Disrupted basal ganglia-thalamocortical loops in focal to bilateral tonic-clonic seizures

Abstract

Focal to bilateral tonic-clonic seizures are associated with lower quality of life, higher risk of seizure-related injuries, increased chance of sudden unexpected death, and unfavourable treatment outcomes. Achieving greater understanding of their underlying circuitry offers better opportunity to control these seizures. Towards this goal, we provide a network science perspective of the interactive pathways among basal ganglia, thalamus and cortex, to explore the imprinting of secondary seizure generalization on the mesoscale brain network in temporal lobe epilepsy. Specifically, we parameterized the functional organization of both the thalamocortical network and the basal ganglia-thalamus network with resting state functional MRI in three groups of patients with different focal to bilateral tonic-clonic seizure histories. Using the participation coefficient to describe the pattern of thalamocortical connections among different cortical networks, we showed that, compared to patients with no previous history, those with positive histories of focal to bilateral tonic-clonic seizures, including both remote (none for >1 year) and current (within the past year) histories, presented more uniform distribution patterns of thalamocortical connections in the ipsilateral medial-dorsal thalamic nuclei. As a sign of greater thalamus-mediated cortico-cortical communication, this result comports with greater susceptibility to secondary seizure generalization from the epileptogenic temporal lobe to broader brain networks in these patients. Using interregional integration to characterize the functional interaction between basal ganglia and thalamus, we demonstrated that patients with current history presented increased interaction between putamen and globus pallidus internus, and decreased interaction between the latter and the thalamus, compared to the other two patient groups. Importantly, through a series of 'disconnection' simulations, we showed that these changes in interactive profiles of the basal ganglia-thalamus network in the current history group mainly depended upon the direct but not the indirect basal ganglia pathway. It is intuitively plausible that such disruption in the striatum-modulated tonic inhibition of the thalamus from the globus pallidus internus could lead to an under-suppressed thalamus, which in turn may account for their greater vulnerability to secondary seizure generalization. Collectively, these findings suggest that the broken balance between basal ganglia inhibition and thalamus synchronization can inform the presence and effective control of focal to bilateral tonic-clonic seizures. The mechanistic underpinnings we uncover may shed light on the development of new treatment strategies for patients with temporal lobe epilepsy.

Keywords: basal ganglia; focal to bilateral tonic-clonic seizures; network neuroscience; resting state functional connectivity; thalamus.

Figure 1

Schematic overview of the analytical pipeline. (A) We applied masked independent component analysis (ICA) (Moher Alsady et al., 2016) on an independent rsfMRI dataset from the Human Connectome Project (HCP, n = 100) to generate functional parcellations of the striatum (10 parcels) and thalamus (seven parcels). In addition, anatomical masks for GPe and GPi, STN, and SN were directly adopted from the ATAG atlas (Keuken et al., 2014), yielding a final parcellation scheme of the basal ganglia–thalamus network with 21 regions of interest (ROIs) per hemisphere. These regions of interest were then used to extract time series from the clinical rsfMRI data collected in this study. (B) Based on the Schaefer Atlas (Schaefer et al., 2018), we estimated thalamocortical functional connectivity between each thalamic parcel and cortical regions of interest, and then sorted them by seven predefined resting state networks (Yeo et al., 2011). We used the participation coefficient to represent the distribution pattern of thalamocortical functional connections across different resting state networks. The more uniform the distribution, the higher the participation coefficient, and vice versa. (C) We also estimated the functional connectivity matrix of the basal ganglia–thalamus network, on which we applied a community detection algorithm, to identify groups of regions of interest with higher preference for interacting with each other (i.e. communities) (Newman and Girvan, 2004; Reichardt and Bornholdt, 2006; Blondel et al., 2008). We used interregional integration to represent the probability of all the regions of interest from two different anatomical origins being assigned to the same community (i.e. allegiance) over iterative applications of this algorithm (specifically, 1000 optimizations of a modularity quality index).

Figure 2

Thalamocortical participation coefficients compared across the three patient groups. (A) A significant difference was found in the ipsilateral medial-dorsal thalamic nuclear group. (B) Atypicality was estimated in reference to data obtained from a matched healthy control group. The remote- and current-FBTCS groups presented Z-scores significantly >0, but not the none-FBTCS group. The full spectrum of Z-scores and data from healthy control group are presented in Supplementary Fig. 2. Ant = anterior; MD = medial-dorsal; LV = lateral-ventral; Post = posterior thalamic nuclear groups; res = residual after confound regression. **P < 0.01; ***P < 0.001. Statistics were obtained via a non-parametric permutation test controlling for multiple comparisons. The central mark indicates the median, and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively.

Figure 3

Pairwise interregional integration of the basal ganglia–thalamus network compared across the three patient groups. (A) Significant differences in interregional integration were found at the striatum–globus pallidus (GP) and globus pallidus–thalamus pairs. (B) No significant differences were found when comparing pairwise functional connectivity instead of integration estimates. (C) No significant differences were found in a null model in which the original functional connectivity matrix was randomly rewired, preserving both the degree and strength distributions. (D) Atypicality was estimated in reference to data from a matched healthy control group. The current-FBTCS groups presented Z-scores significantly different from 0, but the other two groups did not. The full spectrum of Z-scores and data from the healthy control group are presented in Supplementary Fig. 3. FC = functional connectivity; res = residual after confound variable regression; Str = striatum; Tha = thalamus. **P < 0.01; ***P < 0.001. Statistics were inferred with a non-parametric permutation-based method controlling for multiple comparisons. The central mark indicates the median, and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively.

Figure 4

Interregional integration with GPe and GPi compared across three patient groups. (A) No significant differences were observed at the GPe. (B) Significant differences in interregional integration were observed in putamen–GPi, GPi–lateral-ventral thalamus, and GPi–posterior thalamus pairs. (C) Atypicality was estimated in reference to data from a matched healthy control group. The current-FBTCS groups presented Z-scores significantly different from 0, but the other two groups did not. The full spectrum of Z-scores and data from the healthy control group are presented in Supplementary Fig. 4. Ant = anterior; Cau = caudate; LV = lateral-ventral; MD = medial-dorsal; Post = posterior nuclear groups; Put = putamen; res = residual after confound variable regression; Tha = thalamus; VS = ventral striatum. *P < 0.05; **P < 0.01; ***P < 0.001. Statistics were inferred with a non-parametric permutation-based method controlling for multiple comparisons. The central mark indicates the median, and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively.

Figure 5

Simulated ‘disconnection’ analyses on putamen–GPi interregional integration. To delineate specific contributions from different connections of the network, we set functional connectivity value(s) of specific connection(s) to zero (i.e. ‘disconnected’), and then re-estimated the integration. (A) Before simulation, a significant group difference in putamen–GPi integration was found. (B) No significant putamen–GPi integration difference was found when the connections from putamen to GPi (i.e. the ‘direct pathway’) were ‘disconnected’. (C) A significant difference in the putamen–GPi integration was found when the connections from the caudate (Cau) and ventral striatum (VS) to the GPi were ‘disconnected’. (D) A significant difference in the putamen–GPi integration was found when all connections from putamen to GPe, GPe to STN, STN to GPi, and GPe to GPi (i.e. the ‘indirect pathway’) were ‘disconnected’. In the schematics, the orange arrow represents the direct pathway, the blue arrows represent both the long and short routes of the indirect pathway (Smith et al., 1998), the green line represents functional connectivity, and the red dashed double head arrow represents the group difference in interregional integration. Bonferroni corrected post hoc test: **P < 0.01; ***P < 0.001.

Figure 6

Simulated ‘disconnection’ analyses on GPi–thalamus interregional integrations. To delineate specific contributions from different connections of the network, we set functional connectivity value(s) of specific connection(s) to zero (i.e. ‘disconnected’), and then re-estimated the integration. (A) Before simulation, we observed significant group differences in GPi–lateral-ventral (LV) and in GPi–posterior (Post) thalamus integration. (B) Significant integration differences were still observed when the connections from GPi to the posterior thalamic nuclei were ‘disconnected’. (C) No significant integration differences were found when the connections from the GPi to the lateral-ventral thalamic nuclei were ‘disconnected’. (D) Significant integration differences were found at the GPi–lateral-ventral but not the GPi–posterior thalamus pair when all connections between the lateral-ventral and posterior thalamic nuclear groups were ‘disconnected’. In the schematics, thick grey lines represent anatomical connections, the green line represents functional connectivity, and the red dashed double headed arrow represents the group difference in interregional integration. Tha = thalamus. Bonferroni corrected post hoc test: 0.05 < ^{^}P < 0.1; *P < 0.05; **P < 0.01; ***P < 0.001.