Behavior-related pauses in simple-spike activity of mouse Purkinje cells are linked to spike rate modulation

Abstract

Purkinje cells (PCs) in the mammalian cerebellum express high-frequency spontaneous activity with average spike rates between 30 and 200 Hz. Cerebellar nuclear (CN) neurons receive converging input from many PCs, resulting in a continuous barrage of inhibitory inputs. It has been hypothesized that pauses in PC activity trigger increases in CN spiking activity. A prediction derived from this hypothesis is that pauses in PC simple-spike activity represent relevant behavioral or sensory events. Here, we asked whether pauses in the simple-spike activity of PCs related to either fluid licking or respiration, play a special role in representing information about behavior. Both behaviors are widely represented in cerebellar PC simple-spike activity. We recorded PC activity in the vermis and lobus simplex of head-fixed mice while monitoring licking and respiratory behavior. Using cross-correlation and Granger causality analysis, we examined whether short interspike intervals (ISIs) had a different temporal relationship to behavior than long ISIs or pauses. Behavior-related simple-spike pauses occurred during low-rate simple-spike activity in both licking- and breathing-related PCs. Granger causality analysis revealed causal relationships between simple-spike pauses and behavior. However, the same results were obtained from an analysis of surrogate spike trains with gamma ISI distributions constructed to match rate modulations of behavior-related Purkinje cells. Our results therefore suggest that the occurrence of pauses in simple-spike activity does not represent additional information about behavioral or sensory events that goes beyond the simple-spike rate modulations.

Figure 1.

Raw data example of simple-spike activity correlated with licking and respiratory behavior. A, Raw data examples of simultaneous recordings of single-unit Purkinje cell spike activity with licking and respiratory behavior. Under each trace are time markers marking the tongue-to-spout contact times for the licking trace (red), the end-of-expiration and inspiration times for the respiratory trace (blue), and simple-spike and complex-spike activity for the PC spike train recording, respectively. The black triangles mark complex spikes. Expanded portion of raw spike train signal showing simple and complex spikes and the extracted spike shapes used for shape based spike sorting underneath each spike. The histogram on the right shows the interlick interval (red) and interexpiration interval (blue) distribution. B, Combined cross-correlation and autocorrelation histograms showing the cross-correlation between licking and simple-spike activity (black line) and the autocorrelation of licking events (red line). C, Combined cross-correlation and autocorrelation histograms showing the cross-correlation between expiration and simple-spike activity (black line) and the autocorrelation of expiration events (blue line).

Figure 2.

Decomposition of the simple-spike portion of the spike train into interval groups of increasing duration. A, ISI histograms of a single-unit Purkinje cell simple spike train. The ISI distribution for a matched gamma-distributed spike train is shown in the inset. B, Illustration of the decomposition of a simple spike train into groups of interspike intervals of different durations. (Colors correspond to those in the interval histogram in A). Cross-correlation analyses between spike intervals and behavior were performed using the first spikes (i.e., the interval onsets) as the time of interval occurrence.

Figure 3.

Cross-correlation analyses of biological and surrogate Purkinje cell simple spike trains with licking and respiratory behavior. A, Left, Lick–simple spike cross-correlations for all simple spikes (top histograms) and for interspike intervals of increasing duration as denoted on the left of each row. In each histogram, the thin black lines represent the raw cross-correlation histogram. The two gray lines and the center black line represent the ±3 SDs and mean of the shuffled correlations. Right, Same analysis as on the left, but applied to surrogate spike trains and licking intervals. B, Same analysis as in A but for expiration–simple spike cross-correlations in the left column and the matching surrogate spike data on the right. The y-axis scales show z-scores.

Figure 4.

Quantitative analysis of peak spike–lick correlations as a function of interspike interval duration. A, Correlation strength between interspike intervals and fluid licking was quantified using the z-scores of the peak cross-correlation values from 12 Purkinje cell single-unit simple spike trains and matching surrogate spike trains. B, Z-scores of peak cross-correlation values for interspike intervals and expiration from 10 Purkinje cell single-unit simple spike trains and matching surrogate spike trains. ^#Significant differences between all-spikes correlation z-score with correlation z-scores of all other ISI groups. *Significant differences between correlation z-scores for 0–10 ms ISIs with z-scores of all other ISI groups.

Figure 5.

Granger causality analysis of biological and surrogate spike and lick data as a function of interspike interval duration. A, Left column, Perievent raster plots for spikes preceding intervals of durations as indicated at the center above each plot. The black dots represent spikes. Time “0” on the x-axis corresponds to the time at which the mouse's tongue touched the waterspout. Middle column, Firing rate estimation from the perievent time histogram (PETH). The blue traces correspond to firing rate estimation per 1 ms bins. The red traces are an eighth-order polynomial fit for spike rate modulation. Right column, Granger causality analysis for lick to spikes causality. The peak at 6 Hz corresponds to the average lick frequency and represents a causal relationship between licking behavior and simple-spike activity. Note there is a high causal strength for short and long interspike interval durations compared with intermediate ones. B, The same analyses as in A but performed on surrogate spike trains. Left column, Raster plots of surrogate spike activity. Middle column, Firing rate estimation from PETH. Right column, Granger causality analysis for surrogate spike trains. C, D, Same analyses performed for a representative PC recording in relation to breathing.

Figure 6.

Distribution of Granger causality peak values at different ISIs. Five recordings for each behavioral condition with a mean rate close to the overall mean rate were chosen for analysis. This selection ensured that all cases had a representative sample of the different ISI groups chosen. The causality peak distribution plots for recordings and the matched surrogate data are shown side by side. Each black asterisk corresponds to peak Granger causality value of ∼6–7 Hz for datasets associated with licking behavior and ∼3–4 Hz for datasets associated with breathing. The blue bar denotes the mean, and the vertical red line, the SD. The asterisks represent individual data points. The figure clearly shows that the causality prediction is high for short but not for intermediate ISIs, and again increases for long ISIs > 30 ms.