Population calcium responses of Purkinje cells in the oculomotor cerebellum driven by nonvisual input

Abstract

The climbing fiber input to the cerebellum conveys instructive signals that can induce synaptic plasticity and learning by triggering complex spikes accompanied by large calcium transients in Purkinje cells. In the cerebellar flocculus, which supports oculomotor learning, complex spikes are driven by image motion on the retina, which could indicate an oculomotor error. In the same neurons, complex spikes also can be driven by nonvisual signals. It has been shown that the calcium transients accompanying each complex spike can vary in amplitude, even within a given cell, therefore, we compared the calcium responses associated with the visual and nonvisual inputs to floccular Purkinje cells. The calcium indicator GCaMP6f was selectively expressed in Purkinje cells, and fiber photometry was used to record the calcium responses from a population of Purkinje cells in the flocculus of awake behaving mice. During visual (optokinetic) stimuli and pairing of vestibular and visual stimuli, the calcium level increased during contraversive retinal image motion. During performance of the vestibulo-ocular reflex in the dark, calcium increased during contraversive head rotation and the associated ipsiverse eye movements. The amplitude of this nonvisual calcium response was comparable to that during conditions with retinal image motion present that induce oculomotor learning. Thus, population calcium responses of Purkinje cells in the cerebellar flocculus to visual and nonvisual input are similar to what has been reported previously for complex spikes, suggesting that multimodal instructive signals control the synaptic plasticity supporting oculomotor learning.

NEW & NOTEWORTHY It was long known that the climbing fiber input to Purkinje cells in the cerebellar flocculus conveys visual feedback about the accuracy of image-stabilizing oculomotor reflexes. More recently, the same climbing fibers were reported to carry nonvisual signals. Here, we report that both visual and nonvisual inputs can elicit robust calcium responses in the Purkinje cells, suggesting that the instructive signals guiding oculomotor plasticity are multimodal.

INTRODUCTION

The cerebellum uses feedback about behavioral errors to improve subsequent actions. A key source of this feedback is the climbing fiber input to the cerebellum from the inferior olive, which can trigger synaptic plasticity and, hence, learned changes in behavior (1–4). Climbing fibers make unusually powerful synaptic connections with Purkinje cells, so that each discharge of an individual climbing fiber produces a “complex spike” and large calcium transient in its Purkinje cell targets. This, in turn, triggers long-term depression in the recently active parallel fiber-Purkinje cell synapses (4–6). Decades of research have established a contribution of such climbing fiber-triggered plasticity to cerebellum-dependent learning (7–11). Yet, the instructive signals conveyed by the climbing fibers to control learning are not well understood.

The groundbreaking research of Jerry Simpson has made the climbing fiber input to the cerebellar flocculus one of the best understood examples of how behavioral errors are encoded in the cerebellum, and, more generally, in the brain. The flocculus supports oculomotor learning, which calibrates the vestibulo-ocular reflex (VOR) and optokinetic reflex (OKR). The VOR and OKR are reflexive eye movement responses to a vestibular stimulus or visual stimulus motion, respectively, which support vision by stabilizing images on the retina. The amplitude of the VOR and OKR can be adaptively modified to more effectively perform their image-stabilizing function (3, 12, 13).

Seminal work by Simpson and coworkers (14–19) established that the climbing fibers innervating the flocculus carry visual signals that convey information about image motion on the retina, or retinal slip, which could signal a failure of the image-stabilizing reflexes. However, since image stabilization is accomplished through a coordinated set of behaviors (20–22), retinal slip does not necessarily mean that the oculomotor reflexes supported by the flocculus need to be adjusted, since the error also could have been generated by other components of the orienting response, such as the vestibulo-collic (neck) reflex or other postural responses. In addition to this error attribution or “credit assignment” challenge, there is the possibility that image motion on the retina is not the result of a motor error, but rather the unpredictable motion of an external visual stimulus. To distinguish these possibilities, and determine whether retinal slip indicates a need to adjust the oculomotor reflexes would thus seem to require additional information. Intriguingly, Simpson and others (23–32) have found that climbing fibers in the flocculus carry not only information about retinal slip, but also nonvisual signals, which are evident in the complex spike responses of floccular Purkinje cells during vestibular stimulation in the dark. In particular, complex spike rate increases during contraversive head movements in the dark and the associated ipsiversive VOR eye movement responses (30). These results suggest that vestibular signals and/or oculomotor efference copy signals are combined with or can modify the error signals provided by the visual feedback carried by climbing fibers and may even be able to contribute to plasticity in the absence of visual feedback about oculomotor performance.

To evaluate the potential for the nonvisual signals carried by floccular climbing fibers to contribute to the induction of plasticity, the associated calcium responses in the Purkinje cells must be considered. Although the response of a Purkinje cell to its climbing fiber input was previously viewed as binary, with each climbing fiber discharge eliciting a nearly unvarying complex spike (33), more recent work has demonstrated that this is not the case. The number of spikes in the climbing fiber axon burst (34, 35), the amplitude and number of components of the climbing fiber-driven compound excitatory postsynaptic potentials (EPSPs) in a Purkinje cell (36, 37), the number of spikelets in the complex spike of the Purkinje cell (35, 38, 39), and the calcium transient in the Purkinje cell (40–46) are all graded. These graded responses can be modified by several factors, including the intrinsic excitability of and synaptic input to the inferior olive (35, 47–49), the intrinsic excitability of, and synaptic input to Purkinje cells (31, 50–56), sensory input to the circuit (37, 44, 46, 57), and the history of experience and learning (37, 58). These modulatory factors can cause the amplitude of the calcium response in the Purkinje cells to vary in a manner that is not reflected in the rate of complex spikes, which is typically used to assess the instructive signals conveyed by the climbing fibers. Yet, the level of calcium is more closely related to the induction of plasticity (52, 59). Therefore, we assessed whether the nonvisual signals carried by climbing fibers in the flocculus contribute to, not just complex spiking, but also substantial calcium responses in their Purkinje cell targets, comparable to the calcium responses when visual feedback about errors is present.

METHODS

Animals

GCaMP6f was selectively expressed in cerebellar Purkinje cells by crossing B6.Cg-Tg(Pcp2-cre)3555Jdhu/J mice (The Jackson Laboratory RRID:IMSR_JAX:010536), which selectively express Cre in Purkinje cells (60, 61), with Ai95D(RCL-GCaMP6f)-D mice (Jackson Laboratories), which express GCaMP6f in a Cre-dependent manner (see Fig. 1). Mice aged 2–4 mo (n = 7; 4 males and 3 females) were housed individually in standard cages and maintained on a 12-h light-dark cycle with lights on at 9:00 PM. All testing was done between noon and 6:00 PM. Food and water were available ad libitum. All experiments were approved by the Stanford University Administrative Panel for Laboratory Animal Care and carried out in accordance with institutional guidelines.

Figure 1.

A: GCaMP6f labeling of Purkinje cells in the cerebellar flocculus. Maximum intensity projection of a z-stack of images shows uniform labeling of Purkinje cells in one of the male mice used for fiber photometry. The track of the optical fiber used for recording can be seen in the top left. B: representative raw traces, from the subject in A, of calcium-sensitive fluorescence (F, green), calcium-insensitive fluorescence [Isosbestic (G), yellow], eye velocity (red), visual stimulus velocity (gray), and vestibular stimulus velocity (blue) during visual and vestibular stimuli presented alone or in combination. Horizontal dashed lines represent zero velocity. In this and all subsequent figures, positive values for eye and stimulus velocity indicate contraversive movement and negative values indicate ipsiversive movement. The frequency of spontaneous calcium events (orange) mean ± SE is plotted in the bottom left corner. AUs, arbitrary units.

Surgery

Mice were anesthetized with isoflurane and surgically prepared as previously described (62). Briefly, two cylindrical neodymium magnets (each 0.75 × 1 mm and axially magnetized; SuperMagnetMan.com) were stacked and implanted under the conjunctiva on the dorsal side of the left eye with the north-south axis approximately aligned with the horizontal plane. Dental acrylic was used to secure an angular magnetic field sensor (4.8 × 5.8 mm; HMC1512, Honeywell Inc.) to the skull above the magnet, to measure horizontal eye movements. Before implantation, an 8-pin right angle connector was soldered to the sensor, for connection to a preamplifier. A custom head post (Shapeways) was attached to the skull near lambda with dental acrylic. A small craniotomy (∼600 µm) was made in the center of the periotic bone capsule overlying the paraflocculus and flocculus, and a guide cannula (22 gauge; P1 Technologies) was secured to the craniotomy with dental acrylic in the horizontal plane at an angle of 30° from the sagittal plane, with the end of the guide cannula approximately flush with the inner surface of the skull. Mice were allowed to recover for at least 7 days before testing.

Vestibular and Visual Stimuli

During experiments, mice were head-restrained by attaching their surgically implanted head post to a restrainer, which was attached to a vestibular turntable (Carco Electronics, Pittsburgh, PA). Passive sinusoidal vestibular stimuli were delivered by using the turntable to rotate the animal about an earth-vertical axis that passed through the head (1 Hz, ±10°/s peak velocity). Large-field sinusoidal visual stimulus motion was delivered by rotating an optokinetic drum about an earth-vertical axis through the head of the mouse (1 Hz, ±10°/s peak velocity). The drum had alternating black and white vertical stripes, each of which subtended 7.5° and was lit by a string of fiber-optic lights secured around the top. Stimuli were controlled by continuous voltage output signals from a Power1401 board interfaced with Spike2 software (Cambridge Electronic Design, Cambridge, UK), and data were digitized and stored by the Power1401 with a sampling frequency of 1,000 Hz. Eye movements and calcium responses were measured during presentation of the visual stimulus alone with the head stationary (visual), the vestibular stimulus alone in the dark (vestibular-dark), or combined visual-vestibular stimulation with the visual stimulus moving exactly with the head (visual-vestibular in-phase) or exactly opposite to the head (visual-vestibular out-of-phase).

An experimental session comprised three sets of 12 blocks of test stimuli. The first three and last three blocks had alternating tests of 10 s with the head stationary in the dark, i.e., absent any vestibular or visual stimulation, to measure spontaneous calcium signals, and 10 s of vestibular stimulation in the dark (used only in the analysis of stability across the recording session, Supplemental Fig. S1; see https://doi.org/10.6084/m9.figshare.14364698). Each of the other six blocks consisted of 10 s with the head stationary in the dark (spontaneous), followed by 10 s of vestibular stimulation in the dark, and then 20 s of the visual stimulus alone, in-phase visual-vestibular pairing, or out-of-phase visual-vestibular pairing. These six blocks repeated the same visual or visual-vestibular stimulus and the order of sets was randomized and counterbalanced across mice. Population calcium responses were monitored continuously throughout the entirety of each experiment.

Data Acquisition

Eye tracking.

The head-fixed magnetic sensor converted the direction of the magnetic field created by the eye-fixed magnets into an analog voltage signal related to eye position in the head. A custom preamplifier (David Profitt Engineering Services, Los Altos, CA) amplified the voltage signal 60 times. Signals were digitized at 1,000 Hz and stored with the Power1401 interfaced with Spike2 software. Data were digitally filtered with a 100-Hz low-pass Butterworth filter.

Signals from the magnetic sensor were calibrated by simultaneously recording eye movements with the magnetic sensor and dual-angle video-oculography at the end of each experiment, as described by Payne and Raymond (62). Briefly, small infrared LEDs (875 nm, 20 mW, TSHA4400, Digi-Key) were positioned above each of two cameras to create two corneal reflections. The cameras were separated by an angle of 40°, equidistant from the eye and fixed to the vestibular turntable. A third infrared LED was used to illuminate the cornea to create greater contrast between the pupil and iris. Images were acquired at 30 frames per second while vestibular stimulation (1 Hz, ±10°/s peak velocity) was delivered for 3 min to elicit eye movements. The cameras and LEDs for video-oculography were moved into position just before the calibration session, after other data collection was completed.

Fiber photometry recording.

Fiber photometry was used to record calcium responses from a population of Purkinje cells in the cerebellar flocculus. To record fluorescence signals, light from an LED with a 478 nm peak excitation wavelength and from a second LED with a 415 nm peak excitation wavelength were interleaved, bandpass filtered (457–50 nm and 420–10 nm, respectively), reflected by a multipass dichroic mirror (FP3001; Neurophotometrics Ltd., San Diego, CA), and delivered to the flocculus via a 200-µm diameter optical fiber (Doric Lenses) inserted through the implanted cannula. To limit photo-bleaching and avoid saturation of the fluorescent signals, illumination was set to a low level, where the contrast between the illumination within the field of view of the fiber optic and surrounding nonilluminated area outside of the fiber optic was barely detectable to the naked eye. The fiber was initially advanced ∼500–600 µm into the brain and then gradually advanced (∼2–3 µm every few seconds), while monitoring for spontaneous calcium transients occurring at a rate of ∼1 Hz. When such transients were observed, an optokinetic stimulus known to drive the modulation of complex spiking in the flocculus was delivered (1 Hz sinusoidal movement of an optokinetic drum about an earth-vertical axis with peak velocity ±10°/s). If the optokinetic stimulus drove modulation of the recorded calcium signal, then the entire experiment was performed with the fiber optic in that single position, else the fiber was advanced deeper. Recordings were made with the tip of the fiber optic at depths of 782–1,221 µm from the surface of the brain.

Excitation intensity and emission collection efficiency varies with distance from the tip of the fiber, with an estimated 80% of the signal arising from a volume extending 200 µm into the tissue (63). Calcium-dependent and calcium-independent (isosbestic) emission signals were focused by an objective lens (numerical aperture of 0.37), bandpass filtered (514–30 nm), and collected by a CMOS camera (FLIR Blackfly; dynamic range of sensor is 74.35 dB) at a sampling frequency of 40 Hz, with binning of eight detectors for a single pixel. Data were acquired using the open-source Bonsai software, with data acquisition synchronized to behavioral measurements via output from the Power1401. Regions of interest (ROIs) were drawn to capture the full diameter of the fiber optic. All pixels within the ROI were averaged to calculate a single intensity value for each sample time. The calcium responses to repeated presentations of the same stimulus were stable in amplitude throughout the recording session (see Supplemental Fig. S1 and Statistical Analysis).

Immunohistochemistry.

Mice were perfused with 20 µL of 0.1 M PBS, followed by 20 µL of 4% PFA. The brains were extracted, stored in a 4% PFA solution on a rotary shaker at 70 RPM overnight, and then transferred to 30% PB sucrose solution for at least 2 days. Coronal sections (100 µm) including the flocculus were sliced with a vibratome (Leica VT 1200, Leica Biosystems, Buffalo Grove, IL), transferred to a well-plate, and rinsed in 1× PBS and 0.4% Triton X on a rotary shaker at 70 RPM for 3 × 10 min. Slices were then incubated in a blocking solution containing 5% BSA, 0.05% sodium azide, and 0.4% Triton X in 1× PBS for 45 min. Slices were then transferred to blocking solution containing rabbit anti-GFP primary antibody (Molecular Probes, Cat. No. A-11122, RRID: AB_221569) diluted 1:500 and rotated for 16 h. Slices were incubated in Alexa Fluor 594 donkey anti-rabbit IgG secondary antibody (Jackson ImmunoResearch Labs, Cat. No. 711-585-152, RRID: AB_2340621) diluted in blocking solution (1:1,000), followed by a rinse of the tissue in 1× PBS and 0.4% Triton X. Slices were mounted to a glass slide, and then Vectashield antifade mounting medium with DAPI (Vector Laboratories, Cat. No. H-2000, Burlingame, CA) and a coverslip were applied.

Confocal imaging.

Brain sections were imaged with a Leica SP8 confocal microscope. Z-stacks of 30–50 optical sections (2 µm spacing, 60–100 µm) were acquired with a Leica HC APO ×10 magnification air objective lens (0.3 numerical aperture). One-photon excitation (594 nm) was performed with 8 kHz resonant scanning and sixfold line averaging, with each optical section spanning 930 µm × 930 µm. Stacks of images saved as TIFF were imported to ImageJ (Fiji.sc) and the color was converted to green, but with preservation of the relative pixel intensities.

Data Analysis

Eye movements.

Signals from the magnetic sensor, related to eye position, were differentiated and then smoothed using a 50-ms sliding window to calculate eye velocity. Values whose squared deviation from the mean exceeded a manually set threshold were identified as saccades or motion artifacts and were excluded from the analysis, along with 50 ms before and after each data point that exceeded threshold; corresponding segments of calcium imaging data were also excluded.

Eye position, measured with video-oculography during the calibration session, was upsampled to 1 kHz, differentiated, desaccaded, and linearly regressed against the eye velocity signals simultaneously recorded using the magnetic sensor to obtain a calibration factor for converting the signals from the magnetic sensor to eye velocity in degrees/s. Calibrated eye velocity data from the magnetic sensor were aligned on the vestibular or visual stimulus and averaged across stimulus cycles. To calculate retinal slip, the sum of head velocity and eye velocity was subtracted from the velocity of the visual stimulus. The amplitude and timing of peak retinal slip velocity, eye velocity, and head velocity were determined from sinusoidal fits to the cycle-averaged data for each subject.

Fiber photometry.

Baseline fluorescence was calculated as the 30th percentile value of a 1-s (40-frame) sliding window. Baseline values were subtracted from raw fluorescence values (F) to calculate the change in fluorescence (ΔF) at each sample time, then divided by the data obtained from the isosbestic channel (G) to obtain a ΔF/G value. These data were then upsampled to 1 kHz to match eye velocity, visual and vestibular stimulus velocity, and retinal slip data, and averaged across stimulus cycles. The amplitude of the calcium response and timing of the peak calcium response were obtained from a sinusoidal fit to the cycle-averaged data for each subject. To assess stability of the measurements across a session, we compared the calcium responses to the repeated presentations of the vestibular stimulus in the dark, pooled across three consecutive blocks through the entire experiment (Supplemental Fig. S1).

Spontaneous calcium events recorded in the absence of visual or vestibular stimuli were identified by first using a median absolute deviation amplitude threshold taken from the full ΔF/G trace of a recording to identify candidate events. To eliminate artifacts, candidate events with a 10%–90% rise time or decay time constant more than two standard deviations from the mean were then excluded from the analysis of calcium event frequency (TaroTools software, Igor Pro).

Statistical Analysis

Shapiro–Wilk tests indicated a deviation from normality for the visual-vestibular in-phase peak-to-trough calcium response amplitudes (Fig. 3, P = 0.031) and for the timing of the peak calcium responses relative to peak head velocity for vestibular-dark (Fig. 4; P = 0.023); statistical comparisons that included these samples were conducted using Kruskal–Wallis with Dunn post hoc tests. Other data, for which Shapiro–Wilk indicated normality, were analyzed using repeated-measures ANOVA with Tukey post hoc tests, conducted using SPSS.

Figure 3.

The amplitude of the Purkinje cell population calcium response to visual and vestibular stimuli. Calcium responses during vestibular stimulation in the dark are comparable to conditions with retinal slip present (Kruskal–Wallis, P = 0.44). Average peak-to-trough fluorescence amplitude in individual subjects (green open circles; n = 7, 4 males and 3 females) is plotted for each visual and vestibular test stimulus. Boxplots show median, 25th and 75th percentiles. Whiskers extend to the most extreme data points. Notches indicate 95% confidence intervals. After removing the single outlier subject with the highest calcium responses across test stimuli, a repeated measures ANOVA did not indicate significance between conditions (P = 0.093).

Figure 4.

Timing of Purkinje cell population calcium responses relative to sensory and motor signals present during testing. Average time of peak calcium signal, ΔF/G, for each subject (colored circles; n = 7, 4 males and 3 females), plotted relative to peak contraversive retinal slip velocity (A), head velocity (B), or eye velocity (C) (vertical dotted lines at 0 ms), with peak ipsiversive velocity occurring 500 ms earlier and later (vertical dashed lines) during the 1-s sinusoidal stimulus cycle. A: there was no significant difference across test stimuli in the timing of peak calcium responses relative to retinal slip (repeated measures ANOVA, P = 0.39). B: there was a difference in the timing of peak calcium relative to peak head velocity, with visual-vestibular out-of-phase significantly different from visual-vestibular in-phase and vestibular-dark test conditions (Kruskal–Wallis, P = 0.001; Dunn post hoc, *P < 0.001 for each). C: there was a difference in the timing of peak calcium signal relative to eye velocity across test stimuli, with visual-vestibular in-phase and vestibular-dark each significantly different from visual-vestibular out-of-phase and visual (repeated measures ANOVA, P < 0.001; Tukey post hoc, *P < 0.001 for each). N/A indicates testing conditions where there was no visual stimulus and hence no retinal slip (vestibular-dark in A) or no vestibular stimulus (visual in B). Boxplots show median, 25th and 75th percentiles. Whiskers extend to the most extreme data points. Notches indicate 95% confidence intervals.

RESULTS

Fiber photometry was used to record population calcium responses from Purkinje cells in the cerebellar flocculus of awake, head-fixed mice (n = 7) during vestibular and visual stimuli, presented alone or in combination (Fig. 1; see methods). Previous work has documented strikingly uniform responses of floccular climbing fibers to such stimuli, as measured by the complex spike responses in Purkinje cells (15, 30, 64–66, 75). The large majority of cells increase their complex spiking during contraversive image motion and during contraversive vestibular stimulation presented alone in the dark. This uniform direction preference across cells suggested that useful insights could be obtained using a population-level measure of calcium responses provided by fiber photometry, which combines signals from multiple Purkinje cells within a few hundred microns of the tip of the optical fiber (63).

In the absence of visual or vestibular stimulation, the recorded Purkinje cell populations exhibited discrete, spontaneous calcium transients, as reported previously (67). These responses were observed in the calcium-sensitive fluorescence signal recorded from stationary, head-restrained mice in total darkness (Fig. 1, Spontaneous). The spontaneous transients were not observed in the isosbestic, calcium-insensitive fluorescence signal (G), suggesting they were not due to motion or other artifacts. The frequency of spontaneous population calcium events was 1.00 ± 0.027 Hz mean ± SE (Fig. 1, bottom left), which is similar to the 0.5–2 Hz spontaneous frequency of climbing fiber-driven complex spikes recorded electrophysiologically from individual Purkinje cells in mice and other species (34, 68, 69).

The presentation of vestibular and/or visual stimuli drove modulation of the calcium signal in floccular Purkinje cells. Calcium responses were measured during presentation of a visual stimulus alone with the head stationary (sinusoidal image motion, 1 Hz, ±10°/s peak velocity), a vestibular stimulus alone in the dark (1 Hz, ±10°/s peak velocity), or combined visual-vestibular stimulation with the visual stimulus moving exactly with the head (in-phase) or exactly opposite to the head (out-of-phase). For each stimulus condition (visual only, visual-vestibular out-of-phase, visual-vestibular in-phase, and vestibular-dark), modulation of the raw fluorescence signal could be observed with each stimulus cycle (Fig. 1).

The cycle-averaged calcium responses (ΔF/G, see methods) during each visual-vestibular stimulus condition were similar across animals (Fig. 2, top, compare thin traces from individual animals and thick traces showing the average across animals). During presentation of the visual stimulus alone, with the head stationary, the calcium signal rose during contraversive motion of the stimulus, peaking just before the stimulus changed direction, and decreased during ipsiversive visual stimulus motion, (Fig. 2, second column from left, green). The visual stimulus (Fig. 2, gray) elicited tracking eye movements known as the optokinetic reflex (Fig. 2, red), which reduced but did not eliminate image motion on the retina (Fig. 2, purple). The timing of the calcium responses relative to retinal slip during combined visual-vestibular stimulation was similar to that during the visual stimulus alone, with the calcium signal peaking just after peak contraversive retinal slip (Figs. 2 and 4, visual-vestibular in-phase and out-of-phase). This is consistent with electrophysiological studies in multiple species, which have reported the activation of climbing fibers in the flocculus by contraversive retinal slip during the optokinetic response or paired low-frequency visual-vestibular stimulation (23, 28–30, 65, 69–75).

Figure 2.

Purkinje cell population calcium responses are reliably elicited by vestibular stimulation in the dark. Calcium responses (ΔF/G, green), retinal slip velocity (purple), eye velocity (red), visual stimulus velocity (gray), and vestibular stimulus velocity (blue), averaged across stimulus cycles, are plotted for each subject (colored traces; n = 7 for each condition, 4 males and 3 females) and the average across subjects (black traces) for each testing condition. Vertical dotted lines indicate peak retinal slip velocity and horizontal dashed lines indicate zero ΔF/G or zero velocity.

Somewhat surprisingly, the amplitude of the calcium responses across stimulus conditions varied inversely with peak retinal slip speed (Fig. 2, green and purple traces). In-phase visual-vestibular stimulation drove the smallest retinal slip speeds (2.94 ± 0.21°/s) but largest calcium responses (2.61 ± 0.34% ΔF/G), whereas visual stimulation alone drove intermediate-sized retinal slip and calcium responses (6.86 ± 0.48°/s and 1.23 ± 0.49% ΔF/G), and out-of-phase visual-vestibular stimulation drove the highest retinal slip speeds but the smallest calcium responses (11.91 ± 0.70°/s and 0.98 ± 0.28% ΔF/G). This inverse relationship could potentially result from tuning of the climbing fibers to low image speeds (15, 16, 76–78), and/or from an influence of nonvisual signals such as vestibular input or eye movement-related signals, which varied across the three stimulus conditions (Fig. 2, blue and red traces).

To isolate the influence of nonvisual signals on the calcium responses in Purkinje cells, we measured population calcium responses during vestibular stimulation in complete darkness. There was an increase in calcium during contraversive vestibular stimulation in the dark (Fig. 2, vestibular-dark). Calcium responses peaked around the time of the change in direction from contraversive to ipsiversive head rotation, and this timing was consistent across mice. The increasing calcium signal also corresponded to a phase of the stimulus when there was ipsiversive eye movement, driven by the vestibulo-ocular reflex response to the contraversive head movement. The amplitude of the calcium response during vestibular stimulation in the dark was similar to that during the visual stimulus alone, in-phase visual-vestibular and out-of-phase visual-vestibular pairing (Fig. 3; Kruskal–Wallis, P = 0.44). In one mouse, the calcium signals were considerably larger than the others; if we removed this single outlier, a repeated measures ANOVA still found no significant difference in response amplitudes across test stimuli, but there was a trend (P = 0.093). However, this trend was not driven by a difference in amplitude for vestibular-dark compared with the test stimuli with retinal slip (Tukey post hoc: P = 0.99 vs. out-of-phase visual-vestibular, P = 0.14 vs. in-phase visual-vestibular, and P = 0.98 vs. visual), but rather by the tendency for bigger calcium responses during the in-phase visual-vestibular test stimulus compared with the other test stimuli (Tukey post hoc: P = 0.14 vs. out-of-phase visual-vestibular and P = 0.19 vs. visual).

To evaluate the contribution of visual, vestibular, and efference copy signals to the calcium responses of the Purkinje cells, we quantified the time of peak calcium relative to the time of peak retinal slip (Fig. 4A), the time of peak head velocity (Fig. 4B), and the time of peak eye velocity (Fig. 4C) during the four test stimuli. The timing of the calcium response relative to peak retinal slip was similar during the three test conditions with a visual stimulus (Fig. 4A; repeated measures ANOVA, P = 0.38). Peak calcium, averaged across the subject population, lagged peak retinal slip by 201 ± 36.53 ms, 127 ± 40.76 ms, and 176 ± 40.84 ms during the visual stimulus alone, out-of-phase visual-vestibular stimulation, and in-phase visual-vestibular stimulation, respectively. This close correspondence between the timing of the calcium signals and retinal slip is consistent with the idea that this visual signal, when present, is the dominant contributor to the calcium responses.

Unlike the timing of the population calcium responses relative to peak retinal slip, which was consistent across stimulus conditions, timing of population calcium responses relative to eye velocity or head velocity varied considerably across stimulus conditions (repeated measures ANOVA, P < 0.001 for eye velocity and Kruskal–Wallis, P = 0.001 for head velocity). During vestibular stimulation in the dark and in-phase visual-vestibular stimulation, calcium increased during contraversive head movements and ipsiversive eye movements (Fig. 4, B and C, bottom and second from bottom). During both stimulus conditions, peak calcium lagged peak ipsiversive eye velocity by roughly 90°, which is the time the eye switches to moving in the contraverisve direction. During in-phase visual-vestibular stimulation, this also coincides with contraversive retinal slip, potentially enabling these distinct signals to have additive or synergistic effects on the calcium responses in the Purkinje cells, consistent with the tendency for the calcium response amplitude to be larger during in-phase visual-vestibular pairing compared with the other test stimuli (Fig. 3). In contrast, during out-of-phase visual-vestibular stimulation and visual stimulation alone, calcium increased during ipsiversive head and contraversive eye movements (Fig. 4, B and C, second from top and top), a timing opposite what is observed when the vestibular stimulus is presented alone in darkness (Fig. 4, B and C, bottom). The lack of consistent timing of the calcium response relative to head velocity and eye velocity suggests that if vestibular or efference copy signals are contributing to the calcium responses, neither dominates across conditions, but are instead summing with or otherwise modifying the dominant, retinal slip-driven calcium signals.

DISCUSSION

The seminal research of Jerry Simpson and his colleagues established foundational knowledge about the signals conveyed to the cerebellar flocculus by its climbing fiber input during oculomotor learning. He first demonstrated that the floccular climbing fibers carry information about retinal image motion, or “retinal slip,” which could reflect an oculomotor error, i.e., a failure of the VOR or OKR to stabilize images on the retina (14, 15, 79). More recently, his laboratory documented that the climbing fibers also convey nonvisual signals to the flocculus, in addition to retinal slip, as evidenced by the modulation of complex spike rate in floccular Purkinje cells during vestibular stimulation in complete darkness (27, 28, 30). The present study builds directly on Dr. Simpson’s groundbreaking research, by comparing the calcium responses driven in the floccular Purkinje cells by their visual and nonvisual inputs.

A climbing fiber forms between 500 and 1,000 synaptic contacts with the dendritic tree of an individual Purkinje cell (80, 81) and can unleash a massive depolarization and calcium influx (33, 45, 82, 83). However, the amplitude of the depolarization and calcium response can vary considerably, even across complex spikes in the same Purkinje cell (35–38, 44, 45, 56). Multiple neural and behavioral factors have been identified that can influence the size of the calcium response, which, in turn, determines whether plasticity occurs in the Purkinje cell’s synaptic inputs and the direction of the plasticity (31, 35, 44, 51, 52, 55, 84). Therefore, the contribution of the different sensory and motor signals carried by the climbing fibers to the induction of plasticity cannot necessarily be inferred from the complex spike responses. Nevertheless, we found that the calcium responses closely mirrored the previously reported complex spike responses of floccular Purkinje cells to nonvisual, as well as visual input. During vestibular stimulation in the dark, calcium increased during contraversive head movements and the accompanying ipsiversive eye movements. These calcium responses were similar in amplitude to those present during visual stimulus conditions that induce oculomotor learning: the visual stimulus alone induces optokinetic reflex adaptation (13, 85–89) and out-of-phase visual-vestibular pairing with the same parameters used in the present study induces robust VOR-increase learning (90–93). Hence, the calcium responses during vestibular stimulation in the dark may also be sufficient to drive plasticity, potentially contributing to the decrease in VOR gain, i.e., habituation, induced by vestibular stimulation in the dark (90, 94). Moreover, during in-phase visual-vestibular pairing, the nonvisual signals may enhance the calcium response to contraversive retinal slip (Fig. 3).

There are a number of mechanisms through which nonvisual signals could influence the population calcium responses of the floccular Purkinje cells. The nucleus prepositus hypoglossi carries vestibular and efference copy signals (95) and provides inhibitory input to the dorsal cap of the inferior olive, which is the source of the climbing fiber input to the flocculus. The input from the nucleus prepositus hypoglossi is located at the spines and near gap junctions of the dorsal cap neurons (96), which could enable vestibular input and eye velocity to modify the gap junction coupling of olivary neurons and, hence, the population spiking response to the excitatory input from subcortical visual pathways (16, 97). The calcium responses of Purkinje cells to their climbing fiber input also can be regulated postsynaptically. Simultaneous parallel fiber input can have a supralinear effect on the dendritic calcium response of a Purkinje cell to climbing fiber input (52, 56). In addition, parallel fibers drive molecular layer interneurons, whose inhibition of the Purkinje cells can reduce the amplitude of the dendritic calcium response to climbing fiber input (31, 51). This provides two potential mechanisms for vestibular or efference copy signals carried by the parallel fibers in the flocculus to modify the amplitude of the calcium responses elicited in Purkinje cells by climbing fibers.

Excitatory synaptic input from the parallel fibers alone can drive an increase in calcium in a Purkinje cell’s dendrite (44, 45, 52, 98), moreover, calcium in the cell body rises and falls with the simple spike rate (99). However, several considerations suggest that the signals recorded with fiber photometry are dominated by climbing fiber-driven calcium responses in the Purkinje cells, rather than parallel fiber- or simple spike-driven calcium. First, the calcium responses driven by parallel fiber input and simple spiking are generally smaller and more spatially restricted than those driven by climbing fiber input (44, 45, 52, 98). Second, the spontaneous calcium transients observed with no vestibular or visual stimulus present occurred at a rate of 1 Hz (Fig. 1), consistent with the 0.5–2 Hz spontaneous firing rate of inferior olive neurons recorded in vivo (34, 69, 100). Clusters of climbing fibers tend to fire synchronously due to the gap junction coupling in the inferior olive (101), which could create a coordinated rise in calcium across the population of Purkinje cells recorded in the spontaneous condition. The similar rates of spontaneous population calcium events to the spontaneous complex spike rate recorded in individual Purkinje cells suggests that complex spiking may be fairly synchronous across the population. Finally, the calcium responses recorded across the set of visual and vestibular test stimuli closely mirrored the previously reported climbing fiber-driven complex spike responses, but not the parallel fiber-driven simple spike responses of floccular Purkinje cells to these stimuli. During the presentation of a visual stimulus alone or visual-vestibular pairing, the simple spike responses are roughly antiphase with the complex spikes (19, 24, 30, 74, 93), and the calcium responses measured with fiber photometry resembled the complex spike responses, increasing during contraversive retinal image slip and decreasing at the time during the stimulus cycle when parallel fiber-driven simple spiking is highest. This suggests that any contribution parallel fiber or simple spike activity might make to the calcium signals recorded with fiber photometry is considerably smaller than the contribution of complex spike-related calcium. During vestibular stimulation in the dark, simple spike responses are significantly smaller than during a visual stimulus alone or visual-vestibular pairing (70, 93, 102–106), hence any influence on calcium should be even smaller. Thus, the calcium responses in the floccular Purkinje cells during vestibular stimulation in the dark are likely to be driven primarily by the nonvisual complex spike responses.

Our findings reinforce the idea that the signals conveyed by the cerebellar climbing fibers are not simple sensory feedback (8, 30, 43, 107, 108), and that more research is needed to fully understand how the instructive signals they carry are computed. During eye movements, retinal slip could indicate an oculomotor error, but this visual feedback may not always be interpreted as an error. The calcium responses observed in floccular Purkinje cells during vestibular stimulation in the absence of visual feedback about oculomotor performance suggest that the error signals controlling oculomotor learning are computed by integrating visual signals encoding retinal slip with vestibular signals and/or efference copy signals. The integration of these multimodal signals may implement a comparison of a prediction of the sensory consequences of a movement with the actual sensory feedback about the movement (109–111). By comparing oculomotor efference copy signals with retinal slip, visual reafference stemming from eye movements might be cancelled to distinguish unpredicted image motion that is caused by either a faulty internal model, or by movement of the external world.

SUPPLEMENTAL DATA

Supplemental Fig. S1: https://doi.org/10.6084/m9.figshare.14364698.

GRANTS

This work was supported by National Institutes of Health grants R01 NS072406, R01 DC004154, and R01 EY031972 and a grant from the Simons Foundation (SCGB 543031, J. L. Raymond).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors..

AUTHOR CONTRIBUTIONS

A.S.F., A.M.S., and J.L.R. conceived and designed research; A.S.F. performed experiments; A.S.F. analyzed data; A.S.F. and J.L.R. interpreted results of experiments; A.S.F. prepared figures; A.S.F. and J.L.R. drafted manuscript; A.S.F., A.M.S., and J.L.R. edited and revised manuscript; A.S.F., A.M.S., and J.L.R. approved final version of manuscript.