Abstracts

Epileptic EEGs of mesial temporal lobe onset seizures are characterized by a period of prominent organized rhythmic activity of 5 - 8 Hz with a subsequent monotonic decline in frequency, lasting typically less than 60 seconds. In our previous work we have described neural network models of epileptic networks capable of reproducing various rhythmic synchronous bursting activity of 3 - 8 Hz. The objective of the present study was to test the hypothesis, whether changes in the rate of intracellular calcium clearance from neurons can results in changes in the frequency of simulated rhythmic synchronous bursting activity in these networks.
We use a spatially distributed array of interconnected sub-networks of neurons capable of reproducing synchronized bursting activity as a model of epileptic seizures involving multiple adjacent regions of brain. The network was activated by random input (spike train derived from Poisson process). In order to simulate the repetitive rhythmic activity, the strengths of inhibitory synaptic weights in sub-networks have been gradually decreased resulting in recurrent bursts in neurons. After the onset of simulated bursting activity, the value of the intracellular calcium removal rate in neurons was decreased. The local field potential (LFP, an average membrane potential of all neurons in the sub-networks) was calculated.
In these neural network models, induced recurrent rhythmic activity results in higher than normal calcium influx into neurons, which slows the rate at which calcium is cleared from neurons. Decreasing the rate of calcium clearance from neurons results in decreases of the frequency of synchronous bursts in simulated network activities. The time-frequency decompositions of simulated LFPs reveal prominent rhythms with a monotonic decline in frequency from 7.5 to 2.5 Hz. The patterns of changes of the frequency in simulated LFPs are consistent with the patterns of changes of the dominant frequency observed in recordings from depth electrode contacts nearest the region of mesial temporal seizure onset in humans. The dominant frequency of the simulated LFP is correlated with the value of the calcium clearance rate.
The neural network models studied here allow one to modify cellular characteristics to attempt to understand what changes at the cellular level could produce the network behavior observed during seizures. In these network models the rate of intracellular calcium clearance from neurons determines the frequency of simulated synchronous bursting activity. Simulated reductions in the rate of calcium clearance from neurons produce decreases in the frequency of recurrent rhythmic activity in networks. Based on this mechanism, network models can reproduce characteristics of the repetitive synchronous neuronal bursting that are consistent with patterns of organized rhythmic activity occurring in seizures of mesial temporal lobe onset.
[Supported by: Epilepsy Foundation and NIH NS 38958]