Abstracts

Rationale:
The broad impact of epilepsy on the whole brain increasingly demands approaches tapping into network organization. Structural covariance analysis taps into across-subject correlations of MRI-derived morphological measures, and have been shown to be sensitive to maturational, genetic, and disease-related processes of macroscale brain organization. While previously applied to describe topological alterations in samples with epilepsy compared to controls, prior work has generally been restricted to small samples, data from single sites, or restricted to specific syndromes. This multi-site ENIGMA-Epilepsy study assessed large-scale network organization in hundreds of patients diagnoses with common focal and generalized epilepsies. Here, we fingerprinted covariance network topology using network neuroscience approaches, and investigated associations to epilepsy-related gene co-expression patterns.
Method:
Participants. We included two patient cohorts with site-matched healthy controls: temporal lobe epilepsy with neuroradiological evidence of hippocampal sclerosis (TLE; 15 sites, nHC/TLE=1,311/717, 330 right-sided focus) and idiopathic generalized epilepsy (IGE; 10 sites, nHC/IGE=973/309). Covariance networks. Cortical thickness data were measured across 68 brain regions, harmonized across sites, and corrected for age, sex, and intracranial volume using ComBat. Site- and cohort-specific covariance networks were computed from cortical thickness correlations (F1A). Global and nodal network properties. We computed normalized clustering coefficient (?; an index of local network efficiency) and characteristic path length (λ; an index of global network efficiency) in controls, TLE, and IGE. Whole-brain (across all densities) and nodal (set at a density of 10%) network parameters in patients were compared to controls across sites via multivariate linear models. Transcriptomic associations. We performed spatial correlations between a predefined set of epilepsy-related gene expression maps1,2 and multivariate topological alterations.
Results:
Comparing TLE patients to controls, we observed global increases in clustering coefficient and decreases in path length (p < 0.05). Node-level differences and multivariate changes, however, revealed a more regularized, “lattice-like,” arrangement of bilateral temporo-parietal cortices (increased ? and λ) and small-world properties (increased ? and decreased λ) in ipsilateral mesiotemporal and limbic cortices (F1B). Conversely, when compared to controls, IGE showed global decreased clustering and path length (p < 0.05), suggesting a randomized network configuration (F1C). This pattern was consistent across fronto-parietal cortices, while sensorimotor and mesiotemporal cortices exhibited small-world attributes. Moreover, we found significant associations between the spatial patterns of regional topological alterations and the expression of epilepsy-related genes in TLE (F2A) and in IGE, albeit to a lesser extent (F2B).
Conclusion:
Our international multi-site network analyses unveiled marked disruptions of macroscale brain systems across the common epilepsies, suggestive of network regularization in TLE and network randomization in IGE. Transcriptomics provided neurobiological context, linking these in vivo connectome reconfigurations to potential molecular risk factors. 1Abou-Khalil et al., 2018. Nat Commun, 9:1-15. 2Hawrylycz et al., 2012. Nature, 489:391-399.
Funding:
:FRQS, CIHR, NIH