3D tracking and characterization of the structure of mouse spontaneous grooming.

A. Schematic of experimental setup. Mice were placed into an equilateral triangular arena made of transparent acrylic (12-inch sides and height) and their behavior was captured with three side view cameras. Schematic is not to scale. B. Example video frames of mice during spontaneous grooming. C. Overview of our 4-step grooming identification approach. D. Distribution of the percent of each session that mice spent spontaneously grooming (6 mice, 63 sessions, 117 experiment hours, 1,475 grooming bouts, 304.8 total minutes of spontaneous grooming). E. Distribution of spontaneous grooming bout durations in seconds (for the same dataset as in D). F. Distribution of inter-bout intervals in minutes (for the same dataset as in D).

Manual refinement of grooming identification.

A. Example annotations of grooming behavior by four annotators. Green regions in each row denote times that the annotator labeled as grooming and gray regions denote times not annotated as grooming. Vertically aligned white spaces indicate breaks of variable time between individual grooming bouts. Average pairwise Jaccard Index (or Intersection over Union) computed on all annotations from a single 2-hour session was 0.76 ± 0.04.

Striatal cell type specific activity mapped to spontaneous self-grooming.

A. Classification of recorded units. Left: units were categorized as either SPNs, FSIs, or ‘other’ by their firing rates, spike waveform peak width, and duration between spike waveform peak to valley (SPN: 88.2% 2,755 units, FSI: 3.2% 100 units, other: 8.6% 269 units). Right: average +/- SEM spike waveforms for units in each category. B. Activity of 4 example SPNs during grooming. Neural activity are aligned to the start of grooming bouts, denoted by t = 0. Top: spike raster plots for the example neuron during each grooming bout in the given session. Grooming bouts are sorted by grooming bout duration, denoted by the green rectangles. Bottom: unit average firing rate aligned to grooming start and normalized to [0, 1]. C. Activity of 4 example SPNs as in B, but for aligned to groom end. D. Activity of 4 example units as in B, but for FSIs aligned to groom start. E. Activity of 4 example units as in B, but for FSIs aligned to groom end.

Probe placement and firing rate characteristics.

A. Schematic depicting how the four shanks within our recording electrodes were implanted along the anterior-posterior axis (shanks span 600 μm). B. Representative coronal slices of electrode placements. Coronal histological slices were obtained and registered to the Allen Common Coordinate Framework (CCFv3) using the WholeBrain software package. Blue arrows depict the most ventral electrode position for each mouse. Blue lines depict the region covered by sites in one shank of each electrode. Anterior-Posterior (A/P) coordinates from Bregma used for registration are depicted above each slice. (6 mice, scale bar DV and ML = 1 mm). C. Distribution of firing rates for all recorded units (3,124 units. 6 mice). D. Distribution of inter-spike intervals for all recorded units (3,124 units. 6 mice).

Emergent motifs in SPN and FSI activity around grooming transitions.

A. First principal component from decomposing 10 seconds of SPN activity centered around grooming bout start times (1,632 units, explains 9.3% of variance). B. Activity around grooming start for units that had the largest magnitude weight for the first principal component. Units with weight > 0 are shown above the horizontal black line (336 units) followed by units with weight < 0 (640 units). Units are sorted by their weight for PC1. Each unit’s activity was normalized to the range from zero to one. C. Same as A, but for the second principal component (explains 5.1% of variance). D. Same as B, but for the second principal component (364 units with weight > 0, 292 units with weight < 0). E. Donut plot depicting the number of units with positive and negative weights for the first two principal components. F-J. Same as A-E, but for FSIs aligned to groom start (66 units with PC1 weight > 0, 1 unit with PC1 weight < 0, 2 units with PC2 weight > 0, and 19 units with PC2 weight < 0. PC1 explains 21.2% of variance and PC2 explains 7.1% of the variance). K-O. Same as A-E, but for SPNs aligned to groom end (526 units with PC1 weight > 0, 424 units with PC1 weight < 0, 164 units for PC2 weight > 0, and 518 units with PC2 weight < 0. PC1 explains 12.9% of variance and PC2 explains 5.2% of the variance). P-T. Same as A-E, but for FSIs aligned to groom end (22 units with PC1 weight > 0, 44 units with PC1 weight < 0, 5 units with PC2 weight > 0, and 19 units with PC2 weight < 0. PC1 explains 34.6% of variance and PC2 explains 8.6% of the variance).

Identification and characterization of striatal grooming ensembles.

A. Heatmap of activity during grooming for all units in an example session (53 units, 6.55 minutes). Arrows point out a subset of synchronous events. Each unit’s activity is normalized to [0, 1]. B. Correlation matrix for the activity shown in A. C. Distribution of statistical estimate for the number of ensembles within a given session (4 mice, 33 sessions). The estimate is obtained by computing the number of eigenvalues from the data that are above the 99th percentile of the distribution of eigenvalues from 5,000 random shuffles of the data. D. Average cumulative distribution of pairwise unit distances for each pair of units that are within the same cluster (red) and each pair of units that are not within the same cluster (blue). Plot depicts the mean ± SEM of the cumulative distribution across mice (4 mice). E. Distribution of pairwise unit correlations during grooming for each pair of units that are within the same cluster (pink 2,747 pairs) and each pair of units that are not within the same cluster (blue 27,421 pairs). Left: histogram. Right: cumulative distribution. F. Same as in E, but for pairwise correlations computed from activity during the entire session including grooming times. G. Comparison between the difference in pairwise unit correlations for units within and not within the same cluster computed during grooming (green) and during the whole session (orange) (grooming AUC: 0.81, whole session AUC: 0.69).

Statistics of identified clusters and activity patterns.

A. Heatmap of activity during grooming with shuffled time bins for all units in the example session shown in 4A (53 units, 6.55 minutes). Arrows are in the same position as in Figure 4A highlighting the absence of synchrony after shuffling. Each unit’s activity is normalized to the range from zero to one. B. Correlation matrix for the shuffled activity shown in A. C. Total number of SPNs and FSIs that were clustered (pink) and unclustered (blue) (4 mice, 33 sessions, 243 clusters). D. Total number of clusters that comprised of only SPNs (purple), only FSIs (green), and of both SPNs and FSIs (orange) (data as in C). E. Distribution of the number of clusters in each session. F. Relationship between the number of clusters found in each session and the number of units in that session (R2 = 0.239, p = 0.004). G. Distribution of the number of units in each cluster (data as in C). H. Relationship between the average cluster size in each session and the number of units in that session (R2 = 0.01, p = 0.582). I. Cumulative distribution of the percent of clustered units in each session. Dashed lines denote that half the sessions have at least 82% clustered units (data as in C).

Striatal ensembles encode features of self-grooming.

A. Heatmap depicting the grooming-aligned average activity of all striatal ensembles. Ensemble averages are normalized to the range from zero to one. Prior to averaging across grooming bouts, ensemble activity during each grooming bout, excluding the 5 seconds before and after grooming, was linearly time warped to a fixed duration. Ensembles are sorted by peak time (112 ensembles, 4 mice, 33 sessions). B. Distribution of time at which ensemble average activity peaked (data as in A). C. Representative example of an ensemble with increased average activity around grooming transitions. Grey region denotes the range of activity for shuffled ensemble activity. Bottom of the range depicts the 2.5th percentile of the shuffled activity, top of the range depicts the 97.5th percentile of the shuffled activity, and grey line depicts the average shuffled activity. D. Representative example of an ensemble with increased average activity at the start of grooming (grey region as in C). E. Representative example of an ensemble with increased average activity at the end of grooming (grey region as in C). F. Representative example of an ensemble with increased average activity throughout the duration of grooming (grey region as in C). G. Neural trajectories traced out in factor space by the population of units recorded during all grooming bouts in an example session (65 units, 22 grooming bouts). Colors depict different grooming bouts, pluses denote the start of grooming, and asterisks denote the end of grooming. H. Neural trajectories traced out in factor space by the population of clustered units during all grooming bouts in an example session (49 units, 22 grooming bouts). Visualization elements as in G. I. Ensemble trajectories traced out in factor space by the striatal ensembles during all grooming bouts in an example session (13 ensembles, 22 grooming bouts). Visualization elements as in G.

Patterns of cluster engagement.

A. Heatmap depicting the percent of units within each ensemble that are active during grooming averaged across all grooming bouts in a given session. Prior to averaging across grooming bouts, ensemble activity during each grooming bout was interpolated to a fixed duration, excluding the 5 seconds before and after grooming. Ensembles are in the same order as in Figure 5A and data are normalized to the range from zero to one (112 ensembles, 4 mice, 33 sessions). B. Distribution of times when the percent of active units in an ensemble peaked (data as in A). C. Representative example of an ensemble with decreased average activity throughout the duration of grooming (grey region as in Figure 5C). D. Neural trajectories traced out in factor space by the population of unclustered units during all grooming bouts in an example session (16 units, 22 grooming bouts). Visualization elements as in Figure 5G.