Dataset Viewer
prompt
stringclasses 1
value | text
stringclasses 10
values |
|---|---|
Vytvor v slovenčine zhrnutie totho článku. Laicky vysvetli všetky základné pojmy v článku. Ak používaš úvodzovky, nezabudni, že v slovenčine sa prvé úvodzovky píšu dole. Používaj správne slovenské slová, vyhni sa čechizmom . Využi 8000 až 8100 tokenov. Tu je text článku: |
Introduction
The filoviruses (family Filoviridae ) from the genera Ebolavirus and Marburgvirus are etiologic agents of sporadic viral hemorrhagic fever outbreaks in humans with high mortality rates. An unprecedented outbreak of Ebola virus (EBOV; species Zaire ebolavirus ) disease that began in Guinea during December 2013 [ 1 ] subsequently spread into neighboring West African countries of Sierra Leone and Liberia, prompting the World Health Organization (WHO) to declare the epidemic a public health emergency of international concern ( http://www.who.int/mediacentre/news/statements/2014/ebola-20140808/en/ ). Phylogenetic analysis of viral isolates from this epidemic suggests a single transmission event introduced the virus, named the EBOV Makona variant [ 2 ], from an undetermined natural reservoir into humans in Guinea, followed by transmission between humans to spread the virus throughout Guinea and into Sierra Leone and Liberia [ 3 ]. Implementation of containment measures such as patient isolation and improved burial practices eventually controlled the epidemic, which resulted in 28,616 reported cases with a mortality rate of approximately 40% ( http://www.who.int/csr/disease/ebola/en/ ).
The severity of this epidemic and principle transmission from human to human underscored the need for efficacious vaccines (and therapeutics) against EBOV, accelerating the placement of candidate EBOV vaccines into clinical safety trials [ 4 – 6 ]. This need for safe and efficacious vaccines was again evident with the onset of the 10 th and largest outbreak in the Democratic Republic of the Congo (DRC) from 2018–2020. The 11 th outbreak of EVD continues in the Western DRC.
The characteristics of filovirus infection, where infected patients are contagious only after manifestation of symptoms, allows one to use a ring vaccination strategy for disease containment. Ring vaccination strategy relies on the combination of contact tracing for case identification and a rapid effective vaccine for use in contacts and contacts of contacts of infected patients. The application of this strategy led to the approval of rVSV-ZEBOV (ERVEBO ® ), a single dose vaccine, using the safety and efficacy data from the clinical trial during the 2014 outbreak in West Africa by the Food and Drug Administration in December 2019 ( https://www.fda.gov/news-events/press-announcements/first-fda-approved-vaccine-prevention-ebola-virus-disease-marking-critical-milestone-public-health ). The effectiveness of ERVEBO in a ring vaccination response provides an important countermeasure for public health but does not address all unresolved questions in filovirus vaccine utilization including duration of protection, alternate dosing regimens, and the effectiveness of filovirus vaccines based on other viral platforms or alternative strategies. The development of multiple countermeasures against a disease necessitates the use of a common assay based on a surrogate of protection which can be used to compare the elicited immune response between vaccines and provide valuable information as to the effectiveness and durability of protection. Ideally, this assay is not only informative but simple, reproducible, species independent, and transferrable between labs. For example, during the development of countermeasures against anthrax, a lethal toxin neutralization assay was developed and used by many laboratories [ 7 ].
The development of vaccine candidates for Ebola virus disease prophylaxis [ 8 ] continues today, including deployment of a heterologous prime boost vaccine with European Commission Market Authorization during the last outbreak. However, the demonstration of efficacy for new filovirus vaccines will be complicated in the absence of a large outbreak and may require evaluation under the FDA Animal Rule or via non-inferiority trials against ERVEBO. Regulatory evaluation using these approaches is only possible with a correlate of protection and a well-developed assay that can measure the response in well-characterized animal challenge models as well as in human clinical trials. The species-neutral ELISA is ideal for bridging data between humans and animal models. Also, since the assay likely will be utilized in multiple experiments at many sites, it is important to demonstrate that the assay is reproducible among different laboratories.
In order to facilitate the development of additional vaccine countermeasures and to address such questions as the durability of immunity, the FANG has supported the development of a human anti-EBOV GP IgG ELISA. This study describes the FANG efforts to determine if the performance of the human anti-EBOV GP IgG ELISA [ 9 ] is acceptable for sample evaluation across five participating laboratories. Each laboratory used an anti-EBOV GP IgG ELISA to measure levels of binding in human serum samples from a FANG designed human proficiency panel. The panel consisted of ten human serum samples created by the differential dilution of human serum lot number BMIZAIRE105 (pool of serum with an approximate anti-GP IgG concentration of 1,000 ELISA units/mL) with control human serum (BMI529) without antibody activity. The concentration of the proficiency panel samples ranged from 0.00 ELISA units/mL to approximately 800 ELISA units/mL.
Each participating laboratory received sufficient volume of the proficiency panel for initial testing plus repeats and used their own anti-EBOV GP IgG ELISA established assay. The assay was validated at some laboratories and qualified at others [ 9 ]. Data from the participating laboratories were compared by statistical analysis. Both intra-laboratory and inter-laboratory analyses were performed to evaluate repeatability, intermediate precision, dilutional linearity, and accuracy. This paper summarizes both the intra- and inter-laboratory analysis of the results generated in the five separate laboratories. Results from the laboratories are de-identified in the analysis and reported as Laboratory A through E. The repeatability estimate for Laboratory B was greater than the acceptance criteria as established in laboratories that validated the anti-EBOV GP IgG ELISA with human serum, and, as a result, the proficiency panel assay runs were repeated. Results from both the original and repeated runs were included in the analysis and labeled as being from Laboratory B1 and B2, respectively.
Assay method
A common assay method [ 9 ] was tech-transferred to the participating laboratories, but there were minor variations in equipment/materials/procedures between laboratories. The analysis of the proficiency panel in the ELISA was performed similarly at Labs A, B1, and B2. All three used two separate operators on separate days. Samples were analyzed using a starting dilution of 1:62.5 and followed the plate layout as illustrated in Table 1 . These plate layouts represent 15 plates with specific proficiency panel samples on each plate. All 15 plates were run twice for a total of 30 plates for each of Labs A, B1, and B2.
10.1371/journal.pone.0238196.t001
Table 1 Plate layout used at Laboratories A, B1, and B2.
Sample ID
Plate Number
1
2
3
4
5
6
7
8
BMI-ZPP-11
X (3)
X
X
X
BMI-ZPP-12
X
X
X
X
X (3)
BMI-ZPP-13
X (3)
X
X
X
X
X
BMI-ZPP-14
X
X
X
X
X
X
X (3)
BMI-ZPP-15
X
X (3)
X
X
BMI-ZPP-16
X
X
X
X
X
X
BMI-ZPP-17
X
X (3)
X
X
X
BMI-ZPP-18
X
X
X
X
X
BMI-ZPP-19
X
X (3)
X
X
BMI-ZPP-20
X
X
X
X
Sample ID
Plate Number
9
10
11
12
13
14
15
BMI-ZPP-11
X
X
X
X
X
X
BMI-ZPP-12
X
X
X
X
X
BMI-ZPP-13
X
X
X
X
BMI-ZPP-14
X
X
X
BMI-ZPP-15
X
X
X
X
X
X
BMI-ZPP-16
X (3)
X
X
X
BMI-ZPP-17
X
X
X
X
X
BMI-ZPP-18
X (3)
X
X
X
X
BMI-ZPP-19
X
X
X
X
X
X
BMI-ZPP-20
X
X (3)
X
X
X
X
An “X” indicates that sample was analyzed on the indicated plate.
An “X (3)” (shaded) indicates that sample was analyzed on the indicated plate three times.
The analysis of the proficiency panel in the ELISA was performed at Lab C by two separate operators over three days and at Lab D by two separate operators over five days. Samples were analyzed using a starting dilution of 1:50 and followed the plate layout as illustrated in Table 2 . These plate layouts represent 12 plates with specific proficiency panel samples on each plate. All 12 plates were run at least twice for a total of 24 plates for each of Labs C and D.
10.1371/journal.pone.0238196.t002
Table 2 Plate layout used at Laboratories C and D.
Sample ID
Plate Number
1
2
3
4
5
6
BMI-ZPP-11
X (3)
X
X
X
BMI-ZPP-12
X
X
X
X
X
X
BMI-ZPP-13
X (3)
X
X
X
BMI-ZPP-14
X
X
X
X
X
X
BMI-ZPP-15
X
X (3)
X
X
BMI-ZPP-16
X
X
X
X
X
X
BMI-ZPP-17
X
X
X (3)
X
BMI-ZPP-18
X
X
X
X
X
X
BMI-ZPP-19
X
X
X (3)
X
BMI-ZPP-20
X
X
X
X
X
X
Sample ID
Plate Number
7
8
9
10
11
12
BMI-ZPP-11
X
X
X
X
X
X
BMI-ZPP-12
X (3)
X
X
X
BMI-ZPP-13
X
X
X
X
X
X
BMI-ZPP-14
X (3)
X
X
X
BMI-ZPP-15
X
X
X
X
X
X
BMI-ZPP-16
X
X (3)
X
X
BMI-ZPP-17
X
X
X
X
X
X
BMI-ZPP-18
X
X
X (3)
X
BMI-ZPP-19
X
X
X
X
X
X
BMI-ZPP-20
X
X
X (3)
X
An “X” indicates that sample was analyzed on the indicated plate.
An “X (3)” (shaded) indicates that sample was analyzed on the indicated plate three times.
The analysis of the proficiency panel in the ELISA was performed at Lab E by two separate operators over four days. Samples were analyzed using a starting dilution of 1:50 and followed the plate layout as illustrated in Table 3 . This plate layout represents six plates with specific proficiency panel samples on each plate. The six plates were each run four times for a total of 24 plates. For all laboratories, some samples were analyzed three times on the same plate [denoted with “X (3)” in the plate layouts]. These contributed to assay repeatability.
10.1371/journal.pone.0238196.t003
Table 3 Plate layout used at Laboratory E.
Sample ID
Plate Number
1
2
3
4
5
6
BMI-ZPP-11
X
X
X
X
X
X
BMI-ZPP-12
X (3)
X
X
X
BMI-ZPP-13
X
X
X
X
X
X
BMI-ZPP-14
X (3)
X
X
X
BMI-ZPP-15
X
X
X
X
X
X
BMI-ZPP-16
X
X (3)
X
X
BMI-ZPP-17
X
X
X
X
X
X
BMI-ZPP-18
X
X
X (3)
X
BMI-ZPP-19
X
X
X
X
X
X
BMI-ZPP-20
X
X
X (3)
X
An “X” indicates that sample was analyzed on the indicated plate.
An “X (3)” (shaded) indicates that sample was analyzed on the indicated plate three times.
Samples on a given plate were excluded from analysis if the within-assay CV of at least three dilution-adjusted concentrations determined for that sample was greater than 20%. Samples were also excluded if the plate including that sample failed to meet system suitability criteria. Some samples and plates that failed to meet the sample suitability criteria or system suitability criteria were repeated on later days. The ELISA concentrations of each qualification test sample by laboratory are provided in the supplemental information ( S1 – S6 Tables).
This study, and specifically the use of human serum samples, was approved in writing by the Battelle Institutional Review Board in April of 2015 (approval number HSRE 0223–100062052). Human serum samples were collected from subjects by the sponsor (Crucell Holland) via written consent according to their IRB-approved protocol. These samples were not specifically collected for this interlaboratory study but rather for a different study. Battelle nor any authors were affiliated with this initial study. The sponsor subsequently provided Battelle volumes of these samples for the purposes of conducting the study described in this manuscript. Throughout its analysis of human biological materials and reporting, Battelle had no access to volunteer subjects’ identifiers nor any access to any code-key that would allow Battelle researchers to attribute any results of analysis to the original volunteer human research subjects.
Statistical methods
Inter-laboratory analysis was performed using the combined results across all laboratories. A mixed-effects analysis of variance (ANOVA) model was fitted to the base-10 log-transformed concentrations to evaluate both inter-laboratory precision (i.e., between lab precision) and intra-laboratory precision (i.e., within-laboratory precision). The model included a fixed effect for test sample and random effects for laboratory, test date nested within laboratory, and plate nested within day. Here, test operator was excluded as a random effect because this variable was indistinguishable from test day in most laboratories. Because of this confounding of effects, any variability attributable to test day may also be due to the different test operators.
Results were screened for outliers within each laboratory separately. Deleted studentized residuals were computed for each observation. If the absolute value of the deleted studentized residual was greater than four, then the observation was considered a statistical outlier and removed from the inter-laboratory analysis.
Variability associated with the random effects as well as intermediate precision, repeatability, and total assay variability were estimated separately for each lab using model-based percent coefficient of variation (CV). The percent CV for each source of variance was calculated using Tan’s [ 10 ] relative standard deviation as
100 × e ln ( 10 ) 2 × σ 2 − 1
where σ 2 is the model-estimated variance for the specific variance source. The percent CV associated with the residual variance served as an estimate for the assay repeatability. The percent CV associated with the test day and plate effects served as an estimate for the intermediate precision of the assay. Total assay variability was estimated using all variance components from the model (both inter- and intra-run variability).
The model intercept was obtained for each test sample from the mixed effects ANOVA model to serve as test sample consensus values across the laboratories. Agreement among laboratories was evaluated by comparing individual assay results from each laboratory to the consensus values. Boxplots were produced for each test sample to show the distribution of concentrations by laboratory in relation to the corresponding consensus value. The ratio of individual test results to consensus values was calculated by test sample to evaluate the level of agreement among laboratories based on two one-sided tests (TOST) of equivalence.
To assess dilutional linearity, a random coefficients linear regression model was fitted to the log-transformed observed concentrations versus the log-transformed target concentrations. The model included both a random intercept and slope effect for each laboratory, along with random effects for laboratory, test day nested within laboratory, and plate nested within laboratory. The random slope coefficients were modeled as laboratory-specific differences from the overall slope. The overall slope was used to assess the dilutional linearity based on a test of equivalence (TOST) and random slope coefficients were used to evaluate the level of agreement among the laboratories.
Results
Across all six laboratory runs, there were some false positive observations for Sample 18, a sample with a known negative concentration. All reportable values from Sample 18 were excluded from the statistical models. Table 4 lists five outliers that were removed from their respective intra-laboratory analyses that were also removed from this inter-laboratory analysis. One outlier each were removed from Laboratories B1 and B2. Three outliers were removed from Laboratory C. In the final analysis, Lab A contributed 204 reportable values, Lab B1 had 179 reportable values, Lab B2 had 214 reportable values, Lab C had 268 reportable values, Lab D had 216 reportable values, and Lab E had 218 reportable values.
10.1371/journal.pone.0238196.t004
Table 4 Statistical outliers identified during analysis of intra-laboratory data.
Laboratory
Test Sample
Observed Concentration (ELISA Units/mL)
Target Concentration (ELISA Units/mL)
Studentized Residual
B1
BMI-ZPP-17
4.28
200
-9.48
C
BMI-ZPP-13
896.47
300
5.34
C
BMI-ZPP-16
236.02
500
-4.74
B2
BMI-ZPP-19
51.20
100
-4.39
C
BMI-ZPP-14
1845.88
700
4.33
These observations were deleted from both intra- and inter-laboratory analyses.
Table 5 presents ANOVA variance estimates and %CV for each source of variability, intermediate precision, and total assay variability by laboratory. For Laboratory A, the %CV for test date and plate nested within test date were 0.0 and 9.8, respectively. For Laboratory B1, the %CV for test date and plate nested within test date were 10.8 and 15.3, respectively. For Laboratory B2, the %CV for test date and plate nested within test date were 4.5 and 8.9, respectively. For Laboratory C, the %CV for test date and plate nested within test date were 9.8 and 8.5, respectively. For Laboratory D, the %CV for test date and plate nested within test date were 18.9 and 10.5, respectively. Finally, for Laboratory E, the %CVs for test date and plate nested within test date were 7.3 and 5.0, respectively. Laboratory E had the lowest %CV for intermediate precision (8.9) while Laboratory A had the lowest %CV for repeatability (7.2) and total assay variability (12.2). Laboratory B1 had the highest repeatability and total assay variability (23.7%CV and 30.6%CV, respectively) while Laboratory D had the highest %CV for intermediate precision (21.7).
10.1371/journal.pone.0238196.t005
Table 5 Summary of variance components obtained from mixed ANOVA model fit to data from all laboratories (results shown by laboratory).
Laboratory A
Source of Variability
Variance
%CV
Test Date
0.0000
0.0
Plate Nested in Test Date
0.0018
9.8
Intermediate Precision 1
0.0018
9.8
Residual (Repeatability)
0.0010
7.2
Total Assay Variability 2
0.0028
12.2
Laboratory B1
Source of Variability
Variance
%CV
Test Date
0.0022
10.8
Plate Nested in Test Date
0.0044
15.3
Intermediate Precision 1
0.0065
18.8
Residual (Repeatability)
0.0103
23.7
Total Assay Variability 2
0.0169
30.6
Laboratory B2
Source of Variability
Variance
%CV
Test Date
0.0004
4.5
Plate Nested in Test Date
0.0015
8.9
Intermediate Precision 1
0.0019
9.9
Residual (Repeatability)
0.0033
13.3
Total Assay Variability 2
0.0052
16.7
Laboratory C
Source of Variability
Variance
%CV
Test Date
0.0018
9.8
Plate Nested in Test Date
0.0014
8.5
Intermediate Precision 1
0.0031
13.0
Residual (Repeatability)
0.0027
11.9
Total Assay Variability 2
0.0058
17.7
Laboratory D
Source of Variability
Variance
%CV
Test Date
0.0066
18.9
Plate Nested in Test Date
0.0021
10.5
Intermediate Precision 1
0.0087
21.7
Residual (Repeatability)
0.0023
11.2
Total Assay Variability 2
0.0110
24.6
Laboratory E
Source of Variability
Variance
%CV
Test Date
0.0010
7.3
Plate Nested in Test Date
0.0005
5.0
Intermediate Precision 1
0.0015
8.9
Residual (Repeatability)
0.0015
9.0
Total Assay Variability 2
0.0030
12.7
1 . Comprised of test date and plate nested within test date sources of variability.
2 . Comprised of repeatability and intermediate precision.
Table 6 shows the consensus values (geometric means) along with 95% confidence intervals for each test sample generated from the mixed model ANOVA fitted to the data. Boxplots by sample of the reportable values from each laboratory, with each plot including a horizontal line for the consensus value estimate for the given sample, are provided in the supplemental information ( S1 – S9 Figs).
10.1371/journal.pone.0238196.t006
Table 6 Consensus values by test sample generated from intercept of mixed ANOVA model fit to data from all laboratories.
Sample ID
Target Concentration
Consensus Value
95% CI Consensus Value
BMI-ZPP-11
600
695.93
(677.31, 715.06)
BMI-ZPP-12
400
475.20
(462.47, 488.27)
BMI-ZPP-13
300
325.47
(316.72, 334.46)
BMI-ZPP-14
700
844.08
(821.24, 867.55)
BMI-ZPP-15
800
871.34
(847.91, 895.41)
BMI-ZPP-16
500
561.81
(546.71, 577.33)
BMI-ZPP-17
200
226.25
(220.18, 232.49)
BMI-ZPP-19
100
110.63
(107.66, 113.68)
BMI-ZPP-20
50
70.81
(68.89, 72.79)
Table 7 shows the ratio of the mean concentration for each of the six individual laboratory runs to the consensus value for a given sample along with a 90% confidence interval for the ratio. Agreement among laboratories implies that these ratios should be close to one, indicating that the average concentrations are about the same as the consensus values. The ratios range from 0.95 to 1.08 for Laboratory A; from 0.96 to 1.19 for Laboratory B1; from 0.83 to 1.12 for Laboratory B2; from 0.96 to 1.16 for Laboratory C; from 0.71 to 0.97 for Laboratory D; and from 0.90 to 1.06 for Laboratory E. Fig 1 shows a graph of the mean ratio and 90% confidence interval for each test sample by laboratory.
10.1371/journal.pone.0238196.g001
Fig 1
Graph of ratio of laboratory mean concentration to consensus value with 90% confidence intervals for each test sample by laboratory.
Dotted lines show equivalence region (0.80 to 1.25) and perfect agreement with consensus value (1.00). All means and confidence bounds are entirely within equivalence region for Laboratories A, B2, C, and E.
10.1371/journal.pone.0238196.t007
Table 7 Ratio of laboratory mean concentration to overall consensus value with 90% confidence intervals for each test sample.
Sample ID
Laboratory A
Laboratory B1
Laboratory B2
Ratio
90% Confidence Interval
Ratio
90% Confidence Interval
Ratio
90% Confidence Interval
BMI-ZPP-11
1.08
(1.04, 1.13)
1.09
(0.97, 1.23)
0.83
(0.81, 0.85)
BMI-ZPP-12
1.02
(0.98, 1.08)
0.97
(0.87, 1.09)
1.12
(1.08, 1.16)
BMI-ZPP-13
1.02
(0.98, 1.06)
1.15
(1.02, 1.30) *
0.93
(0.89, 0.97)
BMI-ZPP-14
0.99
(0.96, 1.02)
0.96
(0.87, 1.05)
1.09
(1.04, 1.13)
BMI-ZPP-15
1.01
(0.97, 1.06)
1.08
(0.98, 1.18)
0.88
(0.84, 0.91)
BMI-ZPP-16
1.00
(0.96, 1.04)
1.07
(1.00, 1.14)
1.05
(1.00, 1.09)
BMI-ZPP-17
1.06
(1.01, 1.11)
1.01
(0.88, 1.15)
1.12
(1.06, 1.18)
BMI-ZPP-19
0.95
(0.91, 0.98)
0.98
(0.87, 1.09)
0.83
(0.79, 0.87) *
BMI-ZPP-20
0.98
(0.95, 1.02)
1.19
(1.02, 1.39) *
0.92
(0.88, 0.97)
Sample ID
Laboratory C
Laboratory D
Laboratory E
Ratio
90% Confidence Interval
Ratio
90% Confidence Interval
Ratio
90% Confidence Interval
BMI-ZPP-11
1.10
(1.05, 1.15)
0.85
(0.81, 0.88)
0.90
(0.87, 0.94)
BMI-ZPP-12
1.08
(1.03, 1.12)
0.74
(0.72, 0.77) *
0.96
(0.92, 0.99)
BMI-ZPP-13
1.09
(1.04, 1.14)
0.79
(0.75, 0.83) *
0.95
(0.92, 0.99)
BMI-ZPP-14
1.00
(0.96, 1.05)
0.75
(0.71, 0.80) *
1.06
(1.02, 1.10)
BMI-ZPP-15
1.10
(1.04, 1.15)
0.90
(0.84, 0.97)
0.98
(0.95, 1.01)
BMI-ZPP-16
1.06
(1.02, 1.11)
0.76
(0.73, 0.79) *
1.01
(0.97, 1.06)
BMI-ZPP-17
0.96
(0.93, 1.00)
0.71
(0.69, 0.74) *
1.02
(0.99, 1.05)
BMI-ZPP-19
1.16
(1.10, 1.22)
0.97
(0.94, 1.01)
0.92
(0.89, 0.96)
BMI-ZPP-20
1.10
(1.02, 1.19)
0.77
(0.73, 0.82) *
1.02
(0.98, 1.06)
* 90% confidence interval is outside the acceptance bounds of (0.80, 1.25). Therefore, the concentrations for this test sample are not equivalent to those of other laboratories.
An equivalence test was conducted to determine if the mean test sample concentrations for each laboratory were equivalent to the corresponding test sample consensus value. An equivalence interval of 0.80 to 1.25 (representing a difference of 20% on the log scale) for the ratio of laboratory mean concentration to consensus concentration was used. The mean laboratory concentration for a given test sample is said to be equivalent to the consensus value for that sample if the 90% confidence interval for the ratio of these two values falls completely within the interval (0.80, 1.25).
Following this equivalence criteria: two intervals from Laboratory B1 (corresponding to BMI-ZPP-13 and BMI-ZPP-20) had an upper bound greater than the upper acceptance limit of 1.25 (1.30 and 1.39); one interval from Laboratory B2 (corresponding to BMI-ZPP-19) had a lower bound less than the lower acceptance limit of 0.80 (0.79); and six intervals from Laboratory D (corresponding to BMI-ZPP-12, BMI-ZPP-13, BMI-ZPP-14, BMI-ZPP-16, BMI-ZPP-17, and BMI-ZPP-20) had a lower bound less than the lower acceptance limit of 0.80. Furthermore, three of the six intervals are entirely below the lower acceptance bound of 0.80. These findings indicate that mean concentrations observed at Laboratory D are not equivalent to the other laboratories for six of the nine test samples.
Table 8 presents the estimated slope across the five laboratories and the corresponding 90% confidence interval obtained from the random regression model fit to assess the relationship between log 10 (observed concentration) and log 10 (target concentration). The overall slope was estimated to be 0.95 with a 90% confidence interval of (0.93, 0.97). An equivalence test was conducted to determine if the overall slope was equivalent to 1.00 (perfect dilutional linearity). An equivalence interval of 0.80 to 1.25 for the overall slope was used. Because the 90% confidence interval for the overall slope was completely within the interval (0.80, 1.25), the concentrations were found to be dilutionally linear across the laboratories. The slope estimates specific to each laboratory ranged from 0.94 to 0.96 ( Table 8 ) and were consistent with the overall slope.
10.1371/journal.pone.0238196.t008
Table 8 Estimated slope and lower and upper 90% confidence interval bounds by laboratory from random coefficients regression model fit to all data.
Laboratory
Slope Estimate
90% Confidence Interval #
Overall (All Labs)
0.95
(0.93, 0.97)
A
0.96
(0.90, 1.02)
B1
0.94
(0.88, 1.01)
B2
0.96
(0.90, 1.02)
C
0.96
(0.90, 1.02)
D
0.95
(0.89, 1.00)
E
0.95
(0.90, 1.01)
# 90% confidence interval is within the acceptance bounds of (0.80, 1.25). Therefore, the concentrations were dilutionally linear across the laboratories.
Discussion
The value of an assay as a regulatory tool is dependent on its accuracy, consistency, simplicity, and reproducibility. An assay that is relevant, is species independent, and replicable among laboratories is a powerful tool for product development. The data from a number of clinical trials utilizing ERVEBO strongly suggest that the anti-EBOV GP IgG ELISA provides data that correlate with product efficacy against Ebola infection. The development of new vaccines, or the evaluation of durability or alternative dosing regimens will be based on interpretation of data using the human anti-EBOV GP IgG ELISA. Our ability to use, or trust the data generated from non-clinical studies in different laboratories and clinical trials carried out with sera evaluated at different sites will require an understanding regarding the consistency and reproducibility of the assay among laboratories. In particular, assays using material from animal studies may be performed in laboratories different from those where the assay was performed to evaluate clinical trials. If the assay performance is not consistent among species and across laboratories, then data interpretation will not be possible. This interlaboratory study provided a direct head-to-head comparison of the ELISA performance in five different laboratories. The results from this study confirm the assay can be a universal tool for Ebola virus vaccine evaluation since results were similar when using the assay at multiple labs. However, the small differences in assay performance reinforce that for regulatory purposes, it is still ideal to rely on only one test site where the assay is fully validated.
Intermediate precision for the six laboratory runs ranged from 8.9 to 21.7%CV and repeatability ranged from 7.2 to 23.7%CV. The total assay variability %CVs range from 12.2 to 30.6. As a point of reference, laboratories that validated the anti-EBOV GP IgG ELISA have used the following precision acceptance criteria: (1) The intermediate precision of the assay for samples within the analytic range of the assay must be no larger than 25% CV; and (2) the repeatability of the assay for samples within the analytic range of the assay must be no larger than 20% CV. The repeatability estimate for Laboratory B1 was greater than the upper acceptance bound as established in laboratories that validated the anti-EBOV GP IgG ELISA with human serum. However, a repeat of the proficiency panel run at this laboratory following additional training of laboratory staff resulted in a repeatability estimate less than the upper acceptance bound; thus, illustrating the importance of rigorous training of laboratory staff and the strict adherence to assay procedures to ensure consistent results between runs.
Similarly, laboratories that validated the anti-EBOV GP IgG ELISA have used the following dilutional linearity (relative accuracy) acceptance criteria: the 90% confidence interval for the slope from the random regression model fit to data between the limits of quantitation and relating log 10 (concentration) to log 10 (spike level) will be entirely within (-1.20, -0.80). The interlaboratory study models dilutional linearity as log 10 (observed concentration) to log 10 (target concentration) resulting in a positive relationship between the two variables. Therefore, to conclude that dilutional linearity is acceptable in relation to the validation in human serum, the 90% confidence interval for the slope should be positive and fall entirely between 0.80 and 1.20. The overall slope was 0.95 and has a 90% confidence interval estimate of (0.93, 0.97); thus, the dilutional linearity is within the acceptance criteria as established in the assay validation with human serum.
Agreement among laboratories implies that the ratios of the mean concentration for the five individual labs to the overall laboratory consensus value for a given sample should be close to one. The ratios range from 0.95 to 1.08 for Laboratory A; from 0.96 to 1.19 for Laboratory B1; from 0.83 to 1.12 for Laboratory B2; from 0.96 to 1.16 for Laboratory C; from 0.71 to 0.97 for Laboratory D; and from 0.90 to 1.06 for Laboratory E. Equivalence test results showed that the 90% confidence interval for the ratio were within the equivalence bounds of 0.80 to 1.25 for each laboratory except for samples BMI-ZPP-13 and BMI-ZPP-20 in Laboratory B1, BMI-ZPP-19 in Laboratory B2, and six samples in Laboratory D.
The assessment of between-laboratory performance revealed lower observed concentrations at Lab D and greater variability in assay results at Lab B1 relative to the other laboratories. The lower observed concentrations at Lab D illustrate the importance of monitoring assay performance and harmonizing across laboratories. Given the inherent differences from subject-to-subject in clinical trials and animal-to-animal in non-clinical studies, these differences observed at Lab D relative to the other laboratories are not likely to affect interpretation of study results. The variability in assay results at Lab B1 was mitigated by additional laboratory staff training.
The evaluation of the proficiency panel at these laboratories provides a limited assessment of assay precision (intermediate precision, repeatability, and total assay variability), dilutional linearity, and accuracy. This limited evaluation suggests that the within-laboratory performance of anti-EBOV GP IgG ELISA as implemented at the five laboratories is performing consistently with the intended use of the assay based on the acceptance criteria used by laboratories that have validated the assay.
Supporting information
S1 Fig
Observed concentration (ELISA Units/mL) by laboratory for sample BMI-ZPP-11.
(Consensus Concentration = 695.93). Center line in the box depicts the median concentration while the height of the box represents the 25 th and 75th percentile of the concentration distribution. Vertical lines extending above and below the box represent the maximum and minimum concentration values for the laboratory. Open circles show the observed concentrations.
(TIF)
S2 Fig
Observed concentration (ELISA Units/mL) by laboratory for sample BMI-ZPP-12.
(Consensus Concentration = 475.20). Center line in the box depicts the median concentration while the height of the box represents the 25 th and 75th percentile of the concentration distribution. Vertical lines extending above and below the box represent the maximum and minimum concentration values for the laboratory. Open circles show the observed concentrations.
(TIF)
S3 Fig
Observed concentration (ELISA Units/mL) by laboratory for sample BMI-ZPP-13.
(Consensus Concentration = 325.47). Center line in the box depicts the median concentration while the height of the box represents the 25 th and 75th percentile of the concentration distribution. Vertical lines extending above and below the box represent the maximum and minimum concentration values for the laboratory. Open circles show the observed concentrations.
(TIF)
S4 Fig
Observed concentration (ELISA Units/mL) by laboratory for sample BMI-ZPP-14.
(Consensus Concentration = 844.08). Center line in the box depicts the median concentration while the height of the box represents the 25 th and 75th percentile of the concentration distribution. Vertical lines extending above and below the box represent the maximum and minimum concentration values for the laboratory. Open circles show the observed concentrations.
(TIF)
S5 Fig
Observed concentration (ELISA Units/mL) by laboratory for sample BMI-ZPP-15.
(Consensus Concentration = 871.34). Center line in the box depicts the median concentration while the height of the box represents the 25 th and 75th percentile of the concentration distribution. Vertical lines extending above and below the box represent the maximum and minimum concentration values for the laboratory. Open circles show the observed concentrations.
(TIF)
S6 Fig
Observed concentration (ELISA Units/mL) by laboratory for sample BMI-ZPP-16.
(Consensus Concentration = 561.81). Center line in the box depicts the median concentration while the height of the box represents the 25 th and 75th percentile of the concentration distribution. Vertical lines extending above and below the box represent the maximum and minimum concentration values for the laboratory. Open circles show the observed concentrations.
(TIF)
S7 Fig
Observed concentration (ELISA Units/mL) by laboratory for sample BMI-ZPP-17.
(Consensus Concentration = 226.25). Center line in the box depicts the median concentration while the height of the box represents the 25 th and 75th percentile of the concentration distribution. Vertical lines extending above and below the box represent the maximum and minimum concentration values for the laboratory. Open circles show the observed concentrations.
(TIF)
S8 Fig
Observed concentration (ELISA Units/mL) by laboratory for sample BMI-ZPP-19.
(Consensus Concentration = 110.63). Center line in the box depicts the median concentration while the height of the box represents the 25 th and 75th percentile of the concentration distribution. Vertical lines extending above and below the box represent the maximum and minimum concentration values for the laboratory. Open circles show the observed concentrations.
(TIF)
S9 Fig
Observed concentration (ELISA Units/mL) by laboratory for sample BMI-ZPP-20.
(Consensus Concentration = 70.81). Center line in the box depicts the median concentration while the height of the box represents the 25 th and 75th percentile of the concentration distribution. Vertical lines extending above and below the box represent the maximum and minimum concentration values for the laboratory. Open circles show the observed concentration.
(TIF)
S1 Table
ELISA concentration of each test sample—Laboratory A.
(XLSX)
S2 Table
ELISA concentration of each test sample—Laboratory B1.
(XLSX)
S3 Table
ELISA concentration of each test sample—Laboratory B2.
(XLSX)
S4 Table
ELISA concentration of each test sample—Laboratory C.
(XLSX)
S5 Table
ELISA concentration of each test sample—Laboratory D.
(XLSX)
S6 Table
ELISA concentration of each test sample—Laboratory E.
(XLSX)
|
Vytvor v slovenčine zhrnutie totho článku. Laicky vysvetli všetky základné pojmy v článku. Ak používaš úvodzovky, nezabudni, že v slovenčine sa prvé úvodzovky píšu dole. Používaj správne slovenské slová, vyhni sa čechizmom . Využi 8000 až 8100 tokenov. Tu je text článku: |
Introduction
There is a clear link between the combined activity of neurons and specific neural computations [1] , [2] . A common observation from population recordings is that the correlation between the activities of pairs of neurons can be modulated – for instance, by the spatiotemporal structure of stimuli [3] , [4] , the perceptual state of the subject [5] , [6] , or the spatial focus of attention [7] – [9] . Theoretical work has focused on the cellular and circuit mechanisms that both determine and modulate correlation [10] – [20] . However, the general applicability of these theories is unclear [21] , and how neural populations modulate the correlation between their spiking activity remains an open question.
One complication is that spike train correlations reflect common activity that may be measured at different timescales, ranging from a few (synchrony) to hundreds of milliseconds (co-variation of firing rates). For example, pairs of neurons in visual cortex [22] , [23] , olfactory bulb [24] , and attention responsive cortical areas [7] – [9] show increases in spike time synchrony which accompany simultaneous decreases of rate co-variation. To indicate the complex temporal aspects of this modulation, we label a differential change in correlation over distinct timescales correlation shaping [19] , [24] . In this study, we use a combination of in vivo recordings and computational modeling of electrosensory neurons to study how the spatial structure of a stimulus shapes the correlation of primary sensory neurons.
Weakly electric fish detect perturbations of their self-generated electric field through an array of electroreceptor neurons scattered on their skin surface which synapse onto pyramidal neurons within the electrosensory lateral line lobe (ELL) [25] . Relevant stimuli can be broadly categorized as either local , stimulating only a small fraction of the skin, or global , projecting to a broad area of the animal's body. Local inputs are a reasonable approximation to the spatial scale of prey inputs, while global inputs mimic communication calls from conspecifics [26] . We recorded simultaneously from pairs of ELL pyramidal neurons and found that global inputs increased spike train correlations at short timescales while simultaneously decreasing correlations at long timescales, when compared to the spike train correlation induced by local inputs. While there is a general understanding about how local and global stimuli control single neuron responses [26] – [30] , the cellular and circuit mechanisms that allow the spatial extent of stimuli to shape correlated population activity in the electrosensory system are a new area of study.
Based on the well-characterized anatomy and physiology of electrosensory circuits [25] , we developed a spiking network model of ELL pyramidal neurons that captured the experimentally observed correlation shaping. Diffuse inhibitory feedback was activated preferentially by global stimuli and provided a decorrelating signal that reduced correlations at long timescales. Further, global stimuli recruited feedforward circuitry that increased correlations at short timescales which were immune to feedback decorrelation. For sufficiently weak stimuli, we use a linear response framework [28] , [31] to show how correlation shaping is consistent with a shaping of the single neuron stimulus-response gain function. We tested our model predictions experimentally by selectively blocking feedback input, causing spike train correlations at long timescales to increase, rather than decrease. This directly demonstrates how inhibition can be a source of decorrelation to pyramidal neurons, rather than a source of synchrony as described in many previous studies [10] , [11] , [32] – [35] . Finally, we used our understanding of the population's response properties to study how feedback selectively attenuates responses to distractor stimuli, improving the system's ability to represent relevant signals. In total, our results reveal novel principles by which feedforward and feedback neural circuits are differentially activated by stimuli to shape population spike train correlations.
Methods
Ethics Statement
Animals were obtained from local importers and were acclimated to the laboratory as per published guidelines [36] . All experimental procedures were approved by the McGill University Animal Care Committee and have been described in detail elsewhere [37] .
Electrophysiology
Briefly, dual extracellular recordings from the lateral and centrolateral ELL segments of Apteronotus leptorhynchus were made using metal-filled micropipettes [37] . Pyramidal cells within these segments can be distinguished from cells within the centromedial segment based on recording depth, the medio-lateral and rostro-caudal positions of the recording electrode with respect to surface landmarks such as the “T0” vein and its afferent veins [38] , and their responses to sensory input as previously described [39] . Superficial pyramidal cells were identified based on their low ( ) whereas deep cells were identified based on their high ( ) mean firing rates in the absence of EOD modulations [26] , [30] , [40] . All data was sampled at 10 kHz.
Random amplitude modulations of the animal's electric organ discharge (EOD) consisting of white noise low-pass filtered with a cutoff of 120 Hz (8th order Butterworth filter) were presented either globally via two electrodes positioned on either side of the animal or through a dipole located close to the skin surface [37] . The stimulus lasted and consisted of 6 concatenated segments of the same frozen noise epoch that lasted 20 s [37] .
Pharmacological blockade of the indirect feedback from EGp was performed by either applying the non-NMDA glutamate receptor antagonist CNQX within the ELL molecular layer [30] or by applying a 2% lidocaine solution to the praeminential-cerebellar tract (PECB) as done previously [41] . Since both manipulations gave rise to similar effects, the data was pooled.
Data Analysis
Spike train cross-covariance functions
The recorded signals from a pair of neurons in response to the stimulus were thresholded in order to obtain the spike times , where is the number of spikes from neuron ( ). The spike train from neuron is then given by: (1) Here is the discrete approximation of the Dirac delta function with if and is zero otherwise; throughout so that at most one spike was contained in any time window. We note that this is equivalent to discretizing time in bins of width ms and setting the content of bin to when there is a spike time such that and to otherwise, as was done previously [30] .
The firing rate for neuron is then estimated as: (2) where is the duration of a recording (typically 120 s). The spike train covariance at time lag between neurons and is defined as: (3) where the number of time bins in the discrete spike train is . We refer to as the auto-covariance, while for is called the cross-covariance.
Spike count correlations
We also considered the correlations between the spike counts of pairs of neurons. The spike count from neuron is simply defined as the number of spikes occurring in the time window . It is a random integer given by: (4) For a given window size , we computed a sequence of spike counts from neuron as , using overlapping windows to increase the number of estimates. We have that , where denotes the mean value of the sequence . We can also obtain second order statistics from including the spike count variance and co-variance, which are defined by: (5) (6)
From these one can define the correlation coefficient between the spike counts and over a time window : (7) We use to denote the average value of across all pairs and similarly for other pairwise statistics. For small , the correlation coefficient measures the degree of synchrony between the two trains, while, for large , measures the co-variation in the firing rates of a pair of neurons [12] , [13] .
The variance and covariance functions of the spike count and spike train are related by: (8) These equations are the well known relations between second order spike count and spike train statistics [42] , with resulting from the convolution of the windowing function that converts spike trains to spike counts.
Within-trial vs. across-trial covariance functions and correlation coefficients
We note that both the spike train covariance function and correlation coefficient are within-trial measures of co-variability, since they incorporate both signal induced as well as trial-to-trial variable (i.e noise) aspects of common input fluctuations. Since we presented the same (i.e frozen) realization of the signal six times in succession, we were able to compute the spike train covariance and spike count correlation that were due purely to the common signal by computing joint statistics from neuron pairs recorded in different trials (i.e. across-trial). Specifically, denote the spike train of neuron in response to the realization of the stimulus ( ) by . The across-trial spike train covariance between neurons and is then given by: (9) In Eq. (9) , . Eq. (9) measures the joint spike statistics from neuron pairs when the spike trains were not recorded simultaneously but were stimulated with the same signal. This is because the summation runs over all possibly non-repeating combinations ( ) of the responses of each neuron to the six presentations of the frozen stimulus.
Similarly, one can define the spike count sequence for neuron during stimulus realization as . The across-trial spike count correlation coefficient between neurons and is then given by: (10) where Cov with the sequence of spike counts from the realization of the stimulus.
Linear Response Approximation
We use linear response theory in order to derive an expression for the correlation coefficient in terms of the stimulus gain, as done in past studies [12] – [14] , [19] , [28] , [31] , [43] , [44] . We consider the Fourier transform of the spike train covariance function as the length of the trial becomes large and assuming the processes are stationary: (11) Throughout, we will refer to with as the cross spectrum and as the power spectrum. To relate spike count statistics to spike train statistics, we use the Wiener-Khinchin theorem to rewrite Eq. (8) (assuming is large): (12) (13) with . Note that approaches a -function centered at 0 as and a constant function on as . Therefore, for large , only the zero-frequency components of the spectra contribute to the integral, while for small , all frequencies contribute. A similar relation holds between and .
For a fixed stimulus , we assume that [13] , [28] , [31] , [43] : (14) where is the Fourier transform of the mean-subtracted spike train given a particular realization of , is the Fourier transform of the stimulus, and denotes an expectation over repeated presentations of the stimulus. is the single neuron stimulus-response gain of the neuron (which we refer to as the stimulus gain for brevity). It relates the amplitude of the response to that of a signal at a particular frequency. For both experimental data and numerical simulations, we compute as: (15) where is the cross spectrum between and and is the power spectrum of the signal.
Assuming that the spike trains are conditionally independent given the stimulus, we can write , where denotes an expectation over the random stimulus. Substituting Eq. (15) into Eq. (14) , (16) Finally, combining Eqs. (13) and (16) yields the following approximation: (17) Eq. (17) relates the joint spike count variability to the stimulus gain , and has been derived in several past studies [13] , [19] . We can then approximate the predicted across-trial correlation as: (18)
Modeling
ELL anatomy
The neuroanatomy and physiology of the electrosensory system have been extensively characterized [25] . Pyramidal neurons in the ELL are subdivided according to several criteria. Roughly half of all pyramidal neurons have a basilar dendritic tree (BP neurons) and receive direct electrosensory afferent input. The other half lack a basal dendrite (nBP neurons) and receive afferent input only indirectly via interneurons [45] . Both BP and nBP neurons have an apical dendritic arbor; however, the extent of the arbor is variable across neurons. Pyramidal neurons with small apical dendritic trees are called deep neurons and do not receive much feedback input [30] , [45] , [46] . In contrast, pyramidal neurons with large apical dendritic trees are called superficial neurons and receive large amounts of feedback [30] , [45] , [46] . It has been recently shown [45] that the spatial projection of electroreceptor input to individual pyramidal neurons establishes a putative column, composed of BP and nBP deep and superficial pyramidal neurons.
The afferent and efferent projections between the ELL and higher brain structures further distinguish ELL pyramidal neurons. Indeed, only deep pyramidal neurons project to the praeminentialis dorsalis (Pd) [46] , a second order isthmic structure that directly projects to the posterior eminentia granularis (EGp), which in turn projects back to the ELL along the dorsal molecular layer via parallel fibers [25] that make synaptic contact onto the large apical dendritic trees of superficial pyramidal neurons. Thus, the deep ELL EGp superficial ELL feedback pathway can be characterized as open-loop [46] . Electrophysiological studies suggests that EGp granule cells show temporal locking to electrosensory input [46] , [47] and that the indirect feedback input onto ELL pyramidal neurons is in the form of a negative image of the stimulus that is activated by spatially diffuse but not by spatial localized stimuli [30] , [46] .
ELL model description
Our model of the deep pyramidal neuron to superficial ELL feedback via the nP and EGp contained three distinct neural populations: a deep (Dp) ELL population that projected to a population of granule cells in the EGp, which in turn provided feedback to a population of ELL superficial (Sf) neurons. All cells were modeled with leaky integrate-and-fire (LIF) dynamics [48] . Numerical values of model parameters can be found in Table 1 , and a detailed model summary [49] can be found in Table S1 . The membrane potential obeyed linear subthreshold dynamics supplemented with a spike-reset rule so that implied that , and was marked as a spike time. The deep population consisted of neurons, and the membrane potential of the deep neuron obeyed: (19) The first two terms of the right hand side of Eq. (19) model a static rest state and an intrinsic leak process, respectively. The process models Gaussian stimulus locked electroceptor activity, while models stimulus independent activity afferent to neuron in population ( ). As in the experiments, we set , but the temporal structure of the processes was white with , , and for or . The electroreceptor input contrast was set by and the correlation of the stimulus locked component by .
10.1371/journal.pcbi.1002667.t001 Table 1
Parameter values used in numerical simulations.
Parameter
Description
Value
Number of deep neurons
800
Number of EGp neurons
200
Number of superficial neurons
2
Deep membrane time constant
10 ms
EGp membrane time constant
10 ms
Superficial membrane time constant
15 ms
Deep bias
−56 mV
EGp bias
−60 mV
Superficial bias
−56 mV
Threshold voltage
−55 mV
Reset voltage
−65 mV
Noise strength
1 mV
Deep to EGp synaptic strength
mV
EGp to Superficial synaptic strength
mV
EGp to Superficial synaptic time constant
5 ms
Local input correlation
0.1
Global input correlation
0.2
The EGp population consisted of neurons, and the membrane potential of the EGp granule cell followed: (20) Here is the spike train from the deep neuron, and is the strength of excitation from the Deep ELL EGp. The time constant was chosen as 10 ms, based on recent measurements of input resistance for these cells of approximately 2 G [47] and data from cerebellar granule cells indicating typical capacitance values of 3–5 pF [50] – [52] .
Finally, since we are only interested in the pairwise correlation between superficial neurons and because the feedback is open-loop, it is only necessary to consider a pair of superficial pyramidal neurons. As such, we set . The superficial pyramidal cell's membrane dynamics are given by: (21) Here where is the Heaviside function. The operation denotes convolution. The inhibitory coupling from EGp to the ELL was set by .
During local stimulation, a fraction of deep neurons received coherent, stimulus-locked electroreceptor input ( ), while all other deep neurons received uncorrelated input modeling spontaneous afferent activity. During global stimulation, all deep neurons ( ) received stimulus-locked input ( ). The increased value of reflects the fact that global stimuli will spatially saturate the receptive field center and will thus more effectively drive the afferent population [29] , [53] .
In our model, a pair of neurons in a given layer could receive correlated input from the previous layer in two ways. First, a neuron in the previous layer could project to both downstream neurons and thus correlate their input. Second, neurons in the previous layer could become locked to the stimulus and their pooled activity could correlate the downstream neurons, even if their projections did not overlap anatomically. In the linear model, we assumed that the first source of common input is negligible relative to common input from stimulus locked, pooled activity, as is often the case in feedforward networks [54] . Consequently, correlations between model neurons were due only to external signals that synchronously recruited electroreceptors. Therefore, for the model.
To evaluate for our model using the linear response approximation, we computed the superficial neuron stimulus gain . For numerical simulations, we estimated using Eq. (15) . However, following past work [28] , [31] , we derived a theoretical approach to compute . For global stimulation and assuming that both the input correlations and the effective coupling and are sufficiently small, we compute the feedback filter from the Deep ELL EGp Superficial ELL using the serial computation (22) where is the Fourier transform of the exponential synaptic kernel . This result follows simply from the linear convolution of Deep ELL activity to EGp and then from EGp activity to superficial ELL through . Here we have introduced , the single neuron cellular response function (which we refer to as the cellular response for brevity) that measures a neuron's response to an applied current, independent of network feedback. can be computed using standard techniques from statistical mechanics (see Text S1 ).
We note that can be calculated for mixed excitatory and inhibitory feedback to superficial neurons. In this case, the value of should be interpreted as the effective input strength from both excitatory and inhibitory populations. For example, if the fraction of excitatory synapses from EGp to superficial neurons is given by and the synaptic strength of excitation and inhibition are and , respectively, then we have . Previous studies have established that the stimulus-locked EGp feedback is net inhibitory [46] , and we therefore model the pathway as purely inhibitory for simplicity.
With , we calculate the stimulus gain of a superficial ELL neuron as given in Eq. (25) . Further, these techniques also permit a calculation for the power spectrum . With theoretical expressions for and , and assuming the signal is Gaussian white noise with unit variance, we use Eqs. (17) and (18) to obtain a theoretical prediction for the spike count correlation between the two superficial ELL neuron spike trains: (23) Here we have used the homogeneity of the spike trains to set and for all superficial neurons.
Results
Correlation Shaping with Global and Local Stimuli
We examined the response of ELL pyramidal neurons to time-varying electrosensory input. Broadband electrosensory stimuli (Gaussian, 0–120 Hz) were applied to awake, behaving weakly electric fish ( Apteronotus leptorhynchus ; see Methods). Throughout the study, we delivered stimuli in one of two spatial arrangements: a local or global configuration [26] , [27] , [29] . In the local configuration, stimuli were spatially compact, delivered through a small dipole (tip spacing of 2 mm), and excited only a small region of the skin surface ( Figure 1A , left, blue). Local inputs mimic prey stimuli which drive only a spatially localized portion of the receptive field of an ELL pyramidal neuron [55] . In the global configuration, stimuli were spatially broad, delivered through a pair of electrodes located on each side of the animal, and affected the entire surface of the animal ( Figure 1A , left, orange). Global inputs mimic stimuli caused by conspecifics which drive nearly the entire surface of one side of the animal, stimulating both the classical and non-classical receptive field of a target pyramidal neuron [29] , [56] . During both local and global stimulation, simultaneous extracellular recordings of ELL pyramidal neuron pairs were collected ( Figure 1A , right). There was an intentional selection bias for superficial basilar pyramidal (BP) neurons [25] , since these neurons are known to receive feedback projections that shape their responses to sensory input [30] , [45] , [46] . Superficial neuron firing rates in the local and global configurations were similar ( and respectively).
10.1371/journal.pcbi.1002667.g001
Figure 1
The spatial extent of electrosensory stimuli shapes the temporal correlation between the spike times from pairs of ELL pyramidal neurons.
A , Stimulus protocol for local and global stimulation. Left: Gaussian distributed electric field stimuli with broadband spectral content (uniform over 0–120 Hz) were applied in a spatially compact (local) or diffuse (global) manner. Right: Paired extracellular recordings of ELL pyramidal neurons were made during stimulation. B1 , Spike train cross-covariance function in the local and global stimulus configuration for pairs of simultaneously recorded superficial BP neurons (within-trial correlation). Correlation function is normalized by firing rate. B2 , Same as B1 except computed between spike trains recorded during distinct trials. C1 , Within-trial spike count correlation as a function of window length ( ) in the local and global stimulus configuration. C2 , Across-trial spike count correlation as a function of window length in the local and global stimulus configuration. D1 , Ratio of global and local within-trial spike count correlations shown in panel C1. D2 , Ratio of across-trial global and local spike count correlations shown in panel C2. The data set consists of n = 10 pairs of neurons, and all curves are population average quantities. In all panels, shaded regions denote standard error.
We used the simultaneous unit recordings to estimate the spike train cross-covariance function (see Methods Eq. 3 ) for neuron pairs in both the local and global stimulus configurations. Global stimulation set a narrow peak of the cross-covariance function with a high maximum at zero lag, while it was broad with a lower peak value for local stimulation ( Figure 1B1 ), consistent with previous reports [37] .
To quantify this shift in covariance at different timescales, we computed the correlation coeffcient between the spike counts of neuron pairs' outputs [22] , [42] . This provided a normalized measure of the similarity between the two spike trains as observed over windows over a particular length (see Methods Eq. 7 ). At small window sizes ( ), spike count correlation was larger during global stimulation than during local. For large window sizes ( ), this relationship was reversed ( Figure 1C1 ). Correlation is generally a rising function of window size [57] , since for small few spikes will occur in the same window. However, even small values of correlation (e.g. in magnitude) have substantial influence on the propagation of neural information [54] , [58] and neural coding [59] . To provide a relative measure of the shift in correlation between the two states, we considered the ratio of global correlation to local correlation. This was a decreasing function of window size which was substantially greater than 1 for small window sizes and lower than 1 for large window sizes ( Figure 1D1 ).
We performed statistical tests to confirm that the trends observed were significant. Nonparametric tests confirmed that the distributions for the local and global conditions were different ( , evaluated at , , two-sample Kolmogorov-Smirnov test). The trends with timescale were also significant ( , compared with , for local and for global stimulation, two-sample Kolmogorov-Smirnov tests). The means of the distributions were also different ( , evaluated at , , paired t-test). In summary, the spatial extent of the electrosensory signal shaped the timescales over which spike train pairs were correlated.
Shifts in Single-Neuron Response Gain Predict Correlation Shaping
In general, correlated neural activity can be decomposed into stimulus induced and non-stimulus induced components [21] , [60] . Stimulus induced correlations reflect the two neurons locking to a dynamic stimulus, while the non-stimulus induced correlations reflect the neurons sharing a portion of their trial-variable noise, presumably from a common pre-synaptic source. To uncover the cellular and circuit mechanisms underlying correlation shaping, we first determined whether the changes in correlation observed were present across trials and therefore related to how neurons responded to the repeated stimulus. Using spike trains across different trials of identical stimulus presentations, we computed the across-trial spike train cross-covariance functions and spike count correlations ( Figure 1B2,C2 ; see Methods Eqs. 9 , 10 ). The magnitude of these across-trial correlations was less than that of the within-trial correlations, indicating the presence of some trial-variable noise (compare Figure 1C1 and 1C2 ). Nevertheless, the differential shaping of correlations at short and long timescales was still present in the across-trial spike count correlation ( Figure 1C2,D2 ). This suggests that the way stimulus processing shifts between local and global conditions is related to the mechanisms responsible for correlation shaping.
To investigate this relationship, we considered the stimulus gain , which measures a neuron's response to an external electrosensory stimulus at frequency ( Figure 2A , see Methods Eq. 15 ). We computed the gain in the two stimulus conditions and found that during local stimulation, the gain function was low-pass, while during global stimulation, it was high-pass ( Figure 2B ), consistent with previous studies [29] , [30] . We then asked if the observed changes in correlation could be related to this shift in frequency selectivity.
10.1371/journal.pcbi.1002667.g002
Figure 2
Shifts in stimulus gain predict spike train correlation shaping.
A , Schematic illustration of stimulus gain. The gain is described as the ratio of the change in the output firing rate of a neuron that is evoked by an input sine wave stimulus of amplitude . B , Gain for neuron pairs during local and global stimulation. The signal was assumed to have unit amplitude. C , Across-trial spike count covariance (solid) and the prediction from a linear response theory (dashed, see Methods Eq. 17 ), in both global and local stimulus conditions. The data set consists of n = 10 pairs of neurons, and all curves are population averages. In all panels, shaded regions denote standard error.
Motivated by past studies [12] , [13] we assumed that the cross-spectrum between the spike trains was proportional to the product of their stimulus gain functions (see Methods Eq. 16 ). This amounts to assuming that the common stimulus is the only source of correlation in the neural responses. This theory predicts that the correlation for large window sizes is determined by stimulus gain at low frequencies. In contrast, correlation for small windows involves gain at high frequencies. The shift in from low frequency transfer for local inputs to high frequency transfer for global inputs therefore implies global stimulus correlation will be enhanced for small and attenuated for large , with the inverse true for local stimulation. We verified this hypothesis, obtaining a prediction of the spike count correlation in the two states that matched the experimental data (see Methods Eq. 18 ; Figure 2C , solid versus dashed curves). Thus, the shift in the frequency-selectivity of superficial neurons' stimulus gain between the local and global conditions indeed predicted the changes in correlation.
Modeling ELL Pyramidal Cell Responses
To understand mechanisms behind the shift in neuronal responses under the local and global stimulus conditions, we constructed a simplified population model of ELL pyramidal neurons based on known anatomical and functional data as well as our experimental results ( Figure 3A ; for a detailed discussion of the anatomy, see Methods). This model captured two generic circuit features that modulated population responses: feedforward sensory input and feedback inhibition. All pyramidal neurons received feedforward electrosensory input via electroreceptors, but were divided into two classes based on their feedback afferents: deep neurons did not receive feedback from higher regions, but superficial neurons did receive inhibitory feedback. This feedback arrived from the posterior eminentia granularis (EGp), which was in turn innervated by the deep neurons. In total, this structure formed an open-loop inhibitory feedback pathway, from deep neurons to EGp neurons to superficial neurons. Motivated by past studies, ELL pyramidal neurons were modeled as simple leaky integrate-and-fire units [27] , [28] , [46] . Consistent with experimental data [30] , superficial firing rates in the model were lower than deep firing rates (12 Hz and 36 Hz, respectively) in both local and global stimulation conditions.
10.1371/journal.pcbi.1002667.g003
Figure 3
Open loop feedback inhibition in electrosensory neural circuitry.
A , Detailed schematic of peripheral neural circuitry in the electrosensory system. Basilar (BP) and non-basilar (nBP) pyramidal neurons in the electrosensory lateral line lobe (ELL) have their somata located in the Pyramidal cell layer (PCL). Deep pyramidal neurons (green) have small apical dendritic arbors, projecting only to the Ventral Molecular Layer (VML). In contrast, superficial pyramidal neurons (red) have large apical dendritic arbors, projecting to the Dorsal Molecular Layer (DML). Pyramidal neurons receive direct and/or indirect input from feedforward electroreceptor afferent input to the Deep Fiber Layer (DFL). Deep pyramidal neurons excite neurons in the praminentialis dorsalis (Pd), which in turn excite granule cells in the posterior eminentia granularis (EGp). The EGp projects parallel fiber feedback along the DML exclusively targeting ELL superficial pyramidal neurons. In total the deep ELL EGp superficial ELL pathway is an open loop feedback structure. Pyramidal neuron graphics were from example neurolucida traced neurons [46] . B , Stimulus correlation for pairs of experimentally recorded deep pyramidal neurons (n = 45 pairs; 10 neurons were used) that were driven by the stimulus in local and global (bottom). Little correlation shaping is present. For comparison purposes we show the stimulus correlation for pairs of superficial neurons (top, Figure 1C2 ). C , Simplified model of the ELL-EGp circuit. Individual neurons in the deep ELL, EGp, and superficial ELL were modeled with leaky integrate-and-fire neuron dynamics (example realizations on right). Electroreceptor input was modeled as white noise, with 5% of deep pyramidal neurons receiving a stimulus-locked component in local and 100% in global. We studied the spike responses the pair of superficial pyramidal neurons (labeled 1 and 2) that receive both afferent and EGp feedback inputs.
Previous studies have shown that EGp feedback modulates both the static [41] and dynamic [30] gain of single neuron responses. However, how it controls the ELL population response, and in particular correlations between pyramidal neurons, is unknown. To determine whether feedback is responsible for stimulus-dependent correlations, we recorded from deep pyramidal neurons receiving a frozen stimulus and computed stimulus correlations between the pairs of spike trains. Consistent with the lack of feedback projections to this subpopulation, these neurons did not show substantial shaping of correlations between the local and global conditions ( Figure 3B , bottom), in contrast with superficial pyramidal neurons ( Figure 3B , top). The small decrease in correlation for large time windows observed during global stimulation for deep neurons ( Figure 3B , bottom) is consistent with these neurons receiving little feedback input [40] .
Recruitment of Feedback in the Model During Local and Global Stimulation
We used our model to examine the stimulus dependence of EGp feedback. In our model, electrosensory stimulation caused the firing of deep pyramidal neurons to become stimulus-locked. When the stimulus was local, only a small fraction of this population was stimulus-locked, so that the average correlation across the deep population was low ( across the population, Figure 4B1 ). The weak stimulus correlation across the deep population failed to recruit coherent activity in the EGp granule cell population, resulting in a near tonic inhibitory feedback to the ELL ( Figure 4C1 ). In contrast, when the stimulus was global, the entire deep population was correlated by the stimulus ( , Figure 4B2 ). This led to a dynamic, stimulus locked EGp feedback to the superficial neuron pair ( Figure 4C2 ). Thus, our model captured a link between the temporal locking of EGp feedback and the spatial extent of the external stimulus, which has been suggested in past experiments [46] , [47] .
10.1371/journal.pcbi.1002667.g004
Figure 4
Model EGp feedback is stimulus locked for global, but not local, stimulation.
A Low-pass (0–60 Hz) filtered version of the electrosensory stimulus. Filtering was done as a visual aid in relating the stimulus to the feedback in (C2). B1 , Raster plot of the deep neuron population during local stimulation. The signal weakly correlated only a small fraction of the population. B2 , Same as (b1), but during global stimulation. The spatially broad stimulus correlated the entire deep population. C1 , EGp feedback current during local stimulation, showing little stimulus locking. C2 , EGp feedback was stimulus-modulated by the global signal, due to recruitment of the deep population by the stimulus. The inhibitory feedback is a negative image of the stimulus (A2).
Having characterized the EGp feedback, we next determined how it shaped the responses of superficial neuron pairs. The total input to a model superficial pyramidal neuron, from both feedforward and feedback sources, is: (24) Here is the strength of the afferent activity to an ELL pyramidal neuron and and are Gaussian white noise processes modeling stimulus locked and unlocked (noise) afferent inputs, respectively. The parameter is the fraction of receptor afferents that are stimulus-locked, which determines the correlation between the electroreceptor input to neuron pairs. The function is the parallel fiber feedback kernel and involves compound processing of the stimulus by the population of deep ELL neurons, the EGp granule cells, and finally the inhibitory feedback pathway from the EGp to the ELL (see Methods Eq. 21 ). Assuming weak stimulus correlations (small ) and weak EGp feedback, we use linear response theory [28] , [31] , to obtain an expression for the stimulus gain of a superficial pyramidal neuron (see Methods): (25) Here is the Fourier transform of the feedback kernel (see Eq. 22 in Methods), and is the cellular response of a superficial neuron, which measures its response to a fluctuating current applied directly to the neuron (see Eq. 8 in Text S1 ). In contrast to the stimulus gain, the cellular response does not depend on network feedback. The parameter is the spatial extent of the stimulus ( ), with modeling the lack of stimulus-coherent EGp feedback for local stimulation, and the full recruitment of EGp feedback for global stimulation ( Figure 4 ). With this model of how shifts between local and global stimulus configurations, we next build a theory for the correlation shaping within the superficial ELL pyramidal neuron population.
Correlation Shaping in the ELL-EGp Network Model
We used our ELL-EGp network model to relate the spatial extent of an electrosensory stimulus and the timescale of the pairwise correlation between spike trains from superficial BP neurons. During local stimulation, pairs of nearby superficial neurons received correlated electroreceptor input ( Figure 3C ). The degree of correlation between the afferent input to the superficial pair was . The EGp feedback did not exhibit a substantial stimulus-locked component ( ) during local stimulation, and hence did not contribute to common fluctuations ( Figure 4C1 ). Thus, the stimulus gain in the local condition, denoted , reduced to: (26) Our theoretical (see Methods) quantitatively matched estimates from simulations of the ELL-EGp network of leaky integrate-and-fire neurons ( Figure 5a , blue curve and blue dots) and qualitatively matched the low-pass nature of obtained from experiments ( Figure 2B , blue). The calculation demonstrates that the gain to local stimuli of superficial pyramidal neurons is primarily determined by the cellular response , suggesting that feedback network dynamics can be ignored.
10.1371/journal.pcbi.1002667.g005
Figure 5
Model ELL-EGp network captures correlation shaping between local and global stimulation.
A , Stimulus gain of superficial BP neurons in the model (compare to Figure 2B ). Our analytical theory (solid) matches the simulation results from the ELL-EGp network (dots). B , Correlation between superficial BP neuron pairs during local and global stimulation of the model (compare to Figure 1C ). Since our theory predicts a linear relationship between output correlation and input correlation, the output is shown in units of input correlation in the local state , which was 0.1 in simulations. C , Idealized schematic illustrating the effect of feedback on shared fluctuations. Left: local inputs fail to recruit EGp feedback via deep population (see Figure 4 ), so common input arises purely through feedforward stimulus drive. Center: Low frequency global input recruits a negative image of the stimulus, which cancels the common input to the pair of superficial pyramidal neurons. Right: The cancellation signal is weak for high frequency global inputs due to the low-pass nature of the feedback. Hence, the common fluctuations are not cancelled.
The lack of network activity for local stimulation ( ), was contrasted with the recruitment of EGp feedback for global stimulation ( ). During global stimulation, we also assumed that the receptive fields of neurons were fully saturated, rather than being partially driven due to the limited extent of the stimulus, as suggested by experimental estimates [53] . We therefore increased the correlation of electroreceptor afferents in the global state, so that . Combining these two model assumptions, we expressed the gain in the global configuration, , as: (27) If – that is, if the negative feedback were a perfect replica of the feedforward signal – the stimulus gain would be zero, indicating complete stimulus cancellation by the feedback pathway. However, since the negative feedback was low-pass due to neuronal integration and synaptic filtering along the feedback pathway, only the low frequency components of the gain were strongly attenuated. Consequently, for sufficiently low frequencies ( Figure 5A , compare orange and blue curves for ). However, for high frequencies ( Figure 5A , compare orange and blue curves for ), because of the increase in receptive field saturation ( ). Our theoretical matched simulations of the ELL-EGp network ( Figure 5A , orange curve and orange dots). Thus, the combination of feedback recruitment and feedforward saturation during global stimulation captured the experimentally determined shift in stimulus gain known to occur between local and global stimulation ( Figure 2B and see [29] , [30] ).
Next, we examined how this gain shift controlled correlations across the population of superficial pyramidal neurons. Using the linear response theory we used to predict signal correlations in the experimental data ( Figure 2 , see Methods Eq. 23 ), we calculated theoretically the correlations between model pyramidal neurons. Global stimulation simultaneously increased short correlation and decreased long correlation compared to local stimulation ( Figure 5B ). These findings matched the experimental results (compare Figures 1C and 5B ) and are the primary theoretical result of this study.
Our model provides clear intuition for how the combination of receptive field saturation and the recruitment of EGp feedback during global stimulation shapes the correlation of ELL pyramidal neuron activity ( Figure 5C ). During local stimulation, EGp feedback was not recruited and the feedback did not cancel the feedforward signal from the electroreceptors ( Figure 5C , left). This case is contrasted with global stimulation, in which a broad stimulus-induced synchronization of all of the deep ELL neurons recruited a stimulus-locked EGp feedback. This feedback was low-pass, and therefore canceled the low frequency components of the signal ( Figure 5C , middle), but not the high frequency components ( Figure 5C , right). Thus, correlations due to global stimulation were canceled only for sufficiently long timescales ( Figure 5B , ). Furthermore, the saturation of the receptive field input ( ) enhanced the correlation for small ( Figure 5B , ). In total, feedforward and feedback circuitry shaped depending on the spatial profile of the electrosensory signal.
Our ELL-EGp network model distills correlation shaping into two hypotheses that link the spatial properties of an electrosensory stimulus and the timescale of pairwise correlation between the spike responses of ELL superficial pyramidal neurons:
Receptive field saturation for spatially broad signals increases the short timescale correlation between the spike trains from superficial pyramidal neurons.
Recruitment of EGp feedback by spatially broad signals decreases the long timescale correlation between the spike trains from superficial pyramidal neurons.
To study these two components of correlation shaping in isolation from one another, we used a combination of analysis on a subclass of ELL pyramidal neurons and pharmacological blockade of EGp feedback.
Correlation Shaping of nBP Neuron Responses
We first tested how short timescale correlation was affected by receptive field saturation (Hypothesis 1). The ELL has two classes of pyramidal neuron: non-basilar pyramidal (nBP) and basilar pyramidal (BP) neurons, distinguished by the extent of their basilar dendritic arbor ( Figure 3A ). While BP neurons respond to positive deflections of the electric field, nBP neurons are oppositely tuned, due to their afferent inputs arriving solely via an inhibitory interneuron population [25] . This difference in the feedforward afferent architecture to nBP neurons compared to BP neurons produces nBP neuron classical receptive fields that are smaller than those of BP neurons [26] . Despite the difference in feedforward afferent input for BP and nBP neurons, both superficial BP and nBP neurons receive near equivalent feedback from EGp parallel fibers ( Figure 3A ). Thus, a comparison between BP and nBP neurons is sensitive to a difference in feedforward afferent drive, and not to EGp feedback. We hypothesized that global inputs would not drive nBP neurons as strongly as BP neurons because of their smaller classical receptive fields. Hence, short timescale correlation during global stimulation for nBP neurons should be less than for BP neurons.
We first calculated the stimulus gain for nBP neurons. The difference in gain between local and global stimuli for nBP neurons was different than that for BP neurons ( Figure 6A1 ; [30] ). In particular, while nBP and BP neurons both exhibited a reduction in low frequency gain during global stimulation, nBP neurons exhibited little enhancement of high frequency response. Our model network replicated this difference ( Figure 6A2 ) when we assumed that the nBP neurons integrate stimuli over smaller regions of space, such that local inputs saturate the receptive field ( ), in contrast to the BP neuron case ( ). The lack of high frequency shaping of gain for nBP neurons across local and global configurations predicts that the small correlations do not substantially increase in the global state, while EGp feedback still attenuates low frequency gain and hence large correlations. Measurements of for nBP neurons in both the ELL-EGp model ( Figure 6B2 ), as well as nBP neurons recorded in vivo ( Figure 6B1 ) supported this prediction. Thus, the known differences between the receptive field sizes of nBP and BP neurons, provide evidence for the link between the spatial extent of electrosensory stimuli and short timescale correlation observed for superficial BP neurons.
10.1371/journal.pcbi.1002667.g006
Figure 6
Saturation of the receptive field for both local and global stimuli makes short timescale response insensitive to the spatial extent of electrosensory stimuli.
A1 , Experimental stimulus gain for nBP neurons (n = 14) in local and global stimulus configurations. The gain for BP neurons in the global configuration is shown for comparison (see Figure 2B ). A2 , Stimulus gain for model nBP neurons ( ) in local and global configurations, and the model BP neurons ( ) in global for comparison. B1 , Recorded spike count correlation over windows of length for pairs of nBP neurons. As with BP neuron pairs, firing rates in the local and global states were similar ( and , respectively). B2 , Spike count correlation for pairs of model nBP superficial neurons in the ELL-EGp network. For the model results (A2,B2) our analytical theory (solid) matches the simulation results from the ELL-EGp network (dots). Values are shown in units of input correlation in the local state .
Feedback Inhibition Cancels Long Timescale Correlations
We next tested how long timescale correlation is affected by recruitment of EGp feedback by global stimuli (Hypothesis 2). In our model, the EGp feedback was responsible for the decrease in low frequency stimulus gain and long timescale correlation in the global state. To experimentally confirm that this pathway was responsible for these effects, we pharmacologically blocked feedback from EGp to superficial neuron pairs (see Methods). We first tested whether attenuation of low frequency components of the stimulus gain was removed by the block. In experiments with global stimulation, we found that firing rates during the block were decreased significantly from the control condition (block: ; control: , , paired t-test). We remark that while the net impact of EGp feedback may be excitatory, the signal locked components of EGp feedback are thought to be inhibitory [46] , consistent with our model. To correct for the change in firing rates across control and block conditions, we normalized the gain by firing rate to show the relative modulation of firing rate by the stimulus. The normalized gain increased at low frequencies, yet remained unchanged at high frequencies ( Figure 7B1 , compare orange and gray curves), consistent with model predictions ( Figure 7B2 ). This effect was removed after a washout of the drug ( Figure 7B1 , compare orange and light orange curves).
10.1371/journal.pcbi.1002667.g007
Figure 7
EGp feedback reduces correlations on long timescales when stimuli are global.
A Schematic indicating block of feedback with CNQX in the ELL circuit. B1 , Stimulus gain for individually recorded superficial BP neurons in control, block, and recovered conditions. Gain is normalized to output firing rate in the data. B2 , Stimulus gain for model superficial neurons for global stimuli when feedback was intact or absent. C1 , Left: Spike count correlation at and 200 ms for paired recordings of superficial BP neurons. Right: Spike count correlation as a function of for individual recordings with a frozen stimulus in control, block, and recovered conditions. The standard error bars overlap for both the pre-drug and recovery curves, while they do not overlap with those for the block. Differences between control and recovered conditions could be due to incomplete drug washout or the preparation being in different states before and after the application of the drug. C2 , Spike count correlation as a function of for model neuron pairs when feedback was intact or absent. Values are shown in units of input correlation in the local state .
The spike count correlations for simultaneously recorded superficial neurons in the global state with and without pharmacological block of feedback verified its role in shaping long timescale correlations. Specifically, the spike count correlations for showed a significant increase during the block ( , paired t-test), while correlations for were similar ( Figure 7C1 ; left). Due to the difficulty in obtaining paired recordings under pharmacological blockade, we further verified our theory with units recorded individually with frozen noise in the global state with and without pharmacological block of EGp feedback ( Figure 7C1 ; right). Correlations at long timescales were increased during the block compared to control ( Figure 7C1 ; left, compare orange and gray curves) and recovered to control values after drug washout ( Figure 7C1 ; left, compare orange and light orange curves), consistent with our model ( Figure 7C2 ). Thus, despite EGp feedback being a source of common synaptic input to a pair of superficial ELL pyramidal neurons, removing it during global stimulation increased the spike correlation between the neuron pair. In total, these data supported our second hypothesis: stimuli with large spatial extent recruit inhibitory feedback that cancels the input correlation expected from feedforward afferent projections.
Correlation Shaping and Population Coding of Natural Electrosensory Scenes
We have presented a general mechanism for how spike train correlations from pairs of ELL pyramidal neurons are shaped by the spatial extent of an electrosensory signal. We explored the mechanism with simple noise signals categorized into either spatially local or global inputs. However, natural electrosensory scenes are complex, with a broad range of spatial and temporal scales. In this section, we speculate on how correlation shaping influences the population representation of natural electrosensory scenes.
Sensory systems must produce high fidelity representations of biologically relevant signals, while ensuring that distractor inputs do not contaminate the neural code. The ELL pyramidal neuron population is responsible for coding two distinct electrosensory inputs. First, electric fish routinely perform prey detection, tracking, and capture, during which prey organisms produce electric images with low frequency components ( ) that stimulate a limited portion of the animal's electroreceptive field [55] . Second, electric fish use their electric organ to communicate with conspecifics, using signals that contain primarily high frequency components ( ) and drive a large region of the skin [56] , [61] . However, these two signals often coexist with distractor inputs that the electrosensory system must ignore. Natural distractors arise from the superposition of background electric fields from many neighboring fish [62] , or self generated signals from body and tail positioning [47] . These inputs consist of mostly low to mid range frequencies ( ) and drive a broad sensory area. A critical sensory computation in the ELL is the pyramidal neuron population faithfully locking to prey and communication signals, with minimal locking to distractor electrosensory inputs. The linear response analysis of the ELL-EGp network suggests that EGp feedback to the ELL plays an important role in this computation.
Using our linear theory, we calculated the response of a population of superficial BP neurons to mixed signal and distractor input, with and without EGp feedback. The signal was either a local 4 Hz sine wave ( Figure 8A1–D1 ), or a 50 Hz global sine wave ( Figure 8A2–D2 ). In both cases, the distractor input was 0–10 Hz broadband noise. The population response was modulated by the signal and the distractor, with relative strengths determined by the corresponding gain ( Figure 8D ). To test how EGp feedback affects the coding of relevant signals, we computed the signal to noise ratio (SNR) of this population response, defined as the ratio of the signal power integrated over all frequencies to the distractor power integrated over all frequencies. For both the 4 Hz local and 50 Hz global signals, the SNR was greater with feedback than without feedback ( Figure 8B,C . SNR decreased from 2.3 to 0.70 for the 4 Hz local signal and from 2.8 to 0.70 for the 50 Hz global signal when feedback was removed). This is because EGp feedback was recruited by distractor input, attenuating any distractor induced correlation (low gain for distractor inputs in Figure 8D1,D2 ). In contrast, prey inputs lacked sufficient spatial power to recruit EGp feedback, meaning an EGp cancellation signal was not passed and ELL population stimulus gain was high ( Figure 8D1 ). Communication calls have large spatial power, yet their high frequency power cannot be transmitted by the low pass parallel fiber pathway, again meaning ELL population stimulus gain was high ( Figure 8D2 ). The ELL-EGp network was therefore capable of removing spurious correlations due to distractors while still coding for relevant signals.
10.1371/journal.pcbi.1002667.g008
Figure 8
EGp feedback cancels the ELL population response to global distractor inputs but not prey or communication signals.
A1 , Schematic of response to a prey signal, which occupies a limited spatial extent and contains power at low frequencies. A2 , Schematic of response to a communication call from a conspecific, which is a global, high frequency signal. B1 , Average population firing rate for ELL neurons responding to a local, 4 Hz signal (red) and the same signal with 0–10 Hz distractor noise (black). The SNR was 2.3. B2 , Same as B1, but with a global, 50 Hz signal. The SNR was 2.8. C1 , Same as B1, but without EGp feedback. The SNR was reduced to 0.70. C2 , Same as B2, but without EGp feedback. The SNR was reduced to 0.70. D1 , ELL pyramidal neuron stimulus gain for local inputs (which do not recruit feedback) and global inputs with and without feedback. The frequency of the signal is marked. Note that because the distractor is a global 0–10 Hz signal, its transfer will be enhanced by the removal of feedback, reducing SNR (compare gray and orange curves). D2 , Same as D1 but with a global, 50 Hz signal. Since the signal is high frequency, its stimulus gain is not substantially affected by feedback.
Discussion
Temporal shaping of correlated spiking activity has been observed in a variety of systems [7] , [9] , [19] , [22] – [24] . We have shown that the spatial extent of an electrosensory signal controls the timescale of correlation between the spiking outputs of principal neurons in the ELL of weakly electric fish. Specifically, an increase in the spatial extent of a signal increased pairwise spike time synchronization, while simultaneously decorrelating long timescale rate co-variations. Using a combination of computational modeling and targeted physiological analysis, we identified that correlation shaping in the ELL is mediated both by an increase in the strength of feedforward afferent drive and the recruitment of a feedback pathway for spatially broad signals. Electric fish offer a neuroethologically inspired functional context for correlation shaping, where it promotes an accurate population representation of relevant signals, even in the presence of distractor inputs. The generic circuit features that support correlation shaping and its use in feature selective population temporal codes suggest that the basic principles exposed here may be at play in other neural systems.
Correlation Shaping with Neural Architecture in the Electrosensory System
There has been extensive investigation of the gain shifts of single ELL pyramidal neurons between local and global stimulus configurations [26] – [30] , [46] . These studies have shown that both feedforward and feedback mechanisms mediated these shifts. Indeed, pharmacological manipulations of descending feedback to the ELL provided strong evidence for its role in controlling gain shifts of single unit response at low frequencies [27] , [29] , [30] , [46] . However, previous studies have shown that local stimuli only excited a fraction of the receptive field center [26] , [29] and that spatial saturation of the receptive field center mediated the gain shifts of single unit response at high frequencies only by recruiting a greater fraction of feedforward afferent input [29] , [30] . This importance of feedback activity prompted network models of the ELL and higher brain regions, and these models captured the sensitivity of single unit dynamics to the spatiotemporal structure of electrosensory stimuli [27] , . However, the models relied on heretofore untested assumptions about the population spike train statistics of ELL pyramidal neurons. In parallel to these single-unit studies, other work presented simultaneous recordings from pairs of ELL pyramidal neurons showing significant stimulus evoked correlation in spike activity [63] , and that the spike train correlation is sensitive to a stimulus' spatiotemporal structure [37] . However, these studies did not attempt to relate the dependence of pairwise statistics on stimulus structure to the extensive ELL single neuron experimental gain and network modeling literature. Our study merges the two avenues of research and shows that pairwise correlation shaping is related to gain shifts, as our linear response treatment of the ELL-EGp network model predicts. Thus, our analysis directly tests the proposed feedback mechanisms for single neuron response shifts [30] .
Previous studies of the ELL have focused on the generation of oscillations due to feedback from area nP to pyramidal neurons (the direct feedback pathway) [27] . Theoretical studies have demonstrated that such oscillations arise from a combination of spatially correlated noise and delayed inhibitory feedback [28] , [31] . Unlike neurons receiving closed-loop inhibitory feedback from nP, the superficial pyramidal neurons modeled in our study lack input from this direct pathway, and hence do not exhibit oscillations. Superficial neurons were excluded from the analysis in [27] and [28] , so that the results of our study concern a cell class that is distinct from these previous studies. This distinction emphasizes the qualitative differences in the dynamics induced by open- and closed-loop feedback pathways.
We used well-characterized anatomical data and pharmacological manipulation to study the network architecture that codes for time-varying electrosensory stimuli. This is in contrast to techniques such as the generalized linear model [64] that statistically determine the spike response and network filters that generate a response to a sensory signal with fixed statistics. Our approach allowed us to study the response of the system in distinct stimulus conditions, with varying input statistics. Further, network coupling suggested clear architectural predictions for the mechanisms behind correlation shaping (hypotheses 1 and 2). These predictions were validated with a combination of the known heterogeneity of ELL feedfoward architecture ( Figure 3 ), and a pharmacological blockade of feedback activity ( Figure 7 ). Organisms exist in environments with ever-changing sensory statistics yet must code these environments, often with a single neural population. Our study shows how neural architecture can help shift the response dynamics of neural populations as signals change to better meet this computational need.
Our results also highlight how architectural differences may lead to differential population activity in different layers. Recently, it has been shown that synchronization between neurons in visual cortex is layer-dependent [65] . Furthermore, the cognitive demands of a task may control the recruitment of feedback and influence spike train correlations [66] . Our results demonstrate that both layer-specific recruitment of feedback and connectivity profiles influence correlated population activity.
Finally, theoretical communities have recently made some progress in understanding how network architecture combines with cellular dynamics to determine the correlation between pairs of cells [44] , [67] , [68] . However, the work is general, and a clear neural motivation to base a concrete example upon is lacking. Our study demonstrates that the electrosensory system offers a prototypical system where cellular dynamics, a clear feedforward/feedback architecture, and a single stimulus feature (spatial extent) interact to shape the temporal structure of pairwise spike train correlation.
Decorrelating with Inhibition
The role of inhibition in neural circuits is a complex topic of study. Inhibition is linked to rhythmic, temporal locking between pairs of pyramidal neurons [32] . On fast timescales, inhibition is often thought to synchronize the activity of pairs of pyramidal neurons in both recurrent [10] , [27] , [33] – [35] and feedforward architectures [11] , [32] . However, on longer timescales, inhibition mediates competitive dynamics between populations of pyramidal neurons, and as such may be a source of anti-correlated activity [24] . Recently, studies of densely coupled cortical networks with balanced excitation and inhibition [17] , [18] and feedforward inhibitory cortical circuits [20] , [69] have provided new insights into the role of inhibitory dynamics. In these studies, fluctuations in correlated excitation to a pair of pyramidal neurons are cancelled by correlated inhibitory dynamics, yielding a roughly asynchronous cortical state. This cancellation of correlation is similar to the one explored in our study responsible for the reduction of correlation for global stimuli. However, our study was motivated by a primarily feedforward sensory architecture in which an external signal can drive correlated activity.
The strengths of the electrosensory preparation allowed us to extend the correlation cancellation mechanism along two important directions. First, the ease in controlling the spatiotemporal properties of external stimuli allowed an analysis of the limitations of correlation cancellation. The diffuse ELL EGp feedforward path restricts correlation cancellation to signals with broad spatial scale, while the slow filtering by the parallel fiber pathway can only cancel correlations of low frequency stimuli. Second, the well segregated parallel fibers that mediate EGp feedback to the ELL permitted a pharmacological blockade of inhibition, directly providing evidence for correlation cancellation. The parallel fibers are a source of common input to pyramidal neurons, and a naive analysis would predict that their removal would thus decrease pyramidal neuron spike train correlation. Nevertheless, the blockade of parallel fiber inputs increased the spike train correlation, suggesting that the common inhibition interacts with the common feedforward afferent input in a destructive, rather than cooperative, manner.
Studies of neural codes often investigate the distinction between signal evoked, across-trial correlations and additional ‘noise’ induced, within-trial correlations [60] . Across-trial correlations are attributable to a dynamic locking of the spike train pairs to the common signal. Within-trial correlations measure the trial-to-trial co-variability of a pair of spike trains and may be increased relative to across-trial correlations due to common synaptic input to the neuron pair. These common fluctuations are often deleterious to cortical population codes [60] , acting as a source of variability that cannot be removed through population averaging. The majority of our study presented simultaneously recorded spike train data which contains across-trial correlation as well as additional within-trial correlation. However, the shaping of correlation by the spatial profile of a stimulus was explained from knowledge of only of the across-trial correlation ( Figures 1 and 2 ), and thus our ELL-EGp network model ignored other sources of correlation entirely. Our analysis did study the effects of irrelevant distractor inputs which can act as a source of noise, though originating from external signals rather than internal circuit mechanisms. We found that low frequency distractors that drive a substantial portion of the network recruit a cancellation signal. We therefore predict that within-trial correlations may be cancelled by a similar mechanism if they drive a large number of neurons synchronously. This may be the case when within-trial correlations are driven by the local field potential, which is often low frequency and widespread to populations of neurons [70] , [71] .
In summary, we have identified the combination of feedforward and feedback architecture that allows the spatial extent of a stimulus to shape the temporal correlations between the spike trains of pairs of electrosensory principal cells. Furthermore, correlation shaping allows populations of neurons to respond to stimuli that match a specific spatiotemporal profile and ignore distractor inputs. The generic architectural features of our network and the fact that sensory systems must filter irrelevant signals suggest that our findings may generalize to other systems.
Supporting Information
Table S1
Model summary.
(PDF)
Text S1
Computation of cellular response function and power spectrum.
(PDF)
|
Vytvor v slovenčine zhrnutie totho článku. Laicky vysvetli všetky základné pojmy v článku. Ak používaš úvodzovky, nezabudni, že v slovenčine sa prvé úvodzovky píšu dole. Používaj správne slovenské slová, vyhni sa čechizmom . Využi 8000 až 8100 tokenov. Tu je text článku: |
Introduction
Foreseeably, past generations of patients have used their physicians as the key source of health related information, however, there is evidence that people are increasingly turning to the Internet to supplement their information needs [ 1 ]. For example, a Swedish study found that just over three-quarters (76.2%) of people diagnosed with cancer accessed the Internet for cancer-related information and more than one quarter used social media relating to their health [ 2 ]. Patients commonly report using webpages, blogs, interactive forums and social media to obtain information to help make informed decisions, find practical information or answers to health related questions, stay in touch with others, and share experiences [ 2 , 3 ].
Social media has become ubiquitous to our lives where we share, connect and communicate our experiences with friends, family, organisations and people otherwise unknown to us. Worldwide, approximately 2.5 billion people use social media and almost two-thirds of American adults use social networking sites: an almost ten-fold increase over the past decade [ 4 ]. The portability of these websites via mobile applications has no-doubt accelerated their uptake and allows for the capture of life’s most ephemeral events. The differences in user demographics that are seen between platforms (such as age, ethnicity, gender, education or income), lend themselves to being targeted for various health campaigns, health promotions or health research seeking to reach different audiences [ 5 , 6 ]. Social media data collection foreseeably provides large-scale and easily accessible data for patient reported information, particularly when compared with traditional patient-focused data collection methods [ 7 ].
One of the most popular social media platforms, Instagram, has almost one billion active monthly users [ 8 ]. Instagram is differentiated from other social media platforms by user-posts’ being dominated by a photo. Most users choose to add accompanying text to their photos as well as tags or labels, termed ‘hashtags’ denoted as # label , (e.g. #cancer). The accompanying hashtags provide a method of grouping photos to create virtual social communities of similarly themed content or purpose and allows users to easily connect and share content.
Acute myeloid leukaemia (AML) is a relatively rare and aggressive blood cancer that can occur at any age [ 9 , 10 ]. The standard treatment is immediate intensive chemotherapy, requiring lengthy hospital stays [ 11 ]. Additionally, research shows most patients have a reduced quality of life and persistent side effects or symptoms even after the completion of therapy or in remission [ 12 – 14 ].
AML makes up less than 1% of all cancer diagnoses per year, making research challenging to accrue participants, particularly in young adulthood where incidence is at its lowest [ 9 , 10 ]. However due to the popularity of Instagram, particularly in early adulthood [ 15 ] and the search functionality of hashtags, the Instagram platform presents an opportunity for proposing unique research questions, particularly those focused on rare-diseases (as with AML), or research with participants that are traditionally difficult to access. Despite the popularity of Instagram, and the unique participant group it can reach, little health research has been undertaking using this platform [ 1 ].
Given large numbers of people with cancer are accessing online health-related messages and the relative absence of Instagram research, this exploratory study will be the first to characterise AML-related content on Instagram; specifically who is posting AML-related content and what types of content are being posted. Characterising AML-related content on social media could be useful for targeting people most likely to benefit from health messages, interventions, or support. Using Instagram for this type of extant research has the potential to provide unique insights into the lived experience, as well as observing individuals providing or receiving support through virtual communities and the sharing of health-related information. Additionally we detail a method potentially of interest to other researchers.
Materials and methods
The methods outlined in this paper adhered to Instagram’s terms of service at the time of the research and all the content analysed in this study was publicly available on Instagram (available at www.instagram.com/instagram ). However the data is not owned by the authors and they do not have permission for reproduction of the data used in the analysis. Ethical approval for the study was granted from Monash University, Human and Research Ethics Committee (Project ID 18540) and included a waiver of consent (no individual consent was necessary from Instagram users). The ethics approval prohibits the publication of data that may inadvertently identify any individual.
Data collection
Instagram is primarily a mobile application but has a desktop website with limited functionality. Only the website accessible version of Instagram was used in this study, to ensure complete separation between the researchers and their private accounts. This also ensured that only publicly available posts were being accessed (no Instagram account or login required). One hundred posts were chosen as a convenient number for a time-consuming manual exploratory method. The posts were found by using Instagram’s search bar at the top of the webpage using hashtags that had been previously scoped as being used by people with AML: #acutemyeloidleukemia, #acutemyeloidleukaemia and #amlsurvivor. Each hashtag was searched for separately.
Posts were excluded from the study if they were videos (we were unable to extract these using our context extraction method), had non-English accompanying text or the subject matter was focused on children’s cancer. Children’s cancer was able to be determined by accompanying hashtag, such as #childhoodcancersucks or though examining the post photo and/or the accompanying text.
This study was undertaken after Instagram removed the automatic application program interface (API), which allowed for automation in downloads and much of the accompanying meta-data. Therefore, we detail a manual method of data extraction that may be of use to other researchers. This manual method allowed for retrospective capture for all eligible posts made over seven consecutive days in February 2019; eight consecutive days in April 2019 and twelve consecutive days in May 2019, to obtain a consecutive sample of 100 posts during the collection periods. Posts in chronological order (as opposed to most popular) can be found by scrolling past the initial “top posts” to the “most recent”. It is was these most recent posts that were accessed taking note of the date of the post to ensure it complied with our sampling time-frame. This sampling method was used to avoid awareness campaigns or trending content, (which may generate atypical Instagram posts and traffic), cultural and ethnic influences between users’ geographical location and for researcher convenience. One hundred posts was deemed to be sufficient given practicality of methods employed and the rarity of AML for an initial exploratory study.
As a user can modify or delete the content or their Instagram account, a screenshot was taken of the post and the user profile, thereby creating a ‘post-record’, which became the main unit of analysis. The post-record was made using Microsoft Word. We analysed the content of a post to include both the photo and the accompanying text and hashtags but excluded subsequent comments (and hashtags) that were made by either the ‘post-owner’ or other users.
For each post, basic data points were gathered about the user and the post: age and gender of the user (self-reported in the user profile), and country of origin data by using the location specified as part of the post or contained in the user profile, as well as post-specific information (description of the photo/s, accompanying text and hashtags and the number of likes and comments etc.). We also captured the username, but as duplicates became apparent, we adopted our identification system for each post to be able to distinguish between different posts from the same users.
Whilst the Instagram posts are publically available the data cannot be reproduced to comply with the Instagram terms of service, comply with the ethical approval of this study and to protect the privacy of the individuals posting on Instagram.
Data analysis
We used an adapted mixed-method social network model to frame our analysis [ 16 ]. The model describes sourcing data (Instagram), constructing the data (organising and preparing for analysis) and analysing the data (using network analysis or linear modelling). The framework was appropriate as the study was exploratory and observational and employed a content analysis, however it was modified as we did not employ the network analysis or linear modelling. Fig 1 demonstrates our approach.
10.1371/journal.pone.0250641.g001
Fig 1
Method process, adapted from a mixed methods social network analysis framework [ 16 ].
The content analysis is ideal for exploratory research, as this method seeks to unobtrusively explore the explicit description of the communication and the trends, patterns and frequency of this communication found within data [ 17 , 18 ]. No a priori coding was developed owing to an absence of literature relating to the content of AML on Instagram. An inductive approach was employed to identify frequently occurring content categories and themes [ 18 ].
In brief the process included open coding, creating higher headings and then categories. After reviewing the post records multiple times, open codes were developed in a consultative and iterative process of reviewing the first ten post-records, at which time an open coding scheme was generated, that we thought could be applied to the whole data set [ 19 ]. A further ten posts were classified according to our coding scheme and codes were refined as necessary. The first initial ten posts were re-coded as per this scheme. This process was repeated twice, until we had a open coding scheme (after coding 40 posts) that could be applied to the entire data set. Higher order headings were then able to be developed from the open codes using researcher interpretation as to which codes belong in each higher order heading and then into categories ( Fig 2 ) [ 20 ]. The process was undertaken by two reviewers and any discordance in coding between reviewers was discussed to reach consensus [ 18 ].
10.1371/journal.pone.0250641.g002
Fig 2
Process of generating categories.
After the process was finished and the researchers were reflecting on the findings, we went back and coded for a single theme: ‘hope and/or gratitude’, as the researchers felt that even though this was outside the content analysis it was an interesting finding relevant to the research. This theme was based on the researchers’ interpretation of the image and accompanying text.
Most data were expressed as both means with standard deviations, medians with interquartile ranges (IQRs), as well as frequency and range because the content and distribution varied considerably. Microsoft Excel was used for these descriptive statistics.
Results
During the search period, almost all posts were found using #acutemyeloidleukemia (94%). During the window of analysis, 51 unique users posted content and 16 of these posted more than once resulting in the analysis of 100 posts, consisting of 138 photos—one post can contain up to 10 photos.
Age and gender were mostly unavailable. Only two profiles stated age but we have chosen to conceal this for re-identification protection. Gender was rarely specified in the user profile, and we deemed it unreliable to discern gender either, from self-description (e.g. mom, wife etc.), appearance or socially gender-normative names, so this has not been reported. As shown in Table 1 , we were mostly unable to determine the country of the post origin for most users (34/51).
10.1371/journal.pone.0250641.t001
Table 1 Country of post origin of the post or user account (n = 51).
Country
Frequency n (%)
United States
11 (22)
Canada
1 (2)
United Kingdom
3 (6)
Hong Kong
1 (2)
Malaysia
1 (2)
Unknown
34 (66)
We identified three user categories from the data: patient personal stories, personal support networks and professional organisations ( Table 2 ). The most frequent users were patients themselves (66% of the posts), followed by personal support networks that we interpreted as family and friends (24% of the posts) and lastly professional organisations (10% of the posts).
10.1371/journal.pone.0250641.t002
Table 2 Frequency of posts and photos in each user category.
User categories
Number of posts (n = 100) n (%)
Number of photos (n = 138) n (%)
Patient personal stories
66 (66)
99 (72)
Personal support networks
24 (24)
26 (19)
Professional organisations
10 (10)
13 (9)
As shown in Table 3 , the most frequent content posted in the analysis was patients communicating their health update (31% of the whole sample). The majority of posts made by personal support networks was a health update on behalf of a patient (50% of the personal support networks user category). Professional organisations only accounted for 10% of the total sample and the majority of the content was either patient information provision (40% of the posts) or raising disease awareness (50% of the posts).
10.1371/journal.pone.0250641.t003
Table 3 The content classification frequency by user category and content classification.
User category
Content classification
Frequency of posts for each user category n (%)
Frequency of content classification for whole sample (n = 100) %
Patient personal stories (n = 66)
Personal health
31 (47)
31
Reflection
24 (36)
24
Self-care
11 (17)
11
Personal support networks (n = 24)
Someone else’s health
12 (50)
12
Remembrance
4 (17)
4
Raising awareness
8 (33)
8
Professional organisations (n = 10)
Information provision for patients
4 (40)
4
Raising awareness
5 (50)
5
Other
1 (10)
1
The 10 organisational posts comprised of seven users and thirteen photos. Five of the seven users had an unknown country of origin, while one was based in the United States and the other in the United Kingdom as discerned from their profile or dot-org websites.
One-quarter of all posts detailed symptoms that were being experienced by patients and 19/25 posts containing symptoms came from patients with the remaining posts being made by personal support networks. Please note to the protect privacy of individuals (for example via reverse identification), the quotes chosen below have been altered to encompass the overall sentiment of the quote [ 21 ].
“#selfie #nofilter long term chemotherapy effects have mostly subsided. Still can’t shake that #red eye…” (picture of a person smiling into the camera). Patient personal story .
“…Hubby had platelets to fix his bleeding gums…” (picture of a person sitting upright in bed, surrounded by medical equipment) Personal support networks .
Likes and comments were used as a proxy measure for engagement ( Table 4 ). Overall there was between three and 394 likes and between zero and 54 comments. There was little engagement with organisational posts as measured by ‘likes’ and comments. There were between eleven and 41 likes on the posts and five posts had no comments.
10.1371/journal.pone.0250641.t004
Table 4 Engagement with posts by user category as measured by likes and comments.
User category
Likes
Comments
Mean (SD)
Median (IQR)
Range
Mean (SD)
Median (IQR)
Range
Patient personal stories
67.41 (68.61)
36.5 (51.5)
251
7.47 (10.16)
4 (8.75)
54
Personal support networks
79.71 (41)
91.87 (52.75)
391
5.58 (7.9)
3 (4.75)
31
Professional organisations
28.9 (8.64)
31 (12.75)
30
1 (1)
1 (1)
11
All posts
66.51 (73.59)
35 (50)
391
6.43 (9.35)
3(6)
54
Additionally, throughout the analysis, we noticed a prominent theme of hope often accompanied by gratitude, in the posts, either implicitly but commonly through the use of the accompanying text or hashtags (e.g. #gratitude or #grateful or #thankyou or #hopeful). Almost half (49%) of all the posts demonstrated this theme hope and/or gratitude. Thirty-four of these were made by the user category of patient personal stories, eleven by personal support networks and four by professional organisations.
“…Each day has something good in it, even on the toughest of days…” (Image of a motivational meme) Patient personal story .
Discussion
While much of the social media cancer communication research has focused on Facebook and Twitter, very few studies have focused on Instagram, particularly with a focus on such an emotionally and physically burdensome cancer like AML. Instagram differs from Facebook and Twitter by incorporating visual cancer communication and to our knowledge this is the first study to describe the content of Instagram communication concerning AML, thereby addressing this research gap. The novel method we have outlined is most useful for other investigators looking to utilise social media in the their research and our findings should be considered in the context of the limitations of our methods.
Our exploratory descriptive research showed in our sample, that people with AML communicating personal health updates, was the most common content being posted about AML on Instagram. Personal story sharing related to AML was also prominent by the personal support networks user category of people with AML. This finding was congruent with other Instagram disease-related research [ 3 , 22 , 23 ].
Why people tell such personal stories through Instagram may be explained by social media use being linked with patient empowerment through improved self-management and enhanced psychological and subjective well-being [ 1 , 3 ]. These benefits may be obtained through real or perceived social connectedness of users of Instagram where they feel a sense of intimacy through sharing or social support, through community [ 24 – 26 ]. By posting intimate stories, users may also provide and receive social and emotional support through these virtual online communities [ 23 ]. This is further supported by the high prevalence of hope and/or gratitude in our data, where Steffen et al found in a study of advanced lung cancer patients, that hope may be important in providing support to social and role functioning, irrespective of physical symptom severity [ 27 ]. In sentiment analysis, Cho and colleagues also found hope was the most commonly expressed emotion in their melanoma study [ 23 ]. Whether hope is a common finding on social media contained to people with a malignant disease remains unknown.
In contrast to a Facebook content analysis including breast, prostate and other reproductive cancers, Instagram users concerned with AML do not appear to be information seeking, which may be due to the inherent functionalities of the platform [ 28 ]. This means that health professionals, researchers and professional organisations should endeavour to tailor their communication respectively to the most appropriate platform. However, if users are predominately seeking or providing support through personal storytelling, Instagram presents an opportunity for health providers and other organisations tasked with awareness-raising or support and wellbeing. Furthermore, it is likely patients and their friends and family are highly motivated to sustain the engagement with cancer communication initiated by reputable professional organisations [ 28 , 29 ]. It is worth noting, we were unable to identify any health providers (individually or part of a health facility) posting during our data extraction period. The content of what patients communicate via social media outside the immediate doctor-patient consult provides an unique viewpoint unhindered by bustling waiting rooms or the interpretation of clinicians, to contextualise patient experience and decision making [ 7 ]. In our study, only about 10% of posts were organisational suggesting that Instagram may represent an untapped resource for cancer support communities and awareness campaigns. This suggestion possibly holds relevance for all cancer types. Furthermore, public awareness is particularly relevant for malignant haematological diseases where up to 70% of patients need to seek bone marrow transplant donors outside of their family and only 7% of the American population are registered bone marrow donors [ 30 ]. Increasing public awareness through emotional appeal and capitalising on hope as a concept, may increase the number of registered donors to ensure sufficient diversity in the donor pool to meet the patient demand for bone marrow transplant [ 31 , 32 ].
Social media research can complement other research methods: Crawford et al used YouTube to complement a literature review about the patient experience of haematological malignancies and found that YouTube provided supplementary information that highlighted the multifactorial experiences of patients that may not have been otherwise apparent through traditional research methods [ 7 ].
Certainly some individual healthcare professionals can and are using social media. A recent Italian study of neurologists showed that 56% of the sample used social media to have direct contact with patients and most of these health professionals were in favour of this communication method [ 33 ]. Instagram may provide an opportunity for clinican-led content that is trustworthy and appeals to patients, yet clinician-led social media posts are lacking, yet [ 34 ]. Moorhead et al. [ 35 ] suggests that both health professionals and their patients may need training to maximise the use of social media in their healthcare interaction. However, as yet it remains unknown how effective social media can be in its’ perceived role in healthcare [ 35 ] and how this applies to inherently passive platforms such as Instagram where interactivity between user and viewer is limited.
Given the popularity of Instagram and the potential reach of posts, further research is warranted to understand the implications of online visual communication and how this information can be harnessed to improve health communication, patient experience and the experience of healthcare and balancing this with minimising the perpetuation of misinformation to vulnerable individuals.
Our study is not without limitations: critics rightfully observe that Instagram is often curated and may not reflect real life—experiences are complex and Instagram is a snapshot in time. Additionally, our sample may not reflect the breadth of posts due to our sample size, which was limited by the practicality of employing a manual method and resourcing. The manual method we employed and limitations in the search function also meant the study was unable to capture videos and Instagram stories (which are only available for 24 hours from posting). The sample used in this study had many users posting multiple times, potentially meaning our results may be less diverse and biased towards fewer individual experiences of AML. The retrospective nature of this study only allowed for the capture of data about age, gender and location that the user chose to share and it is therefore unknown whether there are dominating age groups, gender or country of origin in our analysis.
The strengths of this study are that we have demonstrated a unique and innovative way to potentially reach and/or observe hard to reach populations or people suffering rare conditions. Additionally, photos are a unique and expressive medium not conventionally used in cancer support services so other researchers with appropriate research question could also choose to employ an interactive image-based study design.
Conclusion
This exploratory study, presents a novel method whereby we have characterised AML-related Instagram content that contributes to the understanding of how social media fits into the lives of people affected by AML. Our results suggest that social media may have a role to play particularly for social connectedness and support and that there is a potential role to play for health professionals and health organisations.
Further research should focus on exploring the feasibility and effectiveness of targeted awareness campaigns, as well as deploying support networks or health interventions to aid people by providing or seeking support.
|
Vytvor v slovenčine zhrnutie totho článku. Laicky vysvetli všetky základné pojmy v článku. Ak používaš úvodzovky, nezabudni, že v slovenčine sa prvé úvodzovky píšu dole. Používaj správne slovenské slová, vyhni sa čechizmom . Využi 8000 až 8100 tokenov. Tu je text článku: |
Introduction
The c-MYC proto-oncogene encodes a transcription factor that plays a central role in cell proliferation, differentiation, apoptosis, metabolism, and survival [ 1 , 2 ]. It can promote tumorigenesis in a variety of human malignancies [ 3 , 4 ]. c-MYC alteration occurs through various mechanisms, including chromosomal translocation, gene amplification, and perturbation of upstream signaling pathways [ 5 , 6 ]. Gene copy-number (GCN) gain or amplification is the most common c-MYC alteration in solid tumors [ 7 ].
Nevertheless, few studies have examined the clinicopathological implications of c-MYC status in colorectal cancer (CRC). Previous reports have shown that c-MYC GCN gain in CRC is found in approximately 10% of patients [ 8 ]. A recent study reported that several significant amplifications were focused on chromosome 8, including the 8q24 region which contains c-MYC , and suggested that c-MYC was a new marker for aggressive disease in CRC [ 9 ]. However, more recently, Christopher et al . reported data obtained by immunohistochemistry (IHC), indicating that c-MYC protein overexpression was significantly associated with improved prognosis in CRC patients [ 10 ]. Consequently, the prognostic value of c-MYC alterations in CRC is controversial.
Recently, the range of options for systemic chemotherapy has expanded and targeted therapy has been used in advanced CRC patients, increasing patient survival [ 11 ]. However, some CRC patients respond poorly to targeted therapy despite showing positive results in targeted therapy-specific mutation studies [ 12 ]. Tumor heterogeneity is a potential cause for failure of targeted therapy and several studies have reported that CRC possess a heterogenic genotype including KRAS , p53 , and BRAF [ 13 – 15 ]. Therefore, genetic variation between the primary tumor and corresponding metastatic sites needs to be clarified to improve the management of CRC patients with metastatic disease.
The heterogeneity of c-MYC and its prognostic implications have not been systematically studied in primary CRC patients. The aim of this study was to evaluate c-MYC gene status and its clinical significance in CRC. We also analyzed the heterogeneity of c-MYC in the primary tumor and distant metastasis.
Materials and Methods
Patients and samples
A total of 519 CRC patients treated with radical surgery at Seoul National University Bundang Hospital were enrolled in this retrospective study. First, to evaluate the clinicopathologic significance of c-MYC gene status, 367 consecutive CRC patients treated between January 2005 and December 2006 were enrolled (cohort 1). Second, to investigate the discordance between the primary and metastatic tumors, 152 advanced CRC patients with synchronous or metachronous metastasis who had undergone surgical resection for primary CRC between May 2003 and December 2009, were enrolled (cohort 2). All the cases were reviewed by two pathologists (K. S. L. and H. S. L.). The clinicopathological characteristics were obtained from the patients’ medical records and pathology reports. Follow-up information including patient outcome and the interval between the date of surgical resection and death was collected. Data from patients lost to follow-up or those who had died from causes other than CRC were censored.
Ethical statement
All samples were obtained from surgically resected tumors examined pathologically at the Department of Pathology, Seoul National University Bundang Hospital. All samples and medical record data were anonymized before use in this study and the participants did not provide written informed consent. The use of medical record data and tissue samples for this study was approved by the Institutional Review Board of Seoul National University Bundang Hospital (reference: B-1210/174-301).
Tissue array method
Surgically resected primary CRC specimens were formalin-fixed and paraffin-embedded (FFPE). For each case, representative areas of the donor blocks were obtained and rearranged into new recipient blocks (Superbiochips Laboratories, Seoul, South Korea) [ 16 ]. A single core was 2 mm in diameter and those containing > 20% tumor cells were considered valid cores.
Dual-color silver in situ hybridization
The c-MYC gene was visualized by using a blue-staining system (ultraView silver in situ hybridization [SISH] dinitrophenol [DNP] detection kit and c-MYC DNP probe, Ventana Medical Systems, Tucson, AZ, USA). The centromere of chromosome 8 (CEP8) was visualized by using a red-staining system (ultraView red ISH digoxigenin [DIG] detection kit and chromosome 8 DIG probe, Ventana Medical Systems). Positive signals were visualized at 60 × magnification and counted in 50 non-overlapping tumor cell nuclei for each case ( Fig 1 ) [ 17 ]. Small and large clusters were scored as 6 and 12 signals, respectively.
10.1371/journal.pone.0139727.g001
Fig 1
Representative figures of c-MYC status detected by dual-color silver in situ hybridization (A and B) in colorectal cancer patients.
(A) c-MYC gene copy number gain (60 × magnification); (B) c-MYC gene disomy (60 × magnification).
Immunohistochemistry
IHC analysis of c-MYC was carried out using a commercially available rabbit anti-c-MYC antibody (clone Y69, catalog ab32072, Abcam, Burlingame, CA, USA). The staining procedures were carried out using the ultraView Universal DAB kit (Ventana Medical Systems) and an automated stainer (BenchMark®XT, Ventana Medical Systems), according to the manufacturer’s instructions. Nuclear immunostaining of c-MYC was negative in normal mucosa. For statistical analysis, c-MYC nuclear staining of any intensity in greater than 10% of neoplastic cells was scored as positive ( S2 Fig ) [ 10 ].
Microsatellite instability
Microsatellite instability (MSI) was assessed in CRC cases with available tissue. MSI results were generated by comparing the allelic profiles of 5 microsatellite markers (BAT-26, BAT-25, D5S346, D17S250, and S2S123) in the tumor and corresponding normal samples. Polymerase chain reaction (PCR) products from the FFPE tissues were analyzed using an automated DNA sequencer (ABI 3731 Genetic Analyser, Applied Bio systems, Foster City, CA, USA) according to the protocol described previously [ 18 ].
KRAS mutation analysis
KRAS mutation detection was achieved by melting curve analysis using the cobas 4800 System (Roche, Branchburg, NJ, USA) with automated result interpretation software. This is a TaqMelt-based real-time PCR assay designed to detect the presence of 21 KRAS mutations in codons 12, 13, and 61. The workflow and testing process have been described previously [ 19 ].
Statistical analyses
The association between the clinicopathological features and c-MYC status was analyzed using the chi-square or Fisher’s exact test, as appropriate. The correlation between the detection methods was examined using the Pearson correlation coefficient. The patients’ survival was analyzed by using the Kaplan-Meier method and the log-rank test was used to determine if there were any significant differences between the survival curves. Univariate and multivariate regression analysis were performed by using Cox’s proportional hazards model to determine the hazard ratio and 95% confidence intervals for each factor. A P- value < 0.05 was accepted as statistically significant. All statistical analyses were performed using the SPSS statistics 21 software (IBM, Armonk, NY, USA).
Results
c-MYC gene status and clinical implications for consecutive primary CRC patients
In consecutive primary CRC cases (cohort 1), the median c-MYC :CEP8 ratio was 1.29 (range, 0.58–5.17). c-MYC gene amplification, defined by a c-MYC :CEP8 ratio ≥ 2.0 and similar to that established for HER2 [ 20 ], was detected in 31 (8.4%) of 367 patients. The mean c-MYC GCN was 2.88 (range, 1.22–13.12). In the present study, we defined the GCN gain as ≥ 4.0 c-MYC copies/nucleus [ 21 ], and this was detected in 63 (17.2%) of 367 CRC patients. All c-MYC amplification was included in c-MYC GCN gain. A c-MYC GCN gain ≥ 4 had the lowest P -value ( P = 0.015) and thus, was observed to be the most predictive cut-off point for patient prognosis ( Fig 2 ); ≥ 5.0 c-MYC copies/nucleus also influenced patient prognosis ( P = 0.026). There was no significant association between patient prognosis and either c-MYC amplification ( P = 0.149) or > 2, ≥ 3, and ≥ 6 c-MYC copies/nucleus ( P = 0.752, P = 0.175, and P = 0.122, respectively).
10.1371/journal.pone.0139727.g002
Fig 2
Kaplan-Meier survival curves illustrating the prognostic effect of c-MYC status in colorectal cancer (cohort 1).
(A) c-MYC gene copy number (GCN) gain; (B) c-MYC GCN gain in the stage II-III subgroup; (C) c-MYC amplification.
Table 1 shows the relationships between c-MYC status and the clinicopathological parameters in consecutive primary CRCs (cohort 1). Amplification of c-MYC correlated with early-stage disease ( P = 0.039). c-MYC GCN gain was frequently observed in sigmoid colon and rectum tumors ( P = 0.034), small tumors ( P = 0.041), and those classified as microsatellite stable or MSI-low ( P = 0.029).
10.1371/journal.pone.0139727.t001
Table 1 The association between clinicopathological parameters and c-MYC status in 367 CRC patients (cohort1).
Total
c-Myc
P -Value
c-Myc
P -Value
c-Myc IHC
P -Value
4 > GCN
4 ≦ GCN
Non-amplification
Amplification
Negative
Positive
Age
0.983
0.383
0.537
mean
64.2
64.2
64.2
64.1
66.0
64.6
63.9
Sex
0.740
0.619
0.431
male
205
171 (83.4%)
34 (16.6%)
189 (92.2%)
16 (7.8%)
89 (43.4%)
116 (56.6%)
female
162
133 (82.1%)
29 (17.9%)
147 (90.7%)
15 (9.3%)
77 (47.5%)
85 (52.5%)
Location
0.034
0.437
< 0.001
Rectum/sigmoid
237
189 (79.7%)
48 (20.3%)
121 (93.1%)
9 (6.9%)
90 (38.0%)
147 (62.0%)
others
130
115 (88.5%)
15 (11.5%)
215 (90.7%)
22 (9.3%)
76 (58.5%)
54 (41.5%)
pT stage
0.692
0.571
< 0.001
0–2
58
47 (81.0%)
11 (19.0%)
52 (89.7%)
6 (10.3%)
14 (24.1%)
44 (75.9%)
3–4
309
257 (83.2%)
52 (16.8%)
284 (91.9%)
25 (8.1%)
152 (49.2%)
157 (50.8%)
Differentiation
0.139
0.055
0.007
LG
331
271 (81.9%)
60 (18.1%)
300 (90.6%)
31 (9.4%)
142 (42.9%)
189 (57.1%)
HG
36
33 (91.7%)
3 (8.3%)
36 (100.0%)
0 (0.0%)
24 (66.7%)
12 (33.3%)
LN metastasis
0.609
0.070
0.058
absent
168
141 (83.9%)
27 (16.1%)
149 (88.7%)
19 (11.3%)
67 (39.9%)
101 (60.1%)
present
199
163 (81.9%)
36 (18.1%)
187 (94.0%)
12 (6.0%)
99 (49.7%)
100 (50.3%)
Lymphatic invasion
0.152
0.896
0.073
absent
158
136 (86.1%)
22 (13.9%)
145 (91.8%)
13 (8.2%)
63 (39.9%)
95 (60.1%)
present
209
168 (80.4%)
41 (19.6%)
191 (91.4%)
18 (8.6%)
103 (49.3%)
106 (50.7%)
Perineural invasion
0.631
0.530
0.025
absent
154
212 (83.5%)
42 (16.5%)
231 (90.9%)
23 (9.1%)
49 (58.7%)
105 (41.3%)
present
113
92 (81.4%)
21 (18.6%)
105 (92.9%)
8 (7.1%)
61 (54.0%)
52 (46.0%)
Venous invasion
0.776
0.999
0.814
absent
296
246 (83.1%)
50 (16.9%)
271 (91.6%)
25 (8.4%)
133 (44.9%)
163 (55.1%)
present
71
58 (81.7%)
13 (18.3%)
65 (91.5%)
6 (8.5%)
33 (46.5%)
38 (53.5%)
Tumor border
0.524
0.327
0.544
expanding
60
48 (80.0%)
12 (20.0%)
53 (88.3%)
7 (11.7%)
25 (41.7%)
35 (58.3%)
infiltrative
307
256 (83.4%)
51 (16.6%)
283 (92.2%)
24 (7.8%)
141 (45.9)
166 (54.1%)
Size (cm)
0.041
0.061
< 0.001
mean
5.3
5.4
4.7
5.3
4.5
5.8
4.8
Distant metastasis
0.123
0.544
0.252
absent
299
252 (84.3%)
47 (15.7%)
275 (92.0%)
24 (8.0%)
131 (43.8%)
168 (56.2%)
present
68
52 (76.5%)
16 (23.5%)
61 (89.7%)
7 (10.3%)
35 (51.5%)
33 (48.5%)
pTNM stage
0.822
0.039
0.050
I, II
162
135 (83.0%)
27 (17.0%)
140 (88.1%)
19 (11.9%)
64 (39.5%)
98 (60.5%)
III, IV
205
169 (82.4%)
36 (17.6%)
193 (94.1%)
12 (5.9%)
102 (49.8%)
103 (50.2%)
MSI status
0.029
0.256
0.490
MSS/MSI-L
323
264 (81.7%)
59 (18.3%)
294 (91.0%)
29 (9.0%)
141 (38.4%)
182 (49.6%)
MSI-H
32
31 (96.9%)
1 (3.1%)
31 (96.9%)
1 (3.1%)
16 (1.6%)
16 (4.4%)
Abbreviations: CRC, colorectal cancer; T, tumor; LG, low grade; HG, high grade; LN, lymph node; MSS, microsatellite stable; MSI-L, microsatellite instability-low; MSI-H, microsatellite instability-high; GCN, gene copy number; IHC, immunohistochemistry
P -values are calculated by using χ 2 -test or Fisher’s exact test
Prognostic significance of c-MYC gene status in CRC patients
All CRC patients were successfully followed up for inclusion in the survival analysis ( Fig 2 ). In cohort 1, the mean follow-up period was 55 months (range, 1–85 months) and 101 (27.5%) patients died during the follow-up period. Kaplan-Meier analysis showed that c-MYC GCN gain was significantly associated with poor survival in CRC patients ( P = 0.015), but c-MYC amplification was not ( P = 0.149). In the stage II-III subgroup, c-MYC -GCN gain also predicted poor prognosis ( P = 0.034). Multivariate analysis of c-MYC status is summarized in Table 2 , and showed that c-MYC -GCN gain independently predicted poor prognosis in the consecutive cohort ( P < 0.001) and in the subgroup of patients with stage II-III CRC ( P = 0.040).
10.1371/journal.pone.0139727.t002
Table 2 Multivariate Cox proportional hazard models for the predictors of overall survival (cohort 1).
Univariate survival analysis
Multivariate survival analysis
Factors
HR (95% CI)
P value
HR (95% CI)
P value
c-MYC GCN SISH (4≦ vs. 4>)
1.756 (1.117–2.763)
0.015
2.350 (1.453–3.801)
<0.001
Age
1.026 (1.008–1.045)
0.005
1.025 (1.007–1.043)
0.006
Size
1.244 (1.059–1.244)
0.001
1.099 (0.995–1.214)
NS (0.062)
Histologic grade (high vs. low)
3.143 (1.904–5.188)
<0.001
2.844 (1.625–4.977)
<0.001
Stage (3/4 vs. 1/2)
6.151 (3.494–10.829)
<0.001
3.069 (1.603–5.878)
0.001
Lymphatic invasion
3.661 (2.242–5.980)
<0.001
1.251 (0.709–2.205)
NS (0.439)
Perineural invasion
3.942 (2.648–5.870)
<0.001
2.325 (1.487–3.636)
<0.001
Venous invasion
3.985 (2.671–5.946)
<0.001
2.304 (1.490–3.676)
<0.001
c-MYC GCN SISH (4≦ vs. 4>) in a subgroup of stage II/III
2.057 (1.039–4.073)
0.038
2.058 (1.032–4.105)
0.040
Age
1.037 (1.009–1.067)
0.010
1.036 (1.007–1.066)
0.014
Stage (3 vs. 2)
2.955 (1.493–5.850)
0.002
1.705 (0.802–3.623)
NS (0.165)
Lymphatic invasion
2.882 (1.456–5.703)
0.002
1.846 (0.887–3.845)
NS (0.101)
Perineural invasion
3.536 (1.952–6.405)
0.001
2.921 (1.558–5.476)
<0.001
Abbreviations: SISH, silver in-situ hybridization; GCN, gene copy number; HR, hazard ratio
P -values are calculated by using χ 2 -test or Fisher’s exact test
Correlation between the c-MYC GCN gain and protein overexpression
Overexpression of c-MYC protein was detected in 201 (54.8%) of 367 CRC patients (cohort 1) and was associated with early pT stage ( P < 0.001), low grade of histologic differentiation ( P = 0.007), absence of perineural invasion ( P = 0.025) and smaller size ( P < 0.001) ( Table 1 ). Overexpression of c-MYC protein was associated with GCN gain (ρ, 0.211; P < 0.001), which was categorized as weakly correlation according to Dancey and Reidy’s categorization (2004) [ 22 ]. Furthermore, only 46 (22.9%) of 201 patients with c-MYC overexpression showed a GCN gain.
c-MYC status and heterogeneity according to tumor location in advanced CRC patients
To evaluate the regional heterogeneity of c-MYC status, we examined tissue from 3 sites including the primary cancer, distant metastasis, and lymph-node metastasis for each patient with advanced CRC (cohort 2). In the primary tumors of cohort 2, the median c-MYC :CEP8 ratio was 1.14 (range, 0.57–2.97). c-MYC gene amplification was detected in 8 (5.3%) of 152 patients. The mean c-MYC GCN was 2.97 (range, 1.40–9.94). c-MYC GCN gain was detected in 48 (31.6%) of 152 CRC patients. In addition, c-MYC GCN gain was found in 33 (21.7%) patients with distant metastatic tumors. Lymph-node metastasis was observed in 79 of 152 advanced CRC patients and c-MYC GCN gain was observed in 18 (22.8%) of these cases. The heterogeneity of c-MYC GCN gain according to tumor location is shown in Table 3 . Of 152 cases, discordance between c-MYC GCN gain in the primary tumor and distant metastasis was noted in 39 (25.7%) cases. Discordance between c-MYC GCN gain in the primary tumor and lymph-node metastasis was detected in 24/79 (30.4%) cases. Thus, regional heterogeneity of c-MYC GCN gain was quite common in advanced CRC. c-MYC GCN heterogeneity was not correlated with clinicopathological factors and prognosis ( P > 0.05; data not shown).
10.1371/journal.pone.0139727.t003
Table 3 Heterogeneity of c-MYC GCN gain with respect to tumor location in advanced CRC (cohort 2).
c-MYC GCN gain (%)
Primary
negative
positive
total
Distant metastasis
negative
92 (60.5)
27 (17.8)
152 (100)
positive
12 (7.9)
21 (13.8)
LN metastasis
negative
44 (55.7)
17 (21.5)
79 (100)
positive
7 (8.9)
11 (13.9)
Abbreviations: GCN, gene copy number; LN: lymph node
P -values are calculated by using χ 2 -test or Fisher’s exact test
There was no statistically significant correlation between the clinicopathological factors and c-MYC GCN gain in primary, distant metastatic, and lymph-node metastatic tumors from cohort 2 CRC patients ( P > 0.05; data not shown). The mean follow-up time was 42 months (range, 1–105 months) and 67 patients (44.1%) died from cancer during the follow-up period. Kaplan-Meier analysis showed that patients with c-MYC GCN gain in the primary tumor had a poor outcome than those without, but this result was not statistically significant ( P = 0.499). However, ≥ 3.0 c-MYC copies/nucleus in the primary tumor was significantly associated with a poor prognosis ( P = 0.044; S1 Fig ). There was no significant correlation between the patients’ prognosis and c-MYC GCN gain in distant or lymph-node metastases ( P = 0.981 and P = 0.417, respectively; data not shown).
KRAS mutations in advanced CRC
The cobas KRAS test was performed on 152 primary tumors from advanced CRC cases (cohort 2). KRAS gene mutations were observed in 84 (55.3%) cases and were associated with tumors located in the right colon ( P = 0.019), but were not correlated with other clinicopathological factors ( P > 0.05; data not shown). Additionally, there was no statistical difference between the survival of CRC patients with mutated or wild-type KRAS ( P = 0.688; data not shown). Of 68 cases with wild-type KRAS , c-MYC amplification was noted in 4 (5.9%) and a c-MYC GCN gain in 28 (41.2%). Of 84 cases with mutated KRAS , 4 showed c-MYC amplification (4.8%) and 20 (23.8%) revealed a c-MYC GCN gain. c-MYC GCN alterations occurred in patients with both wild-type and mutated KRAS. Therefore, c-MYC GCN alterations and KRAS mutations were not mutually exclusive.
Discussion
Although there have been several reports on c-MYC status in human cancers, there are no established criteria for GCN gain. Cancers with a c-MYC GCN gain are often associated with a poor prognosis. A previous study of mucinous gastric carcinoma showed that c-MYC amplification, defined as a c-MYC :CEP8 ratio > 2.0, was strongly correlated with the advanced stages of cancer [ 23 ]. Another report found an association between c-MYC amplification (> 4 copies/cell in a minimum of 10% of tumor cells) and the advanced stages of ovarian cancer [ 21 ]. In a study of prostate cancer, the c-MYC GCN gain included the criterion of a c-MYC/ CEP8 ratio > 1.5, and a poor prognosis was observed for patients in this category [ 24 ]. In recent research on adenocarcinoma of the lung, patients with > 2 c-MYC copies/nucleus were classified as having an increased c-MYC GCN, which was found to be an independent poor prognostic factor [ 25 ]. In CRC patients, it was reported that c-MYC amplification, defined as a c-MYC/ CEP8 ratio > 2, was frequently detected by using fluorescent in situ hybridization (9.0–14.2%), but was unrelated to clinical outcome and pathological data [ 26 ]. Therefore, we have applied diverse criteria to determine c-MYC amplification or GCN gain in this study, and have defined the c-MYC GCN gain as ≥ 4 copies/nucleus, because it had the lowest P -value for disease prognosis ( Fig 2 ). In cohort 1, the large consecutive cohort, CRC patients with a c-MYC GCN gain had a poor survival than those without ( P = 0.015). Furthermore, in multivariate analysis, c-MYC GCN gain was a significant CRC prognostic factor, both in the consecutive cohort and for those with stage II-III disease. The predictive value of the c-MYC GCN was found to be independent of known prognostic factors. The c-MYC GCN gain criteria used in the present study, together with the SISH method, may be useful in assessing CRC patients because it clearly identified patients expected to have poor survival, regardless of the c-MYC :CEP8 ratio.
In cohort 2, we showed that there was c-MYC GCN regional heterogeneity between the primary site and its related metastases. A c-MYC GCN gain (c-MYC GCN ≥ 4.0) in the primary cancer was not significantly associated with poor survival ( P = 0.499; S1 Fig ), which might be because all of cohort 2 consisted of advanced CRC patients with synchronous and metachronous metastasis and cohort 2 was largely comprised of stage IV CRC (98 cases; 64.5%). They received a variety of personalized treatment respectively and these might reflect the statistical insignificance. Interestingly, we applied slightly non-restrictive criteria of GCN gain ( c-MYC GCN ≥ 3.0) and its prognosis was changed to statistically significant ( P = 0.044; S1 Fig ). In a broad sense, c- MYC GCN gain of primary cancer tends to correlated with poor survival in advanced CRC. On the other hand, c-MYC status in distant and lymph-node metastatic lesion was not related to patient prognosis although we tried every possible GCN criteria. Even though, c-MYC heterogeneity was observed frequently in advanced CRC, a c-MYC GCN gain in the primary cancer was often associated with poor survival. Consequently, the c-MYC GCN in the non-metastatic lesion should be used when evaluating prognosis.
In a previous study, overexpression of c-MYC mRNA in CRC was found to be associated with a better prognosis [ 27 ], but this result was contradicted by another study [ 28 ]. Christopher et al . recently demonstrated that c-MYC overexpression determined by IHC alone, was significantly associated with a better survival in CRC patients when assessed by univariate analysis, but not by multivariate analysis [ 10 ]. Interestingly, we found conflicting results in a previous c-MYC overexpression study; presumably, because c-MYC expression is controlled by a complex regulatory pathway involving multiple interactions with other molecules, rather than just simple GCN gain [ 29 ]. Furthermore, we found a weak correlation between c-MYC protein overexpression and GCN gain in CRC patients. c-MYC GCN gain was not observed in most c-MYC protein overexpression cases. Unlike the c-MYC GCN gain, overexpression of c-MYC protein was correlated with less aggressive features ( Table 1 ). These results suggest that c-MYC GCN gain is probably only partly responsible for protein overexpression. As overexpression of c-MYC is not the same as a c-MYC GCN gain, further research is needed to explain the difference of c-MYC overexpression and GCN gain in CRC tumorigenesis.
Mutations in KRAS are evident in 30–40% of colorectal tumors [ 30 – 32 ]. Indeed, previous studies reported that a KRAS mutation was associated with resistance to anti-epidermal growth factor receptor (EGFR) monoclonal therapy and a poor survival [ 33 – 35 ]. In our study, KRAS mutations were present in 55.3% of advanced CRC patients (cohort 2) and were not associated with prognosis. It may be because we investigated KRAS mutation status in advanced CRC patients. Phipps et al. also reported that KRAS -mutation status was not associated with poor disease specific survival in cases who presented with distant-stage CRC [ 33 ]. c-MYC amplification was observed in 5.9% of wild-type KRAS and 4.8% of mutated KRAS CRCs. A c-MYC GCN gain was observed in 41.2% of wild-type KRAS and 23.8% of mutated KRAS CRCs. It is noteworthy that these 2 genetic events were not mutually exclusive. Further studies are required to investigate the possibility of using c-MYC genetic alterations as therapeutic targets in advanced CRC patients with primary and secondary resistance to anti-EGFR therapies.
In summary, we comprehensively analyzed the c-MYC gene status of CRC patients by using SISH. c-MYC GCN gain and amplification were observed in 17.2% and 8.4% of consecutive CRC patients, respectively. The c-MYC GCN gain was an independent poor prognostic factor, both in the consecutive cohort and in the subgroup of patients with stage II-III disease. These findings show that c-MYC status can be used to predict the prognosis of CRC patients, and may inform future studies on the pathogenesis and mechanisms involved in the progression of CRC.
Supporting Information
S1 Fig
Kaplan-Meier survival curves illustrating the prognostic effect of c-MYC status in primary lesions of colorectal cancer (cohort 2).
(A) c-MYC gene copy number (GCN) ≥ 3.0; (B) c-MYC GCN ≥ 4.0.
(TIF)
S2 Fig
Representative figures of c-MYC overexpression by immunohistochemistry (A and B) in colorectal cancer patients.
(A) c-MYC overexpression (40 × magnification); (B) No c-MYC expression (40 × magnification);
(TIF)
|
"Vytvor v slovenčine zhrnutie totho článku. Laicky vysvetli všetky základné pojmy v článku. (...TRUNCATED) | "\n \n Introduction \n Hantaan (HTNV) and Seoul viruses (SEOV), etiologic agents of hemorrhagic feve(...TRUNCATED) |
"Vytvor v slovenčine zhrnutie totho článku. Laicky vysvetli všetky základné pojmy v článku. (...TRUNCATED) | "\n \n Introduction \n Cutaneous leishmaniasis (CL) is an important neglected tropical skin disease (...TRUNCATED) |
"Vytvor v slovenčine zhrnutie totho článku. Laicky vysvetli všetky základné pojmy v článku. (...TRUNCATED) | "\n Introduction \n Cyclodextrins are water-soluble cyclic oligosaccharides with hydrophilic outer s(...TRUNCATED) |
"Vytvor v slovenčine zhrnutie totho článku. Laicky vysvetli všetky základné pojmy v článku. (...TRUNCATED) | "\n \n Introduction \n To colonize the host, bacterial pathogens and commensals are limited to metab(...TRUNCATED) |
"Vytvor v slovenčine zhrnutie totho článku. Laicky vysvetli všetky základné pojmy v článku. (...TRUNCATED) | "\n \n Introduction \n Bacterial pathogens that are resistant to conventional antibi(...TRUNCATED) |
"Vytvor v slovenčine zhrnutie totho článku. Laicky vysvetli všetky základné pojmy v článku. (...TRUNCATED) | "\n \n Introduction \n Colorectal cancer (CRC) is one of the most common malignancie(...TRUNCATED) |
No dataset card yet
- Downloads last month
- 18