Network resonance during slow-wave sleep facilitates memory consolidation through phase-coding

Abstract

Sleep plays a critical role in memory consolidation, however, the exact role that sleep and its effects on neural network dynamics play in this process is still unclear. Here, we combine computational and experimental approaches to study the dynamical, network-wide underpinnings of hippocampal memory consolidation during sleep. We provide data to support a novel hypothesis on the role of cellular resonance with sleep-associated theta band (4-12 Hz) hippocampal oscillations in this process. We show that increases in the stability of hippocampal memory representations after learning (which predicts successful memory consolidation) are mediated through emergent network-wide resonance and locking of neuronal activity to network oscillations. These changes arise in the network as a function of changes to network structure during learning, and mirror experimental findings in the hippocampus. Finally, we show that input-dependent pattern formation (e.g. "replay") in the hippocampus during sleep states, together with spike timing dependent plasticity (STDP)-based memory consolidation, leads to universal network activity reorganization. This reorganization generates heterogeneous changes in neuronal spiking frequency, similar to what has been observed in a variety of brain circuits across periods of sleep. Our results support the hypothesis that sleep plays an active role in memory consolidation by switching the hippocampal network from rate-based to phase-based information representation. The mechanisms through which this occurs supports the integration of heterogeneous cell populations into memory traces. Significance StatementIn this study, we provide a mechanistic explanation of how sleep selectively facilitates memory consolidation, through recruitment of heterogeneous neuronal populations and structural reorganization of the network into an engram. Specifically, we show that emergent theta band oscillations during sleep facilitate stabilization of memory representations via spike timing dependent reinforcement. This stabilization, together with STDP, allows for systematic reorganization of synaptic connections within these populations, universally redistributing firing rates of participating neurons. Simultaneously, network oscillations facilitate a switch from rate-to phase-coding of information among neuronal populations with highly heterogenous firing frequencies, incorporating more neurons into the engram. Our results reconcile discrepant findings on network reorganization during sleep, and demonstrate a clear mechanism for both strengthening and weakening of synaptic efficacy during sleep.