Thalamic Neuromodulation Drives Disordered Sleep Networks in Epilepsy

Introduction: The relationship between epilepsy and sleep is bidirectional and complex. Nocturnal seizures disrupt sleep physiology, while sleep disruptions portend higher next-day seizure risk. Rapid eye movement (REM) sleep rebalances excitation-inhibition, which is believed to lower seizure risk [1]. In epilepsy, N3 sleep is the stage with the highest network synchrony and highest rate of nighttime seizures [2]. Mesial temporal lobe epilepsy (MTLE) is known to produce seizures which preferentially damage subcortical structures leading to cascading interictal neurocognitive deficits. These same structures conduct the sleep stages. Thus, we hypothesized MTLE pathophysiology may similarly damage subcortical structures which orchestrate sleep networks. We analyzed stereo electroencephalography (SEEG) recordings from patients receiving epilepsy surgery and compared networks from patients with MTLE against those of lateral temporal lobe epilepsy (LTLE) and other focal epilepsies (ETLE).

Methods: We studied recordings from 83 patients’ (32 MTLE, 22 LTLE, 31 ETLE) seizure free nights at Vanderbilt EMU. We automatically staged N2, N3, and REM. For each sleep stage, we measured synchrony via phase locking values (PLV). To account for heterogenous implant locations, we grouped contacts into frontal, parietal, temporal, occipital, limbic, and Rolandic regions. We compared regional PLV values across epilepsy types during each sleep stage per spectral band using the Kruskal-Wallis test and used Mann-Whitney U for pairwise post hoc comparisons between groups. For patients with thalamic implants (12), we computed phase amplitude coupling (PAC) between thalamic nuclei (anterior, centromedian, and pulvinar) and the predefined cortical networks.

Results: MTLE patients demonstrate significant global alpha band PLV modulation from N3 to REM whereas patients with MTLE and ETLE epilepsies do not (p=.0011, Fig 1A). When stratifying by subnetworks, frontal network synchrony is higher in patients with LTLE compared to patients with MTLE and ETLE (p=.0282-.0489, Fig 1B). MTLE connectivity is higher in alpha during N3 sleep for the Rolandic regions compared to ETLE (p=.0086, Kruskal Wallis omnibus test with Mann Whitney U post-hoc comparisons, Fig 2A). PAC between the anterior thalamic nucleus and Rolandic network demonstrates above-random modulation of high frequency cortical activity via low frequency thalamic oscillations (Fig 2B).

Conclusion: Lower REM connectivity in MTLE may be evidence of disordered network transitions from N3 sleep. Furthermore, hypersynchrony in N3 Rolandic networks may be driven by high thalamic PAC in N3, implicating thalamic activity as a driver of seizure risk during N3 sleep. Thalamic neuromodulation may correct these network anomalies and reduce nighttime seizure burden in patients with temporal lobe epilepsies.