Hippocampal Spatial Memory Representations in Mice are Heterogeneously Stable Significance Statement

Significance statement: Single photon calcium imaging using head-mounted miniature microscopes in freely moving animals has enabled researchers to measure the long term stability of hippocampal pyramidal cells during repeated behaviors. Previous studies have demonstrated instability of neural circuit components including dendritic spines and axonal boutons. It is now known that single units in the neuronal population exhibiting behaviorally relevant activity eventually become inactive and that previously silent neurons can quickly acquire task-relevant activity. The function of such population dynamics is unknown. We show here that population dynamics differ for cells coding distinct task dimensions, suggesting such dynamics are part of a mechanism for latent memory reorganization. These results add to a growing body of work showing that maintenance of episodic memory is an ongoing and dynamic process. Figure 1. A, Task outline: each trial has a Study and Test Phase, separated by a 20-second delay. Each trial is followed by a ~15s inter-trial interval in the mouse’s home cage, adjacent to the alternation maze (not shown). B, Performance of individual mice (separate colors) over all days of recording. Only sessions with performance above 70% were included, excluded sessions are marked in red. C, Example viral expression and lens placement in dorsal CA1. Green is GCaMP6f-EYFP, blue is DAPI. D, Map showing regions for activity analysis. Purple indicates Central Stem, Orange indicates return arms. E, Top: Activity maps for one cell (a Turn Splitter Neuron; see Figure 2) over five days of recording. Each plot represents the average activity map for one set of task conditions, ordered clockwise from top-left: Study-Left, Study-Right, Test-Right, Test-Left. In each plot, the black trace is the animal’s recorded position, and colored dots indicate frames where the cell was active. Dots are colored based on the local transient likelihood, normalized by local occupancy, where red is the highest transient likelihood within that day and blue is the lowest. Bottom: Cell ROI masks for that recording day. Cell of interest is colored in green, and indicated with red arrow on first day shown. Masks were aligned across days based on relative positions of cells and cells were aligned based on distance between cell centers and correlation of masks. F, Same as E but for a place cell. Figure 2. A, Example activity maps for each type of splitter neuron on the central stem. B, Percentage of cells in each recording that had at least one calcium event on the central stem. Colored lines indicate individual animals, black line is best fit regression. Statistic: Spearman rank correlation. C, Percentage of splitter cells out of the active cell population on each day for all animals. Box shows inter-quartile range and middle line shows median. Statistic: Wilcoxon signed-rank test. D, Percentage of splitter neurons in individual animals (unique colors) and group regression (black) over the course of the experiment. Color of box indicates splitter type as described by y-axis label. Significance calculated with Spearman rank correlation between percentage of splitters and recording day number for all included sessions (n=38). E, Same as B but for percentage of cells that exceed the activity threshold (see Methods). F, Same as C but for reliably active cells. G, Distribution of centers-of-mass of event activity for Turn and Phase splitter neurons. Statistic: Mann-Whitney U-test. H, same as D but for reliably active cells. * p<0.05, ** p<0.01, ***p<0.001 Figure 3. A, Method for making population vector correlations. Calcium event likelihoods for one day, from one trial type, and from one spatial bin are correlated against event likelihoods from another (or the same) day and trial type but in the same spatial bin. Calcium event likelihoods are included for all cells found on both recording days of interest. B, Population vector correlations between trials of the same turn direction and task phase (gray), different turn directions (red) and different task phases(blue). Correlations in this panel B are generated from trials that occur on the same day. Shaded patch indicates 95% of points for the indicated correlation type in that spatial bin, trend line indicates mean. Statistic: Wilcoxon rank-sum test on all points for these groups. C,D, Mean correlation for pairs of spatial bins over the course of recordings. Thin lines indicate individual animals’ correlations, bold lines are best fit regression. Statistic: Spearman rank correlation on points from all recording days. E,F, Correlations between trials on separate recording days for indicated pairs of spatial bins. See text and supplementary data tables for statistics. * p<0.05, ** p<0.01, ***p<0.001 Figure 4. A, Percentage of cells that are still present at increasing day lags. Statistic: Wilcoxon signedrank test. B, Percentage of each splitter type by what that cell was on the prior day of recording. C, Percentage of active cells that were inactive on the prior day of recording. Colored lines are individual animals, black line is best fit regression. Statistic: Spearman rank correlation. D, Percentage of each splitting phenotype among each recording day’s set of previously inactive cells (from second recording day forward). Statistic: Wilcoxon signedrank test. E, Changes in the distribution of splitting phenotypes among previously inactive over the course of recordings. Colored lines are individual animals, black line is best fit regression. Color of box indicates cell type as described by y-axis label. Statistic is indicated at right: Spearman rank correlation. * p<0.05, ** p<0.01, ***p<0.001