Figure 5. CA1 and prefrontal cortex (PFC) cell assemblies show different awake Sharp-Wave Ripple (aSWR) reactivation dynamics.

(A) Average (z-scored) assembly activation triggered by aSWR occurring in the intertrial intervals for CA1 and PFC, pre- and postlearning sessions (top). Mean aSWR-triggered activation over all the assemblies for pre- and postsessions for each area. Shaded areas represent the standard error of the mean (SEM). Black dots represent windows in which pre- and postassembly activity were statistically different (Wilcoxon rank sum test; p < 0.05). Notice the higher aSWR-triggered activation of assemblies in PFC in postsessions. (B) Histogram (left) and cumulative distribution function (CDF; right) of the mean assembly activity on the reactivation window denoted in A. p values refer to a two-sample Kolmogorov–Smirnov test between pre- and postdistributions. (C) Average aSWR reactivation of each assembly per session (top). Sessions were divided into 10 blocks of equal trial length. Mean aSWR reactivation of all positively (reactivation+) and negatively (reactivation−) reactivated assemblies. Asterisks refer to Wilcoxon signed-rank test performed between the first and last three trial blocks (dashed rectangles) of each area/learning condition (n.s.: nonsignificant; *p < 0.05 and shaded areas represent SEM). Note the evident increase in CA1 aSWR assembly reactivation across the session in both pre- and postsessions for positively modulated assemblies (reactivation+). (D) Mean (z-scored) assembly activity triggered by the stimulus onset for the 25 % most strongly aSWR-reactivated assemblies in CA1 (left). Average of the traces over each trial period is shown for CS+ and CS− (right). Notice the initial decrease of assembly activity in CA1 during the stimulus and the posterior separation between CS+ and CS−. (E) The same as in D, but for PFC assemblies. Note the difference between CS+ and CS− assembly activity during the reward period. Asterisks refer to a Wilcoxon signed-rank test comparing CS+ and CS− (*p < 0.05; ***p < 0.001). Error bars refer to SEM and darker bars denote mean assembly activity significantly different from 0 (p < 0.05; t-test).

Figure 5—figure supplement 1. Awake Sharp-Wave Ripple (aSWR) reactivation of assemblies detected during intertrial intervals.

(A) (Left) Mean reactivation around aSWRs of assemblies detected during the intertrial intervals (excluding aSWR events) for pre- and postlearning sessions. (Middle) Histogram of mean assembly aSWR reactivation on the reactivation window (yellow rectangle) for pre- and postlearning sessions. (Right) Cumulative distribution of mean assembly aSWR reactivation. p values refer to a two-sample Kolmogorov–Smirnov test between pre- and postdistributions. (B) Similar to A, but using sham aSWR times to compute the average reactivation (aSWR events were randomly shifted by ~200 ms).

Figure 5—figure supplement 2. Distribution of awake Sharp-Wave Ripples (aSWRs) during trace conditioning.

(A) Example of simultaneously recorded local field potential (LFP) and single-cell activity from CA1 and prefrontal cortex (PFC) during aSWRs. (B) Ripple rate during CS+ (blue) and CS− (red) trials across all conditioning sessions. Average lick rate during CS+ trials is overlaid in green (shaded areas indicate standard error of the mean [SEM]). (C) Average aSWR rate increases from early to late within individual sessions (top) and average aSWR rate does not change between pre- and postlearning sessions (bottom) (error bars represent SEM and ** refers to Wilcoxon rank sum test; p < 0.01).

Figure 5—figure supplement 3. Detecting cell assemblies in neural populations.

(A) The rastergram (left) of each trial was computed and binned in 20-ms bins with no overlap (middle). After concatenating the activity of all trials, the activity of each neuron was z-scored and the correlation matrix was computed (right). (B) The eigenvalues of the correlation matrix were then computed and compared to the analytical (Marchenko–Pastur) distribution to estimate the amount of assembly patterns present in the data (top). After that, independent component analysis was used to extract the assembly patterns (bottom). (C) The patterns in B were then used to project the assembly activity during the trial, using 20-ms bins with steps of 1 ms.