Learning and memory in the hippocampus: Events on millisecond to week time scales

The activity-dependent modification of synaptic efficacy likely plays an important role in the information storage capability of the brain. In this thesis, I explore the interaction between synaptic learning rules and neural activity in the hippocampus through whole-cell recording experiments and computational modeling studies.; In a series of hippocampal brain slice experiments, I examine how spike-timing dependent plasticity at the hippocampal CA3-CA1 synapse is influenced by temporal features of neural activity patterns likely to occur in awake, behaving animals; namely the modulation of activity by the theta rhythm (5–10 Hz) and bursting of the postsynaptic CA1 neuron. LTD was evoked under a wide variety of conditions. It was not sensitive to pairing frequency (0.1–5 Hz), and could be induced by pairing single presynaptic action potentials with both single postsynaptic action potentials, and bursts. The induction of UP required both repeated causal pairings at 5 Hz, and burst firing in the postsynaptic neuron. This suggests that burst firing associated with the theta rhythm may be essential for gating LTP, but not LTD. Therefore bi-directional synaptic plasticity depends not only on spike-timing, but also on the temporal context of neural activity. In a series of computational modeling studies, I explore the effect of synaptic learning rules on memory storage during different behavioral states. In conventional treatments of learning in neural networks, memories go through separate stages of acquisition and retrieval. During the training period, correlations in neural activity lead to synaptic plasticity, while during the retrieval stage, the strengths of network connections remain constant. Another possibility, suggested by recent genetic and behavioral experiments, is that the retrieval process itself may result in synaptic plasticity that effectively reinforces connections over the days and weeks following initial learning. Testing this hypothesis in a firing-rate model of the hippocampus with a correlation-based learning rule, I demonstrate how a single stored memory can be strengthened through retrieval. Furthermore, I show how coupling synaptic plasticity with retrieval may also provide a mechanism for memory processing beyond the initial training period. This can influence which memories are stored stably in the network over long time periods.