Abstracts

Rationale: Seizures are unpredictable events that are rare and frequently clustered at the onset of epilepsy. Ability to reliably predict seizures would improve patients' quality of life and make acute treatments feasible. To this end we must better understand pathophysiology of ictogenesis. To test whether there are distinct network states preceding seizures we used a computational model of posttraumatic epileptogenesis that displays both interictal and ictal activity Methods: We used computational model of hippocampal area to study ictogenesis in post-traumatic epilepsy. The model contains 22374 of integrate'n'fire neurons. They are arranged into two layers of 105 x 105 pyramidal cells and one layer of 18 x 18 interneurons with GABAergic outputs. Connections between neurons were stochastically generated accounting for distance between cells and for the neuron type. Traumatic cell loss was equally and randomly distributed between both pyramidal cells and interneurons by removing a subset of cells and their corresponding inputs and outputs from the network. Recovery after injury was modeled by recurrent axon sprouting and synaptogenesis. New synaptic connections between surviving cells were randomly generated using various strategies. Sprouting of new connections continued until surviving cells reconstituted the original number of synapses onto and from each neuron Results: Result of traumatic cell death was strong decrease in the level of spontaneous activity. Outcome of recovery varied depending on chosen sprouting strategy. Some networks' spontaneous activity returned to the control level, some developed bursts of synchronous activity that increased in frequency, while other developed seizures comprised of sustained (tonic) ictal-like activity that slowly transitioned into intermittent (clonic) activity. We found that recurrent synaptic connections between pyramidal cells were necessary but not sufficient for ictallike activity. In networks with identical connectivity transition from bursting to ictal activity depended on initial variance, but not the mean level of depression at individual glutamatergic synapses. Seizure termination occurred as a spontaneous transition from tonic to clonic activity and eventually back to periodic synchronous bursting. These transitions were not the result of global synaptic depression, but depended instead on changes in the network-wide variance of synaptic depression. We found relatively that a small group of cells and synapses was responsible for dramatic changes in the mode of activity of the whole network. The variation in synaptic depression was most critical for network activity Conclusions: Seizures may be difficult to predict because a relatively small subset of cells and synapses strongly influence network synchronous activity. Thus initial buildup may be difficult to measure until it is too late. Our findings do not suggest a ready path to seizure prediction but are much more promising for prevention: ictogenesis could be disrupted by modifying or disabling the synaptic connectivity of only small subset of the most active cells in the network