text
stringlengths 65
113k
|
|---|
In Parkinson's disease, there is a progressive reduction in striatal dopaminergic function, and loss of neuromelanin-containing dopaminergic neurons and increased iron deposition in the substantia nigra. We tested the hypothesis of a relationship between impairment of the dopaminergic system and changes in the iron metabolism. Based on imaging data of patients with prodromal and early clinical Parkinson's disease, we assessed the spatiotemporal ordering of such changes and relationships in the sensorimotor, associative and limbic territories of the nigrostriatal system. Patients with Parkinson's disease (disease duration < 4 years) or idiopathic REM sleep behaviour disorder (a prodromal form of Parkinson's disease) and healthy controls underwent longitudinal examination (baseline and 2-year follow-up). Neuromelanin and iron sensitive MRI and dopamine transporter single-photon emission tomography were performed to assess nigrostriatal levels of neuromelanin, iron, and dopamine. For all three functional territories of the nigrostriatal system, in the clinically most and least affected hemispheres separately, the following was performed: cross-sectional and longitudinal intergroup difference analysis of striatal dopamine and iron, and nigral neuromelanin and iron; in Parkinson's disease patients, exponential fitting analysis to assess the duration of the prodromal phase and the temporal ordering of changes in dopamine, neuromelanin or iron relative to controls; and voxel-wise correlation analysis to investigate concomitant spatial changes in dopamine-iron, dopamine-neuromelanin and neuromelanin-iron in the substantia nigra pars compacta. The temporal ordering of dopaminergic changes followed the known spatial pattern of progression involving first the sensorimotor, then the associative and limbic striatal and nigral regions. Striatal dopaminergic denervation occurred first followed by abnormal iron metabolism and finally neuromelanin changes in the substantia nigra pars compacta, which followed the same spatial and temporal gradient observed in the striatum but shifted in time. In conclusion, dopaminergic striatal dysfunction and cell loss in the substantia nigra pars compacta are interrelated with increased nigral iron content. |
Aggregation of α-synuclein plays a key role in the development of Parkinson's disease. Soluble aggregates are present not only within human brain but also the CSF and blood. Characterizing the aggregates present in these biofluids may provide insights into disease mechanisms and also have potential for aiding diagnosis. We used two optical single-molecule imaging methods called aptamer DNA-PAINT and single-aggregate confocal fluorescence, together with high-resolution atomic force microscopy for specific detection and characterization of individual aggregates with intermolecular β-sheet structure, present in the CSF and serum of 15 early stage Parkinson's disease patients compared to 10 healthy age-matched controls. We found aggregates ranging in size from 20 nm to 200 nm, in both CSF and serum. There was a difference in aggregate size distribution between Parkinson's disease and control groups with a significantly increased number of larger aggregates (longer than 150 nm) in the serum of patients with Parkinson's disease. To determine the chemical composition of the aggregates, we performed aptamer DNA-PAINT on serum following α-synuclein and amyloid-β immunodepletion in an independent cohort of 11 patients with early stage Parkinson's disease and 10 control subjects. β-Sheet aggregates in the serum of Parkinson's disease patients were found to consist of, on average, 50% α-synuclein and 50% amyloid-β in contrast to 30% α-synuclein and 70% amyloid-β in control serum [the differences in the proportion of these aggregates were statistically significant between diseased and control groups (P = 1.7 × 10-5 for each species)]. The ratio of the number of β-sheet α-synuclein aggregates to β-sheet amyloid-β aggregates in serum extracted using our super-resolution method discriminated Parkinson's disease cases from controls with an accuracy of 98.2% (AUC = 98.2%, P = 4.3 × 10-5). Our data suggest that studying the protein aggregates present in serum can provide information about the disruption of protein homeostasis occurring in Parkinson's disease and warrants further investigation as a potential biomarker of disease. |
Since the first demonstrations of network hyperexcitability in scientific models of Alzheimer's disease, a growing body of clinical studies have identified subclinical epileptiform activity and associated cognitive decline in patients with Alzheimer's disease. An obvious problem presented in these studies is lack of sensitive measures to detect and quantify network hyperexcitability in human subjects. In this study we examined whether altered neuronal synchrony can be a surrogate marker to quantify network hyperexcitability in patients with Alzheimer's disease. Using magnetoencephalography (MEG) at rest, we studied 30 Alzheimer's disease patients without subclinical epileptiform activity, 20 Alzheimer's disease patients with subclinical epileptiform activity and 35 age-matched controls. Presence of subclinical epileptiform activity was assessed in patients with Alzheimer's disease by long-term video-EEG and a 1-h resting MEG with simultaneous EEG. Using the resting-state source-space reconstructed MEG signal, in patients and controls we computed the global imaginary coherence in alpha (8-12 Hz) and delta-theta (2-8 Hz) oscillatory frequencies. We found that Alzheimer's disease patients with subclinical epileptiform activity have greater reductions in alpha imaginary coherence and greater enhancements in delta-theta imaginary coherence than Alzheimer's disease patients without subclinical epileptiform activity, and that these changes can distinguish between Alzheimer's disease patients with subclinical epileptiform activity and Alzheimer's disease patients without subclinical epileptiform activity with high accuracy. Finally, a principal component regression analysis showed that the variance of frequency-specific neuronal synchrony predicts longitudinal changes in Mini-Mental State Examination in patients and controls. Our results demonstrate that quantitative neurophysiological measures are sensitive biomarkers of network hyperexcitability and can be used to improve diagnosis and to select appropriate patients for the right therapy in the next-generation clinical trials. The current results provide an integrative framework for investigating network hyperexcitability and network dysfunction together with cognitive and clinical correlates in patients with Alzheimer's disease. |
Functional neuroimaging has improved pre-planning of surgery in eloquent cortical areas, but remains unable to map white matter. Thus, tumour resection in functional subcortical regions still presents a high risk of sequelae. The authors successfully used intraoperative electrical stimulations to perform subcortical language pathway mapping in order to avoid postoperative definitive deficit, and correlated these functional findings with the anatomical location of the eloquent bundles detected using postoperative MRI. At the same time, this also improved knowledge of fibre connectivity. Thirty patients harbouring a cortico-subcortical low-grade glioma in the left dominant hemisphere were operated on whilst awake using intraoperative electrical functional mapping during surgical resection. Language cortical sites and subcortical pathways were clearly identified and preserved in the 30 cases. The anatomo-functional correlations between data obtained using intraoperative subcortical mapping and postoperative MRI revealed the existence in all patients of common pathways which seem essential to language. This was shown by inducing reproducible speech disturbances during stimulations as follows: the subcallosal fasciculus (initiation disorders), the periventricular white matter (dysarthria), the arcuate fasciculus and the insular connections (anomia). Clinically, all patients except three presented a transient postoperative dysphasia, which resolved within 3 months. On control MRI, 14 resections were total and 16 subtotal due to infiltration of functional bundles described above. It is recommended that the combination of the techniques as described could prove ideal for future non-invasive reliable subcortical mapping both in healthy volunteers and in patients harbouring a (cortico)subcortical lesion in order to optimize surgical pre-planning. |
Experimental transplantation in rodent models of CNS demyelination has led to the idea that Schwann cells may be candidates for cell therapy in human myelin diseases. Here we investigated the ability of Schwann cells autografts to generate myelin in the demyelinated monkey spinal cord. We report that monkey Schwann cells derived from adult peripheral nerve biopsies retain, after growth factor expansion and transduction with a lentiviral vector encoding green fluorescent protein, the ability to differentiate in vitro into promyelinating cells. When transplanted in the demyelinated nude mouse spinal cord, they promoted functional and anatomical repair of the lesions (n = 12). Furthermore, we obtained evidence by immunohistochemistry (n = 2) and electron microscopy (n = 4) that autologous transplantation of expanded monkey Schwann cells in acute lesions of the monkey spinal cord results in the repair of large areas of demyelination; up to 55% of the axons were remyelinated by donor Schwann cells, the remaining ones being remyelinated by oligodendrocytes. Autologous grafts of Schwann cells may thus be of therapeutic value for myelin repair in the adult CNS. |
Neuropsychiatric disorders are one of the main challenges of human medicine with epilepsy being one of the most common serious disorders of the brain. Increasing evidence suggest neuropeptides, particularly the opioids, play an important role in epilepsy. However, little is known about the mechanisms of the endogenous opioid system in epileptogenesis and epilepsy. Therefore, we investigated the role of endogenous prodynorphin-derived peptides in epileptogenesis, acute seizure behaviour and epilepsy in prodynorphin-deficient mice. Compared with wild-type littermates, prodynorphin knockout mice displayed a significantly reduced seizure threshold as assessed by tail-vein infusion of the GABA(A) antagonist pentylenetetrazole. This phenotype could be entirely rescued by the kappa receptor-specific agonist U-50488, but not by the mu receptor-specific agonist DAMGO. The delta-specific agonist SNC80 decreased seizure threshold in both genotypes, wild-type and knockout. Pre-treatment with the kappa selective antagonist GNTI completely blocked the rescue effect of U-50488. Consistent with the reduced seizure threshold, prodynorphin knockout mice showed faster seizure onset and a prolonged time of seizure activity after intracisternal injection of kainic acid. Three weeks after local injection of kainic acid into the stratum radiatum CA1 of the dorsal hippocampus, prodynorphin knockout mice displayed an increased extent of granule cell layer dispersion and neuronal loss along the rostrocaudal axis of the ipsi- and partially also of the contralateral hippocampus. In the classical pentylenetetrazole kindling model, dynorphin-deficient mice showed significantly faster kindling progression with six out of eight animals displaying clonic seizures, while none of the nine wild-types exceeded rating 3 (forelimb clonus). Taken together, our data strongly support a critical role for dynorphin in the regulation of hippocampal excitability, indicating an anticonvulsant role of kappa opioid receptors, thereby providing a potential target for antiepileptic drugs. |
Evidence from magnetic resonance imaging (MRI) suggests early structural and functional brain changes in individuals with the Huntington's disease (HD) gene mutation who are presymptomatic for the motor symptoms of the disorder (pre-HD subjects). The objective of this study was to investigate the functional neuroanatomy of verbal working memory (WM) in pre-HD subjects. By means of event-related functional MRI, we studied healthy controls (n = 16) and pre-HD subjects (n = 16) with a parametric WM paradigm comprising three different WM load levels. Voxel-based morphometry (VBM) was used to control potentially confounding brain atrophy. Although WM performance did not significantly differ between pre-HD subjects and healthy controls, pre-HD subjects showed a significantly decreased activation of the left dorsolateral prefrontal cortex (DLPFC) at intermediate and high WM load levels only. This region was not affected by early cortical atrophy, as revealed by VBM. Pre-HD individuals close to the onset of motor symptoms showed an increased activation of the left inferior parietal lobule and the right superior frontal gyrus compared with both pre-HD subjects far from symptom onset and healthy controls. In addition, the activation level in the left DLPFC was positively correlated with the UHDRS cognitive subscore in pre-HD subjects. Our findings demonstrate that early functional brain changes in pre-HD subjects may occur in the DLPFC before manifest cortical atrophy, and support a role of this region in the expression of clinical symptoms. Compensatory brain responses in pre-HD individuals may occur with closer proximity to the onset of manifest clinical symptoms. |
In the large group of genetically undetermined infantile-onset mitochondrial encephalopathies, multiple defects of mitochondrial DNA-related respiratory-chain complexes constitute a frequent biochemical signature. In order to identify responsible genes, we used exome-next-generation sequencing in a selected cohort of patients with this biochemical signature. In an isolated patient, we found two mutant alleles for EARS2, the gene encoding mitochondrial glutamyl-tRNA synthetase. The brain magnetic resonance imaging of this patient was hallmarked by extensive symmetrical cerebral white matter abnormalities sparing the periventricular rim and symmetrical signal abnormalities of the thalami, midbrain, pons, medulla oblongata and cerebellar white matter. Proton magnetic resonance spectroscopy showed increased lactate. We matched this magnetic resonance imaging pattern with that of a cohort of 11 previously selected unrelated cases. We found mutations in the EARS2 gene in all. Subsequent detailed clinical and magnetic resonance imaging based phenotyping revealed two distinct groups: mild and severe. All 12 patients shared an infantile onset and rapidly progressive disease with severe magnetic resonance imaging abnormalities and increased lactate in body fluids and proton magnetic resonance spectroscopy. Patients in the 'mild' group partially recovered and regained milestones in the following years with striking magnetic resonance imaging improvement and declining lactate levels, whereas those of the 'severe' group were characterized by clinical stagnation, brain atrophy on magnetic resonance imaging and persistent lactate increases. This new neurological disease, early-onset leukoencephalopathy with thalamus and brainstem involvement and high lactate, is hallmarked by unique magnetic resonance imaging features, defined by a peculiar biphasic clinical course and caused by mutations in a single gene, EARS2, expanding the list of medically relevant defects of mitochondrial DNA translation. |
Patients rarely experience visual hallucinations while being observed by clinicians. Therefore, instruments to detect visual hallucinations directly from patients are needed. Pareidolias, which are complex visual illusions involving ambiguous forms that are perceived as meaningful objects, are analogous to visual hallucinations and have the potential to be a surrogate indicator of visual hallucinations. In this study, we explored the clinical utility of a newly developed instrument for evoking pareidolic illusions, the Pareidolia test, in patients with dementia with Lewy bodies-one of the most common causes of visual hallucinations in the elderly. Thirty-four patients with dementia with Lewy bodies, 34 patients with Alzheimer's disease and 26 healthy controls were given the Pareidolia test. Patients with dementia with Lewy bodies produced a much greater number of pareidolic illusions compared with those with Alzheimer's disease or controls. A receiver operating characteristic analysis demonstrated that the number of pareidolias differentiated dementia with Lewy bodies from Alzheimer's disease with a sensitivity of 100% and a specificity of 88%. Full-length figures and faces of people and animals accounted for >80% of the contents of pareidolias. Pareidolias were observed in patients with dementia with Lewy bodies who had visual hallucinations as well as those who did not have visual hallucinations, suggesting that pareidolias do not reflect visual hallucinations themselves but may reflect susceptibility to visual hallucinations. A sub-analysis of patients with dementia with Lewy bodies who were or were not treated with donepzil demonstrated that the numbers of pareidolias were correlated with visuoperceptual abilities in the former and with indices of hallucinations and delusional misidentifications in the latter. Arousal and attentional deficits mediated by abnormal cholinergic mechanisms and visuoperceptual dysfunctions are likely to contribute to the development of visual hallucinations and pareidolias in dementia with Lewy bodies. |
## Introduction and Methods
Hand selection was assessed in preadolescent children (ages 9–11) within a preferential reaching task to delineate the effects of object location, orientation, and task intention on the assessment procedure and compared to data previously acquired from young adults.
## Results
The observed differences support the notion that children are still in a process of refining their movements in attempt to discern the most efficient and effective patterns of behavior. Notwithstanding differences in performance, similarities between preadolescents and young adults also emerged. Greater right‐hand selection in right space and when the handle was oriented to the right indicate that object proximity and orientation influence efficiency and thus constrain hand selection in unimanual object manipulation and role‐differentiated bimanual manipulation.
## Conclusions
Together, findings add to our understanding of hand preference, unimanual and bimanual object manipulation.
## INTRODUCTION
Handedness is traditionally assessed using measures of preference and performance. Preference indicates the hand typically selected for an action, whereas performance differentiates between abilities of the two hands when completing an action (McManus & Bryden, ). Questionnaires are used most frequently to confirm the direction (i.e., left or right) of hand preference, and a plethora of options are currently available (Scharoun & Bryden, ). The most commonly used questionnaires are the Annett Handedness Questionnaire (Annett, ), Edinburgh Handedness Inventory (Oldfield, ), and Waterloo Handedness Questionnaire (Steenhuis, Bryden, Schwartz, & Lawson, ). Performance measures, in comparison, assess the degree (i.e., strength) of hand preference (Provins & Magliaro, ) through manual strength, speed, accuracy, and/or precision (Scharoun & Bryden, ). Peg‐moving tasks, such as the Annett Pegboard (Annett, ) and Grooved Pegboard (Matthews & Klove, ), and manual aiming tasks (e.g., Roy & Elliott, ) are examples of performance assessments. It is generally understood that the preferred hand is more adept in performance, particularly for right‐handers (Annett, ).
Beyond traditional measures, the assessment of manual midline crossing has also been used to quantify handedness. Failure to reach across the midline into contralateral space by the age of 3 or 4, which is part of the typical progression of sensorimotor development, may highlight a delay or problem that might manifest later in life (Michell & Wood, ). As such, findings that the development of hand preference influences reaching across the body into contralateral space (Carlier, Doyen, & Lamard, ; Doyen, Dufour, Caroff, Cherfouh, & Carlier, ) are supported by research with individuals with neurodevelopmental disorders (Gérard‐Desplanches et al., ; Groen, Yasin, Laws, Barry, & Bishop, ; Hill & Bishop, ).
One method of assessing manual midline crossing involves the assessment of hand selection for reaching throughout regions of hemispace. Bishop, Ross, Daniels, and Bright introduced the Quantification of Hand Preference task in 1996 as one method. Here, three playing cards were placed at 30‐degree intervals in hemispace, and participant hand selection was recorded in three tasks (card pointing, reaching, and posting). Findings revealed the ability to discriminate between left and right‐handers as a function of direction and degree of handedness (Bishop, Ross, Daniels, & Bright, ; Calvert, ; Doyen & Carlier, ), with high homogeneity and test–retest reliability (Doyen & Carlier, ). The card‐reaching task also proved to be sensitive to developmental processes. For example, Carlier et al. ( ) revealed significant differences between young children (ages 3–4) and older children (ages 9–10) and recorded fewer reaches into extreme regions of hemispace with the contralateral hand. Doyen et al. ( ) similarly demonstrated that adolescent children (ages 13–14) and adults reached across the body into contralateral space less often than preadolescent children (ages 7–12).
In addition to card‐reaching, researchers have asked participants to grasp and manipulate the same object in different movement contexts to discern how task complexity influences hand selection throughout regions of space. In one example, Bryden and Roy ( ) had right‐handed children (ages 3–10) reach for objects located at 45‐degree intervals in hemispace and perform simple (toss) and complex (place into receptacle of same size and shape) actions. More recent work from Bryden, Mayer, and Roy ( ) required right‐ and left‐handed children (ages 3–12) and young adults (ages 18–22) to pick‐up and use one of five objects (pencil, paintbrush, spoon, toothbrush, and toy hammer) and five identical dowels located in hemispace. Taken together, findings have revealed patterns of handedness concurrent with those that emerge from more traditional methods (e.g., questionnaires, peg‐moving tasks). As summarized in a review from Scharoun and Bryden ( ), young children (ages 3–5) are generally observed exploring the environment. As direction of hand preference is not yet established, the hand closest to the object is typically selected, reflecting a lack of differentiation between the two hands. Between the ages of 6 and 10, children have garnered more experience and have established which hand is more skilled; therefore, the preferred hand is selected, even in cases when it is not necessarily the most efficient. Finally, an “adult‐like” pattern of behavior is evident by approximately age 10 to 12, as preadolescent children learn to be less dependent on the preferred hand, and nonpreferred hand performance increases (Scharoun & Bryden, ).
This study aimed to further investigate the development of manual midline crossing in preadolescent children (ages 9–11) by extending Scharoun, Scanlan, and Bryden's ( ) work with young adults. In that study, a preferential reaching task with coffee mugs was used to assess hand selection. Mugs were placed in three regions of hemispace (right space, midline, and left space), and handle orientation also varied (toward, away, to the left, and to the right of the participant). Hand selection was assessed in four different tasks: (a) pick‐up; (b) pick‐up and pour; (c) pick‐up and pass; and (d) pick‐up, pour, and pass. For adults, a right‐hand preference emerged for unimanual (pick‐up; pick‐up and pass) tasks, especially when reaching for objects in right space. Bimanual tasks (pick‐up and pour; pick‐up, pour, and pass) revealed role differentiation between the two hands. The left hand was selected to reach for and stabilize the mug, leaving to right hand to mobilize the pitcher. Such findings were concurrent with the dynamic dominance hypothesis proposed by Sainburg et al. (Mutha, Haaland, & Sainburg, ; Przybyla, Good, & Sainburg, ; Sainburg, ; Sainburg & Kalakanis, ; Sainburg & Wang, ) and related literature assessing role‐differentiated bimanual manipulation (Babik & Michel, ; Kimmerle, Mick, & Michel, ; Ramsay & Weber, ). Here, the preferred hand is considered role‐differentiated bimanual manipulation hand preference, as it performs the more complex, mobilizing aspect of the task, whereas the nonpreferred hand serves a more subservient, stabilizing role (Guiard, ; Peters, ). In consideration of previous work comparing behaviors of adults and preadolescent children in unimanual object manipulation and role‐differentiated bimanual manipulation (e.g., Bryden & Roy, ; Bryden et al., ; Rudisch, Butler, Izadi, Birtles, & Green, ; Scharoun & Bryden, ), it was hypothesized that hand selection tendencies would differ from those of adults reported by Scharoun et al. ( ).
## METHODS
### Participants
Twenty‐four right‐handed children ages 9–11 ( n = 6 age 9, 1 male, 5 female; n = 9 age 10, 6 male, 3 female; n = 9 age 11, 7 male, 2 female) participated in this research. Age was only recorded in years. Data were compared to 39 right‐handed undergraduate and graduate students (ages 18–30, exact ages not recorded; 14 male, 25 female) from Scharoun et al. ( ). Recruitment and testing procedures were reviewed and approved by the institution research ethics board. Informed consent was obtained from a parent/guardian of all participating children and adult participants. Child participants also provided verbal assent prior to participation.
### Apparatus and procedures
Using the same methods as Scharoun et al. ( ), participants completed: (a) the 32‐item Waterloo Handedness Questionnaire (Steenhuis et al., ); and (b) a Preferential Reaching Task.
#### Waterloo handedness questionnaire (WHQ)
A pen and paper task used to quantify hand preference, participants indicate their preferred hand for 32‐unimanual tasks by circling one of five responses: left always, left usually, both equally, right usually, and right always. A score of −2 (left always) to +2 (right always) is assigned to each response, and a total handedness score is computed (−64 to +64). A negative score is expected for left‐handed participants, and a positive score is expected for right‐handed participants.
#### Preferential reaching task
Participants were seated across from a researcher throughout the duration of the task. Three differently colored, yet identically proportioned coffee mugs were placed within reaching distance (20 cm) in left space (0°), at the midline (90°), and in right space (180°; Figure ). The mug handle was oriented toward, away from, to the left or right of the participant. A water pitcher, without a handle, was placed at the participant's midline. The pitcher was filled with lukewarm water to simulate a warm beverage without the potential risk of burning that the use of boiling water would have introduced.
Study setup. The participant (black) sat directly across from the researcher (grey). Mugs were placed within reaching distance in left‐ and right space and at the midline. A water pitcher was placed at the participant's midline. Handles were oriented to the right, left, toward, or away from the participant
Participants were asked to complete four different tasks: (a) pick‐up (i.e., pick‐up the mug); (b) pour (i.e., pick‐up the mug and pour a glass of water); (c) pass (i.e., pick‐up the mug and pass it to the researcher; and (d) pour and pass (i.e., pick‐up the mug, pour a glass of water, and pass it to the researcher). Each trial started with the participants’ hands in a neutral position. Participants were not instructed how to complete the task (i.e., were not instructed which hand to use, or to grasp the mug by than handle); therefore, hand and grasp selection were freely selected. Each location–handle orientation–task combination was performed twice, for a total of 96 trials. Although trials were blocked by handle orientation, the order of task presentation was randomized. Participant behavior was recorded using a video camera placed in a front view. Hand selection to pick‐up the mug was coded offline, and the percentage of right‐hand selection was computed.
### Data analysis
The percentage of right‐hand selection was the dependent measure. For analysis purposes, tasks were separated into unimanual (pick‐up, pass) and bimanual (pour, pour and pass). Using SPSS© statistics 24 software ( ), data were submitted to a 2 (group) × 2 (task) × 3 (location) × 4 (handle) mixed analysis of variance test with repeated measures.
## RESULTS
Only significant results with large effect sizes ( ≥ 0.14; Lakens, ) will be discussed in detail. All other nonsignificant effects and significant effects with small and medium effect sizes can be found in Table .
Interactions that emerged with small and medium effect sizes
### Unimanual tasks (pick‐up, pass)
A main effect of handle ( F (3, 183), p < 0.001, = 0.166) revealed the right hand was selected more often when the handle faced right ( M = 61.51, SD = 42.08) compared to all other orientations (left: M = 49.47, SD = 44.00, toward: M = 52.25, SD = 44.47, away: M = 50.27, SD = 45.78). Main effects of task ( F (1, 61) = 24.928, p < 0.001, =0.290) and location ( F (2, 122) = 229.288, p < 0.001, = 0.790) will be discussed within the significant two‐way interaction ( F (2, 122) = 28.615, p < 0.001, = 0.319; Figure ). Here, right‐hand selection in both tasks was most prevalent in right space, and least prevalent in left space compared to the midline. Furthermore, right hand selection was greater in right space for pass compared to pick‐up.
Right‐hand selection in right space was greater in pass compared to pick‐up
Two‐way interactions between task and group ( F (1, 61) = 26.943, p < 0.001, = 0.306; Figure ), and handle and group ( F (3, 183) = 8.814, p < 0.001, = 0.126; Figure ) were revealed. Significantly fewer right‐hand selections were displayed by children in pick‐up compared to pass and compared to adults. No difference emerged between children and adults in pass. Although no differences emerged in adults as a function of handle, children displayed an increase in right‐hand selection when the handle faced right, and the least when the handle faced. The proportion of right‐hand selection differed in all handle orientations, for children, except and away. Furthermore, when the handle faced left, adults displayed significantly more right‐hand selection than children.
Children displayed less right‐hand selection in pick‐up compared to adults and compared to pass
Unlike adults, differences in right‐hand selection emerged for children because of handle orientation. Furthermore, children displayed less right‐hand selection than adults when the handle faced left
### Bimanual tasks (pour, pour and pass)
A main effect of location ( F (2, 122) = 41.942, p < 0.001, = 0.407) revealed differences in all locations, with more right‐hand selection in right space ( M = 39.98, SD = 46.38), compared to the midline ( M = 16.23, SD = 34.10) and left space ( M = 11.31, SD = 29.26). Furthermore, the main effect of handle ( F (3, 183) = 11.055, p < 0.001, = 0.153) revealed more right‐hand selection when the handle was oriented to the right ( M = 27.38, SD = 41.81), compared to left ( M = 16.80, SD = 34.58), toward ( M = 20.05, SD = 38.04), and away ( M = 21.43, SD = 38.93).
## DISCUSSION
In line with our hypothesis, differences emerged between the two groups. In particular, right‐hand selection differed as a function of location for adults in both tasks; however, for children, this was only the case for the pass task. In pick‐up, right‐hand selection at the midline and in right space did not differ. Findings support the notion that preadolescent children are still in a process of refining their movements in attempt to discern the most efficient and effective patterns of behavior. For example, Mason, Bruyn, and Lazarus ( ) stated, about unimanual and bimanual object manipulation, that “the transition time between the ages of 7–10 may therefore be used as a testing period to determine the most effective strategies for accomplishing a variety of task goals” (p. 162). Likewise, Hausmann, Waldie, & Corballis ( ) have described the transition from immature to mature motor control to persist through the ages 10–12. Findings revealed in the present study are thus likely attributed to this transition period. Here, hand selection was similar for the most part (i.e., in most trials); however, minute differences prevent the conclusion that adult‐like behavior is evident in this age group.
Role‐differentiated bimanual manipulation emerges early in a child's life, such that partially differentiates roles for each hand are displayed by approximately 13 months (Babik & Michel, ). Throughout development, the preferred hand establishes itself in a holding and stabilizing role, whereas the nonpreferred hand becomes more proficient at object manipulation (Kimmerle, Ferre, Kotwica, & Michel, ). Findings are concurrent with recent work from Rudisch et al. ( ). Here, 5‐ to 16‐year‐olds opened the lid of a transparent box with one hand and used the other hand to press a button inside the box. The leading hand altered between preferred and nonpreferred hand in two task conditions. Although no differences in task performance were revealed based on the leading hand, young children (ages 5–6) were more variable in temporal cooperation compared to older children (ages 7–9) and adolescents (ages 10–16). Additional analyses, which also assessed unimanual subtasks (i.e., opening the lid, pressing a button), revealed a decrease in smoothness of movement across all participants in bimanual actions. The difference was particularly large for young children, and more apparent between ages 5–6 and 7–9. The decrease in variability and increase in smoothness was interpreted as evidence of automatization (Cohen & Sternad, ) with increasing age (Rudisch et al. ( ). With respect to findings of the current study, it can be argued that children are in the process of automatization, albeit not yet to the same extent as would be expected in young adults.
Beyond developmental effects, a general tendency for greater right‐hand selection in right space and when the handle was oriented to the right was revealed in both unimanual and bimanual tasks. Extending the work of Scharoun et al. ( ) to include preadolescent children, findings are concurrent with the kinaesthetic hypothesis (Gabbard & Rabb, ), which states that preferred hand use will be limited in contralateral space, due to biomechanical constraints. As such, object proximity and efficiency constrain hand selection in preferential reaching task.
Findings also support the notion that an object's orientation influences efficiency and thus constrains hand selection (Scharoun et al., ). Previous work has demonstrated that when a handled mug is viewed, the typical response involves grasping the handle (Ellis & Tucker, ; Tucker & Ellis, ). That said, it is important to acknowledge participants in the present study were not explicitly instructed to grasp the mug by the handle and were thus free to grasp the mug as they best saw fit. Lindemann, Stenneken, Van Schie, and Bekkering ( ) argued that the action of grasping a handled mug only occurs when there exists an intention for mug use. More recent work from van Elk, van Schie, and Bekkering ( , ) has proposed that the human capacity to use tools and objects involves “automatic effects of affordances as well as context‐ and intentionally driven effects” (van Elk et al., ; p. 240). Scharoun et al. ( ) assessed the propensity to grasp the mug by the handle, in addition to hand selection. They found that the handle was grasped more often in independent object manipulation. It can thus be argued that the hand selection tendencies, like grasp selection, are influenced by a multitude of different factors, including affordances, and the circumstances in which the action is performed. Together, findings are in line with previous assessments of preferential reaching (e.g., Bryden & Huszczynski, ), which have discussed the influence of object orientation and location, arm position and task complexity on hand selection.
Although we opted to separate unimanual and bimanual tasks for ease of analysis, as evident in results, and in line with previous reports, right‐hand selection was more prominent in unimanual tasks. Results offer support for role differentiation between the two hands in bimanual tasks. Concurrent with the dynamic dominance hypothesis proposed by Sainburg et al. (Mutha et al., ; Przybyla et al., ; Sainburg, ; Sainburg & Kalakanis, ; Sainburg & Wang, ) and the asymmetric division of labor hypothesis proposed by Guiard ( ), the left hand was selected to reach for and stabilize the mug, leaving to right hand to mobilize the pitcher. Findings from the two bimanual tasks (pour, pour and pass) revealed hand selection tendencies to be more similar than different when comparing preadolescent children to you adults. Although interactions involving the “group” factor did emerge, effect sizes were not large and therefore were not elaborated upon (see Table ).
Taken together, findings revealed differences between hand selection patterns of preadolescent children and young adults. Here, it can be argued that children are still in a process of refining their movements in attempt to discern the most efficient and effective patterns of behavior. As such, unlike young adults, automatization of bimanual control is not yet established. Notwithstanding differences in performance, similarities between preadolescents and young adults also emerged. Greater right‐hand selection in right space and when the handle was oriented to the right provides support for the kinaesthetic hypothesis (Gabbard & Rabb, ). As such, object proximity and orientation influence efficiency and thus constrain hand selection in unimanual object manipulation and role‐differentiated bimanual manipulation.
## CONFLICT OF INTEREST
None declared.
|
## Background and purpose
Intravenous thrombolysis (IVT) has become the standard treatment for acute ischemic stroke within 4.5 hr after symptoms onset. However, a fraction of patients would develop early neurological deterioration (END) after IVT. The aim of our study was to explore the utility of neutrophil–lymphocyte ratio (NLR) in predicting END.
## Methods
From October 2016 to March 2018, 342 consecutive patients with thrombolytic therapy were prospectively enrolled in this study. Blood cell counts were sampled in stroke emergency room before IVT. END was defined as a National Institutes of Health Stroke Scale score increase of ≥4 points within 24 hr after IVT. Multiple regression analysis was used to investigate the potential risk factors of END. We also performed receiver operating characteristic curve analysis and nomogram analysis to assess the overall discriminative ability of the NLR in predicting END.
## Results
Of the 342 patients, 86 (25.1%) participants were identified with END. Univariate logistic regression analysis demonstrated that patients with NLR in the third tertile, compared with the first tertile, were more likely to have END (odds ratio, 9.783; 95% confidence interval [CI], 4.847–19.764; p = .001). The association remained significant even after controlled for potential confounders. Also, a cutoff value of 4.43 for NLR was detected in predicting post‐thrombolysis END with a sensitivity of 70.9% and a specificity of 79.3% (area under curve, 0.779; 95% CI, 0.731–0.822). Furthermore, our established nomogram indicated that higher NLR was an indicator of post‐thrombolysis END (c‐index was 0.789, p < .001).
## Conclusions
This study showed that elevated level of NLR may predict post‐thrombolysis END in ischemic stroke patients.
This study showed that elevated level of neutrophil–lymphocyte ratio (NLR) may predict post‐thrombolysis early neurological deterioration (END) in ischemic stroke patients. A cutoff value of 4.43 for NLR was detected in predicting post‐thrombolysis END with a sensitivity of 70.9% and a specificity of 79.3% (area under curve, 0.779; 95% CI, 0.731–0.822). Furthermore, our established nomogram indicated that higher NLR was an indicator of post‐thrombolysis END (c‐index was 0.789, p < .001).
## INTRODUCTION
Stroke is one of leading causes of mortality and long‐term morbidity in China (Feigin et al., ). Intravenous thrombolysis (IVT) with recombinant tissue plasminogen activator administered up to 4.5 hr after onset has demonstrated benefits for acute ischemic stroke (AIS) with proven efficacy in reducing mortality and long‐term morbidity (National Institute of Neurological Disorders & Stroke rt‐PA Stroke Study Group, ). However, within 24 hr after IVT, a fraction of patients may experience neurological deficit worsening, described as early neurological deterioration (END; Seners et al., , ), which has been reported to be associated with poor outcomes (Mori et al., ). Thus, it is of great importance to explore the mechanism and the associated risk factors for post‐thrombolysis END.
Inflammation plays an important role in the pathophysiology of cerebrovascular diseases (Barone et al., ; Zhang et al., ). Except for the cytokines and chemokines released from in situ ischemic tissues, infiltration of peripheral circulating leukocytes, especially neutrophils, has been regarded as an important contributor to brain injury following ischemia (Qun et al., ). Several studies have confirmed the pivotal role of neutrophils in functional outcome after AIS (Gusdon et al., ; Wu et al., ). Neutrophil–lymphocyte ratio (NLR), a novel inflammatory marker that can be easily calculated from the differential white blood cell (WBC) count, was reported to be associated with the mortality and long‐term disability in stroke population (Gusdon et al., ; Köklü et al., ; Qun et al., ; Tokgoz et al., ). Also, high levels of NLR had a predictive value for 90‐day outcome of stroke patients treated with endovascular therapy (Brooks et al., ). However, there is a lack of data regarding the relationship between NLR and END in ischemic stroke patients underwent IVT. Therefore, we performed this prospectively observational study to explore the utility of NLR in predicting END after IVT.
## METHODS
### Study population
This prospective study was performed from October 2016 to March 2018 in Nanjing First Hospital. Patients with first‐ever AIS treated with IVT within 4.5 hr after symptom onset were included in the study. Patients treated with a bridging therapy consisting of IVT followed by endovascular therapy were also included. The exclusion criteria were as follows: (a) age <18 years and (b) unstable medical conditions such as systemic inflammatory disease, renal failure, hepatic failure, brain tumor, and presence of an active infection. Patients with hospital transfer were also excluded. Informed consent was obtained from participants or legal representatives, and the protocol was approved by the Ethical Committee of Nanjing First Hospital.
### Clinical assessments
Clinical assessments were performed within 24 hr after admission. All participants had standard assessments of demographic characteristics, vascular risk factors (including hypertension, diabetes mellitus, dyslipidemia, current smoking, current drinking, previous stroke, atrial fibrillation, and coronary heart disease), stroke severity, stroke subtype, and laboratory data. Symptomatic intracranial hemorrhage (sICH) was defined as any hemorrhagic transformation associated with NIHSS score worsening ≥4 points. Malignant edema was considered if brain was swelling and midline shift was present together with worsening of consciousness. Computed tomography, magnetic resonance and digital subtraction angiography, and electrocardiogram, transcranial Doppler, and carotid ultrasonography were performed for assessing the stroke etiology. Stroke subtype was classified according to Trial of Org 10172 in Acute Stroke Treatment (TOAST) criteria (Adams et al., ). Blood cell counts, including total leukocyte, neutrophil, and lymphocyte counts, and clotting routine were sampled from each participant in stroke emergency room on admission. Then, the cell counts were analyzed by an auto‐analyzer (XE‐2100, Sysmex). NLR was calculated as neutrophil counts/lymphocyte counts.
### Definition of END
The evaluation of neurological deficits was conducted using the National Institutes of Health Stroke Scale (NIHSS) score on admission and continued at the following 24 hr after IVT by two certified neurologists blind to clinical data. Post‐thrombolysis END was defined as a NIHSS score increase of ≥4 points between baseline and the 24 hr after IVT (Mori et al., ; Seners et al., , ).
### Treatment
All patients were treated with IVT within 4.5 hr after the onset of stroke symptoms in stroke emergency room. Once proximal arterial occlusion had been corroborated via magnetic resonance angiography or CTA, the participants would undergo rapid endovascular treatment. Other treatments, such as risk factor management and statin therapy, were also carried out as appropriate.
### Statistical analysis
Statistical analyses were performed with SPSS version 21.0 (SPSS Inc.). Continuous variables that followed normal distribution were expressed as mean ± standard deviation; other continuous variables that did not follow normal distributions were presented as the median and the interquartile range (25th to 75th percentile). Categorical variables are expressed as constituent ratios. Differences in baseline characteristics were tested using the analysis of variance or Kruskal–Wallis test for continuous variables, and Pearson's chi‐square test for categorical variables. We also used binary logistic regression analysis to detect the risk factors of END. Multivariable analysis was adjusted for all potential confounders with statistically significant association at p < .1 in univariate regression analysis (including age, eGFR ≤ 60 ml/min/1.73 m , initial NIHSS score, malignant edema, stroke subtype, hypersensitive C‐reactive protein, and fasting blood glucose level). Receiver operating characteristic (ROC) curve analysis was performed by assessing the overall discriminative ability of the NLR to predict post‐thrombolysis END and to establish optimal cutoff points at which the sum of the specificity and sensitivity was the highest. A MedCalc 15.6.0 (MedCalc Software) packet program was used to obtain ROC. In addition, a nomogram based on the independent predictors was constructed by R software with the package rms. The predictive capacity of the nomogram was determined by Harrell's c‐index. A two‐tailed value of p < .05 was considered significant.
## RESULTS
From October 2016 to March 2018, 386 patients were screened in this study. Forty‐four patients were excluded for the following reasons: systemic inflammatory disease ( n = 9), renal failure ( n = 14), hepatic failure ( n = 8), brain tumor ( n = 3), and presence of an active infection ( n = 10). A total of 342 subjects (233 men; mean age, 68.1 ± 12.3 years) were included for the final analysis. Among these patients, 238 (69.6%) had hypertension, 75 (21.9%) had diabetes mellitus, 94 (27.5%) had dyslipidemia, and 73 (21.3%) had atrial fibrillation.
After admission, END was observed in 86 patients (25.1%). The median NLR was 4.65, with tertile levels as follows: 0.66–2.27 (first tertile); 2.28–4.43 (second tertile); 4.48–35.73 (third tertile). Baseline characteristics of the study population according to the tertile of NLR are provided in Table . Increased NLR was related to post‐thrombolysis END ( p = .001), onset‐to‐treatment time ( p = .001), sICH ( p = .006), malignant edema ( p = .008), and high levels of fasting blood glucose ( p = .008).
Characteristics of subgroups based on the tertile of neutrophil–lymphocyte ratio
Comparisons of baseline characteristics in patients with or without END are shown in Table . Compared with patients without post‐thrombolysis END, patients with post‐thrombolysis END were older ( p = .016) and had higher proportions of proximal arterial occlusion ( p = .001); lower proportions of previous antiplatelet ( p = .020); lower levels of body mass index ( p = .028) and triglyceride ( p = .015); higher levels of initial NIHSS ( p = .001), hypersensitive C‐reactive protein ( p = .001), fasting blood glucose ( p = .001), and NLR ( p = .001).
Characteristics of subgroups based on the presence of post‐thrombolysis early neurological deterioration
Table showed the results of logistic regression analysis for risk factors with post‐thrombolysis END. Univariate logistic regression analysis demonstrated that the highest tertile of NLR, age, previous antiplatelet, eGFR ≤ 60 ml/min/1.73 m , initial NIHSS score, BMI, proximal arterial occlusion, sICH, malignant edema, hypersensitive C‐reactive protein, and fasting blood glucose level were associated with END. After adjusting for all potential confounders, the highest tertile of NLR (first quartile used as the reference value) was identified as an independent predictor for post‐thrombolysis END (odds ratio [OR], 6.406; 95% confidence interval [CI] 2.646–15.510, p = .002).
Logistic regression analysis for risk factors with post‐thrombolysis early neurological deterioration
Of particular interest, IVT patients were further divided into those with and without endovascular therapy. Compared with the first tertile, NLR in the third tertile showed a higher trend for the occurrence of post‐thrombolysis END (Table ; OR, 7.064; 95% CI, 3.009–16.583; p = .001) in patients without endovascular therapy. Furthermore, in bridging therapy group, higher NLR level remained associated with post‐thrombolysis END.
Subgroup analysis according to the patients undergoing intravenous thrombolysis with and without endovascular therapy
As for the clinical value of NLR in predicting post‐thrombolysis END, we performed ROC curve in Figure . The optimal cutoff value for NLR as a predictor of post‐thrombolysis END was determined as 4.43 in the ROC curve analysis, which yielded a sensitivity of 70.9% and a specificity of 79.3%, with the AUC at 0.779 (95% CI, 0.731–0.822).
Receiver operating characteristic (ROC) curve for the value of neutrophil–lymphocyte ratio (NLR) to predict post‐thrombolysis early neurological deterioration (END)
The nomogram is shown in Figure . The concordance index of this model was 0.789 ( p < .001). The novel models indicated that higher NLR was an indicator of post‐thrombolysis END. These findings were similar to those obtained previously in the multivariate logistic models.
Nomograms of acute ischemic stroke patients to predict post‐thrombolysis early neurological deterioration (END)
We divided the patients into group A (NLR < 4.43) and group B (NLR ≥ 4.43). Comparisons of characteristics based on the two groups are shown in Table . Compared with group B, patients in group A had less time from onset to treatment ( p = .003), higher proportions of previous statin ( p = .034) and proximal arterial occlusion ( p = .001), higher levels of triglyceride ( p = .034), lower levels of hypersensitive C‐reactive protein ( p = .033), and fasting blood glucose ( p = .007).
Characteristics of group A and group B
## DISCUSSION
In this study, we unveiled that NLR, an affordable and readily available tool, may be a powerful predictor of END in AIS patients being considered for IVT. Our study showed that patients with elevated NLR levels were at increasing risks of developing post‐thrombolysis END, even when controlling for age, proximal arterial occlusion, sICH, malignant edema, and other potential confounders. Furthermore, the optimal cutoff value of NLR to indicate post‐thrombolysis END was 4.43, and its corresponding sensitivity and specificity were 70.9% and 79.3%, respectively. Furthermore, our established nomogram indicated that higher NLR was an indicator of post‐thrombolysis END.
Clinical evidences have shown that, in AIS patients, high NLR levels are associated with increased infarct volume and mortality (Celikbilek, Ismailogullari, & Zararsiz, ; Gökhan et al., ; Tokgoz et al., ). Furthermore, NLR is a predictor of recurrent ischemic stroke and 90‐day poor functional outcome in AIS patients receiving endovascular stroke therapy or IVT or antiplatelet medications (Brooks et al., ; Duan et al., ; Malhotra et al., ; Qun et al., ). In the present study, it is the first time to investigate the relationship between NLR and early functional deterioration in AIS patients with IVT therapy. In accordance with previous studies (Rajajee et al., ; Seners, Turc, Oppenheim, & Baron, ; Thanvi, Treadwell, & Robinson, ), our cohort reported a prevalence of 25.1% in post‐thrombolysis END. We found that NLR revealed its predictive value in the occurrence of post‐thrombolysis END. It is worthwhile to note that, in bridging therapy group, NLR still has great potential as a predictor of END occurrence, suggesting the types of treatment may have minimal effects on the relationship between NLR and END following AIS.
Mechanistically, END is believed to be resulted from biochemical abnormality such as inflammation (Alawneh, Moustafa, & Baron, ; Zhang et al., ). END in patients with lacunar infarction has reported a correlation with high peripheral concentrations of pro‐inflammatory factors, such as IL‐6, TNF‐α, and intercellular adhesion molecule‐1 (Castellanos et al., ). In AIS, it is reported that the inflammatory process is launched within 24 hr at ischemic site and has an important role in exacerbating ischemic damage (Kim, Park, Chang, Kim, & Lee, ; Zhang, Chopp, Chen, & Garcia, ). The inflammatory cytokines and chemokines released from ischemic tissues guide the infiltration of circulating leukocytes, among which neutrophils are the most recognized mediator in ischemic brain injury (Wu et al., ). Neutrophil, the main inflammatory cell of AIS, on one hand, can release free oxygen radicals propagating secondary brain injury in penumbra regions (Ceulemans et al., ). On the other hand, neutrophils are the source of matrix metalloproteinase‐9 (MMP‐9), which can directly result in blood–brain barrier (BBB) breakdown and hemorrhagic transformation (Duan et al., ). These may be significant contributors in the occurrence of END. Additionally, some subtypes of lymphocyte have been reported to be major cerebroprotective immunomodulators after AIS in response to ischemic injury and are involved in reduced infarct volume and improved neurological function (Kim et al., ; Liesz et al., ). Therefore, higher lymphocyte count may be related to lower risks of END. However, single biomarker is prone to be affected by various physiological and pathological conditions (Nash et al., ). NLR, a composite parameter in the combination of neutrophils and lymphocytes, could serve to better reflect immunological activities of the cells and divide patients into comprehensive inflammatory profiles, playing a better role in predicting post‐thrombolysis END in patients with AIS.
Our study has some potential limitations. Firstly, our study was conducted within participants from one single center via strict exclusion criteria, whose results might not be able to generalize to the general population. Secondly, the sample size of the present study was relatively small. Larger cohorts of subjects are needed. Thirdly, it has been proposed that blood cell counts may change during the recovery of ischemic stroke (Iadecola & Anrather, ). To maximally reduce the possible correlation of this effect with our results, blood cell counts were assessed before IVT to minimize the time interval between the onset of stroke and blood sampling. Fourthly, although we found the relationship between NLR and post‐thrombolysis END, there did not exist dynamic examination of the blood cell count of every patient. Blood cell count needed to be examined dynamically in further studies. Finally, data were observational. We were therefore unable to establish a causal relationship between NLR and END after IVT.
In conclusion, from the present study, NLR levels appeared to be positively correlated with post‐thrombolysis END in ischemic stroke patients and can serve as a useful noninvasive biomarker for assessment of END after IVT.
## CONFLICT OF INTEREST
All the authors declare that there is no conflict of interest.
## AUTHOR CONTRIBUTIONS
All authors listed have contributed significantly and are in agreement with the content of the manuscript. Pengyu Gong was mainly involved in study design, data analysis, data acquisition, data interpretation, and manuscript preparation. Xiaohao Zhang was mainly involved in study design, data analysis, data interpretation, and manuscript preparation. Teng Jiang was mainly involved in data analysis, data interpretation, and manuscript preparation. Yi Xie was mainly involved in data interpretation and manuscript preparation. Yukai Liu was mainly involved in data acquisition and data analysis. Meng Wang, Huanhuan Sun, and Shuting Zhang were mainly involved in data acquisition. Junshan Zhou and Yingdong Zhang were mainly involved in study design, data interpretation, and manuscript preparation.
|
## Objective
Deductive reasoning is a complex and poorly understood concept in the field of psychology. Many cognitive neuroscience studies have been published on deductive reasoning but have yielded inconsistent findings.
## Methods
In this study, we analyzed collected data from 38 articles using a recently proposed activation likelihood estimation (ALE) approach and used conjunction analysis to better determine the intersection of the results of meta‐analyses.
## Results
First, the left hemispheres in the inferior parietal lobule (Brodmann area 40 [BA40]), middle frontal gyrus (BA6), medial frontal gyrus (BA8), inferior frontal gyrus (BA45/46), caudate, and insula (BA47) were revealed to be significant brain regions via simple‐effect analysis (deductive reasoning versus baseline). Furthermore, IFG, insula, and cingulate (the key neural hubs of the cingulo‐opercular network) were highlighted in overlapped functional connectivity maps.
## Conclusion
The findings of the current study are consistent with the view that deductive reasoning requires a succession of stages, which included decoding of linguistic information, conversion and correction of rules, and transformation of inferential results into conclusive outputs, all of which are putatively processed via a distributed network of brain regions encompassing frontal/parietal cortices, as well as the caudate and other subcortical structures, which suggested that in the process of deductive reasoning, the coding and integration of premise information is indispensable, and it is also crucial to the execution and monitoring of the cognitive processing of reasoning.
Deductive reasoning is supported by a distributed network of regions encompassing frontal/parietal cortices and subcortical structures (e.g., caudate). The results of our conjunction analysis highlight the IFG, insula, and cingulate (the key neural hubs of the cingulo‐opercular network) as core locus of deductive reasoning.
## INTRODUCTION
Deductive reasoning is fundamental to science, human culture, and for deriving solutions to problems in daily life (Fangmeier et al., ). The process of deductive reasoning starts with premises and attempts to reach a logically secure conclusion or a series of conclusions from prior beliefs, observations, and/or suppositions that are not explicit in the initial premises. As a higher cognitive activity of human beings, the mental processes underlying reasoning have been the focus of vigorous investigation within psychology and philosophy (Frank et al., ; Gordon, ).
The advent of neuroimaging techniques has increased the number of studies related to the neural basis of deductive reasoning. Cognitive studies on deductive reasoning have debated whether this process relies primarily on visuospatial mechanisms (Mental Model Theory, MMT) (Johnson‐Laird, ) or linguistic models (Mental Logic Theory, MLT) (Rips, ), and the experimental results have been inconclusive and inconsistent due to interference from various experimental factors. Some categories of such experimental caveats are as follows:
Mode of the deductive‐reasoning task . A review by Prado ( ) reported that the engagement of brain regions involved in deductive reasoning depends on the structure of the argument. Relational arguments are associated with activations in bilateral posterior parietal cortex (PPC) and right middle frontal gyrus (MFG), whereas categorical arguments are associated with the left inferior frontal gyrus (IFG) and left basal ganglion (BG). In contrast, propositional arguments are associated with the left PPC, left precentral gyrus (PG), and medial frontal gyrus (MeFG) (Prado et al., ).
Complexity of the reasoning task . Studies have shown that in contrast with simple deductive trials, complex deductive trials result in a pattern of activation across many brain regions, including left dorsolateral prefrontal (Luca et al., ), frontopolar, fronto‐medial, left frontal, and parietal cortices (Coetzee & Monti, ; Monti et al., ).
Stimulus presentation . Visual experiments have found activity in visual and temporal cortices during the premise processing phase, which suggests that the content for premises elicit visual mental images during this early stage (Fangmeier et al., ). However, this pattern of activity is not found in the absence of any correlated visual input; under this circumstance, deductive reasoning activates an occipitoparietal–frontal network, and occipital activation is found in visual association cortex (middle occipital gyrus) but not in primary visual cortex (Fangmeier & Knauff, ; Knauff et al., ; Markus et al., ).
Baseline task . Knauff et al. ( ) used rest intervals as a baseline for analyzing reasoning of a three‐term series and revealed baseline activation in the left posterior‐temporal cortex. An experiment using an unrelated task (only superficial processing of stimuli) as a baseline task evoked activation of the left IFG, left MFG, temporal gyrus, and cingulate gyrus (Goel et al., , ). Moreover, other studies have applied a memory task as a baseline, but the results of these studies have been inconsistent and have been unable to uncover any significant linguistic activations despite these experiments using verbal content (Monti et al., ; Monti et al., ; Parsons & Osherson, ).
A previous qualitative review of the neuroimaging literature on deductive reasoning identified a bilateral but rather left‐centered frontoparietal network for general deductive reasoning across all experiments by multilevel kernel density analysis (MKDA), and subdivided articles into those investigating relational arguments, categorical arguments, and propositional arguments (Prado et al., ), which provided a preliminary explanation regarding the differences in results caused by different reasoning arguments. In addition, a recent study incorporated recent research on conditional and syllogistic reasoning based on Prado's research, and further subdivided the article by their structure, content, and requirement for world knowledge, which found a widespread activation network encompassing the frontal, parietal, sublobar, limbic, posterior lobes and exhibit clear distinctions between the task's type and content (Wertheim et al., ). However, effective reasoning depends more on the integration of logical relations (i.e., premises) and the application of logical rules in deductive reasoning. Hence, it may be most useful to elucidate brain activation patterns that are independent of the external representational form of deductive reasoning, which would likely delineate the core brain regions underlying the fundamentals of deductive reasoning. In this study, we analyzed collected data using a recently proposed activation likelihood estimation (ALE) approach. The ALE results were assessed against a null distribution of random spatial association between experiments (Eickhoff et al., ), and this coordinate‐based meta‐analysis increases the population sample for better generalization by integrating data across several studies. Compared with a previous review (Prado et al., ), we innovatively used MACM analyses (meta‐analytic connectivity mapping), which delineates patterns of coactivation across thousands of studies using neuroimaging databases and produces data‐driven functional connectivity maps based on predefined ROIs (Langner et al., ). Moreover, MACM allows probing coactivation patterns, that is, task‐based functional connectivity, across a wide range of experimental settings (Bellucci et al., ), which may provide a better summary of the deductive reasoning research published over the past few decades from a new perspective.
Despite an increased number of studies on deductive reasoning, results have been inconsistent due to differing experimental conditions across studies. Neuroimaging meta‐analysis combines results of independent experiments to achieve a quantitative summary of the state of research in a specific domain (Turkeltaub et al., ). Here, we used ALE meta‐analysis to summarize patterns of activity related to deductive reasoning. Specifically, first, the meta‐analysis results of Prado ( ) have suggested that deductive reasoning does not rely on a unitary brain system but relies on fractionated neural systems located in both cortical (frontal and parietal cortices) and subcortical (BG) structures (Prado et al., ). Hence, we collectively analyzed 39 fMRI articles (published over the past 8 years) to further elucidate spatiotemporal brain activation patterns during the processing of deductive reasoning. Second, we used conjunction analysis to better determine the intersection of the results of meta‐analyses, which can further clarify relationships among key brain regions implicated in deductive‐reasoning processing. Third, the interaction between language and thought has become a pivotal phenomenon in the study of human cognition (Frank et al., ; Li & Gleitman, ), and there has been little agreement regarding this topic. Taken together, this study may further elucidate the relationship between deductive reasoning and language by determining language‐activated brain regions during deductive reasoning, which may serve as indicators of this fundamental cognitive process.
## METHODS
### Study selection
There were two qualitative reviews of the neuroimaging literature on deductive reasoning available prior to our meta‐analysis. One of these reviews (Goel et al., ), included studies published through April 2007, while the other review (Prado, ) included studies published through September 2010 (on the basis of the previous Goel review). Here, we searched for additional neuroimaging studies of deductive reasoning published from September 2010 to 2019. More specifically, we searched the PubMed, PNAS, SAGE, Oxford Press Wiley, Elsevier Science, and Baidu Scholar databases for studies on several related topics, including the following: reasoning, inference, deduction, deductive reasoning, deductive inference, conditional reasoning, relational reasoning; propositional reasoning, categorical reasoning, thinking, thought, theory of mind, functional magnetic resonance imaging, MRI, fMRI, and PET. Inclusion criteria for the articles were as follows: (a) the paper was written in English; (b) the task in the study involved deductive reasoning; (c) we here used MRI, fMRI, and PET to collect data; (d) all the subjects in the study were healthy without any psychiatric disorders; (e) the coordinates in each of the studies were in the standard Montreal Neurological Institute (MNI) or Talairach space; and (f) all the reported activation coordinates were based on the entire brain. Following these criteria, we ultimately included 38 published, peer‐reviewed fMRI articles on the neural substrates of deductive reasoning in the present meta‐analysis. The specific selection process is shown in Figure . From each study, we selected experiments corresponding most closely to a comparison between a reasoning condition and a baseline condition. A summary of the included details of each study in the meta‐analysis is provided in Table .
Flowchart of the study selection process for the meta‐analysis
Studies Included in the meta‐analysis
Note
### Meta‐analysis algorithm
Meta‐analysis was carried out using the revised version (Simon B. Eickhoff et al., ) of the ALE approach using Ginger ALE 3.0.2 software ( ) for coordinate‐based meta‐analysis of neuroimaging results (Laird et al., ; Turkeltaub et al., ). The key principle behind ALE is to treat the reported foci not as single points, but as centers of three‐dimensional Gaussian probability distributions for capturing the spatial uncertainty associated with each focus (Caspers et al., ). The probabilities of all activation foci in a given experiment were combined for each voxel, resulting in a modeled activation map (MA map). Taking the union across these MA maps yields voxel‐wise ALE scores describing the convergence of results at each particular location (Evans et al., ). The likelihood of activation for each standard‐space voxel was calculated under a null distribution of spatial independence (Fitzgerald et al., ; Sabatinelli et al., ). In brief, the ALE algorithm aims at identifying areas showing a convergence of activations across different experiments, and to determine if the clustering is higher than expected under the null distribution of a random spatial association between the results obtained in the experiments.
In the present study, all Talairach coordinates were transformed to MNI space using the icbm2tal transform, which has been shown to provide an improved fit over the mni2tal transform (Brett et al., ; Matthew et al., ). Based on the MNI stereotactic coordinates reported and transformed by the studies, ALE analysis of single datasets was conducted. ALE maps were created by modeling each focus as a three‐dimensional Gaussian function with a full‐width half‐maximum (FWHM) of 10 mm. Results were thresholded for significance using a cluster‐level family‐wise error (FWE) correction at p < .01 with a cluster defining threshold of p < .0001 (1,000 permutations, 200 mm minimum volume) (Eickhoff et al., ; Eklund et al., ). The results were viewed using MATLAB software (with loaded SPM and DPABI toolboxes) and were overlaid to a standard space using the MNI file.
We defined the brain regions obtained from the meta‐analysis results as our ROIs. To examine the coactivation patterns of these regions commonly recruited by deductive reasoning, we conducted MACM analyses (meta‐analytic connectivity mapping) using the BrainMap Database ( ) (Laird et al. ). MACM delineates patterns of coactivation across thousands of studies using neuroimaging databases and produces data‐driven functional connectivity maps based on predefined ROIs (Langner et al., ). For our analysis, we constrained our analysis to fMRI and PET experiments from “normal mapping” whole‐brain neuroimaging studies in healthy population, which report activation in standard space. While other studies investigating differences in age, gender, interventions, or clinical populations were excluded. For the IPL, 433 experimental contrasts and 7,041 foci from 5,223 participants were identified; for the MFG, 228 experimental contrasts and 3,372 foci from 3,432 participants; for the MeFG, 422 experimental contrasts and 6,769 foci from 5,140 participants; for the IFG (BA45), 135 experimental contrasts and 1,735 foci from 1,782 participants; for the IFG (BA46), 45 experimental contrasts and 662 foci from 755 participants; for the Caudate, 104 experimental contrasts and 1,897 foci from 1,730 participants; and for the Insula, 116 experimental contrasts and 1,641 foci from 1,708 participants. Importantly, MACM analyses cover experiments in the BrainMap database associated with different types of tasks that involve these activations (Laird et al., ). These analyses consisted of the following two steps. First, whole‐brain peak coordinates of all those experiments in the BrainMap database were downloaded if the study reported at least one focus of activation within each ROI. Second, ALE meta‐analyses were conducted over all coordinates of the retrieved experiments to quantify their convergence and coactivation with each ROI. Finally, the ALE maps were family‐wise error (FWE) corrected at a threshold of p < .01 at the cluster‐level. The resulting regions were anatomically labeled by reference to probabilistic cytoarchitectonic maps of the human brain using the DPABI toolbox. We reported the peak coordinates of the significant cluster and demonstrated the brain regions nearest the peak coordinates within ± 5 mm (Xin et al., ).
## RESULTS
Selected contrasts from 38 neuroimaging studies of deductive reasoning comparing a reasoning condition and baseline condition yielded a total of 702 foci. Pooling the results of 69 experiments onto a single brain resulted in a diffuse pattern of activation across all lobes, with some clustering evident across left hemispheres in the inferior parietal lobule (IPL, BA 40), middle frontal gyrus (MFG, BA 6), medial frontal gyrus (MeFG, BA 8), inferior frontal gyrus (IFG, BA 45/46), caudate, and insula (BA 47) (Figure ). Coordinates of the activation maxima of the meta‐analysis on deductive reasoning are provided in Table .
The results of ALE meta‐analysis revealed the key brain regions most consistently activated in neuroimaging studies of deductive reasoning. (IFG, inferior frontal gyrus; IPL, inferior parietal lobule; MeFG, medial frontal gyrus; MFG, middle frontal gyrus)
Brain regions that was significantly activated in all deductive‐reasoning studies
Note
Functional connectivity analysis was conducted to determine the intersection between these key brain regions identified during deductive reasoning. Since the caudate, one of the key brain regions, is a subcortical structure and the other brain regions are cortical structures, two separate conjunction analyses were conducted (one analysis for the cortical brain regions and a separate analysis for the subcortical caudate). The results of the analysis indicated that three common regions existed in the IFG, insula (BA 13), and cingulate gyrus (BA 32) of the left hemispheres among the cortical brain regions. In addition, the bilateral IFG, the bilateral caudate, MFG (BA 6), insula (BA 13), thalamus (BA 32), IPL, superior parietal lobule (SPL), and cingulate gyrus (BA 23/32) of the left hemispheres were revealed as common regions in the conjunction analysis of the caudate (Figure ). Coordinates of the activation maxima of the results of the conjunction analysis are provided in Table .
Functional‐connectivity analysis. (a) The results of the analysis for the cortical brain regions of the key brain areas (all brain regions are located in the left hemisphere). (b) The results of the analysis for the caudate (IFG, Inferior Frontal Gyrus; IPL, Inferior Parietal Lobule; L, left hemisphere; MFG, Middle Frontal Gyrus; R, right hemisphere; SPL, Superior Parietal Lobule)
Results of the meta‐analysis of functional connectivity
Note
## DISCUSSION
Previous studies have claimed that deductive reasoning does not rely on a unitary brain system but relies on a fractionated neural system located in both cortical (frontal and parietal cortices) and subcortical (BG) structures (Prado et al., ). As in previous investigations, our ALE meta‐analysis indicated both unified and fractionated neural systems activated in cortical areas, specifically IPL, MFG, MeFG, IFG, and insula, but we also found activity in subcortical (caudate) structures during deductive reasoning. On the whole, all the brain regions that were significantly activated were located in the left hemisphere, which is consistent with findings from previous studies on the neural basis of deductive reasoning. Monti et al. ( ) revealed a content‐independent network that was responsible for carrying out deductive processes and included the left hemisphere (frontal and parietal cortices). Moreover, experiments with patients suffering from brain lesions in the prefrontal lobe (Goel et al. ), temporal cortex (Langdon & Warrington, ; Read, ), or throughout widespread regions across the entire hemisphere (Golding, ) have supported a left hemisphere dominance for deductive reasoning. One possible explanation for this spatial localization is that when confronted with the premises of a deductive argument, the left hemisphere might recognize logical structures and generate a hypothesis regarding a collective conclusion (Goel, ). It is not surprising to link mental model theory, having a visuospatial nature, with right hemisphere activation and mental logic theory—which have a propositional nature—with left hemisphere activation (Heit, ). Nevertheless, some researchers (e.g., Knauff et al., ) have suggested that left hemisphere activation may be consistent with mental model theory, because comprehension of arguments will recruit linguistic areas of the brain (Monti et al., , ). However, intrahemispheric differences are apparent, and we will discuss relevant regions in the following sections:
For cortical (IPL, IFG and insula) structures, some studies have identified that the IPL is involved in spatial realignment, which governs shifts of spatial attention and target detection (Chapman et al., ; Shulman et al., ). Likewise, a large body of evidence coming from neuroimaging studies has implicated the activation of the left IFG in syntactic processing at the word or sentence level (Friederici & Kotz, ; Grodzinsky & Santi, ), which appears to support a syntactic‐ or rule‐based view of deductive reasoning (Goel. ). More specific, the left IFG (BA 45) is activated during both the encoding and the integration of propositional premises, and the BA46 of the frontal cortex is involved in the control and allocation of attention (Bishop, ; Luks et al., ), Others have also found that BA46 (Dorsolateral prefrontal cortex) may represent the site of meta‐cognition (Nelson, ; Tzu‐Ching et al., ), and is associated with rule‐guided operations in the early stage of conditional‐proposition testing (Jimei et al., ). Additionally, it is noteworthy that both the IPL and IFG are canonical mirror neuron brain regions in humans (Buccino et al., ; Chong et al., ; Iacoboni et al., ; Kilner et al., ), which play a crucial role in associative learning (Heyes, ; Ray & Heyes, ). Associative learning is often considered as a stable adaptation for tracking relationships between events (Heyes, ), and the mirror neuron system plays an important role in the representation and storage of simple and complex relationships (Cook, ; Ferrari & Fogassi, ; Keysers et al., ). Here, the IPL and IFG may be closely related to the activation of logically related experiences, which may indicate that human deductive reasoning is at least partially derived from experience/schema/mental models and that both of these two brain regions might mediate conclusions or activities in the context of processing premises. In other ways, the IPL and IFG—as key components of the mirror neuron system—may help us understand people's actions, words, and intentions, and it's even an important brain region for human empathy (Baird et al., ; Hojat & Cohen, ), which seem to a certain extent that the human reasoning system is an evolutionary system that helps individuals to prove and justify on the basis of understanding external information during evolution or social activities. Moreover, The activation of the insula seems to confirm the participation of higher‐order cognitive control and attention processing during this process (Luca et al., ).
Furthermore, the MFG and MeFG have been connected to working memory in previous studies (Kirschen et al., ; ; Owen, ; Ranganath et al., ; Ricciardi et al., ). An early response in the left MFG potentially reflects semantic comprehension processing (Porcaro et al., ). In this context, the MFG and MeFG may be particularly sensitive to the difficulty of deductive reasoning.
For subcortical (caudate) structures, previous imaging studies of reasoning have reported that activation of the caudate is involved when multiple rules need to be deduced and integrate (Christoff & Prabhakaran, ; Fangmeier et al., ; Prabhakaran et al., ). Furthermore, caudate activation has been linked to planning and executing demands of reasoning in both reasoning and working memory systems (Melrose et al., ). Moreover, in the present study, we found that the frontal lobe (IFG and MFG), parietal lobe (SPL and IPL) cingulate gyrus, insula, caudate and thalamus were all significant regions as indicated by conjunction analysis of the caudate. About frontal lobe, IFG and MFG are mainly related to the representation and integration of premise information and working memory. Similar to the IPL, left SPL occurs across a variety of tasks requiring manipulation and rearrangement of information in spatial working memory (Koenigs et al., ) and allocation of spatial attention (Dehaene et al., ). In regard to the cingulate gyrus, several recent neuroimaging findings suggest that it is associate with general executive, and it also plays a central role in a wide spectrum of highly integrated tasks such as visuospatial imagery and episodic memory retrieval (Cavanna & Trimble, ; Sommer et al., ). More significantly, the dorsal anterior cingulate and insular cortices represent key neural hubs of the cingulo‐opercular network, which is implicated in maintaining and implementing task set (Dosenbach et al., , ). Additionally, Many previous studies have indicated that the thalamus is related to various higher brain functions such as memory and learning (Karussis et al., ; Nagaratnam et al., ; Radanovic & Scaff, ), and may be related to the regulation and/or facilitation of ongoing cortical processing of memory and language (Baker et al., ; Ojemann et al., ; Elizabeth et al., ; Warburton et al., ). As an important nucleus within the basal ganglia, the caudate nucleus has a strong connection with the dorsolateral prefrontal lobe, and its interaction with the prefrontal lobe in deductive reasoning task may be related to the facilitation of inferences and application of rules.
The results of our functional conjunction analysis among the four cortical brain regions indicated that three common regions existed in the IFG, insula (BA 13), and cingulate gyrus (BA 32) of the left hemispheres. According to the brain maps, the activation position of the inferior frontal gyrus is adjacent to the triangle (BA45) (close to insula), and inferior frontal gyrus pars triangularis is a part of the frontoparietal operculum, which covers the upper surface of the insula. Reverberi et al. ( ) demonstrated that the left IFG (BA 45) is activated during both the encoding and the integration of propositional premises, and this region also has been suggested to be involved in the extraction and representation of the superficial structure of a problem (Reverberi et al., , ). About insula, recent evidence from network analysis suggests a critical role for the insula in high‐level cognitive control and attentional processes, and more researches unveiled that the insula is unique in that it is situated at the interface of the cognitive, homeostatic, and affective systems of the human brain, providing a link between stimulus‐driven processing and brain regions involved in monitoring (Luca et al., ; Menon & Uddin, ). In addition, the neuroimaging results converge with the general executive, learning, and conflict resolution processes in anterior cingulate cortices (BA 32) (Botvinick et al., , ). Beyond that, the dorsal anterior cingulate and insular cortices represent key neural hubs of the cingulo‐opercular network, which is implicated in maintaining and implementing task set (Dosenbach et al., , ). Compare to previous findings, our present meta‐analysis revealed that the IFG, insula, and cingulate may represent potential core site to collaborate in deductive reasoning. However, several studies have shown that the left IFG (BA44), which largely overlaps with the traditional Broca's area, is a key area in representing the formal structure of a logical problem during deductive reasoning (Reverberi et al., ; Reverberi et al., ). Here, our IFG cluster was positioned on is adjacent to the triangle (BA45) and closer to insula rather than BA 44 corresponding to Broca's Area. Besides, as mentioned above, the functions of the insula and cingulate are related to high‐level cognitive control, execution, and monitoring, which suggested that in the process of deductive reasoning, the coding and integration of premise information is indispensable, and it is also crucial to the execution and monitoring of the cognitive processing of reasoning.
## CONCLUSION
Overall, our results suggest that deductive reasoning is supported by a distributed network of regions encompassing frontal/parietal cortices and subcortical structures (e.g., caudate), and the results of our conjunction analysis highlights the IFG, insula, and cingulate (the key neural hubs of the cingulo‐opercular network) as core locus of deductive reasoning. These results conflict with the supposition that logic is subserved by a set of language‐independent regions within frontopolar (BA10) and fronto‐medial (BA8) cortices related to the putative core of deductive reasoning (Monti et al., , ). Moreover, a recent study revealed that the transient inhibition of Broca's area disrupted linguistic processing without affecting thought processing (Coetzee & Monti, ). The discrepancies between these results support the view that deductive reasoning requires a succession of stages that progressively transform premises into a conclusion or a series of conclusions; these stages include the decoding of the initial linguistic information, the conversion and correction of rules, and the transformation of inferential results into conclusive outputs (Coetzee & Monti, ; Monti et al., ). However, our study did not have valid classification criteria to sufficiently separate these different stages of deductive reasoning. Hence, future research can use high‐time resolution EEG and NMR techniques to synchronously record the brain signals of participants or inhibit the activity of certain brain regions to further elucidate the activation characteristics of brain regions at different stages of deductive reasoning and the precise relationship between deductive reasoning and linguistic representations. Besides, researchers need to increase the diversity of reasoning materials. in addition to abstract materials, we should also increase concrete materials that exist in real life to help us better understand reasoning.
## CONFLICT OF INTEREST
The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article.
## AUTHOR CONTRIBUTIONS
L. Wang and M. Zhang developed the study concept. Study Selection and Meta‐analysis algorithm were performed by all authors. L. Wang and M. Zhang drafted and revised the manuscript. All authors approved the final version of the manuscript for submission.
### Peer Review
The peer review history for this article is available at
|
## Background
Depression is a chronic mental health condition that affects millions of individuals worldwide. It is well-established that psychological stress plays an integral role in depression and that depression has numerous negative health outcomes. However, a closer look at components of stress vulnerabilities and depression is required to allow for the development and testing of appropriate interventions.
## Aims and Discussion
This article describes a conceptual framework about the complex and bidirectional relationship between stress vulnerability, depression, and health outcomes in women. The authors elucidate how the framework can be applied in clinical research about cellular aging and on the mechanisms of complementary and alternative medicine (CAM) for depression, using yoga as an example of a CAM modality.
## Conclusion
The proposed conceptual framework may be helpful for adding depth to the body of knowledge about the use of mind-body therapies for individuals at high risk of stress vulnerability and/or depression.
## Introduction
Depression is a leading cause of disability and disease burden worldwide and in the United States, affecting millions of individuals worldwide, particularly women. Women have a high risk of experiencing depression with an estimated lifetime risk of 10–25% (Kessler et al. ; Shenal et al. ). This increased vulnerability to depression starts in puberty and continues through menopause (Nolen-Hoeksema ; Deecher et al. ). Depression is of public health concern because of the short and long-term detrimental effects to the woman and her family. Individuals with depression experience high rates of anxiety, suicidality, substance use, and poor spouse/child relations (Kessler et al. ; Sunderland et al. ; Zbozinek et al. ); depression is also highly related to prevalent health outcomes such as cardiovascular disease, one of the five major causes of death in the United States (Minino and Murphy ; Elderon and Whooley ). The core symptoms of a major depressive episode are: a persistent depressed mood, difficulty concentrating or decision making, decreased energy, loss of interest in previously pleasurable activities, weight changes, changes in sleep (insomnia or hypersomnia), psychomotor changes (agitation or retardation), a pessimistic outlook with or without suicidal ideation (American Psychiatric Association ).
Appropriate treatment of depression is essential, yet many depressed women find the usual depression care (e.g. antidepressant medications and/or psychotherapy) to be inappropriate due to various concerns about cost, side effects, or inadequate relief of symptoms (Schreiber and Hartrick ; Lafrance and Stoppard ; Romans et al. ). Many individuals with depression experience persistent depressive symptoms despite the usual depression care (Zajecka et al. ), prompting them to seek additional relief through adjunctive or complementary therapies (Jorm et al. , ). For example, mind-body therapies, such as yoga, have received attention in both the lay and research literature as possible adjunctive therapies for depression (Bussing et al. ; Cramer et al. ). In order to thoroughly examine these and other similar interventions, research should be guided by a conceptual framework which incorporates and evaluates the relationship of individual, social/environmental, biological, and psychobehavioral factors involved in depression and the impact of the intervention on these factors. Although research is promising about nonpharmacologic complementary therapies such as yoga for depression, herein we suggest that use of the proposed conceptual framework may be helpful for adding depth to the body of knowledge about the use of adjunctive therapies for women with depression.
## A Conceptual Framework of the Relationship between Stress Vulnerability and Depression in Women
Depression and stress have a bidirectional relationship whereby depression may be both a cause and an effect of psychological stress (Kinser et al. ). Typically, the brain moderates the effects of stressors to maintain optimal functioning. Microprocesses regulate neurotransmission, endocrine, and immune functioning centrally, and sympathetic and parasympathetic activity in the periphery, all of which maintain allostasis or psychological and physical balance (McEwen and Lasley ; Peters and McEwen ). In the short term, these regulatory functions enhance the individual's response to stressors and the ability to manage negative physiological effects (Epel ). However, when stressors continue unabated, these same processes begin to impair neuronal function and other regulatory systems (Logan and Barksdale ; Kinser et al. ). The cumulative wear and tear associated with these physiological efforts to manage chronic stressors can cause depression and additional comorbidities. Without the availability and use of biopsychosocial resources, long-term exposure to the chronic stress of depression and/or repeated episodic life stressors can overload one's coping capacity; this may place an individual in a continuous cycle of stress response with negative affect states which can decrease quality of life and increase morbidity and mortality (McEwen , ; McEwen and Lasley ; Luyten et al. ; Clark et al. ; Taylor et al. ). It has been suggested that high levels of stress and depression are associated with accelerated cellular aging, a potential biomarker of the overloaded coping capacity of an individual (Kinser and Lyon ).
## Stress Vulnerability: Individual Chronic/Acute Burdens, Biological and Psychosocial Environment
The cyclic interplay of stress vulnerability, depression, and health outcomes is represented in the conceptual framework shown in Figure . The framework represents how stress vulnerabilities play an important role in depression and health outcomes in women (Kinser and Lyon ). Stress vulnerabilities are based upon numerous factors associated with acute and chronic stress, including individual chronic/acute burdens , the biological environment , and the psychosocial environment . The variety of complex and potential stressors in an individual's life may interact and contribute to increased risk of depression. In addition, the experience of depression may heighten an individual's tendency toward experiencing stressful episodes. Persistent and profound stressors may prevent regulatory mechanisms from adjusting appropriately, continuing the cycle of neurobiological dysregulations, poor health outcomes, and potentially advanced cellular aging (Nolen-Hoeksema ; Kinser et al. ; Kinser and Lyon ).
Conceptual framework of individual stress vulnerability, depression, and health outcomes in women. SES= socioeconomic status; ECD= early childhood development; HPAA= hypothalamus-pituitary-adrenal axis; TA= telomerase activity; TL= telomere length; MDD= major depressive disorder; MDEs= major depressive episode; QoL= quality of life.
Individual chronic and acute burdens involve an accumulation of life stressors that may include current or past stressful life events and current or past illnesses (e.g., chronic or acute psychological or physical illnesses). Acute stressful life events and current or past illnesses may precipitate or exacerbate depressive symptoms. For example, rodent studies and preliminary human studies suggest that the quality of the early childhood environment can shape brain development with associated changes to neuroanatomical structure/function and receptor levels/gene expression (Curley et al. ); theoretically, any of these changes may either be adaptive (and lead to adaptive behaviors and decreased risk of depression) or disruptive (and lead to unhealthy behaviors and a high risk for depression) (Garner et al. ). Stressful early childhood experiences can significantly undermine the development of adaptive coping skills required to deal with challenges in adulthood and may also create the foundation for unhealthy lifestyles, negative interpersonal relationship patterns, and poor health outcomes (Garner et al. ; Danese and McEwen ; Shonkoff et al. ). As another example, studies suggest that women with a history of childhood traumas, such as sexual abuse, and low levels of current social support are at higher risk of unintended pregnancies, which are associated with prenatal and postpartum depression (Mercier et al. ; Nelson and Lepore ). To continue this example, a woman with an unintended pregnancy may find herself unprepared to serve in a social role in which she is expected to put others' needs ahead of her own, which can be acutely and chronically stressful and is highly related to both prenatal and postpartum depression (Mercier et al. ).
Vulnerabilities regarding the psychosocial environment that play a role in the impact of depression on health outcomes may be demographics/socioeconomic status, perceived social support, lifestyle, and interpersonal situations. Socioeconomic status is clearly linked to stress vulnerability, as seen in low-income populations which have high levels of stress, impaired coping, and depression. Persistent socioeconomic inequalities are linked with stress vulnerability, particularly with regards to educational and financial opportunities; these inequalities are also linked with health disparities and unhealthy lifestyle choices and poor health outcomes (Shonkoff et al. ; Danese and McEwen ). Negative interpersonal situations, such as intimate partner violence (IPV) and low perceived social support, add additional vulnerability for stress and depression in women (Devries et al. ); fortunately, long-term sequelae of traumatic experiences such as IPV and early childhood abuse may be attenuated by higher perceived social support (Kaufman et al. ; Chuang et al. ). Nonviolent gender-specific interpersonal situations may also place additional wear and tear on women's mental and physical health, particularly if there are feelings of guilt or selfishness associated with participating in healthy self-care activities rather than focusing on the financial or other needs of the family (Hauenstein and Boyd ; Lafrance and Stoppard ; Hauenstein and Peddada ; Hauenstein et al. ; Petterson et al. ). There also appear to be biological vulnerabilities that affect women, in particular, such as temperament, inherited and epigenetic changes and dysregulated stress response systems (Nolen-Hoeksema and Girgus ; Nolen-Hoeksema et al. ; Davidson et al. ; Hammen ; Nolen-Hoeksema ; Halbreich and Kahn ; Deecher et al. ; Gotlib and Hammen ; Taylor et al. ; Young and Korszun ). Epigenetic research suggests that social experiences in both childhood and adulthood may significantly modulate stress reactivity and depression via a multitude of mechanisms (for a more thorough review, see (Curley et al. ).
## Depression: Psychobehavioral and Physiological Factors
Psychobehavioral factors play an important role in the development and maintenance of depressive states. A woman's perception of stress and her response to that appraisal are important moderators in the relationship between stress and depression. With regard to stress perception, a key individual factor related to the effect of stress is the degree to which an individual perceives that stress to be significant and to what degree the individual thinks she, in this case, has control over the situation (Lazarus and Folkman ). Important factors of stress perception are related to individual differences in personality and cognitive styles in the face of stressful life situations; these may either increase or decrease an individual's risk for and experience of depressive illness. For example, two key aspects of these individual styles are relevant to the bidirectional relationship between stress vulnerability and depression. First, every individual has their own sense of control in the face of stress and depression. Extensive cross sectional and longitudinal research has provided evidence to suggest that individuals with stressful life situations, because of environmental factors outside of the individuals' control, have higher stress-related psychoneuroimmunologic changes (Hauenstein ; Geronimus et al. ; Glover et al. ; Kahn and Pearlin ; Clark et al. ; Johansson et al. ). Second, every individual has more or less tendency toward ruminations, or persistent repetitive negative thinking. Ruminative patterns may be normative because, evolutionarily speaking, humans must pay close attention to stressors or interpersonal distress to maintain safety and social relations (Buss ; Seligman et al. ). However, in modern society, those with heightened attention to and perception of stressors may, in fact, have a biased perception toward negative emotions; for example, increased levels of ruminations on stress may affect levels of depression (Seligman et al. ). When experiencing depression, women quite often report negative ruminations, which are in-turn related to low self-esteem, hypersomnia, and anxiety (Kendler et al. ; Keita ; Marcus et al. ; Rochlen et al. ). Ruminations may be particularly problematic for women because they can increase the stress of depression by inducing negative thoughts about the past, present, and/or future. Ruminations are associated with lower levels of social support and increased suicidal ideations, all of which continue the cycle of stress and prolonged depression (Nolen-Hoeksema et al. ).
Another key psychobehavioral factor in depression is whether an individual has personal and environmental resources for appropriate stress management. The ability to respond to stress and depression in a healthy manner is highly dependent on the availability and use of biopsychosocial resources. To maintain health, an individual must be aware and capable of/interested in using relevant resources. In particular, the use of healthy biopsychosocial resources is a protective mechanism essential for the capacity of women to deal with stressors; in essence, the ability to rapidly reach a sense of equilibrium or return to a calm baseline can be highly protective (Danner et al. ). Research suggests that those with depression and anxiety disorders tend to pick ineffective health maintenance strategies, whereas happier individuals have tendencies to reach out for social support and healthy biopsychosocial strategies and are more willing to use positive appraisal techniques about life stressors (Diener et al. ). Despite these tendencies, however, interventions may be effective for assisting individuals in their use of healthy, positive biopsychosocial resources. Biopsychosocial resources may include a wide range of activities and behaviors. For example, positive health behaviors involving healthy nutrition, exercise, relaxation, and healthy sleep patterns may help women to effectively respond to acute or chronic stressors in their lives (Romans et al. ; Hauenstein ; Chang et al. ; Dusek et al. ; Institute of Medicine (US) Committee on Sleep Medicine and Research et al. ; Ruiz-Nunez et al. ; Clark et al. ; Sims et al. ; Dunn et al. ). Mental wellness may be promoted by psychological therapies that emphasize positive emotions, gratitude, personal strengths, and engagement with life (Seligman et al. ; Emmons ). Furthermore, a significant factor in stress moderation may be the participation in intentional activity, or discrete actions that require behavioral and/or cognitive and/or volitional effort (Lyubomirsky et al. ). Intentional activities, particularly those that focus on personal strengths, positive emotions, and mindfulness, may stimulate a healthy stress appraisal and may greatly assist depressed individuals to decrease ruminations that often impact mood (Fredrickson ); for example, yoga may be a reasonable intentional activity that meets these needs (Kinser et al. ). Finally, an accumulation of experiences through the use of healthy biopsychosocial resources that meet the human needs of competence and social relatedness may allow for mental wellness and decrease the risk of accelerated cellular aging and poor health outcomes (Lyubomirsky et al. ).
## Physiological Factors
Extant research suggests that depression is associated with, causes, and/or may be caused by a number of biological perturbations. Support for the interrelationships among biological pathways and depression has been documented via multiple different pathways. Over the past several decades, multiple categories of biological alterations have been associated with depression, including heightened inflammatory activation (Miller et al. ), hyperactivity of the hypothalamic-pituitary-adrenal axis (Pariante and Lightman ), and more recently, genetic and epigenetic alterations (Massart et al. ). Currently, yet another plausible biological link is being explored: the link between the microbiome and depression, focusing on the activation of the central nervous system (CNS) signaling systems by gastrointestinal bacteria (Cryan and Dinan ). The accelerated aging hypothesis combines several biological theories to create a theoretically based model of interactions among multiple biological events. Although these mechanisms have been well supported in animal models, human trials are still underway in establishing these mechanisms.
## Health Outcomes
The ability of an individual to respond to the chronic stress associated with depression and to individual life stressors varies greatly depending upon the availability and use of healthy biopsychosocial resources (Hauenstein ; Kiecolt-Glaser et al. ; McCain et al. ; Uebelacker et al. ). Poor health outcomes may occur if these resources are not available or used; psychological and physical health may decline and lead to decreased psychosocial functioning, decreased health-related quality of life, and increased incidence of comorbid conditions (McEwen and Lasley ; Luyten et al. ; Clark et al. ; McEwen ; Taylor et al. ). There is accumulating research on the relationship of depression with morphologic changes including decline in gray matter density of the hippocampus, anterior cingulum, left amygdala, and right dorsomedial prefrontal cortex (Frodl et al. ) and hippocampal volume loss (McEwen ). In addition to brain-specific alterations, there is accumulating research on the adverse effects of depression on comorbid conditions. The adverse health outcomes of untreated major depression are decrements in health that surpass the effects of chronic diseases angina, arthritis, asthma, and diabetes (Moussavi et al. ). Individuals with depression have higher rates of obesity, cardiac conditions including hypertension, heart disease, and diabetes than the general population (Katon , ; Clarke and Currie ; Pozuelo et al. ).
## Using the Conceptual Framework: Potential for Complementary and Alternative Medicine (CAM) Research
This framework may be useful when considering research studies on CAM for individuals with depression. The intent of the remainder of this article is to discuss yoga as an example of a CAM modality and how this intervention could influence various aspects of the conceptual framework of the complex relationship of stress and depression in women. In order to reduce risk of poor health outcomes, complementary therapies such as yoga may provide a stress-buffering effect to enhance health and relieve effects of chronic stress and depression (Loizzo ). However, stressful experiences and negative ruminations may contribute to the cycle of stress and depression, yoga may allow for a return to balance of the multiple components involved in mental wellness. Yoga may be an effective biopsychosocial intervention for dealing with the cycle of stress and depression because yoga involves components designed to have an effect on key aspects of depression in women, from physical activity to meditative, relaxing practices to rhythmic, soothing breathing practices to social interactions (Dunn et al. , ; Weintraub ; Netz et al. ; Larun et al. ; Marcus et al. ; Tsang et al. ; Saeed et al. ; Uebelacker et al. , ). Yoga philosophy emphasizes the use of yoga for mental and physical wellness personalized to the individual's needs (Kinser and Williams ), and this may appeal to women who are uncomfortable with interventions based solely on the biomedical model that focuses primarily on neurochemistry. Studies have shown that women often attribute their experiences of and recovery from depression to life experiences and social factors rather than to biochemical pathology or medications (Schreiber and Hartrick ; Lafrance and Stoppard ).
The proposed conceptual framework is a natural complement to the biomedical model that is commonly used in western societies because it maintains an equal focus on biology and the particular needs/experiences of women. Furthermore, many researchers suggest that women with depression seek out complementary therapies because the usual allopathic pharmacologic care does not adequately address their individual symptoms or explanatory model of depression or may have unappealing side effects and expense (Hammen , ; Schreiber and Hartrick ; Kessler et al. ; Pirraglia et al. ; Saper et al. ; Kirkwood et al. ; Lafrance and Stoppard ; Gotlib and Hammen ; Uebelacker et al. ). Therefore, a need exists to investigate the effects of complementary interventions in women who have depression using this integrative framework that acknowledges the complex relationship between depression and stress in women.
The conceptual framework is relevant for research studies testing the use of complementary therapies for women with depression because it provides moderating and outcome variables for measurement. In particular, the framework suggests a number of potential moderating variables: stress perception (individual life stressors, ruminations) and responses (use of healthy biopsychosocial resources). The framework provides a number of outcome variables for measurement consistent with the literature on complementary therapies for depression, stress, ruminations, and anxiety. This framework is also suggested for use in future research on yoga because it provides an obvious target for intervention: an individual's response to stress and depression through the availability and use of biopsychosocial resources. Interventions, like yoga, for treating depression are needed that address gender-specific issues and symptoms, and empower women to participate in positive health promoting self-care activities (Schreiber and Hartrick ; Hammen ; Lafrance and Stoppard ; Nolen-Hoeksema ; Nolen-Hoeksema et al. ; Gotlib and Hammen ; Nolen-Hoeksema and Hilt ). However, women with depression may have difficulty sustaining behaviors for personal health and wellness, which highlights the importance of conducting research based upon the proposed conceptual framework (Lafrance and Stoppard ; Gotlib and Hammen ).
It has been theorized that the practice of yoga as a healthy biopsychosocial resource may assist individuals with depression to cope with stress and thus enhance their mood. Studies suggest that individuals participating in various yoga interventions report decreases in psychological and physical symptoms of depression and stress (Uebelacker et al. ). Cramer and colleagues (2013) have recently conducted a meta-analysis of studies on yoga for depression, finding that yoga may be effective for short-term remission of depression (Cramer et al. ). Numerous putative mechanisms have been suggested to explain the beneficial effects of yoga, including those which enhance global regulation of stress response systems (Kinser et al. ). For example, Streeter and colleagues (2012) suggest that yoga corrects underactivity of the inhibitory neurotransmitter gamma amino-butyric acid [GABA] with resultant decreases in depression symptoms (Streeter et al. , ). As another example, emerging evidence suggests that depression is related to alterations in biological markers, such as inflammatory cytokines and related DNA methylation patterns, which might influence mental health outcomes.(Uddin et al. , ; Akbarian and Nestler ; Penninx et al. ) It appears that the impact of depression-related physiologic changes are potentially reversible with interventions such as yoga (Schmidt et al. ; Yehuda et al. ). For example, remarkable findings from a recent study suggest that the therapeutic potential of interventions involving mindfulness may stem partially from the epigenetic control of inflammatory processes (Kaliman et al. ). Yogic practices that enhance stress regulation by inducing relaxation appear also to induce epigenetic changes in the expression of proteins involved in energy metabolism, mitochondrial function, insulin secretion, telomere maintenance, and inflammation (Bhasin et al. ). Despite these interesting findings, a complication in the development of evidence-based recommendations about yoga for depression in women is that many yoga research studies have methodological limitations and do not integrate an evaluation of individual, social/environmental, and physiological factors in their evaluation of outcomes. Using the proposed conceptual framework that is not only specific to the population of interest (women with depression) but also includes relevant variables for measurement may allow researchers to add substantively to the body of knowledge about yoga as an effective complementary therapy.
## Conclusion
The model in Figure elucidates the bidirectional relationship of stress vulnerabilities, depression, and health outcomes and provides a conceptual framework for the conduct of research about CAM modalities and depression in women. This framework is relevant and timely because it integrates multiple models and theories of the etiology of depression within a context of cellular aging. We suggest that, by providing moderating and outcome variables for measurement, this framework may be helpful for researchers interested in testing the use of complementary therapies, such as yoga, for women with depression.
|
## Introduction
Previous studies have shown that gestational inflammation can accelerate age‐associated cognitive decline (AACD) in maternal mice; enriched environments (EEs) have been reported to protect normally aging mice from AACD and improve mitochondrial function. However, it is unclear whether the nitrosative stress‐related proteins tet methylcytosine dioxygenase 1 (TET1) and S‐nitrosoglutathione reductase (GSNOR) are involved in the accelerated aging process of gestational inflammation and whether EEs can slow this process.
## Methods
In this study, CD‐1 female mice on the 15th day of pregnancy were injected with bacterial lipopolysaccharide (50 μg/kg; LPS group) or an equivalent amount of normal saline (CON group) from the abdominal cavity for 4 consecutive days. Twenty‐one days after delivery, half of the LPS‐treated mice were randomly selected for EE until the end of the behavioral experiment (LPS‐E group). When the female rats were raised to 6 months and 18 months of age, the Morris water maze (MWM) was used to detect spatial learning and memory ability; RT‐PCR and Western blots were used to measure the mRNA and protein levels of hippocampal TET1 and GSNOR.
## Results
As for the control group, compared with 6‐month‐old mice, the spatial learning and memory ability of 18‐month‐old mice decreased, and the hippocampal TET1 and GSNOR mRNA and protein levels were decreased. Gestational inflammation exacerbated these age‐related changes, but an EE alleviated the effects. Pearson's correlation analysis indicated that performance during the learning and memory periods in the MWM correlated with the levels of hippocampal TET1 and GSNOR.
## Conclusions
Our findings suggest that gestational inflammation accelerates age‐related learning and memory impairments and that postpartum EE exposure could alleviate these changes. These effects may be related to hippocampal TET1 and GSNOR expression.
Protein cysteine residues can be oxidized by nitric oxide (NO) to form protein S‐nitrosothiols (SNOs) in a process called S‐nitrosylation. S‐nitrosoglutathione reductase (GSNOR) is a highly specific denitrase for GSNO; it controls the intracellular levels of GSNO and protein SNOs and protects the body from nitrosative stress.
## INTRODUCTION
Age‐associated cognitive decline (AACD) is a common symptom of aging (Wadhwa et al., ). Brain aging is a complex multi‐factor process. Negative life events such as material deprivation, bacterial infections, and low education levels are key risk factors, (Weiner et al., ), while positive life events have a protective effect on brain aging. Among them, normal reproductive experience or maternal experience can alleviate AACD (Li et al., ). However, due to significant changes in the endocrine and immune systems of the body, pregnancy also increases the sensitivity to adverse factors and especially increases the risk of infection (Oskvig et al., ). Lipopolysaccharide (LPS) is a powerful endotoxin that is commonly used to simulate bacterial infection in experimental animals. Our previous studies indicated that maternal exposure to LPS during pregnancy accelerated AACD in the middle‐aged mother (Li, Wang et al., 2016; Sun et al., ). However, how gestational inflammation affects AACD in mothers in young and old age has not been elucidated.
The hippocampus plays an important role in normal learning and memory consolidation and AACD, because the structural and functional changes induced by aging in occur earlier than other brain regions (Mendelsohn Larrick, ; Belblidia et al., ). Studies have shown that prenatal stress in the rat causes long‐term spatial memory deficits and hippocampal abnormalities as detected by magnetic resonance imaging (Stamler et al., ). We previously also found that embryonic exposure to LPS‐induced inflammation led to age‐related spatial learning and memory impairment and corresponding hippocampal neurobiochemical changes (Chen et al., ; Li, Wang et al., 2016; Wu et al., ).
Cognitive function depends on neurotransmitter systems such as nitric oxide (NO) signaling, which exerts effects on cognition by means of S‐nitrosylation. Nitrosation reactions can change protein activity, localization, stability, and protein–protein interactions as well as modulate signaling and stress responses (Khan et al., ; Zhang et al., ). At the physiological level, NO generally facilitates neuronal differentiation, development, and survival. However, excessive NO can result in aberrant S‐nitrosylation of proteins and increase the incidence of senescence‐related diseases (Nakamura et al., ). Accumulating evidence indicates that mitochondrial dysfunction can contribute to the pathophysiology of age‐related diseases. Mitochondria are critical for neuronal energy metabolism, calcium regulation, myelination, and neuronal and glial apoptosis (Princz et al., ; Zhuang et al., ). Therefore, mitochondrial decay may be central to neuronal damage and cognitive decline (Cheng et al., ; Lai et al., ). Mitochondria have a large number of protein cysteine residues. These residues can be oxidized by nitric oxide (NO) to form S‐nitrosothiols (SNOs), named S‐nitrosylation. Excessive S‐nitrosylation of related proteins will directly affect mitochondrial dynamic activity, reduce the efficiency of aerobic metabolism, inhibit mitochondrial autophagy, and destroy mitochondrial homeostasis (Rizza & Cardaci et al., ). We previously found correlations between mitochondrial quality control and AACD (Zhuang et al., ). S‐nitrosoglutathione reductase (GSNOR) is a key enzyme that catalyzes denitrosylation to balance S‐nitrosylation and protects the body from nitrosative stress (Beigi et al., ; Liu et al., ). Studies have shown that whole‐brain GSNOR expression declines with age and that its deficiency can impair mitochondrial function due to nitrosative stress (Rizza & Cardaci et al., ).
Gsnor expression is epigenetically controlled by cytosine methylation of CpG islands located in the promoter region (Rizza & Cardaci et al., ). The demethylating activity of tet methylcytosine dioxygenase 1 (TET1), a member of the 2‐oxoglutarate‐dependent dioxygenases, is required to counteract gene silencing due to cytosine methylation, and, in turn, sustain Gsnor transcription(Rizza & Filomeni, ) . Therefore, TET1 is closely related to cell senescence by its regulation of Gsnor expression. However, whether TET1 and GSNOR are involved in the accelerated AACD of gestational inflammation remains unknown.
Enriched environment (EE) usually refers to a larger group living in a larger living space, with toys that change frequently, such as running wheels, experimental mouse tunnels, poplar toys, rings, and so on, and is a classic model used in neuroscience (Nithianantharajah & Hannan, ). Increasing evidence suggests that EEs can promote hippocampal neurogenesis in mice and induce mitochondrial activity, and these results are closely related to the improvement of impaired cognitive function (Zhuang et al., ). Studies have indicated that an EE can alleviate the age‐related cognitive impairment and deficits in learning and memory capabilities that result from gestational exposure to LPS (Sun et al., ).
We hypothesize that TET1 and GSNOR are involved in the accelerated aging process of LPS exposure in late pregnancy and the positive effect of EEs. In this study, we first identified whether gestational inflammation affected learning and memory in young (6‐month‐old) and aged (18‐month‐old) CD‐1 mice. Subsequently, we evaluated whether the protein and mRNA levels of TET1 and GSNOR were altered in the hippocampus of CD‐1 mice based on age and treatment. Finally, we determined the correlations between learning and memory abilities and the measured neurobiological indicators in the different age and treatment groups.
## MATERIALS AND METHODS
### Animals and treatments
CD‐1 mice, 2 months old, 40 male mice, and 80 female mice, were purchased from the Model Animal Research Institute of Nanjing University. They were reared under standard experimental conditions with a 12‐hour light‐dark cycle, suitable temperature and humidity, and free eating and drinking. The model and breeding were based on our previous studies (Li, Cao et al., 2016; Wu et al., ). CD‐1 mice were caged at 1:2 (male: female) at 9:00 p.m. every night until the vaginal embolus was found and immediately separated. If there is no pregnancy in more than two weeks, the sample should be replaced. On the 15th day of pregnancy (28–34 g weight) were injected with bacterial lipopolysaccharide (50 μg/kg; LPS group) or an equivalent amount of normal saline (CON group) from the abdominal cavity for four consecutive days. On postnatal day 21, the mothers were separated from their offspring (this avoided the premature separation of mother and child, which would cause psychological stress to the mother and offspring) and breastfeeding was stopped. After the mother and child were separated, half of the mother mice that received LPS intraperitoneal injections were randomly selected for EE until the end of the behavioral experiment; these mice constituted the LPS‐E group. The mice of the CON group and LPS group were housed in standard plastic mouse cages (25.5 × 15 × 14 cm , 4 mice per cage), and the mice of the LPS‐E group were housed in enlarged cages (52 × 40 × 20 cm , 10−15 mice per cage). 6‐month‐old (6 M; young) and 18‐month‐old (18 M; aged) CD‐1 mice were used to complete the tests described in the sections below (10 per group in Morris Water Maze, six per group in the protein and mRNA markers measured). Mice with obvious defects were eliminated before the experiment. All procedures were approved by the Experimental Animal Ethics Committee of Anhui Medical University. The experimental process is shown in Figure .
Timeline of experimental events. The pregnant mice were injected with LPS or saline from the abdominal cavity for four consecutive days starting from the 15th day of pregnancy. Weaned on the 21st day after childbirth, the mothers were divided into three groups based on whether the mice were exposed to an EE until the end of the experiment. The MWM test was performed at 6 and 18 months of age. Two weeks after the completion of the MWM test, the mice were sacrificed for follow‐up biochemical experiments. EE, enriched environment; CON, control group; LPS, lipopolysaccharide treatment group; LPS‐E, lipopolysaccharide plus enriched environment treatment group; MWM, Morris water maze
### Enriched environment
Right after breastfeeding was stopped, the LPS‐E mice were housed in a larger cage (52×40×20 cm ) to enrich social life in the group. Toys and novelty objects such as running wheels, experimental mouse mazes, swings, and ladders, were placed in the cage and constantly changed until the end of the behavioral experiment. This was done to expand the living space, provide places for exercise and escape as well as to provide stimulation to novel things, and enhance typical species behaviors of mice such as playing, fighting, and group sleep.
### Morris water maze
Put a movable cylindrical escape platform with a diameter of 10 cm and a height of 24 cm into a black circular pool with a diameter of 150 cm and a height of 30 cm. Fill the pool with tap water at a suitable temperature and control the water level to be 1 cm higher than the platform. Surround the pool with a white curtain 50 cm away from the pool to shield the interference of external factors from the experiment, and paste 3 black markers of different shapes at equal distances on the curtain 1.5 m from the bottom of the pool as a space positioning clues for mice. The experiment process is divided into a learning (positioning and navigation) period and a memory (spatial exploration) period. The behavioral parameters (distance swam in the learning period, percent distance swam spent swimming in the memory period) were recorded and analyzed by behavioral analysis software (ANY‐Maze, USA). Learning period: During seven consecutive days of positioning navigation, the escape platform was 1 cm below the water surface and fixed in the center of the target quadrant. Only before the start of the experiment on the first day, the mice were placed on the platform for 30 s to adapt, and the mice facing the outer edge of the pool were randomly thrown into the water from any quadrant every day and were allowed to swim for a maximum of 60 s to find the platform. Each experimental mouse was trained four times a day, 15 min apart. Record the average swimming distance before finding the platform each time. The shorter the distance, the stronger the learning ability. Memory period: On the 7th day of the learning period, after completing four training sessions, the escape platform was taken out. After the mice recovered their strength, they entered the water again from the opposite quadrant of the platform quadrant (target quadrant) and swam arbitrarily in the pool for 60 s. The greater the percentage of the swimming distance in the target quadrant within 60 s of the total swimming distance, the stronger the memory ability.
### Tissue preparation
In order to avoid the effects of the MWM test in the expression of Tet1 and Gsnor mRNA and protein in mouse hippocampus. We chose two weeks as a recovery period to reflect the more realistic results of the molecular experiments. Two weeks after the completion of the behavioral task, the mice were euthanized by cervical dislocation, but some mice whose health status is seriously affected in the MWM test could be excluded. The brain was removed by craniotomy and the hippocampus was then extracted and stored in an ultra‐low temperature refrigerator at −80°C. The right hippocampus was used for Western blotting (WB) and the left hippocampus was used for reverse transcription‐polymerase chain reaction (RT‐PCR).
### Quantitative real‐time RT‐PCR (RT‐QPCR)
The hippocampus was mixed with TRIzol reagent following the manufacturer's instructions to obtain RNA. cDNA was extracted from RNA (1 μg) using The RevertAidTM First‐Strand cDNA Synthesis Kit. Took out cDNA as a template for quantitative real‐time PCR and performed amplification in a 10‐μL reaction mixture containing 5 uL of 2×SYBR Green mixture, 1 uL of each primer (10 uM), 1 uL of cDNA template, and 2 uL of RNase‐free water. The quantitative real‐time PCR reaction condition included one cycle of 95°Cfor 1 min and 40 cycles of 95°C for 20 s and 60°C for 1 min. The mRNA level was quantified using the 2 method. Beta‐actin served as the internal reference. The primer sequences are listed in Table .
Sequences of the primers used for quantitative real‐time PCR
### Western blotting
Tissue was lysed in protein‐neutral lysate, extract protein, and protein concentrations were determined using the bicinchoninic acid assay kit. After dissolving the protein to an equal concentration, add 5×SDS‐PAGE protein loading buffer according to 1:4 and heat in a boiling water bath for 10 min. After cooling at room temperature, SDS‐PAGE was injected into the sample hole according to 5∼10 μl in each hole. The parameters of the electrophoresis instrument were set to 80 V, and 30 min was used to concentrate the gel; then set to 120 V, and 1 h was used to separate the gel. The membrane was transferred to a constant current, rinsed for 5 min, and blocked with 5% skimmed milk powder at room temperature for 2 h. Added primary antibodies include GSNOR (1:800, Proteintech Cat# 11051‐1‐AP, RRID:AB_593422), TET1 (1:2000, Bioworld, BZ12161), and β‐actin (1:1000, Zs‐BIO, TA‐09). After incubation overnight at 4°C, wash with PBST. Blots were subsequently incubated with horseradish peroxidase‐conjugated secondary antibody (goat anti‐rabbit IgG: 1:5000, Zs‐BIO, ZB‐2301) for 2 h at room temperature, and then washed with PBST. The target protein bands of immune response were visualized using the ECL Kit (Thermo, USA), then use Image J software (Media Cybernetics, USA) to analyze the results of protein bands, and calculate the relative expression.
### Statistical analysis
SPSS 22.0 statistical software was used for data analysis, and x ± s was used to describe the continuity data of the normal distribution. Repeated‐measures analysis of variance was used to analyze the data from the MWM learning task, with day, group, or age as the independent variable. The memory percentage of the distance from the MWM test and the target mRNA or protein expression levels were evaluated by one‐way ANOVA with age or treatment as independent variables, and comparison between groups used the LSD test. The correlation between GSNOR content and behavioral data was analyzed by Pearson correlation analysis. p < 0.05 is considered statistically significant.
## RESULTS
### Performance in the Morris water maze
#### Learning phase
When all mice were combined, the rm‐ANOVA results show that as the number of training days increased, the swimming distance ( F = 1183.986, p < .01) gradually decreases; the average speed (F = 0.113, p = .485) decreases with time, but it is not significant. The 18 M mice swam for a significantly longer distance ( F = 39.252, p < .01) than the 6 M mice, the LPS mice swam for a significantly longer distance ( F = 33.147, p < .01) than the CON mice, the LPS‐E mice swam for a significantly shorter distance ( F = 33.147, p < .01) than the LPS mice. Moreover, the 18 M‐CON mice swam for a significantly longer distance ( F = 475.814, p < .01) and had a slower swimming velocity ( F = 803.290, p < .01) than the 6 M‐CON mice (Figure ). LPS and EE had no significant effect on the swimming speed of 6 and 18 M mice (Figure ). Therefore, swimming distance is more suitable as an indicator of learning ability. There were significant differences in the swimming distance among the treatment groups for both the 6 M mice ( F = 299.718, p < .01; Figure ) and the 18 M mice ( F = 190.557, p < .01; Figure ). In the two age groups, the swimming distance of the LPS group mice was significantly longer than that of the CON group mice ( ps < .01). In addition, mice in the LPS‐E group swam a similar distance as the 6 M LPS mice ( p = .159) and the 18 M CON mice ( p = .150). The effect of the interaction of group × day was not significant in the learning phase ( ps > .05).
Performance in the Morris water maze (MWM) test. Average velocity (a,c,e), and distance (b,d,f) during the learning phase; percent distance swam and percent time swam (g,j) during the memory phase. Age and treatment had a significant effect on learning and memory performance in the MWM. Data are expressed as the mean ± SEM ( n = 10 mice/group). Significant differences compared with 6‐ or 18‐month‐old CON mice ( p < .05, p < .01); Significant differences compared with 6‐ or 18‐month‐old LPS‐E mice ( p < .05, p < .01). 6, 6‐month‐old; 18, 18‐month‐old; CON, control group; LPS, lipopolysaccharide treatment group; LPS‐E, LPS plus enriched environment treatment group
#### Memory phase
When all mice were combined, the percentage of swimming distance ( t = 3.716, p < .01) and the percentage of swimming time ( t = 3.829, p < .01) in the target quadrant of 18 M mice was significantly lower than that of 6 M mice, There were significant differences in the percentage of swimming distance ( F = 19.955, p < .01) and the percentage of swimming time ( F = 15.880, p < .01) among different treatment groups. The percentage of swimming distance ( t = 8.173, p < .01) and the percentage of swimming time ( t = 6.425, p < .01) in the target quadrant of 18 M CON mice was significantly lower than that of 6 M CON mice. There were significant differences in the percentage of swimming distance and the percentage of swimming time between the treatment groups for both the 6‐month‐old mice ( F = 43.117, p < .01; F = 29.314, p < 0.01) and the 18‐month‐old mice ( F = 8.323, p < .01; F = 7.806, p < .05).In the two age groups, the swimming distance percentage of the LPS group mice was significantly lower than that of the CON group mice ( ps < .01). Post‐hoc analysis revealed that mice in the LPS‐E groups had similar percent distance swam and percent time swam as the 6‐month‐old mice in the LPS group ( p = .432; p = .850) and the 18‐month‐old mice in the CON group ( p = .241; p = .386; Figure ).
### The mRNA levels of Tet1 and Gsnor in the hippocampus
When all mice were combined, the 18 M group had significantly lower mRNA levels of TET1 ( t = 10.205, p < .001) and GSNOR ( t = 9.550, p < .001) than the 6 M groups, there were significant differences in the mRNA levels of GSNOR ( F = 3.594, p < .05) among different treatment groups.
#### Age effects
The 18 M CON group had significantly lower mRNA levels of TET1 ( t = 4.306, p = .002) and GSNOR ( t = 23.237, p < .001) than the 6 M CON group.
#### Treatment effects
In 6 M mice, only GSNOR mRNA ( F = 36.421, p < .001) levels differed among the groups; the expression of LPS‐E and LPS group was significantly lower than that of the CON group ( ps < .001), and the expression of LPS‐e group was slightly higher than that of LPS group ( p = .053).In 18 M mice, the levels of both mRNAs ( F = 15.411, p < .001 for TET1; F = 14.465, p < .001 for GSNOR) differed significantly among the groups. Compared with the LPS‐E group and the CON group, the LPS treatment group had lower mRNA levels ( ps < .001) of the two genes; no difference was found between the first two groups ( ps >.05; Figure ).
Relative mRNA and protein levels of Tet1 and Gsnor in the hippocampus. (a) The protein bands after Western blotting. (b,c) Differences in Tet1 and Gsnor mRNA levels in the different groups. (d,e) Differences in TET1 and GSNOR levels in the different groups. Significant differences compared with 6‐month‐old CON mice ( p < .01); Significant differences compared with 18‐month‐old CON mice ( p < .01); Significant differences compared with 18‐month‐old LPS‐E mice ( p < .05, p < .01) ( n = 6 mice/group). CON, control group; LPS, lipopolysaccharide treatment group; LPS‐E, LPS plus enriched environment treatment group
### The protein levels of TET1 and GSNOR in the hippocampus
When all mice were combined, the 18 M group had significantly lower levels of TET1 ( t = 9.428, p < .001) and GSNOR ( t = 8.164, p < .001) than the 6 M groups, there were significant differences in the levels of TET1 ( F = 5.114, p < .05) and GSNOR ( F = 6.395, p < .05) among different treatment groups.
#### Age effects
The 18 M CON group had significantly lower protein levels of TET1 ( t = 4.306, p = .002) and GSNOR ( t = 23.237, p < .001) than the 6 M CON group.
#### Treatment effects
Irrespective of age, different treatments significantly affected the levels of hippocampal TET1 ( F = 322.940, p < .001 in 6 M mice; F = 24.474, p < .001 in 18 M mice) and GSNOR ( F = 392.495, p < .001 in 6 M; F = 26.327, p < .001 in 18 M mice). Post‐hoc analysis showed that the levels of TET1 and GSNOR in the 6 M LPS and LPS‐E groups were significantly lower than those in the same‐age control group ( ps < .001); there were marginally higher levels in the LPS‐E group than in the LPS group ( p = .153 for TET1; p = .098 for GSNOR). In the 18 M, the two protein levels in the LPS treatment group were lower ( p s < .001) than the LPS‐E group and CON group; there was no difference between the latter two groups ( ps < .01, Figure ).
### Correlations between performance in the MWM and the protein or mRNA levels of TET1 and GSNOR
The correlations between the hippocampal levels of TET1 and GSNOR and learning and memory in the MWM task were related to age and treatment, so each group was analyzed separately (Table ).
Correlations between performance in the MWM and TET1/GSNOR in the hippocampus
Irrespective of the group, during the learning period the mRNA and protein levels of TET1 and GSNOR in the hippocampus of the 6 M mice were not significantly correlated with the mean swimming distance in the MWM ( p s > .05). In contrast, the levels of mRNA and protein of TET1 and GSNOR in the hippocampus of the 18 M mice in these groups were negatively correlated with the mean swimming distance ( P s < .05).
During the memory period, there was no significant correlation between TET1 and GSNOR mRNA and protein levels in the hippocampus of the 6‐month‐old mice and the percentage of distance swam in the target quadrant ( p s > .05). In 18‐month‐old mice, hippocampal GSNOR protein in the control and LPS‐E groups and TET1/GSNOR mRNA in the LPS group were positively correlated with the percentage of distance in the target quadrant ( p s < .05).
## DISCUSSION
Pregnancy is a special and sensitive period, during which females are vulnerable to infections from bacteria and viruses (Li et al., ). LPS is a powerful endotoxin commonly used to simulate bacterial infections in laboratory animals, which can provide continuous inflammatory stimulation (Li et al., ). Pregnant animals are more sensitive to LPS, and LPS triggers maternal immune activation, leading to abnormal gene expression in key brain regions of the mother and fetus (Oskvig et al., ). Our previous studies suggested that inflammatory insults during pregnancy could be an important risk factor for the development of AACD (Li, Wang et al., ). The results of this experiment showed that exposure to LPS during pregnancy can accelerate cognitive impairment in elderly mice, and EE can alleviate this effect. Therefore, avoiding gestational infection is of great significance for protecting cognitive ability in old age, and long‐term environmental enrichment could mitigate the cognitive impairment induced by gestational inflammation. Moreover, we also found that the decreased expression of GNSOR and TET1 in the hippocampus was associated with impaired cognitive function in different treatment groups. Our results provide new insights into the mechanisms involved in the accelerated cognitive decline that results from gestational inflammation.
### EEs alleviate the accelerated spatial cognition impairment in aged females caused by gestational inflammation
In both humans and rodents, cognitive functions such as spatial learning and memory gradually decline during aging (Yassa et al., ). The current results indicating that learning and memory abilities gradually decrease with age are in line with many previous results (Cao et al., ; Chen et al., ; Duan et al., ; Ennaceur et al., ; Khan et al., ; Wu et al., ). In fact, our previous studies showed that CD‐1 mice began to experience cognitive impairments at 12 months of age (Li, Cao et al., 2016; Wu et al., ).
LPS, the cell wall component of Gram‐negative bacteria, can activate the immune system and induce the expression of proinflammatory cytokines (Dantzer, ; Li et al., ). These cytokines can affect the normal function of the brain through specific signaling pathways and accelerate brain aging (Akbarian et al., ; Patterson, ; Zhuang et al., ). In the present study, mice exposed to inflammation during pregnancy performed significantly worse than normal mice on the MWM test at 6 months and 18 months. The current results suggested that gestational inflammation can accelerate age‐associated cognitive impairment in both learning and memory. Moreover, the LPS and LPS‐E mice had worse learning and memory performance than the CON mice at both ages. Nevertheless, the LPS‐E mice had similar learning and memory performance as the 6 M LPS mice and had significantly better learning and memory performance than the 18 M LPS mice. Therefore, our evidence suggested that living in EE can significantly reduce the accelerated spatial cognitive impairment in both learning and memory resulting from gestational inflammation of elderly LPS mice, but it takes enough time to achieve this effect (at least more than 3 months).
### EEs slow the accelerated decrease of hippocampal TET1/GSNOR in aged mice with gestational inflammation
Oxidative stress is a condition in which endogenously or exogenously produced pro‐oxidant species—for example, reactive oxygen and nitrogen species (ROS and RNS, respectively)—overwhelm the antioxidant defense during basal conditions (Sies et al., ; Sies & Cadenas, ). Dysfunction of the body's response to oxidative stress results in the accumulation of damaged biomolecules and cellular structures (Singh et al., ), which are hallmarks of cell senescence. In this scenario, mitochondria play a fundamental role as they are the main intracellular source of oxygen free radicals along with being one of the main targets of ROS and RNS (Balaban et al., ).
Mitochondrial aging free radical theory is that aging is the result of mitochondrial damage accumulation and that mitochondrial dysfunction is an important process for a variety of age‐related disease pathophysiologies (Montagna et al., , ). Mitochondria have a large number of protein cysteine residues. These residues can be oxidized by nitric oxide (NO) to form S‐nitrosothiols (SNOs) in a process called S‐nitrosylation. The oxidized form of cysteine is reversible, which makes cysteine act as a redox sensor or redox switch (Hess & Stamler, ; Rizza & Cardaci et al., ). This oxidative modification changes the activity of related proteins and regulates mitochondrial function (Ischiropoulos, ). Excessive S‐nitrosylation of related proteins will directly affect mitochondrial dynamic activity, reduce the efficiency of aerobic metabolism, inhibit mitochondrial autophagy, and destroy mitochondrial homeostasis (Liu et al., ). S‐nitrosoglutathione (GSNO), a low‐molecular‐weight SNO, can be reduced back to a sulfhydryl state by denitrosylation reactions and then affect the concentration of protein SNOs (2014). GSNOR is a highly specific denitrase for GSNO; it controls the intracellular levels of GSNO and protein SNOs and protects the body from nitrosative stress. GSNOR expression has been found to reduce the primary aging cells that are accumulated during the rodent and human aging process (Montagna et al., ). Recently, some have reported that GSNOR, by means of regulating the S‐nitrosylation state of proteins involved in mitochondrial dynamics and mitophagy, sustains the removal of irreversibly damaged mitochondria and delays cell senescence (Rizza & Filomeni, ). Mice with Gsnor defects (Gsnor ) have a greatly increased degree of brain protein S‐nitrosylation and show some characteristics of accelerated aging such as decreased muscle mass, neuromuscular dysfunction, and positive α‐synaptic nuclear protein accumulation in the cerebral cortex (Montagna et al., ; Rizza & Filomeni, ). Therefore, decreased expression of GSNOR may be an important indicator of aging.
The Gsnor promoter has a CpG island, which is an epigenetic predictive marker of cytosine methylation (Rizza & Cardaci et al., ). The expression of Gsnor is regulated by TET1. TET1 catalyzes the demethylation of the methylated cytosine of the Gsnor promoters. Oxidation of 5‐methylcytosine (5meC) to 5‐hydroxymethylcytosine (5hmeC) is the first reaction catalyzed by TET1 to reduce methylated cytosine and then promote Gsnor expression. 5hmeC is a molecular marker of active cytosine demethylation. As age increases, an increase in 5mC and a decrease in 5hmC indicate that the TET1 and GSNOR axis may play an important role in aging (Yassa et al., ). TET1 can be inhibited by succinic and fumaric acids, which are metabolites produced by the tricarboxylic acid cycle (Ferrer et al., ; Laukka et al., ; Xiao et al., ). Excessive S‐nitrosylation caused by GSNOR deficiency may lead to the accumulation of fumaric acid and further inhibit TET1 activity (Zeng et al., ), which suggests that reduced activity of TET1 can lead to methylation of the Gsnor promoter and further reduce GSNOR expression. Because senescence is associated with an increase in 5mC and a decrease in 5hmC in the CpG island within the promoter (Rizza & Cardaci et al., ), we hypothesized that TET1 and GSNOR are closely related in the senescence process.
Some studies have shown an age‐related decrease in TET1 and 5hmeC in peripheral blood mononuclear cells and T cells (Petursdottir et al., ; Truong et al., ; Valentini et al., ). These results are consistent with ours, which showed that the mRNA and protein levels of hippocampal TET1 and GSNOR in the aged (18 M) CON mice were significantly reduced compared with those in the young (6 M) CON mice. This suggested that age significantly affects the expression of TET1 and GSNOR. Moreover, gestational inflammation exerted age‐dependent effects on hippocampal Tet1 and Gsnor expression. Specifically, at 6 months of age LPS mice had significantly decreased levels of Gsnor mRNA and protein and Tet1 protein than the CON mice, while at 18 months of age LPS mice had significantly decreased levels of both Tet1 and Gsnor mRNA and protein than the CON mice. These results suggested that LPS exposure during pregnancy appears to decrease Tet1 and Gsnor expression in the hippocampus of mice, especially aged mice. Consequently, decreased expression of Tet1 and Gsnor would increase S‐nitrosylation of mitochondrial protein, which would lead to mitochondrial dynamic disorders and result in abnormal mitochondria morphology and dysfunction as well as inhibit mitochondrial autophagy (Rizza & Cardaci et al., , ).
Interestingly, compared to LPS mice, LPS‐E mice had similar levels of mRNA and protein of Gsnor and Tet1 at 6 months of age and increased levels at 18 months of age. When compared with the CON mice, LPS‐E mice had increased levels of mRNA and protein of Gsnor and Tet1 at 6 months of age and similar levels at 18 months of age. These findings indicated that only long‐term EE exposure can drive the otherwise accelerated decrease in hippocampal expression of Tet1 and Gsnor caused by gestational inflammation to near normal levels, especially in the aged hippocampus. Therefore, EEs need to be provided over a long time period to reverse the future consequences of gestational inflammation.
The mechanisms of how EEs increase the expression of Tet1 and Gsnor are unknown and complex at the very least. The effects of how EEs delay brain aging have been related to many factors. All of these factors may contribute to the increased expression of Tet1 and Gsnor in the mice living in EEs. For example, EEs can increase brain‐derived neurotrophic factors, promote the growth and maturation of neurons, and increase resistance to brain insult (Shlevkov et al., ). EEs can also alter the expression of important genes involved in brain aging through modifications in acetylation or DNA methylation patterns (Ikegami & Narita, ; Zhuang et al., ). In addition, EEs induce the production of proteins that suppress oxyradical production (Fernández CI, ). EEs include adequate conditions and opportunities for getting food and exercise. Studies have shown that going for long periods with no breakfast can significantly inhibit the mRNA of hippocampal memory‐related genes such as Tet1 (Okauchi et al., ) and that a healthy lifestyle from activities such as long‐term regular exercise can increase hippocampal Tet1 mRNA in rats (Sølvsten et al., ). These studies indirectly support the promotion of hippocampal Tet1 and Gsnor expression by EEs.
### Hippocampal reduction of TET1 and GSNOR are associated with impaired cognition
Mitochondria are critical for neuronal energy metabolism, calcium regulation, myelination, and neuronal and glial apoptosis (Princz et al., ; Zhuang et al., ). Therefore, mitochondrial decay may be central to neuronal damage and cognitive decline (Cheng et al., ; Lai et al., ). In the aging process of rodents and humans, Gsnor expression is reduced. This reduction is associated with mitochondrial stress damage, which is characterized by increased S‐nitrosylation of mitochondrial protein (Larrick & Mendelsohn, ). Tet1 knockout mice showed a reduction in multiple neuron activity‐regulating genes and obvious damage to learning and memory (Rudenko et al., ; Yang et al., ) . Therefore, TET1 and GSNOR may play an important role in the age‐related decline in learning and memory.
As expected, the current experiment found that there were close connections between the age‐related decline in learning and memory and TET1 and GSNOR in the aged hippocampus. In senescent (18‐month‐old) mice, these correlations were treatment‐dependent. The aged mice in the CON group showed a negative correlation between the levels of Tet1 and Gsnor mRNA and protein with the learning‐phase distance swam, and a positive correlation between the levels of Tet1 and Gsnor mRNA and protein with the memory‐phase percent distance swam. Moreover, the LPS and LPS‐E mice showed similar correlations in the learning‐phase as the CON mice. However, there was only a positive correlation between the mRNA levels of Tet1 and Gsnor and the memory‐phase percent distance swam in LPS mice and between the protein‐level of Gsnor and the percent memory‐phase distance swam in LPS‐E mice. Therefore, the impaired cognitive performance caused by exposure to inflammation during pregnancy may be related to decreased Tet1 and Gsnor transcription, which then leads to decreased translation in the aged animals. This decreased translation can be modulated by gestational inflammation or subsequent EE exposure.
That the long‐term EE alleviated the accelerated cognitive decline may be attributable to increased Gsnor translation. Of note, in young mice (6 months old), there was no significant correlation between the learning‐phase distance swam or the memory‐phase percent distance swam and the protein and mRNA levels of Tet1 and Gsnor , regardless of treatment. We speculated that the mitochondrial damage with gestational inflammation in 6‐month‐old mice was not enough to cause changes in learning and memory due to unknown compensatory mechanisms in the brain. The impaired spatial learning and memory abilities in young mice (6 months old) caused by gestational inflammation may be related to complex mechanisms such as changes in hippocampal mitochondrial biogenesis and dynamics as well as synaptic plasticity (Li, Cao et al., 2016; Zhuang et al., ).
## SUMMARY
The expression of hippocampal Tet1 and Gsnor in mice showed an age‐related decline. Exposure to LPS during pregnancy accelerated the down‐regulation of hippocampal Tet1 and Gsnor in aged mice, while a long‐term EE had a protective effect on this damaging outcome. Therefore, providing long‐term positive life events such as EE or avoiding negative life events such as pregnancy infection has important clinical significance for the prevention and treatment of age‐related cognitive impairment. Our study had some limitations. First, there were fewer groups of aged mice. Second, this study only determined the expression of Tet1 and Gsnor but did not further explore the associated signaling pathways. However, despite these limitations, we provide new insights into the mechanisms of cognitive impairment caused by gestational infection and the protective effects caused by almost lifelong EE exposure. Future research is needed for further clarification.
## AUTHOR CONTRIBUTIONS
ZYX conceived and designed the study, conducted literature search and data collection, analyzed data, and wrote manuscripts; WQY, WYT, and ZLP assisted in data collection and preliminary analysis; CL and CGH reviewed the manuscript. All the authors read and approved the content of the manuscript.
## CONFLICT OF INTEREST
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
### PEER REVIEW
The peer review history for this article is available at .
|
## Introduction
Little is known about the neural correlates of mood states and the specific physiological changes associated with their valence and duration, especially in young people. Arterial spin labeling (ASL) imaging is particularly well-suited to study sustained cerebral states in young people, due to its robustness to low-frequency drift, excellent interscan reliability, and noninvasiveness. Yet, it has so far been underutilized for understanding the neural mechanisms underlying mood states in youth.
## Methods
In this exploratory study, 21 healthy adolescents aged 16 to 18 took part in a mood induction experiment. Neutral, sad, and happy mood states were induced using film clips and explicit instructions. An ASL scan was obtained following presentation of each film clip.
## Results
Mood induction led to robust changes in self-reported mood ratings. Compared to neutral, sad mood was associated with increased regional cerebral blood flow (rCBF) in the left middle frontal gyrus and anterior prefrontal cortex, and decreased rCBF in the right middle frontal gyrus and the inferior parietal lobule. A decrease in self-reported mood from neutral to sad condition was associated with increased rCBF in the precuneus. Happy mood was associated with increased rCBF in medial frontal and cingulate gyri, the subgenual anterior cingulate cortex, and ventral striatum, and decreased rCBF in the inferior parietal lobule. The level of current self-reported depressive symptoms was negatively associated with rCBF change in the cerebellum and lingual gyrus following both sad and happy mood inductions.
## Conclusions
Arterial spin labeling is sensitive to experimentally induced mood changes in healthy young people. The effects of happy mood on rCBF patterns were generally stronger than the effects of sad mood.
## Introduction
Little is known about the neural correlates of mood states and the specific physiological changes associated with their valence and duration, especially in young people. Here, we investigate these correlates using a magnetic resonance imaging (MRI) technique known as arterial spin labeling (ASL; Detre and Alsop ) following the induction of neutral, sad, and happy moods in a group of healthy adolescents. We exploit the phenomenon of neurovascular coupling (Attwell et al. ) by measuring the changes in regional cerebral blood flow (rCBF) that accompany the onset and maintenance of specific mood states.
One reason for limited research into the neural substrates of mood is methodological. Experimental designs using functional magnetic resonance imaging (fMRI) typically involve stimulus change measured in the timescale of seconds (Matthews and Jezzard ). It is also important to understand what underlies the persistence of mood states over sufficiently long periods given that the diagnosis and monitoring of patients requires measurement of psychopathology in the order of several hours to days and weeks. Time-series fMRI data using the blood-oxygen-level-dependent (BOLD) contrast are sensitive to a low-frequency drift, and thus less reliable when investigating neural activation changes over periods lasting longer than seconds (Smith et al. ). Electroencephalography (EEG), used to track brain activation changes over time, is not directly sensitive to subcortical neural activity (Kennett ) and has poor spatial resolution, making it less suited for investigation of limbic regions that are implicated in mood. Existing evidence on the neurophysiological correlates of mood states comes mainly from positron emission tomography (PET) studies that measured rCBF in participants experiencing experimentally induced sadness or happiness (e.g., George et al. ; Mayberg et al. ; Liotti et al. ; Keightley et al. ). However, the reliance of PET on radio-labeled compounds makes this method unsuitable when studying children and adolescents. This is unfortunate as mood disorders such as depression have their origins in adolescence, with a sharp increase in prevalence reported with the onset of puberty (Maughan et al. ). Recent studies have also provided evidence for a link between adolescent depression and psychopathology later in life (Thapar et al. ). Therefore, discovering the physiological patterns associated with mood states in adolescence is particularly important.
Arterial spin labeling appears especially well-suited to studying the neural signatures of different mood states and their specific features, such as duration and intensity. First, as blood flow contrast is generated by the pair-wise subtraction of successively acquired pairs of images (see Materials and Methods section), the data are substantially free of low-frequency sources of contamination such as physiological noise and scanner drift, and less sensitive to subject movement (Smith et al. ; Aguirre et al. ; Detre and Wang ; Howard et al. ). Second, the ASL pulse sequence acquisition parameters are tailored to maximize blood flow information from tissue capillaries and the data are therefore a more faithful signature of functionally driven changes in neurovascular coupling (see Materials and Methods section). Unlike PET, ASL is noninvasive and has been previously used in children and newborns (e.g., Wang et al. ; Biagi et al. ). Arterial spin labeling was already shown to successfully distinguish between states of depression in adults (Lui et al. ) and between adolescents with and without depression (Ho et al. ). In addition, its excellent interscan reliability (Hermes et al. ; Hodkinson et al. ) makes it suitable for monitoring treatment effects.
Mood induction is an established experimental manipulation where the participant's mood state is temporarily changed. It can be used in within-subject designs, where each participant serves as his or her own control for different mood states, and specific characteristics of the stimulus, such as intensity, can be manipulated. Mood induction is a more tractable method for an experimental study compared to studying patients who are already depressed as it is not confounded by the effects of illness and considerable between-patients heterogeneity (e.g., disorder severity, duration, number of previous episodes, comorbidity, and medication use).
Despite the above advantages, to our knowledge, ASL has only been used once with mood induction (Gillihan et al. ), in an adult sample. To test the feasibility of examining mood states in youth using ASL, we investigated brain perfusion patterns involved in mood changes in a sample of healthy adolescents. We used film clips combined with mood elaboration instructions, a method found to be the most reliable in inducing mood (Westermann et al. ) and compared sad and happy mood conditions against the neutral. The inclusion of happy condition is based on its clinical relevance to depression, a disorder characterized not only by the predominance of sad mood, but also the absence or inability to perceive positive emotions (APA ). fMRI studies show that adults with major depressive disorder (MDD) display a dampened neural activation to positive stimuli relative to healthy controls (Epstein et al. ). It is therefore theoretically important to study the neural correlates of sad and happy moods together, as dysregulation of both these mood states is relevant to depression.
We hypothesized that the mood induction procedure will lead to significant changes in self-reported mood, and that it will generate rCBF changes in areas implicated in mood processing. We used unbiased, voxel-wise, whole-brain analyses as well as predefined, bilateral regions of interest (ROIs): the amygdala, subgenual anterior cingulate cortex (sgACC), dorsolateral prefrontal cortex (dlPFC), ventromedial prefrontal cortex (vmPFC), and the ventral striatum. We expected higher amygdala activation in response to emotional (sad or happy) than neutral conditions, based on its role in encoding emotional significance. We hypothesized that the sgACC would show higher activation following sad versus neutral mood induction, based on previous PET mood induction studies in adults (Mayberg et al. ; Liotti et al. ; Keightley et al. ) and sgACC hyperactivity in patients with depression (Drevets et al. ), recently replicated in adolescents with depression using ASL (Ho et al. ). Prefrontal ROIs were chosen based on their role in regulating limbic activity (e.g., Davidson ; Drevets et al. ); therefore we expected these regions to be more activated in emotional (sad or happy) conditions compared to neutral. We also hypothesized increased rCBF in the ventral striatum following happy versus neutral mood induction, based on the relation between ventral striatal activity and euphoria in healthy adults (Drevets et al. ). Finally, we explored whether the results are dependent on existing depressive symptoms.
## Materials and Methods
### Participants
Twenty-two healthy adolescents aged 16 to 18 (10 males, 12 females) were recruited via adverts on social media websites and Internet forums for teenagers. In addition, one parent/carer of each participant completed a series of questionnaires (see below) about their child. One female participant was removed from subsequent analyses due to persistently high levels of anxiety while in the scanner, leaving a final sample of 21 participants. All participants were right-handed as measured by the Edinburgh Handedness Inventory (Oldfield ). The adolescent participants did not have any serious medical, behavioral, or emotional conditions, had no history of head injuries by self-report, and did not report any contraindication to MRI. Written informed consent was obtained from all participants. This study was approved by the Psychiatry, Nursing & Midwifery Research Ethics Subcommittee at King's College London (PNM/12/13-44).
### Questionnaires
Participants were screened for the presence of behavioral and emotional difficulties before taking part in the study with a series of self- and parent-reported questionnaires. Mood and Feelings Questionnaire (MFQ; Costello and Angold ) was used to measure depressive symptoms present in the previous 2 weeks. Symptoms of trait anger and irritability in the previous 6 months were measured using the Affective Reactivity Index (ARI; Stringaris et al. ). Additional emotional and behavioral symptoms were measured using the Strengths and Difficulties Questionnaire (Goodman ). Participants scoring high on any of these measures were excluded from the study. Individual cases were discussed with an attending physician (A.S.).
### Procedure
#### Stimuli
Based on a meta-analysis of mood induction procedures (Westermann et al. ), emotional film clips coupled with mood elaboration were chosen as a way of inducing mood. Each film clip was approximately 4 min long and depicted the following: neutral clip – a young man describing how to clip in and out of mountain bike pedals; sad clip – a scene from Dead Poets Society (Weir, ) where a teenage boy finds out that his best friend committed suicide; happy – a series of stand-up comedy routines by a British comedian, Michael McIntyre. Before seeing each film clip, the participants were instructed to enter the specified mood state (as used previously by Habel et al. ). The instructions were as follows: “During this task, I would like you to try to become sad/happy. To help you do that, I will show you a video that most people find sad/happy”. After seeing each film clip, the participants saw a message asking them to think about how the film had made them feel (as used previously, e.g., by Furman et al. ). For instance in case of sad mood condition, the instructions were as follows: “Have you ever been in a similar situation? Have you ever lost a loved one and if so, how did it make you feel? How would you feel if you were in the same situation?”
All participants were shown the scanner environment and invited to lay down inside our mock scanner in order to familiarize themselves with the scanning environment and reduce the potential for drop out. After confirming that they were ready to proceed, the participants entered the MRI scanner. First, a structural MRI scan was taken. The participants then rated their mood on a scale from 0 (very sad) to 10 (very happy), followed by the neutral mood induction that served as a baseline. After having watched the film clip, participants rated their mood again. This was followed by the first ASL scan (7:15-min long, see below) during which the participants were instructed to remain still and look at the screen with the following text: “Think about how you felt when watching this neutral film clip. Please try to maintain this feeling while you're being scanned.” The procedure was then repeated for the sad and happy conditions. Mood ratings for each condition were collected immediately after the end of each film clip.
#### MR imaging
The scanning was carried out at the Centre for Neuroimaging Sciences, Institute of Psychiatry, King's College London using a General Electric MR750 3.0T scanner.
In ASL, the MRI signal of endogenous arterial blood water is used as a contrast agent to measure rCBF. The contrast is achieved by “labeling” or “tagging” a bolus of arterial blood, by inverting its magnetization in the region of the carotid arteries with an external (noninvasive) radiofrequency pulse. If two whole-volume images are rapidly acquired in succession (one with and one without labeling of arterial blood), the resultant difference image is proportional to the volume of blood perfused into each unit volume of tissue during the time between the labeling and the acquisition of the image. This time is typically long enough (1.5 sec) so that the contrast is derived from labeled water in the microcirculation (capillaries) and not in any of the larger arterioles. A suitable model is employed to convert the difference image into a map of rCBF in conventional physiological units of mL blood/100 g tissue/min. As stated earlier, the continuous pair-wise subtraction of labeled and nonlabeled images makes ASL suitable for tasks using longer lasting stimuli due to the low sensitivity to signal drift.
Each ASL image volume of 54 slices (3-mm thickness, no interslice gap) was acquired using a pseudocontinuous flow-driven adiabatic inversion scheme (Dai et al. ); TE/TR = 11.088/4901 ms, flip angle (FA) = 111°, postlabeling delay 1525 ms. Acquisition of five control and labeled pairs was done with a 3D FSE, multishot spiral stack, employing eight spiral arms for each interleave in a total of 7:15 min. Spiral k-space data were regridded to a 256 × 256 in-plane matrix prior to Fourier transformation. A single proton density scan with the same acquisition parameters was used as a reference to compute rCBF in standard units. This procedure yielded rCBF maps with a resolution of 2 × 2 × 3 mm. Enhanced fast gradient echo three-dimensional sequence was used to collect T1-weighted images, with TR = 7.312 msec, TE = 3.016 msec, inversion time 400 msec, FA = 11 , field of view = 270 mm, 256 × 256 matrix, 196 sagittal slices 1.2-mm thick.
### Image processing
Image processing and analyses were performed using the Statistical Parametric Mapping suite (SPM, Functional Imaging Laboratory, University College London, London UK, version 8, ; RRID:nif-0000-00343). ASL images were normalized to the standard space of the Montreal Neurological Institute (MNI) by the following procedure: first, raw rCBF maps were coregistered to the high-resolution T1-weighted anatomical volume after coarse alignment of the origin of both images. Segmentation of the T1-weighted image yielded a “brain-only” binary mask which was multiplied by the coregistered rCBF map to produce an image free of extracerebral artifacts. Finally, the T1-weighted image was transformed to the T1-weighted MNI template and the transformation parameters applied to the clean rCBF maps. All normalized rCBF maps were then spatially smoothed with a 8 × 8 × 8 mm kernel.
#### ROI definition
ROIs (all bilateral) were defined using the WFUPickAtlas toolbox (RRID:nif-0000-00358; Maldjian et al. ) available in SPM. The sgACC was defined as Brodmann area (BA) 25 and dilated by 1 voxel. The dlPFC was generated by combining BA 9 and BA 46, and was dilated by 1 voxel. The amygdala was defined using the Automated Anatomical Labeling (AAL) library (Tzourio-Mazoyer et al. ). The vmPFC ROI combined bilateral medial orbital frontal and rectus regions from the AAL atlas. As the atlas does not include a predefined mask for the ventral striatum, this ROI was defined as two 8-mm spheres based on MNI coordinates (right: x = 9, y = 9, z = −8; left: x = −9, y = 9, z = −8) derived from a previous meta-analysis (Postuma and Dagher ) as used by Nusslock et al. ( ).
### Statistical analysis
We first examined the effectiveness of our mood induction procedure in producing stimulus-congruent mood changes using repeated-measures analyses of variance (ANOVAs) and paired-samples t-tests on self-reported mood ratings.
Subsequently, whole-brain analysis of ASL images from the three mood induction conditions was performed using a one-way, within-subjects ANOVA with gender and mean global CBF added as covariates. This was due to small, but significant changes in global CBF that occurred during the time inside the scanner [ F = 8.66, P = 0.001, = 0.313; mean global CBF decrease from 56.2 to 53.9 mL blood/100 g tissue/min]. As this was an exploratory study and to date there is no consensus regarding statistical analysis of ASL “activation” data, we employed two different methods to indicate significance of findings at the whole-brain level. First, we used the stringent, SPM-derived significance of P < 0.05 with family wise error (FWE) correction based on cluster extent. Second, we performed Monte Carlo simulations using the AlphaSim program in Resting-State fMRI Data Analysis Toolkit (Song et al. ) to determine the cluster size (number of voxels) needed in order to achieve a corrected P lower than 0.05; thresholding the statistical images with a cluster-forming threshold of P = 0.01 and clustering with a cluster connection radius of 2 mm. Minimum cluster size for all individual analyses are provided in the results section. For all ROI analyses, small volume correction in SPM was used, FWE corrected at the voxel level.
We then investigated whether the amount of self-reported mood change from neutral to sad/happy correlated with the amount of change in brain perfusion patterns. To do this, we performed a multiple regression analysis with the difference in self-reported mood scores (sad or happy minus neutral) regressed against the difference between respective perfusion images (neutral subtracted from sad or happy). Gender and mean global CBF were added to the model as covariates.
The effects of depressive symptoms on brain perfusion patterns following mood induction were examined using multiple regression. Total MFQ score was added to the model as a predictor, and the “perfusion difference” image (neutral subtracted from sad or happy) as the outcome. Gender and mean global CBF were added to the model as covariates.
## Results
### Behavioral results
#### Mood ratings
As illustrated in Figure , the mood induction procedure led to significant changes in self-reported mood ratings among the participants, F = 143.48, P < 0.001, = 0.878. Compared to the neutral condition, the participants rated their mood significantly lower after seeing the sad film clip, t = 12.02, P < 0.001, d = 2.18, and significantly higher after seeing the happy clip, t = 8.81, P < 0.001, d = 2.10. The difference in ratings between the sad and happy conditions was also significant, t = 13.22, P < 0.001, d = 4.19.
Mean mood ratings (with 95% confidence intervals) from 21 participants after watching a neutral, sad, and happy film clip in the scanner.
#### Questionnaire data
Total SDQ scores indicated average levels of emotional and behavioral difficulties in the sample by self- (mean = 7.7, SD = 3.9) and parent-report (mean = 4.9, SD = 3.9). Anger and irritability levels were low by both self- (mean ARI = 1.6, SD = 1.5) and parent-report (mean ARI = 2.3, SD = 2.5). Lastly, there was a strong cross-informant agreement between depressive symptoms as rated by the young people themselves (mean MFQ = 6.9, SD = 6.1) and by their parents (mean MFQ = 2.8, SD = 2.8); r = 0.50, P < 0.05.
### Neuroimaging results
#### Sad mood
At whole-brain level, sad mood induction led to increases in two clusters that were significant based on cluster size AlphaSim threshold, but not FWE correction (Table , Fig. ), adjusted for the global CBF. The first, larger cluster with a peak in the left middle frontal gyrus also encompassed left postcentral gyrus. The second cluster included left medial superior frontal gyrus and BA 10. In contrast, right superior and middle frontal gyri and right inferior parietal lobule (BA 40) showed decreased rCBF after sad compared to neutral mood induction (Table , Fig. ). No significant ROI results were found.
Whole-brain level analysis results for (a, b) the ANOVA sad versus neutral contrast, and (c) correlation between self-reported mood ratings difference and brain perfusion maps difference for sad minus neutral mood induction conditions
Whole-brain level ANOVA results showing regional cerebral blood flow (rCBF) levels for the contrasts (A) sad versus neutral and (B) happy versus neutral, overlaid on a T1-weighted structural brain image. Orange = increased rCBF relative to neutral, blue = decreased rCBF relative to neutral. All locations are reported in MNI coordinates. For illustration purposes, the cluster-level significance is P < 0.05 (AlphaSim corrected). BA, Brodmann area; L, left; R, right.
We then correlated the difference in brain perfusion patterns between sad and neutral mood induction conditions with the corresponding difference in self-reported mood ratings. We found a significant negative correlation in right precuneus at whole-brain level (Table , Fig. ), suggesting that the decrease in self-reported mood from neutral to sad condition was associated with increased perfusion changes in this region. No significant ROI results were found.
Results of whole-brain level analyses for the regressions between self-reported mood rating differences and regional cerebral blood flow (rCBF) differences for the contrasts (A) sad minus neutral, (B) happy minus neutral, overlaid on a T1-weighted structural brain image. Orange = positive correlation, blue = negative correlation. All locations are reported in MNI coordinates. For illustration purposes, the cluster-level significance is P < 0.05 (AlphaSim corrected). BA, Brodmann area; L, left; R, right.
#### The role of depressive symptoms
Next, we investigated whether the magnitude of neural activation following mood induction depended on the level of current depressive symptoms. We performed a regression analysis of “perfusion difference” images (neutral condition image subtracted from sad or happy) against total MFQ scores.
For sad mood condition, we found negative whole-brain level correlations between self-reported MFQ and the sad minus neutral perfusion difference. As shown in Table and Figure , higher MFQ scores were associated with lower rCBF in bilateral cerebellum, right lingual gyrus, and right BA 18. The results were significant at the cluster size, but not FWE corrected, level. No significant ROI results were found.
Correlation results between self-reported depressive symptoms (MFQ) and brain perfusion maps difference for (a) sad minus neutral, (b) happy minus neutral; all at whole-brain level
Results of whole-brain level analyses for the regressions between self-reported depressive symptoms (MFQ) and regional cerebral blood flow (rCBF) difference: (A) sad minus neutral, (B) happy minus neutral, overlaid on a T1-weighted structural brain image. Orange = positive correlation, blue = negative correlation. All locations are reported in MNI coordinates. For illustration purposes, the cluster-level significance is P < 0.05 (AlphaSim corrected). BA, Brodmann area; L, left; R, right.
We also found whole-brain level correlations between self-reported MFQ and the happy minus neutral image difference. As can be seen in Table and Figure , the higher the MFQ score, the higher the rCBF after watching the happy versus neutral film clip in the supplementary motor area (SMA) and a large cluster encompassing left middle and inferior temporal gyri. In contrast, MFQ scores were negatively correlated with rCBF in a large cluster encompassing the lingual gyrus and cerebellum (Table ). No significant ROI results were found.
#### Happy mood
At the whole-brain level, happy mood induction led to significant increases in rCBF in a large cluster extending from the brainstem via the cingulate gyrus to the medial frontal gyrus (see Table and Fig. ) and a smaller cluster encompassing the subgyral areas of left parietal and frontal lobes. In contrast, a large cluster including the inferior parietal lobule showed decreased rCBF after happy compared to neutral mood induction (Table and Fig. ).
Whole-brain and ROI results for (a, b) the ANOVA happy versus neutral contrast, and (c) correlation between self-reported mood ratings difference and brain perfusion maps difference for happy minus neutral mood induction conditions
ROI analyses revealed increased rCBF in the sgACC and ventral striatum (see Table ). There was also a marginally nonsignificant finding in the amygdala ( P = 0.051).
We then correlated the difference in brain perfusion patterns between happy and neutral mood induction conditions with the corresponding difference in self-reported mood ratings. We found a significant positive correlation in the BA 8, precentral gyrus, cerebellum, and superior temporal gyrus at whole-brain level, as well as the amygdala and dlPFC ROIs (Table and Fig. ), suggesting that the increase in self-reported mood from neutral to happy condition was associated with increased perfusion in these regions.
By contrast, there was a negative correlation between the magnitude of self-reported mood change and rCBF between happy and neutral conditions in right inferior parietal lobule and right BA 11.
## Discussion
This was the first exploratory study to investigate the neural substrates of mood states in young people using ASL, an MRI method that is especially suited to examining prolonged neural activation. We showed that mood changes can be robustly induced in healthy adolescents using our paradigm, as evidenced by significant changes in self-reported mood without significant between-subject variance. We also found rCBF differences following sad and happy mood induction procedures compared to neutral. The amount of rCBF change was affected by the degree of induced mood change and by current depressive symptoms.
Our main finding in the sad versus neutral contrast was a change in perfusion in the middle frontal gyrus (BA 6), with increased rCBF on the left, and decreased rCBF on the right side following sad mood induction. A PET study of adult patients with depression previously showed that decreased perfusion in middle frontal gyrus can be reversed with antidepressant treatment, consistent with the involvement of this region in mood processing (Buchsbaum et al. ). We did not expect lateralized findings, although one previous PET study in healthy adults also found rCBF in left middle frontal gyrus to be increased, and the right decreased, when performing a cognitive task following sad versus neutral mood induction (Baker et al. ). Second, decreased rCBF in the inferior parietal lobule following sad versus neutral mood induction is consistent with this region's role as a component of the default mode network (DMN), a network of brain regions that are active during wakeful rest (Buckner et al. ). Reduction in DMN activity has been associated with self-referential processing (e.g., Sheline et al. ). Crucially, we also observed decreased rCBF in the inferior parietal lobule following happy mood induction, suggesting that the DMN activity was suppressed when participants actively engaged in mood elaboration regardless of mood valence. Lastly, we found a correlation between the intensity of self-reported sadness and increased rCBF in the precuneus during sad mood elaboration, consistent with the role of precuneus in the recall of episodic and self-referential memory (Cabeza and Nyberg ).
None of our sad mood induction findings reached the stringent, FWE-corrected significance level. There are two possible explanations. First, due to paucity of research with pediatric samples, our hypotheses were mainly based on PET mood induction studies with adults. These showed effects of both happy (Schneider et al. ; George et al. ) and sad mood induction on rCBF (Schneider et al. ; Keightley et al. ). It could be that adolescents show a weaker rCBF response to sad mood induction due to their stage of development or that there is higher variance in this response across subjects. A large-scale ASL study recently found that the trajectory of rCBF evolution undergoes dynamic changes across adolescence (Satterthwaite et al. ). Consistent with the developmental hypothesis, Kliegel et al. ( ) found that younger adults show lower emotional reactivity to negative mood induction compared to older adults, consistent with a weaker relation between daily stress and negative affect in younger versus older adults (Mroczek and Almeida ). However, the extent to which our results reflect a developmental effect remains unclear without direct comparison using matched groups. Alternatively, the healthy, never-depressed adolescents included in this study might not be as susceptible to sad mood induction as they are to the happy. Using mood induction and fMRI, Joormann and colleagues found differences in neural activation between healthy girls and girls at risk of depression in areas implicated in negative mood processing (Joormann et al. ). We were unable to test this hypothesis directly since the MFQ scores of our participants were all within the nondepressed range. Future studies should investigate whether rCBF reactivity to sad mood induction is higher in adolescents with than without depression. Nevertheless, even in our nondepressed sample, we did find a negative correlation between the severity of depressive symptoms (self-reported MFQ score) and rCBF in the cerebellum and lingual gyrus following both sad and happy mood inductions. This is consistent with previous fMRI research showing decreased capacity for processing happy faces in the cerebellum and lingual gyrus in adults with depression (Fu et al. ), an effect that was reversed by antidepressant treatment. Decreased activity in the cerebellum and lingual gyrus in response to positive stimuli was also found in euthymic patients with bipolar depression (Malhi et al. ) compared to healthy controls. Together with the recent finding that adolescents with depression show decreased rCBF in the cerebellum compared to healthy controls (Ho et al. ), our results provide some additional evidence for the involvement of the cerebellum in emotional processing (e.g., Konarski et al. ). We also found a positive correlation between depressive symptoms and rCBF in the SMA following happy mood induction, although the role of this region in mood processing remains unclear.
In line with our hypotheses, we found increased rCBF in the limbic regions (including the ventral striatum and a marginally not significant finding in the amygdala) following happy mood induction procedures. Moreover, there was a positive correlation between the self-reported increase in happiness and rCBF change in the left amygdala and left dlPFC. These results are consistent with the role of the frontolimbic circuitry in emotional processing, with the amygdala involved in determining the emotional content of stimuli and frontal regions modulating emotional responses. This is in keeping with mood induction fMRI findings in patients with depression, who show an opposite direction of effects compared with healthy controls. For instance, in adults with MDD, severity of depressive symptoms correlated negatively with activation in the following areas following happy, but not sad mood induction: left putamen, bilateral caudate, left nucleus accumbens, and left amygdala (Keedwell et al. ). Furthermore, decreased ventral striatum activity when processing positive words in depressed versus healthy adults correlated with symptoms of anhedonia (Epstein et al. ), consistent with abnormalities in the reward processing system in depression. Depressed adults also show an opposite pattern of vmPFC activation following happy and sad mood induction, compared to nondepressed adults (Keedwell et al. ). Finally, adolescents with depression show lower rCBF in the dlPFC at rest compared to controls, as measured by ASL (Ho et al. ). Future studies should investigate whether adolescents with depression show dampened rCBF responsiveness to the happy – and heightened responsiveness to sad mood induction. Notably, happy mood induction is an ethically viable way of inducing a mood state, especially in children and those at risk of a mood disorder. Given what we know about decreased positive affect in depression, happy mood induction may function as a helpful probe for detecting depression in youth.
Contrary to our hypotheses, we also found an increase in sgACC perfusion following happy, rather than sad mood induction. sgACC activation to happy (as well as sad) stimuli was previously reported in adults with treatment-resistant depression (Kumari et al. ); where it was suggested that the severely depressed patients responded to happy stimuli as to frustrative nonreward. However, our participants were not clinically depressed and we did not find a correlation between depressive symptoms and rCBF in the sgACC following happy mood induction in our sample. Interestingly however, healthy adults show significant functional coupling between the left amygdala and both the dlPFC and sgACC during emotion regulation (reappraisal of negative emotion; Banks et al. ). It is possible that coactivation of these regions in our study reflects the participants actively maintaining their happy mood following the induction procedure.
The main strength of this study is the combined use of ASL and mood induction to directly examine rCBF patterns associated with three different mood states in adolescents. Importantly, we carried out the scanning after film clip presentation, ensuring that the resulting neural activity reflected the participant's mood state and not film clip characteristics (color, brightness, or sound). This study is limited by the fixed order of mood induction conditions, used deliberately to maximize the power to detect mood-specific rCBF patterns in our sample. A larger study with a randomized order of mood induction conditions is needed to rule out the possibility of an order effect or emotional contagion. Secondly, a general limitation of mood induction methods is that self-reported mood ratings may be influenced by the participants' desire to please the examiner.
## Conclusion
This study offers a crucial starting point for the investigation of mood states, using methodology that bypasses the limitations of conventional fMRI. Although important challenges remain (Savitz et al. ), studying the tonic activation of neural networks involved in mood processing is likely to have important clinical implications for disorders characterized by persistently sad or happy mood, such as unipolar and bipolar depression, across development.
|
## Objectives
To identify and analyze the characteristics of the most influential articles about central nervous system ( CNS ) inflammatory demyelinating disease.
## Materials and Methods
The Institute for Scientific Information (ISI) Web of Science database and the 2014 Journal Citation Reports Science Edition were used to retrieve the top 100 cited articles on CNS inflammatory demyelinating disease. The citation numbers, journals, years of publication, authorships, article types, subjects and main issues were analyzed. For neuromyelitis optica (NMO), articles that were cited more than 100 times were regarded as a citation classic and described separately.
## Results
The top 100 cited articles were published between 1972 and 2011 in 13 journals. The highest number of articles ( n = 24) was published in Brain, followed by The New England Journal of Medicine ( n = 21). The average number of citations was 664 (range 330–3,897), and 64% of the articles were from the United States and the United Kingdom. The majority of the top 100 cited articles were related to multiple sclerosis ( n = 87), and only a few articles reported on other topics such as NMO ( n = 9), acute disseminated encephalomyelitis ( n = 2) and optic neuritis ( n = 2). Among the top 100 cited articles, 77% were original articles. Forty‐one citation classics were found for NMO.
## Conclusions
Our study provides a historical perspective on the research progress on CNS inflammatory demyelinating disease and may serve as a guide for important advances and trends in the field for associated researchers.
## Introduction
Central nervous system inflammatory demyelinating disease (CIDD) is a term that encompasses a broad spectrum of diseases such as multiple sclerosis (MS), neuromyelitis optica (NMO), acute disseminated encephalomyelitis (ADEM), optic neuritis (ON), and transverse myelitis. MS is a prototypic form of CIDD that has garnered great interest by researchers worldwide due to its high prevalence, young age of onset and chronicity, which results in a significant social burden (Adelman, Rane, & Villa, ). Recent understanding of the role of aquaporin‐4 (AQP4) antibodies in NMO has enhanced researcher attention in this field (Lennon et al., ). As a result, numerous papers were published regarding these diseases, and the clinical characteristics and treatment strategies were decided to some degree (Aleixandre‐Benavent et al., ; Wingerchuk & Carter, ). However, there is increased demand to stratify the current literature regarding idiopathic inflammatory demyelinating disease to serve as a guide to researchers.
Bibliometrics is a research method that analyzes citation frequencies and patterns of articles in a category of interest (Moed, ). Although there is some debate as to the association between the number of citations and the quality of the study, highly cited articles can indirectly represent the impact of a particular article on the scientific community and the trends in a specific field of research (Moed, ). The results of a citation analysis can suggest “classic lists” of articles in a specific field and “hints” about trends in citations within an area.
Although a few bibliometric studies have examined CIDD, the research objects were restricted to particular regions, time periods or subtypes of CIDD, particularly MS (Aleixandre‐Benavent et al., , ; Araujo, Moreira, & Lana‐Peixoto, ; Gonzalez de Dios et al., ). The aim of this study was to identify and analyze the characteristics of the top 100 most frequently cited articles under the heading of “CIDD” worldwide. Considering the increased importance of NMO in this field and the relatively short time period in which NMO has been regarded as a distinct entity from MS due to its AQP4 antibody specificity, different treatment response and pathology (Wingerchuk et al., ), we separately suggested citation classics for NMO in this study.
## Materials and Methods
The Web of Knowledge Journal Citation Reports Science Edition 2014 (Thomson Reuters, New York, NY, USA) was used to search for all journals that are listed under the categories “clinical neurology”, “neuroscience” and “medicine, general & internal”. We retrieved all articles that were cited more than 100 times in the selected journals using the Cited Reference Search option of the Science Citation Index Expanded of the ISI Web of Science database (January 1945–February 2016).
To find the CIDD‐related works among all the articles that had been cited more than 100 times in the three categories, we used following search terms: “multiple sclerosis”, “demyelinating disease”, “myelitis”, “optic neuritis”, “clinically isolated syndrome”, “neuromyelitis optica”, “Devic's disease”, “Balo concentric sclerosis”, “Schilder's diffuse sclerosis”, “Schilder's disease”, “diffuse myelinoclastic sclerosis”, “Marburg multiple sclerosis”, “acute disseminated encephalomyelitis”, “solitary sclerosis”, “acute hemorrhagic leukoencephalitis”, “neuromyelitis optica‐immunoglobulin G (NMO‐IgG)”, and “aquaporin‐4 antibody”. The search terms for the citation classics on NMO included “neuromyelitis optica”, “Devic's disease”, “NMO‐IgG”, and “aquaporin‐4 antibody”. The original texts of all the searched articles were evaluated for their applicability. The citation classics on NMO were defined as articles that were cited more than 100 times ( ). The lists of the 100 top cited articles on CIDD and the citation classics on NMO were obtained and analyzed for their characteristics: number of citations, year of publication, published journal, authorship, country and institution of origin, type of article, subject of article (e.g., MS, NMO, ADEM, ON, etc.) and main issues. The country and institution of origin was defined by the affiliation of the first author. If the first author had more than one affiliation, the affiliation of corresponding author was used. Two researchers independently reviewed the data (J.E.K and K.M.P) and any disagreements were decided by further discussion with another neurologist (J.S.B). This study did not need to be reviewed by an ethics committee since it performed a bibliometric analysis of existing published studies.
## Results
### Characteristics of the top 100 cited articles of CIDD
The list of the top 100 cited articles on CIDD are presented in Table . The average number of citations for the top 100 cited articles were 664 (range, 330–3,897). The articles were published between 1972 and 2011, and the majority of the articles were published after 1995 (Figure a). There were 64 articles that were cited more than 400 times, which is the criteria commonly used as the threshold for a citation classic ( ). The most frequently cited article was about the recommended diagnostic criteria for MS by an international panel that was published by McDonald et al. in 2001 [rank 1]. The revision of the “McDonald Criteria”, which was published by Polman et al. in 2005 [rank 2], was second highest most cited article (Table ). Sixty‐four of the 100 articles originated in the United States of America (USA) and the United Kingdom (UK). The Institute of Neurology (London, UK), the Mayo Clinic (Minnesota, USA), the Cleveland Clinic Foundation (Ohio, USA) and the Free University Hospital (Amsterdam, the Netherlands) had the highest rankings for their contribution to the top 100 most cited articles (Table ). The highly cited articles were published in 13 journals, which was led by Brain ( n = 24) and closely followed by The New England Journal of Medicine ( n = 21) (Table ). Fifty‐three authors were found to have contributed more than three articles on the list. Miller DH was the most prolific author for the 100 articles, followed by Weinshenker BG and Polman CH (Table ). Of the top 100 articles, 77 articles were original research articles and 23 articles were review articles or guidelines.
List of the top 100 cited articles on central nervous system inflammatory demyelinating disease
Frequency distribution based on the year of publication for the top 100 cited articles on central nervous system inflammatory demyelinating disease (a) and the citation classics on neuromyelitis optica (b)
Countries and institutions of origin for the citation classics in the field of central nervous system inflammatory demyelinating disease (a) and neuromyelitis optica (b)
The most predominant subject of the 100 articles was MS ( n = 85) and among these, 7 articles focused on evaluating prognostic factors (clinical, magnetic resonance imaging [MRI] features or treatment availability) to determine the conversion of clinically isolated syndrome to MS [ranks 21,30,45,64,70,92,98 in Table ]. Two articles enrolled ADEM patients, which were mainly pediatrics, and discussed the clinical and MRI features of ADEM that can be used to differentiate ADEM from MS [ranks 81a,88b in Table ]. Two articles enrolled patients with several subtypes of CIDD concomitantly and one of these articles was a randomized controlled trial that evaluated the effect of plasma exchange in acute central nervous system (CNS) inflammatory disease [ranks 68,71 in Table ]. Only nine articles were about NMO [ranks 10,12,17,22,34,72b,77,90a,95 in Table ], and two articles about ON [ranks 36,41 in Table ] were also included in the list.
The main issues and their time trends of the highly cited articles are summarized in Table and Figure a. Among the original articles, the majority of the papers ( n = 27) focused on treatment topics such as the effects of immunomodulatory drugs including intramuscular interferon β‐1a, glatiramer acetate, natalizumab, rituximab, fingolimod, mitoxantrone, alemtuzumab, monoclonal anti‐tumor necrosis factor antibody cA2, steroid, and plasma exchange [ranks 6,9,11,15,18,19,23,25,28b,30,33,35,36,46,48,55,56,64,68,70,87,88a,90a,93a,93b in Table ]. Other common issues were neuroimaging features as evaluated by MRI, diffusion tensor imaging, magnetic resonance spectroscopy and positron emission tomography ( n = 17). A considerable portion of the papers also discussed the pathological findings of CIDD ( n = 16).
General issues discussed in the highly cited articles on central nervous system inflammatory demyelinating disease and neuromyelitis optica
Time trends in categories of citation classics on central nervous system inflammatory demyelinating disease (a) and neuromyelitis optica (b). CIDD , Central nervous system inflammatory demyelinating disease; NMO , Neuromyelitis optica
### Citation classics of NMO
Using the cut‐off threshold of 100 citations, we found 41 citation classics for NMO (Table ). The mean number of citations of the citation classics was 291 (range, 102–1,317). All articles were published between 1976 and 2012 and there was a surge in the number of publications after 2006 (Figure b). This surge in publications corresponds to the publication of the top 2 cited articles, in which the NMO‐specific autoantibody (NMO‐IgG) was detected in the serum of NMO patients [rank 2 in Table ] and was subsequently followed by the suggestion of revised diagnostic criteria that included NMO‐IgG positivity in the diagnosis [rank 1 in Table ]. These two articles had a significant role in discriminating NMO from MS. There are eight different countries of origin for the citation classics (Table ). USA ( n = 20) had the largest number of articles, which was followed by the UK ( n = 6), Japan ( n = 5) and Germany ( n = 4). The Mayo Clinic (Minnesota, USA) was the most active publishing institution for the NMO citation classics (Table ). The citation classics were published in 13 journals, which was led by Neurology ( n = 15), followed by Brain ( n = 7), Archives of Neurology ( n = 5) and Annals of Neurology ( n = 3; Table ). Twenty‐four authors contributed 3 or more of the articles on the citation classics list. Weinshenker BG had the highest number of articles, followed by Wingerchuk DM, Pittock SJ, Lennon VA, and Lucchinetti CF (Table ). Among the 41 articles, 5 were review articles or guidelines [ranks 1,4,28,29,40 in Table ], 1 was a case report [rank 37 in Table ], and the remainder of the articles was original research articles. Four articles reported basic science studies [ranks 14,16,17,20 in Table ]. The main issues discussed in the citation classics for NMO and their trends over time are summarized in Table and Figure b. The most frequently discussed topic among the citation classics was the serological markers, NMO‐IgG or AQP4 antibody. The original studies concerning the diagnostic and prognostic values of NMO‐IgG or the AQP4 antibody and their pathogenicity were frequently cited [ranks 2,10,11,14,15,16,17,20,21,22,23,24,31,33,34,35,36,41 in Table ]. The second most commonly cited issues were related to the natural course and clinical features of NMO [ranks 3,12,25,26,27,30,32,37,39 in Table ]. Other frequently cited issues included NMO pathology [ranks 5,7,13 in Table ], the effect of various treatments (rituximab or prednisolone with azathioprine) [ranks 8,18,19 in Table ], and MRI characteristics [ranks 6,9,38 in Table ].
Citation classics on neuromyelitis optica
## Discussion
In the current study, we identified and characterized the top 100 cited articles concerning CIDD. Using this bibliometric approach, we can examine the historical progress within a field of interest and inform researchers of the articles or authors that have had a significant impact on the field.
The majority of the top 100 frequently cited articles on CIDD were published during 1995–2009. This unequal distribution might reflect the natural life span of articles in a specific academic society. Typically, scientific articles begin to be cited 1 to 2 years after publication and reach a maximum between 3 and 10 years (Marx, Schier, & Wanitschek, ). After this period, the number of citations for a particular article gradually decrease due to the “obliteration by incorporation” phenomenon, where the results of a citation classic are absorbed into the current knowledge (Garfield, ). Therefore, similar to the findings from other studies, a citation analysis can accurately capture the true impact of articles published within the last 10–20 years. Only 3 of the top 100 articles were published in 2010–2016. This result does not suggest that the recently published articles were less important, but indicates that insufficient time has elapsed to accumulate a large number of citations. Therefore, it is important to not only examine the absolute number of citations of an article but also the entire list of published articles to avoid overlooking critical works in that field. For example, the article describing the diagnostic criteria for MS that was published in 1976, which ranked 69 in our list, did not have less influence on researchers studying MS than the study on the McDonald criteria, which was published later and had the top rank in our study. The distribution of publication years of the NMO citation classics showed a somewhat different pattern. The number of citation classics abruptly increased after 2006. Before the detection of NMO‐specific antibodies (NMO‐IgG and AQP4‐antibody), NMO was regarded as a rare form of MS. “Optico‐spinal MS” was historical used to describe this unique phenotype (Kuroiwa & Shibasaki, ). However, the publication of revised NMO diagnostic criteria in 2006 that included MRI diagnostic criteria according to the length of the spinal cord lesion and the detection of the AQP4 antibody led many researchers to distinguish NMO from MS (Wingerchuk, Lennon, Pittock, Lucchinetti, & Weinshenker, ), which increased the importance of investigating NMO specifically. This conceptual change in NMO diagnosis for several decades might be indirectly reflected in our study.
About two thirds of articles in both lists (CIDD and NMO) originated from USA and United Kingdom. The overwhelming influence of these two countries can also be found in other clinical disciplines. These findings might be explained by the larger community of specialists, the availability of research funds and organized support in this area and the tendency of American/British researchers to cite local papers. In addition to these two countries, the Netherlands, Italy and Germany produced a relatively large number of highly cited articles on CIDD. Japan was the only non‐western country that had a significant number of NMO citation classics, which might reflect the ethnic predilection of NMO in Asia and the research efforts of the Institute of Tohoku University (Kira, ).
Journal impact factor is considered a representative metric for the influence of that journal in the associated scientific field (Garfield, ). Eighty‐eight percent of the top 100 cited articles on CIDD were published in five major journals with high impact factors: Brain, The New England Journal of Medicine, Neurology, Annals of Neurology, and The Lancet . However, not all of the frequently cited articles were published in high impact factor journals, but were published in journals with rich histories. The lead high impact factor journals were the same for the NMO citation classics list as for the CIDD list. However, the number of the journals in the Web of Science category “medicine, general & internal” was higher in the CIDD list than the NMO list.
Six of the articles among top 10 most cited articles on CIDD were about proposed diagnostic criteria guidelines for MS or NMO or review articles concerning the pathogenesis or clinical definition of these two diseases. The significant influence of these guidelines to individual clinical or experimental studies is expected considering their essential use in clinical practice and research. Common issues among the CIDD articles focused on the results from randomized controlled trials on the use of immunomodulatory drugs, the pathogenesis of MS and the diagnostic and prognostic use of new neuroimaging technologies such as MRI. The most common issues among the NMO articles were related to the pathogenicity of NMO and the diagnostic/prognostic value of the NMO‐IgG or AQP4 antibodies. Articles discussing the treatment and neuroimaging issues surrounding NMO represented only a small portion of the current list, but is likely to grow fast based on the trends observed in MS over the last few decades.
The number of citation classics, which was defined as articles that received more than 400 citations ( ), was 64 for the CIDD list. This number is slightly smaller than other neurological diseases (107 for Parkinson's disease and 89 for epilepsy; Ibrahim, Snead, Rutka, & Lozano, ; Ponce & Lozano, ). This might be influenced by differences in the number specialists who work in specific field or the portion of basic research publications for each disease. A recent paper presented the bibliometrics for all the research studies under the term “MS” (Aleixandre‐Benavent et al., ). This study had a different viewpoint and methodological approach from our study, which might explain the discordance between the results of the two studies. We analyzed the top 100 cited articles on CIDD published during an unlimited time period, but other researchers included whole articles published in the Multiple Sclerosis Journal and articles under the term “MS” in the Web of Science database during 2003–2012.
Our research has some limitations that should be considered. First, as aforementioned, temporal bias is inevitable in a bibliometric analysis. Older articles can be more or less cited due to increased opportunities to be cited over time and the “obliteration by incorporation” phenomenon (Bohannon & Roberts, ; Garfield, ). There were limited Internet resources for the articles published before 1990. Similarly, the impact of recently published compelling articles can be underestimated due to an insufficient amount of time to accumulate citations. Use of other citation indices, such as average citations per year or time duration of high citations, may overcome this temporal bias, although results may still be influenced by the period of analysis. Second, several important articles might have been missed from our lists including articles published in journals in categories such as “immunology” or “pediatrics”. However, the proportion of articles in these categories was modest among the MS original research articles (Aleixandre‐Benavent et al., ). Third, incomplete citation bias might impact the results. Incomplete citations include various conscious or unconscious bias that arise from self‐citations, omissions, native language preponderance and the tendency to cite review articles or high impact factor journals (Ponce & Lozano, ; Seglen, ). Furthermore, there are other databases such as Scopus and Google Scholar that can be used for citation analysis. Selected search engines may have influenced the results (Kulkarni, Aziz, Shams, & Busse, ).
Here, we presented a detailed list of the top 100 cited articles for the topic CIDD and also separately propose citation classics for NMO using bibliometric methods. A strength of our research is that we included all types of articles published worldwide during an unrestricted time period. Although the citation rate does not directly represent the quality of the study, it is one marker used to recognize the importance of studies in the scientific community. We can trace scientific progress and identify seminal articles in a specific field by citation analysis.
## Conflict of Interest
All authors declare that they have no conflict of interest.
## Supporting information
|
## Background and Purpose
This study investigates the prevalence of delirium in acute stroke patients on a primary stroke unit ( SU ) analyzing associated risk factors and clinical outcomes.
## Method
Prospective, 4‐month observational study from 2015 to 2016 on patients aged ≥18 years with stroke at a German university hospital's SU . The presence of delirium as first outcome was rated at three times daily using the Confusion Assessment Method ( CAM ). Secondary outcome measures were duration of delirium, rehabilitation in SU , length of stay in SU and hospital, complications, and mortality. Significant risk factors were used to conduct a confounder‐matched case–control analysis.
## Results
309 patients were included. The overall prevalence of delirium was 10.7% (33 patients) mostly on the first and second hospital day. Duration of delirium on SU was in median 1.0 day (Interquartile range: 0.3–2 days). In 39.4% of patients delirium was present in a short time interval (≤8 hr) and in 24% of patients delirium was diagnosed during nightshifts exclusively. Significant risk factors for delirium were dementia, age ≥72 years, severe neurological disability on admission, and increased C‐reactive protein on admission. The case–control analysis showed that delirious patients had more complications and a trend toward a worse rehabilitation.
## Conclusions
These results underline the importance of delirium screening in stroke patients specifically during the night. Since even short delirious episodes are associated with more complications and increased disability, future studies are needed to find delirium prevention strategies.
## INTRODUCTION
In Germany, 270,000 persons per year experience a stroke (Wiedmann et al., ). Common complications after stroke are dysphagia, aspiration pneumonia, falls, infections, depression, and delirium (Langhorne et al., ). Despite the high occurrence of delirium, a routine assessment of delirium for patients after stroke is not recommended (Norrving et al., ).
Delirium is defined as disorders in awareness and cognition (mainly attention and memory), develops within hours or days, cannot be explained by other cognitive disturbances as dementia and is a direct result of a physical disturbance or medication (American Psychiatric Association, ). Delirium can appear in hyper‐, hypoactive, or mixed forms, and substance withdrawal (Pandharipande, Jackson, & Ely, ). The etiology of delirium is complex, and current theories explain its development by interaction of hypoxia, inflammatory processes, disturbance of neurotransmitter, and the presence of internal or external risk factors (Riedel, Browne, & Silbert, ). The prevalence of delirium in patients after stroke is estimated to average 26% (Carin‐Levy, Mead, Nicol, Rush, & van Wijck, ). A delirium increases the risks for mortality, complications, longer length of hospital stay, and institutionalization (Shi, Presutti, Selchen, & Saposnik, ).
Till today, the prevalence of delirium on primary Stroke Units (SU) in Germany remains unknown. Hence, the purpose of this observational study was to evaluate the delirium prevalence in a German, national certified SU (Nabavi et al., ).
## METHOD
We conducted a prospective, observational study on a SU over 4 months to assess prevalence of delirium in patients after stroke. Primary outcome was the presence of delirium. Stroke patients were examined for the presence of delirium three times a day. Diagnosis of delirium was assessed by using the Confusion Assessment Method (CAM) (Inouye et al., ), which is a screening algorithm, derived from the Diagnostic and Statistical Manual of mental Disorders, 4 edition for the diagnosis of delirium. Secondary outcome parameters were duration of delirium, rehabilitation and first day of out‐of‐bed mobilization, number of delirium‐related pharmacological treatments, complications, length of stay on SU and in hospital, discharge destination, and mortality (all defined below).
### Setting
The study was conducted in a primary, national certified SU (Nabavi et al., ). The SU has got a 24‐hr presence of a neurologist, interprofessional rounds twice a day, 2 weekly visits of pharmaceutics and antibiotica‐stewardship for patients with symptoms of infections. Nurse‐patient ratio is 1:4 in three shifts in 24 hr. 25% of 42 registered nurses joined a further education for specialized stroke care. Patients were cared by a comprehensive stroke treatment according with the German guidelines including regular assessment of the neurological status four times per day (06:00, 12:00, 18:00, and 22:00) by a neurologist with a continuous attendance on the SU; 2‐hourly observation of vital signs and neurological status by nurses and a permanent 24 hr bedside monitoring of vital parameters.
### In‐ and exclusion criteria
Every patient, who was admitted on the SU, was screened for in‐ and exclusion criteria. Inclusion criteria were as follows: present ischemic or hemorrhagic stroke including transient ischemic attacks (TIA) (Wiedmann et al., ) and patients with cerebral venous sinus thrombosis. Exclusion criteria were (1) due to German law of data protection, no consent for research with patient's data by patients themselves or legal representatives; (2) patients with initial stroke‐like symptoms which could not be confirmed as stroke; (3) patients after neuroradiological interventions, because of a longer stay in hospital before the intervention; (4) patients who were admitted >24 hr on other units or hospitals; (5) patients with an age <18 years; (6) patients who were unable to be assessed for delirium; (7) other reasons, for example, foreign language. Criteria were confirmed by control of discharge information.
### Risk factors and data collection
Based on a systematic review and post hoc analysis of risk factors in previous studies (Nydahl, Margraf, & Ewers, ), following patient factors were included: (1) socio‐demographical data: gender, age; (2) the presence of dementia and/or psychiatric disorders prior to admission, as reported by general practitioner; (3) C‐reactive protein (CRP) >0.4 mmol/L (4) admission in a 2‐ or 5‐bed room as environmental factor; and (5) status before admission: housing conditions and level of preexisting physical disability.
Level of disability was assessed using the modified Rankin Scale (mRS) (Banks & Marotta, ) that is a six‐item scale. Values from 0 to 2 are coded as nor or light disability and good outcome, values from 3 to 5 as severe disability and unwanted outcome. mRS was recorded before admission, during admission and at the point of discharge from SU. Rehabilitation on SU was calculated by difference between mRS at the time of discharge from SU and admission. Evaluation of National Institutes of Health Stroke Scale was not assessed routinely at all time points and could not be used for evaluation.
During the stay on the SU, the first day of mobilization was recorded as well as type and number of delirium related, pharmacological treatments were screened. Complications were assessed on a daily basis and categorized as: (1) falls; (2) urinary tract infection; (3) nosocomial pneumonia 48 hr after admission; (4) restraints of at least hands; and (5) unwanted removal of vascular‐, nasal‐, or bladder tubes.
### Delirium assessment
Delirium was assessed using the CAM (Inouye et al., ). Assessment of CAM is based on four criteria: (1) acute onset and fluctuating course; (2) inattention; (3) disorganized thinking and/or (4) an altered level of consciousness (Inouye et al., ). Patients are assessed positive for delirium, if (1), (2) and either (3) and/or (4) are given, as described in detail on . The CAM has been validated for the assessment of delirium (Inouye et al., ), and can be used for patients after stroke (Dahl, Ronning, & Thommessen, ; Lees et al., ; McManus et al., ; Miu & Yeung, ) with a good sensitivity and specifity and has got a strong interrater reliability (Inouye et al., ). In case, patients had a severe aphasia and/or dementia and could not respond to simple questions, the presence of disorientated behavior and its fluctuation in 24 hr was rated as criteria for delirium (Gustafson, Eriksson, Sture, Bucht, & Gösta, ) and confirmed by families by asking them for new onset of such behavior. Assessment of delirium according to the CAM and its subtypes was conducted by nurses in each shift, three times a day covering a 24‐hr period (Lemiengre et al., ). Duration of delirium could be assessed for the stay on the SU. End of delirium was defined as 24 hr without any delirium‐positive assessment. In case, a patient was discharged from SU and not 24‐hr delirium‐free, the time of discharge was counted as end of delirium on SU.
### Delirium management
Delirium screening was introduced in 2013. The interprofessional team was teached using a standardized script, bedside teachings, and case evaluations. Delirium‐Pocketcards and Posters were provided for clinicians, delirium‐information leaflets for families and patients. Families had no restrictions in visiting times. Delirium management included as first choice nonpharmacological interventions, including information, mobilization, reorientation, provision of glasses and/or hearing aids, sleep hygiene, and integration of families; and as second choice: pharmacological interventions.
### Statistics
Nominal data are reported as frequency ( n ) and percentage (%). Metrical, normal distributed data were reported as mean and standard deviation, non‐normal distributed data as median and interquartile range (IQR). Calculated was length of stay by counting full days. Hypothesis was statistically proven by Fisher's Exact test for nominal data and Mann–Whitney U‐test for metrical data. To avoid misinterpretations by multiple testing, a sequential Bonferroni correction was used to correct for a two‐tailed α‐level of p = .05 by sequential division of number of factors included into the analysis(Bortz & Schuster, ). Multicollinearity was tested by Cramer's V and tolerated, if Variance of Inflation Factor <5 (Urban & Mayerl, ). Normal distribution of metrical data was tested by Shapiro–Wilk test (Bortz & Schuster, ). Due to the limited number of delirious patients, a logistic regression analysis could not be performed (Ottenbacher, Ottenbacher, Tooth, & Ostir, ). Hence, a matched case–control comparison was performed. Matching was conducted in a randomly chosen, 1:1 design without tolerance, using factors that were identified as significant after Bonferroni correction in previous bivariate analysis (Armenian, ). Group comparison between delirious cases and nondelirious controls were calculated using McNemar test, Yates correction, and Wilcoxon test, Odds ratios using Chi square (Bortz & Schuster, ). All calculations were done using 22 (IBM Corp. New York).
### Ethical protocol approval
The study was approved by the ethic committee of Christian‐Albrechts‐University, Kiel. Due to the observational character of this study, the study was not registered.
## RESULTS
The observational study covered 4 months from October 14th, 2015, till February 14th, 2016. Out of 464 admissions, 67.5% ( n = 309) patients could be included (Figure ).
Recruitment of patients
### Screening rate
The rate of delirium screenings in 309 patients was 84.3% ( n = 1,747) of 2,071 possible delirium screenings. During the first week on SU, 40.6% ( n = 685) assessments were conducted in morning shift, 28.5% ( n = 482) in afternoon shift, and 30.8% ( n = 520) in night shift. Delirium was not assessable in 5.5% ( n = 17) of patients by several reasons, mostly during the first days. Out of these 17 patients, 70.6% ( n = 12) became better and were assessable and all except one was free of delirium.
### Delirium
Overall prevalence of delirium was 10.7% ( n = 33) of patients. Delirium was assessed in 45.5% ( n = 15) each on first and second day on SU, 9% ( n = 3) of deliriums occurred during third or later days. Most delirium assessments identified a mixed delirium (57.7%, n = 41), followed by 19.7% ( n = 14) in hyperactive form, 18.3% ( n = 3) in hypoactive, and 4.2% ( n = 3) in alcohol withdrawal form.
Duration of delirium was in median 1.0 days. (IQR: 0.3–2.0 days). Most delirious phases were less than 24 hr (45.5%, n = 15), 39.4% ( n = 13) were delirious for only one assessment, hence, equal or less than 8 hr. 30.3% ( n = 10) of delirious patients were discharged from SU, before they were 24‐hr delirium‐free. Delirium‐positive assessments were found in 5.4% ( n = 37) morning shift, 6.8% ( n = 33) in afternoon shift, and 8.6% ( n = 45) in night shift. 24% ( n = 8) of patients was delirious only during the night. Delirium‐related pharmacological treatment of delirium was administered to 69.7% ( n = 23) of delirious patients. Most used medications for this reason were melperone (21.8%, n = 12), haloperidol (20%, n = 11), lorazepam (18.2%, n = 10), and others (40%, n = 22).
### Comparison
Delirious and nondelirious patients were compared for risk factors. Significant risk factors for delirium after sequential Bonferroni correction were dementia (Odds Ratio (OR): 17.29, 95% Confidence Interval (95% CI): 6.745–44.322), severe neurological disability (mRS) on admission (OR: 6.791, 95% CI: 2.715–16.986), increased age ≥72 year. (median) (OR: 5.819, 95% CI: 1,992–17.002), increased CRP on admission (OR: 2.831, 95% CI: 1.338–5.989), and admission in 5‐bed room (OR: 0.216, 95% CI: 0.097–0.484). Multicollinearity was tolerable.
### Case–Control
A randomized, 1:1 case–control design, matched for above listed significant risk factors could include 27 delirious and 27 nondelirious patients. Multicollinearity of risk factors was tolerable. Included patients ( n = 54) differed significant from patients, who were not included in case–control design ( n = 255): they were older (median: 80.5 year. (IQR: 75.0–87.2 year.) vs. 72.0 year. (60.0–81.0), p < .001), had a pronounced neurological deficit during admission (mRS: mean 3.7 ( SD 1.2) vs. 2.3 (1.5), p < .001), had a longer stay on SU (median 3.0 days. (IQR: 2.7–5.0 days) vs. 3.0 days. (2.0–4.0 days), p = .006) and in hospital (8.0 days. (6.0–12.0 days) vs. 6.0 days. (4.0–10.0 days.), p = .008). The case–control analysis revealed that delirious patients showed significant more complications during their stay, but not a delayed mobilization, an increased length of stay on SU or in hospital nor higher mortality, compared to similar nondelirious patients. Compared to controls delirious patients showed a smaller success on rehabilitation with a difference of one point in the mRs on SU (uncorrected p = .017), which is clinically relevant. The difference is not significant after sequential Bonferroni correction, hence giving a trend to a decreased improvement in rehabilitation during their stay on SU (Table ). The negative impact of delirium on rehabilitation is seen in patients with delirious episodes of 8 hr but is still evident in patients suffering from delirium less than 8 hr. (median: ‐0.54, IQR ‐3.0 ‐ 0.0).
Case–control analysis for delirious and nondelirious patients
## DISCUSSION
In this prospective, observational study prevalence of delirium in more than 300 patients after acute stroke was nearly 11%. Screening of delirium was conducted three times a day and achieved a screening rate of more than 80%. Most delirious episodes were detected on first and second day after admission, during the night and lasted less than 1 day. In a case–control analysis, matched for dementia, severe neurological disability on admission, increased age ≥72 year., and increased CRP on admission, delirious patients had more complications but not a worse outcome except a tendency for a reduced rehabilitation improvement, compared to control patients without delirium.
Delirium is related to several risk factors as higher age, dementia, disability on admission, and increased CRP. Higher age was found as a risk factor in other studies, too (Miu & Yeung, ; Oldenbeuving et al., ) and may be explained by changed morphology of the aging brain (Oldenbeuving et al., ) or reduced perception (Dahl et al., ). Patients with dementia show a higher risk for delirium (Holtta et al., ), especially with infections (Simone & Tan, ). A severe disability on admission, assessed by the modified Rankin Scale, had also a higher risk for delirium, what can be explained by disturbed neurotransmitters, leading to delirium (Maldonado, ). An increased CRP is an indicator for infections, which are a trigger for delirium in general (Maldonado, ). After controlling for confounding risk factors (higher age, dementia, disability on admission, increased CRP), delirious patients in our cohort had not a worse outcome, compared to similar patients without delirium. This is in contrast to other studies (Caeiro, Ferro, Albuquerque, & Figueira, ; Gustafson et al., ; Mitasova et al., ). Severely disabled, delirious patients have more complications and might have a worse rehabilitation on SU, even in short episodes of delirium. Rehabilitation requires active involvement and patients’ alertness. Delirium disturbs these factors and may necessitate a prolonged rehabilitation. Beside, other hypothesis as more complicated infarction and reduced number of rehabilitation session due to delirium might also explain a reduced success in rehabilitation, and hence make it difficult to distinguish between cause and effect in this aspect. Due to the early onset of delirious episodes at first or second day, the effect of nonpharmacological and pharmacological prevention strategies on delirium prevalence, and hence rehabilitation, remains unproven. More research is needed to evaluate the impact of delirium on stroke rehabilitation.
Screening rate of delirium was above 80% and covered 24 hr. Only one other study used a 24‐hr screening, too (Gustafson et al., ), but found higher prevalence. Most positive screenings were during the night, hence screening rates of other studies, which used a once per day assessment, might have underestimated delirium prevalence (Caeiro et al., ; McManus et al., ; Miu & Yeung, ; Oldenbeuving et al., ). The assessment instrument CAM was used in other studies with patients after stroke, too with results of 10% (Dahl et al., ), 12% (Oldenbeuving et al., ), 27% (Miu & Yeung, ), and 28% (McManus et al., ). Delirium screened by nurses might be incorrect, especially in hypoactive delirium and dementia, leading to an underestimated prevalence (Lemiengre et al., ), contrary, a false‐positive delirium in one single shift would add another patient with delirium leading to an overestimated prevalence. Nevertheless, nursing staff was educated and cooperated with specialized physicians, leading to an overall correct screening. Due to the high rate of cases of delirium during the night, 24–hr screening is recommended for covering delirious episodes.
Prevalence of delirium is 10.7%, giving a lower prevalence than 26% in a recent meta‐analysis (Carin‐Levy et al., ). There are different hypothesis’ to explain this result. Improvements in stroke care over the last years may contribute to a low prevalence of delirium (Dahl et al., ). Especially close observation of neurological status and vital parameters (e.g., temperature together with fast escalating infection strategies) may reduce incidence of delirium as response to cerebral reaction to infection induced neurotoxic transmitters (Riedel et al., ). Furthermore, on the SU in our study delirium screening and evidence‐based delirium management is part of daily routine, including nonpharmacological interventions and early mobilization (Bernhardt et al., ). Other studies in the field did not reported delirium‐related structures and processes; hence a comparison is not feasible. Another influencing factor may be the inclusion of patients with TIA, who were included in other studies, too with a delirium prevalence of 10% (Dahl et al., ) and 48% (Gustafson et al., ). Patients with TIA are less impaired and this may reduce prevalence’ rates. Overall, we hypothesize that education of the personnel and delirium‐aware structures on a SU may have a positive impact on delirium prevalence, but a final proof in a delirium reducing setting of SU is still missing.
Duration of delirium was in median 1 day, 45% of delirious episodes lasted for 24 hr or less. In other studies, duration of delirium was in mean 4 days (Dostovic, Smajlovic, Sinanovic, & Vidovic, ; Mitasova et al., ) or 4.8 days (Oldenbeuving et al., ) and delirious episodes ≤24 hr were reported in 25% (Mitasova et al., ) to 46% (Sheng, Shen, Cordato, & Zhang, ). This phenomenon in patients with an acute stroke might be caused by cerebral dysregulation and associated with reduced perfusion, hypoxia, and disturbed neurotransmitters (Maldonado, ). Especially delirium during the night is discussed by disturbed melatonin circulation (Oldham, Lee, & Desan, ) and might explain the result of a delirium only during the night. Another hypothesis might be early infections during onset of stroke causing delirious episodes, for example, by aspiration and first cerebral responses, especially in patients with poststroke immunodepression (Famakin, ). Modern stroke care includes stabilization of circulation, narrow observation of vital signs in an appropriate nurse‐patient‐ratio, fast responses in terms of infections and implementation of nonpharmacological delirium prevention that may be an explanation for a short lasting delirium. Contrary, duration of delirium on SU can be influenced by length of stay on SU, too. In this study, 30% of delirious patients was discharged without being delirium‐free for 24 hr, to cover at least one following night without any delirium. Other authors defined the end of a delirium by one delirium‐free assessment (Oldenbeuving et al., ) or up to 48 hr without delirium (Mitasova et al., ). The question of the most appropriate definition of the end of delirium remains unanswered.
This study has different strengths and limits: A benefit of this study is the utilization of strict and explicit statistical methods. Data were collected over an extensive time period of 4 months on a primary stroke center with a high rate of accomplished delirium assessments. A limitation as a single center study is the missing option of a comparison of different SU settings in terms of delirium rates due to different SU concepts. Furthermore, the evaluation of the delirium duration is limited because some stroke patients were transferred to another ward before the end of delirium missing further follow‐up.
## CONCLUSIONS
Delirium in stroke patients is frequent with a prevalence rate of 11%. An increased awareness toward these patients is required because of the significant more complications and their tendency for a decreased improvement in rehabilitation. A standardized screening of delirium, performed three times a day, is recommended to detect also short delirious episodes, especially during the night. Our use of stricter statistical methods helped to identify more reliable risk factors (mainly dementia, severe neurological disability as well as CRP on admission and age) and challenged results of other studies. Since even short delirious episodes are associated with more complications and increased disability, future studies are needed to find delirium prevention strategies.
## DISCLOSURES
Related to this study: none. Other disclosures: PN, AW, AB AE, and GB reported no conflicts of interest. KW reports personal fees from Medtronic, a travel grant from BIAL, grants from the Ministry of Research and the German Research Foundation, outside the submitted work.
|
This article reviews recent developments in the application of cell-free DNA-based liquid biopsies to neurological diseases.
Over the past few decades, an explosion of interest in the use of accessible biofluids to identify and track molecular disease has revolutionized the fields of oncology, prenatal medicine and others. More recently, technological advances in signal detection have allowed for informative analysis of biofluids that are typically sparse in cells and other circulating components, such as CSF. In parallel, advancements in epigenetic profiling have allowed for novel applications of liquid biopsies to diseases without characteristic mutational profiles, including many degenerative, autoimmune, inflammatory, ischaemic and infectious disorders. These events have paved the way for a wide array of neurological conditions to benefit from enhanced diagnostic, prognostic, and treatment abilities through the use of liquid biomarkers: a ‘liquid biopsy’ approach.
This review includes an overview of types of liquid biopsy targets with a focus on circulating cell-free DNA, methods used to identify and probe potential liquid biomarkers, and recent applications of such biomarkers to a variety of complex neurological conditions including CNS tumours, stroke, traumatic brain injury, Alzheimer’s disease, epilepsy, multiple sclerosis and neuroinfectious disease. Finally, the challenges of translating liquid biopsies to use in clinical neurology settings—and the opportunities for improvement in disease management that such translation may provide—are discussed.
Gaitsch et al. review recent developments in the identification and application of cell-free DNA-based liquid biopsies to neurological diseases, including CNS tumours, stroke, traumatic brain injury, Alzheimer’s disease, epilepsy, multiple sclerosis, and neuroinfectious disease.
## Introduction
### Basics of liquid biopsy
Liquid biopsy, an alternative to solid tissue biopsy, involves sampling body fluids, usually for molecular components released from cells. Biofluids used for liquid biopsy commonly include plasma, serum and CSF but also extend to urine, saliva, pleural effusions and others ( ). The liquid biopsy approach has many advantages, including being relatively less invasive and less expensive compared to tissue biopsy and offering a more panoramic view of disease origins via analysis of genetic and epigenetic markers. Liquid biopsies have long been appreciated as potential tools to resolve the genesis of cancers of unknown origin and to detect occult tissue damage stemming from complex pathologies. Additionally, serial biopsies can be taken to determine treatment effects, a feat difficult to achieve with traditional tissue biopsies and nearly impossible in some tissues, such as brain and spinal cord. Liquid biopsies can sample numerous molecular entities within the blood, including free nucleic acids, such as cell-free nuclear DNA (cfDNA), cell-free mitochondrial DNA (cf-mtDNA), circulating tumour DNA (ctDNA) and cell-free RNA (cfRNA); extracellular vesicles, often containing nucleic acid components; proteins; and metabolites. Circulating nucleic acids (CNAs) were first described in the 1940s, though their usefulness was not fully appreciated until several decades later.
Overview of liquid biopsy, existing applications and target molecules. ( A ) Body fluids used for liquid biopsy include whole blood, plasma, serum, urine, CSF, saliva and others. The choice of biofluid depends on the specific clinical application, extent of disease, biomarker target, signal-to-noise tolerability and patient characteristics. ( B ) Biofluid samples can be used to determine the presence and levels of different molecular contents, including specific cell-free circulating nucleic acid fragments, extracellular vesicles, proteins and metabolites. ( C ) The liquid biopsy approach is currently being applied in several fields of medicine, including oncology, obstetrics and transplant medicine. These applications often rely on identification of specific mutations or polymorphisms in circulating DNA fragments, such as tumorigenic mutations. ( D ) The non-genetic features of cfDNA, including CpG methylation patterns, can allow for identification of their tissue and cellular origin. Additionally, these features may permit the application of liquid biopsy in neurological pathologies that are primarily characterized by degeneration, inflammation and ischaemia.
### Existing applications of liquid biopsy
Recent decades have seen an explosion in research investigating CNA-based liquid biopsies in a wide variety of human diseases. The field of obstetrics, for instance, often relies on non-invasive prenatal testing to determine foetal sex and the presence of chromosomal abnormalities. Similarly, the field of oncology has found numerous liquid biopsy applications, such as the United States Food and Drug Administration-approved Epi proColon test that analyses methylation patterns in cfDNA for population-wide colorectal cancer screening. The liquid biopsy approach has also been used for early detection of graft rejection in human organ transplants, and to evaluate tissue damage in patients with COVID-19. While neuro-oncology has made significant strides in developing liquid biopsies for CNS tumours, other neurological diseases are in need of non-invasive methods for diagnosis and prognosis. Many of the applications listed above rely on genetic anomalies in CNAs; however, many neurological conditions are not characterized by DNA sequence alterations.
### Strategies for liquid biomarker identification
Sequence differences in ctDNA, including known tumorigenic mutations, have been of particular interest in oncology for their potential as prognostic indicators and early-stage screening biomarkers. However, the non-genetic characteristics of CNAs, and cfDNA in particular, can provide information applicable to diseases without genetic mutations ( ). Epigenetic signatures, such as DNA methylation patterns, can be used to identify the tissue and even cell type of origin for cfDNA samples. Given that the human genome contains ∼28 million unique DNA methylation sites, the use of methylation array technology and deconvolution algorithms make it possible to identify differentially methylated regions (DMRs) corresponding to various cellular sources and gene expression levels. These approaches open the possibility of detecting cfDNA released from cell types that comprise small percentages of the total circulating cfDNA, such as CNS tissue-derived cfDNA in plasma. Remarkably, even cfDNA derived from minority cell populations in tissues, such as oligodendroglial lineage cells in multiple sclerosis, have become targets for liquid biopsy development.
### Basis of epigenetic liquid biomarkers
DNA methylation is an epigenetic process by which specific genes are silenced and thus prevented from being transcribed. DNA methylation is thought to physically interfere with transcription factor binding and to promote the formation of heterochromatin via recruitment of histone deacetylases and other proteins necessary for chromatin remodelling. The normal physiological functions of DNA methylation are wide-ranging, including embryonic development, suppression of transposable elements, genomic imprinting and X chromosome inactivation. Aberrant DNA methylation is associated with pathological states such as cancer and various age-related diseases.
The primary targets of methylation are dinucleotide segments of DNA, termed ‘CpG’ units, in which cytosine residues are located adjacent to guanine residues and connected via a phosphate bond. Methylation of DNA in promoter regions tends to repress gene transcription, while methylation of DNA in the gene body is associated with transcript splicing alterations. Thus, epigenetic processes allow for selective gene expression, enabling cell types to have unique functions. As DMRs underlie the unique methylation signatures of specific cell types, CpG methylation patterns are a marker of cell identity.
DNA methylation and other epigenetic patterns are important because the genetic sequence alone cannot give information about non-cancer pathologies such as neurodegeneration, inflammation and ischaemia. Epigenetic analysis of cfDNA may also allow for the determination of collateral tissue injury (e.g. off-target drug effects) and monitoring of treatment response during clinical trials. DNA methylation signatures can be used to map the source of the cfDNA released from cells during times of tissue damage, as in the case of tumour metastasis, since these fragments retain the methylation signatures characteristic of their tissue of origin. Recent advancements in methylome analysis have led to high-resolution deconvolution algorithms, which can match cfDNA fragments to specific tissues and cells based on a given set of DMRs.
### Challenges to clinical application of liquid biopsies
Despite the advantages of liquid biopsy over tissue biopsy, several technical hurdles have slowed its translation. ctDNA is released from tumours in minimal amounts in early-stage disease, often rendering it below the limit of detection. Additional roadblocks include the need for standardization of sample collection protocols and target amplification procedures, and more cost-effective molecular profiling methods. Furthermore, epigenetic signatures may change throughout disease, as is seen in CNS tumours, where the signatures used to identify initial disease are not optimal for recurring or treatment-resistant disease. Finally, confounding factors, such as comorbidities or risk factors, may inhibit liquid biopsy development by rendering studies difficult to compare. Ideal biomarker targets for liquid biopsy—indeed, for any biopsy—are stable and consistently present in high enough amounts to achieve detection with current technologies. Identifying markers that fit this profile will ensure the reproducibility needed for clinical application.
## Cell-free nuclear DNA-based targets in liquid biopsy
This review focuses on cfDNA-based liquid biopsies, one of the most heavily studied liquid biomarkers given the plethora of techniques available to analyse nucleic acids and to probe epigenetic features of DNA. However, rapid advances in genomic and transcriptomic techniques, combined with ongoing reductions in sequencing costs, have led to the identification of several additional types of CNA biomarkers. While not covered in this article, cfRNA—including cell-free mRNA, tRNA, long non-coding RNA and microRNA (miRNA)—and non-nucleic acid circulating components, such as proteins, metabolites, tumour-educated platelets, extracellular vesicles and whole tumour cells, provide promising targets for liquid biopsy in neurological and other disease contexts.
### Cell-free nuclear DNA
Cell-free nuclear DNA fragments are primarily derived from intra- and extracellular nuclease cleavage of DNA. They are present in small amounts in healthy individuals, with the majority derived from haematopoietic cells. Although cfDNA consists mainly of linear double-stranded DNA, it can also be present in the form of circular and single-stranded DNA. Double-stranded cfDNA molecules are the best-studied and most prevalent in plasma. Human plasma DNA consists of a mixture of DNA fragments of different sizes, with the modal size being approximately 166 base pairs (bp). The quantity of cfDNA changes during disease, when it is primarily released from dying cells or by active secretion. cfDNA itself can also precipitate tissue injury by serving as a damage-associated molecular pattern, as in renal tubular cells, via the production of mitochondrial reactive oxygen species. Fragments of cfDNA and chromatin can be taken up by mammalian cells and integrate into the genome, thus behaving like mobile genetic elements. This uptake suggests that cfDNA fragments might, in some contexts, serve as endogenous contributors to DNA damage associated with ageing.
In non-disease states, healthy individuals typically have plasma cfDNA levels lower than 10 ng/ml, whereas individuals with acute or chronic disease can have levels that are orders of magnitude higher. Levels of up to 50× the normal range have been reported in patients with a history of myocardial infarction, stroke, diabetes mellitus and cancer. Given that the half-life of cfDNA has been estimated to be between 4 min and 12 h, it can offer a ‘real-time’ snapshot of a tumour profile. In addition to specific disease states, age appears to play a significant role in the total cfDNA present in the bloodstream, likely due to decreased clearance ability with age, rather than increased cfDNA release from dying cells. In either case, the composition of cfDNA appears to be the same in healthy young and old individuals.
### Circulating tumour DNA
When cfDNA is released from tumour cells, it is referred to as ctDNA. While cancer is a disease often characterized by high levels of blood cfDNA, not all cancers shed DNA into the blood at the same rate. Thus, it is likely that disease states differ significantly in their levels of cell death and, therefore, cfDNA release.
### Cell-free mitochondrial DNA
Relative to cfDNA, levels of circulating cf-mtDNA do not appear to vary directly with cell death—in fact, several neurodegenerative diseases feature decreased circulating cf-mtDNA levels—and may play a more complex role in neuroinflammatory pathology.
## Methods used to identify, detect and analyse cfDNA-based liquid biopsy targets
### Biofluid selection and collection
The low concentration of cfDNA in human biofluids, combined with the high potential for contamination by intracellular nucleic acids, necessitates careful handling during sample extraction. Consideration should be given to the biofluid being sampled, as levels of contamination are affected by the specimen type. For instance, while the levels of cfDNA isolated from plasma and serum are similar, serum samples have increased contamination from large DNA fragments and contain lower levels of ctDNA. Overall, plasma appears to be the preferable blood fraction for cfDNA isolation. Additional factors to consider include biofluid volume, accessibility and minimization of potentially painful procedures such as lumbar punctures (LPs).
For all specimen types, it is vital to minimize contamination by sterilizing the collection site and preventing excessive trauma. For blood sampling, anticoagulant-containing blood collection tubes (BCTs) should be employed. Commonly used BCTs include ethylenediamine tetraacetic acid BCTs and cfDNA BCTs (Streck), the latter containing a preservative to prevent cell lysis and allow more extended periods of sample preservation. Following whole blood collection, plasma or serum can be fractioned via centrifugation.
### Cell-free nuclear DNA isolation and purification
A variety of commercial kits and ‘homemade’ bead-based protocols can be used to isolate cfDNA from biofluids. Isolation kits vary in their volume handling abilities and must be chosen to suit the desired application (e.g. array preparation versus PCR amplification). Automated platforms such as QIAsymphony (Qiagen) can be used for high-throughput cfDNA isolation with comparable performance to manual approaches. Concentrations of purified cfDNA can be quantified using fluorometric instruments or assays such as the Quant-iT assay (Thermo Fisher). Instruments such as the 2100 Bioanalyzer (Agilent) can be used to assess sample purity and fragment size distribution.
### Target discovery
Cell-free nuclear DNA expression can be interrogated using microarrays designed to characterize mutational profiles or epigenetic alterations. Commonly used epigenome-wide arrays include the Infinium MethylationEPIC BeadChip (Illumina) and its predecessor, the Infinium Methylation450K BeadChip, covering over 850 000 and 450 000 CpG sites, respectively. These array platforms represent major technical advancements in the field of cfDNA methylation profiling.
While methylation array-based strategies can be appropriate for discovery studies, the low levels of cfDNA in many samples limit such an approach, which requires a relatively high sample input. Additionally, these platforms only cover a small percentage of the total CpG sites in the human genome, and methylation at each site is measured independently, failing to account for methylation haplotype blocks. Indeed, a recent largescale sequencing-based study reported that most cell type-specific, uniquely demethylated CpG regions identified were not represented in the single-CpG Infinium arrays. These characteristics make sequencing-based approaches indispensable, particularly in situations when the CpG target is already known or an exhaustive search for DMRs desired.
In addition to commonly-used next-generation sequencing (NGS) approaches, two novel sequencing techniques employed in epigenetic liquid biopsy studies include whole-genome bisulfite sequencing (WGBS) and cell-free methylated DNA immunoprecipitation and high-throughput sequencing (cfMeDIP-seq). Whole-genome bisulfite sequencing is used to uncover the methylation state of all cytosines. It can be performed on DNA samples as small as 30 ng and provides the methylation state for all cytosines, even in regions of low CpG density and non-CpG sites. The process of WGBS begins when bisulfite salts are used to deaminate unmethylated cytosine residues to uracil, leaving methylated cytosines unchanged; this allows for subsequent PCR amplification and sequencing to identify methylated regions of DNA. One caveat is that most DNA is degraded during this process, though there is no known degradation bias toward methylated versus unmethylated fragments. The newer methodology, cfMeDIP-seq, enriches for methylated cfDNA fragments and allows comprehensive profiling of methylated cfDNA to detect and classify a range of tumours. One advantage of cfMeDIP-seq over WGBS is that it allows for omission of the bisulfite conversion step, thereby preventing significant sample loss.
### Target amplification
#### Quantitative PCR
Quantitative PCR (qPCR) can be adapted to study epigenetic patterns using methylation-specific primer designs targeting loci identified in array and sequencing studies. It can be used for methylation validation by employing techniques such as methylation-specific high-resolution DNA melting (MS-HSM), which quantifies the methylation status of a locus by exploiting the different melting temperatures of methylated and unmethylated DNA.
#### Droplet digital PCR
Droplet-digital PCR (ddPCR) is an ultrasensitive qPCR method that can detect low copy numbers of DNA. This method is unique in that it is based on a system of partitioning DNA molecules into individual droplets before undergoing PCR amplification. Thus, ddPCR is ideal for the analysis of small nucleic acid inputs.
### Cell-free nuclear DNA origin analysis
It is often useful to determine the anatomical origins of cfDNA fragments using epigenetic signatures; examples include investigations of cancers of unknown origin or localization of off-target treatment effects. Unique cellular methylation patterns have been employed to establish rates of death among specific cell types, including oligodendrocytes, from circulating cfDNA. The development of unbiased deconvolution algorithms for DNA and, more recently, cfDNA has been vital to harness the tissue- and cell-specific nature of epigenetic patterning.
In 2018, Moss et al . reported the creation of a methylation atlas for 25 human tissue and cell types of origin for circulating cfDNA. The cfDNA of healthy volunteers was characterized as originating mainly from white blood cells (55%), erythrocyte progenitors (30%), vascular endothelial cells (10%) and hepatocytes (1%). Clinical laboratory tests, namely complete blood count and liver function tests, correlate well with plasma blood cell- and hepatocyte-derived cfDNA proportions, respectively. Focusing on CNS-derived cfDNA, a separate proof-of-concept study showed that increased cfDNA from glia and neurons could be identified in plasma from military personnel trained to work with explosives following occupational training exposure.
Advancing this work, Loyfer et al . recently reported the creation of a comprehensive human DNA methylation atlas, based on deep WGBS, containing a large collection of potential biomarkers that could serve as liquid biopsy targets. Importantly, the resolution of this atlas allows for identification of tissue and cell type transcriptional enhancers and other gene regulatory elements. Thirty-nine cell types were profiled, including neurons and oligodendrocytes. In silico experiments using mixtures of sequenced reads demonstrated that the use of 25 methylation markers per cell type was sufficient to accurately detect DNA from a given source when it constituted only 0.03–0.1% of the total DNA mixture, thus dramatically lowering the signal-to-noise ratio required for target cfDNA detection in biofluids. These results further open the possibilities of applying the liquid biopsy approach to neurological diseases, in which the target cell types of origin contribute to a small minority of overall cfDNA in the bloodstream ( ). Such deconvolution analyses can be reproduced for alternative experimental setups using publicly available GitHub resources.
General strategy for cfDNA-based liquid biomarker discovery. A generalized workflow to identify and validate cfDNA-based liquid biopsy targets consists of the following steps: biofluid selection and collection, cfDNA isolation and purification, target discovery, target amplification and cfDNA origin analysis. While this pipeline is generally followed, a wide variety of specific methods to identify and isolate cfDNA biomarkers are reported in the literature. This variation may account for some of the difficulty in translating cfDNA-based liquid biopsies for neurological diseases to the clinic, especially given the low concentration of CNS-derived cfDNA in blood, the most common substrate for liquid biopsy.
## Recent developments and applications of cfDNA-based liquid biopsies in neurological disease contexts
### CNS tumours
The potential to diagnose and characterize CNS tumours, replace or complement ambiguous imaging studies and monitor treatment response in a minimally invasive manner provides exciting opportunities for liquid biopsy development ( ). In the case of gliomas, the most common group of primary CNS tumours, progress in screening and treatment efficacy remains low. While advances in genetic profiling of tumours allow for intricate diagnosis, a key component of oncologic precision medicine, obtaining a histopathological diagnosis necessitates brain biopsy, a risky and sometimes impossible procedure depending upon the surgical accessibility of the tumour. Brain biopsies have even been implicated in glioma recurrence via disruption of the tumour microenvironment, leading to the initiation of signalling cascades that make the tumour more aggressive. Non-invasive diagnostic approaches, such as CT and MRI neuroimaging modalities, are sensitive to most spinal and intracranial tumours but can be difficult to interpret in certain situations. Pseudoprogression, for instance, occurs when imaging studies suggest progression in patients undergoing other processes such as necrosis or reactive tissue damage; the confusion caused by this phenomenon might be ameliorated by a confirmatory liquid biopsy test. While surgical intervention will remain essential to CNS tumour treatment, liquid biopsies in these contexts have the potential to play a vital role in understanding tumour genetics and spatial heterogeneity, guiding treatment decision-making, monitoring therapeutic response, and surveying for treatment resistance.
Recent developments and applications of liquid biopsy in neurological disease contexts. There is an unmet need for diagnostic, prognostic, and treatment-monitoring liquid biomarkers in a wide variety of neurological diseases. Recent work highlighting potential cfDNA-based liquid biopsy targets in several common neurological conditions is summarized here. Conditions including ( A ) CNS tumours, ( B ) stroke, ( C ) traumatic brain injury, ( D ) Alzheimer’s disease, ( E ) epilepsy, ( F ) multiple sclerosis, and ( G ) neuroinfectious disease each have characteristics that could uniquely benefit from information provided by liquid biopsy.
Given the characteristic mutational burden of many CNS tumours, the literature surrounding liquid biopsies in neuro-oncology primarily focuses on targeted sequencing of known driver mutations in ctDNA. This broad interest in using ctDNA for early tumour detection has prompted investigations of basic ctDNA dynamics, including turnover rates and the relationship between biofluid concentration and tumour location. Stallard et al. demonstrated that CSF-derived ctDNA from patients with diffuse intrinsic pontine glioma (DIPG) could be used to quantify tumour growth by targeting the tumour-specific H3F3A K27M mutation. To simulate ctDNA release into the CSF, a co-culture model of DIPG007 cells and normal human astrocytes was used to show that increased levels of tumour cell proliferation corresponded to increased ctDNA in the culture media, even when the media was changed frequently to approximate the constant production and resorption of CSF. Further experiments found that irradiation of cells resulted in a dramatic increase in ctDNA for ∼72–120 h post-treatment before tapering off, likely due to the rapid death of cells and the concurrent release of ctDNA fragments.
The unique role of the blood–brain barrier (BBB) in CNS tumours and its dynamic permeability in several neurological conditions necessitate a more nuanced understanding of the relationship between tumour location and biofluid cfDNA concentration. Remodelling of the BBB, a known feature of glioma evolution and an outcome of radiation therapy, potentially facilitates the passage of cfDNA into the peripheral circulation. Multifocal CSF sampling from different anatomical reservoirs indicates that closer tumour proximity to the ventricles or subarachnoid cisterns is correlated with higher levels of mutation-containing ctDNA. Correspondingly, direct sampling of CSF from DIPG patients at autopsy found that samples taken directly from the ventricles contained higher levels of K27M-containing ctDNA than samples collected via LP. In comparing the glioma-derived ctDNA content of CSF versus plasma, several studies indicate that CSF is a more sensitive liquid biopsy method for detecting tumour mutations, particularly given the lack of background signal from blood cell-derived cfDNA. However, even if plasma contains a relatively lower concentration of mutation-containing ctDNA, it may be helpful as an alternative sampling route for clinical situations in which an LP is contraindicated (e.g. brain herniation). Urine is another biofluid of interest, with preliminary results suggesting that urine ctDNA fragmentation patterns can be used to detect glioma.
In addition to tumour detection, genotyping of gliomas and identification of mutation-specific therapy options and relevant clinical trials can be accomplished by sequencing CSF-derived ctDNA. Whole exome sequencing of ctDNA collected from glioblastoma patient CSF can be used to detect mutations preoperatively. This is beneficial given that different glioblastoma mutational profiles warrant different treatment strategies; for instance, IDH1 mutant glioblastoma has a survival benefit associated with maximal surgical resection. Mutational characterization of CNS germ cell tumours, paediatric diffuse midline glioma, and paediatric medulloblastoma is also feasible using CSF-derived ctDNA, as is detection of tumour-associated copy number variations. The range of mutations detectable by these methods—including alterations occurring early in tumorigenesis, such as codeletion of chromosome arms 1p/19q, mutations in IDH1 and IDH2 , and amplifications in ERBB2 , MET and EGFR —largely overlaps with those found in traditional tissue biopsy. However, intriguingly, many mutations that are detectable in CSF-derived ctDNA are not regularly detected through tissue biopsy, highlighting the ‘panoramic’ view provided by liquid biopsies, particularly for highly heterogeneous tumours like glioblastoma. The presence of ctDNA in glioma patient CSF is associated with an increased risk of death, independent of IDH status, tumour burden at the time of LP, or extent of resection; however, no significant associations have been observed between ctDNA-positive CSF and tumour grade, disease duration, or prior therapy.
Liquid biopsy of CSF-derived ctDNA has a sensitivity advantage over CSF cytology, given that many glioma patients with detectable CSF ctDNA do not have detectable CSF tumour cells. Additionally, prospective clinical studies have found that CSF-derived ctDNA contains more variants and higher variant frequencies than genomic DNA extracted from circulating tumour cells, making it superior for molecular profiling. Plasma- and serum-based ctDNA sequencing can also provide information about glioma mutational profile, including driver mutations such as BRAF .
Advances in characterizing ctDNA methylation patterns have expanded the possibilities of using liquid biopsies for the clinical management of CNS tumours, especially for tumours with a low frequency of oncogenic mutations. Global methylation markers represent one way of monitoring epigenetic cfDNA alterations in disease. Given its high abundance in DNA, hypomethylation of the Alu element—a short, interspersed genomic element—has been used as a global marker of methylation status. Significantly lower methylation levels in the Alu element can be detected in the serum-derived ctDNA of glioma patients, effectively separating them from healthy controls and patients with benign intracranial tumours using a liquid chip assay. In addition to distinguishing cases of glioma, methylation of the Alu element in ctDNA is negatively correlated with glioma severity and positively correlated with survival. However, a rigorous survey of Alu methylation levels in other diseases has not been completed and therefore it may not be a specific glioma marker.
Disease-specific methylation changes can also be gleaned from ctDNA analysis. For instance, the MCPH1 promotor in serum-derived ctDNA correlates with both glioma size and grade. Epigenetic signatures, based on patterns of DMRs present in CSF-derived ctDNA from medulloblastoma, can be used for tumour detection and monitoring of treatment effects. In addition to ctDNA methylation patterns, hydroxymethylation patterns identified via highly sensitive chemical labelling of plasma ctDNA have shown promise in distinguishing glioblastoma from lower-grade astrocytomas.
Epigenetic characteristics of ctDNA can also be used to discriminate between intracranial CNS tumour types, an impressive feat given that these tumours often have similar cell types of origin. Nassiri et al . used cfMeDIP-seq to characterize the epigenetic cfDNA profiles of plasma samples from patients with diffuse gliomas, patients with extracranial tumours and healthy controls. DMRs identified in this analysis were used to build a classifier capable of assigning a methylation score to each sample. This score could differentiate gliomas from other cancer types and healthy controls and, when calculated for over 100 additional CNS tumour samples, accurately discriminated between gliomas, meningiomas, hemangiopericytomas and brain metastases. A similar study, this time using a methylation array, found that the epigenetic signatures present in glioma-derived ctDNA isolated from patient CSF and plasma can be used to establish an initial diagnosis prior to surgical intervention. To this end, a machine-learning-based model was used to develop a glioma-epigenetic-liquid biopsy (GeLB) score to indicate the likelihood that a given serum sample is from a glioma patient. While this score was shown to have high sensitivity and specificity in recognizing patients during initial disease, most patients had decreased GeLB scores at recurrence, possibly due to a treatment-induced DNA methylation shift in glioma cells. Recent work has also shown that genetic alterations attributable to clonal haematopoiesis can be detected in the plasma cfDNA of patients with a history of temozolomide chemotherapy. These findings highlight another challenge in tumour biomarker development: the continual evolution of tumours themselves.
Liquid biopsy developments in neuro-oncology have been reviewed extensively elsewhere.
### Stroke
The importance of rapid stroke diagnosis cannot be overstated, given that each minute without intervention corresponds to increased brain tissue damage. The two main types of stroke, ischaemic and haemorrhagic, have fundamentally different treatment approaches, highlighting the importance of accurate aetiological determination. Current diagnosis of acute stroke relies heavily on neuroimaging techniques, including CT and MRI. While acute imaging and angiography will undoubtedly remain vital for in-hospital treatment decision-making—for example, to rule out haemorrhagic stroke and to detect large-vessel intracranial occlusions amenable to mechanical endovascular thrombectomy—these strategies may not be sensitive enough to catch early-stage ischaemic strokes when they are most receptive to intervention via intravenous thrombolysis and/or thrombectomy, and to regaining function. Indeed, approximately half of acute stroke or transient ischaemic attack (TIA) patients lack detectable abnormalities in initial imaging studies. While MRI provides a greater sensitivity for detecting ischaemic stroke when compared to CT, it is not used in many emergency room settings due to real or perceived time constraints as well as hardware availability in less resource-rich settings. Additionally, imaging can be costly and may not be suitable for restless or acutely distressed patients. Thus, there is a distinct need for clinically validated, rapidly assayed blood biomarkers to assess ischaemia severity in stroke, similar to those available for assessment of myocardial infarction.
While several serum protein biomarkers have been associated with stroke, including S100, NSE and CRP, among others, these markers only become elevated several hours following symptom onset and are therefore not ideal for rapid biomarkers. Some protein markers, including GFAP in the case of haemorrhagic stroke, are released into the bloodstream much faster; however, variability in diagnostic performance has limited their progression to clinical use. Simple cfDNA-based tests that could be used by emergency personnel in the field or in the ambulance might allow for immediate initiation of treatment, such as administration of blood pressure lowering medications, and more informed decision-making regarding transport to qualified stroke centres. Even when compared to protein biomarkers at the same timepoints post-stroke, cfDNA has been found to provide sensitive, independent prognostic information, especially in the case of severe stroke outcomes.
In acute ischaemic stroke (AIS), the most common stroke subtype, lack of perfusion quickly leads to death of brain cells and disruption of the BBB. Therefore, it has been hypothesized that this post-injury environment is characterized by cfDNA released into the peripheral circulation by cells undergoing necrosis and apoptosis. This hypothesis is supported by the finding that the total amount of circulating cfDNA measured at the time of AIS patient admission correlates with both stroke severity (assessed using the National Institutes of Health Stroke Scale) and medium-term prognosis (assessed using the modified Rankin Scale at three months post-infarction). Furthermore, circulating cfDNA threshold values can be used to differentiate actual AIS patients from stroke mimics and to classify the patients most likely to experience neurological improvement following mechanical thrombectomy or intravenous thrombolysis. Notably, these increased cfDNA levels in stroke patients may signify intraparenchymal brain damage, increased damage to the BBB, or both. Interestingly, the temporal dynamics of circulating cfDNA differ between ischaemic and haemorrhagic stroke. Mirroring the two different pathologies, haemorrhagic stroke results in a rapid and temporary increase in plasma cfDNA, whereas cfDNA levels in ischaemic stroke increase in parallel with hypoxia- and reperfusion-induced cellular damage over the course of days. These differences may be useful in developing rapid cfDNA-based tests for grouping patients with stroke symptoms based on aetiology and prognosis.
Given the frequency of pre-existing cardiovascular conditions in stroke patients, tissue-of-origin analysis of cfDNA samples is particularly relevant. For instance, it is possible that alterations in stroke patient cfDNA levels are driven by nucleic acid release from endothelial cells or activated platelets, as opposed to dying brain cells. The finding of elevated levels of such cfDNA fragments might serve as a ‘proxy’ for the identification of concurrent or subsequent pathological processes in a patient. In this way, stroke presents a particularly challenging scenario for standardized biomarker development, given its many different causes and the frequency of patient comorbidities. Despite these challenges, there remains a clinical need for a rapid (results within minutes) blood test to diagnose stroke and guide treatment decision-making, either in combination with current imaging studies or as a stand-alone test in low-resource settings that lack access to imaging equipment or trained imaging technicians. Additionally, such blood tests may be used to monitor treatment effects as a measure of ongoing CNS cellular damage, to predict long-term outcomes and even to guide the pursuit of rehabilitation regimens.
Additional liquid biopsy developments relating to stroke are reviewed elsewhere.
### Traumatic brain injury
Severe traumatic brain injury (TBI) is a critical neurological condition that contributes to death and disability in all age groups, and patient clinical courses can be highly unpredictable. Secondary brain injuries due to TBI such as elevated intracranial pressure can cause additional long-lasting neurological damage beyond the initial insult. Even patients with mild TBI can develop post-concussion syndrome (PCS) characterized by symptoms that often last for months after the initial insult. These include personality changes and cognitive difficulties that can affect employment and personal relationships. Despite the evaluation of several protein biomarkers for TBI, no broadly applicable TBI damage marker has been identified. Therefore, TBI patient outcomes would benefit from rapid and accurate classification of the extent of tissue damage, allowing for proper resource allocation and appropriate clinical interventions.
Overall levels of post-injury serum cfDNA positively correlate with the severity of TBI. The decrease in overall cfDNA serum levels at 24 h post-admission has been shown to be important for predicting TBI patient outcomes, with the diminution ratio of serum cfDNA being predictive of fatal outcomes. Additionally, cfDNA levels have been shown to predict disability post-injury in mild TBI. A prospective pilot study using a fluorescence assay to measure cfDNA levels (correlating highly with a traditional qPCR assay of β-globin) in the serum of mild TBI patients found that overall cfDNA levels at hospital admission positively correlated with levels of cognitive impairment after adjusting for age and education level. This method has been proposed as a potential rapid screening system in emergency care settings to determine patients at risk of PCS. Another strategy using an alternating current electrokinetic microchip to test a variety of plasma biomarkers, including cfDNA levels, in mild TBI patients has also shown promise in predicting the risk of PCS-associated symptoms. Such real-time screening might enable rapid allocation of medical and social support, including regular cognitive testing and social services, to those patients who need them most. However, the use of such diagnostic tests should also be examined in patients with multitrauma, given that a significant percentage of TBI cases occur alongside other bodily injuries (e.g. car accidents, falls, assaults), leading to widespread tissue damage.
One relatively common complication of TBI is intracranial haemorrhage (ICH). Current methods of monitoring ICH, including clinical exams, neuroimaging and intracranial pressure monitoring, are limited in their ability to accurately assess damage and predict prognosis. Plasma-derived cfDNA levels, measured using β-globin expression, have been found to positively correlate with patient Acute Physiology and Chronic Health Evaluation score and inversely correlate with patient Glasgow Coma Scale, indicating a relationship between overall cfDNA levels and ICH patient clinical status. Additionally, cfDNA samples taken on Day 3 of intensive care unit admission positively correlated with the duration of stay in intensive care, though this marker was not as predictive of patient outcome as the protein biomarker S100β. This increased cfDNA is thought to arise from neurons and glia damaged during CNS injury; indeed, investigation of epigenetic signatures indicates that demethylation at the brain-specific CpG CG09787504 can be readily detected in circulating cfDNA of patients after TBI, thus reporting on the rate of brain cell death.
Additional liquid biomarkers of TBI are reviewed elsewhere.
### Alzheimer’s disease
The leading cause of dementia in the elderly is Alzheimer’s disease, a neurodegenerative disease characterized by loss of function primarily in the cognitive, language and memory domains. Alzheimer’s disease is one of several neurodegenerative conditions characterized by abnormal misfolded protein aggregation. Centering on the proteins highly associated with disease pathogenesis, an array of Alzheimer’s disease imaging and liquid protein biomarkers have been identified and validated. However, nucleic acid-based biomarkers for Alzheimer’s disease are less developed. With escalating pressure to identify a variety of circulating biomarkers of Alzheimer’s disease for use in both drug development and clinical practice, important questions have arisen involving specific biomarker contexts of use and relevance to disease pathology.
Alzheimer’s disease is diagnosed primarily using clinical criteria and often begins with a prodrome involving mild cognitive impairment (MCI). Imaging modalities, including MRI and PET, can be used to confirm Alzheimer’s disease pathogenesis and even to identify cases during the prodromal period. Recent advancements in techniques that can identify presymptomatic disease and predict cognitive decline, including amyloid- and tau-based PET radiotracers, are particularly important given that disease progression may begin decades prior to clinically noticeable symptoms. Protein-based liquid biopsies of the CSF and blood have also demonstrated good diagnostic and prognostic performance by targeting molecules, such as amyloid-beta, tau, NfL and GFAP. Plasma tau is already being investigated in clinical trials and implemented in clinical practice as a predictor for future development of Alzheimer’s dementia in patients with MCI. Given the overall failure of disease-modifying therapies at treating symptomatic Alzheimer’s disease patients, it is likely that any progress to be made in slowing or preventing Alzheimer’s disease pathogenesis will occur by treating patients at the earliest stages of the disease, emphasizing the importance of biomarkers for early detection.
In addition to blood- and CSF- derived protein biomarkers of Alzheimer’s disease, such as the amyloid-β (Aβ) /Aβ ratio, BACE1 enzyme activity, total-tau, phosphorylated-tau and NfL concentrations, putative cfDNA biomarkers of Alzheimer’s disease have begun to be identified. Given that LPs can be particularly challenging and distressing to perform in the elderly and those with severe cognitive deficits, many studies use peripheral blood as the biofluid substrate. Cell-free nuclear DNA with hypermethylated LHX2 , an epigenetic pattern seen specifically in neural tissue, have been put forth as a marker of Alzheimer’s disease. Patterns of 5-hydroxymethylation are less commonly studied than 5-methylcytosine modifications in the context of liquid biopsy; however, their presence in serum-derived cfDNA could distinguish patients with late-onset Alzheimer’s disease from cognitively normal individuals in a recent case-control study using genome-wide profiling.
Contrasting the progress in developing protein-based MCI biomarkers, the search for cfDNA-based MCI biomarkers has proven more difficult. A study of Finnish twin pairs found no significant epigenetic cfDNA markers associated with episodic memory impairment, possibly due to the low genomic coverage provided by plasma cfDNA fragments and the variation in cfDNA concentrations between individuals. The search for markers of MCI is particularly challenging due to the mildness of disease pathology compared to more severe neurodegenerative conditions, which might lead to less cfDNA being released from damaged neurons and, therefore, a blood signal that is harder to detect.
Both positive and negative predictive biomarkers may be helpful during Alzheimer’s disease drug development as part of biomarker-drug codevelopment programs. Such programs are particularly relevant given the high failure rate of putative Alzheimer’s disease drugs during clinical trials. Compared to CSF or neuroimaging markers, the use of easily accessible blood biomarkers might increase the number of sites able to participate in clinical trials and the number of patients able to be enrolled, thus expanding the statistical power of trials. This arrangement would constitute liquid biopsy-guided drug development. Biomarker panels may be used to stratify patients and help account for the heterogeneity of disease present in conditions like late-onset Alzheimer’s disease. Liquid biopsy findings may be further used to ‘stage’ complex neurodegenerative conditions like Alzheimer’s disease through the use of longitudinal studies enrolling cognitively healthy individuals at risk of developing Alzheimer’s disease and following them for several years.
At present, in the case of Alzheimer’s disease, the specificity and clinical utility of protein-based liquid biopsies is greater than those targeting circulating DNA. However, the development of cfDNA-based liquid biopsies in Alzheimer’s disease may still be advantageous in certain circumstances, at minimum due to their potential to serve as auxiliary biomarkers. For instance, disturbances in the homeostatic proportions of brain-derived cfDNA, identified via differential methylation profiling, may precede frank protein accumulation in the CSF or blood and therefore serve as an earlier disease marker. Furthermore, detection of the loss of specific types of neurons using characteristic DNA methylation patterns present in cfDNA may allow for a more nuanced understanding of neuronal vulnerability and disease pathogenesis compared to broad markers of neurodegeneration, such as NfL. This feature could assist in defining subtypes of disease affecting different neuronal populations, for example. Additionally, liquid biopsy assays targeting proteins in the blood can suffer from significant variation due to both degradation of brain-derived proteins by circulating proteases and non-specific assay detection of endogenous antibodies. This variation might be mitigated by the addition of non-protein biomarkers into diagnostic and prognostic frameworks. Finally, cfDNA-based liquid biopsies relevant to other related neurodegenerative diseases, such as frontotemporal dementia, would benefit from further research into identifying cfDNA targets in dementia.
Additional recent liquid biopsy developments relating to Alzheimer’s disease, and MCI are reviewed elsewhere.
### Epilepsy
Epilepsy diagnosis is complex and presents unique challenges due to the temporary nature of seizure activity, which is often unwitnessed. While EEG and MRI assessments can aid diagnosis, molecular biomarkers could provide an effective way to confirm the diagnosis based on the clinical history, determine treatment regimens, and identify patients with potentially drug-resistant epilepsy. Neuronal death has been observed in patients with epilepsy following seizure activity and in animal models of induced status epilepticus, signifying the potential for increased release of cfDNA into circulating biofluids.
Despite its potential as a diagnostic and prognostic biomarker, studies of cfDNA in the context of epilepsy are limited. In one prospective study of focal epilepsy patients, cfDNA levels were reported to be higher in symptomatic refractory focal epilepsy compared to healthy controls. However, in a study of refractory epilepsy patients undergoing video-EEG monitoring, it was found that overall serum-derived cfDNA levels were lower in extratemporal lobe epilepsy compared to controls. Additionally, patients with a longer duration of epilepsy (18+ years) had lower cfDNA levels (measured as the difference between baseline and peak levels) compared to patients with a shorter history of epilepsy. These results indicate that cfDNA levels do not directly correlate with seizure burden and emphasize the need for further studies to clarify cfDNA differences between epilepsy subtypes and over the course of chronic disease.
Liquid biopsy may be useful in the molecular diagnosis of genetic causes of epilepsy, particularly in cases of intractable epilepsy due to brain malformations. It is possible that such an approach would decrease the need for brain tissue biopsy, allow for a more personalized approach to treatment based on patient mutational status, and even provide predictions of seizure relapse. A recent proof-of-principle study shows that somatic brain mutations can be identified by sequencing brain-derived cfDNA isolated from CSF samples. In another study, cfDNA was isolated from CSF collected during epilepsy surgery and used to detect somatic mutations consistent with those found by analysing resected tissue. Droplet digital PCR was used to detect several variants encompassing a wide array of pathological diagnoses; however, only a minority of patients had cfDNA mutations detectable in their collected CSF, indicating the need for substantial technical optimization and testing prior to clinical implementation of such a strategy.
Additional recent liquid biopsy developments in epilepsy are reviewed elsewhere.
### Multiple sclerosis
Multiple sclerosis is a neuroinflammatory and neurodegenerative condition that is a major cause of neurological disability, particularly in young adults. Its diverse symptomatology is caused by the development of demyelinating lesions in various regions of the brain and spinal cord that occur dynamically over time and space. The regenerative process of remyelination is often able to repair these lesions partially or completely in the early stages of the disease; however, this mechanism tends to fail at later stages of the disease, resulting in persistent lesions with increased damage to the surrounding tissue, axonal degeneration and progressive clinical course. Much remains to be understood regarding the molecular and cellular factors contributing to successful remyelination, how the potential for remyelination varies between individual patients and in different neuroanatomical locations, and how age might disrupt this process. Additionally, biomarkers for risk of neurodegeneration in multiple sclerosis patients are lacking.
A major roadblock to evaluating potential regenerative multiple sclerosis therapies is the lack of ability to accurately track remyelination. To overcome this challenge, liquid biopsy tracking of oligodendrocyte lineage cell dynamics may provide a means of monitoring disease burden and even estimating remyelination activity. The development of liquid biopsies targeting molecular signatures of demyelination might also be useful for other demyelinating conditions, such as progressive multifocal leukoencephalopathy and acute disseminated encephalomyelitis. In multiple sclerosis, concurrent surveillance of patients with liquid biopsy may help clarify MRI findings, or vice versa, and reduce the number of imaging sessions needed for the patient, saving expense and time. Future longitudinal liquid biopsy studies of patients with relapsing-remitting multiple sclerosis may be useful for identifying prognostic signatures present in patients who eventually develop progressive disease.
MOG , a gene expressed specifically in mature oligodendrocytes, is demethylated in mouse and human oligodendrocytes when compared with other cell types. Measuring differential methylation of CNS-specific MOG cfDNA in plasma represents a viable way of quantifying oligodendrocyte death in rodent models and relapsing-remitting multiple sclerosis patients, without the need for sampling CSF. Similarly, cfDNA containing unmethylated MBP3 and WM1 , also specific for oligodendrocytes, can be detected in the plasma of about 75% of patients with active (relapsing) multiple sclerosis, but not in patients with stable (remitting) disease. This may be due to the relapsing/remitting nature of multiple sclerosis, in which temporary inflammatory disruptions of the BBB might affect detection of cfDNA released from brain cells into the bloodstream. Differential methylation patterns of genetic elements beyond gene bodies and promotor regions can also be used to differentiate multiple sclerosis patients from healthy controls. For example, long interspersed nuclear element-1 is an autonomous retrotransposon containing CpG sites that are significantly hypermethylated in the serum-derived cfDNA of relapsing-remitting multiple sclerosis patients.
Investigations of cf-mtDNA in the CSF of multiple sclerosis patients have found increased levels in both relapsing-remitting multiple sclerosis and progressive multiple sclerosis, the latter being associated with brain atrophy. In contrast, a separate study of post-mortem progressive multiple sclerosis patient ventricular CSF targeting two regions of cf-mtDNA ( MTND1 and MTND4 ) found that decreased cf-mtDNA copies were associated with progressive multiple sclerosis. This study did not observe a correlation between cf-mtDNA and known neurodegenerative protein markers GFAP and S100β. This discrepancy may be due to living versus post-mortem sampling of CSF or the use of patients at a later stage of disease. However, it highlights the need for longitudinal studies over the course of complex and dynamic conditions like multiple sclerosis, as cf-mtDNA levels may fluctuate significantly during different disease stages.
### Neuroinfectious disease
Because CNS infections are often found in low-resource settings, testing for microbial nucleic acids present in readily accessible biofluids represents a promising strategy for rapid and inexpensive diagnosis. Additionally, this approach may allow for quick tailoring of treatment, lessening the side effects of prolonged empiric treatment. The clinical workflow for managing infectious diseases already employs biofluid collection: for example, the use of blood cultures and CSF analysis. However, cultures can take days to weeks to yield identification of the causative organism, and CSF analysis can fail to detect rare pathogens. Thus, the existing framework may be used to add additional targeted amplification of nucleic acid sequences released into body fluids by neuropathogens.
Parasites are a common cause of endemic CNS infectious disease around the globe. For example, neurocysticercosis is caused by ingestion of the Taenia solium (pork tapeworm) parasite, which forms cysts in the brain that can lead to the onset of seizures and other severe neurologic signs. Current diagnostic approaches for this urgent condition include imaging studies focused on identifying cysts, and detection of circulating parasite antigens or antibodies in the plasma or CSF. T. solium has been detected in patient CSF and urine samples using PCR primers targeting the pTsol-9 gene with subsequent amplicon sequencing to confirm positivity.
Another neurological consequence of parasitic infection is cerebral malaria, a life-threatening condition that can be unpredictable in onset and often affects young children infected with Plasmodium falciparum . A large-scale retrospective study examining plasma from Malawian children identified a straightforward potential predictive biomarker for the development of cerebral malaria: total plasma cfDNA (the sum of host and parasite cfDNA, of which the host makes a much larger contribution). This finding demonstrates that cfDNA released during host response to parasitic infection, as opposed to cfDNA released by the organism itself, can serve as a robust biomarker for disease severity and may allow for triage of at-risk patients.
Even with the current standard of CSF collection to diagnose cases of infectious meningitis and encephalitis, pathogen levels in the CSF can be below the limit of detection. This is often the case in tuberculosis meningitis (TBM), a life-threatening illness and one of the most serious manifestations of tuberculosis. In these circumstances, amplification of nucleic acid fragments like cfDNA can be used to detect microbes. Detection of Mycobacterium tuberculosis cfDNA (specifically, the IS6110 sequence) in the CSF has been reported to be more sensitive than traditional diagnostic methods, including microscopy, mycobacterial culture and the GeneXpert MTB/RIF Ultra assay, with similar specificity. A prospective multicentre study comparing GeneXpert to cfDNA assay for TBM found both to have optimal sensitivities and specificities (>90%), indicating that both approaches are useful for TBM diagnosis.
Circulating microbial cfDNA can be used to diagnose CNS infection when cultures and PCR testing for common organisms fail, as is demonstrated by a recent case study featuring a patient with a rare form of bacterial meningitis. Importantly, the diagnostic method was not impacted by antibiotic exposure, an essential characteristic of diagnostic tests for severe infections requiring immediate treatment. While the methodology used in this instance—applying NGS to cfDNA isolates from blood samples—is too expensive and time-consuming to be translatable to most clinical settings, it demonstrates the potential utility of identifying microbe-specific cfDNA signatures. A cfDNA-based liquid biopsy approach might be useful in cases where culture time is lengthy and causes a delay in the transition from empiric to narrow-spectrum treatment. This point also pertains to the rapidly growing problem of antibiotic resistance worldwide.
Viral diseases affecting the CNS have also been investigated using cfDNA-based liquid biopsy. In a large-scale study of the CNS HIV Antiretroviral Therapy (ART) Effects Research cohort, patients currently on ART or with a history of ART were found to have decreased levels of cf-mtDNA in their CSF samples. While levels of cf-mtDNA in the CSF were not associated with neurocognitive performance in HIV patients, cf-mtDNA levels were inversely correlated with proteins necessary for iron metabolism and angiogenesis, hinting at novel mechanisms of CNS damage due to HIV.
Beyond diagnosis of primary neuroinfectious disease, profiling of molecular by-products from microbes could enhance understanding of the underlying aetiologies of neuroinflammatory and neurodegenerative diseases. For instance, the growing consensus that Epstein–Barr virus (EBV) is part of the causal chain in the development of multiple sclerosis could be further investigated by targeting EBV-derived cfDNA released into patient CSF or plasma and comparing changes in these levels to neuroimaging biomarkers and patient clinical status.
Alzheimer’s disease has also been associated with multiple human herpesviruses, though the burden of proof is exceptionally high for claims that infection underlies the convoluted pathogenesis of Alzheimer’s disease. Indeed, a recent multicohort study found no difference in human herpesvirus 6 expression between healthy and non-Alzheimer’s disease control brains, suggesting that continuous viral pathogenesis is unlikely to be a primary contributor to this disease, but not ruling out other mechanisms by which viruses may influence or underlie Alzheimer’s disease pathology. A deeper understanding of viral cfDNA dynamics in relation to lesion and plaque development, and neuroaxonal degeneration, may prompt new insights into the complex aetiologies of these conditions.
## The future of cell-free nuclear DNA-based liquid biopsy in neurological disease
In the future, combining multiple biomarkers may allow for more specific application of liquid biopsies, for instance, by targeting combinations of hypo- and hyper-methylated cfDNA regions ( ). Adding more loci per target region can enhance the power of the DNA methylation signature to serve as an accurate and stable biomarker of disease; doing so has already been observed to increase discriminative power by reducing interindividual phenotype-independent variation of DNA methylation levels and by increasing the accuracy of the tissue-of-origin assignment.
Future directions for liquid biomarker application in neurology. ( A ) Design considerations, such as combining multiple biomarkers or epigenetic loci, will allow for more specific application of liquid biopsies. ( B ) Logistical considerations, including reducing biofluid volumes necessary for biopsy, cost and turnaround time will allow for broader application of liquid biomarkers. ( C ) Translational considerations, including the execution of well-powered randomized controlled trials with subjects fitting target patient demographics and comorbidity profiles, will be necessary for the safe and efficient application of neurology-focused liquid biopsies. ( D ) Clinical considerations, such as screening and diagnosis, must be defined for each validated liquid biomarker based on proven sensitivity and specificity characteristics.
Future tests must be sensitive enough to use small volumes of biofluids with low levels of cfDNA, and, as with any translational approach, cost and timing must be considered. Tests must have a quick enough turnaround time to make them clinically useful during real-time medical decision-making. For markers that have reached a consensus in the pre-clinical realm, well-powered randomized controlled trials must be conducted for biomarker validation. Real-world patients are often different from those included in initial studies: exclusion criteria must be carefully selected while attempting to replicate real-world scenarios. For instance, many stroke patients also have hypertension or coronary artery disease. Thus, excluding patients who have these comorbidities might lead to wasted time and resources spent studying a stroke biomarker that may not be useful in the majority of stroke patients.
Additionally, it is vital to consider the potential clinical utility of such liquid biopsies: at what point in the process of screening, diagnosis, prognosis and treatment would they be useful? Are some of these markers more suited for screening versus diagnosis? How would the use of these markers be beneficial in place of (or in conjunction with) the current standard of care? Are certain tests better for discriminating between diseased versus healthy controls, whereas others may be more useful for classifying disease subtypes following initial diagnosis by conventional means?
One particularly exciting aspect of liquid biopsies is their ability to probe the anatomical, tissue and cellular origins of disorders with complex aetiologies. As seen in this review, cfDNA-based liquid biopsies for neurological disease need not only be based on material released from neuronal or glial cells themselves; in some cases, specific diseases are characterized by up- or downregulated levels of cfDNA released by other cell types, such as blood cells or endothelial cells. A deeper understanding of the role of cfDNA release during disease states may provide clues to underlying aetiologies and the dynamics of disease progression. Targeted sequencing of known mutations present in head and neck cancer-associated ctDNA has already been leveraged to create a multi-cancer diagnostic panel with the potential for liquid biopsy application, and high-sensitivity, patient-specific, ctDNA fingerprint panels for several different cancers have been developed to monitor treatment response, recurrence and drug-related mutations. Creating multi-disease liquid biomarker panels for other neurological conditions could aid in rapid discrimination between similar disease presentations, allowing for earlier initiation of treatment and preventing unnecessary, expensive and invasive diagnostic procedures. Analysis of serial liquid biopsies has the potential to provide real-time monitoring of disease pathogenesis, which is critical for optimal management of complex illnesses over time. Finally, liquid biomarkers may be leveraged to gain insight into the effects of traditional and experimental treatments on individual patients, providing a valuable method for personalizing medical care.
## Conclusion
In clinical specialties focusing on neurological diseases, there is a significant unmet need for non-invasive, sensitive and specific ways to evaluate tissue damage and, by proxy, the presence of disease, whether occult (as in the case of early diagnosis and screening) or ambiguously symptomatic (as in the case of conditions like headache or dementia). If liquid biopsies, particularly those focusing on epigenetic signatures as targets, live up to their predicted potential, it is likely that they will be able to be applied to numerous neurological diseases at a variety of stages, from early diagnosis to monitoring effects of treatment, to predicting relapse, to serving as outcome measures during clinical trials for novel therapies. Furthermore, liquid biopsies may prove useful in further subtyping complex neurological conditions, such as stroke or Alzheimer’s disease, to better understand their aetiologies and develop more targeted treatment regimens. Lab-on-a-chip technologies could make liquid biopsies even more useful in clinics or low-resource settings, particularly in areas of the world where expensive imaging and tissue biopsy procedures are difficult to obtain. Multiplex assays may result in tests with higher sensitivity, such as in the case of cfDNA targets for which multiple informative fragments can be detected.
As with all efforts to validate diagnostic tests, preliminary studies in patients with neurological disorders must be followed by observational studies and clinical trials to determine whether the test can detect the disease in its prodrome or early stages. Likewise, prognostic tests and those designed to monitor treatment effects must be interrogated in appropriate longitudinal studies. In this way, putative biomarkers discovered in basic science settings can be applied to real-life clinical scenarios. Ultimately, the future of cfDNA-based liquid biopsies in neurological diseases will rely on further developments in the identification, detection and analysis of specific disease-associated genetic and epigenetic patterns along with universal standardization of these workflows. Before these technologies and workflows can be implemented in clinical settings, they must be engineered to account for the complexity of actual patients with comorbid conditions and heterogeneous disease pathologies. If the pace of recent advances in this field continues, liquid biopsies targeting informative cfDNA profiles in various readily accessible biofluids will represent a valuable tool for clinicians in diagnosing and caring for patients with a wide range of neurological conditions.
|
Memory deficits are a debilitating symptom of epilepsy, but little is known about mechanisms underlying cognitive deficits. Here, we describe a Na channel-dependent mechanism underlying altered hippocampal dendritic integration, degraded place coding and deficits in spatial memory.
Two-photon glutamate uncaging experiments revealed a marked increase in the fraction of hippocampal first-order CA1 pyramidal cell dendrites capable of generating dendritic spikes in the kainate model of chronic epilepsy. Moreover, in epileptic mice dendritic spikes were generated with lower input synchrony, and with a lower threshold. The Na 1.3/1.1 selective Na channel blocker ICA-121431 reversed dendritic hyperexcitability in epileptic mice, while the Na 1.2/1.6 preferring anticonvulsant S-Lic did not. We used in vivo two-photon imaging to determine if aberrant dendritic excitability is associated with altered place-related firing of CA1 neurons. We show that ICA-121431 improves degraded hippocampal spatial representations in epileptic mice. Finally, behavioural experiments show that reversing aberrant dendritic excitability with ICA-121431 reverses hippocampal memory deficits.
Thus, a dendritic channelopathy may underlie cognitive deficits in epilepsy and targeting it pharmacologically may constitute a new avenue to enhance cognition.
Masala et al. describe a Na channel-dependent mechanism underlying altered hippocampal dendritic integration, degraded place coding, and deficits in spatial memory. Dendritic channelopathy may underlie cognitive deficits in epilepsy, and targeting it pharmacologically may constitute a new avenue to enhance cognition.
## Introduction
In most CNS pyramidal neurons, dendrites are capable of generating spikes in local membrane potential that are initiated by dendritic voltage gated Na channels. Because dendritic spikes are elicited by spatiotemporally clustered inputs, arising only if specific ensembles of presynaptic neurons are synchronously active, they have been proposed to endow neurons with the capability to act as input feature detectors. Indeed, dendritic spikes have been found to be relevant for triggering place-related firing in CA1 neurons. They thus constitute a key mechanism for dendritic integration and neuronal input–output computations, and have been strongly implicated in spatial navigation.
Dendritic spikes rely on targeted expression of voltage-gated ion channels in dendritic branches. In epilepsy, as well as in numerous other CNS disorders, the expression and function of ion channels are profoundly altered in CA1 neuron dendrites. In chronic epilepsy models, changes in K channels, T-type Ca channels, HCN channels and Na channels have been identified. However, these studies have mainly focused on larger calibre, apical dendrites of pyramidal neurons, primarily because of the difficulties in obtaining direct patch-clamp recordings from thin dendrites. The integrative properties of thin, higher-order dendrites and how they change in chronic epilepsy have not been studied so far.
In this study, we address how dendritic integration via dendritic spikes is altered in chronic epilepsy, how this affects neuronal coding in vivo and how it impacts behaviour. We propose that an upregulation of Na channels in hippocampal pyramidal neuron apical oblique dendrites underlies altered dendritic spikes, degraded place coding and cognitive deficits in experimental temporal lobe epilepsy (TLE).
## Materials and methods
### Animals
All experiments followed institutional guidelines of the Animal Care and Use committee of the University of Bonn. To avoid the documented sex differences in the severity of comorbidities and seizures in the kainate model of epilepsy and thereby reduce animal numbers, all in vitro , and behavioural experiments were performed on C57BL/6J male wild-type mice (Charles River). For in vivo two-photon imaging experiments, we used male Thy1-GCaMP6 mice, which express GCaMP6s in most hippocampal neurons (C57BL/6J-Tg(Thy1-GCaMP6s)GP4.12Dkim/J; Jackson Lab Stock No.: 025776 ). Mice were kept under a light schedule of 12 h on/12 h off, constant temperature of 22 ± 2°C and humidity of 65%. They had ad libitum access to water and standard laboratory food at all times. All efforts were made to minimize animal suffering and to reduce the number of animals used.
### Kainate model of epilepsy
The kainate model of epilepsy was induced largely as described. Specifically, for the experiments 5 week-old mice were injected with analgesic ketoprofen [Gabrilen, Mibe; 5 mg/kg body weight (b.w.); injection volume 0.1 ml/10 g b.w., s.c.] diluted in H O for injections (Ampuwa, Fresenius Kabi Deutschland) 30 min before injecting the anaesthetic. Mice were anaesthetized using a mixture of ketamine (Medistar; 80 mg/kg b.w.) and medetomidinhydrochloride (Domitor, Orion Pharma; 1.2 mg/kg b.w.; injection volume 0.1 ml/10 g, i.p.) and placed in a stereotaxic frame in a flat skull position. Eyes were covered with eye ointment (Bepanthen, Bayer) to prevent drying and body temperature was maintained at 37°C using a regulated heating plate (TCAT-2LV, Physitemp) and a rectal thermal probe. The surface was locally anaesthetized with a drop of 10% lidocaine and after 3–5 min residual soft tissue was removed from the skull bones with a scraper and 3% H O /NaCl solution. After complete drying, the cranial sutures were clearly visible and served as orientation for the determination of the drilling and injection sites. For stereotactic injection, a hole was carefully drilled through the skull with a dental drill, avoiding excessive heating and injury to the meninges. Any minor bleeding was stopped with a sterile pad. A stainless steel cannula (0.5 mm outer diameter) connected to a 10 μl microsyringe (Hamilton) was filled with a 20 mM kainic acid (KA) solution (Sigma) in 0.9% sterile NaCl (Fresenius Kabi Deutschland) and positioned above the right dorsal hippocampus (anteroposterior −1.9 mm; mediolateral −1.5 mm; dorsoventral, −1.7 mm) with bregma as a reference. Injections of dextran-biotin (50 nl) marker showed that these coordinates correspond to upper border of the CA1 region of the dorsal hippocampus. Once the cannula reached the correct depth, it was left in place for 2 min before beginning injection to allow for tissue adjustment. Mice were given injections for ∼2 min (20 nl/min) of 50 nl of the KA solution using a micropump (Micro4 Microsyringe Pump Controller, WPI) operating the microsyringe. After injection, the cannula was left in place for additional 2 min to avoid reflux of injected solution along the needle track. Sham-operated mice were given injections of 50 nl of 0.9% sterile NaCl, but were otherwise treated identically. In mice intended for two-photon in vivo imaging, an additional virus injection and placement of a head fixation was carried out (see below). Mice had their scalp incision sutured and their anaesthesia terminated with atipamezolhydrochloride (Antisedant, Orion Pharma; 2.5 mg/kg b.w.; injection volume 0.1 ml/10 g, i.p.). Diazepam (Ratiopharm, injection volume 0.15 ml/20 g, s.c.) was administered to all mice 4 h after the start of status epilepticus (SE) to terminate the convulsions. At the same time, mice were also injected with glucose monohydrate (Glucosteril, Fresenius Kabi Deutschland; injection volume 0.25 ml, s.c.). Mice were returned to their cages and kept on a heat-pad until they woke from anaesthesia. After recovery from anaesthesia, the animals were kept under observation and were injected with the analgesic ketoprofen to alleviate pain for 4 days. After surgery mice were housed in individual cages. They were subsequently used for in vitro experiments 30 days after kainate/sham injection, or for in vivo two-photon imaging or behavioural experiments.
### Slice preparation and patch-clamp recording
Mice were deeply anaesthetized with isoflurane and then decapitated. Brains were rapidly removed and placed in ice cold (<2°C) sucrose-based artificial CSF (sucrose-ACSF) containing (in mM): 60 NaCl, 100 sucrose, 2.5 KCl, 1.25 NaH PO , 26 NaHCO , 1 CaCl , 5 MgCl , 20 glucose. Slices of 300 μm were cut with a vibratome (Leica) and incubated in sucrose-ACSF at 35°C for 30 min. Subsequently, slices were transferred to a submerged holding chamber containing normal ACSF containing (in mM): 125 NaCl, 3.5 KCl, 1.25 NaH PO , 26 NaHCO , 2.6 CaCl , 1.3 MgCl , 15 glucose at room temperature. All extracellular solutions were equilibrated with 95% O and 5% CO .
Selected CA1 cells were visualized with infrared oblique illumination optics and a water immersion objective (60×, 0.9 NA, Olympus) and somatic whole-cell current-clamp recordings were performed with a BVC-700 amplifier (Dagan Corporation). Data were filtered at 10 kHz and sampled at 50 kHz with a Digidata 1440 interface controlled by pClamp Software (Molecular Devices). Patch-pipettes were pulled from borosilicate glass (outer diameter 1.5 mm, inner diameter 0.8 mm; Science Products) with a Flaming/Brown P-97 Puller (Sutter Instruments) to resistances of 2–5 MΩ in bath and series resistances ranging from 8 to 30 MΩ. The standard internal solution contained (in mM): 140 K-gluconate, 7 KCl, 5 HEPES, 0.5 MgCl , 5 phosphocreatine, 0.16 EGTA. Internal solutions were titrated to pH 7.3 with KOH, had an osmolality of 295 mOsm, and contained 100 µM Alexa Fluor 594 (Invitrogen). Voltages were not corrected for the calculated liquid-junction potential of +14.5 mV. Membrane potential was adjusted to −75 mV for all recordings. To assess somatic action potential firing, current steps (800 ms) of increasing amplitudes were injected via the somatic patch pipette. The analysis of the effects of S-Lic or ICA-121431 on maximal firing rates was done by identifying the current injection at which maximal firing rates were obtained under control conditions. Effects of drugs and washout were quantified using this current injection magnitude. Passive membrane properties, action potential properties and firing patterns were assessed throughout the entire course of the experiment. Cells with unstable input resistances or lacking overshooting action potentials were discarded as well as recordings with holding currents >–200 pA for 60 mV and access resistances >30 MΩ.
For AMPA receptor-mediated miniature excitatory postsynaptic current (mEPSC) recordings, the Na channel blocker tetrodotoxin (1 μM) and GABA receptor antagonist bicuculline (20 μM) were added to the extracellular ACSF at least 15 min before starting recordings. Cells were voltage clamped at −60 mV and synaptic currents were recorded with an Axopatch 200-B amplifier (Axon Instruments), filtered at 2 kHz and digitized at 5 kHz. Analysis was performed with Minianalysis.
### In vitro two-photon uncaging
Two-photon glutamate uncaging at apical oblique dendrites of CA1 was performed using a dual galvanometer based scanning system (Prairie Technologies) to photo-release glutamate at multiple dendritic spines of CA1 neurons. MNI-caged-L-glutamate 15 mM (Tocris Cookson) was dissolved in HEPES-buffered solution (in mM as follows: 140 NaCl, 3KCl, 1.3 MgCl , 2.6 CaCl 20 D-glucose and 10 HEPES, pH 7.4 adjusted with NaOH, 305 mOsmol/kg) and was applied using positive pressure via glass pipettes (<1 MΩ) placed in closed proximity to the selected apical oblique dendrites of CA1 neurons. We used two ultrafast laser beams of Ti:sapphire pulsed lasers (Chameleon Ultra, Coherent) tuned at 860 nm to excite the Alexa 594 and one tuned to 720 nm was used to photo-release at 10–15 dendritic spines (within ∼10 µm in length). The intensity of each laser beam was independently controlled with electro-optical modulators (Conoptics Model 302RM). MNI-glutamate was uncaged at increasing number of spines (2–15) with 0.5 ms exposure times and the laser was rapidly moved from spine to spine with a transit time of ∼0.1 ms. The laser power at the slice surface was kept below 22 mW to avoid photo damage. The glutamate was uncaged onto a sequence of single spines to evoke unitary excitatory postsynaptic potential (uEPSP). To quantify deviations from linearity in dendritic integration, the arithmetic summation calculated from each individual uEPSP was compared to the actually measured EPSP during glutamate uncaging onto the same sequence of spines. The rate of rise of the dendritic spike initial fast phase was calculated from the maximum d V /d t value from dendritic spikes generated at similar number of spines (sham-control versus epileptic, 9.81 ± 0.31 and 9.51 ± 0.29 stimulated spines). The slow-phase NMDA area was calculated from the same dendritic spikes used to quantify d V /d t . The dendritic spike threshold was calculated as the amplitude of the expected EPSP at which dendritic spikes first occurred. All data analyses were done with Clampfit 9.2 software (Molecular Devices), IGOR Pro (Wavemetrics) and GraphPad Prism (GraphPad Software).
### In vitro pharmacology
S-Lic was kindly supplied by Bial–PORTELA & CA. S-Lic or ICA-121431 (Tocris) were dissolved in dimethyl sulphoxide (DMSO), with a 0.1% concentration of DMSO in ACSF. Control ACSF contained concentrations of DMSO equivalent to the drug-containing solution. Drug effects were analysed 15 min after initiating the drug application. We selected the S-Lic concentrations (300 µM) to be at the high end of effective therapeutic concentrations reported in mouse models and patients. This concentration of S-Lic reported maximally blocks Na current amplitude recorded in isolated hippocampal neurons and Na currents mediated by Na 1.2 and Na 1.6 channels expressed in heterologous expression systems. Of note, this therapeutically relevant concentration failed to affect Na 1.1 and Na 1.3, which were insensitive to S-Lic at concentration up to 1000 µM. ICA-121431 is a highly selective and potent inhibitor of Na 1.3 and Na 1.1 with reported IC values of 13 and 23 nM. We selected a concentration of 100 nM to maximally inhibit Na 1.3 channels (Na 1.3-mediated currents were inhibited by 80% at 100 nM) while still being selective for Na 1.3/1.1 channels. Branch strength plasticity induced by repetitive stimulations of dendritic spines potentiated dendritic spike strength during the 15 min timeframe of our pharmacology experiments and precluded an examination of Na channel blocker on the dendritic spike inactivation and input synchrony.
### Virus injections and head fixation
For Thy1-GCaMP6s mice intended for in vivo two-photon imaging experiments, injection of an adeno-associated virus (AAV) leading to labelling of astrocytes with mCherry (rAAV2/1.GFAP.mCherry, total volume 250 nl, 20 nl/min) was carried out in addition for improved motion correction using an orthogonal imaging channel. Following virus injection, Optibond (Optibond 3FL; two component, 48% filled dental adhesive, bottle kit; Kerr) was applied thinly to the skull to aid adhesion of dental cement. Subsequently, a flat custom-made head post ring was applied with the aid of dental cement (Tetric Evoflow) and the burrhole was closed.
### Two-photon in vivo imaging, window implantation procedure
Cranial window surgery was performed to allow imaging from the dorsal hippocampal CA1 region. Thirty minutes before induction of anaesthesia, the analgesic buprenorphine was administered for analgesia (Buprenovet, Bayer; 0.05 mg/kg body weight; injection volume 0.1 ml/20 g b.w., i.p.), and dexamethasone (Dexa, Jenapharm; 0.1 mg/20 g body weight; injection volume 0.1 ml/20 g body weight, i.p.) and ketoprofen (Gabrilen, Mibe; 5 mg/kg body weight; injection volume 0.1 ml/10 g body weight, s.c.) were given to inhibit inflammation/swelling and pain. Mice were anaesthetized with 3–4% isoflurane in an oxygen/air mixture (25%/75%) and then placed in a stereotactic frame. Eyes were covered with eye ointment (Bepanthen, Bayer) to prevent drying and body temperature was maintained at 37°C using a regulated heating plate (TCAT-2LV, Physitemp) and a rectal thermal probe. Further anaesthesia was carried out via a mask with a reduced isoflurane dose of 1–2% at a gas flow of about 0.5 ml/min. A circular craniotomy (Ø 3 mm) was opened within the head fixation ring above the right hemisphere hippocampus using a dental drill. Cortical tissue was aspirated until the alveus fibres above CA1 became visible. A custom-made cone-shaped silicon inset (upper diameter 3 mm, lower diameter 2 mm, length 1.5 mm, RTV 615, Movimentive) attached to a cover glass (Ø 5 mm, thickness 0.17 mm) was inserted and fixed with dental cement. Postoperative care included analgesia by administering buprenorphine twice daily (Buprenovet, Bayer; 0.05 mg/kg body weight; injection volume 0.1 ml/20 g b.w., i.p.) and ketoprofen once daily (Gabrilen, Mibe; 5 mg/kg body weight; injection volume 0.1 ml/10 g body weight, s.c.) on the 3 consecutive days after surgery. Animals were carefully monitored twice daily on the following 3 days and recovered from surgery within 24–48 h, showing normal activity and no signs of pain.
### Two-photon in vivo Ca imaging
We used a commercially available two-photon microscope (A1 MP, Nikon) equipped with a 25 × long-working-distance, water-immersion objective (NA = 1, WD = 4 mm, XLPLN25XSVMP2, Olympus) controlled by NIS-Elements software (Nikon). GCaMP6s was excited at 940 nm using a Ti:sapphire laser system (∼60 fs laser pulse width; Chameleon Vision-S, Coherent) and a fibre laser system at 1070 nm (55 fs laser pulse width; Fidelity-2, Coherent) to excite mCherry (see ). Emitted photons were collected using gated GaAsP photomultipliers (H11706–40, Hamamatsu). Movies were recorded using a resonant scanning system at a frame rate of 15 Hz and duration of 20 min per movie.
### Habituation and behaviour on the linear track
Experiments were performed in head-fixed awake mice running on a linear track. Two weeks before the measurements, mice were habituated to the head fixation. Initially mice were placed on the treadmill without fixation for 5 min at a time. Subsequently, mice were head-fixed, but immediately removed if signs of fear or anxiety were observed. These habituation sessions lasted 5 min each and were carried out three times per day, flanked by 5 min of handling. During the following 3–5 days, sessions were extended to 10 min each. All experimental recordings were from experimental sessions of 20 min duration. After habituation, mice ran well on the treadmill for average distances between 9 and 27 m per 20-min session (see ). The treadmill we implemented was a self-constructed linear horizontal treadmill, similar to Royer et al . The belt surface was equipped with tactile cues (see ). Belt position and running speed were measured by modified optical mouse sensors. All stimulation and acquisition processes were controlled by custom-made software written in LabView. Detailed construction plans and LabView software are available upon request.
### Data analysis, two-photon imaging
All analyses on imaging data and treadmill behaviour data were conducted in MATLAB using standard toolboxes, open access toolboxes and custom-written code. To remove motion artefacts, recorded movies were registered using a Lucas–Kanade model. In most cases, the red (mCherry) channel was used for motion correction. Individual cell locations and fluorescence traces were identified using a constrained non-negative matrix factorization-based algorithm and afterwards Ca events were identified with a deconvolution algorithm. All components were manually inspected and only those kept that showed shape and size of a CA1 pyramidal neuron and at least one Ca -event amplitude three standard deviations (SD) above noise level in their extracted fluorescence trace. We binarized individual cell fluorescence traces by converting the onsets of detected Ca events to binary activity events.
### Spatial tuning
To address spatial tuning of CA1 pyramidal neurons we used spatial tuning vector analysis. We restricted analysis to running epochs, where a running epoch was defined as an episode of movement with a minimal duration of 2.5 s above a threshold of 4 cm/s in a forward direction. Only cells with four or more event onsets during all running epochs in a 20 min session were included in the analysis. Mouse position was represented as vectors pointing towards the position on the linear track occupied by the mouse. We calculated the mean of the vectors at the times of all transient onsets during a session, weighted by the time spent in that bin. We addressed statistical significance by creating the null distribution for every spatially tuned cell. This was achieved by randomly shuffling the onset times and recalculating the spatial tuning vector 1000 times. The P- value was calculated as the percentage of vector lengths arising from the shuffled distribution that was larger than the actual vector length.
### Detection of aberrant synchronized activity during two-photon imaging
In order to detect and further analyse aberrant activity in the two-photon Ca imaging data, we used a custom-written pipeline in Python, mainly using the computer vision library OpenCV. After importing the movies into our pipeline, the median image obtained from 100 randomly selected frames is set to be the background frame. In the next step, while iterating through all frames, each frame is converted to greyscale and is convolved with a Gaussian kernel of size 35 × 35 pixels. The estimated background frame is then subtracted from all frames. To define a detection threshold for the aberrant activity episodes, the SD of the pixel intensities are calculated for each frame of the movie. The 95th percentile of the resulted distribution was taken as an intensity threshold for creating a binary segmented image. The aberrant activity episodes were then detected by searching the binary images for contours. If the size of a detected contour was comparable to the average size of somata in our field of view, this contour was excluded. The duration of activity epochs was calculated from the time stamp of the start and the end frame of the detected activity.
### Histochemistry
To verify successful viral transduction and window position, animals were deeply anaesthetized with ketamine (80 mg/kg b.w.) and xylazine (15 mg/kg b.w.). After confirming a sufficient depth of anaesthesia, mice were heart-perfused with cold phosphate-buffered saline (PBS) followed by 4% paraformaldehyde (PFA) in PBS. Animals were decapitated and the brain removed and stored in 4% PFA in PBS solution for 24 h. Coronal slices of the hippocampus 50–70-µm thick were cut on a vibratome (Leica). For nuclear staining, brain slices were kept for 10 min in a 1:1000 DAPI solution at room temperature. Brain slices were mounted and the red, green and blue channel successively imaged under an epi fluorescence or spinning disc microscope (Visitron VisiScope).
### In vivo pharmacology
In vivo treatment with the selective Na 1.3 antagonist ICA-121431 for behavioural experiments and in vivo two-photon imaging was carried out at a dose of 0.5 mg/kg body weight. ICA-121431 was solubilized in distilled water and applied via gavage with a stainless steel gavage cannula (20 G, 30 mm long, FST).
### Behaviour, object location, object recognition memory and Y-maze spontaneous alternation task
Three weeks after kainate/sham treatment, mice were habituated to handling by the experimenter for 10 min daily on two consecutive days prior to starting behavioural experiments. On experimental days, mice were moved from the animal holding facility to the experimental room at least 45 min before starting the experiments. The experimental room was quiet and had dim light conditions of around 20 lux, achieved by a single light spot pointed towards the ceiling. Object location and object recognition experiments took place in the same acrylic glass arena that was 56 cm × 36 cm in size and had 20 cm high transparent walls without internal visual cues. Between animals, the arena was thoroughly cleaned using 70% ethanol. The arena was placed on an infrared light board and animal activity was monitored with an infrared camera above the arena and animal tracking software (Noldus × 8.5, Ethovision). On the first three consecutive days, mice were placed in the empty arena for 10 min. After 2 days of break, treatment was started by administering animals twice per day via gavage with the drug or control substance. Animals received either 0.5 mg/kg body weight of ICA-121431 dissolved in water (Ampuwa) or plain water, 4 and 2 h before starting experiments on every experimental day. On the first day of treatment, mice were again placed in the empty open field arena for 10 min (Day 4 of open field; see ). On the next day, mice were introduced to two identical objects (blue beaker bottle caps, diameter 5.5 cm) placed on opposite ends of a narrow side of the arena, 11.5 cm away from the walls, for 10 min (Habituation; see ). Twenty-four hours later, one of the objects was displaced along the long side of the arena by 32 cm and mice were given 5 min to explore [object location memory (OLM) test; see ]. On the next day, mice encountered the objects in the same position for 10 min as the day before to habituate them to the displacement (Habituation, Day 7, see ). Twenty-four hours later, a novel object, a transparent Petri dish of the same diameter, replaced one of the familiar objects and mice were given 5 min to explore [novel object recognition (NOR) test; see ]. In separate experiments, we found that naïve mice on average do not express a preference for either object before learning (data not shown).
As an indicator of successful memory formation and recall, we compared exploration times of the objects. Successful memory formation is indicated by an increased exploration of the displaced or novel object compared to the familiar object. We computed a discrimination index percentage according to the following formula (exploration time of displaced/novel object – exploration time of familiar object)/(exploration time of displaced/novel object + exploration time of familiar object) × 100. Exploration times were manually scored by an experienced observer blinded to the treatment condition of the animal. We considered exploration when the nose of the animal was apposed to the object and pointed towards it. Time spent climbing onto the object was not considered exploration. We excluded trials in which mice did not explore the objects for more than 3 s in both the habituation and recall trials, and if mice displayed a strong preference towards one of the two objects in habituation trials, indicated by a discrimination index (%) of >20. Occupancy plots were created using a custom-written MATLAB script using tracking data from the Noldus software.
To investigate spatial working memory, the mice were placed for 10 min in a Y-shaped arena made of red Plexiglas with three arms (40 cm long, 8 cm wide) arranged at 120° angles. Symbols glued to the end of the arms served as internal visual cues. Spontaneous alternation behaviour was analysed by manually counting the number of three consecutive entries into different arms and dividing it by the number of potential alternations according to the following formula: number of successful alternations/(total arm entries – 2). Arm entries were scored when all four paws of an animal were in an arm.
### Chronic local field potential monitoring
Five male C57BL/6J mice (age 13 weeks, Charles River) were induced with kainate according to the methods described above except that a 10 μl Nanofil syringe (World Precision Instruments) was used and the kainate injection coordinate was [−1.9 mm AP, 1.5 mm ML (right side), −1.1 mm DV from bregma]. Seven weeks after status, mice were implanted with microdrives comprising movable tungsten electrodes. Mice were injected with the analgesic buprenorphine a (0.05 mg/kg b.w.) and anaesthetized with 1–2% isoflurane. Mice were fixed in a stereotactic frame and the microdrive was implanted above the right hippocampus (−2 mm AP, 1.5 mm ML). Reference and ground screws were implanted anterior to bregma. The implant was fixed to the skull with stainless screws and dental cement. Two weeks after implantation, local field potential (LFP) was recorded from the hippocampus of mice for 8 days (1 h each day) using a Neuralynx system (Digital Lynx SX, sample rate: 32 kHz, filter: 1–8000 Hz) and Cheetah 6.4.1. Animals were video monitored during the LFP recording. Seizures were detected in LFP recordings using a home-written MATLAB code. First LFP was band pass filtered (1–300 Hz). In order to detect the ictal spikes a threshold was defined manually by the user (2–4 SD > mean of full recording). Peaks that were bigger than threshold were categorized as spikes. Ictal spikes that were <2 s to the previous spikes were grouped together as one ictal event (seizure). Seizures closer than 5 s were merged together and counted as one seizure. Only seizures that lasted for at least 10 s were counted. Video observation facilitated defining seizure as subclinical or behavioural.
### Multiplex fluorescent in situ hybridization (RNAscope)
Multiplex fluorescent RNA in situ hybridization was performed on 40 μm fixed, frozen hippocampal sections using RNAscope Fluorescent Multiplex Detection Reagents v2 (323110, ACDBio) according to the instructions provided by the manufacturer for fixed-frozen tissue (User Manual: 323100-USM). Hybridized target probes were lableled by TSA Plus Cyanine 3 and TSA Plus Cyanine 5 (NEL760001KT, Perkin Elmer). The probes Scn3a (Mm-Scn3a, Cat No. 502 641), Gad1 (Mm-Gad1-C3, Cat No. 400951-C3) and Gad2 (Mm-Gad2-C2, Cat No. 439371-C2) were designed by ACDBio. Sections were counterstained with DAPI (Cat No. 323108, ACDBio), then mounted with Aqua Polymount (Polysciences Inc.). In situ hybridized sections were imaged at an inverted Zeiss AxioObserver equipped with a CSU-W1 confocal scanner unit (50 μm pinhole disk, Yokogawa). At × 40 (C-Apochromat, 40×/1.2 water, Zeiss) magnification, tile images and z-stacks were acquired with laser lines 405, 488 and 561 nm. Images taken with the 40× objective are maximum intensity projections of z-stacks. Tile images were stitched with VisiView software (Visitron Systems). The anatomical region corresponding to the pyramidal layer of hippocampal CA1 was manually outlined in Fiji ( ), based on the Allen Mouse Brain Reference Atlas. A binary mask of Gad1/2 probe-expressing cells was created in Fiji to segment GABAergic neurons and non-GABAergic pyramidal cells of CA1 and analyse them separately. For quantification, a non-biased automated approach was used to detect and quantify probe signals using the open-source image analysis software CellProfiler. The CellProfiler pipeline ‘Speckle Counting' ( ) was optimized to detect nuclei and probe signals (punctae). A threshold was set at 5 SD above mean pixel intensity and probe signals above the threshold were assigned to the closest nucleus and punctae located within the nuclei were quantified.
### Immunoblotting
For Western blots, CA1 hippocampal regions were harvested from sham-control and epileptic brains. The tissue was homogenized with RIPA buffer (182–02451, Wako) containing 10% protease inhibitors (4693116001, Roche) and phosphatase inhibitors (4906845001, Roche). Homogenates were subsequently incubated on ice for 20 min and centrifuged for 10 min at 4°C. The lysates were mixed with 4 × NuPAGE LDS sample buffer (NP0008, Novex) and heated at 95°C for 5 min. After denaturation, each lysate was separated by sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS-PAGE; 8%) and subsequently transferred to a nitrocellulose membrane, which was blocked in 5% milk. The primary and secondary antibodies were the following: anti-Na 1.2 rabbit polyclonal (1:200; ASC-002, Alomone Labs); anti-Na 1.3 rabbit polyclonal (1:200; ASC-004, Alomone Labs); anti-GAPDH mouse monoclonal (1:1000, NB300-221, Novus); IRDye® 800CW goat anti-rabbit IgG secondary antibody (1:10000, 926-32211, LI-COR Biosciences); IRDye® 680LT donkey anti-mouse IgG secondary antibody (1:10000, 926-68022, LI-COR Biosciences).
The primary antibodies were incubated for 2 h at room temperature (RT), while the fluorescently labelled IRDye anti-rabbit 800 nm IgG (LI-COR Biosciences) or IRDye anti-mouse 680 nm were incubated for 45 min at RT and detected with the infrared Odyssey system (LI-COR Biosciences). Band intensity was quantified by Image Studio Lite 5.2 software (LI-COR Biosciences). The Na protein band intensity was determined as a ratio of the housekeeping protein glycerinaldehyde-3-phosphate-dehydrogenase (GAPDH) within the same immunoblot and normalized to the average ratio obtained in sham-control.
### Histopathology
Histopathological evaluation in the kainic acid model was performed on 40 μm fixed, frozen hippocampal slices. Sections were counterstained with DAPI (Cat No. 323108, ACDBio), then mounted with Aqua Polymount (Polysciences Inc.). The thickness of the CA1, CA3 and DG layers was measured in Fiji. For the CA1 region, the layer thickness was obtained by averaging five points 300 µm equally distant, between proximal and distal CA1, based on the Allen Mouse Brain Reference Atlas. For CA3, dorsal DG and ventral DG, the thickness was obtained by averaging three points 300 µm equally distant.
### Statistics
Average values in the text and figures are expressed as mean ± SEM. Significance level was set to P < 0.05. Detailed statistics are described for each individual test performed in the text.
### Data availability
The data that support the findings of this study are available from the corresponding author, upon reasonable request.
## Results
### Altered dendritic integration via dendritic spikes in chronic epilepsy
We examined dendritic integration in the kainate model, a well-established model of chronic temporal lobe epilepsy, characterized by spontaneous recurrent seizures ( ) and hippocampal pathology ( ). The intrinsic firing properties of neurons showed an increased action potential output ( ), as well as an increased fraction of bursting neurons, similar to other mouse TLE models ( and ).
To probe dendritic integration in CA1 pyramidal neurons we used two-photon glutamate uncaging. Responses of single spines to uncaging of glutamate (uEPSPs) were calibrated to ∼1 mV in both sham-control and epileptic animals (1.09 ± 0.42 mV, n = 98 versus 1.07 ± 0.04 mV, n = 85 dendrites, unpaired t -test, P = 0.80; ). The rise and decay kinetics of such single-spine unitary EPSPs elicited by uncaging were slightly but significantly faster in epileptic animals ( ). This was also observed in mEPSC recordings ( ). Spine density in first- and second-order dendrites did not differ between sham-controls and epileptic animals ( ).
We then went on to probe the capability of CA1 dendrites to generate dendritic spikes by stimulating multiple spines quasi-synchronously (interspine stimulation interval, 0.1 ms). As described previously, CA1 dendrites either displayed linear integration only, or were capable of generating sudden supralinear dendritic spikes when increasing numbers of spines were synchronously stimulated (representative examples in ). In linearly integrating dendritic branches, the linear summation of single spine uEPSPs was augmented in epileptic animals in first-order branches emanating from the main dendritic shaft ( n = 35 and 26 in sham-control versus epileptic mice). Linearly integrating second-order dendrites did not show differences in EPSP summation [ and , n = 23 and 15 in sham-control versus epileptic mice, two-way ANOVA main effect, sham-control versus epileptic mice: F (1,95) = 7.38, P = 0.0078; first order versus second order: F (1,95) = 4.33, P = 0.040; interaction: F (1,95) = 0.32, P = 0.57; Bonferroni’s post-test, first-order dendrites sham-control versus epileptic mice P = 0.018; second-order dendrites sham-control versus epileptic mice P = 0.35; ].
Probing dendritic integration in chronic epilepsy (kainate model of epilepsy). ( A ) CA1 pyramidal neurons filled with a fluorescent dye via the somatic patch recording pipette, close up of a dendritic segment with uncaging targeting points at spines. ( B ) Responses to single-spine stimulation with 2P-uncaging of MNI-glutamate measured with somatic patch recording. ( C ) Representative recordings of compound linear and supralinear EPSPs. Grey lines are expected EPSPs from linear summation of single spine responses. Examples of a branch with linear integration ( left ) and a branch capable of supralinear integration ( right ) are shown for sham-control (black) and epileptic mice (red). ( D ) Examples of linearly integrating first-order dendrites in sham-control (black) and epileptic animals (red). ( E ) Examples of first-order dendrites capable of generating supralinear dendritic spikes in sham-control (black) and epileptic animals (red). Occurrences of dendritic spikes and their voltage thresholds are indicated with arrows. ( F ) Quantification of the slope of the linear phase in linearly integrating first- and second-order dendrites in sham-control (black) and epileptic mice (red). Asterisks indicate Bonferroni’s post-tests P = 0.0176. ( G ) Representative dendrite with first- and second-order branches. ( H ) The propensity for dendritic spikes is enhanced in first-order dendrites in epileptic animals (red) compared to sham-controls (black). Asterisks indicate Fisher’s exact test P = 0.0011. ( I ) The threshold for generation of dendritic spikes, measured as indicated with arrows in E is reduced in first-order dendrites of epileptic animals, but not in second-order dendrites. Asterisks indicate Bonferroni’s post-test, P = 0.0017. ( J – L ) The fast phase of dendritic spikes is accelerated in first-order dendrites from epileptic mice ( K , asterisks indicate Bonferroni’s post-test, rate of rise in first-order dendrites in sham-control versus epileptic P = 0.0025; first- versus second-order dendrites rate of rise in epileptic P = 0.0074). The area of the slow phase was unchanged.
The difference in linear summation was normalized by application of the Na channel blocker tetrodotoxin (TTX), indicating that it is caused by increases in voltage-gated Na currents [ ; sham-control and epileptic n = 7 and 6, respectively, two-way repeated measures ANOVA main effect, sham-control versus epileptic: F (1,11) = 0.88, P = 0.37; ACSF versus TTX: F (1,11) = 26.96, P = 0.0003; interaction: F (1,11) = 10.15, P = 0.0087; Bonferroni’s post-test, ACSF versus TTX in epileptic P = 0.0003].
Strikingly, the fraction of dendrites capable of generating dendritic spikes was significantly increased in first-order dendrites, but not in second-order dendrites in epileptic animals ( , and , first-order dendrites 36% versus 60%, Fisher’s exact test P = 0.0011, second-order dendrites 29% versus 46%, Fisher’s exact test P = 0.062). In these experiments, the average distances of the uncaging sites from the somatic region were not different when comparing sham-control and epileptic mice (first-order dendrites sham-control 66.2 ± 2.2 µm, n = 101 versus epileptic mice 64.9 ± 2.3 µm, n = 69; second-order dendrites sham-control 77.5 ± 2.6 µm, n = 54 versus epileptic mice 80.6 ± 2.8 µm, n = 55; unpaired Student’s t -test first-order sham-control versus epileptic mice P = 0.68; second-order sham-control versus epileptic mice P = 0.43).
We next compared the properties of dendritic spikes in those branches capable of generating them. The threshold for generation of dendritic spikes was computed by determining the expected somatic voltage at which a dendritic supralinearity first occurred (indicated by arrows at the x -axis; and ). In epileptic animals, the voltage threshold to generate dendritic spikes was significantly reduced in first-order but not second-order dendrites [ ; first-order dendrites n = 25 and 27, second-order dendrites n = 13 and 13 for sham-control and epileptic mice, respectively, two-way ANOVA main effect, sham-control versus epileptic: F (1,74) = 9.92, P = 0.0024; first-order versus second-order: F (1,74) = 0.012, P = 0.91; interaction: F (1,74) = 0.77, P = 0.38; Bonferroni’s post-test, threshold in first-order dendrites in sham-control versus epileptic P = 0.0017].
The reduction was remarkably pronounced in some first-order dendrites from epileptic animals, with dendritic spikes sometimes being generated with stimulation of as few as 1–3 spines, and somatic voltages of as little as ∼3 mV (see for dendritic spike elicited with a single spine stimulation).
A further characteristic of dendritic spikes is the rate of rise of the initial fast phase, which reflects the contribution of voltage-gated sodium channels. The maximal rate of rise was determined from the first derivation of the voltage trace (indicated in ). In first-order dendrites, but not second-order dendrites, we observed a significant increase in the maximal rate of rise in epileptic animals [ ; first-order dendrites n = 44 and 52, second-order dendrites n = 16 and 30 in sham-control and epileptic, respectively, two-way ANOVA main effect, sham-control versus epileptic: F (1,139) = 1.86, P = 0.18; first-order versus second-order: F (1,139) = 1.96, P = 0.16; interaction: F (1,139) = 5.08, P = 0.025; Bonferroni’s post-test, rate of rise in first-order dendrites in sham-control versus epileptic P = 0.0025; first- versus second-order dendrites rate of rise in epileptic P = 0.0074]. In contrast to the fast phase of the dendritic spikes, the subsequent slower depolarization was not different in sham-control versus epileptic animals [ ; first-order dendrites n = 42 and 47, second-order dendrites n = 16 and 29, in sham-control and epileptic, respectively, area under the curve of slow depolarization two-way ANOVA main effect, sham-control versus epileptic: F (1,130) = 3.51, P = 0.063; first-order versus second-order: F (1,130) = 1.83, P = 0.18; interaction: F (1,130) = 0.17]. These results collectively show a dramatically augmented excitability of proximal, first-order dendrites in epileptic animals, reflected in the prevalence and properties of dendritic spikes.
### Reduced synchrony requirement for dendritic spike generation
Synchronous stimulation is required for generation of dendritic spikes in normal pyramidal neurons, and is thought to be critical to detect the coincident activity of specific presynaptic ensembles. Therefore, we tested how dendritic spike generation depends on input synchrony by varying the inter-spine stimulation interval in control and epileptic animals (representative examples in , sham-control in black, epileptic in red). Epileptic animals exhibited a remarkable loss of their capability to selectively respond to synchronous inputs via generation of dendritic spikes. While sham-control animals exhibited a steep decrease in dendritic spike generation when stimulation was less synchronous, epileptic animals continued to generate dendritic spikes even with very asynchronous stimulations ( ; n = 12 and 11 in sham-control versus epileptic mice, Fisher’s exact test, P < 0.001 indicated with asterisks, see figure legend).
Degraded synchrony requirement in epileptic mice. ( A ) Representative example traces showing input synchrony dependence of dendritic spikes in CA1 neurons from sham-control and epileptic mice. Changing the inter-spine stimulation interval ( bottom rows in Sham-control and Epileptic) systematically affects dendritic spike generation in control and epileptic animals. Upper rows show somatic voltage response, middle rows show first derivation of the voltage trace (d V /d t ). Note that the rightmost response is again elicited with a 0.1 ms inter-spine stimulation interval, applied following the longer intervals. ( B and C ) Rate of spike occurrence in percent of uncaging stimulations (D-spike success) and dendritic-spike strength expressed as the maximal rate of rise (d V /d t ) of the fast phase of the dendritic spike. Asterisks in B correspond to significance in Fisher’s exact test, with the following P -values: 2 ms P = 0.0013; 3 ms P < 0.0001; 5 ms P = 0.0013. Asterisks in C correspond to significances in Bonferroni’s post-tests for inter-spine stimulation intervals 2 ms P = 0.0013; 3 ms P < 0.0001; 5 ms P = 0.0003.
Correspondingly, the maximal rate of rise of the fast phase of dendritic spikes (d V /d t ) also showed a much less pronounced reduction with less synchronous stimulation in epileptic animals [ ; two-way repeated-measures ANOVA main effects, sham-control versus epileptic: F (1,21) = 32.34, P < 0.0001; inter-spine time: F (4,84) = 22.59, P < 0.0001; interaction: F (4,84) = 9.87, P < 0.0001; Bonferroni’s post-tests indicated with asterisks, P < 0.003, individual P -values see legend]. Intriguingly, in some cases, dendritic branches in epileptic animals were capable of generating multiple sequential spikes ( n = 2 of 11 branches; example in , arrow).
### Reduced dendritic spike inactivation in chronic epilepsy
Sparse dendritic spiking is supported by inactivation of dendritic spike generation by prior activity, a phenomenon that relies on inactivation of dendritic sodium channels. In control animals, dendritic spike inactivation was observed similar to published data, as indicated by a progressive reduction of dendritic spike d V /d t until dendritic spike failure (black traces in ; quantification in and ). In epileptic animals, this reduction in dendritic spike d V /d t as well as inactivation of spike generation was significantly less pronounced ( , statistical results see figure legend).
Degraded dendritic spike inactivation in chronic epilepsy. ( A ) Representative somatic voltage recordings from a sham-control (black) and epileptic (red) mouse ( upper trace ) with the corresponding first derivation of the voltage trace (d V /d t ) shown below. Dendritic spikes were repetitively elicited with synchronous stimulation at 10 Hz ( lowermost trace ). ( B ) The fraction of stimuli successfully generating dendritic spikes strongly decreased with repetitive stimulation ( n as in C , asterisks indicate Fisher’s exact test, third stimulation P = 0.001, fourth stimulation P = 0.0001, fifth stimulation P = 0.037). ( C ) Spike strength (d V /d t ) also decreases steeply with repetitive stimulation in control (black) but not in epileptic animals (red). Sham-control n = 15, epileptic mice n = 12. Two-way repeated-measures ANOVA revealed main effects, sham-control versus epileptic: F (1,25) = 12.51, P < 0.0016; repetitive stimulation: F (4,100) = 80.57, P < 0.0001; interaction: F (4,84) = 8.30, P < 0.0001); asterisks indicate Bonferroni’s post-test, d V /d t for second stimulation P = 0.0006, third stimulation P = 0.0001, fourth stimulation P = 0.0002.
Thus, collectively, the results show that dendritic spike generation is strongly enhanced, and that multiple mechanisms that underlie sparse generation of dendritic spikes in normal animals are severely degraded in epilepsy.
### Involvement of Na channels in augmented dendritic integration
The increased incidence of dendritic spikes, as well as the increased rate of rise of the fast phase of the dendritic spikes suggests upregulation of Na channels as a potential mechanism, and raises the possibility that anticonvulsants targeting Na channels might affect this phenomenon. We first tested if the Na channel blocker and anticonvulsant S-Lic (300 µM) affect aberrant dendritic spikes ( ). S-Lic did not affect the properties of uEPSPs ( ). Surprisingly, application of S-Lic also did not alter the threshold for eliciting dendritic spikes in either control or epileptic animals [ and ; spike thresholds indicated with dashed lines, quantification in and , n = 8 and 5 in sham-control versus epileptic; two-way repeated-measures ANOVA main effect, sham-control versus epileptic: F (1,11) = 5.82, P = 0.044; ACSF versus S-Lic: F (1,11) = 0.86, P = 0.37; interaction: F (1,11)= 0.47, P = 0.51, Bonferroni’s post-test n.s.]. Likewise, S-Lic did not affect the rate of rise of dendritic spikes in epileptic animals [ and ; two-way repeated-measures ANOVA main effect, sham-control versus epileptic: F (1,11) = 27.76, P = 0.0003; ACSF versus S-Lic: F (1,11) = 1.30, P = 0.28; interaction: F (1,11) = 2.51, P = 0.14, Bonferroni’s post-test n.s.]. In contrast to the lack of effect on dendritic spikes, S-Lic significantly inhibited somatic action potential generation [ and ; for statistics see legend] as described previously (see e.g. Doeser et al . and Schmidt et al . ).
The Na 1.3/1.1 selective blocker ICA-121431, but not the Na 1.2/1.6 Na channel blocker S-Lic reverses enhanced dendritic excitability. ( A and E ) Representative examples of the effects of the Na 1.2/1.6 Na channel blocker S-Lic (300 µM) on dendritic spikes ( insets show magnification of the fast phase of the dendritic spike), and on the first derivation of the voltage trace (d V /d t ) in sham-control ( A ) and epileptic ( E ) animals. ( B and F ) Effects of S-Lic on the maximal rate of rise of the dendritic spike in sham-control ( B ) and epileptic mice ( F ). ( C and G ) Representative examples of effects on the dendritic spike threshold in sham-control and epileptic mice (arrows indicate occurrence of dendritic spikes and dashed lines indicate thresholds). ( D and H ) Quantification showing the lack of significant effects on the dendritic spike threshold in sham-control ( D ) and epileptic mice ( H ) with S-Lic application. Two-way ANOVA revealed no significant effects of S-Lic on the rate of rise or threshold of dendritic spikes in control ( B and D ) and epileptic ( F and H ) animals. ( I – K ) Effects of S-Lic on somatic action potential generation. Representative examples of repetitive firing evoked by somatic current injection in sham-control and epileptic mice in ACSF and in the presence of S-Lic (violet, I ). Effects of S-Lic on firing induced by current injection at which firing frequency was maximal under ACSF conditions, control and epileptic animals ( I and K, respectively). Asterisks indicate Bonferroni’s post-test P = 0.0034. ( L – S ) Effects of the Na 1.3/1.1 Na channel blocker ICA-121431 (100 nM) in sham-control ( L – O ) and epileptic animals ( P – S ), depicted in the same manner as in panels ( A – H ). Asterisks indicate Bonferroni’s post-test, P = 0.023 ( Q ) and P = 0.0050 ( S ). ( T – V ) Lack of effects of ICA-121431 on somatic action potential generation. Representative examples of repetitive firing evoked by somatic current injection in sham-control and epileptic mice in ACSF and in the presence of ICA-121431 (blue, T ). Lack of effects of ICA-121431 on the maximal firing frequency of CA1 neurons ( U and V ).
At this concentration, S-Lic has been shown to have strong preferential effects on the availability of Na 1.2 and 1.6 channels, with very little effect on Na 1.3 or 1.1 channels at resting membrane voltages. This raises the possibility that one of these latter channel types underlies increased dendritic spiking, even if Na 1.3 or 1.1 channels are normally not expressed at high levels in adult pyramidal neurons. To test this idea, we took advantage of the selective Na 1.3/1.1 blocker ICA-121431 at 100 nM, a concentration which blocks Na 1.3 and Na 1.1 ( ). In these experiments, we specifically selected uncaging sites on dendrites that exhibited robust and noticeable dendritic spikes to assess if ICA-121431 blocked them. Due to this bias, differences in spike properties between sham-control and epileptic animals are not present in this experiment.
ICA-121431 did not alter the properties of uncaging-evoked single-spine EPSPs ( ). However, in epileptic, but not in sham-control animals, ICA-121431 (100 nM) significantly increased the dendritic spike threshold [ and , spike thresholds indicated with dashed lines, and ; in summary, n = 5 and 7 in sham-control versus epileptic, two-way repeated-measures ANOVA main effect, sham-control versus epileptic: F
(1,10) = 0.008, P = 0.93; ACSF versus ICA-121431: F (1,10) = 6.24, P = 0.032; interaction: F (1,10) = 7.10, P = 0.024; Bonferroni’s post-test epileptic ACSF versus ICA-121431 P = 0.0050]. Likewise, ICA-121431 decreased the rate of rise of dendritic spikes only in epileptic animals [see insets for d V /d t traces in and summary in ; two-way repeated-measures ANOVA main effect, sham-control versus epileptic: F (1,10) = 0.050, P = 0.82; ACSF versus ICA-121431: F (1,10) = 5.36, P = 0.043; interaction: F (1,10) = 2.50, P = 0.14; Bonferroni’s post-test epileptic ACSF versus ICA-121431 P = 0.025]. The NMDA-receptor-driven slow phase of the dendritic spike was unaffected by ICA-121431 (two-way repeated-measures ANOVA, n.s.).
We then tested if, in addition to the significant effects on dendritic integration in epileptic animals, blocking Na 1.3/1.1 channels also affects the generation of somatic action potentials. Surprisingly, ICA-121431 had no effects on action potential generation or repetitive firing induced by somatic current injection in either control or epileptic animals [ , and ; n = 6 in both groups, two-way repeated-measures ANOVA main effect, sham-control versus epileptic: F (1,10) = 11.66, P = 0.0066; ACSF versus ICA-121431: F (1,10) = 1.67, P = 0.23; interaction: F (1,10) = 2.28, P = 0.16, Bonferroni’s post-tests, n.s.].
This suggests that the Na 1.3/1.1 blocker ICA-121431 selectively affects dendritic spikes rather than somatic firing, while S-Lic does the converse, and raises the possibility that subtype selective Na channel antagonists may be useful in selectively dampening dendritic hyperexcitability in epilepsy.
Na 1.3 channels have been previously described to exhibit increased expression in acquired experimental epilepsy, while Na 1.1 channels are downregulated. To confirm the upregulation of Na 1.3 in our model, we investigated Scn3a expression at the mRNA level at single-cell resolution. To this end, we used multiplex RNA in-situ hybridization (RNAscope, see ‘Materials and methods’ section; ). Low levels of Scn3a mRNA puncta were present in pyramidal neurons in the adult hippocampus ( ). However, we found a significant upregulation of Scn3a expression in pyramidal neurons from epileptic animals (Kolmogorov–Smirnov test P < 0.0001 sham-control n = 5, epileptic mice n = 6; ). In contrast, GAD-expressing interneurons showed downregulation of Scn3a compared with sham-control animals (sham-control n = 5 epileptic mice n = 5, Kolmogorov–Smirnov two-tailed test, P < 0.001; ).
Scn3a expression is increased at the mRNA and protein level in the hippocampal CA1 region of epileptic mice . ( A – D ) Multiplex fluorescent mRNA in-situ hybridization for Scn3a in hippocampal CA1 region. ( A and B ) Representative fluorescent images showing labelling of nuclei with DAPI (blue), labelling of GABAergic neurons (GAD) with a probe for Gad1/2 mRNA (green) and of Scn3a mRNA (red) in hippocampal slices from sham-control and epileptic animals, scale bar 500 µm. Close up ( right ) of the boxed areas, scale bar = 100 µm. ( C ) Cumulative distribution of Scn3a punctae per nucleus colocalized excitatory pyramidal cells (excluding GAD-labelled putative interneurons). The amount of punctae was significantly higher in epileptic animals (2979 cells in six animals and 3669 cells in five animals, for sham-control and epileptic mice, respectively; asterisks indicate Kolmogorov–Smirnov two-tailed test P < 0.0001). ( D ) Cumulative distribution of Scn3a punctae per nucleus colocalized in GAD+ interneurons. The amount of punctae was significantly lower in epileptic animals (253 cells in five animals and 141 cells in five animals, for sham-control and epileptic mice, respectively; asterisks indicate Kolmogorov–Smirnov two-tailed test P < 0.001). ( E – H ) Western blots for Na 1.3 and Na 1.2 in hippocampal CA1 region. ( E ) Representative Western blots of Na 1.3 (indicated with arrrowhead) and the housekeeping protein GAPDH, used as loading control in sham-control and epileptic mice. ( F ) Quantitative analysis of the Na 1.2 protein obtained by band intensity analysis from E (normalized Na 1.2/GAPDH ratio of 1.0 ± 0.36 in sham-control n = 4 and 3.64 ± 0.85 in epileptic mice n = 4; unpaired Student’s t -test P = 0.029; data represent mean ± SEM). ( G ) Representative Western blots of Na 1.2 (indicated with arrrowhead) and the housekeeping protein GAPDH, used as loading control in sham-control and epileptic mice. ( H ) Quantitative analysis of the Na 1.2 protein obtained by band intensity analysis from G (normalized Na 1.2/GAPDH ratio of 1.0 ± 0.05 in sham-control n = 3 and 0.93 ± 0.052 in epileptic mice n = 3; unpaired Student’s t -test P = 0.38).
We then assessed Na 1.3 protein level in the hippocampus using Western blot analyses. Consistent with the Scn3a mRNA puncta the protein level of Na 1.3 in CA1 subregion was low in sham animals and up to four times higher in tissue from epileptic animals (sham-control n = 4 and epileptic mice n = 4, Student’s t -test, P = 0.02; ). This is in contrast to Na 1.2 protein, which was not significantly different between sham-control and epileptic animals (sham-control n = 3 and epileptic mice n = 3, Student’s t -test two-tailed, P = 0.38; ).
### ICA-121431 normalizes aberrant place-related firing of CA1 neurons in vivo
These data point towards a substantial decrease in specificity of dendritic spikes and degraded input feature detection. Because dendritic spikes are thought to play an important role in shaping place field properties of CA1 neurons, we hypothesized that aberrant dendritic spikes might be associated with degraded place coding in CA1 neurons in vivo . We therefore examined the activity of CA1 neurons using two-photon Ca imaging in head-fixed sham-control and epileptic Thy1-GCaMP6s mice running on a linear track equipped with spatial cues ( and ). As described previously, the activity of CA1 neurons was higher during locomotion in both the sham-control and epileptic mice. Additionally, we observed a markedly increased activity of CA1 neurons in epileptic animals compared to sham control animals, both during immobility and locomotion ( ; n = 1022 CA1 neurons in five sham-control mice and 1697 CA1 neurons in six epileptic mice; for statistics see figure legend).
ICA-121431 normalizes place-related firing of CA1 neurons from epileptic mice in vivo . ( A ) Experimental protocol used to examine the activity of CA1 neurons using two-photon Ca imaging in Thy1-GCaMP6s mice running on a linear track. ( B and C ) Representative fields of view obtained for in vivo imaging. ( D ) Representative examples of activity in a subset of the CA1 neurons from the fields of view shown in B and C . Upper traces indicate ΔF/F traces from a subset of CA1 neurons, lower traces (blue) indicate position of the mouse. ( E ) Analysis of place coding. Spiral plots for three representative CA1 neurons in a sham-control mouse ( left ) and an epileptic mouse ( right ). One 360° pass around the spiral plot corresponds to a complete transition on the 150 cm circular linear treadmill. Grey circles indicate event onsets in the CA1 neuron. Computed place vectors indicated by black straight lines. Distributions on the right show tests versus shuffled distributions for each cell (see ‘Materials and methods’ section for details). The red vertical lines indicate the vector length of the CA1 neuron, shuffled distributions shown in blue. ( F ) All significantly place-coding CA1 neurons from a representative sham-control ( left ) and epileptic mouse ( right ). Lower panels show the same mice following treatment with ICA-121431. ( G ) Panels show all CA1 neurons in the two experimental groups (sham-control and epileptic), both before and during application of ICA-121431. ( H ) Cumulative distributions of place vector lengths for sham-control ( left ) and epileptic mice ( right ). Blue curves indicate cumulative distribution of vector lengths after application of ICA-121431. ( I ) Differences between average vector lengths in sham-control (black) and epileptic animals (red) were stable over imaging sessions on three consecutive days. Two-way ANOVA main effect, sham-control versus epileptic: F (1,27) = 31.12, P < 0.0001; subsequent days: F (2,27) = 0.2623, P = 0.7727; interaction: F (2,27) = 0.05973, P = 0.9421; asterisks indicate Bonferroni’s post-test, Day 1 P = 0.02, Day 2 P = 0.0066 and Day 3 P = 0.0074. ( J ) Average vector lengths calculated per mouse in sham-control (black) and epileptic mice (red). Asterisks indicate Bonferroni’s post-tests sham-control versus epileptic P = 0.0095, epileptic mice before ICA-121431 treatment versus ICA-121431 treated epileptic mice P = 0.0207. ( K ) Percent change in vector length caused by ICA-121431, quantified for each animal and averaged. Asterisk indicates Mann–Whitney U-test, P = 0.030.
CA1 neurons that exhibited significant place-related activity were found in both sham-control and epileptic mice (examples of representative neurons in , spiral plots show place coding, rightmost distributions show tests versus shuffled distributions for each cell). In both sham-control and epileptic mice, place-related firing fields tiled the extent of the linear track ( , upper panels all significantly place-coding CA1 neurons from a representative sham-control and epileptic mouse, respectively; , upper panels for data from all sham-control and epileptic mice). It was apparent from visual inspection of these data that place-related firing of significantly place-coding cells appeared to be less specific in epileptic animals, consistent with published work demonstrating degraded place coding in epileptic animals. We quantified the precision of spatial coding using a spatial tuning vector measure (see also place vectors in spiral plots in ), where higher place coding specificity corresponds to greater vector lengths. Indeed, we found the distribution of place vector lengths was significantly shifted to shorter values in epileptic mice, implying degraded place coding ( , cf. left and right panels). This difference was stable over multiple sessions of imaging ( ), indicating decreased specificity of place coding in epileptic animals.
We next examined if ICA-121431, which normalized the properties of dendritic spikes in vitro , also improves place coding in epileptic mice. Indeed, in the presence of ICA-121431, the precision of place coding assessed by spatial tuning vector lengths significantly increased in epileptic but not in sham-control mice (cumulative distributions of vector lengths for all cells in ). Correspondingly, ICA-121431 also caused an increase of the average vector lengths calculated per mouse in epileptic, but not in sham-control mice [ ; n = 6 and 5 mice, respectively, repeated-measures two-way ANOVA main effects, sham-control versus epileptic: F (1,9) = 6.29, P = 0.033; baseline versus ICA-121431 treatment: F (1,9) = 3.92, P = 0.079; interaction: F (1,9) = 5.64, P = 0.042; Bonferroni’s post-test sham-control versus epileptic P = 0.0095, epileptic mice before ICA-121431 treatment versus ICA-121431 treated epileptic mice P = 0.0207]. When we computed the effects on average vector lengths in each animal, the effect of ICA-121431 was significantly larger in epileptic mice, as expected ( ; Mann–Whitney U-test, P = 0.030). These experiments suggest that reversing dendritic hyperexcitability with ICA-121431 in vivo significantly reverses degraded place coding in epilepsy.
It is known that place coding can be disrupted by interictal synchronous activity. To elucidate if effects of ICA-121431 on place-related firing in epileptic animals could also be due to an inhibition of aberrant synchronous activity, we tested the effects of this compound on population activity in the CA1 region. We found aberrant synchronized activity in epileptic mice, but never in sham-control mice (an example in an epileptic mouse is shown in ). In epileptic mice, we examined the effects of ICA-121431 on short events most likely corresponding to interictal spikes and longer duration events (1–5 s and >5 s). In all three of these categories, the average number and duration of events was not affected by ICA-121431 ( ; see Supplementary material for analysis methods and legend for statistics).
### ICA-121431 normalizes hippocampal-dependent memory in epileptic mice
We next asked if reversing aberrant dendritic excitability with ICA-121431 in vivo restores performance in a hippocampus-dependent spatial memory task. We selected a task in which rodents learn the spatial arrangement of two objects. In a subsequent session, one of the objects is displaced (OLM test), and the increased exploration of the displaced object is measured. In subsequent experiments, one of the objects is exchanged for a novel object (NOR test). After habituation for 3 days, treatment was started with either vehicle or ICA-121431 during the fourth habituation day ( and ). Subsequent sessions for the OLM test were carried out either in the presence of vehicle or ICA-121431. During acquisition of the object position for the OLM test, mice did not discriminate the two objects ( , upper panels; , two-way ANOVA, n.s.). In the OLM trial with a displaced object, vehicle-treated sham-control mice discriminated the displaced object, while vehicle-treated epileptic animals did not ( , lower panels, n = 17 and 15, respectively; filled bars in ). Importantly, administration of ICA-121431 by gavage (see ‘Materials and methods’ section) recovered memory performance to control levels in epileptic animals ( , lower panels, n = 7 and 8 for sham-control and epileptic animals, respectively; empty bars in ). Two-way ANOVA revealed a main effect of vehicle versus ICA-121431 treatment [vehicle versus ICA-121431: F (1,43) = 5.95, P = 0.019; sham-control versus epileptic: F (1,43) = 1.17, P = 0.29; interaction: F (1,43) = 3.91, P = 0.054 Bonferroni’s post-tests indicated with asterisks].
In contrast to the OLM trial, NOR was not impaired in epileptic animals ( ), consistent with previous reports. ICA-121431 treatment did not alter performance in the NOR trial in either sham-control or epileptic mice ( ; n as for OLM trials; statistics, legend).
ICA-121431 normalizes hippocampal-dependent memory in epileptic mice. ( A ) Timeline of the behavioural experiments with in vivo application of ICA-121431. After habituation to the open field, the OLM, NOR and Y-maze spontaneous alternation tests were performed sequentially. ( B ) Four days of habituation in the open field, with application of either vehicle or ICA-121431 on the fourth day. Animals habituated to the open field on Days 1–3, with progressively less exploration on consecutive days. Left panel shows exploration times. Apart from the first day, the exploration times were similar with respect to distance travelled during the session and time spent in the centre of the open field in sham-control versus epileptic groups [ n = 24 and 29 in sham-control and epileptic mice, two-way ANOVA main effect, sham-control versus epileptic: F (1,51) = 8.45, P = 0.0054; Days 1–3 sessions: F (2,102) = 39.27, P < 0.0001; interaction: F (2,102) = 2.14, P = 0.12. Bonferroni’s post-test, sham-control versus epileptic Day 1 P = 0.0015, sham-control Day 1 versus Day 2 P = 0.0053; Day 1 versus Day 3 P < 0.0001; Day 2 versus Day 3 P = 0.62, travel distance in epileptic mice Day 1 versus Day 2 P < 0.0001; Day 1 versus Day 3 P < 0.0001; Day 2 versus Day 3 P = 0.39]. There was no effect of ICA-121431 application when comparing against the values from Day 3 of habituation, after habituation to the open field (two-way ANOVA, n.s.). Right panel shows time spent in the centre, >5 cm away from walls. Two-way ANOVA main effect, sham-control versus epileptic: F (1,51) = 1.2, P = 0.28; Day 1–3 sessions: F (2,102) = 7.32, P = 0.0011; interaction: F (2,102) = 0.59, P = 0.56. Bonferroni’s post-test, sham-control Day 1 versus Day 3 P = 0.02. Epileptic group Day 1 versus Day 3 P = 0.030. ( C – F ) Results of the OLM test. ( C and D ) Representative examples of session with tracking data in sham-control and epileptic animals with vehicle application ( C ) and ICA-121431 application ( D ), respectively. Occupancy times are colour-coded (for calibration see scale bar). ( E ) The discrimination index (DI) indicating if mice preferentially explore an object during the acquisition session. DI was close to zero, indicating that mice did not discriminate the two objects in any group. ( F ) DI in the OLM session. Vehicle-treated sham-control mice discriminated the displaced object, while vehicle-treated epileptic animals did not (black and red filled bars, respectively). Administration of ICA-121431 by gavage recovered memory performance to control levels in epileptic animals (empty bars). Asterisks indicate Bonferroni’s post-test, sham-control vehicle-treated versus epileptic vehicle-treated P = 0.0193; epileptic vehicle-treated versus epileptic ICA-121431-treated P = 0.0057. ( G – J ) Results of the NOR test. ( G and H ) Representative examples of sessions with tracking data in sham-control and epileptic animals with vehicle application ( G ) and ICA-121431 application ( H ), respectively. Occupancy times are colour-coded (for calibration see scale bar). ( I ) The DI indicating if mice preferentially explore an object during the acquisition session. The DI was close to zero, indicating that mice did not discriminate the two objects in any group (no differences between groups, two-way ANOVA, n.s.). ( J ) DI in the NOR session. In all groups, animals discriminated the novel object. There were no differences between groups (two-way ANOVA, n.s.). ( K ) Schematic representation of Y-maze arena. ( L ) the number of arm entries was significantly higher in epileptic animals treated with vehicle [two-way ANOVA main effect, vehicle versus ICA-121431: F (1,44) = 7.46, P = 0.0090; sham-control versus epileptic: F (1,44) = 8.47, P = 0.0056; interaction: F (1,44) = 0.29, P = 0.059, Bonferroni’s post tests indicated with asterisks, sham-control vehicle-treated versus epileptic vehicle-treated P = 0.043]. ( M ) Alternation scores. Significant Bonferroni’s post-tests indicated with asterisks, P = 0.03. For these experiments we used sham-control vehicle-treated ( n = 14), sham-control ICA-121431-treated ( n = 9), epileptic vehicle-treated ( n = 16) and epileptic ICA-121431-treated mice ( n = 9).
We next investigated if ICA-121431 affects performance in a spatial working memory task, spontaneous alternation in the Y-maze. In this task, animals freely explore a three-arm maze. The extent to which animals sequentially explore three different arms is quantified (alternation score), with higher alternation scores indicating a higher level of efficient exploration requiring spatial working memory ( ). ANOVA revealed main effects on the alternation score of kainate treatment [sham-control versus epileptic: F (1,44) = 7.36, P = 0.0095], but the effects of ICA-121431 treatment did not reach statistical significance [ F (1,44) = 2.86, P = 0.098; interaction: F (1,44) = 0.82, P = 0.59; , Bonferroni’s post-tests indicated with asterisks]. Thus, the effects of ICA-121431 seem to be most prevalent for spatial tasks that require hippocampal-dependent memory consolidation.
## Discussion
In this paper, we describe a Na channel-dependent mechanism underlying a prominent change in dendritic supralinear integration in epilepsy that degrades place coding in vivo and causes deficits in spatial memory.
Dendritic spikes are a cornerstone of dendritic integration that enable neurons to generate precise action potential outputs to clustered inputs deriving from specific populations of presynaptic input neurons. They are therefore important in mediating sharply tuned neuronal responses to features of input neurons. We show that the properties of dendritic spike generation are markedly altered in CA1 pyramidal cells in chronic epilepsy. Specifically, the fraction of dendritic branches that can give rise to dendritic spikes is almost doubled in epileptic animals. In addition, in those dendrites that generate dendritic spikes, virtually all properties that confer input-specificity to dendritic spikes are degraded. First, even very asynchronous stimulations can cause dendritic spikes in epileptic animals. Second, the threshold for eliciting dendritic spikes is significantly lowered in epileptic animals. Third, mechanisms that usually attenuate dendritic spikes upon prior occurrence of somatic or dendritic spikes are much less effective in epileptic animals. Fourth, dendritic spikes rise at an enhanced rate in first-order dendrites from epileptic animals. Because dendritic spikes with a high rate of rise (also termed ‘strong’ spikes) are much less affected by concurrently evoked inhibition, this predicts decreased inhibitory control of dendritic spikes in epileptic animals. The changes in dendritic integrative properties were primarily present in first-order and not second-order dendrites. This is reminiscent of the changes observed following environmental enrichment, in which dendritic spikes in first-order branches are selectively strengthened. In this paradigm, strengthening of proximal branches led to an enhanced propagation of distally evoked dendritic spikes into first-order dendrites. Importantly, epilepsy-related changes in dendritic integration are observed in CA1, but not in other hippocampal subregions such as the dentate gyrus.
Together, these findings suggest that input feature detection of CA1 neurons is degraded in epileptic animals, and predict much less selective task- or stimulus-specific response properties. We have used place coding as a model to test this idea. Place cells within the CA1 fire specifically when a rat occupies a particular location in the environment. Dendritic spiking has been proposed to be essential in driving formation of place-related firing in CA1 neurons, and computational studies have suggested that sharp spatial tuning of place cell responses is supported by dendritic spikes. Indeed, we have found significant broadening of place representations in epileptic animals, consistent with previous studies in different models of epilepsy. This idea is also in line with computational studies suggesting that changes causing an increase in dendritic spike generation degrade place tuning of pyramidal neurons. Importantly, the absence of dendritic spikes also degrades spatial tuning, suggesting that the prevalence and properties of dendritic spikes have to be carefully tuned for optimal place coding.
Our data suggest an upregulation of dendritic Na channels, consistent with a somato-dendritic upregulation of persistent Na channels described in a different model of experimental TLE. Our pharmacological data using antagonists with preferential activity on Na 1.2/1.6 (S-Lic) versus Na 1.3/1.1 (ICA-121431) do not allow us to unequivocally pinpoint the identity of the channel mediating enhanced excitability of first-order CA1 dendrites. However, several lines of evidence suggest that Na 1.3 upregulation is the more probable underlying mechanism. First, we and others have shown that upregulation of Na 1.3 occurs in models of acquired and genetic epilepsy, while Na 1.1. expression is reduced. Second, Na 1.3 channels exhibit rapid repriming and recovery from inactivation, as well as particularly slow closed-state inactivation, properties that are well suited to explain repetitive generation of dendritic spikes and loss of dendritic spike inactivation.
Na 1.2 has also been implicated in dendritic excitability, with loss of Na 1.2 channels decreasing action potential backpropagation into pyramidal cell dendrites. Of note, this study has mainly focused on the main apical dendritic trunk, and not the first- and second-order apical oblique dendrites that generate dendritic spikes and that we studied in our experiments. Which sodium channel subtypes drive dendritic spikes in these dendrites in control animals is still unclear. Irrespective of this question, our data show that Na 1.2 channel protein is unaltered in epileptic mice, consistent with the idea that an additional expression of Na 1.3 is the major driver in increased dendritic excitability.
Although the convergent in vitro and in vivo data point to a role of dendritic Na channels, additional changes also influence processing of CA1 neurons during navigation. Inhibition is strongly altered in chronic epilepsy, with changes in the amount and timing of inhibition delivered by inhibitory circuits, a factor that may also contribute to altered place coding. However, it has been argued that the development of interneuron loss and seizures occurs long before instability of place representations is observed, suggesting that interneuron loss alone cannot account for this phenomenon. It is nonetheless likely that changes in inhibitory circuits act together with degraded specificity of dendritic spikes to compromise the precision of CA1 firing.
Could the upregulation of dendritic Na channels be therapeutically targeted to treat cognitive comorbidities of chronic epilepsy? Several avenues have been shown to be potentially feasible. Despite the high degree of structural similarity between individual Na channel isoforms, selective antagonists for specific Na channel subunits have been developed, including for Na 1.3. In addition, antisense oligonucleotides could be used to selectively inhibit Na 1.3 expression, potentially in a cell-type specific manner. The availability of anticonvulsants that inhibit other, non-Na 1.3 Na channel isoforms, such as S-Lic, and have a proven efficacy against partial seizures may also allow us to combine blockers affecting abnormal dendritic spiking versus somatic action potential firing. In addition to applying selective Na channel modulators, the biological mechanisms leading to enhanced expression of specific Na channels could also be targeted. One mechanism underlying upregulation of Na 1.3 in acquired experimental epilepsy is a GAPDH-mediated posttranscriptional regulation. Importantly, this upregulation is ameliorated by a ketogenic diet, which is used to control therapy-refractory epilepsies in children. In addition, the anticonvulsant valproic acid has been shown to epigenetically reduce Na 1.3 expression via promoter methylation. Thus, several avenues for direct modulation of Na channels or regulation of their expression levels may be exploited to correct epilepsy-induced Na channel overexpression, and to reverse cognitive impairments in chronic TLE.
## Supplementary Material
|
Atsushi Kadowaki, Ryoko Saga, Youwei Lin, Wakiro Sato, Takashi Yamamura. Gut microbiota-dependent CCR9 CD4 T cells are altered in secondary progressive multiple sclerosis. Brain 2019; 142: 916–931, doi: .
In the original version of this article, contained incorrect versions of the heat map and IL-10 ratio graph in panel C; this has now been corrected. The authors apologise for this mistake.
Analysis of the cytokine production of CCR9 CD4 memory T cells. ( A ) Cytokine production from CCR9 (9+) or CCR9 (9−) CD4 T cells in healthy controls (HC) were measured (mean and SEM). * P < 0.05, n.s. (not significant) by Wilcoxon signed-rank test (two-sided). ( B ) CCR9+ or CCR9− Tm cells were analysed from 11 healthy controls, six RRMS, six SPMS, and five elderly healthy control subjects among Supplementary Table 2, and cell proliferation was analysed by incorporating 3H-thymidine (c.p.m.) (Supplementary Fig. 5). Ratios were calculated by (c.p.m of CCR9+ Tm cells)/(c.p.m. of CCR9− Tm cells). Box and whiskers are shown. Whiskers are drawn between 2.5–97th percentile. Not significant (n.s.) by one-way ANOVA. ( C ) Ratio of the cytokine production levels from 9+ to 9− CD4+ T cells (9+/9−) were calculated in each healthy control, RRMS, SPMS, and elderly healthy control sample from Supplementary Table 2. Outliers are determined by ROUT analysis and removed. Mean ratios are represented in a heatmap and dot plots of the 9+/9− ratio of some of the cytokines evaluated are displayed below (mean and SEM). * P < 0.01 by two-way ANOVA. P < 0.1, P < 0.05 by Tukey’s multiple comparison post-tests.
|
Spinal and bulbar muscular atrophy (SBMA) is caused by the polyglutamine androgen receptor (polyQ-AR), a protein expressed by both lower motor neurons and skeletal muscle. Although viewed as a motor neuronopathy, data from patients and mouse models suggest that muscle contributes to disease pathogenesis. Here, we tested this hypothesis using AR113Q knockin and human bacterial artificial chromosome/clone (BAC) transgenic mice that express the full-length polyQ-AR and display androgen-dependent weakness, muscle atrophy, and early death. We developed antisense oligonucleotides that suppressed AR gene expression in the periphery but not the CNS after subcutaneous administration. Suppression of polyQ-AR in the periphery rescued deficits in muscle weight, fiber size, and grip strength, reversed changes in muscle gene expression, and extended the lifespan of mutant males. We conclude that polyQ-AR expression in the periphery is an important contributor to pathology in SBMA mice and that peripheral administration of therapeutics should be explored for SBMA patients. |
The basal ganglia play a critical role in shaping motor behavior. For this function, the activity of medium spiny neurons (MSNs) of the striatonigral and striatopallidal pathways must be integrated. It remains unclear whether the activity of the two pathways is primarily coordinated by synaptic plasticity mechanisms. Using a model of Parkinson's disease, we determined the circuit and behavioral effects of concurrently regulating cell-type-specific forms of corticostriatal long-term synaptic depression (LTD) by inhibiting small-conductance Ca(2+)-activated K(+) channels (SKs) of the dorsolateral striatum. At striatopallidal synapses, SK channel inhibition rescued the disease-linked deficits in endocannabinoid (eCB)-dependent LTD. At striatonigral cells, inhibition of these channels counteracted a form of adenosine-mediated LTD by activating the ERK cascade. Interfering with eCB-, adenosine-, and ERK signaling in vivo alleviated motor abnormalities, which supports that synaptic modulation of striatal pathways affects behavior. Thus, our results establish a central role of coordinated synaptic plasticity at MSN subpopulations in motor control. |
The fidelity of inhibitory neurotransmission is dependent on the accumulation of γ-aminobutyric acid type A receptors (GABA<sub>A</sub>Rs) at the appropriate synaptic sites. Synaptic GABA<sub>A</sub>Rs are constructed from α(1-3), β(1-3), and γ2 subunits, and neurons can target these subtypes to specific synapses. Here, we identify a 15-amino acid inhibitory synapse targeting motif (ISTM) within the α2 subunit that promotes the association between GABA<sub>A</sub>Rs and the inhibitory scaffold proteins collybistin and gephyrin. Using mice in which the ISTM has been introduced into the α1 subunit (Gabra1-2 mice), we show that the ISTM is critical for axo-axonic synapse formation, the efficacy of GABAergic neurotransmission, and seizure sensitivity. The Gabra1-2 mutation rescues seizure-induced lethality in Gabra2-1 mice, which lack axo-axonic synapses due to the deletion of the ISTM from the α2 subunit. Taken together, our data demonstrate that the ISTM plays a critical role in promoting inhibitory synapse formation, both in the axonic and somatodendritic compartments. |
The Wnt family contains conserved secretory proteins required for developmental patterning and tissue homeostasis. However, how Wnt is targeted to the endoplasmic reticulum (ER) for processing and secretion remains poorly understood. Here, we report that CATP-8/P5A ATPase directs neuronal migration non-cell autonomously in Caenorhabditis elegans by regulating EGL-20/Wnt biogenesis. CATP-8 likely functions as a translocase to translocate nascent EGL-20/Wnt polypeptide into the ER by interacting with the highly hydrophobic core region of EGL-20 signal sequence. Such regulation of Wnt biogenesis by P5A ATPase is common in C. elegans and conserved in human cells. These findings describe the physiological roles of P5A ATPase in neural development and identify Wnt proteins as direct substrates of P5A ATPase for ER translocation. |
Microglia are implicated in neurodegeneration, potentially by phagocytosing neurons, but it is unclear how to block the detrimental effects of microglia while preserving their beneficial roles. The microglial P2Y<sub>6</sub> receptor (P2Y<sub>6</sub>R) - activated by extracellular UDP released by stressed neurons - is required for microglial phagocytosis of neurons. We show here that injection of amyloid beta (Aβ) into mouse brain induces microglial phagocytosis of neurons, followed by neuronal and memory loss, and this is all prevented by knockout of P2Y<sub>6</sub>R. In a chronic tau model of neurodegeneration (P301S TAU mice), P2Y<sub>6</sub>R knockout prevented TAU-induced neuronal and memory loss. In vitro, P2Y<sub>6</sub>R knockout blocked microglial phagocytosis of live but not dead targets and reduced tau-, Aβ-, and UDP-induced neuronal loss in glial-neuronal cultures. Thus, the P2Y<sub>6</sub> receptor appears to mediate Aβ- and tau-induced neuronal and memory loss via microglial phagocytosis of neurons, suggesting that blocking this receptor may be beneficial in the treatment of neurodegenerative diseases. |
Obesity comorbidities such as diabetes and cardiovascular disease are pressing public health concerns. Overconsumption of calories leads to weight gain; however, neural mechanisms underlying excessive food consumption are poorly understood. Here, we demonstrate that dopamine receptor D1 (Drd1) expressed in the agouti-related peptide/neuropeptide Y (AgRP/NPY) neurons of the arcuate hypothalamus is required for appropriate responses to a high-fat diet (HFD). Stimulation of Drd1 and AgRP/NPY co-expressing arcuate neurons is sufficient to induce voracious feeding. Delivery of a HFD after food deprivation acutely induces dopamine (DA) release in the ARC, whereas animals that lack Drd1 expression in ARC<sup>AgRP/NPY</sup> neurons (Drd1<sup>AgRP</sup>-KO) exhibit attenuated foraging and refeeding of HFD. These results define a role for the DA input to the ARC that encodes acute responses to food and position Drd1 signaling in the ARC<sup>AgRP/NPY</sup> neurons as an integrator of the hedonic and homeostatic neuronal feeding circuits. |
Neurons communicate through excitatory and inhibitory synapses. Both lines of communication are adjustable and allow the fine tuning of signal exchange required for learning processes in neural networks. Several distinct modes of plasticity modulate glutamatergic and GABAergic synapses in Purkinje cells of the cerebellar cortex to promote motor control and learning. In the present paper, we present evidence for a role of short-term ionic plasticity in the cerebellar circuit activity. This type of plasticity results from altered chloride driving forces at the synapses that molecular layer interneurons form on Purkinje cell dendrites. Previous studies have provided evidence for transiently diminished chloride gradients at these GABAergic synapses following climbing fiber activity. Electrical stimulation of climbing fibers in acute slices caused a decline of inhibitory postsynaptic currents recorded from Purkinje cells. Dendritic calcium-gated chloride channels of the type anoctamin 2 (ANO2) were proposed to mediate this short-term modulation of inhibition, but the significance of this process for motor control has not been established yet. Here, we report results of behavioral studies obtained from Ano2 <sup>-/-</sup> mice, a mouse line that was previously shown to lack this particular mode of ionic plasticity. The animals display motor coordination deficits that constitute a condition of mild ataxia. Moreover, motor learning is severely impaired in Ano2 <sup>-/-</sup> mice, suggesting cerebellar dysfunction. This reduced motor performance of Ano2 <sup>-/-</sup> mice highlights the significance of inhibitory control for cerebellar function and introduces calcium-dependent short-term ionic plasticity as an efficient control mechanism for neural inhibition. |
Bilateral volume reduction in the caudate nucleus has been established as a prominent brain abnormality associated with a FOXP2 mutation in affected members of the 'KE family', who present with developmental orofacial and verbal dyspraxia in conjunction with pervasive language deficits. Despite the gene's early and prominent expression in the cerebellum and the evidence for reciprocal cerebellum-basal ganglia connectivity, very little is known about cerebellar abnormalities in affected KE members. Using cerebellum-specific voxel-based morphometry (VBM) and volumetry, we provide converging evidence from subsets of affected KE members scanned at three time points for grey matter (GM) volume reduction bilaterally in neocerebellar lobule VIIa Crus I compared with unaffected members and unrelated controls. We also show that right Crus I volume correlates with left and total caudate nucleus volumes in affected KE members, and that right and total Crus I volumes predict the performance of affected members in non-word repetition and non-verbal orofacial praxis. Crus I also shows bilateral hypo-activation in functional MRI in the affected KE members relative to controls during non-word repetition. The association of Crus I with key aspects of the behavioural phenotype of this FOXP2 point mutation is consistent with recent evidence of cerebellar involvement in complex motor sequencing. For the first time, specific cerebello-basal ganglia loops are implicated in the execution of complex oromotor sequences needed for human speech. |
The variability in motor dysfunction is not uncommon in autoimmune disorders. Antibody-mediated system-wide malfunction or effects on the neural network shared by two independent pathophysiological processes can cause such heterogeneity. We tested this prediction for motor dysfunction during gaze holding in 11 patients with increased titers of glutamic acid decarboxylase (anti-GAD) antibody. High-resolution oculography measured horizontal and vertical eye positions. The analysis was performed with customized signal processing algorithms. Downbeat and gaze-evoked nystagmus commonly coexisted; one patient had a combination of upbeat and gaze-evoked nystagmus. The nystagmus was associated with saccadic intrusions in 10 patients; all had squarewaves, but five also had saccadic oscillations. The nystagmus and saccadic intrusions, both in the same axis of eye rotations, were not uncommon. Tandem appearance of the episodes of nystagmus and saccadic intrusions suggested a coupling between the two abnormalities. We speculated a unifying framework where the anti-GAD antibody inhibited (GAD mediated) conversion of glutamate to gamma-aminobutyric acid (GABA). Paucity GABA and excess of glutamate cause nystagmus (less GABA) and high-frequency saccadic oscillations (excessive glutamate). |
One of the most widely used experimental models for the study of learning processes in mammals has been the classical conditioning of nictitating membrane/eyelid responses, using both trace and delay paradigms. Mainly on the basis of permanent or transitory lesions of putatively-involved structures, and using other stimulation and recording techniques, it has been proposed that cerebellar cortex and/or nuclei could be the place/s where this elemental form of associative learning is acquired and stored. We have used here an output-to-input approach to review recent evidence regarding the involvement of the cerebellar interpositus nucleus in the acquisition of these conditioned responses (CRs). Eyelid CRs appear to be different in profile, duration, and peak velocity from reflexively-evoked blinks. In addition, CRs are generated in a quantum manner across conditioning sessions, suggesting a gradual neural process for their proper acquisition. Accessory abducens and orbicularis oculi motoneurons have different membrane properties and contribute differently to the generation of CRs, with significant species differences. In particular, facial motoneurons seem to encode eyelid velocity during reflexively-evoked blinks and eyelid position during CRs, two facts suggestive of a differential somatic versus dendritic arrival of specific motor commands for each type of movement. Identified interpositus neurons recorded in alert cats during classical conditioning of eyelid responses show firing properties suggestive of an enhancing role for CR performance. However, as their firing started after CR onset, and because they do not seem to encode eyelid position during the CR, the interpositus nucleus cannot be conclusively considered as the place where this acquired motor response is generated. More information is needed regarding neural signal transformations taking place in each involved neural center, and it its proposed that more attention should be paid to functional states (as opposed to neural sites) able to generate motor learning in mammals. The contribution of feedforward mechanisms normally involved in the processing activities of related centers and circuits, and the possible functional interactions within neural systems subserving the associative strength between the conditioned and unconditioned stimuli, are also considered. |
Although the neocortex in awake, adult animals is resistant to the induction of long-term potentiation (LTP), synaptic potentiation may be enhanced by rhythmic patterns of activation that evoke short- term synaptic facilitation effects. The effectiveness of stimulation patterned after the theta (4-12 Hz) EEG rhythm for the induction of LTP of sensorimotor cortex responses to corpus callosum stimulation was assessed in vivo by inducing LTP using either high- frequency (300 Hz) trains or paired trains delivered at a 100 ms (10 Hz) interval. High-frequency trains caused a reduction of the early field potential component, reflecting a potentiation of direct layer V activation, and a potentiation of the late component, reflecting enhanced polysynaptic activation in layer V. Paired trains resulted in a much larger potentiation of polysynaptic responses than was observed following 300 Hz trains. To determine if short-term facilitation effects contributed to the enhanced LTP induction by theta-patterned trains, facilitation effects induced by the trains were challenged with NMDA receptor antagonists. NMDA-receptor antagonism reduced responses to single pulses, and also reduced facilitated responses evoked by theta-patterned stimulation. The effectiveness of theta-patterned stimulation for the induction of LTP of layer V polysynaptic responses is therefore likely due to frequency-dependent synaptic facilitation effects that enhance NMDA receptor activation. |
Sports-related concussion (SRC) is a form of mild traumatic brain injury that has been linked to long-term neurological abnormalities. Australian rules football is a collision sport with wide national participation and is growing in popularity worldwide. However, the chronic neurological consequences of SRC in Australian footballers remain poorly understood. This study investigated the presence of brain abnormalities in Australian footballers with a history of sports-related concussion (HoC) using multimodal MRI. Male Australian footballers with HoC (n = 26), as well as noncollision sport athletes with no HoC (n = 27), were recruited to the study. None of the footballers had sustained a concussion in the preceding 6 months, and all players were asymptomatic. Data were acquired using a 3T MRI scanner. White matter integrity was assessed using diffusion tensor imaging. Cortical thickness, subcortical volumes, and cavum septum pellucidum (CSP) were analyzed using structural MRI. Australian footballers had evidence of widespread microstructural white matter damage and cortical thinning. No significant differences were found regarding subcortical volumes or CSP. These novel findings provide evidence of persisting white and gray matter abnormalities in Australian footballers with HoC, and raise concerns related to the long-term neurological health of these athletes. |
Previous functional magnetic resonance imaging (fMRI) studies have showed obesity (OB)-related alterations in intrinsic functional connectivity (FC) within and between different resting-state networks (RSNs). However, few studies have examined dynamic functional connectivity (DFC). Thus, we employed resting-state fMRI with independent component analysis (ICA) and DFC analysis to investigate the alterations in FC within and between RSNs in 56 individuals with OB and 46 normal-weight (NW) controls. ICA identified six RSNs, including basal ganglia (BG), salience network (SN), right executive control network/left executive control network, and anterior default-mode network (aDMN)/posterior default-mode network. The DFC analysis identified four FC states. OB compared with NW had more occurrences and a longer mean dwell time (MDT) in state 2 (positive connectivity of BG with other RSN) and also had higher FC of BG-SN in other states. Body mass index was positively correlated with MDT and FCs of BG-aDMN (state 2) and BG-SN (state 4). DFC analysis within more refined nodes of RSNs showed that OB had more occurrences and a longer MDT in state 1 in which caudate had positive connections with the other network nodes. The findings suggest an association between caudate-related and BG-related positive FC in OB, which was not revealed by traditional FC analysis, highlighting the utility of adding DFC to the more conventional methods. |
Transient neocortical events with high spectral power in the 15-29 Hz beta band are among the most reliable predictors of sensory perception. Prestimulus beta event rates in primary somatosensory cortex correlate with sensory suppression, most effectively 100-300 ms before stimulus onset. However, the neural mechanisms underlying this perceptual association are unknown. We combined human magnetoencephalography (MEG) measurements with biophysical neural modeling to test potential cellular and circuit mechanisms that underlie observed correlations between prestimulus beta events and tactile detection. Extending prior studies, we found that simulated bursts from higher-order, nonlemniscal thalamus were sufficient to drive beta event generation and to recruit slow supragranular inhibition acting on a 300 ms timescale to suppress sensory information. Further analysis showed that the same beta-generating mechanism can lead to facilitated perception for a brief period when beta events occur simultaneously with tactile stimulation before inhibition is recruited. These findings were supported by close agreement between model-derived predictions and empirical MEG data. The postevent suppressive mechanism explains an array of studies that associate beta with decreased processing, whereas the during-event facilitatory mechanism may demand a reinterpretation of the role of beta events in the context of coincident timing. |
Studying white matter connections with tractography is a promising approach to understand the development of different brain processes, such as proprioception. An emerging method is to use functional brain imaging to select the cortical seed points for tractography, which is considered to improve the functional relevance and validity of the studied connections. However, it is unknown whether different functional seeding methods affect the spatial and microstructural properties of the given white matter connection. Here, we compared functional magnetic resonance imaging, magnetoencephalography, and manual seeding of thalamocortical proprioceptive tracts for finger and ankle joints separately. We showed that all three seeding approaches resulted in robust thalamocortical tracts, even though there were significant differences in localization of the respective proprioceptive seed areas in the sensorimotor cortex, and in the microstructural properties of the obtained tracts. Our study shows that the selected functional or manual seeding approach might cause systematic biases to the studied thalamocortical tracts. This result may indicate that the obtained tracts represent different portions and features of the somatosensory system. Our findings highlight the challenges of studying proprioception in the developing brain and illustrate the need for using multimodal imaging to obtain a comprehensive view of the studied brain process. |
Development and remodeling of synaptic networks occurs through a continuous turnover of dendritic spines. However, the mechanisms that regulate the formation and stabilization of newly formed spines remain poorly understood. Here, we applied repetitive confocal imaging to hippocampal slice cultures to address these issues. We find that, although the turnover rate of protrusions progressively decreased during development, the process of stabilization of new spines remained comparable both in terms of time course and low level of efficacy. Irrespective of the developmental stage, most new protrusions were quickly eliminated, in particular filopodia, which only occasionally lead to the formation of stable dendritic spines. We also found that the stabilization of new protrusions was determined within a critical period of 24 h and that this coincided with an enlargement of the spine head and the expression of tagged PSD-95. Blockade of postsynaptic AMPA and NMDA receptors significantly reduced the capacity of new spines to express tagged PSD-95 and decreased their probability to be stabilized. These results suggest a model in which synaptic development is associated with an extensive, nonspecific growth of protrusions followed by stabilization of a few of them through a mechanism that involves activity-driven formation of a postsynaptic density. |
Dopamine (DA) exerts a strong influence on inhibition in prefrontal cortex. The main cortical interneuron subtype targeted by DA are fast-spiking gamma-aminobutyric acidergic (GABAergic) cells that express the calcium-binding protein parvalbumin. D1 stimulation depolarizes these interneurons and increases excitability evoked by current injection. The present study examined whether this direct DA-dependent modulation of fast-spiking interneurons involves DARPP-32. Whole-cell patch-clamp recordings were made from fast-spiking interneurons in brain slices from DARPP-32 knockout (KO) mice, wild-type mice, and rats. Low concentrations of DA (100 nM) increased interneuron excitability via D1 receptors, protein kinase A, and cyclic adenosine 3',5'-monophosphate in slices from both normal and DARPP-32 KO mice. Immunohistochemical staining of slices from normal animals revealed a lack of colocalization of DARPP-32 with calcium-binding proteins selective for fast-spiking interneurons, indicating that these interneurons do not express DARPP-32. Therefore, although DARPP-32 impacts cortical inhibition through a previously demonstrated D2-dependent regulation of GABAergic currents in pyramidal cells, it is not involved in the direct D1-mediated regulation of fast-spiking interneurons. |
Associative theory postulates that learning the consequences of our actions in a given context is represented in the brain as stimulus-response-outcome associations that evolve according to prediction-error signals (the discrepancy between the observed and predicted outcome). We tested the theory on brain functional magnetic resonance imaging data acquired from human participants learning arbitrary visuomotor associations. We developed a novel task that systematically manipulated learning and induced highly reproducible performances. This granted the validation of the model-based results and an in-depth analysis of the brain signals in representative single trials. Consistent with the Rescorla-Wagner model, prediction-error signals are computed in the human brain and selectively engage the ventral striatum. In addition, we found evidence of computations not formally predicted by the Rescorla-Wagner model. The dorsal fronto-parietal network, the dorsal striatum, and the ventrolateral prefrontal cortex are activated both on the incorrect and first correct trials and may reflect the processing of relevant visuomotor mappings during the early phases of learning. The left dorsolateral prefrontal cortex is selectively activated on the first correct outcome. The results provide quantitative evidence of the neural computations mediating arbitrary visuomotor learning and suggest new directions for future computational models. |
It is unclear what neural processes induce individual differences in perceptual organization in different modalities. To examine this issue, the present study used different forms of bistable perception: auditory streaming, verbal transformations, visual plaids, and reversible figures. We performed factor analyses on the number of perceptual switches in the tasks. A 3-factor model provided a better fit to the data than the other possible models. These factors, namely the "auditory," "shape," and "motion" factors, were separable but correlated with each other. We compared the number of perceptual switches among genotype groups to identify the effects of neurotransmitter functions on the factors. We focused on polymorphisms of catechol-O-methyltransferase (COMT) Val(158)Met and serotonin 2A receptor (HTR2A) -1438G/A genes, which are involved in the modulation of dopamine and serotonin, respectively. The number of perceptual switches in auditory streaming and verbal transformations differed among COMT genotype groups, whereas that in reversible figures differed among HTR2A genotype groups. The results indicate that the auditory and shape factors reflect the functions of the dopamine and serotonin systems, respectively. Our findings suggest that the formation and selection of percepts involve neural processes in cortical and subcortical areas. |
Synapsins (Syn I, Syn II, and Syn III) are a family of synaptic vesicle phosphoproteins regulating synaptic transmission and plasticity. SYN1/2 genes have been identified as major epilepsy susceptibility genes in humans and synapsin I/II/III triple knockout (TKO) mice are epileptic. However, excitatory and inhibitory synaptic transmission and short-term plasticity have never been analyzed in intact neuronal circuits of TKO mice. To clarify the generation and expression of the epileptic phenotype, we performed patch-clamp recordings in the CA1 region of acute hippocampal slices from 1-month-old presymptomatic and 6-month-old epileptic TKO mice and age-matched controls. We found a strong imbalance between basal glutamatergic and γ-aminobutyric acid (GABA)ergic transmission with increased evoked excitatory postsynaptic current and impaired evoked inhibitory postsynaptic current amplitude. This imbalance was accompanied by a parallel derangement of short-term plasticity paradigms, with enhanced facilitation of glutamatergic transmission in the presymptomatic phase and milder depression of inhibitory synapses in the symptomatic phase. Interestingly, a lower tonic GABA(A) current due to the impaired GABA release is responsible for the more depolarized resting potential found in TKO CA1 neurons, which makes them more susceptible to fire. All these changes preceded the appearance of epilepsy, indicating that the distinct changes in excitatory and inhibitory transmission due to the absence of Syns initiate the epileptogenic process. |
Metacognition, the ability to know about one's thought process, is self-referential. Here, we combined psychophysics and time-resolved neuroimaging to explore metacognitive inference on the accuracy of a self-generated behavior. Human participants generated a time interval and evaluated the signed magnitude of their temporal production. We show that both self-generation and self-evaluation relied on the power of beta oscillations (β; 15-40 Hz) with increases in early β power predictive of increases in duration. We characterized the dynamics of β power in a low-dimensional space (β state-space trajectories) as a function of timing and found that the more distinct trajectories, the more accurate metacognitive inferences were. These results suggest that β states instantiate an internal variable determining the fate of the timing network's trajectory, possibly as release from inhibition. Altogether, our study describes oscillatory mechanisms for timing, suggesting that temporal metacognition relies on inferential processes of self-generated dynamics. |
Diffusion magnetic resonance (MR) tractography represents a novel opportunity to investigate conserved and deviant developmental programs between humans and other species such as mice. To that end, we acquired high angular resolution diffusion MR scans of mice [embryonic day (E) 10.5 to postnatal week 4] and human brains [gestational week (GW) 17-30] at successive stages of fetal development to investigate potential evolutionary changes in radial organization and emerging pathways between humans and mice. We compare radial glial development as well as commissural development (e.g., corpus callosum), primarily because our findings can be integrated with previous work. We also compare corpus callosal growth trajectories across primates (i.e., humans and rhesus macaques) and rodents (i.e., mice). One major finding is that the developing cortex of humans is predominated by pathways likely associated with a radial glial organization at GW 17-20, which is not as evident in age-matched mice (E 16.5, 17.5). Another finding is that, early in development, the corpus callosum follows a similar developmental timetable in primates (i.e., macaques and humans) as in mice. However, the corpus callosum grows for an extended period of time in primates compared with rodents. Taken together, these findings highlight deviant developmental programs underlying the emergence of cortical pathways in the human brain. |
Alzheimer's disease (AD) is the most common neurodegenerative disease among the elderly with a progressive decline in cognitive function significantly affecting quality of life. Both the prevalence and emotional and financial burdens of AD on patients, their families, and society are predicted to grow significantly in the near future, due to a prolongation of the lifespan. Several lines of evidence suggest that modifications of risk-enhancing life styles and initiation of pharmacological and non-pharmacological treatments in the early stage of disease, although not able to modify its course, helps to maintain personal autonomy in daily activities and significantly reduces the total costs of disease management. Moreover, many clinical trials with potentially disease-modifying drugs are devoted to prodromal stages of AD. Thus, the identification of markers of conversion from prodromal form to clinically AD may be crucial for developing strategies of early interventions. The current available markers, including volumetric magnetic resonance imaging (MRI), positron emission tomography (PET), and cerebral spinal fluid (CSF) analysis are expensive, poorly available in community health facilities, and relatively invasive. Taking into account its low cost, widespread availability and non-invasiveness, electroencephalography (EEG) would represent a candidate for tracking the prodromal phases of cognitive decline in routine clinical settings eventually in combination with other markers. In this scenario, the present paper provides an overview of epidemiology, genetic risk factors, neuropsychological, fluid and neuroimaging biomarkers in AD and describes the potential role of EEG in AD investigation, trying in particular to point out whether advanced analysis of EEG rhythms exploring brain function has sufficient specificity/sensitivity/accuracy for the early diagnosis of AD. |
The aim of brain glioma surgery is to maximize the quality of resection, while minimizing the risk of sequelae. Due to the frequent location of gliomas in "eloquent areas" and because of major interindividual anatomofunctional variability, the cortical functional organization, effective connectivity and potential for plasticity must be studied for each patient individually. Consequently, in addition to preoperative functional neuroimaging, intraoperative electrostimulation (IES) can be used, under general anesthesia for motor mapping or on awake patient for language and cognitive mapping. This is an easy, accurate, reliable, and safe technique of detection of both cortical and subcortical functionally essential structures. Thus, IES enables: (i) to study the individual cortical functional organization before any resection; (ii) to understand the pathophysiology of areas involved by gliomas; (iii) to map the subcortical structures along the resection, allowing a study of the anatomofunctional connectivity; (iv) to analyze the mechanisms of on-line short-term plasticity, using repeated IES; (v) to tailor the resection according to individual cortico-subcortical functional boundaries, enabling to optimize the benefit:risk ratio of surgery. Moreover, IES can be combined with perioperative functional neuroimaging, before and after surgery, to validate these noninvasive techniques and to better understand the short-term and long-term plasticity mechanisms based on functional cortical reshaping and connectivity changes. Such individual knowledge allows planning multiple-stages surgery. In conclusion, IES enables to increase the impact of surgery on the natural history of gliomas, to preserve the quality of life, and to better understand the dynamic functional anatomy of the brain. |
For many years people have speculated that electroencephalographic activity or other electrophysiological measures of brain function might provide a new non-muscular channel for sending messages and commands to the external world - a brain-computer interface (BCI). Over the past 15 years, productive BCI research programs have arisen. Encouraged by new understanding of brain function, by the advent of powerful low-cost computer equipment, and by growing recognition of the needs and potentials of people with disabilities, these programs concentrate on developing new augmentative communication and control technology for those with severe neuromuscular disorders, such as amyotrophic lateral sclerosis, brainstem stroke, and spinal cord injury. The immediate goal is to provide these users, who may be completely paralyzed, or 'locked in', with basic communication capabilities so that they can express their wishes to caregivers or even operate word processing programs or neuroprostheses. Present-day BCIs determine the intent of the user from a variety of different electrophysiological signals. These signals include slow cortical potentials, P300 potentials, and mu or beta rhythms recorded from the scalp, and cortical neuronal activity recorded by implanted electrodes. They are translated in real-time into commands that operate a computer display or other device. Successful operation requires that the user encode commands in these signals and that the BCI derive the commands from the signals. Thus, the user and the BCI system need to adapt to each other both initially and continually so as to ensure stable performance. Current BCIs have maximum information transfer rates up to 10-25bits/min. This limited capacity can be valuable for people whose severe disabilities prevent them from using conventional augmentative communication methods. At the same time, many possible applications of BCI technology, such as neuroprosthesis control, may require higher information transfer rates. Future progress will depend on: recognition that BCI research and development is an interdisciplinary problem, involving neurobiology, psychology, engineering, mathematics, and computer science; identification of those signals, whether evoked potentials, spontaneous rhythms, or neuronal firing rates, that users are best able to control independent of activity in conventional motor output pathways; development of training methods for helping users to gain and maintain that control; delineation of the best algorithms for translating these signals into device commands; attention to the identification and elimination of artifacts such as electromyographic and electro-oculographic activity; adoption of precise and objective procedures for evaluating BCI performance; recognition of the need for long-term as well as short-term assessment of BCI performance; identification of appropriate BCI applications and appropriate matching of applications and users; and attention to factors that affect user acceptance of augmentative technology, including ease of use, cosmesis, and provision of those communication and control capacities that are most important to the user. Development of BCI technology will also benefit from greater emphasis on peer-reviewed research publications and avoidance of the hyperbolic and often misleading media attention that tends to generate unrealistic expectations in the public and skepticism in other researchers. With adequate recognition and effective engagement of all these issues, BCI systems could eventually provide an important new communication and control option for those with motor disabilities and might also give those without disabilities a supplementary control channel or a control channel useful in special circumstances. |
Paranormal belief and suggestibility seem related. Given our recent findings outlining a putative association between suggestibility and a specific dopaminergic genetic polymorphism, we hypothesized that similar exploratory genetic data may offer supplementary insights into a similar correlation with paranormal belief. With more affordable costs and better technology in the aftermath of the human genome project, genotyping is increasingly ubiquitous. Compelling brain theories guide specific research hypotheses as scientists begin to unravel tentative relationships between phenotype and genotype. In line with a dopaminergic brain theory, we tried to correlate a specific phenotype concerning paranormal belief with a dopaminergic gene (COMT) known for its involvement in prefrontal executive cognition and for a polymorphism that is positively correlated with suggestibility. Although our preliminary findings are inconclusive, the research approach we outline should pave the road to a more scientific account of elucidating paranormal belief. |
Visual neglect is considerably exacerbated by increases in visual attentional load. These detrimental effects of attentional load are hypothesised to be dependent on an interplay between dysfunctional inter-hemispheric inhibitory dynamics and load-related modulation of activity in cortical areas such as the posterior parietal cortex (PPC). Continuous Theta Burst Stimulation (cTBS) over the contralesional PPC reduces neglect severity. It is unknown, however, whether such positive effects also operate in the presence of the detrimental effects of heightened attentional load. Here, we examined the effects of cTBS on neglect severity in overt visual search (i.e., with eye movements), as a function of high and low visual attentional load conditions. Performance was assessed on the basis of target detection rates and eye movements, in a computerised visual search task and in two paper-pencil tasks. cTBS significantly ameliorated target detection performance, independently of attentional load. These ameliorative effects were significantly larger in the high than the low load condition, thereby equating target detection across both conditions. Eye movement analyses revealed that the improvements were mediated by a redeployment of visual fixations to the contralesional visual field. These findings represent a substantive advance, because cTBS led to an unprecedented amelioration of overt search efficiency that was independent of visual attentional load. |
It is often assumed that upright faces are represented in a holistic fashion, while representations of inverted faces are essentially part-based. To assess this hypothesis, we recorded event-related potentials (ERPs) during a sequential face identity matching task where successively presented pairs of upright or inverted faces were either identical or differed with respect to their internal features, their external features, or both. Participants' task was to report on each trial whether the face pair was identical or different. To track the activation of visual face memory representations, we measured N250r components that emerge over posterior face-selective regions during the activation of visual face memory representations by a successful identity match. N250r components to full identity repetitions were smaller and emerged later for inverted as compared to upright faces, demonstrating that image inversion impairs face identity matching processes. For upright faces, N250r components were also elicited by partial repetitions of external or internal features, which suggest that the underlying identity matching processes are not exclusively based on non-decomposable holistic representations. However, the N250r to full identity repetitions was super-additive (i.e., larger than the sum of the two N250r components to partial repetitions of external or internal features) for upright faces, demonstrating that holistic representations were involved in identity matching processes. For inverted faces, N250r components to full and partial identity repetitions were strictly additive, indicating that the identity matching of external and internal features operated in an entirely part-based fashion. These results provide new electrophysiological evidence for qualitative differences between representations of upright and inverted faces in the occipital-temporal face processing system. |
Selective attention is the process of directing limited capacity resources to behaviourally relevant stimuli while ignoring competing stimuli that are currently irrelevant. Studies in healthy human participants and in individuals with focal brain lesions have suggested that the right parietal cortex is crucial for resolving competition for attention. Following right-hemisphere damage, for example, patients may have difficulty reporting a brief, left-sided stimulus if it occurs with a competitor on the right, even though the same left stimulus is reported normally when it occurs alone. Such "extinction" of contralesional stimuli has been documented for all the major sense modalities, but it remains unclear whether its occurrence reflects involvement of one or more specific subregions of the temporo-parietal cortex. Here we employed repetitive transcranial magnetic stimulation (rTMS) over the right hemisphere to examine the effect of disruption of two candidate regions - the supramarginal gyrus (SMG) and the superior temporal gyrus (STG) - on auditory selective attention. Eighteen neurologically normal, right-handed participants performed an auditory task, in which they had to detect target digits presented within simultaneous dichotic streams of spoken distractor letters in the left and right channels, both before and after 20 min of 1 Hz rTMS over the SMG, STG or a somatosensory control site (S1). Across blocks, participants were asked to report on auditory streams in the left, right, or both channels, which yielded focused and divided attention conditions. Performance was unchanged for the two focused attention conditions, regardless of stimulation site, but was selectively impaired for contralateral left-sided targets in the divided attention condition following stimulation of the right SMG, but not the STG or S1. Our findings suggest a causal role for the right inferior parietal cortex in auditory selective attention. |
The coding of complex sounds in the early auditory system has a 'standard model' based on the known physiology of the cochlea and main brainstem pathways. This model accounts for a wide range of perceptual capabilities. It is generally accepted that high cortical areas encode abstract qualities such as spatial location or speech sound identity. Between the early and late auditory system, the role of primary auditory cortex (A1) is still debated. A1 is clearly much more than a 'whiteboard' of acoustic information-neurons in A1 have complex response properties, showing sensitivity to both low-level and high-level features of sounds. |
Recent advances in non-invasive neuroimaging have enabled the measurement of connections between distant regions in the living human brain, thus opening up a new field of research: Human connectomics. Different imaging modalities allow the mapping of structural connections (axonal fibre tracts) as well as functional connections (correlations in time series), and individual variations in these connections may be related to individual variations in behaviour and cognition. Connectivity analysis has already led to a number of new insights about brain organization. For example, segregated brain regions may be identified by their unique patterns of connectivity, structural and functional connectivity may be compared to elucidate how dynamic interactions arise from the anatomical substrate, and the architecture of large-scale networks connecting sets of brain regions may be analysed in detail. The combined analysis of structural and functional networks has begun to reveal components or modules with distinct patterns of connections that become engaged in different cognitive tasks. Collectively, advances in human connectomics open up the possibility of studying how brain connections mediate regional brain function and thence behaviour. |
The ciliated receptive endings of sensory cells and the dendrites of other neurons are shaped by adhesive interactions, many of which depend on machinery also present in epithelia. Sensory cells are shaped by interactions with support cells through adhesion junctions via the Crumbs complex, tight junction components such as claudins, as well as interactions with apical extracellular matrix composed of zona pellucida domain proteins. Neuronal dendrites are shaped by adhesion machinery that includes cadherins, catenins, afadin, L1CAM, CHL1, Sidekicks, Contactin and Caspr, many of which are shared with epithelia. This review highlights this shared machinery, and suggests that mechanisms of epithelial morphogenesis may thus provide a guide to understanding dendrite morphogenesis. |
At present there is no generally accepted theory of how cognitive phenomena arise from computations in cortex. Further, there is no consensus on how the search for one should be refocussed so as to make it more fruitful. In this short piece we observe that research in computer science over the last several decades has shown that significant computational phenomena need to circumvent significant inherent quantitative impediments, such as of computational complexity. We argue that computational neuroscience has to be informed by the same quantitative concerns for it to succeed. It is conceivable that the brain is the one computation that does not need to circumvent any such obstacles, but if that were the case then quantitatively plausible theories of cortex would now surely abound and be driving experimental investigations. |
Hebbian plasticity, a synaptic mechanism which detects and amplifies co-activity between neurons, is considered a key ingredient underlying learning and memory in the brain. However, Hebbian plasticity alone is unstable, leading to runaway neuronal activity, and therefore requires stabilization by additional compensatory processes. Traditionally, a diversity of homeostatic plasticity phenomena found in neural circuits is thought to play this role. However, recent modelling work suggests that the slow evolution of homeostatic plasticity, as observed in experiments, is insufficient to prevent instabilities originating from Hebbian plasticity. To remedy this situation, we suggest that homeostatic plasticity is complemented by additional rapid compensatory processes, which rapidly stabilize neuronal activity on short timescales. |
Learning and memory theories consider sleep and the reactivation of waking hippocampal neural patterns to be crucial for the long-term consolidation of memories. Here we propose that precisely coordinated representations across brain regions allow the inference and evaluation of causal relationships to train an internal generative model of the world. This training starts during wakefulness and strongly benefits from sleep because its recurring nested oscillations may reflect compositional operations that facilitate a hierarchical processing of information, potentially including behavioral policy evaluations. This suggests that an important function of sleep activity is to provide conditions conducive to general inference, prediction and insight, which contribute to a more robust internal model that underlies generalization and adaptive behavior. |
Synapses and neural connectivity are plastic and shaped by experience. But to what extent does connectivity itself influence the ability of a neural circuit to learn? Insights from optimization theory and AI shed light on how learning can be implemented in neural circuits. Though abstract in their nature, learning algorithms provide a principled set of hypotheses on the necessary ingredients for learning in neural circuits. These include the kinds of signals and circuit motifs that enable learning from experience, as well as an appreciation of the constraints that make learning challenging in a biological setting. Remarkably, some simple connectivity patterns can boost the efficiency of relatively crude learning rules, showing how the brain can use anatomy to compensate for the biological constraints of known synaptic plasticity mechanisms. Modern connectomics provides rich data for exploring this principle, and may reveal how brain connectivity is constrained by the requirement to learn efficiently. |
Sex differences in neural and behavioral development are integral to understanding neurodevelopmental, mental health, and neurodegenerative disorders. Much of the literature has focused on late prenatal and early postnatal life as a critical juncture for establishing sex-specific developmental trajectories, and data are now clear that immune signaling plays a central role in establishing sex differences early in life. Adolescence is another developmental period during which sex differences arise. However, we know far less about how immune signaling plays a role in establishing sex differences during adolescence. Herein, we review well-defined examples of sex differences during adolescence and then survey the literature to speculate how immune signaling might be playing a role in defining sex-specific adolescent outcomes. We discuss open questions in the literature and propose experimental design tenets that may assist in better understanding adolescent neurodevelopment. |
Intact phonological processing is crucial for successful literacy acquisition. While individuals with difficulties in reading and spelling (i.e., developmental dyslexia) are known to experience deficient phoneme discrimination (i.e., segmental phonology), findings concerning their prosodic processing (i.e., suprasegmental phonology) are controversial. Because there are no behavior-independent studies on the underlying neural correlates of prosodic processing in dyslexia, these controversial findings might be explained by different task demands. To provide an objective behavior-independent picture of segmental and suprasegmental phonological processing in impaired literacy acquisition, we investigated event-related brain potentials during passive listening in typically and poor-spelling German school children. For segmental phonology, we analyzed the Mismatch Negativity (MMN) during vowel length discrimination, capturing automatic auditory deviancy detection in repetitive contexts. For suprasegmental phonology, we analyzed the Closure Positive Shift (CPS) that automatically occurs in response to prosodic boundaries. Our results revealed spelling group differences for the MMN, but not for the CPS, indicating deficient segmental, but intact suprasegmental phonological processing in poor spellers. The present findings point towards a differential role of segmental and suprasegmental phonology in literacy disorders and call for interventions that invigorate impaired literacy by utilizing intact prosody in addition to training deficient phonemic awareness. |
Trust and cooperation increase from adolescence to adulthood, but studies on gender differences in this development are rare. We investigated gender and age-related differences in trust and reciprocity and associated neural mechanisms in 43 individuals (16-27 years, 22 male). Participants played two multi-round trust games with a cooperative and an unfair partner. Males showed more basic trust towards unknown others than females. Both genders increased trust during cooperative interactions, with no differences in average trust. Age was unrelated to trust during cooperation. During unfair interactions males decreased their trust more with age than females. ROI analysis showed age-related increases in activation in the temporo-parietal junction (TPJ) and dorsolateral prefrontal cortex (dlPFC) during cooperative investments, and increased age-related caudate activation during both cooperative and unfair repayments. Gender differences in brain activation were only observed during cooperative repayments, with males activating the TPJ more than females, and females activating the caudate more. The findings suggest relatively mature processes of trust and reciprocity in the investigated age range. Gender differences only occur in unfair contexts, becoming more pronounced with age. Largely similar neural activation in males and females and few age effects suggest that similar, mature cognitive strategies are employed. |
Adolescence is a period characterised by increases in risk-taking. This behaviour has been associated with an imbalance in the integration of the networks involved in cognitive control and motivational processes. We examined whether the influence of emotional cues on cognitive control differs between adolescents who show high or low levels of risk-taking behaviour. Participants who scored especially high or low on a risky decision task were subsequently administered an emotional go/no-go fMRI task comprising angry, happy and calm faces. Both groups showed decreased cognitive control when confronted with appetitive and aversive emotional cues. Activation in the inferior frontal gyrus (IFG) increased in line with the cognitive control demands of the task. Though the risk taking groups did not differ in their behavioural performance, functional connectivity analyses revealed the dorsal striatum plays a more central role in the processing of cognitive control in high than low risk-takers. Overall, these findings suggest that variance in fronto-striatal circuitry may underlie individual differences in risk-taking behaviour. |
Multiplication tables are typically memorized verbally, with fluent retrieval leading to better performance in advanced math. Arithmetic development is characterized by strategy shifts from procedural operations to direct fact retrieval, which would not necessitate access to the facts' conceptual meaning. This study tested this hypothesis using a combination of event related brain potentials (ERP) and behavioral measures with 3rd-5th grade children and young adults. Participants verified the solutions to simple multiplication problems (2 × 3 = 6 or = 7) and the semantic fit of word-picture pairs, separately. Children showed an N400 effect to multiplication solutions with larger (more negative) amplitude for incorrect than correct solutions, reflecting meaning-level processing. A similar ERP response was observed in the word-picture verification task, with larger negative amplitude for word-picture pairs that were semantically mismatched compared to matched. In contrast, adults showed a P300 response for correct solutions, suggesting that they treated these solutions as potential targets in over-rehearsed mathematical expressions. This P300 response was specific to math fact processing, as the word-picture verification task elicited a classic N400 in adults. These ERP findings reveal an overlooked developmental transition that occurs after fifth grade, and speak to theories of arithmetic that have been based primarily on adult data. |
Highlights
We compare gyrification and surface area in prefrontal and parietal cortex in young cannabis users and controls.
Frequent cannabis use was associated with reduced gyrification in prefrontal subregions.
Reduced gyrification in cannabis users was associated with poorer performance.
Sensitive periods during neurodevelopment may be affected by frequent cannabis use.
## Background
Regions undergoing maturation with CB1 receptors may be at increased risk for cannabis-induced alterations. Here, we examine the relationships between cannabis use and prefrontal (PFC) and inferior parietal gyrification and surface area (SA) in youth.
## Methods
Participants included 33 cannabis users and 35 controls (ages 18–25). Exclusions included co-morbid psychiatric/neurologic disorders and heavy other drug use. Multiple regressions and Pearson r correlations examined the effects of cannabis use on gyrification, SA and cognition.
## Results
Cannabis use was associated with decreased gyrification in: ventral-medial PFC (RH: [FDR corrected p = .02], LH: [FDR corrected p = .02]); medial PFC (RH: [FDR corrected p = .02], LH: [FDR corrected p = .02]); and frontal poles (RH: [FDR corrected p = .02], LH: [FDR corrected p = .02]). No differences were observed in bilateral hemispheres, PFC, dorsolateral, ventrolateral, or inferior parietal ROIs. Cannabis use was associated with marginally decreased SA in left: medial PFC [FDR corrected p = .09], and ventral lateral PFC: [FDR corrected p = .09]. In cannabis users, increased gyrification was associated with improved working-memory performance in right medial ( p = .003), ventral-medial ( p = .03), and frontal pole ROIs ( p = .007).
## Conclusions
Cannabis use was associated with reduced gyrification in PFC regions implicated in self-referential thought and social cognition. Results suggest that these gyrification characteristics may have cognitive implications.
## Introduction
Cannabis is the second most used drug after alcohol, with 22.9% of high school seniors and 20% of college students using in the past month, and perhaps most alarmingly, one in every 15 seniors reporting daily use ( ). Cannabis legislation changes are sweeping across the United States. Policy experts predict that increased access and reduced price will lead to increased usage, especially in young adults who are the heaviest users ( ). Late adolescence and emerging adulthood is a period of ongoing neurodevelopment, with pruning of inefficient gray matter connections ( , ). Healthy adult rats demonstrate enhanced binding of cannabinoid (CB ) receptors within areas such as the prefrontal cortex (PFC) ( ) in comparison to juveniles, suggesting increased reliance upon the cannabinoid system with age. Indeed, converging lines of animal and human evidence have suggested that this is a sensitive period that may be particularly vulnerable to cannabis-induced neurocognitive effects ( , ; see for review).
Preclinical animal models suggest that endogenous endocannabinoid signaling in the PFC influences executive functioning (EF) performance (for review see ). In humans, significant CB receptor density has been measured in the PFC, a region associated with mood regulation and EF, and throughout the cortex ( , ; see ). Therefore, disruption of the endogenous cannabinoid system during adolescence may particularly impact the integrity later developing regions, such as the PFC and parietal lobes ( , ). Indeed, daily cannabis users demonstrate significant, though reversible, downregulation of the CB density in PFC and other cortical regions including the parietal lobes ( ). Further, cannabis-using youth demonstrate impairments in executive functioning, including complex attention, inhibitory control, and working memory ( , , , ).
Previous structural magnetic resonance imaging (MRI) research has demonstrated that regular (weekly or more) cannabis using adolescents demonstrate larger PFC (including orbitofrontal cortex) volume in female cannabis users ( ) and reduced medial orbitofrontal volumes in a primarily male sample ( ). Our group has found reduced medial orbitofrontal (mOFC) and inferior parietal volumes in this same sample of young adults ( ) compared to controls, and other groups have found that earlier age of onset significantly predicted decreased right superior PFC thickness ( ). Recent MRI advances have yielded new measurements of cortical architecture that may be more sensitive to drug effects than volume or cortical thickness. One such candidate is local gyrification index, or a 3-dimensional ratio representing the degree of folding on the outer surface relative to buried cortex within neighboring sulci, which may also be calculated for regions of interest ( , ). Several candidate theories attempt to explain the primary driving mechanisms of gyrification development, including cortico-cortical mechanical tension, morphogenetic, and differential cortical expansion rate influences ( ; see , , , ; see , , for reviews). Another measure is cortical surface area (SA), which is a reflection of the amount of area on the cortical surface represented in mm ( ).
Age-related changes in cortical surface area (SA) and other surface characteristics, including gyral and sulcal shape, have been noted in several preliminary studies. measured SA changes between MRI scans in 504 subjects. Results from the study found age-related changes in SA such that adolescence is a period in which the cortex is greatly expanding and reaches the maximum individual peak in SA during this time. Further, the same study found that those with the highest IQ had the greatest rate of cortical SA change during this period. found a significant relationship between age with gyral and sulcal shape in a sample of 148 participants aged 18–82. A more recent two-year longitudinal study with 52 participants found overall decreases in gyrification index in youth who were between the ages of 11 and 17 at baseline, with significant widening of sulci and loss of SA within the frontal cortex ( ). Other samples have found reduced PFC surface complexity in teens compared to children ( ), and reduced PFC gyrification in young adults compared to early teens ( ). Further, increased gyrification has been associated with enhanced vocabulary knowledge in typically developing youth ( ). In a large cohort of 322 healthy adults spanning ages 20–85, SA decreases were most robust within the dorsomedial frontal, and PFC gyrification decreases were observed with older age ( ). Sex differences in folding have also been noted with females demonstrating greater gyrification in PFC compared to males ( , ). Lastly, a large longitudinal study in 647 participants found an inverted-U shaped trajectory of SA maturation between the ages of 3 and 30 ( ). Changes in SA appeared to peak later than cortical thickness in the large cohort. The same study found that gyrification index ( note : this index differs from the LGI measure) and convex hull area influence SA changes during early to late adolescents; however, late adolescent changes in SA may be most attributed by reductions in gyrification in comparison to reduced convex hull area. Further, SA may peak at later developmental periods compared to other cortical measures such as volume ( ). Preliminary evidence suggests that later developing regions, such as the PFC ( ), continue to undergo gyrification, cortical surface shape, and SA changes during adolescents and young adulthood.
While several studies have demonstrated a great degree of genetic influences on cortical thickness, gray, and white matter volume (see ), studies of gyrification or surface characteristics among small samples of monozygotic (MZ) twins demonstrate observable differences ( , , , ), suggesting that environmental factors may influence the shape of the cortical surface (see ) especially in secondary and tertiary sulci ( ). For example, found that PFC gyrification was no more similar in MZ twins compared to dizygotic twins. Therefore, compared to other brain characteristics, such as gray and white matter volume, surface morphometry values (including gyrification) appear to be significantly influenced by environmental factors compared to genetics, although this needs to be confirmed in larger sample sizes.
Therefore, gyrification may reflect changes sensitive to repeated behavioral or environmental influences, such as substance use, although additional research in emerging adults is needed. To our knowledge, only one study has examined surface morphology in a sample of young cannabis users ( ). examined sulcal concavity, a measure similar yet distinct from a 3-dimensional gyrification value. The study noted decreased sulcal concavity in the left PFC and bilateral temporal lobes of young adult cannabis users compared to controls ( ). The study also failed to find any significant differences in global SA after controlling for potential confounds, suggesting a unique characteristic of sulcal curvature differences in regions undergoing neuromaturation in young cannabis users ( ). The same study did not examine sub-regional differences in SA or how sulcal differences between cannabis users and non-users relate to downstream behavioral phenotypes, such as neuropsychological function.
Because cannabis use has an age of onset (SAMHSA, 2014) that overlaps with continued PFC gyrification development ( , ), examining the impact of cannabis use on gyrification remains an important area to investigate. The current study examined whether cannabis use status predicted PFC or parietal gyrification in a sample of adolescents and emerging adults. Surface morphology may be related to cortical thickness and volume ( ). Given that both reductions in cortical thickness and volume ( , ) and reductions in PFC sulcal concavity ( ) were previously found in young cannabis users, we predicted that cannabis users would demonstrate reduced gyrification and SA in PFC and parietal regions. Reduced SA and gyrification may be most pronounced in both inferior frontal and parietal regions that show reductions in volume ( , ) Within regions that differed between cannabis users and controls, follow-up analyses examined brain–behavior relationships in both groups.
## Materials and methods
### Participants
Participants included 68 (33 cannabis-users, 35 controls) right-handed adolescents and emerging adults between the ages of 18–25 (21 male and 12 female cannabis-users; 15 male and 20 female controls) from a larger imaging genetics study (PI: Lisdahl, NIH R03 DA027457). Exclusion criteria included MRI contraindications; history of chronic medical or neurologic illness or injury (meningitis, HIV, epilepsy, brain tumor, traumatic brain injury, >2 min of unconsciousness and concussion symptoms, stroke, cerebral palsy, Parkinson's disease, Huntington's disease, high blood pressure, diabetes, chronic migraines); history of a learning disability; complications during birth/premature birth; prenatal exposure to alcohol (>4 drinks/day or >7 drinks/week) or illicit drugs (>10 uses); current use of psychoactive medication; preexisting DSM-IV Axis I disorders independent of substance use; current pregnancy; >20 lifetime use occasions of any of the following drug categories (stimulants, ecstasy, inhalants, hallucinogens, sedatives, or opiates); and refusal to remain abstinent from all drugs and alcohol for at least seven days. Individuals that classified as very heavy alcohol drinkers (>8 standard drinks per week on average) were also excluded. Eligible participants consisted of two groups, cannabis users (>25 past year and >50 lifetime joints) and controls (≤5 past year and <15 lifetime joints). Groups were matched as closely as possible on age, education, ethnicity, gender, and verbal IQ.
### Procedures
The Institutional Review Board at the University of Cincinnati approved all aspects of this study. Participants were recruited through advertisements in a local free newspaper and fliers. Those interested were screened by phone for exclusionary criteria, which have been described in further detail elsewhere (see ; ). Briefly, a semi-structured interview based on DSM-IV-TR criteria for Axis I psychotic, anxiety, and mood disorders was administered (see ). Next, eligible participants completed either one or two sessions. Those with moderate or greater substance use completed the questionnaires, drug use interview, neuropsychological battery, and MRI scan in two sessions (typically 2–3 days apart; see biological samples below). Participants were paid $160 for two sessions (5.5 h) ($110 for one, 3.5 h) and received parking reimbursement, local substance treatment resources and images of their brain.
### Screening inventories and questionnaires
#### Demographic information
Participants completed a Background Questionnaire outlining demographic variables (see ).
Demographic, substance use information according to group.
#### Biological samples
Participants were administered a urine toxicology screen using the One Step Drug Screen Test, a breathalyzer test, and female participants were administered a pregnancy test. Those who tested positive for drugs and/or alcohol except cannabis and nicotine were excluded. Metabolite levels were further examined for participants that tested positive for cannabis via mass spectrometry testing. Session 2 total THC metabolite ratios, controlling for creatinine, were subtracted from session 1 total ratios to ensure there were no increases or current use while in the study (see ). Thus, if the difference in ratios reached >50 ng/mL, the participant was excluded. Cotinine levels measured recent nicotine use or exposure.
#### Drug use
Past year drug use was measured using a modified version of the Time-Line Follow-Back ( ) interview (see for categories). Drug use was measured by the number of standard units (cigarettes or cigars for nicotine; standard drinks for alcohol; joints for cannabis; tablets for ecstasy; grams for stimulants; number of hits or pills for inhalants, hallucinogens, and opioids; and pills or hits for sedatives). The Customary Drinking and Drug Use Record (CDDR), measured lifetime and past 3-month substance use, withdrawal symptoms, DSM-IV abuse and dependence criteria, and substance-related difficulties ( , ).
#### Self-reported mood
The Beck Depression Inventory-II ( ) assessed current depressive symptomology.
### Neuropsychological assessments
#### Premorbid verbal intelligence/quality of education
The Wide Range Achievement Test-4th edition (WRAT-4) Reading subtest ( ) measured estimates of verbal intelligence and quality of education for group comparison purposes (see ).
#### Complex attention
Complex attention was assessed using the total correct responses in the Wechsler Adult Intelligence Scale – Third Edition (WAIS-III) Letter Number Sequencing (LNS) and the Paced Auditory Serial Attention Test (PASAT). The LNS is a subscale of the WAIS-III Working Memory Index and measures the ability to retain and manipulate bits of information over several separate trials ( ). The PASAT is a working memory task in which participants must retain two serially presented numbers and perform a summation roughly every 2 s ( ). LNS and PASAT total scores were used for the current study.
#### Cognitive inhibition
The D-KEFS Color Word Interference Test Inhibition condition total completion time assessed inhibitory ability ( ). For this task, participants were required to read the color of ink a color word is printed in (inhibition condition).
### MRI data acquisition
#### Parameters
T1-weighted, 3-D SPGR anatomical brain scans were obtained on a 4T Varian Unity MRI scanner using a modified driven equilibrium Fourier transform (MDEFT) sequence (FOV = 25.6 cm, 256 × 256 × 192 matrix, slice thickness = 1 mm, in-plane resolution = 1 × 1 mm, TR = 13 ms, TE = 5.3 ms, flip angle = 22°). A neuroradiologist at the Center for Imaging Research reviewed anatomical scans, and participants with noted abnormalities were excluded from this sample.
### MRI processing
#### PFC local gyrification analysis
Images were preprocessed in FreeSurfer version 5.3 ( ). Average local gyrification indices (LGI) were created using a radius set to 20 mm for each region listed below, in order to maximize sensitivity ( , ), and cortical surface-based anatomical atlas ( ). Regions of interest (ROIs) included bilateral: dorsal lateral PFC (DLPFC); medial PFC (mPFC); frontal pole; ventral medial PFC (vmPFC); ventral lateral PFC (vlPFC); and inferior parietal (infPariet). Control regions reflecting the average LGI and SA for each of the left and right hemispheres was included to test whether results were diffuse or specific to a priori defined ROIs.
#### Surface area analysis
As part of the FreeSurfer processing stream ( ) SA was computed for each participant. Corresponding with the ROIs listed above in the LGI analysis (see Section ), SA was calculated for all ROIs and the hemisphere control regions.
#### Operating system
Mac Pro with: OS X version 10.6.8, 12GB of memory, and 2×2.26 GHz Quad-Core Intel Xeon.
### Statistical analyses
All analyses were conducted using SPSS. ANOVAs, Mann–Whitney U -test (drug variables), and Chi-square tests were run to examine potential demographic differences as well as differences in past year drug use histories between drug groups. Variables that either significantly differed between groups or may impact neural architecture were entered as covariates ( , ). Covariates included WRAT-4 Reading scaled score, age, gender, past year alcohol use, cotinine levels, and current depressive symptoms.
General linear modeling (GLM) in SPSS was used to examine whether cannabis group status was significantly associated with a priori defined LGI or SA ROIs. Standard least squares multiple regression was used; block one included covariates, and block two included cannabis group status. All dependent variables were normally distributed and there was no evidence of multicollinearity. Significance was determined if p < .05, and correction for multiple comparisons was calculated for each hemisphere's results utilizing Benjamini and Hochberg's False Discovery Rate correction (FDR; ). All FDR corrections were computed for the left and right hemispheres separately.
In the cannabis users, Pearson r correlations were run between cognitive performance (complex attention and cognitive inhibition; see ) and gyrification or SA ROIs that significantly differed between groups. Significance was determined if p < .05 (after FDR correction).
## Results
### Demographic and mood information
#### Demographic and self-report variables
ANOVAs and Chi-squares were run to test differences between cannabis users and controls. There were significant differences in self-reported BDI-II depressive symptoms, with cannabis users reporting on average 2 more symptoms than controls, but still within the minimal range of symptoms [ F (1,66) = 4.24, p = .04]. Groups did not differ in ethnicity [22 Caucasian cannabis users and 23 Caucasian controls [ X (4) = 3.86, p = .43], gender [ X (1) = 2.9, p = .09], past year Cahalan alcohol drinking patterns criteria [ X (5) = 4.3, p = .51], age [ F (1,66) = .02, p = .90], WRAT-4 Reading standard score [ F (1,66) = .39, p = .54], education [ F (1,66) = 3.6, p = .06], annual income [ F (1,66) = .17, p = .68], or body mass index [ F (1,65) = .46, p = .50].
#### Drug variables
Mann–Whitney U tests revealed significant differences between cannabis users and controls in past year nicotine ( U = 244.5, p < .01), recent nicotine use ( U = 199.5, p ≤ .01), past year alcohol use ( U = 334.5, p = .003), past year cannabis use ( U = 0.00, p ≤ .01), and past year other drug use (measured as standardized hits or pills of stimulants, ecstasy, inhalants, hallucinogens, sedatives and opiates; U = 236, p ≤ .01). The cannabis group used more of these substances in comparison to controls, although the other drug use category was relatively low for the vast majority of the cannabis users and our exclusion criteria consistent with ≤ 20 lifetime uses of any drug category.
### Gyrification results
#### Cannabis group
After controlling for WRAT-4 Reading scaled score, age, gender, past year alcohol use, cotinine levels, and current depressive symptoms, cannabis users demonstrated significantly reduced gyrification in bilateral medial PFC (Right: [ t (59) = −2.9, beta = −.41, p = .005; FDR corrected p = .02] and Left: [ t (59) = −3.1, beta = −.45, p = .003; FDR corrected p = .02]); bilateral frontal poles (Right: [ t (59) = −2.7, beta = −.38, p = .009; FDR corrected p = .02] and Left: [ t (59) = −3.1, beta = −.46, p = .003; FDR corrected p = .02]); and bilateral ventral-medial PFC (Right: [ t (59) = −2.8, beta = −.40, p = .006; FDR corrected p = .02] and Left: [ t (59) = −3.0, beta = −.44, p = .004; FDR corrected p = .02]) (see ).
ROI's of cannabis users with significantly reduced LGI and the corresponding p values and FDR corrected p values corrected for multiple comparisons. Note : yellow = medical PFC; blue = frontal pole; orange = ventral medial PFC; Lh = left hemisphere; Rh = right hemisphere; LGI = local gyrification index.
No significant group differences were observed in LGI for total hemisphere (control region) (Right: [ t (58) = −1.6, beta = −.25, p = .11] and Left: [ t (58) = −.62, beta = −.09, p = .54]); dorsolateral PFC (Right: [ t (59) = .05, beta = .007, p = .96] and Left: [ t (59) = 1.5, beta = .2, p = .15]); ventral lateral PFC (Right: [ t (59) = −1.5, beta = −.23, p = .13] and Left: [ t (59) = −.7, beta = −.11, p = .49]); or bilateral inferior parietal cortex (Right: [ t (60) = −1.9, beta = −.28, p = .06] and Left [ t (60) = −.41, beta = −.06, p = .69]). There were no regions where cannabis users showed significant increases compared to controls.
### Surface area results
#### Cannabis group
After controlling for WRAT-4 Reading scaled score, age, gender, past year alcohol use, cotinine levels, and current depressive symptoms, cannabis users demonstrated significantly reduced SA in the left ventral medial [ t (60) = −2.5, beta = −.35, p = .02; FDR corrected p = .09], and left ventral lateral PFC [ t (60) = −2.7, beta = −.32, p = .008; FDR corrected p = .09], although these findings were only marginally significant after correcting for multiple comparisons. No significant group differences were observed in total hemisphere (control region) SA (Right: [ t (59) = −1.1, beta = −.12, p = .26] and Left: [ t (59) = −1.9, beta = −.2, p = .07]); bilateral medial PFC (Right: [ t (60) = −1.0, beta = −.12, p = .30] and Left: [ t (60) = −1.8, beta = −.22, p = .08]); bilateral dorsolateral PFC (Right: [ t (60) = −.65, beta = −.07, p = .52] and Left: [ t (60) = −1.7, beta = −.19, p = .10]); right ventral medial PFC [ t (60) = −1.5, beta = −.21, p = .14]; right ventral lateral PFC [ t (60) = −1.4, beta = −.16, p = .18]; or bilateral inferior parietal (Right: [ t (60) = −1.3, beta = −.15, p = .20] and Left: [ t (60) = −.53, beta = −.07, p = .60]).
### Brain–behavior results
#### Cannabis group
Positive correlations were found between increased gyrification and improved LNS performance in the right medial PFC [ r = .50, n = 33, p = .003], right ventral medial PFC [ r = .38, n = 33, p = .03] and right frontal pole [ r = .46, n = 33, p = .007].
#### Controls
No significant correlations were observed between brain regions that significantly differed between groups and neuropsychological performance in controls.
## Discussion
This study examined whether cannabis use status predicted prefrontal or parietal local gyrification index (LGI) and surface area (SA) in a sample of otherwise healthy adolescents and emerging adults. Consistent with the predicted hypotheses, after controlling for reading ability, age, gender, past year alcohol use, cotinine levels, and current depressive symptoms, cannabis users had reduced LGI in bilateral medial frontal, ventral medial, and frontal poles. No significant differences were found in hemispheric or inferior parietal LGI, suggesting that aberrant gyrification may be localized to particular PFC regions in emerging adults. Further, group differences in SA in orbitofrontal areas were consistent with LGI findings, but did not pass correction for multiple comparisons.
Decreased gyrification in right medial, ventral medial, and frontal pole regions, were associated with poorer performance on complex attention in cannabis users, suggesting that reduced gyrification confers a functional deficit. This is consistent with previous studies suggesting increased gyrification is associated with better cognitive functioning ( ) and may reflect improved cognitive control ( ).
Present findings are consistent with prior research demonstrating unique PFC surface morphology characteristics in cannabis using youth ( ). Specifically, found reduced sulcal concavity in the PFC of cannabis users in comparison to non-users and failed to find global hemispheric differences in SA. In the current study we found significantly reduced LGI in medial, ventral medial, and frontal poles in cannabis users compared to controls. We found no significant differences in inferior parietal LGI and marginal differences in SA, while in an overlapping sample we previously reported subtle volume abnormalities in this region ( ). Though we did not examine the relationship between either LGI or SA and other cortical measures in this study, surface area, gyrification, and cortical thickness appear have distinct patterns in neurodevelopment from ages 6 to 22 ( ). We also found unique patterns in cannabis effects between two cortical morphometry measures; after controlling for covariates including age and gender, results from the current study suggest that frequent cannabis use may influence LGI in a more diffuse PFC distribution compared to SA since we found only marginal reductions of SA in two PFC regions (left: ventral lateral and ventral medial PFC) among cannabis users compared to controls. Therefore, while gyrification may be partially related to gray matter volume and SA, it likely reflects a novel measure of brain maturation ( ). Though did not find global hemispheric group differences in SA, perhaps the influence of frequent cannabis use on SA is restricted to regions with later SA development. Changes in global SA during late adolescents may be primarily driven by reduced global gyrification index ( ) and may differ from influences driving cortical thickness maturation ( ). Future studies may want to examine how cannabis use impacts neurodevelopment utilizing multiple measures of cortical morphometry (LGI, cortical thickness, volume, and SA).
Frequent cannabis-using youth report using cannabis to cope with stressors or relax ( , , , , ), although continued use may negatively impact regions underlying healthy affective processing ( ). For example, the medial portions of the PFC are implicated in self-referential thought, regulation of stress response, autonomic regulation, emotional processing, and social cognition ( , , ; for reviews see , ). Ventral medial portions of the PFC play a role in regulating amygdala activity, contextual decision-making, fear response and extinction, anticipatory responses, and social processing ( , , , , , ). Animal studies suggest that the inferior frontal regions also play a vital role in insight or one's ability to imagine consequences of behavior in new situations ( ). The frontal pole underlies detecting contextual change, and reward-related decision-making ( , ). Therefore, additional studies examining functional consequences of cannabis use in youth may focus on affective processing, reward processing, and mood symptomatology. In addition, given the potential impact of endocannabinoid signaling on PFC activation ( ), future studies may want to examine whether genotypes related to endocannabinoid signaling interact with cannabis exposure to predict frontolimbic structural integrity in youth.
Further, there is also evidence of functional abnormalities as evidenced on fMRI studies of inhibitory control and complex working memory (see , , ). In an overlapping sample, our lab previously reported smaller inferior parietal volumes with small effect size among cannabis using emerging adults, although this finding did not survive multiple corrections (see ). Taken together, later to develop PFC regions may be more consistently susceptible to morphological abnormalities associated with cannabis use during youth.
The current study was cross-sectional; therefore, it is not possible to determine whether the LGI and SA results reflect premorbid characteristics that co-occur with the onset of cannabis use. For example, there is evidence that smaller OFC volumes and increased impulsivity predict earlier age of cannabis use onset during adolescence ( ). However, there are no studies to date that have examined whether other cortical morphometric measures (SA or gyrification) predict the onset of cannabis use, so it is not known whether these measures are sensitive premorbid markers for addiction risk, or cannabis use potential. On the other hand, it is not known whether these morphological features are more sensitive to environmental influence later in the course of development compared to volume or cortical thickness. Therefore, larger prospective longitudinal studies are needed to examine the influence of cannabis use on multiple morphometric measures (volume, LGI, SA, cortical thickness).
There are other limitations that need to be considered. Alcohol and cannabis use among youth are highly comorbid ( ). Although the current study excluded “very heavy” drinkers, statistically controlled for past year alcohol use, and did not find any relationship between past year alcohol use and gyrification or SA, it is possible that some of the findings are associated with combined or simultaneous cannabis and alcohol use. Lastly, due to the potential impact of various workstations on results ( ) results of the current study may be specific to the particular operating system, FreeSurfer version, imaging acquisition and preprocessing. Thus, replication using different imaging parameters or processing techniques is warranted.
## Conclusion
In conclusion, this study found that regular cannabis users had less complex PFC gyrification, especially in medial and ventral medial regions. Cannabis users also demonstrated marginal reductions in orbitofrontal surface area. Reduced gyrification was significantly correlated with poorer working memory. These findings may reflect alterations in synaptic connections, resulting in reduced prefrontal complexity and poorer cognitive functioning in adolescent onset cannabis users. This adds to converging lines of evidence that suggest that adolescence and emerging adulthood is a sensitive period for drug-induced neurocognitive effects. Understanding the impact of regular cannabis use on neurodevelopment during adolescence and emerging adulthood remains a significant public health priority and additional prospective longitudinal studies are warranted.
## Conflict of interest
None declared.
|
Developmental cognitive neuroscience is a truly interdisciplinary field of research that has the potential to answer critical questions about neural plasticity and neural substrates of learning and behavior across cognitive, affective, and social domains of functioning. It therefore has the potential to not only help us understand trajectories and mechanisms of typical development, but also translate this knowledge to the prevention and treatment of emerging psychopathology and health-risking behaviors. However, to reach these goals our field must be able to model how these processes change within individuals across time. Given how central this methodological issue is to our endeavours, it is surprising that there has been relatively little attention paid to integrating neuroscientific methods with cutting edge statistical techniques for modelling longitudinal change, nor have there been published methodological guidelines on many relevant topics. The current special issue sets out to begin to address this lacuna.
Many techniques have been employed to examine the brain across development, including (but not limited to) familiar modalities like structural, functional, and diffusion magnetic resonance imaging (s/f/d MRI), as well as less widespread ones like functional near infrared spectroscopy (fNIRS) or magnetic resonance elastography (MRE). Each of these methods addressed in the special issue have unique strengths and limitations that shape recommendations for data acquisition and analysis, and provide different information about normative and atypical trajectories of brain development. Importantly, researchers are increasingly relying on longitudinal data sets to investigate change within individuals. However, longitudinal studies require special consideration in design, as well as data acquisition, processing, analysis, and interpretation. Despite increasing acknowledgement of methodological issues across modalities and study designs in the cognitive neurosciences, there are relatively limited guidelines available to provide best practices, particularly with developmental populations. This lack of consensus could be contributing to inconsistencies in the literature. For example, recent studies have found that differences in sample composition, quality control procedures, and data analytic approaches affect observed trajectories of brain development ( , ). Further, it is unclear how these factors affect associations between brain development and cognition or other behavior.
This special issue of Developmental Cognitive Neuroscience, “Methodological Challenges in Developmental Neuroimaging: Contemporary Approaches and Solutions,” presents papers that make headway in understanding and overcoming these methodological concerns, as well as shape strategic research priorities, and suggest guidelines that may serve as best practices for study design, data acquisition, analysis, and dissemination of findings.
## Individual Differences, versus Developmental Processes, versus Phasic Responses
One of the enduring challenges in any attempt to characterize longitudinal change within individuals is to understand exactly what kind of change one is observing. This is especially complex with respect to developmental studies. There are a wide variety of processes that will influence not only the measurements taken of an individual at a specific time, but also the particular processes that might be responsible for an observed pattern of change across longitudinal data collection. For example, there are individual differences in the intercept, slope (and shape), and final outcome of developmental growth processes. Indeed, one of the most difficult problems is that it is often impossible to fully characterize a growth process until it is complete. For example, if one is observing two 16 year-olds who are both six feet tall, it is not clear whether this height represents the final outcome of their adolescent growth spurt, or whether it is an intermediate point on that trajectory. Only further longitudinal observations can resolve the issue. Another complexity is that there may be processes that contribute to change that are not developmental per se, such as when a person experiences an environmental exposure such as trauma, or an episode of mental or neurological illness. These processes may also contribute to change across time and it can be extremely difficult within some designs to disentangle these effects. This issue is especially complex in the study of high risk samples where developmental issues are often confounded with stage of illness issues during the emergence of disorders.
An additional notable complexity here derives from the fact that we often probe phasic response processes, such as those associated with fMRI activation tasks, in our studies. This requires us to understand how the developmental dynamics of baseline brain structure and function might determine differences in these phasic responses across time. Indeed, one manuscript in the special issue directly tackles the relationship between tonic and phasic aspects of brain development by characterizing how the brain departs from its baseline functional architecture during task-induced functional connectivity modulations (Chauvin et al., this issue). The authors propose a novel measure called “task potency,” which allows direct comparison between tasks by operationalizing sensitivity to task manipulations. Chauvin et al. show that their potency measure can demonstrate maturational changes in task-dependent functional co-activation over and above maturation in baseline connectivity.
Ultimately, we will not be able to make significant progress on these issues without strong methodology, and here a number of challenges are notable. For example, Herting and colleagues (this issue) take on the fundamental and yet tricky issue of test-retest reliability of fMRI tasks. This is widely and increasingly recognized as an existential issue for the field, and one that is especially critical for developmental science. The authors review the current state of test-retest reliability for child and adolescent fMRI studies, and provide important guidance on the way forward by highlighting ways to improve fMRI test-retest reliability in developmental cognitive neuroscience research, emphasizing the critical role of open platforms for longitudinal fMRI study designs, analyses, and reporting of results.
Finally, a key question that remains unresolved is whether it is ultimately useful to provide generalized normative growth curves when we are trying to understand individual development. There has recently been significant discussion of whether it is even possible to characterize normative patterns that capture meaningful information about individuals given that for many measures, within-individual variation is often significantly greater than between-individual variation ( ). One methodological advancement that appears to be critical to further understanding this issue is the collection of intensive longitudinal data, where measurements are repeated with high frequency within individuals ( ).
## Ecological and Developmental Validity
A methodological challenge that bedevils all of cognitive neuroscience, indeed all of experimental psychology, is that of ecological validity - do our experimental tasks actually probe the processes associated with the issue of ultimate interest? In a previous publication in the journal ( ) we have pointed out the importance of this issue for the future of the field, and suggested it should be a critical criterion on which studies are evaluated. For too long experimental designs have been justified on face validity criteria in the absence of actual empirical data showing that performance or neural responses associated with the experimental paradigm correlate with the outcome of interest (e.g., everyday decision making, mood, interpersonal functioning), or more broadly the psychological and neurobiological processes that are specifically relevant to these functional activities in daily life. Sherman and colleagues (this issue) address this important issue in the context of a set of questions that have been extensively studied in developmental cognitive neuroscience - functional brain responses that are putatively related to vulnerability to engage in risky decision-making. Their findings suggest that region of interest approaches may be particularly problematic in this regard, possibly because neural factors differentiating riskier teens are not localized in specific regions. They suggest that whole brain approaches may therefore provide more ecologically valid conclusions. The field requires similar systematic analyses of other key ecological outcomes, with associated methodological recommendation in order to address this critical challenge.
Relatedly, van den Bos and colleagues (this issue) argue for the advantages of employing existing computational models of cognition to bridge the gap between neurobiological mechanisms identified via traditional neuroimaging approaches, and the descriptive level of psychological processes. The authors propose that computational models will help us build more specific theories about development as well as identify the processes that produce behavioral change across development. Van den Bos et al. then demonstrate the utility of computational modeling for understanding development in the context of risk-taking, strategy selection, and reinforcement learning. For example, heuristic models of risk-taking tend to focus on reward sensitivity and are loosely defined, whereas computational models attribute differences in risk preferences to more specific mechanisms and allow a more accurate characterization of behavior. In each of these contexts, use of computational models has significant implications for imaging. Continuing with the example of risk-taking, because expected utility and expected value differ, the choice of which to enter as a predictor will affect how the model fits particular voxels; and if these predictors also vary across development, the result may be murky or even faulty inferences.
## Integrating Contemporary Statistical Techniques into Neuroimaging
One major goal of this special issue was to facilitate understanding and application of advanced statistical techniques suitable for longitudinal neuroimaging analysis. Over time, varying procedures for modeling, handling missing data, and power calculations have been constructed across laboratories, sometimes with minimal attention to the ways similar issues have been tackled in non-imaging applications. Developmental cognitive neuroscience is currently in need of informed consensus on many of these issues. Although some of the topics in the papers in this special issue are not necessarily new to statistical longitudinal modelling, they are designed to summarize key concepts, suggest best practice guidelines, and perform a didactic role for neuroimaging researchers specifically, often by demonstrating the use of these statistical methods with real or simulated neuroimaging data. We believe that these advanced concepts in longitudinal modelling are critical for developmental neuroimaging researchers to understand at a deep level, in order for the field to develop robust research designs and analytical practices that result in replicable and interpretable data.
### A Priori Theory and Design
Although historically it has probably been common for statistical analysis to occur largely after data has been acquired, planning ahead for optimally appropriate ways to model data should improve study design and measure selection. For example, one’s theoretical model of change during development will have important implications for the number of time points included in a study and the spacing of the observations. King and colleagues (this issue) provide simulations to show that by changing the follow up time in a longitudinal study by just one year, the estimates for the model can change considerably. The authors strongly urge us to design both the frequency and age boundaries of our assessments to reflect our theory and hypothesis of change ( ). This is salient in developmental research; how many of us often group the period of “adolescence” into a vague age range? The authors also remind us not to mark time points in our models as equal intervals if the time periods are not, in fact, equal, and they also illustrate the importance of choosing where to center the models with regards to age or time (i.e., where to place the intercept), which has implications for results and should, again, align with the a priori theoretical model. Similarly, knowledge from clinical developmental research can be tied to neurobiological theory to include appropriate timing in models, as Haller et al. (this issue) illustrate in the case of psychopathological outcomes such as social anxiety disorder.
Another design consideration for our longitudinal studies is to include a measure asking participants (if possible) the reason they may not have come in for a wave of an ongoing study. Matta and colleagues (this issue) review the differences between types of assumed missing data mechanisms and point out that if we have information that is related to both the missingness (i.e., lack of a scan at a certain wave) and the outcome of interest, we can include that data as a covariate in our analytical model. It may be prudent to attempt to gain this information from our participants in cohort studies.
Finally, regarding the a priori power calculation, which is often a requirement of grant applications (sometimes irrespective of its appropriateness to the aims of the study), Kievit and colleagues (this issue) provide a freely available script to simulate a dataset and compare potential statistical models. This powerful tool will provide developmental neuroimaging researchers with more accurate and robust study designs.
### Analysis
There are several modeling strategies to assess developmental change, which are reviewed in King et al. (this issue). Developmental neuroimaging research, especially fMRI, has long suffered from a dearth of studies with more than two time points, making it difficult if not impossible to estimate complex models. Many studies have attempted to fit non-linear trajectories without having three or more time points. However, all is not lost for those studies with only two time points; Kievit et al. (this issue) focus their didactic paper entirely on latent change score modeling (LCSM) and give examples of this analysis using datasets with only two time points. This modelling strategy is also useful for brain-behavior research questions using cross-domain coupling to ask if, for example, changes in cognition depend on initial neural measures such as ROI volume, or if volume depends on initial cognitive measures, or both. However, it is probably of no surprise that answers to more complex questions about development require more than two time points of data, and Kievit et al. explore more advanced techniques possible with these richer datasets, including dual change score modeling and multigroup comparisons.
Haller and colleagues (this issue) provide an excellent illustration of how advanced statistical techniques for longitudinal modeling can be applied to a specific topic area: the development of social anxiety disorder. After reviewing the literature on SAD and its emergence, and arguing for the necessity of conducting longitudinal studies to understand this disorder, the authors step through various analytical options to consider during the design phase. This includes multilevel modeling, parametric versus nonparametric models, and differential equation models.
Another modelling complexity specific to longitudinal research involves decisions about how to treat missing data. Matta et al. (this issue) provide the first sensitivity analyses comparing available and complete data for longitudinal neuroimaging data to illustrate how parameter estimates can change and bias can be introduced if missing data mechanisms are not modelled correctly depending on the assumed missingness mechanism. For example, when exploring a longitudinal fMRI dataset, the authors showed that when using all available data, two additional clusters were identified in a task that were absent when only complete cases were analyzed (i.e.,including only participants who had all waves of data). This is a powerful illustration of how much choices about how to treat missing data in longitudinal neuroimaging studies matter.
Although many of these advanced statistical analysis, design, and modeling tools have been available and used in developmental science for quite some time, longitudinal fMRI research in particular has not benefited from them. This may be partially because of software issues related to multiple model comparisons in whole-brain voxel-wise analysis that are reviewed in Madhyastha et al. (this issue). They provide a useful table summarizing various types of statistical models in longitudinal research and what they are capable of. They also point out that none of the advanced SEM models can be used in any of the current fMRI software (FSL, SPM, and AFNI are reviewed and current longitudinal capabilities described). This is hardly surprising given that voxel-wise analysis in fMRI involves tens of thousands of separate analyses using the GLM framework. However, they also reveal a new, sophisticated software solution (“Neuropointillist”) that allows researchers to interface with R to conduct the types of complex multivariate analyses for voxel-wise modelling that are described elsewhere in this special issue. Of note, Neuropointillist can accept output (i.e., parameter estimates) from the first-level analysis of any other fMRI software. It is also currently the only software that can handle missing data when correlating neural and behavioral data at the whole-brain level without listwise deletion. While Matta et al. (this issue) describe strategies for dealing with subject drop out (i.e., missing a scan/wave of a study completely), until now there was no good solution for missing data at the voxel level (e.g., with movement artifact). Neuropointillist is a promising direction for more flexible processing and analysis in this regard. Although this only begins to address how best to conduct model selection and comparison with voxelwise modelling, it is a flexible and powerful tool for neuroimaging researchers wishing to assess more complex models of developmental change. However, as they point out, we need much more progress in programming technology due to the time and computing power necessary for testing each model for such a large number of voxels.
Furthermore, for missing data mechanisms, the next challenge is data that we consider or assume to be missing not at random (MNAR). Do we, as developmental cognitive neuroscientists, think that there could be a good reason that the probability of someone missing a scan is dependent on the data that went uncollected? In many cases, the answer may be yes, and if so, using all available data could result in biased estimates. But, as discussed above, we could consider asking participants why they cannot or chose not to come in for scans, and include this covariate that is presumably related to both the missingness of the data and the dependent variable in our analytical models. Furthermore, as was demonstrated in Matta et al. (this issue), more research could provide sensitivity estimates comparing complete case and available data analyses.
## Guidelines and Diversity of Methods
Another goal of the special issue was to outline best practice guidelines for the processing and analysis of developmental neuroimaging data, especially for longitudinal study designs. There was significant diversity in the methods covered - sMRI, fMRI, dMRI, fNIRS, and even magnetic resonance elastography (MRE) - although some modalities were noticeably absent from the special issue (such as EEG and MEG). Perhaps experts in those fields will be motivated by this special issue to produce manuscripts to serve similar guiding functions. In developmental cognitive neuroscience, best practices may also vary to some degree by age group, and there was a significant effort in this special issue to address imaging of infants and very young children. It is notable that several of the papers included in the special issue provided tables, checklists, and tools to facilitate decision making at every stage of the research process. For example, King et al. (this issue) distilled their essential messages into a box for easy printing, framing, and hanging next to one’s workstation for quick reference. With so many statistical models and considerations to track, an easy reference guide provides increased utility for developmental cognitive neuroscientists just beginning their journey of longitudinal analysis of developmental change.
### Reviews by Modality
Vijayakumar et al. (this issue) provide a checklist for researchers to use when reporting methodological detail in longitudinal structural brain imaging studies. This checklist was developed after the authors systematically reviewed the existing longitudinal studies of brain structure in developmental samples, and realized that many studies left out essential details needed for comparing results across studies, such as quality control or model selection procedures. It will serve as a valuable resource for developmental cognitive neuroscientists to use when writing and reviewing papers. Standardizing how we report our study design, methods and results will benefit future meta-analyses, systematic reviews, and help us understand how methodological differences could be related to discrepant findings. Vijayakumar et al. also focus extensively on statistical analysis of sMRI data, including analytic methods such as multilevel or spline modeling, considering trajectories and peaks, as well as model selection, among several other issues. This provides excellent concrete translation of many of the concepts in the more theoretical statistical manuscripts from the special issue, such as King et al. (this issue).
Telzer and colleagues (this issue) turn a critical eye to longitudinal fMRI. They provide a complementary overview to Madhyastha et al. (this issue), with respect to commonly used software for longitudinal fMRI, and many insights are shared between the two manuscripts - particularly the limitations in many of the widely available packages. However, one unique aspect of this paper is its focus on the particulars of tasks frequently employed in developmental fMRI studies, and key issues to consider if utilizing these tasks in a longitudinal design. It also provides an overview of recent longitudinal developmental fMRI studies, emphasizing accelerated longitudinal designs using a region-of-interest (ROI) approach, and exploring using neural trajectories to predict outcomes. Readers are then treated to detailed “behind the scenes” coverage of three longitudinal fMRI papers, one from each of the three laboratories represented amongst the authors of the collaborative manuscript. This illustrates the host of decisions made from task design to analysis.
Tamnes et al. (this issue) summarize dMRI approaches, including diffusion tensor imaging (DTI) and other less common but more advanced techniques and metrics. Examples of the latter include high angular resolution diffusion imaging (HARDI) and whole-brain probabilistic tractography of ‘fixels’ (populations of fibers within voxels), as well as diffusion kurtosis imaging (DKI) and neurite orientation dispersion and density imaging (NODDI). These more advanced approaches are promising, but currently difficult to apply with developing populations because of their increased acquisition times relative to DTI. The authors discuss various approaches to maximize image quality with the shortest durations possible, and various trade-offs that researchers must consider. They also compare ROI, voxel-based methods, tractography, and graph theory analysis. A running theme throughout the special issue mirrored in this manuscript is the absolutely essential nature of quality control in dMRI, and yet the failures of the field to converge on consensus or automated approaches to carry out QC. After providing readers with this detailed treatise on the method, Tamnes et al. (this issue) conclude with a systematic overview of longitudinal dMRI studies, as well as developmental effects seen through the lens of sex and individual differences, puberty, and atypical development. However, their review highlights the difficulties associated with ascribing changes in dMRI to specific underlying cellular and molecular events, which is a key challenge for the future.
Perhaps the least familiar method in the special issue received introductory coverage from Johnson & Telzer (this issue). They provide a primer on the technique of MRE, which assesses brain tissue stiffness and viscoelasticity by imaging the shear deformations resulting from light vibration of the head, and then reconstructing the underlying mechanical properties. Although MRE has been around for about a decade, it has almost exclusively been applied to imaging the adult brain, including effects of aging and neurodegenerative diseases. MRE is particularly sensitive to motion as it has an acquisition time ranging from 5-10 minutes, which may pose difficulties in more widespread application in younger children and those with psychological or developmental disorders. Nevertheless, we are interested to see the application of this technique grow.
### Imaging the Early Years
Issard & Gervain (this issue) provide an overview of functional near infrared spectroscopy (fNIRS), and outline important technical and physiological considerations for its use, particularly with infants. They focus extensively on constraints of this modality presented by the considerable variability of the hemodynamic response within the literature. In particular, there are reports of inverted hemodynamic responses, as well as extended durations to peak hemodynamic responses. This obviously creates interpretive challenges. The authors review these effects separately by sensory or cognitive function, providing a kind of roadmap for infant researchers looking to use fNIRS. Canonical response functions are typically seen earlier in temporal than occipital or frontal cortices and follow a more linear developmental trajectory. The authors also identify ways in which the paradigms and stimuli themselves may influence the hemodynamic response, especially in frontal cortex. Stimulus complexity and familiarity, biological development with age, and experimental design may all influence neurovascular responses.
Another article tackling methodological issues in infant neuroimaging turns the spotlight on fMRI (both activation- and connectivity-based approaches). Cusack, McCuaig, and Linke (this issue) focus on challenges specific to comparing activation and connectivity across age groups. Given the dramatic growth in head size, shape, and gyrification, the authors note that inter-subject registration in studies comparing across age groups including infants is best when first registering to an age-specific template. Different hemodynamic responses again asserts itself as a potential concern, as well as physiological noise, brain chemistry, infant behaviors such as motion or sleep, and peripheral sensory changes that affect the way in which the same stimulus is perceived by infants versus other age groups. Cusack et al. (this issue) provide recommendations specific to each of these challenges.
Meanwhile, yet another article in this special issue addresses some of the concerns inherent to the historically sparse literature in young children under the age of six years (Van Phan et al., this issue). fMRI studies in children out of infancy but under age six have only recently begun to increase in number. This is in part due to development of improved methods to acquire high-quality imaging data ( , ). Now that high-quality data are being collected, Van Phan et al. (this issue) argue it is critical to turn our attention to ensuring the data processing techniques applied to these data adjust for the unique population. The authors examine by turn each processing step and outline child-specific considerations and recommendations. Like others in the special issue, they make an essential point that quantifying data quality should be validated and explored in particular across other MRI modalities. They also argue for using age-specific atlases, to improve segmentation and registration. 4D spatio-temporal atlases, which consist of a series of age-dependent averaged 3D atlases that summarize details of brain structures by age, may be an important new frontier in this regard, but are not easily implemented in current neuroimaging software packages. Additionally, multi-atlas based methods may also be a promising new avenue, using learning algorithms to select the best atlas for each participant.
## Open Methods
Several of the papers presented within this special issue have made methodological tools available to facilitate the adoption of advanced data analytic techniques. For example, in addition to describing the theory behind and rationale for using latent change score modeling (LCSM), Kievit et al. (this issue) presented a practical tutorial for applying LCSM to longitudinally acquired MRI data with the open-source software R and Ωnyx. They also made sample data and scripts available for researchers to adapt to their own projects. In a display of high conscientiousness, the authors even created a graphical interactive web application to help researchers understand the impact of changing various model parameters ( ).
Many contributors to the special issue attended the Modeling Developmental Change workshop held in September 2017 in Portland, Oregon, USA, immediately prior to the annual meeting of Flux: The Society for Developmental Cognitive Neuroscience. The purpose of this workshop was to teach best practices for processing, analyzing, modeling, and interpreting longitudinal neuroimaging data in developing populations. In effect, this workshop presented the opportunity for researchers to apply many of the methods discussed in the special issue. All of the resources used in the workshop, including presentations, tutorials, tools, scripts, and example data, were made freely available to researchers through open science repositories (available here: ). Published guidelines and methods papers represent one approach to increase the output of robust and reproducible research in developmental cognitive neuroscience, and hands-on workshops and online tutorials represent a complementary approach to allow researchers to return to their projects with tools in-hand. The two approaches are necessary to give researchers the theoretical knowledge of when and why to apply certain analytic techniques, and practical knowledge of how to do so.
## Conclusion
To close this introduction, we want to strongly encourage readers to keep in mind that the best practices outlined in the special issue are guides rather than constraints, and that our field benefits from the diversity of methods employed for understanding the developing brain just as we do from standardizing our practices. The breadth of manuscripts in the special issue illustrates how guidelines are not meant to constrain the field, but rather draw attention to statistical and methodological considerations when conducting longitudinal and other forms of developmental neuroimaging research. We hope this special issue can play a role in potentiating high-quality research in the field for years to come.
|
According to Dual Systems models (Casey et al., 2008; Luna and Wright, 2016; Steinberg, 2008), a rapidly-developing socioemotional system and gradually-developing cognitive control system characterize adolescent brain development. The imbalance hypothesis forwarded by Dual Systems models posits that the magnitude of the imbalance between these two developing systems should predict the propensity for engaging in a variety of risk behaviors. The current integrative review argues that the excitement generated by the imbalance hypothesis and its implications for explaining adolescent risk behaviors has not been meet with equal efforts to rigorously test this hypothesis. The goal of the current review is to help guide the field to consider appropriate and rigorous methods of testing the imbalance hypothesis. First, we review the analytic approaches that have been used to test the imbalance hypothesis and outline statistical and conceptual limitations of these approaches. Next, we discuss the utility of two longitudinal analytic approaches (Latent Difference Scores and Growth Mixture Modeling) for testing the imbalance hypothesis. We utilize data from a large community adolescent sample to illustrate each approach and argue that Latent Difference Scores and Growth Mixture Modeling approaches enhance the specificity and precision with which the imbalance hypothesis is evaluated.
## Introduction
Adolescence is a developmental period marked by the emergence of a range of risk behaviors and mental health concerns that can persist into adulthood for some youth ( ; ). These behaviors include the initiation and escalation of substance use ( ; ), risky sexual activity and high rates of sexually transmitted diseases ( ), and fatal car crashes ( ). Moreover, rates of externalizing disorders, including oppositional defiant disorder, conduct disorder, and substance use disorders increase from early to late adolescence ( ).
The Dual Systems Model ( ), Maturational Imbalance Model ( ), and Driven Dual Systems Model ( ), provide similar theoretical accounts for the increase in risk behaviors seen during adolescence. A common tenet of these leading developmental neuroscience models is that adolescent risk behaviors results from an imbalance between the development of a cognitive control and a socioemotional neural system (see ). The prefrontal cortex primarily encompasses the cognitive control system, which includes executive functions and other higher-order self-regulatory processes necessary for top-down control of behavior ( ). Behaviorally, cognitive control involves the ability to adjust behavior in response to changing task demands and inhibit behavior that is no longer adaptive ( ). Cognitive control requires integrations of inhibitory control, conflict monitoring, working memory, and attentional control, though most work in this area focuses on tasks that assess inhibition (see ). Common paradigms include the Stop Signal Task (SST), Go/No-Go, and antisaccade tasks. Briefly, all three tasks require participants to withhold a prepotent response and instead execute a subdominant response.
Adapted with permission from . The figure depicts three Dual Systems models and the development of sensation seeking and self-regulation from late childhood to young adulthood according to each of these models. The blue portion in each model represents the imbalance between sensation seeking and self-regulation. The challenge when assessing the imbalance hypothesis is to use a data analytic technique that captures the difference between sensation seeking and self-regulation. Further, each of these Dual Systems model posit systematic changes in sensation seeking and self-regulation across time, therefore, data analytic techniques used to assess the imbalance hypothesis must also be able to capture the proposed developmental differences in sensation seeking and self-regulation from late childhood to young adulthood. A model that captures the dashed line (sensation seeking), either at a single time point or across time, is not assessing the imbalance. Similarly, a model that captures the solid line (self-regulation), either at a single time point or across time, is also not assessing the imbalance. Further, a model that simultaneously models the dashed line (sensation seeking) and solid line (self-regulation), at a singly time point, is not modeling the imbalance. We argue that only data analytic approaches that quantify the imbalance between the dashed and solid lines (the blue portion of each Dual Systems model) and account for developmental changes in imbalance can be rigorous tests of the imbalance hypothesis. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).
Fig. 1
The socioemotional system is mediated by subcortical dopaminergic regions, particularly the striatum, and is responsible for enhancing the motivational salience of rewarding stimuli ( ). Indicators of the socioemotional system at a behavioral level include self-report measures such as the Sensation Seeking Scale ( ) and the BIS/BAS Questionnaire ( ), as well as paradigms such as the Iowa Gambling Task ( ; ) and the Point Score Reaction Time Task for Children-Revised ( ). Throughout the review, we refer to these systems at the level of the psychological, rather than the biological, construct. Accordingly, “sensation seeking” will refer to the socioemotional system, and “self-regulation” will refer to the cognitive control system ( ).
We recognize that Dual Systems models diverge in a number of important ways, including the extent to which the development of sensation seeking and self-regulation should be orthogonal (Dual Systems Model) or interdependent (Maturation Imbalance) and whether the hypothesized trajectories of growth of each system are linear versus curvilinear ( , ). However, despite these differences, a cornerstone of all three models is the imbalance hypothesis – that gradual maturation of self-regulation is outpaced by rapid developments in sensation seeking, particularly during the first half of adolescence ( ; ; ). This imbalance between sensation seeking and self-regulation is hypothesized to result in myriad adolescent risky behaviors (see for a depiction of the imbalance forwarded by Dual Systems models). That is, motivation for risky behavior increases in adolescence, and the systems responsible for inhibiting those behaviors have not yet fully developed.
The Lifespan Wisdom Model ( ) is related to dual systems models and the imbalance hypothesis, yet there are notable differences. For example, the Lifespan Wisdom Model asserts that the imbalance between sensation seeking and cognitive control is best represented by high levels of acting without thinking (poor impulse control), rather than separate measures of each system. Given this feature, it is unclear if the model falls under the rubric of a dual systems model. Furthermore, the model posits that imbalance (poor impulse control) characterizes only a subset of youth rather than being part of a normative developmental trajectory.
The centrality of the imbalance hypothesis to different theoretical approaches and models suggests the importance of using appropriate methods to test this hypothesis. As we argue below, we believe that data analytic methods used to date fail to appropriately capture the difference between sensation seeking and self-regulation and how that difference predicts risk behavior. The primary purpose of the current review is to discuss data analytic methods that provide a rigorous test of the imbalance hypothesis with longitudinal data. For the sake of simplicity, we refer to the Dual Systems Model ( ), Maturational Imbalance Model ( ), and Driven Dual Systems Model ( ) collectively as Dual Systems models.
### Critiques of dual systems models
Dual Systems models have spurred a large volume of research on adolescent brain development that has expanded our understanding of the neural underpinnings of adolescent risk behaviors ( ). These models have become increasingly influential over the last decade, providing a theoretical framework for empirical investigations of substance use, delinquency, and myriad other risk behaviors, and also shaping public policy and juvenile justice decisions regarding adolescent delinquent behaviors ( ; ; ; ).
Although these models have stimulated a productive and informative line of inquiry, researchers have identified several limitations. First, recent papers have noted a lack of specificity (e.g., according to Dual Systems models, how would one go about testing the imbalance and its proposed relationship to risk behavior?) and questionable falsifiability (e.g., how could one provide strong evidence that the imbalance is not implicated in risk behaviors?), suggesting that Dual Systems models of adolescent risk behavior are informative heuristics, rather than testable theories ( ; ; ). In line with this critique, several recent papers have called for the need for greater specificity of Dual Systems models ( ; ).
A second critique of Dual Systems models is that there is limited evidence demonstrating that an imbalance between sensation seeking and self-regulation in adolescence is associated with real-world risk behavior ( ; ; ; ). As the field continues to accumulate neuroimaging data on the development of the socioemotional and cognitive control systems in adolescence, it is critical that the neuroimaging findings are complemented by rigorous behavioral evidence to inform the extent to which patterns of brain activation predict engagement in different types of risk behavior in adolescence ( ).
Although Dual Systems models are developmental in nature, most of the evidence in support of them is cross-sectional (Crone, van Duijenvoorde, & Peper, 2016; ). As shown in , the imbalance between the systems should grow from early to middle adolescence, peak in middle adolescence, and then diminish ( ). Most previous work in this area has quantified the imbalance without regard to its developmental trajectory (e.g., ; ; ). As more longitudinal studies are being undertaken to address this concern, it is important to consider analytic methods appropriate to describing and testing the complex developmental patterns hypothesized in these models ( ).
### Improving the specificity of the imbalance hypothesis
The current integrative review seeks to address these limitations. Our goals are to delineate the specificity issues and the lack of behavioral and longitudinal data in support of this hypothesis ( ) and address key conceptual and statistical considerations involved in operationalizing and testing the imbalance hypothesis with behavioral data in order to facilitate the generation of more precise predictions. We focus on behavioral data because of recent calls for empirical papers that assess Dual Systems models with behavioral data and the importance of demonstrating that the imbalance does in fact impact real-world risk behavior ( ). However, the data analytic methods discussed below could readily be applied to neuroimaging data.
We first review analytic approaches that have been used to date to quantify the imbalance between sensation seeking and self-regulation and describe both conceptual and statistical limitations of these approaches. To begin to address the aforementioned limitations on specificity, we also clearly delineate what information these analytic approaches provide and what questions each approach is able to answer. Next, we present several longitudinal analytic approaches that more appropriately test the imbalance hypothesis than previously-used methods in this area.
We hope the present paper leads the field to further consider the following questions: How do we test the imbalance with longitudinal data? What can the statistical analyses we use tell us about the imbalance and its development over time, as well as the relation of the imbalance to risk behaviors? Do current analytic approaches advance theory and move it in a direction towards falsifiability? Before delving into these issues, we take a short detour to discuss issues of measurement.
### A brief word on measurement of the imbalance
Dual Systems models have repeatedly been criticized for their lack of specificity regarding how to measure the socioemotional and cognitive control systems ( ; ; ). This lack of precision has resulted in the same measures (e.g., delay discounting, acting without thinking) being used as indicators of both cognitive control ( ; ; van den Bos et al., 2015) and socioemotional systems ( ; ). Considering prior reviews and commentaries have discussed these issues at length ( ; ; ), we briefly note the implications of imprecise measurement of the socioemotional and cognitive control systems on testing the imbalance hypothesis.
The Dual Systems Model, Maturational Imbalance Model, and Driven Dual Systems Model all necessitate separate measures of the socioemotional system and cognitive control system to assess the imbalance (e.g., ). In contrast, the Lifespan Wisdom Model asserts that measuring poor impulse control (acting without thinking) captures the imbalance because of its positive association with reward sensitivity and negative association with cognitive control ( ). These different conceptualizations raise the question of how best to distinguish measures of the imbalance from measures/constructs that reflect correlates of the imbalance, an issue that has garnered little attention in the literature. Moreover, these varying conceptualizations of the imbalance have significant implications for the statistical methods needed to assess the imbalance hypothesis. Because the Dual Systems Model, Maturational Imbalance Model, and Driven Dual Systems Model all call for separate measures of the socioemotional and cognitive control systems, statistical techniques must capture the difference between two variables across time (see ). In contrast, ascribing to the Lifespan Wisdom Model’s operationalization of measuring the imbalance only requires the modeling of a single variable across time.
The lack of a clear operationalization of these systems is of concern because it limits research progress for Dual Systems models and undermines efforts towards increased precision and risky tests of the imbalance hypothesis ( ; ). Our review of past analytic methods and presentation of alternative longitudinal data analytic approaches focuses on studies that conceptualize the imbalance as the difference between separate measures of sensation seeking and self-regulation as this seems to characterize most dual systems perspectives. However, where appropriate, we also highlight how our longitudinal methods can be applied to conceptualizations of a single indicator, poor impulse control, as reflecting the imbalance.
### Review of methods used to assess the imbalance
Literature Search and Inclusion Criteria
We do not to provide an exhaustive review of all studies that have attempted to test the imbalance hypothesis (for a comprehensive review see ). Rather, our aim is to bring attention to the conceptual and statistical limitations of methods commonly used to test the imbalance hypothesis and to recommend alternative statistical methods that are better suited to test this tenet of Dual Systems models. Articles were identified through PubMed, Psych Info, and Google Scholar. Searches included keywords such as Dual Systems, Maturational Imbalance, socioemotional system, sensation seeking, reward sensitivity, cognitive control system, self-regulation, inhibitory control and risk behaviors (e.g., substance use, alcohol use, marijuana use, cigarette use, risky driving). Further articles were identified through searching reference sections of relevant articles. For inclusion, studies were required to have adolescent samples and behavioral measures assessing the socioemotional and self-regulation systems, as well as behavioral indicators of risk behavior. We focused primarily on studies that purported to directly test Dual Systems models or were cited as evidence of Dual Systems models. provides a summary of the methods reviewed to test the imbalance hypothesis.
Summary of Current Methods as well as Recommended Alternative Models to Testing the Imbalance Hypothesis.
Table 1
#### Method 1: regression analyses demonstrating unique effects of sensation seeking and self-regulation
Across all methods discussed below, a key limitation is that they are not well suited to model the imbalance as a function of age. The cross-sectional designs used to date to test the imbalance hypothesis preclude examination of developmental changes in the imbalance, which is essential for testing Dual Systems models.
A simple method researchers have used to assess the imbalance hypothesis involves regression analyses, in which indicators of sensation seeking and self-regulation are used to predict a risk outcome, while controlling for the effects of the other process (e.g., ; ). Although finding unique effects of sensation seeking and self-regulation supports the tenet of Dual Systems models that sensation seeking and self-regulation are implicated in risk behavior in adolescence ( ), studies demonstrating unique effects of sensation seeking and self-regulation have been cited as evidence that heightened sensation seeking in the context of low self-regulation is implicated in adolescent risk taking ( ). Yet, a regression approach testing unique effects does not, in fact, take into account one’s standing on one variable relative to the other variable, and therefore provides no information about an imbalance between the two systems. Thus, because the regression approach does not characterize the imbalance between sensation seeking and self-regulation and it does not account for developmental changes, we argue that it is not an informative statistical test of the imbalance hypothesis.
#### Method 2: moderation of sensation seeking x self-regulation
An extension of the regression approach is to include cross-product interaction terms (sensation seeking x self-regulation interaction terms) to predict risk behavior. To date, a number of studies have argued that assessing interactions of sensation seeking x self-regulation is a test of the imbalance hypothesis ( ; , ; ; ; ; ), and we agree that the interaction can provide information regarding the imbalance. For example, examined the interaction between sensation seeking assessed at age 16 and effortful control (a behavioral indicator of the self-regulation system) assessed at age 11 on adolescent drug use at age 16. The authors found support for an interaction, such that sensation seeking was related to alcohol and cannabis use for adolescents characterized by low effortful control. Conversely, there was no relationship between sensation seeking and alcohol or cannabis use for adolescents characterized by high effortful control. A similar analytic approach was used by . These authors used behavioral tasks to assess sensation seeking and inhibitory control and found that high levels of sensation seeking prospectively predicted increases in delinquency, but only at low levels of inhibitory control.
There are several notable limitations of the moderation approach, although it does represent an improvement over the regression approaches that only examine unique effects of sensation seeking and self-regulation. First, as discussed above, assessing moderation using a standard regression approach does not provide any information about development of sensation seeking and self-regulation. Theory argues that the imbalance should be the largest in middle adolescence and decline in late adolescence and young adulthood (see ). Testing this assertion, even with cross-sectional data, would require testing a three-way interaction term between sensation seeking, self-regulation, and age (see ).
A more serious concern regarding the moderation approach is that interaction terms do not uniquely and solely capture the imbalance between sensation seeking and self-regulation. Although moderation can demonstrate that self-regulation may modulate sensation seeking, arguing that moderation is a test of the imbalance hypothesis is statistically inaccurate ( ). We use hypothetical data from three adolescents to illustrate this point. Imagine three adolescents, Bill, Bob, and Barbara, who each completed measures of sensation seeking and self-regulation (higher scores on these hypothetical measures indicate stronger sensation seeking and self-regulation). Bill’s standardized score on the measure of sensation seeking was 1.50, and his standardized score on the measure of self-regulation was 1.20. Bob’s scores on sensation seeking and self-regulation were 0.20 and -0.10, respectively, and Barbara’s scores on sensation seeking and self-regulation were -0.60 and -0.90, respectively. The imbalance, or difference between the strength of sensation seeking and self-regulation, is the same for all three teens (0.30), and suggests that they should have the same risk level for engaging in risk behavior. Despite the imbalance being the same for all three adolescents, their estimated propensity for risk behavior varies considerably in a moderation approach. To illustrate this point, below is the overall regression equation for examining moderation of sensation seeking (SS) x self-regulation (SR) on a risk outcome, as well as the regression equations for Billy, Bobby, and Barbara predicting a risk behavior (Y).
If we plug in plausible regression coefficient values for β (0.50), β (0.30), β (-0.15), β (-0.05), the Y values for Bill (0.68), Bob (0.58), and Barbara (0.43) all differ, despite their equal imbalance between sensation seeking and self-regulation (0.30). This demonstrates that moderation does not solely capture the difference between sensation seeking and self-regulation because individuals with the same imbalance will have different predicted Y values on an outcome with different combinations of X values. For a more thorough explanation of how moderation does not accurately capture an imbalance, please see . Thus, we urge researchers to be mindful in their interpretations of interactions between sensation seeking and self-regulation because these interaction terms do not solely reflect the imbalance. Considering this point, we do not recommend using the moderation approach when testing the imbalance hypothesis.
#### Method 3: observed difference scores between sensation seeking and self-regulation
A useful alternative is to consider a difference score in a regression model. The difference between sensation seeking and self-regulation has the advantage of clear conceptual linkage to the imbalance hypothesis. A greater difference can be conceptualized as a larger imbalance. Below is the regression equation for the difference between sensation seeking and self-regulation as a predictor of risk behavior. When using the same values for β (0.50) and β (0.30) as the previous example and plugging in the values on sensation seeking and self-regulation for Bill, Bob, and Barbara, the risk values are the same across all three adolescents (Y = 0.59), as they should be, given that their imbalance is equal.
The observed difference score approach may be advantageous over moderation because it uniquely informs whether the difference in the two systems accounts for variance in a risk outcome. In their cross-sectional study of individuals ages 12–28, found that a difference score between their measures of sensation seeking and impulsivity predicted deviancy, such that individuals characterized by high levels of sensation seeking relative to levels of impulse control were more likely to engage in deviant behaviors.
However, there are conceptual and statistical concerns under certain circumstances that researchers should be mindful of when using observed difference scores. Several papers have outlined the primary issues with observed difference scores, including ambiguity, confounded effects, untested constraints, dimensional reduction, and, in particular, concerns about reliability ( ; ). We briefly touch on a few of these concerns most central to Dual Systems models but recommend Edwards (2001, 1994) and for in-depth discussions on these issues. Concerns regarding ambiguity stem from collapsing two measures into a single variable, the variance of which is a function of the variances of the two components from whence it came ( ). The first issue related to ambiguity is determining how to create a meaningful difference score when the indicators of sensation seeking and self-regulation are on different metrics ( ). In testing a difference score, standardized their indices of sensations seeking and impulse control. Although they do not provide a rationale for why they standardized, the likely motivation was to create a meaningful difference score by placing the two components on the same standardized metric. In the subsequent section on modeling the imbalance with longitudinal data, we present a method of placing sensation seeking and self-regulation measures on the same metric using longitudinal data.
The second issue concerning ambiguity is whether it is accurate to state that a difference score between sensation seeking and self-regulation accurately reflects the difference (imbalance) between these two variables, as opposed to either reflecting sensation seeking or self-regulation ( ). The relative contribution of each component to a difference score depends heavily on the variance of each component ( ). For example, if a measure of sensation seeking is found to have much greater variance than a measure of self-regulation, then the difference score created by subtracting sensation seeking and self-regulation would primarily reflect sensation seeking and not self-regulation. The variance of measures of sensation seeking and self-regulation would have to be comparable in magnitude to attenuate this concern. Further, establishing a nomological network ( ) where the observed difference score is correlated with other constructs in a theoretically-consistent manner has been suggested as a method of attenuating these ambiguity concerns ( ; ). Specifically, a researcher could correlate the indicator of sensation seeking and self-regulation, as well as the sensation seeking – self-regulation difference score, with a host of other variables (e.g., puberty, risk-taking, peer delinquency, temperament), and the difference score should demonstrate a pattern of correlations that is distinct from the correlation patterns of either sensation seeking or self-regulation alone.
An important assumption of using a difference score is that the components have equal regression weights of opposite signs when predicting the outcome of interest and that they account for roughly equivalent amounts of variance in an outcome. A strength of is that they examined these constraints to determine if their data met this assumption. For example, the authors demonstrated that the effects of impulse control ( β =-0.25) and sensation seeking ( β = .23) were both significantly associated with deviance (their outcome of interest) and were opposite in sign and nearly equivalent in magnitude.
Lastly, reliability has been repeatedly cited as a concern with observed difference scores ( ; Edwards, 1991). As the covariance between components increases, the reliability of observed difference scores decreases ( ). This issue may be particularly problematic for tests of Dual Systems models because most studies find moderate correlations between their measures of sensation seeking and self-regulation ( ; ).
Despite these concerns, we view the observed difference score method as the strongest method used to date to test the imbalance hypothesis when using separate measures of sensation seeking and self-regulation, as long as the aforementioned assumptions are met. In line with , we urge researchers to be mindful of the potential issues of using observed difference scores and test the appropriateness of using observed difference scores when feasible.
### Summary of methods to date
To summarize, the primary methods used to date to assess the imbalance between sensation seeking and self-regulation with behavioral data include 1) regressions, in which a measure representing one system predicts the outcome, while controlling for levels of a measure representing the other system, 2) interactions of indicators of the two systems predicting risk outcomes, and 3) difference scores. Although the regression and moderation approaches are the most common, these statistical methods provide inadequate tests of the imbalance hypothesis. Observed difference scores provide the strongest test to date of the imbalance hypothesis, although researchers should be mindful of the assumptions and potential drawbacks with this approach.
Critically, all three analytic approaches have almost exclusively utilized cross-sectional data, which provides no information about developmental change. Although and had longitudinal samples, sensation seeking and self-regulation were assessed once, precluding analysis of maturation of these processes. As seen in and articulated in theoretical papers on the imbalance hypothesis ( ; ), a central tenet of this hypothesis is that developmental changes in the imbalance between sensation seeking and self-regulation leads to changes in the probability of risk behaviors. Considering this argument in Dual Systems models, we view modeling the development of the imbalance as a critical component of conducting more precise tests of the imbalance hypothesis.
### Longitudinal approaches to modeling the imbalance
Next, we discuss and illustrate two flexible data analytic strategies that provide more precise tests of the tenets of Dual Systems models that the imbalance between sensation seeking and self-regulation, and the development of the imbalance over the course of adolescence, should be associated with risk behaviors. After discussing the conceptual and statistical benefits of Latent Difference Score and Growth Mixture Modeling, we provide examples in which we use these methods to assess the relationship of the sensation seeking – self-regulation imbalance with risk behavior using a large longitudinal community sample. The worked examples focus on modeling the imbalance with separate measures of sensation seeking and self-regulation. For researchers taking the perspective that acting without thinking (impulse control) reflects a direct measure of the imbalance ( ), we note how Latent Difference Score and Growth Mixture Modeling approaches can be used with a single indicator of the imbalance. We hope to stimulate the field to reflect on the statistical complexity of modeling the imbalance and to explicitly discuss which data analytic strategies provide the best approach to modeling the imbalance and its relationship with risk behavior.
#### A latent difference score approach
As noted above, difference scores provide a straightforward operationalization of the imbalance that can serve as the predictor of risk behavior ( ), but have several statistical limitations. Latent Difference Score (LDS) models are an alternative approach to observed difference scores that maintain the theoretical appeal of assessing the imbalance with a difference score, while mitigating concerns about unreliability ( ; ; , ). LDS models impose strict parameter constraints that permit modeling of latent changes that represent the difference of two latent variables (see ). A significant benefit of LDS modeling is that a growth curve can be fit to the difference scores ( ), allowing researchers to examine the developmental changes in the imbalance ( ), as well as the association between growth in imbalance and risk behaviors. Thus, LDS models are particularly well-suited to test the imbalance hypothesis because they (1) explicitly model the imbalance using latent variables , (2) allow for the examination of developmental changes in the imbalance, and (3) allow researchers to assess how growth in the imbalance is related to risk behaviors.
Despite these benefits, there are also several limitations of LDS models. First, it requires researchers to have three waves of data to estimate the growth portion of the model. Because the growth portion of the model is specified as a latent growth curve (see LDS example for greater detail), factors that typically influence performance of latent growth curve models (e.g., misspecification of the growth model, missing data patterns, and individual differences in measurement occasion time points) also impact performance of the LDS growth model ( ; ; ). The LDS approach also requires the use of large sample sizes ( ), which is not always feasible, particularly if constructs are being measured at the neurobiological level. Third, due to the specification of the latent difference scores, it is not possible to examine whether the imbalance accounts for unique variance in risk behaviors above and beyond either sensation seeking or self-regulation.
Finally, this modeling approach does not address how to create a meaningful difference across two variables on separate metrics ( ). As noted earlier, this commonly occurs with behavioral indicators of self-regulation and sensation seeking. One solution is to standardize the indicators, as per ; however, simple standardization will not work with longitudinal data because standardizing indicators at each assessment removes information about the means (means are 0 for standardized variables), which is crucial for understanding growth. An alternative is to collapse data across repeated measures and standardize (i.e., include all participants’ data from all waves of data collection onto the same distribution before standardizing; see for an example of this standardization approach). This method puts each indicator on the same metric and retains information about growth. This method also provides a meaningful value for the intercept in a latent growth curve model when examining growth in the imbalance across time. That is, with standardized difference scores, the latent intercept represents agreement between sensation seeking and self-regulation (e.g., score below zero indicate an imbalance where inhibitory control is greater than sensitivity to reward, scores of zero indicate no imbalance between sensitivity to reward and inhibitory control, and scores greater than zero indicate an imbalance where sensitivity to reward is higher than inhibitory control) at the age where the intercept is specified. However, this approach does have limitations because the value of the difference score is relative and sample-dependent, making comparisons across samples difficult.
In our empirical example we use LDS modeling to demonstrate how this standardization approach can be used to model the relationship of the imbalance with risk behaviors, specifically the probability of alcohol and marijuana use, as well as levels or intensity of alcohol use and the frequency of marijuana use from early through late adolescence (ages 13–20). Our LDS model allowed us to assess the following questions:
Is there significant growth and individual variability in growth of the imbalance across ages 12–14?
Are initial levels of the imbalance or growth in the imbalance associated with alcohol and marijuana use?
#### A growth mixture modeling approach
Growth Mixture Modeling (GMM) is a statistical technique that can be used to identify classes or groups of individuals that share similar patterns of growth in sensation seeking and self-regulation over time (for detailed descriptions of GMM as well as applied papers using GMM see ; ; ). GMM is related to latent growth curve modeling, although it relaxes the assumption of a latent growth curve model that all individuals are drawn from a single population ( ). Applied to Dual Systems models, one might expect a majority of adolescents to be characterized by rapidly-developing sensation seeking and slower-developing self-regulation ( Panel A), and this pattern would be expected to be associated with frequent engagement in risk behaviors ( ). However, other patterns of growth in both sensation seeking and self-regulation for adolescents likely exist, and of interest is how different patterns of growth in sensation seeking and self-regulation and the imbalance they imply are related to risk behaviors ( ; ). For example, one might also expect a class that shows gradual increases in both sensation seeking and self-regulation, and hence a relatively modest imbalance and low levels of risk behavior.
Although no studies, to our knowledge, have used GMM to assess the imbalance hypothesis using separate measures of sensation seeking and self-regulation, applied GMM to a measure of impulsivity, which they posit reflects the imbalance (acting without thinking). Invoking the Lifespan Wisdom Model ( ), hypothesized that only a subset of adolescents would exhibit an imbalance across adolescence and found evidence for two distinct subgroups of trajectories of acting without thinking across adolescence, a low stable trajectory and a high increasing trajectory. Adolescents in the high increasing trajectory had higher levels of Substance Use Disorder severity relative to adolescents in the low stable trajectory. This study highlights the appeal of GMM by demonstrating the ability of this modeling approach to identify distinct developmental patterns of the imbalance, which has been increasingly advocated for in studies of the imbalance ( ; ).
It is important to differentiate GMM from parallel process growth models. GMM simultaneously estimates multiple growth curves, such as sensation seeking and self-regulation, and allows researchers to identify different groups or classes of individuals that share similar trajectories of both sensation seeking and self-regulation. Parallel process growth models allow one to model growth in multiple constructs, such as sensation seeking and self-regulation, and to estimate associations between the growth curves. For example, used a parallel process growth model to estimate trajectories of impulse control and sensation seeking, and found that both increased over time during adolescence, that increases in impulse control and sensation seeking were independent of each other. While parallel process models provide descriptive information regarding growth in sensation seeking and self-regulation, they do not provide much information about imbalance. A unique advantage of GMM is that it allows researchers to identify different patterns of growth in sensation seeking and self-regulation, and thereby different patterns of the imbalance ( ). This feature of GMM aligns with recent calls for research on Dual Systems models, and adolescent development more broadly, to account for the heterogeneity in developmental changes which may help identify adolescents at greatest risk to engage in risk behaviors ( ; ; ; ).
Yet, GMM also has limitations. The reliability and validity of classes obtained in GMM has been questioned ( ; ). Researchers have called for the use of validity analyses, such as creating a nomological network, to provide support for obtained class structures ( ). Further, replication of class structure is important across multiple samples to provide evidence that growth patterns of sensation seeking and self-regulation and observed patterns of imbalance are not sample specific ( ). In addition to concerns pertaining to reliability and validity, growth mixture models can be difficult to estimate and often result in underidentified or inadmissible solutions, especially as more parameters are allowed to vary across classes ( ; ). Acknowledging the difficulties in estimating GMM, recommend a sequential approach to GMM that balances accounting for within-class heterogeneity with the pragmatics of model estimation. This approach is used in our GMM example for the current integrative review. As with the LDS approach, another limitation of this method is that it requires large sample sizes ( ).
In our example using GMM, we demonstrate how this approach can be used to model trajectories of growth in sensation seeking and self-regulation and assess the relationship of the growth trajectories to the probability of alcohol and marijuana use and levels or intensity of alcohol and marijuana use from early through late adolescence. Importantly, as with the LDS approach, measures of sensation seeking and self-regulation will need to be on the same metric in order to make the imbalance between trajectories of sensation seeking and self-regulation interpretable. Once sensation seeking and self-regulation are on the same metric, GMM allowed us to assess the following questions:
How do sensation seeking and self-regulation develop across ages 12–14?
What are the different prototypical patterns of growth for sensation seeking and self-regulation across ages 12–14?
Are different patterns of growth of sensation seeking and self-regulation associated with alcohol and marijuana use?
#### Moderation
A particularly appealing feature of GMM is that it permits assessment of moderation. This is important, as most developmental models of risk behavior emphasize context as an important factor that interacts with individual differences ( ; ; ). Although moderation of dichotomous variables in univariate LDS models is possible, the use of continuous moderators is still being developed (see ). The LDS growth model proposed in the current integrative review deviates from a typical LDS model in that the difference scores are a function of two variables and a latent growth curve is fit to the difference scores at each age. A benefit of this LDS growth model is that the specification of a latent growth curve permits using methods of assessing moderation of the association between latent slopes and intercepts and some outcome (see ). For example, it is possible for researchers to examine how other variables (such as peer delinquency and parental monitoring) might moderate the association between growth in the imbalance, in an LDS model, as well as classes representing different patterns of imbalance, in a GMM, and risk behaviors.
## Materials and methods
### Participants
To illustrate these proposed data analytic approaches to assessing the imbalance hypothesis with longitudinal data, we used data from a community sample of 387 families (1 adolescent and 1 caregiver from western New York state) assessed annually for 9 years. The study examined risk and protective factors associated with the initiation and escalation of early adolescent substance use. The sample was evenly split on sex (55% female) and was predominantly non-Hispanic Caucasian (83.1%) and African American (9.1%). Median family income was $70,000 and ranged from $1500 to $500,000, and 6.2% of the families received public income assistance. The demographic characteristics of our community sample are similar to those from whence the sample came (for more complete details, see ).
Participants had an average age of 12.1, 13.1, 14.1, 15.1, 16.1, 17.1, 18.4, 19.4, and 20.4 ( SD range = 0.59 to 0.67) at waves (W) 1 to W9, respectively. The sample included 387, 373, 370, 368, 361, 349, 352, 349, and 350 adolescents at W1 to W9, respectively. Overall attrition across W1 through W9 was low (9.6%).
### Procedures
For W1 to W3, adolescents and their parents were interviewed in university research offices. Informed consent and assent procedures were completed before the interviews began. Target families were compensated for their participation. W4 to W6 consisted of a brief telephone-based audio-computer-assisted self-interviewing (CASI) survey of substance use that took 10 to 15 min to complete. Parents provided consent over the phone and were given a phone number and PIN for their adolescent to use. Assent from the adolescent was obtained at the initiation of the audio-CASI survey. Procedures at W7 to W9 mirrored those of W1 to W3; however, adolescents and caregivers were provided with the option to complete the questionnaires online. Indicators of sensation seeking and self-regulation in our study were measured from W1 to W3, and substance use was assessed from W1 to W9 .
### Measures
#### Inhibitory control (W1-W3)
The current study assessed inhibitory control as the behavioral indicator of the self-regulation system (see ). The Stop Signal Task (SST; ) is among the most commonly used task of inhibitory control; the SST assesses participants’ abilities to inhibit a dominant response through the use of two concurrent tasks – a go task and a stop task. The integration method was used to compute the stop signal reaction time (SSRT) for each test block and then averaged across blocks ( ; ). To make it easier to interpret this measure, SSRT values were multiplied by a constant (-1) so higher values reflected stronger response inhibition. Cronbach’s alpha for inhibitory control for the current sample computed using the SSRT from each of the three experimental blocks was acceptable (α range = 0.66-0.73). Further, in order for inhibitory control to be on the same metric as sensation seeking, SSRT was collapsed across waves and participants and then standardized. Previous work with this sample has shown that higher SSRT (i.e., worse inhibitory control), in the context of high reward sensitivity, prospectively predicts delinquent behavior in adolescence ( ) and that initial levels of SSRT at age 11 prospectively predicts delinquent behavior in late adolescence ( ).
#### Sensitivity to reward (W1-W3 )
The Point Score Reaction Time Task for Children-Revised (PSRTT-CR; ) was used to assess sensitivity to reward as the behavioral indicator of the socioemotional system. This task starts with a practice block followed by four experimental blocks presented in a fixed order: no reward , reward , punishment , and post-punishment . Of interest in this task was the degree to which reaction times declined (i.e., got faster) during the reward compared to the no reward block (sensitivity to reward). Higher values on this task represented greater sensitivity to reward. In order for reward sensitivity to be on the same metric as inhibitory control, reward sensitivity was standardized across ages 12–14. Prior work with this sample has found faster growth in reward sensitivity, as measured on the PSRT, to be associated with more rapid escalation of substance use ( ). Reward sensitivity measured using the PSRT was also found be associated with parent report of reward sensitivity and physiological reactivity to reward ( ). Cronbach’s alpha for sensitivity to reward was adequate (α range = .73–.79) and computed by dividing the PSRT into three equal blocks at each age.
#### Substance use (W1-W9)
Alcohol use was assessed across W1 to W9 with questions assessing past year frequency of alcohol use, as well as past year typical quantity of alcohol use during a drinking occasion without parental permission. Indicators of past year frequency and quantity of alcohol use were multiplied to get an index of number of drinks consumed in the past year. Past year marijuana frequency was assessed at W1 to W9.
## Results
### Latent difference score growth model approach
A detailed account of the fitting of both the LDS growth model and our two-part growth models with random effects for alcohol use as well as Mplus output files can be found in the Supplemental Materials. As noted in the methods section, in order to create a meaningful difference, inhibitory control and sensitivity to reward were standardized to place them on the same metric ( ). Descriptive information regarding inhibitory control, sensitivity to reward, and alcohol and marijuana use can be found in . Analyses were conducted using Mplus version 8.2 using maximum likelihood estimation with robust standard errors and numerical integration to fit robust chi-square and standard error estimates (Muthén & Muthén, 1998–2017). Full information maximum likelihood estimation (FIML) was used in Mplus to handle missing data. Specification of our LDS model can be seen in .
Mean values for inhibitory control, sensitivity to reward, and alcohol and marijuana use from ages 12 to 20.
Table 2
IC = inhibitory control, SR = sensitivity to reward, Imb = the difference (imbalance) between sensitivity to reward and inhibitory control, I = intercept, and S = slope.
Fig. 2
As outlined by , the first step when estimating an LDS model with separate constructs is to assess for measurement invariance. Longitudinal measurement invariance was tested separately for inhibitory control and sensitivity to reward using the three blocks of the stop signal task as indicators of latent inhibitory control at ages 12, 13, and 14 and trials of the PSRT were broken into three equal blocks and used as indicators of latent sensitivity to reward. Partial residual invariance was supported for both inhibitory control ( χ = 46.18(33), p = .06, CFI=.98, TLI=.98, RMSEA=.03, SRMR=.05) and sensitivity to reward ( χ = 36.73(35), p = .38, CFI=.99, TLI=.99, RMSEA=.01, SRMR=.04). After establishing partial longitudinal measurement invariance for inhibitory control and sensitivity to reward, latent difference scores were specified to represent the imbalance between inhibitory control and sensitivity to reward at ages 12, 13, and 14 ( ).
As seen in , Imb is a LDS that mathematically represents an adolescent’s latent score on sensitivity to reward at age 12 minus their latent score on inhibitory control at age 12 (the imbalance). Thus, Imb , as well as Imb and Imb , represent the imbalance, accounting for measurement error. The LDS model provides information regarding adolescents’ imbalance at a static point in time.
Because a central tenet of the imbalance hypothesis pertains to the development of the imbalance across adolescence, we extended the LDS model to include growth in the latent imbalance difference scores , which we refer to as the LDS growth model. As seen in , a latent growth curve was fit to the latent imbalance difference scores at ages 12, 13, and 14. Through the specification of a latent intercept and slope, this LDS growth model provides information regarding adolescents’ initial levels of the imbalance and permits examination of whether there is significant variability in the imbalance at age 12 (σ I ). The model also includes a latent slope (S ) representing growth in the imbalance across ages 12 to 14 and whether there is significant variability in growth of the imbalance from ages 12 to 14 (σ S ). Through the estimation of a latent intercept and slope of the imbalance, researchers can then examine whether initial levels of the imbalance (in our example age 12) are associated with risk behaviors, and whether growth in the imbalance across adolescence (in our example growth from ages 12 to 14) is associated with risk behaviors.
The LDS growth model provided an adequate fit to the data ( χ = 266.22(148), p < .001, CFI=.92, TLI=.91, RMSEA=.04, SRMR=.12). Linear growth in the latent differences at ages 12, 13, and 14 was compared to a model where the loading of the growth curve at age 13 was freely estimated. Freeing the loading at age 13 led to a significant improvement in model fit. The slope mean of -0.47 (p<.001) indicated that the difference between sensitivity to reward and inhibitory control decreased with age. There was also evidence of significant variability in the intercept (σ = 0.25, p < .001) and slope (σ = 0.18, p < .001) suggesting that initial levels of the imbalance and growth in the imbalance varied significantly across individual adolescents. The covariance between the intercept and slope of the imbalance was also significant (σ =-0.18, p < .001), indicating that higher levels of an imbalance at age 12 was associated with quicker decreases in the imbalance across ages 12–14. Mplus syntax output files for our LDS growth model can be found in Supplemental Materials 2.
Next, we assessed whether the intercept and slope terms from our LDS growth model covaried with the intercept and slope factors from a two-part growth model for both alcohol and marijuana use that spanned ages 13–20 (see ). Two-part models simultaneously model dichotomous past year use (yes/no) as well as continuous levels of past year use (quantity x frequency of use for alcohol and frequency for marijuana). This modeling framework provides an analytic approach to account for the large proportion of zeros often observed when studying risk behaviors ( ) and distinguishes probability of use from intensity or levels of use. Hence, this modeling approach allows for the examination of whether initial levels of the imbalance, as well as growth in the imbalance, were associated with (1) the initial probability of alcohol and marijuana use at age 13, (2) growth in the probability of alcohol use across ages 13 to 20, (3) initial levels of alcohol and marijuana use at age 13, and (4) growth in levels of alcohol and marijuana use across ages 12 to 20 and 13 to 20, respectively. For information regarding model specification and fitting of the two-part growth models with random-effects see Supplemental Materials 1 and 2.
LDS approach to assessing the relationship between the imbalance and the probability of alcohol use and growth in alcohol use across ages 12–20. IC = inhibitory control, SR = sensitivity to reward, Imb = the latent difference (imbalance) between sensitivity to reward and inhibitory control, AU = alcohol use, D = dichotomous use (use vs. no use), C = continuous levels (quantity x frequency) of past year use, I = intercept, and S = slope. Solid two-headed arrows depict significant covariances. Dashed two-headed arrows depict non-significant estimated covariances between the imbalance and alcohol use.
Fig. 3
For our alcohol use two-part growth curve model with random effects, the intercept for the probability of alcohol use, as well as levels of alcohol use, were set to age 13 due to the small number of users at age 12 (N = 15) (see ). A piecewise model provided the best fit for modeling growth in the continuous portion of the model. The first slope represents change from ages 12 to 16 and the second slope represents change from ages 16 to 20. The means for the dichotomous and continuous slopes were all positive and significant indicating significant increases in the probability and levels of alcohol use from ages 12 to 20. The intercept for the dichotomous portion, intercept for the continuous portion of the model, and all slopes for the dichotomous and continuous portions of the model had statistically significant variability. This suggests individual variability in growth in alcohol use.
Non-linear slope factors provided the best fit for the dichotomous and continuous portions of the marijuana two-part growth model. The slope mean for both the dichotomous portion and continuous portions of the marijuana model were statistically significant and indicated significant increases in the probability of marijuana use and frequency of use from ages 13 to 20. Further, there was significant variability in the intercept of the dichotomous portion of the model and all slopes had significant variability, again suggesting individual variability in growth.
Next, the LDS growth model and two-part models for alcohol and marijuana use were combined to assess the relationship between the imbalance and substance use. power tables indicated that these models had sufficient power to detect a small to medium sized effect between the intercept and slope of the imbalance and growth in alcohol and marijuana use. For both the alcohol and marijuana use models, there were no significant covariances between the intercept and slope of the imbalance and the intercepts and slopes of alcohol and marijuana use. This suggests that the magnitude of the imbalance in early adolescence and growth in the imbalance across early-to-middle adolescence was unrelated to substance use in early-to-late adolescence.
### Growth mixture modeling approach
GMM is an analytic technique that allowed us to simultaneously model growth in sensation seeking and self-regulation and to identify classes or groups that shared similar patterns of growth. The classes provide a description of different developmental patterns of imbalance. Detailed information regarding specification of our GMM and determination of class structure can be found in the Supplemental Materials. Analyses were conducted using Mplus version 8 ( ). The three standardized sensation seeking and inhibitory control blocks were averaged at ages 12, 13, and 14, respectively, and used for these analyses using the same standardization procedure discussed above. A sequential process was used to estimate our GMM, where univariate growth models were first estimated for inhibitory control and sensitivity to reward ( ). Next, a series of growth mixture models were estimated: (1) GMM with unique means (also known as the Latent Class Growth Model) where means of the intercepts and slopes of inhibitory control and sensitivity to reward are allowed to vary freely across classes and variances and covariances of the growth factors are fixed to zero, (2) GMM with unique means and shared variance where means are once again allowed to vary across classes and variances and covariances are estimated for the growth factors but constrained to be equal across class solutions, and (3) and the GMM with unique means and variances where both means and variances are allowed to freely vary across classes.
A growth model where the slope loading for inhibitory control at age 13 was freely estimated provided the best fit to the data (χ (0) = 0.00, p = 1.00, CFI = 1, TLI = 1., RMSEA = 0.00, SRMR = .01). The slope mean was significant (M = 0.52, p < .001) indicating significant growth in inhibitory control from ages 12 to 14. The variance of the intercept of inhibitory control at age 12 (σ = 0.37, p < .001) was statistically significant, indicating significant individual differences in initial levels of inhibitory control. The slope variance for inhibitory control was not statistically significant (σ = 0.18, p = .40) limited variability in change in inhibitory control from ages 12 to 14.
When modeling growth in sensitivity to reward, the residual variance for age 12 sensitivity to reward was initially estimated to be negative and was constrained to 0. A model where age 13 sensitivity to reward was freely estimated provided the best fit to the data (χ (2) = 3.33, p = .18, CFI = .98, TLI = .98, RMSEA = 0.04, SRMR = .05). The slope mean was significant (M= -0.09, p = .01), indicating significant declines in sensitivity to reward from ages 12 to 14. The variance of the intercept of sensitivity to reward at age 12 (σ = 0.41, p < .001) was statistically significant, indicating significant individual differences in initial levels of sensitivity to reward. The variance of the slope for sensitivity to reward was also significant (σ = 0.34, p < .001), indicating significant individual differences in declines in sensitivity to reward from ages 12 to 14.
The Akaike information criterion (AIC), Bayesian information criterion (BIC), sample size adjusted BIC, entropy, class size, and bootstrapped likelihood ratio test (BLRT) were all used to determine the number of classes to extract ( ). Relative to a single class solution, a two class solution is supported by lower information criteria and a significant BLRT. These fit criteria were compared both within growth mixture modeling strategies (e.g., comparing fit criteria in the two and three class solutions for the GMM with unique means and shared variances) as well as across growth mixture modeling strategies (e.g., comparing fit criteria for the best class solution for the GMM with unique means to the best fitting class solution for the GMM with unique means and variances) to determine the final class solution.
The two class GMM with unique means and shared variances was selected as the final class solution. Information criteria and the BLRT suggested extracting more than 2 classes; however, we elected to retain a 2-class solution considering there was an extremely small class in the 3 class solution (N = 8, 2% of sample). The GMM with unique means and shared variances had a lower AIC, BIC, and aBIC than the best GMM with unique means solution and a higher entropy than the GMM with unique means and variances (see Supplemental Materials 1 for greater detail and Mplus output files). GMM with unique means and shared variances constrains variances and covariances to be equal across classes while allowing for unique mean patterns of growth across classes. The benefit of this GMM approach is that it allows for variability in growth parameters while being more stable than models with unique variance estimates in each class ( ).
Developmental patterns of the two classes can be found in . The intercept of inhibitory control (σ = 0.37, p < .001), slope of inhibitory control (σ = 0.33, p < .001), intercept of sensitivity to reward (σ = 0.40, p < .001), and growth in sensitivity to reward (σ = 0.34, p < .001) all had significant variability in the two class GMM with unique means and shared variances. Class 1, which consisted of 39 adolescents (11%), was characterized by a larger imbalance at ages 12 and 13 where sensitivity to reward was greater than inhibitory control and inhibitory control became greater than sensitivity to reward at age 14. Class 2 consisted of 323 adolescents (89%) and was characterized by a smaller imbalance, relative to Class 1, at ages 12–13 where sensitivity to reward was higher than inhibitory control, but then inhibitory control was greater than sensitivity to reward at age 14. The patterns of growth found in both Class 1 and Class 2 lend partial support to the imbalance hypothesis that adolescents’ sensitivity to reward was higher than their inhibitory control at ages 12 and 13. However, contrary to the tenet of imbalance hypothesis that argues that the imbalance should be largest in middle adolescence, both Class 1 and Class 2 found higher levels of inhibitory control relative to sensitivity to reward at age 14.
Final two class solution for the growth mixture model with unique means and shared variances.
Fig. 4
Next, we created a categorical variable representing class membership by assigning adolescents to their most likely class using class probabilities. Using the two-part growth models as described above, the probability of alcohol and marijuana use, as well as levels of alcohol and marijuana use, were regressed on the categorical class membership variable. power tables indicated that these models had sufficient power to detect a small to medium sized effect between class membership and growth in alcohol and marijuana use. No significant associations were found between class membership and either the probability of alcohol or marijuana use or levels of alcohol or marijuana use, suggesting that the different patterns of imbalance were not associated with growth in substance use.
## Discussion
Dual Systems models are popular and influential theoretical accounts of adolescent risk behavior ( ). As the field accumulates longitudinal data of the socioemotional and self-regulation systems at both neural and behavioral levels of analysis, there will be exciting opportunities to increase our understanding of the etiology and impact of risk behaviors, such as substance use. In line with recent calls to improve specificity of prediction models ( ), the current paper sought to highlight the need for more rigorous data analytic methods to test the imbalance hypothesis forwarded by Dual Systems models. We have argued that statistical techniques such as Latent Difference Score (LDS) and Growth Mixture Models (GMM) provide more precise and theoretically-consistent tests of the imbalance hypothesis than other approaches that have been used in the literature thus far, such as simple observed difference scores and regression approaches that test multiplicative interactions.
We reiterate that our goal of the current paper was not to provide support for or against Dual Systems Models. Our goal was to bring attention to data analytic methods that more appropriately quantify the imbalance than previously-used methods. Nevertheless, our empirical examples did not demonstrate expected relationships between the imbalance and risk behavior. Although our measures do not perfectly reflect the socioemotional and cognitive control systems ( ), these findings are not due simply to measurement concerns. Indeed, using the same behavioral tasks from the current study, we have previously demonstrated that increases in reward sensitivity predict increases in substance use ( ), and that poor self-regulation in the context of high sensitivity to reward predicts rule-breaking behavior ( ). These two studies utilized data analytic methods most akin to the regression and moderation approaches discussed earlier. Taken together, our findings tentatively suggest that data analytic methods that do not adequately quantify the imbalance can be misconstrued as support for the imbalance hypothesis, whereas methods that are more appropriate for measuring the imbalance do not. As more large-scale, longitudinal studies collect data that will be suitable to the LDS and GMM approaches discussed herein, we look forward to seeing the extent to which this pattern is replicated with other samples.
### Summary of LDS growth model approach
The LDS modeling approach allowed us to assess a number of important tenets of Dual Systems models. First, LDS modeling allowed for a more reliable estimate of the difference between our measures of sensation seeking and self-regulation than could be accomplished with observed difference scores. Second, it allowed for examining change in the imbalance across ages 12 to 14. Another advantage of this modeling approach was that it allowed for the examination of whether initial levels of the imbalance at age 12, as well as growth in the imbalance from ages 12 to 14, were associated with risk behaviors.
Of note, a potential conceptual limitation of the LDS growth model is that it does not provide information regarding what is leading to the imbalance. This is particularly problematic for studies interested in understanding the factors, such as substance use, that are thought to alter socioemotional and cognitive control systems ( ; ; ). For example, growth in the imbalance across adolescence could be a function of increasing sensation seeking and stable self-regulation, rapidly increasing sensation seeking and slowly increasing self-regulation, or stable sensation seeking and decreasing self-regulation. The LDS growth model does not provide information that would distinguish these multiple possibilities for growth in the imbalance across adolescence.
### Summary of GMM approach
GMM provides a flexible statistical approach to model patterns of development in sensation seeking and self-regulation. Similar to the LDS method, this approach allows for heterogeneity in growth patterns. Although the LDS growth model and GMM approaches both account for within -person change, GMM is a person-centered analytic technique that allows for the examination of different patterns or classes of growth in sensation seeking and self-regulation. The classes can provide a description of the different magnitudes and developmental patterns of imbalance, and how distinct growth patterns may be related to risk behavior. This feature of GMM is consistent with current conceptualizations of adolescent brain development that note that there is heterogeneity across adolescents in their patterns of growth in sensation seeking and self-regulation (e.g., ).
### Relevance to neuroimaging data
Although the focus of the current integrative review was on testing the imbalance hypothesis with behavioral data, we see no reason why the data analytic approaches we discussed could not be extended to certain types of neuroimaging data. LDS and GMM approaches could be applied to neuroimaging data that include indicators of both the socioemotional and cognitive control systems (e.g., striatum activation during a rewarding tasking and lPFC activation during an emotionally-salient task). The increased sophistication of the LDS and GMM approaches relative to past methods of assessing the imbalance, in conjunction with their ability to model both behavioral and neuroimaging data, may be helpful in the field’s efforts to more rigorously test the imbalance hypothesis and determine whether findings are consistent across methods.
Further, the ability of the LDS and GMM approaches to model neuroimaging data allows for the examination of competing views of how sensitivity to reward and cognitive control contribute to risk behavior during adolescence. have suggested that examining connectivity between the socioemotional and cognitive control systems may provide important insight into Dual Systems models over examining solely the imbalance hypothesis. Indeed, studies examining connectivity between the socioemotional and cognitive control systems have informed our understanding of how the socioemotional and cognitive control systems are connected, and how differences in their connectivity across adolescence is related to risk behavior (e.g. ; ). Using the LDS approach or GMM approach to model the imbalance, future work could examine whether growth in the imbalance, as forwarded by the imbalance hypothesis, or growth in functional connectivity of the socioemotional and cognitive control systems across adolescence, is more strongly related to risk behavior. This example highlights how improving the specificity and precision of modeling the imbalance can help facilitate the generation of riskier prediction models and allow for testing competing models.
### Conclusion
The maturational imbalance hypothesis is a cornerstone of Dual Systems models, which have shaped conceptualizations of adolescent risk behavior and had a substantial impact on public policy to address adolescent delinquency ( ; ). Although the maturational imbalance hypothesis has been taken as having garnered a lot of support ( ), we have argued that this tenet of Dual Systems models has not been rigorously tested. Commonly-used methods to date, such as the regression approach, moderation approach, and observed difference score approach each fail to adequately test this tenet of Dual Systems models. We proposed two promising techniques to assess the maturational imbalance hypothesis, a Latent Change Score approach and a Growth Mixture Modeling approach, which we argued provide a more rigorous evaluation of this hypothesis. We acknowledge that these are not the only longitudinal data analytic methods that can be used to assess the imbalance with longitudinal data (e.g., multilevel modeling). However, we view these methods as making significant improvements over techniques used to date. We hope this integrative review pushes the field to wrestle with the question of how best to assess the imbalance and its relation to risk behavior with longitudinal data. Doing so will help to refine and constrain theory and advance our understanding of adolescent development.
## Funding
This research was supported by a grant from the (R01DA019631) awarded to Craig R. Colder and a grant from the (F31AA025521) awarded to Samuel N. Meisel.
## Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
|
Prenatal opioid exposure has been linked to altered neurodevelopment and visual problems such as strabismus and nystagmus. The neural substrate underlying these alterations is unclear. Resting-state functional connectivity MRI (rsfMRI) is an advanced and well-established technique to evaluate brain networks. Few studies have examined the effects of prenatal opioid exposure on resting-state network connectivity in infancy. In this pilot study, we characterized network connectivity in opioid-exposed infants (n = 19) and controls (n = 20) between 4–8 weeks of age using both a whole-brain connectomic approach and a seed-based approach. Prenatal opioid exposure was associated with differences in distribution of betweenness centrality and connection length, with positive connections unique to each group significantly longer than common connections. The unique connections in the opioid-exposed group were more often inter-network connections while unique connections in controls and connections common to both groups were more often intra-network. The opioid-exposed group had smaller network volumes particularly in the primary visual network, but similar network strength as controls. Network topologies as determined by dice similarity index were different between groups, particularly in visual and executive control networks. These results may provide insight into the neural basis for the developmental and visual problems associated with prenatal opioid exposure.
## Introduction
Opioid use during pregnancy remains high in the US, with one opioid-exposed infant born every 15 min ( ). The effects of prenatal opioid exposure on the developing brain remain poorly understood. Multiple studies on outcomes in opioid-exposed children show adverse effects of maternal opioid use on neurodevelopment, behavior, and vision ( ; ; ; ; ; ). In addition, animal studies show negative effects of maternal opioids on fetal oligodendrocytes and neurons. However, few human imaging studies have evaluated the effects of opioid exposure on the developing fetal brain.
In experimental animal models, prenatal exposure to opioids is associated with decreased neurotransmitter levels ( ; ; ), decreased neurogenesis ( ; ), increased apoptosis ( ), and altered myelination ( ; ). In addition, several studies in animal models (mostly rats and mice) show impaired learning and memory in the offspring of dams exposed to opioids during gestation ( ; ; ; ), In humans, multiple recent meta-analyses have examined the effects of prenatal opioid exposure on neurodevelopment and behavior ( ; ; ; ; ; ). All found that opioid-exposed infants and children performed more poorly than non-exposed children across all domains examined, including cognition, motor development, behavior, attention, and vision. Most of these meta-analyses commented that the results may be biased, with many studies unable to control for additional risk factors for poor outcome in the opioid-exposed group.
Neonatal imaging studies suggest that prenatal opioid exposure may affect the structure and microstructure of the developing brain. In particular, infants prenatally exposed to methadone show decreased fractional anisotropy on diffusion tensor imaging in the first weeks of life compared with unexposed infants ( ). Infants exposed to opioids prenatally show decreased brain volumes in multiple regions compared to unexposed infants ( ).
Resting-state functional MRI (rsfMRI) is a powerful technique that can be used to gain insight into the brain’s functional organization during infancy. Studies have consistently shown that functional networks in the brain are present at birth and mature over the first years of life ( ), and small-world network topology is also present at birth and becomes more efficient over the first 2 years ( ; ). Normal development of brain networks and topology is known to be altered by prenatal drug ( ; ) and alcohol ( ) exposure. To our knowledge, only one small study has evaluated the functional connectivity of the brain in opioid-exposed infants, focusing on the amygdala ( ). The study found increased functional connectivity from the amygdala to several cortical areas in exposed infants compared to controls.
The aim of our pilot study in infants 4–8 weeks of age was to examine functional connectivity using rsfMRI in infants exposed to opioids prenatally and unexposed controls. We aimed to characterize network-specific differences between the two groups using a whole-brain connectomic approach. In addition, we focused on a specific set of functional networks known to underlie particular sensory and cognitive functions and examined connectivity with these networks as seeds. Previous studies on prenatal substance exposure have consistently shown functional hyperconnectivity in exposed groups compared with controls within pre-selected brain regions and/or networks ( ; ; ). Given these previous results and the consistent effects of prenatal opioid exposure on development, behavior, and sensory functions (vision in particular), we extended this area of research by further evaluating how opioid exposure may impact brain functional development at both the whole brain and network levels and hypothesized that opioid-exposed infants would show unique functional connectivity features when compared to control infants in visual and higher order networks.
## Materials and methods
### Participants
Infants born at ≥37 weeks gestation with prenatal opioid exposure and no other medical problems were recruited from clinics at Cincinnati Children’s Hospital or from birth hospitals in the Greater Cincinnati area. Healthy control infants born at ≥37 weeks gestation were recruited from birth hospitals, from the Pediatric Primary Care clinic at Cincinnati Children’s, or through community research advertisements. Infants in both groups were excluded if they had known chromosomal disorders or congenital anomalies, required any positive pressure ventilation outside of the delivery room, or had any conditions besides neonatal opioid withdrawal syndrome that required a NICU stay. Prenatal opioid exposure was determined by maternal history and/or maternal urine toxicology screen at the time of delivery and confirmed with neonatal toxicology screen (meconium or umbilical cord). All infants in the opioid-exposed group had exposure to opioids throughout pregnancy, but detailed information about the opioid(s) infants were exposed to during each week of pregnancy was not available. Lack of opioid or other drug exposure in controls was confirmed by maternal history and maternal urine toxicology screen at the time of delivery, which is standard clinical practice in our region. Additional information about drug exposure was collected by review of infant medical records and by maternal questionnaire at the time of MRI. Pregnancy and birth history were collected by review of infant medical records. Information about maternal socioeconomic status (including education, employment, and income) and race was collected by maternal questionnaire at the time of MRI. This study was approved by the Institutional Review Boards at Cincinnati Children’s Hospital, Good Samaritan Hospital, and St. Elizabeth Hospital. Written informed consent was obtained from a parent or guardian prior to any study procedures.
### MRI imaging acquisition
Infants were scanned during sleep with no sedation. All infants had been weaned off medications for neonatal opioid withdrawal syndrome by the time of scan. Infants were fed, swaddled, fitted with ear protection, placed in the Med-Vac vacuum bag (CFI Medical Solutions, Fenton MI), and moved into the scanner bore. All infants were scanned on the same Philips 3 T Ingenia scanner with 32 channel receive head coil in the Imaging Research Center at Cincinnati Children’s Hospital. Structural MR imaging included a sagittal magnetization prepared inversion recovery 3D T1-weighted gradient echo sequence (shot interval = 2300 milliseconds, repetition time = 7.6 milliseconds, echo time = 3.6 milliseconds, inversion time = 1100 milliseconds, flip angle = 11 degrees, voxel size 1 mm x 1 mm x 1 mm, acceleration (SENSE) = 1 in plane and 2.0 through plane (slice) phase encode, scan time 3 min 6 s) and an axial 2D T2-weighted fast spin echo sequence (repetition time = 19100–19500 milliseconds, echo time = 166 milliseconds, voxel size 1 mm × 1 mm × 1 mm, acceleration (SENSE) = 1.5 in plane phase encode, scan time 3 min 50 s). Resting state functional MRI was obtained using an axial gradient echo echo-planar imaging sequence with simultaneous multi-slice excitation (multi-band) (repetition time = 1011 ms, echo time = 45 ms, flip angle = 54 degrees, voxel size 2.25 mm × 2.25 mm × 2.25 mm, 60 contiguous slices, multi-band factor = 6, 500 dynamics, scan time 8 min 37 s). Structural images were reviewed by a board-certified pediatric neuroradiologist to confirm there were no significant clinically relevant abnormalities (BK).
### Statistical analysis of demographic/clinical data
This analysis was performed in STATA 16.0 (Stata Corp., College Station, TX). Descriptive statistics for the two groups (opioid-exposed and controls) were computed. Groups were compared using two-sided t-tests for continuous variables and Fisher’s exact test for categorical variables, with a p-value of <0.05 considered significant for demographic variables.
### MRI processing
#### Pre-processing
MRI data were pre-processed using an in-house infant-specific pipeline ( ; ; ) which shares some common steps with the HCP pipeline ( ), including head motion correction, alignment of rsfMRI images to T1 space, and band-pass filtering (0.01 Hz-0.08 Hz), but adds several unique steps tailored to infant functional connectivity MRI ( ). Brain tissue segmentation was first conducted to generate tissue labeling maps (gray matter, white matter, or cerebrospinal fluid) using a multi-site infant-dedicated computational toolbox, iBEAT v2.0 Cloud ( ) ( ). The tissue labeling maps were used to register to the template (Colin 27 atlas) ( ) in MNI space using advanced normalization tools (ANTs) ( ). Automatic noise-related component detection and regression were performed ( ). Specifically, a deep learning-based rsfMRI QC method, the long short-term memory (LSTM) neural network ( , ), was employed, capable of effectively extracting FC quality-related features from the raw data. The deep learning-based approach considered not only head motion indices that have been widely used, but also several other important parameters, including the signal-to-noise of the raw images, temporal signal-to-noise ratio (tSNR) as well as grey-scale differences between neighboring volumes ( ; ). Specifically, the LSTM neural network method for deep time-series learning was implemented, aiming to yield three classes of image quality: passed, questionable, or failed. To achieve this goal, we adopted a two-stage strategy. We first engineered eight diagnostic features (maximum grey-scale difference between neighboring volumes of raw data, tSNR of raw fMRI data, maximum translation and rotation along all directions, mean framewise displacement (FD), percentage of number of the time points with large FD > 0.5 mm, quality of FC in primary visual and sensorimotor network) at different processing stages to generate initial QC labels, including passed, failed and questionable. Subsequently, we leveraged semi-supervised learning LSTM-based QC model to further examine the data that were deemed questionable and reassigned these data to either passed or failed. All preprocessing steps, including resampling and denoising, were conducted in each subject’s native space.
#### Connectome-based analysis
The preprocessed functional data was parcellated into 90 regions of interest (ROIs) encompassing both cortical and subcortical areas in the AAL atlas ( ). First, the atlas in the standard Montreal Neurological Institute (MNI) space was warped to each subject’s native space to extract regional averaged time series for post-processing. Pairwise functional connections were then calculated in the original space and Fisher z-transformed. Positive connections that differed significantly from zero were identified separately for the opioid-exposed and control groups (p < 0.05, corrected for multiple comparisons via false discovery rate (FDR)). We focused on positive connections for this analysis given the controversy in the literature about the meaning of negative connections ( ; ; ) although we have included information about negative connections in the supporting materials. The identified significant connections were further separated into three groups: the connections that overlapped between the two groups as well as unique functional connections in each group.
To better characterize the network architecture present in the opioid-exposed and control groups, we chose to focus on the graph theory metric betweenness centrality as a primary method to determine which regions played the most central role in opioid-exposed and control groups separately. Betweenness centrality refers to the fraction of all shortest paths in a network that pass through a given node, thus identifying the most central nodes in a network ( ). We also calculated normalized rich-club coefficients for each subject. The rich-club coefficient is a quantitative measure to discern the presence or absence of rich-club topology. We next examined the anatomic length of unique connections in each group and those that overlapped, using their anatomical coordinates in MNI space ( ). Finally, we examined how many of the unique and overlapping connections were components of particular known functional networks; specifically, the visual, default mode (DMN), sensorimotor, auditory, and attention networks using the network parcellation proposed by Tao et al. .
#### Seed-based analysis
To further assess if prenatal opioid exposure impacts brain functional networks, a seed-based approach was employed. Eight networks using the regions reported by Smith et al. were used as seeds: V1 (medial visual), V2 (occipital pole), V3 (lateral visual), DMN (default mode network), SM (sensorimotor), AN (auditory), SA (executive control/salience), and FPN (bilateral frontoparietal). First, we transformed each seed into individual original space, then extracted the time series of each seed based on the preprocessed data with spatial smoothing using a 4 mm full-width half-maximum kernel. AFNI software ( ) was used for all seed-based calculations. The connectivity map of each seed was generated in subject space and then transformed to MNI space for subsequent statistical comparison and modeling. Regions of significant functional connectivity within a given seed network in each group were calculated on a voxel-wise basis (p < 0.05, FDR corrected, minimum cluster size: 50). Using the resulting mask of regions of significant functional connectivity, we calculated the cluster volume (number of voxels) and mean connectivity strength (r) for each seed network for each group, as well as the dice similarity index, which measures the spatial similarity between two sets of topologies. A dice index of >0.7 indicates that networks have similar topologies ( ; ).
## Results
Demographic data is shown in . The groups had similar sex and age distributions. Of a total sample of 49 infants (21 opioid-exposed and 28 controls) with completed MRIs, 10 infants were excluded due to excessive motion or poor data quality, leaving a final sample of 19 opioid-exposed and 20 control infants for analysis. Of the 10 excluded infants, six subjects had less than 250 time points after removing time points that did not pass QC and four subjects had incomplete brain coverage. We compared the numbers of time points that were removed between the opioid-exposed (433.70 ± 70.55) and control (427 ± 81.62) groups using a t -test and no significant differences were observed (p = 0.7634). On review by a pediatric neuroradiologist, no infant in either group had overt brain injury. Two infants in the opioid-exposed group and two infants in the control group had incidental findings of small cerebellar hemorrhages. Three infants in the opioid-exposed group each had a single punctate periventricular white matter lesion. Information about neonatal opioid withdrawal syndrome for the 19 opioid-exposed infants is provided in .
Demographics of study population.
Table 1
Information about neonatal opioid withdrawal syndrome symptoms and treatment.
Table 2
### Connectome-based analysis
We found 121 significant positive connections unique to the opioid-exposed group, 135 significant positive connections unique to the control group, and 459 positive connections that were common between the two groups. We compared the connection length among these connections. The connections unique to the opioid-exposed group as well as the connections unique to the control group were longer than those that overlapped between groups, which tended to be shorter ( t (37) = 8.63, t (37) = 9.88, respectively, both p < 0.05) ( ). To determine if these findings depended on the choice of atlas, we performed the same analysis using the Harvard-Oxford atlas. Results using the Harvard-Oxford atlas were similar and are provided in Supplemental materials (A-1: Effects of Atlases, Fig. A1). To ensure that the distance measures were not affected by differences in brain volumes, we compared brain volumes between groups and found no statistical differences (Table A2) between the two groups for total brain, gray and white matter volumes. This is consistent with our previous published work, which found differences in regional brain volumes but no difference in total brain volume or overall white or gray matter volumes. We therefore did not include these parameters as covariates in our analyses (A-2: Comparisons of brain volumetric measures).
Comparison of length of positive connections common to both groups, positive connections unique to the opioid-exposed group, and positive connections unique to controls.
Fig. 1
With the identified significant connections unique to the opioid-exposed group, we further determined the potential associations between the individual connection strengths and the highest Finnegan scores of the opioid-exposed subjects. and show the connections exhibiting significant associations with the highest Finnegan scores (no correction for multiple comparisons). While these associations are marginally significant owing to the limited sample size, they are largely related to subcortical connections (4 out of 11), with the amygdala-insula connection and caudate-anterior cingulum connection among the smallest p-values, consistent with our previous findings showing reduced brain volumes in the subcortical areas in opioid-exposed infants ( ). Associations with the average Finnegan scores were also analyzed (A-4: Associations between connection strengths and Finnegan scores), exhibiting similar findings (Figs. A5 and A6 and Table A4) as those using the highest Finnegan scores.
Connections showing significant associations with highest Finnegan scores. Red lines and points represent a positive association with the highest Finnegan scores, while the blue lines and points represent negative association.
Fig. 2
Connections with significant associations with highest Finnegan scores.
Table 3
Rich-club coefficients with respect to a range of degrees ( k ) for both the opioid-exposed and control groups are shown in Fig. A7 for different sparsities. Both the opioid-exposed and control groups exhibit a rich-club topology span over a large range of k ( Φ > 1). Although there appear to be some differences between the two groups for low sparsity, these were not statistically significant.
We evaluated in which known resting-state networks the unique and overlapping connections were seen and whether the connections were within the network (intra-network) or between networks (inter-network). As shown in , the connections unique to the opioid-exposed group were much more likely to be inter-network connections than the connections in the other two groups.
Distribution of positive connections within networks. a Connections unique to opioid-exposed group b Connections unique to controls c Overlapping connections. Inter-network connections are shown in blue and intra-network connections are shown in red. Def = default mode, Att = attention, Vis = visual, Aud = auditory, SM = sensorimotor.
Fig. 3
The distribution of betweenness centrality was different between the opioid-exposed group and controls ( ). a–c show the brain regions and d–f show the betweenness centrality distribution of the highest 25 % betweenness centrality for the opioid-exposed, controls, and overlapping connections, respectively. Regions with the highest betweenness centrality in the connections unique to the opioid-exposed group were the left inferior frontal gyrus (pars triangularis), left medial orbitofrontal cortex, right paracentral lobule, and left anterior cingulate cortex. In connections unique to the control group, these regions were the right inferior frontal gyrus (all subregions), the right precuneus, and the left superior medial frontal region. Betweenness centrality was overall higher in the regions among which connections overlapped between the two groups. Regions with highest betweenness centrality were bilateral insula, left anterior cingulate, right postcentral gyrus, left inferior orbitofrontal cortex, right middle cingulate, and the left superior occipital region.
Distribution of node betweenness centrality. a Betweenness centrality of unique positive connections in opioid-exposed group b Betweenness centrality of unique positive connections in controls c Distribution of betweenness centrality of overlapping connections seen in both groups d List of regions with the highest betweenness centrality unique to the opioid-exposed group e List of regions with the highest betweenness centrality unique to controls f List of regions with the highest betweenness centrality seen in both groups.
Fig. 4
### Seed-based analysis
The seed-based network topologies of the opioid-exposed and controls in 8 networks – V1 (medial visual), V2 (occipital pole), V3 (lateral visual), DMN (default mode network), SM (sensorimotor), AN (auditory), SA (executive control/salience), FPN (bilateral frontoparietal), are shown in . We found no significant difference in individual strength between the two groups in any of these networks ( a). However, the two groups showed different network topology. Controls had a larger number of connected voxels with all seed networks evaluated except for the default mode network ( b). This is further illustrated by the dice index for each network. The dice index was <0.7 for all but one network (sensorimotor), suggesting that the opioid-exposed group and controls had different network topologies ( c).
Illustration of different network topologies in the opioid-exposed group versus controls. V1 = medial visual, V2 = occipital pole, V3 = lateral visual, DMN = default mode network, SM = sensorimotor, AN = auditory, SA = executive control/salience, FPN = bilateral frontoparietal.
Fig. 5
Network strength and topology in opioid-exposed (blue) and control (red) groups. a Mean functional connectivity strength b Volume c Dice index. Network 1 = medial visual, Network 2 = occipital pole, Network 3 = lateral visual, Network 4 = default mode network, Network 5 = sensorimotor, Network 6 = auditory, Network 7 = executive control/salience, Network 8 = bilateral frontoparietal. FC = functional connectivity.
Fig. 6
## Discussion
The aim of this study was to investigate differences in functional networks and network-specific characteristics in infants with prenatal opioid exposure and controls. We found that connectivity and network topology are different in infants with prenatal opioid exposure. Specifically, we found different patterns of unique functional connections in each group, and different topologies and volumes of resting state networks. Our results suggest that prenatal opioid exposure influences brain connectivity early in life, which may underlie some of the later developmental, behavioral, and visual issues seen in this population.
Our findings are consistent with previous work in neonates exposed to substances prenatally. Although the literature on infants exposed to opioids prenatally is sparse, opioids appear to affect brain volume ( ) and microstructure ( ; ). In one small study of 10 opioid-exposed infants and 12 controls using seed-based analysis, higher resting-state connectivity between the right and left amygdala and several cortical regions were seen in the exposed group ( ). In the broader drug exposure literature, prenatal substance exposure has been associated with altered connectivity. Prenatal marijuana exposure is associated with hypoconnectivity in the insula and connections in the striatum ( ). Multi-drug exposure is associated with hyperconnectivity of the left amygdala to orbitofrontal cortex and hypoconnectivity of the posterior thalamus to the hippocampus ( ). Prenatal cocaine exposure is associated with hyperconnectivity between the thalamus and frontal regions, and polysubstance exposure is associated with hypoconnectivity between the thalamus and motor-related regions ( ). Prenatal substance exposure is associated with connectivity disruptions (both hyper- and hypo-connectivity) within the amygdala-frontal, insula-frontal, and insula-sensorimotor circuits ( ).
Opioid receptors are present in all areas of the brain, but especially widely distributed in the cortex, limbic system, and brain stem ( ). In experimental models, prenatal exposure to buprenorphine or methadone affects neurotransmitter biosynthesis ( ), neurogenesis ( ), and myelination ( ; ). Prenatal opioid exposure induces long-term alterations in brain and behavior in rats, affecting multiple brain sites and multiple neurotransmitter systems ( ). Although the preclinical literature is inconsistent, likely due to different animal species and variations in type and dosing of opioids, it appears that prenatal opioid exposure alters normal modulatory influences in the developmental program of the brain in animal models ( ).
Using a whole-brain connectomic approach, we found a unique pattern of connections in the opioid-exposed group and controls, with connections that were common to both groups. We evaluated betweenness centrality of nodes for overlapped connections and connections unique to each group. Nodes with high betweenness centrality are important for efficient communication across the network as a whole ( ). Higher betweenness centrality means more connections through that node to achieve the shortest path lengths. The regions with the highest betweenness centrality we found that were unique to the opioid-exposed group included a number of nodes in the frontal, central and occipital regions, consistent with previously reported results ( ; ; ; ). It should be noted that there are different approaches to discerning brain functional hubs and brain hubs may vary depending on the approach by which hubs are defined ( ). We did not find a difference in rich-club coefficients between groups.
The connections unique to the opioid-exposed group were more often inter-network connections, while the connections that were unique to controls and those that overlapped were fairly evenly distributed between inter- and intra-network connections. Previous literature has shown that in networks that serve basic brain functions (auditory, visual, sensorimotor), intra-network connections increase with age, to increase functional specialization of those networks ( ). In the higher order networks, inter-network connections increase with age, due to the multimodal nature of these higher order functions ( , ; ). Infants with prenatal opioid exposure who are experiencing withdrawal symptoms tend to be “disorganized” soon after birth and need decreased auditory and visual stimulation. We hypothesize that the increased inter-network connections we see in the connections unique to this group, even in the basic sensory networks, reflect this outward disorganization. We found that the highest Finnegan score (a measure of withdrawal) was in fact correlated with connectivity, especially with those among subcortical brain regions, although the association did not survive correction for multiple comparisons. Further work is needed to investigate the relationship between differences in network organization and behavior.
The connections that were unique to the opioid-exposed group and the connections unique to controls were longer than those that overlapped between groups. Connection length tends to increase with age and myelination ( ). Perinatal opioid exposure is known to adversely impact neurogenesis, programmed neuronal death, and myelination in murine models. Therefore, we hypothesize that the long-distance and inter-network connections are more vulnerable to prenatal opioid exposure, leading to differences between the groups.
Using seed-based analysis, we found that although mean functional connectivity strengths over the networks evaluated were not different between the opioid-exposed and control groups, the network volumes and topologies were different in most networks. The difference between network volumes and topology was particularly striking in the V1 network, with opioid-exposed children having markedly decreased network volume compared to controls. Children with prenatal opioid exposure are known to have a higher risk of visual problems, including reduced visual acuity, strabismus, and nystagmus ( ; ; ; ). After prenatal exposure to methadone, infants show abnormal, smaller, or slower visual evoked potentials relative to controls ( ). Vision is a key function for later cognitive and behavioral development and may provide an early measure of brain functional integrity ( ). Although the neural substrate for the visual abnormalities seen with prenatal opioid exposure is unknown, opioid receptors are present in the cortex, in the lateral geniculate nucleus in the thalamus (visual relay center), and midbrain (responsible for some aspects of eye movements) ( ). Our findings suggest that differences in the primary visual network could also potentially explain visual problems in children with prenatal opioid exposure.
Strengths of our study include rigorous imaging at a defined early time point. Groups were well-defined with results of maternal urine toxicology at the time of delivery for both opioid-exposed newborns and controls, and umbilical cord or meconium toxicology screens for all opioid-exposed infants. We were able to assess a brain-behavior connection by correlating Finnegan scores with connectivity.
Several limitations of our study should be considered. First, the sample size was small, and results should be validated in studies with larger sample sizes. We were unable to control for confounders including maternal smoking, maternal Hepatitis C, and maternal education, all of which have potential effects on brain development ( ; ; ). Due to the small sample size, it was not possible to directly compare infants exposed to opioids in utero with and without these additional risk factors. This will be important in future studies. We chose to focus our analyses on positive connectivity, which may be seen as a limitation. To address this, we conducted additional analyses where absolute correlation coefficients were used, with results provided in the Supporting materials (A-3: Positive vs absolute correlation coefficients). Finally, although we are following this cohort to 2 years of age, in this study we were unable to evaluate neurodevelopmental outcomes.
## Conclusions
Prenatal opioid exposure disrupts the architecture of brain networks. Future research using larger sample sizes and a longitudinal design, such as the upcoming HEALthy Brain and Child Development study, are required to understand how the type, timing, and duration of opioid exposure, as well as other confounding factors, affect the developing brain and subsequent neurodevelopment.
## Funding
This work was supported by , part of NIH via R34 DA050268 (SM, JN, BK, JT) and R34 DA 050262 (WL).
## Data statement
The datasets generated during the current study are available from the corresponding author on reasonable request.
## CRediT authorship contribution statement
Stephanie L. Merhar: Conceptualization, Investigation, Funding acquisition, Project administration, Writing - original draft. Weixiong Jiang: Methodology, Visualization, Writing - review & editing. Nehal A. Parikh: Methodology, Resources, Supervision, Writing - review & editing. Weiyan Yin: Methodology, Writing - review & editing. Zhen Zhou: Formal analysis, Writing - review & editing. Jean A. Tkach: Data curation, Investigation, Methodology, Writing - review & editing. Li Wang: Methodology, Writing - review & editing. Beth M. Kline-Fath: Investigation, Methodology, Writing - review & editing. Lili He: Data curation, Funding acquisition, Writing - review & editing. Adebayo Braimah: Data curation, Formal analysis, Writing - review & editing. Jennifer Vannest: Conceptualization, Funding acquisition, Writing - original draft. Weili Lin: Conceptualization, Funding acquisition, Methodology, Supervision, Writing - review & editing.
## Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
|
Impairments in social relationships and awareness are features observed in autism spectrum disorders (ASDs). However, the underlying mechanisms remain poorly understood. Shank2 is a high-confidence ASD candidate gene and localizes primarily to postsynaptic densities (PSDs) of excitatory synapses in the central nervous system (CNS). We show here that loss of Shank2 in mice leads to a lack of social attachment and bonding behavior towards pubs independent of hormonal, cognitive, or sensitive deficits. Shank2<sup>-/-</sup> mice display functional changes in nuclei of the social attachment circuit that were most prominent in the medial preoptic area (MPOA) of the hypothalamus. Selective enhancement of MPOA activity by DREADD technology re-established social bonding behavior in Shank2<sup>-/-</sup> mice, providing evidence that the identified circuit might be crucial for explaining how social deficits in ASD can arise. |
## Objective
A spectrum of seizure disorders is linked to mutations in Kv7.2 and Kv7.3 channels. Linking functional effects of identified mutations to their clinical presentation requires ongoing characterization of newly identified variants. In this study, we identified and functionally characterized a previously unreported mutation in the selectivity filter of Kv7.3.
## Methods
Next‐generation sequencing was used to identify the Kv7.3[T313I] mutation in a family affected by neonatal seizures. Electrophysiological approaches were used to characterize the functional effects of this mutation on ion channels expressed in Xenopus laevis oocytes.
## Results
Substitution of residue 313 from threonine to isoleucine (Kv7.3[T313I]) likely disrupts a critical intersubunit hydrogen bond. Characterization of the mutation in homomeric Kv7.3 channels demonstrated a total loss of channel function. Assembly in heteromeric channels (with Kv7.2) leads to modest suppression of total current when expressed in Xenopus laevis oocytes. Using a Kv7 activator with distinct effects on homomeric Kv7.2 vs heteromeric Kv7.2/Kv7.3 channels, we demonstrated that assembly of Kv7.2 and Kv7.3[T313I] generates functional channels.
## Significance
Biophysical and clinical effects of the T313I mutation are consistent with Kv7.3 mutations previously identified in cases of pharmacoresponsive self‐limiting neonatal epilepsy. These findings expand our description of functionally characterized Kv7 channel variants and report new methods to distinguish molecular mechanisms of channel mutations.
Key Points
The Kv7.3[T313I] mutation was identified in a family with heritable pharmacoresponsive self‐limiting neonatal seizures
Functional characterization revealed that Kv7.3[T313I] mutant subunits assemble with Kv.7.2 and suppress currents, but have little effect on gating parameters
These findings validate Kv7.3[T313I] as a pathogenic mutation in neonatal epilepsy
## INTRODUCTION
Mutations of the KCNQ2 and KCNQ3 voltage‐gated potassium channel genes have been implicated in the pathophysiology of a spectrum of seizure disorders, ranging from pharmacoresponsive self‐limiting neonatal epilepsy to severe epileptic encephalopathy. , , , , , , The milder end of this spectrum has been historically referred to as BFNE (Benign Familial Neonatal Epilepsy) and is characterized by seizures that begin in the first few postnatal days, which are typically pharmacoresponsive and self‐limiting within a few months and rarely to a few years, and do not interfere with structural or cognitive development. , The term BFNE has been discouraged in recent position statements because the term “Benign” may minimize the impact of the disease on patients and caregivers. One issue often encountered clinically is how to best predict long‐term outcomes and approach management for a neonate with new‐onset seizures. Increased access to genetic sequencing has allowed clinicians to rapidly identify genetic forms of epilepsy, but there is often still insufficient information for an accurate prediction of phenotype from genotype, particularly for mutations in KCNQ2 and KCNQ3, which can present with a wide range of epileptic severity and may have overlapping features. ,
KCNQ2 and KCNQ3 gene products, Kv7.2 and Kv7.3, form heterotetrameric potassium channels that encode the neuronal M‐current, which plays a modulatory role on threshold properties and repetitive burst firing in response to excitatory stimuli. , , , Heteromeric assembly of Kv7.2 and Kv7.3 subunits enhances expression of channels in heterologous systems by roughly 10‐fold relative to homomeric channels of either subtype alone, and so dysfunction in either subunit can be linked to a pathological outcome. In general, Kv7.2 is more frequently implicated in both self‐limiting seizures and epileptic encephalopathy phenotypes. Pathogenic variants are less frequently identified in Kv7.3.
Kv7.3 mutations have been reported throughout the channel sequence, including C‐terminal domains associated with calmodulin or PIP2 association and the voltage‐sensing domain. This is highlighted in Figure , illustrating the location of pathogenic point mutations of Kv7.3 that are reported in clinVar, on the structural template of a recently reported Kv7.1 cryo‐EM structure. There is a cluster of validated disease‐linked mutations in the Kv7.3 selectivity filter region, which underlies ion selectivity and permeation. , , , , , The geometry and stability of the selectivity filter are determined by several intra‐ and intersubunit interactions that are widely conserved among K channels. , , In some channel types prone to selectivity filter mediated inactivation, the disruption of these bonds leads to pronounced alterations of the kinetics of inactivation or a total loss of channel conductance. , , , In contrast, M‐channels exhibit no apparent inactivation, and selectivity filter mutations can lead to a range of effects on overall channel function. , ,
Inheritance of Kv7.2 and Kv7.3 mutations associated with epilepsy. A, Mutations in Kv7.2 or Kv7.3 associated with a documented case of epilepsy (compiled from ClinVar or RIKEE databases) are highlighted on molecular models of each channel. Mutations are color coded based on severity (green = BFNE, red = epileptic encephalopathy or other severe outcomes). Mutations that do not map to structural elements defined in the KCNQ1 cryo‐EM structure have been omitted (VSD, voltage‐sensing domain, SF, selectivity filter, CaM, Calmodulin). B, Pattern of inheritance of a neonatal seizure phenotype in a family carrying the Kv7.3[T313I] mutation. Upper, sequence alignment of the reference KCNQ3 gene and Kv7.3 protein in relation to the proband. The identified mutation [T313I] is highlighted in bold type. Lower, pedigree for the family characterized in our study with filled symbols indicating affected individuals
In this study, we report a family with a pattern of inheritance and symptoms indicating a diagnosis of pharmacoresponsive self‐limiting seizures (BFNE). The proband was identified to have a KCNQ3 variant, p.Thr313Ile, which is a previously unreported selectivity filter mutation in Kv7.3. A mutation has been reported in Kv7.2 channels at a homologous position (T274M), associated with a severe clinical outcome including profound global developmental delay, motor dysfunction, and remitting pharmacoresistant seizures. , , , We undertook the characterization of the electrophysiological effects of the T313I mutation in Kv7.3 homomeric and Kv7.2/Kv7.3 heteromeric channels, to describe in more detail the potential effects of selectivity filter mutations in this channel family. We also developed an assay using a subtype‐specific Kv7.2 activator (ICA‐069673) as a “fingerprinting” tool to distinguish functional Kv7.2/Kv7.3 heteromeric channels from Kv7.2 homomers. , , Our findings highlight variable outcomes of mutations in this critical selectivity filter position on the overall function of Kv7.2/Kv7.3 heteromeric channels. We find that Kv7.2/Kv7.3 heteromers are relatively tolerant of the Kv7.3[T313I] mutation, and this may underlie the mild disease phenotype, in contrast to severe outcomes arising from Kv7.2[T274M].
## MATERIALS AND METHODS
### Molecular biology and plasmid construction
The KCNQ3[T313I][A315T] and KCNQ3[T313I] mutation were generated by site‐directed mutagenesis in pSRC5 (gift from Dr. M. Taglialatela, University of Naples, Italy, and Dr. T. Jentsch, Max‐Delbrück‐Centrum für Molekulare Medizin, Germany) and pTLN plasmids encoding Kv7.3.
Primers used for mutagenesis for the KCNQ3[T313I] construct were as follows (together with their reverse complement primer):
KCNQ3[T313I]: 5′‐GCCTGATCATACTGGCCACCATTGGC‐3′
KCNQ3[T313I]: 5′‐GCCTGATCATGCTGGCCACCATTGGC‐3′
KCNQ3[A315T][T131I]: 5′‐GCCTGATCATACTGACCACCATTGGC‐3′
KCNQ2[T274I]: 5′‐GGGGCCTGATCATCCTGACCACCATTG‐3′
KCNQ2[T274M]: 5′‐GGGGCCTGATCATGCTGACCACCATTG‐3′
### Two‐electrode voltage‐clamp expression and recording
Complementary RNA was transcribed from cDNA to express the monomeric KCNQ3 constructs in Xenopus laevis oocytes using the mMessage mMachine kit (Ambion). Constructs expressed in pSRC5 (KCNQ3[T313I][A315T] or (KCNQ3[A315T])) were linearized with ApaL1 and transcribed using a T7 primer. Constructs in pTLN (KCNQ2, KCNQ3, and KCNQ3[T313I]) were linearized with HpaI and transcribed using an SP6 primer. Oocyte preparation and RNA injection were performed as described previously, under the approval of the University of Alberta Animal Care protocol AUP00001752. After injection, oocytes were incubated for 48 hours at 18°C before electrophysiological recording. Voltage‐clamp recordings were obtained in modified Ringer’s solution (in mmol/L): 116 NaCl, 2 KCl, 1 MgCl2, 0.5 CaCl2, and 5 HEPES, pH adjusted to 7.4 with NaOH, using an OC‐725C voltage clamp (Warner). Microelectrodes were backfilled with 3 mol/L KCl to obtain resistance between 0.1 and 1 MΩ. Data were processed using a Digidata 1440A acquisition system controlled by pClamp 10 software. For experiments using ICA‐069673, oocytes were incubated in 100 μmol/L ICA‐069673 in modified Ringer's solution for 3‐4 minutes before recording.
### Data analysis and statistics
Statistical tests used throughout the manuscript are described in the corresponding figure legends. Gating parameters describing voltage dependence of channel activation were determined by fitting with a standard single‐component Boltzmann equation of the form: where V is the voltage where channels exhibit half‐maximal activation, and k is a slope factor reflecting the voltage range over which an e ‐fold change in open probability (Po) is observed. The extent of tail current deactivation in the presence of ICA‐069673 was inferred by the ratio of the magnitude of instantaneous: peak current after the repolarization interval, rather than fitting tail current kinetics directly (these are often too slow to generate a meaningful fit).
### DNA sequencing
Patient samples were sequenced clinically by an external provider (CeGAT GmbH, Tubingen, Germany). The proband was initially investigated using a metabolic/mitochondrial epilepsy panel in order to rule out potentially treatable metabolic causes. Given the family history, some individual genes associated with pharmacoresponsive self‐limiting seizures were also included in addition to the standard panel. The genes examined as part of the standard metabolic epilepsy clinical panel were as follows: AARS2, ABAT, ABCC8, ACY1, ADCK3, ADK, ADSL, ALDH5A1, ALDH7A1, AMT, ATIC, AUH, BCKDHA, BCKDHB, BCKDK, BCS1L, BTD, C10ORF2, CAD, CARS2, CNNM2, COQ4, COX8A, CPT1A, CPT2, D2HGDH, DARS2, DBT, DHFR, DLD, DNM1L, DPYD, EARS2, ETFA, ETFB, ETFDH, ETHE1, FARS2, FOLR1, FOXRED1, GAMT, GATM, GCDH, GCH1, GCK, GCSH, GFM1, GLDC, GLUD1, GLUL, GPHN, HADH, HLCS, HPD, 1DH2, INSR, ITPA, IVD, KCNJ11, L2HGDH, LIAS, MDH2, MLYCD, MMACHC, MOCS1, MOCS2, MT‐ATP6, MT‐TK, MT‐TL1, MTHFR, NARS2, NDUFA1, PC, PCBD1, PCCA, PCCB, PDHA1, PDHX, PDSS2, PET100, PHGDH, PNPO, POLG, PROSC, PSAT1, PSPH, PTS, QDPR, SDHA, SLC16A1, SLC19A3, SLC1A2, SLC25A1, SLC2A1, SLC46A1, SLC6A8, SLC6A9, SUOX, SURF1, VARS2. In addition, KCNQ2, KCNQ3, and PRRT2 sequences were added for analysis based on a suspicion of a BFNE diagnosis. The T313I mutation was identified in the proband, and in silico predictors suggested the variant was damaging. The T313I variant was classified as a VUS because this position had not been previously associated with disease and was not present in the EVS, gnomAD, or RIKEE databases. Targeted testing was then used to determine the presence of the variant in other family members. Targeted testing confirmed the presence of the variant in the father and paternal grandmother who also have a history of early seizures and absence of the variant in the mother who has no history of seizure.
## RESULTS
### Proband features
The proband was a term infant female born after an unremarkable pregnancy and uncomplicated vaginal delivery. At 4 days of age, she began having spells described as “not breathing properly” while “holding her arms and legs out (flexed or extended) with a mild tremor and at times hyperextension of her back”. The events increased in frequency, beginning to cluster, and she was admitted to the hospital for workup. Seizures were suspected and she was loaded with 20 mg/kg of IV levetiracetam and continued with 10 mg/kg bid of maintenance. Video‐EEG monitoring was initiated. Six discrete, self‐limited electroclinical seizures were recorded in sleep and wakefulness in the first 24 hours of recording. The seizures lasted between 43 and 69 seconds (mean 54 seconds) and while the semiology varied slightly, the typical pattern entailed three clearly discernible stages: (a) wide eye opening, with occasional gaze deviation to the left, (b) tonic posturing of limbs and torso (typically left arm extension followed by arching of back and flection or extension of legs), and (c) clonic movements of bilateral arms and legs until offset. The ictal EEG correlate consisted of symmetric background attenuation with overriding low amplitude fast beta frequencies for ~15 seconds (tonic phase) followed by bilateral high amplitude spike‐sharp delta‐theta frequencies in a crescendo‐decrescendo manner (clonic phase) until offset. Interictally, the EEG showed bilateral symmetric and continuous background activity with mixture of delta‐theta frequencies and rare (<1/min) non‐specific negative sharp waves with independent maximum amplitude over T7, T8, C3, C4, or Cz. Normal wake‐sleep cycling was also seen.
A neurological examination revealed a normal non‐dysmorphic infant without focal neurological deficits. Brain MRI showed normal anatomy. On review of the pregnancy history, the patient’s mother was healthy, with a history of 3 previous early pregnancy losses. There was hyperemesis in the first trimester but pregnancy was otherwise unremarkable with no drug or alcohol exposures, and the patient was delivered at 40 weeks gestation. On review of the family history, the infant’s father had a history of seizures in early infancy (9‐12 months) that were self‐limiting. He also has attention deficit hyperactivity disorder but normal cognition. The patient’s paternal grandmother had seizures in infancy, which were also self‐limiting and normal cognition. There is also a paternal cousin with epilepsy. There was no history of seizure in the patient’s mother or maternal relatives (Figure ).
With the continuation of her seizures, the maintenance dose of IV levetiracetam was increased to 15 mg/kg bid without effect. She was then loaded with 10 mg/kg of IV phenobarbital and the seizures stopped. Given the efficacy of phenobarbital, her levetiracetam was discontinued and she was kept on maintenance phenobarbital (2.5 mg/kg bid). Due to the initial uncertainty of the etiology of this patient’s seizures, effective control only after initiating phenobarbital, and that the patient’s father had sporadic seizures in infancy (9–12 months), a maintenance dose was kept for safety until a clear etiology was found. The patient remained event‐free and a repeat EEG at 6 months of age was normal. With the normal EEG, lack of clinical events, and normal development, phenobarbital was weaned at 8 months of age. Seizures had remained after levetiracetam initiation; therefore, seizure cessation was suspected to be due to phenobarbital effects. Nevertheless, one cannot rule out spontaneous seizure remission. Seizures did not relapse despite being off anti‐seizure therapy, and she remains seizure‐free for more than 3 years. Development also remained normal in all domains, suggesting a clinical presentation consistent with self‐limiting pharmacoresponsive seizures.
Metabolic investigations for this patient were unremarkable. Clinical genetic testing had been initiated for this patient at seizure onset. Array CGH (60K) was negative. A 114 gene epilepsy panel through the clinical commercial laboratory CeGAT (see Materials and Methods) identified a variant of uncertain significance in KCNQ3, c.938C>T, p.Thr313Ile (NM_004519) (Figure ). This variant was not identified in the patient’s mother, but was present in the patient’s father and paternal grandmother (Figure ). This variant is absent from gnomAD and was not previously reported in ClinVar, EVS, or RIKEE databases. ,
### KCNQ3[T313I] mutation eliminates homomeric Kv7.3 function
We were intrigued by both the biophysical and physiological consequences of the T313I mutation. The T313 position is one of several positions identified in patients with self‐limiting neonatal seizures, clustered in the region of the selectivity filter (Figure ). Due to the critical functional importance of this region, variants in the vicinity of the selectivity filter are often linked to disease (D305, A306, W308, W309, G310, I317, Y319, G320 are listed in ClinVar and linked to BFNE). Despite the essential role of the selectivity filter, the pathology arising from these mutants is relatively mild and self‐resolving. A previous study of an analogous mutation in the widely studied Shaker potassium channel, a model system for understanding many basic principles of voltage‐dependent ion channel function, highlighted the importance of this particular site in maintaining the stability of the selectivity filter. Kv7.3 position T313 is analogous to Shaker residue T439, which is predicted to form an intersubunit hydrogen bond with the selectivity filter residue Y445 (G Y G), highlighted in Figure (2 of 4 potential H‐bonds are shown). In Shaker channels, disruption of this H‐bond in even a single subunit caused a profound suppression of overall channel current and acceleration of inactivation by ~100‐fold. These powerful effects were attributed to the intersubunit nature of the interaction, leading to a propagated effect of a mutation from a single subunit (Figure ). Similarly, the Kv7.2[T274M] mutation of an analogous position in Kv7.2 has been associated with severe epileptic encephalopathy and dominant negative effects when co‐expressed with Kv7.3. However, the physiological outcomes in this family carrying the Kv7.3[T313I] mutation were much less severe. We initially characterized the effects of the T313I mutation in homomeric Kv7.3, using the A315T mutation to enable expression of functional homotetramers. , Consistent with prior studies in Shaker channels, we observed that the T313I mutation virtually abolished functional current relative to Kv7.3[A315T] channels (Figure ).
Kv7.3[T313I] abolishes Kv7.3 function. A, Molecular model of essential intersubunit hydrogen bond between conserved residues T313 and Y319 located in the selectivity signature sequence. B, Two‐electrode voltage‐clamp recordings from Xenopus laevis oocytes expressing Kv7.3[A315T] and Kv7.3[A315T][T313I]. Oocytes were held at −80 mV and depolarized for 1.5s to voltages between −140 mV and +40 mV (in 10‐mV steps) followed by repolarization to −20 mV test pulse. Current amplitudes at +20 mV of Kv7.3 [A315T] and Kv7.3[A315T] [T313I] were compared using a student’s t test (* indicates P <0.05 relative to A315T alone)
### Kv7.3[T313I] attenuates currents in Kv7.2/Kv7.3 heteromers
Although this profound effect of Kv7.3[T313I] was consistent with previous reports in Shaker channels and Kv7.2, it does not address the effects of the mutation in heteromeric Kv7.2/Kv7.3 channels that predominate in vivo . , To investigate the effects of the T313I mutation in heteromeric channels, we co‐expressed different ratios of Kv7.2, Kv7.3, and Kv7.3[T313I] to simulate different heteromeric assemblies that may occur in vivo . Wild‐type Kv7.2/Kv7.3 heteromeric channels (1:1 mRNA injection ratio) generated larger currents relative to homomeric Kv7.2 channels. When Kv7.2 and Kv7.3[T313I] subunits were injected in a 1:1 ratio, the current was reduced significantly relative to the wild‐type heteromer (Figure ).
Co‐expression with Kv7.2 and Kv7.3[T313I] reduces heteromeric channel function with no effect on gating. A, Two‐electrode voltage‐clamp sample traces from oocytes expressing various combinations of Kv7.2 and Kv7.3 (injected with a total of 50 ng of mRNA per group). The voltage step protocol is the same as Figure . B, Current amplitudes after a 1.5 s + 20 mV voltage step. Current magnitudes were compared using one‐way ANOVA, followed by Tukey’s post hoc test (* indicates P < 0.05). C, Conductance‐voltage relationships were collected using the protocol in panel (A), using tail current magnitudes (−20 mV) to assess the extent of channel opening during the conditioning step. Fitted gating parameters were (mean ± S.E.M.): for Kv7.2 + Kv7.3 +[T313I], k = 9.4 ± 0.3 mV, V = −30 ± 1 mV; for Kv7.2 + [T313I], k = 9.1 ± 0.3 mV, V = −35.1 ± 0.5 mV; for Kv7.2 + Kv7.3, k = 10.2 ± 0.3 mV, V = −34.2 ± 0.3 mV; for Kv7.2 homomers, k = 9.2 ± 0.2 mV, V = −34.2 ± 0.6 mV. No significant differences in gating parameters were detected
We also tested the mixed expression of WT Kv7.3 and Kv7.3[T313I] channels with Kv7.2, to mimic the heterozygous genotype of the proband (Figure ). A 1:0.5:0.5 (Kv7.2:Kv7.3:Kv7.3[T313I]) RNA injection ratio did not suppress current magnitude relative to WT Kv7.2/Kv7.3 (Figure ). However, current magnitudes in these experiments are variable and there may be assembly of various channel stoichiometries even with fixed RNA ratios. No difference in voltage‐dependent activation was detected in oocytes expressing different ratios of Kv7.2, Kv7.3, and Kv7.3[T313I] (Figure ). Although these experiments illustrate modest current suppression by Kv7.3[T313I] in certain subunit combinations, they do not fully reveal the nature of Kv7.3[T313I] mutant subunits on function of heteromeric channels, particularly in light of the severe phenotype arising from the Kv7.2[T274M] mutation. That is, it was not clear whether currents generated from co‐expression of Kv7.3[T313I] and Kv7.2 were generated by heteromeric channels, or from Kv7.2 homomers that did not assemble with Kv7.3[T313I].
### Heteromeric composition determined by ICA‐069673 sensitivity
We recently recognized that the functional effects of the Kv7 activator ICA‐069673 depend strongly on subunit composition. Kv7.2 homomeric channels exhibit profound deceleration of deactivation in the presence of ICA‐069673, whereas these effects are much weaker in homomeric Kv7.2/Kv7.3 channels (due to weak/absent ICA‐069673 sensitivity of Kv7.3). We used this pharmacological “fingerprint” to investigate whether currents observed in the Kv7.2/Kv7.3[T313I] condition were generated by Kv7.2/Kv7.3[T313I] heteromers or only arise from Kv7.2 homomers (due to strong current suppression by Kv7.3[T313I]).
Using oocytes continuously perfused with a saturating concentration of 100 μmol/L ICA‐069673, we delivered a voltage step protocol that depolarized channels to +20 mV for 1.5 seconds, followed by repolarization to a range of voltages for 12 seconds (Figure ). This was followed by a second step to +20 mV to assess the extent of channel closure during the repolarization interval, based on the instantaneous vs activating current fractions elicited by the voltage step (expanded time scales of the second depolarizing step are illustrated in the right hand panels of Figure ). Kv7.2 homomeric channels in 100 μmol/L ICA‐069673 are characterized by extremely slow channel closure, resulting in a large instantaneous current fraction, even after repolarization at very negative voltages. In contrast, heteromeric Kv7.2/Kv7.3 channels exhibit much more prominent deactivation during this interval, which results in smaller instantaneous currents (Figure ). In 100 μmol/L ICA‐069673, Kv7.2 homomeric channels exhibited ~70% instantaneous current after a 12 second repolarization at −120 mV, whereas heteromeric channels generated <20% instantaneous current (Figure ). Co‐expression of 1:1 Kv7.2:Kv7.3[T313I] channels exhibited an intermediate behavior, but was distinct from WT Kv7.2 homomers. This observation suggests that heteromeric Kv7.2:Kv7.3[T313I] channels retain function, but there is also likely a contribution of homomeric Kv7.2 channels (Figure ). Co‐expression to mimic the heterozygous genotype (1:0.5:0.5, Kv7.2:Kv7.3: Kv7.3[T313I]) displayed a similar ICA‐069673 response to the wild‐type heterozygous control (Figure ).
Reduced ICA‐069673 sensitivity of Kv7.2/Kv7.3[T313I] heteromeric channels. A, Example currents of two‐electrode voltage‐clamp recordings. Oocytes were depolarized to +20 mV and repolarized for 12 s in a step‐down manner (−20 mV per sweep), followed by another +20 mV depolarizing pulse to determine instantaneous current at −100 mV. Currents within the dashed box are illustrated on an expanded time scale in the right panels, showing the assessment of instantaneous current levels in different experimental conditions. B, Fractional instantaneous current after incubation with 100 μmol/L ICA‐069673 was measured as indicated by the arrows in panel (A), data are shown as mean ± SEM. C, Fractional instantaneous current (repolarization voltage of −100 mV) for various combinations of Kv7.2, Kv7.3, and Kv7.3[T313I], shown on a cell‐by‐cell basis (Kruskal‐Wallis one‐way ANOVA on ranks, * indicates P < 0.05 relative to Kv7.2 homomeric channels)
### Distinct effects of Kv7.2 and Kv7.3 mutations
In order to investigate the functional basis for different clinical outcomes of Kv7.2 and Kv7.3 mutations reported at analogous positions, we also tested the effects of the Kv7.3[T313M] mutation (equivalent to Kv7.2[T274M]) and the Kv7.2[T274I/M] mutations (Figure ). We measured current amplitude (Figure ) and also used the ICA‐069673 response described in Figure to confirm assembly of Kv7.2/Kv7.3 heteromers (Figure ). Among these mutant channels, the Kv7.3[T313I] mutant stood out for its comparably mild effects on current magnitude. Co‐expression of Kv7.3[T313M] with Kv7.2 led to a significantly stronger suppression of current relative to Kv7.3[T313I]. Furthermore, co‐expression of Kv7.2[T274M] or Kv7.2[T274I] led to even further suppression of heteromeric Kv7.2/Kv7.3 channel currents. For most of these subunit combinations, we could detect currents resembling Kv7.2/Kv7.3 heteromeric channels based on their weak sensitivity to ICA‐069673 (Figure ), although the Kv7.2[T274I] + Kv7.3 combination did not generate sufficient currents in any oocytes to confidently assess any properties of the elicited currents. Overall, these findings confirm that the various Kv7.2 and Kv7.3 mutants can assemble with their WT partner in heteromeric channels, but with very markedly different outcomes on overall channel function.
Variable outcomes of homologous selectivity filter mutations in Kv7.2 and Kv7.3. A, Current magnitude was measured at +20 mV in Xenopus oocytes injected with Kv7.2 + Kv7.3, or various combinations of channel mutants and their wild‐type counterpart, as indicated. Currents were recording 48‐56 h after injection (n = 10 per condition, one‐way ANOVA on ranks and Tukey post hoc test, # indicates P < 0.05 relative to Kv7.2/Kv7.3, * indicates P < 0.05 relative to Kv7.2/Kv7.3[T313I]). B, Instantaneous tail current magnitudes were recorded as described in Figure in the presence of 100 μmol/L ICA‐069673, for all possible combinations of injected channels (note that Kv7.3 + Kv7.2[T274I] did not yield current sizes that could be confidently assessed or analyzed)
## DISCUSSION
Determining genotype‐phenotype correlation of mutations in Kv7.2 and Kv7.3 is of significant interest in understanding the prognosis and pathogenesis of affected patients. In addition, characterization of variants is valuable for understanding assembly, function, and regulation of Kv7.2/Kv7.3 heteromeric channels, which play a significant modulatory role in the central nervous system. Delineation of how variants impact channels also informs development of targeted drug therapeutics. In this study, we report a novel mutation in the selectivity filter region of Kv7.3 channels, predicted to form an intersubunit interaction that has been previously shown to strongly disrupt the function of Shaker channels, and M‐channels when mutated in Kv7.2. , The T313I mutation in Kv7.3 was identified in a family presenting with neonatal self‐resolving pharmacoresponsive epilepsy (frequently referred to as BFNE) and was not previously reported in gnomAD, clinVar, or the KCNQ disease‐specific RIKEE databases. Although this variant was initially classified as a variant of uncertain significance, we feel our findings classify this mutation as pathogenic given its impact on channel function (Figure ), and due to segregation of the variant with disease in the family.
Previous work in Shaker channels demonstrated the contribution of this position to channel function, based on formation of a H‐bond between T439 (equivalent to Kv7.3[T313I]) and Y445 (in the GYG selectivity filter sequence of an adjacent subunit). An important comparison was drawn to a commonly used non‐conducting mutant of Shaker (W434F), illustrating much greater tolerance of the W434F relative to either T439V or Y445F mutations, which both cause a dramatic loss of function even when expressed in a single channel subunit. , , This difference was attributed to the intersubunit nature of the T439 interaction. This previous finding of an essential structural role for this amino acid position motivated us to investigate the generality of this finding and the differing clinical outcomes in patients with mutations at this position in Kv7.3 in our study, versus previously described mutations at this position in Kv7.2 with severe clinical outcomes.
The overall importance of position T313 (and its likely H‐bond interaction with Y319) is highlighted by the observation that Kv7.3[T313I] subunits cannot form functional homomeric channels, even using a high‐expressing Kv7.3[A315T] background channel (Figure ). However, the effects of the T313I mutation in heteromeric channels (co‐expressed with Kv7.2) are much less pronounced (Figure ). Modestly smaller current magnitudes are observed in the Xenopus oocyte system used for this study, and no significant differences in gating properties were observed between Kv7.2/Kv7.3[T313I] heteromeric channels and Kv7.2/Kv7.3 wild‐type channels (Figure ). This observation is a departure from the powerful effects of disruption of intersubunit selectivity filter interactions in Shaker channels and likely highlights that other intersubunit interactions contribute significantly to stabilizing the selectivity filter of Kv7 channels.
Our findings also indicate variable outcomes of mutations at this position, depending on the amino acid substitution and mutated Kv7 subtype. For example, a Kv7.2 mutation at the analogous position (Kv7.2[T274M]) is linked to severe epileptic encephalopathy and global delay. , , , Functional characterization of this mutation in Xenopus oocytes results in a more pronounced effect on current suppression of heteromeric channels, relative to Kv7.3[T313I] (Figure ). This difference in functional effects in heteromeric channels correlates with the severity of clinical outcomes. We investigated this difference in more detail by comparing equivalent mutations in both Kv7.2 and Kv7.3 and demonstrated non‐equivalent effects of these mutations in these different subtypes. Overall, mutations in Kv7.3 (T313I or T313M) were less severe than their comparators in Kv7.2 (Figure ). In a broad context, this finding is consistent with reports that neonatal epilepsy (including severe epileptic encephalopathy) is far more commonly reported to arise from Kv7.2 mutations as opposed to Kv7.3 (Figure ). , , However, in terms of biophysical effects on channel function, there is not an obvious explanation for these diverging effects of Kv7.2 and Kv7.3. In addition to this apparent subtype‐dependent effect of certain mutations, other differences in physiological regulation of Kv7.2 and Kv7.3 likely contribute to the general differences in severity of diseases between these subtypes. For instance, Kv7.2 expression is more prominent than Kv7.3 in early development and infancy, , and these subtypes may contribute differently to channels with alternative stoichiometries (perhaps Kv7.2 homomeric channels, or other assemblies). , , , Another possibility is more penetrant effects of Kv7.2 versus Kv7.3 mutations on channel function may translate into stronger effects on structural and electrical remodeling of the axon initial segment, as has been observed in the presence of Kv7 inhibitors.
Lastly, we hope to highlight the use of subtype‐selective Kv7 activators such as ICA‐069673 for investigating the assembly of heteromeric channels and inferring the functional impact of disease‐linked mutations in heteromeric channels. Since ICA‐069673 is strongly selective for Kv7.2 over Kv7.3 subunits, Kv7.2/Kv7.3 heteromeric channels have a distinct response from Kv7.2 homotetramers (Figure ). This property reveals the impact of mutated versions of Kv7.3 on heteromeric channel function and clearly distinguishes homomeric Kv7.2 channels versus heteromeric assembly of Kv7.2/Kv7.3. Another useful approach used previously has been to exploit the differential sensitivity of Kv7.2 and Kv7.3 to extracellular tetraethylammonium. , , Our alternative approach using a more specific pharmacological agent is useful for resolution of distinct Kv7.2/Kv7.3 stoichiometries. Our findings also further validate prior reports of the stoichiometry‐dependent effects of ICA‐069673 and related compounds.
In summary, our study reports a novel variant in the Kv7.3 selectivity filter, linked to BFNE. The biophysical consequences of this mutation primarily involve suppression of function, without pronounced effects on voltage‐dependent gating. The assembly of functional channels comprising Kv7.2 and Kv7.3[T313I] subunits is unambiguously demonstrated based on sensitivity to a Kv7.2‐specific activator compound. Interestingly, the biophysical consequences of the Kv7.3[T313I] are less severe than reported in the prototypical Shaker potassium channel.
## CONFLICT OF INTEREST
None of the authors have any conflicts of interest to disclose. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.
## AUTHOR CONTRIBUTIONS
J. Maghera, J. Li, and S.M. Lamothe collected experimental data, performed molecular biology, and analyzed data. M. Braun, J.P. Appendino, and BPY Au collected clinical data and interacted with patients. All authors approved the final version of the manuscript.
|
Ictal vomiting is considered a localizing sign indicating nondominant lateralization in patients with partial seizures of temporal lobe origin. We report a case of ictal vomiting associated with left temporal seizure activity in a left hemisphere language-dominant patient with a left mesial temporal glioma. Bilateral mesial temporal depth electrodes helped verify seizure lateralization. Surgery consisting of tumor resection and a left anterior temporal lobectomy and amygdalohippocampectomy resulted in freedom from seizures and episodes of vomiting. This case indicates that ictal vomiting can occur as a manifestation of left temporal onset seizures in left hemisphere-dominant patients. |
If lamotrigine (LTG) has to be replaced with valproate (VPA), this exchange may be complicated by adverse events that result from the complex interaction of both drugs. We report on two cases in which such problems occurred in spite of a cautious switch considering the VPA induced LTG serum increase. The satisfying outcome after a sudden and complete withdrawal of LTG in both cases encouraged us to perform the switch from LTG to VPA systematically by discontinuing LTG abruptly and building up the VPA maintenance dosage very rapidly in the following five consecutive patients who required this exchange. We recommend our abrupt dosage change-over strategy as an easy, safe and cost-effective option. |
## Objective
The Patient Information Leaflet (PIL) is an authoritative document that all people with epilepsy in the EU receive when prescribed antiseizure medication (ASM). We undertook the first independent, comprehensive assessment to determine how understandable they are. Regulators state that when patients are asked comprehension questions about them, ≥80% should answer correctly. Also, recommended is that PILs have a maximum reading requirement of US grade 8.
## Methods
Study 1: We obtained 140 current ASM PILs written in English. "Readability" was assessed using four tests, with and without adjustment for influence of familiar, polysyllabic words. A total of 179 online materials on epilepsy were also assessed.
Study 2: Two PILs from Study 1 were randomly selected (Pregabalin Focus; Inovelon) and shown to 35 people from the UK epilepsy population. Their comprehension was assessed.
Study 3: To understand whether the student population provides an accessible alternative population for future examination of ASM PILs, Study 3 was completed, using the same methods as Study 2, except that participants were 262 UK university students.
## Results
Study 1: No PIL had a reading level of grade 8. Median was grade 11. Adjusting for context, the PILs were still at grade 10.5. PILs for branded ASMs were most readable. PILs were no more readable than (unregulated) online materials.
Study 2: Users struggled to comprehend the PILs' key messages. The eight questions asked about pregabalin were typically answered correctly by 54%. For Inovelon, it was 62%.
Study 3: Most student participants comprehended the PILs' key messages. The questions about Inovelon were answered correctly by 90%; for pregabalin it was 86%.
## Significance
This is the first independent and comprehensive examination of ASM PILs. It found that PILs being used fail to meet recommendations and regulatory requirements and risk not being understandable to a substantial proportion of users. In finding that people from the epilepsy population differ markedly in comprehension of PILs compared to students, this study highlights the importance of completing user testing with the target population.
Key Points
The PIL, as the only document all PWE in the European Union prescribed ASMs routinely receive, could be key to self‐management
No independent evidence is available on the understandability of ASM PILs
We found none of the 140 PILs for ASMs used in the UK met the recommended maximum for reading age
When PWE were shown two of the PILs, they struggled to comprehend their key messages on how to safely and effectively use the related ASM
Only two of 16 comprehension questions asked of PWE were answered by a sufficient proportion to satisfy the threshold regulators recommend
## INTRODUCTION
People with epilepsy (PWE) and their significant others assume substantial responsibility for the management of epilepsy. To make informed decisions, they need understandable information. The Patient Information Leaflet (PIL), which has accompanied medicines in the EU since 1999, forms an authoritative document all people receive about their antiseizure medication (ASM). How understandable are they?
Most economically developed nations include a significant minority of people with low literacy levels. In England, 15% of adults have a literacy level at or below that of an 11‐year‐old. A further 28% have a reading age of a 12–14‐year‐old. When information exceeds someone's literacy level, there is the potential for misunderstandings. It has been recommended written materials have a maximum required reading age of ~13 years (US grade 8).
We systematically searched for studies examining ASM PILs. Five , , , , were identified (File ). Only one examined European PILs. Conducted by Wong, it focused on the PILs for 12 branded ASMs written in English. The length of the sentences used and the complexity of the words within them were quantified. Based on this assessment and comparison to reference data, Wong concluded the PILs should be relatively understandable to UK adults, with the text being classified as representing "plain English." However, published in 1998 and focusing on only branded ASMs, Wong's study tells us little about the understandability of current PILs.
Within the EU, the European Medicines Agency (EMA), and national competent authorities, approve PILs before use. They state PILs should be designed and worded so a maximum number of people can understand them. Since 2005, manufacturers have also legally been required to engage with users to develop their PILs.
A debate is occurring as to whether the EMA's processes are sufficient. , It would be appropriate for the epilepsy community to contribute. In England alone, in the 12 months from December 1, 2020, there were >30 million ASM prescriptions.
### What evidence is needed on ASM PILs ?
Different methods are available to determine how understandable ASM PILs are.
#### Readability
One way to gauge how understandable ASM PILs are is by using automated readability tests. These use different text characteristics to estimate "reading ease." Word length is used as a proxy for semantic difficulty, and sentence length indicates syntactic complexity. A numeric value is assigned to the text to indicate its "readability."
A standard application of readability tests to ASM PILs would provide evidence on their ease of use in a common format and allow comparison to PILs for other medications. It would be helpful, however, to also apply them while adjusting for context. This is because for many readability tests, the more polysyllabic words present within a document, the higher the judged required reading age. The challenge is that many polysyllabic words within a PIL (e.g., convulsion, levetiracetam) may be uniquely familiar to the target audience and so be poor predictors of readability. Without adjusting for this, a test might artificially inflate required reading age.
Regulators have made a PIL template available to manufacturers and stipulate standard headings. Some commentators contend this inadvertently reduces the readability of PILs. , In assessing ASM PILs, it would thus be insightful to compare their readability to materials on epilepsy written in English for PWE, whose presentation is less regulated. Online materials meet these criteria.
#### Literal comprehension
A second way to assess the PILs would be by determining their comprehensibility. Readability does not guarantee comprehension, as factors beyond text characteristics affect it (e.g., prior knowledge, interest, how information is presented).
No published evidence on the comprehensibility of ASM PILs is available. It could be obtained by completing so‐called "user testing." User testing (as per the Australian‐Sless method ) involves a PIL being given to ~20 individuals from the target population. They are asked questions to assess the PIL's ability to ensure people can find and understand information pertinent to the medicine's safe and effective use. Regulators cite it as a way manufacturers can demonstrate their PIL is ready for use, stating that a PIL should be iteratively refined and retested until each question is answered correctly by ≥80% of users.
User testing is resource intensive, as there is a need to recruit people from the target population. When they have a stigmatizing condition, this can be challenging. Funding for researchers undertaking user testing is also not forthcoming. To position the epilepsy community to independently check ASM PILs in the future, it would be helpful to understand whether PILs could be tested with populations more accessible to academic investigators. One that they can recruit from in large numbers is the student population. To be a suitable alternative for testing, the student population's pattern of comprehension results would need to be broadly indicative of those of the epilepsy population.
#### Objectives of current study
Given the information gaps identified, we conducted a series of studies that sought to:
Describe the readability of current PILs for ASMs prescribed in the UK (Study 1);
Explore factors associated with their readability (Study 1);
Compare the readability of PILs to online epilepsy materials (Study 1);
Complete user testing of a sample of ASM PILs to understand their comprehensibility to persons from the epilepsy population (Study 2); and
Complete user testing of the same ASM PILs, but with a larger sample of persons from the student population (Study 3).
## MATERIALS AND METHODS
### Study 1: Readability of PILs
#### Design
Study 1 was a cross‐sectional assessment of PILs for ASMs.
#### Materials
##### Patient information leaflets
In the UK, 27 active ingredients are approved for epilepsy (Table ). The PILs for the 148 medications containing them were obtained on October 6, 2020 from the Electronic Medicines Compendium; 140 (94.6%) were in a format that permitted testing. File lists them.
Patient Information Leaflets for antiseizure medications that were tested and their readability score by their active ingredient
##### Online materials to compare to PILs
A representative sample of 179 online epilepsy materials encountered by PWE was compiled (median word count = 1179, interquartile range [IQR] = 695–1916). File lists them and their identification. In brief, five internet searches were completed on October 6, 2020 using search terms PWE use (i.e., “epilepsy symptoms”, “what is epilepsy”, “epilepsy seizures”, “epilepsy UK” and “epilepsy medication”). Two reviewers independently screened the first 100 results from each search to identify eligible materials.
##### Preparation of materials for testing
Individual Microsoft Word versions of the PILs and online materials were created. No pictures, symbols, copyright notices, citations, advertisements, or internet addresses were included.
In line with standard practice, the PIL versions included text from five of the six sections that form a PIL in the UK and EU: "1. What X is and what it is used for"; "2. What you need to know before you <take> <use> X"; "3. How to <take> <use> X"; "4. Possible side effects"; and "5. How to store X." We excluded Section 6 ("Contents of the pack and other information"), because patients do not rate the information contained within it as particularly important. ,
The Word versions of the online materials included only text from the landing page.
#### Tests
The following established tests that consider different text characteristics and estimate years of US schooling required were used: Simple Measure of Gobbledygook (SMOG), Flesch–Kincaid (F‐K), and FORCAST. As per previous studies, , a composite score for each document was formed based on its median score on the tests. It can be broa converted to UK reading age by adding 5.
To provide a measure of readability on a continuous scale, the Flesch Reading Ease (FRE) test was also used. It ranges from 0 to 100; higher scores indicate greater readability. File details the formulae of each test.
Tests were completed using Readability Studio Professional Edition (v2019). They were first run in the standard way and then while adjusting for context (see below).
##### Adjustments for context
A list of words was compiled for exclusion from consideration by the testing software. It comprised the n = 59 words that create the generic and branded names of the ASMs, n = 78 key epilepsy terms, and n = 1464 adverse event terms. File details them and the rationale. The testing software was also instructed to exclude proper nouns and to treat all numerals as monosyllabic words.
#### Analysis
As the readability data were not normally distributed (Shapiro–Wilk, p < .01), analyses were completed using nonparametric tests (Mann–Whitney, Wilcoxon signed rank test, Spearman rank test). Central tendency is described according to the median and IQR. The proportion of PILs satisfying the recommended reading rade 8 level is described.
Factors explored for their association with PIL readability were: time since the ASMs focused on had been authorized for use and time since the PIL examined had been revised ; whether the ASM was branded or generic ; and extent to which the ASM was prescribed, with PILs for the three most commonly prescribed ASMs in England (lamotrigine, levetiracetam, valproate) being compared to the others. These analyses were completed using data from when the readability tests were adjusted.
For the main analyses, alpha was set at .05. When exploring factors associated with readability, alpha was Bonferroni adjusted ( p < .006).
Analyses were conducted using SPSS (v27).
### Study 2: User testing with people from the epilepsy population
#### Design
An anonymous, cross‐sectional online survey was run using Qualtrics. To minimize participant burden, we tested comprehension of two PILs. Order of presentation was randomized (1:1).
#### Recruitment
As is standard for user testing, a sample of ~ n = 20 users was sought, while recognizing a need to account for potential missing data.
Between November 2021 and February 2022, a participant advertisement was distributed using different social media platforms by UK epilepsy user groups (see Acknowledgments). Table shows the eligibility criteria and approvals.
Participant inclusion and exclusion criteria
Approval was provided by the University of Liverpool's Health and Life Sciences Research Ethics Committee (Reference: 7766). All participants provided informed consent.
#### Materials
To select the two PILs, we stratified the PILs assessed within Study 1 by their adjusted FRE score. We then randomly selected one PIL from the top quartile (namely, Inovelon film‐coated tablets) and one PIL from the bottom quartile (Pregabalin Focus; Table ).
Details of PILs selected for user testing (Studies 2 and 3)
#### Survey content
Participants were asked brief questions about demographics and epilepsy profile (or that of a person with epilepsy they knew). For each PIL, the participant was then asked eight comprehension questions (Table ).
Questions asked of participants about the different antiseizure medication leaflets to assess comprehension and the scores of the sample
PILs remained available to participants when answering the questions, and no time restrictions were applied. Participants typed their answers to the questions within free‐text boxes.
In developing the comprehension questions, regulatory guidance , was followed. Most were framed as scenarios, asking participants in an open‐ended way what the correct course of action was. Some requested the person to imagine finding themselves in a certain situation, others asked them to imagine someone they knew found themselves in the situation. This approach is consistent with guidance and has been used before. It also permitted the same set of questions to be used with all participants regardless of their characteristics (e.g., questions regarding female birth control and breastfeeding could be asked of all). Questions were phrased differently from the relevant text of the PILs, and the order of the topics asked about differed from the PIL. Face validity was confirmed by a consultant neurologist.
#### Analysis
Responses to the comprehension questions were coded as correct or incorrect by two independent raters based on criteria established a priori. Any discrepancies were resolved through discussion. Raters were trained undergraduate psychology students (N.C., S.H.). To understand interrater reliability, percentage agreement and the prevalence‐adjusted bias‐adjusted kappa (PABAK) were calculated.
The primary analysis focused on participants who completed comprehension questions for both PILs. As comprehension scores were not normally distributed (Shapiro–Wilk, p < .01), central tendency is described according to the median and IQR. The proportion of participants providing a correct response to each question is reported, along with 95% confidence interval (CI). Questions for each PIL were ranked according to the proportion of correct responses elicited.
To understand how participants answering comprehension questions for two PILs (completers) compared to those completing them for only one PIL (noncompleters), the total comprehension scores the two groups achieved on their first allocated PIL was calculated.
Analyses were conducted using SPSS (v27), StatsDirect3 was used for CIs, and PABAK was determined using the calculator at .
### Study 3: User testing with student population
#### Design
An anonymous, cross‐sectional online survey similar to that used for Study 2 was employed.
#### Recruitment
A sample size calculation was completed. It was informed by Biggs et al.’s estimate that 83% of children without epilepsy can potentially answer comprehension questions correctly having read an ASM PIL. This, together with a required confidence level of 95% and precision of ±5%, indicated 214 participants with complete data were required.
Between January and February 2022, participant advertisements were sent by email to students at the University of Liverpool within the schools of engineering, geography, management, and health and life sciences. Table shows the eligibility criteria.
Approval was provided by the University of Liverpool's Health and Life Sciences Research Ethics Committee (Reference: 7766 Amend). All participants provided informed consent.
#### Materials, survey content, and analysis
Materials, survey content, and analysis were the same as for Study 2. The only difference was that the comprehension scores of completers and noncompleters were formally compared (Mann–Whitney, alpha = .05.).
## RESULTS
### Study 1: Readability of PILs
#### Characteristics of PILS
Of the 140 PILs, 79 (56.4%) were for generic ASMs. The median authorization date for the ASMs focused on by the PILs was October 3, 2011 (IQR = December 31, 2005 to December 14, 2015); 106 (75.7%) of the ASMs had been authorized after October 2005. The median date on which the PILs examined had last been revised was November 1, 2019 (IQR = April 1, 2019 to March 1, 2020).
The PILs had a median word count of 2439.5 (IQR = 2116–2958.8), of which 17.5% of the words (IQR = 15.8–19.4) were polysyllabic. Sentences within the PILs had a median length of 14.1 words (IQR = 12.9–14.9).
#### Readability of PILs
##### According to standard test approach
No PIL had a reading grade score at or below grade 8. The estimated median required reading grade of the documents was 11.2 (IQR = 10.9–11.5), equivalent to a UK reading age of ~16 years (Table ). The median FRE score of the PILs was 50 (IQR = 45–55).
Scores on F‐K, SMOG, and FORCAST were all significantly correlated with one another in the expected direction ( r = .629–.969, all p < .001).
##### When adjusting for context
The adjustments reduced the proportion of polysyllabic words within the PILs by a median of 3.6% to 14.3% (IQR = 12.5–15) and led to the median reading grade requirement of the PLS being reduced to 10.5 (IQR = 10.2–10.7, z = −10.296, p < .001). FRE scores also significantly improved to 60 (IQR = 57–64, z = 10.282, p < .001). Nevertheless, only one (.7%) PIL had a reading grade at or below 8.
##### Factors associated with PIL readability
Time since the ASM was authorized and time since the PIL examined had last been revised were not significantly correlated with required reading grade ( r = .04 to −.17) or FRE score ( r = −.05 to .06). Moreover, PILs authorized before and after October 2005 did not significantly differ.
Compared to PILs for generic ASMs, PILs for branded ASMs had a significantly lower required reading grade (median = 10.3, IQR = 10.2–10.6 vs. median = 10.6, IQR = 10.3–10.8; U = 1689, p < .008) and higher FRE score (median = 62, IQR = 59–64 vs. median = 59, IQR = 57–63; U = 3150.500, p < .006). PILs for branded and generic ASMs were similar in word count ( U = 2690, p > .05), but branded PILs included a smaller proportion of polysyllabic words (13.5 vs. 14.6%; U = 1724.000, p < .006).
The required reading grade for the PILs for the most prescribed ASM ingredients was not statistically different from that of PILs for the ASMs with another ingredient ( p = .27). They did have a slightly worse FRE score (median = 58, IQR = 56–63), but this was not significant at the Bonferroni‐corrected level (median = 60, IQR = 58–64.5; U = 1453, p = .01).
##### Comparison of the readability of PILs with online epilepsy materials
No statistically significant differences were found to exist between PILs and online materials (all p > .05). Their unadjusted median required reading grade was 11.1 (IQR = 10.5–11.7), their FRE was 51 (IQR = 44–58), and four (2.2%) items had a reading grade at or below 8.
### Study 2: User testing with people from the epilepsy population
#### Characteristics of participants
Thirty‐five participants from the epilepsy population were recruited. Complete responses to the comprehension questions were provided by 24 (68.6%). It took them a median of 26 min to complete the survey (IQR = 10.1–37.8).
Their median age was 42 years (range = 36–45), most ( n = 22, 91.7%) were female, and most ( n = 21, 87.5%) took part because they had epilepsy (Table ). In terms of education, the highest attainment for 12 (50.0%) participants was a basic school certificate (typically completed at the age of 16 years in the UK), one (4.2%) had completed an advanced school certificate (aged 18 years in the UK), four (16.7%) had completed a university degree, and five (20.8%) had completed a postgraduate degree. For two (8.3%) participants, the education level was not clear.
Characteristics of participant samples for Studies 2 and 3
#### Comprehension
##### Interrater reliability
Rater agreement was excellent (Table ). For the Inovelon PIL, raters agreed between 88.6% and 100% of the time (PABAK = .77–1). For the pregabalin PIL, raters agreed between 85.7% and 100% of the time (PABAK = .71–1).
##### Participant comprehension
######
Completers versus noncompleters
The median number of correct answers that completers (5, IQR = 2.5–6) and noncompleters (5, IQR = 4–6) achieved for their first allocated PIL was similar.
###### Pregabalin
The median proportion of participants providing correct responses to the individual questions was 54.2% (range = 25.0%–83.3%). Only one question (Number 8) satisfied the regulators' ≥80% threshold (Table ). Question 3 elicited the least correct responses.
###### Inovelon
The median proportion of participants providing correct responses to the individual questions was 62.5% (range = 33.3%–83.3%). One question (Number 5) satisfied the ≥80% threshold. Question 2 elicited the least correct answers.
### Study 3: User testing with student population
#### Characteristics of participants
Two hundred sixty‐two participants were recruited; 237 (90.5%) provided complete responses to the comprehension questions. Median age was 20 years (IQR = 19–22), 66.2% were female, and 24 (10.1%) reported English was not their main language. Seven (3.0%) reported having an epilepsy diagnosis (Table ). They took a median of 17.5 min to complete the survey (IQR = 13.8–23.4).
#### Comprehension
##### Interrater reliability
Agreement between raters was excellent (PABAK = .72–1; Table ).
##### Participant comprehension
###### Completers versus noncompleters
The median number of correct answers that completers (7.0, IQR = 6–8) and noncompleters (7, IQR = 6–8) gave for their first allocated PIL did not significantly differ ( U = 2589.5, p > .05).
###### Pregabalin
The median proportion providing correct responses to the individual questions was 86.5% (range = 48.1%–95.4%; Table ). Six had ≥80% of participants providing correct responses to them. The question eliciting the least correct responses was Question 1.
###### Inovelon
The median proportion of participants providing correct responses to the individual questions was 90.9% (range = 70.9%–97.0%). Six had ≥80% of participants providing correct responses to them. The question eliciting the least correct responses was Question 7.
## DISCUSSION
### Main findings
Our comprehensive assessment suggests ASM PILs available in the UK may not be understood by a substantial proportion of the epilepsy population.
We assessed 140 PILs using readability tests. None had a reading age requirement at or below the recommended grade 8 level. Most were grade 11, similar to PILs for other medications. , Based on literacy level data, ~40% of the general adult population in the UK might struggle with the PILs. It could be worse in the epilepsy population, because it is at higher risk of low literacy.
We were cognizant that readability tests might, when applied in a standard way, not offer an accurate assessment. However, even after adjusting for this, the PILs still had too high a reading level (grade 10.5).
Despite all the regulations, templates, and guidelines in place to support PIL development, they performed no better than online materials on epilepsy. By some measures, the latter were marginally better.
There was some evidence that PILs for branded ASMs were more readable than those for generics. However, even branded PILs were written at too high a level (grade 10.3).
To our knowledge, this is first time a difference between branded and generic PILs in Europe has been reported. The practical relevance of the difference is unclear. It is nonetheless concerning. Generic ASMs are commonly prescribed in the UK, and there is momentum to use them more. Why the difference occurred is unknown. It is the case that applications for authorization for generic and branded medications in the EU can be submitted and reviewed slightly differently. This might be relevant.
Although readability tests are helpful, how a document performs with its intended user is the most important test. For Study 2, we recruited 34 people from the epilepsy population and presented them with two PILs. Only two of the 16 questions had sufficient people answering them correctly to meet the ≥80% threshold cited by regulators.
The size of the sample we used for Study 3 was in line with that recommended. Nevertheless, it does lack precision. Thus, it is helpful to consider the CIs for the estimates. Even if the upper bounds of the intervals are used, half of the comprehension questions still fail to satisfy the ≥80% threshold.
The consequences of a person failing to understand a PIL will be context dependent. PILs are also only one way that patients can obtain information about their medications. Deficiencies in the understandability of PILs could, for instance, be mitigated by any counsel the patient receives from their care provider(s). Nevertheless, it is concerning that the questions eliciting the most incorrect responses in Study 2 related to safety warnings released for the two ASMs, namely, potential consequences of taking Inovelon if one has a pre‐existing heart condition and the risks of taking pregabalin with oxycodone.
Only a small number of studies have assessed how well users of other medications comprehend materials written for them about their medication, and variability in the methods used prevents direct comparisons. Nevertheless, the studies do indicate ASM PILs are not unique in their failure to ensure patients consistently comprehend core messages. , , Another important finding from some of these other studies is that they showed how PILs can be successfully modified and patient comprehension improved.
### Findings in relation to regulations
Criticisms of PILs are not new. However, most studies from Europe have focused on PILs developed before the 2005 requirement of manufacturers to demonstrate engagement with users. Most of the PILs we examined had been authorized after 2005. Why then did they perform so poorly?
Were the two PILs we considered outliers? This is unlikely. We randomly selected them and included one from the quartile with the best readability score from Study 1.
A second possibility is our participants were unrepresentative. Our participants did report poor seizure control. However, they had characteristics that should have made comprehending the PILs easier. They were more educated than would be expected (37.5% were working toward/had achieved a university degree compared to 27.1% in England ), and ~12% reported some familiarity with one or more of the ASMs focused on by the PILs.
Third, might the way we conducted the user testing differ from the approach used by manufacturers? This is hard to know. The evidence manufacturers submit to regulators is not publicly available.
If we assume manufacturers all use the Australian‐Sless method, then it is true that some differences existed in how we conducted the user testing. However, these should not account for the PILs performing so poorly.
One difference was (partly because of COVID‐19) that we assessed comprehension via a survey, rather than by face‐to‐face interview. The approach has been used before. , Might it, however, have meant people were less likely to be scored as having given a correct response (e.g., answers could not be explored)? Our findings suggest not, because the answers people typed were clear enough for two raters to consistently agree on their correctness.
People viewed the PILs electronically rather than as paper documents. Could this have made the PILs less easy to comprehend? This is possible. However, PILs are used by people in this form, and doing so can allow them to overcome complaints about paper PILs (e.g., zooming in to increase text size, using word search function).
Finally, the assessment process we used with users was abbreviated. We asked users eight questions regarding each PIL. The Australian‐Sless method involves users being asked more (~15). Half of these typically ask the person to show where specific information in the PIL is; the other half assesses the person's comprehension of that information. Regulators state ≥80% of participants should be able to both find information and answer related comprehension questions. To minimize participant burden, we only assessed participants' comprehension (i.e., we did not ask participants to show where the information was or award marks for this). This difference should not explain why PILs performed so poorly in our study, because we simply described the proportion of participants giving correct answers to the different questions and the number satisfying or exceeding the 80% threshold cited by regulators.
### Implications
Our findings have relevance for both the UK and the EU. All the PILs examined had been approved while the UK was an EU member. Moreover, the processes the UK uses now that it has left the EU remain similar.
One interpretation of our findings is that more regulation and guidance on PIL development is required. We contend there is a need to first determine how well current regulations on involving users are being adhered to by manufacturers and enforced by regulators. User involvement should be meaningful, not a "tick‐box" exercise. Regulators could clarify the situation by including within the Public Assessment Reports they publish detailed evidence on what user engagement manufacturers did. In the meantime, the identified limitations of PILs highlight the importance of pharmacists and other care providers providing comprehensive medication counseling when dispensing any new ASM.
It would be helpful if the epilepsy community could periodically complete independent evaluations of ASMs. Funding for such work is limited. We explored the utility of completing user testing with the student population. Although the student population was straightforward to recruit and assess, its comprehension scores were not indicative of those of the epilepsy population. At least 80% of the student sample answered 12 of the 16 questions correctly. Moreover, the questions they struggled with most differed. Alternative ways to support independent assessments of ASM PILs warrant consideration.
### Strengths and limitations
Our identification of PILs was systematic, the assessment comprehensive, and reporting transparent. The online materials we compared the PILs to were systematically identified and representative. , As shown by our systematic literature search, we are presenting the first published evidence on user testing of ASM PILs with the epilepsy population.
A potential weakness of our study is that the PILs are reflective of those available at one point in time. Some may have since been updated and understandability improved. This seems unlikely, because no substantive changes to how PILs are approved have been introduced. Also, we did not find time since authorization or revision to be related to readability in Study 1.
## CONCLUSIONS
PILs are a mandatory document all people prescribed ASMs in the EU receive. Our independent and comprehensive examination of them suggests those being used in the UK may not be understandable to a sizeable proportion of the epilepsy population.
## AUTHOR CONTRIBUTIONS
Adam J. Noble (Senior Lecturer in Health Services Research) was chief investigator; conceived of the study; led its design; supervised and coordinated the study and the analysis and interpretation of the data; and wrote the final report. Niamh Coleman and Sara Haddad (Research Assistants) contributed to the running of and recruitment for Studies 2 and 3 and scoring of comprehension responses. They also assisted with the systematic review of the literature. Anthony G. Marson (Professor of Neurology and Consultant Neurologist) contributed to the design of Studies 2 and 3 and the interpretation of results, and reviewed the final report.
## CONFLICT OF INTEREST
None of the authors has any conflict of interest to disclose.
## Supporting information
|
Myoclonus is often observed in epilepsy. It is characterized by sudden involuntary shock-like movements of the body (myoclonic jerks, MJs). This study examined whether epileptic myoclonus was under genetic control. Inbred strains of mice were administered eight daily flurothyl exposures, a 28-day rest period, and a final flurothyl retest. For all trials, the latency to the first MJ (threshold) and the number of MJs (MJ#) were recorded. The inbred strains that we examined exhibited significant variability in initial myoclonic response, and myoclonus across the eight flurothyl exposures. C57BL/6J and DBA/2J mice displayed significantly different initial latencies to a MJ, MJ# preceding a generalized seizure (GS), and changes in MJ threshold and MJ# across the eight seizure trials. [C57BL/6J x DBA/2J] F1-hybrid mice showed an initial MJ threshold and decreases in MJ threshold over the eight trials, which were similar to C57BL/6J; however, F1-hybrids had an initial MJ# and trend in MJ# over the eight trials that were similar to DBA/2J. Decreases in MJ threshold and MJ# following multiple seizure trials, observed in C57BL/6J mice, were dependent on the expression of GSs and not on MJ occurrence. Our study is the first to document the potential for genetic heterogeneity of myoclonus in mice; we show that significant alterations in myoclonic behavior occur after GSs. These results indicate that multiple GSs affect MJ thresholds. An understanding of the genetics of myoclonus will be important for determination of the brain areas responsible for myoclonus as well as for identification of candidate genes. |
Here we assessed whether the presence of an aromatic ring as a commonality in chemical structures of AEDs can explain skin reaction. We found that 164 cases of skin reactions associated with the use of AEDs were reported. Aromatic AEDs were suspected in 88.41% (145/164) of patients with skin reactions versus 59.80% (2316/3873) of patients without skin reactions. The presence of an aromatic ring in the chemical structure was associated with a significant increased risk of skin reactions (adjusted ROR 3.50; 95% CI 2.29, 5.35). Among the aromatic AEDs, skin reactions were significantly associated with carbamazepine, lamotrigine, and oxarbazepine. These results confirm that the presence of an aromatic ring as a common feature in chemical structures of AEDs partly explains AED-skin reactions. Skin reactions were reported triple as frequently with aromatic AEDs than with non-aromatic AEDs. |
Despite the best possible medication and treatment protocols, one-third of epilepsy patients have drug resistance which is associated with an elevated risk of mortality and debilitating psychological consequences. P-glycogen encoded by ABCB1 is major drug transporter for a wide variety of AED. To evaluate the complex haplotypic association, genetic and allelic frequency distribution of rs1128503, rs1045642, and rs2032582 polymorphisms of ABCB1 gene with drug resistance in Pakistani pediatric epilepsy patients, we performed this study. A total of 337 individuals including 100 healthy control, 110 drug-resistant patients, and 127 drug-responsive patients were enrolled and genotyped for three polymorphisms. PCR and direct sequencing of DNA were done for genotyping. All the studied SNPs showed a statistically significant association with drug-resistant epilepsy at p < 0.01. In addition, we identified a novel variant at c 0.2678C > A (SCV001712095) position. The haplotype analysis indicated strong linkage disequilibrium between three SNPs. The in-silico analysis indicated that rs2032582 polymorphism at c 0.2677T > A is benign while c 0.2677T > G and c 0.2678C > A are possibly damaging. Our findings showed that pharmacogenetic variants play a key role in disease. Our findings shed light on the pharmacogenomic association of ABCB1 with epilepsy which might facilitate study on pharmacokinetics concerning ethnology. |
Social cognitive neuroscience has highlighted the importance of frontotemporal neurocircuitry for social cognition. Temporal lobe epilepsy (TLE) impacts these brain areas and their functional connections and might therefore specifically affect social perceptual and cognitive skills. In the study described here, an established paradigm was used to evaluate the social cognitive skills of female patients with left TLE. Study participants were shown dynamic animations in which virtual characters either looked at the human observer directly or looked away toward someone else, thus manipulating self-involvement. The virtual characters then exhibited different facial expressions that were either socially relevant or arbitrary. Participants were asked to rate the communicative intentions of the virtual character. Patients' ratings of communicative intent appeared to be linked to their own self-involvement in the interaction, whereas healthy volunteers' ratings of facial expressions were independent of self-involvement. Potential mechanisms for the observed differences are discussed. |
Women with epilepsy (WWE)'s knowledge of the interaction between antiepileptic drugs (AEDs) and oral contraceptives (OCs) and the potential teratogenicity of AEDs has received limited study. We conducted a cross-sectional questionnaire study (English or Spanish) among young WWE (18-44 years) to assess demographic characteristics, current AED use, and knowledge of AED interactions with OCs and teratogenicity. We used the Food and Drug Administration's classification system to categorize each AED's teratogenic potential. Participants (n=148) had a mean age of 32 years (SD 8); 32% spoke Spanish and described themselves as Hispanic. Among women prescribed a cytochrome p450-inducing AED, 65% were unaware of decreased OC efficacy. Forty percent of those prescribed Category D AEDs were unaware of potential teratogenic effects. WWE have limited knowledge of the potential interaction between AEDs and OCs and the teratogenic effects of AEDs. Educational efforts should highlight the reproductive health effects of AEDs in WWE. |
Epilepsy affects approximately 0.5-1% of youth, and challenges for them and their families reach far beyond seizures. Quantitative studies have shown that in addition to increased risk for psychosocial difficulties, many experience stigma and barriers to services and resources. As a complement to quantitative analyses, qualitative research further provides unique insight into understanding the impact of epilepsy on youth and families. In the present study, focus groups were held to discuss families' experiences with epilepsy and access to related services. Qualitative analysis revealed three themes highlighting medical, educational, and social challenges of youth with epilepsy. Implications include recommendations for improvements in public awareness and public policy change. |
The current study examined the specific types of attention-related problems children with childhood absence epilepsy (CAE) experience and the role of disease factors in the development of attention-related problems. Thirty-eight subjects with CAE and 46 healthy controls, aged 6 to 16, participated in the study. The Behavior Assessment System for Children (BASC) was completed by parents, and the Attention Problems and Hyperactivity subscales were used to characterize the problems of children with CAE. Item analysis within the subscales revealed that children with CAE demonstrate higher rates of hyperactive (overactivity and fidgetiness) and inattentive (forgetfulness and distractibility) problems, and require more supervision. Within-CAE-group analyses revealed that those who were actively having seizures were more impatient and those with a longer duration of illness were less proficient in completing homework. Children with CAE are at risk for certain inattentive and hyperactive problems, which can differ depending on duration of illness and active seizure status. |
This study constitutes a preliminary test of a theoretical model proposed by Sexson and Madan-Swain to explain the school status of students with epilepsy. Sixty-six classroom teachers participated in the study, as did 74 of their students with epilepsy. Three predictor variables-teachers' attitude towards persons with epilepsy, teachers' training in instructing students with epilepsy, and students' seizure frequency-were examined. Consistent with the model, the three variables collectively predicted attendance (F = 54.48, p<.001, R2 = 0.70), reading (F = 21.40, p<.001, R2 = .48), math (F = 12.61, p<.001, R2 = 0.35), writing (F = 12.61, p<.001, R2 = 0.35), and special education usage (χ2 = 30.96, p<.001). Moreover, both seizure frequency and teachers' attitude, but not teachers' training, uniquely predicted some outcome variables. Limitations and potential advantages of the model are discussed. |
The new classification of epilepsy stratifies the disease into an acute level, based on seizures, and an overarching chronic level of epileptic syndromes (Berg et al., 2010). In this new approach, seizures are considered either to originate and evolve in unilateral networks or to rapidly encompass both hemispheres. This concept extends the former vision of focal and generalized epilepsies to a genuine pathology of underlying networks. These key aspects of the new classification can be linked to the concept of cognitive curtailing in focal epilepsy. The present review will discuss the conceptual implications for acute and chronic cognitive deficits with special emphasis on transient and structural disconnectivity. Acute transient disruption of brain function is the hallmark of focal seizures. Beyond seizures, however, interictal epileptic discharges (IEDs) are increasingly recognized to interfere with physiological brain circuitry. Both concomitant EEG and high-precision neuropsychological testing are necessary to detect these subtle effects, which may concern task-specific or default-mode networks. More recent data suggest that longstanding IEDs may affect brain maturation and eventually be considered as a biomarker of pathological wiring. This brings us to the overarching level of chronic cognitive and behavioral comorbidity. We will discuss alterations in structural connectivity measured with diffusion-weighted imaging and tractography. Among focal epilepsies, much of our current insights are derived from temporal lobe epilepsy and its impact on neuropsychological and psychiatric functioning. Structural disconnectivity is maximal in the temporal lobe but also concerns widespread language circuitry. Eventually, pathological wiring may contribute to the clinical picture of cognitive dysfunction. We conclude with the extrapolation of these concepts to current research topics and to the necessity of establishing individual patient profiles of network pathology with EEG, high-precision neuropsychological testing, and state-of-the-art neuroimaging. This article is part of a Special Issue entitled "The new approach to classification: Rethinking cognition and behavior in epilepsy". |
This study aimed to compare the self-esteem of Brazilian adolescents with epilepsy and Brazilian adolescents without this condition and the correlations between self-esteem of these adolescents with depression and anxiety symptoms. Study participants were 101 adolescents of both sexes, aged 10-19years old, from elementary and high school education. Fifty patients diagnosed with uncomplicated epilepsy attending the pediatric epilepsy clinic of University Hospital composed the case group. The other fifty-one adolescents without this diagnosis were attending public schools in Campinas-SP region. The instruments used were: identification card with demographics and epilepsy data, Multidimensional Self-Esteem Scale, Beck Depression Inventory and Inventory of State-Trait Anxiety - IDATE. A statistically significant result was found in the Responsibility Self-esteem Dimension favoring the control group. Significant correlations between self-esteem scores and anxiety and depression symptoms were also found. The development of a chronic disease such as epilepsy leads to a change in the way the individual perceives himself and the social environment he is inserted, influencing his behavior. The way people with epilepsy experience their seizures is a subjective measure that will control his/her well-being. Childhood and adolescence form the basis for a healthy emotional development; thus, our results show the importance of studying how subjective variables relate to the physical aspects of a chronic disease in these life stages. |
Renaissance was a period full of religious and supernatural concepts and practices distant from the contemporary scientific world. Some erratic behaviors were considered demonic possessions and treated by exorcisms. It is supported by many sources. However, some important sources of the Renaissance point to a different picture. They show trends towards naturalistic explanations of many diseases, including epilepsy. This critical review discusses this approach, using texts by Mondino de' Luzzi and Leonardo da Vinci. However, more than an historical study, this review considers the passage from religious and supernatural practices to modern science. Contemporary consequences of that passage are considered, considering this Special Issue of Epilepsy & Behavior. |
The dorsal attention network (DAN) is involved in the process that causes wide-ranging cognitive damage resulted in right temporal lobe epilepsy (rTLE). Nevertheless, few studies have evaluated the relationship between DAN and rTLE. There has been little research on alterations in the network homogeneity (NH) of the DAN in rTLE. The aim of the present study was to investigate NH changes in DAN in patients with rTLE. We included 85 patients with rTLE and 69 healthy controls in this study, and resting-state functional magnetic resonance imaging (rs-fMRI) data were acquired. The NH method was used for data analysis. All subjects took the attention network test (ANT). Network homogeneity in the right superior parietal lobule (SPL) and right precuneus (PCU) was significantly higher in patients with rTLE than in healthy controls. The reaction time (RT) was significantly longer in patients with rTLE than in controls. Notably, we observed no significant relationship between the clinical variables and the abnormal NH. These results indicated that abnormal alterations in DAN existed in patients with rTLE and highlighted the crucial role of DAN in the pathophysiology of cognitive damage in rTLE. Our findings suggested that the executive function (EF) significantly weakened in patients with rTLE. |
This study aimed to investigate the trajectory of fibres from the pontine nuclei that reach the two sides of the cerebellum. Injections of biotinylated dextran amine (BDA) were made within the basilar pontine nuclei (BPN) and the nucleus reticularis tegmenti pontis (NRTP) in one side of rats with electrolytic injury of the middle cerebellar peduncle (MCP), ipsilateral or contralateral to the side of injection. Fibres were traced from the pontine nuclei (BPN and NRTP) to both sides of the cerebellum passing through the respective MCPs. The study carried out in rats with injury to one peduncle showed projections segregated to the half-side of the cerebellum innervated by the intact peduncle. The laterality observed was confirmed by a retrograde tracer study. In fact, injections of different fluorescent tracers in rats with injury of single MCP showed that in the pontine nuclei only cell bodies stained by the tracer injected in the half-cerebellum ipsilateral to the intact peduncle. Finally, similar injections (i.e. different fluorescent tracers in symmetric areas of the cerebellar cortex) in the cerebellum of intact brain rats showed that BPN and NRTP differ for the laterality of their projections. In fact, 82% of BPN cells project contralaterally and 18% ipsilaterally, whereas 60% of NRTP cells project contralaterally and 40% ipsilaterally. In conclusion, this study showed that the MCPs receive fibres from the pontine nuclei of both sides and project to the ipsilateral half of the cerebellum and that different contingents of projections to the two sides of the cerebellum arise from BPN and NRTP. |
The visual field is retinotopically represented in early visual areas. It has been suggested that when adult primary visual cortex (V1) is deprived of normal retinal input it is capable of large-scale reorganisation, with neurons inside the lesion projection zone (LPZ) being visually driven by inputs from intact retinal regions. Early functional magnetic resonance imaging (fMRI) studies in humans with macular degeneration (MD) report > 1 cm spread of activity inside the LPZ border, whereas recent results report no shift of the LPZ border. Here, we used fMRI population receptive field measurements to study, for the first time, the visual cortex organisation of one macaque monkey with MD and to compare it with normal controls. Our results showed that the border of the V1 LPZ remained stable, suggesting that the deafferented area V1 zone of the MD animal has limited capacity for reorganisation. Interestingly, the pRF size of non-deafferented V1 voxels increased slightly (~20% on average), although this effect appears weaker than that in previous single-unit recording reports. Area V2 also showed limited reorganisation. Remarkably, area V5/MT of the MD animal showed extensive activation compared to controls stimulated over the part of the visual field that was spared in the MD animal. Furthermore, population receptive field size distributions differed markedly in area V5/MT of the MD animal. Taken together, these results suggest that V5/MT has a higher potential for reorganisation after MD than earlier visual cortex. |
In Parkinson's disease, a loss of dopamine neurons causes severe motor impairments. These motor impairments have long been thought to result exclusively from immediate effects of dopamine loss on neuronal firing in basal ganglia, causing imbalances of basal ganglia pathways. However, motor impairments and pathway imbalances may also result from dysfunctional synaptic plasticity - a novel concept of how Parkinsonian symptoms evolve. Here we built a neuro-computational model that allows us to simulate the effects of dopamine loss on synaptic plasticity in basal ganglia. Our simulations confirm that dysfunctional synaptic plasticity can indeed explain the emergence of both motor impairments and pathway imbalances in Parkinson's disease, thus corroborating the novel concept. By predicting that dysfunctional plasticity results not only in reduced activation of desired responses, but also in their active inhibition, our simulations provide novel testable predictions. When simulating dopamine replacement therapy (which is a standard treatment in clinical practice), we observe a new balance of pathway outputs, rather than a simple restoration of non-Parkinsonian states. In addition, high doses of replacement are shown to result in overshooting motor activity, in line with empirical evidence. Finally, our simulations provide an explanation for the intensely debated paradox that focused basal ganglia lesions alleviate Parkinsonian symptoms, but do not impair performance in healthy animals. Overall, our simulations suggest that the effects of dopamine loss on synaptic plasticity play an essential role in the development of Parkinsonian symptoms, thus arguing for a re-conceptualisation of Parkinsonian pathophysiology. |
Excitability of regenerated fibers remains impaired due to changes in both passive cable properties and alterations in the voltage-dependent membrane function. These abnormalities were studied by mathematical modeling in human regenerated nerves and experimental studies in mice. In three adult male patients with surgically repaired complete injuries of peripheral nerves of the arm 22 months-26 years prior to investigation, deviation of excitability measures was explained by a hyperpolarizing shift in the resting membrane potential and an increase in the passive 'Barrett and Barrett' conductance (GBB) bridging the nodal and internodal compartments. These changes were associated with an increase in the 'fast' K(+) conductance and the inward rectifier conductance (GH). Similar changes were found in regenerated mouse tibial motor axons at 1 month after a sciatic crush lesion. During the first 5 months of regeneration, GH showed partial recovery, which paralleled that in GBB. The internodal length remained one-third of normal. Excitability abnormalities could be reversed by the energy-dependent Na(+)/K(+) pump blocker ouabain resulting in membrane depolarization. Stressing the Na(+) pumping system during a strenuous activity protocol triggered partial Wallerian degeneration in regenerated nerves but not in control nerves from age-matched mice. The current data suggest that the nodal voltage-gated ion channel machinery is restored in regenerated axons, although the electrical separation from the internodal compartment remains compromised. Due to the persistent increase in number of nodes, the increased activity-dependent Na(+) influx could lead to hyperactivity of the Na(+)/K(+) pump resulting in membrane hyperpolarization and neurotoxic energy insufficiency during strenuous activity. |
We tested the predictions of the dynamic reweighting model (DRM) of audiovisual (AV) speech integration, which posits that spectrotemporally reliable (informative) AV speech stimuli induce a reweighting of processing from low-level to high-level auditory networks. This reweighting decreases sensitivity to acoustic onsets and in turn increases tolerance to AV onset asynchronies (AVOA). EEG was recorded while subjects watched videos of a speaker uttering trisyllabic nonwords that varied in spectrotemporal reliability and asynchrony of the visual and auditory inputs. Subjects judged the stimuli as in-sync or out-of-sync. Results showed that subjects exhibited greater AVOA tolerance for non-blurred than blurred visual speech and for less than more degraded acoustic speech. Increased AVOA tolerance was reflected in reduced amplitude of the P1-P2 auditory evoked potentials, a neurophysiological indication of reduced sensitivity to acoustic onsets and successful AV integration. There was also sustained visual alpha band (8-14 Hz) suppression (desynchronization) following acoustic speech onsets for non-blurred vs. blurred visual speech, consistent with continuous engagement of the visual system as the speech unfolds. The current findings suggest that increased spectrotemporal reliability of acoustic and visual speech promotes robust AV integration, partly by suppressing sensitivity to acoustic onsets, in support of the DRM's reweighting mechanism. Increased visual signal reliability also sustains the engagement of the visual system with the auditory system to maintain alignment of information across modalities. |
A large body of data has identified numerous molecular targets through which ethanol (EtOH) acts on brain circuits. Yet how these multiple mechanisms interact to result in dysregulated dopamine (DA) release under the influence of alcohol in vivo remains unclear. In this manuscript, we delineate potential circuit-level mechanisms responsible for EtOH-dependent dysregulation of DA release from the ventral tegmental area (VTA) into its projection areas. For this purpose, we constructed a circuit model of the VTA that integrates realistic Glutamatergic (Glu) inputs and reproduces DA release observed experimentally. We modelled the concentration-dependent effects of EtOH on its principal VTA targets. We calibrated the model to reproduce the inverted U-shape dose dependence of DA neuron activity on EtOH concentration. The model suggests a primary role of EtOH-induced boost in the I<sub>h</sub> and AMPA currents in the DA firing-rate/bursting increase. This is counteracted by potentiated GABA transmission that decreases DA neuron activity at higher EtOH concentrations. Thus, the model connects well-established in vitro pharmacological EtOH targets with its in vivo influence on neuronal activity. Furthermore, we predict that increases in VTA activity produced by moderate EtOH doses require partial synchrony and relatively low rates of the Glu afferents. We propose that the increased frequency of transient (phasic) DA peaks evoked by EtOH results from synchronous population bursts in VTA DA neurons. Our model predicts that the impact of acute ETOH on dopamine release is critically shaped by the structure of the cortical inputs to the VTA. |
Addiction to nicotine is extremely challenging to overcome, and the intense craving for the next cigarette often leads to relapse in smokers who wish to quit. To dampen the urges of craving and inhibit unwanted behaviour, smokers must harness cognitive control, which is itself impaired in addiction. It is likely that craving may interact with cognitive control, and the present study sought to test the specificity of such interactions. To this end, data from 24 smokers were gathered using EEG and behavioural measures in a craving session (following a three-hour nicotine abstention period) and a non-craving session (having just smoked). In both sessions, participants performed a task probing various facets of cognitive control (response inhibition, task switching and conflict processing). Results showed that craving smokers were less flexible with the implementation of cognitive control, with demands of task switching and incongruency yielding greater deficits under conditions of craving. Importantly, inhibitory control was not affected by craving, suggesting that the interactions of craving and cognitive control are selective. Together, these results provide evidence that smokers already exhibit specific control-related deficits after brief nicotine deprivation. This disruption of cognitive control while craving may help to explain why abstinence is so difficult to maintain. |
Behavioral assays in the mouse can show marked differences between male and female animals of a given genotype. These differences identified in such preclinical studies may have important clinical implications. We recently made a mouse model with impaired presynaptic inhibition through Gβγ-SNARE signaling. Here, we examine the role of sexual dimorphism in the severity of the phenotypes of this model, the SNAP25Δ3 mouse. In males, we already reported that SNAP25Δ3 homozygotes demonstrated phenotypes in motor coordination, nociception, spatial memory and stress processing. We now report that while minimal sexually dimorphic effects were observed for the nociceptive, motor or memory phenotypes, large differences were observed in the stress-induced hyperthermia paradigm, with male SNAP25Δ3 homozygotes exhibiting an increase in body temperature subsequent to handling relative to wild-type littermates, while no such genotype-dependent effect was observed in females. This suggests sexually dimorphic mechanisms of Gβγ-SNARE signaling for stress processing or thermoregulation within the mouse. Second, we examined the effects of heterozygosity with respect to the SNAP25Δ3 mutation. Heterozygote SNAP25Δ3 animals were tested alongside homozygote and wild-type littermates in all of the aforementioned paradigms and displayed phenotypes similar to wild-type animals or an intermediate state. From this, we conclude that the SNAP25Δ3 mutation does not behave in an autosomal dominant manner, but rather displays incomplete dominance for many phenotypes. |
The concept of alexithymia has garnered much attention in an attempt to understand the psychological mechanisms underlying the experience of feeling an emotion. In this study, we aimed to understand how the interoceptive processing in an emotional context relates to problems of alexithymia in recognizing self-emotions. Therefore, we prepared experimental conditions to induce emotional awareness based on interoceptive information. As such, we asked participants to be aware of interoception under an anxiety-generating situation anticipating pain, having them evaluate their subjective anxiety levels in this context. High alexithymia participants showed attenuated functional connectivity within their 'interoception network', particularly between the insula and the somatosensory areas when they focused on interoception. In contrast, they had enhanced functional connectivity between these regions when they focused on their anxiety about pain. Although access to somatic information is supposed to be more strongly activated while attending to interoception in the context of primary sensory processing, high alexithymia individuals were biased as this process was activated when they felt emotions, suggesting they recognize primitive and unprocessed bodily sensations as emotions. The paradoxical somatic information processing may reflect their brain function pathology for feeling emotions and their difficulty with context-dependent emotional control. |
Growing evidence indicates that the parasympathetic system is implicated in migraine headache. However, the cholinergic mechanisms in the pathophysiology of migraine remain unclear. We investigated the effects and mechanisms of cholinergic modulation and a mast cell stabilizer cromolyn in the nitroglycerin-induced in vivo migraine model and in vitro hemiskull preparations in rats. Effects of cholinergic agents (acetylcholinesterase inhibitor neostigmine, or acetylcholine, and muscarinic antagonist atropine) and mast cell stabilizer cromolyn or their combinations were tested in the in vivo and in vitro experiments. The mechanical hyperalgesia was assessed by von Frey hairs. Calcitonin gene-related peptide (CGRP) and C-fos levels were measured by enzyme-linked immunosorbent assay. Degranulation and count of meningeal mast cells were determined by toluidine-blue staining. Neostigmine augmented the nitroglycerin-induced mechanical hyperalgesia, trigeminal ganglion CGRP levels, brainstem CGRP, and C-fos levels, as well as degranulation of mast cells in vivo. Atropine inhibited neostigmine-induced additional increases in CGRP levels in trigeminal ganglion and brainstem while it failed to do this in the mechanical hyperalgesia, C-fos levels, and the mast cell degranulation. However, all systemic effects of neostigmine were abolished by cromolyn. The cholinergic agents or cromolyn did not alter basal release of CGRP, in vitro, but cromolyn alleviated the CGRP-inducing effect of capsaicin while atropine failed to do it. These results ensure for a first time direct evidence that endogenous acetylcholine contributes to migraine pathology mainly by activating meningeal mast cells while muscarinic receptors are involved in CGRP release from trigeminal ganglion and brainstem, without excluding the possible role of nicotinic cholinergic receptors. |
The ability of four different brainstem motoneuron pools to perform a newly acquired motor task was studied in alert cats. A classical conditioning of eyelid responses was carried out in (i). unoperated animals, and in animals with (ii). transection, 180 degrees rotation, and re-suture of the zygomatic facial nerve branch, (iii). a crossed anastomosis of the buccal to the zygomatic facial nerve branch and (iv). a hypoglossal-facial nerve anastomosis. Animals were conditioned with a delay paradigm using a tone (350 ms, 600 Hz, 90 dB) as conditioned stimulus, followed 250 ms later by an air puff (100 ms, 3 kg/cm2) as unconditioned stimulus. Animals with zygomatic nerve rotation performed conditioned responses (CRs) at control rate, with significantly larger amplitude, area and velocity, but a de-synchronized oscillatory pattern. Animals with buccal-zygomatic anastomosis acquired CRs at control rate, but these CRs had significantly smaller amplitude than those of controls and a de-synchronized pattern. Animals with a hypoglossal-facial anastomosis were unable to perform CRs. The trigeminal hyper-reflexia triggered by the axotomy was probably the origin of the large CRs after zygomatic nerve rotation. Trigeminal hyper-reflexia could also contribute to generation of the small CRs recorded after buccal-zygomatic anastomosis. Although trigeminal hyper-reflexia was also present following hypoglossal-facial anastomosis, hypoglossal motoneurons did not reach their firing threshold to perform CRs. In accordance with the embryonic origin of involved motoneurons, animals with buccal-zygomatic and hypoglossal-facial anastomoses moved the ipsilateral eyelid synchronously to mouth-related activities. It is suggested that there is a gradient of adaptability in motoneuron pools forced to perform new motor tasks through foreign muscles, which depends on their embryological origins and functional properties. |
The role of adenosine triphosphate (ATP) as a neurotransmitter and extracellular diffusible messenger has recently received considerable attention because of its possible participation in the regulation of synaptic plasticity. However, the possible contribution of extracellular ATP in maintaining and regulating synaptic efficacy during intracellular ATP depletion is understudied. We tested the effects of extracellular ATP on excitatory postsynaptic currents (EPSCs) evoked in CA1 pyramidal neurons by Schaffer collateral stimulation. In the absence of intracellular ATP, EPSC rundown was neutralized when a low concentration of ATP (1 microm) was added to the extracellular solution. Adenosine and ATP analogues did not prevent the EPSC rundown. The P(2) antagonists piridoxal-5'-phosphate-azophenyl 2',4'-disulphonate (PPADS) and reactive blue-2, and the P(1) adenosine receptor antagonist 8-cyclopentyltheophylline (CPT) had no detectable effects in cells depleted of ATP. However, the protective action of extracellular ATP on synaptic efficacy was blocked by extracellular application of the protein kinase inhibitors K252b and staurosporine. In contrast, K252b and staurosporine per se did not interfere with synaptic transmission in ATP loaded cells. Without intracellular ATP, bath-applied caffeine induced a transient (< 35 min) EPSC potentiation that was transformed into a persistent long-term potentiation (> 80 min) when 1 microm ATP was added extracellularly. An increased probability of transmitter release paralleled the long-term potentiation induced by caffeine, suggesting that it originated presynaptically. Therefore, we conclude that extracellular ATP may operate to maintain and regulate synaptic efficacy and plasticity in conditions of abnormal intracellular ATP depletion by phosphorylation of a surface protein substrate via activation of ecto-protein kinases. |
Strong synchronization of neuronal activity occurs in the 8-35 Hz band in the subthalamic nucleus (STN) of patients with Parkinson's disease (PD) and is evident as oscillatory local field potential (LFP) activity. To test whether such synchronization may contribute to bradykinesia and rigidity, we sought correlations between the suppression of synchronization at 8-35 Hz in STN and the reduction in Parkinsonism with levodopa. LFPs were recorded on and off medication from STN deep-brain stimulation electrodes in nine PD patients. LFP power was calculated over the frequencies of the most prominent spectral peak within the 8-35 Hz frequency band on each of 17 sides (off medication), and over the frequencies of any peak in the 60-90 Hz band, if present (seven sides, on medication). Levodopa-induced reduction of LFP power over these two frequency ranges was then correlated with improvement in motor impairment as assessed by the Unified Parkinson's Disease Rating Scale (UPDRS). The reduction in peak activity in the 8-35 Hz band with levodopa positively correlated with the improvement in the contralateral hemibody motor UPDRS score with levodopa (r = 0.811, P < 0.001) as well as with hemibody subscores of akinesia-rigidity (r = 0.835, P < 0.001), but not tremor. A trend for negative correlations was found between peak 60-90 Hz LFP power and UPDRS hemibody score, suggesting that positive correlations were relatively frequency-specific. Our results support a link between levodopa-induced improvements in bradykinesia and rigidity and reductions in population synchrony at frequencies < 35 Hz in the region of the STN in patients with PD. |
Long-term potentiation (LTP), a use dependent long-lasting modification of synaptic strength, was first discovered in the hippocampus and later shown to occur in sensory areas of the spinal cord. Here we demonstrate that spinal LTP requires the activation of a subset of superficial spinal dorsal horn neurons expressing the neurokinin-1 receptor (NK1-R) that have previously been shown to mediate certain forms of hyperalgesia. These neurons participate in local spinal sensory processing, but are also the origin of a spino-bulbo-spinal loop driving a 5-hydroxytryptamine 3 receptor (5HT3-R)- mediated descending facilitation of spinal pain processing. Using a saporin-substance P conjugate to produce site-specific neuronal ablation, we demonstrate that NK1-R expressing cells in the superficial dorsal horn are crucial for the generation of LTP-like changes in neuronal excitability in deep dorsal horn neurons and this is modulated by descending 5HT3-R-mediated facilitatory controls. Hippocampal LTP is associated with early expression of the immediate-early gene zif268 and knockout of the gene leads to deficits in long-term LTP and learning and memory. We found that spinal LTP is also correlated with increased neuronal expression of zif268 in the superficial dorsal horn and that zif268 antisense treatment resulted in deficits in the long-term maintenance of inflammatory hyperalgesia. Our results support the suggestion that the generation of LTP in dorsal horn neurons following peripheral injury may be one mechanism whereby acute pain can be transformed into a long-term pain state. |
Phenylketonuria (PKU) is caused by deficiency of phenylalanine hydroxylase, resulting in an accumulation of phenylalanine in brain tissue and cerebrospinal fluid of phenylketonuria patients. Phenylketonuria is neuropathologically characterized by neuronal cell loss, white matter abnormalities, dendritic simplification, and synaptic density reduction. The neuropathological effect may be due to the "toxicity" of the high concentration of phenylalanine, while the underlying mechanism remains unclear. In this study, we found that cultured cerebral cortical neurons underwent mitochondria-mediated apoptosis when exposed to phenylalanine. We further demonstrated that phenylalanine induced RhoA activation. Phenylalanine also promoted myosin light chain (MLC) phosphorylation, which might be the result of the activation of Rho-associated kinase (ROCK). The RhoA antagonist, C3 transferase (C3), Rho-associated kinase specific inhibitor, Y-27632, and the overexpression of either dominant negative RhoA or dominant negative Rho-associated kinase inhibited phenylalanine-induced caspase-3 activation and rescued neurons from apoptosis, indicating that the RhoA/Rho-associated kinase signalling pathway plays an important role in phenylalanine-induced neuronal apoptosis. |
The CC2D1A/Freud-1 gene has recently been linked to non-syndromic mental retardation and a short isoform of mouse Five prime REpressor Under Dual repression binding protein 1 (Freud-1) can repress the serotonin-1A (5-HT1A) receptor gene in rodent cells. In this study, we addressed the expression, localization and regulation of the human 5-HT1A receptor gene by a long isoform of human Freud-1 protein (Freud-1L). We show that human CC2D1A/Freud-1 RNA is expressed in brain and peripheral tissues and encodes short and long isoforms, which differ by an upstream in-frame translational start site. Whereas previous studies identified the short isoform of Freud-1 as the predominant isoform in rodent cells, we demonstrate that the long isoform is more abundant in human cells, especially in the nuclear fraction. The nuclear localization of Freud-1L was enriched upon inhibition of chromosome region maintenance 1/exportin 1-dependent nuclear export, indicating a dynamic regulation of Freud-1 nuclear localization. Consistent with a functional role in the nucleus, human Freud-1L bound specifically to its dual repressor element in the 5-HT1A receptor gene in vitro and repressed transcription from these sites. Importantly, chromatin immunoprecipitation using antibodies specific for human Freud-1L demonstrated that it is bound to the dual repressor element in chromatin, indicating a functional role in regulating the basal expression of the 5-HT1A receptor gene. Taken together, these results indicate that both the short and long isoforms of Freud-1 are expressed, although Freud-1L is the major isoform that regulates the human 5-HT1A receptor gene. Disruption of transcriptional regulation by mutation of Freud-1 may play a role in abnormal brain function leading to mental retardation. |
Using whole-cell patch-clamp recordings from spinal cord slices of young (10-15 days old) rats, we have characterized and compared the properties of inhibitory synaptic transmission in lamina II and laminae III-IV of the dorsal horn, which are involved in the processing of nociceptive and non-nociceptive sensory information, respectively. All (100%) of laminae III-IV neurons, but only 55% of lamina II neurons, received both gamma-aminobutyric acid (GABA)ergic and glycinergic inputs. The remaining 45% of lamina II neurons received only GABAergic synapses. Neurons receiving only glycinergic synapses were never observed. Among the 55% of lamina II neurons receiving both GABAergic and glycinergic inputs, all displayed a small proportion (approximately 10%) of mixed miniature inhibitory postsynaptic currents (mIPSCs), indicating the presence of a functional GABA/glycine co-transmission at a subset of synapses. Such a co-transmission was never observed in laminae III-IV neurons. The presence of mixed mIPSCs and the differences in decay kinetics of GABAA-type receptor mIPSCs between lamina II and laminae III-IV were due to the endogenous tonic production of 3alpha5alpha-reduced steroids (3alpha5alpha-RS) in lamina II. Stimulation of the local production of 3alpha5alpha-RS was possible in laminae III-IV after incubation of slices with progesterone, subcutaneous injection of progesterone or induction of a peripheral inflammation. This led to the prolongation of GABAergic mIPSCs, but failed to induce the appearance of mixed mIPSCs in laminae III-IV. Our results indicate that, compared with lamina II, inhibitory synaptic transmission in laminae III-IV is characterized by a dominant role of glycinergic inhibition and the absence of a functional GABA/glycine co-transmission. |
The basal ganglia (BG) are involved in numerous neurobiological processes that operate on the basis of wakefulness, including motor function, learning, emotion and addictive behaviors. We hypothesized that the BG might play an important role in the regulation of wakefulness. To test this prediction, we made cell body-specific lesions in the striatum and globus pallidus (GP) using ibotenic acid. We found that rats with striatal (caudoputamen) lesions exhibited a 14.95% reduction in wakefulness and robust fragmentation of sleep-wake behavior, i.e. an increased number of state transitions and loss of ultra-long wake bouts (> 120 min). These lesions also resulted in a reduction in the diurnal variation of sleep-wakefulness. On the other hand, lesions of the accumbens core resulted in a 26.72% increase in wakefulness and a reduction in non-rapid eye movement (NREM) sleep bout duration. In addition, rats with accumbens core lesions exhibited excessive digging and scratching. GP lesions also produced a robust increase in wakefulness (45.52%), and frequent sleep-wake transitions and a concomitant decrease in NREM sleep bout duration. Lesions of the subthalamic nucleus or the substantia nigra reticular nucleus produced only minor changes in the amount of sleep-wakefulness and did not alter sleep architecture. Finally, power spectral analysis revealed that lesions of the striatum, accumbens and GP slowed down the cortical electroencephalogram. Collectively, our results suggest that the BG, via a cortico-striato-pallidal loop, are important neural circuitry regulating sleep-wake behaviors and cortical activation. |
Subsets and Splits