPP aggregates at membrane surfaces based on broad bandwidth SFG spectroscopy \[[@B14], [@B61], [@B62], [@B64], [@B65]\].
Chiral and achiral SFG spectra can be selectively collected using different polarization settings. Typically, one can use*psp*,*spp*, and*pps* to probe the chiral SFG spectra and*ssp*,*sps*, and*ppp* to probe the achiral SFG spectra \[[@B33]\]. When performing the chiral and achiral SFG measurements, one can control the polarization of the beams using appropriate wave-plates and polarizers. Chiral SFG is particularly useful for probing biomolecules because most secondary structures are chiral, such as *α*-helices and *β*-sheets. These chiral macromolecular structures are expected to show a strong chiroptical response in chiral SFG while the solvent and other molecules lacking macroscopic chiral structures are silent in the chiral SFG spectra. Thus, chiral SFG spectra of biomacromolecules are less affected or distorted by signals from solvent or other achiral molecules, simplifying the spectral analysis and interpretation.
The choice of surface platform is another important factor for efficiently probing biomolecules at interfaces. For studies of hIAPP, there are two applicable surface platforms that allow for probing the lipid/aqueous interfacial region \[[@B83]\]. One surface platform is made by spreading a lipid monolayer onto the water surface \[[@B31]\], while the other is made by fabricating a supported lipid bilayer on a solid substrate, using the Langmuir-Blodgett (LB)/Langmuir-Schaefer (LS) method \[[@B51]\]. Both platforms have been widely used as mimics of biomembranes. Here, we will focus on the hIAPP studies that have been carried out using the platform of lipid monolayer. Briefly, the experimental procedure ([Scheme 3](#sch3){ref-type="fig"}) \[[@B62]\] starts with adding hIAPP to an aqueous buffer and then spreading the lipid monolayer at the air/water interface. The hIAPP will adsorb at the interface and interact with the lipid membrane. SFG spectroscopy can then be used to monitor lipid-induced conformational changes in hIAPP*in situ* and in real time at the interface.
3. SFG Probes the Early Stages of hIAPP Aggregation at Membrane Interfaces {#sec3}
==========================================================================
The early stages of hIAPP aggregation at interfaces involve hIAPP-membrane interactions associated with the pathogenic mechanism of type II diabetes \[[@B6], [@B84], [@B85]\]. However, it has been challenging to probe how hIAPP adsorbs onto the interface and whether hIAPP undergoes structural and orientation changes that might induce toxicity to pancreatic *β*-cells. Previous studies of the early stages of hIAPP aggregation mainly focused on bulk detection using methods such as CD, NMR, 2D-IR, and fluorescence spectroscopy that are not surface-sensitive or require spectroscopic labeling \[[@B9]\].
To overcome these challenges, Winter and coworkers applied infrared reflection absorption spectroscopy (IRRAS) and found *β*-sheets structures formed in the process of hIAPP aggregation at the air/water interface with a negatively charged lipid \[[@B31]\]. Inspired by the earlier IRRAS work, our group used SFG spectroscopy to probe hIAPP aggregation at membrane surfaces*in situ* and in real time monitoring the amide I and N-H stretching vibrational modes \[[@B61], [@B62]\]. Protein structures, including *α*-helices and *β*-sheets, are formed by hydrogen bonding interactions between the amide and N-H groups along the protein backbone. Thus, the vibrational frequency and line-shape of the amide I and N-H stretching modes are sensitive markers for distinguishing protein secondary structures \[[@B86], [@B87]\]. In conventional vibrational studies, the O-H bending and O-H stretching of water overlap with the amide I and N-H stretching modes, masking the characteristic bands of secondary structures, the peak assignment, and characterization of the protein secondary structure. In contrast, chiral SFG provides high-quality vibrational spectra revealing conformational changes previously undetectable by using conventional methods, since it probes only the interfacial molecules without any significant vibrational background from water solvent. Below is a summary of such vibrational SFG studies in the amide I and N-H stretching regions probing the early aggregation of hIAPP upon the interaction with membrane surfaces.
3.1. hIAPP Aggregation at Interfaces Probed by Amide I SFG Signals {#sec3.1}
------------------------------------------------------------------
First, our group has focused on experiments for both human IAPP (hIAPP) and rat IAPP (rIAPP) that probed the amide I region in both achiral and chiral SFG spectra \[[@B32], [@B33], [@B61], [@B62], [@B65], [@B88]\]. The two peptides are different by only six amino acids. Remarkably, the rIAPP does not aggregate into amyloids, making it an ideal control system for SFG studies. The achiral SFG spectra ([Figure 1](#fig1){ref-type="fig"}) were collected in the absence and presence of the negatively charged lipid dipalmitoylphosphoglycerol (DPPG). In the absence of DPPG, both hIAPP and rIAPP show amide I peaks at 1650 cm^−1^, suggesting the presence of both peptides at the air/water interface. Moreover, the spectra do not change over 10 hours, suggesting no noticeable structural changes. In contrast, with a DPPG monolayer and after incubation for 10 hours, the amide I band of hIAPP changes dramatically in terms of both the peak position and the line-shape. In contrast, the spectra of rIAPP remained unchanged. Since the frequency of the amide I band varies with protein secondary structures, the different spectroscopic responses from hIAPP and rIAPP indicate that hIAPP exhibits structural changes upon interaction with a DPPG monolayer, while rIAPP does not. [Figure 1](#fig1){ref-type="fig"} shows similar results for experiments in H~2~O and D~2~O, suggesting that isotopic effects are negligible on hIAPP aggregation.
A closer look at the spectral change of hIAPP incubated with DPPG after 10 hours shows that the amide I peak position is blue-shifted by 10 cm^−1^, from \~1650 to \~1660 cm^−1^, and there is an additional peak at 1750 cm^−1^ corresponding to the carbonyl stretch of the DPPG lipid \[[@B87]\]. Nonetheless, it is still challenging to specify what structural changes are involved at the lipid/aqueous interface. To address this question, we applied chiral SFG.
The chiral SFG measurements ([Figure 2](#fig2){ref-type="fig"}) show more interesting phenomena. Without the DPPG lipid, neither hIAPP nor rIAPP shows detectable chiral SFG signal in the amide I region. The lack of signal is not surprising since the native structures of hIAPP and rIAPP are disordered and do not adopt any chiral conformation. However, after incubating with DPPG for 10 hours, hIAPP shows a strong chiral SFG signal at 1622 cm^−1^ with a shoulder at 1660 cm^−1^. The low-frequency amide I band at 1622 cm^−1^ and a shoulder-peak at 1660 cm^−1^ are characteristic of parallel *β*-sheets \[[@B86], [@B87]\]. Thus, these results indicate that hIAPP forms parallel *β*-sheets upon the interaction with DPPG at the lipid/water interface.
Altogether, the achiral and chiral SFG studies of hIAPP in the amide I region demonstrate that both hIAPP and rIAPP can adsorb onto the air/water interface. Furthermore, hIAPP undergoes structural changes from disordered structures to parallel *β*-sheets upon the interaction with the surface of a lipid monolayer. Moreover, as a control system, rIAPP remains unchanged with and without lipid at the surface, revealing clear differences in the behavior of hIAPP and rIAPP at the lipid/aqueous interface. Detailed analyses of these spectral data have provided structural, kinetic, and orientation information, discussed in the following sections.
3.2. SFG Allows for Kinetic Studies of hIAPP Misfolding at the Early Stage {#sec3.2}
--------------------------------------------------------------------------
The chiral and achiral SFG spectra have been collected over time ([Figure 3](#fig3){ref-type="fig"}) \[[@B61], [@B62]\] to explore the kinetics of aggregation of hIAPP at the interface. [Figure 3(a)](#fig3){ref-type="fig"} shows that during the aggregation process the achiral amide I band of hIAPP gradually shifts to higher frequency with increased intensity, suggesting conformational changes in hIAPP. The increase in intensity reflects more ordered structures of hIAPP aggregates because SFG signals are sensitive to the ordering of the molecules at interfaces. [Figure 3(b)](#fig3){ref-type="fig"} shows that the chiral amide I signal at 1622 cm^−1^ starts emerging after three hours and keeps increasing with an appearance of a shoulder-peak at 1660 cm^−1^. This result not only suggests the formation of parallel *β*-sheets but also confirms the ordering of hIAPP during the aggregation process.
The successful use of the amide I band to probe the kinetics of *β*-sheet formation in hIAPP inspired us to use SFG signals from the peptide backbone in other vibrational regions, such as N-H stretching (amide A) \[[@B61]\]. We probed the chiral N-H stretching signals of hIAPP during the aggregation process of hIAPP under the same experimental conditions and observed unique spectral features. The chiral SFG spectra in [Figure 3(c)](#fig3){ref-type="fig"} show that initially there are no N-H stretching signals, but a peak at 3280 cm^−1^ gradually builds up and reaches a maximum value after roughly three hours of interaction with DPPG and then slowly vanishes after 10 hours. This transient chiral N-H stretching signal clearly reveals an intermediate in the hIAPP aggregation process. To investigate further the structure of this intermediate, we obtained the chiral N-H stretching spectra of several model proteins in *α*-helical structures. We concluded that the chiral SFG N-H stretching mode at 3280 cm^−1^ is due to *α*-helical structures \[[@B61]\]. A combination of the results of kinetic studies using the amide I ([Figure 3(b)](#fig3){ref-type="fig"}) and N-H stretching bands ([Figure 3(c)](#fig3){ref-type="fig"}) reveals an important finding: the appearance of the N-H stretching peak reaches a maximum and starts to disappear prior to the accumulation of the chiral amide I signal ([Figure 4(a)](#fig4){ref-type="fig"}), providing a molecular picture of hIAPP misfolding at membrane surfaces, where hIAPP initially adsorbs to the membrane surface as a random coil and then forms *α*-helical intermediates, which subsequently convert into parallel *β*-sheet aggregates.
As the first kinetic study using chiral and achiral SFG to probe conformational changes of proteins at interfaces*in situ* and in real time, the above studies demonstrate SFG as a method of high selectivity and sensitivity not only for characterizing protein secondary structures but also for studying the kinetics of conformational changes at interfaces. The N-H stretching and amide I bands are two well-separated vibrational regions that can be used synergistically for revealing aspects of the molecular mechanism of aggregation at interfaces. The kinetic data obtained at the lipid/water interface by SFG spectroscopy can be compared to measurements in the bulk solution based on conventional physical methods, yielding a better understanding of the role that the membrane surface plays in the amyloid aggregation process. This methodology is expected to find applications in testing the efficacy of drug candidates that inhibit the aggregation of hIAPP at lipid/water interfaces, as illustrated in the following section.
4. Inhibition of the hIAPP Aggregation at Membrane Surface {#sec4}
==========================================================
Several studies have shown that the aggregation of hIAPP is associated with the disruption of membrane integrity and death of pancreas cells \[[@B9], [@B11], [@B12]\]. Hence, the search for inhibitors of hIAPP aggregate formation has been a strategy explored for drug development \[[@B77], [@B89]\]. Previously, most of the screening of drug candidates inhibiting hIAPP aggregation has been tested in the bulk aqueous solution. However, inhibitors that work in the bulk may have low efficacy on membrane surfaces. Bonn and coworkers applied SFG to address this issue \[[@B64]\] and have confirmed that the surface indeed plays an important role in reducing the inhibition of hIAPP aggregation by drug candidates.
4.1. EGCG Shows Less Inhibition of hIAPP Fibril Formation at Interfaces {#sec4.1}
-----------------------------------------------------------------------
Bonn and coworkers studied (−)-epigallocatechin gallate (EGCG) as an inhibitor for hIAPP aggregation and compared its effects in the bulk solution and on membrane surfaces \[[@B64]\]. EGCG is a natural product found in green tea and belongs to a class of inhibitors containing polyphenols. Previous studies showed that EGCG can effectively inhibit the misfolding and fibrillation of hIAPP and even disaggregate *β*-sheet-rich amyloids in the bulk solution \[[@B90], [@B91]\]. However, the inhibitive effect of hIAPP at the interface remains unclear.
Bonn and coworkers combined achiral SFG spectroscopy with thioflavin T (ThT) fluorescence and atomic force microscopy (AFM) for bulk analysis to examine the effect of EGCG on inhibiting hIAPP aggregation at the lipid/water interface \[[@B64]\]. They used the amide I band of hIAPP in the achiral SFG spectra to monitor the kinetics of the formation of *β*-sheet fibrils. In the absence of EGCG at the lipid/water interface, the time-dependent SFG spectra ([Figure 5(d)](#fig5){ref-type="fig"}) show blue-shifts in peak position and increasing intensities of amide I band, similar to the observations made by Fu et al. ([Figure 3(a)](#fig3){ref-type="fig"}). In the presence of EGCG, blue-shifts in peak positions can still be observed ([Figure 5(e)](#fig5){ref-type="fig"}), indicating the formation of *β*-sheets. This is confirmed by the AFM images of the sample at interfaces transferred onto mica. The AFM images show the formation of fibrils on the film made from hIAPP at the air/water interface with lipid in the presence of EGCG after incubating for \~17 hours ([Figure 5(c)](#fig5){ref-type="fig"}). The extent of fibril formation is similar to that in the absence of EGCG ([Figure 5(b)](#fig5){ref-type="fig"}). The results at the interface elicit the hypothesis that EGCG has a reduced inhibitive effect on hIAPP aggregation at the lipid/water interface.
To test this hypothesis, a comparison is made for the inhibitive effect of EGCG on hIAPP aggregation at the interface versus in the bulk. The amount of *β*-sheets formed at the lipid/water interface in both the presence and the absence of EGCG is quantitatively estimated by the deconvolution of each time-dependent spectrum using the Lorentzian line-shape fitting. The time-dependent *β*-sheet component deconvoluted from the SFG spectra is plotted in [Figure 6(a)](#fig6){ref-type="fig"} (⋄ and □ curves), along with the time-dependent fluorescence intensity from hIAPP aggregates in the bulk (× and + curves). The flat + curve in [Figure 6(a)](#fig6){ref-type="fig"} suggests a lack of amyloid fibril formation in the bulk in the presence of EGCG, which is further confirmed by the AFM image in the presence ([Figure 6(c)](#fig6){ref-type="fig"}) of EGCG. Therefore, both the fluorescence and the AFM results indicate a strong inhibition of EGCG on hIAPP aggregation in the bulk. On the other hand, the *β*-sheet component deconvoluted from the SFG spectra is still increasing with time at the lipid/water interface in the presence of EGCG ([Figure 6(a)](#fig6){ref-type="fig"}, □ curve). The above comparison demonstrates that the inhibitive effect of EGCG on hIAPP is indeed reduced at the interface.
4.2. Consideration of Membrane Effects for Drug Design {#sec4.2}
------------------------------------------------------
The reduced efficacy of EGCG as an inhibitor at membrane surfaces may be due to the role of the surface in controlling the structure, orientation, and dynamics of the aggregation process. The proposed inhibitory mechanism for EGCG involves the binding of the phenol groups to the hydrophobic aromatic side chains of hIAPP. In the bulk, both the inhibitor and the hIAPP diffuse freely. Thus, the inhibitor may bind to the hIAPP aromatic groups more effectively. At the amphipathic membrane-water interface, however, hIAPP is anchored with a specific orientation leaving the hydrophobic *β*-strands pointing towards each other, buried inside the membrane phase, while the hydrophilic side chains make contact with water solvent and the lipid polar head groups. This specific orientation suppresses free diffusion of hIAPP and makes it difficult for EGCG to bind. On the other hand, the small EGCG molecules with multiple phenol groups are more soluble in the bulk, potentially leading to low surface population, which further reduce its efficacy.
The SFG study demonstrates that the effect of the surface can be a critical factor to be considered in the drug design for type II diabetes. The pathogenic origin of type II diabetes has been proposed to be linked to the hIAPP aggregation process and the disruption of membrane integrity \[[@B5]\]. Therefore, it is important to screen drug candidates that might affect aggregate/membrane interactions, by changing the hIAPP interfacial orientation, conformation, or dynamics. The resulting conformational changes could be probed by SFG as described in the following section focused on the orientation of aggregates at the lipid/water interface.
5. Orientation of hIAPP Aggregates at Lipid Membrane Surfaces {#sec5}
=============================================================
Understanding, at the molecular level, whether hIAPP aggregates disrupt cell membranes could provide valuable insights into the pathogenic mechanisms of type II diabetes. Disruption due to a specific orientation of the hIAPP aggregates adsorbed on the membrane surface might increase the membrane permeability to water and ions. However, determining the interfacial orientation of complex biomolecules by conventional methods is challenging. In particular, complex hIAPP aggregates in the form of pleated parallel *β*-sheets pose significant challenges for both theory and experiments \[[@B92]--[@B94]\]. We have combined molecular dynamics and divide-and-conquer*ab initio* quantum chemistry calculations of hyperpolarizability derivatives with respect to normal modes to simulate the chiral SFG spectroscopy of hIAPP \[[@B65]\]. Our simulations found that the hIAPP aggregates are neither parallel nor perpendicular to the membrane surface but rather inserted into the lipid membrane tilted at an angle of about 45° relative to the surface normal. The resulting orientation optimizes amphiphilic interactions by exposing hydrophilic domains of the aggregate to the aqueous phase and hydrophobic parts to the lipids. Such "detergent-like" orientation is expected to cause significant disruption of the cell membrane.
This section describes the theoretical and experimental analyses of chiral SFG spectra necessary to retrieve the orientation of the hIAPP aggregate at the interface.
The chiral SFG spectrum of the aggregated hIAPP at the membrane surface shows a dominant peak at 1620 cm^−1^ with a shoulder-peak at 1660 cm^−1^ in the amide I region ([Figure 7(b)](#fig7){ref-type="fig"}). These two peaks correspond to the antisymmetric (*B*-mode) and symmetric (*A*-mode) bands of parallel *β*-sheets. Detailed analyses of the molecular symmetry and vibrational coupling indicate that their relative ratio of intensities (*I* ~*B*~/*I* ~*A*~) is correlated with the orientation of the *β*-sheet$$\begin{matrix}
{I_{B/A}\left| {\frac{\chi_{psp,B}^{|2|}}{\chi_{psp,A}^{|2|}}} \right|^{2}} \\
{= \left| {\left\langle {{\tan^{2}\psi}} \right\rangle\frac{\beta_{bca,B}}{\beta_{acb,A}} + \left| { 1 - \left\langle {{\tan^{2}\psi}} \right\rangle} \right|\frac{\beta_{bac,B}}{\beta_{acb,A}}} \right|^{2},} \\
\end{matrix}$$where *ψ* is the angle between the *β*-strand and the interface, as defined in [Figure 7(a)](#fig7){ref-type="fig"}. The hyperpolarizability elements (*β* ~*acb*,*A*~, *β* ~*bca*,*B*~, *β* ~*bac*,*B*~) provide the specific molecular property of the hIAPP aggregates, with *a*, *b*, *c* referring to the Cartesian coordinates in the molecular frame of the hIAPP parallel *β*-sheet, where *b* is the direction parallel to the *β*-strand and *c* points to the axis of propagation of intermolecular *β*-sheets ([Figure 7(a)](#fig7){ref-type="fig"}). From ([2](#EEq2){ref-type="disp-formula"}), it is clear that the orientation angle *ψ* can be determined as the value that matches the experimental ratio of SFG intensities for the *B* and *A* bands (*I* ~*B*/*A*~). The intensity ratio is measured to be 4.8 from the fitted amide I spectrum ([Figure 7(b)](#fig7){ref-type="fig"}). Consequently, knowing the values of three hyperpolarizability elements (*β* ~*acb*,*A*~, *β* ~*bca*,*B*~, *β* ~*bac*,*B*~) in ([2](#EEq2){ref-type="disp-formula"}) allows the determination of the average orientation (*ψ*) determined by SFG spectroscopy.
We have applied a divide-and-conquer approach that fragments the hIAPP aggregate into domains amenable to quantum chemistry calculations and computes the hyperpolarizability elements of the constituent fragments at the density functional theory (DFT) level. Specifically, the NMR structure \[[@B18]\] was divided into 16 tripeptide pairs in the *β*-sheet region, and the hyperpolarizability of each tripeptide pair was calculated by DFT ([Figure 7(c)](#fig7){ref-type="fig"}). The overall hyperpolarizability of hIAPP aggregates is then integrated from the hyperpolarizability elements of the individual tripeptide pairs. The plot of the intensity ratio of the amide I peaks as a function of the orientation of the parallel *β*-sheet shows that the orientation *ψ* = 45--48° has the best agreement with experimental data, suggesting that the hIAPP aggregates orient with the *β*-strand at *ψ* ≈ 45° from the surface. The tilted orientation of hIAPP *β*-sheet aggregates at the lipid/water interface suggests that significant disruption might be caused on the lipid membrane. These findings are supported by molecular dynamics simulations that analyzed the orientation and stability of hIAPP aggregates at DPPG/water interfaces \[[@B14]\]. The simulation shows that hIAPP gets inserted into lipid monolayers at about *ψ* = 40° and in lipid bilayers at a tilted angle of *ψ* = 60°, as shown in [Figure 8](#fig8){ref-type="fig"}, leading to water permeation and Na^+^ percolation through the membrane supporting the hypothesis of ion-cytotoxicity for islet *β*-cells \[[@B14]\].
These molecular dynamics simulations support the unique capabilities of chiral SFG spectroscopy for probing the orientation of *β*-sheet amyloid aggregates, providing fundamental insights that should be particularly relevant for understanding amyloid diseases at the molecular level. The combination of computational modeling and SFG spectroscopy thus provides a valuable methodology to identify potential noncompetitive inhibitors that might change the conformation and orientation of hIAPP aggregates and consequently reduce their toxicity as implicated in amyloid diseases.
6. Perspectives and Challenges of SFG in Biological and Medical Applications {#sec6}
============================================================================
In summary, we have reviewed the application of SFG spectroscopy to study the amyloidogenesis of hIAPP interacting with membrane surfaces \[[@B14], [@B61], [@B62], [@B64], [@B65]\]. We have shown that SFG can reveal structural and dynamic information characterizing the formation of aggregates*in situ* and in real time by using different polarization settings and probing in different vibrational regions. The versatility of SFG experiments also provides ample information that allows for elucidating the orientation of hIAPP aggregates at the water/lipid interface. With the capabilities of probing and characterizing structure, orientation, and dynamics, studies based on SFG spectroscopy can provide valuable insights into membrane/protein interactions that are critical to a wide range of pathological diseases, including amyloidogenesis and cytotoxicity to pancreas *β*-cells leading to the onset of type II diabetes. Based on these findings, potential drug candidates that specifically target the early aggregation of hIAPP at the membrane/water interface are currently being proposed and tested by using SFG spectroscopy to guide the rational design of drugs for treatment of type II diabetes.
Studies of other amyloid diseases such as Alzheimer\'s, Parkinson\'s, Huntington\'s, and Prion diseases could also benefit from SFG techniques. In fact, Luo and coworkers have already applied SFG to study the membrane-mediated structural change of prion protein fragments and characterized the concentration dependence of structures and orientation for prion oligomers \[[@B95]\]. In addition, Weidner and coworkers have performed SFG studies to investigate oligomerization of lysozyme at membrane surfaces, where they simultaneously monitored conformational states of lysozyme and the organization of lipid molecules in contact with aqueous buffer at various values of pH \[[@B63]\].
It is foreseeable that SFG spectroscopy can be extended to a wider range of studies critical for the mechanistic understanding of amyloid diseases and drug development. Potential applications include studies of the interactions of amyloid proteins with components in cell membranes (e.g., cholesterol, membrane protein, and glycolipid) \[[@B96]\]; other relevant biomolecules (e.g., insulin \[[@B97]\] and sphingolipid); and potential drug candidates (e.g., small aromatic organic molecule and peptide analogue of amyloid protein). Those applications could provide valuable insights into amyloidogenic intermediates during the onset of membrane diseases.
Given the potential applications of SFG in the investigation of amyloidogenesis, two major challenges remain in order to develop SFG into a general biophysical tool for biological and biomedical research. These challenges include the development of efficient methods for acquisition of high-quality spectroscopic data and unequivocal interpretation of the spectra. On one hand, experimentalists in the SFG field have been striving to improve the instrumentation for SFG spectroscopy to enhance the quality of the data. Broad bandwidth and even ultrabroad bandwidth SFG spectroscopy have been developed to cover a wider vibrational frequency range with one-shot measurement \[[@B56]\], shortening the time for spectral acquisition and rendering it possible to monitor kinetics of structural and orientational changes. High-resolution SFG spectroscopy can characterize surfaces and interfaces with unprecedented subwavenumber (\<1 cm^−1^) resolution \[[@B98]\], providing detailed molecular information necessary to understand structure/function relations in physical and biological processes \[[@B99]--[@B101]\]. In addition, heterodyne-detected SFG spectroscopy has been developed to enhance the signal level of SFG spectra \[[@B102]--[@B107]\]. Moreover, 2-dimensional SFG spectroscopy and SFG microscopy have emerged as complementary techniques for the characterization of couplings and interactions in biomolecules \[[@B108]--[@B110]\]. On the other hand, theoretical methods for modeling SFG spectroscopy are critical for the interpretation of the SFG data. Methodologies that combine the essence of traditional theories of vibrational spectroscopy and the power of high-performance computing continue to be improved for more efficient calculations \[[@B111], [@B112]\] that provided rigorous first-principle interpretations of the experimental data in terms of structure/orientation relations of biological systems at interfaces. Efforts have been focused on accurate computations of molecular hyperpolarizabilities for different secondary structures of proteins, critical for extracting information about molecular orientation at membrane surfaces. These advancements made by strong collaborations between experimentalists and theoreticians in the field of SFG spectroscopy are expected to continue to produce valuable insights into a broad range of problems in chemical, biological, and biomedical sciences.
Elsa C. Y. Yan is the recipient of the Starter Grant Award, Spectroscopy Society of Pittsburgh, the National Science Foundation Grant (CHE 1213362), and the National Institutes of Health Grant (1R56DK105381-01). Victor S. Batista acknowledges high performance computing time from NERSC and support from the NSF Grant CHE-1213742.
Conflict of Interests
=====================
The authors declare no conflict of interests.
{#sch1}
{#sch2}
![Illustration of adsorption of hIAPP on a lipid monolayer and the SFG experiment for probing the hIAPP aggregations at the lipid/water interface. Adapted from \[[@B62]\] with permission. Copyright 2010 American Chemical Society.](JDR2016-7293063.sch.003){#sch3}
![The*ssp* (achiral) SFG spectra of IAPPs. Human IAPP without DPPG (*t* = 0 h and *t* = 10 h) and with DPPG at *t* = 10 h at the (a) air/D~2~O and (b) air/H~2~O interfaces; rat IAPP without DPPG (*t* = 0 h and *t* = 10 h) and with DPPG at *t* = 10 h at the (c) air/D~2~O and (d) air/H~2~O interfaces. Adapted from \[[@B62]\] with permission. Copyright 2010 American Chemical Society.](JDR2016-7293063.001){#fig1}
![The*psp* (chiral) SFG spectra of IAPPs. Human IAPP without DPPG (*t* = 0 h and *t* = 10 h) and with DPPG at *t* = 10 h at the (a) air/D~2~O and (b) air/H~2~O interfaces; rat IAPP without DPPG (*t* = 0 h and *t* = 10 h) and with DPPG at *t* = 10 h at the (c) air/D~2~O and (d) air/H~2~O interfaces. Adapted from \[[@B62]\] with permission. Copyright 2010 American Chemical Society.](JDR2016-7293063.002){#fig2}
![Kinetics of human IAPP aggregates at the lipid/H~2~O interface probed by time-dependent SFG spectra in the amide I region using (a)*ssp* (achiral) and (b)*psp* (chiral) polarization in the presence of DPPG and (c) in the N-H stretching region using*psp* (chiral) polarization in the presence of DPPG. Adapted from \[[@B61], [@B62]\] with permission. Copyright 2010 and 2011 American Chemical Society.](JDR2016-7293063.003){#fig3}
![The misfolding pathway of hIAPP on membrane surfaces. (a) Chiral SFG intensities of the chiral N-H stretching (3280 cm^−1^) and amide I signals (1620 cm^−1^) as a function of time from triplicate experiments. (b) The model mechanism on hIAPP aggregation at the membrane surface: adsorption of disordered hIAPP onto membrane leads to formation of *α*-helical intermediates which are then converted to *β*-sheet aggregates. Adapted from \[[@B61]\] with permission. Copyright 2011 American Chemical Society.](JDR2016-7293063.004){#fig4}
![Kinetics study using AFM and SFG measurements during hIAPP aggregation at the phospholipid interface in the presence, or absence, of EGCG. AFM images of hIAPP with lipid for (a) 10 and (b) 1020 min in the absence of EGCG and (c) AFM image after hIAPP aggregation in the presence of EGCG for 1020 min. SFG spectra of hIAPP with phospholipid in the amid I region (d) in the absence of, and (e) in the presence of, EGCG. Adapted from \[[@B64]\] with permission. Copyright 2012 American Chemical Society.](JDR2016-7293063.005){#fig5}
![Bonn and coworkers compared the inhibitory effect of EGCG on hIAPP fibrillation in the bulk and at the phospholipid interface. (a) In the bulk, the formation of hIAPP amyloid fibrils was measured using a ThT fluorescence assay. The sigmoidal increase in fluorescence signal in the absence of EGCG (×) indicates the formation of hIAPP fibrils, and the low fluorescence signal in the presence of EGCG at a 1 : 1 molar ratio (+) suggests the inhibition of fibril formation. At the phospholipid interface, in absence (⋄) and presence (□) of EGCG, SFG spectra measure the formation of *β*-sheets. (b) AFM also shows hIAPP fibrils formed in the absence of EGCG and (c) EGCG completely inhibits fibril formation of hIAPP. Adapted from \[[@B64]\] with permission. Copyright 2012 American Chemical Society.](JDR2016-7293063.006){#fig6}
{ref-type="disp-formula"}), and the red curve is obtained numerically. (e) Chiral SFG spectra of human IAPP aggregates simulated for various orientations at the interface. (f) Visualized orientation of the human IAPP aggregates at the lipid/aqueous interface. Adapted from \[[@B65]\] with permission. Copyright 2012 Elsevier.](JDR2016-7293063.007){#fig7}
![Bilayer thickness around embedded hIAPP for (a) trimer and (b) tetramer aggregates, as described by molecular dynamics simulations. Color key: bilayer thickness (in nm) is mapped to the corresponding colors. (c) Water channel formed by the hIAPP trimer in the DPPG bilayer, showing water molecules (red and white) and Na^+^ ions (blue vdW spheres) percolating through the bilayer. (d) Average particle density of water and Na^+^ ions within the membrane. Adapted from \[[@B14]\] with permission. Copyright 2013 by the Biophysical Society.](JDR2016-7293063.008){#fig8}
[^1]: Academic Editor: Lucie Khemtemourian
| {
"pile_set_name": "PubMed Central"
} |
**Core tip:** Always review a peripheral blood film in all cases of multiple myeloma and be aware of an entity called plasma cell leukemia, a rare and aggressive form of leukemia.
INTRODUCTION
============
Primary plasma cell leukemia (pPCL) is a malignant plasma cell disorder characterized by the presence of 2 × 10^9^/μL peripheral blood clonal plasma cells or \> 20% plasma cells in the peripheral blood. It is very rare, accounting for 0.6%-4% of all plasma cell neoplasms, and is reported to occur in \< 1 in a million. PCL has a relatively poor prognosis, due to its very aggressive nature involving extramedullary organs, lytic bone lesions, destruction of red blood cells, and bone marrow failure. Treatment includes immunomodulators, proteasome inhibitors, and autologous stem cell transplantation. Outcomes are not promising, however, even after treatment; median survival after rigorous chemotherapy and transplant is not more than three years\[[@B1]\]. It is important to consider PCL as a possible diagnosis as well as multiple myeloma (MM), whenever we encounter the typical constellation of "Hypercalcemia, Renal Failure, Anemia, and Lytic Bone lesions", and to contemplate the peripheral smear, a basic test, which, in our case study, showed \> 20% plasma cells, hence leading to the diagnosis of PCL.
CASE PRESENTATION
=================
Chief complaints
----------------
A 56-year-old male with a history of hypertension was referred to the hospital by his primary care physician after a routine, yearly laboratory exam showed abnormal hemoglobin of 5.1 mg/dL.
History of present illness
--------------------------
The patient complained of some pain in the left flank area, which started around four months prior to admission.
History of past illness
-----------------------
He denied any history of smoking and did not have any significant family history.
Physical examination
--------------------
On admission, his vital signs were stable and physical exam was completely benign.
Laboratory examination
----------------------
His laboratory evaluation revealed hemoglobin of 5.1 mg/dL, white blood cell count 6.6 × 10^9^/μL with 16% atypical lymphocytes, and platelet count of 51000/μL. Chemistries were significant for a sodium of 125 meq/L, creatinine of 2.79 mg/dL, calcium of 8.3 mg/dL, and normal liver enzymes with a total bilirubin of 3.1 mg/dL. Total protein was 15.3 g/dL.
Imaging examination
-------------------
A Computed tomography (CT) scan was ordered to evaluate the left flank pain, which showed a lytic lesion at T11 (11^th^ thoracic vertebra), right anterior 7^th^ rib with a pathological fracture, and multiple small lucencies in the vertebrae.
Further diagnostic work-up
--------------------------
Peripheral smear showed more than 10%-15% plasma cells (Figure [1](#F1){ref-type="fig"}), and flow cytometry of peripheral blood confirmed PCL with 24% plasma cells (CD138+). Serum protein electrophoresis showed a kappa/lambda ratio of 43 and a monoclonal spike of 8.44 g/dL in the IgG kappa region on immunofixation. Beta-2 microglobulin was also high at 19 μg/mL. A 24-h urine collection showed a total protein of 1557 mg and urine protein electrophoresis (UPEP) showed two spikes of 15 g/dL (IgG kappa) and 2 g/dL (free kappa). Bone marrow biopsy (Figure [2A](#F2){ref-type="fig"} and B) demonstrated 80% plasma cells (38+, 138+, 117+, 10-, 19-, 20-, 56-) with 90% cellularity. Fluorescence in situ hybridization (FISH) and chromosome analysis showed 13q deletion among many other aberrancies.
{#F1}
{#F2}
FINAL DIAGNOSIS
===============
pPCL.
TREATMENT
=========
VCD therapy was started \[cytoxan (750 mg/m^2^), velcade (0.7 mg/m^2^ on day 1, 4, 8, 11) with dexamethasone (40 mg)\]. The patient received a total of five units of blood transfused during his 15-d hospital stay. After completing therapy at 1, 4, 8, and 11 days, the patient was discharged home.
OUTCOME AND FOLLOW-UP
=====================
The patient responded well to chemotherapy and on discharge he was being considered for bone marrow transplant evaluation within two months.
DISCUSSION
==========
PCL is established by the presence of \> 20% circulating plasma cells or an absolute plasma cell count \> 2 × 10^9^/L on peripheral smear. It is a very uncommon and aggressive form of monoclonal gammopathy characterized by a poor prognosis with a rapidly fatal outcome. Complications usually lead to death within the first few months of diagnosis. Outcome is thought to be poor because of the absence of effective treatment for this condition\[[@B1]\]. PCL is classified as "primary" (occurring de novo), or "secondary" (occurring in patients with MM).
There are a few differences between these two disease processes. Patients with PCL are usually younger (aged 50-59 years). PCL has a predisposition to develop malignant plasma cells, which circulate in peripheral blood and lead to extramedullary spread involving the liver, spleen, lymph nodes, pleura, peritoneum, and less often the bone, resulting in lytic lesions. The extramedullary spread is explained by negative CD56, a cell adhesion molecule which anchors plasma cells to the bone marrow stroma in contrast to MM, where most of the plasma cell population is found in bone marrow\[[@B2]\]. Hypercalcemia, low platelet count, and destruction of erythrocytes is found in both diseases, but it is more pronounced in PCL than in MM. PCL differs in biology compared to MM with more immature cells in the bone marrow. Most PCL patients have abnormal karyotypes. The FISH probes should be directed against genetic abnormalities, such as del(17p13), del(13q), del(1p21), and (1q21) amplification, and chromosome 14 abnormalities, such as t(11;14), t(4;14), t(14;20) and t(14;16). About 87% of the PCL cases can be attributed to IGH (14q32) translocation, which is hence the most common mutation, followed by t(11;14), which accounts for 25%-65%. In our patient, however, 13q deletion was observed, which is seen in \< 20% of patients\[[@B3],[@B4]\]. Plasma cell markers, which can be identified in PCL on immunophenotyping, include CD38 and CD138, which were both present in our patient. Flow cytometry should be performed on peripheral blood to confirm the presence of plasma cells that typically have immunophenotypes of CD138+, CD38+, CD19--, and CD45+/--. Our finding of negative expression of CD10, CD19, CD20, CD56 was consistent with previously published cases\[[@B5],[@B6]\].
Patients having pPCL must undergo a detailed history and physical examination. A comprehensive laboratory evaluation of their blood should be performed, including a complete blood count with differential, peripheral blood smear, electrolyte panel, urea and creatinine levels, liver enzymes, bilirubin, lactate dehydrogenase, uric acid, β2 microglobulin, albumin, serum protein electrophoresis, and serum free light chain analysis. Chest X-ray along with whole body imaging with either MRI or CT should be performed to look for metastasis. Sometimes 18F-FDG PET/CT imaging to look for both lytic bone lesions and extramedullary plasmacytomas is necessary. A 24-h urine collection for electrophoresis and total protein assessment should also be obtained. Finally, it is imperative to perform a bone marrow biopsy and aspiration to assess morphology, proliferation rate, immunophenotyping, and cytogenetic analysis by FISH\[[@B7]\].
Managing patients with PCL requires an intensive risk-adapted approach. Induction therapy with novel triplet therapy using immunomodulators and proteasome inhibitors such as VRd (bortezomib, lenalidomide, and dexamethasone) or KRd (carfilzomib, lenalidomide, and dexamethasone), is usually a satisfactory choice. In some patients with pPCL who have an aggressive form of disease, more aggressive combination regimen, such as VDT-PACE (bortezomib, dexamethasone, thalidomide or lenalidomide, cisplatin, doxorubicin, cyclophosphamide, and etoposide) or HyperCVAD (High dose Cyclophosphamide, vincristine, adriamycin and thalidomide or lenalidomide) should be used, because cyclophosphamide and doxorubicin are particularly effective in proliferative disease. For elderly patients who may not be able to tolerate such an intense regimen, CyBorD (cyclophosphamide, bortezomib, and dexamethasone) or PAD (bortezomib, doxorubicin, and dexa-methasone) can be used as a milder alternative\[[@B8]\].
After induction therapy, autologous stem cell transplant is recommended (for transplant-eligible pPCL patients) to achieve prolonged disease control. Standard antiviral and bacterial prophylaxis should be used mainly because the patients are immunocompromised. The patients may also develop tumor lysis syndrome, so appropriate prophylaxis with allopurinol and appropriate hydration may be indicated. Adequate antithrombotic prophylaxis as well as platelet, red blood cell monitoring, and replacement is also crucial. Finally, although osteolytic bone lesions are less common in pPCL than in MM, all patients with pPCL should be started on bisphosphonate therapy along with vitamin D\[[@B9]\].
Median survival has been estimated to be around 6-7 years with conventional chemotherapy. With autologous or allogeneic stem cell transplantation, survival has shown improvement to around three years. Less than 10% patients survive for more than five years.
CONCLUSION
==========
PCL is the most aggressive variant of monoclonal gammopathy and is a rare form of clonal plasma cell dyscrasia. Around 60%-70% PCL are primary and 30%-40% are secondary. Median age of diagnosis is 52 to 65 years which is ten years younger than the usual myeloma population. PCL and MM are distinct clinico-pathologic entities with different treatment options and most important, different prognoses. Multi-center studies and clinical trials should be conducted to develop accurate criteria for the initial diagnosis and prompt treatment of this disease.
Informed consent statement: We obtained informed consent for publication.
Conflict-of-interest statement: The authors state that they have no conflicts of interest.
CARE Checklist (2016) statement: The manuscript was prepared and revised according to the CARE Checklist (2016).
Manuscript source: Unsolicited manuscript
Peer-review started: November 10, 2018
First decision: December 7, 2018
Article in press: February 27, 2019
Specialty type: Medicine, Research and Experimental
Country of origin: United States
Peer-review report classification
Grade A (Excellent): 0
Grade B (Very good): 0
Grade C (Good): C, C
Grade D (Fair): 0
Grade E (Poor): 0
P- Reviewer: Hosseini M, Ali I S- Editor: Dou Y L- Editor: A E- Editor: Wu YXJ
[^1]: Author contributions: Jain AG, Manoucheri M played significant roles in reviewing the manuscript; Faisal-Uddin M, Khan AK, and Wazir M worked on writing the manuscript and Shen Q worked on the pathology legends.
Corresponding author: Akriti Gupta Jain, MD, Doctor, Resident, Internal Medicine, Florida Hospital, 2501 N. Orange Avenue, Orlando, FL 32804, United States. <[email protected]>
Telephone: +1-330-3229730 Fax: +1-407-3032553
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#sec1-1}
============
Post pandemic 2009 H1N1 viral infection, multiple small epidemic outbreaks have been reported worldwide.\[[@ref1]\] Mortality up to 45% among patients admitted to intensive care units (ICUs) has been reported in a recent epidemic.\[[@ref2][@ref3]\] The main cause of death in patients with pneumonia caused by H1N1 viral infection is multiorgan failure secondary to hypoxemia because of acute respiratory distress syndrome (ARDS) and secondary bacterial infection.\[[@ref4][@ref5]\] As with other causes of ARDS, management of H1N1 viral infection ARDS is focused on lung protective ventilation (LPV) strategies, but despite that mortality remains high.\[[@ref6]\] Other measures to improve survival include prone position ventilation (PPV) and extracorporeal membrane oxygenation (ECMO); although enough evidences showed that PPV reduces mortality (16%) compared to supine position ventilation (32.8%),\[[@ref2][@ref7][@ref8]\] there are underutilization of this strategy, which is found in recent trials.\[[@ref5][@ref9]\] In addition, there is scarcity of literature on the use of prone ventilation in H1N1 ARDS patients.\[[@ref10][@ref11][@ref12]\] We studied the effect of prone ventilation in severe ARDS patients admitted with the diagnosis of H1N1 during an epidemic in the winter months of 2014--2015 in a southern state of India.
METHODS {#sec1-2}
=======
This study was conducted at a tertiary health-care center in southern part of India. All patients admitted to medical ICU with H1N1 viral pneumonia having severe ARDS requiring prone ventilation as a rescue therapy for severe hypoxemia were reviewed prospectively. Polymerase chain reaction-confirmed H1N1 was included in this series. ARDS was defined as per the Berlin definition.\[[@ref13]\] Patients with cardiogenic pulmonary edema and spinal instability were excluded from the study. All the included patients were studied while sedated with fentanyl and midazolam and paralyzed with atracurium. The ventilatory settings were set as per the treating physician. Positive end-expiratory pressure (PEEP) and FiO~2~ were adjusted according to the PEEP-FiO~2~ table of ARDSnet trial.\[[@ref6]\] Measurements were obtained in supine (baseline) and PPV, after 30--60 min and then 4--6 hourly. The patients were considered to turn prone if PaO~2~/FiO~2~ ratio was \<100 cmH~2~O and PaCO~2~ was \>45 cmH~2~O. Patients were classified as nonresponders if the PaO~2~/FiO~2~ ratio does not increase by \>20% and PaCO~2~ decrease by 6 cmH~2~O. In PPV, if no progressive improvement was seen in PaO~2~/FiO~2~ and PaCO~2~ over a period of 4 h, then the patients were considered to turn supine. In this descriptive study, parametric and nonparametric data were presented as means (standard deviation) and medians, respectively.
RESULTS {#sec1-3}
=======
Between September 1, 2014, and March 15, 2015, 93 patients were treated for H1N1 viral pneumonia. Of these, 11 adult patients developed severe ARDS which did not responded to lung-protective ventilator strategies and were ventilated in prone position \[[Table 1](#T1){ref-type="table"}\]. Their age range was 26--59 years (median: 48), with four patients under the age of 40 years. Seven patients were female, including one postpartum. All presented with fever, cough, and dyspnea for 4--12 days. The number of days from symptom onset to invasive ventilation ranged from 7 to 12 days (median: 8). Five patients had comorbid illness, the common being morbid obesity and diabetes mellitus. On the day of ICU admission, the median Acute Physiological and Chronic Health Evaluation-II and Sequential Organ Failure Assessment scores were 20 (range: 16--27) and 7 (range: 5--12), respectively. Noninvasive ventilation trial was given in eight patients before invasive ventilation, out of which two required invasive ventilation within 12 h and the rest in 48 h. The worst PaO~2~/FiO~2~ ratio range on the day of invasive ventilation was 48--100 (median: 79). Out of 11 patients, PPV was started within 24 h in 8 patients and within 48 h in 3 patients. Preprone, ventilator, and blood gas parameters are summarized in [Table 2](#T2){ref-type="table"}; the response of PaO~2~/FiO~2~ ratio to prone ventilation is shown in [Figure 1](#F1){ref-type="fig"}. Four patients had cardiac dysfunction, with one having viral myocarditis (global hypokinesia ejection fraction of 23%), and three patients had pulmonary arterial hypertension with acute right heart failure. A total of 39 PPV sessions were done, with a range of 1--8 prone sessions per patient (median: 3). A total of 580 h of prone ventilation was done with an average of 14 h 52 min/session (range: 8--24) of prone ventilation. Out of the 39 PPV sessions, PaO~2~/FiO~2~ ratio and PaCO~2~ responder were 38 (97.4%) and 27 (69.2%) sessions, respectively.
######
Demographic and clinical characteristic of patients
Patient number Age Gender Comorbid At ICU admission Days of illness before invasive ventilation Severity of ARDS: PaO~2~/FiO~2~ ratio on the day of prone ventilation Intubation to prone ventilation in days Number of prone ventilation sessions MV days ICU days Outcome at discharge
---------------- ----- -------- -------------------------------- ------------------ --------------------------------------------- ----------------------------------------------------------------------- ----------------------------------------- -------------------------------------- --------- ---------- ---------------------- ----------
1 48 Female \- 27 12 8 55 1 4 12 16 Survived
2 55 Female \- 22 7 8 100 2 5 15 19 Survived
3 40 Female DM 20 8 8 57 0 3 14 15 Survived
4 59 Female HTN, COPD 18 5 8 79 1 5 19 22 Survived
5 52 Male \- 19 10 7 75 2 8 16 21 Survived
6 52 Male CKD, Ca tongue on chemotherapy 24 10 8 73 1 3 9 10 Died
7 35 Male \- 19 5 12 79 1 2 7 9 Survived
8 44 Female DM, morbid obesity 22 7 9 90 2 4 22 26 Survived
9 26 Female \- 16 6 9 90 1 2 12 14 Survived
10 54 Male Morbid obesity 23 10 11 48 0 1 3 3 Died
11 26 Female Postpartum 16 6 8 82 1 2 9 12 Survived
APACHE-II score: Acute Physiological and Chronic Health Evaluation score, SOFA score: Sequential Organ Failure Assessment score, ARDS: Acute respiratory distress syndrome, PaO2/FiO~2~: Partial pressure of oxygen in arterial blood/fractional inspired oxygen concentration, MV: Minute ventilation, ICU: Intensive care unit, DM: Diabetes mellitus, HTN: Hypertension, COPD: Chronic obstructive pulmonary disease, CKD: Chronic kidney disease, Ca: Carcinoma
######
Ventilator parameters and blood gases before the first prone session
Patient number PBW Preprone ventilator parameters
---------------- ----- -------------------------------- ----- ---- ---- ---- ---- ----- ------ ------
1 58 90 62 49 12 31 18 360 3.5 10.6
2 60 80 48 58 14 32 16 360 3.75 12
3 52 80 57 38 12 31 18 320 3.5 10
4 56 70 100 56 11 30 28 340 3 13
5 70 80 75 38 14 31 22 420 3 11.8
6 62 70 52 40 12 31 22 370 3 10.6
7 68 65 80 57 14 32 20 400 3 10.2
8 56 80 70 49 14 32 14 340 3 12.4
9 52 75 90 40 12 30 19 320 2.5 10.2
10 60 100 50 60 16 35 14 390 3.25 13
11 52 80 80 44 14 32 14 339 3 12
PBW: Predicted body weight (kg), PaO2/FiO2: Partial pressure of oxygen in arterial blood/fractional inspired oxygen concentration, PEEP: Positive end-expiratory pressure (cmH2O), Tv: Inspired tidal volume (ml), MV: Inspired minute ventilation (L/min), LIS: Lung injury score
{#F1}
The median ICU stay and mechanical ventilation days were 15 (range: 3--26) and 12 (range: 2--22) days, respectively. In five patients, tracheal aspirate showed bacterial growth (3 -- *Acinetobacter baumannii*, 1 -- *Pseudomonas aeruginosa*, and 1 -- *Klebsiella pneumoniae*) and all survived. Apart from respiratory failure, two patients required vasopressor therapy and two had cardiovascular and renal failure, of which both patients died. Seven patients were tracheotomized. The median duration of spontaneous breathing trial from the day of prone ventilation was 8 (range: 5-16) days. The complications were observed as pressure ulcer in four patients; while back pain and critical illness neuromyopathy in one each patient. Two patients died due to shock and multiorgan failure.
DISCUSSION {#sec1-4}
==========
In this single-center, retrospective observational study, all H1N1 infection-associated severe ARDS patients showed improvement in PaO~2~/FiO~2~ ratio (\>20%) after the first prone session, with many of them undergoing multiple prone sessions also. There are multiple factors which may account for the benefit in PPV such as reduction in dependent atelectasis and pleural pressure gradient,\[[@ref14]\] alveolar recruitment,\[[@ref15]\] homogeneous ventilation,\[[@ref16]\] and relief of the compressive segment of the lung by the heart\[[@ref17][@ref18]\] and abdomen.\[[@ref19]\] All these factors cause reduction in ventilation and perfusion mismatch, intrapulmonary shunt, and ventilator-induced lung injury (VILI). These all lead to improved oxygenation and decreased PaCO~2~. The oxygenation and PaCO~2~ response to PPV are variable and depend on the type of lung injury, percentage of recruitable lung tissue,\[[@ref20]\] and duration of PPV.\[[@ref4][@ref21]\] Patients with diffuse lung injury have less nonaerated lung tissue and respond to prone ventilation by increase in oxygenation without decrease in PaCO~2~, whereas lobar lung injury has significant percentage of nonaerated recruitable lung tissue and shows marked decrease in PaCO~2~ along with improvement in oxygenation.\[[@ref15]\] Alveolar recruitment in PPV is time dependent, with different nonaerated lung units having different plateaus for complete alveolar recruitment.\[[@ref22]\]
Till 2013, a large number of randomized controlled trials were published which did not show any mortality benefit of PPV in ARDS patients, although improvement of oxygenation was seen.\[[@ref21][@ref23][@ref24][@ref25][@ref26][@ref27][@ref28]\] Many reasons have been put forth for nonimprovement in mortality in earlier studies such as lack of power in the study, too short PPV, heterogeneous patient population, or improved oxygenation does not translate to decrease mortality as multiorgan failure being the most common cause of mortality followed by hypoxemia in ALI or ARDS patients.\[[@ref29][@ref30]\]
We studied the effect of prone ventilation in severe ARDS due to H1N1, with early initiation, prolonged duration, and multiple sessions. Our study replicated the findings of Guérin *et al*. regarding the beneficial effect of PPV in severe ARDS when started early with prolonged duration.\[[@ref8]\] Worldwide, ARDS is associated with substantial mortality, with unadjusted ICU and hospital mortality of 35.3% (95% confidence interval \[CI\], 33.3%--37.2%) and 40.0% (95% CI, 38.1%--42.1%), respectively.\[[@ref5]\] H1N1 viral infection ARDS causes 36%--38% mortality in accordance with other causes of ARDS.\[[@ref3][@ref31]\]
Overall, worldwide, PPV is used in 7.9% (95% CI, 6.8--9) in all patients and 16.3% (95% CI 13.7--19.2) in severe ARDS patients.\[[@ref5]\] As with other causes of ARDS, management of H1N1 viral infection ARDS is focused on LPV strategies, and PPV is a very safe and effective way of achieving improved oxygenation and preventing VILI.\[[@ref12][@ref32]\] In recent published trials worldwide on severe H1N1 viral ARDS, only 20%, 45%, 34%, 28%, and 47% of patients were subjected to prone position ventilated before starting ECMO in Australian, French, UK, Italian, and German studies, respectively.\[[@ref2][@ref3][@ref7][@ref31][@ref33]\]
Our study differs from that of previous with regard to the optimal duration of PPV. Earlier studies with prone ventilation used short duration PPV (8 h) with no mortality benefit,\[[@ref23][@ref24]\] but later studies which used longer duration PPV showed mortality benefit when used in severe ARDS patients with high percentage of nonaerated lung unit.\[[@ref8][@ref21]\] In our studies, we used variable PPV duration (average: 14 h 52 min, range: 8--24 h) and continued until we did not find any improvement in oxygenation, decrease in PaCO~2~, or increase in tidal volume over a period of 4 h. In the present study, according to clinician\'s decision, multiple prone sessions were considered till PaO~2~/FiO~2~ ratio sustained \>150 in supine position. One patient had only one prone session, and later he died \[[Figure 1](#F1){ref-type="fig"}\].
CONCLUSION {#sec1-5}
==========
Our study demonstrates that PPV improves oxygenation when started early with adequate duration and should be considered in all severe ARDS cases secondary to H1N1 viral infection.
Financial support and sponsorship {#sec2-1}
---------------------------------
Nil.
Conflicts of interest {#sec2-2}
---------------------
There are no conflicts of interest.
Ethical conduct of research {#sec2-3}
---------------------------
This study was approved by the Institutional Review Board / Ethics Committee. The authors followed applicable EQUATOR Network (<http://www.equator-network.org/>) guidelines during the conduct of this research project.
| {
"pile_set_name": "PubMed Central"
} |
Background
==========
MicroRNAs (miRNAs) are endogenous \~22 nt RNAs that can play important regulatory roles in a variety of biological processes. They are genome-encoded, endogenous negative regulators of translation and mRNA stability originating from long primary transcripts with local hairpin structures \[[@B1]\]. Conserved seed pairing indicates that over one third of human genes appear to be conserved miRNA targets \[[@B2]\]. MiRNAs are involved in cell proliferation, intercellular signaling, cell growth, cell death \[[@B3],[@B4]\], cellular differentiation, apoptosis \[[@B5]\] and cellular metabolism \[[@B6],[@B7]\]. Meanwhile they have emerged as key post-transcriptional regulators of gene expression, and their dysregulation may lead to abnormal gene expression, which is associated to human diseases such as cancer. For example, miR-378\* (expressed from the 3\'-arm), which mediates metabolic shift in breast cancer cells, leading to a reduction in tricarboxylic acid cycle gene expression and oxygen consumption as well as an increase in lactate production, via the PGC-1β/ERRγ transcriptional pathway \[[@B8]\].
Recent studies have shown that miRNAs play important roles in energy metabolism, including glucose and lipid metabolism and amino acid biogenesis \[[@B9]\] (Table [1](#T1){ref-type="table"}). Besides, miRNAs are also able to recognize and modulate metabolic factors in transcriptional levels, relevant both in non-neoplastic and in cancer cells \[[@B10]\]. The altered metabolism of tumor cells may be a potential means to evade programmed cell death in order to favor survival and growth. The best characterized metabolic phenotype observed in tumor cells is the Warburg effect, in which the deregulation of miRNAs contributes to high glycolysis \[[@B11],[@B12]\].
######
Summary of miRNA regulation in energy metabolism
**miRNA** **Tissue / cell lines** **miRNA functions** **Target gene/Pathway** **Reference**
-------------- ------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------- -------------------------------------- ---------------------
miR-103/107 obese mice: ob/ob mice and diet-induced-obese (DIO) C57BL/6J mice regulate insulin sensitivity caveolin-1 41
miR-122 primary mouse hepatocytes and AML12 regulator of cholesterol and fatty-acid metabolism 23,26,38,49,5054,60
miR-133 293FT cells decreased GLUT4 expression and reduced insulin-mediated glucose uptake in cardio myocytes 31,34
miR-14 *Drosophila* regulate fat metabolism 36
miR-143 liver of obese mouse models impairs insulin-stimulated AKT activation and glucose homeostasis *Orp8* / *Akt* pathway 12,26,35,36, 37,91
miR-146 diabetic db/db mice islets / MIN6B1 cells cell death *Irak1* &*Traf6*/ AP-1 pathway 52
miR-15a/16-1 leukemic cell line model (MEG-01) and in primary CLL samples directly or indirectly affect apoptosis and cell cycle *MCL1, BCL2, ETS1,or JUN* 27
miR-195-5p bladder cancer T24 cells inhibited cell growth and promoted cell apoptosis through suppression of GLUT3 expression 25
miR-210 human pulmonary arterial endothelial cells (HPAECs) cellular metabolism and adaptation to cellular stress *ISCU1/2* 42
miR-23a/b human P-493 B cells regulate expression of glutaminase and glutamine metabolism *c-Myc* 6
miR-277 *D. melanogaster* a metabolic switch controlling amino acid catabolism 61
miR-27a 3T3-L1 suppress adipocyte differentiation PPARγ 51
Male C57BL/6J mice and 3T3-L1 cells a negative regulator of adipocyte differentiation
miR-29b human kidney cells (HEK293) control metabolic pathway of amino acid catabolism mRNA for DBT 62
miR-335 liver of obese mouse affects adipocyte differentiation and lipid accumulation *PPAR*γ &*aP2* 53
miR-33a/b mouse peritoneal macrophages regulate both HDL biogenesis in the liver and cellular cholesterol efflux ABCA1 58
miR-34a diabetic db/db mice islets / MIN6B1 cells sensitization to apoptosis and impaired nutrient-induced secretion *BclII* / p53 pathway 52,68
miR-370 liver of mouse affects lipid metabolism *Cpt1*α 54
miR-375 pancreatic endocrine cells (MIN6 cells) suppressed glucose-induced insulin secretion *Mtpn* 45,76
miR-378 NMuMG cells and NT2196 reduce tricarboxylic acid cycle gene expression and oxygen consumption as well as increase lactate production ERRγ and GABPA 8
INS-1E cells/primary rat islets decreased glucose-stimulatory action on insulin gene expression and DNA synthesis
Cell growth *Eef1e1* 76
cell growth *Cadm1* 76
negatively regulate cellular growth and proliferation *C1qbp* 76
regulate cell cycle and cellular proliferation *Cav1* / p53 pathway or MAPK pathway 76,79
angiogenesis and cell proliferation *Id3* / VEGF pathway or MAPK pathway 76,79
regulate cell growth/survival *Rasd1 / Ras* pathway 76,81
cell-cell signaling *Rgs16* 76
mitochondrial morphology and cristae structure, cell survival and death *Aifm1* 76
negative regulation of proliferative activity *HuD* 76,83
MiRNAs participate in controlling cancer cell metabolism by regulating the expression of genes whose protein products either directly regulate metabolic machinery or indirectly modulate the expression of metabolic enzymes, serving as master regulators. Generally, miRNA signatures may distinguish physiological, pathologic from cancerous states, which could be useful biomarkers in targeted therapeutic-diagnostics for cancer. Therefore, this review will focus on discussing the important roles of miRNA expression and deregulation in the altered metabolism in cancer cells.
MiRNAs involved in cancer cell metabolism
-----------------------------------------
The biogenesis of miRNAs is tightly associated with their action mechanism (Figure [1](#F1){ref-type="fig"}). Most miRNAs derived from independent transcription units \[[@B13],[@B14]\] and are encoded by a bewildering array of genes. Their transcription is typically performed by RNA polymerase II, with transcripts capped and polyadenylated. The resulting primary or pri-miRNA transcript extends both 5' and 3' from the miRNA sequence. The sequential processing reaction excises the stem-loop from the remainder of the transcript to create a pre-miRNA product, which occurs in the nucleus and is mostly carried out by a nuclear member of the RNase III family (Drosha). The following step excises the terminal loop from the pre-miRNA stem to create a mature miRNA duplex of approximately 22 bp length, which is carried out by the canonical Dicer enzyme in the cytoplasm. Either of the strands becomes stably associated with RNA-induced silenced complex (RISC), which can be called miRISC complex \[[@B15],[@B16]\]. The miRISC complex acts as a regulator of target gene by specially recognizing and regulating particular mRNAs to inhibit target genes \[[@B17]\].
{#F1}
A shift in glucose metabolism from oxidative phosphorylation to aerobic glycolysis was a key biochemical hallmark of tumor cells \[[@B18],[@B19]\]. The altered metabolism was called "Warburg phenomenon", which consists of an increase in glycolysis maintained in conditions of high oxygen tension and gives rise to enhanced lactate production \[[@B20],[@B21]\]. Metabolic shift in cancer cells seems to be influenced by oncogene and tumor suppressor networks \[[@B22]\]. What's more, most of these tumor suppressors are miRNA targets. For example, phosphatidylinositol 3-kinase, a lipid kinase that regulates the levels of phosphorylated phosphatidylinositol at the plasma membrane, plays a key role in cancer cell metabolism, which is targeted by miR-320, miR-123a, miR-422, miR-506 and miR-136.
There are several lines of evidence that many key molecules in cell metabolism are miRNA targets, thus giving a clue that miRNA regulates cell metabolism. Since miRNAs regulate a substantial fraction of genes in animal genomes, Tibiche and Wang systematically analyzed the human metabolic network by integrating miRNA target genes into the network \[[@B23]\]. They performed randomization tests to determine whether a multiple-gene-node is significantly regulated by miRNAs and defined 79 multiple-gene-nodes as miRNA targets. They merged the miRNA targets of single-gene-nodes with the multiple-gene-nodes, and found that 238 (22%) nodes are miRNA targets. The functional association analysis of miRNAs and metabolic pathways uncovered that miRNAs predominantly regulate central metabolic pathways such as amino acid biosynthesis, certain sugar and lipid metabolism (Figure [2](#F2){ref-type="fig"}).
{#F2}
Regulation of metabolic activity by miRNAs
------------------------------------------
MiRNAs regulate cell metabolic processes through complicated mechanisms, including directly targeting key molecules (transporters or enzymes / kinases) of metabolic processes and regulating multiple oncogenic signaling pathways (Figure [3](#F3){ref-type="fig"}). MiRNAs could directly modulate the expression of metabolic transporters or enzyme activities. In addition, MiRNAs also play pivotal roles in the expression level of transcription factors and oncogenes or tumor suppressors, including p53, c-Myc, AMPK and AKT signaling pathway.
{#F3}
The molecular mechanisms driving the Warburg effect in cancer cells were taken as an example to explain miRNA regulation in energy metabolism. As shown in Figure [3](#F3){ref-type="fig"}, several miRNAs affect gene transcription and expression of glucose transporters (GLUTs) which are responsible for transporting glucose into cytoplasm. In the initial step of glucose metabolism, glucose could be transported over a plasma membrane by GLUT3 or GLUT4 which is a target of miR-133 \[[@B24]\] or miR-195-5p \[[@B25]\]. Thus, miRNAs could directly regulate intracellular glucose levels. In the following step, the hexokinase 2 (HK2), the first rate-limiting enzyme of glycolysis, is among the top list of genes predicted and potentially regulated by multiple miRNAs including miR-143 \[[@B26]\]. Along the glycolysis reaction chain, fructose 1,6-bisphosphate is broken down into glyceraldehyde 3-phosphate and dihydroxyacetone phosphate, which is catalyzed by aldolase A (Aldo A) in the reversible aldol reaction. While the enzyme Aldo A is down-regulated by miR-15a/16-1cluster \[[@B27]\]. And the specific miRNAs in glucose metabolism will be summarized into 4 subtitles as follows, including miRNA effects on glucose uptake, glycolysis, tricarboxylic acid (TCA) cycle and insulin regulation.
On the other hand, aerobic glycolysis in tumor cells is driven by multiple miRNA-involved oncogenic signaling pathways. For example, AKT, a cardinal node in diverse signaling cascades, is regulated by miR-21 \[[@B28]\], which stimulates glycolysis by directly regulating glycolytic enzymes and activating downstream mammalian target of rapamycin (mTOR) activity. The regulation of several specifically signaling pathways involved in cancer cell metabolism by miRNAs will be introduced in detail in the second section of this review. Therefore, mRNA dysregulation in any step of metabolic processes contributes to metabolic abnormalities and even cancer development.
MiRNAs regulate glucose metabolism
----------------------------------
### MiRNAs affect glucose uptake
GLUTs (or SLC2A) are a wide group of membrane proteins that facilitate the transport of glucose over a plasma membrane in most mammalian cells. To date, 14 members of GLUTs have been identified \[[@B29]\]. The amounts of the GLUT1, GLUT2, and GLUT3 transcripts were elevated in most cancer tissues, while mRNA levels of GLUT4 and GLUT5 were below sensitivity in these cancer tissues. The potential effects of the GLUTs level seem to facilitate accelerated metabolism, high glucose requirements, and increased glucose uptake in malignant cells. Several factors have been implicated in the regulation of their expressions. Hormonal, for example, ovarian hormones, particularly estrogen, could provide a mechanism of GLUT regulation \[[@B30]\]. In addition, hypoxic also drives GLUT expression \[[@B31]\] as well as metabolic-stress-induced signaling pathways, such as adenosine monophosphate-activated protein kinase (AMPK), triggering upregulation of GLUT receptors \[[@B32]\].
MiRNAs could regulate glucose uptake via altering the GLUTs expressions. MiR-133 has been confirmed to regulate the expression of GLUT4 by targeting KLF15 in a rat model \[[@B31]\]. A study in renal cell carcinoma demonstrated that down-regulated miR-199a, miR-138, miR-150 and miR-532-5p were correlated with an increased expression of GLUT-1, whereas an increased expression of miR-130b, miR-19a, miR-19b and miR-301a can result in the down-regulation of GLUT-1 \[[@B32]\]. MiR-195-5p has been identified as a direct regulator of GLUT3 by targeting GLUT3 3'-untranslated region in bladder cancer T24 cells \[[@B33]\]. Interestingly, miR-19a and miR-133a are altered in colorectal carcinoma \[[@B34]\], and their roles in regulating GLUT expression might explain the disordered metabolism in colorectal carcinoma. In addition, miR-130b is highly down-regulated in pancreatic tumors, and its role in regulating GLUT-1 expression might explain the increased glucose uptake in pancreatic adenocarcinoma \[[@B35]\].
### Functions of miRNAs on glycolysis
Studies show that miRNAs regulate the irreversible steps in glycolysis, especially the key enzymes \[[@B33]\]. For example, miR-143, as an essential regulator of glycolysis, modulates glycolysis via targeting HK2 \[[@B12]\], which phosphorylates glucose to produce glucose 6-phosphate, thus committing glucose to the glycolytic pathway. Recently new protein targets of miRNAs have been identified by sensitive mass spectrometric studies. The oxysterol-binding-protein-related-protein 8 has been revealed as a target of miR-143 by quantitative mass spectrometry analysis \[[@B34]\]. For example, miR-155 could repress miR-143 thereby upregulating the expression of HK2 at the post-transcriptional level, except by activating the signal transducer and activator of transcription 3, a transcriptional activator for HK2 \[[@B35]\]. Besides, miR-143 inhibits the expression of HK2 both in primary keratinocytes and in head and neck squamous cell carcinoma-derived cell lines \[[@B36]\]. What's more, HK2 has been validated as a miR-143 target and thus miR-143 could affect glucose metabolism in colon cancer cells \[[@B37]\]. Likewise, miR-143 has also been identified as an essential regulator of cancer glycolysis via targeting HK2 in human lung cancer \[[@B12]\]. Interestingly, the above articles were published almost at the same time. These reports all illustrated that miR-143 targets HK2 to regulate glucose metabolism in cancer cells, and it is a potential cancer therapeutic target.
Except for targeting the irreversible rate-limiting steps, miRNAs also regulate other important intermediate steps in the glycolysis pathway. The enzyme Aldo A catalyzes a reversible aldol reaction in which fructose 1,6-bisphosphate is broken down into glyceraldehyde 3-phosphate and dihydroxyacetone phosphate. In this process, miR-122 was predicted to target Aldo A \[[@B26],[@B38]\], and the miR-15a/16-1 cluster could reduce the levels of Aldo A \[[@B27]\]. Thus miR-122 and miR-15a/16-1 cluster are involved in glycolysis in cancer cells.
### Roles of miRNAs in TCA cycle
As described before, aerobic glycolysis in tumor cells implies conversion of glucose into pyruvate and subsequently into lactic acid \[[@B39]\]. Acetyl-CoA tends to be introduced into a truncated TCA cycle, with the net result that acetyl-CoA is exported into cytosol. In this truncated TCA cycle, citrate is preferentially exported to cytosol and cleaved by ATP citrate lyase (ACL) to generate oxaloacetate and acetyl-CoA. Oxaloacetate is reduced to malate, then reimported into mitochondria and reconverted to oxaloacetate in the matrix (while generating NADH that represses the TCA cycle), and it reacts with acetyl-CoA to complete the substrate cycle.
A shift in glucose metabolism from oxidative phosphorylation to aerobic glycolysis has been accepted as a common event in cancer. This process implicates different kinds of energy production pathways are mediated by diverse regulators, including miRNAs \[[@B40]\]. For example, miR-103 and miR-107 have been predicted in regulating acetyl CoA and lipid levels in cellular systems \[[@B41]\]. Additionally, a set of miRNAs, including miR-152, miR-148a, miR-148b, miR-299-5p, miR-19b, miR-122a, miR-421, miR-494 and miR-19a \[[@B23]\], regulate the citrate synthase gene which encodes a major enzyme in TCA cycle. Besides, miR-210, a miRNA specifically induced by HIF-1α during hypoxia, represses the iron-sulfur cluster assembly proteins (ISCU1/2) \[[@B42]\]. ISCU1/2 facilitates the assembly of \[4Fe-4S\] and \[2Fe-2S\] iron-sulfur clusters, which are incorporated into the TCA cycle related enzymes, like aconitase. Thereby, the effect of miR-210 on ISCU1/2 leads to decrease the activity of TCA cycle. Furthermore, miRNAs also regulate TCA cycle indirectly by acting on transcription factors Myc and HIF etc.
### Regulation of insulin production by miRNAs
MiRNAs have shown to modulate the secretion, action and sensitivity of insulin to affect glucose uptake and production \[[@B43]\]. Insulin acts in concerting with glucagon to maintain glucose homeostasis. MiR-375 has been identified to be expressed selectively in pancreatic endocrine cell lines. The overexpression of miR-375 results in suppressed glucose-stimulated insulin secretion and its inhibition enhances insulin secretion \[[@B44]\]. Besides, *Myotrophin,* a protein implicated in exocytosis, was a validated target of miR-375. Similarly, a tantalizing new candidate target of miR-375, 3'-phosphoinositide-dependent protein kinase-1, is a key molecule in PI3-kinase signaling in pancreatic β-cells \[[@B45]\].
### MiRNAs affect lipid metabolism
Several miRNAs participate in the regulation of lipid metabolism. The deletion of miR-14 increased the levels of triacylglycerol and diacylglycerol while its overexpression resulted in the converse effect, suggesting that miR-14 acts as a regulator of fat metabolism \[[@B46]\]. Additionally, there's evidence that miRNAs are involved in the development and maturation of adipocytes from precursor cells called pre-adipocytes \[[@B47],[@B48]\]. MiR-122 could act as an important regulator of cholesterol and fatty-acid metabolism in the adult liver \[[@B49]\]. The inhibition of miR-122 in normal mice resulted in reduced plasma cholesterol levels, increased hepatic fatty-acid oxidation, and a decrease in hepatic fatty-acid and cholesterol synthesis rates, but resulted in decreased plasma cholesterol levels, a significant improvement in liver steatosis and reductions in several lipogenic genes. After that, miR-122 has been found as an important regulator in liver lipid metabolism \[[@B50]\]. In the recent article, miR-27a has been revealed to be involved in adipocyte differentiation by binding to the PPARγ 3'-UTR, whose sequence motifs are highly conserved in mammals \[[@B51]\]. All these studies indicate that miRNA plays an important role in lipid metabolism.
The changes of miRNA expression in β-cell under physiopathological conditions have been illustrated before. And at least part of the detrimental effects of palmitate on pancreatic β-cells has been caused by alteration in the levels of specific miRNAs, like miR-34a and miR-146 \[[@B52]\]. Apart from that, the up-regulation of miR-335 has also been found in obesity by microarray analysis \[[@B53]\]. Besides, the expression of miR-335 was up-regulated in liver and white adipose tissue in obese mice, which was associated with an elevated body, liver and WAT weight, and hepatic triglyceride and cholesterol. Additionally, miR-370 acting via miR-122 may accumulate hepatic triglycerides by modulating initially the expression of SREBP-1c, DGAT2, and Cpt1α and, subsequently, the expression of other genes that affect lipid metabolism \[[@B54]\].
Recently, miR-33a/b has been discovered to govern cholesterol / lipid metabolism and energy homeostasis \[[@B55]\]. MiR-33a/b embeds within intron sequences of the human SREBF genes and controls the levels of ATP-binding cassette transporter ABCA1, a cholesterol efflux pump critical for high-density lipoprotein synthesis and reversing cholesterol transport from peripheral tissues \[[@B56],[@B57]\]. MiR-33a/b also acts in the lipid homeostasis pathway by controlling the expression of fatty acid β-oxidation genes including carnitine *O*-octanoyltransferase, hydroxyacyl-CoA-dehydrogenase, and carnitine palmitoyltransferase 1A, as well as energy homeostasis regulators AMPK a1, SIRT6, and insulin receptor substrate 2 \[[@B58]\]. These reports bring us a further view of miRNA function on lipid metabolism.
### Effects of miRNA in amino acid metabolism
Amino acid metabolism is linked to biosynthesis of protein, nucleotide and lipids, redox homeostasis, and energy metabolism. MiR-23b\* (expressed from the 3\'-arm) mediates proline oxidase, the first enzyme in proline catabolism, down-regulation in human kidney tumors \[[@B59]\]. Furthermore, the metabolic link between proline and glutamine afforded by Myc emphasizes the complexity of tumor metabolism. While miR-122 was reported to downregulate the high affinity cationic amino acid transporter CAT-1 \[[@B60]\], thereby regulating amino acid metabolism. Involving evidences have been found in *Drosophila*, where miR-277 plays a role as a metabolic switch controlling amino acid catabolism by bioinformatics approaches \[[@B61]\]. Additionally, miR29b has been identified to control the component of the branched chain a-ketoacid dehydrogenase complex, which catalyzes the irreversible step in branched chain amino acids (including leucine, isoleucine and valine) catabolism \[[@B62]\], suggesting that miR-29b exerts effects of controlling on amino acid catabolism.
### miRNA regulation of signaling pathways in cell metabolism
The intertwined connections between aberrant expression of microRNAs and unbalanced signaling pathways contribute to abnormal cell metabolism and carcinogenesis. The specific p53, c-Myc, AMPK and AKT signaling pathways are included to clarify their roles in miRNA-mediated metabolism.
### p53 pathway
p53, one of the most common tumour suppressor genes, functions to prevent tumour development by inhibiting the outgrowth of stressed or damaged cells. In addition to well established functions to block cell proliferation, recent studies have revealed regulation roles for p53 involved in metabolism \[[@B63]\]. The p53 can inhibit the expression of GLUT-1, GLUT-4, phosphoglyceromutase and TIGAR to affect glycolysis. TIGAR is TP53-induced glycolysis and apoptosis regulator protein, and it inhibits the glycolytic enzyme PFKFB2 \[[@B64]\]. Additonally, p53 could also activate the expression of synthesis of cytochrome c oxidase 2 at transcriptional level \[[@B65]\] and induce the expression of the ribonucleotide reductase subunit p53R2 \[[@B66]\], leading to the restraint on glycolytic rate.
Several miRNAs are able to control p53 activity. The miR- 125b has been identified as a negative regulator of p53 in both zebrafish and human \[[@B67]\]. To date, miRNAs including miR-125b, miR-504, miR-25, miR-30d, miR-34a, miR-122, miR-29, miR-192, miR-194 and miR-215 have been shown to regulate p53 abundance and/or activity. Among these, miR-125b, miR-504, miR-25 and miR-30d negatively regulate p53 by binding to its 3\'UTR whereas the others indirectly influence p53 abundance and/or activity by regulating the regulators of p53 \[[@B68]\]. The functions of these miRNAs on p53 give a clue of their effects in cancer cell metabolism.
### c-Myc pathway
The c-Myc is a transcription factor that regulates the expression of genes involved in nucleotide metabolism, DNA replication, and ribosomal and mitochondrial biogenesis. Studies in the past few years have led to the identification of miRNAs as novel regulators of c-Myc activity. A mutated version of Myc leads to the unregulated expression of many genes, some of which are involved in cell proliferation and results in the formation of cancer \[[@B69]\]. For example, c-Myc has crucial roles in glutamine metabolism mediated by miR-23b \[[@B70]\]. Moreover, in concerts with HIF1 to regulate glucose uptake and glycolytic enzyme expression, thus favouring tumour growth in hostile environments \[[@B71]\].
The regulation of Myc mRNA by let-7a has been confirmed \[[@B72]\]. Similiarly, the overexpression of let-7a can inhibit the growth of lung cancer transplanted subcutaneously in nude mice by suppression of k-Ras and c-Myc \[[@B73]\]. Inspiringly, c-Myc transcriptionally represses miR-23a and miR-23b, resulting in increased expression of mitochondrial glutaminase, enhancing glutamine catabolism through increased mitochondrial glutaminase expression \[[@B6]\].
### AMPK pathway
AMPK acts as a metabolic master switch regulating several intracellular systems including the cellular uptake of glucose, the β-oxidation of fatty acids and the biogenesis of GLUT4 and mitochondria \[[@B74]\]. AMPK controls glucose homeostasis by regulating metabolism in multiple peripheral tissues, such as skeletal muscle, liver, adipose tissues, and pancreatic β cells \[[@B75]\].
The functions of miR-375 on glucose homeostasis have been studied \[[@B76]\]. Total 381 putative direct targets of miR-375 were selected, which contained a miR-375 recognition motif, and confirmed 10 of these genes, involving caveolin1 \[[@B77],[@B78]\], inhibitor of DNA binding 3 \[[@B79],[@B80]\], *Smarca2*, Ras-dexamethasone-induced-1 \[[@B81]\], regulator of G protein signaling 16 \[[@B82]\], eukaryotic elongation factor 1 epsilon 1, apoptosis-inducing factor, mitochondrion-associated 1, cell adhesion molecule 1, HuD antigen \[[@B83]\], and complement component 1 q subcomponent binding protein. Published data have shown that some of these genes play a role in AMPK signaling, inducing apoptosis and inhibiting normal developmental growth processes or the proliferation of tumors.
### AKT pathway
The PI3K/AKT/mTOR pathway is an intracellular signalling pathway, which is important in apoptosis. It has risen to prominence as a key regulator of cell cycle proliferation, growth, survival, protein synthesis, and glucose metabolism \[[@B84]\]. Activation of PI3K leads to the activation of downstream effectors including Akt and mTOR that support cellular biosynthesis \[[@B85]-[@B87]\]. Enhanced PI3K/Akt signal increases the expression of nutrient transporters, enabling increased uptake of glucose, amino acids, and other nutrients. Additionally, Akt-dependent stimulation of hexokinase and phosphofructokinase drives glycolysis. Furthermore, AKT-involved signaling enhances transcription of genes to involve in glycolysis and lipid genesis \[[@B88]-[@B90]\].
Regulation of this pathway by miRNAs mainly results in altered glucose and lipid metabolism. For example, miR-21, which inhibits a negative regulator PTEN of the PIK/AKT pathway, is induced in gemcitabine-resistant pancreatic cancer cells \[[@B28]\]. And AKT pathway can involve in glycolysis by directly regulating glycolytic enzymes and activating downstream mTOR activity. Similiarly, ORP8 has been identied as an miR-143 target and the reduction of ORP8 expression in cultured liver cells impairs the ability of insulin to induce AKT activation, revealing an ORP8-dependent mechanism of AKT regulation \[[@B91]\].
### MiRNAs affect multiple targets in regulatory networks
Certain miRNAs have also been shown to affect multiple targets in linear pathways or interconnected nodes in regulatory networks \[[@B25]\], thereby exerting a larger cumulative effect \[[@B9],[@B55]\]. For example, miR-33a and miR-33b, as described before, interact with the SREBP transcription factors to regulate cholesterol and lipid homeostasis. Furthermore, they may also influence insulin signaling and glucose regulation by targeting IRS2, SIRT6 and AMPK α1 \[[@B58]\]. MiR-34a, a miRNA that may have important function in a network with SIRT1 and p53, has additionally been implicated in cholesterol, lipid and energy homeostasis \[[@B52],[@B68]\]. MiRNAs typically have rather modest effects on target protein levels, and combinatorial actions on multiple functionally related targets are probably required for single miRNAs to significantly influence a complex biological process such as metabolic homeostasis.
### MiRNAs as biomarkers for human cancer
By targeting and controlling the expression of mRNA, miRNAs can control highly complex signal transduction pathways and multiple metabolic processes, which are usually involved in different oncogenic pathways \[[@B92]\]. The knowledge that miRNA expression is frequently dysregulated in cancer has uncovered an entirely new repertoire of molecular factors upstream of gene expression, with exciting potential as novel biomarkers and therapeutic targets in cancer \[[@B93]\]. Exploiting the unique characteristics of these molecules including their stability, tissue specificity, ease of detection and manipulation, will bring clinicians ever closer to achieving the goal of individualized cancer treatment \[[@B94]\].
On the one hand, miRNAs are produced in a tissue-specific manner, and changes in miRNA within a tissue type can be correlated with disease status. The tissue concentrations of specific miRNAs have been associated with tumor invasiveness, metastatic potential, and other clinical characteristics for several types of cancers, including chronic lymphocytic leukemia, and breast, colorectal, hepatic, lung, pancreatic, and prostate cancers \[[@B95]\]. On the other hand, there has been an accumulating body of evidence to support circulating miRNAs as non-invasive, sensitive biomarkers of disease states, particularly cancers (breast, lung, pancreas, ovarian, and prostate) \[[@B96]\]. For example, miR-9 and miR-9\* (expressed from the 3\' mature sequence), mostly neuronal and thus expressed in central nervous system tumors but absent in other tumors, present their potential as tumor markers \[[@B97]\]. In addition, the reduced levels of miR-126, members of the miR-17-92 cluster, inflammation-related miR-155, and smooth muscle-enriched miR-145 in patients with coronary artery disease compared with healthy controls \[[@B98]\]. What's more, published data showed that plasma miR-29a and miR-92a have strong potential as novel noninvasive biomarkers for early detection of colorectal carcinoma \[[@B99]\].
Furthermore, since they are abundant in blood, easy to measure, highly stable and disease associated, serum microRNAs are attractive disease biomarkers \[[@B100]\]. There have been over 200 publications on circulating miRNA in cancers including prostate, breast, colon, lung, ovarian and leukemia since 2008. Considering the sources of variation, state of microRNA in plasma and origin and implications for disease specificity, miRNA expression profiles of potential patients could be assessed by measuring circulating miRNAs in patient serum. This profile could be hopefully used for early detection of cancer.
Conclusion and perspective
==========================
MiRNAs are important regulators of numerous aspects of metabolic homeostasis, physiology and disease. In general, miRNAs could mainly have two ways to regulate cellular metabolism. MiRNAs could regulate transcription factors or signaling proteins, which in turn regulate metabolic enzymes. Alternatively, miRNAs could regulate the production of certain metabolites by directly regulating the genes that encode metabolic enzymes \[[@B101]\]. In addition, miRNAs could regulate mRNAs through chromatin remodeling \[[@B102]\]. The emergence of miRNAs as important regulators of metabolism has garnered much interest not only from a scientific point of view but also from a clinical perspective. The function of miRNAs on cellular metabolism reveals molecular strategies for controlling metabolic flux by miRNAs in living organisms, thus lighting up one aspect of miRNA therapeutics. MiRNAs are promising in the diagnosis of cancer, drug target identification and clinical treatment in the future (Figure [4](#F4){ref-type="fig"}). The use of miRNAs, such as oligonucleotide complementary \[[@B103]\] or antisense oligonucleotides \[[@B104]\] in miRNA inhibition, to suppress cell metabolism altering will hopefully lead to a new therapeutic strategy for malignant cancer \[[@B105],[@B106]\]. For example, endothelial miR-126 is deregulated in patients with type 2 diabetes, which may ultimately lead to novel biomarkers for risk estimation and classification and could be exploited for miRNA-based therapeutic interventions of vascular complications associated with this disease \[[@B107]\].
{#F4}
So far, a variety of new strategies to identify and characterize the targets of individual miRNAs have been developed. Because miRNAs can also regulate other non-coding RNAs, these interactions will increase the complexity of gene regulation. Moreover, cost-effective miRNA profiling strategies and larger studies are needed to determine its advantage for cancer diagnosis. Additionally, a new class of miRNA-based drugs that are capable of targeting molecules outside the range of traditional medicinal chemistry, their clinical implementation will require improvements in drug composition and delivery. Since these challenges lie on the way, molecular strategies for cancer therapy by miRNAs are still in their infancy. Nevertheless, the successful development of miRNA biology technologies could ultimately translate our understanding of miRNA functions in cancer into strategies for the control of cancer.
Abbreviations
=============
ACL: ATP citrate lyase; AldoA: Aldolase A; AMPK: Adenosine monophosphate-activated protein kinase; GLUT: Glucose transporter; ISCU1/2: Iron-sulfur cluster assembly proteins; HIF1: Hypoxia-inducible factor 1; HK2: Hexokinase 2; MiRNA: MicroRNA; MTOR: Mammalian target of rapamycin; ORP: Oxysterol-binding-protein-related-protein; RISC: RNA-induced silenced complex; TCA: Tricarboxylic acid.
Competing interests
===================
The authors declare that they have no competing interests.
Authors' contributions
======================
All authors participated in the preparation of the manuscript, read and approved the final manuscript.
Acknowledgements
================
This work was financially supported by the grants from National Key Basic Research Program of China (2011CB910703, 2013CB911303), National Natural Sciences Foundation of China (30970654, 31071235), and grants for New Century Excellent Talents in University (NCET-10-0595), specialized research fund for the Doctoral Program of Higher Education (20120181110025). This research was also funded by Sichuan Province Program (2010JQ0016, 2012SZ0002) and Chengdu Local Scientific Project (11DXYB356JH-027).
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
In recent years, there has been intense research on the microscopic interactions of energetic radiation with organic and biological relevant molecules. An important motivation is the application of radiation in medical treatment like radiation therapy and the desire to understand the underlying mechanisms and possibly to improve its effectiveness^[@CR1]^. The primary X-rays or swift charged particles, while penetrating biological tissue, produce large numbers of secondary electrons that in turn cause cellular damage either directly via ionization or indirectly by producing radicals such as OH in the aqueous environment^[@CR2]--[@CR5]^. To get insight into the reaction mechanisms, on one hand, gas-phase experiments were conducted on building blocks of macro-molecules such as DNA or proteins, where bulk effects do not mask the intrinsic molecular properties. On the other hand, one has to recognize the influence of the natural environment on biomolecules, in particular through hydrogen bonding, which can modify their structure and functionality. Examples are the hydrogen bonds between DNA base pairs linking both strands and the structural water molecules, which are hydrogen-bonded to the DNA with roughly 22 hydration sites per base pair^[@CR6]^.
Consequently, a number of studies on pure and nano-hydrated biomolecular clusters have been performed using different mass spectrometric techniques for both cations and anions^[@CR7]--[@CR16]^. The effect of solvation on the fragmentation of biomolecules after collisions with various projectiles was studied by measuring the yields of different fragment ions. A general observation was that although monomer ionization results in a large number of fragmentation channels for larger clusters, essentially only cations with integer number of intact molecules were found in the mass spectra. It was concluded that the environment has an overall "protective" effect on the systems and that the cluster environment acts as a buffer that rapidly redistributes excess energy, leading to suppression of molecular dissociation in clusters. On the other hand, for the smallest clusters, i.e., for dimers and trimers, some new molecular fragment species can be identified in the mass spectra, which are not consistent with the aforementioned protection effect^[@CR9]--[@CR12]^. The authors did not discuss or clarify the formation mechanisms of these species. Therefore, in the present study we go beyond pure fragment mass measurements and identify the initial ionized states from which the fragmentation process starts. This information accompanied by high-level quantum-chemistry calculations gives insight into the molecular geometry evolution and intermediate transition states (TS), which must be overcome.
We investigate the electron-collision-induced ionization and dissociation processes in clusters consisting of water and tetrahydrofuran (THF). Here, the THF (C~4~H~8~O) molecule is considered as a molecular analog of the deoxyribose sugar-ring in the DNA backbone (see Fig. [1a](#Fig1){ref-type="fig"}). It provides a simple model to probe possible mechanisms of electron-induced deoxyribose decomposition^[@CR17]--[@CR19]^. Water (H~2~O) is the predominant medium in which biological chemistry takes place^[@CR20]^. An accurate description of the energetic and structural aspects of the interaction of water with biomolecules is essential for a better understanding of their functions in biological processes^[@CR21]--[@CR28]^. The present study aims to understand how the reaction properties of the isolated THF molecule are affected when a water or second THF molecule is attached to it via hydrogen bonding, to mimic a chemical environment. We elucidate the electron-driven fragmentation dynamics of hydrated and pure THF clusters, i.e., H~2~O·THF and THF·THF dimers (Fig. [1b, c](#Fig1){ref-type="fig"}) in comparison with the isolated THF molecule.Fig. 1Chemical structure of the studied systems and schematic of the ring-break process.**a** A section of DNA containing the four bases and the sugar-phosphate backbone. **b**, **c** Chemical structure of H~2~O·THF (**b**) and THF·THF (**c**) dimers for the hydrated and pure THF model systems. **d** Schematic of the electron-induced ionization and ring-break process in the H~2~O·THF dimer. Here ionization of THF initiates significant rearrangement of the dimer structure resulting in THF ring opening and finally in THF ring-break. In **b**, **c**, **d** the white, gray, and red balls represent to hydrogen, carbon, and oxygen atoms, respectively. In **d**, the green balls labeled e~1~ and e~2~, and the green lines indicate electrons and their trajectories, respectively.
Experiments were carried out using a multi-particle (electrons plus ions) imaging spectrometer with a supersonic gas jet target and a pulsed low-energy electron beam^[@CR29],[@CR30]^. The projectile energy of 65 eV was chosen close to the mean energy of secondary electrons produced by high-energy primary radiation in a condensed medium such as water^[@CR4]^. We find that the removal of an electron from the highest occupied molecular orbital (HOMO), which leads to a stable parent ion for the THF monomer, for the dimer initiates a THF ring-break reaction. Our ab-initio calculations show that THF ring-break after HOMO ionization requires geometrical changes via several TSs, including those requiring structural rearrangement like ring-opening and proton transfer (PT). The highest barrier to overcome is a C--C bond-break for the ring-opening. For the THF molecule embedded in a THF·THF or H~2~O·THF dimer, the energy of the respective TS is reduced in comparison with the isolated THF molecule. This reduced barrier can activate ring-break for the dimers, while this channel is closed in the isolated THF molecule. In addition, PT takes place during the molecular relaxation, which releases some amount of internal energy to the system. As a consequence, the cluster cation finally dissociates, i.e., H~2~O·THF^+^ → H~2~O·C~2~H~4~O^+^ + C~2~H~4~, (see in Fig. [1d](#Fig1){ref-type="fig"}). These observations reveal a so-far unnoticed role of the water environment in enhancing the ring-break of the THF molecule after ionization of the HOMO. It can be inferred that noncovalent hydrogen bonding can considerably weaken the covalent bonds in a neighboring molecule. This can be important for a better understanding of the reaction mechanisms concerning ionizing radiation in biological matter^[@CR3],[@CR5]^.
Results {#Sec2}
=======
Sample composition and characterization of reaction products {#Sec3}
------------------------------------------------------------
In our experiments, two kinds of gas jet targets were employed, a pure THF jet containing about 10--15% dimers THF·THF and a mixed THF water jet with about 4% THF dimers and 4% H~2~O·THF dimers. The abundance of larger clusters goes down by a factor of roughly 5 for each additional molecule (these numbers are discussed in Methods). Therefore, most of the ionizing collisions concern monomers and the identification of dimer ionization processes is according to the characteristic mass of the ionic fragment species. For each ionizing collision, the ion and the two outgoing electrons are detected in coincidence. During offline analysis, the mass-over-charge ratios, the momentum vectors and the kinetic energies for all three charged particles are determined (see Methods). In case there is not more than one neutral fragment, its momentum can also be reconstructed from the measured momenta and momentum conservation, and the measurement is kinematically complete. We deduce the correlation of the ionic fragment species with the ionized electron's binding energy (BE) from which the ionized orbital is identified. Here, the BE *E*~*b*~ is determined as the initial projectile energy *E*~0~ minus the sum energy of the two final state electrons *E*~1~ + *E*~2~, i.e., *E*~*b*~ = *E*~0~ − (*E*~1~ + *E*~2~). *E*~*b*~ constitutes the vertical transition energy between the electronic ground state and an ionized state of the molecule.
Fragment mass spectra for monomers and clusters {#Sec4}
-----------------------------------------------
The measured mass spectra of pure and hydrated THF clusters are presented in Fig. [2](#Fig2){ref-type="fig"} in the range from 10 to 150 u (atomic mass units). The spectra are normalized at mass 72 u corresponding to the intact C~4~H~8~O^+^ cation. Hydrogen abstraction from THF monomers gives rise to C~4~H~7~O^+^. Ring-break reactions in THF monomers yield the ions C~2~H~*n*~^+^, C~3~H~*n*~^+^, and C~2~H~*n*~O^+^ assigned in Fig. [2](#Fig2){ref-type="fig"}. The fragmentation of the THF ion was studied before by us and other groups^[@CR12],[@CR31]^. Briefly, ionization of the HOMO gives rise to a stable parent ion while hydrogen abstraction occurs for HOMO-1 ionization which is 1.5 eV above the ionic ground state. The behavior of the ring-break channels complies with an unimolecular statistical decay: electronically excited states produced by ionization of inner orbitals quickly evolve to the ionic ground state (internal conversion) giving rise to vibrational excitation. The excess energy of 2.5 eV is sufficient for ring-opening and subsequent dissociation of the resulting linear molecule^[@CR31]^. For higher internal energies, additional hydrogen atoms can be abstracted which manifests as series of lines in the mass spectrum, e.g., $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\mathrm{C}}_3{\mathrm{H}}_n^ +$$\end{document}$ with *n* = 3--6.Fig. 2Mass-over-charge spectra.The measurements are for pure (green line) and hydrated THF clusters (magenta line) upon electron ionization. The peaks are labeled with the assigned fragment formula. Both spectra include ion fragments from THF monomers. These are the lines in the mass region from 24 to 46 u, as well as the C~4~H~7~O^+^ and C~4~H~8~O^+^ ion signals. The spectra are normalized at mass 72 u corresponding to the intact C~4~H~8~O^+^ cation. The experiment has no detection efficiency in the mass region from 49 to 58 u. The branching ratios of fragment species can be seen from Supplementary Note [6](#MOESM1){ref-type="media"}. Source data are provided as a Source Data file.
The fragments that can be assigned to originate from clusters in the THF jet are the protonated cations (C~4~H~8~O)~*m*~·H^+^ for *m* = 1, 2 and C~4~H~8~O·C~2~H~4~O^+^ from a ring-break reaction. For hydrated clusters, we observe the protonated ions (H~2~O)~*n*~·C~4~H~8~O·H^+^ for *n* = 1--3 and also H~2~O·C~2~H~4~O^+^, which is the equivalent ring-break reaction as the one seen in the THF dimer. These reactions will be discussed in more detail below.
It can also be seen from Fig. [2](#Fig2){ref-type="fig"} that the fragmentation patterns in the mass region from 24 to 46 u and the C~4~H~7~O^+^ ion signals do not change significantly between the results of pure and hydrated THF clusters. This suggests that they are mainly attributed to the fragmentation processes of THF monomers.
Identification of the ionized molecular orbitals {#Sec5}
------------------------------------------------
In the following, we focus on the channels giving rise to the complete ring-break of THF molecule in monomers and clusters, i.e., C~2~H~4~O^+^, H~2~O·C~2~H~4~O^+^, and C~4~H~8~O·C~2~H~4~O^+^ fragments. Although this is the only ring-break reaction found for dimers, it is a rather minor channel for the monomer and superimposed by additional H-loss, namely C~2~H~3~O^+^. Figure [3](#Fig3){ref-type="fig"} presents the measured BE spectra for the three channels. Also included in the figure are the BE spectra of C~4~H~7~O^+^ and C~4~H~8~O^+^ channels from monomers. These show peaks located at 9.8 and 11.6 eV (see Fig. [3d](#Fig3){ref-type="fig"}), respectively. The peak width of about 3.1 eV (full-width at half maximum, FWHM) is mainly attributed to the experimental BE resolution (2.9 eV, FWHM), which was determined by a measurement with He gas. The measured BE peak locations are in good agreement with previous studies of ionization and dissociation of THF monomer^[@CR31],[@CR32]^ showing that the HOMO ionization of THF leads to the intact C~4~H~8~O^+^ ion, whereas the so-called *α*-cleavage reaction channel, i.e., C~4~H~7~O^+^ or (THF-H)^+^, originates mainly from ionization of the THF HOMO-1 orbital. As mentioned above the ring-break reaction in the monomer requires significantly more energy and the BE for C~2~H~4~O^+^ is broader and peaking around 13 eV in Fig. [3c](#Fig3){ref-type="fig"} corresponding to the ionization of the HOMO-3 and HOMO-4 orbitals. Interestingly, the same ring-break reactions in the THF·THF and H~2~O·THF dimers, which give rise to the C~4~H~8~O·C~2~H~4~O^+^ and H~2~O·C~2~H~4~O^+^ ion fragments, show significantly lower BE with peaks located at about 9.8 and 10.4 eV, respectively (Fig. [3a, b](#Fig3){ref-type="fig"}). This indicates that these fragments are the result of HOMO ionization. Here the BE difference for the H~2~O·C~2~H~4~O^+^ channel, which is about 0.6 eV higher than the HOMO ionization energy of THF (\~9.8 eV), is likely due to the modified HOMO BE in hydrated dimers. Our calculated vertical ionization energies (VIEs) are shown in Fig. [4](#Fig4){ref-type="fig"}. The VIE of the H~2~O·THF dimer (9.95 eV) is about 0.43 eV higher than the VIE of THF monomer (9.52 eV) using the same quantum-chemistry method (Fig. [4a and b](#Fig4){ref-type="fig"}). The measured BE spectrum for C~4~H~8~O·C~2~H~4~O^+^ fragment channel is nearly the same as the intact C~4~H~8~O^+^ channel of THF monomer. This is in agreement with our calculation obtaining a VIE of THF·THF dimer (9.07 eV) close to the VIE of the THF monomer (9.15 eV) using the same method as for the THF·THF dimer (see Supplementary Fig. [4](#MOESM1){ref-type="media"}). Thereby, we can conclude that the two ring-break channels, i.e., C~4~H~8~O·C~2~H~4~O^+^ and H~2~O·C~2~H~4~O^+^, are formed upon ionization of the HOMO in the THF site of dimers. Whereas in the monomer such ring-break channel, i.e., C~2~H~4~O^+^, is associated with the ionization of HOMO-3 and HOMO-4 orbitals^[@CR32]^, as seen in Fig. [3c](#Fig3){ref-type="fig"}.Fig. 3Measured binding energy spectra for various fragment species.**a**, **b** ionization of THF·THF (**a**) and H~2~O·THF (**b**) dimers and subsequent dissociation to C~4~H~8~O·C~2~H~4~O^+^ and H~2~O·C~2~H~4~O^+^ channels, respectively. **c**, **d** ionization of THF monomer and subsequent dissociation to C~2~H~4~O^+^ (**c**) and C~4~H~7~O^+^, and non-dissociated C~4~H~8~O^+^ (**d**) channels. The vertical lines on the top of the figure are the valence orbital ionization energies of the THF monomer. HOMO refers to the highest occupied molecular orbital. The statistical error bars shown correspond to the 1*σ* confidence interval. Source data are provided as a Source Data file.Fig. 4Calculated potential energy levels (eV) of ionized pure and hydrated THF systems.**a**--**c** The energetics of THF^+^ (**a**), H~2~O·THF^+^ (**b**), and THF·THF^+^ (**c**) relative to the neutral ground state. Upon ionization of the HOMO, H~2~O·THF^+^ and THF·THF^+^ can relax to the highest transition state (TS), i.e., TS-1, and then rearrange significantly the molecular structures involving particularly proton transfer (PT) and finally dissociate into H~2~O·C~2~H~4~O^+^ and C~4~H~8~O·C~2~H~4~O^+^ ion fragments, respectively, and a C~2~H~4~ neutral part. For THF^+^, the activation energy (*E*~*a*~) from the vertical ionization point to the TS-1 state is much higher compared with the dimer systems. Thereby, the subsequent dissociation processes are not likely for the isolated THF^+^ cation. FC refers to the Franck--Condon region, which is marked by the yellow bar. All energies include zero-point vibrational energy corrections. For comparisons between THF monomer and THF·THF dimer, the energy levels in THF monomer are recalculated using the same quantum-chemistry method as for the THF·THF dimer which is shown in Supplementary Fig. [4](#MOESM1){ref-type="media"}. The white, gray, and red balls represent hydrogen, carbon, and oxygen atoms, respectively. The PT process is shown by the circled hydrogen and arrow.
Theoretical analysis of the relevant reaction pathways {#Sec6}
------------------------------------------------------
To uncover the underlying mechanism of the present observations, the TSs that are passed through in the progressing reaction after ionization of the THF HOMO are analyzed using high-level ab-initio calculations (see Methods and Supplementary Notes [1](#MOESM1){ref-type="media"} and [2](#MOESM1){ref-type="media"}). We calculated the potential energy surface (PES) of various intermediates to determine the reaction pathway using a relaxed PES scan. The reaction pathway is further confirmed by an intrinsic reaction coordinate calculation. The energy levels of various intermediate stationary points are illustrated in Fig. [4a--c](#Fig4){ref-type="fig"} for the three systems that are calculated using the coupled cluster single-double and perturbative triple \[CCSD(T)\] method with aug-cc-pVTZ basis set for the systems of THF monomer and H~2~O·THF dimer. For the THF·THF dimer we had to use a more simple basis set of cc-pVDZ because of the computational complexity of this system. All energies include the zero-point vibrational energy correction and the ionization energies are obtained for the full Franck--Condon (FC) region (see Supplementary Notes [3](#MOESM1){ref-type="media"} and [4](#MOESM1){ref-type="media"}).
The calculations show that the reaction can occur only if the highest TS corresponding to the opening of C~*β*~--C~*β*~ bond, which is in the ring opposite the oxygen atom, is overcome in the first step (TS-1 state, in Fig. [4a--c](#Fig4){ref-type="fig"}). Once the cation reaches this TS-1 state, there is significant rearrangement of the molecular structure during the molecular relaxation leading to the complete breakup of THF ring structure. We emphasize particularly the important role of PT in the ring-break process. After TS-1, PT is taking place from the bond position of C~*α*~ to C~*β*~ of the ionized THF site (see Fig. [4](#Fig4){ref-type="fig"}). This releases some amount of internal energy to the system, which is about 1.74 eV for H~2~O·THF^+^ and 1.0 eV for THF·THF^+^. The released energy accelerates the final dissociation of the cations.
It is shown in Fig. [4](#Fig4){ref-type="fig"} that the energy levels of cations are strongly dependent on whether the THF molecule has a hydrogen-bonded neighbor, i.e., if it is embedded in a chemical environment. The calculated activation energy (*E*~*a*~), corresponding to the required energy from vertical transition point at equilibrium geometry to the highest TS, is about 1.8 eV for the isolated THF^+^. This energy barrier is too high for THF^+^ to reach the TS-1 even considering the energy width of the FC region. Thus, the dissociation channel is completely closed in the THF monomer upon ionization of the HOMO. The activation energy is reduced by a factor of about two in the dimer systems, i.e., *E*~*a*~ = 1.13 eV and 0.85 eV for H~2~O·THF^+^ and THF·THF^+^, respectively. Part of this reduction is due to the energy gain from a rearrangement of the dimer ion geometry following ionization. This is because the THF HOMO corresponds to the lone pair orbital of oxygen which participates in the hydrogen bonding with the neighbor and after ionization the dimer equilibrium geometry changes considerably. Therefore, the rearranging ion is in a vibrationally excited state and this internal energy facilitates reaching the TS-1. For the THF dimer after ionization there is not only rearrangement but also PT to the neighboring molecule, which contributes to the energy gain. Here, one should consider that the internal energy after ionization depends on the considered geometry of the neutral dimer. If we assume for the THF dimer geometry instead of the present stacked arrangement a T-like geometry as it was obtained by some groups^[@CR33]^, then a vertical transition energy of 9.71 eV is obtained which is even closer to the TS-1 energy. The FC region is shown by the yellow bands in Fig. [4](#Fig4){ref-type="fig"}. The vertical transition point at the high edge of the FC region is also comparable in energy with the TS-1 in both H~2~O·THF^+^ and THF·THF^+^ systems, indicating that the ring-break processes are more likely to occur in dimers.
Discussion {#Sec7}
==========
Our experimental technique as well as our calculations do not allow all possible ionization channels of the dimer systems to be examined comprehensively. Nevertheless, the current results demonstrate that embedding a molecule into a chemical environment like water can increase the biological effectiveness of ionization-induced damage to the molecule. In the present work we mainly consider small systems with only one hydrogen-bonded neighboring molecule for which high-level ab-initio calculations can be performed. By using THF as a model system of the DNA sugar-ring structure it is shown that the ring-breaking channel is activated when a water molecule is attached to THF via hydrogen bonding and an electron is removed from the HOMO of THF. This suggests that water acts as a catalyst for damaging the ring structure of THF initiated by electron ionization. Through high-level ab-initio calculations we found that the potential energy levels of various TSs are significantly modified by the water environment. This induces geometrical rearrangement and via several intermediate states finally leads to the dissociation channel. We noticed that during the molecular rearrangement PT takes place and releases internal energy to the system, which is a key point to accelerate the ring-break of the system. Similar mechanisms were found to act also in the pure THF dimer. This implies that the stability of molecular covalent bonds can be seriously affected by adding a hydrogen-bonded molecule to the system. It is also to be noted that our measurements on other hydrogen-bonded organic systems such as ethanol dimers show a similar behavior. Although HOMO ionization in the monomer produces stable cations, in the dimer C--C bond breaking in one ethanol molecule is activated (see Supplementary Note [7](#MOESM1){ref-type="media"}). This indicates that the present observation might be a general phenomenon in the ionization-induced fragmentation of hydrogen-bonded systems.
Finally, we consider a more realistic environment in the energy calculations by adding more water molecules (*n*H~2~O·THF, *n* up to 4) to THF and, furthermore, we include the solvent effect using the polarizable continuum model (see Supplementary Note [5](#MOESM1){ref-type="media"}). It was found that in both cases the activation energy is reduced in comparison with THF monomer and stays almost constant for the larger hydration numbers (*n* \> 2) (see Supplementary Fig. [5](#MOESM1){ref-type="media"}). This means that the catalysis effect energetically may still be open in the micro-solvated environment.
These findings at first sight contradict existing studies, which have found that solvation shells protect biomolecules from fragmentation^[@CR8],[@CR10],[@CR12]--[@CR16]^. However, in the systems studied here, the mechanism reducing the energy barrier to the ring opening partly is the population of vibrationally excited states of the dimer ion. If the THF molecule is embedded in a complete solvation shell, the vibrational energy can be redistributed among more degrees of freedom or even be dissipated by evaporating water molecules such that the ring-break reaction is quenched. Therefore, the protection effect may appear to compete with the catalysis effect and become dominant. The lack of larger hydrated C~2~H~4~O^+^ species such as (H~2~O)~2~·C~2~H~4~O^+^ and (H~2~O)~3~·C~2~H~4~O^+^ at the mass-over-charge values of 80 and 98, respectively, in Fig. [2](#Fig2){ref-type="fig"} is a signature of such protection effect in our experiment. Thus, the occurrence of the THF ring-break reaction strongly depends on the spatial geometry and mobility of the molecules in the local environment. These results could have implications for our understanding of ionization damage in biological matter.
The hydrogen bonding is an important noncovalent interaction ubiquitous in nature from base-pair interactions in DNAs and sophisticated supramolecular assemblies to the dense and cold molecular clouds in outer space and planetary atmospheres^[@CR34]--[@CR36]^. The present electron-collision induced reactions in molecules are common phenomena in many fields of science and technology, in the gas phase and in the condensed phase or at interfaces. Our studies concerning THF·THF and H~2~O·THF dimers clearly demonstrate that hydrogen-bonded molecules can considerably affect the neighboring molecule and play the role of a catalyst for the break-up of its covalent bonds.
Methods {#Sec8}
=======
Coincident ion and electron detection {#Sec9}
-------------------------------------
The experimental data were obtained by crossing an electron projectile beam with a gas target jet and employing a multi-particle coincidence spectrometer (reaction microscope) for detection of the charged collision fragments^[@CR29],[@CR30]^. The pulsed electron beam is generated in an electron gun from a tantalum photocathode, which is irradiated by UV-light pulses of 0.5 ns duration and it is collimated by electrostatic lens elements. The energy width of the electron beam is about 0.5 eV^[@CR31]^. It is guided by an axial magnetic field (0.7 mT) to the crossing zone with the gas jet and further to a beam dump which is a central bore in the electron detector. Secondary electrons and ions are extracted by means of a homogeneous electric field to opposite directions and projected onto two position- and time-sensitive multi-channel plate detectors with 80 mm diameter of the active area. From the impact positions and the times-of-flight, the momentum vectors and consequently the kinetic energies of the particles emerging from the reaction are determined. The acceptance angle for detection of low-energy electrons up to the kinetic energy of 15 eV is almost 4*π* with exception of small forward and backward angles, which are lost due to the detector bore. Projectile electrons that have undergone ionizing collisions are detected for scattering angles up to about 35°. To maximize the acceptance for molecular ion fragments, the electric extraction field of 1.0 V cm^−1^ is ramped up to 20 V cm^−1^ after 400 ns when the electrons have reached the detector. In our experiment, monomer ionization is simultaneously recorded with cluster ionization.
Cluster production {#Sec10}
------------------
The pure and hydrated THF clusters are generated in a supersonic expansion of helium gas (stagnation pressure 1 bar), which is seeded with pure THF vapor or mixed water and THF vapor. To pick up the target molecules, the helium gas is guided through one or two reservoirs containing liquid THF and water. For production of pure THF clusters, only the THF reservoir is used. The hydrated clusters are created with the two reservoirs containing water and THF at temperatures of 60 and 25 °C (room temperature), respectively. The gas mixture is expanded through a 30 μm nozzle, which is heated to 100 °C into the vacuum. The gas beam is collimated by two skimmers with 200 μm diameter aperture at their apex, and located ∼3 mm and 20 mm downstream from the nozzle. The relative fraction of pure and mixed THF dimers in the jet as given in the text was estimated from the relative intensities of the respective ion species produced by HOMO ionization. For the THF monomer, this is C~4~H~8~O^+^. The related line in the TOF spectrum in Fig. [2](#Fig2){ref-type="fig"} is comparatively narrow showing that this ion is not resulting from dissociation of clusters. For the THF~2~ dimer, the ion species produced by HOMO ionization is C~4~H~8~O·C~2~H~4~O^+^ and for the H~2~O·THF dimer this is H~2~O·C~2~H~4~O^+^.
Quantum-chemistry calculations {#Sec11}
------------------------------
The calculations were carried out using the Gaussian package^[@CR37]^. The ground-state equilibrium geometries of the isolated THF molecule and the hydrogen-bonding H~2~O·THF dimer were optimized with the second-order Møller--Plesset method using the aug-cc-pVTZ basis set. For the THF·THF dimer, due to the computational complexity of this system, the Becke's three-parameter hybrid functional combined with Lee--Yang--Parr correlation functional (B3LYP) method was used together with a cc-pVDZ basis set. The neutral and singly charged electronic energies were determined using a CCSD(T) method with a aug-cc-pVTZ basis set for the systems of THF monomer and H~2~O·THF dimer and a cc-pVDZ basis set for THF·THF dimer. The TSs and zero-point energy corrections were determined by the B3LYP method using the aug-cc-pVTZ basis set for THF, H~2~O·THF and the cc-pVDZ basis set for THF·THF. The VIE in the FC region were calculated by outer-valence Green's function method considering the zero-point vibration by quantum harmonic oscillator distribution.
Supplementary information
=========================
{#Sec12}
Supplementary Information Peer Review File
Source data {#Sec13}
===========
Source Data
**Peer review information** *Nature Communications* thanks the anonymous reviewers for their contribution to the peer review of this work. Peer reviewer reports are available.
**Publisher's note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
=========================
**Supplementary information** is available for this paper at 10.1038/s41467-020-15958-7.
This work was supported by the National Natural Science Foundation of China under Grants Numbers 11774281 and 11974272, and by the Deutsche Forschungsgemeinschaft (DFG) under project number RE 2966/3-1. E.W. acknowledges a fellowship from the Alexander von Humboldt Foundation.
X.R. and A.D. conceived and designed the experiments. X.R. performed the experiments and analyzed the data. E.W. carried out the quantum-chemistry calculations. X.R. and E.W. wrote the first draft of the manuscript. W.B., H.R., and T.P. contributed to the interpretation of the data and edited the manuscript. All authors edited and approved the final version of the manuscript.
The source data underlying Fig. [2](#Fig2){ref-type="fig"} and Fig. [3a--d](#Fig3){ref-type="fig"} are provided as a Source Data file. The data supporting this study are also available from the corresponding author upon reasonable request.
The authors declare no competing interests.
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION
============
Helping behavior has been observed throughout the animal kingdom, from social insects to primates ([@R1]). Rescue behavior observed in ants can arise in predator-prey interactions, by rescuing nestmates that have fallen into an antlion trap by digging, pulling the ant out and attacking the antlion, or excavating ants trapped under sand or soil ([@R2]--[@R5]). All hitherto observed types of rescue behavior in social insects were always directed toward individuals under an imminent threat ([@R1], [@R6], [@R7]), that is, suffocation or being eaten. *Megaponera analis* is a strictly termitophagous ponerine ant species, found in sub-Saharan Africa from 25°S to 12°N ([@R8]) that specializes in raiding termites of the subfamily Macrotermitinae at their foraging sites ([@R9]--[@R13]). A scout ant that has returned to its nest after having found an active termite foraging site initiates a raid. It will recruit approximately 200 to 500 nestmates and lead them to the termites in a column-like march formation, which can be up to 50 m away from the nest ([@R11], [@R13], [@R14]). During the raid, division of labor occurs ([@R15]): larger ants (majors) break open the protective soil cover created by the termites, whereas the smaller ants (minors) rush into these openings to kill and pull out the prey ([@R16]). Afterward, the majors collect the dead termites, the column forms again, and the hunting party returns to the nest. These raids occur two to four times a day ([@R9], [@R11]--[@R13], [@R17]). Termites have evolved various ways to defend themselves effectively against predators such as *M. analis*, of which a specialized soldier caste with strong sclerotized heads and big mandibles is the main defensive force ([@R18], [@R19]). Consequently, ants involved in the hunting process incur high injury risks. We observed a unique helping behavior in *M. analis* to compensate for this high injury rate by carrying back injured ants to the nest. The carrying of ants after the hunt was also observed in Kenya ([@R13]) and the Democratic Republic of Congo ([@R20]); however, no attempt was made in those studies to explore the adaptive value of this behavior to the colony or the individual. We further observed the removal of termites, still clinging on to ant extremities in the ant nest, and the rescue behavior toward ants that carry long-term injuries in the form of lost extremities. This specialized rescue behavior is unanticipated in insects, where the value of individuals is generally underestimated, and could provide further proof that empathy is not necessary for helping behavior to emerge in animals ([@R21]).
RESULTS
=======
Injured *M. analis* ants were antennated by their nestmates at the hunting ground, whereupon they adopted a pupal pose, most likely for ease of transportation back to the nest (movie S1 and fig. S1A). On an average raid, a median of 3 ± 2.9 ants (of 416 ± 153 ants) were carried back (*n* = 53 raids with 154 carried ants), for a total of 9 to 15 rescued ants per day (3 to 5 raids per day). Only in 11% (6 of 53) of the raids were no ants carried back to the nest, and in half of those cases, the raid itself was unsuccessful (no encounter with termites at the hunting ground). If we consider a mean estimated birth rate of 13.3 ± 3.8 ants per day (*n* = 5; for estimate calculation, see the "Quantification of model" section in Materials and Methods), the rescued ants make up a large proportion of the daily turnover in the colony.
Value of rescue behavior for the individual
-------------------------------------------
We classified carried ants into three mutually exclusive categories: (i) ants that partially or completely lost an extremity (antenna or leg), (ii) ants that have termites clinging to their bodies, and (iii) ants that appear to carry no obvious injury (fig. S1B). Most carried ants had a termite clinging on an extremity ([Fig. 1A](#F1){ref-type="fig"} and table S1). This handicap reduces the speed of the ant the most (4.5% of the mean speed of a healthy individual; [Fig. 1B](#F1){ref-type="fig"} and table S1) and, if removed successfully, has no long-term consequences. When 20 randomly selected individuals from each of the three categories of carried ants were forced to return alone from the hunting ground, 32% (*n* = 19 of 60) of them died ([Fig. 2A](#F2){ref-type="fig"}), in contrast to 10% of healthy individuals (*n* = 2 of 20). Ants that were carried back to the nest were never observed to be under any threat of predation (*n* = 420 raids observed during the entire field phase), thereby reducing return journey mortality of injured ants from 32% to close to 0%. The main cause of death when ants were forced to return alone was predation by spiders (57.1%: *n* = 12 of the 21 ants killed during the return journey alone from the hunting ground; [Fig. 2](#F2){ref-type="fig"}, B to E). Ants that had a termite clinging on an extremity had the highest mortality rate (50%, *n* = 10 of 20; [Fig. 2A](#F2){ref-type="fig"}). In nature, injured individuals were never observed to return alone without help, but six fatal injuries were observed at the hunting ground (in a total of 53 raids): removed head, thorax, gaster, or multiple legs. These ants were left behind at the hunting ground.
{#F1}
{#F2}
Ants that were carried back to the nest were observed again in subsequent raids 95% of the time (*n* = 38 of 40), sometimes less than an hour after the injury (individuals were marked with acrylic color codes for recognition). Termites clinging onto extremities were removed in 90% of the cases in the following 24 hours without removing the extremity (*n* = 20), thereby completely rehabilitating the handicapped individual. Ants that had lost two randomly selected legs were able to recover in the safe confines of the nest. Twenty-four hours after their injury, they reached mean running speeds 32.1% faster than freshly injured ants, a speed not significantly different from that of healthy individuals ([Fig. 3](#F3){ref-type="fig"} and table S2).
![Speed of injured ants at different times after injury.\
Box-and-whisker plot showing median (horizontal line), interquartile range (box), distance from upper and lower quartiles times 1.5 interquartile range (whiskers), and significant differences (different letters) of the different running speeds 5 min after removing two legs (Fresh injury) and 24 hours later \[Old injury (+24 hours)\] and of healthy ants (Healthy) (Kruskal-Wallis rank sum test, followed by a Dunn's test with Bonferroni correction; *n* = 20 trials with five colonies). See also table S2 for detailed statistical results.](1602187-F3){#F3}
Of the carried ants, 96.1% were minors (*n* = 154 in 20 observed raids). This is also reflected by the fraction of injured individuals in raiding columns before the fight. A significantly larger fraction of intermediates and minors had lost an extremity compared to majors ([Fig. 4](#F4){ref-type="fig"} and table S3). The few majors that were carried either had a termite clinging on them or had lost an extremity; they never appeared unharmed.
![Distribution of long-term injuries in different size classes of *M. analis*.\
Box-and-whisker plot showing median (horizontal line), interquartile range (box), distance from upper and lower quartiles times 1.5 interquartile range (whiskers), and significant differences (different letters) for the percentage of ants that lost an extremity in previous raids for majors, intermediates, and minors \[analysis of variance (ANOVA), followed by Tukey post hoc test; *n* = 20\]. See also table S3 for detailed statistical results.](1602187-F4){#F4}
Focus of rescue behavior
------------------------
To show that this behavior is focused on injured nestmates, we artificially injured individuals by removing one leg on each side. These individuals were then placed at the front of the return column, forcing all ants in the column to walk past the injured individual. Whereas healthy and dead individuals were ignored or disposed of by their nestmates, the artificially injured individuals were picked up and carried back to the nest ([Fig. 5A](#F5){ref-type="fig"}, table S4, and movie S1). Artificially injured individuals from other colonies were always attacked and removed from the column ([Fig. 5A](#F5){ref-type="fig"} and table S4). Rescue behavior occurred both directly at the hunting ground and on the return journey, whereas artificially injured ants on the way to the termites were ignored ([Fig. 5B](#F5){ref-type="fig"} and table S4).
![Behavioral responses of helper ants toward different treatments of injured individuals or dummies.\
Positive values show clear attempts of help by picking up the ant and dropping it again (black) or carrying it back to the nest (gray). Negative values show behavior in which the ant was disposed of (dragged away from the raiding column or attacked) \[Fisher's exact test for count data between neutral treatment (Healthy) and the other categories; *n* = 20\]. (**A**) Response toward different injury states. (**B**) Response at different points of the raid (Way out: on the way toward the termites; Hunting: at the hunting ground; Return: on the return journey after the fight). (**C**) Response toward dummies (dead minors) treated with different glands (Mg dead: mandibular gland applied on a dummy; Mg alive: mandibular gland applied on a healthy/living ant; Dufours: Dufour's gland applied on a dummy; Poison: poison gland applied on a dummy). (**D**) Response toward dummies treated with different synthetic compounds (DMDS/DMTS: 50:50 mixture of DMDS and DMTS; Hexane: pure hexane as control). See also table S4 for detailed statistical results. \**P* \< 0.05; \*\**P* \< 0.01; \*\*\**P* \< 0.001. n.s., not significant.](1602187-F5){#F5}
Gland and pheromone triggering rescue behavior
----------------------------------------------
When looking for the signal triggering this rescue behavior, we first ruled out stridulation ([@R22]), a mechanism known to trigger helping behavior in other ants ([@R1], [@R7]). We observed artificially injured ants, on which stridulation was inhibited, to still be rescued (*P* \< 0.001; *n* = 20; fig. S2, A and B, and table S4). After extensive behavioral experiments on dummies (frozen minors), we were able to identify the mandibular gland as the most likely candidate, while ruling out hindgut content and the Dufour's and poison gland reservoirs as triggers of the behavior. To further test this hypothesis, we applied the mandibular gland contents onto healthy individuals; we found that healthy ants covered with mandibular gland material were then carried back by their nestmates ([Fig. 5C](#F5){ref-type="fig"} and table S4). A gas chromatography--mass spectrometry (GC-MS) analysis identified dimethyl disulfide (DMDS) and dimethyl trisulfide (DMTS) as the main chemical components of the gland, confirming a previous analysis of the gland contents and concentrations (14 ng of DMDS and 5 ng of DMTS per gland) ([@R10]). Although 9 ng of DMDS alone was not enough to trigger the rescue behavior on a dummy, 9 ng of DMTS by itself sufficed. A more pronounced response was achieved with an equal mixture of the two components (9 ng of DMDS and DMTS each) ([Fig. 5D](#F5){ref-type="fig"} and table S4).
Value of rescue behavior for the colony
---------------------------------------
The rescue behavior in *M. analis* reduces the foraging costs through a reduced mortality risk. We provide a simple analytical model (additional information in "Rescue behavior model" and "Quantification of model" sections in Materials and Methods) that identifies critical factors promoting the evolution of this rescue behavior and why it may have evolved so rarely.
We consider that this behavior could only emerge in species that forage or hunt in groups and in a limited spatial domain so that injured individuals are likely detected by other nestmates. Our model identifies three additional key variables that affect the potential benefit of this rescue behavior: (i) the product of the absolute rate at which ants are severely injured (or killed) in conflict with termites (ε~*H*~) and the fraction *f* (0 ≤ *f* ≤ 1) of these ants that could profit from the rescue behavior; (ii) the baseline mortality μ~0~ of ants---helping is more profitable if μ~0~ is small compared to ε~*H*~; and (iii) the future added mortality rate (μ~*J*~) of individuals that were injured and rescued.
The foraging behavior of *M. analis* seems to offer ideal conditions for rescue behavior to arise. Injury rates in combat (ε~*H*~ = 0.17% per day; for detailed calculations, "Quantification of model" section in Materials and Methods) seem to be large compared to the general mortality rate (μ~0~ = 0.76% per day), but injuries are rarely fatal (six observations of fatal injuries in 53 raids). Further, ants that recently lost a leg or had termites clinging onto their extremities are significantly hindered in their movement. This presumably makes returning to the nest on their own costly in terms of energy and time needed, thereby prolonging exposure to potential predators and signaling a vulnerable state. These effects result in a high mortality risk of 32% for injured individuals if not helped ([Fig. 2A](#F2){ref-type="fig"}). Carried injured individuals therefore benefit greatly from the rescue behavior, by reducing that risk to close to 0% (injured ants that were marked and rescued were observed again in subsequent raids in 38 of 40 cases and were never observed to die during the rescue process). Injured ants that are carried back recover from injuries in a short time, that is, parameter μ~*J*~ is close to ε~*H*~, if we conservatively assume a second injury to be fatal. The fact that 21% of all ants carry some type of long-term injury in the raiding column ([Fig. 4](#F4){ref-type="fig"}) substantiates the great value of helping injured nestmates, a conservative estimate because nonpermanent injuries are not included in this estimate. The value of rescue behavior is reflected in the sustainable colony size calculated by our model, which predicts a 28.7% larger colony size compared to colonies without this behavior.
DISCUSSION
==========
This study shows the adaptive value of rescue behavior in a social predator specialized on a highly defensive prey---a behavior specifically focused on rescuing injured and handicapped individuals (remarkably also individuals that have permanent injuries in the form of lost extremities). Furthermore, by showing that this behavior is induced by pheromones, we support the hypothesis that the convergent evolution of rescue behavior in different taxa has led to distinctive triggering mechanisms, such as chemical communication in insects or empathy in humans and possibly other mammals \[([@R23], [@R24]); but see the study of Vasconcelos *et al.* ([@R21]) for other interpretations\].
Rescuing injured individuals
----------------------------
Intermediates and minors carry injuries considerably more often than majors ([Fig. 4](#F4){ref-type="fig"}). The division of labor at the hunting ground could explain this discrepancy. Whereas the smaller ants enter the termite galleries to hunt termites, the majors mostly focus on breaking up the protective soil layer over the hunting ground and carrying back the dead termites ([@R16]). Minors and intermediates are therefore far more exposed to injury risks.
A considerable number of carried ants did not seem to be injured ([Fig. 1](#F1){ref-type="fig"}). Either the injuries were too small to be detected by the naked eye or these ants were truly unharmed. Most of the ants are picked up at the hunting ground after the fight, when the ants are preparing to leave. One possibility could be that the majors running over the hunting ground searching for leftover termites or injured individuals are less selective in what to carry. If they still encounter a minor, which might have lagged behind because it is inside the termite galleries, the major might just pick it up, thereby preventing it from falling behind even further while the column is already leaving.
The fact that experimentally injured ants are not picked up during the outward journey toward the termites seems to suggest that the behavior is context-specific ([Fig. 5B](#F5){ref-type="fig"}). It seems unlikely for the ants to ever encounter this situation naturally. Furthermore, the rescue behavior would have to deviate from the natural one. If the helping ant would carry the injured individual back to the nest, it would expose itself to considerable predation risks by being forced to return alone while the rest of the column keeps marching to the termites. The other possibility would be to carry the injured ant all the way to the hunting ground only to have it carried back to the nest afterward. The different response necessitated by the helper ant in this situation and the very low injury risk on the outward journey most likely prevented the ants from developing a corresponding response.
We were able to show that this behavior is triggered through the chemical compounds DMDS and DMTS harbored in the mandibular gland reservoir; thus, this suite of compounds is a pheromone that seems to be specifically released when the individual is injured, in order to induce rescue behavior. The only other known species harboring this pheromone is the solitary hunting ponerine ant *Paltothyreus tarsatus*, in which the pheromone triggers digging behavior, most likely to rescue trapped nestmates ([@R25]). This species is in the same genus group as *Megaponera* ([@R8]), but being a solitary forager, *P. tarsatus* probably has not evolved the same kind of cooperative foraging-injury rescue behavior as in *M. analis*.
Cooperative self-defense has also been observed in *M. analis*, as a behavior in which nestmates scanned each other's legs and antennae and removed *Dorylus* sp. (driver ants) clinging to their extremities during encounters ([@R26]). The removal of these *Dorylus* ants seems to follow a mechanism similar to the removal of termite soldiers within the nest.
Evolution of helping the injured
--------------------------------
We were able to assess the value of rescue behavior for injured individuals. Because assistance to individual ants is the main benefit of the rescue behavior, understanding the evolutionary benefit of this behavior for the colony as a whole is paramount. Ants that had lost an extremity do not immediately switch to a four- or five-legged locomotion mechanism but keep tripping over their phantom limbs. Ants that had termites clinging on them were even more severely handicapped in their movement ([Fig. 1B](#F1){ref-type="fig"}). These ants were therefore unable to keep up with the returning column, fell behind, and thus became isolated from their nestmates. This retarded movement, on top of reduced dexterity, increased predation risk considerably ([Fig. 2](#F2){ref-type="fig"}, A to E). Once termites clinging on their extremities were removed within the nest, they were able to fully perform again in future raids without any clear handicap. Ants that had lost an extremity had the benefit, after being carried back, to recover from their injury in the safe environment of the nest, allowing them to get accustomed to a four- or five-legged locomotion. Thus, they reached running speeds similar to those of uninjured ants 24 hours later ([Fig. 3](#F3){ref-type="fig"}). Because nearly all injured ants were observed in subsequent raids, we conclude that they carried no obvious long-term handicaps from their injuries and may fully participate again in colony tasks.
This type of rescue behavior, focused specifically on injured and handicapped individuals after hunting, is unique in social insects. Although the benefits seem obvious, there are several reasons as to why this has not yet been discovered in other species. First, because this behavior can only evolve in group-hunting species, where an injured ant can be detected by its nestmates, it excludes all solitary hunting species as potential candidates. Second, it is also essential that hunting occurs in isolated events, thus creating the risk for the injured ant to be separated from the group during the return journey. In a constantly occupied trail between ants and food source, the increased risk carried by an injured ant would be marginal, because it is constantly surrounded by nestmates warding off potential predators. In *M. analis*, the outward and return journeys are conducted as a discrete column with all ants marching together. The fact that the ants wait after the fight so that all ants may gather before returning to the nest ([@R11], [@R15]) exemplifies the importance of returning as a group. Third, the prey species must be able to inflict a high amount of nonlethal injuries from which the ants can recover. With their large soldiers, termites fulfill this criterion as injury-inducing prey. Group foraging ant species that focus on leaf cutting, nectar, seeds, or scavenging are less likely to develop this rescue behavior of termite-raiding ants. Fourth, the benefit to the colony by the rescued ant has to outweigh the cost of help. In *M. analis*, the majors carry back termites and injured individuals; because minors sustain the most injuries ([Fig. 4](#F4){ref-type="fig"}), the additional task for the majors to carry them back seems minimal from an energetic point of view (fig. S1A). Moreover, because on an average raid only 30% of the ants carry back prey ([@R11]), a large part of the workforce is available to help the injured individuals without decreasing the profits of the raid. Because the cost of helping an injured ant is therefore likely to be marginal in *M. analis*, it is thus ignored in our model. Last, the value of an individual for the colony plays an important role. This can be approximately quantified through the mean mortality rate in a colony. For a colony to be in equilibrium, the number of ants being born has to match the mortality rate, and in equilibrium, the population turnover is directly related to the life span of the individual. In *M. analis*, the population turnover is relatively low, with a birth rate of only 13 ants per day, demonstrating again the importance of rescuing the injured. Species with a very high turnover, such as army ants, seem likely to derive less benefit from saving one injured ant, although this hypothesis remains to be tested. The specific biology of *M. analis* therefore provides the right circumstances where the benefit of saving the injured is especially large. We thus argue that this behavior evolved as part of an evolutionary arms race against termites, as a means of minimizing losses during raids and therefore foraging costs.
Rescue behavior has been previously observed in ants ([@R1]) but in very different contexts. Excavating trapped nestmates after a cave-in and rescuing an ant that fell in an antlion trap are both situations in which the individuals are confronted with an imminent danger, that is, suffocation or being eaten ([@R1], [@R3], [@R5]). This is not the case in our situation: Not only are the injured ants in many cases handicapped for life through the loss of extremities, but the immediate danger toward these ants is far less obvious. There is no direct threat to the injured ant but rather an abstract increased predation risk if these ants were to return alone. This study demonstrates that complex rescue behavior can evolve in very unique situations if the necessary drivers are present, even in species that are very likely unable to recognize the increased risks to which they are exposed to.
Outlook
-------
Our observations offer a unique opportunity to experimentally study the evolutionary drivers leading to the emergence of rescue behavior in animals: Injury and predation rates can be manipulated, rescue behavior can easily be prevented, and critical variables and parameters can be measured. The Pan-African distribution of *M. analis* should also allow us to study the degree of fine-tuned adaptations to differing external selection pressures prevailing in different ecosystems and enable identification of the most important potential driving factors for evolution of this behavior. Our model also helps us to identify other candidate species in which this behavior might be found. Other ponerine genera, such as *Leptogenys*, also focus on hunting termites, with some of them hunting in groups ([@R27]); examining their raiding behavior in more detail could be promising. Slave-making ants could potentially also fulfill the criteria, if their prey can inflict a significant amount of nonlethal injuries.
MATERIALS AND METHODS
=====================
Experimental design
-------------------
The study was conducted in a humid savannah woodland located in the Comoé National Park, northern Côte d'Ivoire (Ivory Coast), at the Comoé National Park Research Station (8°46′N, 3°47′W) ([@R28]). Observations throughout several days in April 2013 established that raiding activity was highest in the morning and afternoon hours between 0600 to 1100 and 1500 to 1900 local time, which corresponds to previous observations ([@R11], [@R13], [@R14]). Night raiding was also observed but was not included in this study. Experiments and observations were carried out in the field from 0700 to 1100 and 1500 to 1800 from April to September 2013, August to October 2014, and January to March and July to November 2015. *M. analis* is found throughout sub-Saharan Africa from 25°S to 12°N ([@R8]). We observed *M. analis* in a total of 52 different colonies for a total of 420 raids, in which the predominantly hunted termite species was *Pseudocanthotermes* sp. Living nests of *Macrotermes bellicosus*, which in other areas were often favored prey, could potentially cause a higher and more fatal injury rate due to their stronger soldiers; this species was absent in the vicinity of the study area. Colony size of 10 excavated colonies ranged between 900 and 2300 ants, a result comparable to previous studies in other regions ([@R16], [@R29]). Although *M. analis* is known to show monophasic allometry within its worker sizes ([@R16], [@R30]) for statistical analysis and illustration, the workers were divided into majors (head width, \>2.40 mm), minors (head width, \<1.99 mm), and intermediates (head width, 2.40 to 1.99 mm), as proposed by Villet ([@R16]).
Quantification of carried ants
------------------------------
Experiments and observations were conducted in the field by waiting in front of a colony for a raid to be initiated and then following the raid column to the hunting ground. In total, we observed 420 raids in 52 different colonies. These 420 raids were used in various different experiments and observations.
To quantify the number of ants being carried back from the hunting ground, we counted the number of ants carrying a nestmate during the return journey shortly before arriving at the nest in a total of 53 raids. To classify the type of injuries in the carried ants (that is, with the following categories: lost limb, carried unharmed, and termite clinging), we retrieved the ants with forceps and investigated them. Injury inspections were conducted in 20 raiding columns of 20 different colonies for a total of 154 carried ants ([Fig. 1A](#F1){ref-type="fig"} and fig. S1B).
To quantify the number of long-term injured ants participating in raids, we collected ants of all castes from 20 raids, each from a different colony, when raid columns were leaving the nest (that is, before any new fight could have taken place). In total, we collected 763 minors, 582 majors, and 502 intermediates (total *n* = 1847; [Fig. 4](#F4){ref-type="fig"}).
Velocity and mortality
----------------------
This experiment was conducted 20 times for each of the three categories (lost limb, termite clinging, and carried unharmed) in individual raids, with an additional control test of healthy individuals ([Fig. 1](#F1){ref-type="fig"}, A and B). Individuals for the experiments were randomly selected from the pool of carried ants in a raid, with the control being a healthy ant walking unassisted in the returning raid column showing no sign of injury or handicap. Velocity was measured for the distance the ant followed the pheromone trail back to the nest. If a predator killed the ant during the return journey, the speed was calculated on the basis of the distance covered up to that incident. This allowed us to quantify the handicap and mortality risk that each injury posed during the return journey ([Fig. 2](#F2){ref-type="fig"}, A to E). If the ant stopped moving during the return journey, most likely because of fatigue, the time was also stopped and the velocity was calculated up to that point.
Injury recovery
---------------
To analyze the potential recovery of ants that lost an extremity, we randomly cut off one leg on each side of a healthy ant (with scissors) and picked up during the return journey of the raid. The ant was then released on the return pheromone trail, and the covered distance in 60 s was measured. This experiment was repeated with the same individual 24 hours later and with healthy individuals as a control. These experiments were conducted with laboratory colonies, so that we could easily reproduce experimental parameters with regard to time between experiments and nest conditions. We used acrylic pens to mark individual ants, because previous observations showed no indication of any lasting disturbance to the ants with this marking method (fig. S2).
Ethogram of rescue behavior
---------------------------
Because there was no significant difference in the quantity of ants helped at the hunting ground or on the return journey ([Fig. 5B](#F5){ref-type="fig"}; Fisher's exact test, *P* = 0.33; *n* = 20), we carried out subsequent experiments ([Fig. 5](#F5){ref-type="fig"}, A and C) during return journeys for easier reproductibility of trials. The experiments were repeated 20 times with at least five different colonies, with each return raid used for only one trial. For these experiments, an injured ant (or dummy) was placed at the front of the return column at least 1 m away from the hunting ground. All behavioral reactions by the nestmates were recorded until the entirety of the column passed the study subject. The behavioral reactions of the helping ants consisted of five categories: (i) Ignored: contact with the study subject was less than 2 s; (ii) Investigated: the study subject was antennated for more than 2 s; (iii) Picked up: the study subject was fully lifted from the ground; (iv) Carried back: the study subject was carried back for at least 20 cm toward the direction of the nest; (v) Carried away: the study subject was removed from the return column in a direction that was not the one back to the nest, that is, away from the column. For graphical illustration and statistical analysis, we summarized behaviors (iii) and (iv) as rescue behavior and (iii) in combination with (v) as disposing of the study subject.
Laboratory colonies
-------------------
Ten colonies were excavated and placed in artificial nests in the field station laboratory (colony size, 1373 ± 520 ants), consisting of a 20 × 20 × 10--cm large nest made of polyvinyl chloride connected to a 1 × 1--m arena. For raids, this arena was connected by a 10-m-long corridor to a second arena (1 × 1 m). The ground was covered with earth from the surrounding area. In the second arena, *M. bellicosus* termites were placed, which were collected from the surrounding area with a pot filled with dry grass. These termites were found by scouts and triggered raiding behavior on which we performed the injury recovery experiments. For further details on laboratory keeping, see Yusuf *et al.* ([@R29]).
Pheromone and stridulatory communication
----------------------------------------
To inhibit stridulation, we coated the stridulatory organ, located between the first and second tergites ([@R22]), with black acrylic paint. After the paint dried for 2 min, the experiment was conducted. To confirm that stridulation was truly inhibited, we triggered normal stridulation behavior by exposing the ant to CO~2~, as previously described by Hölldobler *et al.* ([@R22]). During this process, the sound was recorded with an external microphone (Speedlink SL-8703-BK, Jöllenbeck GmbH). To visualize the sound, a sonogram was created with the digital audio editor Audacity version 2.0.5.0 (fig. S1A).
For the pheromone experiments, we dissected a gland and placed it on a glass surface, then pulled the thorax of the study subject three times over the burst gland reservoir (for smaller mandibular glands, two glands were used per experiment). For the experiments with synthetic chemicals, we first diluted the substance in hexane until we reached a concentration of 90 ng/ml. Subsequently, two drops (roughly 9 ng of the substance) were applied on a glass surface. The selected concentrations were similar to the quantities found in a mandibular gland \[14 ng of DMDS and 5 ng of DMTS per gland in a major worker according to Longhurst *et al.* ([@R17])\] and a comparison of the mass spectrometer of the gland reservoir with our solution. After 30 s, most of the hexane evaporated and the thorax of the dummy was pulled over the glass surface three times.
Chemical analysis
-----------------
Foraging *M. analis* workers were collected from various colonies at the Comoé National Park (Côte d'Ivoire). The workers were then transported alive to the University of Würzburg (Germany) and killed with CO~2~ before excision of the mandibular gland reservoirs. The caput and the mandibles, including the mandibular gland, of 20 ants were then soaked in 1 ml of pure pentane for 2 hours (two caputs and six mandibular glands, respectively). These extracts were evaporated to a residue of approximately 100 μl. We used 1 μl extracts for GC-MS analyses, which were carried out on a gas chromatograph (6890) coupled to a mass selective detector (5975) from Agilent Technologies. The GC was equipped with a DB-5 capillary column (0.25 mm inside diameter × 30 m; film thickness, 0.25 μm; J&W Scientific). Helium was used as a carrier gas, with a constant flow of 1 ml/min. A temperature program from 60° to 300°C with 5°C/min and finally 10 min at 300°C was used, with data collection starting 2 min after injection. The mass spectra were recorded in the electron ionization mode, with an ionization voltage of 70 eV and a source temperature of 230°C.
The software ChemStation (Agilent Technologies) for Windows was used for data acquisition. Identification of the components was accomplished by comparison with purchased chemicals and the use of a commercial MS database (NIST 4.0). Because of the very small quantities of DMDS and DMTS within the extracts, we used diagnostic ions and the retention time to confirm the identification.
Rescue behavior model
---------------------
We designed an equilibrium model to quantify the possible benefits of rescue behavior to the colony. Benefit was expressed as the proportional increase in equilibrium worker number of a colony with rescue behavior compared to a colony that would not show this behavior. For the sake of argument, we chose a very simple model that does not account for all the mechanisms that truly regulate worker numbers in ant colonies.
We assume that the worker dynamics of a colony without rescue behavior is described by equation$$\frac{\mathit{d}\mathit{H}}{\mathit{d}\mathit{t}} = \mathit{b} - (\varepsilon_{\mathit{H}} + \mu_{0})\mathit{H}$$where *H* is the number of noninjured (healthy) workers, *b* is the rate at which new workers are added to the colony, ε~*H*~ is the rate at which workers are involved in injuring interactions with termites, and μ~0~ is the base mortality rate of workers.
For this colony, colony size (worker number) will settle into equilibrium$$\hat{\mathit{H}} = \frac{\mathit{b}}{\varepsilon_{\mathit{H}} + \mu_{0}}$$
In addition, a colony that manages to rescue a fraction *f* (0 ≤ *f* ≤ 1) of the workers injured in action will build a "pool" *J* of workers that were injured in previous raids but were rescued; conservatively, we do not separate between injured ants that may ultimately recover (and would thus return to pool *H*) and workers that carry permanent damages, such as a lost extremity. The dynamics of injured ants is described as$$\frac{\mathit{d}\mathit{J}}{\mathit{d}\mathit{t}} = \mathit{f}\varepsilon_{\mathit{H}}\mathit{H} - (\mu_{\mathit{J}} + \mu_{0})\mathit{J}$$where *f* (0 ≤ *f* ≤ 1) is the proportion of ants injured in combats that survive and μ~*J*~ is the added (future) mortality rate of injured compared to noninjured workers. For simplicity and on the basis of empirical observation, we conservatively assume that a second injury sustained in another raid would always be fatal. The equilibrium number of injured ants in a colony is thus$$\hat{\mathit{J}} = \hat{\mathit{H}}\frac{\mathit{f}\varepsilon_{\mathit{H}}}{\mu_{\mathit{J}} + \mu_{0}}$$
The relative size of colonies with rescue behavior compared to a colony not showing this behavior, that is, a total loss of injured individuals (*f* = 0), is thus defined by$$\frac{\hat{\mathit{H}} + \hat{\mathit{J}}}{\hat{\mathit{H}}} = 1 + \frac{\mathit{f}\varepsilon_{\mathit{H}}}{\mu_{\mathit{J}} + \mu_{0}}$$
Quantification of model
-----------------------
The observed survivability of an injured ant not receiving help is 68% ([Fig. 2A](#F2){ref-type="fig"}). Thus, *f* = 0.68 characterizes hypothetical colonies without rescue behavior, whereas in colonies where the behavior is present, *f* = 1, because all rescued ants were observed in later raids. All other parameters stay the same in both cases and were calculated as follows. Because we can only quantify the injury ratio in the colony for ants that lost an extremity, our value ε~*H*~ was defined as the percentage of lost limb injuries per raid (0.21·3) per day ([@R3]) divided by the ratio of healthy ants in a colony (0.79·1373, *n* = 10 excavated colonies); therefore, ε~*H*~ = 0.0017. We conservatively argue that the added mortality of a previously injured ant is the probability of getting injured again; therefore, μ~*J*~ = ε~*H*~, in our scenario. We estimated the birth rate *b* of the colony by observing the callow worker population of excavated nests until they were fully sclerotized (106 ± 30 callow workers per excavated colony, *n* = 5). Sclerotization time was calculated to be 8 days on average (*n* = 5), leading us to an estimate of 13.3 ± 3.8 ants born per day. In [Eq. 1b](#E1b){ref-type="disp-formula"}, we were able to calculate μ~0~ = 0.0076. To test the precision of the parameters estimated from empirically observed data, we compared the empirical ratio of injured ants in the colony (0.21; [Fig. 4](#F4){ref-type="fig"}) to our model prediction from [Eq. 2b](#E2b){ref-type="disp-formula"}: 0.21. The good agreement of predicted and empirical values allowed us to reliably calculate the benefit of the rescue behavior by comparing the calculated colony size (*H* + *J*) of a colony with rescue behavior to one without. Our results indicate that the helping behavior results in a 6.0% larger colony size if we just consider the benefit for the 21% of carried ants that lost an extremity ([Fig. 1A](#F1){ref-type="fig"}). If extrapolated for all injuries, the benefit of the rescue behavior can be estimated to be 28.7%.
Statistical analysis
--------------------
For statistical analysis and graphical illustration, we used the statistical software R version 3.1.2 ([@R31]) with the user interface RStudio version 0.98.501. We tested for deviations from the normal distribution with the Shapiro-Wilk test (*P* \> 0.05). A Bartlett test was used to verify homoscedasticity (*P* \> 0.05). If data were normally distributed and homoscedastic, an ANOVA was used to compare the significance of the results with a Tukey post hoc test for post hoc analysis. If this was not the case, a Kruskal-Wallis rank sum test was used, followed by a Dunn's test with Bonferroni correction. To analyze the ethogram data, a Fisher's exact test with Bonferroni correction was used with a no-help control (0 of 20 helped) compared to our treatments. Median values mentioned in the text are followed by a median absolute deviation.
Supplementary Material
======================
###### http://advances.sciencemag.org/cgi/content/full/3/4/e1602187/DC1
We thank V. Frank, B. Fiala, S. Leonhardt, K. Stein, and the anonymous referees for critical discussions and reading the manuscript. We thank the Comoé National Park Research Station for the use of their facilities for the field research and the park management of Office Ivoirien des Parcs et Réserves for facilitating field research in the park. **Funding:** O.M. and T.H. are funded by the German Research Foundation (DFG), Collaborative Research Center (grant number SFB 1047, "Insect Timing," Project C6). E.T.F. was supported by a grant of the German Excellence Initiative to the Graduate School of Life Sciences, University of Würzburg (grant number GSC106/3). This publication was funded by DFG and the University of Würzburg in the funding programme Open Access Publishing. **Author contributions:** E.T.F. and K.E.L. designed the study and collected and analyzed the field data. E.T.F. wrote the paper. T.S. performed MS analysis and contributed to the study design. J.S. conducted part of the field experiments. T.H. and O.M. developed the model. K.E.L. initiated and supervised the study. All authors discussed the results and commented on the manuscript. **Competing interests:** The authors declare that they have no competing interests. **Data and materials availability:** All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.
Supplementary material for this article is available at <http://advances.sciencemag.org/cgi/content/full/3/4/e1602187/DC1>
fig. S1. Illustration of a helping ant and different injury types, as shown in [Fig. 1](#F1){ref-type="fig"}.
fig. S2. Effect of stridulation on rescue behavior.
table S1. Statistical differences in injury-type frequency ([Fig. 1A](#F1){ref-type="fig"}) and speed ([Fig. 1B](#F1){ref-type="fig"}) in injured *M. analis* ants (Kruskal-Wallis test, followed by pairwise comparisons with Bonferroni-corrected Dunn's test; *n* = 20 per test).
table S2. Statistical differences in running speed of individuals with different stages of injury, as shown in [Fig. 3](#F3){ref-type="fig"} (Kruskal-Wallis test, followed by Dunn's test with Bonferroni correction; *n* = 20 per test).
table S3. Statistical differences in long-term injuries in the different castes, as shown in [Fig. 4](#F4){ref-type="fig"} (ANOVA test, followed by Tukey post hoc test; *n* = 20 per test).
table S4. Statistical differences in significance of rescue behavior compared to behavior of healthy individuals, as shown in [Fig. 4](#F4){ref-type="fig"} (A to D) and fig. S2 \[Fisher's exact tests for count data between treatment healthy (no help) and the other categories with Bonferroni correction\].
movie S1. An injured ant receiving help by its nestmates.
| {
"pile_set_name": "PubMed Central"
} |
The article by Kweon TD et al. entitled, \"Heart rate variability as a predictor of hypotension after spinal anesthesia in hypertensive patients\" (Korean J Anesthesiol 2013 October 65(4): 317-321) contained an error in Accepted date.
**Before correction:**
Received: January 2, 2013. Revised: 1st, February 27, 2013; 2nd, March 18, 2013; 3rd, April 11, 2013. Accepted: April 24, 2012.
**The correct information is found below:**
Received: January 2, 2013. Revised: 1st, February 27, 2013; 2nd, March 18, 2013; 3rd, April 11, 2013. Accepted: April 24, 2013.
The Accepted date was misspelled as April 24, 2012. The correct spelling is April 24, 2013.
We apologize for any inconvenience this mistake may have caused.
| {
"pile_set_name": "PubMed Central"
} |
Background
==========
Alcohol use has been ranked high as a risk factor for death and loss of healthy life years, as well as being a source of personal and social harm \[[@B1]-[@B3]\]. Policies and interventions exist that can reduce alcohol use, but there is still a need to understand more about why alcohol use and abuse change in populations and how alcohol-related harm can be reduced \[[@B4]\].
Studies in Europe have generally found that immigrants from non-Western countries tend to drink less or less often than the host population. In the Netherlands, Turkish and Moroccan immigrants reported less alcohol use than the Dutch population in both the first and second generation, and their alcohol consumption did not converge towards the higher rates in the host population \[[@B5],[@B6]\]. In another study from the Netherlands mono-ethnic immigrants in secondary schools from the Antilles and Surinam were also included in addition to Turks and Moroccans, all reporting a lower prevalence of drinking than the Dutch students \[[@B7]\]. In the United Kingdom, African Caribbeans have reported lower alcohol consumption than their white counterparts in studies from 1986 to 1995 \[[@B8]\]. Asian Muslims reported a very low level of alcohol consumption in the United Kingdom \[[@B9]\]. Muslim 15-16-year-olds in England, as well as Hindus and Sikhs in the same age group, reported a far lower level of consumption than white adolescents \[[@B10]\]. In Germany, among 9^th^ graders, the largest immigrant group (Turks) reported less than half the level of binge drinking in the previous four weeks than those of German descent. The drinking behaviour of the second largest group of immigrants in Germany (from the former Soviet Union) was similar to those of German descent. Otherwise adolescents from Islamic countries living in Germany had a lower lifetime prevalence of drinking than German and Western European adolescents, with the exception of students from Iran \[[@B11]\]. In a study from Oslo, Norway, 15-16- year-olds with an ethnic Norwegian background reported a higher frequency of drinking than their age-peers from Pakistan, Somalia, Turkey and Morocco where Islam was the predominant religion. The frequency of drinking among youth from Iran, another country where Islam is the predominant religion, was also lower than among Norwegians, but not as low as among youth from the Islamic countries already mentioned. A large proportion of youth from Vietnam had tasted alcohol, but fewer drank often and fewer got drunk compared to ethnic Norwegians \[[@B12]\]. Albanians in Florence, Italy, reported more drinking, however, than Albanians in the home country and among Italians \[[@B13]\]. Also, in a study of recent Hispanic immigrants to Spain over the age of 15, there were no significant differences between the previous 12 months' drinking among immigrants compared to the native population. Almost 40% of the immigrants reported a higher level of drinking than in their country of origin \[[@B14]\].
In the United States (US) adolescents of Chinese origin were less likely to be drinkers then Chinese adolescents in Hong Kong and American adolescents in the United States \[[@B15]\]. Among Mexican-heritage youth (12--17 years old) living in the US, 13% of those born in Mexico reported use of alcohol, while 19% of the US-born reported such use \[[@B16]\]. In two studies (one among adolescents and one among those 18 years old and over) comparing alcohol use in white and Asian subgroups, the prevalence of increased drinking behaviour ranked highest for whites, followed by Japanese or Filipinos, Koreans, Chinese and Vietnamese. Other studies show somewhat different ranks for people of Asian background \[[@B17]\]. Numerous studies have been carried out on alcohol use and misuse according to race/ethnicity in the US, but the country of origin is seldom included since this requires very large or country-specific samples \[[@B18]\]. Finally, in Australia, Vietnamese were found to consume alcohol at a lower rate than that of Australians in general \[[@B19]\].
Thus, alcohol use among immigrants to Western countries tended to be less prevalent among non-Western immigrants, but more prevalent and more frequent among those residing in the host country for a longer period, i.e., among second-generation immigrants (as compared to first-generation). In addition those immigrants who spoke the language of the receiving country drank more often \[[@B11],[@B15],[@B20]-[@B22]\].
Drinking is strongly influenced by other drinkers in a person's personal network. It has been argued that society at large, or at least large segments of society, behave collectively regarding drinking \[[@B23]\]. A higher level of contact with one's own dry culture will thus predict a lower level of drinking, while a higher level of contact with the host's 'wet' culture will predict a higher level of drinking. Immigrants from dry alcohol cultures may influence alcohol use in the whole population in a 'wetter' host country, however, not only by being a low-consumption group themselves. In general, the interaction between the host society and immigrant populations has an impact on the attitudes, values and behaviours of both collectives \[[@B24]\]. In Norway and the Netherlands it was found that the higher the percentage of Muslims in a school, the lower the frequency of drinking both among immigrant youth and native youth from the host country \[[@B7],[@B12]\]. A higher level of contact with a 'drier' drinking culture may thus reduce drinking also in the host population.
The pattern of drinking in Norway has been characterised by relatively infrequent consumption, but with a high level of consumption and drinking into intoxication, especially on festive occasions \[[@B25]\]. Alcohol sales increased by 40% from 1995 to 2009, but sales per capita in Norway were still the lowest of the Nordic countries. In 2010 sales per capita in Norway were almost half of sales in Germany, Spain and France. In a population survey from 2008 (15 years and above), 86% of women and 92% of men reported drinking during the last 12 months, while 16% of women and 27% of men reported drinking several times a week. Drinking alcohol every day with meals or in other contexts was thus not common; most alcohol was consumed on weekends \[[@B26]\]. There has been a reduction in drinking among young people in Norway in recent years, but still consumption and drinking into intoxication are widespread. Among 15-16-year-olds in Norway 63% of the boys and 70% of the girls reported in the 2007 ESPAD study that they had used alcohol last year, both figures at the lower end of European measurements. Boys usually drank more on each occasion than girls. Rated by the proportion of drinking into intoxication, however, young Norwegians were found in the mid-range of the European countries \[[@B27]\]. Non-Western immigrants to Norway have thus encountered a much 'wetter' drinking culture than in their country of origin, and social events on weekends included drinking from a young age.
The aim of the Norwegian government's alcohol policies has for many years been to reduce both individual and societal costs. The measures introduced have included imposing a monopoly for the import and sale of wine and stronger alcohol, limiting access to beer sold in shops (restricted hours of sale), imposing high taxes on all sales and disseminating information on the consequences of alcohol use. Such a system has not been given priority in the EU or in single member countries (with the exception of Finland before entering the EU and Sweden) \[[@B4]\]. Norwegian policies have moved in a liberal direction in recent years, however, for example by opening up for personal import of wine and stronger beverages \[[@B28]\].
Immigration to Norway started in the latter part of the 1960s. In 1970 1.5% of the population was registered as immigrants (born outside Norway of non-Norwegian parents or born in Norway with two non-Norwegian parents). Only 0.1% came from non-Western countries at that time (Asia, Africa, South or Central America, Turkey). In 2001, at the time of the study referred to here, comparable figures for immigrants were 6.6% in total and 3.4% non-Western at the national level and 19% in total and 13% non-Western in Oslo \[[@B29]\]. The third generation of immigrants from those countries who came in the late 60s to early 70s had started to enter primary school in 2001. By 2011, the proportion of non-Western immigrants had increased even further \[[@B30]\].
The availability of three health surveys in 2000--2002 covering adolescents, adults, and adults in the largest immigrant groups in Oslo, Norway, enabled us to study alcohol drinking among ethnic Norwegians and three of the five largest non-Western immigrant groups with Islam as the predominant religion in their home country: Iranians, Pakistanis and Turks. The first aim of this study was to describe frequency of drinking in two generations of immigrants in Oslo, comparing the result to drinking among ethnic Norwegians. The second aim was to study how frequency of drinking among adult immigrants was associated with social interaction with their own countrymen and ethnic Norwegians, acculturation factors, age, gender, socioeconomic factors and the Muslim religion.
Methods
=======
Sample surveys
--------------
This analysis was based on three sample surveys included in the Oslo Health Study (HUBRO), which was conducted in joint collaboration with the Oslo City Council, the University of Oslo and the Norwegian Institute of Public Health. The study protocol was approved by the Norwegian Data Inspectorate and the Regional Committee for Medical Research Ethics.
HUBRO was carried out in the city of Oslo from May 2000 to September 2001 \[[@B31]\]. An invitation for participation in the health survey was sent to birth cohorts of 1924, 1925, 1940--41, 1955, 1960 and 1970 who had resided in Oslo on 31 December 1999. The postal invitation included a standardised main questionnaire and an invitation to attend a clinical examination with a second questionnaire, which included acculturation-related questions. In all 18,770 individuals participated in the adult part of HUBRO, 46% of those originally invited. More details on the sampling procedure can be found elsewhere \[[@B31]\]. In the analysis conducted here only persons born in 1940 to 1970 (30-60-year-olds in 2000) and persons born in Norway, the Islamic Republic of Iran (response rate 48%), Pakistan (response rate 40%) and Turkey (response rate 41%) were included. This selection regarding age was imposed to restrict the sample to the parent generation of the 15-16-year-olds (see next paragraph) and other adults who socialised with them. The selection of the three countries made it possible to study drinking among people of the Muslim faith. Missing values for frequency of alcohol drinking were 1.1% for the total sample, and higher for the three immigrant groups: 6.3% for Iranians, 29.9% for Pakistanis and 1.6% for Turks.
In addition, all 15-16-year-olds on class lists in Oslo in the period of January to April in both 2000 and 2001, as well as those who entered the school later, were invited to participate in a school-based survey. A total of 8435 students were registered in or entered schools, including those in special schools for the disabled. Thirty-one were unable to respond to the questionnaire due to impairment and 88 had moved or quit school before the survey was carried out. Thus 8316 students were eligible, among whom 7343 responded to at least one question (response rate 88%). This survey consisted of a questionnaire only. Response rates for Iranians were 82%, for Pakistanis 83% and for Turks 79%. More details on the sampling procedure can be found elsewhere \[[@B32]\]. Missing values for frequency of alcohol drinking were 1.2% for youth with two ethnic Norwegian parents, 1.6% for youth with one ethnic Norwegian parent, 4.1% for Iranians, 2.4% for Pakistanis and 3.7% for Turks.
A third HUBRO study, the Oslo Immigrant Health Study, was conducted among the five largest immigrant groups in Oslo in 2002 \[[@B33]\]. All individuals born in Sri Lanka, Turkey, the Islamic Republic of Iran and Vietnam between 1942 and 1982 were invited to participate, except for birth cohorts who had been previously invited to the HUBRO study mentioned above. A 30% random sample of Pakistanis, the largest immigrant group, was also invited. A postal invitation, a physical examination and a second questionnaire were administered in the same way as for HUBRO. In the age group 30--60 years selected for analysis in this study 7890 persons were eligible for participation (response rate 40%). The response rates for three countries selected for this analysis were: Turkey 33%, Iran 39% and Pakistan 32%. More details on the sampling procedure can be found elsewhere \[[@B33]\]. Missing values for frequency of alcohol drinking were 2.7% for Iranians, 11.9% for Pakistanis and 6.2% for Turks.
Item missing was at 2.2% or lower for gender and country of birth/ethnicity in the three surveys.
Three datasets were established for this study, including ethnic Norwegians and/or Iranians, Turks and Pakistanis who had reported frequency of alcohol drinking:
1\. The 15-16-year-olds in 2000 in HUBRO: 5381 youth with ethnic Norwegian background, 93 Iranians, 559 Pakistanis and 105 Turks. This sample was called the Oslo youth population sample.
2\. The 30-60-year-olds in 2000/2001 in a merged dataset of HUBRO and the Oslo Immigrant Health Study: 12,259 with ethnic Norwegian background and 604 born in Iran, 607 born in Pakistan, and 436 born in Turkey. Age and gender weights were used to estimate the population distribution. This dataset was called the Oslo adult population sample, constituting the parent generation of the 15-16-year-olds plus those adults who socialised with them.
3\. A selection from the Oslo adult population sample called the adult immigrant sample: immigrants who had attended the physical examination and delivered the second questionnaire. This sample included 389 born in Iran, 323 in Pakistan, and 270 in Turkey who also reported acculturation variables and other relevant variables. Item non-response for acculturation variables etc. was between 1.2 and 12.9%.
Survey variables
----------------
Alcohol use among adults varies by gender, age, ethnic or religious groups and socioeconomic status \[[@B4],[@B34]\]. Analyses of differences in drinking patterns between groups of immigrants and the host population should include such variables, partly due to their independent interest and partly due to possible confounding.
How often have you consumed alcohol in the course of the past year?
-------------------------------------------------------------------
The responses were divided into four categories: weekly/monthly/less than monthly/never drank alcohol.
Socio-demographic, socio-economic and religious variables
---------------------------------------------------------
Included gender and age, workforce participation (no, part-time, full-time), education (number of years of education), and being a Muslim.
Country of birth/ethnic background
----------------------------------
The place of birth was recorded as it appeared in the Norwegian Population Register. The parents' place of birth for the 15-16-year-olds was reported by the youth in the questionnaire. The categories by country background were based on the mother's country of birth, while the father could be from any country other than Norway. The parents of all the Iranians were from Iran, while among Pakistanis 98% were mono-ethnic, as were 97% of the Turks.
Aspects of acculturation and interaction among adults
-----------------------------------------------------
Various modes of the use of language and of socialisation/affiliation are frequently used to establish measures of competence and identification regarding both one's own and the host culture \[[@B35]-[@B38]\]. In the HUBRO survey for adults the following seven variables were available:
How often have you in the course of the past year:
· read a newspaper in your own language?
· read a newspaper in Norwegian?
· had a visit from a Norwegian?
· received help/support from a Norwegian?
· taken part in meetings arranged by your own countrymen?
In your opinion, how good is your knowledge of Norwegian? The first five measurements had four response categories: daily/weekly/less often/never, while the sixth had five response categories: very good/good/average/rather poor/poor.
The variable 'Number of years since moving to Norway' was also included in the analysis as an exogenous variable to measure the length of the acculturation process.
Immigrant background information
--------------------------------
Immigrants have tended to settle in Norway\'s capital, Oslo, where they represented almost one-fifth of the population in 2001 \[[@B29]\]. In the HUBRO youth survey 66% of the Pakistani youth were second generation (i.e. born in Norway by two foreign-born parents); among Turkish youth, 42%; and among Iranian youth, only 3%. Immigrants from non-industrialised countries tended to concentrate in some city areas, but no specific enclaves had been established. Table [1](#T1){ref-type="table"} shows immigrant population characteristics related to aspects of integration and participation in Norwegian society \[[@B39]\]. All three countries studied had low levels of recorded alcohol consumption during the period from 1970 to 2000, the period of immigration to Norway. Average recorded adult (15+ years) per capita consumption of pure alcohol 1970--2000 was 0.13 litres for Iran, 0.02 for Pakistan, and 1.19 for Turkey. The comparable figure for Norway was 5.17 litres. Unrecorded consumption is not included and thus all figures may be too low \[[@B40]\].
######
Immigrant group and population characteristics in Norway
**Population characteristics** **Pakistan** **Turkey** **The Islamic Republic of Iran** **Population 2005**
--------------------------------------------------------------------------- ------------------ ------------------ ---------------------------------- ---------------------
Status first generation Migrant workers Migrant workers Refugees \-
Main settlement periods 1970 and onwards 1970 and onwards 1986 -1990, 1997-2003 \-
Marriages with non-immigrant population in 2^nd^ generation Males/females 2.3%/ 4.2%/ 16.6%/ \-
1.3% 3.1% 19.7%
Number of children, 1^st^ generation 3.6 2.6 2.0 1.8
Norwegian citizenship 2005 77% 75% 66% 95.4%
Voting participation 2005 54% 43% 51% 77%
Education participation, 2^nd^ generation 16--18 years. Males/females 88%/90% 82%/79% 89%/95% 90%/92%
Work force participation. Males/females 60%/28% 59%/37% 54%/45% 73%/66%
Source: Kristin Henriksen, Statistics Norway \[[@B38]\].
On the host country national level, the immigrant population studied varied with respect to the status of immigration (migrant workers vs. refugees), main settlement period (indicating differences in the length of the acculturation period), tendency to marry one's own culture members, number of offspring, Norwegian citizenship, participation in voting, participation in upper secondary school level of education and workforce participation (see Table [1](#T1){ref-type="table"}).
Even though the picture is not clear-cut, Pakistanis were farther away from the ethnic Norwegians than the Turks and the Iranians regarding the factors measured (more children, higher level of marrying fellow citizens, lower level of workforce participation by women). Such differences among immigrants are also found in the sample (see Table [2](#T2){ref-type="table"}). Iranians had a shorter mean stay in Norway, a lower mean age and a higher education level than the other immigrants. Pakistanis included the highest proportion of Muslims.
######
Sample characteristics of immigrants in Oslo adult population sample by country background
**Pakistan** **Turkey** **The Islamic Republic of Iran**
------------------------------------------------------------ -------------- ------------ ----------------------------------
Gender,% female 41 42 41
Mean age (SD) 44.2 (9.4) 42.3 (7.7) 41.9 (7.1)
Muslim faith,% 96 85 51
Mean number of years in Norway (SD) 20.1 (8.2) 17.2 (8.0) 12.0 (4.4)
Mean number of years in school (SD) 10.9 (3.6) 9.4 (4.4) 14.5 (3.9)
Workforce participation male, no/part time/full time (%) 22/6/72 32/5/63 30/8/62
Workforce participation female, no/part time/full time (%) 65/19/16 49/16/35 39/20/41
Statistical methods
-------------------
Demographic, social and economic characteristics of the three immigrant groups were reported from population statistics compiled by Statistics Norway and by the samples.
Frequency of alcohol use in the parent generation, 30--60 years of age, and in the young generation, 15--16 years of age, are shown in tables for gender and country background. Chi square tests were used for testing differences in the distribution of alcohol drinking frequency for groups.
A structural equation modelling (SEM) approach was used to model alcohol drinking frequency for adult immigrants. First a factor analysis approach was used to reduce the dimensions of the six acculturation variables and establish the measurement part of the model. The factors found were used as intermediate latent endogenous variables. Exogenous variables were gender, age, Muslim faith, years of education, participation in the workforce and length of residence in Norway. The structural model was developed in AMOS (Analysis of Moment Structures) version 17.0 run within the IBM SPSS Statistics version 19.0, using the modification indices approach \[[@B41]\]. Three goodness-of-fit measures are reported: a Chi^2^ statistic, CFI (Comparative Fit Measure) and RMSEA (Root Mean Square Error of Approximation). An acceptable fit can be based on a value of CFI higher than 0.90 and a good fit higher than 0.95 (possible range zero to one), while a value of RMSEA lower than 0.05 represents a good fit and lower than 0.08 an acceptable fit (zero is the lowest possible value) \[[@B42]\]. The acceptable model was run for the whole dataset using the full information maximum likelihood approach (FIML) option for imputation of missing values. This model was applied to each country, and finally a multi-group analysis was carried out to test for equality of parameters across countries.
In the final, most parsimonious model with best fit, associations between the exogenous variables and the outcome alcohol frequency can be both direct and indirect through one or more latent variables. The indirect effects are calculated as the product of any direct effect of an exogenous variable on the latent variables, multiplied by the direct effect of the latent variables on alcohol frequency. The total effect will be the sum of the direct and the indirect effects.
A significance level of 5% was used for all hypothesis tests. Few p-values for tests are reported in the text, but statements of differences are all based on tests using the 5% level.
Results
=======
Comparisons of drinking
-----------------------
Frequency of drinking among adults was significantly different between ethnic Norwegians and all three immigrant groups, both for males and females. Among the immigrant groups, Iranians reported the highest frequency of drinking, Turks were at a middle level and Pakistanis reported the lowest frequency of drinking both for males and females (see Table [3](#T3){ref-type="table"}).
######
Alcohol frequency in Oslo adult population, by country of birth and gender
**Country of birth** **Drinking weekly** **Drinking monthly** **Drinking seldom** **Never drunk alcohol** **Total** **N**
------------------------------ --------------------- ---------------------- --------------------- ------------------------- ----------- -------
Norway 54 30 14 2 100 12259
Males 62 26 10 1 100 5071
Females 48 33 17 2 100 6188
The Islamic Republic of Iran 18 30 38 14 100 604
Males 24 35 35 6 100 385
Females 7 21 44 28 100 219
Turkey 12 14 25 48 100 436
Males 19 22 30 29 100 261
Females 3 2 18 77 100 175
Pakistan 2 3 12 82 100 607
Males 4 5 21 70 100 335
Females 1 0 1 97 100 272
Age groups for ethnic Norwegians 30, 40, 45, 59--60, age groups for immigrants 30--60 years.
Comparing equality of alcohol frequency distributions, all significantly different with p \< 0.000: Norwegian vs. Iranian vs. Turkish vs. Pakistani males: Chi^2^ = 3157 (12 df), Norwegian vs. Iranian vs. Turkish vs. Pakistani females: Chi^2^ = 4037 (12 df), Norwegian vs. Iranian males: Chi^2^ = 351 (6 df), Norwegian vs. Iranian females: Chi^2^ = 667 (6 df), Norwegian vs. Turkish males: Chi^2^ = 981 (6 df), Norwegian vs. Turkish females: Chi^2^ = 2481 (6 df), Norwegian vs. Pakistani males: Chi^2^ = 3081 (6 df), Norwegian vs. Pakistani females: Chi^2^ = 2481 (6 df), Turkish vs. Iranian males: Chi^2^ = 65.6 (6 df), Turkish vs. Iranian females: Chi^2^ = 98,1 (6 df), Pakistani vs. Iranian males: Chi^2^ = 343 (6 df), Pakistani vs. Iranian females: Chi^2^ = 259 (6 df), Pakistani vs. Turkish males: Chi^2^ = 117 (6 df), Pakistani vs. Turkish females: Chi^2^ = 48.1 (6 df).
Among young people, 15--16 years of age, the reported frequency of drinking was significantly lower among those with an immigrant background than among ethnic Norwegians, both for boys and girls (see Table [4](#T4){ref-type="table"}.) Iranian and Turkish boys reported a higher frequency of drinking than Pakistani boys, while there was no significant difference between Iranian and Turkish boys. For girls we see the same result as for female adults. In addition, there was no significant difference between youth with two vs. one Norwegian parent.
######
Alcohol frequency in the Oslo youth population, 15-16-year-olds, by country background and gender
**Country background** **Drinking weekly** **Drinking monthly** **Drinking seldom** **Never drunk alcohol** **Total** **N**
------------------------------ --------------------- ---------------------- --------------------- ------------------------- ----------- -------
Two Norwegian parents 21 34 32 12 100 4588
Males 21 33 32 14 100 2241
Females 22 35 32 11 100 2340
One Norwegian parent 22 33 33 12 100 804
Males 23 35 28 14 100 409
Females 20 31 38 10 100 391
The Islamic Republic of Iran 12 19 32 37 100 93
Males 14 23 21 42 100 43
Females 10 16 42 32 100 50
Turkey 4 10 15 70 100 105
Males 7 15 20 59 100 46
Females 2 7 12 79 100 58
Pakistan 1 4 6 89 100 559
Males 2 6 10 82 100 287
Females 0 1 1 97 100 268
Comparing equality of alcohol frequency distributions, all significantly different with p \< 0.000: Norwegian vs. Iranian vs. Turkish vs. Pakistani males: Chi^2^ = 646 (12 df), Norwegian vs. Iranian vs. Turkish vs. Pakistani females: Chi^2^ = 1051 (12 df), Norwegian vs. Iranian males: Chi^2^ = 26.7 (6 df), Norwegian vs. Iranian females: Chi^2^ = 28.7 (6 df), Norwegian vs. Turkish males: Chi^2^ = 71.3 (6 df), Norwegian vs. Turkish females: Chi^2^ = 238 (6 df), Norwegian vs. Pakistani males: Chi^2^ = 694 (6 df), Norwegian vs. Pakistani females: Chi^2^ = 1126 (6 df) For the following comparisons the p-value is given explicitly: Turkish vs. Iranian males: Chi^2^ = 3.08 (6 df), p = 0.80, Turkish vs. Iranian females: Chi^2^ = 24.7 (6 df), p \< 0.000, Pakistani vs. Iranian males: Chi^2^ = 40.0 (6 df), p \< 0.000, Pakistani vs. Iranian females: Chi^2^ = 162 (6 df), p \< 0.000, Pakistani vs. Turkish males: Chi^2^ = 14.2 (6 df), p = 0.03 Pakistani vs. Turkish females: Chi^2^ = 30.8 (6 df), p \< 0.000. Finally, the alcohol frequency distribution was not significantly different for those with two vs. one Norwegian born parent: for males Chi^2^ = 2.83 (6 df), p = 0.83 and for females Chi^2^ = 6.16 (6 df), p = 0.41.
Adult females in all groups reported a lower frequency of drinking than males (see Table [3](#T3){ref-type="table"}). For adolescents there were no significant gender differences for ethnic Norwegians or group-wise for those with immigrant backgrounds. But seen as a whole, males with an immigrant background had a significantly different drinking frequency than females with the same background (Table [4](#T4){ref-type="table"}).
Analysis of drinking among adult immigrants
-------------------------------------------
An exploratory factor analysis was run, which included the six variables measuring language/reading competence and social contact with own countrymen and Norwegians. It yielded three factors with eigenvalue larger than 1, explaining 75% of the variance. The variables loaded on the two first factors in a way that was expected from the literature in addition to a third factor, which was called social interaction irrespectively of own or host orientation (see also Figure 1, upper part):
1\. Host culture competence (high level of Norwegian language skills/often read Norwegian newspaper/often had a Norwegian visitor/often had support from a Norwegian)
2\. Own culture competence (often read newspaper in own language/often participated in meetings arranged by own countrymen)
3\. Social interaction (often had a Norwegian visitor/often had support from a Norwegian/often participated in meetings arranged by own countrymen)
Rotation techniques did not improve or change the interpretation of the factors.
{#F1}
The structural equation model with drinking frequency as outcome (four categories) thus included six exogenous variables (gender, age, Muslim faith, years in Norway, years in school and work participation) and the three endogenous variables identified above. In the baseline model all exogenous variables could be correlated, as could the three endogenous variables. In addition all direct and indirect effects were included. Correlation between the exogenous variables in the regression on the latent variables in the structural model was highest between age and time in Norway (0.40) for the total sample. A somewhat higher value was found for this correlation for Pakistanis (0.47), while all other correlations were lower. Thus colinearity was not an issue in the regression part of the structural equation model.
Using the modification indices approach to remove all non-significant correlations and effects, an acceptable fit was obtained for the model shown in the lower section of Figure [1](#F1){ref-type="fig"}. Host culture competence and own culture competence mediated the association between the exogenous variables and frequency of alcohol drinking. The third factor, social interaction, did not mediate any association between the exogenous variables and alcohol frequency, but the variable's presence was decisive for acceptability of the model. The CFI dropped to 0.75 when the social interaction factor was excluded from the model. Thus this factor had an overall association with frequency of drinking, regardless of group characteristics. Host culture competence and social interaction were positively associated with drinking frequency, while own culture competence was negatively associated with drinking frequency. Own and host culture competence were positively correlated (0.295, p \< 0.05), while social interaction was not correlated with those. The regression weights, etc., are shown in the first column of Table [5](#T5){ref-type="table"}.
######
Regression weights with standard error, correlation and fit measures in adult immigrant sample, by country background
**Regressions** **All** **The Islamic Republic of Iran** **Pakistan** **Turkey**
------------------------------------------------ ------------------ ---------------------------------- ------------------ ------------------
N = 982 N = 389 N = 323 N = 270
Alcohol frequency
Host culture competence^1^ 0.293 (0.061)\* 0.175 (0.108) 0.244 (0.077)\* 0.240 (0.122)\*
Own culture competence^1^ −0.409 (0.103)\* −0.055 (0.066) −0.476 (0.164)\* −0.219 (0.207)
Social interaction^1^ 0.190 (0.066)\* 0.176 (0.086)\* 0.089 (0.101) 0.466 (0.154)\*
Gender^2^ 0.678 (0.068)\* 0.653 (0.090)\* 0.356 (0.087)\* 0.913 (0.156)\*
Muslim faith^2^ −0.577 (0.071)\* −0.104 (0.086) −0.831 (0.181)\* −0.503 (0.157)\*
Years in Norway^1^ −0.013 (0.004)\* 0.026 (0.010)\* −0.006 (0.005) 0.003 (0.007)
Years in school^2^ 0.028 (0.009)\* 0.019 (0.013) 0.012 (0.015) 0.046 (0.017)\*
Host culture competence
Gender^2^ 0.117 (0.046)\* −0.018 (0.057) 0.196 (0.101) 0.212 (0.089)\*
Age^2^ −0.030 (0.003)\* −0.025 (0.004)\* −0.039 (0.005)\* −0.034 (0.006)\*
Years in Norway^1^ 0.034 (0.003)\* 0.043 (0.007)\* 0.054 (0.006)\* 0.036 (0.006)\*
Years in school^2^ 0.083 (0.005)\* 0.049 (0.008)\* 0.098 (0.012)\* 0.089 (0.011)\*
Work participation^1^ 0.240 (0.026)\* 0.182 (0.033)\* 0.230 (0.053)\* 0.256 (0.049)\*
Own culture competence
Gender^1^ 0.366 (0.060)\* 0.282 (0.096)\* 0.214 (0.101)\* 0.621 (0.113)\*
Muslim faith^1^ 0.136 (0.068)\* −0.103 (0.093) 0.128 (0.217) −0.043 (0.161)
Years in Norway^1^ 0.011 (0.004)\* −0.001 (0.011) 0.012 (0.006)\* −0.004 (0.007)
Years in school^1^ 0.015 (0.007)\* 0.021 (0.012) 0.064 (0.013)\* 0.046 (0.013)\*
Correlation own and host culture competence^2^ 0.295\* 0.060 0.534\* 0.516\*
Model fit
Chi-square 145\* 92\* 91\* 89\*
CFI 0.959 0.930 0.947 0.936
RMSEA 0.053 0.059 0.064 0.069
^1^ Non-variation (similarity in parameter values) between country background not rejected.
^2^ Non-variation rejected ^\*^Significant at five percent level.
Being male and having attended school for many years were positively associated with drinking frequency, while being a Muslim and having lived in Norway for a long time were negatively associated with this outcome. Those having lived in Norway longest were mainly Pakistanis with low alcohol consumption. No direct association was found between age and drinking frequency, or between work participation and drinking frequency. Indirectly, however, there were associations between those variables and alcohol frequency through host culture competence. Age was negatively associated with host culture competence (−0.030 (se 0.003)), which was positively associated with alcohol frequency (0.293 (se 0.061)) and thus a total negative association was established between age and drinking frequency. Work participation was in the same way positively associated with alcohol frequency.
In addition to the direct negative association between the Muslim faith and drinking frequency, an additional negative association was mediated through a higher level of own culture competence. In addition, the positive association between the number years attending school and frequency of drinking was strengthened by the mediation through own and host culture competence. Regarding gender and years in Norway, the indirect association mediated by host and own culture competence implied a reduction of the direct effects.
When country-specific analyses were run, the CFIs were somewhat reduced, while the RMSEAs were somewhat increased for each country -- indicating a poorer, but acceptable fit for each country. The same structural model was applied for the three countries, i.e. the non-significant parameters were not removed. Fewer significant results were present in the country-specific analyses, as seen by the lower number of significant coefficients in columns 2--4 in Table [5](#T5){ref-type="table"}. As the fit for each country was acceptable, a multi-group analysis for country background was conducted. The measurement part of the model (weights and means) was assumed equal for each country as non-variation of this part of the model could not be rejected (Chi^2^ = 25.7 for weights with 18 degrees of freedom, p = 0.11 and Chi^2^ = 19.9 for means with 14 degrees of freedom, p = 0.13 in a nested model).
Non-variation was rejected for the rest of the model structure, however, indicating significant differences between country backgrounds (Chi^2^ = 372.2 with 26 degrees of freedom, p \< 0.00). Differences in parameters between countries were identified for the association between alcohol frequency and the three variables gender (Chi^2^ = 22.2 with 2 degrees of freedom, p \< 0.00), being a Muslim (Chi^2^ = 20.2 with 2 degrees of freedom, p \< 0.00) and years in school (Chi^2^ = 6.7 with 2 degrees of freedom, p = 0.035). For gender the association (direct and indirect) was strongest for Turks and weakest for Pakistanis (see columns 2--4 in Table [5](#T5){ref-type="table"}). Thus differences in drinking between males and females were greatest for Turks and smallest for Pakistanis when controlled for the other variables included. Being a Muslim showed the strongest positive association for Pakistanis, while no association was found for Iranians. Thus being a Muslim among Iranians did not imply differences in frequency of drinking from non-Muslims. Finally, for years in school the strongest association with alcohol drinking was found for Turks, and no association was found for the two other countries (see columns 2--4 in Table [5](#T5){ref-type="table"}).
Differences in the structural model were also identified for the regression weights for the exogenous variables on the latent variables. Years in school on host culture competence (Chi^2^ = 8.7 with 2 degrees of freedom, p = 0.013) and age on host culture competence (Chi^2^ = 26.1 with 2 degrees of freedom, p \< 0.00) were lowest for Iranians. Gender on own culture competence (Chi^2^ = 6.4 with 2 degrees of freedom, =0.041) was highest for Turks. In addition, the correlation between host and own culture competence was significantly different (Chi^2^ = 7.5 with 2 degrees of freedom, p = 0.023). This correlation was lowest for Iranians.
Among Iranians, direct associations with drinking frequency were found for gender and years in Norway (see Table [5](#T5){ref-type="table"}). In addition, social interaction was associated with drinking frequency. An indirect association through own culture competence for gender and indirect associations through host culture competence for years in Norway and years in school, age and work participation could have been present. But own and host culture competence had no significant association with frequency of drinking for Iranians, and thus indirect associations could not be stated.
Among Pakistanis, direct associations with drinking frequency were found for gender and being a Muslim. Several indirect associations through own and host culture competence were also present, as the associations between those variables and drinking frequency were significant. There were no significant associations between social interaction and frequency of drinking (see Table [5](#T5){ref-type="table"}).
Among Turks, direct associations with drinking frequency were found for gender, being a Muslim and years in school. Several indirect associations through host culture competence were present, as the associations between this variable and drinking frequency were significant. Own culture competence was not significantly associated with frequency of drinking. There was also a significant association between social interaction and frequency of drinking (see Table [5](#T5){ref-type="table"}).
Discussion
==========
Summary of main findings
------------------------
Adults and youth with an ethnic Norwegian background reported more frequent alcohol drinking than immigrants with a background from Iran, Turkey and Pakistan. Iranians reported a higher drinking frequency than Turks and Pakistanis. For the three immigrant countries high host culture competence and social interaction were associated with a higher drinking frequency, while high own culture competence was associated with a lower drinking frequency. Muslim immigrants reported significantly lower drinking frequency than non-Muslims, and those who had lived for a longer time in Norway and were of a higher age drank less frequently. Those with a higher level of education and work participation drank more frequently.
Multi-ethnic comparisons
------------------------
Iranians had the shortest mean stay in Norway (almost all of the 15-16-year-olds were first-generation immigrants), but a higher level of drinking than Pakistanis and Turks. They were closest to the population characteristics of the Norwegian population (Table [1](#T1){ref-type="table"}). Direct associations with drinking frequency were found for gender and years in Norway. In addition, social interaction was associated with drinking frequency, while own and host culture competence were not. Iran had the lowest proportion of Muslims (50%). As there was no significant difference in drinking for Muslims and others among Iranians, this did not have any impact on drinking frequency. A possible interpretation is that the Koran's prohibition of alcohol is not obeyed to the same extent among Iranians as by immigrants from the other two countries. Iranians also had the highest average number of years in school, but this variable also lacked a significant association with drinking frequency. As the variable 'years in Norway' was positively correlated with drinking, Iranians may not have the same strict social norms regarding alcohol consumption as the Pakistanis and Turks, and are thus more prone to increase their drinking in a 'wetter' society \[[@B43]\]. The insignificant difference in drinking frequency between Muslim Iranians and other Iranians is an exception \[[@B7],[@B9],[@B10],[@B12],[@B44]-[@B46]\]. Donath et al. found that adolescents from Islamic countries in Germany had lower lifetime prevalence of drinking than German and Western European adolescents, except for students from Iran \[[@B11]\].
Pakistanis had the longest mean stay in Norway, and the lowest level of drinking. The second generation in Norway often married a cousin or other close family member from Pakistan. Thus the family-oriented Pakistani culture and drinking pattern has continued in Norway. Direct associations with drinking frequency were found for gender and being a Muslim. Several indirect associations through own and host culture competence were also present, and thus such culture competences played a larger role among Pakistanis than among Iranians. The general latent variable 'social interaction' did not play a role in relation to drinking, however. The high percentage of persons of Muslim faith (96%) and the low level of education may have strengthened the Muslim imperative to abstain from alcohol.
Turks were a group in the middle position for almost all measurements: population characteristics (Table [1](#T1){ref-type="table"}), sample characteristics (Table [2](#T2){ref-type="table"}) and drinking frequency (Tables [3](#T3){ref-type="table"} and [4](#T4){ref-type="table"}). The recorded level of alcohol consumption in Turkey was higher than in Pakistan and this may have played a role. Direct associations with drinking frequency were found for gender, being a Muslim and years in school. Several indirect associations between exogenous variables and drinking through host culture competence were present, albeit not through own culture competence. For Turks, taking part in the host culture was thus more important for drinking than taking part in their own culture. The small number of cases in which Turks married non-immigrants may have kept the Turkish cultural expressions at a high level, as it did among Pakistanis, but the impact this had on drinking was lower. Also, a significant positive association between social interaction and frequency of drinking was present. The fact that the proportion of Muslims was high (85%) played a role.
Acculturation and socio-demographic factors
-------------------------------------------
Young immigrants of the first and second generations from Iran, Pakistan, and Turkey drank less than ethnic Norwegians, as did adults from the same countries. This is in line with earlier studies among Turks, Moroccans, Antilleans, Surinamese, Asian Muslims, Hindus, Sikhs, Vietnamese in Europe \[[@B5]-[@B12]\], those of Chinese origin or Mexican heritage, Japanese, Filipinos, Koreans and Vietnamese in the US \[[@B15]-[@B17]\], and Vietnamese in Australia \[[@B19]\]. Although prevalence of alcohol use and frequency of drinking have been generally lower among non-Western immigrants to Western countries, alcohol-related problems may be higher than or equal to such problems in the host population \[[@B9],[@B47],[@B48]\].
Pakistani and Turkish youth adhered to the low drinking frequency of the first generation of adults. In an earlier study in Oslo, young immigrants from Morocco, Pakistan, Turkey and Vietnam reported higher frequency of use in the second generation than in the first, by country and gender, although this was not significant \[[@B12]\]. In the Netherlands, Turkish and Moroccan immigrants reported less alcohol use than the Dutch population in both the first and the second generations, and their alcohol consumption did not converge towards the higher rates in the host population \[[@B5],[@B6]\]. There may be an increase in drinking over time in the host country, but it will take more than two generations before differences disappear between ethnic Norwegians and groups with most non-Western immigrant backgrounds, especially when a high percentage of them are Muslims.
Lower consumption of alcohol among females than males is a worldwide finding. The exception for Norwegian adolescents seen here has been recognised in other studies \[[@B27]\]. The positive association between frequency of drinking and both years of education and work participation (indirectly by host culture competence) is also in line with earlier studies \[[@B4],[@B34]\].
Time in the host country has been used as an indicator of integration or assimilation for immigrants. In this study drinking level was negatively associated with both mean stay in Norway and age. This reflects the fact that those with the longest mean stay were older, low-consuming Pakistanis and Turks. Country-wise analyses showed a positive association between drinking and mean stay only among Iranians. Thus the use of length of stay as an indicator of immigrant behaviour becoming closer to the host behaviour was group-dependent and not dominant.
Social interaction
------------------
The latent variable called 'host culture competence' loading on the four variables 1) Norwegian language skills, 2) reading of Norwegian newspapers, 3) had Norwegian visitors and 4) had support from Norwegians, also had a social interaction component: knowledge of the host language was necessary for reading and for socialising with the host community. The variable 'own culture competence' loaded on two variables: 1) reading newspaper in own language and 2) participating in meetings arranged by own countrymen. Thus this variable also included a social component. The variables were positively correlated, indicating that the same person tended to score high (or low) on both variables. Thus host and own culture competence could not be measured as opposite values along one dimension only. The direction of the association with drinking was as predicted for these factors: high host culture competence was associated with a higher frequency of drinking, while high own culture competence was associated with a lower frequency. The third latent variable, social interaction, loaded on the social interaction variables mentioned above (see also Figure [1](#F1){ref-type="fig"}). The variable did not mediate any association from the exogenous variables on drinking and was not correlated to the two other latent variables, but was necessary for acceptance of the model. The association with drinking was positive for all immigrants, as well as country-wise for Iran and Turkey. This means that social interaction is a common quality, which in the Norwegian setting was linked to a higher level of drinking even when host and own culture competence, which imbedded social aspects, was included in the model. This is in line with Skog's theory based on social interaction, that people will behave collectively with regard to drinking \[[@B23],[@B49]\].
Consequences of low level of drinking among immigrants
------------------------------------------------------
The existence of large immigrant groups of Muslim low-level alcohol consumers in Western cities contributes to a lower level of consumption at the population level and thus to lower levels of alcohol-related harm and disease in the population. First, the contribution is due to a low level of alcohol use among immigrants themselves, which applies to both the first and the second generation. The distance in drinking level to ethnic Norwegians is so long that it may take several generations to reach the host level for two of the three country backgrounds studied here. Second, a low level of drinking among immigrants who interact with ethnic Norwegians may influence the hosts to reduce their level of drinking. Dominguez and Maya-Jariego point out that interaction between members of both immigrant and host groups has an impact on the attitudes, values and behaviours of both collectives \[[@B24]\]. Earlier studies among young people showed that a high percentage of Muslim immigrants were associated with reduced drinking also among young people from the host background \[[@B7],[@B12]\]. A similar study has not been conducted among adults. Based on the lasting low level of drinking among non-Western immigrants, the associations between drinking and the three factors own and host culture competence and social interaction found here, and also theoretical effects of social interaction on drinking, we can expect to observe reduced drinking also among ethnic Norwegians in the future.
Weaknesses
----------
Self-reported measures of alcohol use have demonstrated reasonable levels of reliability and validity, but many factors may influence such reports \[[@B50]\]. One such factor is the social desirability of drinking, which presumably is lower for persons with a 'dry' immigrant background compared to persons with a Norwegian background. On the other hand, in most surveys of alcohol consumption reporting of 40% - 60% can be assumed, but we do not know whether this varies by ethnicity \[[@B51]\]. Thus the distance from the 'true' difference in drinking vs. the measured one is difficult to ascertain.
Response rates for the youth part of the HUBRO were satisfactory (almost 90%), while they were much lower for the adult cohorts (46%) and the immigrant survey (40%). These were response rates based on the whole population of Iranians and Turks, not response rates for a drawn sample from those populations. In the immigrant survey, a 30% sample was first drawn of Pakistanis. Self-selection may be very important, however. Self-selection according to socio-demographic variables was shown to have little impact on several prevalence estimates in the adult survey and social inequality in health by different socio-demographic variables seemed unbiased \[[@B52]\]. A similar investigation of self-selection has not been conducted for the immigrant study, however. Since the setting was the same, it is plausible that the situation could be similar. Analyses that particularly addressed immigrant groups supported this \[[@B53]\].
Cross-sectional studies have limitations and shortcomings in the study of longitudinal social processes. The result of the acculturation process on drinking was measured at the same time as the factors assumed to influence the drinking. This is not a major problem if socialisation patterns are stable, but it is relevant to assume that such patterns develop during the acculturation process. It is a huge challenge to establish prospective studies in immigrant samples, however. It is also seldom that a cross-sectional sample as large as the one used here can be established among immigrants. Even in this study, however, some country-specific results were in line with results for the total sample, but not significantly so. This may partly be a result of less power due to a lower number of observations for each country's background.
The three surveys employed here were conducted in 2000--2002, after 30 years of steady immigration from non-Western countries. During the 10--12 years since then, there has been a further increase in immigration from non-Western countries, and the country background has been extended to Somalia, Iraq and others. Thus the research questions addressed here have not become less topical.
Conclusions
===========
The existence of large immigrant groups with a low frequency of alcohol drinking in Oslo contributed to a lower level of consumption at the population level. Social integration and own and host competence were associated with drinking, as were being a Muslim, years in school and gender. Results were specific to the particular country backgrounds, however. Based on the lasting low level of drinking among non-Western immigrants, the established associations between own and host culture competence and social interaction, and also theoretical effects of social interaction on drinking, we can expect to observe reduced drinking also among ethnic Norwegians in the future. This may imply reduced alcohol-related harm and disease.
Competing interest
==================
There are no competing interests. The study was financed by the Norwegian Institute for Alcohol and Drug Research.
Author contribution
===================
The sole author conceived the idea, carried out the analyses and wrote this manuscript.
Author's information
====================
The author has worked with the topic of substance use among immigrants for several years and has published nationally and internationally plus reviewed articles on the subject for peer-reviewed journals.
Pre-publication history
=======================
The pre-publication history for this paper can be accessed here:
<http://www.biomedcentral.com/1471-2458/12/535/prepub>
Acknowledgements
================
I thank the Norwegian Institute of Public Health for carrying out the Oslo Health Survey and, in addition, I thank all the respondents.
| {
"pile_set_name": "PubMed Central"
} |
Background {#Sec1}
==========
In the past 20 years, enhanced recovery pathways (ERPs) have become increasingly integrated into most surgical fields as standard care in high income countries, as is exemplified by national priority programs \[[@CR1]--[@CR3]\], and the widespread acceptance of the Enhanced Recovery After Surgery (ERAS) society network \[[@CR4]\]. ERPs represent a fundamental shift towards a patient-centred, multidisciplinary-driven continuity of care that aim to attenuate surgical stress and expedite recovery \[[@CR5]\]. Studies on total joint arthroplasty (TJA) for both hips and knees have shown that implementation of an evidence-based, structured approach to patient care decreases postoperative morbidity and consequently length of stay without increasing readmission rate \[[@CR6]--[@CR8]\].
However, in low- and middle-income countries (LMICs), the value of implementing ERPs is yet to be explored. This may be because: i) the perception that current hospital resources may make it difficult to develop and implement structured and sustainable protocols to enhance postoperative recovery, and ii) short and long-term data collection on the quality of the work provided is scarce, inhibiting the ability to benchmark clinical results and improve the service provided to patients. Despite these challenges, a healthcare system in a middle-income country such as South Africa may benefit from the implementation of ERPs through reduced postoperative morbidity and the associated cost reductions, as has been demonstrated in high-income countries (HICs) \[[@CR9]\].
While the goals of implementing ERPs can be expected to be independent of a country's economic status, we believe the differences in patient demographics, healthcare infrastructure and healthcare resources between HICs and LMICs warrants a LMIC derived programme of enhanced care to facilitate practice change and improve patient outcomes in these settings. The aim of our study was therefore to establish multidisciplinary consensus on; i) preoperative risk factors associated with poor outcomes, ii) perioperative interventions considered necessary to improve outcomes, and iii) important postsurgical patient and clinical outcomes. This study was conducted in South Africa, which represents an upper-middle-income country, as defined by the World Bank \[[@CR10]\]. However, as this work was conducted in the public healthcare sector, and South Africa has one of the world's highest levels of inequality \[[@CR11]\], it is likely that this work reflects the state funded healthcare system of a LMIC, as opposed to high-middle-income countries. This assumption is supported by the South African public healthcare service data from the African Surgical Outcomes Study, where the median number of specialists per 100,000 population was 0.9 (IQR 0.2-1.9) (unpublished data) \[[@CR12]\], which is well below the recommended 20--40 specialists per 100,000 population \[[@CR13]\].
Methods {#Sec2}
=======
We conducted a Delphi survey with experts from different fields involved in the care of arthroplasty surgical patients in South Africa. The Delphi study is an accepted method for achieving convergence of opinions concerning knowledge solicited from experts within specific fields, and has been adopted for priority-setting in medicine \[[@CR14]\]. The technique is an iterative process which allows the participant to refine his or her prioritization of items, in an anonymous manner, based on the group's work from round to round and with controlled feedback of opinions \[[@CR15]\].
Participant recruitment {#Sec3}
-----------------------
Participants were recruited from all the hospitals which we knew had a history of performing elective TJAs. This approach was necessary, as currently there is no national arthroplasty database of public hospitals performing TJAs in South Africa. We invited orthopaedic arthroplasty surgeons, anaesthetists and physiotherapists from 18 regional and central hospitals in the public sector covering seven of the nine provinces in South Africa. They were contacted by email and asked to participate in four sequential studies aimed at improving perioperative care for patients scheduled for primary elective unilateral hip and knee TJA in South Africa. The Delphi study is the first of these four studies. For a hospital to participate we required participation of both the Anaesthesia and the Orthopaedic Departments in the project. With the use of telephone calls, face-to-face meetings and further email correspondence, 33 experts in the perioperative management of arthroplasty patients from 10 hospitals representing four provinces accepted the invitation to participate in these four studies. Reasons for exclusion from the study where i) not confirming their participation (5) or ii) declining to participate due to lack of interest or lack of resources to participate in this and future studies (3). Prior to commencement of the Delphi study, the participants were given detailed information of the Delphi process and how consensus would be defined.
The Delphi process {#Sec4}
------------------
This Delphi survey was conducted over 3 months from December 2016 to March 2017. In the first round participants submitted suggestions for; i) risk factors associated with poor outcome, ii) best practices for preoperative, intraoperative and postoperative interventions to improve postoperative outcomes and iii) important patient and clinical outcomes to benchmark care, deemed relevant in the South African context for patients scheduled for primary elective unilateral hip and knee TJA. Participants were encouraged to elaborate on how to quantify these components and provide supporting references. UP and BMB grouped the responses in each category into statements. The category statements and supporting references were shared with all participants. In the second Delphi round, the participants were asked to rank the top-ten statements in each category, and where possible, add further comments or relevant references. Based on participants' responses, statements that overlapped were grouped together prior to the third Delphi round. In the third round the participants were presented with their individual as well as the overall group ranking of the prioritised statements within each category. They were asked to re-evaluate their previous round's ranking, considering the group ranking and where possible when their rankings differed greatly from that of the group, to add further comments or references supporting their decision. In the fourth and final round, participants were given an opportunity to present any strong disagreement with the priority rankings from the third Delphi round with a Skype teleconference. Non-participation in the fourth round indicated agreement with the proposed Delphi priorities from the third round. Following the teleconference, the consensus of the group was taken as final. UP and BMB were neutral in the prioritization of statements throughout the study.
Statistical analysis {#Sec5}
--------------------
The rank order of the research priorities for each round was established using a reverse scoring system i.e. a respondent's rank of 1 received 10 points, down to a rank of 10, which received 1 point. The scores of the respondents were combined for each round to develop the research priority rank order**.**
Results {#Sec6}
=======
Participants and response rate {#Sec7}
------------------------------
The recruited participants included 13 arthroplasty surgeons, 12 anaesthetists and 8 physiotherapists involved in hip and knee arthroplasty. Response rate in the first round was 97% (32/33), 91% (30/33) in the second round and 91% (30/33) in the third round. In the fourth round, all 33 participants accepted the ranking of the prior third Delphi round. However, three participants contributed in the fourth round to a refinement of two of the Delphi statements. The first was an amalgamation of "peripheral nerve blocks" with "multimodal opioid-sparing analgesia regimen" in the postoperative intervention category, which changed the overall ranking in this category. This change clarified that non-opioid analgesic regimens can include regional anaesthesia. The second change was to define "long term survival" in the outcome category as "1-year mortality", to ensure an objective outcome variable.
Preoperative risk factors {#Sec8}
-------------------------
Two hundred forty-seven suggestions were submitted for round 1 for preoperative risk factors believed to be associated with poor outcomes in patients scheduled for primary elective unilateral hip and knee TJA. The suggestions were categorised into 36 broad statements for round 2 which were refined to 28 statements for round 3. The ten prioritised risk factors identified after the second round did not change in the subsequent rounds (Table [1](#Tab1){ref-type="table"}).Table 1The ten prioritised preoperative risk factors considered most important determinants of poor outcomes in patients scheduled for primary elective unilateral hip and knee total joint arthroplasty in South Africa1. Poor general health (ASA 3 and above)2. Impaired cardiovascular functional status3. Advanced age4. Preoperative mobility5. Obesity or chronic malnutrition6. Recent or current sources of infection (e.g. bladder, respiratory, dental etc.)7. Preoperative chronic pain7. Matching surgical complexity with surgical experience or skill9. Psychiatric disorders and/or cognitive impairment10. Preoperative anaemia*ASA* American Society of Anesthesiologists
Preoperative interventions {#Sec9}
--------------------------
Round 1 yielded 166 suggestions of preoperative interventions judged to be important to improve outcomes following primary elective unilateral hip and knee TJA. These were amalgamated into 14 statements in round 2 and further refined to 11 different statements for round 3. The ten priorities identified after the second round did not change in subsequent rounds (Table [2](#Tab2){ref-type="table"}).Table 2The ten prioritised preoperative interventions considered most important determinants to improve outcomes following primary elective unilateral hip and knee total joint arthroplasty in South Africa1. A patient optimisation clinic2. Multidisciplinary planning3. Patient education4. Infection prevention5. Establishing high-volume units6. Smoking cessation7. Optimisation of preoperative analgesia regimen8. Minimise preoperative fasting9. Establish a patient blood management programme10. Alcohol cessation
Intraoperative interventions {#Sec10}
----------------------------
One hundred forty-four suggestions for intraoperative interventions believed to improve postoperative outcomes following primary elective unilateral hip and knee TJA were submitted in round 1. These were amalgamated into 18 statements for the second round and further refined to 11 statements for round 3. The ten priorities identified by the second round, did not change in the fourth round (Table [3](#Tab3){ref-type="table"}).Table 3The ten prioritised intraoperative interventions considered most important determinants to improve outcomes following primary elective unilateral hip and knee total joint arthroplasty in South Africa1. Meticulous surgical technique2. Infection prevention3. Optimisation of prosthesis choice and placement4. Multimodal opioid-sparing analgesia regimen5. Monitoring and optimisation of haemodynamics6. Central neuraxial anaesthesia7. Establish a patient blood management programme8. Temperature regulation9. Glycaemic control10. Deep vein thrombosis prophylaxis
Postoperative interventions {#Sec11}
---------------------------
The first Delphi round yielded 181 suggestions of important postoperative interventions to possibly improve outcomes following primary elective unilateral hip and knee TJA. These were amalgamated into 23 statements for the second Delphi round and further refined to 17 statements for the third Delphi round. The final ten priorities were agreed upon in the fourth round of the Delphi process, following amalgamation of "peripheral nerve blocks" into "multimodal opioid-sparing analgesia regimen" (Table [4](#Tab4){ref-type="table"}).Table 4The ten prioritised postoperative interventions considered most important determinants to improve outcomes following primary elective unilateral hip and knee total joint arthroplasty in South Africa1. Early mobilisation after surgery2. Standardised orthopaedic nursing care3. Multimodal opioid-sparing analgesia regimen4. Active management of medical co-morbidities5. DVT prophylaxis6. A pain management team7. Patient empowerment in his or her recovery8. Patient controlled analgesia9. Multidisciplinary ward rounds10. Postoperative rehabilitation
Important patient and clinical outcomes {#Sec12}
---------------------------------------
One hundred sixty-four suggestions were made in the first Delphi round for important patient and clinical outcomes following primary elective unilateral hip and knee TJA. These were categorised into statements for the second Delphi round and further refined to 23 statements for the third Delphi round. The ten prioritised outcomes did not change after the second round (Table [5](#Tab5){ref-type="table"}).Table 5The ten prioritised patient and clinical outcomes considered most important following primary elective unilateral hip and knee total joint arthroplasty in South Africa1. Patient reported outcome measures2. Postoperative pain at rest and during movement3. 1-year mortality4. Early mobilisation5. Prosthetic joint infection rate6. Joint range of motion7. Major adverse cardiac events8. Hospital length of stay9. Implant longevity10. Cost of care
Discussion {#Sec13}
==========
This study reports a national consensus of the predictors of morbidity, perioperative interventions to improve surgical outcomes, and the clinical outcomes necessary to document perioperative success for patients scheduled for primary elective unilateral hip and knee TJA in South Africa. These findings provide the information necessary to develop a feasible enhanced care programme for South African arthroplasty patients.
The multidisciplinary involvement of regional and central hospitals performing TJAs across South Africa provides a realistic consensus of the factors needed for an enhanced care arthroplasty programme in the public service in South Africa. We believe that the "buy-in" by the participants was high, and this is important for successful organisational change \[[@CR16]\]. Furthermore, consensus on the priorities was established early (within Delphi round 2) in four of the five categories, supporting the validity of the final consensus document \[[@CR17]\].
However, this study also has limitations. Firstly, while expert consensus is the lowest level of evidence, it is an established method to facilitate clinical guidelines when the evidence is limited \[[@CR18]\], particularly when study interventions and study results might not be transferable to settings with a different socio-economic and demographic structure. Furthermore, group consensus studies can expedite the transformation of evidence-based knowledge gained in HICs into practical implementation in LMICs \[[@CR19]\], which is why we believe this process is entirely appropriate for the public health service in South Africa, and may be applicable to other LMICs. In our study we have: i) identified feasible interventions which may improve patient outcomes in a resource limited environment, and ii) prioritized which interventions are preferable for implementation if resources do not allow for adoption of all suggested interventions in clinical practice. We believe this approach will allow all sites to focus their resources on developing a pragmatic multidisciplinary programme of enhanced care.
A second limitation is the possibility that we did not invite all sites which performs TJAs in South Africa to participate in the study, as the public health care sector currently does not have a national arthroplasty database. Nevertheless, we succeeded in enrolling both regional and central hospitals from different provinces, which ensured a broad representation of specialists involved in TJAs in South Africa. Finally, we did not include the full spectrum of stakeholders involved in the perioperative management of joint arthroplasty patients or patients themselves. However, we believe our consensus document does represent stakeholders who were not participants in this study, as patient relevant outcomes and parameters important to nursing care, physicians, nutritionists and geriatricians are included (Tables [1](#Tab1){ref-type="table"}, [2](#Tab2){ref-type="table"}, [3](#Tab3){ref-type="table"}, [4](#Tab4){ref-type="table"} and [5](#Tab5){ref-type="table"}).
Identification of modifiable and non-modifiable risk factors is essential to guide surgical decision making and prepare the patient optimally ensuring safe perioperative care \[[@CR20]\]. This is important in a country such as South Africa, which has a medium Human Development Index (HDI)[1](#Fn1){ref-type="fn"} suggesting a higher risk for perioperative mortality compared to countries with high HDI \[[@CR21]\]. Hence, addressing the prioritised preoperative risk factors (Table [1](#Tab1){ref-type="table"}) may improve patient outcomes \[[@CR22]\]. Additionally, introducing a best practice protocol in the perioperative period (Tables [2](#Tab2){ref-type="table"}, [3](#Tab3){ref-type="table"} and [4](#Tab4){ref-type="table"}) aims to provide continuity of care with emphasis on less variability and better quality of service provided \[[@CR23]\]. Finally, identifying and standardising procedure specific outcomes facilitates benchmarking, which is crucial to improve the quality of patient care \[[@CR24]\]. Only recently have such multinational collaborative efforts been instituted for TJAs to guide future trials towards comparable outcomes \[[@CR25]\]. Importantly, this international group of patient partners, orthopaedic surgeons, physical therapists, rheumatologists and methodologists successfully achieved consensus for six core outcome domains; i) pain, ii) function, iii) patient satisfaction, iv) revision, v) adverse events and vi) death, which are all represented in our consensus document (Table [5](#Tab5){ref-type="table"}). While this similarity provides external validity to the work of our Delphi group, it also suggests that aspirations for best patient practice is independent of a country's income status. However, the novelty of our Delphi study remains with the prioritised preoperative risk factors and perioperative interventions, which we hope will facilitate a pragmatic approach to achieving these postoperative goals in our resource limited settings.
Conclusion {#Sec14}
==========
This national multidisciplinary consensus Delphi study has produced priorities for preoperative risk stratification, perioperative interventions, and outcome assessments necessary for benchmarking, from which a pragmatic enhanced care programme for primary elective unilateral hip and knee TJA in South Africa can be developed. It is anticipated that these priorities may either be applicable or encourage other LMICs to initiate a similar Delphi process. The next phase will involve an audit of current perioperative care addressing the prioritised statements, followed by implementation of the Delphi group's proposed interventions.
ASA
: American Society of Anesthesiologists
DVT
: Deep vein thrombosis
ERAS
: Enhanced recovery after surgery
ERPs
: Enhanced recovery pathways
HDI
: Human Development Index
HICs
: High-income countries
LMICs
: Low- and middle-income countries
PBM
: Patient blood management
TJA
: Total joint arthroplasty
^\*^ Human Development Index (HDI) is an index based on i) life expectancy, ii) education and iii) per capita income indicators, which is used to determine whether a nation is a developed or a developing country.
The datasets used and analysed during the current study are available from the corresponding author on request.
UP and BMB were responsible for overall conception and design of the Delphi study; acquisition, analysis and interpretation of data; drafting the manuscript and critical revising of the work. MBN, LCM, JDJ, RP, NvdW, JFvdM, JM, WVS, GLD, TP, CS, PR, AMT, ZF, RS, CC, HS, SS, AM, HRH, OSP, NET, RES, CvdW, AJT, CAB, LAG, TWM, HKSS, PR, JGvdW, RIN and AT made substantial contributions to acquisition, analysis and interpretation of data and critical revising of the work. All authors approved the final version and agreed to be accountable for all aspects, accuracy and integrity of the work.
Ethics approval and consent to participate {#FPar1}
==========================================
The study was approved by the University of Cape Town, Faculty of Health Sciences Human Research Committee, South Africa; HREC REF: 807/2016. Written consent was obtained as participants consented to take part in the study by replying to each of the Delphi cycles via emails and their responses were stored in a password protected electronic format.
Competing interests {#FPar2}
===================
The authors declare that they have no competing interests.
Publisher's Note {#FPar3}
================
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
| {
"pile_set_name": "PubMed Central"
} |
Background {#Sec1}
==========
Tear film (TF) is a viscous and complex trilaminar fluid composed mainly of lipids, electrolytes, proteins and water \[[@CR1]\]. Normal TF dynamics require adequate tear production, tear retention on the ocular surface and balanced tear drainage. The TF is responsible for lubrication, nutrition and protection against microbial and toxic agents \[[@CR2]\]. Disruption of TF dynamics can lead to dry eye or the ocular surface can become more susceptible to the onset of diseases \[[@CR3]--[@CR5]\].
Different tools can be used to evaluate the ocular surface \[[@CR6]\]. Schirmer tear test-1 (STT-1) is considered the gold standard method for measuring tear production; however, it does not measure tear quality \[[@CR7]\]. The tear ferning test (TFT) is a qualitative test developed for humans, and it has become a useful diagnostic tool in tear ferning research. Various ferning patterns can be observed as a result of tear crystallization after evaporation of lacrimal samples, and these patterns depend mostly on TF composition \[[@CR8]\]. In addition, tear crystallization may be affected by humidity and temperature \[[@CR9], [@CR10]\]. Therefore, changes in ferning patterns are believed to reflect possible changes in both composition and stability of the TF \[[@CR10]\].
In humans, Rolando suggested the first grading scale for the TFT, wherein types I and II indicated normal TFs, while types III and IV indicated abnormal TFs \[[@CR9]\]. Thereafter, Masmali developed a TFT grading scale for humans, with the aim of addressing the gaps in previous classification systems. According to this scale, TFs with grades 0 and 1 were considered normal, while those with grades 2, 3 and 4 were considered abnormal \[[@CR10]\].
In addition to humans \[[@CR6], [@CR10], [@CR11]\], the Rolando and Masmali grading scales have been applied in horses \[[@CR12]\], dogs \[[@CR13], [@CR14]\], camels \[[@CR11]\] and capuchin monkeys \[[@CR15]\]. Although no record in the veterinary medicine literature documents the use of a grading scale specifically for animals, previous studies have successfully used the human grading scales in some animal species, thus showing that the TFT is a feasible and complementary test that is simple and inexpensive for ocular surface assessment \[[@CR12], [@CR13]\].
Increasing evidence suggests that qualitative TF deficiencies are an important cofactor or cause of some of the most common and challenging ocular diseases in cats, including conjunctivitis, corneal ulcer, spontaneous chronic corneal epithelial defects (SCCED), pigmentary keratitis, corneal sequestrum and dry eye syndrome \[[@CR16]--[@CR21]\]. Despite this, the TFT has not yet been performed in cats, even though it is already considered a complementary diagnostic tool in ocular surface research. Therefore, this lack of applied research on tear ferning in cats justifies an investigation of the application of the two TFT grading scales in healthy cats to understand the potential role of tear deficiency and the value of tear testing in cats.
Methods {#Sec2}
=======
Study population {#Sec3}
----------------
The study included 60 mixed breed cats (120 eyes) of which 33 were females (55.0%) and 27 were males (45.0%), aged between 1 and 8 years (mean 2.5 ± 1.86 years), with no complaint of illness. Additionally, they had to be vaccinated, without systemic or ocular signs of disease, without any history of ocular secretion or injury and with STT-1 values within the normal range for the species. The animals also had normal, complete blood counts and biochemical test results, including those for urea, creatinine, alanine aminotransferase and alkaline phosphatase. Moreover, all animals tested negative for feline coronavirus, feline leukemia virus and feline immunodeficiency virus.
Before data collection, the STT-1 (Schirmer Tear Test; Ophthalmos, São Paulo, Brazil) was performed and used as a screening method to measure tear production. The median value and interquartile range (median ± S-IQR) was 20 ± 7 mm/min, and the 95%-confidence intervals (CI) was between 18.4 and 19.8 mm/min.
Evaluations of the ocular adnexa and anterior segment were performed using a slit-lamp biomicroscope (Vision Class II BL IIIB/YZ30T; Ramos Mejia, São Paulo, Brazil). Intraocular pressure (IOP) was measured using a rebound tonometer (Icare tonometer; Icare Finland Oy, Vantaa, Finland); For IOP the median ± S-IQR value was 22 ± 6 mmHg, and the CI was between 22.3 and 24.1 mmHg. The ocular surface was evaluated using fluorescein dyes (Fluorescein test; Ophthalmos), TF breakup time (TFBT) (range 5--9 s) and lissamine green (Lissamine green test; Ophthalmos). The IOP, fluorescein, TFBT and lissamine green tests were performed after tear sampling to avoid any interference with tear crystallization. All data were collected in a room in which the temperature and humidity were controlled.
This study was approved by the Ethics Committee on Animal Experimentation of the Use of Animals from the State University of Santa Cruz, Brazil (protocol no. 003/17). All procedures were conducted in accordance with the Association for Research in Vision and Ophthalmology's (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research and NIH statement.
Sample collection and TFT {#Sec4}
-------------------------
Tear samples were collected between 8:00 a.m. and 11:30 a.m., first from the right eye and then the left eye. Once the tear wetted 30 mm on the Schirmer strips, which were the same ones used for STT-1, the strips were immediately placed in a 0.5 ml microtube (Protein LoBind Tubes; Eppendorf, São Paulo, Brazil) and conditioned in a thermal box until centrifugation. Immediately before centrifugation, the bottom of the 0.5 mL microtube was punctured and it was inserted into a larger 2.0 mL microcentrifuge tube (Protein LoBind Tubes; Eppendorf) for extracting the tear fluid as previously described by Oria et al. \[[@CR13]\]. The tear fluid was obtained through centrifugation (25.830 *g* for 10 min at 4 °C) of the Schirmer strips (Schirmer Tear Test; Ophthalmos).
During sample collection and processing, the temperature and humidity ranged from 20.9 to 27.1 °C and 42 to 62%, respectively. A teardrop of approximately 2 µL was deposited on a glass slide, at the center of a circle made previously and the tear ferning time (i.e., from tear deposition until drying) was measured using a digital timer. After complete drying, the slides were evaluated under a 10× lens magnification polarized light microscope with a camera (Microscope Scope A.1/AX10 Axion Cam ICc5; Zeiss, São Paulo, Brazil). The acquired images were classified and the formation of branches, angulations and zones of transitions were evaluated according to the proposed scales of Rolando et al. \[[@CR9]\] and Masmali et al. \[[@CR10]\].
The images of the ferning patterns were classified by three separate blinded examiners (APO, ACSR and AJL) from the Veterinary Ophthalmology Research Group of the School of Veterinary Medicine and Zootechny of Federal University of Bahia. All the examiners had experience of and expertise in the use of the scales. The final grading was assigned on the basis of agreement between the classifications of at least two of the three examiners. Tear ferning patterns were classified according to the Rolando grading scale (types I, II, III and IV) and according to the Masmali grading scale (grades 0, 1, 2, 3 and 4).
Statistical analysis {#Sec5}
--------------------
Statistical analysis was conducted using IBM SPSS Statistics for Windows, Version 22.0 (IBM Corp.), and the R Software for Windows, Version 3.6.1 with package irr was used for Cohen's kappa coefficient (k). The level of significance was set at 5% (P \< 0.05) and CI at 95%. The Shapiro--Wilk test was used to test data normality of the TFT values. Wilcoxon test was used for comparison of the same variables between eyes on each scale. The Mann--Whitney test was used to compare classification (both scales) with sex. Age was correlated with the classifications obtained through the Spearman test. Cohen's kappa coefficient was used to verify the agreement among the examiners in each scale.
Results {#Sec6}
=======
Crystallization occurred in the tear samples of all animals submitted to the test, with an average time of 14.6 ± 4.3 min. The crystallization patterns that received lower grades showed full crystallization with high density, without gaps between the ferns and branches, forming several nuclei that were easily distinguished. The branches showed medium length and were thin, with well-defined primary and secondary ramifications. As the grade increased, the nuclei lost definition, gaps were observed between the branches and, sometimes, coarse crystals were formed. Nevertheless, crystallization was observed even in the higher grades.
As described in methodology, after the ferning, the images were classified by resemblance to Rolando scale (Fig. [1](#Fig1){ref-type="fig"}a--c) and Masmali scale (Fig. [1](#Fig1){ref-type="fig"}d--f). The results obtained for each grading scale are expressed in Table [1](#Tab1){ref-type="table"}.Fig. 1Examples of tear ferning patterns in cats according to (**a**--**c**) the Rolando and **d**--**f**) the Masmali grading scales. **a** Type I representation: dendritic fern growth is uniform, nuclei (*yellow arrows*) are easily distinguished and no gaps are seen between the branches. **b** Type II: small spaces (*green arrows*) begin to appear between the branches and the ferns are thicker. **c** Type III: incomplete crystallization process; coarse crystals (*red arrows*) are formed in single and small size, and branches are rare. **d** Grade 0 representation: full crystallization without gaps between the ferns and branches, and nuclei well demarcated. **e** Grade 1: branch density is decreased and small spaces appear between them. **f** Grade 2: small branches---sometimes thick and large---with clear gaps between the ferns *(blue arrows*), and nuclei not visible. Note the similarity between panel (**a**) and (**d**); between (**b**) and (**e**); Furthermore panel (**b**) is closer to (**e**), whereas (**b**) and (**f**) appear more similar than (**c**) and (**f**), suggesting that Roland type I is similar to Masmali grade 0, and Rolando type II shares some common features with both Masmali grades 1 and 2Table 1Grading results for Rolando scale and Masmali scale for tear ferning in healthy catsScaleClassificationResultsRolandoI60 (50%)II56 (46.6%)III4 (3.4%)IV0120 (100%)Masmali018 (15%)168 (56.6%)234 (28.4%)3040120 (100%)
The obtained classifications did not show a normal distribution (P \< 0.001). The Wilcoxon test revealed no difference between the right and left eyes for both Rolando (P = 0.225) and Masmali (P = 0.683) scales. The median ± S-IQR value for the Rolando scale was 2 ± 0.5 and the CI was between 1.47 and 1.77. For the Masmali scale, the median ± S-IQR value was 1 ± 0.5 and the CI was between 1.0 and 1.62. No differences were found in relation to sex and there was no correlation between the scores obtained and the age of the study animals. The Kappa coefficient of agreement obtained for each pair of examiners for the Rolando and Masmali scales are presented in Table [2](#Tab2){ref-type="table"}. All concordances were significant (P \< 0.05), except for the comparison between examiner A and B for the Rolando scale. The examiners B and C had strong agreement in their classifications of both scales.Table 2Cohen's kappa agreement coefficient and P value among the examiners (A, B and C) for tear crystallization of cats according to Rolando and Masmali scalesExaminerExaminerRolando scaleMasmali scale*k*P-value*k*P-valueAB0.038^a^0.615^\*^0.141^b^0.038AC0.17^b^0.0180.199^b^0.004BC0.537^c^\< 0.0010.617^d^0^a^ poor correlation; ^b^ slight correlation; ^c^ moderate correlation; ^d^ strong correlation; ^\*^ P \> 0.05, therefore not significant
Discussion {#Sec7}
==========
Qualitative test of the TF is of great importance in animals that are affected by ocular surface disease and is complementary to quantitative assessment of tear production, as the latter does not guarantee a TF of good quality \[[@CR20]\]. In this study, we have shown that the TFT can be easily and safely performed in cats and has promising potential to be considered in the future as a valuable research tool in veterinary ophthalmology, as already reported by other authors \[[@CR11]--[@CR14]\]. Nevertheless, no other study to date has performed the TFT in cats.
STT strips were used to collect the tear samples for the TFT, because previous studies have shown that these strips minimize conjunctival lesions that may modify the crystallization pattern; moreover, studies recommend that the tear collection technique used in cats should be minimally invasive \[[@CR13], [@CR22]\]. Furthermore, this technique can be used without necessitating further manipulation of the eye. The STT strip is made of soft, absorbent and malleable paper, thereby minimizing reflex tearing or damage to the cornea and adjacent structures during tear collection. In contrast, tear collection with capillary microtube can lead to eye damage caused by voluntary and involuntary movements of cats, as they were not sedated in this study \[[@CR5], [@CR22], [@CR23]\]. This is especially important if more than one test is performed during the same examination, as the manipulation of cats' eyes can lead to exfoliation of conjunctival cells and/or release of TF components that are not normally expressed. Therefore, STT-1 tear sample collection was performed before any other clinical and ophthalmic tests \[[@CR24]\].
We found differences when comparing feline TF samples to the descriptions of TF made for other species. The differences observed included branching pattern density when compared to humans \[[@CR6], [@CR10], [@CR11]\], dogs \[[@CR13], [@CR14]\], capuchin monkeys \[[@CR15]\] and horses \[[@CR12]\]; crystals arrangement when confronting with descriptions of horses \[[@CR12]\], capuchin monkeys \[[@CR15]\] and dogs \[[@CR13], [@CR14]\]; and nuclei visibility when compared to all species previously mentioned \[[@CR6], [@CR10]--[@CR15]\] plus camels \[[@CR11]\]. All these pattern differences can be attributed to variations in tear composition \[[@CR10]\], sampling method \[[@CR25]\], temperature and humidity \[[@CR26]\].
Types I and II of the Rolando scale were observed in 96.6% of the evaluated animals, and grades 0 and 1 of the Masmali scale were observed in 71.6%; these findings were similar to those of previous studies on other animals and healthy humans \[[@CR10]--[@CR13], [@CR27]\]. However, the pattern with the second highest frequency the Masmali scale (28.4% in grade 2) is considered unhealthy in humans \[[@CR10]\]. Raposo et al. \[[@CR15]\] attributed the high frequency of Masmali grade 2 (72.7%) in lacrimal film of capuchin monkeys samples to species-specific characteristics, since all the animals were also healthy. Under these circumstances, the occurrence of grade 2 patterns in the cats evaluated in the present study is likely to be a standard of normality for cats, since the animals were clinically healthy, and no statistical difference was observed regarding gender or age.
Another explanation for the TF patterns observed in this study could be the osmolarity of feline tears, which is higher than that of human tears. Electrolyte concentration has already been reported as a cause of changes in crystallization patterns \[[@CR14], [@CR20], [@CR28]\]. Although tear osmolarity was not assessed in this study, it could be a possible explanation for the observed patterns, and hence, more research is warranted on this aspect.
The lack of strong agreement among examiners suggests that observation of crystal's morphology details is observer-dependent and so indicates that this assessment requires training and harmonization between different examiners. Therefore, the adoption of the two scales aimed to minimize the subjectivity of the test in cat samples. A similar observation has already been made in TFT studies for other species, such as dogs \[[@CR13]\] and capuchin monkeys \[[@CR15]\], and as happened in these studies we suggest that the adoption of a species-specific scale can be a way to enhance pattern classification. It is worth noting that the TFT is described as a complementary test to other methods \[[@CR9], [@CR10], [@CR20], [@CR25], [@CR26]\], and will certainly contribute to better understanding of the ocular surface in all species.
Aqueous tear deficiency ("Dry Eye Syndrome") is seldom reported in cats \[[@CR21]\], and none of the cats in this study showed signs of ocular disease, including corneal surface disease. Therefore, these findings suggest that Rolando types I and II and Masmali grades 0, 1 and 2 reflect normal TFT results in cats. While the results of most cats fell within types I and II of Rolando scale the grades 1 and 2 of Masmali scale, this can be attributed to the species-specific differences between human and feline tear film. Masmali grade 2 can be considered a normal tear pattern for the species, because cats studied were clinically healthy. For this reason, future complementary studies are necessary in comparing healthy eyes with different ocular surface disease in cats. Both scales can be feasible options for grading tear crystallization in cats, but as Rolando scale included 96.6% of the samples in the 2 types that are considered normal for humans, we think that this scale seemed to be more precise to classify crystallization pattern in cats. The crystallization patterns observed in this study can form the basis for standardizing the TFT of domestic cats.
Conclusions {#Sec8}
===========
The TFT is a feasible, inexpensive and low-risk test in cats. Our findings revealed high prevalence of the type I pattern according to the Rolando scale and of the grade 1 and 2 pattern according to the Masmali scale. These findings suggest that the TFT can be used as a complementary test for evaluating the ocular surface of cats and can provide insights into qualitative deficiencies of TF. We believe these findings will enhance our understanding of qualitative tear film disease and aid the design of prospective, case-controlled studies to better define the TFT results in various feline ocular diseases, such as conjunctivitis, corneal ulcer, SCCED, pigmentary keratitis, corneal sequestrum and dry eye syndrome.
**Publisher\'s Note**
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We would like to thank the State University of Santa Cruz (UESC) for offering development conditions for this research, such as availability of laboratories and equipment. Also, would like to thank Editage ([www.editage.com](http://www.editage.com)), Ford Hotchkiss Anderson Harvey and Tatiani Vitor Harvey for English language editing, and Anaiá da Paixão Sevá for the statistical consultancy.
All authors contributed in designing the study. JFV, CVBS and LFL sampled from animals, and JFV performed the tear ferning test. All authors contributed in interpretation of data. JFV was the major contributor in writing the manuscript with contribution from all authors. Further revisions of manuscript were done by RSAC, APO, ACSR and AJL. JFV submitted the manuscript. All authors read and approved the final manuscript.
This study was (co-) funded by the Faculty of Veterinary Medicine, State University of Santa Cruz, Foundation for Research Support of the State of Bahia, and by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior---Brasil---Finance Code 001.
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
This study was approved by the Ethics Committee on Animal Experimentation of the Use of Animals from the State University of Santa Cruz (protocol no. 003/17). All procedures were conducted in accordance with the Association for Research in Vision and Ophthalmology's (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research and NIH statement.
Not applicable.
The authors declare that they have no competing interests.
Data have not been published previously.
| {
"pile_set_name": "PubMed Central"
} |
Background {#Sec1}
==========
Genome-wide association studies (GWAS) have been widely used as a reliable method for identifying genetic variants associated with a trait or complex disease. A high density of SNPs increases the chance of finding either a causal mutation for the trait or SNPs close enough to the mutation to confidently suggest a gene or another sequence feature underlying the trait. One way to overcome this problem is using imputation, a process in which samples are genotyped using a low-density SNP array and imputed with information from a reference panel genotyped on a high-density SNP array. This method will also recover genotypes that are missing because of technical issues.
Imputation has successfully helped to identify genetic susceptibilities to various diseases and phenotypes that were not recognized in a genotyped panel \[[@CR1], [@CR2]\]. The method relies on the number of SNPs being shared between the two panels and the amount of linkage disequilibrium (LD) between genotyped and non-genotyped SNPs \[[@CR3]\]. A low average LD will reduce the accuracy and might require more typed SNPs. The quality of imputation also depends on the choice of reference \[[@CR1]\]. If the reference contains genetic variants not present in the actual sample population, it will increase the noise in the data and reduce the usefulness of the imputation. One study of malaria resistance in Gambian children only identified a previously known hemoglobin S variant in the hemoglobin-β gene when a Gambian-specific reference was used \[[@CR1]\]. Although this problem is more likely to occur in Africa, where there is a considerably lower LD compared to Europe and Asia \[[@CR4]\], determining how to choose the best reference is relevant for any study performing imputation with publicly available reference sets.
Many studies have validated the accuracy and reliability of imputation \[[@CR5]--[@CR7]\], but most of these studies focused on populations of European descent \[[@CR5], [@CR7]\]. One study showed that the accuracy of using a publicly available database varied across human populations with Europeans having the highest accuracy and Africans having the lowest \[[@CR6]\]. Because Asian populations have some unique genetic characteristics \[[@CR8]\], it is not always possible to directly adapt information about genetics or genomics from studies in Caucasian populations \[[@CR9]\].
Several types of software are currently available for performing genotype imputation \[[@CR10]--[@CR13]\]. Similarly, many publicly available genetics databases are accessible for public use \[[@CR14], [@CR15]\]. One of these is the Pan-Asian SNP genotyping database (PanSNPdb), which collects SNPs and copy number variations from 1719 samples in 71 populations from China, India, Indonesia, Japan, Malaysia, the Philippines, Singapore, South Korea, Taiwan, and Thailand \[[@CR16], [@CR17]\]. The genotyping process was performed using the Affymetrix GeneChip Human Mapping 50 K Xba Array.
Most of the studies on imputation have looked at the overall outcome of all SNPs \[[@CR5], [@CR18]\], and a few have focused on a particular region within a gene, not the whole genome \[[@CR19], [@CR20]\]. We proposed two objectives for the current study. The first was to identify the most preferred reference for imputation in Southeast Asian populations. Using two publicly available haplotype databases, the International HapMap Project (HMII) and the 1000 Genomes project (1000G), we compared the accuracy and yield of imputation in several Southeast Asian populations. Additionally, we looked at imputed results using genotyped samples from a study of a Thai genome cohort. The second objective was to evaluate the imputation results of different regions in the human genome using a real dataset from the Thai dengue study as a model. This is the first extensive study of imputation in Southeast Asian populations and the first illustration of imputation differences between SNPs in different regions of the genome.
Methods {#Sec2}
=======
This study was divided into two parts. The first part aimed at showing the difference in imputation accuracy by using different criteria for selecting a reference database. Additionally, using data from populations within the Southeast Asian region illustrated the variation in accuracy when going from one population to another. The second part used real genotype data in all autosomes to classify SNPs into different groups according to their location within genes. Imputation accuracy, GWAS significance and allele frequency were then correlated with the classification.
Sample datasets {#Sec3}
---------------
We performed the first part of our analysis using data from PanSNPdb \[[@CR16]\]. To illustrate the imputation accuracy in Southeast Asian populations, we selected all available samples from Indonesia (ID, *n* = 288), Malaysia (MY, *n* = 217), the Philippines (PI, *n* = 125), Singapore (SG, *n* = 90), and Thailand (TH, *n* = 245). Only SNPs that were polymorphic in all populations were used in this study (*n* = 52,160).
The second part was imputation accuracy and yield in a patient dataset in which we had access to phenotypes because the phenotypes allowed us to observe the effect of imputation on subsequent association tests. The subjects were 609 Thai dengue patients who were 1--15 years-old from Siriraj, Ramathibodi, and Khon Kaen hospitals. A total of 468,987 SNPs from Illumina Human Hap610 array (Illumina Inc., San Diego, CA) passed the quality control requirements (QC). The accuracy of imputation was tested for each SNP from the dengue dataset by first randomly choosing half of the SNPs from the genotyping panel to create a mutually exclusive set of SNPs: *Set 1* and *Set 2*. Then, *Set 1* SNPs were used to impute *Set 2* to create a complete SNP panel*. Set 2* was also used to impute *Set 1* to create a complete SNP panel. Based on our results in the previous section, HMII data were used as a reference for imputing the SNPs from the dengue dataset. The total number of SNPs after imputation was 1,417,081. Post-imputation QC reduced these numbers to 858,480.
Quality control and multidimensional scaling {#Sec4}
--------------------------------------------
For all of the sample datasets, QC was performed in PLINK v1.07 \[[@CR12]\] using standard procedures for GWAS \[[@CR21]\]. We included all markers with a call rate \> 0.95, a minor allele frequency (MAF) \> 0.01, and a Hardy--Weinberg equilibrium (HWE) \> 10^− 7^. Samples with call rates \< 0.95 were excluded from the analysis along with samples that had first-degree relationship agreement, as evaluated by expected IBD sharing in PLINK v1.07. Multidimensional scaling (MDS) of Southeast Asian populations from PanSNPdb was performed in PLINK v1.07. This method allowed for visualization of principle components in the admixed population. Plotting of the MDS was conducted in R version 3.0.2 (<http://www.r-project.org/>).
Imputation procedure {#Sec5}
--------------------
In the first part of the study, each population from PanSNPdb was analyzed independently. Five percent of SNPs were randomly selected and removed. The same SNPs set of the removed SNPs were applied to all populations. SHAPEIT version 2 software was used to pre-phase the SNPs \[[@CR22]\]. Imputation was accomplished with IMPUTE2 to recover the removed SNPs \[[@CR23]\]. Each population was phased and imputed using both references in turn. According to guidelines from IMPUTE2, we imputed each chromosome separately and used windows of 5 Mb with an additional 250 kb buffer region on both sides of the analysis interval. The options used in the program were -buffer 1000, −iter 30, −burnin 10, and -k 80. The processes for random removal, phasing, and imputation were repeated five times.
The second part of our study used all the autosomal SNPs from the dengue dataset. Half of the autosomal SNPs from Thai dengue patients were removed by every second SNP (*Set 1*). Another dataset (*Set 2*) was then created in which the other half of the SNPs (*Set 1*) were removed. In this way, all SNPs were imputed once. Imputation was performed with HMII as described above. After imputation, SNPs were filtered using a QC process similar to the initial filtering of raw genotypes. Post-imputation QC excluded SNPs with MAF \< 0.01, call rate \< 0.95, and HWE \< 5 × 10^− 7^. These datasets were used in the GWAS analyses. We then selected only the imputed SNPs from the two datasets and merged them into a single dataset in which all SNPs had been imputed. This dataset was used to compare imputation accuracies of SNPs according to their location relative to known genes.
References used for the imputation were downloaded prior to the imputation process from the Impute website (<http://mathgen.stats.ox.ac.uk/impute/impute.html>). The references were labeled on the website as International HapMap project phase II release \#22 (HMII) and 1000G phase I. A total of 1,417,081 SNPs from 90 Chinese and Japanese samples were used from HMII with an additional 39,343,900 SNPs from 1092 worldwide sample populations in the combined reference from 1000G. PanSNPdb shared 47,870 SNPs with the HapMap reference and 51,849 with the 1000G reference. The Thai dengue dataset shared 493,846 SNPs with the HapMap reference and 565,912 with the 1000G reference.
Imputation yield and accuracy {#Sec6}
-----------------------------
IMPUTE2 gives each imputed genotype a posterior probability score (info score) between zero and one. A higher threshold cut off for the probability score will usually result in higher accuracy but a lower yield. In this study, the posterior probability threshold was set to 0.9 to gain results with a high confidence of accuracy \[[@CR24]\]. Genotypes with posterior probabilities \< 0.9 were set to missing. Yield was reported as the percentage of non-missing genotypes within the removed SNPs and accuracy as the percentage of imputed, non-missing genotypes that matched the original genotypes.
SNP selection and annotation {#Sec7}
----------------------------
All SNPs from the dengue dataset were grouped based on their location in genes. Gene annotation was collected from Illumina sample sheets (Illumina Inc., San Diego, CA) and the NCBI database of genetic variation \[[@CR25]\]. Targeted gene regions included coding, intergenic, intronic, and untranslated regions (UTR). Some SNPs mapped to more than one location and were marked as being in a complex region. There were 53,277 SNPs that were not in any of these 5 groups and were subsequently discarded from further analysis.
The difference between imputed and genotyped data in the dengue dataset was evaluated by looking at several properties of the SNPs. Differences in accuracy and yield for each gene region were measured by varying the posterior probability threshold from 0.5 to 1.0. MAFs and *p*-values from chi-square tests for each SNP were compared between imputed and genotyped datasets. Coefficients of determination (r^2^) were calculated for each comparison to estimate the concordance of imputed and genotyped SNPs. Pairwise LD, which is measured as r-squared, was calculated within a region of \< 1 Mb around each SNP. The calculation was performed in PLINK v1.07 using the option \--r2 with \--ld-window-r2 0 and \--ld-window-kb 1000. Average R-squared values for each region were calculated and plotted in Microsoft Excel. R-squared values were also plotted against distance and averaging over 1 kb bins using R software (<http://www.r-project.org/>).
Results {#Sec8}
=======
Genotype imputation of Southeast Asian populations {#Sec9}
--------------------------------------------------
The accuracy and variability of imputation for Southeast Asian samples were assessed with five populations downloaded from PanSNPdb and two publicly available references from the HapMap project and the 1000 genomes project. All populations had an average imputation accuracy of more than 92% (Fig. [1](#Fig1){ref-type="fig"}), regardless of which reference was used. Imputation with HMII as a reference gave an average accuracy of 96.57%, while for 1000G, the accuracy was 93.98% (Fig. [1a](#Fig1){ref-type="fig"}). The Thai (TH) population had the highest accuracy for both reference panels followed by Indonesia (ID), whereas the population from the Philippines (PI) had the lowest accuracy. The yield for each population was lower when imputation was performed with the HMII reference (average = 59.03%) compared to the 1000G reference (average = 68.44%) (Fig. [1b](#Fig1){ref-type="fig"}). The yields for all populations were similar when using the same reference. The only exception was the TH population imputed with the 1000G reference, which had a higher yield compared to the other populations.Fig. 1Boxplot of accuracies and yields for imputation results across all populations. Five percent of randomly removed SNPs were imputed with IMPUTE2 using either the 1000 Genomes project phase I (1000G) or combined Chinese and Japanese haplotypes from the International HapMap project phase II (HMII) as a reference. The imputed SNPs were tested for accuracy with the previously removed SNPs. The same set of the removed SNPs was applied to all population dataset. The technique was repeated five times. **a** Boxplot of accuracy comparing populations and references. **b** Boxplot of yield comparing populations and references. Abbreviations: Indonesia (ID), Malaysia (MY), the Philippines (PI), Singapore (SG), and Thailand (TH)
Next, we looked at the results from each chromosome separately to demonstrate the variability of accuracies and yields (Fig. [2](#Fig2){ref-type="fig"}). Most results of imputation with HMII as a reference provided more than 95% accuracy (Fig. [2a](#Fig2){ref-type="fig"}). The 1000G reference provided a lower accuracy compared to HMII (Fig. [2b](#Fig2){ref-type="fig"}). The most striking result was that there was a change in standard deviation for imputation accuracy between the chromosomes. In particular, chromosomes 19 showed the lowest accuracy and higher variation in accuracy. Plotting the yield by population and chromosome showed no significant differences (Fig. [2c, d](#Fig2){ref-type="fig"}). However, chromosomes 19 also showed lower yield and chromosome 22 exhibited the highest level of variability. Because the imputation technique is, to a large extent, based on LD, we wanted to see if the higher variance could be correlated to any differences in the LD-pattern. We calculated the LD for each SNP pair with a distance between 10 kb and 1 MB. The number of these pairs that had an LD \> 0.2 was recorded for each chromosome (Additional file [1](#MOESM1){ref-type="media"}: Figure S1). Chromosome 19 and 22 had the lowest values.Fig. 2Imputation accuracy and yield by chromosome. The results derived from 5% randomly removed SNPs. Imputation with IMPUTE2 was accomplished to recover the removed SNPs. The imputed SNPs were tested for accuracy with previously removed SNPs. The same set of the removed SNPs was applied to all population dataset. This process was repeated five times. **a** Imputation accuracy by chromosome using HMII as a reference. **b** Imputation accuracy by chromosome using 1000G as a reference. **c** Imputation yield by chromosome using HMII as a reference. **d** Imputation yield by chromosome using 1000G as a reference. Abbreviations: Indonesia (ID), Malaysia (MY), the Philippines (PI), Singapore (SG), and Thailand (TH)
Trying to explain the differences in imputation accuracy, we investigated the population diversity of the five populations using MDS (Fig. [3](#Fig3){ref-type="fig"}). Whereas most samples were grouped together, all populations except TH showed large internal variation along the primary axis (C1) (Fig. [3a](#Fig3){ref-type="fig"}). The secondary axis (C2) mainly described the difference between 18 samples of the Thai Mlabri ethnic group (from Nan Province, Thailand) to the rest of the Thai samples. The third axis (C3) also mainly described the variation within Thai samples, whereas the fourth axis (C4) showed variation within PI, ID and SG (Fig. [3b](#Fig3){ref-type="fig"}).Fig. 3Multidimensional scaling plot of Southeast Asian populations from PanSNPdb. Genotype data of samples from Southeast Asian populations were downloaded from PanSNPdb. After quality control, multidimensional scaling (MDS) was performed in PLINK v1.07. **a** Plotting of the first (C1) and the second (C2) axes. **b** Plotting of the third (C3) and the fourth (C4) axes. Abbreviations: Indonesia (ID), Malaysia (MY), the Philippines (PI), Singapore (SG), Thailand (TH), China (CHB) and Japan (JPT)
Effects of SNP location in a gene on imputation {#Sec10}
-----------------------------------------------
Using the dengue GWAS dataset to evaluate the effects of SNP locations within a gene on the quality of imputation, we found 415,710 SNPs located within genes (Additional file [2](#MOESM2){ref-type="media"}: Table S1). As expected, intergenic regions contained the most SNPs, while complex regions, where SNPs have been associated with more than one gene, had the fewest. Varying the threshold settings in IMPUTE2 for accepting an imputed genotype showed that increasing the threshold led to an increase in accuracy but a decrease in yield (Additional file [1](#MOESM1){ref-type="media"}: Figure S2). Imputation results for SNPs in coding regions showed the highest yield and accuracy. Intronic and intergenic regions led to the second and third highest yield and accuracy, respectively. At a threshold of 0.5, all regions showed a similar yield. However, with an increasing threshold, the yield of coding regions increased compared to other locations. The opposite effect was observed for accuracy, and all locations approached the same level of accuracy when the threshold approached 1.
We further compared the measured MAF from the initial genotyped data to the MAF of the imputed data (Additional file [1](#MOESM1){ref-type="media"}: Figures S3-S4). Then, any imputed SNP that did not pass the quality-criteria (call rate \> 0.95, MAF \> 0.01, HWE \> 10^− 7^) were removed (Additional file [2](#MOESM2){ref-type="media"}: Table S1). The MAF of this reduced set of SNPs were similarly compared to the same set of SNPs from the initial genotyped data (Additional file [1](#MOESM1){ref-type="media"}: Figure S4). The correlation for each region between the MAF of the genotyped data and both the imputed data and the imputed and post-imputation filtered data were calculated (Table [1](#Tab1){ref-type="table"}). Coding regions had the fewest SNPs failing the QC, while complex regions had the lowest correlation, followed by untranslated regions (UTR), before filtering of the imputed data. After removing low quality SNPs, all regions showed a high correlation in MAF between imputed and actual genotypes. Looking at the direction of change in MAF showed that SNPs with low initial MAF (MAF \< 0.25) predominantly (\> 80%) had even lower MAFs after imputation. SNPs with an initial MAF close to 0.5 had an equal distribution of SNPs that obtained higher and lower post imputation MAF. Imputation also appeared to systematically reduce the allele frequency of the initially low MAF for most imputed SNPs (Additional file [1](#MOESM1){ref-type="media"}: Figure S5). The same analysis was repeated using *p-*values from the binary association test instead of the allele frequency for each SNP (Additional file [1](#MOESM1){ref-type="media"}: Figures S6 and S7). All regions had a generally low correlation for the *p-*values before removing low quality SNPs (*r* \< 0.4). After post-imputation QC, all regions showed an increase in correlation. Coding regions had the highest correlation both before (r^2^ = 0.39) and after (r^2^ = 0.81) removing low-quality SNPs. The lowest correlation for the binary association was found in complex regions before post-imputation QC (r^2^ = 0.27) and UTR regions after the QC (r^2^ = 0.76).Table 1Squared correlation of allele frequencies and chi-square *P-*values from SNPs in different regionsRegionSquared correlation (r^2^) of minor allele frequencySquared correlation (r^2^) *p*-value from chi-squareBefore post-imputation QCAfter post-imputation QCBefore post-imputation QCAfter post-imputation QCCoding region0.8680.9970.3870.813Complex region0.8170.9910.2670.782Intergenic region0.8640.9960.3280.789Intron region0.8630.9950.3400.784UTR region0.8300.9910.3030.756
To test whether coding regions had a stronger LD compared to the other regions, PLINK v1.07 was used to calculate pairwise LD between each SNP and any other SNP within 1 Mb. LD values, presented as r-squared, are shown as averages for each region (Additional file [1](#MOESM1){ref-type="media"}: Figure S8) and as LD vs. distance plot (Additional file [1](#MOESM1){ref-type="media"}: Figure S9). SNPs within coding regions had the highest average LD to neighboring SNPs (r^2^ = 0.253), followed by SNPs in intronic regions (r^2^ = 0.232).
Discussion {#Sec11}
==========
Imputation of genotyped datasets is a common practice when performing genome-wide association studies. This technique is used to fill in missing genotypes and to increase the density by adding information from SNPs that are not present in the original dataset \[[@CR5], [@CR26]\]. Imputation of SNPs that are not available in the dataset serves several purposes. First, if available SNP arrays are designed based on a specific population, such as Europeans, the SNPs may not cover the areas of interest for another population. SNPs important to populations from Southeast Asia might therefore be underrepresented or missing from these arrays. This situation has been reported for populations from Africa \[[@CR27]\] and Mexico \[[@CR28]\]. Second, if data are collected independently between groups of case and control populations, the datasets might have been genotyped on different SNP sets. Third, GWAS usually requires a high number of SNPs to increase chance to detect association signals. Although current genotyping arrays could contain more than a million markers, imputation still adds more SNPs for denser full genome coverage.
The choice of reference panels can affect the accuracy of imputation through the genetic variation of the samples and the genetic relationship between the samples in the reference panel and the imputed references \[[@CR6], [@CR27]\]. We studied these effects in five populations from Southeast Asia. The 1000G reference provided the highest yield, while the HMII reference had the highest accuracy. These results were the same for all Southeast Asian populations. IMPUTE2 software provides the posterior probability score for each imputed genotype. There was increasing accuracy with a decreasing yield when the probability threshold increased (Additional file [1](#MOESM1){ref-type="media"}: Figure S2). At the threshold 1.0, the result showed a large drop in yield but only a limited increase in accuracy. At the threshold 0.9, the slope of the yield and accuracy were significantly changed. This threshold might be a good starting point for using IMPUTE2 for Southeast Asian populations.
Genotype imputation of the Thai population had the highest accuracy in the current study. Previous work has shown that in the PanSNPdb database, the Thai population had the highest relationship to the Chinese and Japanese populations out of the other four study populations \[[@CR16]\]. The increased accuracy can therefore be explained by this closer relationship. To determine if population diversity influenced the average imputation accuracy, classical MDS was used to display the variation within Southeast Asian populations. The main variation in the MDS plot (C1) was related to the sub-populations in 4 of the 5 main populations, and the only exception was the Thai population (Fig. [3](#Fig3){ref-type="fig"}). The Thai samples form a more homogenous group compared to other populations, and this outcome can help explain why they had the most accurate results. Eighteen individuals from the Thai Mlabri group, which is a hunter-gatherer group in Northern Thailand, clustered away from other Thai samples, which is consistent with the findings of previous studies \[[@CR17], [@CR29]\]. Nevertheless, further investigation is necessary to understand the effects of population stratification on imputation results.
Plotting the imputation accuracy and yield for each chromosome revealed that the variation within populations was the highest for the smaller chromosomes, especially chromosomes 19 and 22. The same results were also observed for yield. In particular, the population from the Philippines showed a large increase in variability for these chromosomes when using the 1000G reference. This result might indicate a specific issue with the SNP selection for these chromosomes, such as the LD structure being different or the relationship to the reference being lower in these regions. One possible reason for the increased variability was found by looking at the long distance LD in each chromosome. Chromosome 19 and 22 was shown to have the fewest SNPs connected by LD above 0.2 when comparing inter-SNP distances above 10 kb (Additional file [1](#MOESM1){ref-type="media"}: Figure S1). The accuracy of resulting imputation for these two chromosomes will therefore be more dependent on which SNPs are removed and will show more variance between replications with randomly selected SNPs. The results also clearly show that only taking the average numbers for accuracy or yield into account will result in overlooking potentially important information.
Coding sequences have been the favored area to search for functional mutations because these sequences are more informative and easier to interpret due to the direct link to a protein and the possibility of functional changes \[[@CR30]--[@CR32]\]. This approach made it useful to compare the imputation results between coding and other SNP regions. Our results show that a higher percentage of SNPs in coding regions passed the post-imputation QC than in other regions. SNPs in coding regions also had the highest accuracy. This outcome correlates well with the results showing coding regions to have the highest LD with the surrounding SNPs (Additional file [1](#MOESM1){ref-type="media"}: Figure S8). This is an important factor to consider when discussing significant results because imputed SNPs in coding regions will have a higher accuracy compared to SNPs in less conserved areas. This trend does not mean that we can omit non-coding SNPs because they have been shown to be associated with phenotypes in more than one-third of GWAS \[[@CR33]--[@CR35]\].
Our study also compared GWAS results from a dataset from Thai dengue fever patients to see how imputation affected the reported results. Imputation appeared to systematically reduce the allele frequency of the initially minor allele for most imputed SNPs (Additional file [1](#MOESM1){ref-type="media"}: Figure S5). This effect was more pronounced for SNPs with an initial low MAF and will make imputing low frequency alleles difficult. Our results demonstrated that imputation tended to increase the common allele. This outcome was especially problematic if the genotypes are very rare variants \[[@CR26]\]. Even if the most significant SNPs had higher significance in the imputed dataset, this trend was not seen when comparing *p*-values from GWAS before and after imputation.
The imputed genotypes were also subjected to QC-filtering, which is similar to the QC being performed on the raw genotype data. It was previously shown that post-imputation QC did not influence the imputation outcome \[[@CR36]\]. However, we observed an improvement in the correlation between measured MAF and *p*-values from imputed data and genotyped data. Without the SNPs failing the QC, the correlation between imputed data and genotyped data was close to one for allele frequency and had improved *p*-values. Even if over half the SNPs were removed in this step, the remaining data had higher quality and were more trustworthy. This difference in post-QC improvement could be due to the initial imputation accuracy. Post-imputation QC might be more important if the initial imputation results are less accurate. This result is also supported by a previous experiment that similarly demonstrated Hardy--Weinberg disequilibrium is a crucial step for post-imputation filtering \[[@CR37]\].
This study demonstrates that the expected accuracy and yield of imputation in various Southeast Asian populations varies between populations. Our reference comparison of HMII and 1000G in imputation in Thai GWAS showed that using a larger reference provided a higher yield but caused a reduction in accuracy compared to a smaller but more related reference. We also extensively showed the imputation results with respect to SNP localization near genes using Thai genome-wide genotypes as a model. This study provides crucial information for investigators undertaking imputation, especially in Southeast Asian populations.
Conclusions {#Sec12}
===========
This work provides the first evidence of imputation reference selection for Southeast Asian studies and highlights the effects of SNP locations respective to genes on imputation outcome. Researchers will need to consider the trade-off between accuracy and yield in future imputation studies.
Additional files
================
{#Sec13}
Additional file 1: Figure S1.Long distance LD pr. Mb separated by the chromosome for each population LD for each SNP pair with a distance between 10 kb and 1 Mb taken into account. **Figure S2.** Accuracy and yield of imputation between each location of SNPs. **Figure S3.** Comparing minor allele frequencies of SNPs between imputed and actual genotypes before quality control of imputed results separately plot by each SNP location. **Figure S4.** Comparing allele frequencies of SNPs between imputed and actual genotypes after quality control of imputed results separately plotted by each SNP location. **Figure S5.** Proportion of SNPs with lower AF after imputation vs initial AF. **Figure S6.** Comparing *p*-values between cases and controls of SNPs between imputed and actual genotypes before quality control of imputed results separately plotted by each SNP location. **Figure S7.** Comparing *p*-values between cases and controls of SNPs between imputed and actual genotypes after quality control of imputed results separately plotted according to each SNP location. **Figure S8.** Average linkage disequilibrium as r-squared for each region. SNPs were assigned to gene locations. **Figure S9.** LD, measured as r^2^, plotted against physical distance. (PDF 1286 kb) Additional file 2: Table S1.Summary of SNPs used in study of SNP annotation to imputation outcome. (XLS 26 kb)
1000G
: the 1000 Genomes project
HMII
: the International HapMap Project
ID
: Indonesia
LD
: linkage disequilibrium
MY
: Malaysia
PanSNPdb
: Pan-Asian SNP genotyping database
PI
: The Philippines
QC
: Quality control
SG
: Singapore
TH
: Thailand
**Electronic supplementary material**
The online version of this article (10.1186/s12881-018-0534-8) contains supplementary material, which is available to authorized users.
We thank all team members of the Pan-Asian SNP Genotyping Database especially Dr. Sissades Tongsima and Dr. Chumpol Ngamphiw for collecting and curating PanSNPdb data.
Funding {#FPar1}
=======
This research was supported by National Research Council of Thailand (NRCT) (grant number 92/2550). Additionally, the research was partial funded by the Office of the Higher Education Commission and Mahidol University under the National Research Universities Initiative. W. Lert-itthiporn was supported by a scholarship from the Medical Scholars Program, Mahidol University. P. Malasit was supported by NSTDA Chair Professor Grant. P. Suriyaphol was supported by a National Research University (NRU) Grant through Mahidol University and TRF-office for R&D.
Availability of data and materials {#FPar2}
==================================
The data supporting the conclusions of this article are included within the article and its additional files. The datasets from Thai dengue patients generated and/or analysed during the current study are not publicly available due to ethical restrictions enforced by the Siriraj Institutional Review Board, but are available from the corresponding author on reasonable request.
WL participated in the imputation experiment and analyzed the association study as well as the SNP regional effects on imputation and contributed to the article. HG analyzed pre- and post-imputation results and wrote the article. FM and AS designed the study and participated in genotyping the DNA samples. PM and NT conducted the dengue patient's recruitment and DNA. BS and PS participated in the study design and wrote the article. All authors read and approved the final manuscript.
Ethics approval and consent to participate {#FPar3}
==========================================
The ethics of this study were approved by the Siriraj Institutional Review Board Certificate of Approval. This program is completed to certify that the Siriraj Institutional Review Board is in full compliance with international guidelines for human research protection such as the Declaration of Helsinki, the Belmont Report, CIOMS Guidelines and the International Conference on Harmonization in Good Clinical Practice (ICH-GCP). Written informed consent was obtained from both the participants and legal guardians.
Consent for publication {#FPar4}
=======================
Not applicable.
Competing interests {#FPar5}
===================
The authors declare that they have no competing interests.
Publisher's Note {#FPar6}
================
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
| {
"pile_set_name": "PubMed Central"
} |
1. INTRODUCTION {#sec1-1}
===============
The World Health Organization (WHO) has defined *preterm birth* as delivery before 37 completed weeks of gestation (gestational age is reported in terms of completed weeks (i.e., one never rounds gestational age up, so 36 weeks and 6 days of gestation is 36 weeks and not 37 weeks of gestation)) ([@ref1]). All newborns are subject to decline in hemoglobin levels in the first weeks after birth. In neonatology, this condition is often called physiological anemia of the newborn. In healthy term infants, clinical signs or symptoms of anemia are absent; this normal decline in Hb is referred to as "physiologic" or "early anemia of infancy" ([@ref2]). It was noted that the lowest value of hemoglobin in term infants rarely fall below 100 g/L in age from 10 to 12 weeks ([@ref3]). In contrast, anemia in preterm infants (anemia of prematurity) is the pathophysiological process with larger and faster drop in hemoglobin. Consequently, there is a need for blood transfusion and application of human recombinant erythropoietin. Generally, it is considered that at the age of 4-6 weeks, premature infants of birth weight between 1000 and 1500 grams have hemoglobin value about 80 g/L, while premature infants of birth weight less then 1000 grams have hemoglobin value about 70 g/L ([@ref3]). Preterm infants are faced with the appearance of anemia of prematurity due to several main reasons ([@ref4]). The first is the shortened gestational age - when it is shorter the clinical presentation is more severe. Another reason is the underdevelopment of the hematopoietic system in preterm infants ([@ref2]). The third, very important cause of anemia are repeated acts of taking blood for laboratory analysis.
2. AIM {#sec1-2}
======
The aim of this study was to determine the frequency of clinical manifestations of anemia in premature infants at the Pediatric Clinic, University Clinical Center Sarajevo, accompanied by a drastic drop in hemoglobin and hematocrit in the blood count and a need for treatment in high-risk groups of premature infants of gestational age below 32 weeks, compared to a group of premature infants over 32 weeks and the linkage of perinatal and neonatal risk factors with the development of anemia of prematurity as well.
3. PATIENTS AND METHODS {#sec1-3}
=======================
Research has been set as a retrospective analysis of the characteristics of anemia of prematurity in the period of the first six months of year 2014. The study included 100 patients, gestational age \< 37 weeks (premature infants), who were admitted to, after giving birth, at the Department of Neonatal Intensive Care at the Pediatric Clinic, of Clinical Center of Sarajevo University (UCCS). Inclusion criteria in the study were: the patient is a premature infant gestational age \< 37 weeks and the patient's first admission at the Department of Neonatal Intensive Care. Exclusion criteria were no data on: gestational age of the patient, the blood count at the admission, the number of received transfusions and administration of iron therapy. Analytical and descriptive methods were used. Data were collected by examining the medical records of patients at the Pediatric Clinic, UCCS. Statistical analysis of the collected data was performed through SPSS ver. 21.0. Data obtained after statistical analysis were presented in tables and figures, using Microsoft Office Excel 2007. Statistical analysis was performed by it-test. The level of significance of P \< 0.05 was considered statistically significant.
4. RESULTS {#sec1-4}
==========
There were two groups of patients, the first, children of gestational age ≤ 32 weeks, and the second, children of gestational age 33--37 weeks. In the first group there were 62/100 patients (62 %), while in the second 38/100 (38 %) patients . The larger number of patients in each group were males ([Table 1](#T1){ref-type="table"}). Data analysis demonstrated a statistically significant difference (P \<0.01) between the two groups of patients, in terms of body weight at birth ([Table 2](#T2){ref-type="table"}).
######
The gender structure of patients by the groups
######
The birth weight of patients
There was a statistically significant difference between the infants of gestational age ≤ 32 weeks and gestational age 33-37 weeks, in terms of APGAR score in the first minute (P \<0.001), and after 5 minutes (P \<0.004) ([Table 3](#T3){ref-type="table"}).
######
APGAR score in the first and fifth minute
In the first group, 16.13 % of children were reanimated, and in the second 5.26 %. The statistical analysis did not show any significant difference in the number of resuscitations of children between groups. In terms of deaths, 27.42 % of children in the first group had a lethal outcome, and 10.53 % of children in the other. It is proved a statistically significant difference in mortality between groups (P \<0.04).
######
The number of resuscitation and deaths
Statistical analysis demonstrated a statistically significant difference in the number of erythrocytes between groups of patients gestational age ≤ 32 weeks, and those of gestational age 33-37 weeks ([Table 5](#T5){ref-type="table"}.). However, there wasn't a significant difference in the level of hemoglobin and hematocrit between the two groups ([Table 6](#T6){ref-type="table"}-[T7](#T7){ref-type="table"}).
######
The number of erythrocytes in blood count at admission
######
The value of hemoglobin in blood count at admission
######
The value of hematocrit in blood count at admission
Out of 62 children, 18 children gestational age ≤ 32 weeks received a blood transfusion, while only 3/38 of children gestational age 33--37 weeks needed transfusion ([Table 8](#T8){ref-type="table"}). Statistical analysis showed a significant difference (P \<0.01) between mentioned groups in terms of the need for transfusion.
######
The number of patients with applied blood transfusion
It is observed that the bleeding occurred in patients of gestational age ≤ 32 weeks, and in other patients it did not occur. However, statistical analysis showed no significant difference between groups. [Table 9](#T9){ref-type="table"} provides information about the time of administration blood transfusions to patients which had an intracranial hemorrhage. 50% of these patients received blood transfusion in the first week of life, which is statistically significant (P = 0.000) in terms of association between intracranial hemorrhage and transfusion applications in the first week. However, this pattern is not sufficiently representative for the adoption of concrete conclusions, because of small number of such patients.
######
The connection of bleeding with the application of blood transfusion in the first week
In the group of patients ≤ 32 weeks of gestational age, iron is introduced into therapy on average 34^th^ day of life, while in second group of patients, it began on average on 32^th^ day ([Table 10](#T10){ref-type="table"}). There was no statistically significant difference between the two groups in term of iron therapy.
######
The beginning of iron therapy
There was a statistically significant difference (P \<0.005) in the length of treatment in patients of gestational age ≤ 32 weeks and those of gestational age 33--37 weeks ([Table 11](#T11){ref-type="table"}).
######
The length of treatment
5. DISCUSSION {#sec1-5}
=============
Anemia of prematurity is the focus of clinical trials in the last decade, with still uncoordinated approach, moving from observational approach and occasional controls of hematological status, limited use of erythropoietin, to liberal or today more often, restrictive practice of transfusion of packed RBCs. Based on research conducted in recent years, protocols are established, which include in evaluation the level of hemoglobin, the level of respiratory disease and the traditional signs and symptoms of pathological anemia. In this study, the premature infants are divided into two groups based on gestational age. Premature infants of gestational age ≤ 32 weeks (62 % of patients) were in the first group, while premature infants of gestational age between 33-37 weeks (38 % of patients) were in the second group. The division is modeled on other studies, which state that the anemia of prematurity is significantly more pronounced in infants gestational age ≤ 32 weeks, compared to the rest of preterm infants. Cassady G. and Rosenkrantz T. ([@ref5]) state that a half of prematurely born children, who belong to this group, develop anemia, as it typically does not occur in children of gestational age of 33-37 weeks. In a similar way, a study by Wardrop and associates was done ([@ref6]), with the difference that, they are the second group divided into two, where one did infants gestational age 33-35 weeks, and the other infants gestational age 35-37 weeks.
The birth weight was measured for all premature infants and there was a significant difference among two groups, as it was expected. In the first group average birth weight was 1.394 g, while in the second group, it was 2.242 g ([Table 2](#T2){ref-type="table"}). This parameter was taken into account as one of the risk factors for anemia of prematurity, because there is a connection between low and very low birth weight and early occurrence of serious anemia ([@ref2], [@ref3], [@ref7]).
As regards APGAR score, values in the first and fifth minute were taken into account and significant statistical difference has been noticed. In the first group, average APGAR score in the first minute was 5.56 while in the fifth minute it was 6.5. In the second group, there were less variations, so the average score in the first minute was 7, while in the fifth minute it was 7.67 ([Table 3](#T3){ref-type="table"}).
Among 100 patients, 21 % died. In the first group of patients there were 27.42 % deaths, while in the second group number of deaths was significantly lower. Namely, only 10.53 % of premature infants with gestational age between 33-37 weeks died. Statistical analysis has shown significant difference in terms of mortality among these groups ([Table 4](#T4){ref-type="table"}.). These results correlate with the results of a study by Banarjee J. and associates ([@ref7]), which show a significantly higher infant mortality gestational age ≤ 32 weeks, compared to the rest of preterm infants.
In addition, patients in the first group had, on average, a longer stay at the Pediatric Clinic in relation to another group. The average length of treatment, patients from the first group, was 27 days. Patients in the second group, on average, spent 14 days on treatment ([Table 11](#T11){ref-type="table"}). Hintz S. et al. ([@ref8]) also indicate a significant difference in the length of treatment between infants small for gestational age and very low birth weight, compared to the rest of preterm infants. We found a statistically significant difference in length of treatment between the two groups (P \<0.005).
Focus of the research was hematology status and blood transfusion needs. At the admission, the first and second group had approximately same values of blood count which has been changed significantly during hospitalization, meanwhile blood transfusion needs arise. So the average number of erythrocytes in the first group was 4,39x10^12^, hemoglobin value was 155 g/L and average hematocrit was 0,49 L/L. In the second group results were similar: average number of erythrocytes in the second group was 4,52x10^12^, hemoglobin value was 167 g/L and average hematocrit was 0,51 L/L. Statistical analysis has proven significant difference in number of erythrocytes among patients of gestational ≤ 32 weeks and patients of gestational age between 33-37 weeks (P\<0,003). However, there was no significant difference in hemoglobin and hematocrit level among groups of patients ([Table 5](#T5){ref-type="table"}, 6 and 7). Banerjee J. et al. ([@ref7]) did a study association of hemoglobin after birth with short-term outcome of treatment prematurely born infants. They proved its connection with the subsequent need for transfusions and death.
Among 100 patients, 21 % required blood transfusion during the treatment. In the group of premature infants of gestational ≤ 32 weeks, 29.03 % received blood transfusion. In the second group, 7.9 % of patients required blood transfusion during hospitalization at Clinic ([Table 8](#T8){ref-type="table"}). Statistical analysis showed significant difference in blood transfusion needs among premature infants of gestational age ≤ 32 weeks and premature infants of gestational age between 33-37 weeks. Strauss R.G. ([@ref9]) states that transfusions annually receive even 300.000 premature infants, but mostly, it is about infants with very low and extremely low birth weight.
In the group of premature infants of gestational age \< 32 weeks even 29.3 % received transfusion of erythrocytes, while in the group of premature infants of gestational age between 33-37 weeks, only three of them received transfusion of erythrocytes. Statistical analysis has proven significant difference (P\<0,01) among these groups in terms of need for blood transfusion.
Zuppa and coworkers ([@ref10]) worked on study about connection of gestational age and blood transfusion needs. They have proven that premature infants of gestational age ≤ 32 weeks are the most probably candidates for blood transfusion and that is in accordance to our findings.
Interesting data was arisen during the research of connection between occurrence of intracranial bleeding and blood transfusion application. Namely, among 18 premature infants of gestational age ≤ 32 weeks, in 10 of them intraventricular hemoragy were registered. In contrast, among 3 premature infants of gestational age between 33-37 weeks, there was no such bleeding. However, because of small number of cases, statistical analysis did not show significant difference among groups.
Due to intracranial bleeding associated with the application of transfusions in the first week of life, it is made the examination of these data in the first group of patients. Namely, Christensen R. D. ([@ref11]) associated the occurrence of intracranial bleeding with early application of transfusions, indicating that the subsequent application is associated with the occurrence of necrotizing enterocolitis.
Goldber and coworkers ([@ref12]) show the possibility of bleeding caused by rapid growth of intravascular volume in premature infants of gestational age ≤ 32 weeks. And other studies confirms connection between intraventricular hemoragy and blood transfusion of deplasmated erythrocytes in premature infants ([@ref13], [@ref14], [@ref15]).
Among 10 patients with intracranial bleeding in the group of premature infants of gestational age \< 32 weeks, 50 % of them received blood transfusion in first week of life and other 50 % received blood transfusion later during hospitalization. Statistically there was significant connection (P=0.000) of occurrence of intracranial bleeding and blood transfusion application in first week of life. Since intraventricular hemoragy has multifactor etiology which includes fluctuations of cerebral flow and increase of cerebral venous pressure and taking into account that small number of patients were in place, this sample is not enough representative for making concrete conclusions, although surely indicates the connection noticed by other authors as well as need for further research of this connection in more relevant sample.
Finally, there were collected the data about the time of beginning the iron therapy, during the stay of patients at the Clinic. In the first group of patients, iron therapy was administered on average 34^th^ day of life, and in the second group on average on 32^th^ day. There was no statistically significant difference between the two groups in term of iron therapy ([Table 10](#T10){ref-type="table"}).
Meyers P. et al. ([@ref16]) reported that the timely beginning of iron therapy is necessary to stimulate erythropoiesis in premature infants, since the iron depots are established in the third trimester of pregnancy, in order to avoid anemia due to iron deficiency, since it is known that anemia of prematurity does not correspond to the iron therapy.
6. CONCLUSIONS {#sec1-6}
==============
Frequency of anemia with drastic fall of hemoglobin and hematocrit and need for blood transfusion in premature infants with gestational age \< 32 weeks, is 29.3 %, meaning that almost every third premature infant of gestational age \< 32 weeks develops anemia requiring treatment.
There is statistically significant difference in frequency of anemia in premature infants of gestational age \< 32 weeks compared to premature infants of gestational age \> 32 weeks, meaning that level of anemia negatively correlate with gestational age.
There is a significant difference in birth weight and APGAR score between premature infants of gestational age ≤ 32 weeks and premature infants of gestational age between 33-37 weeks, which indicates worse condition of these children at birth and more aggressive approach during laboratory analysis, which makes influence to earlier development of anemia.
There is no significant difference in the blood count of preterm infants at admission in relation to gestational age.
Premature infants of gestational age ≤ 32 weeks due to frequent development of manifested anemia during hospitalization, more often requires blood transfusion during treatment what is statistically significant.
Premature infants of gestational age ≤ 32 weeks have risk of occurrence of posttransfusion complications, while the most serious one among them is intracranial hemoragy connected to application of blood transfusion in first week of life.
• Conflict of interest: none declared.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-genes-10-00573}
===============
In monocotyledonous cereal crops, such as rice (*Oryza sativa* L.), wheat (*Triticum aestivum* L.), barley (*Hordeum vulgare* L.), and rye (*Secale cereale* L.), as well as several forage grasses, the inflorescence is characterized by bristle-like extensions, called awns, adhered at the tip end of the lemmas in the florets \[[@B1-genes-10-00573]\]. Awn formation and development rely on lemma primordium, which produces awn meristem as the lemma apex continues to elongate \[[@B2-genes-10-00573]\]. The sequence of awn development events is a complex process, involving morphological features, such as inflorescence density and glume length to control awn length \[[@B3-genes-10-00573]\], growing conditions \[[@B4-genes-10-00573]\], and genetic mechanisms \[[@B5-genes-10-00573]\]. Awns are beneficial in wild species because they aid in seed dispersal, as the barbed awns adhere the seed to animal fur \[[@B6-genes-10-00573]\]. They repel seed-eating birds and animals. Long awns are also easily controlled by external dispersal factors such as wind, and their movement may propel the seed into the ground, which enables self-planting \[[@B7-genes-10-00573]\]. Awns are also effective for light interception and CO~2~ uptake because photosynthates travel easily from awns to the growing kernels \[[@B8-genes-10-00573]\]. Evidence in barley \[[@B1-genes-10-00573],[@B5-genes-10-00573]\] and in wheat \[[@B8-genes-10-00573],[@B9-genes-10-00573]\] revealed that awns are coupled with grain yield increase, particularly in drier and warmer environments. Despite the undisputable importance of awns especially in nature, long and barbed awns deter manual harvesting and interfere with seed-processing activities, that is, malting and milling \[[@B10-genes-10-00573]\]. In forage barley cultivars, long and barbed awns increased bulky fibers, hindering forage quality and palatability \[[@B11-genes-10-00573]\].
In transverse sections, the awns of barley, wheat, oat, and rye are triangular-shaped with three vascular bundles and two strands of chlorenchyma tissues, and they may contribute to photosynthesis. Awns of rice, on the contrary, have a round shape in transverse section with single vascular bundles and a lack of chlorenchyma tissues, implying their minor contribution to photosynthesis \[[@B1-genes-10-00573]\]. Supporting this observation, awn removal experiments revealed a minor impact on grain yield and, consequently, rice species with shorter or nonexistent awns were preferred by early cultivators to enhance harvesting and postharvest processing. This is in contrast with some other domesticated cereals, such as barley and wheat, where the removal of awns led to yield loss. Thus, most cultivated barleys and wheats possess long awns \[[@B12-genes-10-00573]\].
Previous genetic studies identified several awn-related quantitative trait loci (QTLs) in several major cereals, and several genes were recently cloned in some. In this respect, the genetic basis of awn development was deeply investigated in rice, and allowed researchers to identify many QTLs associated with awn growth on almost all chromosomes \[[@B13-genes-10-00573],[@B14-genes-10-00573]\]; furthermore, some of their causal genes were dissected molecularly \[[@B12-genes-10-00573],[@B13-genes-10-00573],[@B14-genes-10-00573],[@B15-genes-10-00573],[@B16-genes-10-00573],[@B17-genes-10-00573]\]. In barley, functional advancements regarding genetic mechanisms of awn development emerged recently, with the cloning and characterization of genes such as *Lks1* \[[@B18-genes-10-00573]\] and *Lks2* \[[@B1-genes-10-00573]\]. Although genes causing awn development in wheat are not yet cloned, scientists identified the main inhibitors of awn development in some wheat cultivars \[[@B19-genes-10-00573]\].
Despite the undisputable role of awns for inflorescence architecture and their putative impact for yield determination, details regarding how the awn is formed and how its morphology is variable among species remain elusive, and a clear model explaining the awn developmental process is needed. As such, it demands solid and basic skills of the major plant developmental processes. Our aims in this review are (1) to discuss the phenotypic evolution of awns by illustrating morphological structures of the awn and physiological changes during awn development, and (2) to assess the genetic basis associated with awn development and variation in major cereal crops. Understanding awn development is essential for agriculture, with respect to improving yield and quality traits of grass species.
2. Morphological and Histological Characteristics of the Awn {#sec2-genes-10-00573}
============================================================
2.1. Morphological Features of the Grass Inflorescence {#sec2dot1-genes-10-00573}
------------------------------------------------------
The grass inflorescence is a unique, complex structure, different from those of most other plants. It comprises one to several structural units called florets, packed together in a spikelet \[[@B20-genes-10-00573]\]. The spikelet is the basic unit of the grass floral structure, usually comprising glumes (either empty glumes or rudimentary) and one to several florets. The floret, the spikelet's individual unit, consists of a lemma (which extends in the functional awn in some grasses), a palea, and reproductive organs, including the stamens, pistils, and lodicules \[[@B21-genes-10-00573]\]. The arrangement of spikelets and the number of florets are key determinants of inflorescence type in grass species. Based on the arrangement of spikelets, the three most well-known kinds of grass inflorescence are spike, raceme, and panicle. A spike is an unbranched inflorescence in which spikelets are attached directly to the main axis, as in wheat, barley, and rye. A raceme is an unbranched inflorescence in which spikelets are attached to the main stem by pedicels, as in *Festuca*, some *Danthonia*, and *Avena*. A panicle is a branched inflorescence with spikelets attached to the side branches rather than the main stem with or without pedicels, such as what is found in rice and oats \[[@B22-genes-10-00573]\]. The inflorescence meristem may be classified based on floral meristem determinacy. The inflorescence is determinate, when the main axis always ends in an apical spikelet, as in wheat, rye, and oats. The inflorescence is indeterminate, when the main axis never becomes terminated by a spikelet and continues to initiate branches and a number of fertile spikelets, as in barley and maize \[[@B2-genes-10-00573],[@B23-genes-10-00573]\].
Spikelets are modified units defining the inflorescence structure of grass species. Based on the presence or absence of pedicel, a small stalk or stem connecting the flower to the main axis, spikelets may be sessile or pedicelled. Sessile spikelets are connected directly to the axis, whereas pedicelled spikelets are attached to the axis by the stalk \[[@B22-genes-10-00573]\]. Scientists also classified spikelets referring to the position of the spikelets in the main inflorescence. Basal spikelets are located in the bottom, central spikelets in the center, and apical spikelets on the top of the inflorescence \[[@B23-genes-10-00573]\]. Variation of the grass floral morphology contributes greatly to the diversity of grass spikelets. For instance, the presence of fertile florets (those possessing a caryopsis) and sterile florets (empty glumes) raised arguments in the classification of spikelets. A fertile spikelet contains one to several fertile florets, whereas a sterile spikelet contains only sterile florets \[[@B2-genes-10-00573],[@B23-genes-10-00573]\]. Furthermore, the grass inflorescence is usually composed of both male and female structures. In some cases, the spikelet may possess imperfect flowers with only male or female structures. For instance, the staminate spikelet only contains stamens, whereas the pistillate spikelet only possesses pistils \[[@B24-genes-10-00573]\]. Spikelets can also be classified based on the presence or absence of awns. The awned spikelet possesses at least one but sometimes several awns, while the awnless spikelet possesses no awn in any form \[[@B25-genes-10-00573],[@B26-genes-10-00573],[@B27-genes-10-00573]\].
2.2. Awn Anatomy and Histology {#sec2dot2-genes-10-00573}
------------------------------
Awns are characteristic features of grass inflorescence \[[@B7-genes-10-00573]\]. While awns' morphological features vary among grass species, their description varies among sources. Most cultivated cereals (for example, wheat, barley, oat, and rye) are awned and possess a single awn per floret, but some cereal grasses, such as maize and sorghum, are awnless \[[@B19-genes-10-00573],[@B25-genes-10-00573]\]. In rice, most wild rice species are known as long-awned, whereas cultivated cultivars are awnless or remarkably short-awned \[[@B6-genes-10-00573]\]. In addition, some wild grasses may have two to several awns per floret, as in grasses from the *Aristida* genus \[[@B28-genes-10-00573],[@B29-genes-10-00573]\]. Thus, awn length is one of the main distinguishing features in grasses, and domestication process via awn length led to dramatic phenotypic variations through evolutionary processes \[[@B6-genes-10-00573],[@B15-genes-10-00573]\]. There are several arguments in deciding whether the awn is long or short. In barley, awns that are more than twice the length of the spike are classified as long, whereas awns that are less than twice the length of the spike are classified as short \[[@B30-genes-10-00573]\]. In wheat, there is no rule for awns to be long or short, and these categories are estimated via observations \[[@B31-genes-10-00573]\].
Straightness is also another distinguishing feature of grass awns. Awns that are extremely bent or flex from the middle to the tip end are known as curved, whereas awns that are straight or slightly curved are classified as straight. Also, long awns are usually curved, and short awns are straight \[[@B19-genes-10-00573],[@B32-genes-10-00573]\]. Furthermore, the presence of barbs on the outer layer of the awn is also a distinguishing feature in grasses. Rough (or barbed) awns are entirely covered with barbs, whereas smooth (or barbless) awns have no barbs on the outer surface \[[@B11-genes-10-00573],[@B15-genes-10-00573],[@B32-genes-10-00573]\].
In different species, awns are distinct in cell type, structure, and number. Long awns have more elongated cells in the top and bottom and less in the middle, whereas short awns contain small cells, and most of the cells occupy the middle part \[[@B1-genes-10-00573]\]. Except for rye and its relatives lacking chlorenchyma cells, many species such as wheat and barley have parenchyma cells on both awn sides, and parenchyma cells occupy a big part of the awn transverse section ([Figure 1](#genes-10-00573-f001){ref-type="fig"}). Longitudinal sections of the awn show clear branched and elongated chlorenchyma cells with two or more roundish proteinaceous bodies \[[@B33-genes-10-00573]\]. Moreover, several grasses contain stomata in their awn's external tissues, but some cereals such as wheat and rice contain cuticles in the awn external layers. Long awns also possess more phytoliths (silica depositions) and adaxial and abaxial terminals in comparison to short awns \[[@B34-genes-10-00573],[@B35-genes-10-00573]\].
Awn cells are responsible for several functions assisting in the elongation of the awn, as well as the development of the seed grain. Parenchyma cells and green cells (chlorophyll-containing cells) synthesize photosynthate; vascular bundles work together with parenchyma cells to convey assimilates to the other cells; abaxial and adaxial terminals and stomata regulate water, temperature, and metabolic chemicals through transpiration; proteinaceous bodies mainly store proteins and minerals; phytoliths maintain the rigidity and persistence of the awn \[[@B36-genes-10-00573]\]. It is noteworthy that improvements in the performance of cereal crops and forage grasses require an in depth understanding of plant morphological features including those of seed awns.
3. Awn Development Pathways {#sec3-genes-10-00573}
===========================
Plant development analyses revealed that the reproductive phase is simpler in eudicots, such as *Arabidopsis thaliana* (L.) Heynh, where the inflorescence meristem (IM) directly generates the floral meristem (FM) unlike in monocots, such as grasses, where the process can undergo various intermediate stages that may last several weeks \[[@B2-genes-10-00573]\]. Although the reproductive process is more complex in grasses, the vegetative phase is nearly the same. The plant developmental process in grasses is governed by the shoot apical meristem (SAM) and the root apical meristem (RAM), meristems formed during embryogenesis. The SAM controls the development of leaves, stems, and flowers, whereas the RAM controls the rooting system. During vegetative growth, the SAM generates leaf primordia; the leaf primordia produce the axillary meristems (AMs), and these meristems generate the secondary shoots or tillers. After the vegetative growth, the SAM shifts its focus to the inflorescence meristem (IM) and, thus, the reproductive phase is initiated \[[@B2-genes-10-00573],[@B37-genes-10-00573]\]. The sequence of reproductive events varies in grass species, and it involves various intermediate meristems. To better understand the process of awn development, below, we compare the developmental fate of rice and wheat, two major grasses, exhibiting different developmental processes.
In rice, florets are developed by spikelet meristems (SPMs) which are initiated by the branch meristems (BMs), that is, either by primary branch meristems (PBMs) or secondary branch meristems (SBMs). The IM generates PBMs, which then develop SBMs. The branch meristems (BMs) generate the SPMs including lateral spikelet meristems and terminal spikelet meristems, SPMs generate FMs, and FMs initiate spikelets. The empty glumes begin spikelet development, followed by the lemmas, paleas, and lodicules; then, fertile florets produce the stamens and pistils \[[@B21-genes-10-00573]\].
The development of awns in rice is one of the intermediate floral developmental phases. [Figure 2](#genes-10-00573-f002){ref-type="fig"} is a generalized diagram of awn development in rice. It also demonstrates several parts of the spikelet. The initiation of the awn can be defined as the morphological change of the lemma apex as floral organs continue to grow. The awn primordium is initiated after the lemma primordium \[[@B16-genes-10-00573]\]. This stage is followed by the appearance of a pair of ridges, in which the upper ridge generates the panicle and its parts. As the process proceeds, the lemma continues to grow, and the awn propagates from the terminal part composing the vascular bundle of the lemma \[[@B21-genes-10-00573]\]. Although this process clearly shows that awns protrude from the lemma, problems regarding factors causing awn suppression in some rice species persist. Luo et al. \[[@B12-genes-10-00573]\] used both awned and awnless rice cultivars to illustrate awn development. They found that both awned and awnless cultivars exhibit similar growth before the lemma primordium. After the lemma primordium, the lemma apex of the awned cultivar elongates, and the awn primordium differentiates and extends. In contrast, the lemma apex of the awnless cultivar stops growing and forms a round tip. Thus, awn development in rice is divided into two main stages: awn formation, which begins after lemma primordia and ends at stamen primordia; and awn elongation, which begins during stamen primordia and ends during early maturity, the dough stage \[[@B16-genes-10-00573]\].
In wheat, the floral developmental process is more complex. Unlike rice having a single floret per spikelet, wheat spikelets contain three to six florets \[[@B38-genes-10-00573]\]. This variation in spikelet patterning may affect the developmental processes. After the shift from the vegetative to reproductive phase, the IM initiates the spike meristem (SM). This phase is indicated by the appearance of a pair of ridges. The upper ridge represents the SPM, and the lower ridge forms the leaf primordium. Both ridges continue to differentiate until the upper covers and suppresses the lower. As the process continues, the SM initiates spikelet meristems (SPMs) starting in the middle toward the base and the top of the floral body. After the initiation of FMs by SPMs, two pairs of glumes differentiate. The first pair appears on both sides of the SPM as transverse ridges, and the latter appears above the first pair. The first pair becomes empty glumes, and the second differentiates into the lemmas of the first and second florets. As the development proceeds, the FM elongates upwardly (from the base to the apex) until the initiation of the lemma of the sixth floret. The first two florets generate anthers, followed by the paleas and then the pistils. The third floret generates its parts, followed by the fourth, fifth, and sixth floret, respectively. In the awned species, lemmas continue to elongate and develop the awns. The development of awns within the spikelet follows the fate of their respective florets, that is, starting in the first floret to the sixth \[[@B23-genes-10-00573],[@B37-genes-10-00573]\].
The study by Bonnett \[[@B38-genes-10-00573]\] unveiled that all parts of the wheat floral structure including awns become visible before the differentiation of the last floral organ, the peduncle. Despite this, the initiation time of the awn primordium remains poorly understood. The study by Ponzi and Pizzolongo \[[@B39-genes-10-00573]\] in common wheat cv. Ofando found that SPMs and subtending leaf primordia expand and initiate FMs, the process probably followed by the initiation of awn primordia. The authors revealed that the initiation of awn primordia at this stage was indicated by the presence of the positive starch granules (PSGs) at the base of the developing spikelet. As the development proceeded, FMs initiated floral primordia at the axil of the lemma, while awns expanded rapidly at the top of the lemmas. The authors also found that awns and anthers might be initiated at the same time because PSGs were also found in the anthers, but were absent in other floral parts. This finding is consistent with the study by Bonnett \[[@B40-genes-10-00573]\] in barley and Bremer-Reinders \[[@B41-genes-10-00573]\] in rye. In both species, the basal of the lemma differentiates rapidly and initiates awn primordia during the initiation of anther primordia. After the initiation of floral parts, the awn expands more rapidly than the lemma and palea to attain a considerable length. Furthermore, the elasticity in the duration from awn primordium to the time when the awn is apparent raised several arguments. A study suggests that this duration is controlled by genetic mechanisms and growing conditions such as vernalization, temperature, and photoperiod \[[@B42-genes-10-00573]\].
Awn development varies considerably in grass species despite possible little similarities. In the *Danthonieae* tribe, the lemma is divided at the apex into three lobes, in which the middle one extends into a geniculate awn. In the *Aveneae* tribe, the awn arises on the back of the lemma with oneor two lobes \[[@B43-genes-10-00573]\]. This variation is regulated by several genetic switches that may initiate the awn primordia, inhibit awn differentiation in awnless species, control the duration of the process and, thus, control the interaction of awns and several signaling centers \[[@B12-genes-10-00573],[@B16-genes-10-00573]\]. It is, thus, noteworthy that unveiling the molecular underpinnings of awn development remains important.
4. Genetic Basis of Awn Development in Grasses {#sec4-genes-10-00573}
==============================================
As the awn exhibits complex morphology and development \[[@B14-genes-10-00573],[@B44-genes-10-00573]\], identifying its growth regulators was a challenge for decades \[[@B1-genes-10-00573]\]. To address this, focus should be put on the genetic mechanisms given their considerable influence in the development of awns \[[@B6-genes-10-00573],[@B12-genes-10-00573],[@B37-genes-10-00573]\]. This section reviews strategies that were applied so far to identify quantitative trait loci (QTLs) and causal genes for awn development. It also discusses how these efforts could provide novel insights into understanding awn development and how these insights may trigger ongoing and new research initiatives.
4.1. Awn Development and Recent Genetic and Genomic Advances {#sec4dot1-genes-10-00573}
------------------------------------------------------------
The artificial selection by ancient farmers was considered the root of the reduced awn length in cultivated grasses \[[@B5-genes-10-00573]\]. However, questions on the variability of awn features among and within grass species persist. Moreover, the interaction of awns and other domestication traits such as panicle architecture, grain size, grain weight, and seed shattering obscure the genetic basis of awn development and, correspondingly, may have far-reaching economic implication in several grass species \[[@B45-genes-10-00573],[@B46-genes-10-00573]\]. To resolve this puzzle, scientists adopted several approaches to reveal the genetic basis of awn development.
QTL mapping (traditional linkage mapping) was an important tool in chasing awn development regulators for many years \[[@B47-genes-10-00573]\]. QTL mapping depends on molecular markers such as restriction fragment length polymorphisms (RFLP), amplified fragment length polymorphism (AFLP), sequence-related amplified polymorphism (SRAP), simple sequence repeat (SSR) or single-nucleotide polymorphism (SNP), and mapping populations, including F~1~, F~2~, backcross (BC), recombinant inbred lines (RIL), doubled haploid (DH), and multiparent advanced generation intercross (MAGIC) lines \[[@B48-genes-10-00573]\]. Several studies adopted QTL mapping and identified a huge number of QTLs that control awn development in major grasses such as rice, wheat, and barley \[[@B49-genes-10-00573],[@B50-genes-10-00573]\]. QTL mapping enabled scientists to develop recombinant inbred lines (RILS), which were used to illustrate awn development in grasses (for example, in rice) and disclose the interaction of awns and yield traits \[[@B51-genes-10-00573]\]. However, unlike rice species, compatible with this tool, defining a few megabases and covering thousands of genes was a hindrance in many species, such as polyploidy wheat and barley, having huge and complex genomes \[[@B48-genes-10-00573]\]. To resolve this issue, scientists envisaged several other approaches, such as fine mapping and QTL cloning. These approaches displayed several QTLs underlying awn development in rice \[[@B49-genes-10-00573],[@B51-genes-10-00573]\], barley \[[@B18-genes-10-00573]\], wheat \[[@B19-genes-10-00573]\], and sorghum \[[@B52-genes-10-00573]\] despite being a time-consuming and expensive approach. They also assisted to dissect several causal genes for awn development in some major cereals such as rice \[[@B12-genes-10-00573]\] and barley \[[@B1-genes-10-00573]\].
The interactions among awn development QTLs and grain yield QTLs and the environment were a drawback for breeding strategies in grasses. Therefore, QTL cartographer (that is, a software suite which uses molecular markers to map quantitative traits) and QTL networking (that is, bioinformatics-based software which analyzes QTL data of large genotypes and phenotypes collected from many groups of related individuals) are alternatives to resolving this issue \[[@B53-genes-10-00573]\]. Recently, research by Masoud et al. \[[@B50-genes-10-00573]\] applied these approaches and intriguingly found six additive awn length QTLs using QTL cartographer and three additive and two epistatic QTLs using QTL networking.
Genome-wide association analysis (GWAS) offers a new approach to gene discovery unbiased with regard to presumed functions or locations of the causal variant. It was a powerful tool in displaying QTLs and genes underlying awn development. This approach applies linkage disequilibrium (LD) to unveil polymorphisms associated with a trait of interest \[[@B48-genes-10-00573]\]. Using this approach, Magwa et al. \[[@B46-genes-10-00573]\] identified several QTLs linked to awn length and one QTL linked to *AN1* (the main gene controlling awn development in rice). Interestingly, they also identified an association SNP in a region near the seed shattering gene (*qSH1*), indicating that awn development and seed shattering may exhibit molecular interactions. Finally, these approaches enabled researchers to identify several genes and QTLs underlying seed awning especially in major cereals. Thus, additional advances are required to further explain awn development in several species, particularly forage grasses, and unveil the interaction between awns and other seed traits.
4.2. Molecular Mechanisms Underlying Awn Development in Grasses {#sec4dot2-genes-10-00573}
---------------------------------------------------------------
### 4.2.1. Mechanisms of Awn Development Causal Genes {#sec4dot2dot1-genes-10-00573}
Awn development is a polygenic trait that is regulated by many genes, with several genes identified in rice, along with their role in the initiation and growth of the awn ([Figure 3](#genes-10-00573-f003){ref-type="fig"}; [Table 1](#genes-10-00573-t001){ref-type="table"}). Although awns have a negligible effect on rice performance \[[@B1-genes-10-00573]\], this species received considerable interest in displaying the genetic basis of awn development due to its small genome size and its relationship with several other cereal crops. Additionally, rice species exhibit variation in awn features. For example, wild rice species such as *Oryza rufipogon* Griff. and *Oryza barthii* A. Chev. are long-awned. Most cultivated rice species such as *O. sativa* cv. Indica and *Oryza glaberrima* Steud. are awnless, whereas *O. sativa* cv. Japonica is short-awned \[[@B6-genes-10-00573]\]. This variability reflects that awn development in rice is controlled by drastic genetic changes. Matsushita et al. \[[@B13-genes-10-00573],[@B14-genes-10-00573]\] used the introgression lines generated from wild rice (*Oryza meridionalis* Ng.) and cultivated rice (*O. sativa*) and uncovered five genes (*An6*, *An7*, *An8*, *An9*, and *An10*) for awn development in *O. meridionalis*. The authors discovered that *An6*, *An7*, and *An8* generated longer awns than other genes and *An10* generated a short-awn phenotype.
Several years later, Luo et al. \[[@B12-genes-10-00573]\] used genetic mapping in *O. rufipogon* and *O. sativa* to identify further mechanisms underlying awn development. They found some major genes related to cytokinin regulated awn development. *AN1*, the major domestication gene located on chromosome 4, encodes a basic helix-loop-helix (bHLH) transcription factor linked to awn development in wild rice (*O. rufipogon*). The authors also used in situ hybridization to unveil the cause of awnless habits in cultivated rice (*O. sativa*), and they surprisingly found the *an1* allele in *O. sativa*, which reduces cytokinin in awn primordia and, thus, inhibits awn differentiation. This indicates that the absence of awns in cultivated rice may be caused by the mutation in *AN1* during awn elongation. The authors also observed that the expression of *AN1* in wild rice ceased during the late flowering stage, suggesting that the further elongation of awns after flowering might be induced by other genes. In this respect, Gu et al. \[[@B16-genes-10-00573]\] revealed that the complete action of *AN1* in wild rice relies on the mutual effect with other genes. Intriguingly, they cloned *AN2* (*AWN2*), a major gene that encodes a *Lonely Guy Protein 6* (*OsLOGL6*), which catalyzes the final step of cytokinin synthesis and, thus, promotes awn elongation in *O. rufipogon.* The authors also stated that the mutual effect of *AN1* and *AN2* could remarkably promote awn development in *O. rufipogon*, implying that *AN1* controls the initiation and formation of awns, whereas *AN2* regulates the elongation of awns. The authors also suggested the awnless phenotype in some rice species may be caused by the suppression or mutation in *AN1*, while the short-awned phenotype may be caused by the mutation in *AN2*. In conclusion, cytokinin is one of the important plant hormones promoting cell division and differentiation during plant organ development. Rice *an1* mutant showed a reduced cytokinin level in awn primordia, while *AN2* encodes a cytokinin synthesis protein *OsLOGL6*, which further confirmed the role of cytokinin in awn development at the molecular level.
In barley, awn development is mainly controlled by *Lks1* and *Lks2* genes located on chromosome arms 2HL and 7HL, respectively. *Lks1* (*Awnless1*) is a dominant gene, which causes the awnless phenotype in barley \[[@B59-genes-10-00573]\]. *Lks2* (*LONG AWN2*) encodes the SHORT INTERNODES (SHI) family transcription factor that promotes awn elongation \[[@B1-genes-10-00573]\]. By contrast, its mutant gene, *SHORT AWN2* (*lks2*), reduces number of longitudinal cells in the awn and, thus, shortens awn length up to 50% \[[@B5-genes-10-00573]\].
In wild wheat (*Aegilops tauschii* Coss.), Nishijima et al. \[[@B18-genes-10-00573]\] identified the awnless locus (*Antr*) on chromosome arm 5D, which inhibits awn formation and, thus, affects the awnless phenotype in wild wheat. The identification of this novel gene provided intriguing insights into the genetic basis of awn development in wheat species. However, further investigation is needed to unveil whether this gene is involved in the development of awns in cultivated wheat.
### 4.2.2. Functions of Floral Meristem Genes {#sec4dot2dot2-genes-10-00573}
The shift from vegetative to reproductive development in grass species is under strict genetic control, in which each pathway involves genetic factors that affect the architecture of secondary inflorescence meristems. Genetic advances reported ample evidence that gene mutations may obstruct the growth of floral meristems and thus transform floral parts into functional awn-like structures. In rice, the growth of shoot apical meristem is regulated by many genes, including *SHOOTLESS2* (*SHL2*), *SHOOT ORGANIZATION1 (SHO1*), and *SHOOT ORGANZATION2* (*SHO2*). Genetic experiments revealed that these genes have strong mutant alleles (*shl2*, *sho1*, and *sho2*) that encode proteins acting in trans-acting small interfering RNA, acting in the verification and silencing of some plant genes and, thus, inhibiting the growth of apical shoot meristem. This process may cause plant abnormalities, including the transformation of the lemma and floral meristems into awn-like structures \[[@B63-genes-10-00573],[@B64-genes-10-00573]\]. *TOB1* (*TONGARI-BOUSHI1*), *DL* (*DLOPPING LEAF*), and several other members of the *YABBY* gene family encode transcription factors that are involved in the upholding of plant meristems and development of floral organs in rice. Interestingly, *DL* interacts with *OsETT2* of the *ARF* gene family to initiate cell divisions in the awn primodium and, thus, regulate awn length; *tongari-boushi1* (*tob1*) mutants possess similar functions, but also control the formation and elongation of lemma and palea \[[@B65-genes-10-00573],[@B66-genes-10-00573]\].
Evidence also shows that the overexpression of genes that are involved in the expansion of floral meristems can control awn development. In wheat, for example, *Wknox1a*, *Wknox1b*, and *Wknox1d* genes are located on the 4AS, 4BS, and 4DS chromosome arms, respectively. In some wheat cultivars, such as Chinese spring, the overexpression of these genes in the floral meristem may actuate awn formation \[[@B67-genes-10-00573]\]. This supports the idea that, although the causal genes for awn development arenot yet cloned in cultivated wheats, this process is controlled by several interacting genes. The current data suggest that variation in awn length and hood phenotype in wheat is a result of alterations in genes underlying floral meristems \[[@B19-genes-10-00573]\].
### 4.2.3. The Role of Interacting Genetic Factors {#sec4dot2dot3-genes-10-00573}
Several molecular experiments showed that awn development is controlled by multiple interacting genetic factors. In barley, for instance, *LONG AWN2* (*Lks2*) and *HvKNOX3* genes and *suK* alleles (*suKD*, *suKB*, *suKC*, *suKE*, and *suKF*) exhibit molecular interactions, resulting in the variation of awn length. The *HvKNOX3* and *suK* alleles possess a dominant mutant K, which either shortens awn length or inhibits awn formation \[[@B62-genes-10-00573],[@B68-genes-10-00573]\]. This dominant mutant K may also divert the initiation and growth of awn meristem and forms a hooded (flower-like) phenotype \[[@B60-genes-10-00573]\]. Other studies observed that the mutation in *suK* alleles and *Lks2* gene may inhibit the expression of *HvKNOX3* and, thus, promote awn development \[[@B19-genes-10-00573],[@B21-genes-10-00573],[@B62-genes-10-00573]\]. Furthermore, recombinant inbred lines (RIL) from two barley accessions, Azumamugi and Kanto Nakate Gold, displayed a QTL for awn length near the *vrs1* locus, which is known to be involved in regulating spike development and plant statute traits. Molecular analyses revealed that interactions between these genetic factors have pleiotropic effects that result in reduced awn length \[[@B67-genes-10-00573]\]. Genomic studies in barley also detected molecular interactions between *vrs1* and *Awnless1* (*Lks1*) locus and mutants for reduced awn length, which caused awnless and short-awned phenotypes, respectively \[[@B67-genes-10-00573]\]. Analysis of the gene controlling the spike density in barley, *dsp1*, revealed that interaction of this gene with the *short awn 2* (*lks2*) gene underlays short and thin awns in Aizu Hadaka 3, a six-rowed barley variety \[[@B69-genes-10-00573]\]. These observations support the view that awn length in barley is controlled by a wide variety of molecular interactions. They also leave open interesting questions on the molecular interactions among *vrs1*, *lks2*, *dsp1*, and *suK* alleles. Furthermore, a recent study detected *Vrs2*, the gene responsible for the floral organ patterning in barley, in several floral parts including the lemma \[[@B20-genes-10-00573]\]. Identifying whether this gene is expressed in the awn remains a goal for future studies.
### 4.2.4. Regulation of Awn Formation and Length by Dominant Inhibitors {#sec4dot2dot4-genes-10-00573}
In wheat, awn development is strictly controlled by three dominant inhibitors: *Hooded* (*Hd*), *Tipped1* (*B1*), and *Tipped2* (*B2*). Molecular advances detected *Hd* on chromosome arm 4AS, *B1* on chromosome arm 5AL, and *B2* on chromosome arm 6BL \[[@B19-genes-10-00573]\]. Full-awned phenotypes carry three homozygous recessive (*hd*, *b1*, and *b2*) alleles. Awnless phenotypes possess two dominant alleles and one recessive allele, that is, either *HdHdb1b1B2B2* as in the Chinese spring cultivar or *hdhdB1B1B2B2* as in the Federation cultivar \[[@B19-genes-10-00573],[@B70-genes-10-00573]\]. Furthermore, Yoshioka et al. \[[@B19-genes-10-00573]\] used RILs derived from short-awned common wheat cultivars (Chinese Spring and Mironovskaya 808) to investigate the further impact of these inhibitors on the variability of awn traits. Intriguingly, they found that wheat RILs that possess *Hd* developed very short awns that are bent at the base. Wheat RILs with *B1* generated short awns at the base and middle of the spike, but awn length rose at the top of the spike reaching 1cm. RILs with *B2* produced short, curved awns that were not bent at the base as in *Hd*. Despite the undisputable role of these inhibitors during awn development in wheat, none were characterized molecularly. A study attempted to associate *Hd* with the *Wknox1a* gene. Also, *Hd* and *B2* were found near the *DL* and *RAE2* genes, respectively; however, neither of these genes is directly coupled with awn development in wheat \[[@B19-genes-10-00573]\]. Thus, future investigations should continue to improve our limited knowledge on molecular genetic factors coupled with awn development in wheat.
5. Conclusions and Future Perspectives {#sec5-genes-10-00573}
======================================
Awn development was described morphologically and genetically, especially in major cereal crops like rice, wheat and barley. These advancements not only offer an intriguing paradigm for studying the evolution and domestication of grasses, but also have significant implications for future breeding programs. From a breeding standpoint, identifying the interactions between seed characters and agronomic traits, including those determining yield and quality potential, remains an essential goal to realizing the potential of agricultural research and innovation. In this respect, combining morphological- and molecular-based approaches will yield comparative data that will help to unveil the pleiotropic impacts of awns in grass species. All these essential perspectives should enable us to unravel the link between awn morphology and several genetic factors in major cereal crops, but the question regarding how we can achieve the same level of understanding in forage grasses remains. In the future, these important genes for awn development identified in major crops, such as *AN1*, *AN2*, *GAD1*, *LABA1*, *Lks1*, and *Lks2*, can effectively be used as a source of candidate genes for other grass species, and they can also facilitate comparative genome research between crops and forage grasses. The application of the information presented here should considerably aid in this quest.
We thank Ding, X.S. and Zulfi Jahufer for their constructive suggestions and remarks. We also thank the reviewers for their insightful comments on an earlier version of this manuscript.
Both authors conceived the idea, designed the study, discussed the content, and wrote the manuscript. F.N. designed the graphics and diagrams. Both authors revised the manuscript.
This work was supported by the Fundamental Research Fund for the Central Universities (LZUJBKY-2019-37), the Chinese National Basic Research Program (2014CB138704), and the program for Changjiang Scholars and Innovative Research Team in University (IRT13019).
The authors declare no competing interests.
{#genes-10-00573-f001}
{#genes-10-00573-f002}
{#genes-10-00573-f003}
genes-10-00573-t001_Table 1
######
Genes that underlie awn development in major grass species.
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Species Locus Name Locus Symbol Associated Makers Physical Position ^a^ Chr. ^b^ Phenotypic Function Reference
---------------------- ----------------------------------------------------- -------------- ------------------------ ----------------------- ---------- --------------------------------- -----------------------------------------------------
*Oryza sativa* *AWN1* *AN1* *M6298*, *RM6285* 5.88 Mb (region) 4 Awn formation Luo et al. \[[@B12-genes-10-00573]\]
*O. sativa* *AWN2* *AN2* *FM5*, *FM6* 56 kb (region) 4 Awn elongation Gu et al. \[[@B16-genes-10-00573]\]
*O. sativa* *AWN3* *AN3* \- \- 3 Awn elongation Takamure et al. \[[@B54-genes-10-00573]\]
*O. sativa* *AWN4* *AN4* *RT1-8a\*T65*\ \-\ 8 Awn elongation,\ Sato et al. \[[@B55-genes-10-00573]\]
*RT1-8c\*T65*\ -\ inflorescence\
*RT8-10\*T65* - development
*O. sativa* *AWN5* (*An5*) *AN5* *RZ740*\ 21.2 cM\ 4 Awn elongation Xiong et al. \[[@B56-genes-10-00573]\]
*RG122* 24.5 cM
*O. sativa* *An-10* \[*An5(t)*\] *An10* *C1677*\ 2.4 cM\ 3 Awn elongation Kubo et al. \[[@B57-genes-10-00573]\]
*G1316* 2.2 cM
*Oryza glumaepatula* *An-11*(*An7*) *An11* *RM3419*\ 3.6 cM\ 5 Awn elongation Matsushita et al. \[[@B13-genes-10-00573]\]
*RM289* 3.6 cM
*O. glumaepatula* *An-12*(*An8*) *An12* *RM261*\ 12.2 cM\ 4 Awn elongation Matsushita et al. \[[@B13-genes-10-00573]\]
*RM1359* 7.1 cM
*Oryzameridionalis* *An-13*(*An6*) *An13* *RM3496* 11.7 cM 8 Awn elongation Matsushita et al. \[[@B14-genes-10-00573]\]
*O. meridionalis* *An-14*(*An9*) *An14* *RM8111*\ 3.0 cM\ 1 Awn elongation Matsushita et al. \[[@B14-genes-10-00573]\]
*RM8051* 3.6 cM
*O. meridionalis* *An-15*(*An10*) *An15* *RM237*\ 15.1 cM\ 1 Awn elongation Matsushita et al. \[[@B14-genes-10-00573]\]
*RM265* 3.0 cM
*O. sativa* *Chromogen for Anthocyanin* *C* *XNpb165-1*\ 16.4 cM\ 6 Lemma, palea, and\ Kishimoto et al. \[[@B58-genes-10-00573]\]
*XNpb200* 17.1 cM awn color
*O. sativa* *Long and Barbed Awn 1* *LABA1*\ *M3*, *RM17242* 34.6 kb (region) 4 Awn elongation Hua et al. \[[@B15-genes-10-00573]\]
(*RAE1*)
*O. sativa* *Grain Number, Grain Length, and Awn Development 1* *GAD1*\ *MX14*, *MX16* 6 kb (region) 8 Awn elongation,\ Jin et al. \[[@B17-genes-10-00573]\]
(*RAE2*) grain number
*O. glabberima* *Regulator of Awn Elongation 3* *RAE3* *6KG28331*, *6KG30196* 1.9 Mb (region) 6 Awn elongation Furuta et al. \[[@B6-genes-10-00573]\]
*Triticum aestivum* *Hooded* *Hd* *WABM229716*,\ 0.9 cM 4AS Awn suppression Yoshioka et al. \[[@B19-genes-10-00573]\]
*WABM117400*,\
*WABM233735*17.1
*T. aestivum* *Tipped 1* *B1* *Xgwm291* 1.3 cM 5AL Awn suppression Yoshioka et al. \[[@B19-genes-10-00573]\]
*T. aestivum* *Tipped 2* *B2* *WABM232658*\ 1.3 cM\ 6BL Awn suppression Yoshioka et al. \[[@B19-genes-10-00573]\]
*WABM243094* 1.9 cM
*Aegilops tauschii* *Anathera* *Antr* *S57615-1*\ 1.3 cM\ 5DS Awn suppression Nishijima et al. \[[@B18-genes-10-00573]\]
*Xctg211719* 3.9 cM
*Hordeum vulgare* *Awnless 1* *Lks1* *SNP 1_0619*\ 133.59 cM\ 2HL Awn suppression Franckowiak and Lundqvist \[[@B59-genes-10-00573]\]
*SNP 1_1533* 141.56 cM
*H. vulgare* *Long Awn2* *Lks2* *k06123*\ 0.27 cM\ 7HL Awn elongation Yuo et al. \[[@B1-genes-10-00573]\]
*k04151* 1.0 cM
*H. vulgare* *Hooded lemma 1* *Kap1*\ *glf3* 25.1 cM 4HS Hood formation, awn suppression Müller et al. \[[@B60-genes-10-00573]\]
(*Knox3*)
*H. vulgare* *Six-rowed spike 1* *vrs1* *e40m36-1110S*\ 0.01 cM\ 2H Reduces awn length Komatsuda et al. \[[@B61-genes-10-00573]\]
*BC12348* 0.06 cM
*H. vulgare* *K Suppressor* loci *suKB* *suKB* *E34M46*\ 16.9 cM\ 7H Reduces awn length,\ Roig et al. \[[@B62-genes-10-00573]\]
*E35M39* 9.9 cM Hood formation
*H. vulgare* *K Suppressor* loci *suKC* *suKC* *E35M44*\ 0.0 cM\ 7H Reduces awn length,\ Roig et al. \[[@B62-genes-10-00573]\]
*E41M46* 7.8 cM Hood formation
*H. vulgare* *K Suppressor* loci *suKD* *suKD* *E37M38*\ 15.9 cM\ 5H Reduces awn length,\ Roig et al. \[[@B62-genes-10-00573]\]
*E41M36* 1.1 cM Hood formation
*H. vulgare* *K Suppressor* loci *suKE* *suKE* *E40M43*\ 5.6 cM\ 7H Hood formation Roig et al. \[[@B62-genes-10-00573]\]
*E36M44* 0.0 cM
*H. vulgare* *K Suppressor* loci *suKF* *suKF* *E35M40*\ 5.3 cM\ 7H Hood formation Roig et al. \[[@B62-genes-10-00573]\]
*E43M32* 22.4 cM
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
^a^ Physical position: bp, base pair; cM, centimogan; kb, kilobase; Mb, megabase. ^b^ Chr, chromosome.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Patients with chronic kidney disease (CKD) are usually recommended to maintain low protein diet to slow down renal function deterioration [@B1]. It is well-recognized that progressive decline of renal function with aging is common [@B2]. However, higher protein intake can prevent protein-energy malnutrition in the elderly. Therefore, how to adjust protein intake appropriately for the elderly with CKD is an important issue. For elderly population without CKD, the recommended protein intake is over 0.8 g/kg/day [@B3]. It has been estimated that 10 to 35 % of elderly people take protein below minimal requirement (0.7 g/kg BW/day) [@B4]. In order to minimize the progression of sarcopenia, increased protein intake to 1.0-1.3 g/kg/day was suggested [@B5]. In a national-wide study, glomerular filtration rate (GFR) less than 30 mL/min/1.73 m^2^was an independent factor associated with malnutrition for older adults [@B6]. Collectively, it would be better to individualize the amount of protein intake by close monitoring renal function and muscle wasting status in the elderly. Previous studies of body compositions of patients with CKD are usually with small numbers and mostly included patients with age less than 60 [@B7], [@B8]. The concern on safety of low protein diet for elderly patients is raised but only little information is available.
Anorexia, dietary restriction, acidosis, and inflammation in CKD patients can increase the risks of cachexia and protein-energy wasting syndrome [@B9]. Muscle wasting is associated with increased mortality in patients with chronic illness [@B10]-[@B12]. Therefore, it is indicated to assess body composition and monitor muscle mass in these patients. Serial body composition measurements can detect changes in muscle mass and provide additional information of nutritional status than common nutritional markers, such as body weight, body mass index (BMI), and serum albumin [@B12], [@B13]. Dual energy X-ray absorptiometry (DXA) is the gold standard for body composition assessment. However, the machine occupies large space with high cost, and is not recommended for routine clinical use.
In the present study, bioelectrical impedance analysis (BIA) with tetra-polar impedance meter was employed for the determination of body composition. We analyzed the effects of low protein diet on body compositions of CKD patients. We also compared the alterations of body composition between elderly and non-elderly patients.
Patients and methods
====================
Patients with eGFR ≤ 45mL/min/1.73m^2^ (CKD stage 3b) regularly followed up in nephrology clinics were recruited. Patients were excluded if they had chronic heart failure (New York Heart Association Functional Classification System, ≥ stage III) or active infection, and any of which might affect dietary intake, such as swallowing difficulty or cancer under treatments. Subjects with contraceptive devices, metallic transplant, liquid filled catheter, or pregnancy were excluded as well. This study was approved by Chang Gung Medical Foundation Institutional Review Board (101-3599B). All participants involved gave written informed consent.
Demographic data including gender, age, body weight, body height, BMI were collected. Diabetes mellitus (DM) was defined as patients who were receiving oral anti-diabetic or insulin treatment; with fasting blood sugar ≥ 126 mg/dL or random blood sugar ≥ 200 mg/dL with associated symptoms. Blood pressure was measured at every visit. Laboratory data including serum creatinine, albumin, hemoglobin, glycosylated hemoglobin, total cholesterol, high-density lipoprotein, low-density lipoprotein, and triglyceride were measured at baseline and one year later. The eGFR was calculated by using Modification of Diet in Renal Disease (MDRD) formula [@B14]. The participants received dietary counselling and their body compositions were measured every three months for one year. The registered dietitians calculated the energy and protein intake of these CKD patients from each interview. Dietary counselling was individualized and focused on educating and advising patients about food portions, selections and preparations. For participants\' understanding and encouraging them doing exercise, the registered dietitians interpreted the results of body composition measurement to all participants. The low protein group was defined as average protein intake ≤ 0.8 g protein /kg/day [@B15]. The rest of enrolled patients were defined as non-low protein group. Age greater than 60 was defined as the elderly group in the present study.
Waist circumference was measured at the midway between the lowest rib and iliac crest. The participants were instructed to fast for 4-hours before body composition measurement. The assessment of body composition followed the manufactory\'s protocol of the bioelectrical impedance analysis (BIA) (ioi 353, Jawon Medical, S. Korea). The BIA device measured five body segments (right arm, right leg, left arm, left leg, and trunk) via tetra-polar electrode method using 8 touch electrodes. Appendicular skeletal muscle mass (ASM) index is calculated as muscle of limbs measured by BIA divided by height squared (kg/m^2^) [@B16], [@B17].
Statistical methods
-------------------
All statistical analyses were performed using statistical SPSS version 19 software (IBM Corporation). Data were presented as mean ± standard deviation or percentage as appropriate. Continuous variables were compared using ANOVA or the Mann-Whitney U test. Comparison of body compositions at baseline and every 3 months was analyzed by paired t test or Wilcoxon test. A p value \< 0.05 was considered as statistically significant.
Results
=======
A total of CKD patients including 103 elderly patients and 56 non-elderly patients were recruited. Table [1](#T1){ref-type="table"} displays their baseline characteristics of non-low protein and low protein groups in different age groups. The mean age of elderly CKD patients was significantly greater than the non-elderly group (70.2 ± 6.8 vs. 50.7 ± 8.9 and 70.2 ± 7.3 vs. 46.8 ± 9.5 in non-low and low protein groups respectively, both p \< 0.001). Diabetes accounts for 23% of enrolled patients. In elderly patients, protein and energy intake were significantly lower in low protein group than non-low protein group (0.71 ± 0.06 g/kg and 23.3 ± 2.5 kcal/kg vs. 1.01 ± 0.17 g/kg, 29.0 ± 4.2 kcal/kg, both p \< 0.001). There were no significant differences in blood pressure, BMI, waist circumference and eGFR. The biochemical data was similar between two groups. Elderly patients in low protein group had higher body fat percentage and lower muscle percentage than non-low protein group (p \< 0.05). No difference was noted in their ASM index. In the non-elderly patients, low protein group had lower protein intake and energy intake (both p \< 0.001). Their body compositions did not differ between two protein groups. We further compared elderly and non-elderly patients in either non-low or low protein groups. In non-low protein patients, diastolic blood pressure was higher in non-elderly patients (p \< 0.05). In the low protein groups, non-elderly patients had higher serum albumin levels and lower total cholesterol levels than the elderly patients (both p \< 0.05). Comparison in body composition revealed non-elderly patients had lower body fat percentage and higher muscle percentage than the elderly (both p \< 0.05).
Table [2](#T2){ref-type="table"} represents the baseline and 1-year follow-up data of non-elderly patients. In one year, we found there was significant decline of eGFR in non-low protein group while the eGFR was not influenced in low protein group. The biochemical data and body composition did not change significantly in 1-year follow up either in non-low or low protein groups. Table [3](#T3){ref-type="table"} presents the changes in elderly CKD patients. There was significant decrease in BMI and eGFR in the non-low protein group after 1-year follow-up. Modest but significant increase in albumin level was noted. Their hemoglobin level was decreased. Measurement in body composition indicated that a significant decrease in fat and increase in muscle component after 1 year (both p \< 0.05). In low protein group, their BMI was decreased and levels of serum albumin and triglyceride were increased significantly. Comparison in body composition revealed decrease in fat percentage, including total body and trunk fat. The muscle component was increased (p \< 0.05). Similar to the results of comparison in baseline, after 1 year, there were significant differences between non-low and low protein groups in fat and muscle distribution. Patients in low protein group had higher percentage of fat and lower percentage of muscle (both p \< 0.05). There was no significant change in ASM index after 1-year follow-up in both groups.
We further analyzed the serial changes of muscle percentage in every 3 months body composition measurements. As shown in figure [1](#F1){ref-type="fig"}, elderly CKD patients had lower muscle percentage than the non-elderly CKD patients. The percentage did not change significantly in non-elderly patients in one- year follow-up. There was significant increase at 12 months measurement in the elderly patients.
Discussion
==========
Our study clearly demonstrated that diet intervention with low protein therapy did not affect nutritional status in CKD patients. Furthermore, in elderly CKD patients, despite their progressive decrease in BMI, low protein diet was associated with increased serum albumin level and their muscle mass were preserved. In 1-year follow-up, there was a significant decline of eGFR in patients with non-low protein intake.
Therefore, low protein diet is the recommended nutritional therapy for CKD patients especially for those with eGFR less than 45 mL/min/1.73m^2^. However, the potential risk of protein-energy wasting from dietary protein restriction prompted researchers to investigate the effect of low protein diet on nutrition status. Most studies analyzed the effect of low protein diet on body composition focused on middle-aged patients. These studies indicated that low protein diet did not have adverse effects on body composition despite patients usually had weight loss in the first six months then recovered eventually [@B18], [@B19]. With constant BMI, body fat percentage increased with aging [@B20]. In our study, comparing with non-elderly CKD patients, the elderly had higher percentage of fat. We further compared the alterations of body composition and found no significant change in the non-elderly patients. In the elderly CKD patients, low protein diet preserved muscle mass, and serum albumin was even increased.
It has been reported that elderly with higher BMI such as 25-35 kg/m^2^had lower mortality [@B21], indicating elderly should maintain higher BMI. In another study, Lu *et al.* found the beneficial effect of high BMI was attenuated in patients with eGFR \< 30 mL/min/1.73 m^2^ [@B22]. Therefore, whether higher BMI is associated with better outcome in CKD patients remains inconclusive. In non-CKD population, elderly with higher skeletal muscle mass index rather than BMI were associated with lower mortality [@B23]. In CKD patients, decreased abdominal adiposity together with lower waist circumferences and lower trunk fat, were associated with improved systemic inflammation and lower mortality [@B24]-[@B26]. In a longitudinal follow-up study on healthy elderly with stable energy intake and body weight, decrease in physical activity can cause progressive decrease in fat-free mass and increase in fat mass [@B27].
Previous studies have shown for the elderly with age greater than 70, unintended weight loss occurred even with disease absent [@B28], [@B29]. In our study, BMI was decreased significantly in elderly CKD patients in one-year follow-up. Both non-low protein and low protein group had body fat percentages decreased and muscle percentages increased. Apparently, these changes were not observed in the non-elderly patients. This finding indicated that aging process plays a key role in affecting body composition irrespective of protein intake. Nevertheless, whether the decline of renal function during aging contributes to the above change in BMI as well as body composition is unclear. Determination of body composition helps providing important information that elderly CKD patients both in non-low or low protein group can maintain muscle mass. Hence, patients with low protein intake still can preserve their muscle mass and serum albumin level was not reduced.
In the present study, we found patients with non-low protein diet were associated with significant decrease in eGFR during 1-year follow-up irrespective of their age. This finding highlights the importance of low protein intake in CKD patients. Compared with low protein group, patients with non-low protein diet had a drop of 3.5 mL/min/1.73 m^2^in non-elderly CKD patients and it was 1.2 mL/min/1.73 m^2^ in the elderly group. In the low protein group, their eGFR change was decreased minimally. How to retard the progressive loss of residual renal function of CKD patients is primarily the utmost goal of CKD care [@B30]. Our results highlight the important role of diet intervention among CKD patients.
There are several limitations in our study. First, underestimation of energy intake may occur with the method of dietary history. Even with our method of estimating food portions during the dietary interview, calculation of fat intake may be imprecise. Second, the treatment of blood pressure, lipid profile and glycemic control of diabetic patients were not included for detail analysis. Thirdly, patients without diet intervention were not enrolled as control group. Lack of control group may underestimate the effect of low protein diet intervention. Lastly, 1-year follow-up period is rather short and a longer observation may help provide longitudinal changes in more aspects of CKD patients.
In conclusion, Low protein diet did not affect the nutritional status of elderly CKD patients. Their muscle mass was preserved with decreasing fat component. With the addition of body composition information provided by BIA device, diet intervention therapy can offer beneficial effects more effectively and appropriately in CKD patients.
The study was funded by research grants from Kaohsiung Chang Gung Memorial Hospital, CMRPG8B1121.
Authors\' contributions
=======================
K-Y Hung, Terry Chiou, and C-T Lee, study design, data analysis and manuscript writing; C-H Wu, K-T Hsu, Y-C Liao, C-N Chen, P-H Yang, and H-J Wan, clinical work and data collection. All authors read and approved the final manuscript.
{#F1}
######
Comparisons of baseline characteristics and body composition between non-low protein and low protein CKD patients in different age groups.
Elderly Non-elderly
------------------------------ -------------- ---------------- ---------------- ------------------
Age 70.2 ± 6.8 70.2 ± 7.3 50.7 ± 8.9^\*^ 46.8 ± 9.5^\*^
Male (n, %) 57 ( 72 % ) 15 ( 63 % ) 23 ( 54 % ) 7 ( 54 % )
DM( n, %) 21 ( 27 % ) 5 ( 21 % ) 9 ( 21 % ) 2 ( 15 % )
Systolic BP (mmHg) 132 ± 15 131 ± 18 127 ± 13 125 ± 16
Diastolic BP (mmHg) 66 ± 9 67 ± 9 71 ± 7^†^ 67.7 ± 8.0
Protein intake / IBW(g/kg) 1.01 ± 0.17 0.71 ± 0.06^‡^ 0.95 ± 0.13 0.71 ± 0.05^‡^
Energy intake / IBW(kcal/kg) 29.0 ± 4.2 23.3 ± 2.5^‡^ 27.9 ± 2.9 22.9 ± 2.4^‡^
BMI (kg/m^2^) 24.1 ± 3.1 25.4 ± 2.9 23.9 ± 3.9 23.4 ± 4.4
waist circumference (cm) 85.9 ± 9.0 84.7 ± 11.6 83.3 ± 12.1 83.6 ± 15.7
eGFR (mL/min/1.73m^2^) 25.7 ± 11.9 23.9 ± 11.8 24.5 ±10.2 19.5 ± 11.1
Serum albumin (g/dL) 4.2 ± 0.3 4.2 ± 0.3 4.3 ± 0.3 4.5 ± 0.3^†^
Hemoglobin level ( g/dL) 11.6 ± 1.9 11.4 ± 1.7 11.6 ± 2.1 10.4 ± 1.7
HbA1c ( % ) (Diabetes) 6.7 ± 0.6 6.7 ± 0.9 7.1 ± 1.5 6.0 ± 0.4
Total cholesterol (mg/dL) 168\. ± 25.6 170.4 ± 30.9 175.3 ± 33.2 147.9 ± 26.5^§†^
HDL (mg/dL) 57.3 ± 16.5 53.6 ± 14.8 54.0 ± 15.3 52.0 ± 11.6
LDL (mg/dL) 88.7 ± 21.5 92.6 ± 32.7 93.8 ± 29.4 78.1 ± 24.9
TG (mg/dL) 128.8 ± 82.3 96.6 ± 29.7 154.7 ± 118.4 116.0 ± 42.9
**Body composition**
Fat % 26.0 ± 6.0 29.7 ± 5.3^§^ 24.8 ± 5.9 24.1 ± 6.5^†^
Trunk fat% 13.4 ± 3.1 15.3 ± 2.7^§^ 12.8 ± 3.0 12.4 ± 3.4^†^
Muscle% 68.2 ± 5.9 64.6 ± 5.2^§^ 69.4 ± 5.9 70.2 ± 6.4^†^
Leg-muscle% 12.6 ± 1.1 11.8 ± 1.0^§^ 12.8 ± 1.1 12.8 ± 1.2^†^
Trunk muscle% 33.9 ± 2.9 32.3 ± 2.5^§^ 34.6 ± 2.9 35.1 ± 3.5^†^
ASM index 8.24 ± 0.99 8.17 ± 1.04 8.29 ± 1.28 8.13 ± 1.24
Data were expressed as mean ± standard deviation. eGFR, estimated glomerular filtration rate; HbA1c, glycosylated hemoglobin; HDL, high-density lipoprotein; LDL, low-density lipoprotein; TG, triglyceride; BMI, body mass index; ASM index, appendicular skeletal muscle mass index.
\* p \< 0.001 elderly vs. non-elderly. ‡ p \< 0.001 non-low protein vs. low protein group. † p \< 0.05 elderly vs. non-elderly. § p \< 0.05 non-low protein vs. low protein group.
######
Comparisons of characteristics and body compositions of non-elderly patients in different protein intake groups at baseline and 1-year follow-up.
Non-low protein group, n = 43 Low protein group, n = 13
--------------------------- ------------------------------- --------------------------- -------------- --------------
BMI (kg/m^2^) 23.9 ± 3.9 24.0 ±3.7 23.4 ± 4.4 23.3 ± 4.6
waist circumference (cm) 83.3 ± 12.1 81.6 ± 9.4 83.6 ± 15.7 85.1 ± 15.8
eGFR (mL/min/1.73m^2^) 24.5 ±10.2 21.9 ± 11.3^\*^ 19.5 ± 11.1 19.3 ± 14.0
Serum albumin (mg/dL) 4.3 ± 0.3 4.5 ± 0.4 4.5 ± 0.3 4.4 ± 0.3
Hemoglobin level ( g/dL) 11.6 ± 2.1 12.7 ± 5.8 10.4 ± 1.7 10.9 ± 2.3
HbA1c ( % ) (DM only) 7.1 ± 1.5 6.6 ±1.1 6.0 ± 0.4 6.0 ± 0.6
Total cholesterol (mg/dL) 175.3 ± 33.2 161.5 ± 47.9 147.9 ± 26.5 142.5 ± 28.0
HDL (mg/dL) 54.0 ± 15.3 53.6 ± 15.1 52.0 ± 11.6 46.2 ± 6.1
LDL (mg/dL) 93.8 ± 29.4 86.0 ± 33.9 78.1 ± 24.9 84.0 ± 30.7
TG (mg/dL) 154.7 ± 118.4 182.8 ± 241.7 116.0 ± 42.9 105.8 ± 35.8
**Body composition**
Fat % 24.8 ± 5.9 24.7 ± 5.9 24.1 ± 6.5 24.0 ± 7.2
Trunk fat % 12.8 ± 3.0 12.7 ± 3.0 12.4 ± 3.4 12.3 ± 3.7
Muscle % 69.4 ± 5.9 69.5 ± 5.9 70.2 ± 6.4 70.3 ± 7.2
Leg-muscle % 12.8 ± 1.1 12.8 ± 1.2 12.8 ± 1.2 12.9 ± 1.3
Trunk muscle % 34.6 ± 2.9 34.6 ± 2.8 35.1 ± 3.5 35.1 ± 3.7
ASM index 8.29 ± 1.28 8.54 ± 1.85 8.13 ± 1.24 8.08 ± 1.08
Data were expressed as mean ± standard deviation. ASM index, appendicular skeletal muscle mass index; BMI, body mass index.
\* p \< 0.05 compared with baseline.
######
Comparisons of characteristics and body compositions of elderly patients in different protein intake groups at baseline and 1-year follow-up.
Non-low protein group, n = 79 Low protein group, n = 24
--------------------------- ------------------------------- --------------------------- -------------- ------------------
BMI (kg/m^2^) 24.1 ± 3.1 23.9 ± 3.1^\*^ 25.4 ± 2.9 24.8 ± 2.8^\*^
waist circumference (cm) 85.9 ± 9.0 85.3 ± 9.0 84.7 ± 11.6 86.2 ± 10.5
eGFR (mL/min/1.73m^2^) 25.7 ± 11.9 24.5 ± 13.2^\*^ 23.9 ± 11.8 23.2 ± 13.6
Serum albumin (mg/dL) 4.2 ± 0.3 4.3 ± 0.4^\*^ 4.2 ± 0.3 4.3 ± 0.3^\*^
Hemoglobin level ( g/dL) 11.6 ± 1.9 11.3 ± 2.0^\*^ 11.4 ± 1.7 11.2 ± 1.5
HbA1c ( % ) (DM only) 6.7 ± 0.6 6.5 ± 0.7 6.7 ± 0.9 6.9 ± 0.9
Total cholesterol (mg/dL) 168\. ± 25.6 168.3 ± 32.0 170.4 ± 30.9 153.4 ± 28.8
HDL (mg/dL) 57.3 ± 16.5 56.5 ± 20.8 53.6 ± 14.8 53.8 ± 17.8
LDL (mg/dL) 88.7 ± 21.5 90.6 ± 25.6 92.6 ± 32.7 78.3 ± 26.3
TG (mg/dL) 128.8 ± 82.3 114.0 ± 61.4 96.6 ± 29.7 126.9 ± 57.0^\*^
**Body composition**
Fat % 26.0 ± 6.0 25.4 ± 6.6^\*^ 29.7 ± 5.3 28.6 ± 5.1^\*†^
Trunk fat % 13.4 ± 3.1 13.1 ± 3.4^\*^ 15.3 ± 2.7 14.7 ± 2.6^\*†^
Muscle % 68.2 ± 5.9 68.8 ± 6.5^\*^ 64.6 ± 5.2 65.7 ± 5.0^\*†^
Leg-muscle % 12.6 ± 1.1 12.7 ± 1.3 11.8 ± 1.0 12.0 ± 1.0^\*†^
Trunk muscle % 33.9 ± 2.9 34.2 ± 3.1^\*^ 32.3 ± 2.5 32.8 ± 2.3^\*^
ASM index 8.24 ± 0.99 8.23 ± 1.05 8.17 ± 1.04 8.14 ± 1.18
Data were expressed as mean ± standard deviation. ASM index, appendicular skeletal muscle mass index; BMI, body mass index.
\* p \< 0.05 compared with baseline.
† p \< 0.05 non-low protein vs. low protein group after 1-year follow-up.
[^1]: Competing Interests: The authors have declared that no competing interest exists.
| {
"pile_set_name": "PubMed Central"
} |
1. Background {#sec1}
=============
The toxicity of anticancer drugs to noncancer cells is an important barrier that limits the efficacy of anticancer drugs \[[@B1]\]. In addition, drug resistance of cancer cells due to mechanisms such as increased drug efflux, alteration or mutation of drug targets, alterations in DNA repair, and evasion of apoptosis \[[@B2]\] often limits the efficacy of anticancer drugs. The presence of a subpopulation of cancer stem cells (CSCs) or cancer stem-like cells (CS-LCs) associated with chemoresistance and tumor relapse has been also linked to poor response to chemotherapy in many cancers \[[@B3]\]. Novel therapeutic options that selectively target cancer cells, especially those with high resistance to anticancer drugs, with little or no toxicity to normal cells have been the focus of intensive research but the success has been limited. For instance, the success of targeted therapies that interfere with specific proteins involved in tumorigenesis rather than using broad base cancer treatments has been limited by the difficulty in identifying specific cancer biomarkers \[[@B4]\] and to the development of acquired drug resistance through mutations in targeted proteins or through the adaptation of alternate cancer cell survival strategies \[[@B5]\]. Drugs that more selectively target CSCs/CS-LCs have been identified but once again toxicity to normal cells limits the clinical application of these drugs. For instance, Salinomycin has been identified as a highly specific drug toward cancer stem cells \[[@B6]\] but its use in humans has been limited probably due to the considerable toxicity observed in mammals \[[@B7]\].
Tumorspheres are useful model for screening of drugs since they are enriched in cancer stem cells (CSCs) or cancer stem-like cells (CS-LCs) that are usually more resistant compared to non-CSCs/CS-LCs \[[@B8]\], and it is thought that the ability to form clonal spheres is a unique characteristic of CSCs \[[@B9], [@B10]\]. The ability to sustain proliferative signaling and divide in the absence of exogenous mitogenic stimulation leading to unregulated proliferation is considered one hallmark of cancer cells \[[@B11]\]. This has been demonstrated for glioma \[[@B12], [@B13]\], lung \[[@B14]\], and breast CSCs/CS-LCs \[[@B15]\] that can form spheres in serum-free media without exogenous mitogens. Lung tumorspheres (LTs) and mammospheres (MSs) obtained in the absence of any external mitogenic stimulation showed increased resistance to conventional anticancer drugs such as Paclitaxel (PX), hydroxyurea (HU), Colchicine (CX), and Obatoclax (OBT). We have also reported that adherent H460 lung and breast cancer cells that survive prolonged periods of serum starvation divide slowly and become highly resistant to PX, HU, CX, OBT, and the PI3 kinase inhibitors Wortmannin (WT) and LY294002 (LY) \[[@B16], [@B17]\]. LTs showed elevated expression of stemness-associated markers that may contribute to the multiresistant phenotype associated with CSCs/CS-LCs. On the other hand, the multiresistant phenotype of cells growing under PPSS is likely the result of extensive rewiring of signaling pathways rather than increased stemness \[[@B16]\]. These traits make cells growing under PPSS and tumorspheres useful complementary models to screen drugs able to overcome multidrug resistance as well as to identify the underlying mechanism(s). VP is a calcium channel blocker that has been shown to inhibit the activity of the MDR1 protein and has shown potential as a sensitizing agent to overcome the chemoresistance of CSCs/CS-LCs in a variety of cancers including lung \[[@B18]\], pancreatic \[[@B19]\], and breast \[[@B20]\] cancer cells. Sorafenib (SF) is a multikinase inhibitor that also inhibits the activity of the ABGC2 multidrug-resistant protein. However, combinatorial treatment using VP and SF has not been extensively characterized.
The aim of this study was to evaluate the effect of VP in combination with SF in lung cancer cells growing under PPSS as well as tumorspheres. We found that short term-exposure to VP + SF selectively and irreversibly decrease the viability, likely by activating necroptotic cell death, of cancer cells growing under PPSS or as tumorspheres but have little or negligible effect on noncancer cells or in cancer cells growing under RCCs.
2. Methods {#sec2}
==========
2.1. Chemicals and Reagents {#sec2.1}
---------------------------
### 2.1.1. Drugs {#sec2.1.1}
Verapamil (VP), z-VAD-FMK (zVAD), chloroquine (CQ), poly-HEMA (poly(2-hydroxyethyl methacrylate)), and MTT (thiazolyl blue tetrazolium bromide) were purchased from Sigma-Aldrich (St. Louis, MT). Sorafenib (SF) necrostatin 1 (Nec1), and 1-methyl-D-tryptophan (1-D-M-T) were purchased from VWR (Radnor, PA). Stock solutions of SF (10 mM), Nec1 (10 mM), and zVAD (10 mM) were in DMSO and stored in aliquots at −20°C. CQ was prepared as stock solution (10 mM) in distilled sterile water and filter sterilized and stored in aliquots at −20°C. VP (50 mM) was freshly prepared in distilled sterile water and filter sterilized. 1-D-M-T (20 mM stock solution) was prepared by dissolving in 0.1 N NaOH, and the pH was adjusted to 7.5 using hydrochloric acid, filter sterilized \[[@B21]\], and stored in aliquots at −20°C. Final dilutions were freshly prepared in culture media before use.
### 2.1.2. Cell Culture {#sec2.1.2}
The human lung epithelial cancer cell line NCI-H460 and the noncancerous cell line Beas-2B were obtained from American Type Culture Collection (Manassas, VA). Beas-2B cells are epithelial cells that were isolated from normal human bronchial epithelium obtained from the autopsy of noncancerous individuals ([http://www.atcc.org](http://www.atcc.org/)). For routine culture conditions (RCCs), cells were plated and propagated in complete media (CM) = RPMI 1640 (for NCI-H460) or DMEM/high glucose (for Beas-2B) supplemented with 5% FBS, L-glutamine, 100 U/ml penicillin, and 100 mg/ml streptomycin. Glutamine concentrations in RPMI-1640 and DMEM/high-glucose media were 2 or 4 mM, respectively. All cells were cultured in a 5% CO~2~ environment at 37°C. For cells growing under routine culture conditions (PPSS) or growing as floating tumorspheres, cells were maintained in serum-free media (same as CM but without FBS, see details below).
### 2.1.3. Generation of Lung Tumorspheres (LTs) {#sec2.1.3}
A detailed protocol for the generation of floating tumorspheres grown in the absence of any external mitogenic stimulation can be found in Yakisich et al. \[[@B15]\]. Briefly, H460 cells grown in CM (70--80% confluency) were cultured overnight in serum-free media (SFM, same as CM but without FBS). Then, cells were trypsinized and incubated in SFM for at least 14 days in poly-HEMA-coated plated to prevent attachment. For maintenance of LTs, the SFM was replaced every 3-4 days. LTs grown in SFM for 14--21 days were used for subsequent experiments.
### 2.1.4. Short-Term Antiproliferative Assay (MTT Assay and CCK Assay) {#sec2.1.4}
For routine culture conditions and adherent cultures (parental H460 and Beas-2B), cells were plated in 96-well cell-culture microplates (Costar, USA) at \~2000 cells per well and incubated overnight in CM. For cells growing under prolonged periods of serum starvation (PPSS), cells (\~500 cells/well) were plated in 96-well cell-culture microplates and incubated overnight in CM to allow them to adhere and then maintained in SFM for 7--12 days. Then, the cells were exposed to the appropriate concentration of drug or vehicle for 24--72 h. Cell viability for adherent cells was evaluated by the MTT assay. The absorbance of solubilized formazan was read at 570 nm using Gen 5 2.0 All-In-One microplate reader (Bio-TEK, Instruments Inc.). For floating LTs and MSs, cells growing in poly-HEMA plates were collected in 15 ml Falcon tubes, centrifuged at 700 rpm × 3 min, and resuspended in fresh SFM. In order to plate the same number of cells, this cell suspension was split in 1 ml aliquots. Vehicle or drugs were added to each aliquot and then 150 *μ*l cell suspension was loaded into each microwell (in a 96-well plate) and incubated for 72 h. For floating LTs, cell viability was evaluated by the CCK-8 assay (Dojindo Laboratories).
In all cases, the highest concentration of DMSO was used in the control and this concentration was maintained below 0.01% (*v*/*v*). This DMSO concentration did not show any significant antiproliferative effect on the cell lines or tumorspheres in a short-term assay.
### 2.1.5. Western Blotting {#sec2.1.5}
Preparation of cell lysates and Western blotting were performed as described previously \[[@B22]\]. Antibodies for PARP, cleaved PARP, caspase 3, caspase 9, RIP1, MLKL, Beclin, p62, and peroxidase-conjugated secondary antibody were purchased from Cell Signaling (Danvers, MA). Antibody for GAPDH was purchased from Santa Cruz Biotechnology (Dallas, TX). The blotting membranes were probed with 1 : 1000 diluted primary antibody and 1 : 4000 for the peroxidase-conjugated secondary antibody. Immune complexes were detected by chemiluminescence using SuperSignal™ West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific, Grand Island, NY) and photographed using myECL imager instrument (Thermo Fisher Scientific, Grand Island, NY).
### 2.1.6. Statistical Analysis {#sec2.1.6}
All pairwise multiple comparison procedures (ANOVA, Student-Newman-Keuls method) have been done using SigmaPlot (V. 11.0) software.
3. Results {#sec3}
==========
3.1. Short-Term Exposure to Verapamil in Combination with Sorafenib Inhibits the Viability of Highly Resistant Cancer Cell {#sec3.1}
--------------------------------------------------------------------------------------------------------------------------
We first investigated the ability of VP + SF to inhibit the viability of human lung H460 cancer cells growing under culture conditions that promote stemness and make cells highly resistant to anticancer agents: (1) cells growing under prolonged periods of serum starvation (PPSS) and (2) cells growing as floating tumorspheres (FTs). Cells growing under PPSS for 8 days were incubated for 24 hours with VP (100 *μ*M), SF (5 *μ*M), VP (50 *μ*M) + SF (2.5 *μ*M), or VP (100 *μ*M) + SF (5 *μ*M), and viability was measured by the MTT assay. [Figure 1(a)](#fig1){ref-type="fig"} shows that VP or SF alone or low concentration of VP (50 *μ*M) + SF (2.5 *μ*M) has no significant effect on cell viability but high concentration of VP (100 *μ*M) + SF (5 *μ*M) significantly decrease the viability of H460 cancer cells. A similar effect was observed when these drugs were tested in H460 LTs, and viability was measured by the CCK assay ([Figure 1(b)](#fig1){ref-type="fig"}). As it can be observed in Figure S1 available online at <https://doi.org/10.1155/2017/5987015>, in LTs treated for 24 h with DMSO alone in 96-well uncoated microplates, the cells are able to reattach. In contrast, VP (100 *μ*M) + SF (5 *μ*M) treated cells fail to reattach and lose integrity (Figure S1) indicating that this drug combination induces a rapid cell death that is in agreement with the massive decrease in cell viability measured by the MTT and CCK assays in PPSS and FTs, respectively.
3.2. Short-Term Exposure to Verapamil in Combination with Sorafenib Has Little Effect on the Viability of Cancer and Noncancer Cells Growing under Routine Culture Conditions {#sec3.2}
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
In order to evaluate the effect of VP, SF, and VP + SF on cancer cells (H460) and noncancer cells (Beas-2B) growing under RCCs, a culture condition in which cancer cells are relatively highly sensitive to anticancer drugs and have low expression levels of stemness-associated markers, cells growing under RCCs were incubated for 24 hours with VP (100 *μ*M), SF (5 *μ*M), VP (50 *μ*M) + SF (2.5 *μ*M), or VP (100 *μ*M) + SF (5 *μ*M), and viability was measured by the MTT assay. [Figure 2(a)](#fig2){ref-type="fig"} shows that both cancer and noncancer cells growing under RCCs are more resistant to VP + SF compared to cells growing under PPSS or as FTs (see [Figure 1](#fig1){ref-type="fig"}). In parallel, we tested the effect of 72 hours of exposure to VP or SF alone or in combination ([Figure 2(b)](#fig2){ref-type="fig"}).
3.3. Short-Term Exposure to Verapamil in Combination with Sorafenib Irreversibly Inhibits the Viability of Lung Cancer Cells Growing under Prolonged Periods of Serum Starvation (PPSS) {#sec3.3}
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
To evaluate if the effect of short-term exposure to VP + SF can irreversibly decrease the viability of cancer cells, we performed "recovery" experiments and compared to "continuous treatment" experiments as indicated in Figure S2. For "recovery" experiments, cells growing under RCCs and cells growing under PPSS for 8 days were treated with different concentrations of VP + SF (VP 100 *μ*M + SF 2.5 *μ*M; VP 50 *μ*M + SF 5 *μ*M; or VP 100 *μ*M + SF 5 *μ*M). After 24 h treatment, the drug was removed, cells were allowed to recover in drug-free media for 48 h, and viability was evaluated by the MTT assay. [Figure 3](#fig3){ref-type="fig"} (left panels) shows that cells treated for 24 h and then allowed to recover for 48 h in drug-free media cannot recover showing decreased viability compared to control cells. In parallel experiments ("continuous treatment" experiments), cells were treated with VP + SF for 72 hs ([Figure 3](#fig3){ref-type="fig"}, right panels). Overall, these results indicate that short exposure (24 h) to VP + SF is able to irreversibly induce cell death in cancer cells growing under PPSS but both cancer cells and noncancer cells (Beas-2B) can recover significantly from short-term exposure to VP + SF. In addition, we performed the same "recovery" experiments but allowing cells to recover for extended periods (up to 5 days for cells treated with VP100 + SF5 for 24 h) in drug-free media to ensure that cells growing under RCCs continue to recover and are not irreversibly damaged as PPSS cells. In these set of experiments, the MTT assay will not be reliable due to the high number of cells expected in control wells after \~6 days of culture. Instead, at the end of the experiments, cells were stained with Hema 3® Stain Set according to manufacturer\'s instructions (Fisher Diagnostics, Middletown, VA) and evaluated microscopically. [Figure 3(a)](#fig3){ref-type="fig"} shows that when cells growing under RCCs were treated for only 24 h with VP100 + SF5 and then allowed to recover for 5 days they were able to grow continuously although at lower density compared to control cells. Longer treatment (for 72 h) completely eliminated both Beas-2B and H460 cells. In contrast, in cells growing under PPSS for 9 days both, short (24 h) and long (72 h) treatments, completely eliminated H460 cells ([Figure 3(b)](#fig3){ref-type="fig"}). These results are consistent with the MTT data shown in [Figure 3](#fig3){ref-type="fig"}.
3.4. VP + SF Modulates the Expression of Key Proteins Involved in Apoptosis, Autophagy, and Necroptosis in a Cell Type-Dependent Manner {#sec3.4}
---------------------------------------------------------------------------------------------------------------------------------------
To gain insight into the mechanism by which VP + SF eliminates cancer cells, we evaluated the expression of key proteins involved in apoptosis (PARP, caspase 3, and caspase 9), autophagy (Beclin-1 and p62), and necroptosis (RIP1 and MLKL). Protein lysates were collected from floating and attached H460 cells grown under PPSS for 8 days that were exposed for 12 or 18 hs to VP 100 *μ*M + SF 5 *μ*M, and the expression of proteins was evaluated by Western blots. Control cells were treated with equivalent concentrations of vehicle (H~2~O + DMSO for 18 hs). For comparison, the expression of these protein markers was also evaluated in untreated cells growing under RCCs. [Figure 4](#fig4){ref-type="fig"} shows that VP100 + SF5 modulates the expression of these key proteins in cells growing under PPSS. For instance, the apoptosis markers (cleaved caspase 3 and cleaved PARP) were found to be elevated. VP100 + SF5 significantly decreased the expression of the autophagy markers p62 and Beclin-1. The necroptosis markers\' RIP1 levels were reduced significantly. Collection of protein lysates for Western blots was not performed at later times (e.g., 24 h) since microscopic observation showed extensive cell death (in agreement with the viability data presented in [Figure 1](#fig1){ref-type="fig"}) and loss of cellular integrity.
3.5. Necrostatin 1 (Nec1) Partially Prevented the Effects of VP + SF on Cell Viability {#sec3.5}
--------------------------------------------------------------------------------------
Due to the difficulty in monitoring the expression profiles of proteins at later times, we used pharmacological inhibitors of apoptosis (zVAD-FMK (zVAD)), necroptosis (necrostatin 1 (Nec1)), or autophagy (chloroquine (CQ)) to further elucidate the mechanism of cell death triggered by VP + SF. H460 cancer cells growing under PPSS for 8--10 days were incubated with VP 100 *μ*M + SF 5 *μ*M alone or coincubated with Nec1 (50 *μ*M) and zVAD (10 *μ*M). Nec1 and zVAD partially rescued the viability cells when incubated for 24 hours with VP 100 *μ*M + SF 5 *μ*M. This protective effect was observed in cells growing under PPSS ([Figure 5(a)](#fig5){ref-type="fig"}). In cells growing as FTs, Nec1 but not ZVAD partially prevented the effect of VP + SF on cell viability ([Figure 5(b)](#fig5){ref-type="fig"}). The autophagy inhibitor CQ had no protective effect on the decrease of cell viability induced by VP + SF in either cell growing under PPSS ([Figure 6](#fig6){ref-type="fig"}) or as FTs (Figure S4). In fact, in cells growing under PPSS, CQ significantly enhanced the effect of VP 100 *μ*M + SF 5 *μ*M ([Figure 6](#fig6){ref-type="fig"}). Since Nec1 is also an IDO inhibitor, the effect of 1-methyl-D-tryptophan (1 mM), a classical IDO inhibitor that does not affect necroptosis \[[@B23]\], was tested and showed no protective effect in cells growing under PPSS ([Figure 6](#fig6){ref-type="fig"}) or as FTs (data not shown).
4. Discussion {#sec4}
=============
Lung cancer is a leading cause of cancer-related deaths \[[@B24], [@B25]\], and resistance to chemotherapy is a major challenge to treat these tumors. Therefore, a drug or treatment that can selectively kill cancer cells with no harm to normal cells has been considered the magic bullet to treat these malignancies. In this study, we evaluated the anticancer effects of Verapamil in combination with Sorafenib (VP + SF) in lung cancer cells growing under three different culture conditions: routine culture conditions (RCCs), prolonged periods of serum starvation (PPSS), and cell growing as floating tumorspheres (FTs). FTs growing in absence of external mitogenic factors showed elevated resistance to conventional anticancer drugs such as PX, CX, and HU \[[@B14]\] which is a trait usually found in CSCs/CS-LCs \[[@B9]\]. Lung CSCs are known to be resistant to PX \[[@B26]\] and other conventional anticancer drugs such as Cisplatin, Doxorubicin, and Etoposide \[[@B27]\]. In the present study, we found that both VP and SF, even at high concentrations (100 *μ*M and 5 *μ*M, resp.) were almost ineffective when used as a single drug. When used in combination, only a high concentration of VP + SF (100 *μ*M + 10 *μ*M, resp.) showed a potent inhibitory effect on cell viability in cells growing under PPSS ([Figure 1(a)](#fig1){ref-type="fig"}) as well as in cells growing as floating tumorspheres ([Figure 1(b)](#fig1){ref-type="fig"}). This effect was observed within 24 hours of exposure. More importantly, (1) cancer cells and noncancer cells growing under RCCs were much more resistant to 24 hours exposure to VP (100 *μ*M) + SF (5 *μ*M) and (2) the effect of 24 h exposure to VP (100 *μ*M) + SF (5 *μ*M) on cell viability was found to be irreversible in cancer cells growing under PPSS. In contrast, in cancer cells and noncancer cells growing under RCCs, the effect was reversible since cells were able to recover once the VP + SF was removed from the media ([Figure 3](#fig3){ref-type="fig"}). This indicates that VP (100 *μ*M) + SF (5 *μ*M) triggers cell death only in cells growing under PPSS in an irreversible manner within 24 h. For technical reasons, the reversibility was tested only in adherent cells (RCCs and PPSS) where media can be easily replaced without the need of centrifugation steps that would be required for FTs.
In our study, we used a very high concentration of VP: 50--100 *μ*M. The average steady-state plasma levels measured for Verapamil was approximately 0.5 *μ*M \[[@B28]\]. However, Verapamil has been used *in vitro* as a classical inhibitor of MDR1 at 50 *μ*M \[[@B18], [@B29]\] and up to 200 *μ*M \[[@B19]\]. In humans, SF achieves drug levels of about 10 *μ*M \[[@B30]\] and this concentration is enough to inhibit ABCG2 \[[@B31]\]. It is important to mention that the mechanism by which elevated concentrations of VP + SF when used in combination decrease the viability of highly resistant cancer cells may be unrelated to their ability to downregulate the expression of MDR1 or ABCG2, respectively. The explanation behind this assumption is because we previously reported that cancer cells growing under PPSS respond aberrantly and sometimes paradoxically to a variety of pharmacological agents. For instance, in H460 cells growing under PPSS, we found that VP and SF (alone or in combination) induce rather than decrease the expression of MDR1 and ABCG2, respectively. Similarly, Obatoclax (a Bcl-2 inhibitor) downregulated the expression Bcl-2 in cells growing under RCCs but induced a paradoxical increase in the levels of this protein in cells growing under PPSS \[[@B16]\]. We attributed this response to a "rewiring" of signaling pathways as an adaptative response of cells to prolonged serum starvation, and it is likely that similar changes may occur in FTs that typically grow in serum-free conditions. Additional experiments behind the scope of this manuscript are needed to better understand the biology of these models of chemoresistance in order to elucidate the mechanism by which VP + SF exerts the potent effects reported in the present study. This new knowledge will be valuable for the identification or development of new toxic compounds with similar activities against highly resistant cancer cells. Regardless of the high concentration of VP used in our study that may limit its clinical use, our data provide significant novel insight into the biology of multiresistant cancer cells. First, we have identified a combination of drugs that does not need prolonged exposure to selectively and irreversibly eliminate chemoresistant cells (such as CSCs/CS-LCs) while having little or no effects in non-CSCs/CS-LCs as well as in noncancerous cells (e.g., Beas-2B). This finding can be exploited to develop chemotherapy regimens to irreversibly induce cell death in chemoresistant cancer cells by short-term treatment (24 h). Second, at the molecular level, we showed that VP + SF modulated the expression of key proteins involved in apoptosis, autophagy, and necroptosis ([Figure 4](#fig4){ref-type="fig"}) but no conclusive data could be obtained regarding the type of cell death triggered by this compound combination by analyzing the protein profile. While VP + SF increased the expression of cleaved PARP and cleaved caspase 3 in H460 cells, this treatment reduced the levels of P62 and Beclin-1 and RIP1 suggesting that autophagy may play a protective role. However, pharmacological evidence indicates that necroptosis and, to a lesser extent, apoptosis play a major role in cell death induced by VP + SF since (1) the inhibitory effect of VP + SF on cell viability could be partially prevented by Nec1 in both cells growing under PPSS and tumorspheres. zVAD, a pan caspase inhibitor, partially prevented the effect of VP + SF in cells growing under PPSS but had no effect in FTs (Figures [5(a)](#fig5){ref-type="fig"} and [5(b)](#fig5){ref-type="fig"}), (2) the autophagy inhibitor CQ did not prevented the effect of VP + SF neither in cells growing under PPSS nor in cells growing as FTS. Moreover, the significant enhancement on the decrease of cell viability when CQ was added to VP + SF ([Figure 6](#fig6){ref-type="fig"}) strongly suggests that induction of autophagy plays a protective role in VP + SF-induced cell death. Finally, the lack of effect of the IDO inhibitor 1-M-D-T indicates that Nec1 exerts its protective effect likely via inhibition of RIP1. Overall, our results reveal a complex crosstalk between apoptosis, necroptosis, and autophagy and support a model in which necroptosis has an important role in the cell death triggered by VP + SF. The differential protective effect of zVAD in cells growing under PPSS compared to FTs needs further evaluation. One possible explanation is that in our experimental system for cells growing under PPSS and cells growing as FTS, we use only serum-free media without any external mitogenic stimulation (e.g., EGF or bFGF) and the length of the lack of mitogenic factors rewires the cell death machinery to a different status depending on how long the cells have been adapted to SFM conditions. This hypothesis is supported by the differential protein levels of PARP and procaspase 9 observed in cells growing under PPSS compared with cells growing under RCCs ([Figure 4](#fig4){ref-type="fig"}). We have previously demonstrated that cells growing under PPSS rewire signaling pathways associated with multidrug resistance and respond aberrantly to inhibitors of multidrug resistance proteins such as MDR1 \[[@B16]\].
Our results clearly demonstrate that VP + SF by their ability to eliminate highly resistant cancer cells can be a leading combination to elucidate the underlying mechanism(s) that is necessary to selectively eliminate highly resistant cancer cells responsible for chemoresistance and tumor relapse.
5. Conclusion {#sec5}
=============
We report for the first time that cancer cells growing under PPSS or growing as FTs display a multidrug-resistant phenotype are, compared to cells growing under culture conditions, highly sensitive to a combination of VP + SF. We presented pharmacological evidence that short exposure to VP + SF irreversibly triggers necro/apoptotic cell death in cells growing under PPSS and necroptotic cell death in cells growing as FTs. More importantly, noncancer cells can almost fully recover from short-term exposure to VP + SF. Therefore, we have identified a novel therapeutic opportunity that can be considered an "Achilles\' heel" and can be targeted to selectively kill highly resistant cancer cells including CSCs/CS-LCs responsible for tumor resistance and tumor relapse while sparing noncancer cells.
Supplementary Material {#supplementary-material-sec}
======================
######
Figure S1. Representative images of LTs treated for 24 h with DMSO alone (control) or VP (100 µM) + SF (5 µM). Magnification: 20X (Bar = 200 µM). The inset shows an example of cell that reattach to the plate when treated with DMSO alone. In contrast, in the drug-treated plate, cells fail to reattach and clearly shows loss of cellular integrity. Figure S2. Simplified schema for the "Recovery" (A) and "Continuous" treatment (B) experiments performed for figure 3. For recovery experiments control or experimental cells (Exp.) were treated with DMSO or Verapamil+Sorafenib (VP+SF), respectively. After 24 h the media was changed (MC), incubated with drug-free media (Media lone) for 48 h and cell viability was measured at 72 h. For "Continuous" treatment" control or experimental cells (Exp.) were treated with DMSO or Verapamil+Sorafenib (VP+SF), respectively and cell viability was measured at 72 h. Figure S3a. Representative images of Beas-2B and H460 cells growing under RCCs and then treated for 72 h with DMSO alone (control, top pictures) or VP (100 µM) + SF (5 µM) (VP100+SF5) for 24 h (middle pictures) or 72 h (bottom pictures) followed by incubation in drug-free media for up to 5 days (for cells treated for 24h). Magnification: 20X. The results clearly shows that Beas-2B and H460 cells treated for 24 h with VP100+SF5 are able to recover while treatment for 72 h is toxic leaving only cellular debris. Figure S3b. Representative images of H460 cells growing under PPSS for 9 days and then treated for 72 h with DMSO alone (control, top picture, left) or VP (100 µM) + SF (5 µM) (VP100+SF5) for 24 h (middle picture, left) or 72 h (bottom picture, left) followed by incubation in drug-free media for up to 5 days (for cells treated for 24h). Magnification: 20X. The insets show examples of control cells (top picture, right) and cell debris (bottom picture, right) showing that treatment with VP100+SF5 for 24 h or 72 h irreversible eliminates H460 cells growing under PPSS. Figure S4. *CQ does not prevent VP+SF decrease on cell viability in FTs*. Cells growing as FTS for 14-16 days were incubated with VP (100 µM) + SF (5 µM) alone or in the presence of CQ for 24 h. Cell viability was measured by the CCK assay. Results (X±SD) are representative of two independent experiments performed in sextuplicates.
This study was supported by the grant CA173069 from the National Cancer Institute (NIH/NCI) and HL112630 (NIH/NHLBI).
RCCs:
: Routine culture conditions
PPSS:
: Prolonged period of serum starvation
FTs:
: Floating tumorspheres
VP:
: Verapamil
SF:
: Sorafenib
CQ:
: Chloroquine
zVAD:
: z-VAD-FMK
Nec1:
: Necrostatin 1
1-M-D-T:
: 1-Methyl-D-tryptophan.
Conflicts of Interest
=====================
The authors declare that they have no competing interests.
{#fig1}
{#fig2}
{#fig3}
{#fig4}
{#fig5}
{#fig6}
[^1]: Academic Editor: Andrzej Lange
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
The placental cells arising from the outer layer of the blastocyst, the trophoblast, differentiate along either the villous or extravillous pathways. Extravillous trophoblast (EVT), invade into the pregnant uterus (decidua) where they interact with maternal cells, including decidual natural killer (dNK) cells. These comprise ∼70% of the decidual immune cell population and are a distinct subset of NK cells. They display a large granular lymphocyte morphology and are CD56^bright^CD16^−^, as opposed to peripheral blood natural killer cells, where the predominant population is CD56^dim/−^CD16^bright^ ([@DEU017C21]). During early pregnancy, maternal uterine spiral arteries are remodelled from low-flow, high-resistance vessels into higher flow vessels with low-resistance. The extent of EVT invasion is critical for implantation and remodelling of the uterine spiral arteries ([@DEU017C32]), and EVT have been shown to play an active role in regulating the remodelling events ([@DEU017C2]; [@DEU017C19]; [@DEU017C14]). However, there is now considerable interest in the role that other cells, particularly the dNK cells, may have in regulating trophoblast invasion.
dNK cells produce a number of soluble factors such as cytokines, growth factors and pro-and anti-angiogenic proteins, in contrast to peripheral blood NK cells, which have a more cytotoxic role in defence ([@DEU017C39]). Trophoblast invasion and spiral artery remodelling are complex processes, with many interactions taking place between the various cell types in the decidual environment. The appropriate regulation of these processes is likely to be influenced by the levels of dNK cell-derived factors since they are located in close proximity to both invading EVT and remodelling vessels and several dNK-derived cytokines, chemokines and growth factors have been identified at the fetal--maternal interface ([@DEU017C13]; [@DEU017C25]). For example, interferon (IFN)-γ may modulate both chemokine expression and trophoblast activity to limit invasion ([@DEU017C29]).
In pregnancies complicated by pre-eclampsia and intrauterine growth restriction, shallow trophoblast invasion and insufficient spiral artery remodelling have been observed ([@DEU017C4]; [@DEU017C31]; [@DEU017C26]). Poor spiral artery remodelling is established in the first trimester of pregnancy. However, human studies are restricted by a lack of access to first trimester tissue with a known pregnancy outcome or a known stage of spiral artery remodelling. Doppler ultrasound scanning can be used to measure the resistance index (RI) of uterine artery blood flow, reflecting the level of remodelling in the spiral arteries, and therefore can be used as a proxy measure of the remodelling process. In this study, we have used this technique to identify pregnancies at the highest risk (21%) and at lowest risk (\<1%) of developing pre-eclampsia ([@DEU017C35]), with the highest and lowest evidence of first trimester poor spiral artery remodelling.
In this study, the factors produced by isolated dNK cells were examined and the role that two of these factors, angiogenin and endostatin, may play in modulation of trophoblast activity during pregnancy was investigated.
Materials and Methods {#s2}
=====================
Doppler ultrasound of uterine artery resistance and ethical approval {#s2a}
--------------------------------------------------------------------
Maternal uterine artery Doppler ultrasound scans were conducted on women attending a clinic for elective termination of pregnancy as previously described ([@DEU017C27]; [@DEU017C36]). Wandsworth Local Research Ethics Committee approval was in place for both the Doppler ultrasound before surgical termination and the use of first trimester tissue after the termination, and all women gave informed written consent. Terminations of pregnancy were carried out at 9--14 weeks of gestation. All were singleton pregnancies, with no known pre-existing medical conditions. High-RI cases were defined as those presenting with bilateral uterine diastolic notches and a mean RI above the 95th percentile, while normal-RI cases were defined as presenting with no diastolic notches and a mean RI below the 95th percentile. These two RI groups represent cases with the most (21%) and least (\<1%) likely chance of developing pre-eclampsia, had the pregnancy not been terminated ([@DEU017C27]).
Positive selection of dNK cells {#s2b}
-------------------------------
Products of conception were obtained immediately after surgical termination of pregnancy. NK cells were isolated from decidual tissue using positive selection with anti-CD56 antibody coated magnetic beads (Miltenyi Biotec, Surrey, UK) as previously described ([@DEU017C9]). Purity was 93.6 ± 1.3% (mean ± SEM, *n* = 19 patients).
Cell culture {#s2c}
------------
The well-characterized human EVT cell line, SGHPL-4, is derived from primary human first trimester EVT ([@DEU017C6]; [@DEU017C5]). SGHPL-4 cells were cultured in Hams F10 media supplemented with 10% (v/v) fetal bovine serum (FBS), containing [l]{.smallcaps}-glutamine (2 mmol/l), penicillin (100 IU/ml) and streptomycin (100 µg/ml). All cells were incubated with 95% air and 5% CO~2~ at 37°C in a humidified incubator.
The isolated CD56^+^ NK cells were centrifuged at 400*g* for 10 min at 22°C and cultured for 24 h at 6 × 10^6^cells/10 ml in RPMI 1640 medium (Invitrogen, Paisley, UK) with 10% FBS, containing 2.5 µg/ml amphotericin B (Sigma Aldrich, Dorset, UK), 2 mM [l]{.smallcaps}-glutamine, 50 µg/ml penicillin and 50 µg/ml streptomycin (Invitrogen), 50 ng/ml stem cell factor and 5 ng/ml IL-15 (Peprotech, London, UK) at 37°C in a 5% CO~2~ humidified incubator. There was no significant difference between the gestational ages of the patients in either of the two groups (*P* = 0.235, high-RI group: *n* = at least 18 per test, mean gestational age 74.8 ± 2.4, normal-RI group: *n* = at least 18 per test, mean gestational age 77.48 ± 1.87).
Multiplex array {#s2d}
---------------
Factors secreted by dNK cells were quantitatively analysed by bead-based multiplexing \[angiogenin, endostatin, placental growth factor (PLGF); R&D Systems, Abingdon, UK, all other factors; Invitrogen, Life Technologies Ltd\] according to manufacturer\'s protocols with detection on a Luminex 100 system (Luminex, Austin, TX, USA). Culture supernatants from dNK cells isolated from individual patients were examined (*n* = at least 18 normal- and high-RI samples per test). Supernatants were tested at three concentrations; concentrated 5-fold (Vivaspin columns, Sartorius Stedim UK Ltd, Surrey, UK), neat and diluted 3-fold in serum-free medium. Results were corrected according to the cellular protein concentration determined by Bradford assay (Bio-Rad, Hemel Hempstead, UK) of the pelleted cells after collection of the culture supernatant. In the case of a factor being undetected in \>15% of the culture supernatants, this factor was reported but not included in the analysis. Statistical comparisons were made between the patient groups for the remaining factors.
Motility assay {#s2e}
--------------
SGHPL-4 motility in response to endostatin and angiogenin was assessed as previously described using time-lapse microscopy ([@DEU017C12]). SGHPL-4 cells were seeded in Hams F10 media supplemented with 10% (v/v) FBS before overnight incubation in Hams F10 media supplemented with 0.5% (v/v) FBS. Recombinant human endostatin (Peprotech) was incubated with the SGHPL-4 cells in serum-free media alone or in the presence of 10 ng/ml epidermal growth factor (EGF) at concentrations of 50, 500 and 5000 ng/ml for 24 h (*n* = 4); angiogenin (R&D Systems) was incubated with the SGHPL-4 cells alone or in the presence of 10 ng/ml EGF at concentrations of 10, 100 and 1000 ng/ml for 24 h (*n* = 4). Cells were randomly chosen at the beginning of the experimental sequence and their movement was tracked manually using Image-J software (version 1.47d, National Institutes of Health, USA).
Invasion assay {#s2f}
--------------
Invasion of SGHPL-4 cells in response to recombinant endostatin and angiogenin was measured using a spheroid invasion assay as previously described ([@DEU017C45]). A volume of 100 µl of control medium or recombinant endostatin at 50, 500 and 500 ng/ml (*n* = 4) or angiogenin at 10, 100 and 1000 ng/ml (*n* = 4), with or without 10 ng/ml EGF, was added in serum-free media and spheroids were visualized after 24 h incubation using an Olympus 1X70 inverted microscope. Images were captured using a Hamamatsu C4742-95 digital camera. Invasion was measured as the average number and length of all invasive processes from each spheroid using Image-J software (version 1.47d).
Tube formation assay {#s2g}
--------------------
The ability of SGHPL-4 cells to form endothelial-like tube structures on Matrigel (BD, Oxford, UK) in serum-free media in the presence of angiogenin (10, 100 or 1000 ng/ml, *n* = 3) or endostatin (50, 500 or 5000 ng/ml, *n* = 5) was assessed using a µ-slide Angiogenesis Assay (Ibidi, Planegg, Germany) according to the manufacturers\' instructions.
Western blot analysis {#s2h}
---------------------
SGHPL-4 cells were serum starved for 24 h in Hams F10 containing 0.5% FBS (v/v) before incubation of cells with 100 ng/ml angiogenin (*n* = 4) or 500 ng/ml endostatin (*n* = 4) for 0, 5, 15, 30 or 60 min in Hams F10 containing 0% FBS (v/v). SGHPL-4 cells were then lysed in Radio-Immunoprecipitation Assay buffer containing a protease inhibitor cocktail of aprotinin (60 µg/ml), phenylmethylsulfonyl fluoride (1 mM) and sodium orthovanadate (1 mM). Western blotting was then performed as previously described ([@DEU017C12]) using rabbit anti-phospho-Akt^Ser473^ (1/2500 dilution, Cell Signalling Technology, Hertfordshire, UK), rabbit anti-phospho-ERK 1/2 (1/1000 dilution, Cell Signalling Technology), mouse anti-phospho-FAK (pY937, BD Biosciences, Oxford, UK) or mouse anti-α-tubulin (1/10 000 dilution, Abcam, Cambridge, UK). Western blots were scanned and the integrated density of each band determined using Image-J software (version 1.47d). Results are expressed as a ratio to the loading control within the same sample.
Statistical analysis {#s2i}
--------------------
Data were log transformed to stabilize variance and a Student\'s *t*-test was used to compare differences between patient groups using GraphPad Prism v5 (GraphPad Software, San Diego, CA, USA). A *P*-value of \<0.05 was considered to be statistically significant. Non-parametric Freidman one-way analysis of variance with Dunns *post hoc* test was used to compare trophoblast invasion, tube formation and alterations in phosphorylation of signalling proteins.
Results {#s3}
=======
Levels of angiogenic factors in normal-RI and high-RI dNK cell conditioned media {#s3a}
--------------------------------------------------------------------------------
dNK cell culture supernatants from individual patients were examined after 24 h of culture, by multiplex bead-based assays. Analysis of the various cytokines showed a significantly higher production of angiogenin (*P* \< 0.05; Fig. [1](#DEU017F1){ref-type="fig"}A), soluble interleukin-2 receptor (sIL-2R, *P* \< 0.01; Fig. [1](#DEU017F1){ref-type="fig"}B), endostatin (*P* \< 0.05; Fig. [1](#DEU017F1){ref-type="fig"}C) and PLGF (*P* \< 0.01; Fig. [1](#DEU017F1){ref-type="fig"}D) by high-RI dNK cells in comparison with normal-RI dNK cells. The following cytokines were detectable in \>85% of the samples assayed, however, were not significantly different between the high-RI and normal-RI dNK groups (*P* \> 0.05, data not shown): IL-1RA, monokine induced by gamma interferon (MIG), macrophage inflammatory protein 1α (MIP1α), MIP1β and regulated upon activation normal T-cell expressed and secreted (RANTES). Figure 1Comparison of factors secreted by normal-RI dNK cells and high-RI dNK cells. Measurement by multiplex assay of (**A**) Angiogenin (\* *P* \< 0.05) (**B**) sIL-2R (\*\**P* \< 0.01), (**C**) Endostatin (\**P* \< 0.05) and (**D**) PLGF (\*\**P* \< 0.01) in dNK cell culture supernatants incubated for 24 h. N = at least 18 normal and high-RI samples. Data points presented are from individual patients and include the mean ± SEM. RI, resistance index.
Several secreted factors were only detected in \<15% of the total samples, and these were: fibroblast growth factor-β, granulocyte colony stimulating factor, granulocyte and macrophage colony stimulating factor, IFN-α, IFN-γ, IL-1β, IL-2, IL-7, hepatocyte growth factor, IL-10, IL-12, tumour necrosis factor α and vascular endothelial growth factor (VEGF). The factors, EGF, eotaxin, IL-4, IL-5, IL-13 and IL-17 were not detectable in any culture supernatants. When regression analysis was carried out across gestation for angiogenin, sIL-2R, endostatin, PLGF, IL-1RA, MIG, MIP1α, MIP1β and RANTES, there was no relationship between the levels of factors detected and gestational age (in either high-RI or normal-RI groups, data not shown).
Effects of endostatin on trophoblast function {#s3b}
---------------------------------------------
To transform spiral arteries, trophoblast must perform the varied functions of invasion and migration into the decidua and the transformation into an endothelial-like phenotype, and these functions may be altered by secreted factors from dNK cells. We measured the effect of recombinant endostatin on known trophoblast functions using the extravillous-like trophoblast cell line, SGHPL-4. Recombinant endostatin at concentrations of 50, 500 and 5000 ng/ml did not significantly affect basal or EGF-induced motility of SGHPL-4 cells (Fig. [2](#DEU017F2){ref-type="fig"}A). Figure 2The effect of endostatin on basal and EGF-stimulated trophoblast function. Recombinant endostatin was added to SGHPL-4 cells at the concentrations listed. (**A**) SGHPL-4 cell basal and EGF-stimulated motility was measured over a 24 h period in the presence of recombinant human endostatin. Motility was not significantly different compared with the control, *n* = 4. (**B**) SGHPL-4 cells were cultured to form spheroids, embedded in fibrin gels and the length and (**C**) the number of invasive processes was measured. The average length and number of invasive process outgrowths of SGHPL-4 cell spheroids in the presence of 10 ng/ml EGF were decreased in response to 5000 ng/ml endostatin when compared with control culture media (\**P* \< 0.05, *n* = 4). (**D**) SGHPL-4 cells were cultured on Matrigel to induce endothelial-like tube formation, which was assessed by counting the number of branching points between tubes. The total number of branching points was decreased in response to 5000 ng/ml endostatin when compared with control culture media (\**P* \< 0.05, *n* = 5).
Using a fibrin gel invasion assay, we investigated both the number of SGHPL-4 cells invading from a 3D spheroid and the average length of the invasive process. EGF-induced invasion of SGHPL-4 cells was significantly decreased in the presence of 5000 ng/ml endostatin, as measured by both the average length of the invasive process (Fig. [2](#DEU017F2){ref-type="fig"}B, control mean length: 165.2 ± 12.6, 5000 ng/ml mean length: 132.4 ± 8.4, *P* \< 0.05) and number of cells invading (Fig. [2](#DEU017F2){ref-type="fig"}C, control mean number: 34.6 ± 1.8, 5000 ng/ml mean number: 24.6 ± 2.2, *P* \< 0.05).
The ability of SGHPL-4 to form endothelial-like tube structures on Matrigel was also examined in the presence of endostatin. Endostatin at a concentration of 5000 ng/ml significantly decreased the number of branching points between trophoblast tubes (Fig. [2](#DEU017F2){ref-type="fig"}D, control mean no. branching points 513.5 ± 70.2, 5000 ng/ml mean no. branching points 336.8 ± 40.8, *P* \< 0.05).
Effect of angiogenin on trophoblast function {#s3c}
--------------------------------------------
To examine the importance of increased angiogenin expression by dNK cells from high-RI patients, we measured the effect of angiogenin on the same trophoblast parameters. Recombinant angiogenin at concentrations of 10, 100 and 1000 ng/ml did not significantly affect basal or EGF-induced motility of SGHPL-4 cells (Fig. [3](#DEU017F3){ref-type="fig"}A). Recombinant angiogenin at concentrations of 10, 100 and 1000 ng/ml did not significantly affect the length of SGHPL-4 cell invasive processes into fibrin gels (Fig. [3](#DEU017F3){ref-type="fig"}B), however, at concentrations of 10 and 100 ng/ml, angiogenin did significantly inhibit the number of EGF-induced invasive outgrowths, while 1000 ng/ml did not (Fig. [3](#DEU017F3){ref-type="fig"}C, control mean number: 41.3 ± 5.5, 10 ng/ml mean number: 32.4 ± 3.5, 100 ng/ml mean number: 33.0 ± 4.9, *P* \< 0.05). At a concentration of 1000 ng/ml, recombinant angiogenin significantly increased the number of branching points between SGHPL-4 tube-like structures formed on Matrigel (Fig. [3](#DEU017F3){ref-type="fig"}D, control no. branching points: 160.3 ± 52.3, 1000 ng/ml no. branching points 224 ± 62, *P* \< 0.05). Figure 3The effect of angiogenin on basal and EGF-stimulated trophoblast function. Recombinant angiogenin was added to SGHPL4 cells at the concentrations listed. (**A**) SGHPL-4 cell basal and EGF-stimulated motility was measured over a 24-h period in the presence of recombinant human angiogenin. Motility was not significantly different compared with the control, *n* = 4. (**B**) SGHPL-4 cells were cultured to form spheroids, embedded in fibrin gels and the length and (**C**) number of invasive processes was measured. The average length of invasive process outgrowths of SGHPL-4 cell spheroids alone or in the presence of 10 ng/ml EGF was not altered in response to angiogenin when compared with control culture media. The average number of invasive process outgrowths of SGHPL-4 cell by spheroids in the presence of 10 ng/ml EGF was significantly decreased in response to 10 ng/ml and 100 ng/ml angiogenin when compared with control culture media (\**P* \< 0.05, \*\* *P* \< 0.01, *n* = 5). (**D**) SGHPL-4 cells were cultured on reduced-growth-factor Matrigel to induce endothelial-like tube formation, assessed by counting the number of branching points between tubes. The total number of branching points was increased in response to 1000 ng/ml angiogenin when compared with control culture media (\*\**P* \< 0.01, *n* = 3).
Signalling pathways affected by endostatin and angiogenin {#s3d}
---------------------------------------------------------
To identify a mechanism for the effects of endostatin and angiogenin on SGHPL-4 cells, we investigated signalling pathways known to be activated by these proteins and important in motility and invasion of trophoblast cells ([@DEU017C20]; [@DEU017C34]; [@DEU017C12]). SGHPL-4 cells were incubated for 0--60 min with recombinant 500 ng/ml endostatin or 1000 ng/ml angiogenin, concentrations previously found to influence trophoblast behaviour, and the signalling pathways activated were examined by western blot analysis. Endostatin led to a temporary de-phosphorylation of Akt^ser473^ at 15 and 30 min (Fig. [4](#DEU017F4){ref-type="fig"}A, 15 min: 2.95-fold, 30 min: 2.65-fold, *P* \< 0.05), however, phosphorylation of ERK 1/2 and FAK was not significantly altered at any time point. Angiogenin significantly decreased phosphorylation of Akt^ser473^ at 60 min (Fig. [4](#DEU017F4){ref-type="fig"}B, 1.9 ± 0.17-fold decrease, *P* \< 0.05), however, did not significantly affect ERK1/2 phosphorylation or phosphorylation of phopsho-FAK at any time point. Figure 4Signalling pathways affected in SGHPL-4 cells by recombinant human endostatin and recombinant human angiogenin. (**A**) SGHPL-4 cells were cultured in the presence of 500 ng/ml recombinant endostatin for 0 to 60 min. Cell lysates were collected, and equal amounts of total protein were subjected to western blot analysis for the phosphorylation of phosphorylated (p)-Akt, p-ERK 1/2 or p-FAK. Data presented are p-Akt relative to the α-tubulin loading control, mean ± SEM, \**P* \< 0.05. (**B**) SGHPL-4 cells were cultured in the presence of 1000 ng/ml recombinant angiogenin for 0 to 60 min. Cell lysates were collected, and equal amounts of total protein were subjected to western blot analysis for phosphorylated Akt, ERK 1/2 and FAK. Data presented are p-Akt relative to the α-tubulin loading control, mean ± SEM, \**P* \< 0.05, *n* = 4.
Discussion {#s4}
==========
The major maternal immune cell component of the maternal decidua is made up of dNK cells, but the role that they play in controlling trophoblast function is still unknown. Decidual NK cells were isolated from first trimester termination of pregnancy samples, and separated into two groups based on uterine artery Doppler resistance index, of which a high-RI is a proxy measure of the extent of poor spiral artery remodelling. We examined the profile of cytokine and angiogenic factor secretion in the culture media of dNK cells from high-RI and normal-RI pregnancies. Angiogenin and endostatin were produced at a higher level by dNK cells from high-RI pregnancies, and endostatin was found to inhibit trophoblast invasion and endothelial-like trophoblast tube formation, while angiogenin was found to inhibit trophoblast invasion but promote tube formation.
In our initial screen, no difference was found between normal-RI and high-RI dNK cell secretion of a number of factors. With the exception of IL-1RA, these have all previously been confirmed as produced by dNK cells ([@DEU017C44]). The angiogenic factors and cytokines secreted by dNK cells at the maternal--fetal interface are thought to have crucial roles in controlling trophoblast functions prior to the remodelling of spiral arteries. For example, dNK are capable of chemoattraction of EVT via IL-8 and interferon induced-protein 10 (IP-10; [@DEU017C13]). Decidual NK secreted cytokines can also alter the differentiation of EVT into the invasive phenotype ([@DEU017C17]) and support the development of EVT into 'endothelial-like' structures through secreted factors including VEGF ([@DEU017C16]).
Four factors were found at significantly different concentrations in the culture media of our two groups of dNK cells: sIL-2R, PLGF, endostatin and angiogenin. The concentration of soluble IL-2R was increased in the culture supernatant of dNK cells from high-RI pregnancies. Secretion of this cytokine is often preceded by increased cell-surface expression of IL-2R ([@DEU017C42]), which can occur due to increased IL-2 and IL-15 signalling, two cytokines present at the maternal--fetal interface ([@DEU017C43]; [@DEU017C40]). This may suggest an elevated activation status of dNK cells from high-RI pregnancies, which would also indicate altered proliferation and survival ([@DEU017C22]). It is therefore an important future study to investigate the activation status of dNK cell-associated IL-2R in the normal- and high-RI populations, in combination with the receptor repertoire of dNK cells in each group.
The angiogenic factor PLGF was also secreted at increased levels by dNK cells from high-RI pregnancies. PLGF at the maternal--fetal interface has been shown to have both pro- and anti-angiogenic effects via the receptors it shares with VEGF, VEGFR1 and the fms-like tyrosine kinase-1 (flt-1) receptor ([@DEU017C10]). Altered PLGF expression has been associated with pre-eclampsia ([@DEU017C1]). PLGF promotes trophoblast proliferation ([@DEU017C3]) and survival ([@DEU017C7]), however, the role of dNK cell secreted PLGF may be difficult to determine as the interaction between PLGF, VEGF and flt-1 may be more important in the function of PLGF as opposed to the concentration of PLGF alone ([@DEU017C10]).
The functions of endostatin and angiogenin are unknown in the decidua. Endostatin is the C-terminal cleavage product of Collagen XVIII, an extracellular matrix protein associated with the basement membrane. The endostatin present in the dNK culture medium may therefore be a result of dNK produced proteases cleaving collagen produced by other cells in our cultures, or any bound extracellular matrix still present after dNK isolation, which could indicate higher protease levels in the high-RI dNK group. Endostatin is an anti-angiogenic factor in cancer, and most likely works through several modes of action including direct interactions with the basement membrane, inhibition of MMPs and interaction with endothelial cell receptors ([@DEU017C18]). Endostatin has been shown to be expressed in the decidua ([@DEU017C33]) and cultured endometrial stromal cells ([@DEU017C30]). Endostatin has been demonstrated to reduce invasion of the trophoblast cell line SGHPL-5 as well as the outgrowth of EVT from placental villous explants ([@DEU017C34]). In the present study, we determined that endostatin reduced EGF-stimulated trophoblast invasion and reduced the basal capacity of SGHPL-4 cells to form endothelial-like tubes, and may function in SGHPL-4 cells by inhibition of phosphorylation of Akt^ser473^. However, it is likely that endostatin is also able to function by alternative signalling, for example endostatin can interact with α~v~ integrins to alter cell migration and survival ([@DEU017C38]), and expression of the same integrin receptors on trophoblast are key to their invasion ([@DEU017C17]) and the adoption of an endothelial-like phenotype such as the endovascular trophoblast ([@DEU017C47]).
The expression of angiogenin has been demonstrated by dNK cells ([@DEU017C8]), trophoblast ([@DEU017C37]) and decidual stromal cells ([@DEU017C23]). It is a pro-angiogenic secreted ribonuclease that binds to an unknown receptor leading to endocytosis and pro-angiogenic effects in the cell nuclei, including proliferation ([@DEU017C28]). Despite the known role of angiogenin in vessel remodelling, its actions during pregnancy on spiral arteries or trophoblast are unknown. In our invasion assay, angiogenin inhibited, rather than promoted, EGF-stimulated SGHPL-4 invasion. However, the formation of endothelial-like trophoblast tubes was promoted by angiogenin, indicating that differentiation into endovascular trophoblast would be promoted by this factor. We examined three signalling pathways known to be involved in angiogenin or SGHPL-4 signal transduction, including phosphorylation of Akt^ser473^, ERK 1/2 and FAK ([@DEU017C11]; [@DEU017C24]; [@DEU017C46]) and found that Akt^ser473^ phosphorylation was time-dependently decreased by angiogenin. It is possible that the increased levels of angiogenin in the high-RI dNK group may contribute during pregnancy towards preventing vessel breakdown in the initial stages of spiral artery remodelling and maintaining vessel stability rather than direct effects on trophoblast.
The role of dNK cell-secreted factors during the first trimester of pregnancy is now recognized to be important in signalling not only to invasive trophoblast but also to decidual spiral arteries and during remodelling of the decidua ([@DEU017C41]; [@DEU017C9]). Dissecting the role of dNK cells is complex because of the number of cytokines and angiogenic factors produced by these cells, some of which can have opposing effects, as evidenced by the inhibition of tube formation by endostatin and promotion by angiogenin. The overall outcome in the decidua will depend on the balance of these interactions, and also the concentration of these factors; the concentrations of these secreted factors in the decidua are unknown, and we have therefore used a range of concentrations commonly used in the literature to examine both angiogenin and endostatin. However, it is important to highlight that the interpretation of these results may be altered in the context of the correct concentration of angiogenin and endostatin, as it is possible that local concentrations between cells in the decidua may be within a different range to those described here. Future functional co-culture studies may be able to tell us more about the specific effects of these dNK cell-derived factors and the implications of these effects in normal and pre-eclamptic pregnancies. Our use of Doppler ultrasound to differentiate between pregnancies most and least likely to display poor spiral artery remodelling will further aid the discovery of altered signalling in pathological pregnancies.
Authors\' roles {#s5}
===============
R.F., A.E.W., A.P.J., G.S.W and J.E.C. conceived all of the experiments. A.E.W., R.F., J.E.C., S.G. and S.S.G. carried out all of the experiments. The manuscript was prepared by A.E.W., R.F. and J.E.C. and all authors critically revised the manuscript and approved the final version.
Funding {#s6}
=======
This study was supported by the Wellcome Trust (project reference 091550). R.F. was a recipient of a PhD studentship from the Division of Biomedical Sciences, St. George\'s, University of London. Funding to pay the Open Access publication charges for this article was provided by The Wellcome Trust.
Conflict of interest {#s7}
====================
No conflicts of interest are declared.
The authors acknowledge Prof. Baskaran Thilaganathan and the staff of the Fetal Medicine Unit at St. George\'s Hospital for their assistance with sample collection.
[^1]: The authors consider that the first two authors should be regarded as joint First Authors.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction
===============
According to the WHO \[[@B1-molecules-16-02726]\], about 450 million people in the entire world have suffered mental, neurological, or behavioral problems at some time in their life. Extensive research on plants and their derivatives has taken place in recent years that could provide some new alternative treatments and therapeutic uses for diseases of the central nervous system (CNS).
Epilepsy is the term used for a group of disorders characterized by recurrent spontaneous seizures that apparently result from complex processes involving several neurotransmitter systems such the glutamatergic, cholinergic, and gabaergic system. Actual estimations of the prevalence rate for epilepsy are 1--2% of the world population. Although a considerable number of classic and more modern anticonvulsant drugs are available for the pharmacological treatment of epilepsy patients worldwide, seizures remain refractory in more than 20% of the cases. In addition, all current antiepileptic drugs, which belong to several quite different chemical classes such as hydantoins, deoxybarbiturates, succinimides, benzodiazepines, iminostilbenes and carboxylic acids, have been obtained through chemical synthesis \[[@B2-molecules-16-02726]\].
Several species of aromatic plants are used medicinally because of their volatile oils or chemical components. In particular, some of them possess certain CNS properties, including antiepileptic action and have been traditionally used for a long time in folk medicine. Recent studies on essential oils and their main components have attracted the attention of many scientists and encouraged them to screen any of these natural products to study their chemical and pharmacological aspects that might potentially may lead to lead to the development of new anticonvulsant-like compounds with advantages over current therapeutic drugs \[[@B3-molecules-16-02726]\].
1.1. Chemistry of Essential Oils and Main Chemical Constituents
---------------------------------------------------------------
Aromatic plants are at present widely studied for their large therapeutic potential and benefits. These benefits depend largely on essential oils which, in general terms, occur in many herbs. The essential oils of the plant are the essence of their fragrance. They are called essential oils, ethereal oils, or volatile oils because they evaporate quickly when exposed to the air at ordinary temperatures. In general, the essential oils consist of chemical mixtures involving several tens to hundreds of different types of molecules. Only a few have a high percentage of a single component. These chemical constituents are divided into two broad classes: terpenes and phenylpropanoids. However, many volatile oils consist largely of monoterpenes, which are a group of terpenes having ten carbon atoms in the carbon skeleton and, therefore, are composed by two isoprene units \[[@B4-molecules-16-02726]\].
Essential oils are distilled from different parts of the plants including flowers, stem-bark, seeds, leaves, roots, and the whole herbs. They are used to give flavor to foods and drinks and as fragrances in the food and cosmetics industries, where numerous herbal plant and spice ingredients are components in the manufacture of skin creams, lip balms, shampoos, soaps and perfumes.
2. Methodology
==============
The present study was carried out based on the literature review of plants and their essential oils with anticonvulsant activity. All the information about 30 species with anticonvulsant activity is given in [Table 1](#molecules-16-02726-t001){ref-type="table"}. The list of plants is organized by family and botanical name, parts used and pharmacological activity, as described in the literature. Compounds isolated and references are also provided. The scientific names of the plants were based on W3Tropicos Database (<http://mobot.mobot.org/W3T/Search/vast.html>), and the abbreviations of author names are according to Brummitt and Powell \[[@B5-molecules-16-02726]\].
The plant species presented here were selected based on the effects shown by their essential oils in specific animal models used for evaluation of anticonvulsant activity and/or by complementary studies, aimed at elucidating the mechanism(s) of action of the oils or individual components.
The essential oils or the main constituents were deemed to display anticonvulsant activity when they had shown effects in one or more different seizure model, including the maximal electroshock (MES) model, the pentylenetetrazole seizures model (PTZ), the pilocarpine model and the prolonged PTZ-kindling model.
Some scientific publications that were excluded from this study because it was not possible to access their full text or because their abstracts were in a language different from English, included the following species, citing their psychopharmacological activities: *Apium graveolens* Cham., *Aralia continentalis* Kitag., *Asarum heterotropoides* F. Schmidt*, Asarum himalaicum* Hook. F. & Thomson ex Klotzsch*, Asarum ichangense* C.Y. Cheng & C. S. Yang, *Cumimum cyminum* L, *Eugenia uniflora* L*, Gardenia augusta* (L) Merr., *Gardenia jasminoides* J. Ellis, *Ligusticum sinense* Oliv., *Ocimum basilicum* L*. Oplopanax elatus, Radix bupleuri* and *Salvia sclarea.*
3. Results and Discussion
=========================
The plant diversity with confirmed activities in the central nervous system is dominated by higher plants, mainly by dicotyledons. In this review 30 species belonging to 13 families and 23 genera have been reported to possess anti-seizure activity.
The families in decreasing order of predominance of species with activity are Myrtaceae and Lamiaceae with five species each; Apiaceae with four species; Asteraceae and Poaceae with three species each and Araceae and Lauraceae with two species each. Six families, corresponding to 46 % of the total, are represented by only one species. Some of these species have other biological activities and are used for different purposes, like *Egletes viscosa* (L.) Less \[[@B6-molecules-16-02726]\], also mentioned for their antimicrobial activity.
The predominance of higher plants used for medicinal purposes confirms the results obtained in other ethno-medicinal surveys reported by Agra \[[@B7-molecules-16-02726],[@B8-molecules-16-02726]\]. This has also been documented by authors in different countries around the world such as Brazil \[[@B9-molecules-16-02726]\], Saudi Arabia \[[@B10-molecules-16-02726]\], Bolivia \[[@B11-molecules-16-02726]\] or Italy \[[@B12-molecules-16-02726]\], *inter alia*.
Several aromatic species had been employed since ancient times for their medicinal properties and also as aromatic agents and to give flavor to foods. The pharmacological uses of the plants are mainly attributed to their essential oils having a great variety of pharmacological activities such as prevention and treatment of cancer, against cardiovascular diseases and diabetes. They are also used as sources of gastro-protective, anti-inflammatory, antioxidant, and antibacterial agents. These varied effects are probably due to the high structural diversity of the essential oil constituents. The study of each individual chemical component is critical to understanding its mechanism of pharmacological action and toxicity, including potentially beneficial clinical effects on human health.
However, according to Lahlou \[[@B13-molecules-16-02726]\], the diversity of biological activities presented by the same essential oil has also stimulated discordance between researchers. Many reasons have been proposed for this variability, for example: (a) all the factors that have influence in the chemical composition; (b) the plant's state of maturation; and (c) the chemotypic difference, among others.
It is accepted that a refined assessment of the chemical composition of tested essential oils/constituents should be performed using GC/MS to perform a quantitative analysis, which would provides additional information of their contents and, consequently, confirmation of their therapeutic effects.
Thirty anticonvulsant chemical constituents of essential oils were mentioned. Most of these compounds are monoterpenes or phenylpropanoids ([Figure 1](#molecules-16-02726-f001){ref-type="fig"}). They are effective in several experimental models of seizure. These constituents must contribute to the anticonvulsant activity of bioactive essential oils, as presented in [Table 1](#molecules-16-02726-t001){ref-type="table"}. This pharmacological activity may be due to action of a major component or the effect of various bioactive components found in essential oil.
molecules-16-02726-t001_Table 1
######
Species and respective essential oils that showed anticonvulsant activity organized by botanical family, botanical name, part used, activity as described in the literature, compound isolated and references.
FAMILY *Species* PART USED ACTIVITY OF ESSENTIAL OILS (as described in the literature) MAIN COMPOUNDS ISOLATED/REFS.
----------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------
**APIACEAE**
***Cuminum cyminum* Linn. (syn. *Cuminum odorum* Salisb)** Fruits The oil showed protection against pentylenetetrazole-induced epileptic activity by increasing the duration, decreasing the amplitude of after hyperpolarization potential (AHP) following the action potential, the peak of action potential, and inhibition of the firing rate. \[[@B14-molecules-16-02726]\]
***Ferula gumosa* Boiss.** Fruits The essential oil protected mice against pentylenetetrazole-induced tonic seizures. The protective dose produced neurotoxicity. Moreover, this dose was too close to the LD~50~ of the essential oil. The anticonvulsant and toxic effects of the essential oil may be related to the compounds pinene and α-thujene respectively. Pinene andα-thujene \[[@B15-molecules-16-02726]\]
***Heracleum crenatifolium*** Fruits The essential oil protected mice against maximal electroshock (MES)-induced seizures. Octyl acetate and octanol \[[@B16-molecules-16-02726]\]
***Pimpinella anisum* L.** Fruits Inhibitor of tonic convulsions induced by high doses of pentylenetetrazole and electroshock trans-corneal. \[[@B17-molecules-16-02726]\]
**ARACEAE**
***Acorus calamus* L.** Not available Anticonvulsant action against experimental electroshock. Neither essential oil nor diphenylhydantoin were effective in modifying convulsions produced by metrazole. \[[@B18-molecules-16-02726]\]
***Acorus gramineus* Aiton** Rhizomes Inhibited the specific bindings of a use-dependent NMDA receptor-ion. Neuroprotective effects on cultured cortical neurons through the blockade of NMDA receptor activity. \[[@B19-molecules-16-02726]\]
Rhizomes Anticonvulsant effects, both *Acorus gramineus* and α-asarone can enhance the reactivity and convulsive threshold of immature rats to electric stimulation. α-asarone \[[@B20-molecules-16-02726]\]
Rhizomes Pre-inhalation of the oil markedly delayed the appearance of pentylene-tetrazole-induced convulsion. Furthermore, inhalation impressively inhibited the activity of gamma-aminobutyric acid (GABA) *trans*-aminase, a degrading enzyme for GABA as the inhalation period was lengthened. The GABA level was significantly increased and glutamate content was significantly decreased in mouse brain by pre-inhalation of the essential oil. \[[@B21-molecules-16-02726]\]
***Acorus tatarinowii* Schott.** Rhizomes The volatile oil (1.25 g/kg) decreased the convulsive rate significantly in the maximal electroshock model. But failed to prevent seizures in the dose range tested, although prolonged seizure latency and decreased mortality were found at a dose of 1.25 g/kg. The oil can prevent convulsions as well as convulsion-related GABAergic neuron damage in the brain in the prolonged pentylenetetrazol kindling model. \[[@B22-molecules-16-02726]\]
**ASTERACEAE**
***Artemisia annua* L.** Fresh leaves The essential oil (AEO) obtained by hidrodestilation increased the latency time to convulsions induced by picrotoxin and pilocarpine but prevented the onset of pentylenotetrazol and strychnine induced seizures. \[[@B23-molecules-16-02726]\]
***Artemisia dracunculus* L.** Aerial parts The oil exerted dose and time-dependent antiseizure activity reported in both maximal electroshock and pentylenetetrazole models. *trans*-Anethole, α-*trans*-ocimene, limonene, α-pinene, cymene, eugenol, β-pinene, α-terpinolene, bornyl acetate, and bicyclogermacrene \[[@B24-molecules-16-02726]\]
***Egletes viscosa* (L.) Less.** Inflorescences Activity against convulsion induced by PTZ in mice. \[[@B6-molecules-16-02726]\]
**EUPHORBIACEAE**
***Croton zehntneri*** **Pax & K. Hoffm.** Not available Elevation of the threshold for starting a minimal convulsions induced by pentylenetetrazol. \[[@B25-molecules-16-02726]\]
**FABACEAE-MIMOSOIDEAE**
***Tetrapleura tetraptera*** **(Schumach. & Thonn.) Taub.** Fruits Inhibitor of convulsions induced by PTZ and electroshocking in mice of both sex. \[[@B26-molecules-16-02726]\]
Fresh fruits The fresh oil given intraperitoneally offers some protection against leptazol-induced convulsions. A dose of 0.4 ml of the oil per mouse protected 78% of the animals when administered 30 min prior to leptazol \[[@B27-molecules-16-02726]\]
**LAMIACEAE**
***Aeollanthus suaveolens* Mart. ex Spreng.** Not available Inhibitor effect of convulsions induced by pentylenetetrazol and maximal electroshock in mice Linalool\[[@B28-molecules-16-02726]\]
Leaves The anticonvulsant properties of γ-decanolactone was observed in mice. The neurochemical essay with linalool in cortical membranes of rats showed a dose-dependent, not competitive of *binding* of \[^3^H\] MK 801 -- an antagonist of receptor NMDA. Also observed the effects of linalool on glutamatergic system in the rat cerebral cortex and that linalool modifies the nicotinic receptor -- ion channel kinetics at the mouse neuromuscular junction. γ-Decanolactone and linalool \[[@B29-molecules-16-02726],[@B30-molecules-16-02726],[@B31-molecules-16-02726],[@B32-molecules-16-02726]\]
Not available The oil has an inhibitory effect of linalool on glutamate binding in rat cortex was observed. Linalool \[[@B30-molecules-16-02726]\]
Not available Sedative properties on mice and has dose-dependent marked effects on the central nervous system, including hypnotic, anticonvulsant and hypothermic activity γ-Decanolactone \[[@B29-molecules-16-02726]\]
Not available An inhibitory effect of linalool on the acetylcholine (ACh) release and on the channel open time in the mouse neuromuscular junction. Linalool \[[@B31-molecules-16-02726]\]
***Hyptis suaveolens* (L.) Poit.** Leaves Anti-convulsive in tests induced by pentylenetetrazol and electroshock in rats of both sex \[[@B26-molecules-16-02726]\]
***Lavandula stoechas* L.** Not available Used as inhalator showed anti-convulsive activity similar to the above. Moreover, was verified a higher level of latency and a reduction of level of the severity of convulsions. Complementary test suggest this activity maybe is related with the blocking of canals of calcium. The inhaling lavender oil vapor blocked pentylene-tetrazole- and nicotine-induced convulsion and electroshock convulsion in mice. \[[@B33-molecules-16-02726]\]
***Ocimum gratissimum* L.** Not available Essential oil obtained in Spring was able to protect animals against tonic seizures induced by electroshock (MES, 50 mA, 0.11 s). Eugenol \[[@B34-molecules-16-02726]\]
***Ocimum basilicum*** Aerial part When tested in mice, the essential oil, higher doses, produced significantly increased in a dose-dependent manner the latency of convulsion and percent of animals exhibiting clonic seizures. Likewise, it reduced lethality in response to different convulsive stimulus used in this study. \[[@B35-molecules-16-02726]\]
***Ocimum basilicum*** Leaves Essential oil increased the latency for development of convulsions in pentylenetetrazol and PIC tests. For pentylenetetrazol, the effects of EO were reversed by flumazenil. EO did not interfered with the convulsions induced by strychnine. 1.8-Cineole, linalool, and geraniol were the main components, comprising 92.9% of the oil. \[[@B36-molecules-16-02726]\]
Not available Essential oil blocked the clonic seizures induced by pentylenetetrazole, picrotoxin and strychnine \[[@B37-molecules-16-02726]\]
**LAURACEAE**
***Laurus nobilis* L.** Leaves Anticonvulsant activity was observed against pentylenetetrazole- and maximal electroshock-induced seizures. At anticonvulsant doses, the essential oil produced sedation and motor impairment. Methyleugenol, eugenol and pinene \[[@B38-molecules-16-02726]\]
***Myristica fragrans*** Seed Nutmeg oil was found to possess significant anticonvulsant activity against electroshock-induced hind limb tonic extension. It exhibited dose dependent anticonvulsant activity against pentylenetetrazole-induced tonic seizures. It delayed the onset of hind limb tonic extensor jerks induced by strychnine. Also it was anticonvulsant at lower doses, whereas weak proconvulsant at a higher dose against pentylenetetrazole and bicuculline induced clonic seizures. \[[@B39-molecules-16-02726]\]
**MYRTACEAE**
***Eucalyptus citriodora* Hook** Leaves Not available Citronellal, citronellol, and citronellyl acetate \[[@B40-molecules-16-02726]\]
***Eucalyptus urophylla*** Leaves The oil increased the number of mice protected against pentylenetetrazole-induced death \[[@B41-molecules-16-02726]\]
***Eucalyptus camaldulensis* var. camaldulensis Dehn.** Leaves Not available *p*-Cymene, spathulenol, cryptone, thymol, and linalool \[[@B42-molecules-16-02726]\]
***Syzygium aromaticum* (L.) Merr. & L.M. Perry** *=Eugenia caryophyllata* Thunb. Flowers Inhibition of tonic convulsions induced by electroshock in rats \[[@B43-molecules-16-02726]\]
***Psidium persoonii* McVaugh** ***=*** *Psidium guianense* **Pers.** Leaves The doses of 100, 200, and 400 mg/kg, by via oral reduced as dose-dependent the severity of convulsions induced by pentylenetetrazole. Moreover, induced the depressor effect of spontaneous movement. \[[@B44-molecules-16-02726],[@B45-molecules-16-02726],[@B46-molecules-16-02726]\]
**POACEAE**
***Cymbopogon winterianus*** **Jowitt** Leaves Anticonvulsant activity was observed in pentylenetetrazole, pilocarpine, and strychnine tests. Anticonvulsant effect was blocked by flumazenil in pentylenetetrazole model. The EO showed presence of geraniol, citronellal, and citronellol as the main compounds. \[[@B47-molecules-16-02726]\], \[[@B48-molecules-16-02726]\]
***Cymbopogon citratus*** **(DC)Stapf** Leaves Anticonvulsant activity was observed in pentylenetetrazole, pilocarpine, strychnine, and maximal electroshock tests. \[[@B47-molecules-16-02726]\], \[[@B49-molecules-16-02726]\]
***Cymbopogon proximus*** Plant Administration of the oil to mice before induction of convulsions with electroshock, resulted in complete protection. There was partial protection in pentylenetetrazole, picrotoxin and strychnine tests. Piperitone, elemol, and eudesmol. \[[@B50-molecules-16-02726]\]
**RANUNCULACEAE**
***Nigella sativa*** **L.** Not available The oil was the most effective in preventing pentylenetetrazole-induced seizures relative to valproate; also showed significantly decreased oxidative injury in the mouse brain tissue in comparison with the pentylenetetrazole-kindling group Thymoquinone \[[@B51-molecules-16-02726]\]
**RUTACEAE**
***Citrus aurantium* L.** Peel from fruits EO from peel increased the latency period of tonic seizures in pentylenetetrazol and maximal electroshock models. Effect was not dose-dependent \[[@B52-molecules-16-02726]\]
**VALERIANACEAE**
***Nardostachys jatamansi*** **(D. Don) DC.** Not available A ketonic principle, jatamansone, isolated from oil of *N. jatamansi*, is more effective than quinidine and essential oil of jatamansi in suppression of ectopic ventricular activity in unanesthetized dogs produced by 2-stage coronary ligation, equal to quinidine in combating auricular flutter induced by injury stimulation, and more effective than Na diphenyl-hydantoin and essential oil of jatamansi in maximal electroshock seizures. Jatamansone \[[@B53-molecules-16-02726]\]
**VERBENACEAE**
***Lippia alba* (Mill.) N.E. Brown.** Not available The anticonvulsive effects of the essential oils (EOs) from three chemotypes of *Lippia alba* was observed. The animals were treated with citral (100 mg/kg, i.p.), α-myrcene or limonene (200 mg/kg, i.p.), EOs chem. constituents, presented significant increases in the latency of convulsion and percentage of survival. The association of EOs with diazepam significantly potentiated their effects, suggesting a similar mechanism of action. Citral β-myrceneLimonene \[[@B54-molecules-16-02726]\]
{#molecules-16-02726-f001}
Considering the thirty species selected in this study, anticonvulsant effect was observed in twenty four of them using the PTZ model. The prevention of seizures induced by PTZ in laboratory animals is the principal protocol used to characterize a potential anticonvulsant drug. The PTZ test represents a valid model for human generalized myoclonic seizures and also by the absence of seizures. On other hand, seventeen species showed anticonvulsant effect by blocking tonic convulsions induced by MES. This test is considered to be a predictor of anti-epileptic drug activity against generalized tonic-clonic seizures.
The more frequent studies involved substances tested via an intraperitoneal route. However, essential oils of *Acorus gramineus* Aiton and *Lavandula stoechas* L. only showed anticonvulsant activity when administered by inhalation. This pharmacological result is very interesting as it represents a non invasive route. In a recent review, Edris \[[@B55-molecules-16-02726]\] also presented some comments about the effects of substances administered by inhalation. Additionally, have been suggestions that the inhibitory effect on the central nervous system occurs via gamma-aminobutyric acid (GABA)-ergic neuromodulation system.
The anticonvulsant effect of the main chemical constituents of essential oils ([Figure 1](#molecules-16-02726-f001){ref-type="fig"}) have also been studied. The natural abundance of these compounds allowed systematic studies on their pharmacological properties. Most studies report antiseizure activity of these compounds in animal models of convulsion. The results showed that common essential oil constituents such as terpinen-4-ol \[[@B56-molecules-16-02726],[@B57-molecules-16-02726]\], citral, β-myrcene, limonene \[[@B53-molecules-16-02726],[@B58-molecules-16-02726]\], safranal \[[@B59-molecules-16-02726],[@B60-molecules-16-02726]\], linalool \[[@B27-molecules-16-02726],[@B61-molecules-16-02726]\], γ-decanolactone \[[@B28-molecules-16-02726],[@B62-molecules-16-02726]\], α-terpineol \[[@B63-molecules-16-02726]\], (-)-isopulegol \[[@B64-molecules-16-02726]\], citronellol \[[@B65-molecules-16-02726]\], thymoquinone \[[@B66-molecules-16-02726]\], α,β-epoxycarvone \[[@B67-molecules-16-02726]\], (*S*)-(+)-carvone \[[@B68-molecules-16-02726]\], eugenol, methyleugenol, isoeugenol, estragole, safrole \[[@B69-molecules-16-02726]\], *trans*-anethole \[[@B70-molecules-16-02726]\], (-)-borneol and carvacrol \[[@B58-molecules-16-02726]\] are all bioactive in experimental models. Interesting, (*R*)-(-)-carvone had no anticonvulsant effect. The study showed that the chiral center at carbon 4 in the carvone molecule is important in the interaction with the receptor. The molecule with isopropenyl group in *S* configuration at carbon 4 is clearly capable of reducing the convulsive effect of PTZ and PIC in terms of onset time. A comparative analysis of the constituents cannot however be done due to the different experimental conditions and variability of methods and routes of administration used in the evaluation of these compounds.
Anticonvulsant activity was also identified in some sesquiterpenes such as valeranone \[[@B52-molecules-16-02726]\] and β-eudesmol \[[@B71-molecules-16-02726]\]. Others compounds found in essential oils also showed this effect. For example, some phthalides such as butylidenephthalide, ligustilide, butylphthalide, neocnidilide, and senkyunolide \[[@B72-molecules-16-02726]\]. α-Asarone, a phenylpropanoid, also presented effective anticonvulsant activity \[[@B19-molecules-16-02726]\].
The present study reports a limited number of plant species based on a set of papers which the molecular mechanisms of anticonvulsant action of the volatile oils/constituents was possible to investigate. Of these, two species are prominent, *Acorus gramineus* Aiton and *Aeollanthus suaveolens* Mart. ex Spreng, belonging to the Araceae and Lamiaceae families, respectively.
The activity of *Acorus gramineus* was suggested by the essential oil having anticonvulsant effect through the blockade of NMDA receptor. On other hand, the psychopharmacological effect of *Aeollanthus suaveolens* was attributed to the monoterpene compound linalool, reported to be a major component of essential oils in several aromatic plants, including this species. Recent findings provide evidence that the anticonvulsant effects of monoterpenes could be modulated by Ach mechanisms. Moreover, other studies suggest a direct interaction with the NMDA receptor complex \[[@B31-molecules-16-02726]\].
4. Conclusions
==============
In conclusion, the thirty aromatic species, listed in the present paper and represented by their essential oils/chemical constituents, appear to be promissory as sources of anticonvulsant agents. Thirty anticonvulsant chemical constituents of essential oils also were related. Nevertheless, the data show that most chemical classes of compounds found in essential oils are effective in convulsion models. They must act by different mechanisms of action that merit further investigation. The future outlook for the development of new antiepileptic drugs derived from essential oils is therefore positive.
This research was supported by Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and Fundação de Apoio a Pesquisa e Inovação Tecnológica do Estado de Sergipe (FAPITEC).
*Sample Availability*: Not available.
| {
"pile_set_name": "PubMed Central"
} |
{
"pile_set_name": "PubMed Central"
} |
|
INTRODUCTION {#sec1-1}
============
Indoor air pollution is one of the environmental risk factor affecting mainly the rural population of developing countries. Comparative risk assessment by WHO attributes 1.6 million premature deaths annually to indoor smoke in developing countries.\[[@CIT1]\] The range of effects due to exposure to particulate matter due to combustion of biofuels is broad, affecting the respiratory and cardiovascular systems and extending to children and adults and to a number of large, susceptible groups within the general population.\[[@CIT2]\]
The mucociliary escalator is the primary defense mechanism against inhaled particulate matter. Mucociliary clearance comprises of the cephalad movement of mucus caused by the cilia lining the conducting airways until it can be swallowed or expectorated thereby protecting the human upper and lower airways from deleterious effects of inhaled pollutants, allergens, and pathogens.\[[@CIT3]\] The nasal mucociliary clearance (NMC) system transports the mucus layer that covers the nasal epithelium toward the nasopharynx by ciliary beating at a frequency of 7--16 Hz at body temperature and is controlled by certain physiological, anatomic, and biochemical variables.\[[@CIT4]\] Physiological factors such as age, sex, posture, sleep, exercise, temperature (\<10 °C and \>45°C) also influence the duration of NMC. When disruption of NMC occurs, respiratory secretions accumulate and impair pulmonary function, reduce lung defenses, and increase the risk for infection.\[[@CIT5]\] Tobacco smoke and environmental pollution due to combustion of biomass fuel are suspected to have a depressant effect on NMC and may lead to the development of various respiratory diseases. This also depends on factors such as pollutant concentration and the duration of exposure.\[[@CIT6]\] The underlying pathophysiology is stasis of sinonasal secretions due to ineffective sinonasal mucociliary clearance followed by subsequent bacterial overgrowth, frank infection, and/or inflammation. Nasal mucociliary clearance, the mirror image of bronchial mucociliary clearance is thus a biomarker of nasal mucosal function.\[[@CIT7]\]
Limited information is available regarding the effect of biomass fuel smoke on NMC in our country. Moreover, understanding the effect of biomass fuel smoke on NMC may advance our understanding of pathogenesis of chronic effects of such exposures. The mucociliary clearance of tracheobronchial tree can be assessed using bronchial spray mixed with radioactive compound and following with gamma camera, but it is a costly and cumbersome procedure and is not suited for field studies. Hence, this simple noninvasive method was chosen for this cross-sectional study to evaluate the mucociliary clearance in apparently healthy women dwelling in a periurban area using biomass fuel and clean fuel for cooking.
MATERIALS AND METHODS {#sec1-2}
=====================
This cross-sectional study was conducted in 30 apparently healthy biomass fuel users (life time users of wood, dung cake, crop residues) and 30 clean fuel users (life time liquefied petroleum gas users). Women with history of cooking exposure to either biomass fuel alone or clean fuel alone for a minimum period of 2 years were included in the study. These female subjects of age group ranging from 18 to 45 years were randomly selected from a periurban area in Chennai. Informed written consent was obtained from all the study subjects.
Details regarding cooking fuel, duration of exposure were collected using a pretested validated household questionnaire. Multiple fuel users (*n*=18), smokers/passive smokers (*n*=15), and tobacco chewers/snuff users (*n*=7), were excluded. A complete ear, nose, and throat examination was also performed to rule out diseases (sinusitis, nasal polyps, allergic rhinitis, and deviated nasal septum), which are known to affect the mucociliary clearance. Thereby, women with history of (h/o) deviated nasal septum/nasal polyp (*n*=2), allergic disorders (*n*=3), h/o intake of any medications particularly antihistaminics (*n*=2), respiratory or nasal symptoms within the preceding 2 weeks (*n*=2) were also excluded. Moreover, self-reported diabetes (*n*=3), and other factors such as primary ciliary dyskinesia, bronchiectasis, valvular heart diseases, bleeding diatheses, h/o exposure to formaldehyde, ammonia, phenols were also excluded from the study. Ten women refused to participate in the study in spite of their eligibility. Thus, 122 women were contacted in order to reach the desired sample size of 60 study subjects.
The nasal mucociliary clearance was studied using the saccharine method of Anderson *et al*.\[[@CIT8]\] A 1 mm particle of saccharine was placed on the floor of the nose, just behind the anterior end of the inferior turbinate and the test was carried out in sitting position with head flexed about 10° to avoid particle falling backwards and the time required by the subject to perceive the sweet taste was noted. The test was carried out on both nostrils with an interval of half an hour. The time of mucociliary clearance of each nostril was noted separately. Nasal mucociliary clearance time is the average time of the mucosal clearance of the two nostrils. The subjects were advised to avoid nasal manipulation, sniff, cough, inhale or exhale forcefully during the test, and were simply told to report any change in taste. Subjects were blinded about the nature of particle. (The subjects were informed that some harmless edible particle will be placed in the nostril and they were not informed about its nature.) A single examiner performed the test in all subjects to avoid inter observer variability. Peak Expiratory Flow Rate (PEFR) was recorded using the Wright's Peak flow meter. At least three readings were taken and the maximal value was noted down.
Access template was used for data entry and the analysis was performed using *R* statistical software version 2.8.1. Saccharin Transit Time (STT) and PEFR are expressed in terms of mean and standard deviation. Comparisons between the groups were analyzed by *t*-test and ANOVA and the significance was taken at 0.05 level.
RESULTS {#sec1-3}
=======
This study has compared NMC and PEFR between 30 clean fuel using women and 30 biomass fuel using women. Both STT and PEFR were considered as the outcome variables. The descriptive characteristics of the study population is given in [Table 1](#T0001){ref-type="table"}. NMC time was significantly (*P* = 0.007) prolonged in biomass fuel users (765.8 ± 378.16 s) in comparison to clean fuel users (545.4 ± 215.55 s) [Figure 1](#F0001){ref-type="fig"}. PEFR was significantly (*P* = 0.002) reduced in biomass fuel users (319.3 ± 67.21 l/min) when compared with LPG users (371.7 ± 59.49 l/min) \[[Figure 2](#F0002){ref-type="fig"}\]. In addition, both STT and PEFR were also compared across the several subcategories as shown in [Table 1](#T0001){ref-type="table"}. The prolongation of STT increased with increasing years of exposure (*r* = 0.47). NMC time was significantly prolonged in illiterates (*P* = 0.014), and the PEFR was also significantly reduced in this group (*P* = 0.000). Women from lower socioeconomic status (total family income \<Rs. 25 000 per annum) and lower literacy status (not able to read and write), women dwelling in kutcha houses, older women (40--50 years), undernourished women (BMI\<18), and women cooking for \>15 years had prolonged STT and reduced PEFR.
######
Comparison of STT and PEFR among different age, BMI, demographic and other sources of particulate matter categories of the study group
Study variables N STT (s) *P* value PEFR (l/min) *P* value
----------------------------- ------------------- ---- ------------------------------------------------- ----------- ------------------------------------------------- -----------
Age (20--29) 28 625.6 ± 352.00 0.586 363.9 ± 60.45 0.134
(30--39) 17 638.6 ± 237.99 334.1 ± 68.65
(40--50) 15 731 ± 364.09 324 ± 76.51
Income ≥25000 54 647.3 ± 339.8 0.556 351.5 ± 66.6[\*](#T000F1){ref-type="table-fn"} 0.041
\<25000 6 730.5 ± 115.07 291.7 ± 63.7
BMI ≥18 50 651.9 ± 323.93 0.844 348.2 ± 71.08 0.497
\<18 10 674.3 ± 345.98 332 ± 52.66
House type Pucca 20 586.6 ± 189.29 0.247 367.5 ± 69.2 0.077
Kutcha 40 690.1 ± 372.08 334.5 ± 65.87
Kitchen type Indoor 44 601.1 ± 258.61 0.029 350.5 ± 64.62 0.35
Outdoor 16 805.6 ± 436.39 331.9 ± 77.822
Ventilation type Well ventilated 52 647.3 ± 316.6 0.616 354.6 ± 66.05 0.007
Poorly ventilated 8 709.8 ± 393.48 286.2 ± 53.16
Education Literate 44 594.4 ± 295.9[\*](#T000F1){ref-type="table-fn"} 0.014 365.2 ± 64.75[\*](#T000F1){ref-type="table-fn"} 0.000
Illiterate 16 824 ± 350.1 291.2 ± 44.85
Family size (no. of persons 1--4 37 604.2 ± 268.55 0.12 352.2 ± 70.71 0.341
at home) 5--8 23 738.4 ± 391.46 334.8 ± 64.09
Incense usage Never 4 725.8 ± 650.28 0.582 347.5 ± 65 0.257
Irregular 39 623.3 ± 234.71 335.4 ± 70.74
Regular 17 713.4 ± 414.12 368.2 ± 60.75
Mosquito coil usage Never 14 605.9 ±218.89 0.129 365±72.93 0.431
Irregular 27 747.6 ± 334.03 335.6 ± 62.47
Regular 19 561.6 ± 355 345.3± 73.06
Data expressed as Mean ± SD
*P* \< 0.05; STT: Saccharin transit time; PEFR: Peak expiratory flow rate
{#F0001}
{#F0002}
DISCUSSION {#sec1-4}
==========
This cross-sectional study has evaluated both NMC and PEFR in nonsmoking women using biomass fuel and compared it with those women using clean fuel for cooking. PEFR, a lung function parameter and NMC were significantly altered in women using biomass fuel. The significantly prolonged STT in biomass fuel users has highlighted the deleterious effects of biomass fuel usage on respiratory system. Tobacco smoke is known to cause depression of NMC.\[[@CIT9]\] Very few studies have been conducted on effects of biomass smoke on mucociliary clearance. The prolongation of STT in these nonsmoking study subjects could be probably due to mucus abnormality and ciliary malfunction. Earlier studies also indicate that chronic exposure to biomass fuel smoke could result in structural changes of the respiratory mucosa including epithelial sloughing, intracellular edema, and mitochondrial swelling.\[[@CIT10]\] The prolongation of STT increased with increasing years of exposure. The increase in STT could also be due to the additive effect of aging. As this is a preliminary study, additional statistical analysis to assess the contribution of other risk factors such as age could not be performed due to small sample size. The STT in clean fuel users was similar to normal mucociliary clearance time as reported by other studies such as Golhar and Arora from Chandigarh in which STT was 6.2 min and Golhar from Nagpur in which it was 7.2 min.\[[@CIT11][@CIT12]\]
This study has also evaluated PEFR, a lung function parameter. Biomass users had reduced PEFR compared to women using LPG. The reduction in lung function due to exposure to environmental pollution can be due to several inflammatory processes. For example, upregulation of P-selectin expression in platelets following activation plays an important proinflammatory role in mediating interactions among neutrophils, platelets, and the vascular endothelium.\[[@CIT13]\] Studies conducted from rural India has reported increased leukocyte aggregates and increased CD11/CD18 expression on PMN and CD62P expression on platelets in women exposed to biomass fuel.\[[@CIT14]\] In addition, compromised lung is prone for repeated respiratory viral infections leading to desquamation of epithelial cells of the lung, microvascular dilation, edema, and an inflammatory cell infiltrate. The lung damage caused by exposure to particulate matter emitted by combustion of biomass fuel will predispose the respiratory tract to bacterial infection by interfering with mucociliary clearance and by reducing the bacterial killing by alveolar macrophages.\[[@CIT15]\] Thus, the significant reduction in PEFR in biomass fuel users could be attributed to the toxic effects of the components of smoke, free radicals,\[[@CIT16][@CIT17]\] acute neutrophilic airway inflammation.\[[@CIT18]\] A decline in pulmonary functions was also observed in our earlier study conducted in biomass fuel users in Tamil Nadu.\[[@CIT19]\]
This study has highlighted the effects of biomass fuel exposure in apparently healthy adult women in whom STT has been used as an indicator for assessing the functional status of NMC. The early effects of biomass on lung function indicates the need for intervention in this vulnerable category as chronic exposure can lead to development of Chronic Obstructive Pulmonary Disease (COPD), asthma and lung cancer, increasing the morbidity and mortality in women. Large scale studies and experimental studies are required in this area of research to establish the pathogenesis of the lung damage caused by exposure to biomass fuel. STT can be used as an economical biomarker in large scale epidemiological studies in developing countries where resources for research are limited.
In conclusion, this preliminary study has shown that the NMC time is prolonged in women using biomass fuel. PEFR is reduced significantly in biomass fuel using women. This study has shown the effects of indoor air pollution on respiratory defense mechanism. This study highlights the importance of this simple non invasive, inexpensive, screening test that can be used as an early indicator of respiratory damage caused by exposure to air pollutants. In addition, dissemination of the findings of this study to the general public and the health officials regarding the rising burden of indoor air pollution and its respiratory health effects will pave way in creating awareness among the young and middle aged women about the implications of indoor air pollution so that proper intervention strategies can be implemented in order to reduce the morbidity and mortality.
**Source of Support:** Nil
**Conflict of Interest:** None declared.
| {
"pile_set_name": "PubMed Central"
} |
Background
==========
In China, economic transition, urbanization, industrialization and an aging population have quickly increased the incidence and prevalence of coronary heart disease (CHD) in the past decades \[[@B1]\]. CHD has been ranked among the top three causes of death in China \[[@B2]\]. Anxiety and depression are common psychological problems associated with a diagnosis of CHD \[[@B3]-[@B5]\]. Importantly, depression and anxiety have been linked with the morbidity and mortality of CHD \[[@B6]\]. Therefore, valid and reliable screening for clinically significant anxiety and/or depression is paramount in this clinical group.
The Hospital Anxiety and Depression Scale (HADS) \[[@B7]\] is a widely used, self-administered questionnaire specifically developed to detect anxiety and depression states in hospital and medical out-patient clinic settings. It is composed of two 7-item scales, one for anxiety and one for depression.
The original English version has been translated into and validated in many languages, including Chinese \[[@B8]-[@B12]\]. The Chinese version of HADS is a popular instrument for assessing psychological distress in clinical studies in China \[[@B5],[@B12]-[@B14]\]. A number of studies have validated the Chinese version of this questionnaire both in Hong Kong (HK) and China \[[@B8],[@B12],[@B15]\]. Leung and colleagues \[[@B8]\] evaluated the psychometric properties of the Chinese version of HADS in 100 medical students. The results indicated factorial inconsistency with the English language version with a three-factor solution emerging. Despite this anomaly, the authors concluded the instrument was a valid Chinese translation. A more recent study of the Chinese version of the HADS \[[@B12]\] across a broad clinical range of in-patients again found three underlying factors to the instrument. These were interpreted as depression and two distinct factors of psychic anxiety and psychomotor agitation \[[@B12]\]. However, the utility of the instrument is based on the theoretical assumption of an underlying bi-dimensional (anxiety and depression) factor structure. Moreover, a underlying tri-dimensional structure of the HADS may have significant implications for both scoring and case detection accuracy \[[@B16],[@B17]\].
A recent review \[[@B16]\] of the HADS has suggested that the instrument may in reality have an underlying tri-dimensional factor structure in CHD patients and other clinical groups. Recent investigations of the psychometric properties of the HADS in CHD patients support the notion that the instrument essentially comprises three, rather than two sub-scales \[[@B18]-[@B20]\]. One study \[[@B20]\] of Cantonese-speaking Chinese CHD patients in Hong Kong has also furnished compelling evidence for the tri-dimensionality of the HADS. Hong Kong Chinese invariably speak Cantonese whereas in the Xi\'an province of China Mandarin is spoken. Due to the pictorial nature of Chinese writing, both Cantonese and Mandarin-speaking Chinese would be able to read the Chinese version of the HADS.
In Europe, the HADS has been applied extensively in the studies of patients with CHD as an index of both outcome and the effect of therapeutic intervention \[[@B21]-[@B23]\]. There have also been reports that the Chinese version of HADS may have some utility as a screening and assessment tool in patients with CHD \[[@B5]\]. However, the factorial structure of the Chinese version of the HADS has not been established in this clinical group with consequent implications for screening and case detection utility and accuracy \[[@B16]\]. The present study was designed to examine the underlying factor structure of the Chinese version of the HADS in a mainland Mandarin-speaking population of patients admitted to hospital with CHD.
Methods
=======
Design
------
The study used a cross-sectional design. To address the research question confirmatory factor analysis methods were used on a pooled HADS data set from mainland Mandarin-speaking patients admitted to hospital with CHD.
Statistical analysis
--------------------
The factor structure of the HADS was determined using confirmatory factor analysis using Mplus version 3 \[[@B24]\]. The weighted least-square with mean and variance correction estimator (WLSMV) was used to evaluate model fit as this estimation method can be both used reliably with ordered categorical level data, and be used dependably with modest samples sizes. Seven models developed from HADS validity and psychometric studies were tested \[[@B7],[@B17],[@B25]-[@B28]\]. The characteristics of each model tested and item-factor loading characteristics are shown in Table [1](#T1){ref-type="table"}. Confirmatory factor analysis represents a powerful statistical technique used to determine whether the number of factors and pattern of item-factor loadings is consistent with what would be expected by a priori theory. This represents a significant methodological advance over the more commonly used exploratory factor analysis where no prior assumptions of structure are explicitly made. Confirmatory factor analysis is a special case of structural equation modelling and is statistically and methodologically distinct from exploratory factor analysis. Though strictly speaking exploratory factor analysis should be used to determine the original factor structure of an instrument, and confirmatory factor analysis used to determine how well an a priori-defined factor structure fits data, it is common in the literature to see exploratory factor analysis used to investigate the factor structure of an instrument that has previously been investigated using exploratory factor analysis. Consequently, exploratory factor analysis is often found to be used where confirmatory factor analysis would be more desirable and more appropriate, this situation being widely observed in many previous studies of the factorial structure of the HADS.
######
Characteristics of each factor model tested
**Model** **Number of factors** **Clinical population** ***n*** **Factor extraction method**^**\#**^
--------------------------- ----------------------- ------------------------- ------------------- --------------------------------------
Zigmond and Snaith (1983) 2 Medical 100 No factor analysis
Moorey et al. (1991) 2 Cancer 568 PCA
Dunbar et al. (2000) 3 Non-clinical 2,547^+^ CFA
Friedman et al. (2001)\* 3 Depressed 2,669 PCA
Razavi et al. (1990) 1 Cancer 210 PCA
Caci et al. (2003)\*\* 3 Non-clinical 195 CFA
**FLI1**\*\* **FLI2** **FLI3**
Zigmond and Snaith (1983) 1,3,5,7,9,11,13 2,4,6,8,10,12,14 \-\-\-\-\-\-\-\--
Moorey et al. (1991) 1,3,5,9,11,13 2,4,6,7,8,10,12,14 \-\-\-\-\-\-\-\--
Dunbar et al. (2000) 1,5,7,11 2,4,6,7,8,10,12,14
Friedman et al. (2001)\* 1,7,11 2,4,6,8,10,12,14 3,5,9,13
Razavi et al. (1990) All items \-\-\-\-\-\-\-\-- \-\-\-\-\-\-\-\--
Caci et al. (2003)\*\* 1,3,5,9,13 2,4,6,8,10,12 7,11,14
\*The three-factors are correlated in this model. \*\* Two models based on Caci et al. are tested, the second model removing item 10. ^+^Based on CFA of three independent samples of N = 894, 829 and 824, the total cohort in this study is 2,547. ^\#^PCA: Principal Components Analysis; CFA: Confirmatory Factor Analysis. \*\*FLI: Factor Loading Items. The HADS items loading on each model tested.
Procedure
---------
The first author administered the Chinese version of HADS to patients for self-completion. Demographic data and medical history were also obtained from patients and medical charts.
Participants
------------
One hundred and sixty patients with CHD were initially enrolled into the study of which 154 completed the questionnaires. The patients ranged in age from 38 to 86 years with a mean age of 60 years (SD = 10.37). One hundred and twenty (77.9%) patients were males. In terms of the clinical data, over two thirds of subjects were angina patients. The study was conducted in the general cardiovascular wards of two large university-based teaching hospitals in Xi\'an City of China. Inclusion criteria were diagnoses of angina pectoris or myocardial infarction, no known psychiatric problems, and could understand Chinese. The study was conducted over a four-month period. The clinical ethical committee of the two university affiliated hospitals in Xi\'an approved the study. Informed consent was obtained from all patients prior to commencement of the study.
Results
=======
The mean HADS anxiety (HADS-A) sub-scale score was 6.16 (SD 3.86) and the mean HADS depression (HADS-D) sub-scale score was 6.43 (SD 4.12). Based on Snaith and Zigmond\'s \[[@B29]\] interpretation of HADS-A and HADS-D scores of 8 or over, approximately one-third of the patients screened positive for anxiety (32%) and/or depression (35%). The results of the CFA are summarised in Table [2](#T2){ref-type="table"}. and reveal that the two-factor model of Moorey et al. \[[@B25]\] and the three-factor model of Dunbar et al. \[[@B17]\] offered the best fit to the data. The model fit indices, which revealed identical values for both models, was highly acceptable by conventional model-fit criteria \[[@B30]-[@B33]\].
######
Factor structure of the HADS determined by testing the fit of models derived from factor analysis. All χ^2^analyses were statistically significant at p \< 0.01 (χ^2^degrees of freedom in parentheses).
**Model** **WLSMVχ**^**2**^ **CFI** **TLI** **RMSEA**
--------------------------------- ------------------- ---------- ---------- -----------
Zigmond and Snaith (1983) 65.61 (33) 0.95 **0.97** 0.08
Moorey *et al.*(1991) 57.79 (32) **0.96** **0.97** **0.07**
Caci *et al.*(2003) model 1 71.04 (32) 0.94 0.96 0.09
Caci *et al.*(2003) model 2^\#^ 73.92 (29) 0.93 0.95 1.00
Dunbar *et al.*(2000) 60.19 (33) **0.96** **0.97** **0.07**
Friedman *et al.*(2001) 62.99 (32) 0.95 **0.97** 0.08
Razavi *et al.*(1990) 137.54 (30) 0.84 0.89 0.15
Note: Bold indicates best model fit as a function of model fit index criterion. Abbreviations: Weighted least-square with mean and variance correction (WLSMV); Comparative fit index (CFI); Tucker-Lewis Index (TLI); Root mean squared error of approximation (RMSEA). ^\#^Three-factor model 2 excludes item 10.
Discussion
==========
The finding of high proportions of Chinese CHD patients screening positive with anxiety and depression using the HADS is consistent with previous studies conducted in China or in western countries \[[@B3],[@B5],[@B34],[@B35]\]
The CFA findings are in many respects, rather surprising. It has been suggested that the bi-dimensional underlying factor structure of the HADS observed in many early psychometric evaluations of this instrument is an artifact of the factor extraction method (exploratory factor analysis) and that the instrument is indeed tri-dimensional, a perspective supportive by more recent studies using CFA \[[@B16]\]. In studies of the HADS in patients with CHD using CFA, a clear advantage of three-factor models over two-factor models is consistently observed \[[@B18]-[@B20]\]. In the current study the best performing three-factor and two-factor models were equivalent in terms of model fit indices and offered a good fit to the data. It is interesting however to reflect that the best-performing two-factor model of Moorey et al. \[[@B25]\], represents a modification of the original model proposed by the instrument developers \[[@B7]\], suggesting that even in the context of a two-factor model, the traditional HADS scoring system may not be optimal. The best-performing three-factor model of Dunbar et al. \[[@B17]\] is an important observation because it represents a good performing model based on a cogent theoretical model of anxiety and depression. The observation of a well-fitting theoretical model when applied to clinical data is deemed to be a good test of the model.
Given the observation that tri-dimensional models offered a similar fit to the data as bi-dimensional models, the issue of scoring the instrument as a tri-dimensional instrument is worthy of discussion. It has been previously suggested that the HADS could be scored as a three sub-scale instrument \[[@B17]\]. However, such tri-dimensional scoring approaches that have been proposed are complex and time-consuming for busy practitioners to use routinely since they require factor scores to be regressed to calculate sub-scale scores \[[@B17]\]. A key operational rubric of the HADS is that it is a quickly administered and easily scored measure, therefore implementation of a complicated scoring system would be highly undesirable. Moreover, the extensive use of the HADS in clinical research over the last 20 years has led to the dissemination of several hundred publications reporting the HADS sub-scale means for a broad range of clinical groups. Adopting a tri-dimensional scoring approach would essentially remove this valuable reference data for comparative purposes in new research. Finally, the principle finding of the current study of no clear advantage of tri-dimensional models over bi-dimensional models would suggest that consideration of tri-dimensional scoring approaches is at the very best, highly premature.
The finding of virtually identical fit characteristics of the best performing two and three-factor models also raises the issue of conclusively defining the underlying factor structure of the HADS. The HADS clearly cannot be both bi-dimensional and tri-dimensional within the same data set and further clarification of the structure of the instrument is desirable since this may provide additional evidence not only on the limitations of the HADS, but also the development, enhancement and possible future revision of this widely used measure. The current study was limited by sample size and this may be an important factor in clarifying the relative performance of the competing models tested. It is worthy of note that a number of studies that have utilised factor analysis with large sample sizes have found a clear advantage in model fit of tri-dimensional models over the traditional anxiety/depression bi-dimensional model of the HADS \[[@B17],[@B18],[@B26]\]. Large sample sizes are generally desirable in confirmatory factor analysis and the conservative sample size of the current study may have contributed to the absence of differences in model fit between the best-fit two and three-factor models. Further research is necessary to address this particular issue conclusively.
Previous findings of the factor structure of the Chinese version of the HADS in Cantonese-speaking Chinese CHD patients in Hong Kong indicates clear and consistent superiority of three-factor models in fits to data \[[@B20]\]. One possibility that may account for the ambiguous factorial conclusions in the current investigation concerns the issue of translation. Translating English language instruments to Chinese language versions can be problematic in terms of establishing cultural and semantic equivalence \[[@B36],[@B37]\].
The original validation of the Chinese version of the HADS \[[@B8]\] identified potential issues of case detection accuracy with the instrument. It is conceivable that problems of case detection accuracy may be artifactual of the original translation process, which may also explain inconsistencies between the underlying factor structure of the instrument between Mandarin-speaking Chinese CHD patients in the current study and those reported in Cantonese-speaking Chinese CHD patients.
It should be acknowledged that the current study had a number of limitations, in particular, the modest sample size and the absence of a comparison to a \'gold standard\' such as a structured clinical interview to assess for anxiety and depressive disorder. Further research addressing these limitations is recommended.
Authors\' contributions
=======================
WW participated in the design of the study and assisted in the drafting of the manuscript. VL participated in the design of the study and assisted in the drafting of the manuscript. CM participated in the design of the study, performed the statistical analysis and assisted in the drafting of the manuscript.
Acknowledgements
================
All the authors would like to thanks the patients who took part in the study for their assistance. The authors are also grateful to two anonymous reviewers for their very helpful comments on a previous version of this manuscript. CM would also like to thank his esteemed colleague Dr Hervé Caci, for invaluable discussion and debate on the issues of factor analysis in clinical research.
| {
"pile_set_name": "PubMed Central"
} |
All relevant data are within the paper and its Supporting Information files.
Introduction {#sec007}
============
The nematode *Anisakis* spp. is a parasite of marine mammals that can parasitize humans when a raw or undercooked fish containing live *Anisakis* spp. L3 is consumed. Ingestion of L3 causes an acute and self-limiting infection that can manifest with abdominal pain, nausea, vomiting or diarrhoea. Infection causes a strong polyclonal humoral immune response, and IgM, IgA, and IgG antibodies are detected after one month of infection \[[@pntd.0004864.ref001]\]. In some patients, an IgE-mediated immune response is also triggered, and in those patients, allergic symptoms, such as urticaria, angioedema and anaphylaxis can develop after sensitization and re-exposure to the allergens of this parasite. The rise in specific IgE is usually accompanied with an increase in total IgE in the first month after the presentation of allergic symptoms, and serial serological analysis of both specific and total IgE values have been proven useful in the diagnosis of gastro-allergic anisakiasis \[[@pntd.0004864.ref002]\]. To avoid the appearance of symptoms, sensitized patients are advised to consume frozen or heat-treated fishery products because these treatments kill larvae to prevent new parasitism \[[@pntd.0004864.ref003]--[@pntd.0004864.ref005]\].
Several groups have investigated the kinetics of specific antibody production in experimental animal models \[review in [@pntd.0004864.ref006]\], but the results of those studies may not be applicable to the human immune responses to this parasite. Studies of the changes over time of the level of specific IgE to *Anisakis* spp. in sensitized patients have shown the persistence of IgE sensitization up to 38 months after the onset of symptoms \[[@pntd.0004864.ref001], [@pntd.0004864.ref002], [@pntd.0004864.ref004], [@pntd.0004864.ref007]\]. However, those studies did not report variations in the specific IgE levels at different follow-up time points.
The aim of this study was to analyse the changes in *Anisakis* spp.-specific IgE levels through repeated measures during a longer follow-up period than previously reported and to compare IgE sensitization between patients whose diets did not include fishery products and subjects who regularly consumed fishery products.
Methods {#sec008}
=======
Patients {#sec009}
--------
To analyse the kinetics of the IgE response, *Anisakis* spp.-allergic patients with at least 30 months of follow-up after symptom presentation were selected for this study. A total of 17 patients (six males) with a median age of 53 years (IQR = 45--57 years) were diagnosed as being allergic to *Anisakis* spp. because they reported allergic (urticaria, angioedema or anaphylaxis) and/or gastrointestinal (vomiting, diarrhoea, or abdominal pain) symptoms within 24 h after eating raw or undercooked fish or seafood. One patient reported symptoms after eating cooked fish (scorpion fish cake). Five patients had grass pollen allergy and one of them had dog dander allergy. The data collected at the first visit are shown in [Table 1](#pntd.0004864.t001){ref-type="table"}. Allergy was confirmed by a positive prick test and/or detection of specific IgE to *Anisakis* spp. and undetectable levels of IgE to shrimp, *Ascaris lumbricoides*, fish and mites. To assess new sensitization to these allergens, specific IgE levels were quantified at the last visit, and they remained undetectable in all patients. Measurements of the levels of total and specific IgE to *Anisakis* spp. and clinical evaluations were performed during successive visits. All patients were advised at the first visit to avoid consumption of raw and undercooked fish and to eat farmed fish and deep-frozen fishery products; however, patients with levels of specific IgE to *Anisakis* spp. higher than 100 kU/L were initially instructed to consume a fish-free diet for six months.
10.1371/journal.pntd.0004864.t001
###### Patients' data at first visit and total and specific IgE values at first and last visits of follow-up.
{#pntd.0004864.t001g}
**Patient** **Sex** **Age** **Symptoms** **Te (months)** **Previous allergic episodes** **Fish** **Detection of rAni s 1/rAni s 4** **Follow-up (months)** **First/last total IgE (kU/L)** **First/last specific IgE (kU/L)**
------------- --------- --------- -------------- ----------------- -------------------------------- --------------------- ------------------------------------ ------------------------ --------------------------------- ------------------------------------
1 M 55 U, GI 0.6 n anchovy +/+ 96 658/605 \>100/23.7
2\* M 48 AN, GI 12 n anchovy, anglerfish +/- 71 28/20 5/2.1
3 M 63 AG, U, GI 0.6 n anchovy +/- 118 1740/437 32.4/21.7
4\* F 30 U, GI 4 n anchovy +/- 43 341/41 43.9/3.18
5 M 62 GI 0.7 y anchovy +/- 49 541/70 \>100/15.7
6 F 60 GI 1 n hake +/- 95 1196/384 \>100/52.5
7\* F 50 U, GI 1 y anchovy +/+ 50 1750/71 \>100/8.95
8 F 54 U 1 n anchovy +/- 61 786/338 \>100/37.8
9 F 64 AG, U, GI 5 y anchovy +/+ 38 213/70 72.6/13.3
10 F 53 AN, U 1 n anchovy +/- 44 2958/258 \>100/12.1
11\* F 50 U 12 n anchovy +/+ 38 57/45 4.3/1.1
12 M 27 U 1 n anchovy +/+ 37 1557/440 82.8/20.1
13 F 54 U, GI 4 y anchovy +/- 58 327/103 \>100/17.4
14 F 43 U 12 n scorpion fish cake +/+ 112 557/347 16.1/0.6
15 F 32 U 0.6 n anchovy +/- 31 271/478 9.5/6.1
16\* F 53 U, GI 3 n anchovy +/- 45 91/11 24.9/2.9
17 M 51 GI 9 n anchovy +/- 35 2423/184 79.2/20.6
U: urticaria; GI: gastrointestinal; AN: anaphylaxis; AG: angioedema; Te: Time elapsed from the allergic episode to the first visit; M: male; F: female; n: no; y: yes. Asterisks indicate patients who did not include fish or seafood in their diet during the follow-up period
This study was approved by the Ethics Committee of the Hospital Carlos III (Madrid, Spain), and all included subjects were asked to sign an informed consent form.
Total and specific IgE {#sec010}
----------------------
The serum total and specific IgE measurements were performed with a Phadia 250 instrument (Thermo Fischer Scientific, Phadia, Madrid, Spain) according to the manufacturer's instructions. The detection range for total IgE was 2--5000 kU/L. Regarding positivity for specific IgE antibodies, values \>0.7 kU/L \[[@pntd.0004864.ref008]\] were considered positive for IgE to *Anisakis* spp., and values \>0.35 kU/L were considered positive for IgE to the other allergens.
*Anisakis* spp. antigens {#sec011}
------------------------
Live *Anisakis* spp. larvae in the third stage of the life cycle (L3) were obtained from parasitized hake (*Merluccius merluccius*) at local markets in Madrid, Spain. L3 were extracted from fish tissue, washed in PBS and immediately frozen at -20°C until use. Then, L3 were ground in a Potter-ELV homogenizer and sonicated at 18 w for 5 s. Protein extracts were obtained after centrifugation at 16,000 g and 4°C for 10 min.
Recombinant (r)Ani s 4 and rAni s 1 were obtained as previously reported \[[@pntd.0004864.ref009], [@pntd.0004864.ref010]\].
IgE immunoblotting {#sec012}
------------------
In addition to the total and specific IgE measurements, IgE immunoblotting was performed with the parasite crude extract, recombinant (r) Ani s 1 and rAni s 4.
Proteins extracted from L3 (15 μg), rAni s 1 (3 μg) and r Ani s 4 (3 μg) were subjected to electrophoresis at 120 V on a 4%-20% Tris-glycine gel (Bio-Rad, Hercules, CA, USA). Thereafter, proteins were transferred to nitrocellulose membranes by applying a constant current of 1.3 A for 7 min in a Trans-Blot Turbo Instrument (Bio-Rad). Membranes were blocked with PBS, 0.05% Tween 20 and 1% BSA for 1 h at room temperature and then incubated with 10 mL of the sera of *Anisakis* spp. -allergic patients (1/20) overnight. After washing with PBS, the membranes were incubated for 2 h with 10 mL of a 1:1000 dilution of a monoclonal anti-IgE antiserum (1 mg/mL; Ingenasa, Madrid, Spain). After additional washes, the membranes were incubated with 10 mL of a 1:20,000 dilution of an alkaline phosphatase--labelled goat anti-mouse antiserum (Sigma-Aldrich, St. Louis, MO, USA). Finally, the membranes were washed and incubated with the BCIP-NBT (Sigma-Aldrich) substrate for 30 min.
Statistics {#sec013}
----------
Statistical analyses were performed using SPSS 20.0 software (IBM Corporation, NY, USA). Quantitative variables are described as medians and interquartile ranges (IQR). The Mann-Whitney U test was used to compare the values of total IgE and specific IgE quantified at the first and last visits. A paired-samples comparison between the baseline and last visit values was performed using the Wilcoxon signed-rank test. Regression was used to analyse the trends in the changes in specific IgE values. Changes in specific IgE values over time were estimated with linear and non-linear regression models, and the best fit was selected. A p--value of \<0.05 was considered statistically significant. For the statistical analysis, all values of *Anisakis* spp.-specific IgE \>100 kU/L were assigned a value of 101 kU/L \[[@pntd.0004864.ref011]\].
Results {#sec014}
=======
At first visit, three patients reported gastrointestinal, allergic (n = 6) or allergic and gastrointestinal symptoms (n = 8) after eating raw or undercooked fish, except P14 who reported urticaria after eating cooked fish (scorpion fish cake). Raw anchovies in vinegar sauce were the most frequently consumed fish related to the onset of symptoms ([Table 1](#pntd.0004864.t001){ref-type="table"}). The median time elapsed from the allergic episode to the first visit was one month (IQR = 0.85--7 months). The median follow-up duration was 49 months (IQR = 38--83 months). P5, P7, P9 and P13 reported previous allergic episodes associated with raw fish consumption. At the first visit, using IgE immunoblotting, in addition to *Anisakis* spp.-specific IgE, specific IgE to rAni s 1 was detected in all patients, and specific IgE to rAni s 4 was detected in six patients ([Table 1](#pntd.0004864.t001){ref-type="table"}). Although patients were advised to eat aquaculture and previously frozen fish, five patients did not include fish or seafood in their diet during the follow-up period (P2, P4, P7, P11 and P16), and they were considered to have not been re-exposed to *Anisakis* spp. antigens or allergens.
Positive specific IgE to *Anisakis* spp. values (\>0.7 kU/L) were detected in all patients throughout the follow-up period, except for P14, who had a specific IgE to *Anisakis* spp. value of 0.6 kU/L at the end of a 112-month follow-up period ([Table 1](#pntd.0004864.t001){ref-type="table"}). Baseline total IgE (557 kU/L, IQR = 242--1648 kU/L) and *Anisakis* spp.-specific IgE values (79 kU/L, IQR = 20--101 kU/L) were significantly higher than the final total IgE value (184 kU/L, IQR = 57--410 kU/L; p \<0.01) and the *Anisakis* spp.-specific IgE titre (13 kU/L, IQR = 3--21 kU/L, p\< 0.01) ([Table 1](#pntd.0004864.t001){ref-type="table"}). The median value of the specific IgE decrease was 76% over the follow-up period (IQR = 60%-88%). The paired-samples comparison between the measurements at baseline and the last visit showed significant decreases in both total IgE (p \< 0.01) and *Anisakis* spp.-specific IgE titres (p \< 0.01). No significant differences were found between the patients who were and were not re-exposed to *Anisakis* spp. in the decrease in the *Anisakis* spp.-specific IgE level over the follow-up period (76%, IQR = 51%-84% and 88%, IQR = 66%-92%, respectively; p = 0.28) or in the follow-up duration (53 months, IQR = 37--96 and 45 months, IQR = 40--60, respectively; p = 0.72).
Regression was used to analyse the trend in the changes in *Anisakis* spp.-specific IgE values. The *Anisakis* spp.-specific IgE values underwent an exponential or polynomial decay trend in 13/17 patients, including the patients who had not included fish in their diet during the follow-up period ([Fig 1](#pntd.0004864.g001){ref-type="fig"} and [S1 Fig](#pntd.0004864.s002){ref-type="supplementary-material"}). However, the changes in the *Anisakis* spp.-specific IgE levels from baseline to last visit in four patients (P1, P3, P6 and P15) did not fit a regression model ([Fig 2](#pntd.0004864.g002){ref-type="fig"}). The *Anisakis* spp.-specific IgE level in P1 decreased to 1 kU/L at 10 months, and then increased and remained at \> 12 kU/L throughout the remainder of the follow-up period. The *Anisakis* spp.-specific IgE level in P3 initially showed a 70% decrease (from 50 kU/L to 14 kU/L) and, then, increased up to 37 kU/L at one year later. The changes in the *Anisakis* spp.-specific IgE level in P6 were similar to those in P1, with a decrease of specific IgE titre until reaching 3 kU/L at 12 months and an obvious increase at the next visit (50 kU/L). P15 showed a moderate level of *Anisakis* spp.-specific IgE (10 kU/L) at the first visit; it decreased at one month (6 kU/L) and increased at the following visit. The *Anisakis* spp.-specific IgE increases in these four patients were coincident with the introduction of aquaculture and frozen fish into their diet ([Fig 2](#pntd.0004864.g002){ref-type="fig"}). The paired samples comparisons in these patients revealed no significant differences between baseline total IgE (927 kU/L, IQR = 367--1604 kU/L) and last visit total IgE (457 kU/L, IQR = 397--573 kU/L; p = 0.27) or between baseline *Anisakis* spp.-specific IgE (66 kU/L, IQR = 15-\>100 kU/L) and last visit *Anisakis* spp.-specific IgE (23 kU/L, IQR = 10--45 kU/L; p = 0.07). When the total and specific IgE titres and the duration of follow-up for the patients whose data fit a regression model were compared to those of these four patients, only total IgE at last visit was found to be higher in the latter group (184 kU/L, IQR = 57--410 kU/L vs. 457 kU/L, IQR = 397--573 kU/L, p \< 0.01).
{#pntd.0004864.g001}
{#pntd.0004864.g002}
Three patients reported symptoms during the follow-up period. P3 experienced localized acute urticaria in the hands after eating cod that had been frozen for 72 h at home ([Fig 2](#pntd.0004864.g002){ref-type="fig"}). P5 reported diarrhoea after eating aquaculture sea bass ([Fig 1](#pntd.0004864.g001){ref-type="fig"}). In these two patients, the appearance of symptoms was coincident with an increase in *Anisakis* spp.-specific IgE level. P3 and P5 did not report new episodes in the subsequent visits. P13 showed generalized acute urticaria after eating commercial frozen hake ([Fig 1](#pntd.0004864.g001){ref-type="fig"}). Unfortunately, this episode occurred at the end of the follow-up period, and we could not assess if the level of IgE against *Anisakis* spp. had varied. Other common causes of acute urticaria were discarded.
To determine if the changes in specific IgE over time are related to changes in the recognition pattern of allergens, IgE immunoblotting was performed at different time points. As expected, baseline IgE immunoblotting showed different patterns for the parasite crude extracts \[[@pntd.0004864.ref012]\]. rAni s 1 was detected by all patients, and rAni s 4 was detected by six patients (Figs [1](#pntd.0004864.g001){ref-type="fig"} and [2](#pntd.0004864.g002){ref-type="fig"} and [S1 Fig](#pntd.0004864.s002){ref-type="supplementary-material"}). The intensities of the bands corresponding to rAni s 1 paralleled the *Anisakis* spp.-specific IgE levels. During the follow-up period, the rAni s 4 bands disappeared prior to the rAni s 1 bands, suggesting that Ani s 4 is an early marker of *Anisakis* spp. infection. On the other hand, the increase in *Anisakis* spp.-specific IgE over time observed in some patients was associated with an increase in the intensity of some proteins in the parasite crude extract and rAni s 1 ([Fig 2](#pntd.0004864.g002){ref-type="fig"}).
Discussion {#sec015}
==========
Our study analysed the changes in the values of total IgE and specific IgE to *Anisakis* spp. over time using a longer follow-up period (31--118 months) than previous studies (6--38 months) \[[@pntd.0004864.ref001], [@pntd.0004864.ref002], [@pntd.0004864.ref004], [@pntd.0004864.ref007]\]. Repeated measures were performed throughout the study period and included detection of Ani s 1 and Ani s 4. Ani s 1 is a major and heat stable allergen \[[@pntd.0004864.ref012]\] and Ani s 4 is a pepsin and heat-resistant allergen, and its clinical relevance has been verified because it is associated with anaphylaxis \[[@pntd.0004864.ref009]\]. Exposure to the live fish-borne parasite *Anisakis* spp. L3 can produce an acute and self-limiting infection in humans with allergic and gastrointestinal symptoms. To avoid the appearance of symptoms, patients are advised to consume frozen or heat-treated fishery products because these treatments kill larvae and thus prevent new parasitism \[[@pntd.0004864.ref003]--[@pntd.0004864.ref005]\]. An alternative is to consume aquaculture fish because the risk of exposure to *Anisakis* spp. larvae in farmed fish has been shown to be minimal, because no *Anisakis* spp. larvae have been found in their viscera or flesh \[[@pntd.0004864.ref005], [@pntd.0004864.ref013]\]. Twelve patients regularly consumed aquaculture and frozen fish during the study period, and five patients decided to stop consuming fish throughout the follow-up period. The decrease in the *Anisakis* spp.-specific IgE levels in these patients, who were considered to have not been re-exposed to parasite material, was higher than those found previously in patients on a fish-free diet (57%) with a follow-up period of 13 months, and this discrepancy was probably due to our longer follow-up period \[[@pntd.0004864.ref004]\]. On the other hand, we did not observe significant differences in the total or *Anisakis* spp.-specific IgE values between the patients who did and did not consume fish during the follow-up period \[[@pntd.0004864.ref004]\]; thus, the rate of *Anisakis* spp.-specific IgE decay does not seem to be influenced by the consumption of previously frozen fish.
Our results show that IgE sensitization to *Anisakis* spp. allergens persists over several years *since specific IgE was detectable in some patients after more than 8 years from the allergic episode*. Similar results were observed in a nine-year follow--up study of adult subjects sensitized to food allergens \[[@pntd.0004864.ref014]\]. In our study, specific IgE to Ani s 1 was detectable up to 118 months from the onset of symptoms, and this result agrees with that obtained in a study with 6--38 months of follow-up \[[@pntd.0004864.ref007]\] and suggests that Ani s 1 can be detected in both recent and old *Anisakis* spp. infection cases. Therefore, the persistent IgE sensitization observed in *Anisakis* spp. allergic patients who were not exposed to parasite allergens for several years indicates that *Anisakis* spp. allergens induce long-lived IgE responses.
The analysis of the changes over time in specific IgE level in the patients that were not re-exposed to *Anisakis* spp. allergens and in some patients who regularly consume aquaculture and frozen fish shows that the decrease in specific IgE titres fit a non-linear regression model. However, we observed that the *Anisakis* spp.-specific IgE level in four patients (P1, P3, P6 and P15) increased at some time points during follow--up. This increase in *Anisakis* spp.-specific IgE could be due to sensitization to other allergen sources that have been reported to cross-react with *Anisakis* spp. allergens, i.e., other parasite nematodes, mites and crustaceans \[[@pntd.0004864.ref015]--[@pntd.0004864.ref017]\]. However, according to the available patient data, this explanation does not seem to be valid because our patients were not sensitized to allergens that cross-react with *Anisakis* spp. allergens at baseline, and no new onset hypersensitivity to those allergens was found during the follow-up period. In addition, the increase in *Anisakis* spp.-specific IgE was parallel to the increase in Ani s 1 detected by immunoblotting, which supports the hypothesis that the observed changes in *Anisakis* spp.-specific IgE were actually due to exposure to *Anisakis* spp. allergens. Another explanation for this observation is re-infection with live L3. However, we do not believe this is the case because the patients reported symptoms after eating frozen or aquaculture fish, and it is unlikely that an acute parasitism episode went unnoticed by these patients, as they had experienced one previously. On the other hand, previously unrecognized allergens have been detected in the course of gastro-allergic anisakiasis \[[@pntd.0004864.ref001]\], which was not observed in our patients according to the results of IgE immunoblotting. Accordingly, it has been shown that adult patients allergic to aeroallergens did not acquire sensitization to new allergens, but they exhibited a pre-established profile of allergens after antigen exposure. The levels of allergen-specific IgE in previously sensitized allergic patients decreased in the absence of allergen exposure and increased upon allergen exposure \[[@pntd.0004864.ref018]\]. Because the raise in specific IgE levels is coincident with the introduction of fishery products to the diet, a more plausible explanation for the changes in IgE sensitisation is that exposure to allergens present in those fishery products is involved in the increase in the *Anisakis* spp.-specific IgE level. Some *Anisakis* spp. antigens have been shown to be stable and to maintain their capacity to bind IgE after different freezing and heat treatments \[[@pntd.0004864.ref019], [@pntd.0004864.ref020]\]. Furthermore, *Anisakis* spp. antigens have been detected in farmed salmon and processed fish products \[[@pntd.0004864.ref021]\], which could contribute to the variations in the trends of IgE sensitization observed in this study.
The clinical significance of these results is difficult to determine because very few patients (n = 3) reported symptoms after the initial acute parasitism episode in our study. It has been proposed that the high rate of *Anisakis* spp. parasitism of fish in our region would result in frequent contact with parasite material, which would cause symptoms in sensitized subjects exposed to it when previously frozen or heat-treated fish is included in their diet \[[@pntd.0004864.ref022]\]. However, our results suggest that the exposure to *Anisakis* spp. antigens present in fishery products may contribute to the persistence and even to the increase in *Anisakis* spp.-specific IgE level. The increase in the *Anisakis* spp.-specific IgE level associated with the appearance of symptoms indicates that exposure to *Anisakis* spp. material could be involved in the onset of symptoms in some *Anisakis* spp. allergic patients \[[@pntd.0004864.ref012]\]. However, more studies are required to evaluate the clinical relevance of the re-exposure to *Anisakis* spp. in sensitized patients.
In conclusion, we have shown that IgE sensitization to *Anisakis* spp. allergens can last more than 8 years. Specific IgE monitoring showed that specific IgE titres increase in some allergic patients and that allergen contamination of fishery products can account for the observed increase in *Anisakis* spp.-specific IgE level.
Supporting Information {#sec016}
======================
###### STROBE checklist.
(DOCX)
######
Click here for additional data file.
###### Changes in IgE sensitization to *Anisakis* spp. allergens during follow-up.
Letters indicate the time points during follow-up when IgE immunoblotting was performed. Lane 1: *Anisakis* spp. crude extract; lane 2: rAni s 1; lane 3: rAni s 4. The shadow box indicates the follow-up period during which the patients were consuming fish. sIgE: specific IgE to *Anisakis* spp.
(PDF)
######
Click here for additional data file.
###### Accession numbers.
(DOC)
######
Click here for additional data file.
[^1]: The authors have declared that no competing interests exist.
[^2]: Conceived and designed the experiments: MC AN IM MGM. Performed the experiments: NCS AIRM. Analyzed the data: NCS AIRM MC AN IM MGM. Contributed reagents/materials/analysis tools: NCS AIRM MC AN. Wrote the paper: IM MGM.
| {
"pile_set_name": "PubMed Central"
} |
{#sp1 .697}
{#sp2 .698}
{#sp3 .699}
{#sp4 .700}
{#sp5 .701}
{#sp6 .702}
{#sp7 .703}
{#sp8 .704}
{#sp9 .705}
{#sp10 .706}
{#sp11 .707}
{#sp12 .708}
{#sp13 .709}
{#sp14 .710}
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s0005}
============
In 2017, pancreatic cancer (PC) represents 3.0% of all new cancer cases in the United States [@bb0005]. Compared to other cancers, PC is relatively rare. However, it is more common with increasing age and has an aggressive behavior with poor prognosis, resulting in an estimated 44,330 deaths in 2017 and making it the third leading cause of cancer death in the United States [@bb0005]. Because of its high frequency of chemoresistance, PC is relatively insensitive to conventional chemotherapy. Gemcitabine was recommended as the first first-line drug for chemotherapy of PC and chemotherapy using gemcitabine alone was the standard for about a decade, as a number of trials testing it in combination with other drugs failed to demonstrate significantly better outcomes [@bb0010], [@bb0015], [@bb0020]. Hence, how to further improve the sensitivity of PC cells to gemcitabine will provide clues for new targeted therapies.
Polo-like kinase 1(Plk1), a serine/threonine kinase that plays an essential role in cell mitosis, spindle assemble, DNA damage and so on, is a member of polo-like kinases family [@bb0025], [@bb0030]. It is an early trigger for G2/M transition and localizes to centrosomes during interphase. Plk1 is closely related with the occurrence of tumor development. Compared to normal tissues, it is over-expressed in a broad range of tumors, such as ovarian carcinoma, colorectal carcinoma, prostate cancer, skin cancer, and others [@bb0035], [@bb0040], [@bb0045], [@bb0050], and has been implicated in tumorigenesis and progression [@bb0055], [@bb0060]. Overexpression of Plk1 can inhibit the activity of p53 through phosphorylation p53, which cause the apoptosis process failure, then cancerous cells survival which bring out the occurrence of cancer eventually [@bb0055], [@bb0065]. As reported, Plk1 is over-expressed in PC, invasive pancreatic adenocarcinomas were Plk1 positive in 47.7% of cases [@bb0070], which means Plk1 overexpression is likely to be related to biological behavior of PC. Therefore, targeting Plk1 in developing small molecule inhibitors as anti-cancer drugs becomes a hotspot in the recent years. Such as Plk1 inhibitor BI2536 has been evaluated for patients with various cancers in clinical trials [@bb0075], [@bb0080], [@bb0085], [@bb0090].
The deregulation of PI3K/Akt pathway has been confirmed that plays an crucial role in human cancers including PC [@bb0095], [@bb0100]. PI3K/Akt pathway includes a series of cascade: PI3K generation PI3P through phosphorylation PI2P, and PI3P combination with the N-terminal of Akt, then activation of Akt, which activate or inhibit its downstream substrates, such as mTOR, Caspase-9, Bad, etc. [@bb0105]. Akt, a serine/threonine kinase which plays a core role in the PI3K/Akt signal pathway. Meanwhile, Akt contributes to cell plasticity in pancreas as a regulator and its overexpression has been proved to be a common phenomenon in PC [@bb0110], [@bb0115], [@bb0120]. Until now, several Akt inhibitors have been evaluated in clinical trials [@bb0125], [@bb0130]. Hence, the PI3K/Akt pathway plays a crucial role in the development of PC, and it can be a potential therapeutic target for PC.
Cell cycle is crucial for proliferation, differentiation, and growth both in normal cells and tumor cells. Hence, the factors affecting cell cycle can be used as a potential target of tumor therapy. Whether PI3K/Akt signaling pathway or Plk1 in cell cycle is the indispensable factors, keep unclear. It is reported PI3K/Akt-dependent phosphorylation of Plk1-Ser99 is required for metaphase-anaphase transition, and Plk1-dependent phosphorylation of IRS2-S556 inhibits mitotic exit through reducing Akt activity [@bb0135], [@bb0140]. Yu et al. showed that up-regulation of Plk1 were related to chemoresistance of PC, and Kim et al. found that inhibited PI3K/Akt pathway could increase the chemosensitivity of PC to gemcitabine [@bb0145], [@bb0150]. However, the mechanism of cell apoptosis induced by PI3K/Akt pathway and Plk1 still remains unclear. In current study, we combined the inhibition of PI3K/Akt pathway with down-regulation of Plk1 and observed that the chemosensitivity of PC to gemcitabine further increased. In addition, LY294002 and BI2536 (an inhibitor of Plk1), together with gemcitabine significantly suppressed the growth of xenografts in nude mice model, which made PI3K/Akt pathway inhibition as well as Plk1 down-regulation become a potential target to increase the chemosensitivity of PC.
Materials and Methods {#s0010}
=====================
Cell Culture and Reagents {#s0015}
-------------------------
PC cell lines were obtained from ATCC and were maintained in Dulbecco\'s Modified Eagle Medium supplemented with 10% fetal bovine serum (FBS) (BxPC-3 and PANC-1), in RPMI 1640 supplemented with 10% FBS (AsPC-1). Gemcitabine was purchased from Sigma (Sigma-Aldrich Co. LLC, USA), it was dissolved in 100% dimethyl sulfoxide (DMSO, Muskegon, MI, USA) to a stock concentration of 10 mM and stored at −20 °C. The PI3K inhibitor LY294002 was purchased from CST (Beverly, MA, USA). BI2536 (Plk1 inhibitor) was gained from Selleck (Houston, USA), it was dissolved in DMSO and stored at −80 °C. The following antibodies were purchased for Western blot: Plk1, p-Akt, Akt (Cell Signaling Technology, MA, USA); Bcl-2, BAX (Abcam, UK); and glyceraldehyde 3-phosphate dehydrogenase (GAPDH, Santa Cruz, CA, USA). Plk1 and Akt antibody for immunohistochemistry (IHC) was gained from Abcam (Abcam, UK).
Immunohistochemistry (IHC) and Clinical Specimens {#s0020}
-------------------------------------------------
The human pancreatic tissues and related clinical data were purchased from Xian Ailina Biotechnology Co. Ltd. (Xian Ailina Biotechnology Co. Ltd., China). Paraffin-embedded sections of human pancreatic tissues and mouse xenografts were subjected to specific antibodies for Plk1, Akt, Cleaved caspase 3, or isotype-matched controls at appropriate dilutions. IHC staining in human tissues were scored independently by two pathologists, by evaluating a semiquantitative immunoreactivity score (IRS) as described [@bb0155]. Then, tissues with IRS 0--5 and IRS 6--9 were defined as low and high expression of Plk1, respectively.
Construction of Recombinant Adenoviral rAd-EGFP and rAd-Plk1-shRNAs (rAd-shPlk1) {#s0025}
--------------------------------------------------------------------------------
Based on gene sequence (GenBank: [NM_005030](ncbi-n:NM_005030){#ir0005}), four shRNAs targeting different regions of the Plk1 transcript were synthesized with the vector pYr-1.1 (hU6/EGFP/Neo) (Changsha Yingrun Biotechnology Co. Ltd., China). After package and screen, both rAd-Plk1-shRNA2 and rAd-Plk1-shRNA4 worked effectively, especially rAd-Plk1-shRNA4 so that the latter was finally used for the following experiment and was set as rAd-shPlk1. The empty vector rAd-EGFP was constructed as an experimental control.
Western Blotting Analysis {#s0030}
-------------------------
Cells were lysed in RIPA Lysis Buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 1 mM PMSF) and resolved by SDS-PAGE. Samples were analyzed as described [@bb0160]. Phospho-specific antibody to Akt(S473) was detected to determine the level of activated protein, with antibody recognizing total Akt to control for total protein expression. Antibody to Plk1 was used to monitor total protein expression. Antibodies for BAX and Bcl2 were used to monitor apoptosis. Antibody for GAPDH was used to verify equivalent loading of total cellular protein.
RNA Isolation and Quantitative Real-Time PCR {#s0035}
--------------------------------------------
The mRNAs were extracted with Trizol reagent (Life Technologies, Ltd). The cDNAs were prepared with Superscript III (Life Technologies, Ltd) Takara Kit (Dalian, China)as indicated by the manufacturer. The cDNAs were amplified by PCR in an iQ5 Multicolor Real Time Detector System (Bio-Rad) with the fluorescent double-stranded DNA binding dye SYBR Green (Bio-Rad). The relative amounts of gene expression were calculated with GAPDH expression as an internal standard (calibrator). Primers used for the Akt gene: forward primer 5′-TCACCATCA CACCACCTG AC-3′ and reverse primer 5′-CTCAAATGCACCCGAGAA AT-3′. Primers used for the Plk1 gene: forward 5′-ACC AGC ACG TCG TAG GAT TC-3′ and reverse 5′-ATA ACT CGG TTT CGG TGC AG-3′. Primers used for the GAPDH gene were 5′-AAC GGA TTT GGT CGT ATT GG-3′ (forward) and 5′-GGA TCT CGC TCC TGG AAG AT-3′ (reverse) (Invitrogen, USA). The results, expressed as N-fold differences in target gene expression, were determined as follows: N \*target = 2 (Ct sample Ct calibrator).
Apoptosis Detection by Flow Cytometry (FCM) {#s0040}
-------------------------------------------
Apoptosis was detected by flow cytometry. The cells were incubated with gemcitabine for 48 h, LY294002 for 48 h or recombinant adenovirus for 24 h, 48 h, 72 h and 96 h; then washed twice with ice-cold PBS and resuspended in 1× Binding Buffer(BD Pharmingen, USA) at a concentration of 1× 105 cells/ml. Added 5 μl of APC Annexin V(BD Pharmingen, USA) and 5 μl of 7-AAD(BD Pharmingen, USA).Then, samples were analyzed by flow cytometer(BD FACSCalibur equipped with CellQuest Pro).
Determination of IC50 {#s0045}
---------------------
To determine the half maximal inhibitory concentration (IC50), 1--2 × 10^3^ cells were seeded in 96-well plates and treated with concentrations of gemcitabine from 0 μM to 20 μM). After 48 h of treatment, cell proliferation was analyzed with CCK8 and read at an absorbance of 450 nm. IC50 values were calculated with CalcuSyn or CompuSyn software.
PC Cell Line Xenografts {#s0050}
-----------------------
Female athymic nude mice (nu/nu, 6 to 8 weeks of age, gained from Chinese Academy of Sciences, Shanghai, China) were maintained under pathogen-free conditions in micro isolator cages. Animal procedures were performed in accordance with the rules set forth in the NIH Guide For The Care And Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee at Merck. All animals received food and water ad libitum. BxPC-3 cells (1 × 10^7^/120 μl in PBS:Matrigel = 1:1) were injected s.c. into the right flank region of the mice. Tumors were allowed to reach 80--120 mm^3^ before randomization to treatment groups. For continuous dosing studies, gemcitabine was administered intraperitoneally (i.p.) at 100 mg per kilogram body weight (mpk) once daily (qd) for 14 to 21 days, LY294002 (25 mg/kg) and BI2536 (30 mg/kg) were administered i.p. twice a week. Mice in control group received 1%DMSO on the same schedules that gemcitabine, LY294002 and BI2536 were administered. Six mice were in each treatment group. Tumor volumes were measured twice weekly using calipers and calculated by the formula (length × width^2^)/2. Animal body weights were measured on the same days twice weekly. Statistically significant differences between the treatment groups were determined by the Student\'s *t* test and 2-way ANOVA using GraphPad Prism. Six hours after their last dose, mice were euthanized by carbon dioxide inhalation. The tumors were then removed and cut into two pieces. One piece was frozen in liquid nitrogen while the second piece was fixed in 10% formalin for future analysis.
Terminal Deoxynucleotidyl Transferase dUTP Nick End Labeling (TUNEL) Assay {#s0055}
--------------------------------------------------------------------------
Cut the paraffin blocks into slices with 4 μm thickness. The TUNEL Brighted Apoptosis Detection Kit, POD (Vazyme, Nanjing, China) was used to detect cellular apoptosis in tumor tissues according to the instructions.
Statistical Analysis {#s0060}
--------------------
Data were expressed as mean ± SD unless otherwise specified and each group was statistically analyzed using t-test or ANOVA. All statistical analyses were performed using GraphPad Prism 5.0 and SPSS 20.0 software. Each experiment was performed three times. Differences were considered statistically significant at *P* \< .05.
Results {#s0065}
=======
Plk1 is Overexpressed in Human Pancreatic Carcinoma Versus Normal Tissue {#s0070}
------------------------------------------------------------------------
We purchased pancreatic tissue microarray (including 59 cases of normal pancreas and 88 cases of PC), and IHC staining was used to detect the expression of Plk1. The staining showed that 83% of normal tissues were Plk1 with low expression (IRS 0--5), 17% of normal tissues were Plk1 with high expression (IRS 6--12); whereas 66% of PC tissues were low expression (IRS 0--5), 34% of PC tissues were high expression (IRS 6--12). The above results mean that the expression of Plk1 in PC tissues was higher than that of the normal pancreatic tissues (*P* = .022) ([Figure 1](#f0005){ref-type="fig"}, *A* and *B* and [Table 1](#t0005){ref-type="table"}).Figure 1Plk1 is over-expressed in human pancreatic carcinoma versus normal tissue. (A) Immunohistochemical staining for Polo-like kinase 1(brown) and counterstaining with hematoxylin illustrates Plk1 is over-expressed in the pancreatic cancer. Single cells positive for Plk1 are shown in high magnification. N: Normal pancreatic tissue; T: pancreatic cancer tissues. (B) Statistical analysis of immunohistological staining for Plk1, including 59 normal tissues and 88 pancreatic cancers(\*,*P* = .022). Error bars represent standard deviations. (C) The expression of Plk1 is higher with an increased pathologic grade.Figure 1Table 1Expression of Plk1 in pancreatic tissue was detected by immunohistochemical staining and illustrates Plk1 is over-expressed in the pancreatic cancerTable 1Expression of Plk1 in pancreatic tissueGroupnLow expressionHigh expression*P*Normal tissue5949 (83%)10 (17%).022^⁎^Pancreatic cancer8858 (66%)30 (34%)
Next, we analyzed the expression of Plk1 in PC tissues and the clinical characteristics of these patients to explore their relationship. The study showed that the expression level of Plk1 in PC tissues has no significantly difference among gender, age or clinical stage (*P* \> .05). Nevertheless, Plk1 expression level was correlated with pathologic grade. Pathologic grade I-II with 74% low expression and 26% high expression of Plk1; Pathologic grade III-IV with 48% low expression and 52% high expression of Plk1. These analysis result imply that more PC tissues with high expression of Plk1 in Pathologic grade III-IV than those in Pathologic grade I-II (*P* = .019) ([Figure 1](#f0005){ref-type="fig"}*C* and [Table 2](#t0010){ref-type="table"}).Table 2Expression of Plk1 in pancreatic cancer was detected by immunohistochemical staining and illustrates the expression of Plk1 is higher with an increased pathologic grade (*P* = .019) while there is no significant difference among gender, age and clinical stage (*P* \> .05)Table 2Expression of Plk1 in Pancreatic cancerCharacteristicsnLow expressionHigh expression*P*Gender.173 M4734 (72%)13 (28%) F4124 (59%)17 (41%)Age.728 \<605537 (67%)18 (33%) ≥603321 (64%)12 (36%)Pathologic grade.019^⁎^ 1--26145 (74%)16 (26%) 32713 (48%)14 (52%)Clinical Stage.719 I-II7550 (67%)25 (33%) III-IV138 (62%)5 (38%)
Targeted Depletion of Plk1 via Recombinant Adenoviral shRNA Causes Apoptosis in PC Cells {#s0075}
----------------------------------------------------------------------------------------
To further explore the role of Plk1 in PC cells, we constructed the recombinant adenoviral to depletion of Plk1 (rAd-shPlk1). QRT-PCR and Western blotting analysis showed that, under the condition of the same amount of the total sample, GAPDH strips were basically same when we used GAPDH as internal control, while the Plk1 strips were significantly different, the Plk1 mRNA and protein relative quantity in rAd-shPlk1 group was markedly lower than those in the rAd-EGFP groups (*P* \< .001) and control groups (*P* \< .01) ([Figure 2](#f0010){ref-type="fig"}, *A* and *B*). From our data above we believed that both Plk1 mRNA and protein levels were significantly down-regulated by rAd-shPlk1 in PC cells. We used Western blotting to detect apoptosis-related proteins Bcl-2 and BAX levels after down-regulation of Plk1 expression. In result, we found that Bcl-2 was down-regulated and BAX was up-regulated, which means depletion of Plk1 could induce apoptosis in PC cells ([Figure 2](#f0010){ref-type="fig"}*B*). Then, we examined the cell apoptosis of each group by flow cytometry. As shown in [Figure 3](#f0015){ref-type="fig"}*E*, we checked the apoptosis rates of the control group, the rAd-EGFP group and the rAd-shPlk1 group at 24 h, 48 h, 72 h and 96 h, respectively. Obviously, during the first 24 h, apoptosis rates among these three groups showed no significant difference (*P* \> .05); while after 48 h, the rAd-shPlk1 group had the highest cell apoptosis rate among them (*P* \< .01) and comparing to the control group and rAd-EGFP group, the viable cell number of the rAd-shPlk1 group was significantly lower than that of the rAd-EGFP group at each time point ([Figure 3](#f0015){ref-type="fig"}, *C*--*E*). Hence, we inferred that rAd-shPlk1 could induce the apoptosis of PC cells.Figure 2Targeted depletion of Plk1 via recombinant adenoviral shRNA causes apoptosis in PC cells. (A) QRT-PCR was performed to detect Plk1 mRNA level and reveals lowest Plk1 mRNA level in rAd-shPlk1 group. Control: no treatment; rAd-EGFP: adenoviral vector which carries enhanced green fluorescent protein (EGFP) alone; rAd-shPlk1: adenoviral vector which carries EGFP-Plk1-shRNA. (B) The Plk1 and apoptosis-related proteins Bcl-2 and Bax protein expression levels in human pancreatic cancer cell lines were shown by Western blotting analysis. GAPDH expression was used as internal controls. The Plk1-siRNA reduces the expression of the Plk1 protein expression significantly and more cell apoptosis is observed in the rAd-shPlk1 group. (C) Pancreatic cancer cell lines were treated with recombinant adenovirus and harvested at 72 h. Apoptosis in different groups after infection was detected by Flow Cytometry. (D) Statistical analysis of apoptosis rate detected by Flow Cytometry at 72 h after recombinant adenovirus infection shows that the rAd-shPlk1 group cell apoptosis rate is significantly higher than that of the other two groups (*P* \< .001). (E) Cell apoptosis rate detected by Flow Cytometry at 24 h, 48 h 72 h and 96 h after recombinant adenovirus infection revealsrAd-shPlk1 has a long-term apoptosis-inducing effect on pancreatic cancer cells and it works after 24 h post-infection. All the experiments were performed in triplicates. Error bars represent standard deviations.Figure 2Figure 3Gemcitabine-resistance in PC is closely related to PI3K/Akt signal pathway. (A) Immunohistochemical staining for Akt(brown) and counterstaining with hematoxylin illustrates Akt is over-expressed in the pancreatic cancer. At the meantime, Akt expression is higher in the pancreatic cancer tissue after the patient accepted gemcitabine treatment. N: Normal pancreatic tissue; T1: pancreatic cancer tissue without gemcitabine treatment; T2: pancreatic cancer tissue with gemcitabine treatment. (B) MTT assay was performed to detect cytotoxicity of gemcitabine on pancreatic cancer cell lines and the appropriate concentrations of Gemcitabine in different cell lines are IC50~AsPC-1~ = 15uM, IC50~BxPC-3~ = 7.5uM and IC50~PANC-1~ = 12.5uM. The relative cellular viability was measured by setting untreated cells as 100% cell viability. (C) Akt mRNA level tested by qRT-PCR illustrates gemcitabine up-regulates the Akt expression in pancreatic cancer (*P* \< .001) while PI3K inhibitor LY294002 can reverse this process (*P* \< .001). (D) Akt protein level tested by Western Blotting shows the same results as that of qRT-PCR. (E) Pancreatic cancer cell apoptosis was detected by Flow Cytometry at 72 h after gemcitabine or LY294002 or both drugs treatment. (F) Statistical analysis of apoptosis rate detected by Flow Cytometry at 72 h after drugs treatment reveals combination of gemcitabine and LY294002 enhances chemosensitivity of gemcitabine on pancreatic cancer cells (*P* \< .001). All the experiments were performed in triplicates. Error bars represent standard deviations.Figure 3
Gemcitabine-Sensitivity in PC is Closely Related to PI3K/Akt Signal Pathway {#s0080}
---------------------------------------------------------------------------
As mentioned earlier, the PI3K/Akt signal pathway are closely related to cell cycle, proliferation, differentiation, apoptosis and so on, and the development and drug-resistance of most malignant tumors are correlation with the deregulation of PI3K/Akt pathway [@bb0100], [@bb0165]. Therefore, we explore the influence of PI3K/Akt pathway on resistance of PC-chemotherapy. We got normal pancreas tissues, PC tissues and PC tissues after the patient accepted gemcitabine treatment, then we used IHC to detect the Akt level in three tissues. The data shows the Akt is over-expressed in PC tissues than normal tissues, at the same time, Akt level is higher in PC tissues after the patient appeared gemcitabine-resistance ([Figure 3](#f0015){ref-type="fig"}*A*), which suggested that the abnormal activation of PI3K/Akt pathway might be associated with chemoresistance in PC. Next, we used cellular experiment to validate our hypothesis. Cell proliferation influenced by different concentrations of gemcitabine was detected by CCK-8 assay ([Figure 3](#f0015){ref-type="fig"}*B*). The optimal concentrations of gemcitabine for different PC cells were IC50AsPC-1 = 15uM, IC50BxPC-3 = 7.5uM and IC50PANC-1 = 12.5uM, respectively. Then, we disposed PC cells with gemcitabine or LY294002 (PI3K/Akt inhibitor) or both combined. And the qRT-PCR and Western blotting showed gemcitabine could activate PI3K/Akt pathway through up-regulating Akt and p-Akt level. While LY294002 had no effect on mRNA level of Akt, and protein level of p-Akt was down-regulated, which imply that LY294002 influenced PI3K/Akt pathway through decreasing phosphorylation of Akt ([Figure 3](#f0015){ref-type="fig"}, *C* and *D*). After the PC cells were treated 72 h, we examined the cell apoptosis of each group by flow cytometry. Compared to the control group, gemcitabine group (Gem), and LY294002 group (LY), using LY294002 combined gemcitabine (Gem+LY) could induce the highest apoptosis rate (*P* \< .001) ([Figure 3](#f0015){ref-type="fig"}, *E* and *F*). Hence, the results confirmed our previous hypothesis: the gemcitabine resistance to PC may be associated with abnormal activation of PI3K/Akt pathway, and LY294002 could increase chemosensitivity of PC cells to gemcitabine.
LY294002 Promotes Chemosensitivity of PC Through Downstream Plk1 Inactivation {#s0085}
-----------------------------------------------------------------------------
Next, we ask did LY294002 increase chemosensitivity of PC cells to gemcitabine? It is reported that PI3K/Akt signal pathway are closely related to resistance to apoptosis signal transduction and chemoresistance of PC [@bb0170], [@bb0175]. On the other hand, Plk1 is essential for cell cycle, our previous study showed Plk1 is correlated to cell proliferation and chemoresistance in PC, and PI3K/Akt-dependent phosphorylation of Plk1-Ser99 is required for metaphase-anaphase transition [@bb0135], [@bb0145]. So, it is possible that PI3K/Akt and Plk1 have interaction role in chemoresistance of PC. In order to test this point, we disposed PC cells with gemcitabine or LY294002 (PI3K/Akt inhibitor) or both combined and detected the Plk1mRNA and protein level. As shown in [Figure 4](#f0020){ref-type="fig"}, *C* and *D*, the Plk1 expression level in PC cells with gemcitabine treatment showed up-regulated compared to control cells. While Plk1 expression level in LY294002 treatment showed down-regulated compared to control group, partly reversal gemcitabine up-regulating Plk1. And gemcitabine combined LY294002 could markedly up-regulating BAX and down-regulating Bcl-2, showed both combined can increase cell apoptosis ([Figure 4](#f0020){ref-type="fig"}, *A* and *B*). To further explore Plk1 in the process, we used rAd-shPlk1 to infect PC cells. The qRT-PCR and Western blotting showed Plk1 suppression has no influence on Akt and p-Akt ([Figure 4](#f0020){ref-type="fig"}, *C* and *D*), and as [Figure 2](#f0010){ref-type="fig"} showed suppression Plk1 can induce PC cells apoptosis. Thus as a conclusion, we believed that LY294002 promotes chemosensitivity of PC to gemcitabine through down-regulating Plk1.Figure 4PI3K/Akt signal pathway promotes chemoresistance of PC through downstream Plk1 activation. (A) Plk1 mRNA level detected by qRT-PCR illustrates gemcitabine up-regulates the Plk1 expression in pancreatic cancer (*P* \< .001) while PI3K inhibitor LY294002 can reverse this process (*P* \< .001). (B) The Plk1 and apoptosis-related proteins Bcl-2 and Bax protein expression levels in human pancreatic cancer cell lines were shown by Western blotting analysis. GAPDH expression was used as internal controls.PI3K inhibitor LY294002 reduces Plk1 expression and can lower the effect of Plk1 up-regulating by gemcitabine. Combination of gemcitabine and LY294002 enhances the cytotoxicity of gemcitabine. (C) The Plk1 and Akt protein expression levels in human pancreatic cancer cell lines were detected by Western blotting analysis. GAPDH expression was used as internal controls. Plk1 knockout through shRNA shows no influence in Akt expression. (D) QRT-PCR illustrates the same result as that showed in the previous Western Blotting analysis and proves Plk1-shRNA does not influence the Akt expression.Figure 4
Targeted Depletion of Plk1 Enhances Chemosensitivity Induced by PI3K Inhibitor LY294002 in PC {#s0090}
---------------------------------------------------------------------------------------------
Based on the above research, we reasoned that PI3K/Akt and Plk1 are both essential for chemoresistance in PC, and Plk1 as the downstream of PI3K/Akt regulate the chemosensitivity of PC. Hence, we further explored inhibition of PI3K/Akt and depletion of Plk1 on the effect of PC chemosensitivity. We treated PC cells with gemcitabine or LY294002 or rAd-shPlk1 or rAd-shPlk1 + Gem+LY. QRT-PCR and Western blotting showed rAd-shPlk1 + LY can reverse the role of gemcitabine up-regulating Plk1 obviously, and rAd-shPlk1 + Gem+LY can up-regulating BAX and down-regulating Bcl-2 significantly ([Figure 5](#f0025){ref-type="fig"}, *A* and *B*). Flow cytometry showed rAd-shPlk1 + Gem+LY could induce significant apoptosis in PC cells than the other groups (*P* \< .001) ([Figure 5](#f0025){ref-type="fig"}, *C* and *D*). As a result, we think that the targeted depletion of Plk1 could enhance chemosensitivity induced by LY294002 in PC.Figure 5Targeted depletion of Plk1 enhances chemosensitivity induced by PI3K inhibitor LY294002 in PC. (A) The Plk1 mRNA level detected by qRT-PCR shows Plk1 knockout significantly reduces the effect of Plk1 up regulating by gemcitabine in pancreatic cancer cell lines (*P* \< .001). (B) The Plk1 and apoptosis-related proteins Bcl-2 and Bax protein expression levels in human pancreatic cancer cell lines were shown by Western blotting analysis. GAPDH expression was used as internal controls. The Plk1 protein level under the same conditions is similar as its mRNA level. The Bcl-2 and Bax protein levels reveal that Plk1 knockout through shRNA enhances the chemosensitivity of combined chemotherapy and results in more cell apoptosis. (C) Pancreatic cancer cell apoptosis was detected by Flow Cytometry. Apoptosis rate in rAd-shPlk1 + Gem+LY group is higher than it in other four groups. (D) Statistical analysis of apoptosis rate detected by previous Flow Cytometry shows targeted depletion of Plk1 enhances chemosensitivity induced by PI3K inhibitor LY294002 in pancreatic cancer and results in higher cell apoptosis rate (*P* \< .01).Figure 5
Inhibition of PI3K/Akt Signal Pathway and Down-regulation of Plk1 Combined with Gemcitabine Inhibit the Growth of PC Xenografts in Nude Mice {#s0095}
--------------------------------------------------------------------------------------------------------------------------------------------
Finally, we investigated the effect of PI3K/Akt, Plk1 and gemcitabine in vivo through tumor xenograft model. We treated PC xenografts in nude mice with gemcitabine or LY294002 or BI2536(Plk1 inhibitor) or Gem+LY + BI2536. As shown in [Figure 6](#f0030){ref-type="fig"}, *A*--*C*, the tumor volume and the weight of Gem+LY + BI2536 group were significantly less than the other groups. Hematoxylin--eosin and IHC staining showed more necrosis and higher expression of Cleaved caspase-3 in Gem+LY + BI2536 group ([Figure 6](#f0030){ref-type="fig"}*E*). Meanwhile, Western blotting and TUNEL assay revealed Gem+LY + BI2536 induced PC cells apoptosis in the tumor xenograft model ([Figure 6](#f0030){ref-type="fig"}, *D* and *E*). Therefore, all the data above indicated that inhibition of PI3K/Akt signal pathway and down-regulation of Plk1 combined with gemcitabine could suppress PC growth and induced tumor necrosis and cell apoptosis in vivo.Figure 6Inhibition of PI3K/Akt signal pathway and down-regulation of Plk1 combined with gemcitabine inhibit the growth of PC xenografts in nude mice. (A, B and C) Gem+LY + BI2536 could diminish the tumor size and weight according to measurement of the resected tumors(^⁎⁎^, *P* \< .01). (D and E) Western blotting and Hematoxylin--eosin staining showed more necrosis and the highest expression of Cleaved caspase-3 in Gem+LY + BI2536 group. TUNEL assay showed the highest cell apoptosis rate in Gem+LY + BI2536 group than that in the control group, Gem+LY group and BI2536 group.Figure 6
Discussion {#s0100}
==========
PC is one of the cancers with the highest mortality due to its difficulty in detection at early stage and lack of effective treatment at middle and late stage [@bb0180]. Although being the first-line chemotherapy drug for PC, gemcitabine still has some disadvantages, such as low drug sensitivity and significant side effects. Thus, how to further improve the sensitivity of PC cells to gemcitabine still focuses and difficult subject in the field of PC-treatment. In recent years, deeply investigating the relationship between the key regulatory proteins of the DNA damage checkpoints and the signal pathways has become an import part in reversing the chemoresistance of tumors.
Plk1, as a crucial role in the regulation of cell proliferation, is up-regulated in the majority of PC cell lines and other human tumors [@bb0185]. Many studies have proved that even tumor metastasis and patients\' survival are correlated with its overexpression [@bb0190], [@bb0195], [@bb0200]. Virginie et al. found Plk1 is overexpressed in triple-negative breast cancer and high Plk1 expression is associated with poor prognosis within the entire population, Plk1 is a potential therapeutic option combine with chemotherapy for triple-negative breast cancer patients [@bb0205]. On the other hand, many chemotherapeutic drugs via p53-dependent cell cycle and apoptosis act, and Plk1 could inactivate of p53, then Plk1 contribute to chemoresistance [@bb0065], [@bb0210]. Our study showed Plk1 is overexpressed in PC tissues compare to normal tissues, and Plk1 expression level was correlated with pathologic grade, there were more PC tissues with Plk1 high expression in Pathologic grade III-IV than those in Pathologic grade I-II. Also, Weichert et al. observed that Plk1 is overexpressed in high-grade PanIN, the most important precursor lesion of PC and suggested that Plk1 overexpression is an early event in pancreatic cancer [@bb0070]. Therefore, it is not hard to imagine that Plk1 is associated with PC formation and progression. Then we constructed rAd-shPlk1, and found that depletion of Plk1 causes apoptosis in PC cells. Our earlier research showed that the Plk1 overexpression promoted cell proliferation, as well as an increase in G1/S phase cell percentage but a reduction in G2/M phase cell population [@bb0145]. Therefore, Plk1 is a potential therapeutic target for PC.
It is reported PI3K/Akt signal pathway controls G1 phase progression of PC cells regulating p21, SKP2, cyclin A, CDK2 etc. and after inhibition PI3K/Akt, SKP2 shows the most down-regulation, and PI3K/Akt inhibitor, LY294002, reduces Plk1 dephosphorylation following mitotic DNA damaging treatments, which suggesting that the PI3K pathway maybe involved in regulating Plk1 activity [@bb0205], [@bb0215]. Schlieman et al. found activation of Akt existed in half of PC cases [@bb0220]. Our present study found Akt level is the highest in PC tissues after the patient appeared gemcitabine-resistance, so the gemcitabine resistance to PC may be associated with abnormal activation of PI3K/Akt pathway, and LY294002 could increase chemosensitivity of PC cells to gemcitabine.
Both the PI3K/Akt pathway and Plk1 play a major role in regulating the chemoresistance of PC at the level of DNA damage checkpoint. Zhang et al. reported Plk1 acts upstream of the PI3K/Akt/mTOR pathway during oxidative stress [@bb0160]. However, in cycle, PI3K/Akt-dependent phosphorylation of Plk1-Ser99 is required for metaphase-anaphase transition [@bb0135]. During our study we observed PI3K/Akt inhibition promotes chemosensitivity of PC through downstream Plk1 inactivation. Liu et al. HEATR1 regulates Akt phosphorylation at Thr308 by promoting Akt-PP2AB56β interaction, then regulates chemotherapy [@bb0225]. Our previous study showed suppress Plk1 expression can induce cancer cell apoptosis through activating caspases pathway and decreasing Bcl-2/BAX ratio [@bb0230].
Next we aims to investigate the influence of PI3K/Akt pathway combine Plk1 on sensitization of PC-chemotherapy, and it represent suppression of Plk1 could enhance chemosensitivity induced by inhibition of PI3K/Akt in PC cells in vitro. At last, we verify our hypothesis in vivo, we used LY294002 (PI3K inhibitor), BI2536 (Plk1) inhibitor or gemcitabine treat PC xenografts nude mice, and the result shows inhibition of PI3K/Akt and down-regulation of Plk1 combined with gemcitabine inhibit the growth of PC xenografts in nude mice. However, a phase II/III randomized study showed the combination of rigosertib(a first-in-class Ras mimetic and small-molecule inhibitor of multiple signaling pathways including Plk1 and PI3K) + gemcitabine failed to demonstrate an improvement in survival or response compared with gemcitabine in patients with metastatic PC [@bb0235]. Both LY294002 and BI2536 phase I trials revealed that the drug was well tolerated with minor antitumor responses, and they also can enhance chemosensitivity to gemcitabine in PC cells [@bb0090], [@bb0240]. In summary, our findings confirm that suppression of PI3K/Akt and Plk1 combined gemcitabine could be a potential therapeutic schedule for PC patients, and further efforts need to be done.
Disclosure of Interest {#s0105}
======================
The authors declare that they have no conflicts of interest concerning this article.
This work was supported in part by grants from the National Natural Science Foundation of China (No. 30972910, 81172269) and Natural Science Foundation of Jiang Su Province, P. R. China (No. 050104313).
| {
"pile_set_name": "PubMed Central"
} |
As with most other athletic movements, the biomechanics of baseball pitching is studied to improve performance and prevent and/or rehabilitate injury. As technology in the sports science field has developed over the past 20 years, the interest has skyrocketed in using these advancements to the benefit of athletes. The initial studies provided accurate descriptions of the pitching kinematics and kinetics,^[@bibr7-1941738109338546],[@bibr9-1941738109338546][@bibr10-1941738109338546][@bibr11-1941738109338546][@bibr12-1941738109338546]-[@bibr13-1941738109338546],[@bibr25-1941738109338546]^ which helped athletes, coaches, medical professionals, and scientists understand the demands of pitching. Subsequent research has analyzed factors that correlate to performance enhancement and/or injury. The purpose of this review is to assimilate all the available scientific research on baseball pitching biomechanics related to performance and injury. This information is grouped into 5 areas: kinematics and its relationship to velocity; the association among kinematics, kinetics, and injury; the effects of fatigue; the development of a pitcher from youth to adult; and the effect of pitch types on mechanics. Over the years, research has been collected from different institutions with assorted methodologies, thereby making it difficult to compare numbers directly. Despite variance in numbers, the commonalities among pathomechanical patterns are most interesting.
Kinematics and Velocity {#section1-1941738109338546}
=======================
If you ask baseball coaches what elements make a pitcher effective, their responses will be "velocity" and "accuracy." Pitching coaches and biomechanists have studied the motion of elite pitchers to discern how they consistently throw fast pitches in the strike zone. Limited scientific research exists on the biomechanical factors that affect accuracy, but a lot is known about kinematic measures that improve ball velocity. Implicit in higher ball velocity are higher kinetic values for the elbow and shoulder.^[@bibr9-1941738109338546]^ Pitching kinematic variables affecting velocity are found in upper and lower body measures.
Much of the focus in the literature has been on the upper body, but the lower body is the foundation for baseball pitchers; pitching utilizes the kinetic chain to transfer energy from the lower body to the upper body. MacWilliams et al^[@bibr25-1941738109338546]^ performed one of the first biomechanical studies to examine the contributions of the lower body to pitching. They found that maximum linear wrist velocity (used as an indicator of ball velocity) correlated highly with the maximal push-off force of the throwing leg in the direction of the pitch. Montgomery and Knudson^[@bibr29-1941738109338546]^ demonstrated that decreases in stride length lowered velocity whereas increases in stride length increased velocity without affecting accuracy. The underlying mechanism was unknown. The push-off force supplies the initial forward momentum of the body, whereas the braking force that is applied by the lead leg during and after lead foot contact (FC) is actually the source of the energy that is transmitted up the body to maximize power output.^[@bibr25-1941738109338546]^ Matsuo et al^[@bibr26-1941738109338546]^ compared high- and low-velocity groups of pitchers and found significantly more lead knee extension angular velocity near the time of ball release (BR) in the high-velocity group. They hypothesized that a properly flexed lead knee at FC, approximately 38° to 50°,^[@bibr8-1941738109338546],[@bibr10-1941738109338546],[@bibr14-1941738109338546],[@bibr17-1941738109338546],[@bibr37-1941738109338546]^ stabilizes the lead leg for trunk rotation.
Assuming that the lead leg adequately flexes at FC and extends thereafter, the next links in the kinetic chain are the rotations of the pelvis and upper trunk. Escamilla et al^[@bibr9-1941738109338546]^ found that Americans had significantly greater maximum pelvis rotation velocity and ball velocity, compared to Korean pitchers. A critical component to maximizing the contribution of each link of the kinetic chain is the proper timing between the rotation of the pelvis and the rotation of the upper trunk. If too much lag or not enough occurs between the movements, the unique contributions of the 2 segments are lost.^[@bibr16-1941738109338546]^ If pitch cycle time is normalized such that 0% represents FC and 100% represents BR, the instant of peak pelvis rotation velocity is between 28% and 35%, and the instant of peak upper trunk rotation velocity is between 47% and 53%, with a separation of approximately 18% to 22%.^[@bibr6-1941738109338546],[@bibr9-1941738109338546],[@bibr17-1941738109338546],[@bibr26-1941738109338546]^ Although Matsuo et al did not directly measure this separation timing,^[@bibr26-1941738109338546]^ the high-velocity group had a separation-timing mean difference of 23%, whereas the low- velocity group had a mean difference of 17%. Stodden et al^[@bibr38-1941738109338546]^ also found, when analyzing pitcher variations, that the pelvis orientation at the times of maximum shoulder external rotation (MER) and BR and the proper rotational velocities of the pelvis and upper trunk translated into higher ball velocities.
The shoulder and elbow are the 2 joints that channel the significant power created by the lower body and trunk through the pitching arm. Because the shoulder complex has 3 degrees of freedom and the elbow, forearm, and wrist have 2 degrees of freedom, the throwing arm has many unique positional combinations. Finding the optimal arm path for a dynamic, explosive movement such as a baseball pitch becomes a daunting task for any athlete. Ball velocity has been correlated with shoulder positioning at the instant of FC. Increased ball velocity correlates with increased horizontal abduction^[@bibr9-1941738109338546],[@bibr37-1941738109338546]^ and decreased external rotation.^[@bibr9-1941738109338546],[@bibr42-1941738109338546]^ These correlations apply only within a reasonable range. Excessive horizontal abduction puts additional strain on the anterior capsule of the glenohumeral joint, and late external rotation may disrupt the timing of the arm path. For the higher-velocity group in a study by Escamilla et al,^[@bibr9-1941738109338546]^ horizontal abduction was 27° ± 10° and external rotation was 45° ± 19°. As the delivery moves into the arm-cocking phase, the amount of MER is linked to increased ball velocity.^[@bibr9-1941738109338546],[@bibr26-1941738109338546]^ During arm acceleration, pitchers with higher velocity reach peak shoulder internal rotation velocity closer to the instant of BR (102.3% time versus 104.4% time), optimizing the timing of arm acceleration and BR to maximize ball velocity.^[@bibr26-1941738109338546]^ At the instant of BR, the combination of shoulder abduction and lateral trunk tilt creates the pitcher's arm slot. Matsuo et al^[@bibr28-1941738109338546]^ conducted simulations based on biomechanical data to determine the optimal shoulder abduction angle at BR. It was traditionally taught that 90° maximizes functional stability.^[@bibr33-1941738109338546]^ Matsuo et al suggested a fairly narrow range centered on 90° that was self-optimized by selecting a comfortable lateral trunk tilt angle to maximize wrist velocity and, therefore, ball velocity.
Kinematics, Kinetics, and Injury {#section2-1941738109338546}
================================
Performance enhancement and injury prevention often go hand-in-hand in biomechanics. Pitchers occasionally sustain groin and abdominal muscle strains, as well as knee and back soreness, but the overwhelming number of injuries have been at the elbow and shoulder.^[@bibr5-1941738109338546]^ The instants of maximum shoulder external rotation and BR are critical for upper extremity kinetics analysis during pitching.^[@bibr13-1941738109338546]^ At least 7 kinetic variables have been implicated as mechanisms of injury.^[@bibr13-1941738109338546]^ During the arm-cocking phase, which ends at maximum shoulder external rotation, the throwing arm produces maximum anterior shoulder force, horizontal adduction torque, internal rotation torque, and elbow varus torque. During the arm acceleration phase (between MER and BR), maximum elbow flexion torque is achieved. Immediately after BR, when the arm begins to decelerate, maximum proximal shoulder force and proximal elbow force occur.^[@bibr13-1941738109338546]^
Injuries are most likely when high forces and/or torques are repeatedly applied to vulnerable tissue and when the pitcher transitions through susceptible positions. Fleisig^[@bibr12-1941738109338546]^ hypothesized 8 mechanisms that increase kinetic values and the risk of injury. Five of these mechanisms had significant correlations to increased kinetics. An open lead foot angle (for a right-handed pitcher, foot pointing toward left-handed batter) or an open foot position (for a right-handed pitcher, foot landing toward first-base side) at FC can cause the pelvis to rotate too soon. At FC, the normative mechanics are 19° ± 11° closed for foot angle, 19 ± 14 cm closed for foot position, and 30% ± 17% for the timing of maximum pelvis rotation velocity.^[@bibr17-1941738109338546]^ These improper lead foot mechanics and pelvis rotation produce additional anterior shoulder force and medial elbow force. The timing of shoulder rotation is also important. If there is insufficient or excessive shoulder external rotation at FC, the throwing arm may not be in correct position (thereby adding shoulder stress), or it will lag behind stressing the elbow. In either case, compensations can increase shoulder and elbow kinetics. During arm cocking, a pitcher who excessively adducts the shoulder horizontally (ie, leads with the elbow) increases anterior shoulder force, medial elbow force, and horizontal adduction shoulder torque. This pathomechanical pattern is seen in pitchers who have a compromised ulnar collateral ligament. Increased varus torque leads to increased ulnar collateral ligament strain.^[@bibr2-1941738109338546],[@bibr20-1941738109338546],[@bibr21-1941738109338546]^ As such, leading with the elbow lowers varus elbow torque, effectively reducing such strain. These elements may explain the development of shoulder injuries in pitchers with previous elbow injuries.
Although Newton's second law of motion dictates that increasing the acceleration of a constant mass will require an application of more force, the temporal sequencing of the kinetic chain during pitching makes this simple concept much more complex. Ascertaining the safe limit for each kinetic value is also difficult because many factors are dependent on one another (eg, anthropometrics, strength, flexibility, medical history). When assessing kinetic changes due to kinematic variability, the most reasonable approach is to determine which kinematic variables unnecessarily increase kinetic values. Changes in kinematics can increase or decrease velocity or not affect it at all. Clearly, any kinematic pattern that significantly increases kinetic values without increasing velocity is pathomechanical.
A simulation of shoulder abduction and lateral trunk tilt (the pitcher's arm slot) showed that if the values deviated from approximately 10° of lateral trunk tilt and 100° of shoulder abduction, maximum varus elbow torque increased.^[@bibr27-1941738109338546]^ With a shoulder abduction of 90°, linear wrist velocity was maximized and varus torque stayed relatively low.^[@bibr28-1941738109338546]^ Aguinaldo et al^[@bibr1-1941738109338546]^ showed that professional pitchers generated significantly less normalized shoulder internal rotation torque than that of college, high school, and youth pitchers. The researchers hypothesized that the professional pitchers were able to maximize their efficiency by rotating their upper trunks at the appropriate time, allowing the energy to pass from the trunk to the shoulder at precisely the right sequence. Although both pathomechanical patterns (pelvis and upper trunk rotation) make intuitive sense, Fleisig et al^[@bibr14-1941738109338546]^ did not find significant differences in the timing of the maximum pelvis or upper trunk rotation velocity. However, there were clinically significant separation timing differences between pelvis and upper trunk rotation: 11% and 12% in youth and high school pitchers, 17% and 18% in college and professional pitchers.
In a study comparing American and Korean pitchers,^[@bibr9-1941738109338546]^ there were no significant differences in shoulder and elbow force and torque, despite the fact that Americans threw the ball significantly faster (38.0 m/s to 34.6 m/s). The Koreans displayed 2 significant pathologic kinematic differences: greater shoulder external rotation at FC (68° to 45°) and less forward trunk tilt at BR (26° to 36°).
Fatigue {#section3-1941738109338546}
=======
The game of baseball has evolved in many ways, including the use of pitchers. In the early days of baseball, pitchers threw the entire game, regardless of score, inning, or number of pitches thrown. After generations of this approach, managers realized that it was counterproductive to continue to use a pitcher who was fatigued. Epidemiological and biomechanical studies have analyzed pitching to determine the effects of fatigue on performance and injury. The former have focused on pitch counts, whereas the latter have focused on kinematics and kinetics. Combining these 2 factors may be an effective strategy to determine exactly when a pitcher is fatigued.
Research has shown that several factors increase the risk of pain and injury in pitchers. Much of this work has focused on youth pitchers because arm pain is common at that level. In fact, roughly one-half of the 476 participants in a 2002 study of youth pitchers^[@bibr23-1941738109338546]^ reported elbow or shoulder pain at least once during a season. Lyman et al^[@bibr24-1941738109338546]^ found an increased risk of elbow and shoulder pain by pitchers with self-reported fatigue (5.94 times the risk for the elbow and 4.14 times the risk for the shoulder). The risk of pain increased if they threw more than 75 pitches per game (2.48, shoulder) and more than 600 pitches per season (3.44, elbow). This study supports the theory that high pitch counts lead to fatigue, which can in turn lead to injury. Olsen et al^[@bibr32-1941738109338546]^ reported compelling results of a direct comparison between adolescent pitchers who had elbow/shoulder surgery and those who did not. In this study, pitchers who averaged more than 80 pitches per appearance were nearly 4 times likely to require surgery. Those who pitched competitively more than 8 months per year were 5 times more likely to require surgery. Last, those who occasionally pitched with a fatigued arm were 4 times more likely to undergo surgery, whereas those who regularly pitched with a fatigued arm were 36 times more likely to have an injury that required surgery. Olsen et al concluded that overuse was the overriding factor in the development of arm pain.
Like most athletes, pitchers are generally reluctant to tell coaches that they feel fatigued, even when not telling might be detrimental to both the team and the player. A pitching coach's observational skills and judgment may be best suited to detect fatigue in the pitcher's mechanics and performance. A decline in ball velocity is typically seen in fatigued pitchers.^[@bibr8-1941738109338546],[@bibr31-1941738109338546]^ Specific mechanical flaws are also usually present in a fatigued pitcher. Murray et al^[@bibr31-1941738109338546]^ filmed pitchers during the first and last innings of games. In the last inning, pitchers achieved significantly less maximum shoulder external rotation and knee flexion at BR, and they threw 2 m/s slower (5 mph). Significantly less shoulder and elbow proximal force and shoulder horizontal abduction torque were applied. Escamilla et al^[@bibr8-1941738109338546]^ found that during simulated games, fatigued pitchers had a slightly more upright trunk position at BR. No significant differences in kinetics were found, although there was a drop in velocity of 1 m/s (2 mph) from the first inning to the last. During simulated pitching, Hirayama et al (unpublished data, 2008) found (1) a negative correlation between the number of pitches thrown in a game and lead hip extension work and (2) a positive correlation between the number of pitches thrown and shoulder horizontal adduction work---all of which suggests that pitchers rely less on the lower body and more on the arm as they fatigue.
Fatigue can affect motor control with losses in proprioception visualized by a significantly different arm path while throwing a baseball.^[@bibr41-1941738109338546]^ Significant losses of arm strength have been seen after pitching approximately 7 innings and throwing 100 pitches (shoulder flexion, 10%; internal rotation, 14%; humeral adduction, 12%; and grip strength, 8%).^[@bibr30-1941738109338546]^ Like pain, fatigue is generally difficult to quantify because it is a subjective measure that varies among persons. Therefore, pitch counts, ball velocity, ball location, pitching mechanics, and strength may be better guides in determining fatigue.
Development {#section4-1941738109338546}
===========
A majority of baseball players in the United States are younger than 18 years.^[@bibr36-1941738109338546]^ Therefore, the study of baseball pitching mechanics should be rooted in the development of youth pitchers from the time that they first pick up a baseball through high school. Collegiate and professional pitchers represent the most advanced in both skill and talent. Stodden et al^[@bibr39-1941738109338546],[@bibr40-1941738109338546]^ demonstrated that the initial acquisition of the overhead throwing skill is progressive, advancing from a single movement to a sequence of movements utilizing the body as a kinetic chain, which suggests that coordination is essential for developing throwing talent. Ishida et al^[@bibr19-1941738109338546]^ studied youth aged 6 to 12 years and demonstrated that players 9 and older displayed many of the biomechanical features of adult throwers. The researchers recognized that one possible limitation for the young thrower may be the weight of the baseball. Fleisig et al,^[@bibr18-1941738109338546]^ however, did not find significant differences between the arm paths of youth pitchers between 9 and 12 years who used lightweight and standard-weight baseballs. But the researchers' study did report significantly lower kinetic values with lightweight balls, suggesting that they may lessen the risk of an overuse injury. In another study, Fleisig et al^[@bibr14-1941738109338546]^ found few kinematic and temporal differences among youth, high school, college, and professional pitchers. Nearly all the kinetic values increased at each level of development.
Most complex skills such as baseball pitching take years of practice and thousands of repetitions to master. Fleisig et al^[@bibr15-1941738109338546]^ measured the change in variability in pitching biomechanics at different levels of development by comparing the standard deviations of relevant parameters. Their study found that variability of several kinematic parameters decreased as the level of development increased: foot placement, knee flexion at FC, maximum upper torso angular velocity, maximum elbow flexion, maximum shoulder external rotation, and forward trunk tilt at BR. Differences in kinetic parameters, however, were not significant. The researchers concluded that there was no increased risk of arm injury due to variability in pitching mechanics. They also noted that the largest changes occurred between youth and high school pitchers, and they emphasized the importance of teaching proper mechanics at an early age.
As youth pitchers approach physical maturity, their growing bodies are susceptible to a multitude of pathologies; therefore, the need to refine mechanics through repetition remains constant. During puberty, bones grow rapidly at the physes, which are weaker and therefore more susceptible to avulsion fractures. These fractures are most common during youth because ligaments and tendons are stronger than bones at the attachment sites.^[@bibr34-1941738109338546]^ Sabick et al^[@bibr34-1941738109338546]^ examined the stress that the adolescent proximal humeral epiphysis endures during the pitching motion. Their biomechanical testing suggested that the humeral epiphysis may experience more than 4 times the tolerable load of epiphyseal cartilage.
An extremely common diagnosis in young throwers is "Little League elbow." This inflammation at the medial epicondyle apophysis is the result of repetitive valgus overload during the cocking phase of pitching.^[@bibr4-1941738109338546]^ The medial epicondyle is the proximal attachment site for the ulnar collateral ligament, which can tear from repetitive stress.^[@bibr35-1941738109338546]^ These injuries have led youth organizations to adopt strict pitching policies to reduce the risk of injury to athletes.^[@bibr22-1941738109338546]^
Pitch Types {#section5-1941738109338546}
===========
Baseball legend Ted Williams is credited with saying that hitting a baseball is the most difficult thing to do in sports. One factor that makes hitting so challenging is the variety of pitches that a hitter must recognize---fastballs, cutters, curves, sliders, sinkers, changeups, and knuckleballs, to name a few. Theoretically, each has a unique trajectory that is controlled by pitching mechanics. Researchers in recent years have attempted to differentiate kinematic patterns among pitch types, as well as assess the kinetic values associated with each. A few studies compared the kinematics of common pitches in collegiate pitchers.^[@bibr3-1941738109338546],[@bibr10-1941738109338546],[@bibr17-1941738109338546]^ The largest differences were found between the fastball and the curveball, and the fewest were between the fastball and the slider. The curveball had significantly more forearm supination (32°) than that of the fastball (17°) and the changeup (18°).^[@bibr3-1941738109338546]^ Fastballs had significantly greater pelvis and upper trunk rotation velocities (600 and 1120 degrees per second, respectively) than curveballs (560 and 1070 degrees per second) and changeups (540 and 1020 degrees per second).^[@bibr17-1941738109338546]^ The lead knee extended 9° from FC to BR in the fastball and 5° in the curveball, flexing 4° during the changeup deliveries. Pitchers landed with their lead foot 4 cm more closed (for a right-handed pitcher, toward the third-base side of the mound) when throwing curveballs versus fastballs. Maximum elbow extension and shoulder internal rotation velocity (ie, arm speed) were similar between the fastball (elbow, 2210 degrees per second; shoulder, 6520) and curveball (2160 and 6480 degrees per second) but significantly slower in the changeup (1970 and 6360 degrees per second).
The forces and torques experienced by the elbow and shoulder joints during various pitch types are important for understanding the mechanics and potential injury risk of each pitch type. In collegiate pitchers, unique kinetic patterns for the fastball, curveball, changeup, and slider have been detected.^[@bibr17-1941738109338546]^ Six of the 9 kinetic variables were significantly lower in the changeup versus the fastball. The fastball was kinetically similar to the curveball; only proximal elbow force was significantly higher in the fastball. Kinetics were also similar between the fastball and the slider, although sliders were thrown with significantly higher horizontal shoulder adduction torque. These kinetic investigations were partially based on anecdotal evidence targeting the curveball as a dangerous pitch for younger pitchers. Dun et al^[@bibr6-1941738109338546]^ analyzed the kinematics and kinetics of 10- to 14-year-olds throwing fastballs, curveballs, and changeups. The fastball had significantly higher values than those of the curveball for elbow varus torque (35 to 32 N·m), shoulder internal rotation torque (35 to 32 N·m), elbow flexion torque (16 to 14 N·m), proximal elbow force (462 to 428 N), and proximal shoulder force (466 to 433 N). Neither the curveball nor the changeup had significantly higher values than those of the fastball, implying that curveballs were not more stressful than fastballs.
At all levels of competition, a good fastball is the foundation for successful pitching; thus, the young baseball pitcher should master the fastball first. The changeup seems to be a good choice for a second pitch, given that it produces lower kinetics in the elbow and shoulder.
######
Summary of pathomechanics associated with increased kinetics and decreased ball velocity.^[a](#table-fn1-1941738109338546){ref-type="table-fn"}^
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Phase / Event Proper Mechanics Pathomechanics → Consequences
----------------------------------------------------------------- -------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Windup Lift front leg.
Maximum knee height Pitcher is balanced.
Stride Front leg goes down and forward.\ ↓ Push off rubber → ↓ Ball velocity^[@bibr24-1941738109338546]^
Arms separate, swing down, and up.
Foot contact Front foot is planted slightly to third-base side (for a right-handed pitcher).\ ↓ Stride length → ↓ Ball velocity^[@bibr28-1941738109338546]^\
Front foot is pointed slightly inward.\ Front foot open (position or angle) → ↑ Shoulder and elbow force^[@bibr12-1941738109338546]^\
Shoulder is abducted approximately 90°, with approximately 60° of external rotation. Improper shoulder external rotation → ↑ Shoulder and elbow kinetics^[@bibr12-1941738109338546]^\
Excessive shoulder external rotation → ↓ Ball velocity^[@bibr9-1941738109338546],[@bibr40-1941738109338546]^\
↓ Shoulder horizontal abduction → ↓ Ball velocity^[@bibr11-1941738109338546],[@bibr36-1941738109338546]^
Arm cocking Pelvis rotation, followed by upper trunk rotation.\ Early pelvis rotation → ↓ Ball velocity^[@bibr35-1941738109338546],[@bibr40-1941738109338546]^\
Shoulder externally rotates, and trunk arches. Late pelvis rotation → ↑ Shoulder and elbow kinetics^[@bibr40-1941738109338546]^\
↓ Pelvis rotation velocity → ↓ Ball velocity^[@bibr9-1941738109338546],[@bibr40-1941738109338546]^\
Poor timing between pelvis rotation and upper trunk rotation →↓ Ball velocity^[@bibr23-1941738109338546],[@bibr35-1941738109338546]^\
Poor timing between pelvis rotation and upper trunk rotation →↑ Shoulder internal rotation torque^[@bibr1-1941738109338546]^
Maximum external rotation Shoulder external rotation is approximately 180°.\ ↓ Shoulder external rotation → ↓ Ball velocity^[@bibr9-1941738109338546],[@bibr23-1941738109338546],[@bibr29-1941738109338546]^\
Elbow flexion is approximately 90°. Excessive shoulder horizontal adduction and elbow flexion → ↑ Shoulder kinetics^[@bibr12-1941738109338546]^
Arm acceleration Elbow extends, followed by shoulder internal rotation.\
Front knee extends.
Ball release The throwing shoulder is abducted approximately 90° ↓ Knee extension velocity → ↓ Ball velocity^[@bibr23-1941738109338546]^\
Improper shoulder abduction → ↓ Ball velocity^[@bibr25-1941738109338546]^\
Improper shoulder abduction → ↑ Elbow varus torque^[@bibr26-1941738109338546]^\
↓ Forward trunk tilt → ↓ Ball velocity^[@bibr8-1941738109338546],[@bibr9-1941738109338546]^
Arm deceleration Shoulder internal rotation and front knee extension continue.\
Trunk tilts forward.
Maximum internal rotation Shoulder external rotation is approximately 0°.
Follow through Arm crosses in front of body.Trunk flexes forward.
Ball release The throwing shoulder is abducted approximately 90° ↓ Knee extension velocity → ↓ Ball velocity^[@bibr23-1941738109338546]^Improper shoulder abduction → ↓ Ball velocity^[@bibr25-1941738109338546]^Improper shoulder abduction → ↑ Elbow varus torque^[@bibr26-1941738109338546]^↓ Forward trunk tilt → ↓ Ball velocity^[@bibr8-1941738109338546],[@bibr9-1941738109338546]^
Arm deceleration Shoulder internal rotation and front knee extension continue.Trunk tilts forward.
Maximum internal rotation Shoulder external rotation is approximately 0°.
Follow through Arm crosses in front of body.\
Trunk flexes forward.
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Definition of symbols in chart: ↓, decreased; ↑, increased; →, correlates with.
Summary {#section6-1941738109338546}
=======
Knowledge of the mechanics that can improve performance and prevent injury is an invaluable resource for doctors, athletic trainers, therapists, coaches, and athletes. Increased ball velocity has been seen with proper separation timing between the pelvis and upper trunk,^[@bibr26-1941738109338546],[@bibr38-1941738109338546]^ with greater maximum shoulder external rotation,^[@bibr9-1941738109338546],[@bibr26-1941738109338546],[@bibr31-1941738109338546]^ with greater knee extension velocity,^[@bibr26-1941738109338546]^ and with more forward trunk tilt at BR.^[@bibr8-1941738109338546],[@bibr9-1941738109338546]^ Some of the mechanics that lead to additional stress on the arm include an open foot position or angle,^[@bibr12-1941738109338546]^ too much or too little shoulder external rotation at FC,^[@bibr12-1941738109338546]^ poor timing between the separation of the pelvis and upper trunk rotations,^[@bibr26-1941738109338546],[@bibr38-1941738109338546]^ and shoulder abduction angle deviating from 90° at BR.^[@bibr27-1941738109338546]^
Complementing this knowledge of mechanics with the known effects of fatigue, growth and development, and pitch types enhances our understanding of the demands of pitching. Executing proper, repeatable fastball mechanics is the fundamental skill that all pitchers must learn first. Although curveballs have not been shown to have increased kinetic values over fastballs, it may be wisest to teach the changeup as a second pitch because of the reduced amount of stress that it places on the arm.^[@bibr6-1941738109338546],[@bibr10-1941738109338546]^ Pitchers need to be conscious of their fatigue and pain levels at all times and make a concerted effort to avoid pitching when either of these become uncomfortable. Resting between 3 and 4 months between seasons is also advised.^[@bibr32-1941738109338546]^
***NATA Members*:** *Receive* ***3 free CEUs*** *each year when you subscribe to* Sports Health *and take and pass the related online quizzes! Not a subscriber? Not a member? The* Sports Health*--related quizzes are also available for purchase. For more information and to take the quiz for this article, visit [www.nata.org/sportshealthquizzes](http://www.nata.org/sportshealthquizzes)*.
No potential conflict of interest declared.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-ijerph-15-01734}
===============
Poultry slaughterhouses discharge a significant volume of highly polluted wastewater, principally during the slaughtering process and the periodic washing of residual particles, which cause a significant variation in the biodegradable organic matter concentration. Organic matter is considered the primary pollutant in the effluents of slaughterhouses \[[@B1-ijerph-15-01734]\]. The contribution of organic load to these effluents usually comes from different materials such as undigested food, blood, fat and lard, loose meat, paunch, colloidal particles, soluble proteins, and suspended materials \[[@B2-ijerph-15-01734],[@B3-ijerph-15-01734]\]. Due to the mentioned components in the slaughterhouses wastewater, these wastewaters have a high concentration of organics such as chemical oxygen demand (COD), biochemical oxygen demand (BOD), phosphorous, and nitrogen \[[@B4-ijerph-15-01734]\]. Therefore, before discharging these wastewaters into receiving water bodies, an efficient treatment process should be carried out to prevent severe environmental pollution. In the last few decades, several treatment methods for the slaughterhouse wastewater have been reported. Biological (aerobic and anaerobic) treatment methods have been traditionally used for slaughterhouse wastewater treatment. However, both biological techniques have some limitations. For example, aerobic treatment processes require high energy consumption for aeration and generate a high amount of sludge \[[@B1-ijerph-15-01734]\]. The anaerobic treatment process of the poultry slaughterhouse wastewater is often impaired or slowed down because of the accumulation of suspended solids and floating fats in the reactor, which in turn leads to reduction in methanogenic activity and biomass washout \[[@B2-ijerph-15-01734]\]. Moreover, the anaerobic treatment process is more suitable in treating high organic loading wastewater \[[@B5-ijerph-15-01734],[@B6-ijerph-15-01734]\].
Sequential Batch Reactors (SBR) are one of the biological processes applied to remove several types of pollutants. The SBR process is different from conventionally activated sludge techniques, because SBR merges all treatment units and operations into a single basin or tank, whereas traditional systems rely on various tanks. SBR has been successfully used for the treatment of domestic, municipal, industrial, dairy, synthetic, toxic and slaughterhouse wastewaters, swine manure, and landfill leachates \[[@B7-ijerph-15-01734],[@B8-ijerph-15-01734],[@B9-ijerph-15-01734],[@B10-ijerph-15-01734],[@B11-ijerph-15-01734],[@B12-ijerph-15-01734]\].
Recently, the application of biomass carriers in the SBR process has been investigated by various researchers \[[@B13-ijerph-15-01734],[@B14-ijerph-15-01734],[@B15-ijerph-15-01734]\]. Fiber-based biomass carriers exhibit a good performance in removing pollutants, especially nitrogenous substances \[[@B16-ijerph-15-01734],[@B17-ijerph-15-01734]\]. Previous studies that applied the swim bed technologies in SBR using bio-fringe (acryl fiber) revealed high treatment efficiency in removing pollutants, especially nitrogenous substances \[[@B17-ijerph-15-01734]\]. Several types of fibers have been used previously in wastewater treatments, such as plastic fibers \[[@B18-ijerph-15-01734],[@B19-ijerph-15-01734]\], geotextiles \[[@B20-ijerph-15-01734]\], bio fringe acryl fiber \[[@B17-ijerph-15-01734]\], fibrous packing \[[@B21-ijerph-15-01734]\], and polyester fiber \[[@B22-ijerph-15-01734]\]. However, the application of fibers as attachment materials in SBR for poultry slaughterhouses wastewater treatment has not been well investigated.
The aim of this paper is to examine the potential use of various types of fibers as biomass carriers for slaughterhouses wastewater treatment by evaluating the removal efficiency of the pollutants with and without fiber in the reactor. The fibers involved are natural white Jute fiber (JF), synthetic siliconised conjugated polyester fiber (SCPF), bio-fringe (acrylic fiber) (BF), and the combination of three fibers in the reactor, called composite fiber (CF). The treatment efficiency of the different reactors with and without fibers on BOD, COD, ammonia-nitrogen (NH~3~-N), phosphorus (P), nitrite (NO~2~-N), nitrate (NO~3~), TSS, and oil-grease were evaluated. Parameters, such as BOD, COD, and NH~3~-N, were monitored every day during the experiments. However, the other parameters were evaluated based on the optimum value obtained.
2. Materials and Methods {#sec2-ijerph-15-01734}
========================
2.1. Wastewater Source and Characteristics
------------------------------------------
The wastewater used in this study was collected from a local poultry slaughterhouse plant with a 13,000 birds per day capacity, located in the city of Nibong Tebal, Penang state, Malaysia, generating approximately 140 tons of wastewater daily. This wastewater, which is produced from different operations such as chickens cutting, chilling, scalding, packing and plant cleanup, was collected from the final collection tank after the screening of internal organs and feathers (partially treated using physical treatment). Wastewater samples of 150 to 200 L were collected twice per week, during the period from 23 May 2012 to 11 March 2013. Following the sampling procedure, the wastewater samples obtained were characterized based on pollutant concentration. Samples were preserved by storing in a cold room at 4 °C and were only taken out to room temperature 2 h before the experiment began. Characteristics of the raw wastewater are shown in [Table 1](#ijerph-15-01734-t001){ref-type="table"}.
2.2. Activated Sludge and Characteristics {#sec2dot2-ijerph-15-01734}
-----------------------------------------
The activated sludge (AS) used in this study was collected from the sludge dewatering system at the Jelutong Sewerage Treatment Plant (JSTP), Penang State, Malaysia. The AS in this study acts as microorganisms that are responsible for transforming the pollutants into acceptable end products. The AS also followed the poultry slaughterhouses wastewater storing procedures. Characteristics of the AS are shown in [Table 1](#ijerph-15-01734-t001){ref-type="table"}.
2.3. Fiber Preparation {#sec2dot3-ijerph-15-01734}
----------------------
Three types of fibers were used in this study as mentioned earlier. The first type was bio-fringe (BF) fiber made of acrylic fiber and imported from Japan. The other two types were Jute fiber (JF) and siliconised conjugated polyester fiber (SCPF). Composite fibers (CF) are a combination of these three fibers where all types were put together in the reactor. Both JF and SCPF were prepared similar to the size of the ready-made BF. The fibers were sewed neatly into pieces of yarns. [Table 2](#ijerph-15-01734-t002){ref-type="table"} shows the physical properties of the fibers.
2.4. Reactor Setup {#sec2dot4-ijerph-15-01734}
------------------
Two identical, laboratory scale Plexiglas reactors were used as SBR reactors for this study. Each reactor has the following dimensions: 80 cm × 40 cm × 25 cm with a total volume of 80 L. However, the experimental volume of the liquid for each reactor was 60 L. The first reactor was only operated with activated sludge without adding the fibers, while the other reactor was operated with activated sludge in the presence of fibers. [Figure 1](#ijerph-15-01734-f001){ref-type="fig"} shows the schematic diagram of the SBR reactor. The first cycle started with seeding of the AS collected from the JSTP. Following this, the reactor was fed with the collected raw poultry slaughterhouse wastewater during the filling phase and was aerated and mixed for a certain period of time during the aerating phase. The pH was adjusted approximately to 7.0 ± 0.5 and the mixed liquor suspended solids (MLSS) were maintained at a minimum range of 1500 mg/L to 4000 mg/L during the whole experiment. The adjustments were conducted before the aeration phase. The pH value was adjusted by adding either acid (0.5 M of H~2~SO~4~) or base solutions (0.5 M of NaOH). A 24-h cycle was selected, and the wastewater was operated for 20 h with the aeration rate of 60 L/min to make sure the wastewater and AS were mixed homogeneously. The final MLSS was 3782 mg/L. An air pump was used for the aeration and water circulation in the reactors. The aerated phase was stopped at the end of the aeration phase (after 20 h) and before the start of the settling phase (3 h). The decanting and discharging phase was the last process in the cycle, which meant that a cycle had been completed. After the first cycle was completed, the SBR reactor was filled with raw poultry slaughterhouse wastewater, aerated, settled, and decanted to repeat the second day treatment. [Table 3](#ijerph-15-01734-t003){ref-type="table"} summarizes the operation design parameters of SBR reactor.
2.5. Operating Conditions {#sec2dot5-ijerph-15-01734}
-------------------------
To maintain 1500 mg/L of MLSS, the poultry slaughterhouse wastewater feed was set at 21 L/day. The concentration of MLSS was checked during every aeration phase (1.00 p.m.). The ratio of food to microorganism (F/M) was set at 0.2, where F refers to BOD (mg) applied per day to the reactor and M refers to TSS (mg) in the reactor. The F/M ratio is an important parameter that represents the amount of substrate available for the microorganisms in activated sludge. A typical F/M ratio for SBR ranges from 0.04 to 0.1 \[[@B23-ijerph-15-01734]\], whereas for SBR nutrient removal it ranges from 0.2 to 0.6 \[[@B24-ijerph-15-01734]\]. Either a too low or too high value of F/M may cause filamentous bulking or foaming, which leads to poor settlebility. AS was added if the concentration of MLSS was below 1500 mg/L, which meant that the F/M ratio during the reactors operation was maintained below 0.25. Filamentous bulking might occur if the MLSS exceeded 4000 mg/L. In the case where the MLSS exceeded 4000 mg/L, some sludge would be wasted until the MLSS dropped to the desired level (1500--4000 mg/L). Meanwhile, the hydraulic retention time (HRT) and sludge retention time (SRT) have been calculated and kept at 72 h and 176 h, respectively. The reactors were operated at room temperature (25 °C) without any temperature controlling system. Both Filling and decanting processes were also conducted manually without any pumping system. Several experiment runs with the different types of fibers were carried out in order to achieve the objectives of the study. The filling period was 30 min, the aeration period was 20 h, and the settling period was 3 h.
2.6. Analytical Methods {#sec2dot6-ijerph-15-01734}
-----------------------
In this study, the performance of the reactors was evaluated based on the values of BOD, COD, NH~3~-N, NO~2~, NO~3~, Phosphate, Oil-grease, TSS, and color. All of the mentioned parameters were determined and carried out as described in the Standard Methods for Water and Wastewater Examination \[[@B25-ijerph-15-01734]\]. The COD concentration was measured using a DR 2800 Spectrophotometer while NH~3~-N and PO~4~^3−^ concentration were calculated using a Hach DR 2500 Spectrophotometer. The removal efficiency of BOD, COD, and color was calculated using the following formula: where *C~i~* and *C~f~* are the initial and final concentrations of parameters, respectively. Results from this study were analyzed using the Statistical Package for Social Sciences (SPSS) Version 17, via a one-way analysis of variance (One-Way ANOVA).
3. Results and Discussion {#sec3-ijerph-15-01734}
=========================
3.1. BOD and COD Removal {#sec3dot1-ijerph-15-01734}
------------------------
During the experiment, the pH of the activated sludge reactor was maintained at around 7 ± 0.5, the optimum range of pH for microbial growth. The MLSS was maintained in the system in the range between 1500 to 4000 mg/L.
The initial concentration of COD was 950 mg/L before treatment. Maximum removal efficiency of COD was achieved in day 12 with 69% and 293 mg/L COD. For the BOD value, the initial concentration was 350 mg/L where the maximum removal was achieved in day 13 with 88% removal efficiency with 105 mg/L BOD after treatment in the reactor without fibers. The growth of microbes started to become slow and stable (between days 11 to day 16) because the microbes did not have any shelter to regenerate before the cycle completes. As compared to the reactor with fibers, the reactor without fibers does not have shelter for microbes to attach.
On the other hand, after using the different types of fibers in the reactor, the removal efficiency of BOD and COD increased over the time in all reactors. As an overall result, the performance of SBR using fibers as an attachment material has demonstrated better results compared to SBR without fibers in the reactor. In general, the BOD removal for each type of fiber was efficient, with an average removal efficiency of higher than 90%. The observations showed that the CF reactor gives the higher removal efficiency of BOD with 96% (40 mg/L). For the JF and BF reactors, the optimum BOD removal was achieved on day 14 and day 12 with 62 mg/L (93%) and 45 mg/L (95%), respectively. Apart from that, the SCPF reactor showed optimum removal on day 13 with 50 mg/L BOD (94%).
For COD, the BF reactor showed a higher performance in the removal efficiency of COD with 93%. The COD removal achieved maximum removal on day 12 at 93% with 45 mg/L COD as shown in [Figure 2](#ijerph-15-01734-f002){ref-type="fig"}.
3.2. Ammonia-Nitrogen Removal {#sec3dot2-ijerph-15-01734}
-----------------------------
The daily observations of the NH~3~-N showed that there was no removal for the first two days of treatment as shown in [Figure 2](#ijerph-15-01734-f002){ref-type="fig"}. This observation was due to the increase in the concentration of ammonia-nitrogen due to the occurrence of nitrification \[[@B26-ijerph-15-01734]\]. According to the Fontenot et al. \[[@B27-ijerph-15-01734]\], in SBR, the nitrogen (protein or lipid) removal process was designed for the aerobic carbon removal and nitrification followed by an anoxic de-nitrification with the addition of an external carbon source.
Before using the fiber in reactors, the maximum removal was achieved on day 13, with the removal of 77% and concentration pollutant of 20 mg/L. It was observed that when the pH of the wastewater in the system was not basic, an oxidation of ammonia took place.
When the oxidation occurred, the pH of the wastewater quickly dropped, and this simultaneously produced nitrite. Ammonia can exist as molecular ammonia or ammonium gas. These two forms in water are strongly dependent on pH and temperature. Nitrification and de-nitrification occurred in good condition due to the aerobic and anaerobic zones inside the same system. The higher concentrations of ammonia were shown to have inhibitory effects of pH level during the anaerobic process \[[@B28-ijerph-15-01734]\].
On the other hand, and after using the fiber with SBR reactors, the NH~3~-N removal was considered to be high at 94%, 94%, 93% and 94% for CF, JF, BF and SCPF reactors, respectively. In the case of CF reactor, during the first and second day of treatment, no removal due to the increase in ammonia-nitrogen concentration from 85 mg/L to 98 mg/L (data not shown). However, in day 3 and 4 of treatment process, a fluctuation in removal efficiency was observed. The maximum removal efficiency was obtained at day 13 at 94%. The initial concentration of ammonia-nitrogen before treatment was 106 mg/L and diminished 5 mg/L after treatment. In general, biological nitrogen removal can be categorized into two separate steps: nitrification and denitrification. In the nitrification process, ammonium is usually converted to nitrate under aerobic condition, whereas the de-nitrification process converted the nitrate into nitrogen gas (N~2~) \[[@B29-ijerph-15-01734]\]. Therefore, when a higher aeration was used, a better removal efficiency of NH~3~-N was observed due to the fact that more DO was provided to the nitrifying bacteria in order to convert ammonium to nitrate.
3.3. Nitrite (NO~2~) and Nitrate (NO~3~) Removal {#sec3dot3-ijerph-15-01734}
------------------------------------------------
As can be seen in [Figure 2](#ijerph-15-01734-f002){ref-type="fig"}, the maximum removal efficiency for NO~2~ was achieved on day 8 with the removal of 45% in the reactor without fiber. According to Erses et al. \[[@B28-ijerph-15-01734]\], when oxidation ammonia takes place, nitrite is produced simultaneously.
However, the maximum removal efficiency for NO~3~ was achieved on day 13 with the removal of 85% when the remaining concentration in the reactor was 12 mg/L. After using the fibers in the reactors, the maximum removal efficiency for NO~2~ obtained in the JF reactor was 84%, which is approximately twice the removal efficiency of the reactor without fibers. In addition, the maximum removal efficiency for NO~3~ was 94%, which was achieved using SCPF reactor.
3.4. Phosphate Removal
----------------------
The maximum removal efficiency for phosphate was achieved on day 13 with a removal efficiency of 61%, where the remaining concentration of 15 mg/L in the case of reactor without fiber, as shown in [Figure 2](#ijerph-15-01734-f002){ref-type="fig"}. The removal showed a fluctuated trend on day 6, due to the increased concentration of nitrate in the anaerobic zone and phosphate in the aerobic zone (data not shown). The phosphorus content in the sample may also be affected by the apparatus used during the experiment if the apparatus was contaminated with detergent \[[@B30-ijerph-15-01734]\]. However, after using the fibers in reactors, the maximum removal efficiency of phosphate was 85% for the BF reactor during the experiment. No removal was obtained during the first day due to the increase of phosphate concentration from 39 mg/L to 65 mg/L. This was owing to the occurrence of the polyphosphate bacteria which started to accumulate large quantities of phosphate within their cells \[[@B31-ijerph-15-01734]\].
3.5. Oil-Grease and TSS Removal
-------------------------------
In the reactor without fiber, the maximum removal efficiency for oil-grease was achieved on day 13 with 57% (117 mg/L). In addition, the same trend was observed for the TSS values. The maximum removal of TSS was recorded on day 13 with the remaining concentration at 72 mg/L (84%) from 532 mg/L. After applying the fibers in reactors, the maximum removal efficiency of the oil-grease and TSS was obtained using a JF reactor and BF reactor with 86% and 97% removal efficiency, respectively, as shown in [Figure 2](#ijerph-15-01734-f002){ref-type="fig"}. The pattern of removal showed the fluctuated trend until day 4 of the treatment. Starting from day 5, the treatment slowly showed an increase in the removal over time. Maximum removal efficiency for oil-grease was obtained on day 11. The TSS removal efficiency showed an increase trend over that time. A maximum removal efficiency of TSS was achieved on day 13. The fluctuated trend that occurred in the early part of the treatment may be due to the fact that a pump was not used when the sample was taken. This led to the additional concentration of oil and grease and TSS. From the observation and the properties of the raw wastewater, the value of oil-grease increased because of not using the pump during the sample collection. The oil-grease may have been partially filtered using the pump.
3.6. Statistical Analysis {#sec3dot6-ijerph-15-01734}
-------------------------
[Table 4](#ijerph-15-01734-t004){ref-type="table"} summarized the maximum removal efficiency for all SBR. A One-way (ANOVA) test was used for multiple comparisons between the different types of the reactors. The test shows a significant difference between mean concentration BF and AS, CF, JF, and SCPF for all parameters tested. For example, for the BOD concentration after treatment, the ANOVA test shows that the concentration of BOD in the CF reactor was small as compared to the other reactors, which emphasize the experimental results. In addition, the BF reactor has a lower mean concentration for COD as compared to other reactors, which means that the BF reactor achieved the higher removal efficiency, proving the experimental results. For other parameters, similar results with experiments were obtained. The characteristics of final effluents for all reactors are summarized in [Table 5](#ijerph-15-01734-t005){ref-type="table"}.
4. Conclusions {#sec4-ijerph-15-01734}
==============
This paper investigates the potential application of different types of fibers as biomass carriers in SBR reactors for slaughterhouses wastewater treatment, by evaluating the removal efficiency of the pollutants with and without fibers in a SBR reactor. The study showed that the removal efficiency of the SBR with fibers achieved a higher performance for all tested parameters as compared with the SBR without fibers. The SBR reactor with fibers as attachment materials enabled the attachment of suspended solids to the fibers, which increased the biomass concentration in the reactor and provided a better treatment efficiency. The treated effluent had 40 mg/L BOD and 45 mg/L COD with an average removal efficiency of 96% and 98%, respectively, which meet the discharge limits stated in the Environmental Quality Act in Malaysia. Moreover, all other parameters also satisfied the limits of the Environmental Quality Act in Malaysia.
This work was funded by Universiti Sains Malaysia under Iconic grant scheme (Grant No. 1001/CKT/870023) for research associated with the Solid Waste Management Cluster, Engineering Campus, Universiti Sains Malaysia.
H.A.A. conceived and designed the experiments; N.N.A.P. performed the experiments and data analysis; M.Y.D.A. wrote and revised the paper, Y.-T.H. revised and proofread the paper.
This research was funded by \[University Sains Malaysia\] grant No. \[Grant No. 1001/CKT/870023\].
The authors declare no conflict of interest.
{#ijerph-15-01734-f001}
######
Daily monitoring for the removal efficiency of (**a**) BOD, (**b**) COD, (**c**) ammonia-nitrogen, (**d**) phosphate, (**e**) NO~2~, (**f**) NO~3~, (**g**) Oil-grease, (**h**) TSS, in all reactors.
ijerph-15-01734-t001_Table 1
######
Characteristics of raw wastewater and activated sludge (No. of samples = 7).
Parameter Min Max Average Std. Dev.
---------------------- -------- -------- ---------- -----------
**Raw wastewater**
BOD (mg/L) 573 1177 875 427.09
COD (mg/L) 777 1825 1301 741.04
NH~3~-N (mg/L) 56.7 104 80.35 33.44
Nitrite (mg/L) 45.3 80 62.65 24.53
Nitrate(mg/L) 52.6 178.4 115.5 88.95
Oil-grease (mg/L) 2361.5 3616 2988.75 887.06
TSS (mg/L) 395 783 589 274.35
pH 6.3 6.9 6.6 0.4242
**Activated sludge**
BOD (mg/L) 1246 1548 1397 213.54
COD (mg/L) 51,248 59,345 55,296.5 5725.44
DO (mg/L) 0.65 0.68 0.665 0.0212
MLSS (mg/L) 47,000 59,000 53,000 8485.28
pH 6.75 6.85 6.8 0.0707
BOD: biochemical oxygen demand; COD: chemical oxygen demand; TSS: total suspended solids; DO: Dissolved oxygen; MLSS: Mixed liquor suspended solids.
ijerph-15-01734-t002_Table 2
######
Fiber characteristics.
Characteristic Types of Fibers
------------------------------ ----------------- ---------- ---------- ----------
Support filament length (cm) 50 50 50 50
Yarn length (cm) 10 ± 0.5 10 ± 0.5 10 ± 0.5 10 ± 0.5
Yarn diameter (cm) 0.1--0.2 0.5--1.0 1.0--1.5 0.1--1.0
Yarn per string 65 20 20 20
Total weight (g) 19.3 52.65 42.15 114.1
BF: bio-fringe; JF: Jute fiber; SCPF: siliconised conjugated polyester fiber; CF: composite fibers.
ijerph-15-01734-t003_Table 3
######
Summary of operating design parameters.
Parameters Value
--------------------------------------- ------------------
Volume (L) 80
Operating liquid volume (L) 60
Dimension (m) 0.4 × 0.4 × 0.25
Hours per cycle (h) 20
F/M ratio (day^−1^) 0.2
MLSS (mg/L) 1500
Feeding rate (L/day) 21
Hydraulic Retention Time (HRT) (days) 2.9
Sludge Retention Time (SRT) (days) 7.5
Temperature (°C) 25
Sludge Volume Index (SVI) (mg/L) 50
MLSS: mixed liquor suspended solids.
ijerph-15-01734-t004_Table 4
######
Summary of maximum removal efficiencies for all reactors.
Max. Removal Efficiency (%)
----------------------------- ---- ---- ---- ---- ----
BOD 88 95 93 94 96
COD 69 98 93 91 90
NH~3~-N 77 93 94 94 96
Phosphate 61 85 72 74 79
NO~2~ 45 81 84 69 71
NO~3~ 86 92 93 94 93
TSS 84 97 93 92 96
Oil-grease 57 73 86 74 77
BF: bio-fringe; JF: Jute fiber; SCPF: siliconised conjugated polyester fiber; CF: composite fibers.
ijerph-15-01734-t005_Table 5
######
Summary of final effluents of all parameters for all reactors.
Parameters Without Fiber BF JF SCPF CF
---------------- --------------- ----- ----- ------ -----
BOD (mg/L) 120 70 65 65 65
COD (mg/L) 309 74 75 112 150
NH~3~-N (mg/L) 23 7 10 7 7.9
P (mg/L) 17 10 14 16 16
NO~2~ (mg/L) 51 30 20 37 40
NO~3~ (mg/L) 14 12 10 10 12
TSS (mg/L) 75 25 39 55 30
Oil-grease 855 801 431 751 723
BF: bio-fringe; JF: Jute fiber; SCPF: siliconised conjugated polyester fiber; CF: composite fibers.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
Articular cartilage is a highly specialized and resilient connective tissue that functions to distribute physiological loads to the underlying bone without developing unacceptably high stresses^[@CR1]^. Cartilage function can be altered by degenerative changes in its structure and composition, often as a consequence of injury. Cartilage injuries may result in acute lesions, which without intervention may progress to post-traumatic osteoarthritis (PTOA)^[@CR2]^. Several techniques are available for cartilage repair, with recent research suggesting that pharmaceutical interventions may be effective in preventing the onset or halting the progression of PTOA if the injury is detected early^[@CR2],\ [@CR3]^. Thus, characterization of cartilage integrity and disease progression at the early disease stages is crucial for effective management and treatment of PTOA^[@CR2]^.
Current clinical diagnosis of joint pathologies often involves clinical examination, with radiographic (X-ray) examination and/or magnetic resonance imaging (MRI) conducted to verify diagnosis. This is then followed by repair surgery via arthroscopic intervention. Arthroscopy enables detailed description of lesion size and severity; however, the method is ineffective in detecting early degenerative changes in cartilage. In addition, the reproducibility of arthroscopy has been reported to be poor^[@CR4],\ [@CR5]^ due to its subjective nature. Thus, appropriate diagnostic methods capable of detecting the onset and progression of cartilage degeneration, both objectively and in real-time, is required.
Cartilage integrity can be characterized histologically using the Mankin grading system^[@CR6]^. Although this method is effective for overall tissue matrix characterization, it requires destructive (biopsy excision) and time-consuming protocols for histological evaluation of cartilage health. Direct application of this technique is therefore not feasible in surgical applications. Consequently, non-destructive approaches for determining articular cartilage health indirectly via the Mankin score have been proposed, including near infrared (NIR) spectroscopy^[@CR7]--[@CR14]^, mid-infrared spectroscopy^[@CR15],\ [@CR16]^ and optical coherence tomography (OCT)^[@CR17],\ [@CR18]^. This study investigates the capacity of NIR spectroscopy for detecting and characterizing progressive degenerative changes in articular cartilage.
NIR spectroscopy is a vibrational spectroscopic technique that is sensitive to specific molecular species containing CH, NH, OH and SH bonds, which constitute the fundamental chemical structure of biological tissues. NIR has been shown to be sensitive to micro- and macroscopic properties of cartilage^[@CR10],\ [@CR11],\ [@CR19],\ [@CR20]^, and a typical spectrum incorporates latent information on structural, compositional and morphological properties of the tissue. In addition, NIR spectroscopy is a rapid, non-destructive optical technique that penetrates deep into soft tissues^[@CR21]^, permitting full-depth cartilage probing^[@CR22],\ [@CR23]^. The potential of NIR spectroscopy for non-destructive probing of articular cartilage is currently gaining research attention^[@CR7]--[@CR14],\ [@CR19],\ [@CR24],\ [@CR25]^, and its capacity for evaluation of engineered cartilage has been demonstrated^[@CR26],\ [@CR27]^.
Earlier, we demonstrated the potential of NIR spectroscopy for estimating articular cartilage Mankin score from its spectral response, with respect to differentiating between types and severity of cartilage degeneration^[@CR7]^. However, no study has assessed the capacity of this optical method for monitoring progressive degenerative changes in cartilage. In this study, we hypothesized that NIR spectroscopy is capable of detecting and characterizing degenerative changes in articular cartilage, evaluated histologically and biochemically via Mankin score and glycosaminoglycans (GAG) content analyses, respectively. Multivariate techniques for classification (principal component analysis -- PCA, and support vector machines -- SVM) and regression (partial least squares regression -- PLSR) combined with variable selection were utilized for investigating changes/differences in the NIR spectrum associated with disease progression relative to the Mankin score and GAG content of articular cartilage.
Methodology {#Sec2}
===========
Animals {#Sec3}
-------
All animal experiment and protocols were approved by the Ethics Committee of Queensland University of Technology. The animal experiment and protocols were performed in accordance with relevant guidelines and regulations of the aforementioned committee. Male Wistar rats (n = 12, 8--10 weeks old) were purchased from the Animal Resource Centre (Perth, Western Australia, Australia), each animal weighing approximately 320 g. The animals were housed under conditions that included a controlled light cycle (light/dark: 12 h each) and controlled temperature (23 ± 1 °C), and were allowed to habituate themselves to the housing facilities for at least 7 days before surgeries.
Rat OA model and sample preparation {#Sec4}
-----------------------------------
Experimental osteoarthritis (OA) was induced in the rats by surgical removal of the medial meniscus (meniscectomy, MSX) of the right knee as described in our previous studies^[@CR7]^. The left knees were left intact and used as controls (sham). The animals were euthanized at four time points: 1, 2, 4, and 6 weeks (n = 3 animals/week) post-surgery and both knee joints, injured OA and control (sham) were removed by dissection. Subsequently, NIR spectral measurements were acquired from the tibial and femoral medial and lateral condyles of each joint. The first animal in week 1 post-injury was excluded from the study as a result of experimental error, resulting in a total of 58 measurement locations (**n** ~**sham**~ = 14; **n** ~**w1**~ ** = **8; **n** ~**w2**~ = 12; **n** ~**w4**~ = 12; **n** ~**w6**~ = 12). Although 44 sham sample locations were available, only 14 randomly selected locations were used in order to have similar number of samples from both sham and different diseased joints. The sham samples were randomly selected to include at least one sample location per animal. Following NIR spectral acquisition, the knee joints were processed for histological staining and sulphated glycosaminoglycan (sGAG) assay analysis.
NIR spectroscopy {#Sec5}
----------------
Diffuse reflectance NIR spectroscopy was performed using a Bruker MPA™ FT-NIR (Fourier Transform NIR) spectrometer (Bruker Optics, Germany), with detector spanning the full NIR spectral range (4,000--12,500 cm^−1^). The spectrometer was equipped with a custom-made fibre optic probe (dia. = 5 mm, optical window = 2 mm) consisting of seven 600 µm fibres: six peripherally positioned for transmitting the NIR light, and one centrally placed for collecting the diffusely reflected light from the tissue. The spectrometer was connected to a PC running OPUS 6.5 software (Bruker Optics, Germany) for data acquisition.
In preparation for spectral measurement, each joint is firmly held in a custom-built rig as described in our previous study^[@CR7]^. Prior to sample scanning, a reference spectrum was taken from a spectralon reflectance standard -- *SRS-99* (Labsphere Inc., North Sutton, USA). Spectral data was then obtained over the full wavelength range at 8 cm^−¹^ resolution, with each spectrum consisting of 16 co-added scans. The location of NIR measurement was visually noted to ensure that further analyses were conducted on tissue extracted from the same region where spectral data were acquired.
Morphological and histological characterization of OA samples {#Sec6}
-------------------------------------------------------------
After NIR spectroscopy, the joints were fixed in 4% paraformaldehyde and decalcified in 10% EDTA over a period of 2--3 weeks. Following decalcification, cartilage sections (surface-to-bone) from the region that was subjected to NIR spectroscopic probing were carefully extracted for histological evaluation based on Mankin score. After dehydration and paraffin embedding, two serial 5 μm sagittal sections obtained at 100 μm intervals from non-weight-bearing and weight-bearing regions, were cut from the joints and stained with safranin O--fast green. For Safranin-O/Fast Green staining, the paraffin-embedded sections were counterstained with Haematoxylin before being stained with 0.02% aqueous Fast Green for 4 min (followed by 3 dips in 1% acetic acid) and then 0.1% Safranin-O for 6 min. The slides were then dehydrated and mounted with crystal mount medium. OA severity in the joints was evaluated according to the modified Mankin histological grading system^[@CR6]^, and Mankin score was assigned for each sample location by three independent assessors. The Mankin score assesses structural integrity (0--6 points), cellularity (0--3 points), matrix staining (0--4 points), and tidemark integrity (0--1 points), with a maximum score of 14 points. The final score for each sample location was determined as the most severe histological change observed in multiple sections. In the case where the scores were different among the assessors, the highest Mankin score was selected. The inter-assessor agreement was assessed using the kappa (κ) coefficient, a chance-corrected estimate of agreement. κ values of 1.00--0.81 indicate excellent agreement, 0.80--0.61 substantial agreement, 0.60--0.41 moderate agreement, 0.40--0.21 fair agreement and 0.20--0.00 slight agreement^[@CR28]^.
Biochemical quantitation of GAG content {#Sec7}
---------------------------------------
Sulphated GAG (sGAG) assay was performed based on a protocol provided in the Blyscan sulfated Glycosaminoglycan assay kit (Biocolor Life Science Assays; Labtek, West Ipswich, QLD, Australia). sGAG content was measured at a wavelength of 595 nm by a microplate spectrophotometer (Benchmark Plus, Tacoma, WA, USA). Three measurements were obtained and repeatability coefficient calculated from Bland--Altman analysis as 1.96 times the standard deviation of the differences.
Statistical evaluation and spectral analyses {#Sec8}
--------------------------------------------
Statistical comparison of the reference histological and biochemical measurements between joints in the different groups was performed using GraphPad Prism (ver. 6, GraphPad Software Inc.). The data were expressed as mean ± SD and were compared using one-way ANOVA, a *p*-value of less than 0.05 was considered for statistical significance.
Principal component analysis (PCA), a feature extraction and classification method^[@CR29],\ [@CR30]^, was employed to identify inherent patterns in the spectra of samples in the different groups, and also to reduce data dimensionality to a few uncorrelated variables (principal components -- PC). Scores representing the first and second PCs, which explain variations in the original spectral data, were obtained and used to characterize/group the samples. Subsequently, support vector machines (SVM), a machine learning technique based on statistical learning, was utilized to determine decision boundaries that optimally demarcate and classify any inherent classes/groups from the PCA score plot of the samples.
Multivariate prediction models for correlating the spectral data to the samples' Mankin score and GAG content were developed using partial least squares (PLS) regression. This multivariate technique generates predictive models that optimizes the variance in both predictor variables (NIR data) and response variables (Mankin score and GAG content)^[@CR31]^. The performance of the developed model is then evaluated using validation techniques, such as cross-validation. Prior to PLS regression, variable selection techniques and spectral pre-processing were applied.
Variable selection algorithms based on shaving and genetic algorithm were applied to determine the most informative spectral variables. In shaving variable selection method, a model is first fitted to the entire data, the variables are then sorted according to their selectivity ratio, a filter method for extracting a set of important variables. The selectivity ratio (SR) is defined as the ratio between explained and residual variance of the spectral variables on the target-projected component^[@CR32]^. It provides a numerical assessment of the usefulness of each variable in a regression model; the larger the SR, the more useful the given variables are for the prediction. Secondly, a threshold is used to eliminate a subset of the least informative (worst performing) variables. Then a model is fitted again to the remaining variables and performance is measured. The procedure is repeated until maximum model performance is achieved^[@CR33]^. Genetic algorithm is a subset search algorithm inspired by biological evolution theory and natural selection^[@CR34]^. In this variable selection algorithm, an initial population of variable sets are built by random selection. A PLS model is then fitted to each variable set, and performance computed using cross-validation procedure. A collection of variable sets with higher performance are then selected to survive until the next "generation". Subsequently, new variable sets are formed by crossover of selected variables between the surviving variable sets, and by changing (mutating) the random selection for each variable. The procedure is then repeated with the surviving and modified variable sets until optimal model performance is achieved.
Following variable selection, spectral pre-processing methods including multiplicative scatter correction (MSC) and derivative pre-treatment, as well as combinations of both methods, were applied in order to reduce spectral non-linearities resulting from light scattering variations in reflectance spectroscopy^[@CR35]^ and to eliminate unwanted baseline variations, respectively. Subsequently, the resulting data (pre-processed most informative variables) are subjected to PLS analysis and then model performance is measured. The procedure is applied to both raw pre-processed spectral data.
Leave-one-out (LOO) cross-validation was employed to determine the optimal number of PLS components, and to estimate the performance of the predictive models developed. This validation method was adopted because it is effective for small sample sizes. To potentially minimise under- or over-fitting, optimal model selection was based on the lowest root mean square error of cross-validation (RMSECV), minimum number of components and high coefficient of determination (R^2^). For multivariate spectral analyses, R statistical software^[@CR36]^ (ver. 3.3.1) was employed with the following packages: **stats** ^[@CR36]^ for PCA analyses, **e1071** ^[@CR37]^ for SVM, **pls** ^[@CR38]^ for PLS analyses, and **plsVarSel** package^[@CR39]^ for shaving and genetic algorithm variable selection.
Results {#Sec9}
=======
Variation of NIR spectra with tissue degeneration {#Sec10}
-------------------------------------------------
Variations in the optical response of the injured joints, as tissue degeneration progresses, is mostly characterized by an overall increase in absorbance across the NIR spectral range (Fig. [1](#Fig1){ref-type="fig"}). This is consistent with morphological and histological changes in the samples, which can be observed to change consistently with progression of tissue degeneration (Fig. [2](#Fig2){ref-type="fig"}). Spectral data below 5250 cm^−1^ were excluded due to spectral saturation in the 1^st^ overtone water peak and combination region of the NIR spectral range.Figure 1Representative raw (**a**), 1^st^ derivative (**b**) and 2^nd^ derivative (**c**) cartilage NIR spectra from control (sham) and progressively degenerating joint samples. The derivative spectra show consistent changes with progression of tissue degeneration. Figure 2Progressive degenerative changes in an animal model of OA. **(a)** Representative histology sections of the tibial condyles for the sham (control) joint and at 1, 2, 4 and 6 weeks post-injury. \[Distal tibia was sectioned coronally and stained with Safranin-O. Scale bar = 200 μm\]. The most degenerated area (medial part) of each sample was selected. **(b)** Quantitation of safranin-O stained sections using the Mankin histopathology scoring system for the progressively degenerated cartilage samples. **(c)** Quantitation of the GAG content of the samples. Values are the mean ± SD; \*p \< 0.001.\[w1 = week 1; w2 = week 2; w4 = week 4; w6 = week 6\].
The severity of cartilage injury with progressive tissue degeneration, evaluated using the modified Mankin score, were observed to increase consistently (mean ± sd): 0.0 ± 0.0 for sham, 0.3 ± 0.5 for samples after 1 week (w1), 0.9 ± 0.7 2 weeks (w2), 4.7 ± 0.9 4 weeks (w4), and 6.8 ± 1.0 6 weeks (w6) post-injury (Fig. [2b](#Fig2){ref-type="fig"}). Substantial agreement was observed between the independent assessors of Mankin score (κ = 0.65). The GAG content obtained biochemically were observed to consistently decrease with progression of tissue degeneration, with repeatability coefficient of 1.24 μg/mg and the following average values: 32.1 ± 0.7 μg/mg for sham, 29.3 ± 1.4 μg/mg for w1, 25.2 ± 1.2 μg/mg for w2, 20.6 ± 1.1 μg/mg for w4, and 10.5 ± 0.9 μg/mg for w6 (Fig. [2c](#Fig2){ref-type="fig"}). This is supported by the consistent decrease in staining intensity in the corresponding histological sections of the samples due to progressive loss of proteoglycans as injury severity progresses with time (Fig. [2a](#Fig2){ref-type="fig"}).
In general, reduction in cartilage proteoglycan content, extensive alterations characterized by marked hypercellularity due to cloning, numerous osteophytes on the margins, and decrease in the cartilage thickness were observed to increase with disease progression (Fig. [2](#Fig2){ref-type="fig"}). Statistically significant difference (p \< 0.001) in Mankin score and GAG content were observed between the samples at different weeks post-injury.
Multivariate classification of tissue degeneration {#Sec11}
--------------------------------------------------
Variable reduction and classification based on PCA show that the samples can be grouped into two linearly separable classes based on variations in their NIR spectral data using the 1^st^ and 2^nd^ principal components scores (PC~1~ and PC~2~). The scores can be observed to group the samples according to level of degeneration along PC~1~, while samples within each group cluster along both PC~1~ and PC~2~.
"Class 1" consists of samples with low Mankin score (\<=2) and relatively high GAG content (\>23 μg/mg), and is representative of cartilage with mild degeneration (sham, weeks 1 and 2); while "class 2" consists of samples with relatively high Mankin score (=\>3) and low GAG content (\<23 μg/mg), indicative of advanced tissue degeneration (weeks 4 and 6) (Fig. [3](#Fig3){ref-type="fig"}). Although PCA was performed on the spectra with and without pre-processing (multiplicative scatter correction (MSC) and derivative (1^st^ and 2^nd^)), optimal classification was obtained without pre-processing. SVM shows that a decision boundary that optimally demarcates both classes in the PCA score plot can be obtained, since the classes are linearly separable (Fig. [3a](#Fig3){ref-type="fig"}). The SVM model classified all samples with advanced degeneration correctly, but misclassified two samples with mild degeneration (Fig. [3b](#Fig3){ref-type="fig"}). No significant difference (p = 0.0588) in tissue degeneration (via the Mankin score) was observed between the samples in class 1; however, statistically significant difference (p = 0.009) was observed between the samples in class 2.Figure 3(**a**) PCA score plot of the NIR spectral data of the samples showing classification into two distinct groups. "Class 1" consists of sham and samples from weeks 1 and 2 post-injury, class 2 consists of samples from weeks 4 and 6 post-injury. (**b**) SVM decision boundary showing the optimal demarcation of both classes.\[w1 = week 1; w2 = week 2; w4 = week 4; w6 = week 6\].
Multivariate analyses: spectral pre-processing, variable selection and PLSR {#Sec12}
---------------------------------------------------------------------------
A combination of MSC spectral pre-processing and variable selection based on shaving resulted in the best models for estimating the Mankin score, while variable selection alone, based on genetic algorithm, was optimal for estimating the GAG content from the NIR spectra of the samples (Table [1](#Tab1){ref-type="table"}). The most informative variables for estimating the Mankin scores from the spectral data consisted of 493 variables from two main spectral regions: 8547--10361 cm^−1^ and 6411--6496 cm^−1^ (Fig. [4a](#Fig4){ref-type="fig"}), while the model for estimating GAG content required 192 variables (Fig. [4b](#Fig4){ref-type="fig"}) broadly dispersed across the whole spectral range, with sparse variables around the 2^nd^ overtone water peak (around 6860 cm^−1^).Table 1Multivariate analyses assessment of the relationship between the optical characteristics of articular cartilage and its matrix integrity during progressive tissue degeneration.**MANKIN SCOREPre-pro + VSn** ~**comp**~**R** ^**2**^ ~**cv**~**RMSECV% error\*n** ~**vars**~none981.231.1814.75allMSC1064.21.6320.38allShaving987.640.9511.921503MSC + Shaving886.221.0112.60493GA1080.201.1814.81211MSC + GA1085.401.0413.01201**GAG CONTENTPre-pro + VSn** ~**comp**~**R** ^**2**^ ~**cv**~**RMSECV (µg/mg)% error\*n** ~**vars**~none995.171.717.46allMSC1061.574.8120.94allShaving894.471.837.97616MSC + Shaving887.172.7812.10493GA9951.747.57192MSC + GA1069.754.2618.53204\[Pre-pro = spectral pre-processing; VS = variable selection; n~comp~ = number of PLS components; RMSECV = root mean square error of cross-validation; n~vars~ = number of spectral variables used in the multivariate analyses. \*Error is estimated relative to the range of the reference data, all = 1879 variables\]. Figure 4Spectral plots showing regions with the most informative variables obtained from variable selection algorithms for optimal PLS regression models for (**a**) Mankin score and (**b**) GAG content.
High correlation with relatively low error were obtained with the optimal models for Mankin score (R^2^ = 86.2%, error = 12.6%) and GAG content (R^2^ = 95%, error = 7.6%). The relationship between the measured and NIR predicted Mankin scores and GAG contents are presented in Fig. [5](#Fig5){ref-type="fig"}. The significantly high correlation between the optical response and both the tissue's structural integrity (Mankin score) and composition (GAG content) demonstrates the capacity of NIR to monitor progression of cartilage degeneration, and ultimately OA progression.Figure 5Relationship between NIR spectral predicted and measured (**a**) Mankin score and (**b**) GAG content of the progressively degenerated cartilage samples.
Discussion {#Sec13}
==========
In this study, we demonstrate the capacity of NIR spectroscopy to monitor changes in articular cartilage matrix during degeneration. In earlier studies^[@CR7],\ [@CR9]^, we demonstrated the potential of NIR to identify, classify and predict different types and severity of cartilage degeneration using rat models of OA. In these studies, only one time-point (8 weeks post-injury) was considered. However, in the present study 4 time-points (1, 2, 4 and 6 weeks post-injury) were considered, with assessment of overall tissue integrity (Mankin score) and GAG contents, thus extending the approach to monitor progression of tissue degeneration based on the hypothesis that NIR spectroscopy is capable of detecting and characterizing histological and biochemical changes in cartilage during degeneration. The results show that this optical technique is capable of classifying the samples based on severity of degeneration, as well as predict their Mankin scores and GAG contents with relatively low error.
Since absorptions in the NIR spectral range arise mainly from OH, CH, NH and SH bonds, which characterize the fundamental molecular make-up of biological materials, NIR spectroscopy is capable of detecting both micro- and macroscopic changes in articular cartilage structure and composition^[@CR10],\ [@CR11],\ [@CR14],\ [@CR19],\ [@CR20]^. Thus, the capacity of NIR to effectively monitor cartilage degeneration can be attributed to the characteristics of the spectrum, which embeds information on key chemical^[@CR19]^, physical^[@CR11],\ [@CR22]^ and morphological properties of the tissue^[@CR7],\ [@CR9]^. The overall increase in absorption across the spectrum as cartilage degeneration progresses (Fig. [1](#Fig1){ref-type="fig"}) is due to increase in the matrix water content, which is an expected consequence of degenerative changes in cartilage matrix composition and structure (Fig. [2](#Fig2){ref-type="fig"}), and is among the earliest indicators of cartilage degeneration.
Classification based on PCA provides an insight into the relationship between the optical response of articular cartilage and the state of its matrix, and suggests that there are subtle spectral similarities between the samples, giving rise to the two classes that cluster according to level of tissue degeneration. Separation between the classes occurs along the 1^st^ PC axis (Fig. [3a](#Fig3){ref-type="fig"}), indicating that this axis is associated with tissue integrity. Although there is clear separation between both classes, no obvious grouping can be observed within class 1, suggesting that the effect of degenerative changes on the optical response of cartilage within the first two weeks post-injury is not substantial. In addition, no statistically significant difference (p = 0.0588) in the Mankin score of the samples in this class was observed, possibly due to very early sample-specific matrix changes. However, this was not the case for class 2, where statistically significant difference (p \< 0.001) was observed between samples from week 4 and week 6. A potential trend in grouping within this class can be observed, albeit with some overlap (Fig. [3a](#Fig3){ref-type="fig"}). The SVM decision boundary shows the region in space where the PC scores of new samples is likely to be located following PCA analysis. This could be an important preliminary diagnostic check to classify a sample prior to further diagnostics.
The optimal predictive model for estimating Mankin score from the NIR spectra in this study was obtained with a combination of MSC spectral pre-processing and variable selection based on shaving (Table [1](#Tab1){ref-type="table"}). This is different from our previous study^[@CR7]^ where the relationship was optimized with a combination of SNV and 1^st^ derivative pre-processing. This is probably due to the statistical variable selection technique employed in the present study, as opposed to the manual approach adopted in our previous study. In addition, MSC enables separation of tissue light absorption from physical light scattering, and has been shown to correct the undesirable effect of light scatter in NIR diffuse reflectance spectroscopy^[@CR35]^. The coefficient of determination (R^2^) obtained after cross-validation is similar in both studies, with slightly higher error in the present study.
The most informative spectral variables for predicting the Mankin score are confined to two main spectral regions: 8547--10361 cm^−1^ and 6411--6496 cm^−1^ (Fig. [4a](#Fig4){ref-type="fig"}), which are indicative of the solid extracellular matrix components of cartilage. The first region (8547--10361 cm^−1^) is associated with the 2^nd^ overtone --CH2 stretch vibration, which is due to the combined proteoglycan and collagen spectral absorption. The second region (6411--6496 cm^−1^) can be attributed to proteoglycan-related spectral absorptions associated with 1^st^ overtone --NH stretch vibrations. It is worth noting that although the water content is consistently changing with tissue degeneration, spectral data associated with the main (1^st^ overtone) OH absorption region (6700--7200 cm^−1^) and directly related to cartilage water content were not optimal for predicting the tissue's Mankin score. However, the interaction between the tissue's water content and its changing solid matrix components is the likely driver of the relationship obtained with data from the variable selected regions.
The variables selected based on genetic algorithm for estimation of cartilage GAG content are dispersed across the whole spectrum (Fig. [4b](#Fig4){ref-type="fig"}), arguably due to the close relationship between the proteoglycan and water contents of cartilage, as the latter can be observed to increase consistently with degeneration and affects overall spectral absorption (Fig. [1](#Fig1){ref-type="fig"}). Only very few variables were required around the water peak (Fig. [4b](#Fig4){ref-type="fig"}), suggesting that data within this region does not add much information to the predictive model. The model based on the whole spectrum (without variable selection: 1879 variables) is similar to that obtained with genetic algorithm (variable selection: 192 variables) in terms of coefficient of determination and error. This is due to the dispersive nature of the genetic algorithm selected variables, which spread across the whole spectrum. However, the model based on genetic algorithm is preferable over that based on the full spectrum, since only about 10% of the variables is required to estimate the GAG content, resulting in a simpler and more robust model.
The strong correlation (Fig. [5](#Fig5){ref-type="fig"}) between the measured and predicted tissue properties (Mankin score and GAG content) demonstrates the potential of NIR spectroscopy for detecting and monitoring cartilage degeneration, as well as a tool for clinical diagnosis of cartilage condition post-injury during surgery. However, clinical adaptation of the outcome of this study would need to be preceded by extensive and rigorous pre-clinical validation and optimization using human tissue samples.
This study contributes to knowledge that could advance understanding of the degenerative process of cartilage post-injury, as well as potential clinical monitoring of the extent of cartilage degeneration during surgery, particularly in relation to PTOA, thus enabling more focused and optimized treatment regiment. Application of variable selection techniques, which significantly reduces the number of spectral data required for prediction of cartilage properties, presents a computationally faster approach compared to using the whole spectrum. This translates to less time for analyses and better support for real-time application, such as mapping tissue properties around lesions during surgery. In addition, the outcome of this study presents potential applications beyond cartilage degeneration, to monitoring the progression of other soft tissue pathologies.
Although the Mankin scoring system is not widely used clinically for assessing tissue degeneration, it provides detailed histological information on the tissue health. Nevertheless, the analytical techniques applied in this study can be extended to clinically applicable scoring systems, such as OARSI system. Another limitation of the current study is that tissue degeneration in humans may not follow the exact pattern observed in the rat samples, as osteoarthritic degeneration tend to be complex in nature. However, the injury model employed, which was surgically initiated, is effective in modeling the characteristics of natural OA^[@CR40]^. Finally, the present study did not evaluate the capacity of NIR spectroscopy to estimate collagen-related information, such as cross-linking and fibril orientation, which are not accounted for in the Mankin scoring system.
**Publisher\'s note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Dr Afara acknowledges funding from Finnish Cultural Foundation - Suomen Kulttuurirahasto (00160079). We thank the National Health and Medical Research Council (NHMRC) Australia, and Queensland University of Technology, Brisbane, Australia for partly funding this project.
Contributions by the authors to the preparation of this manuscript are as follows: Isaac Afara: Study conception and experimental design, conduction of laboratory experiment and data acquisition, multivariate and statistical analysis, and preparation of the manuscript. Indira Prasadam: Experimental design, conduction of laboratory experiment, histological evaluation and analysis, and preparation of the manuscript. Zohreh Arabshahi: Data acquisition, conduction of laboratory experiment, and approval of final manuscript draft. Yin Xiao: Guidance in experimental design and analysis, writing and editing manuscript. Adekunle Oloyede: Guidance in experimental design and data analysis, writing and editing manuscript.
Competing Interests {#FPar1}
===================
The authors declare that they have no competing interests.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
Food poisoning of bacterial origins is widespread and a considerable public health concern both in the US and worldwide. In the US alone, the Centers for Disease Control and Prevention estimates that 9.4 million Americans contract foodborne illness each year \[[@B1]\]. Foodborne bacterial contaminants such as *Salmonella spp.*, *Listeria* *spp.*, *Campylobacter* *spp.*, and *Escherichia coli* (*E. coli*) account for a considerable portion of these cases, and these four pathogens are responsible for many of the most deadly outbreaks \[[@B2],[@B3]\]. *E. coli* strains expressing Shiga toxin (Stx), known as STEC (Shiga toxin-producing *E. coli*), can result in a wide variety of clinical manifestations, ranging in severity from innocuous diarrhea to hemorrhagic colitis and life-threatening hemolytic uremic syndrome (HUS) \[[@B4]\]. Phage encoded Stx is among the most important virulence factors for enterohaemorrhagic *E. coli* (EHEC) \[[@B5],[@B6]\] and enteroaggregative hemorrhagic *E. coli* (EAHEC) \[[@B7],[@B8]\]. Many serotypes of EHEC, a class of pathogenic *E. coli* that can cause bloody diarrhea, possess one or several *stx* genes. Enteroaggregative *E. coli* (EAEC) are characterized by their ability to attach to cells which line the intestine; EAHEC additionally have the ability to cause bloody diarrhea \[[@B9]\]. The O104:H4 strain of *E. coli* responsible for the most deadly recent outbreak of STEC in Germany (2011) is classified as EHAEC (or EAEC) and possesses a *stx* gene (*stx2*) \[[@B10]\].
The diversity of STEC strains, both in the genes repertoires they possess and the virulence factors they encode, is considerable. Although the *E. coli* O157:H7 serotype is the most infamous, non-O157 serotypes are responsible for a considerable number of STEC outbreaks. Six O groups (O26, O45, O111, O121, O103, and O145) cause approximately 71% of non-O157 outbreaks \[[@B11]\]. The EAEC strain O104:H4 caused one of the worst *E. coli* incidents in history, a mass outbreak of STEC in Germany in 2011, affecting 3816 people and resulting in 845 cases of HUS and 54 deaths \[[@B7],[@B8],[@B12]\]. These serotypes can harbor one or more *stx* genes, of which there are many varieties. These genes are carried by lambdoid bacteriophages, which can facilitate the transfer of *stx* sequences between STEC serotypes, non-pathogenic *E. coli* \[[@B13]\], and possibly other close relatives to *E. coli* in Enterobactericiae \[[@B14],[@B15]\]. The two main types of Stx include Stx1, which is nearly identical to the toxin from the *Shigella* genus, and Stx2, which is considerably different from Stx1 (only 56.6% amino acid identity between A subunits without signal sequences).
Like several other bacterial toxins, Stx has an AB~5~ structure: the catalytic A subunit is delivered to target cells by a B subunit pentamer. The B subunit pentamer binds the glycolipid receptors globotriaosylceramide (Gb3Cer) and/or globotetraosylceramide (Gb4Cer) on the surface of target cells, allowing entry of the A subunit which then inactivates ribosomes via its *N*-glycosidase activity \[[@B16],[@B17]\]. Although Stx1 seems to be more toxic to Vero cells \[[@B18]\], Stx2 is the much more potent toxin *in vivo*: Stx2 is around 100 times more toxic to mice than Stx1. Stx2 seems to have comparable catalytic activity to Stx1 \[[@B19]\]. The A and B subunits of Stx1 and Stx2 possess N-terminal signal sequences which facilitate their transport to the periplasm, where they assemble into mature toxin \[[@B20],[@B21]\]. Expression of Stx is driven by a late-phase phage promoter, which is strongly activated upon induction of the bacterial SOS response. Expression of Stx1 is additionally dependent upon a bacterial promoter that is responsive to iron concentration \[[@B22]\]. The SOS response also initiates lysis of *E. coli* cells by the phage, resulting in release of the toxin. Some antibiotics, such as the quinolones (e.g., ciprofloxacin), exacerbate the effects of Stx toxicity, presumably by inducing and releasing large amounts of toxin at once \[[@B23],[@B24]\]. Treatment of STEC by these antibiotics might actually worsen the symptoms of STEC infections \[[@B25]\]. Because of this, there are currently no widely accepted antibiotic treatments of STEC infections, although proper antibiotic treatment may ultimately improve the prognosis of patients with the potentially life-threatening HUS \[[@B26]\].
Within each Stx type (Stx1 and Stx2), there are a number of subtypes which vary in sequence, specificity, and toxicity. There are 3 characterized subtypes of Stx1 (Stx1a, Stx1c, and Stx1d) and 7 subtypes of Stx2 (Stx2a, 2b, 2c, 2d, 2e, 2f, and 2g) \[[@B27]\]. The subtypes of Stx1 are relatively conserved at the amino acid level, whereas those of Stx2 can be more diverse. However, the Stx2a, Stx2c, and Stx2d subtypes are very similar to each other, and these subtypes are typically associated with HUS \[[@B18],[@B28]\]. Stx2b, Stx2e, Stx2f, and Stx2g are less commonly found in serious human disease, although Stx2e can cause edema disease in neonatal piglets \[[@B29]\]. Stx2f (found mostly in avian isolates) \[[@B30]\] is the most unique of the Stx2 subtypes (73.9% identity to Stx2a in the A subunits), followed by 2b (93.3%), Stx2e (93.9%), and finally Stx2g (94.9%). Differences among the B subunits determine each subtype's receptor specificity. Stx2a, Stx2c, and Stx2d bind preferentially to Gb3Cer, while it has been reported that Stx2e prefers Gb4Cer (but can also bind Gb3Cer) \[[@B31]\]. Several amino acids in the C-terminus of the B subunit are critical for determining receptor preference. When the double mutation Q64E/K66Q is made to the Stx2e B subunit, it loses its ability to bind Gb4Cer, and has a receptor preference analogous to Stx2a \[[@B32]\]. The B subunit of Stx2f has Q64/K66 like Stx2e, and can bind both Gb3-LPS and Gb4-LPS, which are mimics of Gb3Cer and Gb4Cer, respectively \[[@B33]\].
Most Stx2 detection kits (both PCR and immunoassays) are optimized to Stx2a, and cross-react with closely related Stx2c and Stx2d. However, many do not recognize the divergent Stx2b, Stx2e, and Stx2f subtypes. Antibodies that recognize Stx2f have been reported, but few are commercially available and they are generally sold only as components of an assay kit, making them difficult to use as research tools and very expensive. Whether there is a reliable immunological method for detecting Stx2f is still a matter for debate. One of the primary means for detecting Stx1 and Stx2, the Premier EHEC kit from Meridian Biosciences, has been reported to detect Stx2f in two studies \[[@B30],[@B34]\] but is insensitive to Stx2f in another \[[@B6]\]. A reverse passive latex agglutination assay (VTEC-RPLA) has repeatedly been shown to recognize Stx2f, but the sensitivity of this assay to Stx2f is unknown (Denka Seiken, Japan) \[[@B30]\]. In this study, we detail and characterize a group of novel monoclonal antibodies (mAbs) that react robustly and uniquely to Stx2f. With these antibodies, we have developed an immunoassay for simple detection of the Stx2f subtype.
Materials and Methods {#s2}
=====================
Ethics Statement {#s2.1}
----------------
All procedures with animals were carried out according to institutional guidelines for husbandry approved by the Animal Care and Use Committee of the U.S. Department of Agriculture, Western Regional Research Center (USDA ACUC Protocol 09-J-10). Mice were euthanized using rapid cervical dislocation to minimize suffering.
*E. coli* strains and growth conditions {#s2.2}
---------------------------------------
Strains expressing Stx2a (RM10638) and Stx2f (RM7007) as well as a control strain (K12) were grown as previously described \[[@B33]\]. Briefly, *E. coli* strains were inoculated into 10 mL of LB overnight at 37°C with agitation, then diluted 1/10 into 500 mL LB with 50 ng/mL mitomycin C (MMC) (Sigma-Aldrich, St. Louis, MO) and grown in a shaking incubator for 24 hours at 37°C. Cells were centrifuged for 15 minutes at 5000xG, the cell pellet was autoclaved, bleached, and discarded, and the media was sterile filtered. Stx2b (RM7005), 2c (RM10058), 2d (RM8013), 2e (RM7988), and 2g (10468)-expressing stains were also grown in this manner. Cells expressing the His-tagged Stx2f A subunit were also grown as described, as were Gb3-LPS- and Gb4-LPS-expressing strains \[[@B33]\]. FSIS EC465-97 is a wild-type *E. coli* O157:H7 strain transformed with pGFP which produces green fluorescent protein. It was provided by Todd J. Ward at the USDA-ARS, NCAUR, Peoria, IL 61604. All strains used in this study are listed in [Table 1](#pone-0076563-t001){ref-type="table"}.
10.1371/journal.pone.0076563.t001
###### *E. coli* strains used in this study.
Strain Other names Serotype Biomolecule expressed Origin Reference
----------------------- ------------- ---------- ----------------------------- -------------- ------------
RM10638 O157:H7 Stx2a Cow (2009) \[[@B43]\]
RM7005 EH250 O188:H12 Stx2b Clinical \[[@B43]\]
RM10058 O157:H7 Stx2c Bird (2009) \[[@B43]\]
RM8013 ND^a^ Stx2d Cow (2008) \[[@B43]\]
RM7988 ND^a^ Stx2e Water (2008) This study
RM7007 T4/97 O128:H2 Stx2f Feral pigeon \[[@B43]\]
RM10468 ND^a^ Stx2g Cow (2009) \[[@B43]\]
RM5034 K12 \[[@B43]\]
CWG308 pJCP-Gb3 Gb3-LPS \[[@B44]\]
CWG308 pJCP-*lgt*CDE Gb4-LPS \[[@B45]\]
CWG308 \[[@B44]\]
TOP10 Invitrogen
TOP10 pTrcHis2-Stx2fA Stx2f A subunit +6xHis \[[@B33]\]
FSIS EC465-97 O157:H7 *GFP-positive* Stx-negative USDA, FSIS \[[@B42]\]
^a^ Not determined.
Purification of Stx2a and Stx2f {#s2.3}
-------------------------------
Purifications were conducted using cell-free supernatants of Stx2a (RM10638) and Stx2f-expressing (RM7007) *E. coli strains* and previously published protocols \[[@B33]\]. Recombinant His-tagged Stx2f A subunit was purified as previously described \[[@B33]\]. Partially purified (^≈^50% pure) Stx1 was purchased from Toxin Technologies (Sarasota, FL).
Cell culture {#s2.4}
------------
Complete hybridoma media (cHM) used for the culturing of SP2/0 mouse myeloma cells and hybridoma cell lines consisted of Iscove's modified Dulbecco's Minimal medium (Sigma-Aldrich) containing NaHCO~3~ (36mM) and 1x Glutamax (Invitrogen, Carlsbad, CA), supplemented with 10% heat-inactivated fetal calf serum (FCS) (Invitrogen). Incomplete hybridoma media (iHM) is cHM without FCS. HAT (hypoxanthine, amniopterin, and thymidine) selection medium was prepared as 1x HAT supplement (Sigma-Aldrich) dissolved in cHM. Macrophage conditioned medium (MPCM) was prepared as previously described \[[@B35]\]. cHM was supplemented with 50% MPCM for the initial HAT selection and 10% MPCM for the hybridoma cloning steps. cHM with 1x HT (Hypoxanthine and thymidine, Sigma-Aldrich) was used during the first and second cloning steps. Cells were maintained at 37°C, 5% CO~2~.
Immunization and polyclonal serum production {#s2.5}
--------------------------------------------
Mouse immunizations were conducted using His-tagged Stx2f A subunit as described previously \[[@B36]\]. Briefly, female Balb/cJ mice were injected intraperitoneally three times with 5 µg of Stx2f A-subunit in Sigma adjuvant system (Sigma-Aldrich) at two-week intervals, then bled (using the tail vein procedure) to collect polyclonal serum and confirm that the serum possesses antibodies that recognize Stx2f by direct ELISA using Stx2f purified from a bacterial strain as an antigen. The mouse with the highest anti-Stx2f serum titre was then boosted once with 1 µg Stx2f A subunit without adjuvant a week later. Three days later, the spleen was excised aseptically after euthanasia.
Hybridoma development, cloning, and screening {#s2.6}
---------------------------------------------
Monoclonal antibodies (mAbs) were produced as described \[[@B36]\]. Briefly, cell fusions were achieved using SP2/0 myeloma cells, splenocytes extracted from the inoculated mouse spleen, and polyethylene glycol. Following fusion, the cells were diluted into ten 96-well plates and allowed to recover for 12 days in 50% MPCM/HAT/cHM medium. The hybridomas were then screened for antibodies recognizing Stx2f by ELISA and positive wells were transferred to 24-well plates in 10% MPCM/HT/cHM media to recover. Following recovery, the hybridomas were diluted to 500 cells/mL then serial diluted (2-fold) across a 96-well plate. This cloning step was repeated two additional times, with the final cloning being conducted in 10% MPCM/cHM. After clonal hybridoma lines were isolated, cells were grown in cHM media.
Monoclonal antibody preparation {#s2.7}
-------------------------------
Around 400 mL of antibody-containing media (hybridoma cells grown in cHM for 2-3 days) was passed through a Protein G column (GE Healthcare). Antibody was eluted with 0.1 M glycine (pH 2.7), resulting in 4-6 mg of purified Stx2f antibody. Protein concentration was determined using the BCA Protein Assay Kit (Thermo Scientific, Rockford, IL). Biotinylation of antibodies was performed using the Lightning-Link Biotin Conjugation Kit (Innova Biosciences, Cambridge, UK). Antibody isotyping was conducted by ELISA using Stx2f and horseradish peroxidase (HRP) -conjugated isotype-specific antibodies (Southern Biotech, Birmingham, AL).
Enzyme-linked immunosobent assays (ELISA) {#s2.8}
-----------------------------------------
For hybridoma screening, Stx2f (50 ng/mL in Phosphate buffered saline \[PBS\]) was bound to the wells of a black NUNC Maxisorb 96-well plate overnight at 4°C. The plates were washed twice with PBS/0.05% Tween 20 (PBST) (using a BioTek ELx405 plate washer) and blocked with 200 μL/well 5% nonfat dry milk in PBST (blocking solution) for 1 hour at room temperature (RT). The plates were then washed twice with PBST, then 50 μL/well blocking solution was added to 50 μL/well hybridoma culture media. This was incubated for 1 hour at RT, followed by six washes with PBST. A 1/5,000 dilution of HRP-conjugated goat anti-mouse IgG antibody (GAM-HRP) antibody (Promega) in blocking solution was then dispensed into the plates, and incubated for 1 hour at RT. The plates were washed a further six times with PBST, then 100 μL/well Pico chemiluminescent substate (Thermo Scientific) was added, and 5 minutes later, luminescence was measured using a Victor II plate reader (PerkinElmer). Direct-well binding ELISAs ([Figure 1B](#pone-0076563-g001){ref-type="fig"}) were conducted in the same manner, except that 250 ng/mL Stx2f, Stx2a, and Stx1 was used to coat ELISA plates.
{#pone-0076563-g001}
For sandwich ELISAs, purified capture antibody at 1 µg/mL in PBS was incubated in black Maxisorb 96-well plates overnight at 4°C. The plates were blocked and washed as with the hybridoma screening ELISA, except using 3% BSA in lieu of 5% nonfat dry milk. Plates were washed twice with PBST, then Stx2f (diluted in PBS to various concentrations) was then added at 100 μL/well, incubated for 1 hour at RT, and washed six times with PBST. Biotinylated antibody was diluted to 1 µg/mL, added to the plate at 100 μL/well, and incubated for 1 hour at RT, then the plates were washed six times with PBST. 1 mg/milliliter streptavidin-HRP conjugate (Invitrogen) was diluted to 1/10,000 in PBS, added at 100 μL/well, and incubated for 1 hour at RT. Following another six washes with PBST, the plates were developed and read like the hybridoma screening ELISAs. Limit of detection (LOD) was determined by extrapolating ng/mL of Stx2f from the background luminescence plus 3 standard deviations of the background. For chicken breast extract ELISAs, 0.25 g chicken breast was combined with 0.5 mL PBS and homogenized using a pestle in a microfuge tube. Debris was removed by centrifugation (12 kG, 5 min.), and the resulting suspension was sterile filtered (0.2 µm). This chicken breast extract was diluted 10-fold in PBS, then used to dilute Stx2f during the toxin binding step.
Western blots {#s2.9}
-------------
Western blots were conducted as previously described \[[@B33]\]. Samples were incubated at 72°C for 10 minutes in 1x NuPage SDS loading buffer before being run on a 4%--12% NuPAGE Novex Bis-Tris mini gel (Invitrogen). Then the proteins were transferred to a PVDF membrane (pore size, 0.45 µm; Amersham Hybond-P), blocked with 2% ECL Prime blocking agent (GE Healthcare) in PBST for 1 hour at RT, and washed thrice with PBST (3 minutes each). Antibodies were diluted 1/1000 in blocking solution and incubated with the blots for 1 hour at RT, then the blots were washed thrice again in PBST. GAM-HRP antibody (Promega) at a 1/10,000 dilution was incubated on the blot for 1 hour at RT, the blots were washed four more times with PBST, and developed using Lumigen TMA-6 (Lumigen) substrate. The blots were visualized with a 2 minute exposure using a FluorChem HD2 (Alpha Innotech).
Gb3/4-LPS binding assays {#s2.10}
------------------------
Mouse mAb against Stx2a B-subunit (VT136/8-H4 from Sifin Institute, Berlin, Germany) or mAb Stx2f-1 (for Stx2f) at 1 µg/mL in PBS were bound to black Nunc Maxisorp plates and incubated overnight at 4°C. The plates were washed twice with PBST and blocked with 200 μL/well 3% BSA in PBST for 1 hour at room temperature (RT). During the blocking step, 125 ng/mL Stx2a and Stx2f toxin was incubated in microfuge tubes with varying amounts of Gb3-LPS or Gb4-LPS formalin-fixed cells diluted in PBS for 1 hour at RT. The toxin/cell complex was then spun down (12 k RPM for 2 minutes), and the liquid portion (containing unbound free toxin) in the microfuge tubes was then dispensed onto the blocked plates (100 μL/well) and incubated for 1 hour at RT. Plates were washed six times with PBST, then 1 µg/mL biotinylated detection antibody diluted in BSA blocking solution was added (biotinylated mAb Stx2f-4 was used for detection of Stx2f; biotinylated mAb VT135/6-B9 was used for detection of Stx2a) and allowed to incubate for 1 hour at RT. After another six washes, streptavidin-HRP conjugate was added at 1/10,000 in BSA blocking solution for 1 hour at RT. Following another six washes with PBST, the plates were developed and read in the same way as the hybridoma screening ELISAs above. There is an inverse relationship between the ELISA signal obtained and the amount of Stxs bound to Gb3-LPS and Gb4-LPS cells, since these cells remove the toxin from solution. The ELISA signal for Stx2a and Stx2f incubated without the presence of cells was initially set to 100% because all toxins were available to bind to the capture and detection antibodies for signal development. The ELISA signal for Stx2a and Stx2f incubated with Gb3 or Gb4 cells at an A~600~ of 0.2 was initially set to 0% because at this condition there is almost no toxin left to bind to the detection antibody; all the toxin bound to cells and was removed by centrifugation. Since we wanted to display binding of Stx to Gb3/4-LPS, we then flipped the values (100% signal became 0% binding, 0% signal became 100% binding, etc.). 50% binding was determined by calculation off a three point linear curve (points at A~600~ 0.067, 0.022, and 0.0074).
Neutralization of Stx2f mediated cytotoxicity in Vero cells {#s2.11}
-----------------------------------------------------------
Vero (African green monkey kidney) cells \[[@B18]\] were prepared as previously described \[[@B33]\]. Briefly, Vero cells were dispensed into 96-well cell culture plates at 10^5^ cells/mL overnight. The media used was Dubecco's Modified Eagle's Medium (DMEM) plus 1x Glutamax (Invitrogen) and 10% FBS (Invitrogen). Cells were treated at 4°C for 1 hour with 100 µl/well of Stx2f (5 ng/mL) or Stx2f pre-incubated with mAbs (100 µg/mL in Vero cell media) for one hour at RT. The media containing unbound toxin was then removed and replaced by fresh media, and cells were shifted to 37°C to grow for 24 hours. The cells were then lysed using 100 µl/well 1/5 dilution of CellTitre-Glo reagent (Promega), and luminescence was measured using a Victor II plate reader. The CellTiter-Glo Assay relies on the properties of a thermostable luciferase, which generates a stable luminescent signal in the presence of ATP and luciferin. The luminescent signal is proportional to the amount of ATP present, while the ATP is directly proportional to the number of metabolizing cells present in culture. The wells containing only 5 ng/mL toxin (without mAbs) were defined as 100% cytotoxicity or 0% neutralization and the negative control (no antibody or toxin) was set to 0% cytotoxicity (100% cell viability). Photos were taken using a Leica DM IL microscope at 200x magnification ([Figure S1B](#pone.0076563.s001){ref-type="supplementary-material"}). For Stx2 subtype treatments of Vero cells ([Figure S2](#pone.0076563.s002){ref-type="supplementary-material"}), strains expressing all seven subtypes of Stx2 were induced with 50 ng/mL MMC. The media was centrifuged to remove bacterial cells, then filter-sterilized (0.2 µm). Cell-free media (5 μL/well) containing Stxs was added to Vero cells, incubated for 1 hour at 4°C, and replaced with fresh media. Photos were taken using a Leica DM IL microscope at 200x magnification ([Figure S2](#pone.0076563.s002){ref-type="supplementary-material"}).
Antibody affinity measurement {#s2.12}
-----------------------------
Antibody affinity to Stx2f was measured using an Octet QK system (Forte-bio, Menlo Park, CA) as described previously \[[@B36]\]. The biotinylated antibodies were coupled to streptavidin biosensors at 10 µg/mL in PBS. Probes coupled to antibody were incubated with Stx2f at four different concentrations (142, 71, 36, and 18 nM), then allowed to dissociate in PBS. Binding kinetics were calculated using the Octet QK software (Data Acquisition 7.0).
Colony immunoblots {#s2.13}
------------------
Strains RM7007 (Stx2f) and FSIS EC465-97 (GFP-labeled O157:H7, Stx-negative) were grown in LB broth for 12 hours at 37°C with agitation. RM10638 (Stx2a) was additionally grown for [Figure S3](#pone.0076563.s003){ref-type="supplementary-material"}. Following this, the A~600~ of each of these cultures was set to 2 and 100 µL of each culture was combined (for a total volume of 200 µL \[or 300 µL for [Figure S3](#pone.0076563.s003){ref-type="supplementary-material"}\]). The mixture was diluted 10^6^ times in LB broth and 100 µL of this dilution was plated on LB agar plates supplemented with 50 ng/mL MMC, using sterile glass beads for distribution. The LB agar plates were incubated for 12 hours at 37°C. A rectangular cut of PVDF membrane was then wetted in methanol and incubated in water for 5 minutes. After blotting the membrane dry, it was placed upon the LB plate and incubated at 4°C for 2 hours. It was then incubated in a boiling hot 2% SDS solution for 5 minutes, and this step was repeated to kill all residual bacteria. The membrane was then rinsed three times in PBS, for 5 minutes each time with agitation to remove cell debris. The membrane was then blocked in 2% ECL Prime blocking agent/PBST for 1 hour at RT. Afterwards, the membrane was incubated with a solution of 1 µg/mL mAb Stx2f-4 in blocking solution for 1 hour at RT. Following this, the membrane was washed thrice (3 minutes each) with PBST then incubated with a 1/10,000 dilution of GAM-HRP (Promega) in blocking solution for 1 hour at RT. After 4 washes with PBST (5 minutes each), the blots were developed with Lumigen TMA-6 (Lumigen) substrate. Colony blots were visualized with a 2 minute exposure using a FluorChem HD2 (Alpha Innotech). Photos for plates were taken using an iPhone 4S and GFP-labeled control cells were illuminated on a UV box (U: Genius, Syngene, Cambridge, UK). The colony immunoblot was false-colored (red) in Photoshop (Adobe) to enhance contrast for an overlay picture. For plates supplemented with chicken breast extract (see the [Enzyme-linked immunosobent assays]{.ul} section), 50 μL/plate extract was dispensed on LB plates containing 50 ng/mL MMC and allowed to absorb before plating 50 µL of the bacterial mixture (diluted 5 x 10^5^ in LB broth).
Stx2a and Stx2f PCR {#s2.14}
-------------------
Diagnostic colony PCR ([Figure S3B](#pone.0076563.s003){ref-type="supplementary-material"}) was performed to confirm the specificity of the colony immunoblot assay for [Figure S3A](#pone.0076563.s003){ref-type="supplementary-material"}. Colonies were tapped with a pipet tip before performing the colony immunoblot, and the bacteria was suspended in 100 µL sterile water. PCR was performed using previously described primers and protocols \[[@B27]\], with a few modifications. The Stx2a PCR used the primers stx2a-F2 and a 1:1 combination of stx2a-R2 and stx2a-R3, for an amplicon of 347 or 349 base pairs. The Stx2f PCR used the primers stx2f-F1 and stx2f-R1, for an amplicon of 324 base pairs. A mixture of 20 µM primers (1.25 µL per 25 µL reaction), 2x GoTaq master mix (Promega) (12.5 µL per 25 µL reaction), bacterial suspension (1 µL of the suspension per 25 µL reaction), and water (up to 25 µL) was cycled 35 times, with an annealing temperature of 64°C.
Results {#s3}
=======
Producing Stx2f mAbs {#s3.1}
--------------------
To generate high-affinity mAbs against Stx2f, we immunized mice with purified recombinant His-tagged Stx2f A subunit (a listing of all bacterial strains used is included in [Table 1](#pone-0076563-t001){ref-type="table"}) \[[@B33]\] and fused the resulting splenocyes to SP2/0 myeloma cells. Splenocyte/myeloma hybridoma fusions, plated into 96-well culture plates (960 wells total), were screened using purified Stx2f \[[@B33]\]. Thirty-seven wells were chosen for further analysis. After repeated expansion and isolation of cells by limiting dilution, four hybridoma cell lines were selected. The antibodies purified from these hybridoma cell lines are designated mAbs Stx2f-1, Stx2f-2, Stx2f-3, and Stx2f-4, and all possessed IgG2 except mAb Stx2f-2 which has an IgG1 isotype ([Table 2](#pone-0076563-t002){ref-type="table"}). All these antibodies bound specifically to the Stx2f A subunit (^≈^32kD) on a western blot and had no discernable affinity to the B subunit (^≈^5kD) ([Figure 1A](#pone-0076563-g001){ref-type="fig"}). All four antibodies bound strongly to purified Stx2f but not to purified Stx2a or partially purified Stx1 in a direct ELISA ([Figure 1B](#pone-0076563-g001){ref-type="fig"}). Mouse mAb VT135/6-B9 (for Stx2a) and VT109/4-E9b (mouse mAb against Stx1 B-subunit from Sifin) were included as controls to confirm the presence of these toxins. Dissociation constants for mAbs Stx2f-1 and Stx2f-2 (0.52 and 0.53 x 10^-9^ M, respectively) were considerably lower than that of Stx2f-4 (8.4 x 10^-9^ M), while no dissociation constant could be calculated for Stx2f-3 despite repeated attempts ([Table 2](#pone-0076563-t002){ref-type="table"}).
10.1371/journal.pone.0076563.t002
###### Properties of Stx2f monoclonal antibodies.
Antibody Isotype K~D~ (x 10^-9^ M)
---------- ------------- -------------------
Stx2f-1 IgG2, kappa 0.516 ± 0.14
Stx2f-2 IgG1, kappa 0.533 ± 0.39
Stx2f-3 IgG2, kappa nd\*
Stx2f-4 IgG2, kappa 8.35 ± 1.1
\* Not detectable.
Developing a sensitive and specific sandwich ELISA for Stx2f {#s3.2}
------------------------------------------------------------
In order to establish a sensitive immunoassay, all possible capture/detector combinations of mAb pairs were evaluated in a sandwich ELISA format using a biotinylated antibody as a detector. The following capture/detector antibody pairs are highly effective at detecting Stx2f purified toxin: mAb Stx2f-1/2, Stx2f-1/4, Stx2f-2/1 and Stx2f-2/4 ([Figure 1C](#pone-0076563-g001){ref-type="fig"}). mAbs Stx2f-1 and Stx2f-2 were very effective as capture antibodies, Stx2f-4 was the best detector, and Stx2f-3 was not compatible with any of the other Stx2f antibodies, either as a capture or detection antibody. The most sensitive antibody pair employed mAb Stx2f-1 as a capture antibody and Stx2f-4 as a detector. This pair detected purified toxin down to 0.123 ng/mL ([Figure 2A](#pone-0076563-g002){ref-type="fig"}). In addition, over the range of toxin tested (0-60 ng/mL), the assay was linear, with an R^2^ value of 0.9979. The specificity of the mAb Stx2f-1/Stx2f-4 pair was evaluated using filtered cell culture media containing different Stx2 subtypes, induced with MMC (purified toxin is not available for many of these subtypes). The MMC-induced media from all seven subtypes of Stx2 was toxic to Vero cells, confirming the presence of toxin ([Figure S2](#pone.0076563.s002){ref-type="supplementary-material"}). The mAb Stx2f-1/4 ELISA did not recognize any other Stx2 subtype tested, suggesting that this combination of antibodies is specific to Stx2f ([Figure 2B](#pone-0076563-g002){ref-type="fig"}).
{#pone-0076563-g002}
Poultry is an emerging source of *E. coli* contamination \[[@B37],[@B38]\]. Since Stx2f-producing *E. coli* has been isolated from an avian vector (pigeon), it is reasonable to assume that Stx2f-producing bacteria may soon be associated with poultry meat. We therefore examined the effect of chicken extract on the Stx2f sandwich ELISA assay (mAb Stx2f-1/4). Stx2f was spiked in chicken breast extract (1/10 dilution in PBS). Although the overall signal of the sandwich ELISA was reduced somewhat, this had little detrimental impact upon the limit of detection for Stx2f, which rose to 0.210 ng/mL ([Figure 2A](#pone-0076563-g002){ref-type="fig"}).
Examining the preference of Stx2f to receptor using Stx2f mAbs {#s3.3}
--------------------------------------------------------------
Our previous results suggest that Stx2f is able to bind both Gb3- and Gb4-LPS receptors. Using the Stx2f-1/4 antibody pair, it was possible for us to perform sandwich ELISAs to confirm the receptor preference of Stx2f. Gb3-LPS or Gb4-LPS-expressing *E. coli* cells were pre-incubated with Stx2a or Stx2f toxin, the cells were then removed, and the remaining toxin was quantified by ELISA using the corresponding coating/detection antibody combination (mAb Stx2f-1/4 for Stx2f and mAb VT136/8-H4/VT135/6-B9 for Stx2a). With 50% of toxin bound at an A~600~ of 0.017 for Gb3-LPS and 0.018 A~600~ for Gb4-LPS, Stx2f bound strongly to both Gb3-LPS and Gb4-LPS receptors. Stx2a only bound to Gb3-LPS in this assay, with 50% bound at an A~600~ of 0.03 ([Figure 3](#pone-0076563-g003){ref-type="fig"}). Control cells (CWG308, [Table 1](#pone-0076563-t001){ref-type="table"}) did not bind either Stx2a or Stx2f (data not shown \[[@B33]\]).
{#pone-0076563-g003}
*In vitro* toxin neutralization {#s3.4}
-------------------------------
Antibodies against Stx2 B subunits tend to possess stronger neutralizing potential in cell-based assays than those against the A subunit \[[@B36]\], presumably by disrupting the binding of the toxin to Gb3/4 binding sites. However, antibodies against the A subunit that can reduce the *N*-glycosidase activity of Stx2 and provide some toxin neutralizing activity have been reported \[[@B39]\]. Therefore, we investigated whether our panel of Stx2f antibodies, administered at a 100 µg/mL concentration, can protect Vero cells from Stx2f toxicity. Though none of our antibodies conferred full protection from Stx2f, three of the four antibodies partially mitigated toxicity, with the best being mAb Stx2f-4 at 43% neutralization ([Figure 4](#pone-0076563-g004){ref-type="fig"}). These antibodies were about two-thirds as effective at neutralizing toxin at a lower concentration (10 µg/mL) (data not shown). In some circumstances, different partially neutralizing antibodies can synergize and strongly neutralize when combined \[[@B40]\]. While the combination of mAbs Stx2f-1 and Stx2-4 (the best sandwich ELISA combination) did not fully neutralize Stx2f, it did protect better than either of these antibodies separately, at 62% neutralization ([Figure S1A](#pone.0076563.s001){ref-type="supplementary-material"}, S1B).
{#pone-0076563-g004}
Identification of Stx2f-producing *E. coli* present in chicken extracts by colony immunoblot assay {#s3.5}
--------------------------------------------------------------------------------------------------
Detecting STEC in environmental or clinical samples is often a lengthy process, involving selection and isolation, usually followed by PCR or immunological confirmation of a pure culture of the organism. Since there is no accepted immunological assay for Stx2f and PCR does not reveal expression of the toxin, we sought to provide a plate-based method of detecting Stx2f-expressing *E. coli* using colony immunoblot assays. Growing STEC on agar plates supplemented with mitomycin C (MMC) is a way to maximize sensitivity in this type of assay \[[@B41]\]. We used a GFP-tagged O157:H7 marker strain (FSIS EC465-97 \[[@B42]\]), which has a genetic background analogous to STEC strains except it does not contain any functional *stx* genes, as a control and mixed it with the Stx2f-producing strain. Under our experimental conditions, all negative control colonies (fluorescent green) were not detected by mAb Stx2f-4 in the colony blot ([Figure 5A](#pone-0076563-g005){ref-type="fig"}-GFP), whereas all Stx2f-producing colonies which did not fluoresce were positive for Stx2f ([Figure 5A](#pone-0076563-g005){ref-type="fig"}-Stx2f blot and Overlay). To verify that the Stx2f-immunoblot assay does not cross-react with Stx2a-expressing colonies, we mixed the Stx2f-strain with the Stx2a- and GFP-strains and plated the mixture of these three strains on MMC plates. We then performed the Stx2f colony immunoblot, along with a colony PCR to detect the presence of the *stx2f* and *stx2a* genes. As we expected, every colony that was positive by Stx2f immunoblot was also positive for *stx2f* -PCR. Additionally, no Stx2a or GFP-O157:H7 colonies were detected by Stx2f immunoblot, meaning that this assay is specific to Stx2f colonies ([Figure S3A](#pone.0076563.s003){ref-type="supplementary-material"}, S3B).
{#pone-0076563-g005}
Since poultry is a likely source of future contamination by Stx2f-expressing *E. coli*, sterile homogenized chicken breast extract was added to a subset of immunoblot plates to test for matrix effects. Surprisingly, both *E. coli* strains tested (Stx2f-producing *E. coli* and GFP-O157:H7) had smaller colonies, suggesting that the chicken breast was inhibiting their growth on plates ([Figure S4](#pone.0076563.s004){ref-type="supplementary-material"}), although this extract didn't inhibit their growth in liquid media (data not shown). The Stx2f-producing bacteria could unambiguously be identified by colony immunoblot using our mAb Stx2f-4, ([Figure 5B](#pone-0076563-g005){ref-type="fig"}), however, suggesting that this colony immunoblot assay could be applied to poultry samples.
Discussion {#s4}
==========
Since STEC are among the most dangerous foodborne pathogens, their detection and study is critical to maintaining a safe food supply and preventing deadly outbreaks. Stx2 is a particularly important virulence factor in EHEC and EHAEC, and is thought to be a causative agent in the development of HUS. The seven Stx2 subtypes, unlike those of Stx1, are very diverse, both at the amino acid and biochemical levels. Certain subtypes, like Stx2e and Stx2f, are not commonly associated with serious human disease, but this may be due to their current serotype distribution. Only a few immunoassays claim to detect Stx2f and the sensitivity and specificity of these assays is unclear. Here, we detail new mAbs for detecting the Stx2f subtype. The four antibodies developed in this study are specific to Stx2f and do not cross-react with Stx2a or Stx1. A sensitive new immunoassay for Stx2f was developed using a compatible pair of these Stx2f mAbs (Stx2f-1/Stx2f-4) in a sandwich ELISA format. The limit of detection for this new Stx2f immunoassay is 0.123 ng/mL and the specificity of this ELISA is to Stx2f uniquely: it does not cross-react with any of the other Stx2 subtypes. Using this ELISA, Stx2f was confirmed to bind to both Gb3-LPS and Gb4-LPS with similar affinities to each. Three of the four antibodies partially neutralize purified Stx2f toxin in a Vero cell viability assay, and the combination of Stx2f-1 and Stx2f-4 neutralize Stx2f even more effectively. Additionally, one of these new Stx2f antibodies (Stx2f-4) was validated for the detection of Stx2f-expressing colonies in a colony immunoblot assay, a technique that may prove useful in characterizing new Stx2f-expressing STEC isolates.
Although Stx2f-expressing STEC is not commonly detected in human disease, if and when the phage carrying the *stx2f* gene establishes itself in an *E. coli* strain that is capable of causing human disease, it could become a serious human pathogen. This new sandwich ELISA will be extremely valuable for future diagnosis and source tracking of contaminations. Some additional studies are needed to fully evaluate this assay and prepare it for use in food safety. A full evaluation of the effect of various common matrices (beef, milk, etc.) upon the sensitivity and reliability of this assay is currently underway. However, we predict that chicken or turkey meat is the most likely immediate source of contaminations, as Stx2f-producing strains were first isolated from an avian species (pigeon). Thus, chicken breast was used as a matrix for the Stx2f colony immunoblot and mAb Stx2f-1/4 sandwich ELISA, and both assays appear to be compatible with this matrix.
Ideally, ELISA assays for determining antibody sensitivity and specificity should be conducted using pure antigen, but bacterial supernatant containing Stx2 subtypes is frequently used as a toxin source. This is mainly due to the lack of pure toxin for some of the Stx2 subtypes. There are several drawbacks of using crude toxins in such studies. Most importantly, different STEC isolates and strains could exhibit different expression levels of Stx subtypes, which could easily explain contradictory results in previous studies \[[@B6],[@B34]\]. Additionally, Stx-expressing isolates are very diverse, and some may possess molecules that interfere with an assay, generating false-positive or false-negative results. Western blotting is a highly informative assay that should be included in assay evaluations if pure antigen is unavailable.
The Stx2f colony immunoblot could greatly improve detection and isolation of any Stx2f-expressing bacteria, such as STEC, *Enterobacter cloacae*, which may express Stx as well \[[@B15]\], and *Escherichia* *albertii*, which may be a reservoir for the phage carrying Stx2f \[[@B14]\]. Strategies such as this and colony DNA hybridization are helpful for pinpointing specific STEC colonies in complex environmental or clinical samples and recovering it. An advantage of Stx colony immunoblotting over colony DNA hybridization is that the colony immunoblot can determine whether an STEC strain is actually expressing Stx, rather than just carrying the gene. Additionally, the signal of a colony immunoblot can be amplified using the appropriate culture conditions, in this case, supplementation with 50 ng/mL MMC \[[@B41]\]. Our Stx2f immunoblot detected only Stx2f-expressing colonies, not Stx2a-expressing colonies or toxin-free O157:H7 control strains ([Figure 5A](#pone-0076563-g005){ref-type="fig"}, [Figure S3](#pone.0076563.s003){ref-type="supplementary-material"}), even in the presence of chicken extract ([Figure 5B](#pone-0076563-g005){ref-type="fig"}). This suggests that the mAb Stx2f-4 colony immunoblot may be uniquely specific to Stx2f, and could be used to isolate Stx2f-producing bacteria regardless of the species, matrix, or presence of other Stx. Ultimately, we envision an all-encompassing colony immunoblot, ELISA, and/or lateral flow assay where all Stx subtypes can be detected separately from each other, in a wide variety of matrices. Although assays with wide ranges of specificity are very useful for monitoring the safety of food sources, an analogous subtype-specific kit would be even more valuable for tracking the migration prevalence of Stx subtypes and Stx-encoding phages.
Supporting Information
======================
######
A. Neutralization of Stx2f in a Vero cell assay with different combinations of mAbs against Stx2f. All neutralizations were conducted using 5 ng/mL purified Stx2f (except for the "No toxin" \[PBS\] control) and 100 µg/mL total concentration of mAbs. B. Microscope photographs are displayed for these assay wells with the indicated treatments.
(TIF)
######
Click here for additional data file.
######
**Stx2 subtype cytotoxicity.** Vero cells (seeded at 10^5^ cells/well and grown for 12 hours at 37°C) were treated with 5 μL/well bacterial cell-free supernatant (induced by 50 ng/mL MMC) containing the indicated Stx2 subtype for 16 hours at 37°C. All seven subtypes are expressed and are toxic to Vero cells.
(TIF)
######
Click here for additional data file.
######
A. Stx2f colony immunoblot with Stx2f- and Stx2a-expressing strains, as well as GFP-labeled control cells. The same plate portion is displayed for all four panels. B. Confirmation of the presence of the *stx2a* and *stx2f* genes by colony PCR. The Stx2a-specific PCR band is \~347 base pairs; the Stx2f-specific band is 424 base pairs. All colonies that are neither green (GFP) nor red (Stx2f-producing) are Stx2a-producing, confirmed by colony PCR (colony no. 1, 4, 7, 9, 12, and 14).
(TIF)
######
Click here for additional data file.
######
**The effect of chicken extract on colony size.** Adding 50 µL of chicken breast extract slows the growth of the Stx2f-expressing strain and the FSIS EC465-97 fluorescent control strain. The colonies on these plates are derived from the same two strain mixture as in [Figure 5B](#pone-0076563-g005){ref-type="fig"}.
(TIF)
######
Click here for additional data file.
We would like to acknowledge Robert Mandrell, Anne Bates, and Michael Cooley for providing STEC subtype *E. coli* strains, as well as James Paton for the Gb3-LPS- and Gb4-LPS-expressing *E. coli* strains (CWG308:pJCP-Gb3, CWG308 pJCP-lgtCDE, and CWG308 control cells). We would also like to thank Todd J. Ward for the fluorescent FSIS EC465-97 strain and Paul Merrill for his assistance in the Stx2f hybridoma development.
[^1]: **Competing Interests:**The authors have declared that no competing interests exist.
[^2]: Conceived and designed the experiments: XH CS. Performed the experiments: CS SP. Analyzed the data: XH CS SP LS. Contributed reagents/materials/analysis tools: XH LS. Wrote the manuscript: XH CS.
| {
"pile_set_name": "PubMed Central"
} |
Background
==========
Every scientific field that processes information with the aid of computers needs to maintain and preserve its technical illustrations in a machine-readable fashion for later reuse. For this, the structure of an illustration has to be decomposed into its geometric primitives. The more previous knowledge of the image content is available to the computer, the less errors will occur during the decomposition process. Computer programs today digitally capture and archive decades-old architectural drawings, for which the usage of symbols, icons and font types is standardized. Conversely, no computer program can automatically convert arbitrarily shaped phylogenetic trees from an illustration into a machine-readable expression, e.g., the Newick format \[[@B1]\]. The styles in which phylogenetic trees have been published are as manifold as are the software packages used for the creation of the trees and the pictures. A comprehensive list of such programs is published in \[[@B2]\]. Because there are no strict design rules, a program intended for the recognition of arbitrary trees must not assume any previous knowledge beyond the existence of a depicted phylogeny.
TreeSnatcher Plus (Figure [1](#F1){ref-type="fig"}) is not the first program aimed at the digitization of phylogenies. Indeed, TreeThief \[[@B3]\] was the first application that converts a tree image into a computer-readable representation of the tree. It allowed the user to digitize a tree by clicking on each of its nodes in turn. It is restricted to Apple Macintosh computers running Mac OS 9. In 2007, we presented TreeSnatcher \[[@B4]\], an application that identified the topology of an arbitrarily shaped tree (e.g., a figure from a publication) semi-automatically with user interaction. However, it required the user to pre-process an image using an external drawing package, to follow a strict succession of program stages and lacked any Undo functionality. Finally, the program TreeRipper by Joseph Hughes \[[@B5]\] automatically converts images of rectangular trees that fulfil a strict set of criteria into the Newick format. However, TreeRipper's success rate is relatively low, with only about one third of sample images converted correctly \[[@B5]\].
{#F1}
Using already published trees in research projects would be trivial if all phylogeny related data were also published in open-access online repositories, e.g., TreeBASE \[[@B6]\], MorphoBank \[[@B7]\], or Dryad \[[@B8]\]. Leebens-Mack et al. \[[@B9]\] propose a roadmap for the development of minimal reporting standards for phylogenetic analyses, MIAPA (Minimal Information about a Phylogenetic Analysis). They maintain a website on which they discuss potential barriers to re-use data from scientific analyses \[[@B10]\]. In principle, electronic data could also be obtained from the authors. This obvious approach appears not to be practical. In an example from another field of study, 73% of the authors refused to share their data when approached \[[@B11]\].
Thus, to reuse most published phylogenetic results, it appears that reliable digitization of tree images is currently the only realistic option.
Implementation
--------------
TreeSnatcher Plus is an extended and fully re-conceptualized version of TreeSnatcher; it is both easier to use and more accurate than its predecessor. The new program features a graphical user interface that is based on the JAVA Swing API. A more flexible workflow is complemented with multiple Undo functionality and the possibility to restore the program state (\'snapshot\'). The user can now pre-process any image within TreeSnatcher Plus, selecting from a full range of pre-processing tools. The current state of processing can be saved as an image that may contain different layers of visual information. The program calculates the branch lengths in freeform and skewed rectangular trees and can mix calculated and user defined branch lengths. Additionally, the user can modify an existing tree or to construct a new tree.
The application opens image files in the formats PNG, JPG/JPEG or GIF. The PDF format is currently not supported, but tools for the extraction of images from PDF documents are readily available (e.g., the Xpdf suite for Linux operating systems \[[@B12]\]). The program offers the following pre-processing tools, most of which were modified from standard algorithms \[[@B13]\]: pencil, rubber, line, fill, stencil, histogram stretch, colour reduction, gray-scale conversion, local and global thresholding, colour manipulation, inversion, median and minimum filter, blurring and sharpening, lightening and darkening, and thinning \[[@B14]\].
Prior to automated node placement, the user has to prepare the image within TreeSnatcher Plus, analogous to the requirements of its predecessor \[[@B4]\]. In particular, the tree has to be converted into a line drawing without intersections with text or graphics unrelated to the tree topology. If the image does not meet those requirements when the automatic node placement is issued, the tree topology is unlikely to be identified correctly.
Working with TreeSnatcher Plus takes place along a general succession of global tasks, which are executed at least once, either on the whole image or on parts. The user supervises all image manipulations and recognition tasks performed by the program, makes corrections and repeats steps if necessary. This process is explained in detail in the tutorials accompanying the program.
The workflow is thus as follows:
1\. The program reads the specified image file. The user trims and cuts the image at will. In this way, one can select sub-trees or a subset of taxa from the image.
2\. Image pre-processing: The user prepares the image with the pre-processing tools.
3\. Binarization: The user thresholds the image to ensure that the foreground is black and the background is white and both are clearly separated.
4\. Skeletonization: The user semi-automatically thins the foreground of the image portion that contains the tree. This is necessary to enable the program to find the paths between the line intersections (step 8).
5\. Foreground flooding: The user marks a position in the tree. The program colours (\'floods\') the foreground reachable from there. In subsequent steps, the flooded area will be treated as the tree. Everything else is ignored.
6\. Inner nodes and outer nodes placement: The program suggests locations for line intersections and end of lines. These represent branching locations and tips. A logical node is assigned to each location. The user can move, remove and add nodes. In the thinned image, black pixels adjacent to exactly one other black pixel become a tip location. Black pixels adjacent to at least three black pixels are candidates for a branching location. If several candidate pixels are adjacent to each other, the branching location is averaged from their positions.
7\. Choice of tree type: The program can distinguish and calibrate freeform and rectangular trees. The choice of the tree type influences how the program treats branch lengths. The tree type must be chosen prior to step 8.
8\. Recognition of branches: The program traces gapless foreground paths between each pair of nodes in order to find the branches of the tree. If there are several candidate paths, the shortest is selected. If a branch is missing or wrong, the user either modifies the image with the drawing tools and repeats step 8, or he/she drags a new branch and manually specifies its length.
9\. Determination of branch lengths: The accuracy depends on the congruence of thinned tree structure, node placement, and original tree. For freeform trees, the branch length in pixels is based on the entire foreground path between the two defining nodes. For rectangular trees, the branch length is the sum of the lengths of the horizontal path segments. The user may type in self-defined branch lengths and mix them with the calculated lengths. The tree can be scaled using a line of known length in the image, e.g., a scale bar.
10\. Assignment of species names: The user right-clicks on each leaf node in turn in order to type in the corresponding species name.
11\. Choice of the tree root: The program, assisted by the user if necessary, chooses the inner node based on which the rooted Newick expression is calculated.
12\. Construction of the Newick string: The program calculates and displays the Newick tree code for the tree depicted. The user may save it to the clipboard or export it into a text file.
Results
=======
An image that shows a uniformly dark, rectangular phylogenetic tree on a uniformly light background in sufficient resolution, without foreground elements overlapping with the tree, will need almost no pre-processing. If the user then settles for consecutively numbered tip labels, the whole recognition process can be finished within minutes. However, in general there will be image portions which require manual correction, e.g., a branch of the tree is not clearly separated from other foreground elements such as lettering. TreeSnatcher Plus offers a special tool that surrounds black drawings with a white border.
For the determination of branch lengths, the program needs to assess the path length in pixels between branching positions on the skeletonized foreground. Additionally, it must reliably detect bends in a branch and horizontal branch portions. These tasks work better if the structures come with a sufficiently high number of pixels. The better the branches in the original image and those in the skeletonized image align, the more accurate are the branch lengths.
The time needed for the complete tree digitization depends on several factors, among them image size and quality, tree size and complexity, whether or not branch lengths are desired and species names are typed in, how much pre-processing is necessary, and how experienced the user is in the usage of TreeSnatcher Plus.
We benchmarked TreeSnatcher Plus using the set of 100 rectangular trees published by Hughes \[[@B5]\], nine additional images with non-standard tree styles, and another image modified from the benchmark set of Hughes. The topologies of all trees and their branch lengths were recognized correctly, with necessary user interaction ranging from minor to extensive (Additional file [1](#S1){ref-type="supplementary-material"}, table \'BenchmarkTreeSet\'). All tip labels were typed manually. As TreeSnatcher Plus is not meant to work autonomously, the user is required to initiate all pre-processing and analysis steps. For different users, the time needed for a particular task may vary. For the benchmark set \[[@B5]\], the time needed for topology reconstruction was on average 165 s, ranging from 30 to 1,800 s on a PC equipped with a Core i7-960 processor. Typing the tip labels required on average 4.4 min, the minimum time was 0 min for an image without labels and the maximum time 35 min. The shortest tree digitisation job finished within one minute, the longest job required 45 min. Recognizing the topology of a tree with 100+ tips (image 1471-2148-6-93 in the tree set by Hughes) and entering the tip labels required 17 min. Using the TreeRipper web frontend \[[@B5]\], we obtained the topology and the tip labels for the same tree within five minutes, the processing time was 272 s. However, all tip labels needed manual correction.
Additionally, we processed nine images of non-rectangular trees (Additional file [1](#S1){ref-type="supplementary-material"}, table \'InternetTrees\'). Again, we obtained the topologies and the branch lengths for all trees. TreeSnatcher Plus can measure the branch lengths in freeform trees but not in circular and polar trees. For three circular trees (\'bustard\', \'TreeofLife\' and \'Phylogenetic_Tree_of_Life\'), we therefore approximated the lengths of the horizontal branch portions with the \'Line Selection\' tool. In one case (\'vert_tree\'), we needed to retrace the tree manually.
We rotated image 1471-2148-6-99-1 from the benchmark set by 6° in order to test how well the program can recognize the rectangular branch portions in the new image (\'RotatedTree\'). While the editing steps were similar for the original and the modified image, TreeSnatcher Plus failed to recognize 10 out of a total 81 horizontal branch portions, compared to 3 in the original image.
In general, the correction of overlapping structures in a tree and typing the branch lengths are the most demanding tasks. For some images, we needed several attempts until we found a suitable order of processing steps.
For all jobs, we provide TreeSnatcher Plus snapshot data that can be reloaded into the program in order to reproduce the recognition results (Additional file [2](#S2){ref-type="supplementary-material"}, Additional file [3](#S3){ref-type="supplementary-material"}, Additional file [4](#S4){ref-type="supplementary-material"}, Additional file [5](#S5){ref-type="supplementary-material"}, Additional file [6](#S6){ref-type="supplementary-material"}.
On the TreeSnatcher Plus project website, we offer tutorials and sample recognition projects for the application. They provide a comprehensive overview about the performance of the program and the expected effort with different illustration styles and topologies.
Discussion
==========
Given the performance of today\'s image analysis and OCR methods, two distinct strategies are feasible: either a fully-automated approach requiring one specific tree type and a fixed illustration style, or a semi-automated approach for arbitrary trees.
TreeRipper, which to our knowledge is the only currently available program with comparable functionality, aims at a fully automatic recognition. To achieve this, it restricts itself to rectangular trees fulfilling a set of stringent criteria. For the trees fulfilling these criteria in the same benchmark set analysed here, TreeRipper was able to recognise the topology correctly in 32% of cases without any manual pre-processing.
TreeSnatcher Plus, on the other hand, accepts the necessity for manual image pre-processing and thus achieves a 100% success rate. It allows the user to process virtually any phylogenetic tree, albeit sometimes with extensive user interaction. On average, extraction of one phylogenetic tree topology required less than three minutes. Currently, tip names have to be entered by hand - using OCR here is very difficult to implement, as the tip labels can be anywhere on the image and in any orientation. TreeSnatcher Plus is easy to install, the single prerequisite being a working Java 1.6 Runtime Environment.
Further improvements are planned. The next version of TreeSnatcher Plus will be able to compute branch lengths for circular tree topologies. Moreover, it is planned to include an experimental OCR option for the recognition of tip label names in rectangular trees. The thinning technique could be improved as it is not tailored specifically to TreeSnatcher Plus.
Conclusions
===========
Although TreeSnatcher Plus does not work fully automatically, it can be used to preserve virtually any phylogenetic tree for future research. Today, automatic digitization, let alone batch processing, of even a subclass of phylogenetic tree images including labels seems hardly realizable. We are nevertheless convinced that novel programs will recognize a large number of different tree topologies in diverse styles. They will combine classical methods from the field of pictorial pattern recognition and image segmentation with new approaches. Until then, TreeSnatcher Plus can be used, with an acceptable effort for the user, for the semi-automated recognition of arbitrarily shaped trees in images. If the images to be digitised fulfil its requirements, TreeRipper may instead be used for the automatic recognition of topologies.
A greater acceptance of increased minimum data reporting standards in phylogenetic research would guarantee that phylogenetic data communicated in future research papers is machine-readable and available to other scientists. It has been mandatory for years to publish DNA sequences electronically in one of the universally accepted formats, and one may hope that a similar requirement will be enforced by scientific journals for phylogenies.
Availability and requirements
=============================
Project name: TreeSnatcher Plus: capturing phylogenetic trees from images
Project home page: <http://www.cs.uni-duesseldorf.de/AG/BI/Software/treesnatcher/>
Operating Systems: Windows, Mac OS X, Linux
Programming language: Sun/Oracle Java 1.6
Other requirements: Sun/Oracle Java Runtime Environment Version 1.6 or higher
License: GNU General Public License
Competing interests
===================
The authors declare that they have no competing interests.
Authors' contributions
======================
TL developed the idea, realized the program, tested the software and drafted the manuscript. MJL and AVH helped to shape the features of the application and revised and approved the manuscript.
Supplementary Material
======================
###### Additional file 1
MS Excel XLS table with Benchmark tree set results and results on non-rectangular trees.
######
Click here for file
###### Additional file 2
ZIP files containing several folders, each of which with TreeSnatcher Plus snapshot files, the original image and a text file.
######
Click here for file
###### Additional file 3
ZIP files containing several folders, each of which with TreeSnatcher Plus snapshot files, the original image and a text file.
######
Click here for file
###### Additional file 4
ZIP files containing several folders, each of which with TreeSnatcher Plus snapshot files, the original image and a text file.
######
Click here for file
###### Additional file 5
ZIP files containing several folders, each of which with TreeSnatcher Plus snapshot files, the original image and a text file.
######
Click here for file
###### Additional file 6
ZIP files containing several folders, each of which with TreeSnatcher Plus snapshot files, the original image and a text file.
######
Click here for file
Acknowledgements
================
I am very thankful to Martin J. Lercher for making it possible to develop TreeSnatcher into TreeSnatcher Plus and for supervising the development process. I am also very thankful to Arndt von Haeseler who supervised the design of the first incarnation of the program, TreeSnatcher. I thank Yannick Schrader-Schilkowsky who proofread and followed the tutorials and pointed out discrepancies.
This work was supported by the Wiener Wissenschafts-, Forschungs- und Technologiefonds awarded to Arndt von Haeseler; and the DFG (Deutsche Forschungsgemeinschaft) through a Collaborative Initiative (SFB 680).
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-marinedrugs-15-00196}
===============
Under exposure of environmental ultraviolet (UV) radiation, certain living organisms have developed characteristic defense mechanisms to diminish the adverse effects of UV radiation, including DNA damage and the production of reactive oxygen species (ROS) \[[@B1-marinedrugs-15-00196],[@B2-marinedrugs-15-00196],[@B3-marinedrugs-15-00196]\]. Recent studies have reported that many photosynthetic marine organisms can synthesize secondary metabolites with UV-absorbing capacity such as mycosporine-like amino acids (MAAs), as one of the most effective UV protection mechanisms \[[@B4-marinedrugs-15-00196],[@B5-marinedrugs-15-00196]\]. These compounds have characteristic chemical structure with either an aminocyclohexenone or an aminocycloheximine ring, which provides them with ability to absorb UV radiation and be resilient to DNA damage by harmful UV radiation. MAAs have been found in a wide range of marine organism exposed to environmental UV radiation, and the chemical structures of more than 30 different MAAs have been characterized \[[@B6-marinedrugs-15-00196],[@B7-marinedrugs-15-00196],[@B8-marinedrugs-15-00196],[@B9-marinedrugs-15-00196]\]. Although a number of recent reports for MAAs in various marine organisms refers mostly to their UV protective ability, scientific evidence is increasingly accumulating that MAAs contribute other functional roles such as antioxidant activity and osmotic regulation. For example, porphyra-334, which has characteristic chemical structure for UV absorption ([Supplementary Materials](#app1-marinedrugs-15-00196){ref-type="app"}, [Figure S1](#app1-marinedrugs-15-00196){ref-type="app"}), exerts potent antioxidant activity and prevents cellular damage caused by UV-induced ROS with free radical scavenging capacity \[[@B10-marinedrugs-15-00196],[@B11-marinedrugs-15-00196],[@B12-marinedrugs-15-00196]\]. This fact suggests that MAAs can play crucial roles as antioxidant molecules to modulate cellular processes affected by ROS, such as DNA damage and apoptosis. In addition to antioxidant activity of MAAs, these compounds could regulate osmotic pressure within the cell \[[@B13-marinedrugs-15-00196]\]. To maintain the essential osmotic balance, most organisms accumulate small weight molecules without charge into cells, which can function as so-called "osmotic solutes". In this aspect, MAAs, which are small uncharged organic molecules, can contribute to intracellular osmotic pressure and provide organism with capability to adapt to extreme environments with high salt concentrations. Despite abundant ecological and physiological studies available on the functional roles of MAAs, our understanding of their roles at the molecular level remains poor. Recently, molecular biological studies on the roles of MAAs have been just begun. For example, MAAs from green algae protect skin against UV-induced skin damage through recovery of UV-suppressed expression of skin aging-related genes \[[@B14-marinedrugs-15-00196],[@B15-marinedrugs-15-00196],[@B16-marinedrugs-15-00196],[@B17-marinedrugs-15-00196],[@B18-marinedrugs-15-00196],[@B19-marinedrugs-15-00196]\]. In addition, they function as effective drugs on immunomodulatory effects \[[@B20-marinedrugs-15-00196]\] and the wound healing process in human keratinocytes through the activation of the various genes, such as focal adhesion kinases (FAK), extracellular signal-regulated kinases (ERK), and c-Jun *N*-terminal kinases (JNK) \[[@B21-marinedrugs-15-00196]\]. Nevertheless, there is a need for extensive molecular mechanism studies to explore the photoprotective roles of MAAs so that a range of industrial and pharmaceutical applications can be found. Many studies have focused on marine organisms as a source of natural bioactive molecules having a photoprotective role, their biosynthesis and commercial application \[[@B4-marinedrugs-15-00196],[@B5-marinedrugs-15-00196]\]. In this aspect, MAAs have attracted considerable research attention in both industrial and pharmacological fields. Recently, we have demonstrated that the expression of skin aging-related genes such as procollagen C proteinase enhancer (PCOLCE), elastin and involucrin were modulated by porphyra-334 in a dose dependent manner \[[@B11-marinedrugs-15-00196]\], and porphyra-334 significantly attenuates UV-induced apoptosis in HaCaT cells through the activation of caspase pathway \[[@B22-marinedrugs-15-00196]\]. In the present study, we firstly investigated the comprehensive molecular networks that are associated with the functional roles of porphyra-334 as an UV-absorbing substance in human keratinocyte.
2. Results and Discussion {#sec2-marinedrugs-15-00196}
=========================
2.1. Porphyra-334-Modulated Differentially Expressed Genes (DEGs) {#sec2dot1-marinedrugs-15-00196}
-----------------------------------------------------------------
Gene expression profiling by high throughput sequencing was performed to investigate the effect of porphyra-334 on UV-modulated transcriptome in UV-exposed HaCaT cells. We focused on the identification of genes whose expression was significantly altered in the porphyra-334 treated group compared with the non-treated group (*p* value \< 0.05). The application of this threshold led to the identification of 447 DEGs, of which 267 were over-expressed and 180 were under-expressed in porphyra-334 treated group ([Figure 1](#marinedrugs-15-00196-f001){ref-type="fig"}). The largest number of up- and down-regulated genes, 103 and 71, exhibited about two-fold increase in their transcriptional levels (2 ≤ x \< 3) or less than threefold decrease (0.3 ≤ x \< 0.4 or 0.4 ≤ x \< 0.5), respectively ([Figure 2](#marinedrugs-15-00196-f002){ref-type="fig"}). To gain a better understanding about their biological function, gene set enrichment analysis was performed to identify significantly over- and under-represented gene ontology (GO) categories ([Supplementary Materials](#app1-marinedrugs-15-00196){ref-type="app"}, [Tables S1 and S2](#app1-marinedrugs-15-00196){ref-type="app"}, respectively). Approximately 17% of up-regulated genes in porphyra-334 treated cells were in the top five canonical biological processes ([Figure 2](#marinedrugs-15-00196-f002){ref-type="fig"}B) which were strongly associated with the immune response, regulation of transcription and RNA metabolic process ([Figure 2](#marinedrugs-15-00196-f002){ref-type="fig"}A). Genes highly activated in response to the treatment of porphyra-334, including *Pla2g7*, *Pitx2*, *Gpr34* and *Fbf1* genes, are specifically expressed in the immune response or Wnt signaling pathway \[[@B22-marinedrugs-15-00196],[@B23-marinedrugs-15-00196],[@B24-marinedrugs-15-00196],[@B25-marinedrugs-15-00196]\] which is necessary for proper development and regeneration of various tissues including bone, heart and muscle. In addition, it has been known that this pathway was clinically important because its dysregulation can lead to various diseases, including breast, prostate, glioblastoma, and diabetes \[[@B26-marinedrugs-15-00196],[@B27-marinedrugs-15-00196],[@B28-marinedrugs-15-00196],[@B29-marinedrugs-15-00196]\]. The expression of these genes was verified by qRT-PCR analysis ([Supplementary Materials](#app1-marinedrugs-15-00196){ref-type="app"}, [Figure S2](#app1-marinedrugs-15-00196){ref-type="app"}). In contrast, approximately 14% of the total down-regulated genes in porphyra-334 treated cells involved in the top five biological process ([Figure 2](#marinedrugs-15-00196-f002){ref-type="fig"}A) which were strongly related to cell-cell adhesion, biological adhesion and regulation of transcription ([Figure 2](#marinedrugs-15-00196-f002){ref-type="fig"}B). Among them, the significantly down-regulated genes, such as *Rasd1*, *Fgf12*, *Nkx2-5* and *Cpeb1*, were involved in these categories. Notably, they are directly or indirectly involved in Notch signaling pathway \[[@B30-marinedrugs-15-00196],[@B31-marinedrugs-15-00196],[@B32-marinedrugs-15-00196],[@B33-marinedrugs-15-00196],[@B34-marinedrugs-15-00196],[@B35-marinedrugs-15-00196]\], which is evolutionarily conserved and responsible for cell fate determination in the developing embryo and mature tissue in a highly tissue context- and cell- type-dependent manner. For example, Rasd1 has been shown to interact with EAR2 \[[@B30-marinedrugs-15-00196]\], the orphan nuclear receptor, which can activate Notch signaling \[[@B31-marinedrugs-15-00196],[@B32-marinedrugs-15-00196]\]. In addition, Notch signaling can be promoted downstream of FGF in developmental processes such as stem cells \[[@B33-marinedrugs-15-00196]\], suggesting that FGF12, a component in the FGF signaling pathways, can be involved in Notch signaling. Many studies demonstrate that Notch signaling increases tumor cell proliferation and is activated in the cancer stem-cell pool \[[@B36-marinedrugs-15-00196],[@B37-marinedrugs-15-00196]\]. Thus, pharmacological inhibition of Notch pathway can improve the effectiveness of cancer treatment and patient survival. Recently, many studies report that Wnt and Notch pathways interact either in synergistic or antagonistic manners to exert their biological roles \[[@B36-marinedrugs-15-00196],[@B37-marinedrugs-15-00196]\].
For example, Notch and Wnt signals play essential roles in cell proliferation and tumorigenesis through the synergetic effects between Notch and Wnt signaling; the dramatic proliferative effect was observed when Notch and Wnt signals are normal in the intestinal development \[[@B38-marinedrugs-15-00196]\]. In contrast, in response to UV-radiation, keratinocyte-derived Wnt signaling inhibits Notch signaling through the activation of the Notch inhibitor such as Numb via Wnt pathway-dependent manner \[[@B39-marinedrugs-15-00196]\]. In addition, Notch-dependent suppression of *Pitx2* gene transcription, which consists in Wnt signaling, was demonstrated in cardiac/laterality defects \[[@B40-marinedrugs-15-00196]\]. These observations are consistent with our results that, at least in part, the expression of porphyra-334-regulated genes could be controlled by Wnt or Notch signaling in antagonistic manners in UV-exposed keratinocyte. Our data show that the biological antagonism of the Wnt/Notch signaling pathways may be relevant to the homeostasis of keratinocyte in the human skin microenvironment against the harmful UV-radiation, implying that the functional disturbance between two pathways may contribute to cause cell damage which may potentially induce skin carcinogenesis. In fact, it has been demonstrated that Notch signaling is frequently dysregulated, most commonly by over-activation, across many cancers including melanoma \[[@B41-marinedrugs-15-00196]\]. According to our data, porphyra-334 considerably decreased the expression level of genes related with Notch signaling. This observation suggests that, at least in part, porphyra-334 can inhibit progression of skin carcinogenesis against UV-radiation. In contrast, it has been known that the inhibition of Notch pathway impairs cell growth and induces apoptosis in UV-exposed cells \[[@B38-marinedrugs-15-00196],[@B42-marinedrugs-15-00196]\]. Thus, the significant suppression of Notch-associated genes in UV-exposed HaCaT cells may induce apoptosis or cell death, even in the presence of UV-absorbing compound porphyra-334, implying that porphyra-334 cannot completely protect from the harmful effect of UV-radiation.
2.2. Porphyra-334-Modulated miRNAs {#sec2dot2-marinedrugs-15-00196}
----------------------------------
MicroRNAs (miRNAs), small noncoding RNAs of \~22 nts that mediate posttranscriptional silencing of specific target mRNAs, are increasingly being recognized as an important determinant in a variety of biological processes \[[@B43-marinedrugs-15-00196]\]. Deregulated miRNAs were suggested to exert their function in organism through silencing of key cell fate regulators by directly binding their 3′ UTR \[[@B44-marinedrugs-15-00196]\]. Recently, many studies have been demonstrated that miRNAs play a critical role in molecular regulatory mechanism of UV-induced cellular processes such as DNA damage, apoptosis, deregulation of cell cycle and generation of ROS \[[@B45-marinedrugs-15-00196]\]. To determine the effect of porphyra-334 on the expression pattern of UV-affected miRNAs, we used the nanoString nCounter platform for profiling expression of miRNAs in the porphyra-334 treated cells and control cells; miRNAs up- or down-regulated by \>1.5-fold change and an expression level higher than 100 code counts were further analyzed.
Unexpectedly, widespread repression of miRNAs expression in this model system was observed in response to porphyra-334 treatment (66 p53-repressed miRNAs of 84 total p53-responsive miRNAs; *P* \< 0.05) ([Figure 3](#marinedrugs-15-00196-f003){ref-type="fig"}A). We identified differentially expressed 84 miRNAs (DEmiRNAs) when comparing porphyra-334 treated group and control ([Figure 3](#marinedrugs-15-00196-f003){ref-type="fig"}A), of which 18 were over-expressed and 66 were under-expressed in the porphyra-334 treated group. To further characterize the porphyra-334-regulated miRNAs in UV-exposed cells, we focused on the most robustly-changed miRNAs; among the DEmiRNAs, the highest ranked 18 up- or down-regulated miRNAs stood out as attractive candidates for a role in porphyra-334-related function. Consistent with previous studies \[[@B46-marinedrugs-15-00196],[@B47-marinedrugs-15-00196]\], among the up-regulated miRNAs, hsa-miR-92a-3p, -21-3p and -27a-5p were identified as UV-induced miRNAs in human keratinocyte. For example, has-miR-92a-3p and -27a-5p were highly expressed in UV-exposed human dermal papilla cells, while hsa-miR-21 was activated in mouse fibroblast upon UV exposure. In addition, some up-regulated miRNAs such as hsa-miR-30c, -23a, -148a, -200c, and -181d have been identified as p53-regulated miRNAs \[[@B48-marinedrugs-15-00196],[@B49-marinedrugs-15-00196]\]. In general, p53 can play a key role in response to DNA damage caused by various types of cellular stresses including UV-radiation \[[@B50-marinedrugs-15-00196]\]. In fact, we observed that UV-elevated level of p53 mRNAs was significantly decreased by porphyra-334 ([Supplementary Materials](#app1-marinedrugs-15-00196){ref-type="app"}, [Figure S3](#app1-marinedrugs-15-00196){ref-type="app"}). Therefore, it can be supposed that the above miRNAs are involved in UV-mediated cellular functions in human keratinocyte. Next, to gain insight into the biological functions of 18 porphyra-334-regulated DEmiRNAs, we predicted miRNA target genes using widely accepted miRNA-target-predicting software PicTar and TargetScan 5.1 algorithms which are the most popular miRNA target gene prediction tools. To reduce false positives, target genes represented in both databases were chosen for further analyses. Finally, the number of total putative target genes of the 18 up-regulated miRNAs was 2890 genes, while the down-regulated miRNAs had 1983 genes. Notably, hsa-miR-92a-3p (among the up-regulated miRNAs) and hsa-miR-6807-5p (among the down-regulated miRNAs) had the most target genes, 1329 and 439, respectively ([Figure 3](#marinedrugs-15-00196-f003){ref-type="fig"}). The expression of both miRNAs was verified by qRT-PCR analysis ([Supplementary Materials](#app1-marinedrugs-15-00196){ref-type="app"}, [Figure S3](#app1-marinedrugs-15-00196){ref-type="app"}).
To gain a better understanding of the functional implications of these miRNAs, we performed enrichment analysis for their putative target genes using the DAVID tools ([Supplementary Materials](#app1-marinedrugs-15-00196){ref-type="app"}, [Table S3](#app1-marinedrugs-15-00196){ref-type="app"}). Approximately 24% of putative target genes for the up-regulated miRNAs were strongly associated with the transcription, macromolecular complex subunit organization, macromolecular complex assembly, cell cycle and translational elongation ([Figure 3](#marinedrugs-15-00196-f003){ref-type="fig"}B). In contrast, about 26% of putative target genes for the down-regulated miRNAs were mainly related to protein localization, regulation of transcription, and regulation of RNA metabolic process ([Figure 3](#marinedrugs-15-00196-f003){ref-type="fig"}C). The full list of significant GO terms is given in [Supplementary Materials](#app1-marinedrugs-15-00196){ref-type="app"}, [Table S4](#app1-marinedrugs-15-00196){ref-type="app"}. Next, to examine the biological pathways enriched in the data set in an unbiased, systematic method, gene set enrichment analysis (GSEA) was performed. This software examines gene expression data at the level of gene sets, which are based on existing biological pathway or co-expression data from published research within the Molecular Signature Database \[[@B51-marinedrugs-15-00196]\]. GSEA results were applied to Enrichment Map in Cytoscape to generate a large network of enriched gene sets, and then the large network was clustered to generate sub-networks of interrelated gene sets ([Figure 4](#marinedrugs-15-00196-f004){ref-type="fig"}).
2.3. Target DEGs were Inversely Correlated with Functionally Enriched miRNAs {#sec2dot3-marinedrugs-15-00196}
----------------------------------------------------------------------------
Next, to further understand the role of DEmiRNAs, we searched for their putative target genes from DEGs profiling data. In the DEGs data, genes up- or down-regulated by \>1.2-fold change were selected for further analysis because miRNAs can decrease the mRNA levels of target genes at the different degree. For example, despite a \>95% decrease in protein expression, only small decrease (approximate 20% or less than) in mRNA abundance was observed \[[@B52-marinedrugs-15-00196]\]. Therefore, genes with fold change above 1.2 were considered as their putative target genes. Application of this threshold led to the identification of 3406 DEGs, of which 1774 were under-expressed ([Figure 5](#marinedrugs-15-00196-f005){ref-type="fig"}A) and 1632 were over-expressed ([Figure 5](#marinedrugs-15-00196-f005){ref-type="fig"}B). To build a comprehensive list of all the putative validated target genes, we compared them with the predicted target genes using miRNA prediction programs such as TargetScan and Pictar software. In total, these DEmiRNAs regulated 4097 putative target genes against 18 up- or down-regulated miRNAs. We looked at whether the DEGs that were significantly and negatively correlated with DEmiRNAs were in fact miRNA targets; of which 2394 genes were on the predicted target genes for the up-regulated miRNAs ([Figure 5](#marinedrugs-15-00196-f005){ref-type="fig"}A), while 1703 genes were for down-regulated miRNAs ([Figure 5](#marinedrugs-15-00196-f005){ref-type="fig"}B). Ultimately, we considered a total of 209 (for 18 up-regulated miRNAs) ([Figure 5](#marinedrugs-15-00196-f005){ref-type="fig"}A) and 211 (for 18 down-regulated miRNAs) ([Figure 5](#marinedrugs-15-00196-f005){ref-type="fig"}B) enriched RNAs, showing that their expression patterns were inversely correlated with DEmiRNAs in response to porphyra-334. Among the DEmiRNAs, has-miR-92a-3p and -6807-5p have the most number of target genes, 131 and 53, respectively ([Figure 5](#marinedrugs-15-00196-f005){ref-type="fig"}). To identify the most relevant cellular activities controlled by each anti-correlated target DEGs, we analyzed overrepresented GO biological process terms. The most significant GO terms for target genes of up-regulated miRNAs were related to regulation of ligase activity, regulation of ubiquitin-protein ligase activity, cell cycle, translational elongation and translation ([Figure 5](#marinedrugs-15-00196-f005){ref-type="fig"}A; [Supplementary Materials](#app1-marinedrugs-15-00196){ref-type="app"}, [Table S4](#app1-marinedrugs-15-00196){ref-type="app"}). In contrast, in the case of target genes of down-regulated miRNAs, they were discovered to participate in biological processes associated with protein amino acid phosphorylation, transcription, DNA-dependent regulation of transcription, regulation of RNA metabolic process and regulation of transcription ([Figure 5](#marinedrugs-15-00196-f005){ref-type="fig"}B; [Supplementary Materials](#app1-marinedrugs-15-00196){ref-type="app"}, [Table S5](#app1-marinedrugs-15-00196){ref-type="app"}). Notably, between two different groups of target DEGs, we found that there is inverse correlation with respect to UV-related biological processes including apoptosis, cell death, cell cycle, and regulation of cell proliferation; target genes for the up-regulated miRNAs are involved in apoptosis, mitotic cell cycle and translational elongation, whereas those for the down-regulated miRNAs function as anti-apoptosis, anti-cell death and positive regulators of cell proliferation. In the previous study \[[@B53-marinedrugs-15-00196]\], following UV DNA damage, there is an overall inhibition of protein synthesis and translational reprogramming. This reprogramming allows the cells to repair DNA damage and counter with apoptosis through the selective synthesis of specific proteins including elongation factors and DDR proteins. Next, to examine the biological pathways enriched in the data set in an unbiased, systematic method, GSEA was performed to generate a large network of enriched gene sets, and then the large network was clustered to generate sub-networks of interrelated gene sets ([Figure 6](#marinedrugs-15-00196-f006){ref-type="fig"}). Taken together, these data suggest that porphyra-334 may protect the cells from harmful UV-radiation through the modulation of genes associated with cell survival processes such as apoptosis, cell death and translational elongation.
3. Experimental Section {#sec3-marinedrugs-15-00196}
=======================
3.1. Cell Cultures and Irradiation {#sec3dot1-marinedrugs-15-00196}
----------------------------------
HaCaT cells were maintained in Dulbecco's modified Eagle's medium (DMEM; Gibco/Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA) with 10% Fetal Bovine Serum (FBS) in 37°C humidified incubator containing 5% CO~2~. Cells were plated at a density of 0.5 × 10^6^/well in 6-well plates and grown overnight. They were incubated with porphyra-334 (1.0 mg/mL) for 30 min prior to UV irradiation, which was carried out as described previously \[[@B11-marinedrugs-15-00196]\]. Briefly, a Philips Original Home Solarium sun lamp (model HB 406/A; Philips, Grogningen, Holland) equipped with a UV lamp delivering a flux of 23 mW/cm^2^ between 300 and 400 nm was used as a UV radiation source. Cells were irradiated for 15 min (275 kJ/m^2^). This UV dose is equivalent to \~90 min of sunshine on the French Riviera (Nice, French) in the summer at noon \[[@B54-marinedrugs-15-00196]\].
3.2. RNA extraction {#sec3dot2-marinedrugs-15-00196}
-------------------
Total RNA was extracted using TRIzol Reagent (Invitrogen, Carlsbad, CA, USA) following the manufacture\'s instruction. Specifically, the pellet obtained from 5 × 10^6^ cells was lysed 1 mL of TRIzol solution. At the end of the extraction the isolated RNA was dissolved in 35 µL in RNase-free water and incubated for 10 min at 55 °C. An aliquot of 5 µg RNA was then used for cDNA synthesis using the SuperScript first strand cDNA synthesis kit (Invitrogen).
3.3. MicroRNA Microarray Analysis {#sec3dot3-marinedrugs-15-00196}
---------------------------------
For control and test RNAs, the synthesis of target miRNA probes and hybridization were performed using Agilent's miRNA Labeling Reagent and Hybridization kit (Agilent Technologies, Palo Alto, CA, USA) according to the manufacturer's instructions. Briefly, each 100 ng of total RNA were dephosphorylated with \~15 Units of calf intestine alkaline phosphatase (CIP), followed by RNA denaturation with \~40% DMSO and 10 min incubation at 100 °C. Dephosphorylated RNA were ligated with pCp-Cy3 mononucleotide and purified with MicroBioSpin 6 columns (Bio-Rad, Hercules, CA, USA). After purification, labeled samples were re-suspended with Gene Expression blocking Reagent and Hi-RPM Hybridization buffer, followed by boiling for 5 min at 100 °C and 5 min chilled on ice. Finally, denatured labeled probes were pipetted onto assembled Agilent Human miRNA Microarray (Human miRNA Microarray Release XX, AXBK) and hybridized for 20 hours at 55 °C with 20 RPM rotating in Agilent Hybridization oven (Agilent Technologies, Santa Clara, CA, USA). The hybridized microarrays were washed as the manufacturer's washing protocol (Agilent Technologies).
3.4. Data Acquisition and Analysis {#sec3dot4-marinedrugs-15-00196}
----------------------------------
The hybridized images were scanned using Agilent's DNA microarray scanner and quantified with Feature Extraction Software (Agilent Technologies). All data normalization and selection of fold-changed probes were performed using GeneSpringGX 7.3 (Agilent Technologies). We performed data transformation (set measurements less than 0.01 to 0.01) and per chip (normalize to 75th percentile) normalization. Probes which were changed not less than 2.0-fold of ratio between test and control samples were selected and considered as differentially expressed probes.
3.5. Library Oreparation and Sequencing {#sec3dot5-marinedrugs-15-00196}
---------------------------------------
For control and test RNAs, the construction of library was performed using SENSE mRNA-Seq Library Prep Kit (Lexogen Inc., Vienna, Austria) according to the manufacturer's instructions. Briefly, each 2 µg total RNA are prepared and incubated with magnetic beads decorated with oligo-dT and then other RNAs except mRNA was removed by washing solution. Library production is initiated by the random hybridization of starter/stopper heterodimers to the poly (**A**) RNA still bound to the magnetic beads. These starter/stopper heterodimers contain Illumina-compatible linker sequences. A single-tube reverse transcription and ligation reaction extends the starter to the next hybridized heterodimer, where the newly-synthesized cDNA insert is ligated to the stopper. Second strand synthesis is performed to release the library from the beads, and the library is then amplified. Barcodes were introduced when the library is amplified. High-throughput sequencing was performed as paired-end 100 sequencing using HiSeq 2000 (Illumina, Inc., San Diego, CA, USA).
3.6. Data Analysis {#sec3dot6-marinedrugs-15-00196}
------------------
mRNA-Seq reads were mapped using TopHat software tool (<https://ccb.jhu.edu/software/tophat>) in order to obtain bam file (alignment file). Read counts mapped on transcripts region were extracted from the alignment file using bedtools (v2.25.0, <http://bedtools.readthedocs.io/en/stable/>) and Bioconductor that uses R (version 3.2.2, <https://cran.r-project.org/bin/windows/base/old/3.2.2/>) statistical programming language (R development Core Team, 2011). The alignment file also was used for assembling transcripts, estimating their abundances and detecting differential expression of genes or isoforms using cufflinks. We used the FPKM (fragments per kilobase of exon per million fragments) as the method of determining the expression level of the gene regions. Global normalization method was used for comparison between samples. Functional gene classification was performed by DAVID (<http://david.abcc.ncifcrf.gov/>). Cytoscape (version 2.7, <http://www.cytoscape.org>), an open source bioinformatics platform, provided by the U.S. National Institute of General Medical Sciences (NIGMS) (<https://www.nigms.nih.gov>) was used to construct network diagrams.
3.7. qRT-PCR Analysis {#sec3dot7-marinedrugs-15-00196}
---------------------
First-strand cDNAs were synthesized from 1 μg of total RNA using a high capacity cDNA reverse transcription kit (ThermoFisher Inc), and the samples were analyzed with sybr green real-time PCR master mixes (ThermoFisher Inc) with gene expression primers. The oligonucleotide primers used were as follows: p53, 5'-CCACCATCCACTACAACTACAT-5' and 5'-AGGACAGGCACAAACACG-5'; Nkx2-5, 5'-CTCCCAACATGACCCTGAGT-3' and 5'-CTCATTGCACGCTGCATAAT-3'; CPEB1, 5'-GTGATCCCCTGGGTATTAGC-3' and 5'-CAGATATGACACAGAGAATCTT-3'; FGF12, 5'-AGCTTGGTTTCTGGGACTCA-3' and 5'-CTCTTCTGAGATGAGTTTCTGCTC-3'; GPR34, 5'-GGGACTGGTTGGGAACATAA-3' and 5'-GAAAGGGAGGCAGAAGATGA-3'; Pitx2, 5'-TCGTCCATGAACTGCATGAAAG-3' and 5'-ATGTCATCGTAGGGCTGCATG-3'; Pla2g7, 5'-TAATGATCGCCTTGACACCCT-3' and 5'-TACAGCAGCAACTATAAACCC-3'; Fbf1, 5'-TGTCAGCTCGGTATCTGTCG-3' and 5'-GTGGTCTCTGCGCATGTCTA-3'; Rasd1, 5'-GCGGCGAAGTCTACCAGTTG-3' and 5'-TGTCTAAGCTGAACACCAGAATGA-3'; GAPDH, 5'-GAAGGTGAAGGTCGGAGTC-3' and 5'-GAAGATGGTGATGGATTTC-3'. For miRNAs, qRT-PCR primers for has-miR-92a-3p (product no: 204258) and has-miR-6807-5p (product no: 2107829) were purchased from Exiqon (Vedbæk, Denmark). As a control, U6 snRNA PCR primer (product no: 308006) was used (Exiqon).
4. Conclusions {#sec4-marinedrugs-15-00196}
==============
In the present study, we showed that porphyra-334-regulated genes were involved in UV-affected biological processes such as Wnt and Notch pathways. In addition, porphyra-334-regulated miRNAs can also target a variety of genes associated with UV-affected biological processes such as apoptosis, cell proliferation and translational elongation. Notably, we observed good inverse correlation respect to biological functions between two different target groups for up- or down-regulated miRNAs. Taken together, these comprehensive molecular data indicate that porphyra-334 can be mostly involved in UV-affected biological pathways to protect organisms from harmful UV radiation, and that provide a new insight to understand its functional molecular networks.
This research was supported by the research projects (PE17180 and PE17100) of Korea Polar Research Institute, Incheon, Korea.
The following are available online at [www.mdpi.com/1660-3397/15/7/196/s1](www.mdpi.com/1660-3397/15/7/196/s1).
######
Click here for additional data file.
Sung-Suk Suh, Il-Chan Kim, and Sanghee Kim designed the research; Sung-Suk Suh, Sung Gu Lee, Ui Joung Youn, and Se Jong Han performed the research; Sung-Suk Suh, Il-Chan Kim, and Sanghee Kim analyzed the data; and Sung-Suk Suh and Sanghee Kim wrote the manuscript. All authors discussed the results and commented on the manuscript.
The authors have declared that no conflict of interest exists.
{#marinedrugs-15-00196-f001}
{#marinedrugs-15-00196-f002}
{#marinedrugs-15-00196-f003}
{#marinedrugs-15-00196-f004}
{#marinedrugs-15-00196-f005}
{#marinedrugs-15-00196-f006}
| {
"pile_set_name": "PubMed Central"
} |
Background
==========
Treatment of children with the acquired immunodeficiency syndrome (AIDS) using antiretroviral drugs (ARVs) has been a major challenge in the fight against the human immunodeficiency virus (HIV), especially in resource-constrained settings. In many African countries, the shortage of human resources for health is one of the major barriers to achieve universal access to HIV treatment and care. In particular, reliance on doctor and hospital-centered care hampers the ability to scale-up antiretroviral treatment (ART), and task shifting, the process of delegation of tasks to health workers with lower qualifications, has become a recognized strategy \[[@B1]-[@B4]\].
While there have been a number of studies showing good outcomes in treating pediatric populations with ARVs in resource-constrained settings \[[@B5]-[@B16]\], these programs have essentially been carried out in hospital settings and/or with significant physician involvement. Although nurse-based ART provision in health centers could increase the coverage of ART, it is not clear whether this compromises quality in terms of outcomes. There are no published studies demonstrating detailed methods and treatment outcomes of children in health centers staffed by nurses. In particular, there is limited published information on how to address the psychosocial issues related to HIV, a major challenge for pediatric ART programs \[[@B17]\], in settings with few specialized personnel available.
As in most African countries, Rwanda has an acute shortage of physicians and is relatively better resourced with nurses \[[@B18]\]. In a recent simulation model, it was estimated that, relying on a physician-centered service provision model, 51% of the total physician capacity of the government of Rwanda would be absorbed by HIV care and treatment by the end of 2008 \[[@B19]\]. Thus, it is completely logical to make greater use of nurses to manage HIV care, but the question remains whether doing so leads to good quality outcomes, especially for pediatric patients.
This report describes the nurse-centered pediatric ARV program implemented in two government health centers in Kigali, Rwanda, with details of its psychosocial aspects and treatment outcomes.
Methods
=======
Design
------
Retrospective analysis of routinely collected outcomes from the ARV program in two health centers in Kigali, combined with interviews with key health center and Médecins Sans Frontières (MSF) staff. MSF reports since program inception were also reviewed.
Setting
-------
Rwanda, with a population of around 9 million inhabitants, has an overall HIV prevalence of 3% and more than 7% in urban areas \[[@B20]\]. The national ART program, launched in 2003, was first established mostly in the district hospitals, with subsequent decentralization to the health centers. From 2004 on, a gradual scaling-up was seen. The latest estimate in 2006 by TRAC (Treatment and Research AIDS Center) of the number of HIV-infected children was 13,901 with half of them (6951) in need of ARVs \[[@B21]\]. By the end of May 2007, almost half (3255) had benefited from ART.
Health worker distribution in Rwanda and their roles within the HIV care program
--------------------------------------------------------------------------------
With one physician/50,000 habitants and one nurse/3900 habitants, Rwanda is clearly short of physicians but is relatively better resourced with nurses \[[@B22]\]. In addition, 80--90% of the population is living in rural areas and mainly relies on care from primary health centers, staffed by nurses. Up to now, HIV/ART care delivery has essentially been provided by physicians, with nurses only playing a supporting role (see Table [1](#T1){ref-type="table"}). In the traditional model, physicians were responsible for all tasks. Equally, at primary health centers, the bulk of medical care was provided by the visiting physician, who was based in the district hospital. When relating to HIV care for children, this reliance on physicians/hospital-based care was even more pronounced as it was believed to be more difficult care.
######
Traditional and modified tasks for nurses and physicians within the HIV/ART care program
**MD-CENTERED** **NURSE-CENTERED**
------------------------------------ ----------------- -------------------- --- -------
**Pre-ART care**
Initial physical exam/staging X X
Ordering CD4 count X X X
Assessment of ART eligibility X X
FU of non-eligible patients X X
CTX refill X \-^b^
Complex medical cases X X
**ART care**
Ordering lab tests X X X
Interpretation of lab tests X X
ART initiation and FU
Non-complex cases X X
Complex cases^a^ X X
ART refill X \-^b^
Register keeping/reporting X \-^b^
Filing of results/medical records X \-^b^
Training/mentoring X
Supervision X
^a^Complex cases included those with advanced HIV disease, severe/persistent opportunistic infections, severe ART side-effects, suspicion of ART failure, severe or recurrent non-adherence to ART.
^b^Activities taken over from the ARV nurse by other staff.
ART: antiretroviral therapy; FU: follow-up; CTX: cotrimoxazole; MD: medical doctor
In the Rwandan health system, nurses are classified into three levels according to their level of training: 1) A3 nurses may have no/minimal secondary training and minimal health training; 2) A2 nurses (the bulk of the health workforce) have two years of secondary education and two years of nursing training; 3) A1 nurses have received two additional years of nursing training after finishing secondary school \[[@B22]\]. Whereas initially A1 nurses were rarely found at health centers, this is gradually changing. Currently, no other non-physician clinician cadres are being trained.
Study sites
-----------
The two clinics in this report were Kimironko and Kinyinya health centers, located in Kigali. Kimironko was an urban government health center with a catchment area of about 75,000 people while Kinyinya was semi-rural, being located at the outskirts of Kigali, with an estimated population of 17,000.
In addition to routine health care, the health centers provided comprehensive HIV care and started offering ARVs at a decentralized level, beginning in October 2003 in Kimironko and followed in January 2004 in Kinyinya. They were among the first services in the country to offer ART. By July 2007, 3252 patients had been started on ART within these two clinics and of these, 332 children were enrolled in the ART program. These two clinics have been supported by MSF since 2002.
Study population
----------------
The analysis included all children enrolled in the HIV program who qualified for ARVs from the launch of the ART program from October 2003 till Jan 1/2007. With data collected until June 30/2007, all children had been on ART for at least six months. See Figure [1](#F1){ref-type="fig"} for details about how children were selected for ARVs and their treatment paths.
{#F1}
Description of the Pediatric HIV program
----------------------------------------
### National government\'s commitment and external support
The pediatric program in the two clinics was part of the Rwandan national ART program, essentially run by government health care staff. The national program has seen a successful scaling-up over the last years, organized through the National AIDS Control Commission (NACC) and TRAC. A constructive collaboration with various international partners has taken place, with substantial financial support provided by the Global Fund to fight AIDS, Tuberculosis and Malaria. The government\'s commitment was demonstrated by providing additional nursing and laboratory staff as well as ongoing laboratory services, training and ARV procurement. MSF\'s contributions were essentially aimed at increasing the clinics capacity for ARV care and included training and supervising staff, refurbishing clinic buildings, upgrading the labs and providing financial incentives.
### Nurse-based treatment and care
#### Increase and retention of nursing staff
At the launch of the HIV program in 2002, the health centers were poorly staffed. The increased staffing by the Ministry of Health was a critical factor. Whereas the health center\'s staff included 7 trained nurses before the launch of the HIV program in 2002, this had gradually increased to 28 by the end of 2004 (mainly A2 nurses), when scaling-up began. Overall, around 50% of the entire staff\'s time was dedicated to HIV care. To compensate for the substantial increase in work-load associated with the fast scaling-up of the ART program, a performance-based financing mechanism was put in place, initially backed by MSF.
To avoid overloading the nurses, most of the tasks traditionally performed by nurses were taken over by new or reinforced cadres in the health centers: receptionists for administrative work and data collection/monitoring; counsellors and community support groups for counselling; and lab-staff for blood collection.
#### Training for nurse-based treatment and care
When the ART program was launched, each health center was supported by a HIV physician. Given the high number of patients in (urgent) need of ART, one of their main duties was to build local capacity, facilitating a rapid scaling-up of HIV treatment and care within the health facilities. From the onset, nursing staff received theoretical and bed-side training in comprehensive HIV care in general, and in pediatric aspects in particular, from the HIV physician, gradually increasing their knowledge and confidence. Using this knowledge, they were able to initiate and change ARV treatment (for toxicity or contra-indications), with medical confirmation, and perform routine follow-up of the children, aided by simplified treatment protocols and growth curve charts provided in the consultation rooms. Protocols for recognizing side-effects of ARVs were standardized and indications for referral to the physician were defined. Ongoing training and supervision was ensured by having one full-time equivalent physician per health center, who was ultimately responsible for the medical care. As the program matured, physician involvement decreased to a presence two to three times a week. A great deal of emphasis was placed on comprehensive care with a family-centered approach including methods to address the psychosocial issues of HIV. ARVs were dispensed by a nurse with extensive practical experience in ARVs and additional training through the national program.
#### Mentoring and supervision
Besides consulting the complex/referred cases, the physician played a major role in assuring the quality of care provided in the program (Table [1](#T1){ref-type="table"}). The physician did not have a separate consultation office, but reviewed patients together with the nurses, allowing for ongoing training and skills-building. In addition, this permitted ongoing evaluation of the quality of care provided by the nurses, and to identify areas where additional training was needed. At the same time, a sample of patient files was reviewed to assess completeness of care and to trace medical problems. Only if the physician was confident in their knowledge and skills, were nurses allowed to consult in the ART program, but they still continued to receive ongoing training and supervision.
### Voluntary Counseling and HIV testing (VCT) for children: overcoming barriers among caregivers
Early in the program, we observed that barriers existed among the caregivers towards testing their children. Caregivers themselves were distressed by their own recent HIV diagnosis. They were reluctant to discuss testing of their children, since they felt guilty, fearing the reaction of the child to their own and the child\'s disclosure, and being worried about the health and future of the child \[[@B23]-[@B25]\]. Thus, from the start of the program, support and discussion groups were organized for the caregivers, designed to increase their acceptance of their HIV-positive status, a prerequisite for discussing testing of their children. In addition, prior to starting ARVs, individual in-depth counseling sessions were held with the caregivers to discuss testing of children in detail \[[@B23],[@B25]\]. Subsequently, the health center-based support groups were transformed into community-based support groups. Within these, the group leaders, serving as role-models of positive living with HIV, helped to make HIV more easily discussed within the community and to raise issues like testing of partners and children.
### Timely and careful disclosure and ongoing psychosocial support
In our experience, children were generally not well prepared when they came for testing. Few children knew exactly why they were there, had little knowledge of HIV and rarely knew about the status of the caregiver \[[@B25],[@B26]\]. However, the children preferred to be informed about their and their caretaker\'s status from the caretaker and felt cheated when they were not told the truth. Consequently, we tried to involve the caretaker as much as possible during disclosure. Caretakers were first counseled on why it was important to talk openly with their children about HIV, why their active participation was important for the child, how the child might react, and how they should respond to questions. During disclosure, we tried to have the caregiver explain their and the child\'s status, while being supported by the counselor \[[@B26],[@B27]\]. The use of a booklet explaining HIV provided a common language around HIV (Figure [2](#F2){ref-type="fig"}).
{#F2}
Testing of children was done on a dedicated day, ensuring that enough time was available for every child, making the environment/event more child-friendly and facilitating on-site training of health care staff in counseling of children. Disclosure was considered when the child was \> 6 years old \[[@B27],[@B28]\].
Within the child support groups, an environment was created by the counselors where the children could express themselves, raise their questions and worries and develop a positive attitude towards life with HIV. Most of the issues discussed were raised by the children themselves and reflected their deeper feelings: HIV (what, why and how?), life and death, sexuality, manipulation in the caregiver-child relationship and discrimination. The issues were addressed in several ways using open discussions, games, fairy-tales and drawings \[[@B27],[@B28]\]. These groups (consisting of 12--15 children) were organized according to age and were open to all children from the age of 7 years on, irrespective of taking ARVs. Our approach was based on general recommendations available within MSF \[[@B27]\] that were adapted to the local context and taken into account the capacity/possibilities within the program.
Initially, children were followed medically together with the adult patients, but had their support groups on separate days. Subsequently, a special consultation for children, integrating both medical and psychosocial aspects, was introduced once a week, reducing the amount of traveling and out-of-school time for the children. It also allowed them to receive care and follow-up in the presence of other children, making the whole experience more child-friendly \[[@B27]\].
### Who were the caregivers ?
Close to 65 % of the children were taken care of by one or both parent(s). Twenty-five percent received care from the extended family (grand-parents, brothers/sisters, cousins,\...), and about 10 % of children were orphans with adoptive caregivers. In general, no child was tested without the caretaker being tested.
ART eligibility, regimens and safety
------------------------------------
Children were started on ART according to the national guidelines, based on the World Health Organization (WHO) immunological and clinical eligibility criteria and using WHO-recommended first line regimens \[[@B29]\]. To simplify treatment regimens and limit calculation errors, four weight categories were used to choose the appropriate dosing of the ARVs, as has been described elsewhere \[[@B9],[@B30]\]. Generic, scored fixed-dose combinations (FDC) were prescribed as much as possible (Triviro 30/40, Ranbaxy), containing stavudine/lamivudine/nevirapine. No quarter-tablets were used (as use of these had not been validated), only whole or half tablets were prescribed. When scored FDCs were not available, children were given either separate medications or were switched to a zidovudine-containing FDC (Duovir-N, Cipla). Initially, small children needing syrup formulation were referred to the central referral hospital in Kigali (CHUK) but from mid-2005, the syrup was available at the health center level.
For most children (orphans in particular), a home-visit was done before ARV initiation to assess the socio-economic situation of the family and to provide support where necessary. Adherence to treatment was evaluated by self-report using calendars, regular clinical attendance (monthly pharmacy refills/visits after start-up period) and pill counts. As well, viral load counts were performed routinely after the first year of treatment from 2005 \[[@B31]\]. HIV care, including ARVs, consultations, lab tests and OI medications, were provided free of charge. Transport costs for the children were covered for families in need.
Side effects were evaluated at each follow-up visit, and treatment changes were made for grade 3 and 4 events \[[@B29]\]. Liver function tests were performed at baseline, 2, 4, 8 weeks and then every 6 months and on clinical grounds. A full blood count was done at baseline, 12 weeks and then every 6 months and on clinical grounds. For children on zidovudine, an additional control was done at 4 and 8 weeks.
At the outset, absolute CD4 counts and percentages, checked every 6 months, were measured with the FACSCalibur™ Flow Cytometry System (Becton-Dickenson) at the National Reference Laboratory (NRL). From 2003--2004, the FACSCount apparatus (Becton-Dickenson) was used and percentage counts were calculated based on total lymphocyte counts of a separate sample.
CD4 percentage counts at baseline were missing for several reasons (n = 33). First, due to limitations in the number of samples that could be handled at the NRL in the initial stages of the program, some children had no baseline CD4 counts because they met clinical criteria for starting ART. Second, for some children transferred from other sites, the information was unavailable. Finally, if the total lymphocyte count was not known, the CD4 percentage could not be calculated, and absolute CD4 counts were used for consideration of ART. The median time from baseline CD4 sample collection to initiation of ARVs was 53 days (interquartile range 30--79). Viral loads were analyzed at the NLR by the Ultrasensitive Amplicor assay, version 1.5 (Roche). All other biochemical tests were performed at Kinyinya health center.
Data collection and statistical analysis
----------------------------------------
Data were routinely collected during the consultations using Access^®^software. Treatment and safety outcomes of the cohort were analyzed by Stata software 9.0^®^(STATA Corp., College Station, Texas, USA). Kaplan-Meier survival methods were used to estimate the probability of remaining in care (failure defined as death or lost to follow-up). CD4 percentages were used for children under 5 years to evaluate immunological status and both CD4 counts and absolute counts used for children above 5 years. Weight-for age Z-scores (WAZ) were calculated with Epi-Info (version 4.3.1; Centers for Disease Control and Prevention). Data for this study were collected from October 1/2003 until June 30/2007.
Ethics
------
This study analyzed routinely collected data from the pediatric HIV program. Approval for publication of this data and report was received from the NACC and TRAC. Confidentiality was maintained with no patient identifiers being used. The Ethics Review Board of MSF and the Rwandan National Ethics Committee gave exemption from formal ethical review. No informed consent was obtained for this study.
Results
=======
Characteristics of the study population
---------------------------------------
Over a period of about 3 years, 315 children were started on ARVs, with a median age of 7.2 years. Table [2](#T2){ref-type="table"} describes their characteristics. Approximately 25% were less than 5 years of age, 7 were started before the age of 18 months. Sixty percent were in WHO stages I and II at treatment initiation. Fifteen tuberculosis (TB) cases were recorded of which 4 had extra-pulmonary manifestations. Baseline CD4 counts percentages were available for 282 children with a median value of 14%. The median WAZ-score was -- 1.9. The vast majority of children, 89%, started on stavudine/lamivudine/nevirapine, with adult FDC tablets used for 282 children. The remaining 33 commenced syrup formulation of whom, by the time of analysis, all but five had changed to tablets. The median time on ART was 2 years. Four fifths of all children were followed for more than one year with a total of 589 patient years of follow-up.
######
Characteristics of children started on antiretroviral treatment (n = 315)
------------------------------------------ -------------------------
Age at start^a^ 7.2 (4.5 -- 10.4)
\< 3 years^b^ 38 (12%)
3 -- 4.9 years^b^ 51 (16%)
5 -- 14.9 years^b^ 226 (72%)
Sex (male/female)^b^ 157/158 (50/50%)
Clinical WHO-stageq^b^
WHO stage I 43 (13.7%)
WHO stage II 145 (46.0%)
WHO stage III 115 (36.5%)
WHO stage IV 12 (3.8%)
Weight for age (Z-score) (n = 293)^a^ -1.9 (-3.0;-0.9)
Baseline CD4 count % (n = 282)^a^ 14 (9--18)
\< 15%^b^ 158 (56.0%)
15--25%^b^ 118 (41.8%)
\> 25%^b^ 6 (2.1%)
Baseline absolute CD4 counts (n= 302)^a^ 345 (229--572)
Baseline hemoglobin (mg/dl)(n = 268)^a^ 11.0 (10.3--11.8)
ART regimen^b^
d4T/3TC/NVP 281 (89.2%)
d4T/3TC/EFV 6 (1.9%)
AZT/3TC/NVP 19 (6.0%)
AZT/3TC/EFV 9 (2.9%)
Time on ART (years)^c^ 2.0 (1.2--2.6)
Total patient years of follow-up 598
On tablets (FDC) 577
On syrup 21
\< 1 year vs. ≥ 1 year^b^ 59 (19 %) vs 256 (81 %)
------------------------------------------ -------------------------
^a^Values are expressed as median (interquartile range)
^b^Values are expressed as n (%)
^c^As of June 30/2007 (closing date of dataset)
WHO: world health organization; ART: antiretroviral treatment; d4T: stavudine; 3TC: lamivudine; NVP: nevirapine; EFV: efavirenz; AZT: zidovudine; FDC: fixed-dose combination
Clinical, immunological and virological outcomes
------------------------------------------------
By June 2007, 84% (265) were still alive and being followed-up in the clinics. Thirty (9.5%) had been transferred to another health facility offering ART. Eight (2.6%) had died, only 12 (3.8%) were lost to follow-up (defined as not coming to their last scheduled visit for more than 2 months). Of the 8 deaths, 2 were clearly not linked to HIV. Kaplan-Meier estimates (Figure [3](#F3){ref-type="fig"}) showed a probability of remaining in care of 94.7% and 92.9% at 12 and 24 months respectively (with failure defined as death or lost-to follow-up). CD4 results, hemoglobin and WAZ all showed progressive improvement over the treatment interval (Table [3](#T3){ref-type="table"}). Most viral load results were obtained between and 15 and 23 months and were available for 87% by 18 months on treatment (n = 174). Viral load was less than \< 400 copies/ml in 82.8% of children and showed satisfactory viral suppression in 86.8%. For 13 children among whom viral loads were detectable, a second sample was made available after adherence counseling. The median time from first viral load collection to the repeat sample was 5.8 months (interquartile range 2.7--8.8). Four out of the 13 positive viral loads became undetectable and 1 decreased significantly. Two children were switched to second line treatment. For the remainder, adherence problems were still being addressed or criteria to switch were not met.
{#F3}
######
Clinical, immunological and virological outcomes for children on antiretroviral treatment (n = 315)
**Baseline** **6 months** **12 months** **24 months** **36 months**
---------------------------- ------------------------- ---------------------- ----------------------- ------------------------ ---------------
CD4 count
[\< 5 years]{.ul} (%)^a^ 16 30 32 33 35
IQR (12--19) (23--35) (26--34) (28--36) (28--40)
N 84 59 48 39 7
[≥ 5 years]{.ul} (%)^a^ 13 25 26 29 29
IQR (9--17) (18--29) (21--32) (24--34) (19--33)
N 198 175 117 80 24
[≥ 5 years]{.ul} (abs)^a^ 297 550 624 704 616
IQR (181--405) (337--747) (459--840) (562--866) (492--865)
N 217 183 120 80 24
Hb (mg/dl)^b^ 11.0 12.1 12.6 12.9 13.7
WAZ^a^ -1.9 -1.6 -1.6 -1.5 -1.5
IQR (-3.0;-0.9) (-2.6;-0.8) (-2.6;-0.7) (-2.5;-0.6) (-2.7;-0.6)
Viral load [Months on ART]{.ul}^a^ [\< 40 c/ml]{.ul}^c^ [\< 400 c/ml]{.ul}^c^ [\< 5000 c/ml]{.ul}^c^
(n = 174) 18 (15--23) 127 (73.0%) 144 (82.8%) 151 (86.8%)
^a^Values are expressed as median (interquartile Range (IQR));
^b^Values are expressed as median;
^c^Values are expressed as n (%)
abs: absolute CD4 count (in cells/μL); Hb: hemoglobine; WAZ: Weight-for-age Z-score; ART: antiretroviral treatment
We only had complete data on pharmacy refill as an indirect measure of adherence to therapy. Allowing a delay of up to two days (accounting for the security stock), we defined excellent, good and poor adherence as being punctual for \> 95%, 80--95% or \< 80% of the visits respectively. Poor adherence was observed for 5% of the children, with the majority having excellent (49%) or good (46%) adherence.
Safety and tolerance
--------------------
Among all children placed on ART, a change in therapeutic regimen was required for 46 children (14.6%), out of whom only 26 (8.3%) were due to toxicity. The start of TB treatment provoked 18 switches. Toxicity was mainly related to nevirapine, requiring a treatment change to efavirenz (n = 24). Sixteen cases were due to grade 3--4 skin manifestations of which two were Stevens-Johnson syndromes occurring within 1 month of therapy. Both recovered. Five early changes (within 3 months) were made for severe hepatitis, with 4 children showing clinical signs, and 1 child having asymptomatic grade 3 liver toxicity. An additional three children were changed late (median 9 months) due to symptomatic hepatitis. We observed an additional 14 children with grade 2 (n = 11) or grade 3 (n = 3) liver toxicities, mostly occurring within the first months of ART. In general, these were transient and well tolerated and did not require treatment change.
Stavudine was changed for one child with severe neuropathy, and two cases of lipoatrophy were reported (one requiring treatment change). No lactic acidosis or anemia requiring treatment change were observed. Thirteen children were found with a hemoglobin level \< 7 mg/dl of which 5 were present at baseline and generally improved on stavudine-containing ART. Of the others, none were related to zidovudine and were transient and/or related to intercurrent infections.
Discussion
==========
This report offers reassuring evidence that health center/nurse-based ART delivery for children is both feasible and very effective. The morbidity and mortality results are comparable to or better than those of other reported studies, which are essentially hospital-based \[[@B5]-[@B16]\]. One of the strengths of the study is that the therapeutic responses were confirmed with virological evidence in the majority of children. This is very important since it suggests that clinic-based programs can provide high quality care without over-reliance on the most sparse human resources in the health system and is thus more likely adapted to and sustainable within this context. In addition, decentralized/nurse-based care might have potential advantages over hospital/physician-centered care in providing accessible and comprehensive HIV care to children.
Reinforcing nurse-based care
----------------------------
The shortage of physicians in many African countries requires a more strategic use of their skills, without compromising quality of care. The health center-based care described here was provided to a large extent by nurses, with appropriate physician supervision. A most important factor in the success of the program was the increase in number of nurses, a human resource pool relatively more available in Rwanda and many African countries, in the health facilities by the Ministry of Health. A substantial part of the physician\'s role was to ensure ongoing on-site training, mentoring and supervision, to gradually increase the capacity and skills of the nurses. Although considerable energy and time were spent on this, especially early in the program, in our experience it clearly paid off in the end. Capitalizing on the potential of the nurses allowed for a rapid scaling-up that would have been difficult, if not impossible, to achieve when relying on physicians only for ART care. At the same time, we would like to caution against over-reliance on nursing skills for ART care in the absence of adequate supervision \[[@B32]\]. We believe that the complementary interplay between nurses\' and physicians\' skills has been important to achieve good results.
Another important element of success has been the high retention of nurses. We think that the main contributing factors have been the continuous training and skills building, the high quality of care provided and patients\' satisfaction, the increase in nursing staff to meet the demand, the promotion of responsibilities for the staff in the project and the performance-based financial incentives \[[@B4],[@B33],[@B34]\]. In addition, the creation of new support staff that took over some nursing duties has avoided overloading the nurses. This has allowed a steady increase in the pool of nurses skilled in HIV care. Over time, consulting nurses with 3--5 years experience in ART care have become more involved in the training of other nurses, and currently function as a source of reference. Their experience has also allowed a gradual decrease of physician involvement to a presence two to three times a week.
Decentralized care: increased accessibility and acceptability
-------------------------------------------------------------
Centralized HIV care at the hospital level tends to create additional bottlenecks in treatment initiation, and creates difficulties with transport and time off work for the patients \[[@B29],[@B35],[@B36]\]. It has been observed in other settings that, relative to nurse/health center based ART delivery, the overall capacity of hospital/physician based ART care in number of patients that can be treated is far more limited, which obviously raises concerns about access to and quality of care provided \[[@B34]\]. The fact that this program was based at a decentralized level might have contributed to earlier diagnosis and good adherence in the program in a number of ways. First, although the distance to the closest hospital was not enormous given the urban setting (1--2 hours walk), geographical proximity is still likely to have facilitated increased access and acceptability for both caregivers and children, and this is consistent with findings from other studies \[[@B37]\]. This will obviously be more important in rural settings. In addition, we were able to provide comprehensive and patient-centered care, within a family-based approach, closely linked with the community. Having all staff trained in HIV/ARV care meant that they were sensitized to the issues in all service areas. Care and counseling could be provided at every patient encounter, and were not limited to the physician\'s consultation or a specific service as is often the case in the hospital setting. Convenient scheduling of visits also facilitated accessibility as did comprehensive care that was provided free of charge \[[@B37]\]. The very low rate of lost to follow-up observed in our program seems to confirm that services were \'user friendly\'.
Nurse-based care -- multiplying care capability
-----------------------------------------------
By 2005, physicians constituted 4.2% of the total health work force in Rwanda, 52.2% were nurses \[[@B22]\]. Recent estimates from Rwanda suggest that approximately one full-time equivalent physician would be needed for 500 patients on ART within the traditional ART care model \[[@B19]\]. A simulation model for ART care delivery in Rwanda estimated that allowing nurses to initiate and follow the ART treatment for non-complex cases would see a 76% reduction in physician demand for HIV care or an 170% increase in physician capacity for non-HIV care \[[@B19]\]. Within the nurse-centered model, as described here, we equally observed a gradual decrease in the need of physician time to approximately one full-time equivalent physician for 1500 patients after the first capacity had been built within the health centers and one equivalent for 3000 patients at later stages. Although investment in human resources at all levels remains a priority, nurse-centered ART care constitutes a more rational use of the human resources currently available.
Psychosocial Care
-----------------
We started ART in a relatively healthy cohort of children, as seen from the clinical and immunological baseline data. This, in turn, was due mostly to our identifying children earlier in their disease. From the outset, a family-centered approach was implemented with a strong focus on the psychosocial aspects of HIV care (Table [4](#T4){ref-type="table"}). This played a crucial role in overcoming barriers to routine testing of children. Dealing with the HIV-infected adults\' psychosocial issues allowed the staff to introduce the idea of testing children, and helped them prepare the children for disclosure \[[@B25],[@B38]\]. Although it remains hypothetical, we think that this sensitive approach, combined with the goal of routinely testing all children of adult HIV-infected patients, facilitated the early diagnoses reflected here. Our experience also demonstrated that there was a significant burden of HIV illness in children that was readily identifiable through HIV-infected adult caregivers. Over 90% of the children in this cohort came from this source.
######
Psychosocial aspects of the pediatric HIV program
---------------------------------------------------------------------------------
**Caregiver-centered approaches**
- organization of discussion/support groups, at health center and in community
- family-based approach to identify eligible children
- individual counseling (pre-ART)
- psychosocial issues addressed in follow-up care
**Child-centered approaches**
- adapted counseling for children for disclosure and ART (child-adapted tool)
- designated days for children\'s clinics
- child support groups
- integrated care, including disclosure, with their caregivers
**Health-staff centered approaches**
- discussion groups for health care staff
- training on psychosocial implications of HIV
- practical training by psychosocial team (check-lists,\...)
- supervision and mentoring
---------------------------------------------------------------------------------
ART: antiretroviral treatment
HIV: human immunodeficiency virus
In addition, once children tested positive, the psychosocial emphasis provided effective means for handling disclosure and subsequent commencement of ARV treatment. Preparation of both caregivers and children contributed to good adherence and treatment outcomes \[[@B24],[@B39]\]. This supports previous research that emphasizes that HIV care is more than getting CD4 counts and prescribing ARVs. Addressing the many psychosocial aspects of the illness is crucial to effective treatment programs \[[@B23],[@B38],[@B40]\]. Whereas this has been effectively implemented in high-income countries and in some hospital-based settings in low-income countries, we think this aspect of pediatric ART has not received sufficient attention in decentralized contexts \[[@B17]\]. Our program experience emphasizes the need and feasibility of providing psychosocial support within health center-based pediatric ART programs.
Safety
------
In our program, ART was safe and very well tolerated, with few severe side-effects reported. In concordance with others \[[@B5],[@B12]\], tolerance of ART seemed to be as good or better in children than in adults, arguing against the need for specialist care for pediatric ART and in favor of providing ART treatment to children at a decentralized level. In addition, treating a healthier population was also likely to reduce side-effects.
Virtually all treatment changes for side-effects or contra-indications were initiated by the nurses, and confirmed by the physicians, in patients with clinically recognizable symptoms. This is reassuring for decentralized pediatric treatment in rural areas where laboratory monitoring might not be available. Few children had to change therapy due to rashes or hepatitis. No adverse effects resulted in death. No cases of severe anemia or lactic acidosis reported, while neuropathy and lipoatrophy were extremely rare. One could argue that the low rate of side-effects could be explained by their having been missed by the nurses. However, within this program, side-effects were documented and treated by the same nurses who managed our adult cohort \[[@B41],[@B42]\], at a frequency similar to other studies \[[@B35],[@B43]-[@B45]\]. Although diagnosing ART toxicity in children might be more challenging, children were additionally monitored through chart review and virtually all were seen by a physician at some point in time. Side-effects missed by the nurses were extremely rare.
Replication/sustainability
--------------------------
We are confident that this kind of program could be replicated in other settings and is sustainable. The program was run in health centers that were already providing a full range of primary care services. The success of the program depended, to a large extent, on the commitment of the government. The increased staffing was a critical factor. External support provided by MSF was essentially directed at capacity building (training, infrastructure, logistics) and has gradually been taken over through the maturing national ART program. The financial incentives provided by MSF were subsequently replaced by a performance-based pay initiative of the Ministry of Health, gradually implemented since 2005 \[[@B46]\]. In addition, all nurses in the public sector have seen a substantial increase in salary. The supervising and mentoring role of the health center-based physician has been gradually taken over by the visiting district physician. This transition has been aided by additional training and quality assurance provided by the national ART program and performance-based financing. The external support decreased gradually over the years, and MSF handed over the project to the Ministry at the end of 2007. Although a follow-up visit by mid 2008 could not reveal any major problems subsequent to the hand-over, this should ideally be reassessed in a few years time.
Another program in Rwanda recently reported excellent patient outcomes for their adult population and good compliance with national guidelines using a health center/nurse-based model \[[@B47]\]. The national program is currently considering expanding this model of care at the national level.
It is important to note that this program, especially the psychosocial elements, was developed over several years and in response to perceived needs of the children and caregivers. This helps explain the ongoing changes we describe, reflecting the subsequent barriers we encountered regarding testing and treatment of children. It is not necessary that the full structure described here be operational at the outset of a program, but could be developed over time, depending on facilities and staffing availability and on the evolving challenges.
Remaining challenges
--------------------
While the program successfully managed to build nursing skills and to redefine the role of the physician, high quality of care will require further training for nurses and redefinition of the tasks and responsibilities of the nurses and physicians, based on the evolving challenges. Treatment failure, an inevitable clinical challenge for the future, in particular, will require a careful mix of physicians\' and nurses\' skills. In addition, increasing problems of acceptance and adherence can be expected in the long run, especially during adolescence. To what extent these will be successfully managed within the program remains to be seen.
Given the severe lack of human resources for health care at all the levels in most low-income countries and the pronounced increase in need to achieve universal access to HIV care, task shifting can only be successful in the long-term when combined with investment in human resources at all levels \[[@B32]\].
Limitations
-----------
Given our relatively healthy cohort, we are cautious in suggesting that similar outcomes would be possible with a sicker population. Since CD4 counts were not available for all children, and some children were started on clinical grounds, the missing baseline CD4 counts might have biased the baseline values to some extent. Consequently, it is possible that the comparison between baseline and follow-up CD4 counts is not entirely accurate, but we believe the differences would be minor. Similarly, we only recorded 87% of viral loads by 18 months. Given the similar baseline characteristics, we have no reason to believe that the results of patients with missing values would be substantially different. We also note that this cohort included few young children and we cannot be sure if similar outcomes would apply to them. At the early stages of the program, the youngest children depended on the university hospital to receive their ARVs in syrup formulation before it become available at the health center level. Since this only involved 9 children, we do not think this could significantly affect the program outcomes. The health centers received targeted support from MSF that has been gradually withdrawn. Although the program seems to be sustainable, with quality of care remaining high, its future success will depend on the governments\' ongoing commitment and resources.
To what extent local factors influenced the outcomes of the program remains to be determined. This program was part of and has benefited from the successful scaling-up of the Rwandan National HIV program. The HIV prevalence rate in Rwanda is lower than most Sub-Saharan African countries. The context of HIV has certain particularities linked to the history of genocide.
Finally, although the program appeared to have been successful, at least in treatment outcome terms, caregivers\' and children\'s perceptions and satisfaction were not known and should be assessed. This information could potentially be an important determinant of long-term success of pediatric ART programs, where the psychosocial consequences of HIV and the overall impact on childrens\' lives might be even more pronounced than for adults \[[@B17]\].
Future research
---------------
This study points to further research to see whether this health center model could be successful with sicker and/or younger populations. Whether this approach of decentralized nurse-based pediatric ART programs can be replicated in other contexts remains to be established. Qualitative research into understanding how the psychosocial aspects of care were received would be useful to refine this part of the program.
Conclusion
==========
Our program shows that, given a strong psychosocial emphasis, health center/nurse-based ART delivery for children is both feasible and very effective, and can at the same time avoid barriers to care related to the scarcity of physicians. Providing comprehensive care close to the community can play an important role in achieving universal access to HIV care for all children.
Competing interests
===================
The authors declare that they have no competing interests.
Authors\' contributions
=======================
All authors read and approved the final manuscript. JVG and LDN were physicians treating the children, CG was coordinator of the MSF HIV project. JU worked as psychologist within the pediatric program. JVG, LDN, JU and CG conceived the project. JVG performed the data analysis and the program evaluation/description; JVG and TR co-drafted the manuscript. LDN, AA, CG and JU critically reviewed the manuscript and improved the intellectual content; TR performed the editing.
Pre-publication history
=======================
The pre-publication history for this paper can be accessed here:
<http://www.biomedcentral.com/1471-2431/8/39/prepub>
Acknowledgements
================
We are grateful to the staff of Kinyinya and Kimironko health centers. We are particularly grateful to all children and caregivers in our program. This program was part of and has benefited from the successful scaling-up of the Rwandan National HIV program, organized through the National AIDS Control Commission (NACC) and TRAC (Treatment and Research AIDS Center) and in collaboration with various international partners. The NACC is mainly focussing on general policy, coordination and resource mobilization. TRAC is mainly working on medical and technical issues (protocols, training, monitoring, surveillance). We acknowledge the financial support for the HIV/ARV program provided by MSF-Brussels operational centre, the Global Fund to fight AIDS, Tuberculosis and Malaria, the Belgian Development Cooperation (DGCD) and the European Union. The World Food Program provided nutritional support. We wish to thank Line Arnould, Myrto Schaeffer, David Olson, Rony Zachariah and Ana Corthouts for their useful comments on the manuscript.
| {
"pile_set_name": "PubMed Central"
} |
SEVERAL STUDIES IDENTIFY that the root cause of crowding in the emergency department is "boarding of patients." ([@R1]) This prolific problem involves "holding" admitted or pending transfer patients in the emergency department, resulting in loss of precious bed capacity. A 2009 white paper ([@R2]) by the American College of Emergency Physicians noted that 62.5% of emergency departments (and 85% of those with annual visit volumes more than 50,000) reported boarding ED patients for more than 2 hours. A 2012 American Hospital Association survey ([@R3]) revealed that 38% of hospital emergency departments were operating "at" or "over" capacity.
Numerous downstream effects result from the boarding burden, including loss of capacity to treat the queue of patients who predictably arrive in the emergency department, longer cycle times for emergency department patients, decreased bed utilization, suboptimal quality and safety for boarded patients, and poor perception of quality and confidentiality by boarded patients housed in hallways ([@R4]).
Although flow models such as split-flow can significantly reduce cycle times and throughput intervals ([@R5]), they do not displace or lessen the gravity of the boarding problem. To improve service and quality in the emergency department, hospitals must confront the institutional, systemic problem of boarding to free the emergency department of the undue burden of caring for inpatients. An essential means to achieving this is through the development of hospital-wide flow teams that engage inpatient unit leaders, physicians, and executives to promote the use of validated real-time systems, principals, and processes.
TIME MATTERS
============
Efficient disposition is critical for many reasons. ED consumers desire timely care. When ED beds are occupied by patients holding for admission, front-end flow is impacted. Patients then leave without treatment, perhaps the ultimate quality indicator for emergency care. In fact, many patient complaints across the industry are often rooted in excessive wait times to see a provider.
In addition, time-sensitive treatments are dependent on efficient ED throughput. As a result, the Centers for Medicare & Medicaid Services (CMS) now collects and reports throughput metrics, such as NQF 497 (Admit Decision to ED Departure Time), and has already begun to link throughput performance to reimbursement. (Data collection now will impact 2014 reimbursement.) The current benchmark for best practice is 97 minutes (based upon [hospitalcompare.gov](http://hospitalcompare.gov) benchmarking data between second quarter 2012 to first quarter 2013 ED visits).
The CMS also requires reporting time-sensitive clinical treatments dependent on ED throughput, such as acute myocardial infarction (door to balloon time less than 90 min), acute stroke (time to tissue plasminogen activator less than t3--4 hr), and sepsis: early goal-directed therapy. By connecting improvement efforts to clinical quality to meet best practice standards on these metrics, ED leaders can effectively engage inpatient clinical staff to improve throughput. In addition, physicians and hospitalists are increasingly contractually obligated to meet flow metrics, incentivizing their active partnership in improving throughput in the emergency department.
Facilitating Admissions From the Emergency Department
-----------------------------------------------------
Best practices for facilitating timely admissions from the emergency department include active identification and communication of potential admissions, the use of no-delay nurse reports, and faxed admission reports. However, before requesting assistance from inpatient leaders on back-end flow, it is critical that the emergency department has first addressed front-end and middle flow---those issues that it can address on its own---to ensure credibility with inpatient leaders.
### Identify and Communication Potential Admissions
Once the emergency department has made significant progress with throughput challenges it can impact on its own, it must next focus on proactive communication with inpatient leaders. The ED staff and providers must ensure timely communication to bed management (e.g., patient access, nursing supervisor, director on call) when a decision is made to admit a patient.
It is the emergency department\'s responsibility to ensure that information on the status of admitted ED patients is communicated through the appropriate bed management channels and current in the electronic health record or other bed management systems as applicable. To ensure accurate communication and anticipation of admissions, the ED attending physician and charge nurse should round at least every 4 hr at the ED tracking board to review patients\' status, identify need for admission, and assess ED crowding as well as the need to implement a surge protocol, if applicable.
### Utilizing No-Delay Nurse Reports
No-delay nurse reports reduce time from admit orders to arrival on inpatient units, decrease the potential for handoff errors (because they ensure that both the ED and inpatient nurses have the same information), and increase patients\' perception of care due to timely transfer of care.
A sample process might resemble the following: After the order is written and a request for a bed is made, the ED nurse opens a "transfer of care" report, which is also accessible to the floor nurse. The ED nurse then awaits the patient bed assignment (by monitoring the tracking board). Fifteen minutes after the bed has been assigned, either the unit clerk calls down to the emergency department to accept the patient or the accepting inpatient primary nurse or charge nurse will call to ask questions. In either case, the ED patient is transferred within 15 min of a bed becoming available. No-delay nurse reports are best included in the electronic health record, but a faxed paper report may also be used.
Best Practices for an Effective Hospital-Wide Throughput Committee
------------------------------------------------------------------
Beginning January 1, 2014, The Joint Commission (TJC) revised standards LD.04.03.11 and PC.01.01.01 ([@R7]; which govern patient flow through the emergency department) to include a requirement for goal setting and measurement to mitigate and better manage the boarding of patients. This includes a requirement for individuals to manage patient flow processes to review measurement results to goals and take action to improve patient flow when goals are not achieved.
An effective hospital-wide throughput committee addresses this need by meeting monthly and typically convening 12--15 individuals who can best drive process improvement and remove barriers for admitting patients quickly from the emergency department. It is an interdisciplinary team that typically includes hospitalists, leaders from the emergency department (e.g., director, manager, flow coordinator), and inpatient leaders from critical care, telemetry, and med-surg. "Although we may consider the operational components from the emergency department to inpatient to be a care transition, we must evolve our processes and mindset to one of a team approach to support a patient-centered continuum of care," notes Dan Smith, MD, FACEP, one of StuderGroup\'s medical directors and an expert on ED flow efficiency.
This group may include the environmental services leader, transport leader, admissions leader, administrative supervisor (nursing supervisor), risk manager, and discharge planner/case manager. Attendance by a senior leader sponsor is also important for buy-in and accountability.
The charter of the hospital throughput committee should clearly define its scope and focus. The scope and focus of this committee should promote the creation and monitoring of measurable outcomes that address the above mandates from CMS and TJC. The committee members must not lose sight of the importance to address back-end ED flow as a hospital-wide problem that only this type of interdisciplinary team can correct. The meeting agenda flows from these goals with a review of "wins," priority throughput indicators, clinical quality issues, financial impact from reducing length of stay, and review of the hospital\'s current "throughput dashboard." The purpose of this hospital-wide throughout dashboard is to collect and track current data on key metrics so that the team can identify trends and performance gaps, and when centered around it, teams can ensure that meetings are productive and actionable.
Dashboard data typically include measures such as disposition to admission by unit, disposition to admission by time of day, hospital discharge by time of day, staffing by hour (for ED nursing, environmental services, and hospitalists), and other inpatient admission process times. Process times may include total turnaround time for admitted patients, time of physician order to discharge to patient departure, patient departure to notification of housekeeping, notification of housekeeping to bed clean, bed clean to assignment of new patient to bed, and time of bed assignment to new patient in bed.
The team can also track bed assignment time to report received to ensure that no-delay nurse reports are occurring consistently. To ensure compliance with and tracking of NQF measures as required by CMS, the dashboard should include these metrics as well. Other optional metrics for this dashboard may include environmental service turnaround times, post anesthesia care unit (PACU) hold hours, and ED hold hours/diversion hours (see Figure [1](#F1){ref-type="fig"}).
.](aenj-37-65-g001){#F1}
As each leader reports out on next action steps with dates due, these are captured in the meeting minutes for follow-up. Goals to consider include times for inpatient to discharge, room assigned to occupied, discharge order goals, environmental service turnaround times, PACU hold hours, and ED hold hours.
It is recommended that this committee also assess and evaluate the effectiveness of the surge plan at least every 3 years to ensure that it is producing desired results on the basis of the charter and goals identified by this team. A strong and effective surge plan identifies specific actions for ED, inpatient, administrative, and ancillary leaders through a tiered response. Each of these tiered responses should have specific action items within them for all departments that impact transition of patients out of the ED. Each of these responses should be precisely designed steps to prevent escalation to the next level.
By collaborating closely with inpatient units, the emergency department at UConn Health, an integrated academic medical center in Farmington, Connecticut, has made dramatic gains in ED throughput in just 2 yr. Over 24 months, the team has reduced ED arrival to departure time from 461 to 336 min and ED admit decision to departure time for admitted patients from 225 min to 143 min.
"Having physicians write transition orders has also really reduced bottlenecks. In fact, we\'ve reduced the time from admit decision to transition orders from 102 minutes to 22 minutes in just the last six months by coordinating this process closely through the ED physician, hospitalist and an inpatient resident," explains AnnMarie Capo, associate vice president for clinical effectiveness and patient safety. "Instead of having one medical officer of the day visit patients in the emergency department, transition orders move patients to the inpatient unit quickly where inpatient nurses can more efficiently divide the work and expedite quality patient care." Moving a case manager into the emergency department full-time has also ensured that all stakeholders are together to assign bed placements.
Tactics to Support an Effective Hospital-Wide Throughput Committee
------------------------------------------------------------------
### Inpatient Bed Huddles
These must occur at least daily and they must be actionable to achieve goals the hospital-wide flow committee has set. Each unit should be represented in the bed huddle and leave with individual action items to discharge patients, prevent delays, and address surges. Expedited admissions to inpatient units and the timely transition of patients to admission should be everyone\'s priority as it produces the best possible outcome for patients.
During huddles, best practices such as early inpatient discharge (before noon), unit discharge rounds, a "pull until full" philosophy with "zero tolerance" for hiding beds, and elimination of unit-specific bed control should be monitored and reinforced. High-performing organizations utilize an afternoon bed huddle to follow up on action items identified earlier in the day and the results they have generated or the need to escalate to others. During times of surge, these same organizations move their inpatient bed huddles to the ED to generate urgency in patient transition and to facilitate inpatient leader rounding.
### Inpatient Leader Rounding
Inpatient leader rounding, in which the inpatient leader rounds on boarded ED patients, is another key tactic to support the actions of the hospital-wide flow committee. Before implementing inpatient leader rounding, ED leaders must generate engagement by sharing the "why,"---the benefits for inpatient leaders to rounding with their patients, before emphasizing the "what" and "how."
Inpatient leader rounding ensures both safety (e.g., a safe handoff and safe acceptance of inpatients) and quality (e.g., improved morbidity and mortality). By connecting inpatient leaders with patients holding in the emergency department, we also establish ownership for the transition, building trust. (Staff rounding on patients reduces anxiety around unexplained and uncertain waits for better clinical outcomes; [@R6].) Frequency of inpatient leader rounding on ED patients is tied to the hospital surge plan with clear guidance and instruction for inpatient units and defined roles and expectations for staff and physicians at tiered surge levels. Inpatient leader rounding is driven by the chief nursing executive or other senior leader sponsor.
The inpatient nurse manager or charge nurse rounds on ED patients awaiting admission in the emergency department. By role modeling rounding, patients are introduced to the rounding process they will experience on the inpatient unit. In starting this practice, it is recommended to begin on one or two pilot units to conduct a test of change using the institution\'s change management philosophy (e.g., plan-do-study-act).
When inpatient leaders round, the goal is to use good listening skills (e.g., refrain from interrupting or rushing the patient), repeat back concerns to ensure that the patient was heard, express a genuine apology where appropriate, thank the patient for entrusting the hospital with his care, express empathy, and explain specifically what the leader will do to address a concern without blaming others or offering excuses.
The inpatient leader should leave a business card and write his or her name and contact information on the ED communication/care board. In addition, the unit phone number is provided to family and visitors. To ensure both effective collaboration and a positive patient experience, the inpatient leader must never "manage down" the admissions process. Rather, it is imperative to reinforce the excellent care patients are currently receiving in the emergency department and share positive steps occurring in process for the transition to inpatient care.
There are many direct benefits to inpatient leaders who round on patients in the emergency department. They ensure that the patient is appropriate for the unit and nursing assignment as well as proactively addressing special needs before transitioning to the inpatient unit. In this way, the inpatient unit can exceed patient\'s expectations with respect to pain, cultural issues, and spiritual care. In short, inpatient leader rounding on ED patients improves patients\' perception of care, which is reflected on patients\' ratings on the Hospital Consumer Assessment of Healthcare Providers and Systems survey.
CONCLUSION
==========
Over the years, emergency departments have tried many tools and tactics to improve back-end flow. These range from expanding the emergency department by building more rooms; using special discharge or observation units when at full capacity; utilizing ED nurses as inpatient nurses with boarded patients; ambulance diversions; and in the emergency department. However, boarding is not an ED problem. It requires effective, consistent hospital-wide collaboration to facilitate timely admission to inpatient units as well as ongoing monitoring, measurement, and real-time process improvement by a hospital-wide throughput committee.
When ED leaders focus their efforts on partnership and collaboration rather than isolation and finger-pointing, many barriers to resolve the issues before them can quickly erode. As difficult as it can sometimes be, challenging ourselves to elevate above the potential for conflict and reach across boundaries to the inpatient world will deliver the best results for our patients. To accomplish this and to gain and maintain credibility with our inpatient colleagues, we must ensure that we have first addressed internal ED flow issues to the best of our abilities. This will generate evidence of our own ability to change, improve care quality, and generate results that impact all patients. With this type of "you go first" change leadership, anything is possible!
Disclosure: Stephanie Baker and Angie Esbenshade are employees of StuderGroup, a national consulting firm that coaches health care organizations on how to accelerate service and operational excellence, including tools to improve patient flow in the emergency department as described in this manuscript.
| {
"pile_set_name": "PubMed Central"
} |
{.258}
{.259}
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
Thoracic Insufficiency Syndrome (TIS) is defined as impairment in breathing and/or postnatal lung growth due to spine and thoracic cage deformity in children who are skeletally immature ([@B1]). It is defined by abnormal respiratory function but the specific indices used to identify impairment have not been standardized. Skeletal maturity is defined radiographically by the closure of growth plates in certain bones and indicates that further growth and hence spine and thoracic deformities that progress with growth are less likely to worsen. There is an international registry of more than 8,000 children at risk for TIS maintained by the Pediatric Spine Study Group ([@B2]). As of 2019, \<5% of the patients in this registry had interpretable pulmonary function data. Pediatric pulmonologists are not often intimately involved with the clinical decision making in the management of these patients. Yet their input can be of great value to surgeons, primary care providers, and families regarding the pulmonary status and functional impact of spine and chest wall deformities over time and with treatment.
This chapter addresses the input that pulmonologists can provide at all ages in children with EOS as an example of TIS, as part of an multi-disciplinary team, which ideally includes spine surgeons, pulmonologists, general surgeons, neurosurgeons, pediatric sleep specialists, nutritionists, genetic counselors, physical therapists, and bioethicists. The management of children with TIS begins on first encounter and ends for many with transition to adult care, i.e., over 15--20 years with multiple surgical procedures throughout that time. Pediatric pulmonologists are ideally involved with management questions that arise throughout this time period.
TIS and Etiologies of EOS {#s2}
=========================
TIS is often recognized in the newborn period by respiratory distress associated with hypoplastic thoraces due to inherited skeletal dysplasias. Small chest walls are classified by surgeons as thoracic "volume depletion" disorders due to the small size of the chest wall. Classic examples are Jeune\'s syndrome (asphyxiating thoracic dystrophy) and Jarcho-Levin syndrome (which includes both spondylocostal dysostosis and spondylothoracic dysplasia). Khombourlis has reviewed the more than 100 different skeletal dysplasias that can produce respiratory impairment in infancy ([@B3]). Many of these are lethal without invasive respiratory support in the first year of life. Several surgical techniques to enlarge the chest wall, such as sternal struts, rib-to-rib attachments, and rib-based expandable titanium arcs have been used in small case series to improve lung volumes. A recent review highlights the surgical options for children with hypoplastic thoraces ([@B4]). However case series reporting long-term outcomes are rare and not generalizable to all conditions.
A common condition producing TIS is early onset scoliosis (EOS) which presents before 10 years of age with a coronal curve spine curve \>10 degrees. The coronal degree of spine curvature is measured using the Cobb angle, illustrated in [Figure 1](#F1){ref-type="fig"}. The Cobb angle (and its progression over time) is the primary structural measure used by spine surgeons to make decisions about surgical and non-surgical interventions such as casting and bracing.
{#F1}
The etiologies for EOS include congenital, syndromic, thoracogenic, neuromuscular, and idiopathic scoliosis. *Congenita*l scoliosis is defined by the presence of vertebral and rib structural deformities such as vertebral hemi-vertebrae and failure of segmentation with fused or block vertebrae. These may be associated with multiple fused ribs which either constrain chest wall motion or provide large gaps between ribs that may produce a flail chest syndrome. Many *syndromes*, e.g., VACTERL syndrome, have multiple organ involvement and decisions must be made what to treat first. Up to 12% of children with congenital scoliosis also have congenital heart disease, and these children represent an overlap between congenital and syndromic categories ([@B4]). Children with *thoracogenic* scoliosis are those receiving thoracic surgery at an early age, either for diaphragmatic hernias, rib resections with tumors, pneumonectomies, or even cardiac repair ([@B5]). Up to 30% of children with congenital diaphragmatic hernias will develop subsequent scoliosis ([@B6]). This is compounded by pulmonary hypoplasia, which is worse on one side. Up to 10% of children undergoing thoracotomy for congenital heart disease will develop scoliosis.
The most common form of TIS due to EOS is scoliosis associated with *neuromuscular* diseases that produce spasticity or weakness. More than 90% of children with spinal muscular atrophy types I and II will develop scoliosis ([@B7], [@B8]). The frequency of scoliosis among children with cerebral palsy varies among reports from 5 to 80% ([@B7]). In one large series of 666 children, 17% had mild scoliosis and another 17% had moderate to severe scoliosis ([@B9]). Those with more severe cerebral palsy, based on GMFCS levels of 3--5, had a 50% prevalence with age of onset at 8 years. EOS accounts for \<10% of all scoliosis in childhood, with the vast majority presenting after age 10 years with adolescent idiopathic scoliosis.
*Infantile idiopathic* scoliosis begins before age 3 and can vary in severity on presentation. The posterior rib hump or abnormal posture is often the first finding identified by parents. Respiratory concerns are rare initially but may cause failure to thrive due to increased respiratory work with feeding. In young children (\<2 years) with infantile idiopathic scoliosis and small coronal spine curves (\<30 degrees) there is high rate of reversal to an almost normal spine shape with serial casting treatment over a period of years ([@B10], [@B11]). In these cases, there may be no pulmonary sequelae after orthopedic treatment. However, with larger spine curves, kyphosis, and etiologies for scoliosis other than idiopathic scoliosis, casting/bracing may not be sufficient to correct the deformity, or prevent curve progression and respiratory impairment. Casting is then used as a tactic to delay surgical intervention until the child is older. Thereafter, surgical use of "growth-friendly" expandable distraction rods are used to control scoliosis until pre-adolescence when spine fusion is often undertaken.
The etiology of scoliosis is important as it is one factor that dictates the risk and rate of progression of a spine/thorax deformity over time. Children with fused vertebrae and fused ribs are most likely to progress due to structural deformities ([@B12]). Duchenne\'s muscular dystrophy will produce scoliosis in 90% of boys but does so in young adulthood instead of early adolescence if they are treated early with steroid therapy ([@B13]). SMA I and II have new "natural histories" with the advent of nusinersin and adenoviral gene therapy. Scoliosis may present later in these children than previously reported as SMA-specific treatment becomes more common.
The heterogeneity of etiologies producing EOS has made assessments of management strategies, such as surgical and non-surgical interventions difficult to assess and predict. A classification system to address this heterogeneity has been devised for spine surgeons which includes age of onset, etiology of scoliosis, degree of coronal curve magnitude, degree of kyphosis, and rate of progression of the coronal curve deformity over time ([@B14]). Use of this classification system has enabled surgeons to estimate risk of post-operative complications ([@B15]). However, there are no functional elements in this system, such as pain, nutritional status, or lung function measures. This reflects the frequent lack of effective input by pediatric pulmonologists for these children.
Pathophysiology and Clinical Consequences of EOS and Other TIS Etiologies {#s3}
=========================================================================
The number of children with EOS that have TIS is uncertain due to the vague nature of the definition of TIS. However, there are common pathophysiologic processes regardless of the etiology. The majority have clinical evidence of restrictive respiratory disease. This is manifested earliest as tachypnea associated with activity or exercise and often is identified by families as exercise intolerance or easy fatigability with exertion. Dyspnea, used in one study to determine the functional consequences of EOS among children old enough to perform spirometry, did not occur frequently until FVC as a % predicted using arm span for height fell below 50% ([@B16]). Children with restrictive chest wall disease often adapt by reducing the intensity of their daily activities. They instead become sedentary to avoid dyspnea with activities. Restrictive changes have been measured both as reduced FVC with a normal FEV1 or alternatively by measuring Total Lung Capacity (TLC). Restrictive respiratory mechanics are produced by reduced chest wall compliance due to deformity and perhaps an additional reduction in lung compliance based on reduced lung volume in various lung regions. The reduction in chest wall compliance seems intuitive but data demonstrating this is rare. Motoyama measured total respiratory compliance in children with EOS during anesthesia and found it to be low prior to surgery and lower after titanium struts were attached to the chest in an effort to straighten the spine ([@B17]).
Decreases in total respiratory compliance leads to an increase in respiratory work and caloric expenditure and hence a reduced weight for age or body mass index (using arm span for height). Children will often avoid large meals and graze, eating frequent small volume snacks. Failure to thrive is reported in 50% of children with EOS and responds in most cases to caloric supplementation either by mouth or with a feeding tube or gastrostomy tube ([@B18]).
Obstructive lung disease occurs in 10--30% of children with EOS ([@B19], [@B20]). In children with EOS who demonstrated reduced FEV1/FVC or reduced FEV1 as a percent of normal, one third had airway reactivity due to asthma ([@B21]). Consequently, if obstructive lung disease is present in TIS, a bronchodilator challenge is in order. In at least 2/3 of cases, the airway obstruction is not reversible with a bronchodilator and usually reflects compression of a lobar or mainstem bronchus by intruding spinal elements and mediastinal structures, as illustrated in [Figure 2A](#F2){ref-type="fig"}. where the right lower lobe bronchus is compressed. Distal fixed tracheal obstruction also occurs by this mechanism. These central airway narrowings are documented by direct visualization during bronchoscopy or alternatively by CT scan images. If a subcarinal airway is compressed, the degree of obstruction described by spirometry will underestimate the severity of the local airway obstruction.
{#F2}
Distortion of the thorax and spine deformity can alter the distribution of ventilation and perfusion quantitated by nuclear lung scans. Unless there is a history of past lung or airway insults, such as pneumonia, the distribution of ventilation and perfusion closely track one another in children with EOS. However, more than half of children with EOS have asymmetric function with one lung functioning better than the other ([@B22]). This can be extreme, i.e., 90 vs. 10% in the two hemi-thoraces ([@B22]). The function on the concave side of the spine curvature is reduced 60% of the time and relative function of the right and left lung cannot be discerned from a chest radiograph. [Figure 2B](#F2){ref-type="fig"} depicts the ventilation lung scan and loss of ventilation due to right lower lobe compression by the rotated vertebrae portrayed in [Figure 2A](#F2){ref-type="fig"}. Surgical management in this case focused on recovery of lung function in the lower lobe with movement of the vertebra posteriorly with a growth friendly expandable titanium rod. The implications of asymmetric lung function have not been reported. However, surgeons are aware that interventions on the side with the greatest function, particularly if it requires a thoracotomy, predispose the child to respiratory failure post-operatively. In addition, there is speculation that the poorly ventilated lung may be predisposed to atelectasis if lung inflation during sighing is compromised. This could predispose children with severe scoliosis to prolonged atelectasis and need of augmented mucus clearance therapy with infections.
There is also evidence of reduced respiratory muscle force generation involving both inspiratory and expiratory muscles in children with EOS. As the ribs become crowded due to spine deformities and the costo-vertebral joints become ankylosed with prolonged immobility, the function of some intercostal muscles is reduced and patients become more dependent on diaphragm function to accomplish inspiratory work. There is less sharing of elastic loads among all the respiratory muscles during time of illness and therefore an increased risk of respiratory muscle fatigue and hypercarbic ventilatory failure. A similar event occurs during REM sleep when intercostal tone and function are diminished. Like children with primary neuromuscular weakness, children with EOS experience periods of hypercarbia during sleep before demonstrating daytime hypercarbia. The mechanisms for poor respiratory muscle performance are unclear. Respiratory muscle weakness involves both inspiratory and expiratory muscles as manifested by reduced Maximum Inspiratory pressures (MIP) and Maximum Expiratory pressures (MEP) in children with EOS and correlation with loss of vital capacity ([Figure 3](#F3){ref-type="fig"}) ([@B23]). Reduced MIP is thought to occur due to malposition of the diaphragm and perhaps tethering of the diaphragm as the spine and ribs distort and rotate. It is unknown if the diaphragm muscle fibers *per se* are atrophied or adapt to chronic mechanical loads. In animal models of EOS produced early in life, the diaphragms in adulthood have reduced cross-sectional surface area ([@B24]).
{#F3}
The reduced MEP is explained in part by the reduced total lung capacity, the lung volume at which MEP is generated. Reduced TLC is the hallmark of restrictive respiratory disease and the further it is reduced, the more MEP is compromised. Of interest, MEPs tend to be reduced to a greater degree than MIPs although both correlate with diminished vital capacity, as illustrated in ([Figure 3](#F3){ref-type="fig"}). Reductions in both MIP and MEP occur in the absence of underlying neuromuscular weakness conditions but are aggravated when underlying neuromuscular weakness disorders are present. The reduced MEP may lead to reduced force generation during cough and impaired mucus clearance when airway secretions increase, as with infection. Therefore, mucus clearance therapies become part of care when children with severe EOS become ill.
The conditions in which respiratory reserve is needed include sleep, exercise, and illness. Sleep related breathing disorders have been reported in over 90% of 63 children studied in one spine center ([@B25]). This high prevalence likely reflects underuse of polysomnograms to assess breathing during sleep in children with EOS and a lack of standardized criteria to refer children with EOS to assess breathing during sleep. The most common features on polysomnograms in this group of children were prolongation and recurrence of hypopneic events during REM sleep and not obstructive apnea *per se* ([@B26]). These hypopneic events are often associated with hypoxemia and hence counted in the Apnea Hypopnea Index used to score "obstructive sleep apnea." Loss of upper airway dilator muscle tone during REM sleep may also contribute to hypopneas. The tendency to become hypoxemic with these events may be related to low lung volumes due to small chest wall size and/or regional chest wall deformity with lung distortion. Regional lung constraint at low lung volumes may lead to increased airway closure and thus predispose to hypoxemia during sleep. Up to 25% of children with EOS in one study were found to have higher than normal hemoglobin concentrations ([@B27]). It is likely that sleep related recurrent hypoxemia and increased erythropoietin account for these findings.
Sleep related breathing disorders in children with EOS are aggravated by the adenoid and tonsillar enlargement that occur in normal children. An upper airway examination should not be overlooked in these children. Although few patients have been reported, BIPAP ameliorates sleep-disordered breathing more often than oxygen treatment alone in children with EOS ([@B25]).
Exercise tolerance is also impaired by TIS. In one report of 35 children with EOS, the distance walked over 6 min was 10% of predicted values ([@B28]). During formal cardiopulmonary exercise testing in more functional patients with EOS, the maximum level of work and maximum oxygen consumption were reduced in proportion to FVC. Sense of dyspnea during exercise also correlated with FVC although children often complained of leg fatigue more than shortness of breath ([@B29]). Both tests are impacted by more than respiratory function and include the effects of nutritional status, balance, deconditioning, and cardiac function.
Pulmonary hypertension has been described in adolescents dying of severe scoliosis in the last century. Cor pulmonale is rare in children in whom breathing during sleep has been assessed and treated. Presence of pulmonary hypertension is reported as \<10% of children with EOS but should be screened in those with severe spine and thoracic deformities.
These physiologic changes do not correlate with structural features of EOS. Cobb angle correlates poorly with measures of vital capacity, apnea-hypopnea index during sleep, and MIP ([@B23], [@B25], [@B26]). Glotzbecker also evaluated spine height and thoracic width as predicted by pelvic width among 121 children with EOS and found that these structural elements only explained 25% of the variation in forced vital capacity ([@B30]). There is a long-standing interest in developing an algorithm from thoracic and spine structure that could be used as a surrogate for lung function. To date, this has not happened and consequently surgeons are encouraged to measure lung function directly. This does not occur without some pulmonary guidance as to the meaning of the results in a clinical context and the quality of the measurements. Many pulmonologists are content to simply interpret the spirometric results without relating them to the clinical status of the patient and this is a disservice to the spine surgeon community.
Half of EOS spine registry patients were initially encountered at \<5 years of age ([@B2]). All of the aforementioned pathophysiologic features can occur in children who are too young to perform lung function tests. Respiratory rates, lung perfusion scans, polysomnograms, blood gas tensions, and echocardiograms do not require cooperation and are equally useful in children \<5 years old. However, these are not surrogates for measures of lung mechanics which are followed with spirometry. Additional clinical features, based on the physical exam may also prove useful. These include failure to thrive when other etiologies are excluded, asymmetric breath sounds on physical exam, and serial respiratory rates over time. Presence of hypotonia or spasticity are useful indicators of neuromuscular scoliosis. Infant lung functions have been performed and reported in a small series of infants using raised volume forced expiration methods under sedation and under anesthesia ([@B2], [@B31]). Reduced lung volumes have been reported in children \<2 years of age with EOS ([@B31]). However, these methods are not available in many pulmonary and spine centers. Also, passive lung mechanics do not account for reduced respiratory muscle function and may under-represent the degree of lung function impairment when awake. The relative frequency and sequence of changes in various measures of lung function among children \<5 years old with TIS and EOS have not been correlated in a systematic way. It is likely that the cumulative number of pulmonary abnormalities that are identified by physical examination an laboratory tests indices that the spine and chest wall deformities are causing more respiratory impairment but this has not been studied. Children with EOS \<5 years old remain a challenging group and no serial measures of objective findings have been described to estimate progressive impact of the thoracic deformity on respiratory function.
Specific Pulmonary Questions Relevant to Children With TIS and EOS {#s4}
==================================================================
How Severely Is Breathing Impaired on First Encounter?
------------------------------------------------------
Lung function testing is of great use during the initial encounter of the child with a spine and thoracic cage deformity. FVCs on initial encounter among 100 children with EOS from 5 to 15 years of age ranged from 18 to 89% of predicted using arm span or ulnar length for predicted height to normalized spirometric values ([@B32]). This range of values was not related to age. The initial lung functions prior to any surgical treatment are extremely significant as they predict eventual values in adulthood and decline in lung function later in adulthood ([@B33]). To date, there are no spine surgical strategies that predictably improve lung function for children with EOS ([@B34], [@B35]). This in part reflects the lack of pulmonary data before and after spine growth modulation treatments and before and after non-invasive spine distraction methods using expandable titanium rods. There is data that among children with TIS due to EOS who require invasive or non-invasive mechanical ventilation for all or some part of the day/night. Twenty-four percent of 77 children in one series were able to reduce their respiratory support after spine surgery ([@B36]).
What Other Pulmonary Conditions Are Impacting Pulmonary Function Results in Children With EOS?
----------------------------------------------------------------------------------------------
Surgeons assume that lung functions performed by children with TIS represent the impact of the spine and chest wall deformity they are asked to treat. However, to do this, the pulmonologist must first identify confounding conditions such as recurrent aspiration syndrome, asthma and lung injury from previous pulmonary infections and minimize their impact on lung function. Aspiration syndromes due to dysphagia associated with fatigue and work of breathing during eating should be considered in infants with TIS and failure to thrive. Among children with underlying neuromuscular weakness, MIP and MEP measures can be reduced by both the neuromuscular weakness and by the subsequent development of a chest wall deformity. In many cases of spinal muscular atrophy I and II, progressive restrictive chest wall disease occurs even after the deformity is controlled by spine fusion due to further collapse of the chest wall. This has been described as a "parasol chest," based on the shape of an umbrella or parasol as it is closed ([@B37]).
In boys with Duchenne muscular dystrophy followed longitudinally, lung function may decline faster among those developing scoliosis, which imposes restrictive forces on a respiratory system already impaired by weak respiratory muscles ([@B38]). Whether spine fusion in this group of children slows the decline in respiratory function due to progressive muscle weakness is unclear. There is no evidence that spine fusion will actually improve lung function in these children with TIS but it may prevent a more rapid progression of the restrictive process ([@B39]). The challenge for the pulmonologist is to maximally reverse any process that could affect lung function apart from the structural deformity of the spine and chest wall before any surgical spine procedure is performed. This may not be possible. A good example is the child with unilateral lung hypoplasia associated with congenital diaphragmatic hernia. Up to 30% of children after will develop scoliosis years after repair of the diaphragmatic hernia ([@B6]). One lung will function worse than the other due to the relative degree of unilateral hypoplasia. Scoliosis produces a spine curve that is usually concave to the hypoplastic side. Whether the spine shape worsens the function in the small lung or stretches the convex contralateral lung is difficult to ascertain. Serial lung scans may be of help in conjunction with spirometry to describe why lung function worsens over time and how each lung responds to surgical treatment of the scoliosis. The goal of management of children with this problem is to preserve function of the most functional lung.
There are two conditions that are particularly challenging. The first is those children with underlying skeletal dysplasias who have extremity and pelvic deformities. This precludes the use of arm span or ulnar length to normalize lung function when children perform spirometry. We have used absolute values for serial measurements and are exploring whether pelvic inlet width might be a useful normalizing method for children of different size and age. Pelvic width correlates well with spinal height in normal children but may not apply to all children with primary skeletal dysplasias ([@B5]).
The other "condition" where pulmonary impairment is difficult to assess is in children between 1 and 3 years of age with EOS treated with serial casting as a way to reverse or hold in check infantile idiopathic scoliosis. Children in casts to treat EOS typically have the entire thorax covered in plaster. The cast itself can limit mobility, modify sleep patterns, and limit caloric intake if the abdominal portion of the cast is not cut out and removed. Placement of the cast is performed under anesthesia and may restrict breathing considerably until the cut-out is performed. Physical examination of the heart, chest, and lungs of children in thoracic casts is limited. Body weight is difficult to interpret especially when a new cast is re-applied every 3--4 months. The pulmonary assessment reflects both the skeletal deformity and the impact of the cast. Ideally the child with EOS receiving casting treatment is assessed in between casting procedures.
What Respiratory Support and Management Will Improve Lung Function in Children With TIS?
----------------------------------------------------------------------------------------
Children with TIS can have failure to thrive, poor sleep quality, and poor mucus clearance when ill. Half of the children with TIS will have a Body Mass Index (BMI) that is \<10% of predicted norms using arm span to calculate BMI rather than height ([@B18]). This requires attention to nutritional status and the safest way to augment caloric intake long-term. Malnutrition can compromise respiratory muscle function, even in normal individuals. It may also increase wound dehiscence and post-operative infections following spine surgery. In children with TIS and failure to thrive, inadequate caloric intake is common and caloric supplements and/or long-term nighttime drip feedings via a gastrostomy tube are needed to produce a normal weight velocity. Halo traction and spine surgery for EOS can also improve weight gain pre-operatively ([@B40]).
Supportive care is also important to assure sufficient high-quality sleep. Arousals from sleep are increased in 45% of children with TIS during overnight polysomnography ([@B25]). BIPAP improves both AHI values and oxygenation in children with TIS during sleep. Further study is needed to describe the impact of oxygen, CPAP or BIPAP on sleep fragmentation and arousal frequency in children with TIS. In a small case series of children with TIS, 54% qualified for BIPAP based on an AHI of 5 or more events/hour ([@B25]). The impact of scoliosis surgery *per se* on sleep quality in children with EOS has not been reported. As the sleep quality in children with TIS is understudied, it is likely that a portion of these children are also undertreated at night both before and after surgical spine treatment.
Does Lung Function Predict Post-operative Pulmonary Complications or the Duration of Intensive Care After Surgery?
------------------------------------------------------------------------------------------------------------------
There is published data that pre-operative pulmonary function and etiology for scoliosis can predict the risk of post-operative complications after spine fusion surgery. Among 298 patients who underwent spine fusion for idiopathic and/or congenital scoliosis, a pre-operative value for FVC \<40% tripled the risk for post-operative pulmonary complications ([@B41]). Yuan et al. found that among children with EOS, the presence of underlying neuromuscular weakness increased the frequency of post-operative complications requiring \>3 days of respiratory support from 15% found in other categories of EOS to 50% ([@B42]). Other risk factors include duration of anesthesia (\>4 h) and volume of blood transfusions during surgery ([@B43]). Post-operatively the need for analgesics to control post-operative pain and thoracoplasty are also risks for longer length of stay after spine fusion in adolescents ([@B44]). Certain risks are specific to subpopulations with EOS. For example, pulmonary hypertension is a risk for prolonged intubation following spine fusion for scoliosis among children with congenital heart disease ([@B45]).
What Is the Impact of Spine Surgery on Lung Function?
-----------------------------------------------------
The pulmonologist must monitor the impact of different surgical strategies using different surgical devices to know when surgery improves or worsens lung function. Such monitoring before and after spine fusion surgery in adolescent idiopathic scoliosis demonstrated that the posterior surgical approach had smaller adverse effects of breathing post-operatively than did an anterior approach or combined anterior-posterior approach to spine fusion ([@B46]). This has led to primarily posterior surgical fusion for children with EOS who have more compromised lung function pre-operatively. There is very little data on the long-term changes in lung function during spinal surgical and non-surgical treatments of EOS. In one study, serial measurements of vital capacity in the operating room under anesthesia on children with TIS demonstrated that vital capacity measurements fell 28% despite insertion and subsequent expansions of growing rods over a 6 years period ([@B47]). The role of weakened respiratory muscles was not assessed due to the passive pulmonary function testing methods.
In some children, lung function will progressively decline despite spine surgery with growing rods, particularly if progressive spine rotation is a prominent feature of the deformity. Demonstration of progressive loss of lung function may require an alternative surgical strategy, e.g., spine fusion, and this will be dictated by serial lung function measurements interpreted by the pulmonologist.
What Is the Status of Patients With EOS After All Spinal Surgery Has Been Performed?
------------------------------------------------------------------------------------
Another role of the pulmonologist is to assure that the restrictive pulmonary disease that persists after all surgical treatment has been completed is sufficiently supported medically and that a transition plan is in place as the adolescent with TIS enters adulthood. With the advent of new prosthetic devices for the growing spine in use since 2002, a new population of patients with chronic severe restrictive respiratory disease has emerged in need of adult care. The lung function among EOS "graduates" of spine surgery is worse than among young adults who undergo surgery for adolescent idiopathic scoliosis ([@B48]).
Children with severe TIS and/or EOS previously did not survive to adulthood. Adult issues such as occupational risks and risks during and after pregnancy have not yet been addressed in this population. Referrals to adult pulmonologists who are aware of these risks are important. Based on one series of patients with EOS that were re-evaluated 25 years after spine fusion during adolescents, progressive decline in lung function during adulthood is more likely in those with a pre-operative FVC \<70% of normal ([@B33]).
Final Thoughts {#s5}
==============
The field of spine surgery for severe spine and thoracic cage deformities which present early in life is rapidly evolving. There are now spine procedures that modulate or direct spine growth, e.g., spine tethering, staples, and the Shilla procedure ([@B49]). These differ from growth friendly expandable rods that reduce spine curvature and foster spine growth through distraction. No pulmonary function testing has been completed with these new spine strategies to know their impact on breathing. The magnetically controlled growing rod which is extended non-invasively is now the most popular method for surgically straightening the spine without resorting to spine fusion in an actively growing child ([@B50]). Yet, no lung functions before and after insertion and expansion of these rods has been undertaken. All of these devices make intrathoracic volume larger but do so at the cost of chest wall compliance, which declines with stiff metal rods are attached to the chest ([@B17]). The reduced chest wall compliance that normally occurs during childhood, the reduced chest wall compliance of the spine deformity, and the additionally reduced chest wall compliance produced by insertion of growing rods into or next to the spine combine to produce progressive restrictive lung disease over time.
In addition, the timing of surgery depends not only on progressive structural deformity of the spine but also its impact on respiratory function over time. Progressive restrictive lung disease may be an indication for intervention even if the structural progression is not dramatic. Surgical decisions among pre-adolescent children with EOS and progressive spine deformities are also controversial. Spine fusion in this age group curtails further spine growth and hence intrathoracic volume during adolescence but also reduces the frequency of complications related to further spine surgeries if growing devices are used. There is no published pulmonary data to answer this question for each of the etiologies of EOS in early childhood.
There is no surgical procedure yet to salvage lung function impairment in children with TIS or EOS. Doing so remains a goal of care and may require attention to diaphragm function and position and also anterior chest wall reconstruction. Insertion of a Nuss bar to increase chest wall depth in a teenager with severe inoperable congenital scoliosis effectively relieved pulmonary artery compression and pulmonary hypertension but did not improve spirometry ([@B51]). New strategies are needed if lifelong restrictive lung disease that may worsen in adulthood is to be avoided. New materials for spine devices may provide more flexibility to maintain chest wall motion, but measurements before and after inserting such devices will be needed to demonstrate improvement. In the interim, there is a new population of children who are emerging with substantial respiratory impairment that require respiratory care along with spine care from the initial presentation, often before age 5 years and extending into young adulthood. This population will require pulmonary expertise to optimize management, assess current and improved spine care techniques, and assure transition to adult care in the future.
Author Contributions {#s6}
====================
GR was responsible for the entry of this manuscript.
Conflict of Interest {#s7}
====================
The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We thanks Ms. Holly Kaopuiki for assistance in preparing this manuscript.
[^1]: Edited by: Mark Lloyd Everard, University of Western Australia, Australia
[^2]: Reviewed by: James Y. Paton, University of Glasgow, United Kingdom; Shannon Sullivan, Stanford University, United States; Colin Wallis, Great Ormond Street Hospital for Children NHS Foundation Trust, United Kingdom; Steve Cunningham, University of Edinburgh, United Kingdom
[^3]: This article was submitted to Pediatric Pulmonology, a section of the journal Frontiers in Pediatrics
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#S1}
============
All living organisms rely on cellular and physiological mechanisms of homeostasis in order to maintain an internal environment optimal for life and function. Mitochondria are the foundation of cellular homeostasis, *via* their multiple roles in energy production, biosynthesis, calcium regulation and signaling, redox balance, and generation of reactive oxygen species. Not surprisingly, cells have evolved multiple mechanisms of quality control to ensure that mitochondria function at their best. These include protein import ([@B1]), folding and degradation ([@B2]), antioxidant defense mechanisms ([@B3]), mitochondrial turnover *via* autophagy ([@B4]), mitochondrial biogenesis ([@B5]), mitochondrial shape changes and cristae remodeling ([@B6]), and communication with the nucleus to coordinate transcriptional responses ([@B7]).
Emerging evidence indicate that mitochondrial dysfunction is associated with disparate diseases, including aging ([@B8]), neurodegenerative diseases ([@B9]), mitochondrial diseases ([@B10]), obesity ([@B11]), diabetes, and cancer. Although some controversies remain regarding whether functional or dysfunctional mitochondria are responsible for metabolic disorders, there is a resurgence of interest in understanding the mechanisms responsible for such mitochondrial alterations in disease. This review focuses on the molecular regulators of mitochondrial dynamics (organelle's shape and localization) in cancer and metabolic pathologies.
Regulation of Mitochondrial Dynamics {#S2}
====================================
Mitochondria constantly undergo shape and number changes thanks to the two opposing processes of fission and fusion ([@B12]). In turn, changes in gross mitochondrial morphology and the interconnectivity of the mitochondrial network impact on energy production ([@B13]), calcium signaling, mitochondrial DNA distribution, apoptosis, mitophagy, and segregation of mitochondria between daughter cells ([@B6]). The fine-tuning of the fusion--fission balance is crucial for cellular fitness in response to extracellular stimuli and environmental stress ([@B14]). Thus, alterations of the fission--fusion balance lead to oxidative stress, mitochondrial dysfunction, and metabolic alterations.
At the molecular level, dynamin-like GTPases orchestrate mitochondria shape changes. The fission protein dynamin-related protein 1 (DRP1) assembles into ring-like structures to constrict mitochondrial membranes in a GTP-dependent manner ([@B6]). DRP1 is recruited to mitochondria by fission protein 1 (FIS1), mitochondrial fission factor (MFF), and the mitochondrial dynamic proteins of 49 (MiD49) and 51 kDa (MiD51). On the other hand, the fusogenic proteins mitofusin 1 and 2 (MFN1/2) are located in the outer mitochondrial membrane, and tether two mitochondria through homo- and hetero-typic dimerization ([@B13]). A single GTPase, optic atrophy protein 1 (OPA1), achieves fusion of the IMM.
An expanding number of degenerative disorders are associated with mutations in the genes encoding MFN2 and OPA1, including Charcot--Marie--Tooth disease type 2A and autosomal dominant optic atrophy ([@B15]). Defective mitochondrial dynamics seem to play a more general role in the molecular and cellular pathogenesis of common neurodegenerative diseases (Alzheimer's and Parkinson's) ([@B14]), as well as in cardiovascular disease ([@B16]), type 2 diabetes (T2D), and cancer.
Mitochondrial Dynamics in T2D {#S3}
=============================
The clinical complications of T2D include dyslipidemia, hyperglycemia ([@B17]), insulin resistance, and defects in insulin secretion from pancreatic beta cells ([@B18]). A major cause of such clinical complications is the increased production of mitochondrial ROS by hyperglycemia ([@B17], [@B19]). A common feature of mitochondrial morphology in T2D is an increased fragmentation (Figure [1](#F1){ref-type="fig"}), achieved *via* activation/upregulation of DRP1 and/or downregulation of MFN2 levels. In turn, increased fission and fragmentation of mitochondria was linked to HG-induced overproduction of ROS ([@B20]) and insulin secretion in mouse and human islets ([@B21]). Importantly, both HG-induced ROS and insulin secretion were blocked by inhibiting DRP1-induced fission. Furthermore, impaired mitochondrial fusion has been associated with insulin resistance in skeletal muscle ([@B22]) and with glucose intolerance and enhanced hepatic gluconeogenesis in a liver-specific MFN2 knockout (KO) mice ([@B23]). Interestingly, MFN2 KO led to increased ROS production, activation of JNK and endoplasmic reticulum (ER) stress response. Studies in rat models show that MFN2 overexpression improved insulin sensitivity and reduced lipid intermediates in muscle ([@B24]) and liver ([@B25]). At the molecular level, liver expression of MFN2 was associated with increased expression of the insulin receptor and the glucose transporter GLUT2, and activation of the PI3K/AKT2 pathway.
{#F1}
In addition, dyslipidemia models of T2D show increased mitochondrial fission (Figure [1](#F1){ref-type="fig"}). Excess palmitate (PA)-induced mitochondrial fragmentation and increased mitochondrion-associated DRP1 and FIS1 in differentiated muscle cells ([@B26]). In addition, PA induced mitochondrial depolarization, lower ATP synthesis and increased oxidative stress, and reduced insulin-stimulated glucose uptake (Figure [1](#F1){ref-type="fig"}). Both genetic and pharmacological inhibition of DRP1 attenuated PA-induced mitochondrial fragmentation and insulin resistance. In another study, DRP1 was induced in rat islets after stimulation by free fatty acids (FFAs), and this DRP-1 upregulation was accompanied by increased pancreatic β cell apoptosis ([@B27]).
Mitochondrial fission is associated with various processes that contribute to atherosclerosis in T2D (Figure [1](#F1){ref-type="fig"}), including endothelial dysfunction ([@B28]), collagen matrix alteration ([@B29]), and motility and proliferation of vascular smooth muscle cells ([@B30]). From a therapeutic standpoint, silencing FIS1 or DRP1 in venous endothelial cells isolated from patients with T2D blunted HG-induced mitochondrial fission and ROS production ([@B28]). Furthermore, metformin attenuated the development of atherosclerosis in diabetic mice by reducing DRP1-mediated mitochondrial fission in an AMP-activated protein kinase (AMPK)-dependent manner ([@B31]). Mitochondrial fission induced by DRP1 also plays a critical role in the pathogenesis of microvascular \[nephropathy ([@B32]), retinopathy ([@B33]), and neuropathy\] and macrovascular \[stroke and myocardial ischemia ([@B34])\] complications of diabetes.
In summary, we know that many of the clinical complications of T2D are associated with mitochondrial fragmentation. We also know that tipping the balance toward increased mitochondrial fragmentation in mice leads to models of T2D. Furthermore, blocking DRP1 (or increasing MFNs) ameliorated hyperglycemia, dyslipidemia, and atherosclerosis in T2D models. Less clear are the mechanisms of alterations in expression and/or activity of DRP1/MFNs. Up to date, most of the studies have shown correlation between the hallmarks of T2D and increased fragmentation of mitochondria (Table [1](#T1){ref-type="table"}). However, more studies should focus on understanding the spatiotemporal regulation of DRP1 and MFN1/2 levels during the natural progression of T2D. In this context, there are a number of open questions. For example, are there alterations on the regulation of DRP1/MFNs at the transcriptional, translational, or posttranslational level? Are DRP1/MFNs regulated by insulin, glucose, FFA signaling pathways? What are the tissue- and cell-specific differences in the regulation of mitochondrial shape in T2D? Identifying such molecular pathways controlling DRP1/MFN alterations in T2D might enable therapeutic efforts in prediabetic patients to prevent full-blown settlement of the disease.
######
Mitochondrial dynamics in T2D and cancer.
Disease Regulatory event Molecular pathway Cell function Reference
----------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------- -----------
T2D DRP1 enrichment in calcified human carotid arteries DRP1 controls matrix mineralization, cytoskeletal rearrangement, mitochondrial dysfunction, and reduced type 1 collagen secretion and alkaline phosphatase activity Extracellular matrix changes in cardiovascular complications ([@B29])
FFA DRP1 leads to cytC release, caspase-3 activation, and generation of ROS Apoptosis ([@B27])
Hyperglycemia ROCK1 phosphorylates DRP1 Nepropathy ([@B32])
PA Fragmentation was associated with increased oxidative stress, mitochondrial depolarization, loss of ATP production, and reduced insulin-stimulated glucose uptake Insulin stimulated glucose uptake in skeletal muscle ([@B26])
FIS1 and DRP1 increased in T2D patients DRP1 induced ROS, and nitric oxide synthase activation Endothelial dysfunction ([@B28])
Hyperglycemia HG leads to DRP1-mediated fragmentation and ROS Cellular respiration ([@B20])
Inflammatory signaling (TNF-α) TNF-α induced MiR-106b which led to MFN2 downregulation Insulin resistance ([@B23])
Insulin Unknown Unknown ([@B30])
Dyslipidemia MFN2 prevents accumulation of lipid intermediates, including diacylglycerol and ceramides Insulin resistance in skeletal muscle ([@B24])
Dyslipidemia MFN2 promotes the insulin signaling pathway (INSR/IRS2/GLUT2PI3K/AKT) Insulin resistance in liver ([@B25])
Hyperglycemia MFN2 deficiency impaired insulin signaling in muscle and liver, induced ER stress, ROS production, and JNK activation Insulin and glucose homeostasis ([@B23])
Cancer Oncogenic MAPK signaling RasG12V or BRAF^V600E^ activate ERK1/2, which then phosphorylates and activates DRP1 Mitochondria function and cell survival ([@B56])
mTOR mTORC1/4E-BP-dependent translation of MTFP1 leads to activation and recruitment of DRP1 to mitochondria Cell survival ([@B58])
Nestin Nestin binds DRP1 and enhances DRP1 recruitment Proliferation and invasion ([@B59])
EHD1 EHD1 and Rabankyrin-5 interact with the retromer complex and induce VPS35-mediated removal of inactive DRP1 from mitochondrial membranes Unknown ([@B60])
AMPK AMPK phosphorylates MFF, which increases DRP1 recruitment to mitochondria Unknown ([@B61])
SPOP loss-of-function mutants SPOP mutations allow localization of INF2 to mitochondria, where it recruits DRP1 Cell migration and invasion ([@B62])
SIRT4 SIRT4 inhibited Drp1 phosphorylation and weakened Drp1 recruitment to the mitochondrial membrane *via* an interaction with FIS1 Cell migration and invasion ([@B63])
Estradiol Estradiol stimulates mitochondria fission by decreasing MFN1/2 levels Cell migration and proliferation ([@B66])
Androgen Androgens increase DRP1 expression *via* the AR Cell proliferation ([@B65])
*The upstream regulators of mitochondrial shape are presented along with the molecular mechanisms at play*.
*SPOP, speckle-type POZ protein; FFA, free fatty acid; FIS1, fission protein 1; DRP1, dynamin-related protein 1; AMPK, AMP-activated protein kinase; MFF, mitochondrial fission factor; T2D, type 2 diabetes; ER, endoplasmic reticulum; MFN1/2, fusogenic proteins mitofusin 1 and 2; AR, androgen receptor*.
Another question that warrants further investigation is whether genetic susceptibility variants of DRP1 or MFNs are associated with T2D. A recent study in type 1 diabetes patients identified genetic factors associated with kidney disease ([@B35]). We propose that a similar approach in T2D patients could address to what extent genomic alterations of the mitochondrial shape genes are associated with disease. A potential association between genomic alterations of mitochondrial shaping genes and T2D might allow for better screening of susceptibility and/or risk prediction of certain T2D complications.
Mitochondrial Dynamics in Cancer {#S4}
================================
Recent evidence indicates that mitochondrial shape, size, and localization regulate several of the hallmarks of cancer. For instance, mitochondrial shape dynamics have been linked to metabolic adaptation, cell cycle progression ([@B36]), necroptosis ([@B19]), apoptosis ([@B37]--[@B39]), autophagy ([@B40]), tumor growth, tumor cell motility ([@B41], [@B42]), invasiveness, and metastasis ([@B43]). The role of mitochondrial shape changes as regulators of cancer biology is reviewed in Ref. ([@B44]). Here, we will discuss recent insights into how mitochondrial dynamics are regulated in cancer.
When considering the common alterations in mitochondria shape, we find a dichotomy between tumors with enhanced mitochondrial fragmentation versus tumors with enhanced mitochondrial fusion. For instance, hepatocellular carcinoma ([@B45]), osteosarcoma ([@B46]), medulloblastoma ([@B47]), thyroid ([@B42]), colorectal ([@B48]), endometrial ([@B49]), and breast cancer ([@B43]) show increased mitochondrial fragmentation, due to upregulation of DRP1 levels and a concomitant reduction in MFN1/2 levels. On the other hand, tumors of the prostate ([@B50]), neuroblastoma ([@B51]), leukemia ([@B52]), glioblastoma ([@B53]), and lung ([@B54]) are associated with downregulation of DRP1 and increased MFN1/2 levels. What could be driving these contrasting preferences of fission versus fusion of the mitochondrial networks in cancer? Plausible explanations could lie on the genomic landscape, hormonal/growth factor context, tumor microenvironmental conditions, and therapy responses of the tumors in question.
Oncogenic and tumor suppressor signaling converge on mitochondria to reprogram cellular metabolism ([@B55]); thus, the particular genomic events driving a tumor might favor mitochondrial shape changes to meet the metabolic demands of the tumor cells. According to this hypothesis, oncogene-induced metabolic reprogramming should induce changes in mitochondrial shape. Indeed, recent studies show that oncogenic RasG12V, BRAF^V600E^ and MAPK/ERK ([@B56], [@B57]), mTOR ([@B58]) Nestin ([@B59]), and the endocytic protein EDH1 ([@B60]) increase DRP1-mediated mitochondrial fission. Similarly, the energy-sensing AMPK increased recruitment of DRP1 to mitochondria *via* phosphorylation of the MFF and ([@B61]). Speckle-type POZ protein loss-of-function mutations commonly found in primary prostate cancer were associated with increased DRP1 activation, mitochondrial fission, and prostate cancer cell invasion ([@B62]). Recently, loss of expression of the sirtuin SIRT4 was shown to lead to increased mitochondrial fragmentation ([@B63]). The signaling events that lead to DRP1 activation downstream of genomic and epigenetic alterations are summarized in Table [1](#T1){ref-type="table"}.
In addition to the increasing number of oncogenes and tumor suppressors, growth factors and hormones regulate mitochondrial shape. Examples include Sonic Hedgehog ([@B47]), non-canonical Wnt ligands, pro-inflammatory cytokines, transforming growth factor-β, estradiol ([@B64]), and androgens ([@B65]). Estradiol promotes mitochondrial fragmentation through a reduction of MFN2 with parallel increase of FIS1 levels in ER+ breast cancer ([@B66]). From a translational standpoint, overexpression of MFN2 prevented estradiol-induced cell proliferation and motility ([@B66]). On the other hand, DRP1 is a transcriptional target of the androgen receptor, and androgen-stimulated DRP1 expression sensitizes prostate cancer cells to therapy-induced apoptosis ([@B65]). The possibility that other hormone-related malignancies exploit similar mechanisms of mitochondrial shape awaits further confirmation.
Tumor microenvironmental conditions exert yet another layer of regulation of mitochondrial shape. For instance, mitochondrial elongation is induced by nutrient deprivation in cancer cells ([@B67]). A hypoxic environment enhances mitochondrial fission in breast cancer ([@B68]) and glioblastoma ([@B69]). In this context, DRP1 was essential for hypoxia-stimulated cell motility. Indeed, silencing or expression of a dominant-negative mutant of DRP1 inhibited hypoxia-induced migration in both tumor cell models.
Finally, cancer cells also remodel their mitochondrial network in response to therapy. For instance, DRP1-mediated mitochondrial fragmentation is associated with cisplatin ([@B68], [@B70]), cytarabine and methotrexate ([@B71]), and tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) ([@B70]) treatment among others. However, other therapeutic agents such as histone deacetylase inhibitors ([@B72]) produce the opposite effect, namely increased elongation of mitochondria. These opposite effects of therapy upon mitochondria morphology can be reconciled when considering the divergent signaling pathways elicited by the drugs. In the case of HDAC inhibitors, a decreased expression of FIS1 impaired DRP1 recruitment to mitochondria. These effects were independent of apoptosis induction. On the other hand, increased mitochondrial fragmentation on cisplatin and TRAIL-treated cells is coupled to apoptosis. Also worth considering, HDAC inhibitors could have additional roles in regulating mitochondrial morphology, due to non-histone-acetylating activity (acetylation of non-histone proteins, regulation of signaling kinases). A final consideration is the influence of the genomic background and tumor microenvironment on eliciting fission versus fusion upon therapy.
In summary, emerging evidence suggests that the contribution of the mitochondrial shaping genes to tumor cell biology is tumor type dependent and may reflect the genetic makeup, hormonal/growth factor context, tumor microenvironment conditions, and therapy responses of the tumor. Future efforts should aim to integrate these novel regulatory pathways and reach a comprehensive picture of the regulation of mitochondrial shape and function in cancer. Second, more emphasis should be directed toward identifying metabolic-dependent versus -independent functions of DRP1 and MFNs in cancer. For instance, which of the phenotypes associated with DRP1 activation in cancer are explained on basis of metabolism (increased glycolysis versus respiration)? Is it DRP1's function on apoptosis (or mitochondrial localization) also important? A third area of interest for future research would be the development of anti-cancer therapies targeting mitochondrial dynamics. Encouraging fresh evidence indicates that modulating mitochondria morphology enhances anti-cancer therapies ([@B73]), particularly death receptor ligands ([@B74]--[@B76]) and antimitotic drugs ([@B77]).
Targeting Mitochondrial Dynamics {#S5}
================================
The involvement of DRP1-mediated fission in disparate diseases settings has fueled the development of pharmacological approaches to inhibit mitochondrial fission. Mitochondrial division inhibitor-1 (mdivi-1) selectively impairs the GTPase activity of DRP1, without affecting the activity of dynamin-1, MFN1/2, or OPA1 ([@B78]). The mechanism of action of mdivi-1 involves allosteric binding and stabilization of a conformational form of unassembled DRP1 that cannot polymerize. mdivi-1 treatment induces rapid mitochondrial fusion, dampens ROS production and increases ATP production. Interestingly, the original report described a second function of DRP1 in mitochondrial outer membrane polarization (MOMP). DRP1 facilitated BAX/BAK-dependent MOMP in response to C8-BID or staurosporine, independently of mitochondrial fragmentation. Thus, mdivi-1 impaired staurosporine-induced apoptosis ([@B78]). Interestingly, mdivi-1 can induce apoptosis in DRP1-KO cells ([@B79]), suggesting that mdivi-1 has off-target effects. In contrast to these initial studies in which mdivi-1 prevented apoptosis, later studies showed that mdivi-1 sensitized cells to TRAIL-dependent apoptosis ([@B74]). This potentiation of apoptosis by mdivi-1 occurred through activation of mitochondrial and ER apoptosis pathways. Thus, these controversial results suggest that mdivi-1 can act either as pro- or anti-apoptotic pharmacologic agent, depending on the cell types and apoptotic stimuli in question ([@B80]).
In T2D models, mdivi-1 prevented mitochondrial fragmentation, oxidative stress and inflammation, and improved endothelial cell function ([@B31]). Another study showed that mdivi-1 prevented HG-stimulated insulin secretion in mouse and human islets ([@B21]). Furthermore, mdivi-1 rescued palmitate-induced mitochondrial dysfunction and ROS generation, as well as insulin resistance in skeletal muscle ([@B26]). Inhibition of Drp1 with mdivi-1 improved mitochondrial function and cardiac function in a model of myocardial ischemia/reperfusion of diabetic hearts ([@B34]).
In cancer cells, DRP1 inhibition has been shown to modulate therapy sensitivity, tumor metabolism, growth, and invasiveness. For instance, mdivi-1 suppressed mitochondrial autophagy, metabolic reprogramming, cancer cell viability, and motility of breast cancer cells ([@B81]). In regards to therapy modulation, mdivi-1 potentiated TRAIL-induced apoptosis in melanoma ([@B74], [@B76]) and ovarian cancer models ([@B75]). Furthermore, mdivi-1 induced cell death ([@B75]) and synergized apoptotic effects of platinum agents in drug resistant ovarian tumor cells ([@B79]). However, mdivi-1 prevents apoptosis induced by cisplatin in breast cancer ([@B68]) and leukemia ([@B52]). As discussed above, these controversial results suggest that mdivi-1 can act either as pro- or anti-apoptotic agent, depending on the cell types and apoptotic stimuli in question \[reviewed in Ref. ([@B80])\]. Further investigations should address the precise mechanisms dictating the differential effects of mdivi-1 on cell survival.
Regarding the potential utility of mdivi-1 in the clinic, a number of questions remain open. For instance, what are the consequences of sustained *in vivo* inhibition of mitochondrial fission? What are the pharmacokinetics and cytotoxicity profiles for mdivi-1? Another point to consider is that mdivi-1 has poor solubility in water ([@B80]). This fact might limit the utility of mdivi-1 and might open the door for the design of new DRP1 inhibitors with improved solubility, specificity, and potency. In this regard, another pharmacological agent targets the recruitment of DRP1 to mitochondria *via* its interaction with FIS1. The small peptide inhibitor P110 blocks DRP1/FIS1 binding ([@B82]) and has shown promising results in neurodegenerative disease models. When tested in hepatocellular carcinoma, P110 blocked cell proliferation *in vitro* and *in vivo* ([@B83]). Future research will be needed to evaluate the utility of P100 both in T2D and cancer models.
Conclusion {#S6}
==========
Given the metabolic alterations that are a hallmark of both T2D and cancer, it is not surprising that mitochondrial alterations are a shared feature in these disparate diseases. Over the past few years, we have learnt that mitochondria are not static, solitary organelles, but they rather undergo constant changes in morphology and subcellular distribution to meet the metabolic demands of the cell. Defects in mitochondrial dynamics play a role in the molecular and cellular pathogenesis of both T2D and cancer. Now, how similar or different are these two pathologies in regards to mitochondrial dynamics? In T2D, the literature unanimously reports an increase of mitochondrial fission mediated by DRP1. In cancer, most tumors follow this same pattern of increased DRP1-mediated mitochondrial fission. However, although less frequently, tumors might display augmented mitochondrial fusion *via* an increase of MFN1/2 levels and/or activity. How are these differences and similarities in the mitochondrial network explained at the molecular level? Up to date, most of the studies have shown correlation between T2D and altered mitochondrial shape. More studies should focus on understanding the spatiotemporal regulation of DRP1 and MFN1/2 levels and activity during the natural progression of T2D. Likewise, there is limited information on how the genetic, epigenetic, and microenvironmental factors influence mitochondrial dynamics, or which signaling pathways integrate extracellular stimuli with mitochondrial shape in T2D. Thus, due to this limited information, is not possible to conclude if T2D and cancer utilize similar or divergent mechanisms of control of mitochondrial shape. In this regard, it would be interesting to address how metabolic pathways commonly altered both in T2D and cancer impinge on mitochondrial morphology. Examples of such pathways include PI3K/AKT and AMPK. Another question that warrants further investigation is whether other aspects of mitochondrial biology are dysregulated in these diseases. For instance, are there alterations in mitochondrial quality control, mitochondria crosstalk to other organelles, or mitochondrial localization present in both T2D and cancer?
Regarding the use of DRP1 inhibitors as anti-T2D and -cancer agents, further studies should determine long-term effects of targeting mitochondrial dynamics *in vivo*, and establish the pharmacokinetics and cytotoxicity profiles for mdivi-1. In addition, the involvement of potential compensatory or resistance mechanisms to mdivi-1 has not been explored yet and should be addressed in the future. An area in need of further investment is the development of selective MFN1/2 inhibitors. Despite the existence of a few DRP1 inhibitors, there is no equivalent therapeutic agent to target fusion. The fact that several tumors show increased fusion might warrant further effort in this area.
Author Contributions {#S7}
====================
MW performed literature search and review; MC conceived the project, designed the figures, and wrote the paper.
Conflict of Interest Statement {#S8}
==============================
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The authors would like to apologize to those colleagues whose work could not be cited or discussed in sufficient detail due to space limitation. MC is supported by ACS IRG \#16-184-56 from the American Cancer Society. The content is solely the responsibility of the author and does not necessarily represent official views of the ACS. The funders had no role in decision to publish or preparation of the manuscript.
[^1]: Edited by: Che-Pei Kung, Washington University in St. Louis, United States
[^2]: Reviewed by: Dhanendra Tomar, Temple University, United States; Thibaud T. Renault, Humboldt-Universität zu Berlin, Germany; Eirini Lionaki, Foundation for Research and Technology Hellas, Greece
[^3]: Specialty section: This article was submitted to Cancer Endocrinology, a section of the journal Frontiers in Endocrinology
| {
"pile_set_name": "PubMed Central"
} |
{#sp1 .150}
{#sp2 .151}
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Diamond-Blackfan Anemia (DBA) is a congenital disease, charactereized by a defective erythroid progenitor maturation and is associated with physical malformations.
Majority of cases are sporadic and dominant with 10% of the patients demonstrating recessive inheritance.
Mutations in the gene encoding for ribosomal protein RPS19 (DBA1) have been found in 25% of patients with either the dominant or the sporadic traits \[[@B1],[@B2]\].
It is noteworthy that these mutations are associated with mental retardation as well as with learning disabilities in DBA patients \[[@B1],[@B3]\].
Somatic abnormalities have been found in 47% of the patients registered with the DBA Registry of North America \[[@B4],[@B5]\].
Associated physical anomalies and growth retardation are common and outstanding even in patients with multifactorial etiology such as long term steroid treatment \[[@B3]\].
The combination of clinical and molecular findings suggests a contiguous gene syndrome with a gene focus for mental retardation and skeletal malformations.
Repetitive and stereotyped behaviors are as common as mental retardation and in some cases their manifestations reach the threshold for diagnosis of obsessive compulsive disorder (OCD) (according to the Diagnostic Statistical Manual of Mental Disease IV edition Text Revised DSM-IV TR criteria) \[[@B6]\].
In the following case we present a DBA patient with comorbidity of OCD. This case has tremendous significance due to the demonstration of the clinical and the pathophysiologic as well as therapeutic implications, involved in the assessment of behavioral abnormalities in DBA.
Case presentation
=================
L. is a 22 year old Italian male, diagnosed with DBA at the age of two. Since being diagnosed with DBA, L has been treated with monthly blood transfusions and subcutaneous injections of deferoxamine mesylate. Years later, he developed iatrogenic hepatitis due to multiple blood transfusions. Despite attending a special education program for children with learning disabilities, the patient has experienced major difficulties in carrying out daily activities since the age of six. He showed attention deficit at school, social isolation and, since the age of 12, verbal and motor repetitive behaviors, apparently cyclically worsening during mood instability episodes.
L was reluctant to speak about his repetitive behaviors. L\'s parents attributed their child\'s behaviours to the developmental disabilities. A standard psychiatric diagnosis was not reached, no treatment was established during childhood.
L\' s grandfather was diagnosed with OCD (checking compulsions) in comorbidity with an Impulsive Control Disorder (Intermittent Explosive Disorders), his grandmother was depressed and alcohol addict.
At the age of sixteen the patient had an episode of herpetic encephalitis with symptoms of delirium and therefore, he was treated for two years with carbamazepine. One year later he was diagnosed with polyendocrinopathy of hypothyroidism, hypoparathyroidism and hypogonadism. The encephalitis process had no consequences. Neither mental state nor cognition alteration were reported following the episode.
At the age of twenty one repetitive behaviors increased in frequency to a level that required psychiatric attention and pharmacological management. Trying to address both mood symptoms and repetitive behaviors, a treatment with low doses of olanzapine and venlafaxine was established, with no improvement in symptoms and a strong deterioration of patient\'s anemia. L was hospitalized and a more thorough psychiatric assessment was conducted.
The patient is in the lower normal range of height (164 cm) and IQ (87). He complained of impulsive sexual and aggressive thoughts that were intrusive, repetitive and distressing. He also complained of compulsive behaviors and rituals, such as hoarding, arranging, ordering, preoccupations with symmetry, exactness, rewriting and doubting. Interrupting the patient while carrying out his rituals lead to violence. The patient had moderate insight of his illness.
He fulfilled the DSM-IV TR criteria for OCD and diagnosis was established by the Structural Clinical Interview for DSM-IV Axis I Disorders (SCID-I) \[[@B7]\]. In order to determine the severity level of obsessive compulsive symptoms, the Yale Brown Obsessive Compulsive scale (YBOCS) \[[@B8]\], a clinician rated 10 item scale, each rated from 0 (no symptoms) to 4 (extremely severe symptoms), was performed on him and revealed a score of 28, which corresponds to a severe form. A treatment with sertaline 200 mg/day (addressing OCD symptoms) and valproic acid 600 mg/day (with the aim of reducing the impulsive features linked to obsessions and according to its efficacy reported in treatment of DBA) \[[@B9]\] has been started.
In the meantime an MRI exam was done as well and it showed low signal areas due to accumulation of paramagnetic substances in the right temporal lobe and in the ventricular choroid plexus, asymmetric sphenoid sinus and hypoplastic pituitary gland. Sellar region and parasellar structures appeared in a regular pattern. No anomalies in encephalic parenchyma were demonstrated after contrast medium injection.
The patient was assessed 12 weeks after the administration of the YBOCS and demonstrated an improved total score of 15, which corresponds to a mild form of OCD with a reduction of more than 45% of the symptoms.
Discussion
==========
For the first time we have described DBA with comorbid OCD. The above described case could demonstrate heuristically valuable clinical, therapeutic and pathophysiological implications if more DBA patients with comorbid OCD would be screened by hematologists and therefore deserves further discussion.
There is a great importance in the assessment of obsessive-compulsive symptoms in DBA patients with mental or behavioral disturbances. Since OCD often goes undiagnosed in the presence of more pervasive disturbances \[[@B10]\], our report assumes a \"Caveat\" value.
Distinguishing between mental retardation, learning disabilities, Asperger Syndrome and OCD can be challenging, especially when treating children. Precocious diagnosis of OCD can make a tremendous difference in terms of evolutionary trajectory and improved life quality of patients and their families.
Pediatricians should bear in mind the possibility of OCD when treating DBA children with behavioralor learning disabilities even in the absence of other malformations.
It has been shown that when adequate screening tools were adopted in clinical disciplines other than psychiatry (for instance in dermatology and immunology), a larger than expected number of undiagnosed OCD patients was revealed \[[@B11]\].
Also, OCD is potentially linked to brain iron accumulation in DBA. Studies done in animals demonstrated that brain iron accumulation leads to damage of neuronal dopaminergic function.
Intranigral iron injection in rats have shown a detrimental effect on dopamine (DA) release and concentration in the caudate putamen (CPu) as well as selective decrease of striatal dopamine (95%), 3,4-dihydroxyphenylacetic acid serotogenic activity (82%), and homovanillic acid (45%) with related behavioral changes, characterized by increased repetitive and compulsive behaviors.
Thus, hemosiderosis might be contributing to psychiatric symptoms in DBA patients \[[@B12]\]. In fact, OCD symptoms may be linked to hemosiderin deposition in the brain and the pituitary gland, just as hypopituitarism has been shown to be linked to hemosiderin deposition in the pituitary as it was hypothesized previously by Berdel \[[@B13],[@B14]\].
Moreover, since OCD has been related also to multiple regions of cortical thinning \[[@B15]\], the MRI imaging in our case of paramagnetic substance accumulation in the right temporal lobe, ventricular plexus and the hypoplastic pituitary gland is suggestive of the importance of neuroimaging assessment and recognition of complications caused by iron deposition due to long term blood transfusions in the management of DBA.
However a conclusion can not be drawn without a verification through larger studies on populations that have undergone blood transfusions at a young age.
Presenting this case report we have shown that OCD symptoms are treatable in DBA as effectively as in other conditions such as mental retardation \[[@B16]\] and Down Syndrome \[[@B17]\].
Despite the fact that some psychiatric medications have shown a worsening of the symptoms of anemia, SSRIs and valproate have been extremely beneficial and safe.
Authors\' contributions
=======================
SP established the treatment of the patient, conceived the case-report and drafted the manuscript, SM drafted the manuscript, SB drafted the manuscript, MM drafted the manuscript, AI collected information about the case and drafted the manuscript, EH drafted the manuscript. All authors read and approved the final manuscript.
Acknowledgements
================
Written informed consent was obtained from the patient for publication of this Case Report.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
CCTGA has wide spectrum of structural and clinical features. The clinical presentation and prognosis of patients with CCTGA vary depending on the severity of the associated cardiac anomalies, the development of systemic ventricular dysfunction, and the development of arrhythmias. The patients with CCTGA have a progressive risk of spontaneous complete AV block throughout life (2% per year) \[[@R1]\]. The incidence of sudden cardiac death (SCD) in CHDs is approximately 1:1000 patients per year, which is 25-100 times greater than in the general population \[[@R2]\]. Despite the relatively common occurrence of SCD, ventricular tachycardia (VT) has been rarely described in the natural history of CCTGA.
Case Report {#s2}
===========
A 56-year-old male was referred to the emergency department (ED) for fatigue and shortness of breath on exertion, for 3 - 4 days. The patient was bradycardic, and his blood pressure was 160/70 mmHg. Electrocardiography (ECG) showed an atrial rhythm with 2:1 AV block ([Figure 1](#F1){ref-type="fig"}). The patient had no history of syncope. He was hospitalized and monitored in the Coronary Care Unit (CCU). His serial cardiac markers and laboratory data were normal, and there was no reversible cause to explain AV block. A chest X-ray showed slight cardiomegaly. A transthorasic echocardiographic evaluation detected an apically localized systemic AV valve, a dilated left atrium, a parallel arrangement of great arteries, significant systemic AV valve insufficiency, and ejection fraction (EF) of 32%. Multi-slice CT and MR angiography revealed AV and ventricular arterial discordance with persistent left superior vena cava and situs inversus abdominalis ([Figure 2](#F2){ref-type="fig"}). Cardiac catheterization was performed. Selective coronary arteriography demonstrated a well-developed right coronary system.
The patient experienced a sudden cardiac arrest during his follow-up in the CCU. The patient was monitored and intubated immediately. Pulseless ventricular tachycardia was detected ([Figure 3A](#F3){ref-type="fig"}), direct electrical cardioversion was accomplished, and rhythm was restored. The ECG showed atrial tachycardia ([Figure 3B](#F3){ref-type="fig"}) after the successful resuscitation. An electrophysiologic study was performed, but ventricular tachycardia was not induced again. The patient underwent dual-chamber ICD implantation with active fixation ([Figure C](#F3){ref-type="fig"}). The patient was then discharged in clinically stable condition with spironolactone 25 mg, lisinopril 10mg, and bisoprolol 5mg treatment. There were no complaints during routine follow up at one month.
In a routine follow up in the sixth month, the patient was admitted with palpitation. Paroxysmal atrial fibrillation was detected by 24-hour continuous ambulatory ECG. The ICD was checked, and atrial tachycardias with variable conductions were detected. Amiadarone 600 mg and warfarin 5mg daily were started. We did not detect any recurrent attack of atrial fibrillation during the 6 months follow-up after the amiodorane therapy.
Discussion {#s3}
==========
CCTGA, or synonym l-transposition, is a rare (less than 1% of all CHD) and complex heart defect. The characteristic feature of this CHD is AV and ventriculoarterial discordance. The great arteries are generally parallel to each other. The aorta is located closer to the anterior and more to the left than the pulmonary artery. The AV valves follow their respective ventricles. Because of the displacement of the AV node and the abnormal course of conduction tissue, there is an increased risk of spontaneous complete AV block. One study examined a series of patients with CCTGA, with complete AV block detected at 8% of diagnosis; the follow-up showed that 38% patients had been documented as experiencing atrial arrhytmias \[[@R3]\]. Associated cardiac defects are common; isolated CCTGA is an exception.
For a number of patients with CHD, life expectancy increases with the development of diagnostic and interventional techniques. SCD is still the leading cause of death in patients with CHD. CCTGA has the highest mortality among all CHD patients \[[@R4]\]. Despite the relatively common occurrence of SCD, VT has rarely been described in the natural history of CCTGA. Although the mechanism of VT in patients with CCTGA is not clear, it is likely related to triggered automaticity and reentry because of the progressive systemic ventricular dysfunction. There is a general acceptance of ICD implantation in patients with severe ventricular dysfunction for primary prevention. Patients with CHD, however, have not yet been shown to benefit from ICD placement for primary prevention. The Toronto study shows that sudden death is the most common mode of mortality in Tetralogy of Fallot, Ebstein's anomaly, CCTGA, and congenital aortic valve anomaly, suggesting that these groups may benefit even more from primary prevention ICD implant \[[@R5]\]. The main indications for ICD implantation, according to the existing guidelines for patients with CHD, are resuscitated cardiac arrest, sustained VT in the absence of a reversible cause, and syncope with inducible sustained ventricular arrhythmia at electrophysiological testing. On the other hand, 2008 ACC/AHA/HRS guidelines recommend ICD implantation as a Class 1b indication for primary prevention in patients with CHD.
To our knowledge, there are only a few documented cases of VT in a patient with CCTGA ([Table 1](#T1){ref-type="table"}) \[[@R6]-[@R9]\]. An alternative therapy is ablation; Baral et al. reported the successful ablation of monomorphic ventricular tachycardia in a 48-year-old woman with CCTGA, Ebstein\'s malformation of the tricuspid valve, and incessant VT \[[@R6]\]. On the other hand, atrial fibrillation or supraventricular tachycardia is very rare in the follow-up of patients with CCTGA.
Because of the increasing numbers and survival of patients with CHD, physicians will continue to encounter rhythm problems. ICDs and pacemakers do not solve all these problems, and medical therapies like beta-blockers or amiodarone are sometimes needed. Our patient is one of the oldest patients with CCTGA that has been documented in the literature, and interestingly the patient suffered from multiple arrhythmic problems on follow-up.
{#F1}
{#F2}
{#F3}
######
Documented cases of VT in patients with CCTGA in the literature
VT: ventricular tachycardia, CCTGA: Congenitally corrected transposition of the great arteries, ICD: implantable cardioverter defibrillator, AV: atrioventricular
| {
"pile_set_name": "PubMed Central"
} |
{.861}
{.862}
{.863}
{.864}
{.865}
{.866}
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1}
===============
CIDP was first described in the mid-1900s as a progressive motor-dominant disease causing severe weakness ([@bib4]). It was later identified as an autoimmune disorder targeting myelin sheaths rather than axons themselves due to its lack of associated muscle atrophy ([@bib4]). While CIDP triggers have not been identified, the pathogenesis results from segmental inflammatory infiltrates in the perivascular space of nerves ([@bib16]). Its prevalence ranges from 1 to 8.9/100,000, and is typically higher in the male population and can occur at any age ([@bib2]; [@bib8]).
A wide range of symptoms exist, but common hallmarks of CIDP include relapsing-remitting, progressive motor and/or sensory loss of distal nerves ([@bib4]; [@bib16]; [@bib5]). Progressive and ongoing proximal or distal motor weakness typically worsens over 8 weeks ([@bib4]). Other symptoms with variable severity include sensory ataxia, areflexia, and decreased sensation ([@bib4]; [@bib16]).
The variable presentation of CIDP has several shared diagnostic findings, including elevated cerebrospinal fluid (CSF) protein levels and slowed conduction velocities on electrophysiological testing ([@bib16]; [@bib5]). The most commonly affected nerves are typically large fibers with plentiful myelin, like the sural, superficial peroneal, or gracilis motor nerves ([@bib2]; [@bib5]). As with most inflammatory conditions, steroids and intravenous immunoglobulins (IVIg) are the mainstay of treatment ([@bib4]; [@bib2]; [@bib5]).
Central nervous system involvement in CIDP may mimic multiple sclerosis (MS) ([@bib4]). However, MRI criteria supportive of MS classically involves periventricular white matter lesions ([@bib18]; [@bib17]). Whereas, MRI in CIDP demonstrates hypertrophy outside of the brain, most commonly the cauda equina and lumbosacral or cervical nerve roots ([@bib5]). Several reports of hypertrophic cranial nerves in CIDP as a result of blood-nerve barrier breakdown have been described ([@bib3]). These reports have involved a variety of cranial nerves, including the optic nerve ([@bib20]), oculomotor nerve ([@bib1]; [@bib11]), trigeminal nerve ([@bib3]; [@bib1]; [@bib11]; [@bib10]; [@bib7]; [@bib14]), and vestibulocochlear nerve ([@bib20]; [@bib9]; [@bib19]; [@bib15]; [@bib13]; [@bib6]). CSF analyses also differentiate the two: MS is supported by 2 or more oligoclonal bands ([@bib17]) and CIDP displays elevated protein and a leukocyte count less than 10/mm^3^ ([@bib2]; [@bib5]). These distinguishing characteristics are essential to prescribe proper treatment for each since IVIg is not the mainstay of treatment for MS ([@bib12]). This case recognizes a constellation of these findings in the setting of CIDP: bilateral cranial polyneuropathy with hearing loss.
2. Case study {#sec2}
=============
A 35-year-old male Iraqi immigrant was referred to our neurotology clinic for persistent conductive hearing loss despite multiple sets of tympanostomy tubes for serous otitis media. His hearing loss began seven years prior, with the left ear worse than right. He had been diagnosed with CIDP by a local neurologist three years prior based on fluctuating episodes of bilateral lower extremity weakness and numbness since age seven, elevated CSF protein levels, and slowed conduction velocities. He was actively receiving IVIg and steroids for the past two years.
He denied otorrhea, tinnitus, vertigo, or aural fullness. He had no history of frequent ear infections, previous ear surgery, head trauma, use of ototoxic medications or a family history of deafness. His only other surgery included bilateral lateral orbital wall decompression for Graves ophthalmopathy. On exam, he had bilateral ptosis, proptosis and myringosclerosis posteriorly with tympanostomy tubes in place. Audiogram revealed an air-bone gap at all frequencies, left worse than right ([Fig. 1](#fig1){ref-type="fig"}). The pure tone averages were 40 dB on the left and 32 dB on the right.Fig. 1Audiogram demonstrating conductive hearing loss. "O" = right unmasked air. "X" = left unmasked air. "\[" = right masked bone. "\]" = left masked bone.Fig. 1
An MRI ordered by his neurologist demonstrated multiple enlarged cranial nerves, including the tympanic and mastoid segments of both facial nerves ([Fig. 2](#fig2){ref-type="fig"}; [Fig. 2](#fig2){ref-type="fig"}a; [Fig. 2](#fig2){ref-type="fig"}b). Other findings included bilateral enlargement of the oculomotor and trigeminal nerves, extraocular muscles, foramen ovale ([Fig. 2](#fig2){ref-type="fig"}a), foramen rotundum and stylomastoid foramen. His extraocular muscle and oculomotor nerve enlargement likely both contributed to his ptosis and proptosis.Fig. 2Post contrast axial T1-weighted MRI demonstrating bilaterally thickened tympanic segments of the facial nerve (arrows). a: Non-contrast axial CT scan showing bilaterally enlarged fallopian canals in the mastoid segment (white arrows), as well as dilated foramen ovale (red arrow). b: Non-contrast axial CT scan showing ossicular erosion secondary to enlarged tympanic segment of the facial nerve.Fig. 2
The patient underwent left middle ear exploration with intraoperative facial nerve monitoring. After elevating a tympanomeatal flap, a significant soft tissue mass was identified just medial to the chorda tympani nerve ([Fig. 3](#fig3){ref-type="fig"}), suspicious for a grossly enlarged facial nerve sheath. This was confirmed with a stimulus from the facial nerve monitor. There was erosion of the incus and partial erosion of the stapes superstructure. The hypertrophic nerve was gently manipulated in an attempt to identify the stapes footplate, but it could not be well visualized, and given these findings, further ossicular chain reconstruction was contraindicated and the procedure was terminated. It was recommended that he pursue bilateral amplification.Fig. 3Intraoperative finding of the soft tissue mass (\*) just medial to the chorda tympani nerve (Chorda). The photo is in surgical position, with the left side of the photo as inferior and top half of the photo as anterior. There is a small piece of surgical packing anteroinferiorly. "RW" = round window. "Promontory" = cochlear promontory.Fig. 3
3. Discussion {#sec3}
=============
CIDP is rarely associated with hearing loss and only a few cases of sensorineural loss have been described ([@bib9]; [@bib19]; [@bib15]; [@bib13]). These reports describe the onset of hearing loss concurrent with the initial CIDP symptoms and were attributed to demyelination of segments of the vestibulocochlear nerve considering its proximity in presentation to CIDP onset ([@bib9]; [@bib19]; [@bib15]; [@bib13]). None had other cranial neuropathies ([@bib9]; [@bib19]; [@bib15]; [@bib13]). However, there is one report of primarily vestibular involvement, rather than cochlear ([@bib6]).
Two of the four reported cases completely recovered hearing with steroids and IVIg ([@bib9]; [@bib15]). The other three patients described in the literature did not have improvement in hearing with steroids or IVIg ([@bib19]; [@bib13]). This paper describes a novel case of conductive hearing loss and bilateral cranial polyneuropathy. His hearing loss persisted despite two years of steroids and IVIg therapy. Initially, the hearing loss was likely secondary to a mass effect on the ossicles with subsequent ossicular erosion from pressure necrosis. After further discussion with his neurology team, further treatment was not pursued and deemed unlikely to reduce facial nerve hypertrophy.
Facial nerve hypertrophy has rarely occurred in the setting of CIDP ([@bib21]). These findings of hypertrophy have been attributed to the frequent demyelination-remyelination cycles causing Schwann cell proliferation ([@bib10]). Our patient\'s symptoms had some similarities and differences compared to previous reports. He lacked facial paresthesias despite his thickened trigeminal nerves while another case reported mandibular hypesthesia ([@bib10]). However, he presented with ptosis and proptosis as described in previous case reports with oculomotor nerve or extraocular muscle thickening ([@bib3]; [@bib1]; [@bib10]). Some of these cases also involved Graves ophthalmopathy which may suggest an association between CIDP and increased incidence of autoimmune disease ([@bib1]; [@bib10]).
4. Conclusion {#sec4}
=============
CIDP has a wide range of symptoms with varying courses and severity of disease that should be distinguished from other autoimmune demyelinating disorders like MS. The patient presented in this case report exhibits a unique cranial polyneuropathy with resultant hearing loss from mass effect secondary to facial nerve hypertrophy and subsequent ossicular erosion.
Declarations of interest {#sec5}
========================
None.
Funding {#sec6}
=======
None.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Plants have to face a broad range of invading pathogens. In response, they can deploy a large set of defense responses including constitutive pre-existing physical and chemical barriers as well as an innate immunity activated after pathogen perception ([@B68]; [@B5]). The first line of recognition is based on the detection *via* pattern recognition receptors (PRRs) of evolutionarily conserved elicitors, also called microbe-, pathogen-, or damage-associated molecular patterns (MAMPs; PAMPs; DAMPs) ([@B14]; [@B31]). MAMP-triggered immunity (MTI) is characterized by early and long-term physiological responses including reactive oxygen species (ROS), ethylene (ET) production, MAPK activation, reprogramming of transcriptome and metabolome (e.g., production of phytoalexins and SA), and callose deposition ([@B5]). One of the earliest responses at the time of pathogen assault is the production of ROS, which plays a crucial role to restrain pathogen development through programmed cell death at the site of infection ([@B58]; [@B38]). The second stage of perception corresponds to the direct or indirect recognition of pathogen effectors by intracellular immune receptors leading to effector-triggered immunity (ETI; [@B31]). MTI and ETI will answer to activate early signaling events in plant defense ([@B62]). Plant hormones, salicylic acid (SA), jasmonic acid (JA), ET, and abscisic acid (ABA) appear as key regulators in defense-signaling networks ([@B42]).
Pathogen attack not only affects plant defenses reactions but can also lead to changes in photosynthesis rates and consequently carbohydrates metabolism. Indeed, during the resistance response, the production of defense-related compounds becomes the high priority of the plant leading to reduced photosynthetic rates until the end of the pathogen growth ([@B46]). As plant defense responses may alter the pool size of a range of metabolic intermediates, photosynthetic metabolism likely will be influenced as it will be regulated to meet the cell/plant requests. The photosynthesis decreases through the infection process as a result of leaf metabolism perturbation attributed to sugar-mediated repression of photosynthetic gene expression ([@B7]). Cell wall invertase (Cw-Inv) catalyzes the cleavage of the sucrose into glucose and fructose, and supply sink organs with carbohydrates, playing thus a crucial role in the regulation of carbohydrate partitioning ([@B44]; [@B45]; [@B55]). Additionally, starch reserves may also be converted to soluble sugars ([@B11]) that may act as signals to induce pathogenesis-related (PR) protein genes and to increase plant resistance ([@B52]).
The use of plant growth-promoting rhizobacteria (PGPR) to induce plant resistance is one of the alternatives developed to protect crops against damages caused by various forms of stress ([@B67]). Induced resistance in grapevine against *Botrytis cinerea*, the gray mold agent, by beneficial rhizobacteria has been reported ([@B2]; [@B65], [@B64]; [@B24]). Among the plant-growth promoting bacteria, *Burkholderia phytofirmans* strain PsJN is able to colonize a variety of genetically unrelated plants such as potato and tomato ([@B13]; [@B40]), maize, switchgrass ([@B32]), both endophytically and at the rhizoplane. In addition to colonize grapevine tissues ([@B12]), *B. phytofirmans* PsJN promotes the growth of roots and also of the aerial part after root inoculation ([@B1]). In addition, during the interaction between *B. phytofirmans* PsJN and grapevine, the bacterium triggers a transient extracellular alkalinization, the production of SA and accumulation of defense-related transcripts, suggesting that this PGPR is perceived by grapevine cells potentially *via* MAMP detection ([@B8]). Moreover, [@B59] showed that flagellin from *B. phytofirmans* PsJN induced resistance against *B. cinerea* and suggest its implication to evade from plant's immune recognition system. The endophytic presence of *B. phytofirmans* PsJN leads to protection against abiotic stresses including cold in grapevine ([@B17]; [@B56]), drought in wheat ([@B39]) or salt and freezing in *Arabidopsis* ([@B43]; [@B53]). It has also been shown that this strain reduces damages caused by chilling in grapevine through a priming of plant defense responses and changes in primary metabolism, particularly an increase of soluble sugars concentration and an accumulation of proline ([@B3]; [@B17],[@B18]; [@B56]). In addition, the bacterium improves tolerance against biotic stress as *Verticillium* sp. in tomato ([@B50]) or *B. cinerea* in grapevine ([@B1]; [@B2]). However, the mechanisms involved beyond the observed induced resistance are not elucidated.
To decipher the mechanisms induced by *B. phytofirmans* PsJN to confer grapevine resistance against *B. cinerea*, we determined (i) the direct antimicrobial effect of PsJN on *B. cinerea* growth; (ii) the effect of *B. phytofirmans* PsJN on the early signaling events (callose, ROS), and on the induction of defense response signaling pathway (gene expression); and finally (iii) changes in leaf carbohydrate metabolism including gene expression, sugar levels and chlorophyll fluorescence imaging after *Botrytis* challenge.
Materials and Methods {#s1}
=====================
Plant Material
--------------
Plantlets of *Vitis vinifera* cv. Chardonnay (clone 7535) were micro-propagated by nodal explants grown on 15 ml of agar medium in 25 mm-culture tubes as described by [@B3]. Cultures were performed in a growth chamber under white fluorescent light (200 μmol/m^2^ s), with 16 h/8 h day/night photoperiod at a constant temperature of 26°C.
Microorganisms
--------------
*Burkholderia phytofirmans* strain PsJN tagged with GFP was cultivated in King's B liquid medium supplemented with kanamycin (50 μg/ml) for 24 h with agitation of 180 rpm at 28°C. Bacteria were collected after centrifugation at 4500 g at 4°C for 15 min and suspended in phosphate-buffer saline (PBS 10 mM, pH 6.5). The concentration of bacteria was determined by spectrophotometry (600 nm) and adjusted to 10^9^ CFU/ml with PBS (*D*~0~ = 0.8).
*Escherichia coli* was cultivated in LB liquid medium for 24 h with agitation of 180 rpm at 37°C. The concentration of bacteria was determined by spectrophotometry (600 nm) and adjusted to 10^9^ CFU/ml with PBS (*D*~0~ = 1).
*Botrytis cinerea* strain 630 was grown on solid medium tomato juice \[33% (v/v), agar 5% (w/v)\] at 20°C. For the inoculum preparation, conidia of *B. cinerea* were collected from 20-day-old culture plates by scratching the Petri dishes surface with sterile potato dextrose broth (PDB 12 g/l) and filtered to remove hyphae. Conidial concentrations were measured and the final density was adjusted to 10^5^ conidia/ml. After incubation during 3 h at 20°C and 150 rpm, germinated spores were used for plant inoculation.
Inoculation of *In vitro*-Plantlets with *B. phytofirmans* Strain PsJN and Infection by *B. cinerea*
----------------------------------------------------------------------------------------------------
Roots of 4-week-old grapevine plantlets were inoculated with 200 μl of bacterial (*E. coli* or PsJN) inoculum (10^9^ CFU/ml). Control and bacterized plantlets were then grown for an additional week before their transfer aseptically into sterile Magenta boxes containing 60 g of soil. After 3 days, each leaf of the plantlet was covered by 2 drops (5 μl each) of suspension of *B. cinerea* germinated spores. This protocol was used for measures of necrosis diameter.
For H~2~O~2~ production, callose deposition, analysis of gene expression, sugar/starch measures, and IMAGING-PAM analysis, plantlets were sprayed with the germinated spore suspension of *B. cinerea* in order to have a homogenous application. Plantlets were placed in growth chamber at 20°C. Leaves were then sampled at different time points after *B. cinerea* challenge.
Observation of *B. cinerea* Mycelium Development after Trypan Blue and Aniline Blue Staining
--------------------------------------------------------------------------------------------
Leaves of control and root-bacterized plantlets collected 24, 48, 72 hpi with *B. cinerea* were stained with lactophenol--trypan blue and destaining in saturated chloral hydrate as described in [@B33] or with 0.05% aniline blue. The mycelium development was then observed using 3D (Keyence, France) or epifluorescence microscope.
Rhizoplane and Endophytic Colonization
--------------------------------------
To determine rhizoplane colonization of *B. phytofirmans* PsJN in the roots, the samples were removed from soil and vortexed (240 rpm) with PBS for approximately 1 min. The homogenate was serially diluted in 10 fold steps and cultured on King's B medium plates (in triplicates) supplemented with kanamycin (50 μg/ml). For endophytic colonization, roots were surface sterilized with 70% ethanol for 3 min, followed by 0.01% commercial bleach and a 0.01% Tween 20 solution for 1 min, and then washed four times in distilled water (1 min each time). Leaves were surface sterilized with 0.01% commercial bleach and a 0.01% Tween 20 solution for 3 min, and then washed four times in distilled water (1 min each time). The samples were then ground with 1 ml of PBS. The homogenate was serially diluted in 10 fold steps and cultured on King's B medium plates (triplicates) supplemented with kanamycin (50 μg/ml). The bacterial colonies were counted after 3 days of incubation at 28°C.
Spore Germination Assay
-----------------------
*B. cinerea* spore germination with 10^2^, 10^4^, or 10^6^ CFU/ml of *B. phytofirmans* PsJN was assessed in 96-well microplates. *B. cinerea* were collected in Potato Dextroxe Broth (PDB) and were added in each well to a final concentration of 5,000 spores, in triplicate, in a total volume of 100 μl. The plates were incubated at 20°C in the dark. Germ tube growth was observed 24 h after challenge using inverted light microscopy (Leica, Wetzlar, Germany).
To test the effect of different soluble sugars (sucrose, glucose, fructose) on *B. cinerea* spore germination, conidia were collected in PDB supplemented with different sugar concentrations (0.1, 1, or 2%). Germ tubes were observed by inverted light microscopy (Leica, Wetzlar, Germany) 24h later. Both experiments were repeated twice (each in triplicates).
Direct Effect of *Burkholderia phytofirmans* PsJN on *B. cinerea* Growth
------------------------------------------------------------------------
The 4-week-old grapevine leaves of plantlets were sprayed with different concentration of *B. phytofirmans* PsJN (0, 10^2^, 10^3^, 10^4^, and 10^6^ CFU/ml) and after 30 min infected with 2 ml of *B. cinerea* (10^5^ conidia/ml). Plantlets were placed in the growth chamber at 20°C. Leaves sampled 0, 2, 24, 48, and 72 h after challenge were immediately frozen in liquid nitrogen and stored at --80°C. Frozen samples were used for RNA extraction and *BcActin* gene analysis. The means ± standard deviations originated from two independent experiments realized in duplicates, each replicate consisted of a pool of six plantlets.
Detection of H~2~O~2~
---------------------
For histochemical detection of H~2~O~2~, the 3,3-benzidine-HCl (DAB, Sigma--Aldrich) method was used according to [@B57]. Fresh entire leaves from control and bacterized plantlets were immersed in the DAB solution (1 mg/ml for 6 h at 37°C), at 8 h post infection with *B. cinerea*, before observations under an optical microscope. The H~2~O~2~ content was also evaluated according to [@B56], with some modifications. Briefly, leaf powder (100 mg) was homogenized in 200 μl of ice-cold acetone and the mixture was centrifuged at 13,500 × *g* for 10 min. Cold water (100 μl) and 40 μl of 5% titanyl sulfate were added to the supernatant, followed by 200 μl of 1 N NH~4~OH solution to precipitate the peroxide-titanium complex. After centrifugation at 6,000 × *g* for 5 min, the supernatant was discarded and the pellet was washed with cold acetone. The precipitate was then dissolved in 600 μl of 2 N H~2~SO~4~ and the final volume adjusted to 800 μl with cold water. The absorbance of the solution was read at 415 nm, and H~2~O~2~ content was calculated according to a standard curve. The means ± standard deviations originated from three independent experiments realized in duplicates, each replicate consisted of a pool of six plantlets.
Callose Deposition
------------------
Callose deposition was observed as described in [@B48]. Detached leaves were collected at 24 h post infection with *B. cinerea* and then incubated overnight in 95% ethanol. De-stained leaves were washed in 150 mM K~2~POH~4~ for 30 min and thereafter stained for 2 h with 0.01% aniline blue in 150 mM K~2~POH~4~. Micrographs were taken by epifluorescence microscope with UV filter (BP, 340-380; LP, 425 nm). This experiment was repeated twice and each replicate consisted of six leaves.
RNA Extraction and Real-Time Quantitative RT-PCR
------------------------------------------------
For each sample, 50 mg of leaves were ground in liquid nitrogen. Total RNA was isolated using Extract'All (Eurobio) and 250 ng was used for reverse-transcription using the Absolute MAX 2-Step QRT-PCR SYBR^®^ Green Kit (Thermo Electron)^TM^ according to the manufacturer's instructions. The transcript levels were determined by real-time quantitative PCR using the Chromo4 system (BIO-RAD, Hercules, CA, USA) and the SYBR Green Master Mix PCR kit as recommended by the manufacturer (Applied Biosystems). PCR reactions were carried out in duplicates in 96-well plates (15 μl per well) in a buffer containing 1x SYBR Green I mix (including Taq polymerase, dNTPs, SYBR Green dye), 280 nM forward and reverse primers and 1:10 dilution of reverse transcript RNA. After denaturation at 95°C for 15 min, amplification occurred in a two-step procedure: 15 s of denaturation at 95°C and 1 min of annealing/extension at 60°C, with a total of 30 cycles. Identical thermal cycling conditions were used for all targets. Specific primers were designed using the Primer Express software (Applied Biosystems, Foster City, CA, USA) and are presented in the Supplementary Table [S1](#SM1){ref-type="supplementary-material"}. PCR efficiency of the primer sets was calculated by performing real-time PCR on serial dilutions. For each experiment, PCR reactions were performed in duplicate and 3 independent experiments were analyzed. Results correspond to means ± standard deviation (SD) of duplicate reactions of three independent experiments. Relative gene expression was determined with the formula fold induction: 2^-ΔΔCt^, where ΔΔ*C*t = (*C*t GI \[unknown sample\]--*C*t GI \[reference sample\])--(*C*t reference genes \[unknown sample\]--*C*t reference genes \[reference sample\]). GI is the gene of interest. *EF1a* and *60RSP* are used as internal controls. The reference sample is the "control+buffer" sample, chosen to represent 1x expression of the gene of interest. The means ± standard deviations originated from three independent experiments realized in duplicates, each replicate consisted of a pool of six plantlets.
Transmission Electron Microscopy of Grapevine Leaf Cell Structure
-----------------------------------------------------------------
Fresh leaves were collected at 24 h post infection with *B. cinerea* and fixed for 20 h at room temperature in 1% (w/v) glutaraldehyde in 0.1 M phosphate buffer (pH 7.2) with 0.5% (w/v) sucrose and 0.2% (v/v) Tween 20. After three rinses (5 min each) with the phosphate buffer containing 0.3% (w/v) sucrose, samples were fixed for 4 h in 1% (w/v) osmium tetroxide in phosphate buffer with 0.5% (w/v) sucrose. The samples were then dehydrated in an alcohol series, transferred to acetone, and finally, they were embedded in Araldite. Transverse ultrathin sections (80 nm nominal thickness) were cut (Reichert Jung Ultracut E) from the Araldite-embedded block and mounted on 200 mesh copper grids. Sections were observed under a JEM2100F TEM (JEOL) without post-staining. Micrographs were recorded using an Orius 200D CCD camera (Gatan). For this experiment, 5 leaves from five plants were used.
Sucrose, Glucose, and Fructose Analysis
---------------------------------------
Fifty mg of frozen leaves powder was mixed with 500 μl of 0.1 M potassium phosphate buffer (pH 7.5) and centrifuged at 1000 g at 4°C for 15 min. The supernatants were recuperated and an aliquot (50 μl) was used to measure the concentration of sucrose, glucose, and fructose. The analysis was performed using enzymatic kits (Megazyme) according to manufacturer's protocol. The means ± standard deviations originated from two independent experiments realized in duplicates, each replicate consisted of a pool of six plantlets. Results were expressed in mg/g dry weight.
Starch Extraction and Analysis
------------------------------
For starch analysis, the pellets from soluble sugar extraction were re-suspended and vortexed in dimethyl sulfoxide-8 M hydrochloric acid (4/1/ v/v). Starch was dissolved over 30 min at 60°C with constant agitation. After centrifugation for 5 min at 5000 g, 100 μl supernatant were added with 100 μl iodine-HCL solution (0.06% KI and 0.003% I~2~ in 0.05 M HCL) and 1 ml distilled water and incubated at room temperature for 15 min. The spectrophotometer was zeroed with the control (blank), and absorbance was read at 600 nm. The means ± standard deviations originated from three independent experiments realized in duplicates, each replicate consisted of a pool of six plantlets. The results were expressed in mg/g dry weight.
IMAGING-PAM Analysis
--------------------
Chlorophyll fluorescence parameters and the redox change of P700 were measured with an IMAGING-PAM measuring system (Heinz Walz, Germany) using the saturation pulse method. Control and bacterized plantlets were dark adapted for 30 min to determine the minimal level of fluorescence (*F*~0~) and the maximal fluorescence (*F*~m~) after a saturating flash (1 s; 13,000 μmol/m^2^ s). Leaves were then exposed to an actinic illumination of 79 μmol/m^2^ s. After fluorescence stabilization, a second saturating flash was imposed to determine the maximal fluorescence (*F*~m~′) of light-adapted inflorescences. Removal of the actinic light and exposure to a short period of far-red light allowed measurement of the zero level of fluorescence (*F*~0~′). The electron transport rate of PS II is calculated according to the equation \[ETR = Y (II) × PAR × 0.5 × PAR absorptivity\]. The effective PSII quantum yield, Y(II), is calculated according to the equation of [@B22]. The quantum yield of regulated energy dissipation in PSII, Y(NPQ), and the quantum yield of non-regulated energy dissipation in PSII, Y(NO), is calculated according to [@B34]. Note that Y (II) + Y (NPQ) + Y (NO) = 1. The data were collected out of necrosis area. The means ± standard deviations originated from three independent experiments realized in duplicates, each replicate consisted of six plantlets.
Statistical Analysis
--------------------
Statistical analyses were carried out using the statistical software SISVAR. Shapiro-Wilk test (α \> 0.05) was used for normality test, and Levene's test (α \> 0.05) for homogeneity of variances test. The data of gene expression, sugars/starch, H~2~O~2~, Imaging PAM was analyzed using two-way analysis of variance (ANOVA). When differences in the means were significant, Tukey test (α = 0.05) was applied to determine which treatments were significantly different from others. For lesion diameter Student's *t*-tests (α \> 0.05) was used to compare lesion area between inoculated and non-inoculated plants.
Results
=======
*Burkholderia phytofirmans* PsJN-Triggered Immunity in Grapevine against *B. cinerea*
-------------------------------------------------------------------------------------
In order to test the capacity of *B. phytofirmans* PsJN to protect grapevine in our system, we performed infection on leaves with *B. cinerea* strain 630 on control or root-bacterized plantlets. Assays performed on detached leaves from bacterized plantlets inoculated with drops of *B. cinerea* strain 630 conidia showed that *B. phytofirmans* significantly reduced *Botrytis*-related necrosis by approximately 50% 72 hpi (**Figure [1A](#F1){ref-type="fig"}**). In addition, whole potted-plant infection was carried out to quantify the gray mold disease symptoms in control *versus* bacterized plants. Therefore, whole plants were sprayed with *B. cinerea* spores suspension and development of decay was monitored 24, 48, 72, and 96 hpi. As for detached leaves, disease symptoms were significantly reduced in bacterized plants, confirming the protective impact of *B. phytofirmans* against *B. cinerea* (**Figure [1B](#F1){ref-type="fig"}**). Further, fungal growth was monitored *in planta* at 2, 8, 24, 48, and 72 hpi by analyzing the transcript levels of the *B. cinerea* actin gene (*BcActin*) by qRT-PCR. While no significant differences were observed between control and PsJN root-inoculated plants at 2, 8, and 24 hpi (Supplementary Figure [S1A](#SM1){ref-type="supplementary-material"}), the *Bc-Actin* transcript level in bacterized plantlets was approximately 250-fold and 800-fold lower compared to non-bacterized plantlets at 48 and 72 hpi, respectively. In order to monitor that the induced resistance of plantlets toward *B. cinerea* was not due to a general effect of bacterization, we carried out the same *B. cinerea* infection procedure using *E. coli*-bacterized plantlets. Results showed that no protection was conferred by the presence of *E. coli* against the fungus (Supplementary Figure [S1A,B](#SM1){ref-type="supplementary-material"}). These data clearly indicated that the significantly enhanced resistance toward *Botrytis* infection is related to the presence of *B. phytofirmans* PsJN.
{#F1}
In addition, *B. cinerea* development was also visualized *in planta* by microscopy (3D and epifluorescence) after trypan (Supplementary Figure [S1C](#SM1){ref-type="supplementary-material"}) or aniline blue (Supplementary Figures [S1D](#SM1){ref-type="supplementary-material"}) staining. As shown in Supplementary Figure [S1](#SM1){ref-type="supplementary-material"}, the development of the fungus was moderately reduced in bacterized plantlets at 24 hpi compared to non-bacterized ones. However, fungal hyphae growth was clearly inhibited in bacterized plantlets 72 hpi with *B. cinerea*. Interestingly, while PsJN was not observed at the leaves surface in the absence of the pathogen, the bacteria were detected at the surface, surrounding the fungal mycelium in botrytized leaves (**Figure [2](#F2){ref-type="fig"}**).
{#F2}
Implication of the *Burkholderia phytofirmans* PsJN-Direct Antifungal Effect on Grapevine Protection
----------------------------------------------------------------------------------------------------
To test if the bacterium could act *via* an antimicrobial effect, we estimated its effect on fungal spore germination and also the ability of *B. cinerea* to develop in the host tissues when inoculated rapidly after direct leaves bacterization.
For spore germination assay, the conidial concentration was adjusted to 5.10^4^ conidia/ml and incubated during 3 h at 150 rpm. *B. phytofirmans* PsJN was then added at 0, 10^2^, 10^4^, or 10^6^ CFU/ml and *B. cinerea* germ tubes growth was observed 24 h after. Germ tubes growth assay showed an inhibition of 32%, 62%, and 88% 20 h after addition of 10^2^, 10^4^, and 10^6^ CFU/ml of bacteria, respectively (**Figure [3A](#F3){ref-type="fig"}**).
{#F3}
To establish the role of the direct antifungal effect of *B. phytofirmans* PsJN in the inability of *B. cinerea* to grow *in planta*, leaves of 6-week-old plantlets were sprayed with *B. phytofirmans* PsJN at different concentrations and then infected by *Botrytis* (10^5^ conidia/ml) 30 min later to prevent the full establishment of plant defense responses. The quantification of *BcActin* in leaves, as an indicator of the rate of fungal growth *in planta*, was then realized 2, 24, 48, and 72 hpi with *B. cinerea*. Our results showed a net dose dependent impact of *B. phytofirmans* PsJN on fungal development (**Figure [3B](#F3){ref-type="fig"}**). Indeed, no significant difference with control was observed at 10^2^ CFU/ml. However, at highest concentrations (10^4^ and 10^6^ CFU/ml), a very low level of *BcActin* gene expression was detected, indicating a strong protective effect of *B. phytofirmans* PsJN at these concentrations.
*Burkholderia phytofirmans* PsJN Primes H~2~O~2~ Production and Callose Deposition after Pathogen Challenge
-----------------------------------------------------------------------------------------------------------
H~2~O~2~ production is an important part of grapevine defense system ([@B9]) and is considered as a signal molecule to activate disease resistance ([@B41]). No significant H~2~O~2~ production was observed in response to bacterium or fungus inoculation (**Figure [4A](#F4){ref-type="fig"}**). However, when bacterized plantlets were inoculated with *Botrytis*, the H~2~O~2~ production was primed. In addition, the H~2~O~2~ generation was localized *in situ* after DAB (3,3-diaminobenzidine) staining. In control and bacterized plants without the pathogen, no H~2~O~2~ generation was visualized except in veins (**Figures [4B,C](#F4){ref-type="fig"}**), which may probably correspond to the lignification process. In botrytized plantlets, quite staining spots were sporadically observed on leaves (**Figure [4B](#F4){ref-type="fig"}**) corresponding most likely to the *Botrytis* conidia-generated ROS (**Figure [4C](#F4){ref-type="fig"}**). On the opposite, DAB deposits observed in bacterized plantlets following *Botrytis* infection were due to H~2~O~2~ production by plant cells and cover the whole leave surface, indicating that *B. phytofirmans* PsJN primed an oxidative burst in grapevine leaves after challenge by the fungus.
{#F4}
Callose depositions are an important characteristic of defense mechanisms and are thought to reinforce the cell wall at fungal penetration sites to impede infections ([@B63]; [@B37]). Therefore, the impact of *B. phytofirmans* PsJN on callose synthesis in grapevine plantlets was monitored after aniline blue staining at 24 hpi (**Figure [4D](#F4){ref-type="fig"}**). The results indicate the absence of callose deposition in the control plant. A similar result was observed after botrytis challenge. In contrast, an obvious callose deposition was observed in stomata of *B. phytofirmans* PsJN-inoculated plantlets. The observed callose accumulation has been strengthened after infection by *Botrytis*. These data indicated that the presence of *B. phytofirmans* PsJN in grapevine plantlets primed callose deposition after pathogen challenge.
Priming of Both SA- and JA-Related Genes by the PGPR
----------------------------------------------------
To further elucidate possible mechanisms contributing to *B. phytofirmans* PsJN-IR against *B. cinerea*, the expression of SA- (*PR1, PR2, PR5*, and *VvWRKY* transcription factor 3), JA- (*HPLA, JAZ*, and *AOC1*) ([@B21]) ET (*ETR1*) and ABA-related genes (*VvZEP, VvNCED*) ([@B26]) was monitored in control and root-bacterized plantlets, 24 h after challenge by *B. cinerea*. In bacterized plantlets, no significant difference was observed in transcript accumulation except a slight repression of *VvZEP* and a slight induction of *AOC* (**Figure [5](#F5){ref-type="fig"}**). These results indicated that bacteria, alone, modulate slightly the gene transcript levels. The *Botrytis* infection alone induced a significant increase in *PR5, WRKY, AOC, HPLA*, and *JAZ* gene expression levels. In response to *Botrytis*, the bacterized plantlets exhibited a stronger expression of *PR* genes (*PR1, PR2*, and *PR5*), and *JAZ*. Taken together, these data suggest that, in response to a subsequent infection by *B. cinerea, B. phytofirmans* PsJN potentiates the simultaneous induction of the SA- and JA- related genes.
{#F5}
*Burkholderia phytofirmans* PsJN Alleviates Photosystem Irreversible Damages
----------------------------------------------------------------------------
The activation of plant defenses requires an increased energy supply that ultimately must come from photosynthesis ([@B6]; [@B51]). In order to evaluate the effect of root-inoculation with *B. phytofirmans* PsJN on photosynthesis before and 24, 48, and 72 h post-infection with *B. cinerea*, changes in excitation flux at PSII were monitored. Photosynthetic parameters including effective PSII quantum yield Y(II), quantum yield of non-regulated energy dissipation Y(NO), quantum yield of regulated energy dissipation Y(NPQ), relative photosynthetic electron transport \[ETR\] and maximum PSII quantum yield (*F*~v~/*F*~m~) were evaluated (**Figure [6](#F6){ref-type="fig"}**). The false color scales shown at the bottom of the fluorescence images indicate the amplitude of the particular parameter (**Figure [6A](#F6){ref-type="fig"}**). Before infection with the pathogen, no significant difference between bacterized and non-bacterized plantlets was observed regarding the monitored photosynthetic parameters. However, 24 h after inoculation with *B. cinerea*, bacterized plantlets exhibit faint symptoms (**Figure [6A](#F6){ref-type="fig"}**) and a significant increase of Y(II) in comparison to control (**Figure [6B](#F6){ref-type="fig"}**). In addition, ETR was significantly improved in bacterized plantlets 48 h after pathogen challenge (**Figure [6D](#F6){ref-type="fig"}**). Further, the efficiency of PSII quantum yield Y(II) decreased in control plantlets 48 h after infection with *B. cinerea* in parallel to a significant increase of Y(NPQ). These changes were accompanied by a significant decrease in both *F*~v~/*F*~m~ and ETR (**Figures [6C,D](#F6){ref-type="fig"}**). At 72 hpi, Y(II) value was lower in non-bacterized plantlets compared to bacterized ones probably due to the significant increase of quantum yield of non-regulated energy dissipation Y(NO). The latter parameter indicates an irreversible damage of photosynthetic apparatus as confirmed by the decreased *F*~v~/*F*~m~ ratio (**Figure [6C](#F6){ref-type="fig"}**). Interestingly, photosynthetic parameters were not affected by *B. cinerea* infection in bacterized plantlets.
![**Fluorescence parameters from grapevine leaves inoculated or not with *B. phytofirmans* PsJN 24, 48, and 72 hpi with *B. cinerea*. (A)** Images of the effective PSII quantum yield Y(II), the quantum yield of regulated energy dissipation Y(NPQ) and of non-regulated energy dissipation Y(NO). The pixel value display is based on a false-color scale ranging from black (0.000) via red, yellow, green, blue, to purple (ending at 1.00). The figure shows representative images of one from three independent experiments. **(B)** Changes in excitation flux at PSII in infected leaves of plantlets root-inoculated or not with *B. phytofirmans* PsJN 24, 48, and 72 hpi with *B. cinerea*. Data presented in **(B)** are the means ± SD from three independent experiments and asterisks above each bar indicate significant differences (*P* ≤ 0.05) as determined by Tuckey analysis. **(C)** Maximum PSII quantum yield (*F*~v~/*F*~m~) and **(D)** relative photosynthetic electron transport \[ETR\]. Results represented the means ± SD from three independent experiments.](fpls-07-01236-g006){#F6}
Modifications of Soluble Sugar and Starch Contents
--------------------------------------------------
Sugars are the final products of photosynthesis and have been reported to be involved as a signal in plant defense mechanisms ([@B46]). Herein, gene expression of α and β-amylase, a *CwINV*, a sucrose synthase (*SUSY*), three hexoses (*HT1, HT3, HT5*), one putative polyol/monosaccharide transporters (*PMT5*), and two hexokinases (*HXK1, HXK3*) were analyzed. Analysis of gene expression showed that β-amylase expression was slightlty repressed while expression of *PMT5* was induced by *B. cinerea* (**Figure [7](#F7){ref-type="fig"}**). The expression of the remaining genes was not affected. When lonely inoculated, *B. phytofirmans* PsJN repressed the expression of β-amylase while *PMT5* was induced. When bacterized-plantlets were infected with *B. cinerea* infection, the expression of β-amylase was significantly repressed. Moreover, significant increases were found at transcriptional levels for α-amylase, *CwINV* and *HT5* upon challenge with *B. cinerea* compared to non-bacterized ones.
{#F7}
Sucrose, glucose, and fructose contents in leaves were measured at 0, 24, 48, and 72 hpi in bacterized and non-bacterized plantlets upon challenge with *B. cinerea*. While sucrose, glucose, and fructose remain constant during kinetics in non-bacterized plantlets, a moderate decrease in sucrose content was observed at 72 h in bacterized plantlets, supporting that *B. phytofirmans* PsJN might consume this sugar (**Figure [8](#F8){ref-type="fig"}**). In the same time, glucose content increased at 48 h whereas no significant fluctuation was occurred for fructose. In response to *B. cinerea*, the non-bacterized plantlets exhibit a constant decrease in sucrose content during the infection process in parallel with an increase in glucose and fructose contents but with a more pronounced effect for the latter (**Figure [8](#F8){ref-type="fig"}**). When bacterized plantlets were infected with the fungus, sucrose content increased transiently at 24 hpi, then decreased during the infection process. In meanwhile, the level of glucose increased to reach the maximum level at 48 hpi then slightly decreased. Nevertheless, no significant change was observed for fructose content (**Figure [8](#F8){ref-type="fig"}**).
{#F8}
Since starch constitutes the main carbohydrate reserve of plants, the starch content was monitored in leaves after *B. cinerea* challenge in bacterized and non-bacterized plantlets (**Figure [9A](#F9){ref-type="fig"}**). Twenty-four hours after *B. cinerea* infection, the starch content was enhanced in leaves of bacterized-plantlets, while a decrease was observed in response to *Botrytis*. Interestingly, after pathogen challenge, starch level was primed in bacterized-plantlets. These results were confirmed by TEM observations (**Figure [9B](#F9){ref-type="fig"}**).
{#F9}
Discussion
==========
*B. phytofirmans* PsJN Reduced *B. cinerea* Growth Development
--------------------------------------------------------------
During the interaction between grapevine plants and *B. phytofirmans* PsJN, the bacterium is able to colonize and diffuse inside plant tissues through xylem vessels ending at stomata chamber of leaves ([@B12]). This study provided new insights in deciphering the mechanisms of *B. phytofirmans* PsJN-IR against *B. cinerea* in grapevine by reporting for the first time that following root inoculation, *B. phytofirmans* PsJN is able to colonize the entire plant before exiting through the leaf stomata, and then forms a biofilm, at leaves surface, around mycelium of *B. cinerea*. This result indicates that the bacterium behaves firstly as endophyte, then as an epiphyte since it goes out to the leave surface to fight against the invader, suggesting the attractive chemotaxis of *B. phytofirmans* PsJN by the pathogen, which leads to an antibiosis effect of the bacterium. As far we know, this is the first time that such behavior was reported *in vivo* for a PGPR. Related to antibiosis impact, our results showed that *B. cinerea* development was inhibited dependently on *B. phytofirmans* PsJN concentrations. But, considering the endophytic level of PsJN in leaves tissues (Supplementary Figure [S2](#SM1){ref-type="supplementary-material"}), we might assume that the observed *B. phytofirmans* PsJN protection against *B. cinerea* could not be explained solely by its direct antifungal effect.
*B. phytofirmans* PsJN Potentiated Grapevine Defense Mechanisms
---------------------------------------------------------------
Microorganisms, for instance rhizobacteria, have been found to prime defense reactions against *B. cinerea* ([@B65], [@B64]). The H~2~O~2~ production, considered as a signal molecule for activating disease resistance ([@B41]), is an important part of grapevine defense system ([@B9]). After pathogen challenge, we observed an early H~2~O~2~ accumulation (8 hpi) only in bacterized plantlets, indicating a priming effect of the bacterium. These data are in agreement with the faster accumulation of ROS in potato plantlets inoculated with *B. phytofirmans* PsJN after *Phytophthora* infection reported by [@B30]. This early ROS production may explain partly the observed restriction of *Botrytis* in bacterized plantlets since [@B29] have reported the direct toxic effects of ROS on pathogens.
Several studies underlined the importance of stress-induced callose synthesis in defense mechanisms ([@B16]). Herein, a callose deposition was observed in stomatal guard cells of bacterized plantlets confirming earlier reports indicating that callose deposition is triggered by classical bacterial MAMPs, flg22 ([@B36]), EF-Tu ([@B35]), LPS ([@B54]), and peptidoglycans hairpins ([@B16]). Moreover, the callose deposition was primed in PsJN-bacterized plants suggesting the involvement of callose deposition in grapevine disease resistance toward *B. cinerea*, as reported previously against downy mildew ([@B61]).
Our data demonstrated that *B. phytofirmans* PsJN-induced resistance against *Botrytis* is mediated by enhanced expression of defense genes solely after pathogen inoculation. These results are in accordance with the previous study on abiotic stress, which showed that PsJN acts as a priming agent for defense responses in grapevine plantlets against low temperatures (4°C) rather than wastefully activating defenses ([@B56]). If SA signaling sector is generally associated with immunity to biotrophs while JA and ET are important for immunity to necrotrophs ([@B23]), there are plenty of exceptions to this rule ([@B19]; [@B47]). In this way, herein a concomitant enhancement of expression of both SA- (*PR1, PR2, PR5*, and *WRKY*) and JA-related genes (*AOC* and *JAZ)* has been observed in *B. phytofirmans* PsJN-bacterized plants after the challenge by *B. cinerea*.
*B. phytofirmans* PsJn Modulated the Carbohydrates Metabolism
-------------------------------------------------------------
During their co-evolution, plants and pathogens participate in a metamorphic tug-of-war, in which the plant limits pathogen access to nutrients and initiates immune responses, while pathogen develops adaptive approaches to redirect for their own nutrient flux and suppress plant immunity ([@B10]). The induced source-to-sink transition is not without consequences for photosynthesis and primary metabolism. In this way, several reports have postulated a relationship between sugar regulations, the expression of defense genes, and the activation of systemic resistance ([@B27]). Twenty hours post-infection, when the symptoms of *B. cinerea* infection did not appear yet, an increase of Y(II) was observed in bacterized plantlets, suggesting the protective role of bacteria on grapevine photosynthetic apparatus during the first step of infection. A significant decrease of effective quantum yield of PSII (Y(II)) accompanied by a quantum yield of regulated energy dissipation \[Y(NPQ)\] increase were observed in non-bacterized plantlets 48 h after inoculation with *B. cinerea*. These changes could result in lower efficiency of PSII photochemistry and increased heat dissipation ([@B28]). At 72 hpi, the increase in Y(NO) and the decline of *F*~v~/*F*~m~ induced by *B. cinerea* in non-bacterized plantlets indicated that the regulation mechanisms of non-photochemical dissipation of energy were blocked, making grapevine-plantlets unable to protect themselves against damage from excessed illumination. This oscillation was accompanied by a decrease in the electron transport flux, indicating that probably, *B. cinerea* damaged irreversibly the PSII and photosynthetic electron transport chain. In line with our results, several reports on photosynthesis and plant defense have indicated that photosynthesis rates are altered after infection with several plant pathogens ([@B7]; [@B4]). Our data indicated also that *B. phytofirmans* PsJN is able to prevent these irreversible damages probably by restricting mycelial development.
As indicated above, the decrease in photosynthetic metabolism, in parallel with an enhanced cellular demands during the resistance response, initiate the transition from source status to sink status in infected tissues. This transition is often accompanied by intensified gene expression and activity of invertases ([@B44]; [@B45]). CwINV regulates phloem unloading in some sink organs ([@B44]) and produces hexose substrates that may be acquired by hexose transporters (HTs) ([@B25]). In this context, our results exhibit that in response to *B. cinerea*, the up-regulation of *CwINV* and *HT5* was stronger in PsJN-bacterized plantlets compared to control ones, indicating a priming effect of the bacterium. Identical coordinated up-regulation of these two genes (*CwINV* and *HT5*) was previously described in grapevine leaves in response to both biotic (powdery and downy mildew) and abiotic (wound) stresses ([@B26]). These changes may reflect the higher demand for assimilates for defense reactions and the withdrawal of assimilates by the pathogen ([@B4]). A significant increase of sucrose content was observed only in bacterized-plantlets after pathogen challenge, which may be linked to the better photosynthetic capacity observed with *B. phytofirmans* PsJN. Indeed, sucrose is of central importance as a product of photosynthesis but also the form in which most carbohydrates are transported between cells and throughout the plant. Sucrose hydrolysis is catalyzed by invertases and the consequence is the shifts of the apoplastic sucrose/hexose ratio in favor of hexoses. A constant decrease of sucrose level was observed 24 hpi, whereas a slight increase of glucose level with a constant level of fructose were observed in PsJN-bacterized plantlets. It is well known that leaf-associated microbes use plant resources such as carbohydrates, amino acids and organic acids ([@B60]). Our results indicated that *B. phytofirmans* PsJN is able to use glucose and fructose, as described previously by ([@B49]). A clear accumulation of fructose was observed at 48 and 72 hpi in control plantlets infected by the fungus. The capacity of fungi to use soluble sugars to germinate is strain dependent ([@B15]). In this study, the effect of different concentrations of sucrose, glucose, and fructose on spore germination showed the inability of *Botrytis* strain 630 to use fructose (Supplementary Figure [S3](#SM1){ref-type="supplementary-material"}). After pathogen challenge, *B. phytofirmans* PsJN was able to induce *HT5*, which is the only HT able to bind fructose ([@B66]; [@B25]). Interestingly, these data suggest that *B. phytofirmans* PsJN redirects carbohydrates in favor to fructose, that is not useful for *B. cinerea*.
Gene expression analysis revealed a down-regulation of β-amylase and an up-regulation of α-amylase in response to *B. cinerea* in bacterized plantlets. A similar pattern was reported in grapevine leaves infected by *Plasmopara viticola* ([@B20]). The latters argued that α-amylase replaces β-amylase for starch degradation in *Plasmopara*-infected leaves. Starch is the main carbon reserve in plants and may be converted to soluble sugars. In our study, a significant increase of starch content was observed in *B. phytofirmans* PsJN-bacterized plantlets compared to control ones, confirming previous results obtained during grapevine-*B. phytofirmans* interaction ([@B3]). Herein, we showed for the first time a priming effect of *B. phytofirmans* PsJN on starch accumulation after *Botrytis* challenge, indicating that PsJN may help plants to respond more effectively and more rapidly to fungal attack by a better sugar mobilization.
Author Contributions
====================
LM-V, CJ, EB, and LS designed the research. LM-V, CJ, BC, LW, JM, EB, and LS carried out the experiments and analysis/interpretation of data. LM-V, CJ, CC, EB, and LS wrote the manuscript with contributions and discussion from all of the co-authors. All authors have given approval to the final version of the manuscript.
Conflict of Interest Statement
==============================
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
This work was supported by the Brazilian National Council for Scientific and Technological Development (CNPq) under research grants N. 246725/2012-5.
Supplementary Material
======================
The Supplementary Material for this article can be found online at: <http://journal.frontiersin.org/article/10.3389/fpls.2016.01236>
######
Click here for additional data file.
[^1]: Edited by: *Laure Weisskopf, University of Applied Sciences Western Switzerland, Switzerland*
[^2]: Reviewed by: *Birgit Mitter, Austrian Institute of Technology, Austria; Gerardo Puopolo, Fondazione Edmund Mach, Italy*
[^3]: ^†^*These authors have contributed equally to this work.*
[^4]: This article was submitted to Plant Biotic Interactions, a section of the journal Frontiers in Plant Science
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
A severely injured extremity poses difficult decisions for the patient and the treating surgeon. The degree of soft tissue injury, neurovascular damage, bone loss, presence of other injuries, patients' general physical and psychological condition are all important factors in decision-making. The majority of the patients were previously treated by amputation. However, with advances in surgical methods of fracture stabilization, soft tissue reconstruction and microsurgical techniques, some mangled limbs can now be salvaged.
Adequate soft tissue reconstruction is of utmost importance. Various treatment options have been suggested for managing the associated bone loss. These include internal fixation with bone grafting, distraction osteogenesis through circular external fixators; primary shortening followed by staged lengthening and allografts. However, these are associated with very prolonged recovery and the functional outcomes are unpredictable. We report a case of a woman presenting with severely injured lower limb and bone loss which was managed using a custom-made endoprosthetic replacement for successful functional outcome.
Case report {#Sec2}
===========
A 52-year-old woman sustained a severe crushing injury to her left lower limb having been trapped between a wall and a rapidly reversing car. She was conscious, coherent and haemodynamically stable. Her other injuries included a deformed right ankle and some bruising over the right cheek. There was no other major system injury. She was resuscitated according to the advanced trauma life support protocol and the bleeding from the leg was attended to. After she was clinically stabilized, a secondary survey was conducted to asses the limb status. Extensive soft tissue injury with degloving of skin and muscle damage from mid-thigh to lower calf was noted.
There was a significant comminution of the lower end of femur with loose fracture fragments visible. Distal pulses were palpable. Neurological assessment was found to be difficult at this stage. The mangled extremity severity score (MESS) was calculated to be 6 (3 for high energy injury, 1 for transient hypotension and 2 for age) and was utilised to asses the suitability for limb salvage. There was no significant past medical history or medication and allergy history of note.
The wound was covered with sterile dressings and the limb was splinted. Antibiotic and tetanus prophylaxis were administered. Adequate analgesia was provided. The right ankle and foot were found to be swollen, bruised, but with intact skin and were extremely tender. Trauma series radiographs and limb radiographs confirmed a comminuted distal end left femur fracture with bone loss (Fig. [1](#Fig1){ref-type="fig"}), a right ankle bi-malleolar fracture and an undisplaced base of right first metatarsal fracture.Fig. 1Soft tissue wound depicting the stretched common peroneal nerve
An emergency debridement and stabilization of the femur fracture were obtained using a bridging Hoffman uniaxial external fixator (Fig. [2](#Fig2){ref-type="fig"}). The devitalized bone fragments were debrided. Operative findings included loose femur fragments with contamination from fragments of plastic material, partially intact lateral gastrocnemius, partially detached iliotibial band and an avulsed biceps. The knee extensor mechanism and the neurovascular bundle were found to be intact. The right ankle fracture was also internally fixed on the same day.Fig. 2Plain radiograph confirming a comminuted distal end fracture of left femur with bone loss
At this stage, the patient was referred to us due to extensive soft tissue defect and a 12 cm femur bone loss. The aim was to reconstruct these defects using an endoprosthetic replacement. Option of an amputation and limb preservation was discussed with the patient. The patient preferred limb salvage. The plastic surgeon was also involved in the decision-making process and the patient was informed of the treatment plan. Over the next few days, a relook debridement was done and was followed by repeat debridement. The vascular status of the limb was normal, but there was a foot drop on clinical assessment. The wound exploration revealed that the common peroneal nerve was found to be intact, but had been stretched. A cement spacer to allow for the bone loss and a lateral gastrocnemius flap were harvested with input from the plastic surgeons. Unfortunately, this got infected with bacillus species. The wound was debrided, an across knee intramedullary rod and cement spacer was used to achieve stability. The infection was controlled by intravenous vancomycin.
In view of the ongoing infection and soft tissue problems, it was necessary to pursue on a cautious note. The next few weeks involved multiple wound washouts in the operating theatre to control infection. The wound swabs grew *Acinetobacter Baumannii* sensitive to the antibiotic merapenem. About 5 weeks after the initial injury, the leg was re-debrided and the cement spacer was revised. A left latismus dorsi free-flap was used for soft tissue reconstruction. The wound was assessed and redressed a few times in the operating theatre subsequently. With continued antibiotic treatment for 4 months, the infection was under control and repeated aspirated from the knee were negative.
The leg wound had healed well. However, the knee had only a jog of movement. A distal femur endoprosthetic replacement was planned for about 11 months from the initial injury (Figs. [3](#Fig3){ref-type="fig"}, [4](#Fig4){ref-type="fig"}). A custom-made distal femoral endoprosthesis, the SMILES implant (Stanmore Modular Individualised Lower Extremity System), Stanmore Implants Worldwide, UK, was used for reconstruction. Quadricepsplasty was needed to achieve sufficient access and the range of movement on table was 0°--60°.Fig. 3Long leg radiographs for preoperative planning for the endoprosthetic replacementFig. 4Manufacturer's proof of the custom-made distal femoral endoprosthesis
The patient underwent staged physiotherapy and gradual rehabilitation. The range of knee movement was of some concern and a manipulation under anaesthesia was performed 6 months after the reconstructive arthroplasty.
Current status {#Sec3}
==============
The patient is now 3 years from the endoprosthetic reconstruction. She is independently mobile and able to drive a car without limitations. The knee flexion remained limited to 0°--30° only. There is no evidence of infection. The prosthesis remained stable and well cemented. The limb is now 4 cm short. Functional score according to the Musculoskeletal Tumor Society-International Symposium on Limb Salvage System \[[@CR1]\] was 70% (21 out of a possible 30 points) and the Toronto Extremity Salvage Score \[[@CR2]\] was 62% (93 out of a possible 150 points). Figure [5](#Fig5){ref-type="fig"} shows the latest radiographs at 3 years after surgery.Fig. 5Plain radiograph of the endoprosthetic replacement at 3 years follow-up
Discussion {#Sec4}
==========
Mangled limbs with severe bone and soft tissue loss present a clinical challenge in terms of reconstructive surgery, prolonged rehabilitation, physical and psychological demands on the patient, and with no guarantee of a successful outcome. A question of primary amputation or limb reconstruction is a difficult one. Various scoring systems are available to asses the suitability for limb salvage in such situations such as the MESS; the limb salvage index; the predictive salvage index; the nerve injury, ischaemia, soft tissue injury, skeletal injury, shock and the age of patient score; and the Hannover fracture scale-97. Although a recent prospective study questions the clinical usefulness of any of these scores \[[@CR3]\], the most commonly used in the MESS system. Helfet et al. \[[@CR4]\] have reported that an MESS score of greater than or equal to 7 had a 100% predictable value for amputation.
The extents of soft tissue damage and neurovascular status are the important local factors in making the correct initial decision. Hence, significant experience with managing such patients and input from other relevant specialists are essential from the onset.
The goal of limb salvage surgery in such extremity injuries is to provide adequate soft tissue cover, skeletal stabilization, restore adequate length and alignment and most importantly, result in a functioning limb. Although various treatment modalities have been advocated in the management of limb injuries with bone loss, the common element in all these are the multiple reconstructive procedures, long term to recovery and the less than satisfactory functional results. Acute shortening followed by progressive lengthening \[[@CR5], [@CR6]\], bone transport using Ilizarov technique \[[@CR7]\], allografts to replace bone defects \[[@CR8], [@CR9]\] and free vascularised bone grafts \[[@CR10]\] have all been advocated for the reconstruction of the defects. However, the complications are many and the success limited. The major advantage of cemented endoprosthesis is early stabilization and recovery of function.
We have reconstructed a mangled extremity with 12 cm bone loss in the femur using a custom-made distal femoral endoprosthetic replacement. To our knowledge, this has so far not been reported in literature. Traditionally, endoprosthetic replacements are used for managing skeletal and soft tissue defects following the tumour resection and have had good long-term results \[[@CR11]\]. However, there seem to be additional applications afforded by these systems. Failed internal fixations of fracture, severe fractures with bone defects, failed joint arthroplasty with insufficient bone stock are the non-tumour indications for endoprosthesis' \[[@CR12]\]. The advantages with endoprosthetic replacement are the ability to reconstruct massive skeletal and soft tissue defects, ready availability and are relatively inexpensive \[[@CR13]\]. The patients can recover rapidly and weight bear early \[[@CR14], [@CR15]\]. The major complications are infection, aseptic loosening, mechanical failure and fracture. Many of the complications can be readily addressed, provided the patient is regularly followed up. Also the surgeon needs to be fully acquainted with the prosthesis and be fully aware of the limitations and dangers of the procedure.
Endoprosthetic replacement for the management of trauma may have a limited, but a defined role in addressing skeletal defects. However, the principles of managing an acute open fracture with bone loss remain the same, whatever the final choice of reconstruction. The initial sequence of management is resuscitation, wound debridement and skeletal stabilization. Appropriate soft tissue cover must be obtained early and any infection be adequately controlled. Only then can secondary osseous reconstruction proceed. This process may take many months, especially if complicated by infection. The limb although having survived the initial trauma can become stiff and functionless. Endoprosthetic replacement with adequate soft tissue procedures can aim to provide a more functional limb that is stable, pain free and has satisfactory range of movements. All through the treatment plan, the patient needs to be fully informed and involved in the process. The treating surgeon and the patient must be able to evaluate the progress, the treatment and change the course of treatment when necessary.
Patients suffer from both physical and emotional consequences following the severe extremity trauma. The goal of limb salvage is to provide a satisfactory limb function at the end of the arduous treatment journey. Endoprosthetic replacements may have a limited role in managing selected patients with mangled extremity and can lead to a good functional outcome to these patients. We recommend that this type of treatment be performed only by specialist teams consisting of surgeons experienced with endoprosthetic surgery, plastic surgeons, specialist microbiologists, wound care nurses, physiotherapists and occupational therapists.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
The Ebola virus continues to re-emerge in lethal outbreaks, the most recent occurring in the Democratic Republic of the Congo, Africa in May 2018^[@CR1]^. As of December 11, 2018, the World Health Organization reported that the Ebola outbreak in the North Kivu and Ituri provinces of the DRC has included a total of 505 cases, with 457 confirmed and 48 probable, and has resulted in 296 deaths^[@CR2]^. Overall, to date there have been 34 Ebola virus disease outbreaks, 18 of which have involved Zaire ebolavirus since the initial emergence of this strain in 1976. This most recent outbreak, and the outbreaks occurring between 2014--2016, have refocused efforts of public health agencies such as the World Health Organization^[@CR2]^ on identifying approaches to reduce the spread of the disease from community to community and from nation to nation. Ebola virus disease is included in the World Health Organization's List of Blueprint Priority Diseases^[@CR3]^, a list of diseases for which: "... given their potential to cause a public health emergency and the absence of efficacious drugs and/or vaccines, there is an urgent need for accelerated research and development..."^[@CR3]^.
It is known that the Ebola virus may be transmitted by contact with infected corpses, infected environmental surfaces (fomites), and secretions and excretions from infected individuals^[@CR4]^. It has also been shown that fomites in the vicinity of infected patients may be contaminated with Ebola virus RNA^[@CR5]^. Environmental persistence of infectious EBOV Makona (EBOV/Mak) suspended in an organic soil load have been reported at eight days from experimentally contaminated surfaces^[@CR6]^. An important intervention approach might therefore involve the use of an effective virucidal agent for disinfecting surfaces and spills potentially contaminated with Ebola virus, thereby mitigating the risk of transmission to healthy individuals, including health-care providers.
There are relatively little suspension inactivation data for the Ebola virus itself. The efficacies of microbicides (disinfectants and antiseptics) for inactivation of Ebola virus typically been determined through evaluation of the inactivation, by such products, of appropriate surrogate viruses such as bacteriophages, enveloped viruses (animal coronaviruses, influenza viruses), or non-enveloped viruses such as caliciviruses or picornaviruses. The Ebola virus is a member of the *Filoviridae* family, and being an enveloped virus should be relatively susceptible to a variety of microbicidal inactivation approaches. In view of the lethality of the virus, however, the United States Centers for Disease Control and Prevention (CDC) has provided the following guidance^[@CR7]^: "... selection of a disinfection product with a higher potency than what is normally required for an enveloped virus is being recommended at this time. EPA-registered hospital disinfectants with label claims against non-enveloped viruses (noroviruses, rotavirus, adenovirus, poliovirus) are broadly antiviral and capable of inactivating both enveloped and non-enveloped viruses." Per the United States Environmental Protection Agency (EPA), in order to claim efficacy of a product for an emerging enveloped virus, the product needs to be approved by EPA for inactivating at least one large or one small non-enveloped virus^[@CR8]^.
Efficacy testing of microbicides through the study of inactivation of surrogate non-enveloped viruses theoretically should ensure their efficacy for inactivation of the Ebola virus. However testing conducted specifically with the high-risk pathogens including viruses is also needed, to provide assurance to critical facilities and personnel. In the present effort, efficacy studies were performed at the Canadian Science Centre for Human and Animal Health Biosafety Level 4 (BSL-4) facility.
In the present study, we evaluated the efficacy of a commercially available hygiene product, Dettol Antiseptic Liquid (DAL), for inactivating EBOV/Mak is suspension. Dettol is used in Europe, Africa, and Asia in homes and healthcare settings for various first aid antiseptic purposes, including wound cleansing^[@CR9]^. It is also used for personal hygiene purposes, and as a microbicide for decontaminating environmental surfaces, objects, and equipment. The microbicidal active ingredient in DAL is 4-chloro-3,5-dimethylphenol (chloroxylenol), also known by its non-systematic name para-chloro-meta-xylenol or PCMX^[@CR10]^. The concentration of the active in DAL is 4.8% (weight to volume) PCMX. Three concentrations each of DAL lots 15083E, 16004E, and 16005E were evaluated for inactivation of EBOV/Mak suspended in an organic soil load using the method specified in the ASTM 1052-11 international standard^[@CR11]^. This test method was developed by the American Society for Testing and Materials International (ASTM) to standardize the evaluation of virucidal activity of microbicidal products in suspension. Organic soil loads are added to the study design in order to better model viral inactivation by microbicides in relevant matrices such as human sputum or blood. Use of hard water as diluent was included in the study design to simulate water hardness in the field. Four contact times (0.5, 1, 5, and 10 min) were evaluated in triplicate for each of three independent lots of DAL. In addition to the methodology described in the ASTM standard, we also evaluated any residual infectious virus following exposure to DAL through inoculation of 500 µL of undiluted neutralized test sample into T-75 flasks of Vero E6 indicator cells. This was done to evaluate the possibility of virus being present at levels lower than the limit of detection of the tissue culture infectious dose~50~ (TCID~50~) assay performed in Vero E6 cells per the ASTM standard.
Results {#Sec2}
=======
Neutralization Effectiveness Evaluation {#Sec3}
---------------------------------------
During the evaluation of possible neutralizing agents, it was determined that 100% fetal calf serum (FCS) and 100% virus culture medium (VCM) failed to adequately terminate the viral inactivating effects of DAL. On the other hand, 1× Letheen broth in VCM (10× Letheen broth diluted 1:10 in VCM), added to the DAL dilutions prior to introduction of the EBOV/Mak in tripartite soil load^[@CR12],[@CR13]^, prevented inactivation of the virus. As shown in Supplemental Figs [S1](#MOESM1){ref-type="media"} and [S2](#MOESM1){ref-type="media"}, no statistically significant (*P* \< 0.05) differences in the viral titers obtained for the virus positive controls, the virus + DAL dilution + neutralizer, and virus + neutralizer conditions were observed when 1× Letheen broth was evaluated. The disinfectant neutralizing agent that was used in each of the inactivation efficacy studies described below was 1× Letheen broth.
Virucidal Efficacy Results {#Sec4}
--------------------------
Three different lots of DAL were evaluated at three dilutions of DAL each (1:10, 1:20, and 1:40 in hard water, corresponding to 0.48%, 0.24%, and 0.12% of PCMX active, respectively) in triplicate. Contact times of 0.5, 1, 5, and 10 min at ambient temperature were evaluated in a BSL-4 facility. An initial EBOV/Mak titer of 1.7 × 10^8^ log~10~ TCID~50~/mL in tripartite soil load was exposed to the various DAL dilutions and contact times using the methodology depicted in Fig. [1](#Fig1){ref-type="fig"}. The post-exposure/neutralization titers for the positive virus controls (virus alone) and the DAL test conditions were calculated. The log~10~ reduction values for each time point were calculated by subtracting the titers obtained for the DAL test conditions from the titers of the corresponding positive virus controls.Figure 1Schematic representation of the inactivation efficacy testing methodology employed. The entire procedure was performed once for each DAL lot.
For the 1:10 dilution of the three DAL lots at 0.5, 1, 5, and 10 min contact time, the mean ± standard deviation values for the log~10~ EBOV/Mak titers measured for the positive virus control condition (virus alone) and the post-exposure titers are displayed in Fig. [2](#Fig2){ref-type="fig"}. Complete inactivation (4.8 to 5.4 log~10~) of EBOV/Mak to the limit of detection of the assay (1.8 log~10~ TCID~50~/mL virus) was observed for all replicates and contact times.Figure 2Time kinetics for EBOV/Mak inactivation efficacy results for the 1:10 dilution of DAL lots 15083E, 16004E, and 16005E at ambient temperature. The values represent the mean ± standard deviation (n = three replicates, one for each DAL lot) of the log~10~ titer of the positive control (0 minutes contact time) and the post-neutralization samples (0.5, 1, 5, and 10 minutes contact time). Individual viral titers were calculated based on three replicate wells per dilution. The limit of detection (LOD) of the assay was 1.8 log~10~ TCID~50~/mL (shown in the plot as a blue line extending from y = 1.8 log~10~ TCID~50~/mL). This was due to the cytotoxicity of the DAL dilution to the Vero E6 cells.
For the 1:20 dilution of the three DAL lots at 0.5, 1, 5, and 10 min contact time, the mean ± standard deviation values for the log~10~ EBOV/Mak titers measured for the positive virus control condition (virus alone) and the post-exposure titers are displayed in Fig. [3](#Fig3){ref-type="fig"}. Virus was detected at the 0.5 min contact time for DAL lot 16004E and at the 1 min time for DAL lot 15083E. Complete inactivation (4.8 to 5.3 log~10~) of EBOV/Mak to the limit of detection of the assay (1.8 log~10~ TCID~50~/mL virus) was observed for all replicates following contact times of 5 min or greater.Figure 3Time kinetics for EBOV/Mak inactivation efficacy results for the 1:20 dilution of DAL lots 15083E, 16004E, and 16005E at ambient temperature. The values represent the mean ± standard deviation (n = three replicates, one for each DAL lot) of the log~10~ titer of the positive control (0 minutes contact time) and the post-neutralization samples (0.5, 1, 5, and 10 minutes contact time). Individual viral titers were calculated based on three replicate wells per dilution. The limit of detection (LOD) of the assay was 1.8 log~10~ TCID~50~/mL (shown in the plot as a blue line extending from y = 1.8 log~10~ TCID~50~/mL). This was due to the cytotoxicity of the DAL dilution to the Vero E6 cells.
For the 1:40 dilution of the three DAL lots at 0.5, 1, 5, and 10 min contact time, the mean ± standard deviation values for the log~10~ EBOV/Mak titers measured for the positive virus control condition (virus alone) and the post-exposure titers are displayed in Fig. [4](#Fig4){ref-type="fig"}. Surviving virus was detected at the 0.5-min contact time for each of the three DAL lots. Complete inactivation (4.8 to 5.3 log~10~) of EBOV/Mak to the limit of detection of the assay (1.8 log~10~ TCID~50~/mL virus) was observed for all replicates following contact times of 1 min or greater.Figure 4Time kinetics for EBOV/Mak inactivation efficacy results for the 1:40 dilution of DAL lots 15083E, 16004E, and 16005E at ambient temperature. The values represent the mean ± standard deviation (n = three replicates, one for each DAL lot) of the log~10~ titer of the positive control (0 minutes contact time) and the post-neutralization samples (0.5, 1, 5, and 10 minutes contact time). Individual viral titers were calculated based on three replicate wells per dilution. The limit of detection (LOD) of the assay was 1.8 log~10~ TCID~50~/mL (shown in the plot as a blue line extending from y = 1.8 log~10~ TCID~50~/mL). This was due to the cytotoxicity of the DAL dilution to the Vero E6 cells.
The possible presence of infectious EBOV/Mak below the limit of detection of the TCID~50~ assay (1.8 log~10~ TCID~50~/mL) was evaluated by inoculating T-75 flasks of Vero E6 cells with 500 µL of undiluted neutralized test sample after each contact time and then evaluating the cultures microscopically for viral cytopathic effects (CPE) after 14 days of incubation. The results of this flask safety testing for the 1:10, 1:20, and 1:40 DAL dilutions (n = 1 flask per condition and DAL lot) are displayed in Table [1](#Tab1){ref-type="table"}. Infectious EBOV/Mak was detected for the 1:10 dilution of DAL lot 16005E at the 0.5-min contact time. No infectious virus was detected for the 1:10 dilution of either DAL lot following contact times of 1 min or greater. The results of the flask safety testing for the 1:20 DAL dilution (n = 1 flask per condition and DAL lot) are also displayed in Table [1](#Tab1){ref-type="table"}. Infectious EBOV/Mak was detected for DAL lots 15083E and 16005E at the 0.5-min contact time and for DAL lot 16004E at the 1-min contact time. No infectious virus was detected for the 1:20 dilution of the three DAL lots following contact times of 5 min or greater.Table 1Flask safety test results for Dettol Antiseptic Liquid (DAL) lot 1 (16004E), lot 2 (15083E), and lot 3 (16005E).Test Condition (contact time)1:10 DAL1:20 DAL1:40 DALLot 1Lot 2Lot 3Lot 1Lot 2Lot 3Lot 1Lot 2Lot 3Negative control000000000Neutralizer + DAL000000000DAL (0.5 min)00**++**0**++++**DAL (1 min)0000**+**0000DAL (5 min)000000000DAL (10 min)000000000+, Viral CPE observed; 0, viral CPE not observed, or no cytotoxicity observed for N + DAL; N + DAL, neutralizer + DAL (cytotoxicity) control. One replicate flask per DAL lot and test condition was evaluated.
The results of the flask safety testing for the 1:40 DAL dilution (n = 1 flask per condition and DAL lot) (Table [1](#Tab1){ref-type="table"}) demonstrate that infectious virus was detected for each DAL lot at the 0.5-min contact time only. No infectious EBOV/Mak was detected for the 1:40 dilution of the three DAL lots following contact times of 1 min or greater.
Discussion {#Sec5}
==========
Due to the need for a BSL-4 containment facility for conduct of inactivation studies on the Ebola virus, there is little information on the efficacy of microbicidal products for suspension inactivation of this filovirus. Efficacy data for microbicides against small or large non-enveloped viruses have been used in support of efficacy claims against emerging enveloped viruses such as Ebola virus^[@CR4],[@CR8],[@CR14]^.
In view of the paucity of available information on suspension inactivation of Ebola virus and other hemorrhagic fever viruses, the CDC has recommended^[@CR15]^ the use of a 1:10 dilution of household bleach (0.5% \[5000 ppm\] final chlorine concentration) for disinfection of infected excreta and corpses. Reproducible inactivation of viruses by chlorine is dependent upon free chlorine concentration, which is dependent both upon the chlorine demand of the inactivation solution as well as the free chlorine concentration of the bleach solution being used. The efficacy of free chlorine for inactivation of EBOV/Mak in sterilized municipal wastewater was investigated by Bibby *et al*.^[@CR16]^. The inactivation resulting from 1 ppm chlorine at 20 °C was incomplete but substantial (3.5 log~10~ by 0.3 min contact time, but little additional activation was observed over contact times up to 60 min). Higher chlorine concentrations (5 and 10 ppm chlorine, corresponding to initial free chlorine concentrations of 0.5 and 1.1 ppm, respectively) resulted in complete inactivation of the EBOV/Mak (4.2 log~10~ reduction for 0.3 min contact time).
Bleach, while an effective microbicide, is not practical for use on intact or wounded human skin. In addition, as mentioned above, the free chlorine levels of commercial bleach products may vary and effect the final free chlorine of the final inactivation solutions. On the other hand, the halophenolic compound PCMX, which is the active microbicidal ingredient in DAL, is suitable for use for preoperative skin antisepsis^[@CR17]^ and for disinfecting surgical instruments^[@CR10],[@CR18]^. It is an active ingredient used in a number of household microbicidal products, including DAL. As with other phenolics, the mechanism of viral inactivation for PCMX is likely related to disruption of lipid envelopes^[@CR19]^. The virucidal effectiveness of PCMX formulations is therefore expected to be greatest for enveloped viruses.
It has been proposed that the detergent Triton-X 100 be used at 0.1% to decontaminate clinical specimens (e.g., blood serum) in order to reduce (albeit not completely) the titer of Ebola virus in such specimens^[@CR20]^. As is implicit in this guidance, the inactivation afforded by 0.1% Triton X-100 is not to be considered complete and, in fact, van Kampen *et al*.^[@CR21]^ have shown that the percentage of serum in a test sample determines the extent of inactivation. In a sample matrix consisting of 98% fetal bovine serum, there was no statistically significant log~10~ reduction of Ebola virus resulted from a 1-hour exposure to 0.1% Triton X-100. A similar result was obtained when 0.1% sodium dodecyl sulfate was used as the inactivating agent. In samples containing lower percentages (9.8%, 0.98%) of fetal bovine serum, the inactivation caused by these agents was \>3 log~10~, but still not complete. The inability of Triton X-100 at up to 1% to completely inactivate Ebola virus in blood serum has also been reported by Burton *et al*.^[@CR22]^ and Colavita *et al*.^[@CR23]^.
Efficacy for viral inactivation is typically expressed in terms of an expected log~10~ reduction value. For instance, the US Environmental Protection Agency (EPA) stated in its 2012 disinfectant product guidance^[@CR24]^ that "The product should demonstrate complete inactivation of the virus at all dilutions. If cytotoxicity is present, the virus control titer should be increased to demonstrate a ≥3 log~10~ reduction in viral titer beyond the cytotoxic level." For disinfectants that are non-cytotoxic to the cellular infectivity assays used for demonstrating efficacy, a 4-log~10~ reduction in viral titer is considered to be effective. These EPA requirements for demonstrating acceptable efficacy were recently modified in the 2018 revision^[@CR25]^. In the revised guidance, a valid test requires (1) that at least 4.8 log~10~ of infectivity per carrier be recovered from the dried virus control film; (2) that a ≥3 log~10~ reduction in titer must be demonstrated in the presence or absence of cytotoxicity; (3) if cytotoxicity is present, at least a 3 log~10~ reduction in titer must be demonstrated beyond the cytotoxic level; and (4) that the cell controls be negative for infectivity. In the revised guidance, therefore, an efficacious product does not need to demonstrate complete inactivation at all dilutions.
The significance of Ebola virucidal efficacy determined in terms of a log~10~ reduction value must, however, be considered against the backdrop of the mortality associated with the virus, the estimated infectious dose, and the expected viral load within a contaminated bodily fluid. The mortality associated with Ebola virus disease is high. For example, the overall case fatality rate for outbreaks occurring between 1976 and 2017 has been calculated as 67%^[@CR26]^. The infectious dose of Ebola virus is believed to be relatively low in humans (between 1 and 10 infectious units)^[@CR16],[@CR27]^. The median plasma viral load in confirmed Ebola virus disease patients during the 2014--2015 Sierra Leone outbreak was found to be 6.7 log~10~ genomic copies per mL (range: 5.5 to 7.8 genomic copies/mL)^[@CR28]^. The average peak blood serum titer in cases from the 2000--2001 Uganda outbreak with a fatal outcome was 3.4 × 10^9^ genomic copies per mL, corresponding to 3.4 × 10^5^ to 3.4 × 10^6^ infectious units per mL^[@CR29]^. Human-to human transmission of Ebola virus disease has been attributed to direct contact with infected blood (including reuse of contaminated needles) or other secretions or bodily fluids of infected persons, and to contact with infected corpses^[@CR30],[@CR31]^. For example, infected patients have been reported to produce large volumes (5 L or more) of watery diarrhea over a period of 7 days or longer^[@CR32]^. The point of this discussion is to stress that the Ebola virucidal efficacy needs to be as high as possible, and certainly should be greater than the 3 or 4 log~10~ reduction routinely expected of microbicidal products.
Efficacy and lot-to-lot consistency of commercially available DAL for inactivating Ebola virus in suspension in the presence of a tripartite organic load has been demonstrated in the present studies. Due to the lethality of the Ebola virus and the low infectious dose for humans noted above, the studies were performed to determine log~10~ reduction in titer, as typically is done in demonstrating virucidal efficacy for microbicidal products. In addition, however, studies were conducted in a manner intended to assure that no infectious virus remained following application of the microbicidal product. The latter studies, referred to here as flask safety testing, was based on the assumption that any residual infectious viral particles following exposure to DAL should be able to infect the susceptible Vero E6 cells in the inoculated T-75 flasks and to amplify over a 14-day observation period, causing detectable viral CPE in the flasks. The greatest assurance of complete inactivation of EBOV/Mak was demonstrated for contact times of 5 min or greater (DAL dilutions of 1:20 and 1:40) or 1 min or greater in the case of the 1:10 DAL dilution. Under these conditions, over 5 log~10~ reduction in titer of infectious Ebola virus in tripartite soil load was demonstrated in the quantitative TCID~50~ assay and no detectable virus was detected in the non-quantitative flask safety tests.
These results suggest that DAL used at a 1:10 dilution in water should afford complete and rapid (within 1 min contact time) inactivation of EBOV/Mak in suspensions at ambient temperature. The data from the TCID~50~ assay and the flask safety testing performed for this DAL dilution made in hard water indicate that over 7 log~10~ inactivation of EBOV/Mak was achieved under these conditions.
Materials and Methods {#Sec6}
=====================
Cell Line, Virus, and Medium {#Sec7}
----------------------------
African green monkey Vero E6 cells (ATCC CRL-1586) were maintained at 37 °C/5% CO~2~ in Dulbecco's modified Eagle medium (DMEM, HyClone) supplemented with 10% FCS (Gibco) and 10 units/mL penicillin/streptomycin (Gibco). Ebola virus (Makona C07 variant) (Ebola virus/H. sapiens-tc/GIN/2014/Makona-C05) was obtained from a clinical isolate. All manipulations involving EBOV/Mak were carried out in a BSL-4 laboratory at the Canadian Science Centre for Human and Animal Health, Winnipeg, Manitoba, Canada.
Stock Virus Preparation {#Sec8}
-----------------------
A characterized stock of EBOV/Mak was prepared by infecting ten T-75 flasks of African green monkey Vero E6 cells (ATCC CRL-1586) at \~80% confluency at a multiplicity of infection (MOI) of 0.01. At approximately 9 days post-infection, marked cytopathic effects were evident in the Vero E6 cells, at which time the flasks were frozen at −70 °C. The flasks were thawed the following day and the conditioned medium from the flasks was clarified by low-speed centrifugation (4500 × g) for 10 min. The supernatants were pooled and overlaid onto 20% w/v sucrose cushions prepared in Tris- NaCl-EDTA buffer. After centrifugation at \~130000 × g for 2 h, the resulting viral pellets were resuspended in VCM (DMEM containing 2% FCS and 10 units/mL penicillin/streptomycin) overnight at 4 °C. The resuspended virus was pooled and aliquoted into usable amounts and frozen at −70 °C until needed. Stock virus titers were determined to be \>9 log~10~/mL by TCID~50~ assay, with titer calculation following the Reed-Muench method^[@CR33]^.
Dettol Antiseptic Liquid Product Dilution {#Sec9}
-----------------------------------------
Dilutions of DAL lot\# 15083E, 16004E, and 16005E were prepared in hard water^[@CR12]^ (prepared as 1 L deionized water supplemented with 0.4 g calcium carbonate) on the day of assay performance. The resulting solutions were inverted to mix and used within 2.5 to 4 h of preparation.
Assessment of Product Neutralization {#Sec10}
------------------------------------
A neutralization assay was performed to evaluate the ability of candidate neutralizing reagents to neutralize the virucidal effects of DAL post-exposure to the neutralizers. The reagents evaluated included 100% FCS, VCM, 1× Letheen broth in VCM (10× Letheen broth, BD Difco, diluted 1:10 in VCM). Ebola virus (Makona C07 variant) was diluted to approximately 10^4^ to 10^6^ TCID~50~ per mL with 10 µL virus evaluated per control in replicates of three. The candidate reagents were evaluated for neutralization efficacy (by scoring wells for CPE) and cytotoxicity to Vero E6 cells (by microscopic evaluation of monolayers in wells for morphology and confluency of Vero E6 cells) using the following conditions:
### Negative Control {#Sec11}
Cells were cultured in VCM and used as a baseline of comparison for evaluation of cytotoxicity or CPE.
### Neutralizer Control {#Sec12}
The candidate neutralizers to be evaluated were diluted in VCM using a 10-fold serial scheme from 10^0^ (undiluted) to 10^−3^. Neutralizer dilutions were added (50 µL) to Vero E6 cells (n = 5 replicates per dilution). Cells were scored for cytotoxicity 14 days post-inoculation.
### Positive Control {#Sec13}
The virus control was prepared by adding 10 µL of EBOV/Mak in a tripartite soil load^[@CR12],[@CR13]^ (10^2^ to 10^4^ TCID~50~ virus; 0.25% bovine serum albumin \[BSA, Sigma\], 0.35% tryptone \[Becton Dickinson\], 0.08% bovine mucin \[Sigma\]) to 990 µL of VCM. Final concentrations in the control were: virus (10^2^ to 10^4^ TCID~50~/mL), BSA (0.0025%), tryptone (0.0035%), and mucin (0.0008%). The positive control were diluted in VCM using a ten-fold dilution scheme from 10^0^ (undiluted) to 10^−3^, and 50 µL of the resulting solutions were added to Vero E6 cells. Cells were scored for CPE 14 days post-inoculation.
### Neutralizer + Virus Controls {#Sec14}
To account for the effect of the neutralizer acting on the virus, neutralizer + virus controls were prepared. These were prepared exactly the same way as the positive virus control, except that the 10 µL of virus in tripartite soil load was added to 990 µL of candidate neutralizer instead of VCM. The neutralizer + virus controls were diluted in VCM using a ten-fold dilution scheme from 10^0^ (undiluted) to 10^−3^, and 50 µL of the resulting solutions were added to Vero E6 cells. Cells were scored for cytotoxicity and CPE 14 days post-inoculation.
### Neutralizer + Disinfectant + Virus Controls {#Sec15}
To demonstrate the efficacy of the candidates for neutralizing viral inactivation by DAL, 50 µL of diluted DAL were added to 940 µL of the candidate neutralizers. Shortly thereafter, 10 µL of virus in tripartite soil load (concentrations of components given above) were added and the resulting mixtures (neutralizer + disinfectant + virus or controls), were diluted in VCM using a ten-fold dilution scheme from 10^0^ (undiluted) to 10^−3^, and 50 µL of the resulting solutions were added to Vero E6 cells. Cells were scored for cytotoxicity and CPE 14 days post-inoculation.
DAL Efficacy Testing {#Sec16}
--------------------
Inactivation efficacy testing for DAL was performed in suspension studies (Fig. [1](#Fig1){ref-type="fig"}) conducted at ambient temperature per ASTM E1052-11^[@CR11]^. Stocks of EBOV/Mak in tripartite soil load were prepared on the day of assay as follows. A single tube of stock virus was removed from frozen storage and mixed with a tripartite soil load^[@CR12],[@CR13]^ (\~1.7 × 10^7^ TCID~50~ virus, 0.25% bovine serum albumin, 0.35% tryptone, 0.08% mucin). The virus in tripartite soil load (10 µL) was added to 90 µL of diluted DAL or to 90 µL VCM (positive virus control). The resulting solutions were mixed and held at room temperature for 30 seconds, 1 min, 5 min and 10 min contact times. At the end of each exposure time point, the DAL was neutralized by adding 900 µL of prepared 1× Letheen broth to the test solutions and pipetted repeatedly to mix. A 500-µL portion of each neutralized test solution was diluted in VCM using a ten-fold dilution scheme, and 50 µL of the resulting dilutions were added to 96-well plates containing Vero E6 cells monolayers (n = 5 replicates per dilution). After a 45-min adsorption period, 150 µL of VCM were added to each well. The Vero E6 cell monolayers in the wells were scored 14 days post-infection for CPE and virus titers (TCID~50~) were calculated according to the Reed-Muench method^[@CR33]^.
Since the undiluted post-neutralization DAL displayed cytotoxic effects in the Vero E6 cells and therefore could not be evaluated for viral inactivation, an alternative method (flask safety test) was employed to evaluate the undiluted test solutions for the presence of any remaining infectious EBOV/Mak. In this test, 500 µL aliquots of undiluted neutralized inactivation test solution were added to single T-75 flasks (n = 1 replicate per DAL lot and neutralization time point) of Vero E6 cells monolayers at \~80% confluency, along with 20 mL of VCM. The flasks were incubated at 37 °C/5% CO~2~ for 14 days and then scored for presence or absence of CPE.
In addition to the efficacy test conditions described above, a single cytotoxicity control was performed in parallel with each assay. This was done to control for observable CPE or cytotoxicity caused by the neutralizer and DAL dilution being tested. This involved combining 900 µL of 1× Letheen broth and 100 µL of diluted DAL. A 500-µL portion of the resulting solution was diluted in VCM using a ten-fold dilution scheme, and 50 µL of the resulting dilutions were added to 96-well plates containing Vero E6 cells (n = 5 replicates per dilution). The Vero E6 cell wells were scored for cytotoxicity 14 days post-inoculation.
Analysis of Inactivation Efficacy {#Sec17}
---------------------------------
To score the test plates post-incubation, TCID~50~ titers for positive virus controls and neutralized DAL test conditions were determined by the method of Reed and Muench^[@CR33]^. The log~10~ reduction values achieved by the various DAL lots and exposure time points were calculated by subtracting the post-virucidal efficacy test of DAL, log~10~ TCID~50~ values from the log~10~ titers obtained for the corresponding positive virus controls. Statistical comparison of the mean (n = 5 replicates) viral titers obtained in the neutralization effectiveness studies (Supplemental Fig. [S1](#MOESM1){ref-type="media"}) was performed using a non-parametric unpaired *t*-test, with statistical significance set at *P* \< 0.05.
Supplementary information
=========================
{#Sec18}
Supplemental information
**Publisher's note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
=========================
**Supplementary information** accompanies this paper at 10.1038/s41598-019-42386-5.
This work was funded by RB, through a Collaborative Research Agreement with the Public Health Agency of Canada.
T.A.C., S.S.T., J.R.R. and M.K.I. designed and approved the project and experimental design; T.A.C. performed the experiments in the BSL-4 facility and aided in assembling the experimental data; R.W.N., T.A.C., M.K.I. and S.S.T. contributed to data analysis and interpretation, preparation of the figures, and to authorship of the manuscript. The Canadian Science Centre for Human and Animal Health and RB provided funding and test products for the project.
Competing Interests {#FPar1}
===================
J.R. Rubino and M.K. Ijaz are engaged in R&D work at RB, which provided funding for the experimental work described herein. The other authors have no financial interest in RB or the DAL product under investigation. The authors T.A.C., S.S.T., J.R.R., R.W.N. and M.K.I. declare no financial or non-financial conflicts of interest in this work.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Epilepsy is a demanding neurological condition that affects many people worldwide. Selecting an appropriate antiepileptic drug (AED) is still challenging, because the selected drug should be effective, safe, and tolerable. Older generation of AEDs, such as phenobarbital and phenytoin are not widely accepted as a primary monotherapy and also long-term therapy for focal seizures, because of their side effects ([@B1]).
This problem is more common in pediatrics, particularly those over the age of one year. Only topiramate and oxcarbazepine are approved as monotherapy despite their side effects, such as leukopenia, aplastic anemia and drug-induced hepatitis. Because these drugs have the potential for drug interactions, reducing the serum level of other AEDs, and producing drug - drug interaction, it is important to consider the safety and efficacy of an AED separately for monotherapy and adjunctive therapy([@B2]).
The newer generation of AEDs have often more favorable side effects, including lesser somnolence and blood dyscrasia than the traditional AEDs. However, no comparative study has demonstrated the improved efficacy over carbamazepine (CBZ), phenytoin or valproic acid ([@B3]).
Levetiracetam (LEV), the *S*-enantiomer of alpha-ethyl-2-oxo-1-pyrrolidine acetamide, is a novel AED that has been approved for use as an add-on therapy for partial-onset seizures in children older than one year. In addition, LEV may provide effective seizure control when used as monotherapy ([@B4]). No serious toxicity has been reported for LEV ([@B5]). LEV does not affect the liver enzymes, like CYP450. Hence, there is no report on its major interaction with other AEDs ([@B2]).
Little evidence is available for LEV monotherapy in children younger than 16 years ([@B6](. Although several other studies have demonstrated successful conversion to monotherapy in a small number of children, the response rate with various durations of treatment in children with refractory epilepsy was as high as 66% ([@B7],[@B8]).
To date, there are limited comparative findings regarding older and newer generations of AEDs ([@B9]) and there is no prospective study for this comparison.
This study aimed at comparing the effects of LEV and CBZ as monotherapy in children with focal seizures.
Materials & Methods
===================
This Single-blind, randomized, prospective study (data recorder was blind to the drug administration) was conducted among 50 newly diagnosed children having focal epilepsy and referring to the Quaem Hospital Pediatric Neurology Ward, Mashhad, Iran from May 2013 to March 2014. The age range of patients was 1-16 years.
Research protocol was approved by the Ethics Committee of the Mashad University of Medical Sciences (t-3181) and the written informed consent was obtained from the parents of the subjects.
***Patients.***The age range of 1-16 years, newly diagnosed focal epilepsy, no history of refractory seizures, the lack of other systemic underlying disorders, especially renal, hepatic, or brain diseases, such as cerebral palsy and no history of previous AED use were the inclusion criteria. Those with pseudo- seizures, drug reaction and major side effects, such as Stevens-Johnson syndrome, drug-induced hepatitis, psychosis, renal disorders, severe agitation or any other minor problems, the lack of parents' willingness to participate in the study and clinical or electroencephalographic findings suggestive of idiopathic generalized epilepsy were the exclusion criteria.
***Study design.***The study participants were randomly assigned to the two treatment groups receiving either LEV or CBZ. LEV (Levebel) was initiated at an initial dose of 10mg/kg/d and increased by 10mg/kg weekly until it reached the usual dose of 30mg/kg/d and continued. In the other group, CBZ (Loqman) was initiated at an initial dose of 5mg/kg/d and increased by 5 mg/kg weekly until it reached the usual dose of 15mg/kg/d and then continued.
At first, all participants were subjected to electroencephalography (EEG). To evaluate hepatic and hematologic side effects, complete blood count (CBC), alkaline phosphatase (ALKP) and aminotransferases (AST and ALT) tests were done one month later. Participants were assessed for side effects, such as somnolence, agitation, urticaria or skin itching. The patients then were divided into two 25-member groups, one group received LEV and the other one received CBZ. These groups were then sub-divided into two groups: responsive and non-responsive to therapy. Patients who completed the trial were considered to receive the allocated treatment until data analysis.
***Statistics***
Due to the lack of relevant study, this study was done as a pilot research. The subjects were initially dichotomized into two groups: those who treated with LEV and those with CBZ. The Student *t-*test and Chi-square test were used to compare continuous parametric and nonparametric data, respectively. The Fisher\'s exact test was used for categorical variables. The seizure-free period was calculated for each subject. The occurrence of adverse events was compared between the two treatment groups using the dosage at the onset of the adverse events and the interval between the initiation of the AED administration and the occurrence of the adverse events.
Results
=======
The initial evaluation sowed 25 patients with seizures who were younger than 16 years treated with LEV and 25 cases treated with CBZ who met inclusion criteria. The demographic characteristics of the two treatment groups were comparable and all patients were followed for 6 months after the initiation of monotherapy ([Table 1](#T1){ref-type="table"}). Two participants receiving LEV were excluded from the study because they developed severe agitation. The final analysis was done on 48 participants. No other case was excluded to follow-up or discontinued taking the medication. There was no need to add adjunct AEDs during the follow-up period.
The mean age of the participants was 7.32±3 years in the CBZ group and 7.89±2.5 years in the LEV group. Based on the Independent sample *t*-test, there was no significant difference in terms of age between the groups (*P*value: 0.516). Twenty-three participants (47.9%) had normal EEG (12 participants in the CBZ group and 11 in the LEV group), whereas 25 participants (52.1%) had abnormal EEG (13 patients in the CBZ group and 12 patients in the LEV group). Chi-square test revealed no significant difference in the frequency of participants with normal and abnormal EEG between the two groups (*P* value: 0.990(. In the CBZ group, 10 participants (40%) (or 20.8% of the total participants) did not respond to the therapy and had one or more seizures during the follow-up period. In the LEV group, only three (13%) participants (6.3% of the total participants) did not respond to the therapy. Fifteen (60%) participants receiving CBZ and 20 participants (87%) receiving LEV responded to the therapy.
There was no significant difference between the participants who were free of seizure attacks during a six-month follow-up and those who had seizure attacks. Regardless of the seizure type, Chi-squared test revealed a statistically significant difference in the response to the therapy between the CBZ and LEV groups (*P* value: 0.035). The participants on LEV monotherapy had a significantly higher response rate. Moreover, in the LEV group, there was no significant difference between the participants who were free of seizure attacks through a six-month follow-up and those who were not.
In the CBZ monotherapy group, five participants (20%) had a complex partial seizure and 20 subjects (80%) had secondary generalized seizures. Moreover, in the LEV monotherapy group, 5 participants (21.7%) had complex partial seizures and 18 subjects (78.3%) had secondary generalized seizures during the follow-up period. Chi-square test revealed no statistically significant difference in the frequency and type of seizure between the two groups (*P* value: 0.882).
Totally, 16 subjects (32%) out of the 50 participants \[9 (36%) on CBZ and 7 (28%) on the LEV\] experienced at least one adverse event and none of the adverse events were life-threatening. Of the total participants, 68% subjects (34/50) did not show any adverse events \[16 cases (64%) in the CBZ group and 18 cases (72%) in the LEV group\]. in addition, the Chi-squared test revealed no statistically significant difference in the occurrence of complications between the two groups (*P* value 0.853).
Six (12.5%) participants in the CBZ group reported somnolence and impaired consciousness; however, no somnolence sign was reported in the LEV group. The Chi-squared test results showed a statistically significant difference between the two groups (*P* value: 0.012).
There were no reported agitation signs in the CBZ group, whereas 7 cases (30.4% of the total) in the LEV group reported agitation signs. Considering those who were excluded, 7 out of the 25 participants (36%) in the LEV group had agitation signs. The Chi-squared test results revealed a statistically significant difference in agitation signs between the two treatment groups (*P* value: 0.003).
On the other hand, dermatologic and hepatic complications were reported only in the CBZ group. However, there were no statistically significant differences in the occurrence of these side effects between the two groups.
Only one participant in the CBZ group developed a hepatic complication. Liver enzymes were three times more than the upper limit. The liver function test repeated one week later revealed that the level of the enzyme had returned back to the normal range, thus, the treatment was continued. No hepatic complication and skin reactions were seen or reported in the LEV group. Only two participants in the CBZ group developed skin complications (itching and redness) who were treated with anti-histaminic medications. There was no hematologic complication among the participants in both groups. Furthermore, data analysis showed no statistically significant in all complications between the two groups.
Discussion
==========
It is difficult to design trials on the AED monotherapy to demonstrate its use in clinical practices in children with focal seizure ([@B10]). LEV is a novel medication that has been approved by the Food and Drugs Administration (FDA) for adjunctive therapy in children older than one month with focal seizure ([@B11]).
Few studies have specifically examined LEV monotherapy in children with newly diagnosed focal epilepsy, whereas many studies have suggested the efficacy of LEV as add-on therapy in adults and children ([@B4],[@B6]).
This was the first randomized clinical trial that compared the efficacy of LEV with CBZ monotherapy in children. In this study, we investigated the efficacy of LEV compared with CBZ regarding the newly diagnosed focal epilepsy. The main goal of the study was the elimination of the seizure for at least 6 months after treatment initiation and continuing the therapy with LEV or CBZ, which can indicate the efficacy of the medication. The standard formulation of the CBZ and LEV were used with a fixed starting dose, slow titration, and the possibility for patients to remain on the modest effective dose.
The primary outcome of this study was seizure remission for at least 6 months after the initiation of therapy with LEV or CBZ. More than 70% of the participants responded positively to these monotherapies and had no seizure through a six-month follow-up, which is consistent with the results of Perry et al. (2008) study ([@B11]). Our results showed more effectiveness in participants receiving LEV with 87% reduction in seizure frequency than to those receiving CBZ (*P* value: 0.035). Seizure freedom may be a more reliable measure for a prospective review, as it is often more clearly documented in the chart and is less susceptible to the recall bias resulting from the reliance on seizure-frequency counts provided by the parents. In the present study, all participants were subjected to monotherapy for at least six months. The majority of participants (87%) receiving LEV, (20/23 participants) experienced seizure freedom through a six-month follow-up. This proportion was significantly higher than the participants in the CBZ group (60%) who were seizure-free during the follow-up period. A previous study on 18 children treated with LEV did not report information about the conversion, treatment initiation, duration of treatment, and clinical response ([@B12]).
Another relevant study reported that 73% of the participants receiving LEV and 65% treated with CBZ were free of seizures. However, the proportion of those treated with CBZ who were free of seizures was slightly higher than our findings ([@B11]). Another study reported no statistically significant difference in seizures freedom between participants treated with LEV and CBZ ([@B3]). In contrast, our findings revealed a statistically significant difference in seizure remission between the participants in the LEV and CBZ groups (*P* value: 0.035). Other studies also reported a higher proportion of seizure-free response among participants with partial seizure who received LEV monotherapy ([@B8],[@B9]). Furthermore, 57% of the participants (children younger than 4 years) who were treated with LEV or CBZ monotherapy for focal seizures have been reported to be free of seizures during the first six months of follow-up ([@B13]).
We prescribed LEV at an initial dose of 10 mg/kg/d that was increased to 30 mg/kg/d and continued. Another study also suggested a lower dose of LEV (≤30 mg/kg/d). Ben-Menachem et al. reported that changing the LEV therapy from adjunctive therapy to monotherapy at a dose of 1500mg twice a day was effective in reducing or ceasing the seizures attack ([@B15]).
Studies on LEV monotherapy in adults and children have reported the effectiveness at relatively lower doses of LEV ([@B12],[@B13]).
These studies also suggested LEV as an effective medication for seizure remission in childhood focal epilepsy. Among the patients treated with LEV ([@B12],[@B13],[@B15]), there were no significant differences in seizure type. Both CBZ and LEV were well tolerated as an initial monotherapy ([@B12],[@B13]).
Only 30% of the participants receiving LEV and 28% treated with CBZ had side effects. The findings showed no statistically significant difference in the occurrence of side effects in the studied groups. This may be due to the small sample size in our study. Besides, only two participants in the LEV group discontinued the monotherapy because of severe agitation. There were no other side effects leading to the discontinuation of therapy among the treatment groups. Similarly, another study reported that agitation was the most common side effect that caused treatment cessation ([@B15]).
In our study, all other side effects, such as hepatic dermatologic diseases and somnolence were only seen in those receiving CBZ. Somnolence was a more frequent complication in patients receiving CBZ that occurred at relatively normal daily doses. However, serum CBZ levels were not routinely available. As a result, we could not determine the effect of dosages on plasma concentrations in participants reporting these complications. Similarly, somnolence was the more frequent side effect among patients treated with CBZ ([@B11]). However, the side-effect profile of a particular formulation, such as the controlled-release formulation of CBZ, which may be better tolerated, was not determined in this study. A statistically significant difference in agitation side effect was also found between the two groups (*P* value: 0.012), which is consistent with the findings of other studies ([@B16]-[@B20]).
It is recommended to conduct more studies using a larger sample size to compare the presence of intolerable side effects more accurately between the two groups receiving the specified monotherapies. It can provide an appropriate finding regarding the effectiveness of a specific therapy more specifically . Our study indicated a significant difference in the effectiveness of the LEV as the first-line medication compared with the CBZ. This implies the noninferiority to CBZ and a more favorable effectiveness of the LEV as a monotherapy for focal seizures. However, data analysis showed no statistically significant difference between the two groups.
To select a treatment for a patient with newly diagnosed epilepsy, the side effects and long-term safety should be considered. LEV can probably be a proper substitution for CBZ because it is associated with lower side effects, more safety, tolerability, and simpler pharmacokinetics that makes it a promising AED to be used as initial monotherapy in newly diagnosed epilepsy cases. The current study showed that children with newly diagnosed focal epilepsy treated with LEV had a better positive prognosis than those receiving CBZ.
**Strengths and Limitations**
The use of seizures remission as an ultimate goal of treatment and as a marker of efficacy, the relative homogeneity between the groups, and using prospective design (randomized controlled trials) were the strengths of this study. However, this study was a pilot study using a small sample size. Hence, the direct comparison between the two groups with the antiepileptic monotherapies was limited. This study can provide strong findings for future trials on LEV monotherapy with a larger sample size in children younger than 16 years with focal epilepsy.
######
Characteristics and adverse Event Profile of the patients receiving Levetiracetam and Carbamazepine
--
--
**In conclusion:**
To select a treatment for a patient with newly diagnosed epilepsy, the side effects and long-term safety should be considered. LEV can probably be a proper substitution for CBZ because it is associated with lower side effects, more safety, tolerability, and simpler pharmacokinetics that makes it a promising AED to be used as initial monotherapy in newly diagnosed epilepsy cases. The current study showed that children with newly diagnosed focal epilepsy treated with LEV had a better positive prognosis than those receiving CBZ.
The use of seizures remission as an ultimate goal of treatment and as a marker of efficacy, the relative homogeneity between the groups, and using prospective design (randomized controlled trials) were the strengths of this study. However, this study was a pilot study using a small sample size. Hence, the direct comparison between the two groups with the antiepileptic monotherapies was limited. This study can provide strong findings for future trials on LEV monotherapy with a larger sample size in children younger than 16 years with focal epilepsy
This study was conducted in the Mashhad University of Medical Sciences. It was supported by Dr. Akhondian through an investigator-initiated grant and also the CobelDarou Company assisted us in the provision of the drugs.
(IRCT No is IRCT2015030821355N1)
Author Contribution:
====================
Hossein Eslamieh: Following-up with patients and writing the article
Farah Ashrafzadeh: Referring the patients to the research team.
Javad Akhondian: Referring the patients to the research team.
Conflict of interest
====================
The authors declare that there is no conflict of interest
| {
"pile_set_name": "PubMed Central"
} |
Spins misaligned to the magnetic field relax into the field direction by transferring their energy to the environment. The relaxation rate is governed by the magnetic damping constant (α) which reveals the interaction between the spins and the environment around the spins. Though α is a material-specific value, it can be actively controlled by the spin transfer torque[@b1][@b2]. The active reduction of α gives such interesting phenomena as magnetization reversal[@b3][@b4][@b5][@b6] and steady spin precession[@b7][@b8][@b9][@b10][@b11][@b12], which have given birth to the novel devices like the spin transfer torque magnetoresistive random access memory (STT-MRAM)[@b13][@b14] and the spin torque nano-oscillator (STNO)[@b15]. In general, these spintronic devices consist of two main spin layers called as free and pinned layer. Although each layer is coupled to the other by the dipolar interaction and the spin transfer torque, the spin dynamics driven by their coupling has not been considered seriously. Only recently, the spin dynamics coupled by the mutual spin transfer torque has been studied by micromagnetic calculation[@b16][@b17]. But the dipolar interaction was assumed to give only marginal effects because the dipolar field from the pinned layer is generally compensated by the antiferromagnetically coupled additional layer, i.e. synthetic antiferromagnet (SAF)[@b17][@b18][@b19]. Here in this report, however, we show that the dipolar coupling actually has a serious impact on the current driven spin dynamics. We observed that the current driven spin oscillation disappears at a specific magnetic field. The micromagnetic calculation and the numerical estimation of the eigenmodes on the model systems with different dipolar interactions show that the observed behavior originates from the dipolar coupling. As will be described later, the current driven magnetic damping of the dipolar-coupled spin system is closely related to the transfer of energy between the two layers. By changing the energy transfer rate with the external magnetic field, one can adjust the damping of the coupled spin system.
Results
=======
Anomalous discontinuity in spin oscillation mode observed in magnetic tunnel junction
-------------------------------------------------------------------------------------
The sample for our study is a typical magnetoresistive tunneling junction (MTJ), which consists of the bottom layer \[Si/SiO2/TiN(60)\], SAF \[PtMn(15)/Co~90~Fe~10~(1.5)/Ru(0.8)/Co~40~Fe~40~B~20~(1.5)\], tunneling barrier \[Mg(0.3)/MgO(0.53)\], magnetic free layer \[Co~20~Fe~60~B~20~(2.0)\], and capping layer\[Ta(5)/Ti(15)/Ru(15)\] as shown in [Fig. 1](#f1){ref-type="fig"}. Here the numbers in the parenthesis represent the nominal thickness in the unit of nanometer. In order to study the coupled motion between the free layer (FL) and the top pinned layer of the SAF (TPL), the thickness of the pinned layer was thin enough to be excited by the spin transfer torque. In previous studies that focus only on the dynamics of the FL spin, the thickness of the TPL was much thicker[@b20][@b21] or wider[@b22] than the FL, inhibiting the excitation of the TPL. The MTJ cell has a circular shape with a diameter of 90 nm, minimizing the in-plane shape anisotropy. The current driven spin precession yields a resistance oscillation in the circular MTJ cell, which has been measured under different external fields (*μ~0~H*). The polarity of *μ~0~H* was defined as positive when the magnetization of the FL is antiparallel with the TPL. The current bias polarity was defined as positive when the electron moves from the TPL to the FL.
[Figure 2a](#f2){ref-type="fig"} shows the resistance of the MTJ cell as a function of *μ~0~H* in the current density . Under the negative (positive) field, the spins in the FL are parallel (antiparallel) to the spins in the TPL, yielding the lower (higher) magnetoresistance level. The spins in the TPL begin to be tilted from the spin flop-field *μ~0~H~sf~* = 90 mT, where the external field becomes comparable to the exchange field between the TPL and bottom pinned layer (BPL) of the SAF. The alignments of the spins in the three layers are schematically shown in [Fig. 2a](#f2){ref-type="fig"} as arrow marks for the comparison with the resistance curve.
The color-coded power spectrums of the MTJ cell in the current biases and are shown in [Fig. 2b and 2c](#f2){ref-type="fig"}, respectively. In the field range 0 \< *μ~0~H* \< *μ~0~H~sf~*, one can observe distinctive dependence of the oscillation mode on the current bias polarity. In positive (negative) current bias, the spins in the FL (TPL) are dominantly excited, yielding a blue (red) frequency shift with the increasing field. Above the spin-flop transition *μ~0~H* \> *μ~0~H~sf~*, both the free and pinned layer show blue frequency shifts. These general features agree with the previous reports on the eigenmodes of the FL and TPL in the MTJ cell[@b23][@b24][@b25].
There are several interesting points in [Fig. 2b and 2c](#f2){ref-type="fig"}. A discontinuity is clearly observed in the oscillation mode. In [Fig. 2b](#f2){ref-type="fig"}, the free layer oscillation observed at low fields disappears at around *μ~0~H* = 40 mT, and reappears at *μ~0~H* \> 60 mT. One should note that the point by point discontinuity in the field range *μ~0~H* \< 30 mT is just due to the large interval of the sampling fields in our experiment. This is different from the real discontinuity of the oscillation mode.
One can also find a gap in the oscillation frequency at around the disconnected point. The extension of the lower field mode by the Kittel formula[@b26] does not matches well with the higher field mode. The oscillation mode of the TPL in [Fig. 2c](#f2){ref-type="fig"} also shows similar discontinuity and frequency gap, which implies that the observed feature is independent of the current polarity. The explanation of the observed anomalous breakdown of the oscillation mode is the main subject of this article.
Micromagnetic calculation : effect of dipolar coupling on spin oscillation
--------------------------------------------------------------------------
The magnetization dynamics of the coupled spin layers have been calculated using the Landau-Lifshits-Gilbert-Slonczewski-Slonczewski (LLGS) equation given by[@b3][@b27][@b28] where *γ*, *α*, *a~J~*, and ***p*** represents the gyromagnetic ratio, the intrinsic damping constant, the amplitude of the spin-torque in the unit of magnetic field, a unit vector parallel to the electron spin polarization. In order to study the effect of the dipolar coupling between the layers, the micromagnetic calculations have been carried out on several different separations (*s*) between the TPL and FL from 0.8 nm to 50 nm. Although electrons cannot tunnel through the oxide barriers in case of large *s* value in reality, we assumed the same tunneling current could pass through the MTJ in our calculations.
The color-coded microwave power calculated on the model system with *s* = 0.8 nm, 10 nm, 50 nm under the positive and negative current biases are depicted in [Fig. 3a--3c](#f3){ref-type="fig"} and in [Fig. 3d--3f](#f3){ref-type="fig"}, respectively. The case of *s* = 0.8 nm shows nearly the same oscillation feature with the experiment. There is a clear discontinuity and frequency gap in the oscillation mode at around 45 mT. As *s* increases, i.e., the dipolar interaction between the FL and TPL becomes weaker, the gap between the higher frequency mode and lower frequency mode decreases. At the separation of 50 nm, the oscillation mode is nearly connected with a negligible gap like the oscillation mode of a 'non-coupled' single spin system. This reveals that the observed breakdown and frequency gap in the oscillation mode at around 45 mT comes from the dipolar coupling between the FL and TPL.
Normal modes of dipolar coupled spin oscillation
------------------------------------------------
To deeply understand the coupled dynamics, we calculated the eigenmodes of the model system by solving the LLGS equation in a small excitation limit. The azimuthal and polar angles of the magnetization in each layer have been defined as and *θ~i~*, respectively. Here *i* represents the layer number of the BPL (*i* = 1), TPL (*i* = 2), FL (*i* = 3). The LLGS equation in Eq. 1 can be described according to the above six variables in the form A linearized form of the above equation around the stationary point ***X*~0~** = (*x*~10~, *x*~20~, *x*~30~, *x*~40~, *x*~50~, *x*~60~) following *F~i~*(***X*~0~**) = 0 can be written as Here **δX** = ***X***−***X*~0~**(*δx~i~* = *x~i~*−*x~i~*~0~) represents a small deviation from the stationary point. If the six eigenvalues and eigenvectors of the 6 × 6 matrix are respectively *ε~k~* and ***U~k~*** (*k* = 1\~6), i.e., , then **δX** in Eq. 3 can be written as Here *c~k~* is a numerical constant determined by the initial condition of the motion. The motion of each variable is the superposition of the six eigenmodes weighted by exp(*ε~k~t*). The imaginary part of the eigenvalue Im(*ε~k~*) corresponds to the angular frequency of the oscillatory eigenmode, and the real part Re(*ε~k~*) reveals the decay rate of the oscillation. Negative Re(*ε~k~*) implies a mode whose oscillation decays out even though it has an initial oscillation amplitude. In the positive Re(*ε~k~*), on the other hand, the spin oscillation is feasible to be excited[@b29]. In this sense, the value of Re(*ε~k~*) is the crucial factor that determines the damping rate of the spin oscillation.
The matrix has been numerically calculated under different external fields and current biases from the LLGS equation. It gives six eigenvalues, three of which are the complex conjugates of the others. The three real and imaginary part eigenvalues Re(*ε~k~*) and Im(*ε~k~*) are depicted as a function of the external field on the positive and negative current biases of *J* = ±3×10^7^ A/cm^2^ in [Fig. 4a--4d](#f4){ref-type="fig"}. The mode 1 (*k* = 1) which has the highest frequency corresponds to the excitation of the BPL, and the mode 2 (*k* = 2) and 3 (*k* = 3) are respectively the optical and acoustic modes between the FL and TPL. One can note that the optical mode has the higher frequency than the acoustic mode because the optical mode has the more dipolar energy. Each mode has been identified through the corresponding eigenvector as described in the [supplementary information A](#s1){ref-type="supplementary-material"}.
Considering that the spin oscillation is the more heavily excited in the mode where Re(*ε~k~*) is the larger, the eigenmode with the maximum Re(*ε~k~*) among the three modes corresponds to the dominant spin precession mode. The dominant precession mode is the acoustic (optical) mode at a lower (higher) field of *μ~0~H* \< 30 mT (*μ~0~H* \> 60 mT) on the case of positive current bias. On the negative current bias, it is reversed, i.e., the optical mode is dominant at the lower field and the acoustic mode is dominant at the higher field. Compared to the coupled motion of the FL and TPL, the motion of BPL is negligible. Although BPL is strongly coupled to the TPL by antiferromagnetic exchange coupling, the dipolar field is not fully compensated due to the different dynamic motions in the TPL and BPL. This reveals that the dipolar interaction should not be neglected in the dynamic case.
The eigenvalue of such a dominant precession mode under several different current densities is depicted as a function of the external field in [Fig. 5](#f5){ref-type="fig"}. The real parts (imaginary parts) on the positive and negative current biases are shown in [Fig. 5a and 5b](#f5){ref-type="fig"} (in [Fig. 5c and 5d](#f5){ref-type="fig"}), respectively. The real part representing the decay rate of the oscillation becomes smaller and goes lower than zero around the field 45 mT in both current polarities. This agrees with the experimental results that the steady state oscillation breaks down around 45 mT. The higher current density lifts up the Re(*ε*) values, which can be understood as the active reduction of the magnetic damping by the current. However, such a general behavior does not occur around the breakdown field, where the steady spin precession is inhibited no matter how high the current density is.
The imaginary part depicted in [Fig. 5c and 5d](#f5){ref-type="fig"} shows a frequency gap around the breakdown point of 45 mT. The dominant oscillation mode changes from the acoustic (optical) mode to the optical (acoustic) mode with the increasing field in the positive (negative) current bias as depicted in [Fig. 5c](#f5){ref-type="fig"} ([Fig. 5d](#f5){ref-type="fig"}).
In order to investigate the effect of the dipolar coupling strength, we estimated the Re(*ε*) and Im(*ε*) values of the dominant precession mode for several different *s* values. The results of Re(*ε*) as a function of external field are shown in [Fig. 6a](#f6){ref-type="fig"}. [Figure 6b](#f6){ref-type="fig"} shows the dependence of the frequency gap upon the *s* values, which was obtained from the results of Im(*ε*). The strong reduction of Re(*ε*) around the breakdown field observed at small *s* becomes weaker and weaker as *s* increases, resulting in a negligible dip of Re(*ε*) at *s* = 1000 nm. The case of *s* = 1000 nm can be considered as a system where each layer is coupled to the others only through the spin transfer torque without any dipolar interactions. In this case, we could not observe any special feature on Re(*ε*). This reveals that the anomalous magnetic damping observed around the breakdown field is strongly related to the dipolar coupling. The coupling only through the spin transfer torque cannot make such a behavior.
Recent micromagnetic calculations on a similar spin system have considered the dynamic coupling only by the spin transfer torque without any dipolar interaction[@b16][@b17]. Unlike our results, they showed several anomalous features such as the reduction of the spectral line-width and frequency jump exactly at the same breakdown field. But this behavior was observed only at a large current density *J*\>4.0×10^7^ A/cm^2^. The estimation of eigenvalues and eigenmodes assumes a spin motion with small amplitude around the stationary points. It cannot cover the dynamics of large amplitude. In this sense, the anomalous behaviors in purely spin transfer torque coupled system must come from the nonlinear dynamics in large amplitude. In fact, we carried out micromagnetic calculation at large current density, and it showed the same behaviors as the previous reports as described in the [supplementary information B](#s1){ref-type="supplementary-material"}. The dipolar-coupling affect the damping property even at small spin amplitudes, making the anomalous damping behavior be observed experimentally in moderate current range. In contrast, the anomalous effect by the dynamic coupling through the mutual spin transfer torque appears only at large spin amplitude.
Damping control through external magnetic field
-----------------------------------------------
According to the behavior of Re(*ε*) shown in [Fig. 5a and 5b](#f5){ref-type="fig"}, the damping of the coupled spin oscillation can be controlled by the external magnetic field. The real time motions of the spins in the FL and TPL have been obtained from the numerical calculation of Eq. 1 under the following condition. Initially the model system was under the field of 27 mT far smaller than the breakdown point in the current bias of . At the time τ = 0 ns, the magnetic field suddenly jumped from 27 mT to the breakdown field. The azimuthal angle motions of the spins in the FL and TPL have been calculated in different *s* values of 0.8 nm and 20 nm. [Figure 7a](#f7){ref-type="fig"} shows the results obtained on the case of *s* = 0.8 nm. To compare the results with the eigenmodes that is valid only small spin oscillation limit, the results at a larger τ are separately depicted in [Fig. 7b](#f7){ref-type="fig"}. [Figure 7c and 7d](#f7){ref-type="fig"} show the results similarly obtained on the case of *s* = 20 nm.
As expected from the Re(*ε*) behavior shown in [Fig. 5](#f5){ref-type="fig"} and [Fig. 6](#f6){ref-type="fig"}, the steady state precession below 0 ns decays out above 0 ns. Comparing the decaying region (τ \> 0) with the steady precession region (τ \< 0), one can find that the amplitude ratio between the FL and TPL is not constant but changes rather periodically. Dashed blue lines marked on the points where the ratio of TPL amplitude over the FL amplitude has a local maximum show this clearly. This beating pattern becomes even clearer when the amplitude of spin oscillation is small enough at a larger τ, as depicted in [Fig. 6b and 6d](#f6){ref-type="fig"}. The period of beating becomes longer as the distance between the FL and TPL becomes larger as one can compare [Fig. 6d](#f6){ref-type="fig"} with [Fig. 6b](#f6){ref-type="fig"}. For the comparison of the beating frequency with the frequency gap of the eigenmodes, scale bars of 1/Δf~1~ and 1/Δf~2~ have been drawn in [Fig. 7](#f7){ref-type="fig"}, which represent the inverses of the frequency gaps Δf~1~ and Δf~2~ obtained respectively from the Im(ε) values on the two cases *s* = 0.8 nm and 20 nm in [Fig. 6b](#f6){ref-type="fig"}. One can find that the beating frequency identified by the dashed blue lines coincides with the frequency gap, i.e., the frequency difference between optical and acoustic mode at the breakdown field. This clearly reveals that the beating pattern comes from the interference of the two eigenmodes. Considering that the spin oscillation in the FL or TPL is expressed as a superposition of the acoustic and optical modes by Eq. 4, the observed beating pattern is not surprising. The beating pattern appears only near the breakdown field because the real part eigenvalues Re(*ε*) of the two competing modes are comparable only at this point. Away from this field, the difference of Re(ε) values is so large that only one eigenmode appears dominantly. In this sense, the increased damping at the breakdown point is closely related to the beating of the two eigenmodes.
Since the amplitude of the spin oscillation represents the energy stored in the spin motion, the beating pattern of the spin oscillation in each layer implies the energy transfer from the FL to TPL, and vice versa. In the field away from the breakdown point, both amplitudes of the FL and TPL remain the same without any energy transfer between the two layers. In the breakdown point, on the other hand, the FL and TPL change their amplitudes alternately, exchanging energies with each other.
Recent study on the two dipolar coupled magnetic disks has also shown similar energy transfer between the two disks[@b30]. The energy transfer in dipolar coupled system is in itself lossless and has no relation with the damping of the system. However, if the spin transfer torque is involved in the dipolar coupled spin layers like our sample, the energy transfer between the coupled layers affects the damping property. The current driven active damping on one layer is reversed on the other layer. If the FL is excited or anti-damped by the current bias, the TPL is the more effectively damped by the same current. Thus the transferred energy does not fully come back to the original layer due to the intrinsic and current driven active damping of the other layer. The energy transferred from one layer with a steady state spin precession supported by the current driven active anti-damping is more heavily dissipated by the active damping on the other layer. A steady spin precession is impossible if only the energy is transferred to the other layer. This is the case observed on the breakdown field. In spintronic devices like MRAM and STNO, the spin transfer torque is considered only to excite the spin state by the active anti-damping. But this is true only when the energy exchange between the FL and TPL is negligible. At the breakdown field, the FL transfers its energy to the TPL through the dipolar coupling, and the current driven spin precession ceases to exist.
Discussion
==========
To summarize, we observed an anomalous breakdown of the current driven spin oscillation in the two spin layers, each of which is antiferromagnetically coupled to the other by the dipolar interaction, and one of which has an additional exchange bias field. From the micromagnetic calculations, the mechanism of the enhanced damping could be understood as the energy transfer between the two layers and the spin transfer torque acting reversely on each other. This damping mechanism gives another way to control the magnetic damping. We can control the damping of the coupled system by changing the resonant frequencies of the layers, i.e., by changing the external field.
Besides, the magnetic field acts as a gate to control the energy transfer between the two layers. When the field is away from the breakdown point, the oscillation energies in the two layers are rather isolated. But by putting the field on the breakdown point, the energy flows to the other layer.
Methods
=======
Sample Preparations
-------------------
The multilayer film has been grown in the sputtering system with the base pressure around 10^−7^ Pa and post-annealed at 275^o^C under a magnetic field of about 1 T for two hours in order to align the spin orientations of the pinned layers in a specific direction. The sample has been patterned into a cylindrical nanopillar structure with a diameter of 90 nm by the electron beam lithography followed by ion beam etching. Subsequently, the sample was completely covered by the thick SiO~2~ film followed by the chemical mechanical polishing (CMP) to open only the top contact area. The bottom electrode contact area was patterned through the photolithography and ion beam etching up to the bottom layer (TiN). After opening the top and bottom electrode contact through an additional photolithography, the Ti/Au electrodes have been formed through lift-off process. The bottom and top electrode have been formed in a form accessible by GSG type radio-frequency (RF) probe.
Micromagnetic calculations
--------------------------
In the micromagnetic calculation, we did not take into account the spin transfer torque between the TPL and BPL. For the calculation of the microwave power shown in [Fig. 3](#f3){ref-type="fig"}, the micromagnetic calculation was integrated with a model circuit that mimics the realistic measurement circuit. The calculation was carried out with the single domain model for each layer, but no significant changes were observed from the results using multi-domain model, because the applied current density through the MTJ is low enough. The following parameters were used for the MTJ: Layer structure is PtMn layer/ BPL(1.5 nm)/ Ru(0.8 nm)/ TPL(1.5 nm)/ MgO(0.8 nm\~1000 nm)/ FL(2.0 nm). The layer has a circular shape with diameter 100 nm. The material parameters are as follows. The saturation magnetization *M~s~* = 1,300 emu/cc, perpendicular interface anisotropy *K~s~* = 1.0 mJ/m^2^ for FL, *M~s~* = 1240 emu/cc, *K~s~* = 0.9 mJ/m^2^ for TPL, and *M~s~* = 1430 emu/cc, exchange bias field from the PtMn layer *μ~0~H~ex~* = 150 mT were used for BPL. Interlayer exchange coupling constant between TPL and BPL is −1.3 μJ/m, damping constant *α* = 0.01, spin polarization *P* = 0.4, TMR = 30%, and *RA* value in parallel state of 10 ohm μm^2^ were used. All parameters were chosen to mimic the experiments. In-plane external fields were applied along the easy axis with a tilt angle of 10°. Positive external fields favor the anti-parallel configuration of magnetizations. For the stochastic calculation, the Gaussian-distributed random fluctuation fields (mean = 0, standard deviation = \[*2k~B~T/(M~S~VΔt*)\], where *Δt* is the integration time step, *V* is the volume of unit cell)[@b31] have been added to the effective fields of LLGS equation.
Author Contributions
====================
S.C.Lee and U.H.Pi contributed equally to this work especially on the analysis. S.C.Lee executed the micromagnetic calculation based on LLGS equation. U.H.Pi carried out the measurements and wrote the manuscript. K.Kim designed and grew the magnetic film, and K.S.Kim fabricated the device and directed the project. J.Shin and U.Chung supervised all the works.
Supplementary Material {#s1}
======================
###### Supplementary Information
Current Driven Magnetic Damping in Dipolar-Coupled Spin System
{#f1}
{#f2}
{#f3}
{#f4}
{#f5}
{#f6}
{#f7}
| {
"pile_set_name": "PubMed Central"
} |
Prolonged symptom duration and disability are common in adults hospitalized with severe coronavirus disease 2019 (COVID-19). Characterizing return to baseline health among outpatients with milder COVID-19 illness is important for understanding the full spectrum of COVID-19--associated illness and tailoring public health messaging, interventions, and policy. During April 15--June 25, 2020, telephone interviews were conducted with a random sample of adults aged ≥18 years who had a first positive reverse transcription--polymerase chain reaction (RT-PCR) test for SARS-CoV-2, the virus that causes COVID-19, at an outpatient visit at one of 14 U.S. academic health care systems in 13 states. Interviews were conducted 14--21 days after the test date. Respondents were asked about demographic characteristics, baseline chronic medical conditions, symptoms present at the time of testing, whether those symptoms had resolved by the interview date, and whether they had returned to their usual state of health at the time of interview. Among 292 respondents, 94% (274) reported experiencing one or more symptoms at the time of testing; 35% of these symptomatic respondents reported not having returned to their usual state of health by the date of the interview (median = 16 days from testing date), including 26% among those aged 18--34 years, 32% among those aged 35--49 years, and 47% among those aged ≥50 years. Among respondents reporting cough, fatigue, or shortness of breath at the time of testing, 43%, 35%, and 29%, respectively, continued to experience these symptoms at the time of the interview. These findings indicate that COVID-19 can result in prolonged illness even among persons with milder outpatient illness, including young adults. Effective public health messaging targeting these groups is warranted. Preventative measures, including social distancing, frequent handwashing, and the consistent and correct use of face coverings in public, should be strongly encouraged to slow the spread of SARS-CoV-2.
Prolonged illness is well described in adults with severe COVID-19 requiring hospitalization, especially among older adults ([@R1],[@R2]). Recently, the number of SARS-CoV-2 infections in persons first evaluated as outpatients have increased, including cases among younger adults ([@R3]). A better understanding of convalescence and symptom duration among outpatients with COVID-19 can help direct care, inform interventions to reduce transmission, and tailor public health messaging.
The Influenza Vaccine Effectiveness in the Critically Ill (IVY) Network, a collaboration of U.S. health care systems, is conducting epidemiologic studies on COVID-19 in both inpatient and outpatient settings ([@R4],[@R5]). Fourteen predominantly urban academic health systems in 13 states each submitted a list of adults with positive SARS-CoV-2 RT-PCR test results obtained during March 31--June 4, 2020, to Vanderbilt University Medical Center. Site-specific random sampling was then performed on a subset of these patients who were tested as outpatients and included patients tested in the emergency department (ED) who were not admitted to the hospital at the testing encounter and those tested in other outpatient clinics. At 14--21 days from the test date, CDC personnel interviewed the randomly sampled patients or their proxies by telephone to obtain self-reported baseline demographic, socioeconomic, and underlying health information, including the presence of chronic medical conditions. Call attempts were made for up to seven consecutive days, and interviews were conducted in several languages ([@R4]). Respondents were asked to report the number of days they felt unwell before the test date, COVID-19--related symptoms experienced at the time of testing ([@R6]), whether symptoms had resolved by the date of the interview, and whether the patient had returned to their usual state of health. For this data analysis, respondents were excluded if they did not complete the interview, if a proxy (e.g., family member) completed the interview (because of their incomplete knowledge of symptoms), if they reported a previous positive SARS-CoV-2 test (because the reference date for symptoms questions was unclear), or (because this analysis focused on symptomatic persons) if they did not answer symptoms questions or denied all symptoms at testing.
Descriptive statistics were used to compare characteristics among respondents who reported returning and not returning to their usual state of health by the date of the interview. Generalized estimating equation regression models with exchangeable correlation structure accounting for clustering by site were fitted to evaluate the association between baseline characteristics and return to usual health, adjusting for potential a priori-selected confounders. Resolution and duration of individual symptoms were also assessed. Statistical analyses were conducted using Stata software (version 16; StataCorp).
At least one telephone call was attempted for 582 patients (including 175 \[30%\] who were tested in an ED and 407 \[70%\] in non-ED settings), with 325 (56%) interviews completed (89 \[27%\] ED and 236 \[73%\] non-ED). Among 257 nonrespondents, 178 could not be reached, 37 requested a callback but could not be reached on further call attempts, 28 refused the interview, and 14 had a language barrier. Among the 325 completed interviews, 31 were excluded: nine (3%) because a proxy was interviewed, 17 (5%) because a previous positive SARS-CoV-2 test was reported, and five (2%) who did not answer the symptoms questions. Two additional respondents were called prematurely at 7 days and were also excluded.[\*](#FN1){ref-type="fn"} Among the 292 remaining patient respondents, 274 (94%) reported one or more symptoms at testing and were included in this data analysis. Following outpatient testing, 7% (19 of 262 with available data) reported later being hospitalized, a median of 3.5 days after the test date. The median age of symptomatic respondents was 42.5 years (interquartile range \[IQR\] = 31--54 years), 142 (52%) were female, 98 (36%) were Hispanic, 96 (35%) were non-Hispanic white, 48 (18%) were non-Hispanic black, and 32 (12%) were other non-Hispanic race. Overall, 141 of 264 (53%) with available data reported one or more chronic medical conditions. The median interval from test to interview date was 16 days (IQR = 14--19 days); the median number of days respondents reported feeling unwell before being tested for SARS-CoV-2 was 3 (IQR = 2--7 days).
Return to Usual State of Health
===============================
Among the 270 of 274 interviewees with available data on return to usual health,[^†^](#FN2){ref-type="fn"} 175 (65%) reported that they had returned to their usual state of health a median of 7 days (IQR = 5--12 days) from the date of testing ([Table 1](#T1){ref-type="table"}). Ninety-five (35%) reported that they had not returned to their usual state of health at the time of interview. The proportion who had not returned to their usual state of health differed across age groups: 26% of interviewees aged 18--34 years, 32% aged 35--49 years, and 47% aged ≥50 years reported not having returned to their usual state of health (p = 0.010) within 14--21 days after receiving a positive test result. Presence of chronic conditions also affected return to health rates; among 180 persons with no or one chronic medical condition, 39 with two chronic medical conditions, and 44 with three or more chronic medical conditions, 28%, 46%, and 57%, respectively, reported not having returned to their usual state of health (p = 0.003) within 14--21 days after having a positive test result. Among respondents aged 18--34 years with no chronic medical condition, 19% (nine of 48) reported not having returned to their usual state of health. Adjusting for other factors, age ≥50 versus 18--34 years (adjusted odds ratio \[aOR\] = 2.29; 95% confidence interval \[CI\] = 1.14--4.58) and reporting three or more versus no chronic medical conditions (aOR = 2.29; 95% CI = 1.07--4.90) were associated with not having returned to usual health ([Table 2](#T2){ref-type="table"}). Obesity (body mass index ≥30 kg per m^2^) (aOR 2.31; 95% CI = 1.21--4.42) and reporting a psychiatric condition[^§^](#FN3){ref-type="fn"} (aOR 2.32; 95% CI = 1.17--4.58) also were associated with more than twofold odds of not returning to the patient's usual health after adjusting for age, sex, and race/ethnicity.
###### Characteristics of symptomatic outpatients with SARS-CoV-2 real-time reverse transcription--polymerase chain reaction (RT-PCR)---positive test results (N = 270)[\*](#FN1){ref-type="fn"} who reported returning to usual state of health or not returning to usual state of health at an interview conducted 14--21 days after testing --- 14 academic health care systems,[^†^](#FN2){ref-type="fn"} United States, March--June 2020
Characteristic Total Returned to usual health, no. (row %) P-value^§^
------------------------------------------------------ --------- --------------------------------------- ------------ -------
**Sex** 0.14
Women **140** 85 (61) 55 (39)
Men **130** 90 (69) 40 (31)
**Age group (yrs)** 0.010
18--34 **85** 63 (74) 22 (26)
35--49 **96** 65 (68) 31 (32)
≥50 **89** 47 (53) 42 (47)
**Race/Ethnicity** 0.29
White, non-Hispanic **94** 58 (62) 36 (38)
Black, non-Hispanic **46** 26 (57) 20 (43)
Other race, non-Hispanic **32** 24 (75) 8 (25)
Hispanic **98** 67 (68) 31 (32)
**Insurance (14 missing)** 0.69
No **46** 31 (67) 15 (33)
Yes **210** 135 (64) 75 (36)
**No. of medical conditions (7 missing)** 0.003
0 **123** 87 (71) 36 (29)
1 **57** 41 (72) 16 (28)
2 **39** 21 (54) 18 (46)
≥3 **44** 19 (43) 25 (57)
**Individual medical conditions (7 missing all)**^¶^
Hypertension **64** 33 (52) 31 (48) 0.018
Obesity (body mass index \>30 kg/m^2^) **51** 23 (45) 28 (55) 0.002
Psychiatric condition **49** 23 (47) 26 (53) 0.007
Asthma **36** 23 (64) 13 (36) 0.99
Diabetes **28** 16 (57) 12 (43) 0.43
Immunosuppressive condition **15** 6 (40) 9 (60) 0.047
Autoimmune condition **13** 7 (54) 6 (46) 0.44
Blood disorder **8** 4 (50) 4 (50) 0.47
Chronic kidney disease **7** 3 (43) 4 (57) 0.26
Chronic obstructive pulmonary disease **7** 4 (57) 3 (43) 0.71
Liver disease **6** 4 (67) 2 (33) 1.00
Neurologic condition **6** 3 (50) 3 (50) 0.48
Coronary artery disease **4** 3 (75) 1 (25) 1.00
Congestive heart failure **2** 2 (100) 0 (0) 0.54
\* 294 patients responded to an interview 2--3 weeks after testing, did not report a previous positive SARS-CoV-2 test before the reference test, and answered questions about symptoms. Of these, 276 (94%) reported one or more symptoms at the time of SARS-CoV-2 RT-PCR testing, with 272 (99%) reporting whether they had returned to their usual state of health by the time of the interview. Two additional patients excluded who were called at 7 days, with 270 included here.
^†^ Patients were randomly sampled from fourteen academic healthcare systems in 13 states (University of Washington \[Washington\], Oregon Health and Sciences University \[Oregon\], University of California Los Angeles and Stanford University \[California\], Hennepin County Medical Center \[Minnesota\], Vanderbilt University \[Tennessee\], Ohio State University \[Ohio\], Wake Forest University \[North Carolina\], Montefiore Medical Center \[New York\], Beth Israel Deaconess Medical Center and Baystate Medical Center \[Massachusetts\], Intermountain Healthcare \[Utah/Idaho\], University of Colorado Hospital \[Colorado\], and Johns Hopkins University \[Maryland\]).
^§^ Respondents who reported returning to usual health and respondents who reported not returning to usual health were compared using the chi-square test or Fisher\'s exact test.
^¶^Excluding seven (3%) patients who did not answer questions about chronic underlying medical conditions; for those who answered questions about underlying conditions, some respondents were missing data on obesity (two), neurologic conditions (one), and psychiatric conditions (one).
###### Characteristics associated with not returning to usual health among symptomatic outpatients with SARS-CoV-2 real-time reverse transcription--polymerase chain reaction (RT-PCR)--positive test results (N = 270)[\*](#FN1){ref-type="fn"} reported at an interview conducted 14--21 days after testing --- 14 academic health care systems,[^†^](#FN2){ref-type="fn"} United States, March--June 2020
Characteristic Odds of not returning to "usual health" at 14--21 days after testing
--------------------------------------- ---------------------------------------------------------------------- --------------------
**Age group (yrs)**
18--34 Referent Referent
35--49 1.40 (0.73--2.67) 1.38 (0.71--2.69)
≥50 2.64 (1.39--5.00) 2.29 (1.14--4.58)
**Sex**
Women Referent Referent
Men 0.68 (0.41--1.13) 0.80 (0.46--1.38)
**Race/Ethnicity**
White, non-Hispanic Referent Referent
Black, non-Hispanic 1.23 (0.60--2.53) 1.13 (0.53--2.45)
Other, non-Hispanic 0.53 (0.21--1.31) 0.63 (0.24--1.61)
Hispanic 0.74 (0.40--1.34) 0.83 (0.44--1.58)
**No. of medical conditions**
0 Referent Referent
1 0.94 (0.47--1.89) 0.74 (0.35--1.55)
2 2.09 (1.00--4.38) 1.50 (0.68--3.33)
≥3 3.19 (1.56--6.50) 2.29 (1.07--4.90)
**Individual medical conditions\*\***
Hypertension 1.98 (1.12--3.52) 1.30 (0.67--2.51)
Obesity (BMI \>30 kg/m^2^) 2.65 (1.42--4.95) 2.31 (1.21--4.42)
Psychiatric condition 2.42 (1.29--4.56) 2.32 (1.17--4.58)
Asthma 1.00 (0.48--2.08) 1.02 (0.47--2.20)
Diabetes 1.38 (0.62--3.05) 1.06 (0.46--2.44)
Immunosuppressive condition 2.84 (0.98--8.26) 2.33 (0.77--7.04)
Autoimmune condition 1.55 (0.51--4.76) 1.05 (0.32--3.46)
Blood disorder 1.82 (0.45--7.45) 1.43 (0.33--6.24)
Chronic kidney disease 2.42 (0.53--11.05) 2.36 (0.48--11.51)
Chronic obstructive pulmonary disease 1.34 (0.29--6.12) 0.70 (0.14--3.48)
Liver disease 0.88 (0.16--4.90) 0.72 (0.12--4.25)
Neurologic condition 1.78 (0.35--9.01) 1.23 (0.23--6.62)
Coronary artery disease 0.58 (0.06--5.70) 0.48 (0.05--4.92)
Congestive heart failure --- ---
**Abbreviations**: BMI = body mass index; CI = confidence interval.
\* 294 patients responded to 14--21-day interview, did not report a previous positive SARS-CoV-2 test before the reference test, and answered questions about symptoms; 276 (94%) of these reported one or more symptoms at the time of SARS-CoV-2 RT-PCR testing, with 272 (99%) reporting whether they had returned to their usual state of health by the time of the interview. Two additional patients who were called at 7 days were excluded, with 270 included here.
^†^ Patients were randomly sampled from academic healthcare systems in 13 states (University of Washington \[Washington\], Oregon Health and Sciences University \[Oregon\], University of California Los Angeles and Stanford University \[California\], Hennepin County Medical Center \[Minnesota\], Vanderbilt University \[Tennessee\], Ohio State University \[Ohio\], Wake Forest University \[North Carolina\], Montefiore Medical Center \[New York\], Beth Israel Deaconess Medical Center and Baystate Medical Center \[Massachusetts\], Intermountain Healthcare \[Utah/Idaho\], University of Colorado Hospital \[Colorado\], and Johns Hopkins University \[Maryland\]).
^§^ For this analysis, generalized estimation equation (GEE) models with exchangeable correlation structure were used to estimate the association between characteristics and the odds of not returning to usual health by the date of the 14--21-day interview. GEE models were used to account for clustering of cases by site. 95% CIs including 1.00 are not considered statistically significant.
^¶^ In adjusted GEE models for age, sex, race/ethnicity, and number of chronic medical conditions, the other variables were used to adjust for potential confounders. Models for individual conditions (e.g., hypertension) were adjusted for age, sex, and race/ethnicity.
\*\* Medical conditions are not exclusive and individual patients could have more than one chronic medical condition.
Resolution of Symptoms and Duration
===================================
Among the 274 symptomatic outpatients, the median number of symptoms was seven of 17 listed in the interview tool (IQR = 5--10), with fatigue (71%), cough (61%), and headache (61%) those most commonly reported ([Figure](#F1){ref-type="fig"}). Among respondents who reported fever and chills on the day of testing, these resolved in 97% and 96% of respondents, respectively. Symptoms least likely to have resolved included cough (not resolved in 43% \[71 of 166\]) and fatigue (not resolved in 35% \[68 of 192\]); among 90 who reported shortness of breath at the time of testing, this symptom had not resolved in 26 (29%). The median interval to symptom resolution among those who reported individual symptoms at the time of testing but not at the time of the interview ranged from 4 to 8 days from the test date, with the longest intervals reported for loss of smell (median = 8 days; IQR = 5--10.5 days) and loss of taste (median = 8 days; IQR = 4--10 days). Among respondents who reported returning to their usual state of health, 34% (59 of 175) still reported one or more of the 17 queried COVID-related symptoms at the time of the interview.
{ref-type="fn"} --- 14 academic health care systems,[^†^](#FN2){ref-type="fn"} United States, March--June 2020\
\* 294 patients responded to 14--21-day interview, did not report a previous positive SARS-CoV-2 test before the reference test, and answered questions about symptoms; 276 (94%) of these reported one or more symptoms at the time of SARS-CoV-2 RT-PCR testing; those who were interviewed at 7 days were excluded, with 274 included here.\
^†^ Patients were randomly sampled from 14 academic health care systems in 13 states (University of Washington \[Washington\], Oregon Health and Sciences University \[Oregon\], University of California Los Angeles and Stanford University \[California\], Hennepin County Medical Center \[Minnesota\], Vanderbilt University \[Tennessee\], Ohio State University \[Ohio\], Wake Forest University \[North Carolina\], Montefiore Medical Center \[New York\], Beth Israel Deaconess Medical Center and Baystate Medical Center \[Massachusetts\], Intermountain Healthcare \[Utah/Idaho\], University of Colorado Hospital \[Colorado\], and Johns Hopkins University \[Maryland\]).](mm6930e1-F){#F1}
Discussion
==========
Most studies to date have focused on symptoms duration and clinical outcomes in adults hospitalized with severe COVID-19 ([@R1],[@R2]). This report indicates that even among symptomatic adults tested in outpatient settings, it might take weeks for resolution of symptoms and return to usual health. Not returning to usual health within 2--3 weeks of testing was reported by approximately one third of respondents. Even among young adults aged 18--34 years with no chronic medical conditions, nearly one in five reported that they had not returned to their usual state of health 14--21 days after testing. In contrast, over 90% of outpatients with influenza recover within approximately 2 weeks of having a positive test result ([@R7]). Older age and presence of multiple chronic medical conditions have previously been associated with illness severity among adults hospitalized with COVID-19 ([@R8],[@R9]); in this study, both were also associated with prolonged illness in an outpatient population. Whereas previous studies have found race/ethnicity to be a risk factor for severe COVID-19 illness ([@R10]), this study of patients whose illness was diagnosed in an outpatient setting did not find an association between race/ethnicity and return to usual health although the modest number of respondents might have limited our ability to detect associations. The finding of an association between chronic psychiatric conditions and delayed return to usual health requires further evaluation. These findings have important implications for understanding the full effects of COVID-19, even in persons with milder outpatient illness. Notably, convalescence can be prolonged even in young adults without chronic medical conditions, potentially leading to prolonged absence from work, studies, or other activities.
The findings in this report are subject to at least three limitations. First, nonrespondents might have differed from survey respondents; for example, those with more severe illness might have been less likely to respond to telephone calls if they were subsequently hospitalized and unable to answer the telephone. Second, symptoms that resolved before the test date or that commenced after the date of testing were not recorded in this survey. Finally, as a telephone survey, this study relied on patient self-report and might have been subject to incomplete recall or recall bias.
Nonhospitalized COVID-19 illness can result in prolonged illness and persistent symptoms, even in young adults and persons with no or few chronic underlying medical conditions. Public health messaging should target populations that might not perceive COVID-19 illness as being severe or prolonged, including young adults and those without chronic underlying medical conditions. Preventative measures, including social distancing, frequent handwashing, and the consistent and correct use of face coverings in public, should be strongly encouraged to slow the spread of SARS-CoV-2.
###### Summary
What is already known about this topic?
---------------------------------------
Relatively little is known about the clinical course of COVID-19 and return to baseline health for persons with milder, outpatient illness.
What is added by this report?
-----------------------------
In a multistate telephone survey of symptomatic adults who had a positive outpatient test result for SARS-CoV-2 infection, 35% had not returned to their usual state of health when interviewed 2--3 weeks after testing. Among persons aged 18--34 years with no chronic medical conditions, one in five had not returned to their usual state of health.
What are the implications for public health practice?
-----------------------------------------------------
COVID-19 can result in prolonged illness, even among young adults without underlying chronic medical conditions. Effective public health messaging targeting these groups is warranted.
Two patients interviewed early at 12 days and three interviewed at 13 days after testing were included. Two patients who requested interview after 21 days because they were unavailable at 14--21 days were included (interviews were conducted at 25 and 26 days). All other included respondents were interviewed 14--21 days after testing.
Patients were asked the question "Would you say that you are feeling back to your usual health?"
Psychiatric conditions included anxiety disorder (38), depression (21), posttraumatic stress disorder (two), paranoia (two), obsessive-compulsive disorder (one), schizophrenia (one); some patients reported more than one condition.
All authors have completed and submitted the International Committee of Medical Journal Editors form for disclosure of potential conflicts of interest. Christopher J. Lindsell reports grants from National Institutes of Health and Department of Defense, and contracts with the Marcus Foundation, CDC, Endpoint Health, Entegrion, bioMerieux, and Bioscape Digital, outside the submitted work. Daniel J. Henning reports personal fees from CytoVale and grants from Baxter, outside the submitted work. Akram Khan reports grants from United Therapeutics, Actelion Pharmaceuticals, Regeneron, and Reata Pharmaceuticals, outside the submitted work. Samuel M. Brown reports grants from National Institutes of Health, Department of Defense, Intermountain Research and Medical Foundation, and Janssen, consulting fees paid to his employer from Faron and Sedana, and royalties from Oxford University Press, outside the submitted work. Ithan D. Peltan reports grants from National Institutes of Health, Asahi Kasei Pharma, Immunexpress Inc., Janssen Pharmaceuticals, and Regeneron, outside the submitted work. Todd W. Rice reports personal fees from Cumberland Pharmaceuticals, Inc., Cytovale, Inc., and Avisa, LLC, outside the submitted work. No other potential conflicts of interest were disclosed.
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION
============
MicroRNAs (miRNAs) belong to a recently identified group of the large family of noncoding RNAs ([@B1]). The mature miRNA is usually 19--27 nt long and is derived from a larger precursor that folds into an imperfect stem-loop structure. The mode of action of the mature miRNA in mammalian systems is dependent on complementary base pairing primarily to the 3′-UTR region of the target mRNA, thereafter causing the inhibition of translation and/or the degradation of the mRNA.
According to recent estimates, while over 30% of vertebrate genomes is transcribed ([@B2]), only 1% consists of coding genes, suggesting that the rest must be various types of noncoding RNA genes. In addition, 701 human miRNA hairpin sequences are currently contained in the miRNA registry (miRBase, release 12.0), of which 92% have been experimentally verified, and it is anticipated that there may be thousands more. A recent estimate of the total number of miRNA genes in the human genome provided by the study of Miranda *et al*. ([@B3]) is in the range of ∼55 000, a number significantly larger than the experimentally verified human miRNAs currently in the registry. Searching through the entire genome of human and/or other species for novel miRNA genes is a complicated task for which fast, flexible and reliable identification methods are required. Currently available experimental approaches working towards this goal are complex and sub-optimal ([@B4]). Inefficiencies result from various sources, including difficulty in isolating certain miRNAs by cloning due to low expression, stability, tissue specificity and technical difficulties of the cloning procedure while selecting the right genomic region to investigate is often a very challenging task of its own. Computational prediction of miRNA genes from genomic sequences is an alternative technique which offers a much faster, cheaper and effective way of identifying putative miRNA genes. Moreover, by predicting the location of miRNA genes, these methods enable experimentalists to concentrate their efforts on genomic regions more likely to contain novel miRNA genes, thus facilitating the discovery process.
Accurate prediction of new miRNAs requires the consideration of certain characteristic properties of these molecules based on either experimental ([@B5; @B6; @B7]), or computational evidence ([@B8; @B9; @B10; @B11; @B12]) which can be used to build a classification scheme or predictive model. These general features include sequence composition, secondary structure and species conservation. MiRNA gene prediction can be achieved via the use of supervised algorithms that are trained on known miRNA biological features and then used to identify putative miRNAs, or un-supervised algorithms such as alignment or conservation. The prediction methodology can also vary significantly between different studies. It can be performed by: scanning for hairpins within sequences that are conserved between closely related organisms like *Caenorhabditis elegans* and *C. briggsae* ([@B10],[@B13]), looking for regions of homology between known miRNAs and other sites within aligned genomes, as for example between human and mouse ([@B14]) or looking for conserved regions of synteny---conserved clustering of miRNAs in the genomes of closely related organisms ([@B14]). Profile-based detection ([@B15]) and secondary structure alignment ([@B16]) of miRNAs have also been suggested using sequences across multiple, highly divergent, organisms (i.e. mouse and fugu). Support vector machines that take into account multiple biological features such as free energy, paired bases, loop length and stem conservation have also been used to predict novel miRNAs ([@B8],[@B9],[@B17]). Many of these prediction methods undertake a pipeline approach, whereby cut-offs are assigned and sequences are eliminated as the pipeline proceeds ([@B10],[@B13]). The drawback of these approaches is that they lose numerous true miRNAs along the line due to stringent cut-offs. Other approaches use homology to detect novel miRNAs based on their similarity to previously identified miRNAs ([@B14; @B15; @B16]). These methods obviously fail when scanning distantly related sequences and when novel miRNAs lack detectable homologs. Two studies ([@B12],[@B18]) used Hidden Markov Models (HMMs) and Bayesian classifiers, respectively, to simultaneously consider sequence and structure information for the identification of miRNA precursors (pre-miRNAs). However, conservation information, a very important characteristic of the majority of miRNA precursors, was not integrated in those algorithms. Finally, in a more recent study ([@B19]), an HMM approach that simultaneously considered structure and conservation features of miRNA genes was shown to achieve very high performance on identifying miRNAs in the human genome.
In addition to computational tools, large scale, high throughput methods such as tiling arrays or deep sequencing have recently been used for the identification of novel miRNA genes ([@B20; @B21; @B22]). These methods are particularly useful as they can provide a very sophisticated and accurate expression map for small RNAs in the genome. Moreover, if such data is coupled to computational tools, it can facilitate rapid and precise detection of novel miRNAs, while at the same time giving greater credence to computational predictions.
MiRNAs have been suggested to play a key regulatory role in numerous processes, including cancer ([@B23],[@B24]). For example, the expression levels of let-7 ([@B25]), miR-15a/miR-16-1 cluster ([@B26]) and neighboring miR-143/miR-145 ([@B27]), are found to be reduced in some malignancies, while other miRNAs such as the miR-17-92 cluster ([@B28; @B29; @B30]) and miR-155/BIC ([@B31]), are overexpressed in various cancers. Additionally it was recently shown that a high percentage of miRNA genes are located in cancer-associated genomic regions (CAGRs), thus implicating miRNAs in tumorigenic events ([@B32]). CAGRs take the form of (i) minimal regions of loss of heterozygosity (LOH), suggestive of the presence of tumor suppressor genes; (ii) minimal regions of amplification, suggestive of the presence of oncogenes; and (iii) common breakpoint regions in or near possible oncogenes or tumor suppressor genes. The identification of novel miRNA genes within these regions is very important as it may reveal putative gene players that exert a regulatory effect on different types of cancer, contribute to the better understanding of molecular pathways responsible for oncogenesis and provide potential targets for therapeutic intervention.
In this work, we present an efficient and freely available prediction tool (SSCprofiler) where *Profile* HMMs are trained to recognize key biological features of miRNAs such as sequence, structure and conservation in order to identify novel miRNA precursors. We first use our method to learn with high accuracy the characteristic features of 249 human miRNA precursors and then apply the trained model on CAGRs in search of novel miRNA genes. Predictions are ranked according to expression information from a recently published full genome tiling array ([@B21]) and the top four scoring candidates are verified experimentally using northern blot.
MATERIALS AND METHODS
=====================
Datasets
--------
The sequences of human pre-miRNAs used to train and test the HMMs were downloaded from the miRNA registry (version 12.0) (<http://microrna.sanger.ac.uk/sequences/>). For the training/validation sequences BLASTclust ([@B33]) was initially performed to cluster all miRNA sequences into groups by precursor similarity and the most conserved member (according to multiz files) was used to represent the cluster. This procedure was done to eliminate redundant pre-miRNAs and avoid over-representation of similar miRNA precursors. Following a set of filtering criteria detailed below, a total of 249 sequences (originally listed in version 8.0) were used for training/validation while a total of 219 sequences (not in version 8.0) were used as a blind test set. The negative miRNA sequences were derived from 3′-UTR regions of the human genome (release---May 2004) since no true miRNA has yet been reported to reside within these regions. They were generated by using a sliding window of 104 nt, shifted 11 nt at a time, over the 3′-UTR regions. RNAfold was executed for every shift and the free energy of the secondary structure was noted. Only sequences whose energy did not exceed a threshold of --14.44 kcal/mol and had at least 14% of their nucleotides conserved, were selected. This generated over 35 000 negative sequences.
Biological features
-------------------
SSCprofiler takes into account three different biological features: sequence, structure and conservation of miRNA precursors. In this study, conservation was retrieved from the multiz ([@B34]) full genome alignment files of the human May 2004 hg17 genome assembly and seven other vertebrate genomes: Mouse May 2004 (mm5), Rat June 2003 (rn3), Dog July 2004 (canFam1), Chicken February 2004 (galGal2), Fugu August 2002 (fr1), Zebrafish November 2003 (danRer1). Chimp data were not included due to high percentage similarity (∼95%) with humans. RNA secondary structure prediction was performed using the RNAfold function of the Vienna-RNA ([@B35]) package. A fixed window (104 nt) was used to align all sequences in order to generate a multiple sequence alignment (msa) required to train the HMM (see Training and Validation of the HMMs). This was achieved by enlarging sequences that fell shorter than this window using flanking genomic nucleotides and trimming sequences that exceeded the defined msa window. The window length was consequently used as the length of the training model and as the window size for querying genomic sequences.
Filtering
---------
To minimize the search space and reduce computational load, the data were first filtered using various secondary structure features of miRNA precursors. Filtering results were displayed as histograms that show the relative distributions of the positive and negative data with respect to eight features: Hairpin---the number of hairpinsBulges---the number of bulgesLoops---the number of loopsAsymmetry---difference in loops + bulges on either side of the hairpin.Bulges-loops---sum of loop and bulge countHairpin length---length of the hairpinFolding min energy---min energy as defined by RNAfoldConservation---according to multiz full genome alignment files
Illustration of data distributions for the various filtering parameters was done to facilitate the filtering process by enabling the adjustment of cut-off values according to the specific dataset. Cut-off values for each of these features are modifiable both prior and after the training procedure (see 'Results' section, [Figure 4](#F4){ref-type="fig"}). Figure 4.Histograms of the distributions of human miRNA (Red--Positive) and negative sequences (Blue--Negitive), as displayed by SSCprofiler. Only three of the eight filtering parameters are shown here. (**A**) Hairpin length, (**B**) Asymmetry and (**C**) Bulges-loops count. Looking at the distributions of positive and negative data, it is possible for the user to select cut-offs that separate the two distributions which can be used for filtering the data.
Combining sequence, structure and conservation
----------------------------------------------
In order to simultaneously consider multiple biological features, a 16 character code was developed that integrates sequence, structure and conservation information for every nucleotide position in a given genomic sequence. Specifically, each position in the genomic sequence is replaced by 1 of 16 letters, depending on three factors: (i) Sequence (A, C, U, G), (ii) Structure, (M = match, L = loop) and (iii) Conservation (\* = conserved, ''= not-conserved) as detailed in [Table 1](#T1){ref-type="table"}. Table 1.The 16-letter code that was used to integrate sequence, structure and conservation informationConservation and structureSequenceACGU**\*M**LMNP**''M**CDEF**\*L**QRST**''L**GIHK
Profile HMMs
------------
The HMMER ([@B36]) software package was used to build a HMM capable of predicting RNA or DNA *Profiles*. HMMs are *generative probabilistic models* which are frequently used to address serious theoretical problems. For correct statistical inference, it is necessary to be able to calculate a probability distribution *P*(*S\|M*) for the probability of sequences *S* given a model *M*, and have this quantity sum to one over the 'space' of all sequences. Generative models work by *recursive* enumeration of possible sequences from a finite set of rules---rules that in an HMM are represented by states, state transitions and symbol emission probabilities. HMMER uses a *Profile* HMM architecture called Plan 7 which is illustrated in [Figure 1](#F1){ref-type="fig"}. *Profile* HMMs are statistical models of multiple sequence alignments. They capture position-specific information about how conserved each column of the alignment is, and which residues are most likely. Figure 1.The HMMER Plan 7 architecture. Squares indicate match states (modeling consensus positions in the alignment). Diamonds indicate insert states (modeling insertions relative to consensus) and special random sequence emitting states. Circles indicate delete states (modeling deletions relative to consensus) and special begin/end states. Arrows indicate state transitions. Figure was adopted from Eddy SR, 1998.
Training and validation of the HMMs
-----------------------------------
Machine learning algorithms such as HMMs require carefully chosen training and validation data sets in order to achieve maximum performance. SSCprofiler allows for a user-defined partitioning of imported data into training and validation sets in order to perform a boosting validation. This is done by randomly dividing the positive data into *K* subsets, some of which are used for training and others for validation. The negative data is only used for validation purposes and is not included in the training sets. This partitioning is repeated 100 times and an average validation performance is reported. The training/validation results are displayed as sensitivity and specificity plots in order to obtain an indication of how well the trained HMMs perform on the specific dataset ([Supplementary Figure S1](http://nar.oxfordjournals.org/cgi/content/full/gkp120/DC1)). The *x*-axis of these plots displays the HMM score threshold and the *y*-axis is the average sensitivity and specificity for every score over the 100 validation runs. Training is performed on the biological feature(s) selected before hand. An overview of the training procedure is shown schematically in the flowchart of [Figure 2](#F2){ref-type="fig"}. At the end of the training/validation procedure all true miRNAs are combined to train a final model which is subsequently used for scanning genomic regions. The HMM score at which average sensitivity and specificity values are 'optimal' can be selected by the user and it is used as a cut-off or threshold for classifying sequences as positives (true miRNAs) or negatives. Figure 2.The supervised procedure of training HMMs for miRNA precursor identification. Biological features of miRNA biogenesis and conservation across other organisms are used as input for training. Initially, secondary structure prediction is performed using the program RNAfold. Every nucleotide position is henceforth represented by an 'M' for match and an 'L' for loop. This information is aligned with conservation and sequence information for every nucleotide position. The 16-character code shown in [Table 1](#T1){ref-type="table"} is then used to represent each position in this alignment with a single letter. The resulting strings of characters for true miRNAs are aligned with respect to their hairpins and used as a training set for the HMM. Once trained, the HMMs, can be utilized to analyze sequences of desired length and assign a likelihood score. The higher the score the greater the chances of a candidate sequence being a true miRNA precursor.
Assessing the expression level of predicted candidates using tiling array data
------------------------------------------------------------------------------
To provide additional support for computational predictions, SSCprofiler enables the detection of regions in the candidate(s) that are expressed in HeLa and/or HepG2 cells according to the recently published full genome tiling array that provides an expression map at 5-nt resolution in these two cell lines ([@B21]). SSCprofiler allows for the expression threshold to be adjusted ranging from 1 to 2000 in order to retain candidates which exceed a given expression cut-off. For the results reported here, a value of 200 was used as a cut-off.
Scanning genomic regions for profiles
-------------------------------------
The process of scanning genomic regions for miRNA precursor profiles involves six steps, illustrated in [Figure 3](#F3){ref-type="fig"}. Step 1: A sliding window of selected length is passed along the genomic sequence shifting 1 nt at a time. Step 2: For every window shift, sequence structure and conservation information is retrieved according to the selected training features; i.e. structure prediction is performed and conservation is obtained from the multiz files. Step 3: Each sequence within the sliding window is passed through the filters utilizing the pre-defined filtering parameters (i.e. hairpin length, asymmetry). Step 4: For each sequence, the features used during training (sequence, structure and/or conservation) are generated according to the 16-letter key described earlier. This allows the simultaneous consideration of information for every nucleotide position in the genomic sequence. Step 5: The trained HMM is used to assign a likelihood score to each genomic sequence within the sliding window. The HMM score threshold can be selected by the user. It is usually defined as the score where sensitivity and specificity from the training/validation process were optimal. Step 6: Candidates that overlap by ≤50 nt were grouped and the candidate with the highest score is used to represent the cluster. Thereafter, the candidates are assessed according to their expression in HeLa or HepG2 cells using tiling array data. Figure 3.Flowchart of the scanning procedure.
RNA extraction and northern blot analysis
-----------------------------------------
Total RNA was extracted from HeLa cells grown in culture using Trizol. Eighty micrograms of total RNA was analyzed on a 15% denaturing polyacrylamide gel containing 8 M urea and transferred to Nytran N membrane (Schleicher & Schuell, Germany). Membranes were probed with standard DNA oligonucleotides, complementary to both polarities. Due to the difficulty in predicting accurately the location of the mature on the pre-miRNA, we select both stem sequences (maximum size 50 nt) from the stem-loop structure of the miRNA gene candidates ([Table 2](#T2){ref-type="table"}). Ten picomoles of each oligonucleotide probe was end-labeled with \[γ-32P\]ATP by using T4 polynucleotide kinase. Pre-hybridization of the filters was carried out in 7% SDS, 5× SSC, 1× Denhardt\'s solution and 0.02 M Na~2~HPO~4~ pH 7.2. Hybridizations were performed in the same solution at 50°C after the addition of the radiolabeled DNA oligonucleotide. Followed an overnight hybridization, the membranes were washed at 50°C in low stringency buffer \[2× SSC, 0.3% SDS\] twice for 30 min ([@B37]). The membranes were stripped by washing in a high stringency buffer (0.1× SSC and 0.5% SDS) for 30 min at 80°C and reprobed with the negative polarity oligonucleotides. Table 2.DNA oligonucleotidesOligo stem 1Oligo stem 2Candidate 15′-ACCTCTCCCCCTGCCAGGTTCCACCAGGGGACACCGTGTGTGT-3′5′-CGAGCAGGGCTCCCCCACCTGAGTACCTGACCATGGGCTTTGGAGAGGC-3′Candidate 25′-TACGCCCACAGCCCCCAGGCCCCCGAAGACAGGTGTCATGGA-3′5′-TCCAAGAGCATCAAGCAGCAGGGGCTGGGGGAGCCAGCAGG-3′Candidate 35′-ATCTACCAGGTCCTGGGCTTCGGGCCGCGTTCCCAAGGCAAGC-3′5′-ACCCGCGGCGAGGACACGGCCGACCGCCCGCCTGCGCCC-3′Candidate 45′-GCCCAGGAGGAGGTGGCACATCTGGGCTCCAGTCCTCGCAC-3′5′-GTGCGGGCACCGCGCGAGCCTCGCCCCTTCCCACCTGCGC-3′
RESULTS
=======
Learning characteristic features of human precursor miRNAs
----------------------------------------------------------
Human miRNA precursors from the RNA registry version 12.0 were used to assess the performance accuracy of the SSCprofiler while over 35 000 negative miRNA sequences were used for evaluation purposes (see 'Materials and Methods' section). Negative sequences were hairpin structures derived from 3′-UTR regions. These regions were selected because they have not yet been documented to contain miRNA genes. To obtain a reliable control set, negative sequences were filtered according to free energy and conservation criteria (see 'Materials and Methods' section) to ensure their resemblance with true pre-miRNAs in both structure and conservation. Prior to training, all positive and negative examples were filtered as follows: initially, filtering exclusively by a minimum energy threshold of −25.44 kcal/mol resulted in 258 true miRNAs and ∼8000 negative sequences. Consequently, seven additional filtering parameters were used to further eliminate false positives. [Figure 4](#F4){ref-type="fig"} shows the histogram distributions of the sequences prior to filtering as generated by the SSCprofiler interface with respect to three filtering parameters: Hairpin Length, Asymmetry and Bulges-Loops. Similar distributions were generated for all eight filtering parameters in order to determine the respective cut-off values that were optimal for discriminating true miRNA genes from negative data. Sequences were only retained if they met the following criteria: Hairpin = 1Bulges ≤16Loops ≤32Asymmetry ≤13Bulges-loops ≤37Hairpin length ≤16Folding min energy ≤−25.44 kcal/molConservation ≥25% of nucleotides conserved
The above-mentioned filtering procedure resulted in 249 true miRNAs and 2330 negative sequences. Subsequently, HMMs were trained solely on the true miRNAs using a 5-fold (three-fifths for training, two-fifths for validation) boosting validation procedure (as described in the 'Materials and Methods' section). The procedure was repeated for different combinations of biological features and the HMMs average performance accuracy was reported for each case. ROC curves showing the average validation performance of HMMs that utilize all possible combinations of sequence, structure and conservation information are shown in [Figure 5](#F5){ref-type="fig"}. There was a significant improvement in prediction accuracy for the validation set when certain features were combined, highlighting the importance of simultaneously incorporating additional biological information during the training procedure. The best results were obtained when all three features were used to train the HMMs, achieving on average 88.95% sensitivity and 84.16% specificity in the validation set for a score threshold of 3 ([Figure 5](#F5){ref-type="fig"} and [Supplementary Figure S1](http://nar.oxfordjournals.org/cgi/content/full/gkp120/DC1)). Once a good performance on the training/validation sets was achieved, all 249 true miRNA precursors were pooled together and used to train the final HMM taking into account the same feature combination and filtering parameters. This final model was used to build the scanning interface of the SSCprofiler. Figure 5.ROC curves for all possible combinations of sequence (Se), structure (St) and conservation (Co) features for the validation set averaged, over 100 repetitions. As evident from the figure, the area under the curve is maximized when all three features are combined. Note that conservation alone significantly outperforms sequence, structure and sequence + structure (SeSt).
To demonstrate the ability of SSCprofiler to generalize on unseen data we used 373 recently identified miRNAs from miRBase version 12.0 that were not contained in our training/validation sets. Of these, 219 precursors passed the SSCprofiler filters. [Table 3](#T3){ref-type="table"} shows the classification performance obtained by SSCprofiler on the 249 training/validation and 219 unseen test precursors for different HMM thresholds. Classification of the 219 unseen precursors was performed using the scanning interface of SSCprofiler and precursors were considered as 'identified' by the model if a significant hit was observed at their respective genomic coordinates. For this reason we only report prediction accuracy for the test set. As evident from the table, generalization performance is maximum for an HMM threshold of 1. Table 3.SSCprofiler prediction accuracy on validation and blind test sets for three different thresholdsThresholdValidationTestSensitivity (%)Specificity (%)Pred. Acc (%)388.9584.1672.15290.0781.7778.08191.378.985.84
Predicting miRNA genes in cancer-associated genomic regions (CAGRs)
-------------------------------------------------------------------
According to Calin *et al*. ([@B32]), there is a large probability that cancer-associated genomic regions contain miRNA genes. This hypothesis is based on the finding that at least 98 known miRNA genes reside in CAGRs, including 80 miRNAs that are located exactly in minimal regions of LOH or minimal regions of amplification described in a variety of tumors such as lung, breast, ovarian, colon, gastric and hepatocellular carcinoma, as well as leukemias and lymphomas. To investigate this hypothesis, we used the final trained SSCprofiler to search for novel miRNA candidates within these regions. Both positive and negative DNA strands were scanned for a number of regions which represent over 350 MB of the human genome and are known to be deleted or amplified in over 20 different types of cancers ([Supplementary Table S1](http://nar.oxfordjournals.org/cgi/content/full/gkp120/DC1)). Filtering parameters and conservation retrieval were the same as described in the previous section. The scanning procedure (see Materials and Methods section, [Figure 3](#F3){ref-type="fig"}) lasted ∼8 days (real-time) using a parallel PC cluster with 10 dual opteron processors. An example of the SSCprofiler output for this scanning is shown in [Supplementary Figure S2](http://nar.oxfordjournals.org/cgi/content/full/gkp120/DC1). [Figure 6](#F6){ref-type="fig"} shows the conservation of all predicted miRNA candidates for an HMM threshold of 3. As shown in [Figure 6](#F6){ref-type="fig"}A, the majority of predicted miRNA candidates had a high degree of conservation (over 50%) across the seven different species. Moreover, the conservation for each nucleotide position along the 104--nt long predicted sequence dropped significantly near the loop. Figure 6.Conservation of all 10‱511 predicted miRNA candidates (HMM threshold = 3) across seven species. (**A**) Histogram of conserved miRNA candidates. As evident from the figure, the majority of candidates are more than 50% conserved across the seven species. (**B**) Distribution of conserved nucleodites for all candidates along the 104-nt positions of the scanning window. As evident from the figure, there is a large drop in conservation near the loop (middle area) while positions around the loop are higly conserved in a symmetrical way.
Identification of the 98 known miRNAs in the CAGRs regions that were scanned was assessed as a function of the HMM score as shown in [Table 4](#T4){ref-type="table"}. As expected, the number of miRNA gene candidates decreases with increasing HMM score. Consequently, as the HMM score becomes larger, the sensitivity drops while the specificity increases. According to the training and testing procedures discussed previously, the HMM score threshold for which both sensitivity and specificity values were maximized ranged between 1 and 3 ([Table 3](#T3){ref-type="table"}). However, when scanning large genomic sequences multiple false positives tend to accumulate, even for an average specificity value of ∼85% (threshold of 3). Since experimental verification is an expensive and time consuming process, we chose candidates attaining a significantly higher HMM score in order to obtain the most probable miRNA gene candidate. Table 4.Predicted miRNA genes as a function of the HMM score[^1]
Experimental verification of top scoring candidates
---------------------------------------------------
According to sensitivity and specificity measures, the prediction accuracy of SSCprofiler with respect to the identification of novel miRNA genes is very high. However, these statistical evaluation criteria depend highly on the specific data sets used to train and evaluate the computational model. Experimental verification of predicted miRNA genes is the optimal way to assess the model\'s performance. Towards this goal, we experimentally validated a few of our top scoring precursor candidates. The HMM threshold score for selecting these candidates was set according to the following criteria: (i) high enough to decrease the number of false positives and at the same time and (ii) low enough to capture many of the true miRNAs. A threshold of 21 was finally selected, at which 421 candidates were predicted with 65.31% (64/98) accuracy for the true miRNAs. At this threshold, the predicted list included candidates with only partial conservation when compared to higher scoring candidates. The expression of all 421 candidates in HeLa and HepG2 cells was assessed using recent data from a full genome tiling array ([@B21]) which provides a small-RNA expression map. A total of 38 candidates whose expression at the stem region was above a threshold of 200 were retained (see tinted grey in [Table 4](#T4){ref-type="table"}). Of these, only 20 were expressed in HeLa cells. The top four of these 20 candidates, according to their expression value (listed in [Table 5](#T5){ref-type="table"}), were tested experimentally using northern blot analysis on cultured HeLa cells (see 'Materials and methods' section for details). Northern blot analysis concurred with the tiling array expression data in all four of the candidates tested. As shown in [Figure 7](#F7){ref-type="fig"}, all candidates produced specific signals whose size is within the mature miRNA range (19--27 nt) while, in some cases, the pre-miRNA was also detected. Moreover, we found that in all four candidates only one strand of the predicted precursor produced a specific signal, further suggesting that our candidates are likely to be true miRNA precursors. Table 5Candidate miRNA genes verified by northern blot analysis located in minimal deleted regions involved in human cancersCandidateCandidate Information[^a^](#TF1){ref-type="table-fn"}CAGRType of cancerClosest miRNAExpression in HeLa1chr9:123327358-123327460 st−chr9:121153509-128793509Bladder cancermiR-181a; miR-199b1667.52chr5:148958951-148959053 st−chr5:144121683-156051683Prostate cancer aggressivenessmiR-145/miR-143363.52chr5:148958951-148959053 st−chr5:148181683-151101683Myelodysplastic syndromemiR-145/miR-143363.53chr22:40863894-40863996 st+chr22:31530000-43583971Colorectal cancermiR-33a345.03chr22:40863894-40863996 st+chr22:31530000-42193557AstrocytomasmiR-33a345.04chr5:149984684-149984786 st−chr5:144121683-156051683Prostate cancer aggressivenessmiR-145/miR-143264.04chr5:149984684-149984786 st−chr5:148181683-151101683Myelodysplastic syndromemiR-145/miR-143264.0[^2][^3] Figure 7.Northern blot analysis shows a specific signal for all 4 miRNA gene candidates. The let7 probe hybridizes on multiple members of the let7 miRNA gene family, accounting for the three bands shown on the reference membrane. The membranes labeled 1--4 represent the miRNA gene candidates predicted by SSCprofiler in the same order as shown in Table 5. The band patterns of four miRNA gene candidates resemble those of known miRNAs; displaying a band at the range of 19--27 nt. The higher signals represent the unprocessed precursors (range ∼70 nt). Structures for the four miRNA gene candidates as predicted by RNAfold are also shown. The strand on each precursor that produces a signal is shown in red.
Blat analysis ([@B38]) against the human genome provided additional supporting evidence for our four miRNA gene candidates. We found that all candidates are more than 45% conserved across eight other organisms and are located within expressed intergenic regions, consistent with the majority of miRNAs ([@B39],[@B40]). Moreover, Blat search produced 100% identity hits at the level of 20--26 nt in other regions of the genome and secondary structure prediction for the genomic sequences flanking these regions resembled that of true miRNA precursors in five out of nine cases. Since the mature miRNA is the main unit of regulation for the miRNA gene it may be more conserved than the rest of the precursor, suggesting the existence of additional miRNA genes. Blast analysis of the oligonucleotides concurs with the results from the Blat search and reveals significant hits in other regions of the human genome.
Tool comparison
---------------
Finally, to assess the prediction accuracy of our tool compared to existing algorithms, we used the four verified candidates (the predicted precursors) as a query set in four existing miRNA gene prediction tools. Interestingly, all of these tools failed to identify our candidates as likely miRNA genes. MiRRim ([@B19]), ProMir II ([@B41]) (HMM algorithms) and BayesMiRNAfind ([@B12]) (Bayes classifier) were not able to identify any of our four candidates while TripletSVM ([@B17]) (SVM classifier) predicted one out of four candidates (candidate 1). It is important to note that all of these tools are considered as highly accurate with respect to traditional sensitivity/specificity measures, which in some cases outperform that of SSCprofiler ([@B24]). However, only for one tool (ProMir II) the authors performed experimental testing of their predicted miRNA gene candidates.
The most recent of these tools \[MiRRim ([@B19])\] is similar to SSCprofiler as it also uses an HMM algorithm that considers structure and conservation features for predicting novel miRNA genes. The main conceptual differences between the two tools include: ([@B1]) the selection of negative data, ([@B2]) the size of the sliding window and ([@B3]) the ability of the user to take into account information from large scale tiling arrays. A detailed comparison between the two tools is provided in [Supplementary Table S2](http://nar.oxfordjournals.org/cgi/content/full/gkp120/DC1). Briefly, for SSCprofiler both positive and negative sequences are similar in their structure and degree of conservation while for miRRim negative data are filtered according to conservation alone. The latter may bias results and make the discrimination between the two classes an easier task, resulting in higher sensitivity/specificity measures on the training and validation sets but not necessarily on a blind test set. Moreover, differences in the scanning procedure utilized by each tool can affect the total number of predicted miRNA candidates. A larger window such as the one used in miRRim, translates into a substantially smaller search space and consequently a smaller number of genomic regions that could be identified as potential miRNA genes. Thus, the number of predicted candidates is not directly comparable between the two tools. An important advantage of SSCprofiler is that it allows the user to further filter resulting candidates using a large scale tilling array data set ([@B21]). Such a feature is not provided in miRRim. Overall, we believe that the differences described above, together with the sequential consideration of sequence, structure and conservation at the nucleotide level offer an important advantage to our tool compared to existing ones. It is the combination of all of these factors that allowed SSCprofiler to identify four novel miRNA gene candidates, which were experimentally verified but could not be identified by four other prediction tools. We firmly believe that for a computational tool to prove its value, especially for the biology oriented user group it is designed for, it must provide a full prediction pipeline from computational identification to experimental verification for at least a few top scoring candidates.
DISCUSSION
==========
In this study we introduced an efficient miRNA gene prediction tool (SSCprofiler) which is based on Profile HMMs and evaluated its performance against a blind set of recently identified miRNAs as well as via the experimental verification of four top scoring candidates. Our tool is provided both as a user friendly trainable interface and a web-based scanning application which can be used for querying genomic regions. In both cases, the user has a large degree of flexibility in terms of dataset specification and parameter tuning.
Our algorithm works by combining sequence, structure and conservation information taken at the nucleotide level throughout the length of miRNA precursors. We show that multiple feature integration is advantageous with respect to prediction accuracy and argue that this type of combination is more effective than other approaches. Incorporation of expression information for predicted candidates is another important advantage of our tool. The use of full genome tiling array data ([@B21]), which provide a small-RNA expression map in HeLa and HepG2 cells at 5-nt resolution, increases the reliability of model predictions and can be extremely useful when selecting miRNA gene candidates for experimental verification due to the tissue specific expression of miRNA genes.
The effectiveness of SSCprofiler in recognizing human miRNA genes was demonstrated using a blind set of 219 recently identified human miRNAs from the latest version of miRBase (version 12). For an HMM threshold of 1, the method reached a prediction accuracy of 85.84% similar to its training/validation performance (91.3% sensitivity, 78.9% specificity). The tool\'s ability to identify novel miRNA genes located within 350 MB of human cancer-associated genomic regions ([@B32]) was also assessed. For an HMM threshold of 11 (73.44% sensitivity, 96.13% specificity) SSCprofiler predicted a total of 5862 novel miRNA candidates within these regions. Assuming an analogy between CAGRs and the whole human genome, SSCprofiler would predict approximately 58 000 new miRNA genes, in agreement with a recent estimate provided by the study of Miranda *et al*. ([@B40]). However, it should be noted that CAGRs are known to contain a disproportionably large number of miRNAs (over 20% of all miRNAs in miRBase 12.0); therefore an analogy between those regions and the entire human genome might not be valid. Taking into account the high costs of reagents and the time consuming nature of experimental procedures we decided to select candidates that were more likely to be successful. For this reason we used a higher threshold (HMM score of 21) for which fewer candidates were predicted. Of the 421 predicted candidates only 20 were highly expressed in HeLa cells. Northern blot analysis of the top four of these candidates verified the presence of a specific RNA molecule at the miRNA range ([@B19; @B20; @B21; @B22; @B23; @B24; @B25; @B26; @B27]) which strongly suggests the presence of a small noncoding RNA. In future efforts, additional predicted candidates for lower HMM thresholds should be analyzed experimentally in order to obtain a more accurate cut-off value for reliably predicting novel miRNAs.
Our findings regarding miRNA gene identification are in accordance to the uniform system of miRNA annotation ([@B42]). The miRNA biogenesis criterion is satisfied by the prediction of a potential fold-back precursor structure that contains the ∼22-nt miRNA sequence within one arm of the hairpin. The hairpin displays a very low free energy, as predicted by RNAfold and only one stem of the precursor shows a northern blot signal. Our candidates do not contain large internal loops or bulges, particularly large asymmetric bulges and all fall within the miRNA precursor range of ∼60--100 nt reported in animals. Phylogenetic conservation of the whole miRNA precursor sequence for all four candidates across seven other organisms is also observed. The miRNA expression criterion is also met by our candidates. A distinct ∼22-nt RNA transcript is detected by hybridization to a size-fractionated RNA sample by northern blot analysis for all four candidates. In addition, expression of ∼22-nt RNA transcripts from the active stem region of the candidates is observed in HeLa cells using tiling arrays. These criteria provide strong evidence that our top scoring candidates are likely to be true miRNA precursors and consequently, that SSCprofiler is a reliable and efficient tool for predicting novel miRNA genes. Interestingly, a comparison study with other miRNA gene prediction tools reveals that three out of four of our verified miRNA gene candidates are not predicted by four other published tools ([@B12],[@B17],[@B19],[@B41]). Only one out of four tools predicted one-fourth of our verified miRNA candidates ([@B17]). This finding highlights the superior prediction capacity of SSCprofiler and further substantiates its significance as a miRNA gene prediction tool. The tool\'s availability in both a trainable and a web-based scanning version can further facilitate its use as a part of a prediction pipeline for novel miRNA genes, hence allowing for minimization of time, cost and effort.
An important finding of this work is the identification of four novel miRNA gene candidates residing within genomic regions which are implicated in numerous cancers (CAGRs). Although a detailed experimental characterization of the mature miRNA function is pending, these molecules are likely to play an important role in regulating carcinogenesis, possibly by acting as 'oncogenes' or 'tumor suppressors'. The CAGRs corresponding to each miRNA candidate are commonly deleted in various types of cancers ([Table 5](#T5){ref-type="table"}). Deletion of a region containing a miRNA gene prohibits the expression of the functional miRNA. As a result, gene(s) regulated by this miRNA will function uncontrollably, a process which may result in a cascade of events that triggers oncogenesis. Since the mode of action of mature miRNAs usually results in downregulation of targeted genes, one possibility is that our candidates play a tumor suppressor role perhaps by stopping a major tumorigenic turning point in the cell. However, since cumulating data support both a negative and a positive regulatory role for miRNAs ([@B43]) it is hard to predict the effect of these potential miRNAs on their target genes. Target prediction programs can provide a starting point for identifying possible target genes for these candidates, thus providing more insights into their potential role in specific types of cancer. Towards this goal, an RNA--RNA duplex prediction algorithm will be incorporated in future versions of the SSCprofiler. This can be achieved using programs such as RNAcofold ([@B35]) which calculate secondary structures of two RNA sequences in the form of a hybrid duplex. This extension will enable the user to train profile HMMs that can recognize the pairing rules between two hybrid RNA molecules, thus allowing the prediction of new miRNA--mRNA interactions that obey similar rules. Our ultimate goal is to develop a stand alone application which will be able to predict novel miRNA genes as well as their probable targets. Such a tool may provide a more concise biological picture of the pathways and genes regulated by our four novel miRNA gene candidates. Experimental verification will ultimately be needed to characterize the production of a mature miRNA, show that predicted interactions take place in the system of interest and that functional interactions are strongly associated with the emergence of a cancerous phenotype.
SUPPLEMENTARY DATA
==================
[Supplementary Data](http://nar.oxfordjournals.org/cgi/content/full/gkp120/DC1) are available at NAR Online
FUNDING
=======
Action 8.3.1 (Reinforcement Pro-gram of Human Research Manpower---'PENED 2003', \[03EΔ842\]) of the operational program 'competitiveness'of the Greek General Secretariat for Research and Technology; INFOBIOMED NoE \[FP6-IST-2002-507‱585\] EU funded project; Marie Curie Outgoing Fellowship \[PIOF-GA-2008-219‱622\] of the European Commission. Funding for open access charge: PENED 2003 \[03EΔ842\].
*Conflict of interest statement*. None declared.
Supplementary Material
======================
###### \[Supplementary Data\]
We would like to thank Prof. Angelos Bilas for the use of his PC cluster.
[^1]: For each HMM score in the range of 3--41 (first column), the table shows: (a) the number of predicted miRNA precursors (second column); (b) the number of true precursors included in the predicted list as a function of all true precursors within CAGRs (third column); (c) the number of predicted candidates (fourth column) and true miRNAs (fifth column) that passed the 200 expression threshold in HeLa and/or HepG2 cells and (d) the respective sensitivity and specificity values (sixth column). The 421 sequences that were predicted for an HMM threshold of 21 were selected for further processing, which is tinted grey in the table.
[^2]: Positions are according to build 35 (hg17) version of the Human Genome at <http://genome.ucsc.edu/>
[^3]: ^a^Chromosomal location and strand (st+ or st−).
| {
"pile_set_name": "PubMed Central"
} |
All relevant data are within the paper.
Introduction {#sec001}
============
Abscisic acid (ABA) is a phytohormone regulating fundamental physiological functions in plants \[[@pone.0140588.ref001], [@pone.0140588.ref002]\]. ABA is also an endogenous hormone in humans, regulating different cell responses and functions, including activation of innate immune cells and stimulation of insulin release and glucose uptake \[[@pone.0140588.ref003]--[@pone.0140588.ref006]\]. The signaling cascade of ABA in mammalian cells involves ABA binding to lanthionine synthetase C-like protein 2 (LANCL-2) and cAMP production \[[@pone.0140588.ref007]--[@pone.0140588.ref009]\]. Pro-inflammatory stimuli induce ABA production and release from human granulocytes, monocytes, keratinocytes and fibroblasts \[[@pone.0140588.ref003], [@pone.0140588.ref010]--[@pone.0140588.ref012]\] and ABA stimulates cell-specific functional activities in granulocytes (chemotaxis, phagocytosis, release of NO and reactive oxygen species), monocytes (chemotaxis, release of TNF-α, monocyte chemoattractant protein-1, metalloprotease 9 and prostaglandin E2), vascular smooth muscle cells (cell proliferation and migration), keratinocytes (release of NO, PGE2, and TNF-α) and fibroblasts (migration) \[[@pone.0140588.ref003], [@pone.0140588.ref010]--[@pone.0140588.ref012]\].
Several observations indicate that ABA is also involved in the regulation of glucose homeostasis in mammals as an endogenous hormone: i) ABA is released by human and murine pancreatic β-cells in response to high glucose, and nanomolar ABA triggers glucose-independent and potentiates glucose-dependent insulin secretion from these cells \[[@pone.0140588.ref004]\]; ii) oral glucose administration increases plasma ABA concentration (\[ABA\]~p~) in healthy human subjects \[[@pone.0140588.ref005]\]; iii) ABA stimulates glucose uptake by rodent adipocyte and myoblast cell lines \[[@pone.0140588.ref005]\]. In line with these data, Guri et al. observed that a chronic oral administration of exogenous ABA reduced the fasting plasma glucose concentration and ameliorated glucose tolerance in leptin receptor-deficient (db/db) mice \[[@pone.0140588.ref013]\].
Interestingly, the increase of \[ABA\]~p~ in response to an oral glucose load in healthy subjects was less consistently observed when the same subjects were administered glucose intravenously \[[@pone.0140588.ref005]\]. Oral, but not intravenous, glucose administration is followed by the release of the incretin glucagon-like peptide 1 (GLP-1), a gastrointestinal hormone secreted by enteroendocrine L-cells in response to nutrients, hormones and neurotransmitters. GLP-1 stimulates insulin and inhibits glucagon release, thereby contributing to the regulation of glycemia \[[@pone.0140588.ref014]--[@pone.0140588.ref016]\]. A possible explanation for the different effect of intravenously or orally administered glucose on \[ABA\]~p~ could come from the observation that GLP-1 stimulates ABA release by insulin-secreting cells, both in the presence of low- (2 mM) or of high- (25 mM) glucose concentrations \[[@pone.0140588.ref005]\].
In this study, we investigated whether ABA affects GLP-1 secretion by enteroendocrine cells, a process known to be regulated by the \[cAMP\]~i~ \[[@pone.0140588.ref014]\], thereby addressing the possible existence of a positive feed-back mechanism between ABA and GLP-1, regulating glucose homeostasis.
Methods {#sec002}
=======
hNCI-H716 cell culture and GLP-1 secretion studies {#sec003}
--------------------------------------------------
The human L cell line hNCI-H716, derived from a poorly differentiated adenocarcinoma of the cecum, was obtained from the American Type Culture Collection (Manassas, VA). Cells were grown in suspension in RPMI-1640 (Sigma, Milano, Italy), supplemented with 10% fetal bovine serum (FBS), 50 U/ml penicillin and 50 μg/ml streptomycin.
For GLP-1 secretion assays, a protocol similar to the one described in \[[@pone.0140588.ref017]\] was followed: briefly, hNCI-H716 cells were seeded on Matrigel matrix (Becton Dickinson, Bedford, MA), at the density of 2x10^5^ cells/well in 24-well plates, in DMEM medium supplemented with 10% FCS, 50 U/ml penicillin, and 50 μg/ml streptomycin. After 48 h, cells were washed in Hank's Balanced Salt Solution (HBSS) and then incubated for 2 h in Krebs Ringer Hepes buffer (KRH buffer: 130 mM NaCl, 5 mM KCl, 1.3 mM CaCl~2~, 25 mM HEPES, 10 mM Na~2~HPO~4~, 1.3 mM MgSO~4~, 0.2% BSA), in the presence or absence of the different treatments: glucose (200 mM), or glutamine (10 mM), or ABA (0.1, 10 or 200 μM).
After treatment~~s~~, medium and cells were collected separately: GLP-1 content in the supernatant was analyzed by GLP-1 Total ELISA Kit (Merck Millipore, Vimodrone, MI, Italy); total protein content in cells was analyzed by Bradford assay (Bio-Rad, Milano, Italy).
Quantitative real time-PCR {#sec004}
--------------------------
Total mRNA was extracted from hNCI-H716 using Qiazol (Qiagen, Milan, Italy) according to the manufacturer\'s instructions. Quality and quantity of RNA were analysed using a NanoDrop spectrophotometer (Nanodrop Technologies, Wilmington, DE). The cDNA was synthesized by the iScriptTM cDNA Synthesis Kit (Bio-Rad, Milan, Italy) starting from 1 μg of total RNA. PCR primers were designed through Beacon Designer 2.0 Software and their sequences were as indicated in [Table 1](#pone.0140588.t001){ref-type="table"}.
10.1371/journal.pone.0140588.t001
###### Primers.
{#pone.0140588.t001g}
Human gene Sequence, 5'-3'
-------------- ---------------------------- --------------------------
**GLP-1** Forward `GCTGAAGGGACCTTTACCAGT`
Reverse `CCTTTCACCAGCCAAGCATG`
**GLUCAGON** Forward `ATTCACAGGGCACATTCACCA`
Reverse `GGTATTCATCAACCACTGCAC`
**ACTIN** Forward `GCGAGAAGATGACCCAGATC`
Reverse `GGATAGCACAGCCTGGATAG`
**HPRT-1** Forward `GGTCAGGCAGTATAATCCAAAG`
Reverse `TTCATTATAGTCAAGGGCATATCC`
qPCR was performed in an iQ5 real-time PCR detection system (Bio-Rad) using 2× iQ Custom Sybr Green Supermix (Bio-Rad). Values were normalized on mRNA expression of human β-actin and HPRT. Statistical analysis of the qPCR was performed using the iQ5 Optical System Software version 1.0 (Bio-Rad) based on the ^2−^ΔCt method \[[@pone.0140588.ref007]\]. The dissociation curve for each amplification was analysed to confirm absence of unspecific PCR products. Experiments were repeated three times in triplicate.
Measurement of the intracellular cAMP concentration {#sec005}
---------------------------------------------------
hNCI-H716 cells were seeded at the density of 5x10^5^/well in 12-well, Matrigel matrix-coated plates. After 24 h, cells were washed with HBSS, pre-incubated for 10 min in HBSS containing 10 μM IBMX, an inhibitor of phosphodiesterases, and then stimulated with 10 mM glutamine or 200 μM ABA for 2.5 and 5 min.
Supernatant was removed and cells were lysed in 0.6 M PCA. Intracellular cAMP content was evaluated by EIA (Cayman, Ann Arbor, MI, USA) on neutralized extracts \[[@pone.0140588.ref018]\].
Vector construction {#sec006}
-------------------
The full length LANCL2 cDNA was amplified by PCR using cDNA obtained with reverse transcription of total RNA from human granulocytes and using the following primers: `5’-CACCATGGGCGAGACCATGTCAAAG-AG-3’`(foward); `5’-ATCCCTCTTCGAAGAGTCAAGTTC-3’` (reverse).
The PCR was performed in 25 μl containing undiluted reaction buffer, 200 μM dNTP, 5 pmol of primers and using 1.25 U of Herculase HotStart DNA polymerase. The PCR reaction profile was 1 cycle at 94°C for 2 min, 35 cycles at 94°C for 15 s, 62°C for 30 s and 72°C for 1 min with a final extension for 5 min at 72°C. The PCR product was purified with Nucleospin^®^ Extract Kit (Macherey-Nagel) and cloned into pcDNA3.1/V5-His-TOPO^©^. This vector allows the synthesis of the recombinant protein as a C-terminal fusion to the V5 epitope and a Histidine tag. The LANCL2 plasmid was purified using PureLink™ HiPure Plasmid Filter Kit (Invitrogen) and sequenced by TibMolbiol (Genova, Italy).
LANCL2 overexpression {#sec007}
---------------------
hNCI-H716 cells were transfected in parallel with pcDNA3.1(+) (control plasmid) or with the plasmid containing the full-length LANCL2 cDNA, LANCL2-pcDNA3.1(+) (LANCL2 plasmid). Transient transfection of hNCI-H716 cells (1.5x10^6^) was performed using the Nucleofector System (Amaxa GmbH, Köln, Germany), program X-005, solution T, with 3 μg LANCL2-plasmid or control plasmid. hNCI-H716 cells were then resuspended in DMEM and seeded in Matrigel-coated 24-well plates. Experiments were performed 48 h after transfection.
Western blot analysis {#sec008}
---------------------
hNCI-H716 cells (2.5x10^5^) were lysed in 50 μl HES lysis buffer (20 mM Hepes, pH 7.4, 1 mM EDTA, 250 mM sucrose) containing a protease inhibitor cocktail (Sigma), and LANCL2 expression was analyzed by Western blot, using a monoclonal antibody against LANCL2 \[[@pone.0140588.ref019]\]. LANCL2 expression was normalized on vinculin levels, detected with a goat polyclonal antibody against actin (Santa Cruz Biotechnology, Dallas, TX). Appropriate HRP-conjugated secondary antibodies (Cell Signaling, Danvers, MA) and enhanced chemiluminescence reagents (GE Healthcare, Little Chalfont, Buckinghamshire, UK) were used to detect antigens after transfer to a nitrocellulose membrane.
In vivo experiments {#sec009}
-------------------
Two-months old female Wistar rats weighing 160 to 198 g (obtained from Charles River Laboratories Italia, Calco, LC, Italy) were housed singly under a 12 h/12 h light/dark cycle under free feeding conditions, in temperature- and humidity-controlled rooms.
After an overnight fast, the DPP4 inhibitor Sitagliptin (Januvia^®^, 10 mg/Kg) \[[@pone.0140588.ref020]\], was orally administered 30 min prior to ABA (50 mg/Kg) or vehicle (water) gavage. After anesthesia with ketamine/xylazine, blood samples were collected at 0, 20, 40 and 60 min by orbital sinus bleeding in heparin and plasma aliquots were stored at -20°C.
The dose of ABA was chosen based on the effect of dietary ABA supplementation \[[@pone.0140588.ref013]\].
In other experiments, where animals were not pre-treated with Sitagliptin, GLP-1 concentration was also evaluated in the portal vein blood, as in \[[@pone.0140588.ref021]\], 10 min after intragastric vehicle or ABA administration.
Measurements of plasma GLP-1, glucose, insulin and ABA {#sec010}
------------------------------------------------------
GLP-1 concentrations were determined by ELISA (Merck Millipore; the kit detects the total GLP-1 levels). Glycemia was measured with a glucometer (Bayer, Milano, Italy) and insulinemia by ELISA (Bertin-Pharma, Montigny, France). ABA plasma concentrations were determined by ELISA, as in \[[@pone.0140588.ref005]\].
Ethics statement {#sec011}
----------------
Animal rearing conditions were consistent with the guidelines of the Italian Ministry of Health and the study was approved by the IRCCS AOU San Martino-IST Ethical Committee (Genova, Italy).
Results {#sec012}
=======
ABA stimulates GLP-1 secretion {#sec013}
------------------------------
hNCI-H716 cells were challenged for 2 h with different ABA concentrations and GLP-1 levels were measured in the supernatants. The basal GLP-1 concentration was 347±106 pM. As shown in [Fig 1A](#pone.0140588.g001){ref-type="fig"}, 200 μM ABA approximately doubled the extent of the GLP-1 secretion. 10 μM ABA was sufficient to trigger a statistically significantly higher GLP-1 release, compared to the untreated control. No stimulation of GLP-1 secretion was obtained in the presence of 100 nM ABA. The calculated EC50 for the ABA-induced GLP-1 release was 23±3 μM (not shown). To compare the effect of ABA on GLP-1 release with that of other secretagogues, cells were also incubated in the presence of 200 mM glucose or 10 mM glutamine \[[@pone.0140588.ref022], [@pone.0140588.ref023]\]: GLP-1 secretion was increased by approximately 1.4-fold with both stimuli ([Fig 1A](#pone.0140588.g001){ref-type="fig"}).
{#pone.0140588.g001}
Interestingly, ABA treatment also significantly increased preproglucagon mRNA levels, as demonstrated by qPCR using two different sets of primers, specific for GLP-1 and glucagon, respectively, yielding a similar result ([Fig 1B](#pone.0140588.g001){ref-type="fig"}).
The ABA-induced GLP-1 secretion is mediated by a cAMP-dependent mechanism {#sec014}
-------------------------------------------------------------------------
In different human cell types, the cell-specific ABA-induced response is mediated by an increase of the second messenger cAMP \[[@pone.0140588.ref003], [@pone.0140588.ref004], [@pone.0140588.ref007], [@pone.0140588.ref009], [@pone.0140588.ref024]\], and by the consequent PKA activation \[[@pone.0140588.ref003], [@pone.0140588.ref004], [@pone.0140588.ref025]\]. Since GLP-1 release is regulated by the \[cAMP\]~i~ \[[@pone.0140588.ref014]\], we verified whether ABA was able to induce an increase of the \[cAMP\]~i~ in hNCI-H716 cells. As a positive control, cells were incubated with glutamine, which is known to determine a \[cAMP\]~i~ increase in hNCI-H716 cells \[[@pone.0140588.ref026]\]. As shown in [Fig 2A](#pone.0140588.g002){ref-type="fig"}, a 2.5-min incubation in the presence of 200 μM ABA induced a 2-fold increase of the \[cAMP\]~i~, while 10 mM glutamine increased the \[cAMP\]~i~ approximately 1.4-fold.
![ABA induces the increase of the \[cAMP\]~i~ in hNCI-H716 cells.\
(A) hNCI-H716 cells were incubated for the indicated time in the absence or presence of 200 μM ABA (squares), or of 10 mM glutamine (rhombi); \[cAMP\]~i~ was then measured on cell extracts. Data are mean±SD of at least 3 different experiments; \*, p\<0.05 compared with untreated cells; \#, p\<0.05 compared to glutamine-treated cells (for the same time). (B) hNCI-H716 cells were transfected with an empty plasmid (control) or with a LANCL2-containing plasmid (LANCL2). After 48 h from transfection, cells were lysed and a Western blot analysis was performed using an anti-LANCL2 monoclonal antibody \[[@pone.0140588.ref019]\]; a representative blot is shown, confirming LANCL2 overexpression after transfection. LANCL2 expression was normalized on vinculin levels. (C) After 48 h from transfection, cells were stimulated for 2.5 min in the absence or presence of 200 μM ABA. \[cAMP\]~i~ was measured on cell extracts and data, expressed as fold increase over values in unstimulated cells, are expressed as mean±SD of at least 3 different experiments; basal cAMP values were not significantly different upon transfection. \*, p\<0.05 compared to control. (D) After 48 h from transfection, cells were incubated for 2 h in the absence or presence of 200 μM ABA. GLP-1 levels in the culture media were then estimated with an ELISA kit. Data, expressed as fold increase over values in unstimulated cells, are expressed as mean±SD of at least 3 different experiments. \*, p\<0.05 compared to untreated cells.](pone.0140588.g002){#pone.0140588.g002}
In mammalian cells, the ABA-induced cAMP increase is mediated by the protein LANCL2 \[[@pone.0140588.ref007], [@pone.0140588.ref009]\]. hNCI-H716 cells were transfected by electroporation with an empty plasmid, or with a plasmid containing the full-length cDNA for human LANCL2. LANCL2 overexpression, confirmed by Western blot analysis with a specific monoclonal antibody ([Fig 2B](#pone.0140588.g002){ref-type="fig"}), was accompanied by a significant increase in ABA-induced cAMP accumulation, as compared to cells transfected with an empty plasmid ([Fig 2C](#pone.0140588.g002){ref-type="fig"}), as well as by a significant increase in ABA-induced GLP-1 release ([Fig 2D](#pone.0140588.g002){ref-type="fig"}). The ABA-induced \[cAMP\]~i~ increase and GLP-1 release were approximately 1.4-fold in cells transfected with the empty plasmid (control bars in [Fig 2D](#pone.0140588.g002){ref-type="fig"}), and not 2-fold as observed in untransfected cells (Figs [1A](#pone.0140588.g001){ref-type="fig"} and [2A](#pone.0140588.g002){ref-type="fig"}), indicating that cell responsiveness was slightly affected by the transfection procedure *per se*.
In order to verify whether the ABA-induced \[cAMP\]~i~ increase mediates the ABA-stimulated GLP-1 release, hNCI-H716 cells were pre-incubated in the presence of a specific adenylyl cyclase inhibitor (2′,3′-Dideoxyadenosine), or a cell permeable PKA inhibitor: both inhibitors abrogated the GLP-1 release stimulated by 200 μM ABA ([Fig 1A](#pone.0140588.g001){ref-type="fig"}).
ABA increases plasma GLP-1 in rats {#sec015}
----------------------------------
First, we examined the effect of a single-dose oral administration of ABA (at 50 mg/Kg) on GLP-1 levels in normal rats (6 animals per experimental group) pre-treated with Sitagliptin. 20 min after ABA administration, plasma GLP-1 (GLP-1p) increased by approximately 50%, whereas the vehicle alone had no effect on GLP-1p levels ([Fig 3A](#pone.0140588.g003){ref-type="fig"}). The area under the curve of GLP-1p (GLP-1p AUC) over the entire time frame was calculated from GLP-1p values relative to time zero: the GLP-1p AUC was significantly higher in the ABA-treated compared to the control animals ([Fig 3B](#pone.0140588.g003){ref-type="fig"}).
{#pone.0140588.g003}
GLP-1 levels also significantly increased in the portal vein blood of rats not pre-treated with Sitagliptin 10 min after ABA administration ([Fig 3A](#pone.0140588.g003){ref-type="fig"}, inset), indicating that ABA alone is capable of increasing plasma GLP-1. ABA concentration in the portal vein blood was in the low nM (4.2±1.9 nM) range in the vehicle-treated animals and in the μM range (3.9±0.4 μM) in the ABA-treated animals.
The observation that ABA induced an increase of GLP-1p, together with the fact that exogenous ABA is known to directly stimulate insulin release from β-cells *in vitro* \[[@pone.0140588.ref004]\], prompted us to measure insulin levels in the ABA-treated rats. As shown in [Fig 3C and 3D](#pone.0140588.g003){ref-type="fig"}, insulinemia indeed significantly increased after ABA administration and the plasma insulin AUC was consequently higher in the ABA-treated than in the vehicle-treated group.
Glycemia was slightly increased in vehicle-treated animals: this increase was not observed upon oral ABA administration ([Fig 3E and 3F](#pone.0140588.g003){ref-type="fig"}). The increase of glycemia observed in the control animals can be attributed to anesthesia: indeed, ketamine/xylazine have been shown to induce hyperglycemia in fed rats and, to a lower extent, also in fasted animals \[[@pone.0140588.ref027]\], as in our experimental protocol.
In conscious rats, oral ABA administration at the same dose used in the anesthetized animals (50 mg/Kg) resulted in a slight, yet significant reduction of blood glucose after 60 min (81±6 mg/dL, n = 6) compared with time zero values (92±10, n = 12, p = 0.03) and with values measured at the same time point in the vehicle-treated controls (99±14, n = 6; p = 0.02).
Discussion {#sec016}
==========
We had previously demonstrated that GLP-1 stimulates ABA release from β-pancreatic cells \[[@pone.0140588.ref005]\]. In this study, we show that ABA can induce GLP-1 release, indicating a positive feed-back between these two molecules, possibly relevant to glycemia regulation.
The molecular mechanism by which ABA induces GLP-1 release by hNCI-H716 cells is similar to the one described in several other cell systems (including immune cells, β-pancreatic cells and endothelial cells), i.e. through the cAMP/PKA axis \[[@pone.0140588.ref003], [@pone.0140588.ref004], [@pone.0140588.ref006], [@pone.0140588.ref007], [@pone.0140588.ref009], [@pone.0140588.ref024], [@pone.0140588.ref025]\]. Indeed, this signaling pathway is known to regulate GLP-1 release also in response to other stimuli, such as glucose and glutamine \[[@pone.0140588.ref014], [@pone.0140588.ref015], [@pone.0140588.ref026]\].
In the *in vivo* experiments, we chose to administer a dose of ABA of 50 mg/Kg, based on the results obtained by Guri et al. \[[@pone.0140588.ref013]\], showing that dietary ABA at 100 mg/Kg, introduced over a 24-h period, was effective in reducing glycemia in db/db mice fed a high-fat diet. We hypothesized that a smaller dose could also be effective, if bolus-administered by gavage.
The plasma insulin increase upon ABA administration was expected, based on the *in vitro* effect of ABA on insulinoma cells and on murine and human β-cells: exogenous ABA, added at concentrations in the low nM range, stimulated insulin secretion both in the presence and absence of glucose \[[@pone.0140588.ref004]\]. The plasma ABA concentration measured in the rats 10 min after oral ABA administration was in the micromolar range and could thus be responsible for the observed increase of plasma insulin ([Fig 3C](#pone.0140588.g003){ref-type="fig"}). The increase of plasma GLP-1 peaked at 20 min ([Fig 3A](#pone.0140588.g003){ref-type="fig"}), preceding the insulin increase (which conversely was maximal at 40 min, [Fig 3C](#pone.0140588.g003){ref-type="fig"}). This timing of events suggests that ABA stimulated the release of both GLP-1 and insulin. Since GLP-1 can stimulate ABA release from β-cells *in vitro*, both in the presence or absence of glucose \[[@pone.0140588.ref005]\], one might envisage a feed-back mechanism whereby GLP-1 contributes to stimulate endogenous ABA release, which in turn further stimulates insulin secretion.
In control animals, glycemia was significantly higher at both 20 and 40 min compared with time zero. The increase of glycemia observed in the control animals might be attributed to anesthesia, as described by Saha JK et al \[[@pone.0140588.ref027]\], who showed that ketamine and xylazine significantly altered glycemia in fed and, to a much lower extent, also in fasted rats, as occurred in our study ([Fig 3E](#pone.0140588.g003){ref-type="fig"}, white squares).
In ABA-treated rats, the increase of glycemia was not observed, neither at 20 nor at 40 min ([Fig 3E](#pone.0140588.g003){ref-type="fig"}, black squares): while, at 40 min, a significant increase of insulinemia ([Fig 3C](#pone.0140588.g003){ref-type="fig"}) may have contributed to glycemia control, at 20 min insulinemia was not (yet) increased. Thus, two mechanisms may be responsible for the maintenance of normal blood glucose levels in the ABA-treated, anesthetized animals: i) an "early" (0--20 min), insulin-independent glycemia lowering effect of ABA, followed by, ii) a "late" (20--40 min) glycemia reducing effect, attributable to the increase of insulinemia. We previously reported that ABA can trigger glucose uptake in myocytes and in adipocytes \[[@pone.0140588.ref005]\], to a similar extent as that observed with insulin at the same concentration (i.e. 100 nM). Thus, the normalization of glycemia observed in the ABA-treated animals compared with the controls could result from stimulation by ABA of both glucose transport and insulin release. The plasma concentration of ABA measured 10 min after its administration, which was in the micromolar range, was markedly higher than the one capable of stimulating myoblast glucose uptake *in vitro* \[[@pone.0140588.ref005]\].
Type 2 diabetes treatments targeting the GLP-1 axis by either inhibiting its clearance by DPP4 or using GLP-1 mimetics \[[@pone.0140588.ref015]\] are currently used. More recently, treatments aimed at stimulating GLP-1 release from L cells have been considered as an alternative approach and our finding that ABA increases GLP-1 release may indicate ABA and/or ABA analogs activating LANCL2 as potential anti-diabetic treatments, alone or in combination with DPP4 inhibitors. Indeed, detection of high GLP-1 levels in the portal vein upon ABA administration demonstrates that oral ABA alone is capable of increasing plasma GLP-1 ([Fig 3A](#pone.0140588.g003){ref-type="fig"}). Identifying LANCL2-activating compounds might indeed prove a successful strategy, in view of the multiple anti-diabetic effect that they might trigger. Indeed LANCL2 can: i) mediate the ABA-induced insulin release \[[@pone.0140588.ref007]\]; ii) facilitate Akt phosphorylation \[[@pone.0140588.ref028]\], and increase GLUT4-mediated glucose uptake \[[@pone.0140588.ref005]\], which is an Akt-dependent mechanism \[[@pone.0140588.ref029]\]; iii) mediate the ABA-stimulated GLP-1 release from L cells ([Fig 2D](#pone.0140588.g002){ref-type="fig"}). The potential role of LANCL2 as a new drug target in diabetes has been also suggested by other authors \[[@pone.0140588.ref030]\] and efforts at discovering LANCL2-targeting drugs have been reported \[[@pone.0140588.ref031], [@pone.0140588.ref032]\].
Besides its effects on glycemia regulation, GLP-1 also exerts protective effects on the cardiovascular system \[[@pone.0140588.ref014]\]: thus, ABA administration, resulting in GLP-1 release, might be beneficial also in this respect, together with improving glycemic homeostasis. Indeed, ABA administration has been reported to improve atherosclerosis in ApoE^-/-^ mice \[[@pone.0140588.ref024]\].
So far, it is not possible to exclude that, when administered *in vivo*, ABA can also trigger GLP-1 release from organs/tissues other than L-type cells, e.g. α-pancreatic cells \[[@pone.0140588.ref015]\]. Moreover, it remains to be defined whether ABA stimulates GLP-1 release acting on the intestinal lumenal side of L-cells, like a nutrient, or whether plasma ABA can stimulate GLP-1 secretion also acting as a hormonal stimulus on the vascular side of L-cells, as happens for glucose \[[@pone.0140588.ref014]\]. Indeed, GLP-1 release is evoked in response to multiple paracrine, neural and hormonal stimuli \[[@pone.0140588.ref014], [@pone.0140588.ref015]\]. Finally, future studies should explore the effect on GLP-1 release by ABA administered together with nutrients.
We thank Prof. Emilia Turco (MBC, Torino, Italy) for kindly providing the monoclonal antibody against vinculin and LANCL2.
ABA
: abscisic acid
\[ABA\]~p~
: plasma ABA concentration
LANCL2
: lanthionine synthetase C-like protein 2
GLP-1
: glucagon-like peptide 1
FBS
: fetal bovine serum
HBSS
: Hank's Balanced Salt Solution
KRH
: Krebs Ringer buffer
PCA
: perchloric acid
qPCR
: quantitative real time PCR
GLP-1p
: plasma GLP-1
AUC
: area under the curve
DPP4
: dipeptidyl-peptidase IV
[^1]: **Competing Interests:**The authors have declared that no competing interests exist.
[^2]: Conceived and designed the experiments: SB MM ADF EZ. Performed the experiments: SB EM MM GS LS CF VB LE. Analyzed the data: SB EM EZ. Wrote the paper: SB ADF EZ.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction and Epidemiological Data {#sec1-ijms-21-03377}
========================================
In December 2019, the Chinese Government officially announced a severe form of pneumonia caused by a new coronavirus. It started in Wuhan, in the province of Hubei, and spread rapidly throughout China and then all over the world.
The World Health Organization (WHO) named the syndrome CoronaVirus Disease-2019 (COVID-19), but it was later renamed "severe acute respiratory syndrome" (SARS) Coronavirus (CoV)-2-related (SARS-CoV-2) by the coronavirus Study Group of the International Committee on Virus Taxonomy \[[@B1-ijms-21-03377],[@B2-ijms-21-03377]\]. SARS-CoV-2 is one of the seven beta coronaviruses belonging to the coronavirus family \[[@B3-ijms-21-03377],[@B4-ijms-21-03377]\], which are common in humans and other mammals \[[@B5-ijms-21-03377]\].
The WHO General Director, Tedros Adhanom Ghebreyesus, declared this infection pandemic in the press conference on 11 March 2020 (at [www.who.int/emergencies](www.who.int/emergencies)).
Although most human coronavirus infections are mild, before the current COVID-19 two severe coronavirus outbreaks affected humans in the past two decades: (1) the severe acute respiratory syndrome (SARS) that was caused by the SARS-CoV virus in 2002 \[[@B6-ijms-21-03377],[@B7-ijms-21-03377],[@B8-ijms-21-03377]\] and (2) the Middle East respiratory syndrome (MERS) that was caused by MERS-CoV in 2012 \[[@B9-ijms-21-03377],[@B10-ijms-21-03377]\], being responsible for more than 10,000 cumulative infected cases with 10% and 37% mortality rates, respectively ([www.who.int/csr/sars](www.who.int/csr/sars) and [www.who.int/emergencies/mers-cov](www.who.int/emergencies/mers-cov)). The SARS-CoV and SARS-CoV-2 strains use the same region, referred to as spike, to bind the same receptor, namely the angiotensin converting enzyme-2 \[[@B11-ijms-21-03377],[@B12-ijms-21-03377]\]. Their spike regions differ in terms of only few amino acids \[[@B13-ijms-21-03377],[@B14-ijms-21-03377]\].
Since its outbreak, the SARS-CoV-2 virus infection has spread rampantly, infecting 2,029,930 confirmed cases worldwide to date, and causing 136,320 deaths, in more than 200 countries (<https://gisanddata.maps.arcgis.com>, 1 April 2020).
At the time we write, the USA situation dominates the world scenario, with 639,644 clinically and laboratory confirmed cases and 30,985 deaths, followed by Spain (180,659 cases) and Italy, with 165,155 confirmed cases and the highest number of deaths, now 21,6454, then France, Germany, the United Kingdom, and China, with a prevalence rate between 0.2--0.3%. In Europe, of 978,632 confirmed cases, 84,628 have died (8.6% case fatality rate and 1,6 mortality rate) (<https://gisanddata.maps.arcgis.com/>, 16 April 2020).
A report on 30 March 2020, related to the 10,026 Italians who had died of coronavirus infection (<https://www.epicentro.iss.it/coronavirus/>), described a median age of 78 (range 30--100, InterQuartile Range---IQR 73-85; 30.8% females, median age 82). The median age was 15 years higher than that of the general SARS-CoV-2-positive population (median age 63 years). Of these 10,026 patients, 74% were aged between 74 and 89 years. Only 112 (1.1%) were younger than 50 years old and 23 patients were under 40. The latter included 15 patients with serious co-existing pathologies, six with no other comorbidities, while no clinical records were available for the remaining two patients.
In a subgroup of 909 (of the 10,026) deceased patients, for whom complete clinical records were available, 51.7% had more than three diseases, including arterial hypertension (73.5%), diabetes mellitus (31.5%), ischemic heart disease (27.4%), chronic renal failure (23.8%), atrial fibrillation (23%), active cancer in the last five years (16.5%), and heart failure (16.4%). In this group, death was caused by the acute respiratory distress syndrome (ARDS) (96.5% of cases) that was associated to acute renal failure (25.7%), acute cardiac injury (11.6%), and/or superinfections (11.2%) ([www.epicentro.iss.it/coronavirus](www.epicentro.iss.it/coronavirus)).
2. Clinical Features {#sec2-ijms-21-03377}
====================
Clinical presentations of COVID-19 range from asymptomatic (81.4%), through mildly symptomatic with or without seasonal flu-like symptoms, to severe pneumonia (13.9%) \[[@B15-ijms-21-03377]\]. Usually, respiratory problems manifest about one week after virus entry and dyspnea ranges from effort dyspnea to dyspnea occurring at rest \[[@B16-ijms-21-03377],[@B17-ijms-21-03377]\]. Patients with dyspnea can revert to an asymptomatic phase or progress to ARDS, requiring positive pressure oxygen therapy and intensive care therapy \[[@B18-ijms-21-03377]\] in 17--19.6% of symptomatic patients \[[@B19-ijms-21-03377],[@B20-ijms-21-03377]\]. ARDS, in turn, can progress to multi-organ failure \[[@B21-ijms-21-03377]\] and, in this phase, disseminated intravascular coagulation (DIC) can also be observed \[[@B22-ijms-21-03377]\]. The main cause of death worldwide in infected patients is a combination of both ARDS and DIC in 13.9% of cases \[[@B23-ijms-21-03377]\].
The ARDS-stage is preceded by a marked rise of inflammatory parameters, such as serum ferritin, C-reactive protein (CRP) levels, d-dimers, and the erythrocyte sedimentation rate, and it is characterized by severe edema of the alveolar wall and lung interstices, responsible for the ground glass picture seen at chest high resolution CT scan. When DIC occurs, d-dimers levels further increase, while increased liver and skeletal muscle enzymes and/or serum urea and creatinine indicate ongoing multiorgan failure.
Clinical recovery is possible at any of the above-mentioned stages, and it is generally associated to a complete clearance of the virus, rather than to its persistence. In the latter, rarer condition, according to preliminary studies \[[@B24-ijms-21-03377]\], the virus can be detected for a period of up to one month. Lastly, clinical recovery from the ARDS stage is rarely achieved (2.9%) \[[@B16-ijms-21-03377]\].
Thus, at a certain stage of the infection, in some individuals the virus becomes a powerful stimulator of inflammation at alveolar levels, leading to an alveolar capillary leak-like syndrome (CLLS), with edema, and a marked impairment of gas exchange requiring assisted ventilation.
3. Pathology and Laboratory Evidence of CLLS and Inflammation {#sec3-ijms-21-03377}
=============================================================
Autopsy studies of the lungs of patients in advanced stages of the disease are not yet available. However, histology revealed a pattern of alveolar wall edema, proteinaceous exudates, and focal reactive hyperplasia of pneumocytes with vascular congestion, patchy inflammatory cellular infiltration, and multinucleated giant cells in two patients who underwent lung lobectomy for adenocarcinoma and subsequently tested positive for SARS-CoV-2 \[[@B25-ijms-21-03377]\]. Post-mortem biopsies of a fifty-year-old Chinese patient who died of SARS-CoV-2 with severe acute respiratory syndrome showed evident desquamation of pneumocytes, bilateral pulmonary edema with hyaline membrane formation, interstitial mononuclear inflammatory infiltration, and multinucleated syncytial cells with atypical pneumocytes \[[@B26-ijms-21-03377]\]. These histological findings resemble those found at histological post-mortem examination in patients during the 2002 SARS infection \[[@B27-ijms-21-03377]\], as regards diffuse alveolar wall and airspace edema, and the presence of multinucleated cells. They suggest a common mechanism(s) underlying the clinical picture of SARS \[[@B27-ijms-21-03377]\] and COVID-19 \[[@B25-ijms-21-03377]\], culminating in ARDS, which is likely mediated by massive cytokines release.
Indeed, high levels of several pro-inflammatory cytokines, including IL-6, IL-1, TNF-α, have been demonstrated in advanced stage patients \[[@B18-ijms-21-03377],[@B28-ijms-21-03377]\], supporting the hypothesis that the onset of ARDS is driven by pro-inflammatory cytokines, which are responsible for the histological changes and clinically full-blown ARDS. Among pro-inflammatory cytokines, IL-6 appears to be heavily involved, as indicated by the constantly elevated levels of CRP detected.
The detection of the aforementioned DIC supports the conclusion that the resulting disease is a potent SARS-CoV-2-induced inflammation, which can be associated to alveolar CLLS. Thus, it resembles the inflammatory syndromes that may complicate anti-phospholipid antibody syndrome (APS), namely the catastrophic anti-phospholipid syndrome (CAPs), characterized by disseminated intravascular microthrombosis \[[@B29-ijms-21-03377]\]; or, the systemic CLLS characterized by altered capillary wall permeability \[[@B30-ijms-21-03377]\]. Besides CAPs and CLLS, the SARS-CoV-2 inflammation, leading to ARDS, shares pathogenic mechanisms and clinical-radiological aspects with other auto-immune/-inflammatory diseases, such as juvenile idiopathic arthritis and its adult form \[[@B31-ijms-21-03377],[@B32-ijms-21-03377],[@B33-ijms-21-03377]\] and Kawasaki disease \[[@B34-ijms-21-03377]\]. Systemic CLLS has also been observed in Kawasaki disease \[[@B35-ijms-21-03377]\].
These considerations have prompted clinicians to resort to off-label use of pro-inflammatory cytokine-targeted reagents to treat SARS-CoV-2-infected patients with ongoing ARDS.
4. Immunological Rationale for Targeting the Immune System to Fight SARS-CoV-2 {#sec4-ijms-21-03377}
==============================================================================
Besides antipyretics, oxygen, antibiotics, and a number of SARS-CoV-2 non--specific antiviral drugs, mostly protease inhibitors (i.e., Lopinavir-Ritonavir, Remdesivir, Favipiravir), which are all still under clinical trial, only one drug with immunomodulatory properties, namely hydroxy-chloroquine, is currently being used in SARS-CoV-2 infected patients in the pre-ARDS phase, following an open-label non-randomized clinical trial \[[@B36-ijms-21-03377]\]. In fact, the drug can inhibit natural and adaptive immunity through different, albeit not well-defined, mechanisms \[[@B37-ijms-21-03377]\]. In addition, hydroxy-chloroquine can reduce the expression of the phosphatidyl-inositol-binding clathrin assembly protein, essential for SARS-CoV-2 uptake \[[@B37-ijms-21-03377],[@B38-ijms-21-03377]\]. Finally, preliminary studies showed that the drug accelerates body virus clearance, due to its ability to alter intracellular pH at lysosomal levels, hence reducing the intracellular viral replication rate \[[@B39-ijms-21-03377],[@B40-ijms-21-03377]\].
All of the remaining therapeutic approaches are aimed at preventing and/or reversing the dramatic inflammatory syndrome triggered by the virus, which ultimately leads to ARDS.
In some case series, anti-IL6 therapy was successful in stabilizing the alveolar capillary membrane, reducing alveolar wall edema, and preventing/reversing ARDS \[[@B41-ijms-21-03377],[@B42-ijms-21-03377],[@B43-ijms-21-03377],[@B44-ijms-21-03377]\], shortening the intensive care unit stay \[[@B45-ijms-21-03377],[@B46-ijms-21-03377]\]. The currently available reagents that can neutralize IL-6 activity are the humanized monoclonal antibody (mAb) Tocilizumab, and the fully human mAb Sarilumab. Tocilizumab is the only one being tested so far. Even so, definitive proof of the efficacy of this therapy will only be obtained from the results of currently ongoing controlled clinical trials. These, with their main operative lines, are reported, as follows: 1. Tocilizumab mAb Vs continuous renal replacement therapy (CRRT) in the management of cytokine release syndrome in COVID-19 (TACOS) (NCT04306705) (last update posted: 17 March 2020; last access date: 9 May); 2. Tocilizumab in COVID-19 Pneumonia (TOCIVID-19), (NCT04317092) (last update posted: 7 April 2020; last access date: 9 May); 3. Favipiravir (an inhibitor of virus RNA polymerase) combined with Tocilizumab in the treatment of corona virus disease 2019 (NCT04310228) (last update posted: 10 April 2020; last access date: 9 May); and, 4. Tocilizumab in the treatment of patients with COVID-19 pneumonia (TOCIVID-19) (EudraCT Number: 2020-001110-38).
However, it cannot be excluded that small molecules, given per os, like the Janus kinases (JAK) inhibitor Baricitinib, may eventually be used instead of mAbs for two reasons: (1) the drug, at a dosage of 4 mg/day, is a selective down-modulator of JAK-1 and JAK-2 and, consequently, of the expression of the JAK down-stream proteins, called "signal transducer and activation of transcription" (STATs), both of which protein families are crucial for the IL-6-driven intracellular signal. (2) Baricitinib has shown affinity for the adaptor protein (AP2)-associated protein kinase 1 (AAK1) (AP2-AAK1), and cyclin G-associated kinase, both regulators of endocytosis and, hence, of the virus uptake \[[@B47-ijms-21-03377]\]. Thus, the ability to block IL-6 triggered intracellular signal and virus uptake could indicate Baricitinib as a suitable drug for treating the infection.
Even so, concerns have been raised regarding the mechanism of action of this class of drugs. Indeed, JAK-1 and JAK-2 are also involved in type I and type II-IFN intracellular signals, which play a pivotal role in host defense against viruses, including SARS-CoV-2 \[[@B48-ijms-21-03377]\]. Consequently, the inhibition of JAK-1 and JAK-2 would also decrease the host defense against the virus, as a side effect, facilitating virus spread in the host. Moreover, the optimal timing for the use of JAK-1 kinase inhibitors is a matter of speculation. It is not clear whether Baricitinib should be given in the early stage of virus infection, together with hydroxy-chloroquine and/or anti-viral drugs, or at the stage of onset of ARDS \[[@B49-ijms-21-03377],[@B50-ijms-21-03377]\].
Ongoing clinical trials will address these issues, as well as the efficacy of Baricitinib alone or in combination with other drugs (1. Baricitinib in symptomatic Patients Infected by COVID-19: an Open-label, Pilot Study, <https://clinicaltrials.gov/NCT04320277> (last update posted: 22 April 2020; last access date: 9 May); and, 2. Treatment of moderate to severe COVID-19 in hospitalized patient with Lopinavir/Ritonavir, hydroxychloroquine sulfate, Baricitinib, and Sarilumab, NCT04321993) (last update posted: 24 April 2020; last access date: 9 May).
As in juvenile idiopathic arthritis, an autoinflammatory-mediated disease, IL-1β also seems to play a strategic role in SARS-CoV-2 infection, by activating the NLRP3 inflammasome, as documented by the increased IL-1β levels in lymphocytes and in the sera of infected patients \[[@B18-ijms-21-03377],[@B51-ijms-21-03377]\]. In a mouse model of MERS-CoV infection, expressing the human dipeptidyl peptidase 4 (hDPP4) receptor, MERS-CoV caused pyroptosis and increased IL-1β expression in macrophages \[[@B52-ijms-21-03377]\]. IL-1 blockers, such as Anakinra (and possibly mAb Kanakinumab), may be an additional weapon against the respiratory distress syndrome in COVID-19, in view of its effectiveness in patients with severe forms of sepsis and with an inflammatory profile resembling the "macrophage activation syndrome" (MAS) \[[@B53-ijms-21-03377],[@B54-ijms-21-03377]\]. An open label, controlled, multicenter study is analyzing the therapeutic efficacy of Anakinra, together with an anti-IFN-γ mAb Emapalumab, in order to reduce the hyper-inflammation that is caused by SARS-CoV-2 Infection (Efficacy and Safety of anti-IFN-γ and Anakinra (antiIL-1) in Reducing Hyperinflammation and Respiratory Distress in Patients With COVID-19 Infection (NCT04324021) (last update posted: 9 April 2020; last access date: 9 May).
The potential utility of therapy with human immunoglobulin for intravenous use (IVIG) has been recently reported by our group \[[@B55-ijms-21-03377]\], based on the similar etiology and inflammatory pathogenesis of SARS-CoV-2 infection to diseases for which the use of IVIG has already been approved by the FDA or EMA ([Table 1](#ijms-21-03377-t001){ref-type="table"}) \[[@B56-ijms-21-03377],[@B57-ijms-21-03377],[@B58-ijms-21-03377],[@B59-ijms-21-03377],[@B60-ijms-21-03377],[@B61-ijms-21-03377],[@B62-ijms-21-03377],[@B63-ijms-21-03377]\], or for which IVIG are employed off-label ([Table 2](#ijms-21-03377-t002){ref-type="table"}) \[[@B30-ijms-21-03377],[@B64-ijms-21-03377],[@B65-ijms-21-03377],[@B66-ijms-21-03377],[@B67-ijms-21-03377],[@B68-ijms-21-03377],[@B69-ijms-21-03377],[@B70-ijms-21-03377],[@B71-ijms-21-03377],[@B72-ijms-21-03377],[@B73-ijms-21-03377],[@B74-ijms-21-03377],[@B75-ijms-21-03377],[@B76-ijms-21-03377],[@B77-ijms-21-03377],[@B78-ijms-21-03377],[@B79-ijms-21-03377],[@B80-ijms-21-03377],[@B81-ijms-21-03377],[@B82-ijms-21-03377],[@B83-ijms-21-03377],[@B84-ijms-21-03377],[@B85-ijms-21-03377],[@B86-ijms-21-03377],[@B87-ijms-21-03377],[@B88-ijms-21-03377],[@B89-ijms-21-03377],[@B90-ijms-21-03377]\], including viral or bacterial septic shock \[[@B91-ijms-21-03377]\] and the autoinflammatory syndrome with MAS \[[@B91-ijms-21-03377],[@B92-ijms-21-03377],[@B93-ijms-21-03377]\]. To dam inflammation, IVIG should be given at a dosage not exceeding 0.4 g/kg/day for three to five consecutive days (maximum total dosage is 2 g/kg) and the patient should be well hydrated. Higher IVIG dosage and/or poorly hydrated patients may increase the blood hyperviscosity, favoring coagulation, self-Ig aggregation with a transient complement activation, and/or the risk of the onset of rare side effects, such as acute aseptic meningitis \[[@B94-ijms-21-03377]\].
5. Passive Immunotherapy {#sec5-ijms-21-03377}
========================
The infusion of plasma serum obtained from PCR-negative, recovered patients, containing IgG anti-SARS-CoV-2 (hyperimmune IgG-containing plasma; HIgCP), could be an attractive approach in newly infected subjects, based on previous experiences that are related to other viral infections, namely SARS-CoV, H5N1 avian influenza, and H1N1 influenza \[[@B95-ijms-21-03377],[@B96-ijms-21-03377],[@B97-ijms-21-03377],[@B98-ijms-21-03377]\]. The administration of HIgCP could be useful to treat or prevent ARDS that is induced by SARS-CoV-2 infection and to accelerate virus clearance \[[@B99-ijms-21-03377],[@B100-ijms-21-03377],[@B101-ijms-21-03377]\]. Even so, the need for a blood group match between donor and recipient makes the use of HIgCP less suitable than IVIG for large scale therapy. Preliminary clinical experiences are promising \[[@B102-ijms-21-03377]\], but ongoing clinical trials are assessing the true effectiveness of this therapeutic strategy (NCT04321421) (last update posted: 27 April 2020; last access date: 9 May).
Finally, the administration of fully human mAbs specific to the receptor-binding domain (RBD) of the virus (responsible for its entry in the cell) \[[@B103-ijms-21-03377],[@B104-ijms-21-03377]\] might be a valid alternative to HIgCP in controlling SARS-CoV-2 infection, because mAbs can be produced in large amounts, reproducible manner, with the same specificity, and they do not require a pre-evaluation of ABO blood groups match \[[@B103-ijms-21-03377],[@B104-ijms-21-03377]\]. Recently, a SARS-CoV2-RBD-specific human neutralizing mAbCR3022, isolated from a single-chain variable antibody fragment (scFv) phage display library, which was constructed with lymphocyte RNA from convalescent SARS patient \[[@B105-ijms-21-03377]\] has been reputed a promising therapeutic candidate for the treatment of COVID-19 infection \[[@B106-ijms-21-03377]\]. Another mAb with potential therapeutic utility is the humanized mAb47D11 generated through hybridoma technology from splenocytes of H2L2 human Ig transgenic mice immunized with the spike protein. This mAb is capable of neutralizing both SARS-CoV and SARS-CoV-2 "in vitro", by the binding to a conserved epitope on RDB, but the results are still preliminary (<https://www.biorxiv.org/content/10.1101/2020.03.11.987958v1>).
6. Active Immunotherapy Approaches {#sec6-ijms-21-03377}
==================================
Multiple attempts are being made to develop an effective vaccine against SARS-Cov-2. To date, there are 78 confirmed active projects, 53 of which are in exploratory development, 18 are at a preclinical stage, and five are being tested in phase I clinical trials \[[@B107-ijms-21-03377]\]. Most of them target the surface-exposed spike protein in order to induce neutralizing antibodies.
The current vaccine platforms employ live attenuated viruses, inactivated viruses, non-replicating or replicating viral vectors, recombinant proteins, virus-like particles involved in DNA replication, and nucleic acid (DNA or mRNA) \[[@B108-ijms-21-03377]\].
The only vaccine platforms that have been moved into phase I clinical trials include (1) the mRNA-based vaccine, from Moderna (NCT04283461) (last update posted: 4 May 2020; last access date: 9 May); (2) the DNA plasmid encoding the spike protein delivered by electroporation, from Inovio Pharmaceuticals (NCT04336410) (last update posted: 24 April 2020; last access date: 9 May); (3) the Adenovirus type 5 (Ad5) expressing the spike protein, from CanSino Biologicals (NCT04324606) (last update posted: 8 May 2020; last access date: 9 May); (4) the lentiviral vector system expressing viral proteins and immune modulatory genes to modify dendritic cells and activate T cells, from the Shenzhen Geno-Immune Medical Institute (NCT04276896) (last update posted: 19 March 2020; last access date: 9 May); (5) the lentiviral vector system expressing viral proteins and immune modulatory genes to modify artificial antigen presenting cells and to activate Tcells, also from the Shenzhen Geno-Immune Medical Institute (NCT04299724) (last update posted: 9 March 2020; last access date: 9 May) \[[@B107-ijms-21-03377]\]. However, anti-SARS-CoV-2 vaccines need to be available very soon, and in large amounts, because the COVID-19 pandemic is still ongoing.
Among the vaccines in a clinical phase as candidates for SARS-CoV-2, them RNA-based vaccine seems to be more promising than other more traditional ones, because it is amenable to low-budget production of large quantities of vaccine. Although safe in pilot clinical trials, the antigen-specific immune response that is induced by RNA-based vaccines is lower than that observed in animal models \[[@B109-ijms-21-03377]\].
Like RNA-based vaccines, DNA vaccines are easy to produce, cost-effective, and present a good safety profile and long term persistence of the immunogen. Even so, they have not been approved for human use due to their inability to evoke a strong enough immune response to be protective. Instead, viral vector-based vaccines, which are highly immunogenic, have been shown to generate efficient humoral and cell-mediated immune responses. However, the use of an adenovirus as a vector to carry the gene encoding the target protein, as in the case of Ad5 expressing spike, raises some concerns that are related to pre-existing immunity against adenovirus in the human population \[[@B110-ijms-21-03377]\]. Lastly, the use of viral protein-derived lentiviral vectors raises some safety concerns that are related to the potential risk of mutagenesis \[[@B111-ijms-21-03377]\].
It is not possible to predict which vaccine will be more effective in preventing SARS-CoV-2 infection. Only results from phase II/III trials will be able to suggest the more effective vaccine formulation. In this context, considerable attention should be paid to the remote possibility of a vaccine-triggered disease activation, as previously observed with animal models of anti-CoV vaccines \[[@B112-ijms-21-03377]\].
7. Conclusions {#sec7-ijms-21-03377}
==============
Many antiviral and immunomodulant drugs are currently available for COVID-19, to be used alone or in combination, either at the initial disease stage or to protect patients from an inflammatory syndrome, preventing ARDS. However, most of these drugs are not virus-specific, with the exception of the hyperimmune Ig-containing sera obtained from individuals who recovered from the infection, but the effectiveness of this is still under study. The previous outbreaks, like the 2002 SARS and 2012 MERS experiences, have generated high expectations for a rapidly available vaccine. Unfortunately, and unexpectedly, all of the ongoing vaccine programs are still in clinical phase I, and most of them are employing platforms that have never been used before rather than the well-known platforms employed for the vaccines that are already commercially available, which have as of yet proven ineffective.
The authors thank Maria Daniele and Giuseppina Dammacco for their excellent secretarial assistance. Mary V.C. Pragnell provided language editing.
Conceptualization: F.P., M.P; original draft preparation, F.P., M.P. and E.F.; review and editing, G.C. and V.R. All authors have read and agreed to the published version of the manuscript.
The authors declare no conflict of interest.
APS
antiphospholipid antibody syndrome
ARDS
acute respiratory distress syndrome
CAPs
catastrophic anti-phospholipid syndrome
CLLS
capillary leak-like syndrome
CoV
coronavirus
COVID-19
coronavirus disease-2019
HIgCP
Hyperimmune IgG-containing plasma
RBD
receptor-binding domain
SARS
severe acute respiratory syndrome
WHO
World Health Organization
ijms-21-03377-t001_Table 1
######
List of diseases for which treatment with human normal immunoglobulin for intravenous administration (IVIG) has been approved by the Food and Drug Administration (FDA) and/or European Medicines Agency (EMA).
Disease Denomination IVIG Use Approved by Rationale and/or Mechanism of Action References
------------------------------------------------------------------------------- ---------------------- -------------------------------------- ------------------------------------------------------ -----------------------------------------------
Primary immunodeficiencies (PID) Yes Yes IgG replacement \[[@B56-ijms-21-03377]\]
Clinically manifest secondary immunodeficiencies (HIV, CLL, B cell depletion) Yes Yes IgG replacement \[[@B57-ijms-21-03377],[@B58-ijms-21-03377]\]
Idiopathic thrombocytopenic purpura (ITP) Yes Yes Fc receptor saturation \[[@B59-ijms-21-03377]\]
Kawasaki disease Yes Yes Anti-inflammatory, binding to virus or superantigens \[[@B60-ijms-21-03377]\]
Chronic Inflammatory demyelinating polyneuropathy (CIDP) Yes Yes Anti-inflammatory \[[@B61-ijms-21-03377]\]
Multifocal motor neuropathy Yes Yes Not defined \[[@B62-ijms-21-03377]\]
Guillain-Barré Syndrome (GBS) Yes \- Anti-inflammatory \[[@B63-ijms-21-03377]\]
^a)^ Diseases for which treatment with IVIG has been approved by the EMA are continually being updated, as reported at "<https://www.ema.europa.eu/en/> (last access date: 28 June 2020). ^b)^ Diseases for which treatment with IVIG has been approved by the FDA are continually being updated, as reported at "<https://www.fda.gov/> (last access date: 3 May 2020).
ijms-21-03377-t002_Table 2
######
Diseases considered for the off-label use of intravenous human normal immunoglobulin (IVIG).
Disease Rationale and/or Mechanism of Action References
-------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------
Prophylaxis in hematopoietic stem cell transplantation Ig replacement \[[@B64-ijms-21-03377]\]
Infection disease conditions (toxemia, parvovirus 19) Neutralization of pathogenic exogenous antigen, anti-inflammatory effects \[[@B65-ijms-21-03377],[@B66-ijms-21-03377],[@B67-ijms-21-03377]\]
Infections in solid organ transplantation, surgery, trauma, burns Ig replacement \[[@B68-ijms-21-03377]\]
Idiopathic arthritis (especially the juvenile inflammatory form) Fc-mediated \[[@B69-ijms-21-03377]\]
Systemic lupus erythematosus and lupus nephritis Fc- and Fab- mediated \[[@B70-ijms-21-03377],[@B71-ijms-21-03377],[@B72-ijms-21-03377]\]
Autoimmune cytopenia (etc. autoimmune hemolytic anemia, Immune-mediated neutropenia) Fc-mediated saturation of FcγRs, ADCC and CDC inhibition \[[@B73-ijms-21-03377]\]
Dermatomyositis and polymyositis Fc-mediated \[[@B74-ijms-21-03377],[@B75-ijms-21-03377]\]
Catastrophic antiphospholipid syndrome Fc- and Fab-mediated \[[@B76-ijms-21-03377]\]
Systemic capillary leak-like syndrome Fc- and Fab-mediated \[[@B30-ijms-21-03377],[@B77-ijms-21-03377]\]
Vasculitides (ANCA associated) Fc- and Fab-mediated \[[@B78-ijms-21-03377],[@B79-ijms-21-03377]\]
Skin autoimmune diseases (pemphigo, epidermolysis bullosa, atopic dermatitis, chronic urticaria) mostly Fc-mediated, anti-inflammatory, \[[@B80-ijms-21-03377],[@B81-ijms-21-03377],[@B82-ijms-21-03377],[@B83-ijms-21-03377],[@B84-ijms-21-03377]\]
Myasthenia gravis Fab-mediated \[[@B85-ijms-21-03377]\]
Small-fiber polyneuropathy Not defined \[[@B86-ijms-21-03377]\]
Epilepsy Not defined \[[@B87-ijms-21-03377]\]
Acquired factor VIII inhibitors Fc- and Fab-mediated \[[@B88-ijms-21-03377]\]
Asthma Anti-inflammatory \[[@B89-ijms-21-03377]\]
Recurrent pregnancy loss Fc- and Fab- mediated \[[@B90-ijms-21-03377]\]
ADCC, antibody-dependent cell-mediated cytotoxicity; CDC, complement-dependent cytotoxicity.
| {
"pile_set_name": "PubMed Central"
} |
Subsets and Splits