Memories are stored in distributed networks across the cortex. Retrieval of memories, for example vivid visual recollection of a sensory scene, activates a wide cortical network ranging from cognitive areas in the frontal lobe all the way to early sensory regions. The formation of the distributed neuron networks underlying these memories requires precise coordination of pre- and post-synaptic spikes across long distances in the brain. Plasticity of long-range excitatory connections linking distant neuron groups occurs through spike-time dependent plasticity (STDP), for which presynaptic vesicle release and postsynaptic spiking must occur with a precision of a few milliseconds. Sleep oscillations have long been known to aid memory consolidation: for example, the number of 11-15 Hz sleep ?spindles? during a nap or over a whole night correlates with improvements in memory performance after sleep. Further studies have recently demonstrated causal evidence for the involvement of spindles in human sleep-dependent memory consolidation. Spindles are known to group pyramidal and interneuron spiking activity at the peak of their neural population rhythm, but the specific mechanism by which spindles coordinate distant populations in neocortex remains unknown. For example, if spindles were fully synchronous across neocortex, vesicle release caused by spikes from distant populations would occur well after spikes in a local group, because of the time delays for white-matter connections. Recent work has shown that, instead of being synchronous, spindles robustly appear as neural traveling waves (nTWs), a smooth flow of activity in this case from temporal cortex, to parietal lobe, on to frontal lobe, and back to the temporal lobe. Because they travel at the speed of white-matter fibers in cortex, these rotating spindle nTWs can align spikes from distant neuronal populations. Further, spindle nTWs can repeat precisely, with up to millisecond temporal precision over several hours of sleep. These results lead to the general hypothesis that spindle nTWs could organize neural activity for storing memories in the distributed synaptic architecture of neocortex. In this proposal, we aim to test this hypothesis through application of computational techniques for signal processing of multichannel neural data and a novel computational model. We will develop a novel model for spindle nTWs based on a population description from physics (Aim 1). We will apply our computational techniques and predictions from our computational model to electroencephalogram (EEG) recordings in the infant (Aim 2, in collaboration with Dr. April Benasich, Rutgers University). We will then apply our computational techniques to test whether broad spindle nTWs appear during targeted memory reactivation (TMR) in human cortex (Aim 3, in collaboration with Dr. Sara Mednick, University of California-Irvine). Results from this research will provide insight into the basic mechanisms for human memory, in particular how the sleep rhythms can coordinate broad, distributed neuron networks in the face of significant time delays for axonal conduction, and how these mechanisms are altered in disease.
Recent work has shown that the 11-15 Hz human sleep ?spindle?, rather than being synchronous throughout neocortex or isolated into local cortical patches, is robustly organized into a neural traveling wave (nTW), influencing spike activity across the brain. The functional and computational role for this spatiotemporal pattern of activity remains unknown. We propose detailed computational modeling and analysis of multichannel recordings to test the hypothesis that spindle nTWs can align spikes across distant cortical regions for storing long-term memories in distributed cortical networks through synaptic plasticity.
