Abstract
A fundamental question in neuroscience is how memory formation shapes brain activity at the level of populations of neurons. Recent studies of hippocampal ‘engram’ cells, identified by immediate-early genes (IEGs) induced by learning, propose that these populations act as a neuronal substrate for memory storage. The current framework for engram formation proposes that cells join ensembles based on increased intrinsic excitability, and that after initial learning, they co-activate to support memory retrieval. However, direct evidence of how engram population dynamics evolve across learning is limited. Here we combined activity-dependent genetic tagging and two-photon calcium imaging to characterize CA1 engram population activity before and after learning. We observed that spontaneous activity two days before learning predicted genetic tagging, consistent with a model in which spontaneous fluctuations bias cells into forming engram assemblies. Surprisingly, we were unable to detect increased spontaneous activity rates or pairwise correlations amongst tagged CA1 neurons after learning. These results were consistent with computational network models that incorporate strong and specific inhibitory connections, supporting the idea that excitatory/inhibitory balance in CA1 may play a key role in engram dynamics. Together these results highlight a potential role for slow time scale excitability fluctuations in driving engram formation and suggest that excitatory-inhibitory balance may regulate engram cell co-activation.