Investigating Epileptic Seizures in Mouse and Drosophila Fly Models: Circuit Mechanisms and Behavior

Abstract

Epilepsy, which is characterized by unprovoked recurrent seizures, is a neurological disorder that affects more than 50 million people globally. Unfortunately, about one third of patients are resistant to the current anti-epileptic drugs (AEDs). Our understanding of epileptic seizures and epileptogenesis is still very limited. Epilepsy involves aberrant changes in the brain at molecular, cellular, circuitry and network levels. Researchers have developed multiple preclinical animal models to study the mechanisms of epilepsy at each level and have been working to discover new therapies. In this thesis, I use mouse and Drosophila fly models to study the circuit mechanism and behavior of epileptic seizures.

Temporal lobe epilepsy (TLE) is the most common form of focal epilepsy. Seizures in TLE initiate in temporal lobe, typically in hippocampus or surrounding areas. In Chapter 2 of this thesis, I use a mouse model of TLE – intrahippocampal kainic acid (KA) model. The model shows pathological changes and spontaneous recurrent seizures (SRSs) akin to those observed in patients with TLE. Many studies of TLE have focused on the hippocampus. Anterior piriform cortex (APC) is also a limbic area that is closely associated with TLE but is understudied. In the APC, parvalbumin-positive (PV⁺) interneurons are distributed throughout all three layers and can provide strong inhibition and maintain local excitation-inhibition balance. The fast-spiking PV neurons project onto about a thousand principal neurons, which enable them to control neuronal synchrony and regulate brain oscillations. However, whether APC^PV neurons play a causal role in seizure expression in TLE is unclear. Here, I use chemogenetics and multi-site local field potential (LFP) recording to examine the efficacy of APC^PV neuronal activation on hippocampal SRSs. I found a reduction of PV⁺ synapses and a dispersion of principal neurons in both sides of the APC with KA injection in one side of the hippocampus. Selective activation of APC^PV neurons reduced the frequency and duration of SRSs in the hippocampus. Further, by quantifying power and interregional synchronization, I found that APC^PV neuronal activation altered the brain-wide dynamics in seizure state as well as interictal state and afforded differential regulation according to brain regions and frequency bands. In conclusion, activating PV⁺ inhibitory interneurons in the APC can alleviate SRSs in the hippocampus. This long-range effect may be due to large-scale changes in epileptogenic network dynamics. In the seizure state, the decrease of power and increase of functional connectivity may contribute to cutting down the seizure durations. In the interictal state, network dynamic also changes albeit with a different pattern from seizure state, which may increase the seizure threshold and reduce the probability of seizure recurrence. These results reveal how a cortical microcircuit can alter neural activity in multiple brain regions and exert antiepileptic benefits. These findings strengthen the idea that large-scale connections from APC play a key role in maintaining the balance of the entire seizure network and thus provide the basis for future circuit-based therapies.

In addition to the rodent model, Bang-sensitive (BS) Drosophila flies also represent an important model for studying epilepsy and neuronal excitability. BS mutants show striking seizure behavior and a refractory period after mechanical shock. The brief and strong mechanical shock is termed ‘Bang’. BS seizures, akin to the electric kindling model in rodents and pharmacologically resistant epilepsy in humans, underscore their significance in epilepsy research. Our previous work identified that mutations in the julius seizure (jus) gene can induce BS seizures. However, the behavioral manifestations of the seizure phenotype of jus mutants have not been fully characterized yet. Although computational tools that leverage deep learning algorithms have been developed to track movement accurately, the subsequent behavioral sequence classification remains a challenge. In Chapter 3, I developed a novel machine learning approach for behavioral sequence classification termed LASC (Long short-term memory and Attention mechanism for Sequence Classification). This approach consists of three main steps: (1) acquire frame-by-frame motion features from the tracking data of body parts, (2) predefine behavioral syllables and create manually labelled dataset, and (3) train a LASC model. Adopting LASC on jus seizures, we achieved 85% classification accuracy for five pre-identified stages: paralysis (P), tonic seizure (T), spasm (S), clonic seizure (C), and recovery episode (R). Different jus alleles showed a shared repertoire of stages in seizure. Furthermore, jus mutants followed a consistent, stereotypical stage progression within a session: P, T, S, C, R. However, analysis of behavioral structures revealed that genotypes could be decoded from their distinct stage usage. Surprisingly, knocking down jus specifically in the thoracic abdominal ganglion, while leaving the brain unaffected, resulted in severe seizures indistinguishable from those observed in one of the jus alleles impacting the entire CNS. This study provides a framework for exploring the causal link between genetic mutations and precise, rapid motor behaviors. The comprehensive pipeline outlined here has the potential to enable high-throughput, behavior-based drug screening for epilepsy.

In summary, this thesis demonstrates the anti-seizure effect of APC^PV neurons and its potential network mechanism in KA mouse model and delineates the seizure behavior in jus alleles in Drosophila flies.

Date of Award	19 Mar 2025
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Chun Yue Geoffrey LAU (Supervisor) & David DEITCHER (External Co-Supervisor)