Abstracts

Rationale: The hallmark of temporal lobe epilepsy (TLE) is hippocampal sclerosis (HS), a lesion characterized by cell loss and gliosis [1].  Increasing evidence from morphometry and connectivity studies suggests that anomalies often extend beyond to impact large-scale networks [2]. To assess whether the severity of HS pathology is mirrored in whole-brain network anomalies, we combined diffusion connectomics with high-resolution hippocampal subfield analysis in a cohort of TLE patients with variable degrees of histologically-verified HS.  Methods: We studied 44 drug-resistant TLE patients and 25 age- and sex-matched healthy individuals. All patients had been operated at the time of study, and histological exam revealed marked hippocampal cell loss and gliosis in 24 (TLE-HS) and isolated gliosis in 20 (TLE-G). Patient subgroups had similar age, sex, age at seizure onset, and duration of epilepsy. Hippocampal subfield segmentations were based on high-resolution T1-weighted and co-registered T2-weighted data at 3T. For hippocampal in-vivo phenotyping, we applied surface-based techniques that measure atrophy (a proxy for cell loss) and T2 hyperintensity (indexing gliosis) along individual subfields [3]. Connectomes were derived from whole-brain diffusion MRI tractography.  We assessed group differences in network topology using graph theory (Fig. 1A). Moreover, we used a recent network control theoretical framework to calculate average controllability, which captures whether a given area could drive the network’s entire dynamics [4]. Multivariate analysis evaluated the modulation of network markers (clustering, path length, controllability) by in-vivo markers of hippocampal subfield pathology. Finally, we employed a geodesic distance mapping paradigm to assess associations between network alterations and spatial proximity to the hippocampus.  Results: We observed gradual alterations of network topology in patients compared to controls (Fig. 1B), with marked increased path length and decreased clustering (indicative of reduced global/local efficiency) in TLE-HS (p < 0.001) and subtle changes in TLE-G (p < 0.1). Group comparison of overall controllability showed also a gradient (Fig. 1C), with more severe decreases in TLE-HS (p < 0.001) than TLE-G (p < 0.05). Regional analysis indicated controllability reductions in posterior default mode regions known to be densely connected to the mesiotemporal lobe in TLE-HS, while changes did not survive correction for multiple comparisons in TLE-G. Across the TLE cohort, multivariate analysis revealed a consistent relation between network anomalies and subfield markers of abnormal volume/T2 intensity, particularly in ipsilateral CA1-3 (FDR < 0.1; Fig. 2A). Distance-based mapping demonstrated that network-level disruptions are most marked in the proximity of the hippocampus, tapering off with increasing distance (Fig. 2B).  Conclusions: Combined high-resolution analysis of hippocampal subfields and whole-brain connectomics provides evidence for a positive association between the severity of microscopic tissue anomalies and structural network configurations in TLE. Hippocampal pathology may exert a cascading impact on large-scale networks, the severity of which depends on the degree of pathology, likely via retrograde degeneration.  References:[1] Bluemcke et al. 2013, Epilepsia, 54(7): 1315-29 [2] Bernhardt et al. 2015, Epilepsy Behav, 50: 162–170[3] Bernhardt et al. 2016, Ann Neurol, 80(1): 142-53.[4] Gu et al. 2015 Nat Commun, 6: 8414. Funding: Canadian Institutes of Health Research (CIHR) and Fonds de la reserche du quebec - santé (FRQS)