Abstract
We aimed to better understand the dynamics of cortical neurons during nonREM sleep-a state in which neuronal populations are silenced for ∼100 ms of every second due to delta wave fluctuations. This alternation between periods of population spiking ("UP states") and silence ("DOWN states") generally synchronizes populations at the 1 s timescale, although some prior work has shown that anticorrelations in nonREM can occur in pairs of neurons that are anticorrelated in wake. We used 24 h recordings of frontal cortical neurons in rats to measure cross-correlation between pairs of neurons in wake, nonREM, and REM. Surprisingly, while most pairs of neurons were synchronized, we found a minority of pairs that showed significant nonREM-induced desynchronization, as indicated by negative cross-correlations in nonREM without equivalent anticorrelation in wake or REM. Interestingly, the degree of anticorrelation within NREM epochs was positively modulated by oscillations in the low-frequency (i.e., "delta" or 1-4 Hz) range, meaning anticorrelation between some pairs increases when correlation increases between other pairs. Furthermore, this effect was mediated by firing during the nonsilent UP state phase of the delta cycle, indicating it is not due to neurons active in the DOWN state. Finally, high-variance spike timing between pairs of neurons and burst spiking during UP states are shown to specifically contribute to the anticorrelation. This state-specific desynchronization during the "synchronized" state represents a new phenomenon that can lead to new understanding of network dynamics during sleep.