Water drop stimulus paradigm evokes grooming behaviors with a rich diversity of movements.

A) Experimental setup. B) Trial structure for an evoked grooming session. C) Experiment timeline. D) Visualization of paw trajectories in UMAP space. Each point represents an individual behavior. Points are colored by cluster. E) Example behaviors from clusters in (D). Colored lines show paw trajectories for 200 individual example behaviors in each behavior cluster. F) Rasters of grooming events during spontaneous sessions without a water drop stimulus (white), and evoked sessions with 20 water drop stimuli (blue). Probability of grooming (bottom) for spontaneous and evoked sessions. G) Quantification of number of grooming episodes. p=2.1e-12, t-test. H) Relative frequency of each grooming behavior during spontaneous and evoked sessions. Images in (A) and (E) are adjusted for brightness to enhance visibility.

Methods for detecting grooming behaviors.

A) Workflow for identifying and validating grooming behaviors. In the unsupervised method, features were generated from DeepLabCut tracks from keypoints on the paws (Figure 1A). B) Manual labels plotted in the same UMAP space as shown in Figure 1D. In the manual labelling scheme, large bilateral grooming events were classified as a separate group. These behaviors were extremely rare and were not categorized by the unsupervised method described in (A). C) Paw trajectories (top) and position probability density (bottom) for the right (red) and left (blue) paw, for all behaviors across all mice. The number of behaviors are displayed above.

Lateralized water drop stimulus elicits lateralized responses.

Drops directed towards the right side of the mouse significantly increased unilateral and asymmetric actions on the right side, but not elliptical asymmetric actions. Drops directed on the left side significantly increased unilateral actions on the left side, but did not result in significantly increased asymmetric or elliptical asymmetric actions directed towards the left.

Temporal presentation of grooming behaviors.

A) Example grooming episodes. Note that the examples shown here are longer than the average grooming events (see panel B). These examples are intended to showcase the variability in constitution of grooming episodes. B) Histogram of episode duration. C) Proportional composition of grooming episodes of various durations. Data compiled across all mice. D) Average transitions between grooming behaviors across all mice. E) Directed network representation of transition matrix in (D). Nodes are arranged using MATLAB’s ‘layered layout’ organization. Edge thickness is proportional to transition probability between nodes. Self-connections are omitted for clarity. Nodes are colored by Louvain community assignment performed on the transition matrix.

Licking and grooming behaviors are temporally overlapping.

A) Percentage of lick events that overlap with each grooming behavior. B) Percentage of grooming behaviors that overlap with lick events. Data are compiled across all mice and sessions (n=7877 licks). N for individual grooming behaviors can be found in Figure 1 - figure supplement 1C.

Mesoscale cortical widefield activation during grooming.

A) Example image of GCaMP6s fluorescence in a Thy1-GCaMP6s mouse with the Allen Institute cortical atlas overlaid. (Abbreviations - MOs: secondary motor area; MOp: primary motor area; SSp-m: primary mouth somatosensory area; SSp-n: primary nose somatosensory area; SSp-ul: primary upper limb somatosensory area; SSp-ll: primary lower limb somatosensory area; SSp-tr: primary trunk somatosensory area; SSp-bf: primary barrel field somatosensory area; VISa: anterior visual area; VISrl: rostrolateral visual area; VISam: anteromedial visual area; VISpm: posteromedial visual area; VISp: primary visual area; RSPd: dorsal retrosplenial area) B) Averaged activity dF/Fo across the entire cortical surface. Grooming behaviors indicated at the top (black bars). C) Ethogram (top) and mean cortical calcium activity (bottom) during the grooming episode indicated in (B). Best fit line through cortical activity shown in red. D) Heatmap (top) of averaged activity across the entire cortical surface for all grooming episodes which exceeded 10s. The dotted line at ∼21s is drawn at the 50th percentile duration. The mean cortical response over rows in the heatmap. Shading shows the standard deviation. E) Comparison between averaged cortical signals at various time points. Lines indicate significance. p<0.05 repeated measures ANOVA with post-hoc Bonferroni correction. F) Slope of the best fit line for all grooming episodes exceeding 10s. Slope is less than 0. p<0.05 t-test. Red marker indicates example shown in (C). G) Response maps for a single mouse averaged across all frames in which the behavior is expressed (top). Averaged response maps for each mouse binarized at >80th percentile to show maximally activated networks (bottom). Each colored contour reflects a different mouse.

Cortical activation maps during spontaneous and evoked grooming behaviors.

Binary response maps for each behavior are shown in the same format as in Figure 3G. Averaged response maps for each behavior are binarized at >80th percentile to show the maximally activated networks during evoked (top) and spontaneous (bottom) grooming behaviors. The total number of events across all animals are indicated below. Maps with only one colored contour indicate that only one mouse exhibited this behavior.

Regression analysis finds similar network involvement.

A) Example ethogram of a single grooming event. The entire binary vectors for all behaviors are used as regressors in the ridge regression model. B) Time-varying coefficients solved for by the regression model. Each map shows the weights of the kernel at various time points. Time series data on the right show the weights as a function of time for cortical regions corresponding to those shown in (C). Vertical lines correspond to time points at which images are shown. C) Top: Example dFF traces from cortical areas (black) and model estimation (color). Middle: Cortical regions from where example dFF traces are shown in top. Bottom: Model explained variance across the cortical window over all sessions. D) Top: Unique contribution of each model variable to the total explained variance for an example mouse. Bottom: Unique contributions for each mouse and behavior binarized at >80th percentile to show maximally contributing cortical networks (bottom).

Transient mesoscale cortical network activation observed during continuous drinking behavior.

A) Experimental setup. Mesoscale cortical calcium activity was imaged in water-restricted mice during continuous bouts of drinking. Movements of the paws and body were assessed with DeepLabCut. B) Trial structure. Following a brief baseline period, mice were given water for an extended period of time. Example licking rates from a single mouse over 12 sessions. Line and shaded areas indicate mean and standard error of the mean. C) Cortical activation during licking behavior in an example tTA-GCaMP6s mouse (top) and the maximally activated network across all mice (bottom, n=4). D) Averaged activity across the dorsal cortex from a tTA-GCaMP6s mouse over 4 example sessions. Licking and movement behaviors are annotated with magenta and blue shading, respectively. E) Time-varying coefficients solved for by ridge regression model for licking (top) and movement behaviors (bottom) in an example mouse. Each map shows the weights of the kernel at various time points. Time series data on the right show the weights as a function of time for the indicated cortical areas. F) Unique explained contribution for licking and movement variables to the total explained variance. Each mouse’s unique contribtions were binarized at >80th percentile to show the maximally contributing cortical networks.

Grooming-responsive neurons observed across cortical areas with 2-photon imaging.

A) Example session showing neuron locations in the cortical atlas. This session corresponds to examples shown in panel (B) and (G). Neurons are color coded by the populations highlighted in (B). Black neurons are unhighlighted population. B) Top: Rastermap embedding of neuronal fluorescence activity with behavior ethogram shown above. Colored lines on the left side of the rastermap show highlighted neuronal populations. Average of the highlighted populations are shown below. Bottom: Expanded view of the data. C) Percentage of imaged neurons that belong to the grooming-responsive population observed in each cortical region. D) Top: Heatmap showing averaged fluorescence across the grooming-responsive population of neurons, for each episode exceeding 10s in duration. Dotted line is drawn at 50th percentile of episode duration. Bottom: Averaged fluorescence across all grooming episodes. E) Comparison between averaged cortical signals at various time points. All time points significantly different. n=18 grooming episodes across 3 mice. p<0.05 repeated measures ANOVA with post-hoc Bonferroni correction. F) Slope of the best fit line for all grooming episodes exceeding 10s. Slope is less than 0. p<0.05 t-test. G) Heatmap of averaged neuronal fluorescence during each behavior. Cells are sorted by cortical region and responses to non-grooming forelimb movements. Data are from the example mouse shown in panels (A) and (B). Top: Dendrogram showing cosine distance between neuronal responses across behaviors. H) Averaged cosine similarity matrix across all mice. Louvain community groupings are denoted by colored squares.

Alignment of 2-photon images to the Allen Institute cortical atlas.

A) Example 2-photon resonant scan image showing neuron regions of interest detected with Suite2p. B) Example 2-photon linear scan image with larger field of view. Image from (A) was registered to (B). C) High resolution single-photon template image. Image from (B) was registered to (C). D) Protocol for mapping somatosensory cortex. E) Example response maps (left) and time courses within the indicated region from forelimb and hindlimb stimulation. F) Sensory response maps were overlaid on the high resolution template image, and used to register the template map to Allen Institute cortical atlas.

Correlation coefficients for all neurons across all mice with grooming behaviors and non-grooming forelimb movements.

A) Pearson correlation coefficients for all neurons (gray) and the grooming neurons identified with Rastermap (blue) with grooming behaviors. All neurons, n=2198, and grooming neurons n=418 over 7 sessions across 4 mice, Wilcoxon rank-sum test, p<0.01. B) Pearson correlation coefficients for each of the grooming neurons within the population highlighted in A (blue), with grooming and non-grooming forelimb movements. n=356 neurons, paired t-test, p<0.05