Model- and scale-independent performance of a hippocampal CA3 network architecture

Abstract

The hippocampus is known to be a critical structure for memory function, but how its anatomical, physiological, and cellular elements endow it with this function remains a mystery. As such, even less is known about how neuropathological changes associated with Alzheimer's disease, epilepsy or traumatic brain injury lead to the clinical manifestations of disease. To begin to address these issues, we have constructed a biophysically detailed model of the CA3 region of the hippocampus that incorporates many of the well-characterized but, from a functional standpoint, poorly understood properties and phenomena into a functioning network. These include cholinergic neuromodulatory regulation, theta (4–6 Hz) oscillations, gamma (20–80 Hz) oscillations, diversity of interneuron classes, regular spiking vs. bursting pyramidal cells, and lamina-specific inputs and characteristic anatomy. We have found that the integration of these hippocampal elements results in a biological analog to the “connectionist” autoassociative attractor network where memories are represented as spatial patterns of temporally precise spikes across CA3. In addition, simulating the cholinergic modulation of this network model and its constituent cellular components suggests novel mechanisms for memory dysfunction in Alzheimer's disease E.D. Menschik, L.H. Finkel, Artificial Intell. Med. 13 (1994) 99–121; E.D. Menschik, L.H. Finkel, Progress in Brain Research, Elsevier, New York, in press. Here we demonstrate such networks to be robust and accurate in memory function, insensitive to the choice of neuronal model from a 64-compartment cell with numerous ionic conductances R.D. Traub et al., J. Physiol. 481 (1994) 79–95; to a reduced dual-compartment model P.F. Pinsky, J. Rinzel, J. Compu. Neurosci. 1 (1994) 39–60; and scalable from small to large networks.