Closed-loop neuromodulation in a mouse model of temporal lobe epilepsy

Abstract

Standard treatment options for temporal lobe epilepsy (TLE) are restricted to pharmacological therapies with limited efficacy across patients, and high-risk surgical resection that many patients do not qualify for. Further understanding of neuronal circuitry in epilepsy and targets for seizure-control are needed. Using a mouse model of chronic TLE and closed-loop seizure-intervention, I investigated two potential targets for hippocampal seizure-suppression. In one set of experiments, I used optogenetics to activate a novel inhibitory cell population, LINCs, at the onset of hippocampal seizures. LINCs provide strong, widespread inhibition in healthy tissue, and were hypothesized to provide robust seizure-control. Unexpectedly, meaningful LINC-mediated seizure-suppression was not observed, with either of two sets of light stimulation parameters tested. Additionally, a substantial reduction of LINCs was observed in epileptic mice, compared to saline-injected controls. This effect was observed even when the initial site of insult (e.g. KA-injection) was outside of the hippocampus, in intra-amygdala kainate mice. Thus, LINCs appear to be vulnerable to epileptic insult. To determine whether long-term circuitry changes observed in TLE patients and the IHKA model, such as axonal sprouting, would influence the effect of LINC-activation, on-demand testing continued once-a-month, for up to six months post epilepsy-induction. Across all time points substantial LINC-mediated seizure-suppression was never observed, indicating that LINCs remain an ineffective target for seizure-control, despite providing strong inhibition in healthy tissue. Electrical stimulation of the cerebellum has provided some promising results, including seizure-freedom in some patients, but mixed results overall in epilepsy patients and models has tempered interest in the cerebellum as a target for seizure-control. In a second set of experiments, I tested whether stimulation parameters are a key factor of effective cerebellar-mediated seizure-intervention. I used closed-loop electrical stimulation of the cerebellar cortex and Bayesian optimization with Gaussian process regression to efficiently search over one thousand unique parameter sets in data-driven manner, and identify effective stimulation parameters for cerebellar-mediated seizure-control. In IHKA mice with either one of two electrode orientations and across four fitting conditions, Bayesian optimization identified parameter sets of relatively high frequency (~256 to 512 Hz) and charge amplitude (~50 to 75 nC). In the majority of animals, pulse width did not appear to be a critical factor, although relatively short (~140-220 µs) were favored at the group-level across experimental conditions. When applied in a closed-loop fashion, optimized stimulation of the cerebellum significantly reduced hippocampal seizure duration, up to 96%, compared to an internal control of no stimulation. Thus, my thesis work provides evidence that the cerebellum should be reconsidered, as it can be a powerful target for seizure-control if the correct parameters are used. Our approach using Bayesian optimization was able to identify effective parameters, and could be applied for data-driven parameter selection in other contexts.