Comparing the Representation of a Simple Visual Stimulus across the Cerebellar Network

Abstract

The cerebellum is a conserved structure of the vertebrate brain involved in the timing and calibration of movements. Its function is supported by the convergence of fibers from granule cells (GCs) and inferior olive neurons (IONs) onto Purkinje cells (PCs). Theories of cerebellar function postulate that IONs convey error signals to PCs that, paired with the contextual information provided by GCs, can instruct motor learning. Here, we use the larval zebrafish to investigate (1) how sensory representations of the same stimulus vary across GCs and IONs and (2) how PC activity reflects these two different input streams. We use population calcium imaging to measure ION and GC responses to flashes of diverse luminance and duration. First, we observe that GCs show tonic and graded responses, as opposed to IONs, whose activity peaks mostly at luminance transitions, consistently with the notion that GCs and IONs encode context and error information, respectively. Second, we show that GC activity is patterned over time: some neurons exhibit sustained responses for the entire duration of the stimulus, while in others activity ramps up with slow time constants. This activity could provide a substrate for time representation in the cerebellum. Together, our observations give support to the notion of an error signal coming from IONs and provide the first experimental evidence for a temporal patterning of GC activity over many seconds.

Keywords: cerebellum; imaging; larval zebrafish; luminance; temporal patterning.

Figure 1.

Experiment description and anatomy. A, Experimental setup: a two-photon microscope was used to image 7 dpf head-restrained larvae while flashes of different luminance were projected on a screen below. B, Composite image showing stacks from GCs (red), IONs (in blue), and PCs (green) registered on a reference anatomy with the whole brain of a 7 dpf larva (in gray). In the enlargements, examples of the raw anatomy (in gray) and the segmented ROIs from a two-photon imaging session (in red, blue, and green for GCs, IONs, and PCs, respectively). See Extended Data Figure 1-1 for more details on the behavior elicited.

Figure 2.

Responses of GCs and IONs to the “luminance steps” protocol. A, Left, Histogram of the average intertrial correlation (average correlation between activity recorded in different trials). The calculated threshold and the relative fraction of active cells are reported on the histogram. Center, Average traces grouped after hierarchical clustering. Cutting the dendrogram at the height marked by the line (on the left) resulted in eight different clusters, whose average activity is shown on the right, overimposed on a shade matching at each timepoint the brightness level displayed. B, Same plot as in A, for IONs. See Extended Data Figure 2-1 for more details.

Figure 3.

GC and IONs respond to stimulus intensity and derivative. A, Regressor-based analysis was used to calculate the correlation coefficient between each cell’s fluorescence (lines) and a panel of regressors (shades). The plot shows the cells with the highest correlation values for luminance-related (first two rows) and transition-related (second two rows) regressors. B, Histogram of the distribution of best-fitting regressors for GCs (top) and IONs (bottom). Regressors on the left of the dashed line are luminance related, and regressors on the right are transition related. GCs score higher in luminance-related regressors compared with IONs. C, Scatter plot of the best transition-related coefficient and the best luminance-related coefficient for GCs (red) and IONs (blue) and their relative marginal distributions. The GC population has been downsampled randomly to match the number of IONs, while the marginal distributions refer to the entire population. GCs cluster in the bottom right quadrant of the plot (high luminance–regressor correlation, low transition–regressor correlation), while IONs show higher correlation coefficients for transition-related regressors. D, Performance of a nonlinear decoder used to predict luminance values from GCs (left) and IONs (right) activity. Each point in the swarm plot is one frame of the protocol with the corresponding luminance level (horizontal line). For GCs, 20 iterations of the decoding analysis were performed, with the number of GCs downsampled to match that of IONs. The violin plots in the left panel show the average distribution of predictions across all iterations, while the dots correspond to a single representative iteration. See Extended Data Figure 3-1 for more details.

Figure 4.

Responses of GCs and IONs to the “flashes” protocol. A, Top-left, Intertrial correlation histogram of GCs for the “flashes” protocol. Center, Cells sorted after hierarchical clustering and (right) average activity for each cluster. B, Same plot as in A, for IONs. See Extended Data Figure 4-1 for more details.

Figure 5.

Granule cell activity is temporally patterned. A, Single traces for early- (green), sustained- (gray), and late-responding (brown) ROIs. The average response (thick line) is superimposed on all single repetitions (thin lines) of stimulus presentation. B, GCs included in any of the luminance-excited clusters were then sorted based on their center of mass (COM) during the longest flash, with earlier-responding neurons on the top and late-responding ones on the bottom (bottom panels). The above traces represent the average response for ROIs binned according to the time at which their COM was reached (green for early-responding neurons, brown for late-responding ones). C, Same plot as in B, for IONs. In this case, traces in the top panels correspond to single ROI responses. D, Probability density function describing the time points at which responses of neurons from the two imaged populations reach their maximal response. E, Scatter plot of predicted versus actual time elapsed stimulus onset. Prediction was performed at each frame from ION activity (blue dots) and activity of a random subset of GCs (red dots). The red line and shaded area show average ± standard deviation of predictions from 200 GCs subsets. See Extended Data Figure 5-1 for more details.

Figure 6.

PC responses to luminance stimuli. A, Top-left, Histogram of the average intertrial correlation and its relative threshold. Center, Hierarchical clustering of PC responses to the “steps” protocol, and average activity of the clusters selected after cutting the dendrogram at the height marked by the lines. The shade matches at each timepoint the brightness level displayed. B, Same as in A, for the “flash” protocol. See Extended Data Figures 6-1 and 6-2 for more details.