Extended Data Fig. 2: Dynamic and unselective inhibitory engrams control network activity.

a-c, Post-encoding evolution of inhibitory engram cells in Fig. 1b. Means and 99% confidence intervals are shown. n = 10 trials. a, Ensemble of inhibitory engram cells as a fraction of all inhibitory neurons (see Methods). Dashed line indicates inhibitory engram cell ensemble at the end of training. b, Ensemble overlap between probing-activated inhibitory engram cells at consolidation time = t and t-1 h as a fraction of inhibitory engram cells at consolidation time = t-1 h. Dashed line indicates ensemble of inhibitory neurons that remained part of the inhibitory engram in all sampled time points (that is, consolidation time = 0, 1, …, 24 h) as a fraction of inhibitory engram cells at consolidation time = 0 h (that is, training-activated inhibitory engram cells). c, Ensemble overlap between inhibitory engram cells activated during both probing and training as a fraction of training-activated inhibitory engram cells (top), probing-activated inhibitory engram cells (middle) and all inhibitory neurons in the network (bottom). d-g, Analysis of inhibitory engram cells during recall in Fig. 1b. Means and 99% confidence intervals are shown. n = 10 trials. Color denotes stimulus as in Fig. 1c. d, Recall rate of inhibitory engram cells as a function of consolidation time. Dashed line indicates threshold ζ^thr = 10 Hz for engram cell activation. e, Recall of inhibitory engram cells as a function of consolidation time. f, Discrimination index between recall of inhibitory engram cells evoked by cues of the training stimulus and individual novel stimuli as a function of consolidation time. g, Fraction of probing-activated inhibitory engram cells reactivated during recall as a function of consolidation time. h, Analysis of excitatory network activity in a 60 s interval during consolidation in Fig. 1b. From top to bottom: training stimulus reactivation times (in red), spike raster of 256 randomly-chosen excitatory neurons (for clarity, only every fifth spike is shown), population activity (that is, average firing rate) of all excitatory neurons (dashed lines indicate target activity level γ = 4 Hz and threshold ζ^thr = 10 Hz for engram cell activation) and network statistics for the interval marked at the top (from left to right: histograms of firing rates, interspike intervals (ISI) and coefficient of variation (CV) of ISI). Population activity is shown without smoothing or convolution (see Methods). Representative trial is shown. i-n, Analysis of inhibitory engram dynamics in Fig. 1b using non-negative matrix factorization to identify engram cells (see Methods). Means and 99% confidence intervals are shown. n = 10 trials. i, Ensemble overlap between probing-activated inhibitory engram cells at consolidation time = t and t-1 h as a fraction of inhibitory engram cells at consolidation time = t-1 h. Dashed line indicates ensemble of inhibitory neurons that remained part of the inhibitory engram in all sampled time points (that is, consolidation time = 0, 1, …, 24 h) as a fraction of inhibitory engram cells at consolidation time = 0 h (that is, training-activated inhibitory engram cells). j, Ensemble overlap between inhibitory engram cells activated during both probing and training as a fraction of training-activated inhibitory engram cells (left), probing-activated inhibitory engram cells (middle) and all inhibitory neurons in the network (right). k-n, Color denotes stimulus as in Fig. 1c. k, Recall rate of inhibitory engram cells as a function of consolidation time. Dashed line indicates threshold ζ^thr = 10 Hz for engram cell activation. l, Recall of inhibitory engram cells as a function of consolidation time. m, Discrimination index between recall of inhibitory engram cells evoked by cues of the training stimulus and individual novel stimuli as a function of consolidation time. n, Fraction of probing-activated inhibitory engram cells reactivated during recall as a function of consolidation time.