Time course of reactivation of memory-related cell ensembles in the rat medial prefrontal cortex during sleep.
A. Peyrache1, K. Benchenane1, M. Khamassi12, V. Douchamps1, P. L. Tierney3 S. I. Wiener1, F. P. Battaglia4,
1LPPA, CNRS Colllege de France, Paris, France, 2Inst. des Systemes Intelligents et Robotiques, Univ. Pierre et Marie Curie, Paris, France, 3Inserm U667, College de France, Paris, France, 4Univ. van Amsterdam, Amsterdam, The Netherlands

The prefrontal cortex is implicated in the flexible learning of stimulus-outcome associations, which are consolidated in memory during offline periods. Reactivation of memory traces, in the form of the reinstatement of experience-related activity in prefrontal cell assemblies during sleep could be the basis for such a consolidation process. To study this, we developed a novel analysis which allows to follow the time course of task-related reactivation in simultaneously recorded cell ensembles. The correlation matrix of binned spike trains from multiple cells is decomposed in its principal components, the largest of which represents groups of cells whose activity was highly correlated during the reference recording period. The instantaneous cell pair co-activation matrix during sleep, weighted by the coefficients in a given principal component, and averaged over all cell pairs, can then be taken as a measure of the reactivation of the cell assembly corresponding to that principal component at a given time.

We analyzed medial prefrontal ensembles from five rats while learning a set-switching task on a Y-maze, and in rest sessions preceding and following the task. In 62 out of 86 sessions, cell assembly reactivation was significantly greater (p<0.05) during slow wave sleep (SWS) after the session than in SWS before. There was a significant correlation (Pearson's correlation test, p<0.05) between the eigenvalues associated with the principal components during task performance (indicating the strength of the encoding) and the increased re-activation in post-task SWS (compared to pre-task SWS). Moreover, in 67 out of 86 sessions, co-activation was correlated (p<0.05) with the power of both delta and spindle cortical oscillations, and it was much weaker during rest periods that were classified as non-sleep. The increased co-activation in the post-experience sleep was attributable to discrete bouts of activation, typically 2-5 seconds in duration. This new technique permits to precisely follow the time course of neural ensemble re-activation. These data demonstrate that theta, ripple-sharp waves and spindles are important for prefrontal post-task SWS reactivation, a possible neural ensemble basis for memory consolidation.