Chronic Neuropixels probe recordings across mPFC subregions.

(a) Left: an image of the Neuropixels probe, right: the probe mapping from one example animal where a part of the trace is seen going through the hippocampus (top) and mPFC (bottom). (b) An example segment of data with the animals running speed (top) and spiking activity recorded from the hippocampus, dorsal mPFC and ventral mPFC (top to bottom). (c) An overview of the behavioral task. Each day starts with a period of rest, followed by rule learning in the maze. Each trial is interleaved with a short period (∼45 seconds) of rest, and the day ends with a period of rest or sleep of up to 60 minutes. Four different rules have to be learned over the course of multiple days, based on allocentric and cue-based strategies. (d) The number of sorted neurons per session for each (sub-)region. The different markers indicate different animals.

Neurons in the dorsal and ventral mPFC are differentially modulated by hippocampal SWRs and theta phase.

(a) Six example neurons in the mPFC that are SWR-excited (top), SWR-inhibited (middle) and SWR-unmodulated (bottom; dorsal mPFC in olive, ventral mPFC in blue). (b) The average normalized firing rates around SWR onset of all SWR-excited and SWR-inhibited neurons in the dorsal and ventral mPFC. Data is presented as mean of all neurons ± SEM. (c) The mean of the SWR modulation per subregion during periods of exploration in the maze. The p-value indicates the result of an independent samples t-test between the absolute modulation scores in the dorsal versus ventral mPFC. (d) The percentage of neurons that is SWR-excited (solid fill) and SWR-inhibited (dashed fill) per mPFC subregion. The p-value indicates the result of an independent samples t-test between the dorsal and ventral mPFC. (e) The mean overall firing rates per subregion of SWR-excited (solid fill), SWR-inhibited (dashed fill) and SWR-unmodulated (no fill) neurons in the maze. The p-values indicate an ANOVA between the different SWR modulation categories. (f) The percentage of neurons that is phase-locked to hippocampal theta per brain region. The p-values indicate the results of Mann-Whitney U test between the hippocampus and mPFC, and between the dorsal and ventral mPFC. (g) The percentage of neurons that is phase-locked to hippocampal theta, divided between mPFC neurons that are SWR-excited (solid fill), SWR-inhibited (dashed fill) and SWR-unmodulated (no fill). The p-values indicate an ANOVA between the different SWR modulation categories. (h) Left panels: the mean angle (-π to π) and (0.0 to 0.6) for all dorsal mPFC clusters that were significantly phase locked to hippocampal theta, divided between SWR-excited clusters (top), SWR-inhibited clusters (middle) and SWR-unmodulated clusters (bottom). Middle panels: the mean angle and for all phase locked clusters in the same configuration as the left panels, but for the ventral mPFC. Right panels: histograms of the mean angles of all theta phase-locked clusters that were SWR-excited, SWR-inhibited or SWR-unmodulated (top to bottom). The p-values indicate the result of a Kolmogorov-Smirnov test to test the difference in distribution between the dorsal (olive) and ventral (blue) mPFC.

SWR-unmodulated neurons are spatially tuned, directionally selective and show theta cycle skipping.

(a) The firing patterns of eight example neurons in the maze, showing a range of spatial tuning scores (in bits/spike) sorted from highest to lowest. (b) Left panel: the spatial tuning score per subregion (p-values indicate a Mann-Whitney U test). Right panel: the spatial tuning score per subregion divided between mPFC neurons that are SWR-excited (solid fill), SWR-inhibited (dashed fill) and SWR-unmodulated (no fill; p-values indicate a Kruskal-Wallis test). (c) Three example trials where quadrants on the maze are categorized as upcoming or alternative choice, North or South, and left or right. (d) Left panel: the percentage of neurons that fire significantly higher than the shuffled population for at least one of the head direction categories. The chance level is indicated at 15%, as an alpha of 0.05 was maintained for each of the three direction categories (future versus alternative, North versus South, and left versus right; p-values indicate the results of a Mann-Whitney U test). Right panel: the absolute maximum directional selectivity index per subregion for mPFC neurons that are SWR-excited (solid fill), SWR-inhibited (dashed fill) and SWR-unmodulated (no fill; p-values indicate a Kruskal-Wallis test). (e) Spike time auto-correlations of two example neurons where the peak of the smoothed signal (in black) at 250 ms is subtracted from the peak at 125 ms and normalized to the highest of the two peaks to obtain the cycle skipping index (left panels). To determine the statistical significance, the score is compared to a shuffled population (right panels). (f) Left panel: the percentage of neurons that showed significant theta cycle skipping properties per subregion (p-values indicate a Mann-Whitney U test). Right panel: the theta cycle skipping index per subregion divided between mPFC neurons that are SWR-excited (solid fill), SWR-inhibited (dashed fill) and SWR-unmodulated (no fill; p-values indicate a Kruskal-Wallis test).

Non-local representations in the mPFC are predictive of the animals’ upcoming choice.

(a) Three examples of trials where the decoded positions (grey scale) and real positions (red) are shown per maze segment. Segments of upcoming and alternative choice are highlighted with light and dark brown spans. (b) The prediction index per session based on the decoded non-local representations while the animal was at the start and center of the maze, separated for when the rule was not yet learned and when the rule was learned, and for the allocentric (navy) and cue-based (orange) rules. Shown in light grey is the prediction accuracy based on the shuffled decoded posteriors, and p-values indicate the result of a paired samples t-test between the real and shuffled predictions. (c) Left panel: an example learning curve (in black) with a rule switch from South to North after 10 trials, and the predicted directions based on the non-local representations overlayed (navy). Right panel: the lag (in number of trials) between the real direction and decoded direction per session for allocentric (navy) and cue-based (orange) strategies. A negative lag indicates that the decoded direction preceded the real direction over the course of learning. P-values indicate the result of a one-sample Wilcoxon test to test the deviation from zero. (d) The distribution of decoded probabilities at the center of the maze for all mPFC neurons (top panel), SWR-modulated neurons (middle panel) and SWR-unmodulated neurons (bottom panel). P-values indicate the results of a two-way repeated measures ANOVA between the maze bins and prediction categories (alternative versus upcoming choice). (e) The overlap between representations in the hippocampus and mPFC (all neurons, SWR-modulated neurons and SWR-unmodulated neurons) for the local representations (left panel), the representations of alternative choice (middle panel), and the upcoming choice (right panel). P-values indicate the results of a paired samples t-test of the percentage of overlapping decoded time bins for each session compared to the overlap of the shuffled data for that session.

Non-local representations in the mPFC are not linked to hippocampal SWRs and theta phase.

(a) The occurrence of representations of alternative choice (left) and upcoming choice (right) in the mPFC in temporal proximity to the closest SWR. The p-values indicate the results of a Wilcoxon signed-rank test between real and shuffled data at bin 0-20 ms from SWR onset. (b) The firing rates of mPFC neurons that are SWR-excited, SWR-inhibited and SWR-unmodulated, during representations of upcoming and alternative choices, normalized to the average firing rate at the start of the maze. The p-values indicate the results of a Kruskal-Wallis test between the different SWR modulation categories. (c) A histogram of the phase of hippocampal theta oscillations at which the start of each representation of upcoming and alternative choices in the mPFC occurred. The p-values indicate the results of a Kolmogorov-Smirnov test for difference in distributions between the representations of alternative and upcoming choices compared to shuffled decoded probabilities. (d) Left: circular histograms of the phase locking shown in (c). Right: The phase locking score of representations of upcoming and alternative choices (Rayleigh’s Z, dark and light brown dots) to hippocampal theta oscillations compared to a randomized population (gray histograms). P-values indicate the result of a left-tailed Monte Carlo test of the Rayleigh’s Z compared to the shuffled population.

Non-local representations in the hippocampus are linked to SWRs and theta phase.

(a) The occurrence of representations of alternative choice (left) and representations of upcoming choice (right) in the hippocampus in temporal proximity to the closest SWR. The p-values indicate the results of a Wilcoxon signed-rank test between real and shuffled data at bin 0-20 ms from SWR onset. (b) The firing rates of mPFC neurons that are SWR-excited, SWR-inhibited and SWR-unmodulated, during representations of upcoming and alternative choices, normalized to the average firing rate at the start of the maze. The p-values indicate the results of a Mann-Whitney U test for the firing rates during upcoming and alternative representations for each of the SWR modulation categories. (c) A histogram of the phase of hippocampal theta oscillations at which the start of each representation of upcoming and alternative choices in the hippocampus occurred. The p-values indicate the results of a Kolmogorov-Smirnov test for difference in distributions between the representations of alternative and upcoming choices compared to shuffled decoded probabilities. (d) Left: circular histograms of the phase locking shown in (c). Right: The phase locking score of hippocampal representations of upcoming and alternative choices (Rayleigh’s Z, dark and light brown dots) to hippocampal theta oscillations compared to a randomized population (gray histograms). P-values indicate the result of a left-tailed Monte Carlo test of the Rayleigh’s Z compared to the shuffled population. Due to less multi-unit activity recorded in the hippocampus, fewer sessions were available with good decoding quality, and figure is made with N=12 sessions from 3 animals.