1. A computer model was constructed of the guinea-pig hippocampal region in vitro, containing 100 pyramidal neurones. This approach has contributed to the understanding of brief (usually less than 100 ms) epileptic events known as 'interictal spikes'. The present study addresses the cellular mechanisms of more prolonged epileptic events, lasting 200 ms and more, that may represent short-duration seizures. Each neurone was simulated with a nineteen-compartment model using six voltage-dependent ionic conductances. The neurones were randomly interconnected with excitatory synapses, each synapse exerting a fast voltage-independent alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) component and a slower voltage-dependent N-methyl-D-aspartate (NMDA) component. Each neurone received input from twenty other neurones. 2. This model was able to generate, in response to synaptic noise or to stimulation of one neurone, a series of synchronized population bursts, the initial (primary) burst being longer than later (secondary) bursts, terminating in a prolonged after-hyperpolarization. The simulated after-discharge potentials resemble those recorded experimentally from pyramidal neurones during perfusion of the hippocampal slice with media containing picrotoxin, a blocker of synaptic inhibition mediated by GABA(A) receptors. 3. Simulated after-discharges agree with the following experiments: over a certain range of total NMDA conductance, blockade of AMPA receptors will prevent the occurrence of synchronized firing, whereas, blockade of NMDA receptors will, in contrast, abolish the secondary bursts, leaving a shortened and somewhat smaller primary burst. Dendritic potential oscillations occur in phase with somatic oscillations. When interneurones (some generating GABA(A)-mediated IPSPs, others generating GABA(B) IPSPs) are included in the model, the occurrence of synchronized events was suppressed, the most significant suppressant effect coming from GABA(A) IPSPS. 4. The model predicts that: a dendritic calcium spike occurs during each secondary burst; AMPA receptors serve to maintain the synchrony of secondary bursts, as well as to initiate the primary burst; and that with sufficient total NMDA conductance, synchronized firing can occur even with AMPA receptors blocked. 5. The model suggests, in addition, that the duration of the initial burst is determined in part by the experimentally observed delay between Ca2+ entry and peaking of the after-hyperpolarization (AHP) conductance, and hence reflects properties of the individual pyramidal neurones. Specifically, a pattern of a long initial burst followed by brief secondary bursts is elicited in single-cell simulations by injection of a steady depolarizing current into the apical dendrite. The same pattern is produced when the single-cell model includes only calcium and calcium-dependent conductances. Modulating the activation time constant for the AHP conductance modulates the length of the initial burst. The long latency to the peak of the GABA(B) IPSP may also contribute to determining the duration of the initial burst. 6. Secondary bursts appear to be more sensitive to antagonists of excitatory amino acids because they occur when the cell membrane is shunted by intrinsic and synaptic conductances and consequently require more current for their generation. 7. In conclusion, recurrent excitatory synapses terminate on dendritic regions of CA3 pyramidal cells which generate repetitive calcium spikes in response to a sustained depolarization. During after-discharges, NMDA receptor activation provides a sustained inward current. The resulting dendritic calcium potentials initiate somatic bursts. These bursts propagate axonally to reactivate non-NMDA receptors at recurrent synapses with a timing that helps to maintain after-discharge synchrony. Thus epileptic discharges involve intrinsic neuronal properties that are activated and synchronized by recurrent excitatory synaptic circuitry.
