Hyperconnectivity in juvenile myoclonic epilepsy: a network analysis.

OBJECTIVE: Juvenile myoclonic epilepsy (JME) is a common idiopathic (genetic) generalized epilepsy (IGE) syndrome characterized by impairments in executive and cognitive control, affecting independent living and psychosocial functioning. There is a growing consensus that JME is associated with abnormal function of diffuse brain networks, typically affecting frontal and fronto-thalamic areas.
METHODS: Using diffusion MRI and a graph theoretical analysis, we examined bivariate (network-based statistic) and multivariate (global and local) properties of structural brain networks in patients with JME (N = 34) and matched controls. Neuropsychological assessment was performed in a subgroup of 14 patients.
RESULTS: Neuropsychometry revealed impaired visual memory and naming in JME patients despite a normal full scale IQ (mean = 98.6). Both JME patients and controls exhibited a small world topology in their white matter networks, with no significant differences in the global multivariate network properties between the groups. The network-based statistic approach identified one subnetwork of hyperconnectivity in the JME group, involving primary motor, parietal and subcortical regions. Finally, there was a significant positive correlation in structural connectivity with cognitive task performance.
CONCLUSIONS: Our findings suggest that structural changes in JME patients are distributed at a network level, beyond the frontal lobes. The identified subnetwork includes key structures in spike wave generation, along with primary motor areas, which may contribute to myoclonic jerks. We conclude that analyzing the affected subnetworks may provide new insights into understanding seizure generation, as well as the cognitive deficits observed in JME patients.

Introduction

Juvenile myoclonic epilepsy (JME) is the most common idiopathic (presumed genetic) generalized epilepsy (IGE) syndrome and represents 5–10 % of all epilepsies (Berg etal., 2010). It is characterized by an age-related onset of upper limb myoclonic seizures in the mid-teens, followed in the majority of cases by generalized tonic–clonic seizures (Janz, 1985). There is increasing evidence of cognitive dysfunction in these patients, with deficits reported on tests of frontal lobe function (Wandschneider etal., 2012).

Brain structure in JME appears normal to visual inspection on routine clinical MRI; though a substantial literature reports quantitative differences in morphometric and diffusion MRI parameters in patients with JME compared to controls, (Kim etal., 2012; O'Muircheartaigh etal., 2012; Vollmar etal., 2012; Vulliemoz etal., 2011; Vollmar etal., 2011) and for review see Seneviratne etal. (2014). Relevant to our study, Kim etal. (2012) showed a reduced fractional anisotropy and increased mean diffusivity in the white matter of the frontal lobe and corpus callosum. Previous studies have focused on a region of interest approach around frontal motor and pre-motor areas, rather than an unbiased whole brain analysis (Kim etal., 2012; O'Muircheartaigh etal., 2012; Vollmar etal., 2012; Vulliemoz etal., 2011; Vollmar etal., 2011). As brain function depends on the coherent activity of widely distributed networks, such an approach may limit the structural extent of changes in JME.

The current study aimed to characterize structural network alterations in JME at a multivariate (Rubinov and Sporns, 2010) (complex network characteristics) and bivariate level (Zalesky etal., 2010) (connectivity between pairs of regions). Using this multilevel approach, we: (1)describe structural connectivity alterations in 34 patients with JME; and (2)establish whether the patterns of altered connectivity are associated with the cognitive impairments seen in JME patients.

Methods

Participants

Patients were recruited from a specialist epilepsy clinic under the supervision of one of the authors (KH). Diagnosis was based on clinical history and an EEG showing generalized spike wave discharges. Visual inspection of routine MRI was normal in all patients. Demographic and neurologic variables are provided in Table 1. Detailed neuropsychometric testing was available for 14 cases (indicated by an asterisk in supplementary table 1), including a full-scale IQ.

Table 1

Summary of demographic characteristics.

Demographic data	JME patients (n = 35)	Controls (n = 35)
Age (years)	26.8 ± 7.8	28.5 ± 7.7
Sex (F:M)	25:10	25:9
Seizure semiology	MJ 82%Abs 60%GTCS 94 %	−
Age at onset	15 ± 3.5	−
EEG	PPR 34 %GSW 100%	−
Duration of epilepsy	15.2 ± 8.8	−
AED	Monotherapy 43%LEV 49%, VPA 43%, LTG 26%, ZNM 17%, TPM 14%, CLB 14%

Abbreviations: PPR — photoparoxysmal response, GSW — generalized spike wave, LEV — levetiracetam, VPA — sodium valproate, LTG — lamotrigine, ZNM — zonisamide, TPM — topiramate, CLB — clobazam.

Standard protocol approval and patient consents

The study was approved by the NHS Research Ethics and local Research and Development Committees. Written informed consent was obtained from all participants.

MRI data acquisition

We acquired 3 T MRI (General Electric, Signa HDx) T1-weighted and diffusion-weighted data for 35 patients (one was excluded from further analysis due to poor data quality) with JME and 34 healthy controls. Diffusion-weighted data were acquired using a cardiac-gated echo-planar imaging sequence using the following parameters (Jones and Leemans, 2011): 60 slices, slice thickness 2.4 mm, echo time (TE) 87 ms, number of diffusion directions 30 (and three non-diffusion weighted scans using an optimized gradient vector scheme), b-value 1200 s/mm2, field of view (FOV) of 230 × 230 mm2, and acquisition matrix 96 × 96 mm2. High resolution T1-weighted data were acquired using magnetization prepared rapid gradient echo (MPRAGE; TR = 2300 ms, TE = 2.98 ms, 1 × 1 × 1.1 mm3 voxels, FOV: 240 × 256 mm2, 160 sagittal slices).

Neuropsychometric testing

Fourteen patients underwent a detailed assessment of frontal lobe cognitive and executive function as part of a separate study (Walsh etal., 2014) including (1)the Wechsler Adult Intelligence Scale (WAIS III), which provides IQ data; (2)the Wechsler Memory Scale (WMS III), a test of immediate and delayed memory; (3)the Delis–Kaplan Executive Function System (D-KEFS), which is an executive function test battery (Delis et al., 2001); (4)four elements from the BADS (Behavioral Assessment of the Dysexecutive Syndrome) which aims to use real world tests of executive function; and (5)the Boston Naming Test (BNT), which measures the naming of common and rare pictures.

Four D-KEFS subtests were administered: (a)the Color-Word Interference Test (CWIT, inhibition), which is a variant of the traditional Stroop procedure; (b)the Trail Making Test (TMT, cognitive flexibility), (c)Verbal fluency (VF, fluency), sometimes called the ‘FAS’ task, and (d)the D-KEFS Proverbs test that consists of eight sayings (5 high frequency or ‘common sayings'; 3 low frequency or ‘uncommon’ sayings), assessing one's ability to interpret nonliteral language. The common proverbs are high-frequency sayings with which most people are likely to be familiar; whereas the uncommon proverbs are comprised of lower frequency and less familiar sayings. Response options include both a free inquiry phase and multiple-choice phase. All tasks were scored in accordance with the manual and the participant's performance was statistically compared to the published normative data for these tests using a one sample t-test.

Diffusion MRI data processing and network construction

The diffusion MRI data were analyzed and processed in ExploreDTI. For each data set the diffusion-weighted MRI images were corrected for subject motion and eddy-current induced geometrical distortions (Leemans and Jones, 2009). During this processing step, we adjusted the B-matrices with the appropriate reorientations and included the required signal intensity modulation with the Jacobian determinant of the spatial transformation (Leemans and Jones, 2009; Jones and Cercignani, 2010). The diffusion tensor was estimated using a non-linear regression procedure (Veraart etal., 2013; Basser etal., 2000). Whole-brain tractography was performed using a deterministic algorithm (Basser and Pierpaoli, 1996). Fibers were reconstructed by placing seed points on a uniform grid across the data set at 2 mm isotropic resolution and by following the main diffusion direction (as defined by the principal eigenvector) until the fiber tract entered a voxel with a FA < 0.2 or when it made a high angular turn considered to be not anatomically plausible (angle > 30°). The step size was set at 0.5 mm. The whole-brain fiber tracts were then parcellated using the automated anatomical labeling atlas (AAL), which is a commonly used atlas to derive nodes in graph theoretical analyses of neuroimaging data. Using this procedure, we obtained 116 cortical and subcortical regions (58 for each hemisphere with the cerebellum included, see Fig. 1). The AAL atlas (and labels/masks) was then registered to the MRI data using a non-linear transformation (Klein et al., 2010). All reconstructed data were visually checked for registration accuracy for each subject. We reinspected the data in three orthogonal planes to ensure that the registration has been performed correctly and that no additional artifacts have been introduced into the data. The numbers of streamlines connecting each pair of AAL regions were aggregated into a 116 × 116 connectivity matrix. Additional constraints were introduced to ensure minimal contamination from spurious streamline trajectories through gray matter: (1)Streamlines were forced to terminate in the gray matter. In other words, if only one of its endpoints reached an AAL region, then the streamline was not included in the computation of the connectivity matrix. (2)Self-connections of nodes (i.e. endpoints of a streamline residing in the same AAL region) were not included in the analyses. (3)Tracking was terminated based on the threshold criteria during fiber tract propagation (as described above) irrespective of the underlying AAL boundary. By doing so, the length of the tracts connecting two regions was not biased.

Fig. 1

Cortical and subcortical regions (58 in each hemisphere; 116 in total) as anatomically defined by a prior template image (the AAL template) in standard stereotaxic space.

Multivariate connectivity analysis: complex network measures

Single-subject connectivity matrices were binarized, whereby we only considered the existence/absence of fiber pathways. More specifically, the network edges were defined as 1 if there was at least one connection between both regions and as 0 otherwise (Hagmann etal., 2008). The topological organization of the resulting binary networks was characterized using both global and local network measures of the Brain Connectivity Toolbox (Rubinov and Sporns, 2010) (). We provide brief definitions for each of the network properties used in this study in Supplementary material.

Two-sample t-tests were used to assess the significance of any between-group differences in each of the network measures (normalized path length, normalized clustering coefficient, small worldness, global efficiency, betweenness centrality) investigated. For local efficiency, a two-sample t-test was used for each of the 116 regions. To correct for the 116 independent tests, an alpha level of 1/116 (p = 0.004) was used to declare significance for the local measures.

Bivariate connectivity analysis: network-based statistic

The network based statistic (NBS) was used to identify pairs of regions between which the structural connectivity was altered in the JME group () (Zalesky etal., 2010). This approach was applied to the nonbinarized connectivity matrices for each participant. In brief, a two-sample t-statistic was first calculated for each pair of regions of the AAL template, to test the null hypothesis of equality in the mean value of structural connectivity between groups. This was repeated independently for each pair of regions. Pairs of regions with a t-statistic exceeding a set threshold of 2.5 (reflecting a p-value of 0.01) were systematically searched for any interconnected networks that may yield evidence of a between-group difference (referred to as connected components in graph theory). In other words, topological clusters among the set of supra-threshold connections were identified. Finally, a family wise error (FWE)-corrected p value was then ascribed to each network using permutation testing. For each permutation, participants were randomly exchanged between the JME patients and controls. The NBS was then applied to the randomized data, and the size of the largest connected component was recorded. A total of 5000 permutations were generated in this manner to yield an empirical null distribution for the size of the largest connected component.

Brain–behavior associations

Dependent variables of the cognitive and executive control tests were examined against structural connectivity using nonparametric (Spearman) correlations within the JME group.

Results

Neuropsychological test results

The subgroup with psychometric test results displayed a normal full scale IQ (average = 98.6, range = 69–131) but a reduction in processing speed (Table 2). There were significant deficits in immediate and delayed visual memory (which survived Bonferroni correction), despite preserved auditory memory. Executive function deficits were also seen in verbal fluency, inhibition switching, the BADS tests and the trail making task. Performance on the naming task was significantly impaired (Table 2).

Table 2

Neuropsychometric characteristics of the 14 individuals who underwent cognitive testing. *Significant at p = 0.05, **Significant after strict Bonferroni correction.

	Mean	Range	p value	Significant
WAIS III	Index scores
Verbal IQ	98.6	(69-122)	0.3821
Performance IQ	98.5	(74-125)	0.3356
Full scale IQ	98.0	(69-131)	0.2053
Processing speed	95.9	(76-122)	0.0174	*
Working memory	97.6	(73-117)	0.1294
WMS III	Index scores
Auditory immediate	100.4	(89-123)	0.7796
Visual immediate	91.8	(75-106)	0.0001	**
Auditory delayed	103.1	(86-132)	0.0563
Visual delayed	92.1	(68-115)	0.0002	**
DKEFS	Scaled scores
Verbal fluency
Letter fluency	9.1	(4-14)	0.0135	*
Category switching	9.4	(3-15)	0.5099
Colour word interference
Verbal inhibition	9.4	(5-14)	0.5288
Inhibition switching	7.8	(3-12)	0.0424	*
Trail making task	7.5	(1-13)	0.0232	*
Proverbs
Free inquiry	10.4	(6-14)	0.6589
Multiple choice	64.5	(5-100)	n/a
BADS	Scaled scores
Rule shift	3.6	(2-4)	0.9784
Key search	3.1	(1-4)	0.5745
Zoo map	2.0	(0-4)	0.5544
Temporal judgement	1.7	(0-4)	0.5112
Total	16.0	(10.5-22.5)	0.0135	*
Boston Naming Test	Total score
Attainment (max 60)	51.7	(43-58)	0.0001	**

Using graph theoretical analysis, we showed that the WM structural networks of both groups exhibited a much higher local interconnectivity of the nodes (normalized clustering coefficient, γ ≫ 1) (JME group: mean = 3.72, SD = 0.38; control group: mean = 3.86, SD = 0.60) and an equivalent shortest path length between any pair of nodes (normalized path length, λ ≈ 1) (JME group: mean = 1.03, SD = 0.04; control group: mean = 1.03, SD = 0.05), compared with the matched random networks. The small-worldness (σ = γ/λ) calculated from these indices was also larger than 1 (JME group: mean = 3.61, SD = 0.44; control group: mean = 3.77, SD = 0.85). Furthermore, these three metrics did not differ between patients and healthy controls (all p's > 0.10). In summary, JME patients displayed prominent small-world values close to the values of the brain network of healthy controls.

Structural network analysis estimated for patients and controls revealed that both groups exhibit hubs. In particular, 21 hub regions were shared by both groups, including the precuneus, superior parietal gyrus, postcentral gyrus, superior frontal gyrus, superior occipital gyrus, and left middle occipital gyrus. These hubs are located predominantly in regions of association cortex that receive convergent inputs from multiple cortical regions. Of note, more brain regions were identified as hub regions in the patient group. These regions comprised the left orbitofrontal gyrus, left paracentral lobule, left inferior temporal gyrus, right middle temporal gyrus, left putamen, and left cerebellar lobule Crus I, and right cerebellar lobule Crus 2.

Using two-sample t-tests, no significant between-group differences were identified for the four global network measures investigated (clustering coefficient, global efficiency, characteristic path length, and betweenness centrality) (all p's  > 0.05). The absence of these group effects suggests that global connectivity is relatively intact in JME patients. Between group-differences were identified in the nodal (region-specific) efficiency, indicating local connectivity alterations in the JME group. Specifically, compared with the controls, the nodal efficiency in the left postcentral gyrus (pcorr < 0.004) was significantly increased in JME patients.

The NBS method identified one subnetwork (p < 0.05 FWE corrected), consisting of 8 nodes and 7 connections, which demonstrated significantly increased connectivity in patients with JME compared to the control group. The subnetwork encompassed primary motor regions (left precentral gyrus), parietal cortical regions (bilateral postcentral gyrus, right precuneus), subcortical regions (left putamen, left pallidum), left cerebellar lobule IV−V, and the right hippocampus. The involved cortical and subcortical regions are shown in Fig. 2. All of the connections exhibited increased values in the patients compared with the controls (as shown in Fig. 3). No regions were found showing increased connectivity in controls relative to JME patients.

Fig. 2

The bivariate method identified one subnetwork consisting of 8 nodes and 7 connections, which demonstrated significantly increased connectivity in patients with JME (lower panel) compared to the control group (upper panel). The edge widths represent the number of tracts between nodes. The figure is made in ExploreDTI ().

Fig. 3

Group results (mean and standard deviation) of the streamline counts for JME patients and controls in each of the connections comprising the implicated subnetwork.

Possible associations between the impaired subnetwork(s)revealed by the NBS procedure and neuropsychological scores were tested within the JME group using nonparametric measures of statistical dependence. For each of the 7 significantly different connections found above (i.e. the lines in Fig. 2), we assessed whether structural connectivity of the connections was correlated with performance on the neuropsychological tests within the JME patients. Bonferroni corrections for multiple comparisons were made, hence pcorr < 0.01 was considered significant.

A significant correlation was demonstrated between the streamline count of the connection between the left postcentral gyrus and right precuneus and verbal IQ(r = 0.54, p < 0.05), WMS auditory delayed memory (r = 0.65, p < 0.01), WMS auditory recognition delayed (r = 0.61, p < 0.05), D-KEFS proverbs multiple-choice phase (r = 0.60, p < 0.05), and BNT(r = 0.71, p < 0.01), with better performance on neuropsychological tests being associated with a higher structural connectivity.

In addition, the connections between the right hippocampus and right postcentral gyrus, and right and left postcentral gyrus correlated positively with performance on elements of the D-KEFS proverbs task: the total achievement score for the multiple-choice phase (r = 0.60, p < 0.05) and free inquiry phase (r = 0.62, p < 0.05) respectively. In summary, increased structural connectivity was associated with better performance in specific cognitive subtests in JME patients.

Discussion

Whole brain functional and structural connectivity studies in epilepsy are changing our conceptualization of epilepsy from generalized or focal disorders to those of a network or system brain disorder. This is particularly true in the idiopathic generalized epilepsies, where a number of studies have shown focal and distributed structural (O'Muircheartaigh etal., 2012; Vollmar etal., 2012; Kay etal., 2013) and functional (Killory etal., 2011; Luo etal., 2011; McGill etal., 2012; Moeller etal., 2010; Zhang etal., 2011) changes, despite the ‘normal’ appearing clinical brain scans, ‘generalized’ EEG discharges and seizures. Using an unbiased, whole brain bivariate network analysis of diffusion MRI data we found a subnetwork of increased connectivity in people with JME compared to a control group. The nodes of the network comprised the primary motor cortex, precuneus, bilateral parietal /postcentral gyrus, subcortical regions and right hippocampus. Testing this in cases who had neuropsychological testing we found a significant association between this subnetwork and auditory memory, verbal fluency, and executive function tasks. We did not find any subnetworks with decreased connectivity, nor did we find any connectivity changes in other frontal regions aside from primary motor cortex.

Notably, we found precuneus involvement in the affected subnetwork and its correlation with a number of the psycho-behavioral measures. Differences in precuneus structure between JME and controls are reported by others, (O'Muircheartaigh etal., 2012; Vollmar etal., 2012; Vulliemoz etal., 2011) though to date most studies have focused more on frontal lobe changes. The precuneus, part of the association cortex, is highly connected with other association areas and subcortical structures, and has a role in a wide range of higher order cognitive functions including visuo-spatial imagery, episodic memory retrieval, self-processing and consciousness (Raichle, 2003). The precuneus makes up part of the functional ‘default mode’ network, areas of decreased activity during task directed behavior (Laufs etal., 2006), but also sleep, anesthesia (Gotman etal., 2005) and spike wave activity in IGE (Gotman etal., 2005; Hamandi etal., 2006; Vaudano etal., 2009). Functional studies suggest a key role for precuneus in generalized spike wave (GSW) discharges in IGE. Dynamic causal modeling of EEG–fMRI (Lee etal., 2014) and an application of Granger causality in EEG have shown a causal link between neural activity in the precuneus and the onset and offset of GSW discharges (Szaflarski etal., 2010). Importantly we found primary motor areas within the altered network described here, an area that would be activated during myoclonic jerks, a defining feature of JME.

Within our significant subnetwork we also found nodes in cerebellum lobules IV and V and the basal ganglia. Cerebellar activations have been reported in two EEG–fMRI studies of spike wave discharges in IGE (Hamandi etal., 2006; Li etal., 2010). A diffusion MRI study using both a voxel-based and TBSS (tract based spatial statistics) approach found a decrease of FA in patients compared to controls in AAL regions, including the right cerebellar lobule IV−V and lobule VI in IGE (Luo etal., 2012). A subsequent connectivity analysis found a negative correlation (uncorrected) between the right cerebellar lobule IV−V and right lingual gyrus on one hand and duration of epilepsy on the other hand (Luo etal., 2012). Moreover, a resting state EEG−fMRI study found increased connectivity between bilateral caudate nucleus and putamen and decreases in cerebellum and supplementary motor areas from recorded sessions that contained spike wave discharges compared to those that did not (Keller etal., 2011).

The predominant focus and findings of structural MRI studies in JME have been in frontal lobe and thalamus, using a region of interest approach (O'Muircheartaigh etal., 2012; Vollmar etal., 2012; Vollmar etal., 2011; Kay etal., 2013; Deppe etal., 2008). For example, Vollmar etal. (2011) found increased functional and structural connectivity between prefrontal ‘cognitive’ cortex and motor cortex, but used an ROI seed-based approach centered on the prefrontal areas. We only partially replicated these findings; our unbiased whole brain approach found a subnetwork around more posterior brain regions. Similarly other studies have reported structural changes in IGE outside frontal lobe regions, including the precuneus (Ronan etal., 2012; Grasby etal., 1994).

The implication of our findings is that structural connectivity changes in JME are not just seen in frontal and thalamic regions, but can also involve more widespread brain areas, and in our bivariate analyses, posterior subnetworks. We provide further support for the precuneus, through its connections and influence on brain network activity, being intricately linked to the genesis of epileptic discharges and neuropsychological performance.

The most frequently observed cognitive deficits in patients with JME are seen on tests of executive function, which supports the thalamo-fronto-cortical model of JME (Wandschneider etal., 2012). Our data confirm this correlation with poor performance across a range of tests of executive function — particularly the trail-making task. However the most striking finding is the difference between visual and auditory memory (Table 2) and the poor performance on the naming task despite intact verbal and performance IQ scores. Verbal memory is a widely distributed cognitive function which relies upon an intact left temporal and parietal lobe. A superior performance on auditory-verbal tasks has been shown to be dependent not only on the anterior and thalamic regions, but also on the cerebellar vermis and hemispheres and the precuneus (Koepp etal., 2014). Hyperconnectivity between the left postcentral gyrus and right precuneus is associated with a sparing of this verbal memory deficit. Similarly hyperconnectivity elsewhere in the right hemisphere appears to spare the left hemisphere language functions and some of the executive function tests that rely on language, such as the proverb testing. There is increasing evidence that JME is not a homogenous electro-clinical syndrome (Thomas etal., 2013; Mori etal., 1999) and so it will be important to repeat these experiments to establish whether network-based analyses can identify these subgroups.

Limitations

Despite the present findings, this study on structural network connectivity in JME patients is still preliminary, and further studies are needed. First, we observed that correlations were present between cognitive task performance and structural connectivity within the subgroup of 14 patients. Further analyses of the relationship between network parameters and performance on behavioral parameters should be carried out within larger groups of JME patients. Secondly, testing patients with JME introduces a number of methodological factors that may affect interpretation of results. JME is a heterogeneous disease with multiple deficits that evolve over time. Also, findings will likely be affected by the age of onset, antiepileptic drug (AED) treatment, and number of seizures. For example, not all our patients fulfill diagnostic criteria for JME in that they did not give a history of myoclonic jerks (N = 6, see Supplemental Table 1), which may be a limitation in diluting the homogeneity of our patient group. However, we feel that the inferences we have derived still relate primarily to JME, even with this small number of IGEsub-syndromes. Six patients are however a too small size to use for detailed subgroup analyses. In other words, our group was relatively homogeneous in terms of neuropathology. Moreover, each control was carefully selected to match the patient's demographics (age and gender). Finally, we employed a DTI based streamline tractography approach to define the edges of the structural network (Basser etal., 2000; Mori etal., 1999). This is by far the most widely applied tractography method, mainly for its simplicity, robustness and speed (Cheng etal., 2012; Griffa etal., 2013). Such a tractography method, however, is not able to resolve crossing fiber bundles (Mori and van Zijl, 2002; Tournier etal., 2011; Jones etal., 2013). Many other algorithms could be used to develop the structural network, but choosing one is not a trivial matter, because different tractography algorithms for analysis of the same imaging data can lead to subtly different graph theoretical results (Bastiani etal., 2012). Further studies should reconstruct anatomical networks with diffusion tractography methods that account for fiber crossings and are more robust (Tuch, 2004; Wedeen etal., 2008; Behrens etal., 2007; Jeurissen etal., 2011).

Conclusion

In summary we believe that this is the first whole brain graph theory analysis of diffusion MRI data in patients with JME. The study shows that network-based analysis of brain white matter connections provides a novel way to reveal structural changes within subnetworks and the structural basis of cognitive dysfunction in JME.

Contributor's statements

Caeyenberghs K: performed the graph theoretical analyses and statistical analyses, drafted the manuscript, and approved the final manuscript as submitted.

Powell HWR: conceptualized and designed the study, drafted the manuscript, and approved the final manuscript as submitted.

Thomas RH: performed the neuropsychological assessment and statistical analyses of the behavioral data, drafted the manuscript, and approved the final manuscript as submitted.

Brindley L: coordinated and carried out the data collection, revised the manuscript, and approved the final manuscript as submitted.

Church C: coordinated and carried out the data collection, revised the manuscript, and approved the final manuscript as submitted.

Evans J: coordinated and carried out the data collection, revised the manuscript, and approved the final manuscript as submitted.

Muthukumaraswamy SD: coordinated and carried out the data collection, revised the manuscript, and approved the final manuscript as submitted.

Jones DK: conceptualized and designed the study, critically reviewed the manuscript, and approved the final manuscript as submitted.

Hamandi K: conceptualized and designed the study, drafted the manuscript, and approved the final manuscript as submitted.

The following are the Supplementary data related to this article

Supplementary material. Description of the network properties.

Supplemental Table 1

Summary of demographic and clinical characteristics for the JME group.

Financial disclosures

Caeyenberghs K has no financial disclosure. The remaining authors have no financial relationships relevant to this article to disclose.

Powell HWR, Thomas RH, Brindley L, Church C, Evans J, Muthukumaraswamy SD, Jones DK, and Hamandi K report no disclosures.

Conflicts of interest

Karen Caeyenberghs has no conflicts of interest. The other authors have no conflicts of interest to disclose.

Source of funding

Support for this study was provided through a grant from NISCHR AHSC, Welsh Government to Hamandi K. Caeyenberghs K was funded by a travel grant of the Research Foundation — Flanders (FWO) (V416712N). RHT is funded by a Welsh Clinical Academic Track fellowship through NISCHR, Welsh Government.