Jackson_2013_J.Neurophysiol_110_2511

Nervous systems are thought to encode information as patterns of electrical activity distributed sparsely through networks of neurons. These networks then process information by transforming one pattern of electrical activity into another. To store information as a pattern, a neural network must strengthen synapses between designated neurons so that activation of some of these neurons corresponding to some features of an object can spread to activate the larger group representing the complete object. This operation of pattern completion endows a neural network with autoassociative memory. Pattern completion by neural networks has been modeled extensively with computers and invoked in behavioral studies, but experiments have yet to demonstrate pattern completion in an intact neural circuit. In the present study, imaging with voltage-sensitive dye in the CA3 region of a hippocampal slice revealed different spatial patterns of activity elicited by electrical stimulation of different sites. Stimulation of two separate sites individually, or both sites simultaneously, evoked "partial" or "complete" patterns, respectively. A complete pattern was then stored by applying theta burst stimulation to both sites simultaneously to induce long-term potentiation (LTP) of synapses between CA3 pyramidal cells. Subsequent stimulation of only one site then activated an extended pattern. Quantitative comparisons between response maps showed that the post-LTP single-site patterns more closely resembled the complete dual-site pattern. Thus, LTP induction enabled the CA3 region to complete a dual-site pattern upon stimulation of only one site. This experiment demonstrated that LTP induction can store information in the CA3 region of the hippocampus for subsequent retrieval.