The intracortical neuronal connectivity subserving focal epileptiform activity in rat neocortex

In the anaesthetized rat, regions of the somatosensory cortex have been subpially isolated, leaving intact the cortical blood supply and the connectivity via the white matter. Application of penicillin or strychnine into layer IV of intact cortex resulted in enhancement of amplitude and prolongation of evoked potentials together with the appearance of spontaneous epileptiform discharges. Within a partially isolated region of cortex, spontaneous and evoked potentials occurred as in normal cortex, but application of convulsant drug resulted in no changes in evoked potentials and in no spontaneous spiking. With incisions for which the surface profile measured 0.9 x 0.9 mm, full-depth isolation resulted in interruption of the propensity for epilepsy, whereas half-depth incisions left epileptic manifestations unimpaired. With the surface profile measuring 0.5 x 0.5 mm, half-depth isolation was sufficient to prevent epileptic activity. Results from isolated regions of various geometries and sizes indicated that the ability of cortical neurones to generate epileptic activity depends on the amount of connectivity with surrounding cortex. The propensity of cortex to become epileptic is thus a mass action effect and the 'epileptic neuronal aggregate' is operationally different from anatomically based modular organizations such as thalamo-cortical or cortico-cortical columns. In the small barrel field of the somatosensory cortex, partial isolations that prevented the appearance of spontaneous epileptiform spiking contained many barrels, indicating that a single thalamo-cortical module contains insufficient inherent lateral connectivity to support epileptiform activity. Theoretical considerations indicated that the excitability of a neurone depends both on its monosynaptic connections with other neurones and on the connectivity of these latter with neurones further afield. The interruption of epileptiform activity by partial isolation could be mimicked by a computer model in which connectivity was mediated via short synaptic paths. The model exhibited self-sustaining synchronized neural activity that could be prevented by interruption solely of polysynaptic paths.