Cholinergic axons in visual cortex are activated during locomotion independent of visual stimuli.

(A) GCaMP6s was expressed by injecting an AAV2/5-hSyn1-FLEx-axon-GCaMP6s vector in basal forebrain of ChAT-IRES-Cre mice. We then imaged calcium activity of cholinergic axons projecting to visual cortex.

(B) Example two-photon image of cholinergic axons in visual cortex.

(C) Schematic of the virtual reality system. Mice were head-fixed on a spherical treadmill and surrounded by a toroidal screen on which we presented visual stimuli in different visuomotor conditions.

(D) Example trace of the calcium activity of one cholinergic axon in visual cortex (red) and the corresponding locomotion velocity of the mouse (blue).

(E) Average locomotion onset activity of all cholinergic axons (5048 axons in 14 mice), sorted by their average response during locomotion onset.

(F) Average response of cholinergic axons to locomotion onset over all the cholinergic axons, of the data shown in E. Shading indicates SEM over axons. Due to SEM being very small, the shading is fully obscured by the red line.

(G) As in F, but for grating onsets. Green shading marks the duration of grating presentation.

(H) As in F, but for visual flow onset during the open loop condition.

(I) As in F, but for visuomotor mismatch onset. Orange shading marks the duration of visuomotor mismatch. As mismatch events occur only during times of locomotion, and locomotion itself is a strong driver of cholinergic activity, we would expect to find an increase in cholinergic activity by chance at mismatch. To correct for this, we quantified the distribution of cholinergic activity on random triggers during locomotion (95% confidence interval (CI), gray shading).

(J) Locomotion onset activity in different visuomotor conditions.

(K) The fraction of cholinergic axons responsive to locomotion, grating, and visuomotor mismatch onset, quantified for each imaging site. Each datapoint is one imaging site. Boxes show 25th and 75th percentile, central mark is the median, and the whiskers extend to the most extreme datapoints not considered outliers. Dashed line marks chance level. n.s.: not significant; *: p < 0.05; **: p < 0.01; ***: p < 0.001; see Table S1 for all statistical information.

Activity in cholinergic axons was better explained by a binary locomotion state signal than by locomotion velocity.

(A) Average locomotion velocity profiles aligned to locomotion onset and offset. The isolated datapoint between the two traces is the average locomotion velocity over the locomotion bout. Shading marks SEM. Here and elsewhere, n.s.: not significant; *: p<0.05; **: p<0.01; ***: p<0.001; see Table S1 for all statistical information.

(B) As in A, but for activity of all cholinergic axons in visual cortex, for the data shown in D. Shading marks SEM over axons.

(C) Bout-by-bout correlation coefficient between the locomotion onset and offset changes for average cholinergic activity and locomotion velocity. Each datapoint is one imaging site. Boxes show 25th and 75th percentile, central mark is the median, and the whiskers extend to the most extreme datapoints not considered outliers.

(D) Average onset and offset responses for all cholinergic axons sorted by strength of locomotion onset response.

(E) Average activity of cholinergic axons as a function of locomotion velocity. We found no evidence of a difference in activation between low and high locomotion velocities (indicated by gray shading). Red shading marks SEM.

(F) Left: For all axons, the correlation coefficient of their calcium activity with locomotion velocity plotted against the correlation coefficient with a binarized version of the locomotion velocity. Right: Distributions of correlation coefficients between calcium activity and locomotion velocity and binarized velocity. Same data as shown on the left. Boxes show 25th and 75th percentile, central mark is the median, and the whiskers extend to the most extreme datapoints not considered outliers.

Average activity of cholinergic axons was more strongly correlated with locomotion velocity than with pupil diameter.

(A) Correlation of average activity of cholinergic axons activity with locomotion velocity in closed loop, open loop, dark, and grating conditions. Each point represents data from one imaging site, error bars indicate SEM. Here and elsewhere, n.s.: not significant; *: p < 0.05; **: p < 0.01; ***: p < 0.001; see Table S1 for all statistical information.

(B) As in A, but for correlation between average activity of cholinergic axons and pupil diameter.

(C) As in A, but for correlation between locomotion velocity and pupil diameter.

Optogenetic activation of cholinergic axons in visual cortex primarily enhances responses of layer 5, but not layer 2/3 neurons.

(A) Average calcium response of layer 2/3 neurons in visual cortex to full field drifting gratings in the absence (light brown) or presence of locomotion (brown). Green shading indicates duration of grating stimulus. The responses are compared bin-by-bin using a nested hierarchical bootstrap test. Here and in subsequent panels, bins with a significant difference (p < 0.05) are indicated by a black line above the plot; those with p > 0.05 are marked gray. Shading marks SEM.

(B) As in A, but for layer 5 neurons in visual cortex in the absence (light blue) or presence (dark blue) of locomotion.

(C) Average calcium response of layer 2/3 neurons in visual cortex to full field drifting gratings while the mice were stationary, without (light brown) or with (pink) optogenetic activation of cholinergic axons in visual cortex. Duration of optogenetic stimulation is marked by a pink bar. Green shading indicates duration of the grating stimulus.

(D) As in C, but for layer 5 neurons in visual cortex.

(E) Average calcium response to visuomotor mismatch in layer 2/3 neurons in visual cortex, without (brown) and with (pink) optogenetic stimulation of cholinergic axons in visual cortex. Orange shading indicates mismatch duration, and pink bar indicates optogenetic stimulation.

(F) As in E, but for layer 5 neurons in visual cortex.

(G) Average calcium response of layer 2/3 neurons in visual cortex to locomotion onset in closed loop, without (brown) or with (pink) optogenetic stimulation of cholinergic axons in visual cortex.

(H) As in G, but for layer 5 neurons in visual cortex.

Optogenetic activation of cholinergic axons in visual cortex did not drive responses in either layer 2/3 or layer 5 excitatory neurons.

(A) Average calcium response of layer 2/3 neurons in visual cortex to locomotion onset in open loop. Shading indicates SEM.

(B) As in A, but for layer 5 neurons in visual cortex.

(C) Average calcium response of layer 2/3 neurons in visual cortex during local optogenetic stimulation of cholinergic axons while the mice were stationary. Pink bar marks duration of the optogenetic stimulus.

(D) As in C, but for layer 5 neurons in visual cortex.

Cholinergic axon activation reduced pairwise correlations between neurons in visual cortex.

(A) Average pairwise correlation in calcium activity of layer 2/3 neurons while the mouse was stationary and while the mouse was locomoting. Each dot is the mean of the pairwise correlations for one neuron to all others in the same field of view.

(B) As in A, but for layer 5 neurons.

(C) As in A, but comparing the correlation while the mice were stationary, to that while cholinergic axons were stimulated optogenetically.

(D) As in C, but for layer 5 neurons. Note, for this experiment the imaging plane was in L2/3 to simultaneously image layer 2/3 neurons and apical dendrites of layer 5. The latter were traced to their soma in layer 5 offline and were used to compute layer 5 activity correlations.

(E) Average change in pairwise correlations of layer 2/3 neurons on locomotion (gray) or on optogenetic stimulation of cholinergic axons during stationary periods (pink), compared to the correlations during stationary (stat.) periods at baseline. Here and elsewhere, n.s.: not significant; *: p < 0.05; **: p < 0.01; ***: p < 0.001; see Table S1 for all statistical information.

(F) As in E, but for layer 5 neurons of visual cortex.

(G) Change in the average pairwise correlations of layer 2/3 neurons during locomotion, upon DREADD activation (pink) or inhibition (orange) of basal forebrain cholinergic neurons.

(H) As in G, but for layer 5 neurons.

Cholinergic input reduced response latency of layer 5 neurons in visual cortex to grating stimuli.

(A) Locomotion (brown) or optogenetic stimulation of cholinergic axons (pink) induced change in the response latency of layer 2/3 neurons in visual cortex to full field drifting grating onset. Here and elsewhere, n.s.: not significant; *: p<0.05; **: p<0.01; ***: p<0.001; see Table S1 for all statistical information.

(B) As in A, but for layer 5 neurons.

Statistical information on all analysis.

For statistical comparisons we used paired or unpaired t-tests, or hierarchical bootstraps as indicated in the table below. The units (ROIs, sites, or mice) over which testing was done are boldfaced. For hierarchical bootstraps the two levels of units used are boldfaced.

Number of mice in different experimental groups

Cholinergic axons in visual cortex were activated after layer 2/3 but before layer 5 visual cortex neurons. Related to Figure 1.

(A) Distribution of time to first significant response after locomotion onset for cholinergic axons in visual cortex. Axons without any response to locomotion onset are omitted.

(B) As in A, but for layer 2/3 neurons in visual cortex.

(C) As in A, but for layer 5 neurons in visual cortex.

(D) Nested bootstrap resampling of the distributions shown in A-C. Here and elsewhere, n.s.: not significant; *: p<0.05; **: p<0.01; ***: p<0.001; see Table S1 for all statistical information.

Responses of GRAB-ACh3.0 in visual cortex are similar to the calcium responses in cholinergic axons. Related to Figure 2.

(A) Average change in GRAB-ACh3.0 fluorescence during locomotion onset. We used the fluorescence change measured in blood vessels (purple) to estimate hemodynamic occlusion and subtracted this from the raw activity (black) to generate a corrected response (red). Shading marks SEM.

(B) As in A, but for the GRAB-ACh3.0 response to grating onsets. The green shading marks the duration of the grating stimulus.

(C) As in A, but for the GRAB-ACh3.0 response to visuomotor mismatch. The orange shading marks the duration of the visuomotor mismatch.

(D) To estimate hemodynamic occlusion effects while two-photon imaging, we measured fluorescence changes on locomotion onset in layer 2/3 neurons in visual cortex that express GFP during locomotion onset.

(E) As in D, but for responses to grating onsets.

(F) As in D, but for responses to visuomotor mismatch.

(G) The correlation coefficient between the population activity vectors during locomotion onset (on), offset (off), and during locomotion bout (btw), compared to shuffle (rnd.) controls. Here and elsewhere, n.s.: not significant; *: p<0.05; **: p<0.01; ***: p<0.001; see Table S1 for all statistical information.

(H) GRAB-ACh3.0 fluorescence as a function of locomotion velocity. We found no evidence of a difference in fluorescence between low and high locomotion velocities. Red shading indicates SEM.

Cholinergic axon activity correlated better with a binarized locomotion vector than with the unfiltered locomotion vector over a large range of binarization thresholds. Related to Figure 2.

The correlation between activity of cholinergic axons in visual cortex and locomotion velocity was higher when we used a binarized version of the locomotion velocity trace. This difference was present over a large range of thresholds. The black bar at the top of the plot indicates threshold values for which the correlation with binarized locomotion velocity was higher than that with raw locomotion velocity.

Activity of cholinergic axons correlated better with locomotion velocity than with facial movement. Related to Figure 3.

(A) Standard deviation of pupil diameter as a function of visuomotor condition. Here and elsewhere, n.s.: not significant; *: p < 0.05; **: p < 0.01; ***: p < 0.001; see Table S1 for all statistical information.

(B) Average correlation of activity of cholinergic axons in visual cortex with facial movements as a function of visuomotor condition, compared with the same activity correlation to locomotion velocity (same data as Figure 3A). Error bars indicate SEM. Individual data points are imaging sites.

(C) Average correlation of activity of cholinergic axons in visual cortex and locomotion velocity (blue), pupil diameter (red), or facial movement (black) as a function of smoothing window. The black arrows marks the smoothing window used for facial movements.

Optogenetic activation did not increase grating responses during locomotion and had no effect in opsin-negative control mice. Related to Figures 4 and 6.

(A) Average calcium response of layer 2/3 neurons in visual cortex to full field drifting gratings (green shading) presented while the mice were locomoting, without (brown) and with (pink) optogenetic stimulation of cholinergic axons in visual cortex. The duration of optogenetic stimulation is marked by a pink bar. The responses are compared bin-by-bin using a nested hierarchical bootstrap test. Here and in subsequent panels, bins with a significant difference (p < 0.05) are indicated by a black line above the plot; those with p > 0.05 are marked gray. Shading marks SEM.

(B) As in A, but for layer 5 neurons.

(C) Average calcium response of layer 5 neurons in visual cortex to full field drifting gratings (green shading) presented while the mice were stationary, without (pale blue) and with (pale pink) optogenetic stimulation in mice that did not express ChrimsonR in cholinergic axons. The duration of stimulation is marked by a pink bar.

(D) Average calcium response of layer 5 neurons in visual cortex to visuomotor mismatch (orange shading) without (blue) and with (pale pink) optogenetic light stimulation in mice that do not express ChrimsonR in cholinergic axons. The duration of stimulation light is marked by a pink bar.

(E) As in D, but for locomotion onset responses. Note, the locomotion onset threshold was relaxed to 0.025 cm/s to increase the number of onsets we could analyze.

(F) Change in pairwise correlations in layer 2/3 neurons and in layer 5 neurons in visual cortex during sham stimulation while the mice were stationary. Here and elsewhere, n.s.: not significant; *: p < 0.05; **: p < 0.01; ***: p < 0.001; see Table S1 for all statistical information.

Locomotion increases the correlation between those layer 5 neurons in visual cortex that on average have a higher correlation. Related to Figure 6.

(A) Change in the pairwise correlation between pairs of layer 2/3 neurons in visual cortex on locomotion and their average correlation over the entire imaging session.

(B) As in A, but for layer 5 neurons in visual cortex.

(C) Related to A, with all neuron pairs binned by their average correlation. Bars indicate SEM.

(D) As in C, but for layer 5 neurons in visual cortex.

Chemogenetic manipulation of basal forebrain cholinergic neurons affects locomotion behavior of mice. Related to Figure 6.

(A) Average locomotion velocity before and after administration of CNO in ChAT-Cre mice that expressed a DREADD activator (pink) or a DREADD inhibitor (orange) in basal forebrain. Each dot is an experimental session.

(B) Quantification of the data shown in A. Here and elsewhere, n.s.: not significant; *: p < 0.05; **: p < 0.01; ***: p < 0.001; see Table S1 for all statistical information.

(C) As in A, but for fraction of time spent running.

(D) Quantification of the data shown in C.

(E) CNO induced change in the total distance traveled.