Abstract
Hippocampus is a brain region critical for the formation of new long-term declarative memories. It transmits and processes memory information with its distinct feedforward trisynaptic pathway. Identifying functional properties of the hippocampal circuits is important for understanding the mechanisms of memory formation and building hippocampal prostheses for restoring memory functions lost in diseases or injuries. In hippocampal slices, trisynaptic responses can be elicited and recorded using conformal multi-electrode arrays. A proof-of-principle hippocampal prosthetic system has been successfully developed based on a computational model that accurately describes the input-output properties of the hippocampal circuit. In behaving animals, hippocampal functional connectivities are analyzed with a nonlinear dynamical multi-input, multi-output (MIMO) model using behaviorally-driven spiking data. Results show that the hippocampal CA3-CA1 functional connection is diffusive along the septo-temporal axis, as opposed to strictly laminar. There are strong causal relations between the CA3 and CA1 spiking activities. The MIMO model can accurately predict the spatio-temporal patterns of the CA1 output spikes based on the ongoing spatio-temporal patterns of the CA3 input spikes. MIMO model-based electrical stimulation to the CA1 region effectively restores the hippocampal memory function by reinstating the CA1 activities. The recording component, the nonlinear dynamical MIMO model, and the stimulation component essentially constitute a closed-loop prosthetic system that bypasses the impaired hippocampal region.