Identifying Network Correlates of Memory Consolidation

Neuronal spiking activity carries information about our experiences in the waking world but exactly how the brain can quickly and efficiently encode sensory information into a useful neural code and then subsequently consolidate that information into memory remains a mystery. While neuronal networks are known to play a vital role in these processes, detangling the properties of network activity from the complex spiking dynamics observed is a formidable challenge, requiring collaborations across scientific disciplines. In this work, I outline my contributions in computational modeling and data analysis toward understanding how network dynamics facilitate memory consolidation. For experimental perspective, I investigate hippocampal recordings of mice that are subjected to contextual fear conditioning and subsequently undergo sleep-dependent fear memory consolidation. First, I outline the development of a functional connectivity algorithm which rapidly and robustly assesses network structure based on neuronal spike timing. I show that the relative stability of these functional networks can be used to identify global network dynamics, revealing that an increase in functional network stability correlates with successful fear memory consolidation in vivo. Using an attractor-based model to simulate memory encoding and consolidation, I go on to show that dynamics associated with a second-order phase transition, at a critical point in phase-space, are necessary for recruiting additional neurons into network dynamics associated with memory consolidation. I show that successful consolidation subsequently shifts dynamics away from a critical point and towards sub-critical dynamics. Investigations of in vivo spiking dynamics likewise revealed that hippocampal dynamics during non-rapid-eye-movement (NREM) sleep show features of being near a critical point and that fear memory consolidation leads to a shift in dynamics. Finally, I investigate the role of NREM sleep in facilitating memory consolidation using a conductance-based model of neuronal activity that can easily switch between modes of activity loosely representing waking and NREM sleep. Analysis of model simulations revealed that oscillations associated with NREM sleep promote a phase-based coding of information; neurons with high firing rates during periods of wake lead spiking activity during NREM oscillations. I show that when phase-coding is active in both simulations and in vivo, synaptic plasticity selectively strengthens the input to neurons firing late in the oscillation while simultaneously reducing input to neurons firing early in the oscillation. The effect is a net homogenization of firing rates observed in multiple other studies, and subsequently leads to recruitment of new neurons into a memory engram and information transfer from fast firing neurons to slow firing neurons. Taken together, my work outlines important, newly-discovered features of neuronal network dynamics related to memory encoding and consolidation: networks near criticality promote recruitment of additional neurons into stable firing patterns through NREM-associated oscillations and subsequently consolidates information into memories through phase-based coding.PhDBiophysicsUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/162991/1/qmskill_1.pd