Abstracts

Rationale: Interictal spikes (IIS) are initiated by a rapid switch in the mode of network activity to pathologically synchronous activity. Understanding the network conditions that lead to this switch could lead to better seizure prediction as well as treatments of epilepsy, but such studies are not yet experimentally tractable. We therefore ran simulations on a large-scale computational model of the CA3 hippocampal network, stepping backward in time from the moment of spike initiation to assess the conditions that predispose to synchronization.Methods: The large-scale CA3 computer model included 10,000 pyramidal cells and 1000 interneurons with dense synaptic connections. Connectivity between cells was random and decreased with distance. The network had real edges. The model was designed as a scale-free network including a limited number of hub cells with extra strong synaptic inputs and outputs to other cells. Both excitatory and inhibitory synapses included activity-dependent modulation of transmitter release probability and spontaneous release. We examined the fraction of active interneurons and GABA synapses, the ratio of active excitatory and inhibitory synapses, and the anatomic pattern of activity during the transitions to synchronous activity. Control conditions were defined in nonepileptic networks, and then the networks were rendered epileptic by instantiating one of the following conditions: neuronal death and subsequent homeostatic axon sprouting, reductions in GABAergic inhibition, and action potential propagation speed.Results: The 10:1 ratio of pyramidal cells to interneurons creates the possibility of saturation of GABAergic output at sufficiently high levels of network activity. When GABAergic strength was reduced, saturation of GABAergic outputs was a common condition leading to network-wide synchronization. However, not all episodes of synchronization are preceded by saturation of GABA output, and for some ictogenic network conditions, no synchronizations are preceded by such saturation. For example reduction in action potentials propagation speed (a model of SCN1A mutations) had much stronger effects on inhibition than on excitation because interneurons have longer axons (interneurons are sparsely distributed), resulting in delayed inhibitory responses. Delays in inhibitory responses also appear to underlie synchronization after cell loss and sprouting.Conclusions: Conditions leading to synchronization can be unexpectedly complex, even in simple network models. This is especially true of conditions that arise in the models of chronic epilepsy after cell loss and sprouting, and epilepsy arising from genetic abnormalities in sodium channel function. The complexity of the synchronization conditions are likely to be increased in larger, 3 dimensional networks. We will test whether the efficacy of periodic stimulation may arise from reductions in the probability of the conditions that permit synchronization.