Human GABRG2 generalized epilepsy: Increased somatosensory and striatothalamic connectivity.

OBJECTIVE: To map functional MRI (fMRI) connectivity within and between the somatosensory cortex, putamen, and ventral thalamus in individuals from a family with a GABAergic deficit segregating with febrile seizures and genetic generalized epilepsy.
METHODS: We studied 5 adults from a family with early-onset absence epilepsy and/or febrile seizures and a GABAA receptor subunit gamma2 pathogenic variant (GABRG2[R43Q]) vs 5 age-matched controls. We infer differences between participants with the GABRG2 pathogenic variant and controls in resting-state fMRI connectivity within and between the somatosensory cortex, putamen, and ventral thalamus.
RESULTS: We observed increased fMRI connectivity within the somatosensory cortex and between the putamen and ventral thalamus in all individuals with the GABRG2 pathogenic variant compared with controls. Post hoc analysis showed less pronounced changes in fMRI connectivity within and between the primary visual cortex and precuneus.
CONCLUSIONS: Although our sample size was small, this preliminary study suggests that individuals with a GABRG2 pathogenic variant, raising risk of febrile seizures and generalized epilepsy, display underlying increased functional connectivity both within the somatosensory cortex and in striatothalamic networks. This human network model aligns with rodent research and should be further validated in larger cohorts, including other individuals with generalized epilepsy with and without known GABA pathogenic variants.

We previously reported an Australian family with a GABA type A receptor subunit gamma2 pathogenic variant (GABRG2[R43Q]) presenting with febrile seizures, febrile seizures plus, and absence epilepsy.[1] This family has decreased cortical inhibition, demonstrated by transcranial magnetic stimulation,[2] and decreased cortical benzodiazepine receptor binding on PET.[3] Aligning with human data, mice with the same Gabrg2[R43Q] pathogenic variant have reduced cortical inhibition shown by deficits in GABA-mediated synaptic currents in the somatosensory cortex and anatomic changes in cortical interneuron positioning.[4,5]

Although these investigations provide clues about the mechanisms underlying the GABRG2 pathogenic variant, it is unknown whether brain connectivity is affected in humans with the R43Q pathogenic variant. To study this issue, we quantified changes in functional MRI (fMRI) connectivity in individuals with this GABRG2 pathogenic variant. We focused on regions previously studied in Gabrg2 mice: the somatosensory cortex, the ventral thalamus, and the putamen (as a recent fMRI study in rats demonstrated that GABAA antagonists enhance connectivity of the somatosensory cortex and striatum[6]). Together, these 3 brain regions comprise a well-known pathway where the cortex sends excitatory signals to the putamen, which in turn exerts strong inhibitory control over the ventral thalamus via the globus pallidus and substantia nigra pars reticulata before relaying excitatory information back to the cortex.

Based on previous research in humans with the GABRG2[R43Q] pathogenic variant[2,3] and mice with the Gabrg2[R43Q] pathogenic variant,[4,5,7] we hypothesize that fMRI connectivity within the somatosensory cortex and between subcortical regions is altered in people with a GABRG2 pathogenic variant.

Methods

Participants and clinical information

We recruited 5 adults from a previously reported family with the GABRG2 [R43Q] pathogenic variant[1] (mean age [SD]: 36.4 ± 4.2 years; clinical information in the table). They were compared with 5 age- and sex-matched controls (mean age [SD] 36.8 ± 4.1 years).

Table

Clinical, EEG and imaging phenotype of individuals with the R43Q GABRG2 pathogenic variant[1]

Standard protocol, approvals, and patient consents

This study was approved by the Austin Human Research Ethics Committee, and participants gave written informed consent to participate.

fMRI preprocessing

We acquired 10 minutes of resting-state fMRI data on a Siemens 3T Skyra scanner with a voxel size of 3 × 3 × 3 mm and repetition time of 3 seconds. The fMRI data were slice-timing corrected, realigned, coregistered to T1 images, tissue segmented, spatially normalized, and filtered between 0.01 and 0.08 Hz. See reference 8 for further details about our fMRI methods.

fMRI connectivity analysis

We used 2 different estimates of fMRI connectivity. (1) Regional homogeneity: to calculate voxel-averaged local connectivity within each node of our brain network model (somatosensory cortex, ventral thalamus, and putamen). This measures Kendall W correlations between a single voxel and its 26 nearby voxels in three-dimensional space. Its values range between 0 (minimal local connectivity) and 1 (maximal local connectivity). (2) Partial correlation: to calculate voxel-averaged Pearson correlations of fMRI time series between each node pair, while regressing out indirect correlation effects from all other connection pairs. Partial correlation values range between −1 (maximal negative correlation) and 1 (maximal positive correlation).

Effect size analysis

We used Hedges' g standardized effect sizes and 95th percentile confidence intervals (95% CIs) to quantify mean differences in brain connectivity between individuals with the GABRG2 pathogenic variant and controls (as Hedges' g is recommended over Cohen's d in studies with a limited sample sizes). Hedges' g effect sizes = 0.2 (small); 0.5 (moderate); 0.8 (large); and 1.2 (very large).

Data availability

Anonymized original data will be shared by request from any qualified investigator.

Results

GABRG2: fMRI connectivity is increased within the somatosensory cortex

We observed stronger fMRI connectivity within the somatosensory cortex in all individuals with the GABRG2 pathogenic variant compared with controls (Hedges' g = 1.46; 95% CI = 0.43–2.41—figure, A). Hedges' g effect sizes were lower between GABRG2 and healthy controls for fMRI connectivity within the putamen (Hedges' g = 0.55; CI 95 = −0.85–1.84) and ventral thalamus (Hedges' g = 0.40; CI 95 = −0.96 to 1.70).

Figure

Functional connectivity within and between brain nodes in GABRG2 participants displayed by purple magenta color bars and healthy controls displayed by blue cyan color bars

Peak regional homogeneity across all voxels within the somatosensory cortex (A), putamen (B), and ventral thalamus (C), and partial correlations averaged across all voxels between the somatosensory cortex-putamen (D), somatosensory cortex-ventral thalamus (E), and ventral thalamus-putamen (F). Error bars denote the standard error across participants. There was little difference in head movement between groups (mean movement [SD] in GABRG2 = 0.12 mm ± 0.05; mean movement [SD] in controls = 0.16 mm ± 0.04). Our 3 brain regions of interest—(1) bilateral somatosensory cortex (yellow); (2) putamen (orange); and (3) ventral thalamus (red) (G)—were delineated using the NeuroSynth database (neurosynth.org) incorporating findings from a multitude of previous fMRI studies describing these brain regions and their behavioral correlates, as well as their functional connectivity patterns. *A and F had large effect size.

GABRG2: fMRI connectivity is increased between the thalamus and putamen

We observed stronger fMRI connectivity between the putamen and ventral thalamus in all individuals with the GABRG2 pathogenic variant compared with controls (Hedges' g = 1.98; 95% CI = 0.11–3.45—figure, F). It is worth noting that the GABRG2 participant with strongest connectivity also experienced seizures into adulthood (see participant 2 in the table for more clinical information). There were no large effect sizes between GABRG2 and controls for fMRI connectivity between the putamen and somatosensory cortex (Hedges' g = 0.68; 95% CI = −0.75–1.98) or the ventral thalamus and somatosensory cortex (Hedges' g = 0.26; 95% CI = −1.09 to 1.55).

GABRG2: post hoc analysis shows less fMRI connectivity changes of the visual cortex and precuneus

To test whether our primary result was specific to the somatosensory cortex, thalamus, and putamen or it reflected a “whole-brain process”, we analyzed 2 additional brain nodes: (1) primary visual cortex and (2) precuneus. We observed no large effect size in fMRI connectivity for these 2 brain regions between GABRG2 and control participants. fMRI connectivity within the visual cortex: Hedges' g = 0.45; 95% CI = −0.93 to 1.74. fMRI connectivity within the precuneus: Hedges' g = 0.25; 95% CI = −1.12 to 1.56. fMRI connectivity between the visual cortex and precuneus: Hedges' g = 0.43; 95% CI = −0.95–1.72.

Discussion

We found that 5 individuals with a GABRG2 pathogenic variant and a history of seizures (5/5 febrile seizures; 3/5 absence seizures) have increased fMRI connectivity within the somatosensory cortex and between the putamen and ventral thalamus compared with 5 healthy controls. This finding suggests a “network-specific” rather than a “whole-brain” effect, as our post hoc analysis revealed no difference in connectivity between GABRG2 and control participants of either the primary visual cortex or the precuneus.

Emerging animal research has highlighted the important role of specific brain areas in generalized epilepsy, in particular absence epilepsy where seizures are thought to originate in the somatosensory cortex[9] and are modulated by thalamostriatal circuits.[10] Although it is not straightforward to compare animal and human studies, our findings provide preliminary evidence that somatosensory cortex and subcortical structures are hyperconnected in people with a genetic predisposition to develop febrile and also absence seizures. This finding is further supported by our previous GABRG2 transcranial magnetic stimulation study—in the same family—showing neuronal hyperexcitability of the perimotor cortex.[2]

It is nontrivial to quantify whether microscale neuronal dysfunction and macroscale fMRI connectivity are related because of their vast difference in spatial scales. However, this study presented us with an opportunity to (indirectly) assess whether inhibitory neuronal dysfunction is reflected in fMRI connectivity, as we know that people with a GABRG2 pathogenic variant have abnormal inhibitory GABAergic neuronal function. We postulate that our fMRI connectivity findings are related to inhibitory neuronal abnormalities, especially hyperconnectivity between the thalamus and the putamen, as the latter brain region consists almost exclusively of medium spiny inhibitory GABAergic neurons.

Our small sample size is a consequence of the difficulty of recruiting a single family with a genetically homogenous disorder to the demands of an imaging study that requires travel and attendance at a single site. Despite this limitation, our preliminary findings give insight into network changes that may underlie human genetic epilepsy caused by a GABRG2 pathogenic variant, and they align with results from the Gabrg2 animal model, which also shows increased activity of the somatosensory cortex.[4] Nevertheless, our human network model based on this family with a GABRG2 variant should be further validated in larger generalized epilepsy cohorts with and without known GABA pathogenic variants.