Purkinje Cell Activity Determines the Timing of Sensory-Evoked Motor Initiation

Abstract

Cerebellar neurons can signal sensory and motor events, but their role in active sensorimotor processing remains unclear. We record and manipulate Purkinje cell activity during a task that requires mice to rapidly discriminate between multisensory and unisensory stimuli before motor initiation. Neuropixels recordings show that both sensory stimuli and motor initiation are represented by short-latency simple spikes. Optogenetic manipulation of short-latency simple spikes abolishes or delays motor initiation in a rate-dependent manner, indicating a role in motor initiation and its timing. Two-photon calcium imaging reveals task-related coherence of complex spikes organized into conserved alternating parasagittal stripes. The coherence of sensory-evoked complex spikes increases with learning and correlates with enhanced temporal precision of motor initiation. These results suggest that both simple spikes and complex spikes govern sensory-driven motor initiation: simple spikes modulate its latency, and complex spikes refine its temporal precision, providing specific cellular substrates for cerebellar sensorimotor control.

Keywords: purkinje cells, simple spikes, complex spikestwo-photon imaging, optogenetics, neuropixels, cerebellum, multisensory, timing, motor initiation.

Graphical abstract

Figure 1

A Multisensory Association Task for Probing the Role of the Cerebellum in Rapid Sensory-Driven Behavior (A) Schematic of the multisensory association task and experimental setup. (B) Task structure. Air puff + tone (Go, magenta), air-puff-only (No-go 1, orange), and tone-only (No-go 2, green) trials were provided. During optogenetics experiments (under gray line), Purkinje cells were photostimulated in a subset of Go trials (Go+S, blue). (C) Representative performance in a subset of a single session. Colors represent trial types. Dots represent first licks for Hit trials. Crosses represent first licks for FA trials. (D) Lick latency distribution for Hit, FA1, and FA2 trials pooled across animals. (E) Lick latency from cue onset across trial types. (F) Same as (E) but for lick latency residuals. (G) Same as (C) but during optogenetics experiments. (H) Lick latency distribution after Go cues in the presence (Go+S) or absence (Go) of photostimulation in a representative session. (I) Hit rate in the absence (−) or presence (+) of LED photostimulation. (J) Same as (I) but for lick latency in Hit trials. (K) Same as (I) but for lick latency residuals in Hit trials. See also Figure S1 and Table S1.

Figure 2

Heterogeneous Sensorimotor Representations in Purkinje Cell Simple Spikes (A) Simultaneously recorded Purkinje cells (PCs, red dots) on a Neuropixels probe. The putative boundary between lobules is indicated by a dashed line. (B) Representative raster plots of simple spikes (black) and licks (light blue) during the task. Note the pause and increased simple spike firing during licking. Shaded areas represent the duration of sensory stimuli (500 ms) for the corresponding trial type. (C) Single-trial simple spike firing rate during Hit trials from a representative Purkinje cell (PC 2). (D) Same as (C) but for PC 5. (E) Single-trial lick rate during Hit trials. (F) Trial-averaged simple spike firing rate (n = 72, 33, 29, 9, 67, and 90 trials for Hit, FA1, FA2, Miss, CR1, and CR2) aligned to the sensory cue onset from PC 2 overlaid by the trial-averaged lick rate for each trial type. (G) Spatially aligned heatmaps for trial-type-averaged simple spike firing rate from clusters in (A). Simple spike firing rate of −100 to 200 ms time window from the onsets of sensory stimuli are Z scored and color coded for each cell per trial type. (H) GLM coefficients for the air puff, tone, and lick initiation fit to simple spike modulation at 0–100 ms from the onset of sensory stimuli in individual Crus I PCs. Gray lines represent individual cells. Red circles represent significant contributions of the predictors to the model, and black circles represent non-significant ones. See also Figures S2 and S3 and Table S1.

Figure 3

Simple Spike Modulation Contributes to Sensory-Driven Lick Initiation and Its Timing (A) Single-trial simple spike (SS) firing rate after Go cues in the absence of optogenetic stimulation from PC 5 (from Figure 2). The red dotted line represents the onset of the sensory cue. (B) Same as (A) but for PC 8 (from Figure 2). (C) Single-trial lick rate after Go cues without optogenetic stimulation. (D) Same as (A) but for Go cues in the presence of optogenetic stimulation (Go+S). (E) Same as (D) but for PC 8. (F) Same as (C) but for Go+S. (G) Trial-averaged SS rate of PC 5 during Hit trials, Hit + photostimulation (Hit+S), and Miss + photostimulation (Miss+S). A dotted line represents the onset of the sensory cue. (H) Same as (G) but for PC 8. (I) Optogenetically induced absolute changes in trial-averaged SS modulation at 0–100 ms from the onset of sensory stimuli during Hit+S and Miss+S trials compared with that during Hit trials. (J) Linear regression of single-trial lick latency to SS modulation of PC 5 during Hit and Hit+S trials. The SS modulation for Miss+S trials is shown on the right. (K) Same as (J) but for PC 8. (L) Volcano plot for the slope of the linear regression and the significance level individually fit for each PC in Crus I. A horizontal dotted line represents the Bonferroni-corrected threshold p value for significance (p = 0.0017). Red dots represent PCs with significant correlation with lick latency. See also Figure S2 and Table S1.

Figure 4

Task-Related Complex Spike Signals Are Organized into Conserved Alternating Parasagittal Stripes (A) Two-photon imaging field of view showing extracted Purkinje cell dendritic regions of interest (ROIs, pseudo-colored). (B) Representative fluorescence traces from selected ROIs (indicated by yellow numbers in A; ROIs are numbered from lateral to medial). Red dots represent extracted complex spike (CS) events. Vertical lines represent the sensory cue onset, and their colors represent corresponding trial types. (C) Single-trial fluorescence traces from a single ROI (ROI 219 in A and B) aligned to the onset of sensory stimuli (dotted lines) for each trial type. Thin lines represent single trials, and thick lines represent trial averages. (D) Trial-type-averaged CS event rate heatmap of ROIs in (A). White dotted lines represent the onset of sensory stimuli. (E) ROIs colored based on CS event probability for 0–250 ms from the onset of sensory stimuli during Hit trials. (F) Correlation matrix of fluorescence traces for a whole imaging session, clustered and sorted based on correlation similarity. White dotted lines represent boundaries of zones. Zonal identity is represented by the thick colored lines at the bottom and right (zone 1, red; zone 2, green; zone 3, yellow; zone 4, blue). (G) Co-activation traces from zones 1–4. Dots represent extracted co-activated complex spike (CoCS) events. (H) Spatial arrangement of three imaging fields across the expanse of Crus I. Six zones are identified and numbered from lateral to medial (zones 1–6). (I) CoCS event probability of zones 1–6 across trial types from a representative animal. (J) Assignments of functionally defined zones in Crus I across fields of view and across mice. (K) Widths of zones 1–6 from all mice. (L) Same as (K) but for CoCS event probability during Hit trials. See also Figures S4 and S5.

Figure 5

Enhancement of Coherence in Sensory-Evoked Complex Spike Signals Contributes to Motor Initiation (A) Trial-type-averaged CS event rate heatmap of ROIs for non-licking air puff (CR1) and tone (CR2) trials, from an example session. Vertical dotted lines represent the sensory cue onset, and horizontal lines represent zonal boundaries. (B) Single-trial co-activation traces of zone 3 for CR1 and CR2 trials in (A). Thin colored lines represent single trials, and thick lines represent trial averages. Vertical dotted lines represent the cue onset. (C) Comparison of coherence level within co-activation (CoCS) events with a 0–250 ms latency for CR1 and CR2 trials in (B). (D) Probability of CoCS events in zones 1–6 during CR1 and CR2 trials in single sessions pooled across mice. (E) Same as (A) but comparing licking trials and non-licking trials after sensory stimuli containing air puff stimulation. (F) Same as (B) but for licking trials and non-licking trials after sensory stimuli containing air puff stimulation. (G) Comparison of coherence level within CoCS events with a 0–250 ms latency for Hit+FA1 and Miss+CR1 trials in (F). (H) Differences in coherence levels of CoCS events in zones 1–6 across licking trials and non-licking trials after air puff stimulation (Hit+FA1 − Miss+CR1) in single sessions pooled across mice. (I) Example table of values used for the GLMM analysis. Rows represent individual trials with CoCS events. (J) GLMM coefficient for lick initiation fit to the trial-by-trial coherence level of CoCS events for each zone. (K) Same as (J) but for air puff stimuli. (L) Same as (J) but for tone stimuli. See also Figures S4, S6, and S7 and Table S1.

Figure 6

Coherent Complex Spike Signals in Alternating Zones Result in Temporally Precise Lick Initiation (A) Trial-averaged CS event rate heatmap for trials with or without co-activation (CoCS) events in zone 3 from an example session. Vertical dotted lines represent the sensory cue onset, and horizontal lines represent zonal boundaries. (B) Single-trial co-activation traces of zone 3 for trials in (A). Thin lines represent single trials, and thick lines represent trial averages. Vertical dotted lines represent the cue onset. (C) Probability of lick initiation after CoCS events (+) or no CoCS events (−) in zones 1–6 for all trials in single sessions pooled across mice. (D) Distribution of the latency of the first lick in the licking trials (Hit, FA1, and FA2 trials combined) with or without CoCS events in zone 5 pooled across animals. (E) Same as (C) but for lick latency residuals in Hit, FA1, and FA2 trials. (F) Same as (C) but for lick latency in Hit, FA1, and FA2 trials. See also Figure S7 and Table S1.

Figure 7

Coherent Complex Spike Signals Are Acquired, Together with Precisely Timed Motor Initiation (A) Lick latency distribution for Hit trials at early (apprentice) and late (expert) stages of task learning pooled across 11 mice. (B) Lick latency during Hit trials at early and late stages of task learning. (C) Same as (B) but for lick latency residuals. (D) ROIs colored based on CS event probability for 0–250 ms from the onset of sensory stimuli during Hit trials across early and late learning stages. (E) Comparisons of coherence level of CoCS events for zones 1–6 during Hit trials across learning. A, apprentice; E, expert. (F) GLMM coefficients for the presence of lick initiation fit on the trial-by-trial coherence level of CoCS events for zones 1–6 across learning. (G) Same as (F) but for air puff stimuli. (H) Same as (F) but for tone stimuli. See also Figure S7 and Table S1.