Abstracts

Rationale: Seizures occur without warning at seemingly random times. Previous work in an animal model of temporal lobe epilepsy has shown that firing rates of some dentate granule and CA3 pyramidal cells increase before seizure onset ('preictal' cells), others increase after seizure onset ('ictal' cells) and others do not change ('unchanged' cells). It is unclear whether seizures arise from activation of random groups of neurons or a fixed seizure-initiation network. If the latter is true, then the seizure-related behavior of individual neurons should be preserved across multiple seizures.Methods: Tetrodes were used to record action potentials from multiple, single granule cells or CA3 pyramidal cells prior to and during spontaneous seizures in pilocarpine-treated, epileptic rats 2-6 months post-status. Seizure onset time was identified electrographically using EEG data. Anatomical locations of tetrodes in the granule or CA3 pyramidal cell layers were verified histologically. Action potentials were sampled at 32 kHz, filtered between 600-6,000 Hz and associated with individual cells by standard cluster cutting techniques. Granule (N=64) or pyramidal cells (N=126) were distinguished from interneurons based on established criteria. 'Preictal' cells showed a significant increase in firing rate in the two minutes preceding seizure onset compared to a baseline period 10-5 minutes prior to seizure onset (ANOVA, p<.05). 'Ictal' cells did not meet preictal cell criteria, but showed a significant increase in firing rate in the first 30 sec after seizure onset. All other cells were labeled 'unchanged'. Seizure-related firing patterns were compared for individual cells across sequential pairs of seizures within the same recording session.Results: If random groups of neurons participate in the buildup to each seizure, then seizure-related changes in firing rate for an individual neuron should vary across seizures. Conversely, if seizures arise through the activation of fixed networks, the same seizure-related changes in firing rate for each neuron should be seen across all seizures. For 70 CA3 pyramidal cells recorded during sequential pairs of seizures (N=3 rats; 5, 1 and 2 seizure pairs), seizure-related changes prior to one seizure were more similar to changes observed in the subsequent seizure than would be expected by chance (p=.0067, chi-square test). Similar results were observed for 33 granule cells recorded during sequential pairs of seizures (N=3 rats; 2, 3 and 4 seizure pairs; p=.044). For all ictal cells, 53% remained ictal when only 19% would be expected by chance. For all preictal cells, 40% remained preictal when only 26% would be expected.Conclusions: These findings suggest that seizure buildup may occur across a nonrandom network of neurons. If single neurons in one or multiple brain structures are activated reliably prior to most or all seizures (i.e., 'preictal' cells), then these units may be particularly useful for real-time seizure prediction. (Supported by NIH/NINDS)