Plasticity-modulated seizure dynamics for seizure termination in realistic neuronal models

In previous studies we showed that autonomous absence seizure generation and termination can be explained by realistic neuronal models eliciting bi-stable dynamics. In these models epileptic seizures are triggered either by external stimuli (reflex epilepsies) or by internal fluctuations. This scenario predicts exponential distributions of the duration of the seizures and of the inter-ictal intervals. These predictions were validated in rat models of absence epilepsy, as well as in a few human cases. Nonetheless, deviations from the predictions with respect to seizure duration distributions remained unexplained. The objective of the present work is to implement a simple but realistic computational model of a neuronal network including synaptic plasticity and ionic current dynamics and to explore the dynamics of the model with special emphasis on the distributions of seizure and inter-ictal period durations. We use as a basis our lumped model of cortical neuronal circuits. Here we introduce 'activity dependent' parameters, namely post-synaptic voltage-dependent plasticity, as well as a voltage-dependent hyperpolarization-activated current driven by slow and fast activation conductances. We examine the distributions of the durations of the seizure-like model activity and the normal activity, described respectively by the limit cycle and the steady state in the dynamics. We use a parametric γ-distribution fit as a quantifier. Our results show that autonomous, activity-dependent membrane processes can account for experimentally obtained statistical distributions of seizure durations, which were not explainable using the previous model. The activity-dependent membrane processes that display the strongest effect in accounting for these distributions are the hyperpolarization-dependent cationic (I(h)) current and the GABAa plastic dynamics. Plastic synapses (NMDA-type) in the interneuron population show only a minor effect. The inter-ictal statistics retain their consistency with the experimental data and the previous model.