Encoding of whisker input by cerebellar Purkinje cells

Abstract

The cerebellar cortex is crucial for sensorimotor integration. Sensorimotor inputs converge on cerebellar Purkinje cells via two afferent pathways: the climbing fibre pathway triggering complex spikes, and the mossy fibre–parallel fibre pathway, modulating the simple spike activities of Purkinje cells. We used, for the first time, the mouse whisker system as a model system to study the encoding of somatosensory input by Purkinje cells. We show that most Purkinje cells in ipsilateral crus 1 and crus 2 of awake mice respond to whisker stimulation with complex spike and/or simple spike responses. Single-whisker stimulation in anaesthetised mice revealed that the receptive fields of complex spike and simple spike responses were strikingly different. Complex spike responses, which proved to be sensitive to the amplitude, speed and direction of whisker movement, were evoked by only one or a few whiskers. Simple spike responses, which were not affected by the direction of movement, could be evoked by many individual whiskers. The receptive fields of Purkinje cells were largely intermingled, and we suggest that this facilitates the rapid integration of sensory inputs from different sources. Furthermore, we describe that individual Purkinje cells, at least under anaesthesia, may be bound in two functional ensembles based on the receptive fields and the synchrony of the complex spike and simple spike responses. The ‘complex spike ensembles’ were oriented in the sagittal plane, following the anatomical organization of the climbing fibres, while the ‘simple spike ensembles’ were oriented in the transversal plane, as are the beams of parallel fibres.

Introduction

Integration of sensory and motor information is one of the most important tasks of the brain. The rodent whisker system combines relatively simple movements with direct sensory feedback (Vincent, 1913; Welker, 1964; Brecht et al. 2006; Kleinfeld et al. 2006), making it an ideal model system to study sensorimotor integration. The cerebellum, being centrally located in both sensory and motor pathways, is essential for sensorimotor integration (Ito, 2000; De Zeeuw & Yeo, 2005; Krakauer & Shadmehr, 2006). Both afferent pathways to the cerebellar cortex, the mossy fibre–parallel fibre and the climbing fibre pathway, convey sensory information from the whiskers (Kleinfeld et al. 1999), and cerebellar output can affect whisker movements (Esakov & Pronichev, 2001; Lang et al. 2006). The mossy fibre–parallel fibre pathway and the climbing fibre pathway converge on Purkinje cells, which are the sole output neurons of the cerebellar cortex.

Thus, Purkinje cells respond to mechanical stimulation of the whiskers (Axelrad & Crepel, 1977; Brown & Bower, 2001; Loewenstein et al. 2005; Holtzman et al. 2006; Ozden et al. 2009), but hardly anything is known about the functional properties of whisker encoding by Purkinje cells. We do not know, for instance, which parameters of whisker movement, like direction or speed, are encoded by Purkinje cells. Furthermore, while it has been shown that there are several areas of the cerebellar cortex that are devoted to the processing of sensory information from the whiskers (Shambes et al. 1978), it is unclear whether there is a somatotopic representation on the single-whisker level within these areas. Such an organization is prominent in other brain regions involved in whisker sensation, including the trigeminal nuclei, the thalamus and the somatosensory cortex (Woolsey & Van der Loos, 1970; Petersen, 2007).

Sensory input from the whiskers reaches the cerebellar cortex via several pathways. It enters the brainstem via the trigeminal nuclei (Torvik, 1956; Clarke & Bowsher, 1962). The inferior olive, where the climbing fibres originate, receives input directly from the trigeminal nuclei (Yatim et al. 1996). In addition, the output of the trigeminal nuclei enters the mossy fibre system via three pathways. There is a direct pathway (Yatim et al. 1996), a short pathway via the pontine nuclei (Swenson et al. 1984), and an indirect connection via the thalamo-cerebro-pontine loop (Kleinfeld et al. 1999) (Fig. 1A). While climbing fibre activity evokes complex spikes in the postsynaptic Purkinje cells, parallel fibre activity can modulate simple spike firing by Purkinje cells.

Figure 1. Multiple single-unit recordings of Purkinje cells in mouse crus 1 and crus 2 in vivo.

A, summary of the main neuronal pathways involved in the integration of tactile input from the whiskers and the coordination of whisker movements (see also Kleinfeld et al. 1999). The cerebellum is centrally located in both the sensory and the motor pathways. B, overview of the recording setup. A set of quartz–platinum recording electrodes placed in crus 1 and crus 2 can be seen in the craniotomy. In the upper part of the photograph are the aluminum guide tubes, allowing the electrodes to be placed individually with an inter-electrode distance of 305 μm. C, raw trace of an extracellular recording from a Purkinje cell in vivo. Purkinje cells produce two kinds of spikes: infrequent complex spikes (*) and frequent simple spikes (•). D, overlay of 312 traces aligned on a complex spike. Simple spike firing is absent for several milliseconds following a complex spike – the ‘complex spike pause’– which is the hallmark of a Purkinje cell single-unit recording. E, complex spike-triggered simple spike histogram. Following a complex spike (at t= 0 ms), simple spikes are completely absent for 5 ms. During the first 20 ms following a complex spike, only very few simple spikes were present. In many Purkinje cells, as in this example, the complex spike pause is followed by a transient increase in simple spike firing rate. The data in the panels C, D and E originate from the same Purkinje cell recording.

Here we studied Purkinje cell responses to mechanical whisker stimulation. We show that sensory input from the whiskers can evoke both complex spike and simple spike responses in Purkinje cells. Remarkably, in the anaesthetised mice the receptive fields of the complex spike and the simple spike responses of a Purkinje cell proved to be different. Simultaneous recordings of multiple Purkinje cells showed that functional ensembles of Purkinje cells cooperate in encoding sensory input. These cells were grouped in two different ways: in sagittal groups based on their complex spike pattern and in transverse groups based on their simple spike pattern.

Methods

Animals

Seven male C57BL/6 mice (6–31 weeks old; Harlan, Horst, The Netherlands) were anaesthetised with isoflurane (2% in O₂) and received lidocaine (∼1 μg) subcutaneously at the surgical location. After reaching a surgical level of anaesthesia, a pedestal was attached to the skull with Optibond adhesive (Kerr Corporation, Orange, CA, USA). After recovery from anaesthesia, they were familiarised with the head-restraint and the recording setup during two or three sessions of 1 h each. On the day of the experiment, the mice were anaesthetised again with isoflurane and a craniotomy was made above crus 1 and crus 2 on the right side. Recordings started at least 30 min after recovery from anaesthesia.

For the electrophysiological experiments involving mechanical whisker stimulation, 41 male C57BL/6 mice (6–31 weeks old; Harlan) were initially anaesthetised with isoflurane (4% in O₂) and subsequently deeply anaesthetised with a mixture of ketamine and xylazine (approx. 110 and 20 mg kg⁻¹, respectively, administered intraperitoneally). At the surgical location, lidocaine (∼1 μg) was given subcutaneously. After reaching a surgical level of anaesthesia, the skin, skull and dura were removed above crus 1 and crus 2 of the right hemisphere. A small recording chamber was attached to the skull with dental cement and filled with 0.9% saline to avoid dehydration of the brain. Throughout the surgery and the subsequent experiment, the mice were kept under anaesthesia. Body temperature, heart rate (ECG) and respiratory frequency were monitored continuously (PowerLab 4/30, ADInstruments, Bella Vista, Australia). The body temperature was kept constant at 37°C and subsequent doses of anaesthesia were given intraperitoneally (approx. 60 mg kg⁻¹ h⁻¹ ketamine and 3 mg kg⁻¹ h⁻¹ xylazine). At the end of the experiments, the mice were killed by cervical dislocation under anaesthesia. All experimental procedures were in accordance with the ethical policy of The Journal of Physiology (Drummond, 2009). As required by Dutch legislation, the experiments were approved by the institutional animal welfare committee (Erasmus MC, Rotterdam, The Netherlands).

Neuronal tracing

Retrograde tracing was performed following injections in various regions of crus 1 and crus 2 of either a gold–lectin conjugate (n= 4) or the cholera toxin b subunit (CTb, n= 3) as described previously (Ruigrok & Apps, 2007). Briefly, mice were anaesthetised with ketamine–xylazine as described above. After making a craniotomy, the tracers were injected in the area where electrophysiological recordings were made, at a depth of approximately 500 μm. For gold–lectin, we used pressure injection of 50–100 nl suspension. CTb was applied by iontophoresis (10 min with 4 μA, anodal, pulsed current). After tracer injections, the mice were allowed to recover for 1 week. Subsequently, they were anaesthetised with pentobarbital (80 mg kg⁻¹, administered intraperitoneally), and fixed by transcardial perfusion with 4% paraformaldehyde. The brains were removed, sliced (40 μm thick) and further processed histologically. From the labelling pattern in the inferior olive, coarse reconstructions of the sagittal zones in crus 1 and crus 2 were made (Apps & Hawkes, 2009).

Whisker stimulation

Whisker stimulation in awake animals was done by repeated air puffs. The air puffs (∼300 mbar (∼30 kPa), 5 ms) were given in a randomly timed manner using a MPPI-2 pressure injector (Applied Scientific Instrumentation, Eugene, OR, USA). The air puff was delivered by a small tube (2 mm diameter), placed approximately 3 cm above the whiskers at a 50 deg angle (with respect to the body axis) to deflect the whiskers in the caudal direction. The air stream was directed away from the face, in order to avoid stimulation of the eye and/or other facial structures.

In anaesthetised mice, mechanical stimulations of the whiskers were applied using a combination of two programmable piezo linear drives (M-663, Physik Instrumente, Karlsruhe, Germany) placed perpendicular to each other, allowing accurate (0.1 μm precision) 2-D manipulation of the whiskers. The piezo drives were controlled using a custom-written LabView (National Instruments, Austin, TX, USA) algorithm. For single-whisker stimulation, one whisker was placed in a small capillary attached to the piezo drives. For multiple-whisker stimulation, the whiskers were placed in a fine mesh attached to the piezo drives. Whisker stimulation occurred at 0.5 Hz (whisker mapping protocol, Fig. 7A–G) or 0.25 Hz (all other experiments). Unless stated otherwise, the extreme position (on average 7 deg deflection; range: 6–9 deg) was reached following a sine waveform in 62.5 ms, after which the whisker was kept at the extreme position for 250 ms and moved backed again in 62.5 ms. The piezo sliders were mounted on a 3-D translation stage (Luigs & Neumann, Ratingen, Germany), so that they could be positioned in such a way that the resting position(s) of the inserted whisker(s) were very close to the native position(s). Movements were either in a fixed direction (caudal) or in a random direction (chosen out of eight possible directions at 45 deg intervals).

Figure 7. Receptive fields of Purkinje cells in anaesthetised mice.

A, the follicles of the large vibrissae are ordered in a grid on the mystacial pad. For the experiments presented in this figure, we confined ourselves to the 14 whiskers depicted here. ‘Rows’ are lines of whiskers ordered in the rostro-caudal plane, ‘arcs’ are lines of whiskers ordered in the dorso-ventral plane. B, during a recording of a Purkinje cell, we tested one by one which of the whiskers elicited a complex spike response. For each responsive whisker, we tested whether the neighbouring whiskers also elicited complex spike responses. We discriminated between direct neighbours and neighbours two or three whiskers away, as well as between neighbours in the same row and in the same arc. C, as for B but for the simple spike responses. D, fraction of Purkinje cells in which a given whisker could elicit a complex spike response. For whiskers in arc ‘3’ we did not have enough data. E, as for D but for the simple spike responses. F, for each of the 28 Purkinje cells tested, the whisker that elicited the largest complex spike response is depicted as a coloured circle on the approximate location, as projected to the brain surface, in crus 1 or crus 2. Colour coding is the same as in A and D. (Near) overlapping locations have been displaced minimally to increase visibility. Purkinje cells that did not have a complex spike response to any of the whiskers tested are not shown. G, as for F but for the simple spike responses. Here, we illustrated the early-negative responses. H, surface-projected locations of Purkinje cells showing a complex spike response to whisker stimulation (filled symbols). Purkinje cells that did not show a complex spike response are indicated by open symbols. Circles: single-whisker stimulation (C2); bars: multiple-whisker stimulation (C-row). Most complex spike responses were found centrally in crus 1 (see also Supplemental Fig. 2). I, as for H but for the simple spike responses. The sagittal and transversal axes are indicated by R (rostral) and C (caudal), and by M (medial) and L (lateral), respectively.

Whisker position tracking

Whisker movement was monitored using a fluid-cooled, high-speed CCD camera (operating at 1 kHz full-frame (480 × 500 pixels) rate; A504k, Basler Vision Technologies, Ahrensburg, Germany). The whiskers were illuminated from below using a custom-designed LED panel. This LED panel worked at a wavelength (640 nm) that is hardly visible to mice (Jacobs et al. 2007). The positions of the whiskers were tracked using a custom-written LabView routine on a dedicated computer running under a real-time operating system (National Instruments).

Electrophysiology

Recording electrodes (2–5 MΩ) were made from quartz-coated platinum–tungsten fibre (outer diameter, 80 μm; Thomas Recording, Giessen, Germany) according to the design of Eckhorn (Mountcastle et al. 1991). The electrodes were placed in a 8 × 4 matrix (Thomas Recording), with an inter-electrode distance of 305 μm. Each electrode was placed individually in the vicinity of a Purkinje cell. The position of the electrode was reconstructed from photographs made of the site of entry and the actual depth (approx. 200–2000 μm) in the tissue. We measured the transversal distance between the site of entry of each electrode and the border between the hemispheres and the vermis as well as the sagittal distance between the site of entry and the border between crus 1 and crus 2 from the photographs (see Fig. 1B). These coordinates were stored and adjusted to a typical microscopical image of a mouse for visualization of the recording sites. Up to 24 electrodes were used simultaneously. The electrophysiological signal was digitised at 12–25 kHz, using a 20–6000 Hz band-pass filter, amplified and stored using a RZ2 multi-channel workstation (Tucker-Davis Technologies, Alachua, FL, USA). Typical recordings were stable for several tens of minutes, while occasionally stable recordings of more than 2 h could be obtained (see also Supplemental Fig. S1).

Electrophysiological data analysis

Neuronal recordings were analysed online using OpenEx software suite (Tucker-Davis Technologies), and later re-analysed offline using custom-written software in LabView. Putative events were first detected using a combination of an amplitude threshold and a time-variant filter (‘Gabor’ method). Next, all putative events were superimposed and a waveform analysis under full visual control was used to select true action potentials. For each Purkinje cell, we constructed a complex spike-triggered simple spike histogram to test for the presence of a pause in simple spike firing following each complex spike (Granit & Phillips, 1956; Bell & Grimm, 1969). Only recordings that had a complex spike pause of at least several milliseconds (complete pause >4 ms, see Fig. 1E) were considered to be ‘single units’ and only these recordings were used for further analysis. When, during a recording, the signal-to-noise ratio decreased, analysis was stopped when there was either no longer a complex spike pause, or when the signal-to-noise ratio no longer allowed reliable detection of spikes.

Statistical significance of peri-stimulus histograms and cross-correlograms was evaluated by defining a control period, during which we calculated the average and the 99% confidence interval. A response or correlation was considered ‘significant’ when it exceeded the threshold during at least one bin of the ‘responsive period’, as defined below. In order to control for ‘false positives’, we searched for threshold crossings during a time interval, equally long as the responsive period, during the control period. Under most conditions, the 99% confidence interval was not associated with any ‘false positives’, but for a subset of experimental groups, the 99% confidence interval was not satisfactory and we adjusted the threshold as described below.

Purkinje cell responses to whisker stimulation

From each recording, two peri-stimulus histograms were made: one for the complex spikes and one for the simple spikes, both with 20 ms bin width. The onset of the whisker movement was at t= 0 ms. The 1 s pre-stimulus interval was considered as the control period, and a response was marked as significant if the peri-stimulus histogram exceeded the 99% confidence interval during at least one 20 ms bin during a defined post-stimulus interval: complex spike response, 0–140 ms after the onset of the stimulus; early-positive simple spike response (0–20 ms); early-negative simple spike response (0–60 ms); late-positive simple spike response (20–200 ms); and late-negative simple spike response (60–200 ms). No threshold crossings were detected during the pre-stimulus interval.

In order to test for direction selectivity, we tested more trials (up to around 1000) and first tested for each Purkinje cell whether the complex and/or simple spikes were significantly modulated by whisker stimulation. Next, we defined the responsive period per Purkinje cell as the time during which the response exceeded the threshold. In these initial steps, we used the aggregate of all directions. Subsequently, we made peri-stimulus histograms for all eight directions and measured the response during the responsive period as a percentage of the firing rate during the 1 s pre-stimulus interval. The responses during the responsive period were plotted in a polar plot, which was evaluated by calculating the circularity of the acquired octagon and divided the obtained value by the circularity of a regular octagon. This ratio was considered the ‘octogonality’. Circularity was defined as 4π* (area/perimeter²). As control, the array containing the directions applied during the experiment was shuffled randomly and used to produce randomised data. Purkinje cells that had less than 50 spikes during the responsive period were excluded from this analysis.

Receptive fields

We measured the receptive field of Purkinje cells by first establishing which individual whiskers evoked a complex spike and/or simple spike response upon mechanical stimulation in a given Purkinje cell. To this end, we tested at least six individual whiskers per Purkinje cell (average, 8; range, 6–12). Per whisker, 200 deflections in the caudal direction were applied at 0.5 Hz. We considered a response significant if it exceeded the 99% confidence interval of the 1 s pre-stimulus interval. To avoid ‘false positives’ as a consequence of binning artefacts, we introduced the additional criterion that the peak of the complex spike response had to exceed three times the baseline average. For the early-negative and late-positive simple spike responses, there were many experiments in which the response was almost exactly 3 s.d. below or above the baseline average (early-negatives: 17 out of 228 whiskers tested (7.5%) had a response between 2.8 and 3.2 s.d.; for late-positives this number was 8 out of 228 (3.5%)). In order to exclude ‘false negatives’ for these categories we lowered the threshold to 2.8 s.d., but rejected experiments in which the threshold was crossed during the baseline recording.

We characterised the ‘receptive field’ per Purkinje cell and response category by starting at the first responsive whisker and checking whether a response was also recorded from the neighbouring whiskers, lying in the same row. Next, the whiskers, still in the same row, but now two places further on were tested, etc. The same procedure was used for the arcs (see Fig. 4A). In all cases, we divided the number of ‘positive’ whiskers for each location (e.g. ‘2 whiskers away in an arc’) by the total number of whiskers tested for that condition. For instance, if we were able to test A1, B1, C1 and D1, and only B1 gave a significant response, this would lead to the following score (for the arc): 1 whisker away: 0 positive, 2 tested (A1 and C1), 2 whiskers away: 0 positive, 1 tested (D1), 3 whiskers away: 0 positive, 0 tested.

Figure 4. Single-whisker stimulation affects both complex spike and simple spike firing in anaesthetised mice.

A, schematic drawing of the organization of the mouse mystacial pad, showing the relative positions of the whiskers used in this study. B, upper trace: programmed trajectory of the C2 whisker, which was attached to a piezo drive. Lower trace: extracellular recording of a Purkinje cell, showing both complex and simple spikes. Note that complex spikes (*) occur shortly after the start of the whisker movement. C, whisker movement tracked with a high-speed CCD camera (sample frequency, 1.0 kHz). Time scale as in D–I. D, raster plot showing the timing of complex spike firing. During each trial, we stimulated whisker C2 according to trajectory depicted in C, but in a random direction. E, as in D but for the simple spikes. Note that the average firing frequency showed some long-term changes, possibly related to variations in the state of anaesthesia (see also Supplemental Fig. 1). The simple spike response was, however, present in periods with higher as well as with lower basal firing rate (data not shown). F, peri-stimulus histogram of the complex spike firing in a representative experiment following stimulation of whisker C2 (777 trials). The largest response was during the movement from the resting position to the extreme position (1). In some experiments, including this one, a second peak was observed around 100 ms later (2). The backward movement evoked only a very small response (3). G, as in F but for the simple spikes. In this experiment, an early-positive simple spike response was present following the forward movements (1), followed by an early-negative response (2). The backward movements triggered an early-negative response (3). H, average peri-stimulus histogram of 7 Purkinje cells showing complex spike responses to C2 whisker stimulation. Only the initial complex spike response to the forward movement is consistent over all experiments (cf. arrow 1 in F). I, average peri-stimulus histogram of 15 Purkinje cells showing simple spike responses to C2 whisker stimulation. Simple spike responses occurred both in response to the forward and to the backward movement, and consisted of two phases: an early and a late phase. Each phase, in turn, had an initial positive and a later negative simple spike modulation. The late-negative simple spike modulation was not observed in single-whisker stimulation trials (see Table 1).

Coherent complex spike firing and synchrony

For pairs of Purkinje cells recorded simultaneously for at least 600 s, we constructed cross-correlograms of the complex spike times using a custom-written LabView routine. The threshold was set at the mean number of complex spikes + 5 s.d. of the −3 to −2 s interval. A pair of Purkinje cells was considered to fire coherently if any of the 10 ms bins between −100 and +100 ms exceeded this threshold. If one, or both, of the bins next to 0 ms exceeded the threshold, we considered that Purkinje cell pair to fire synchronously. We discriminated between 10, 5 and 2 ms synchrony, according to the bin size at which synchrony could still be measured. At 1 ms bins, binning artefacts were so prominent that we refrained from a quantitative analysis of ‘1 ms synchrony’. The synchrony index was calculated as described in De Zeeuw et al. (1997).

Synchrony and sensory input

We selected all Purkinje cell pairs of which both Purkinje cells showed strong complex spike responses to stimulation of the same whisker (response > 5 s.d. of baseline). For these pairs, joint peri-stimulus histograms of the complex spike times were constructed with a custom-written LabView algorithm based on the method described by Aertsen et al. (1989). The histogram over the 45 deg line (containing the synchronous events) was normalized for the changes in complex spike frequency during whisker movement as described by Gerstein (1998).

Simple spike synchrony

Cross-correlograms were made of the simple spike times of pairs of Purkinje cells (>600 s of recording), using 99 bins of 10 ms. The firing pattern was considered to be synchronous if the centre bin exceeded the average baseline firing rate + 2 s.d., corresponding to the 95% confidence interval. As baseline, we considered the interval from −500 to −150 ms.

Statistics

Unless stated otherwise, the data are presented as mean value ±s.d. The statistical tests and threshold for significance used are mentioned in the text and/or figure legends where applicable.

Results

Multiple single-unit recordings of Purkinje cells in vivo

We studied Purkinje cell responses to sensory stimulation of the whiskers. To this end, we made extracellular multiple single-unit recordings of cerebellar Purkinje cells in 7 awake, head-restrained adult, male C57BL/6 mice and in 41 mice under ketamine–xylazine anaesthesia. Quartz-coated platinum–tungsten electrodes were placed in crus 1 and crus 2 of the cerebellar cortex ipsilateral to the stimulated whisker(s) (Fig. 1B). Purkinje cells produce two types of spikes: relatively rare complex spikes (awake mice: frequency (f) = 0.8 ± 0.4 Hz; with ketamine–xylazine anaesthesia: f= 0.6 ± 0.5 Hz) and frequent simple spikes (f= 50 ± 12 Hz and f= 40 ± 21 Hz, respectively) (Fig. 1C). Complex spikes are followed by a pause in simple spike firing of at least several milliseconds (Fig. 1D and E). The occurrence of such a ‘complex spike pause’ is the hallmark of a single-unit recording of a Purkinje cell (McDevitt et al. 1982; Simpson et al. 1996). Only single-unit recordings of Purkinje cells were used for this study. We were able to measure up to around 15 Purkinje cells simultaneously.

Purkinje cell responses to whisker stimulation in awake mice

Air puffs applied to the whisker pad elicited neuronal responses in 20 out of 24 (83%) Purkinje cells in ipsilateral crus 1 and crus 2 of awake mice (Fig. 2). Of these 20 responsive Purkinje cells, 14 (70%) showed both a complex spike and a simple spike response to whisker stimulation, 2 (10%) only a complex spike response and 4 (20%) only a simple spike response.

Figure 2. Whisker stimulation triggers Purkinje cell responses in awake mice.

A, photograph of the whiskers of a mouse. The whiskers located outside the focal plane were cut. The movement of the whiskers along the green line was tracked at 1 kHz. Whisker position was defined as the intersection point of the green line and the whisker. In this experiment, 6 whiskers could be tracked. B, top: positions of the 6 whiskers indicated in A. The colours of the traces correspond to those of the circles in A. The ‘yellow’ whisker moved out of the field of view during air puffs. The timing of the air puffs is indicated by the two black lines. Bottom: the corresponding electrophysiological recording of a Purkinje cell. Complex spikes are indicated by an asterisk. Raster plots of the time stamps of complex spikes (C) and simple spikes (D) during 556 trials. Time t= 0.0 s indicates the onset of the air puffs. Peri-stimulus histograms of the complex spikes (E) and simple spikes (F). In this Purkinje cell, the complex spike response was mono-phasic, while the simple spike response was tri-phasic: an early-positive response (1) was followed by an early-negative response (2). Finally, there was a late-positive response (3). The data in panels A–F originate from the same experiment. G, average peri-stimulus of histogram of all 16 Purkinje cells showing a complex spike response to air puff whisker stimulation. Inset: of the 24 Purkinje cells recorded in crus 1 and crus 2, 16 showed a complex spike response to whisker stimulation. H, average peri-stimulus of histogram of all 18 Purkinje cells showing a simple spike response to air puff whisker stimulation. Inset: of the 24 Purkinje cells recorded in crus 1 and crus 2, 18 showed a simple spike response to whisker stimulation.

In order to increase the visibility of the whisker movements, we trimmed the out-of-focus whiskers prior to the experiment. The positions of the remaining whiskers (on average 6 ± 2, mean ±s.d., Fig. 2A) were tracked automatically (Fig. 2B top). The air puffs were directed at the whisker pad, away from the face to avoid stimulation of the eye and/or other facial structures. In most Purkinje cells (13/16 (81%)), the complex spike response had a latency of maximally 40 ms. The other three cells had a latency of 100 ms. Of the 13 ‘short-latency’ cells, 5 showed an additional second peak. Taken together, the average peri-stimulus histogram shows a clear bi-modal distribution (Fig. 2G).

Simple spike responses were even more variable between individual Purkinje cells than were complex spike responses. Some cells (4/18 (22%)) showed a decrease in simple spike firing, others (11 (61%)) an increase and the rest (3 Purkinje cells (17%)) a combination of both. Particularly striking was the response pattern of the Purkinje cell illustrated in Fig. 2D and F, which showed a tri-phasic response. An initial increase in simple spike firing was directly followed by a period of decreased simple spike firing, after which there was a second, prolonged period of increased simple spike firing. The kinetic properties of the Purkinje cell responses have been summarized in Table 1.

Table 1.

Response kinetics

Category	Anaesthesia	PCs tested	Responsive PCs	Latency (ms)	Duration of response (ms)	Response amplitude (% of baseline)
Complex spike response (0–140 ms)
C2	yes	74	7 (9.5%)*	59 ± 27	30 ± 17	459 ± 211
C-row	yes	68	22 (32.4%)	45 ± 16	40 ± 15	702 ± 449
Air puff	no	24	16 (66.7%)	51 ± 38	27 ± 13	994 ± 1093
Early-positive simple spike response (0–20 ms)
C2	yes	74	6 (8.1%)*	0 ± 0	20 ± 0	122 ± 7
C-row	yes	68	0 (0.0%)	—	—	—
Air puff	no	24	1 (4.2%)	0	20	108
Early-negative simple spike response (0–60 ms)
C2	yes	74	12 (16.2%)	27 ± 10	25 ± 9	78 ± 9
C-row	yes	68	5 (7.4%)	36 ± 9	28 ± 11	85 ± 4
Air puff	no	24	6 (25.0%)	13 ± 10	33 ± 16	58 ± 27
Late-positive simple spike response (20–200 ms)
C2	yes	74	2 (2.7%)*	100 ± 28	20 ± 0	112 ± 3
C-row	yes	68	9 (13.2%)	78 ± 21	40 ± 26	117 ± 8
Air puff	no	24	14 (58.3%)	56 ± 20	127 ± 243	148 ± 35
Late-negative simple spike response (60–200 ms)
C2	yes	74	0 (0.0%)*	—	—	—
C-row	yes	68	5 (7.4%)	100 ± 28	48 ± 23	80 ± 8
Air puff	no	24	5 (20.8%)	96 ± 36	64 ± 36	62 ± 28

Purkinje cell responses to single- and multiple-whisker stimulation (of C2 and C-row (consisting of γ, C1, C2 and sometimes C3), respectively) in mice under anaesthesia. The stimulation protocol was the same as shown in Fig. 4: 350 trials at 0.25 Hz in random directions. Responsive periods were considered to be the interval during which the spike rate passed the threshold (see Methods). Statistical comparisons between single- and multiple-whisker stimulations were made:

^*P < 0.05, Fisher's exact test; the other parameters were not significant (Student's t test). For comparison, the response parameters to air puff stimulation in awake mice are also shown. Because of the qualitative differences in stimulation protocols, we did not test for significant differences between responses in awake and anaesthetised mice. Values are the mean ±s.d.

Whisker-induced decrease in simple spike firing can be independent of complex spikes

In response to air puff stimulation of the whiskers in awake mice, we found in eight Purkinje cells both a complex spike response and a decrease in simple spike firing (Figs 2C–H and 3A). We did not find a Purkinje cell showing a dip in the simple spike rate without also showing a complex spike response. Since complex spikes cause a temporary cessation of simple spikes (Fig. 1C–E), we wondered whether the sensory-induced decreased simple spike firing could be attributed to complex spike firing. To answer this question, for each Purkinje cell recording we separated the trials during which there was a complex spike during the responsive period. For this analysis, the responsive period was defined as the period during which the complex spike response exceeded the threshold (see Fig. 2E). Next, we constructed raster plots of the simple spike times of all trials with a complex spike (Fig. 3B) and of all trials without a complex spike (Fig. 3C). As can also be seen in the associated peri-stimulus histograms (Fig. 3D), the Purkinje cell in the left column shows a whisker stimulation-induced decrease in simple spike firing also in the absence of a complex spike. In contrast, the decrease in simple spike firing observed in the Purkinje cell in the right column was present only in the trials with a complex spike. Of the eight Purkinje cells that had a whisker stimulation-induced decrease in simple spike firing, six cells showed whisker stimulation-induced simple spike inhibition uncorrelated to the occurrence of complex spikes. In the other two recordings, the simple inhibition was tightly correlated to the timing of the complex spikes. Thus, we conclude that, while the complex spike pause can contribute to sensory-induced simple spike inhibition, it cannot explain the majority of the inhibition. Possibly, the inhibitory interneurons of the molecular layer are important for this simple spike inhibition.

Figure 3. Sensory induced simple spike inhibition can be independent of climbing fibre activity in awake mice.

A, peri-stimulus histograms of two representative Purkinje cells. Both Purkinje cells responded to repeated air puff stimulation (at t= 0.0 ms) with complex spike (blue) and simple spike (red) responses. The left Purkinje cell received 338 stimuli, the right Purkinje cell 490. They were recorded from the same mouse and for a large part simultaneously. B, raster plots of simple spike times of all trials with a complex spike response, i.e. with a complex spike during the responsive period (cf. A). C, raster plot of an equal amount of trials as in B, but now from trials without a complex spike response. It can be seen that the simple spike inhibition seen in B is still present in the left Purkinje cell, but not in the right Purkinje cell. Thus, the complex spike pause cannot be the sole cause of sensory-induced simple spike inhibition in the left Purkinje cell. D, peri-stimulus histograms of the simple spike times, comparing the trials with (filled boxes) and without (open boxes) a complex spike response. In 6 out of 8 Purkinje cells, the whisker stimulation-induced simple spike inhibition was not correlated to the complex spike response, when studied at a trial-by-trial base (left column), as it was in the other 2 Purkinje cells (right column).

In conclusion, most Purkinje cells in crus 1 and crus 2 responded to air puff stimulation of the whiskers. However, there was a large variation in neuronal responses, possibly due to the relatively poor reproducibility of air puff stimulation. Therefore, we decided to continue studying Purkinje cell responses to whisker input using accurate piezo-actuators to stimulate single whiskers in mice under ketamine–xylazine anaesthesia.

Complex spike responses to whisker stimulation in anaesthetised mice

We made single-unit recordings of 74 Purkinje cells in ipsilateral crus 1 and crus 2 during single-whisker stimulation (of whisker C2; in random directions). In seven Purkinje cells (9%) a significant complex spike response was recorded (Fig. 4, Table 1). On average, the complex spike response started 59 ± 27 ms after the onset of the movement. In responsive Purkinje cells, the chance of complex spike firing upon whisker stimulation was 4.6 ± 2.1 times the baseline frequency. Even in responsive cells, complex spike firing occurred only in a subset of trials (9 ± 10%). Typically, a response to whisker stimulation consisted of only one complex spike. Sometimes a second, smaller peak was observed in the peri-stimulus histogram, around 100 ms after the first peak. Since this time interval is much larger than the duration of the movement (62.5 ms), it is unlikely that it reflects an off-response. A similar, rhythmic response has, however, been reported previously for forepaw stimulation in cats (Bloedel & Ebner, 1984) and eye-blink stimulation in mice (Van Der Giessen et al. 2008).

Strikingly, the backward movement generally did not evoke a complex spike response (Fig. 4F and H). Since we used eight different directions of whisker movement, alternated in a random fashion, the lack of a complex spike response to the backward movement cannot be explained by a preference for a certain direction of movement (see Fig. 5). Thus, we conclude that repeated stimulations, shortly after each other, fail to elicit complex spike responses.

Figure 5. Direction selectivity of Purkinje cell responses in anaesthetised mice.

A, complex spike responses depended on the direction of whisker movement. Eight peri-stimulus histograms show the complex spike responses for each direction of a typical Purkinje cell. For this Purkinje cell, stimulation of whisker C2 in the dorsal direction had the largest impact. The polar plot shows the number of complex spikes during the response period (the light blue bars in the peri-stimulus histograms) per direction (light blue line), as well as the complex spikes during the same time interval before the onset of the whisker movement (dark blue line). B, the dorso-caudal direction was most often found to be the ‘favourite’ direction of the Purkinje cells. The arrow indicates the number of Purkinje cells for which each direction evoked the largest complex spike response. C, while complex spike responses depended on the direction of the movement, simple spike responses did not. This proved to be true for all four kinds of simple spike responses: early-positive, early-negative, late-positive and late-negative responses. An ‘octogonality’ value (see Methods section) of 1.0 implies no direction selectivity at all. For this analysis, we measured complex spike responses from 15 Purkinje cells, and simple spike responses from 7, 9, 7 and 5 Purkinje cells for the early-positive, early-negative, late-positive and late-negative responses, respectively.

For comparison, we also recorded 68 Purkinje cells during which multiple whiskers were stimulated simultaneously: γ, C1 and C2 (and sometimes also C3). Remarkably, the percentage of responsive Purkinje cells was between three and four times as high as for single-whisker stimulation (22 out of 68 (32%) vs. 7 out of 74 (9%) Purkinje cells, P= 0.0008, Fisher's exact test). Neither the latency (45 ± 16 vs. 59 ± 27 ms, P= 0.1129, Student's t test) nor the percentage of successful trials (13 ± 11 vs. 9 ± 10%, P= 0.3474, Student's t test) differed significantly between multiple- and single-whisker stimulation. This suggests that stimulating more whiskers increases the number of responsive Purkinje cells, but not the amplitude nor the kinetics of the complex spike response in a given Purkinje cell.

Simple spike response to whisker stimulation in anaesthetised mice

Next, we characterised the simple spike response to single-whisker stimulation. We found that, just as air puffs in awake mice (Fig. 2), single-whisker stimulation in anaesthetised mice could induce both increased and decreased simple spike firing. Of both types of modulation, there were an early and a late phase (Fig. 4E, G and I, Table 1). Overall, the sequence was as follows: early-positive, early-negative, late-positive and late-negative response. Although most Purkinje cells showed a combination of some of these four response types, none showed all four. The fraction of Purkinje cells with simple spike responses to single-whisker stimulation was, as it was for complex spike responses, rather low (Table 1). The chance of finding a simple spike response in a given Purkinje cell did not depend on whether or not there was a complex spike response in that Purkinje cell.

The early-positive simple spike response typically consisted of a short response with a short latency, always occurring during the first 20 ms bin of the peri-stimulus histogram. The variation in latency and duration of the later phases of simple spike responses was considerably larger. Remarkably, the early responses were more prominent upon single-whisker stimulation, while the late responses were mainly observed during multiple-whisker stimulation (Table 1).

Direction selectivity

Next, we further characterized which parameters of whisker movement were relevant for the Purkinje cell response. The direction of whisker movement conveys important information about the environment. Hence, in several brain regions, neuronal responses relate to the direction of whisker movement (Bale & Petersen, 2009). We tested whether complex spikes and simple spikes also conveyed information about the direction of whisker movement. To this end we deflected one (C2) or multiple (C-row) whiskers in eight directions (see Fig. 4A).

We recorded the complex spike response to whisker stimulation, and constructed polar plots showing the complex spike response for each of the eight directions used (Fig. 5A). In total, we tested 15 Purkinje cells with a significant complex spike response. The polar plots of these cells had on average an octagonality value of 0.79 ± 0.08, which was significantly smaller than that of a regular octagon (P < 0.001, one-sample t test). This implies that the occurrence of a complex spike response to mechanical whisker stimulation depended on the direction of movement. Of the 15 Purkinje cells tested, 6 ‘favoured’ the dorso-caudal direction (Fig. 5B). In contrast, the simple spike responses did not show any directional selectivity (Fig. 5C).

Activation threshold for complex spike responses

Next, we wondered to which extent amplitude and speed of the whisker movement contributed to the complex spike response. We employed an experimental protocol in which we systematically varied the amplitude of the whisker movement and the time interval needed to reach the maximal amplitude. We used 16 different stimuli, with four values for the amplitude and four for the time interval. A representative example is shown in Fig. 6A. Statistical analysis (Fig. 6B and C) was performed on the data from seven Purkinje cells.

Figure 6. Complex spikes encode the amplitude and speed of whisker movement in anaesthetised mice.

A, complex spike peri-stimulus histograms of a representative Purkinje cell recording during which the amplitude and time to maximal deflection of the whisker movement were systematically varied. Going from left to right, each next column represents data obtained with a doubled stimulus amplitude. Going from top to bottom, each next row represents data obtained with stimuli using half the time to reach the maximal deflection. Consequently, the diagonals (from bottom-left to top-right) have the same angular speed (ω). The green lines show the programmed whisker trajectories. In each histogram, the red line depicts the threshold (upper border of the 99% confidence interval). In the histograms marked with a coloured background, the response threshold was exceeded. From this experiment, it can be concluded that complex spike firing depends on a velocity threshold rather than an amplitude threshold. B, for shorter time intervals, the increase in time to maximal angular velocity (ω_max) is linear with the increase in latency time. For longer time intervals, the latency does not increase linearly anymore. This indicates that for relatively fast movements, the complex spikes encode the maximal angular velocity, but for very long-lasting movements, complex spikes also occur before the maximal angular velocity is reached. C, doubling both the amplitude of a movement and the time to complete that movement does not result in a change in ω_max. Accordingly, the number of complex spikes evoked was unchanged (left bar). Doubling the amplitude, but keeping the time interval constant, was more effective in evoking complex spikes (middle bar) than halving the time, but maintaining the same amplitude (right bar). The analysis shown in B and C is based on the data from 7 Purkinje cells.

An increase in the time to reach the maximal angular velocity (ω_max) postponed the onset of the complex spike response (Fig. 6B). For relatively short times to ω_max, the latency of the complex spike response was changed proportionally to the time to ω_max, irrespective of the absolute amplitude of the response. During very slow movements, the latency of the complex spike response was no longer fully linear to that of the time to ω_max: the complex spike response often occurred before the ω_max was reached (Fig. 6B). The (angular) velocity depends both on the distance (angle) travelled and the time required to do so. A doubling of the amplitude, within the same time interval, increased the number of complex spikes during the response by 87 ± 25% (mean ±s.e.m., Fig. 6C). Halving the time interval, but maintaining the same amplitude, led to an increase of the number of complex spikes during the response of 41 ± 16% (mean ±s.e.m., Fig. 6C). Thus, the number of complex spikes increased with larger angular velocity, whether this increase in velocity was achieved by a larger amplitude or by a shorter time of movement. However, the number of complex spikes was more strongly modified by changes in the amplitude than by changes in the interval time. In conclusion, a complex spike response was produced when the whisker movement exceeded a (velocity) threshold. Once the threshold had been exceeded, the chance of evoking a complex spike depended both on the velocity and the amplitude of the whisker movement. Of these latter two parameters, the amplitude had the strongest impact.

Receptive fields of Purkinje cells

To study the receptive fields of the Purkinje cells, we made recordings of Purkinje cells and stimulated individual whiskers consecutively. When we found a complex spike response to a certain whisker, we tested whether other whiskers also evoked a complex spike response in that same Purkinje cell. In most Purkinje cells, we could find only a single whisker that was able to generate a complex spike response, in line with Axelrad & Crepel (1977). If there was a second ‘active’ whisker, it was always located in the same row as the first whisker (Fig. 7B). The finding that most Purkinje cells reacted only to a single whisker is in line with our finding that multiple-whisker stimulation increased the number of responsive Purkinje cells, but not the amplitude of the response, when compared to single-whisker stimulation (Table 1). We conclude that with respect to complex spike responses, whiskers are largely independent of each other.

For simple spikes, the situation is clearly different. Most Purkinje cells, if they showed a simple spike response to one whisker, also showed simple spike responses to another whisker (Fig. 7C). Remarkably, we could not find a patch of adjacent ‘active’ whiskers. On the contrary, the chance of getting a simple spike response from the direct neighbour of an ‘active’ whisker was almost identical to that of getting a simple spike response from a whisker at the other end of the whisker pad. Furthermore, the whiskers inbetween might perhaps not evoke a simple spike response at all. Thus, while Purkinje cells had a small and clearly defined complex spike receptive field, simple spike receptive fields were larger and did not seem to have a strict structural organization.

In the abovementioned simple spike analysis, we did not take the different types of simple spike modulation (Table 1) into account. Therefore, we repeated the receptive field analysis for each of the four types of simple spike modulation. Late-negative responses were very rare (2 out of 228 whiskers tested (0.9%)), so they were not further analysed. For each of the three other response types, early-positive, early-negative and late-positive responses, we found qualitatively the same results as with the initial simple spike analysis: each Purkinje cell has a large simple spike receptive field without a clear ordering pattern.

The receptive fields of the three types of simple spike modulation were highly correlated, but not identical. This implies that different types of simple spike modulation to stimulation of a certain whisker were often, but not always, manifest together in a single Purkinje cell. For instance, of all 228 whiskers tested, 42 (18%) evoked an early-positive simple spike response and 65 evoked an early-negative response (29%); 26 whiskers evoked both responses, a fraction much larger than expected by chance (P= 0.0007, Fisher's exact test). In contrast, the incidence of complex spike responses was not related to the simple spike response (P= 0.1430 when compared to the incidence of early-positive responses, P= 0.6074 (early-negative responses) and P= 0.2303 (late-positive responses), Fisher's exact test). Thus, complex spike and simple spike receptive fields were markedly different and not related to each other. Subsequently, we calculated the fraction of Purkinje cells in which we could elicit a response by stimulating the individual whiskers (Fig. 7D and E).

Finally, we projected the principal whisker for each recorded Purkinje cell to the surface of crus 1 and crus 2. The principal whisker was defined as the whisker eliciting the largest response. For the simple spike response, we took the early-negative response as a measure for the principal whisker, since this was by far the most abundant type of simple spike response. For both complex spike and simple spike responses there was a highly fractured map without any clear pattern (Fig. 7F and G). Hence, we conclude that the organization of the whisker-sensitive areas in crus 1 and crus 2 does not follow a somatotopic organization, as for instance found in layer IV of the barrel cortex (see Petersen, 2007).

Purkinje cell locations in relation to sagittal zones

In order to further characterize the locations of Purkinje cells receiving sensory input from the whisker system, we made recordings of 134 Purkinje cells during single-whisker (C2) or multiple-whisker stimulation (C-row). For each Purkinje cell, we made 350 whisker movements (at 0.25 Hz) in randomised directions. In a surface projection of the locations of these Purkinje cells in crus 1, it can be seen that Purkinje cells showing complex spike modulation to whisker stimulation are predominantly found in the central area of crus 1 (Fig. 7H, Supplemental Fig. S2). In the medial part of crus 1, i.e. the part bordering the vermis, only a few Purkinje cells showed complex spike modulation. Purkinje cells showing simple spike modulation were distributed more evenly along the medio-lateral axis of crus 1 (Fig. 7I, Supplemental Fig. S2). The distribution of responsive Purkinje cells in crus 2 lacked obvious clustering, both for the complex spike and the simple spike responses (Fig. 7H and I).

Next, we wanted to know how the locations of Purkinje cells responding to whisker stimulation relate to the organization of the cerebellar cortex. It is well known that the organization of cerebellar climbing fibres follows a sagittal zonal pattern, which has been most thoroughly described for the rat (e.g. see Sugihara et al. 2004; Voogd & Ruigrok, 2004; Pijpers et al. 2005), but can also be distinguished in several other species including the mouse (Groenewegen et al. 1979; Garwicz, 1997; Edge et al. 2003; Schonewille et al. 2006b; Sugihara & Quy, 2007). Based on these studies, a general scheme of the sagittal zones was recently provided by Apps & Hawkes (2009). The position of the recordings sites was verified by analysing the labelling patterns from a series of seven micro-injections with retrograde tracers at different locations in crus 1 and crus 2 (Fig. 8A). The labelling pattern obtained in the contralateral inferior olive (Fig. 8C and D) as well as the general scheme of olivo-cerebellar projections served as a tool to define the zonal boundaries within the crura (Fig. 8B). The patterns of retrograde olivary labelling showed that the central regions of the crura receive their climbing fibre afferents mostly from the principal olive (D-zones), whereas the more medially located Purkinje cells received their climbing fibres from the rostro-medial or rostral half of the dorsal and medial accessory olives, respectively (C-zones).

Figure 8. Anatomical connections to crus 1 and crus 2.

A, in 5 animals, we made in total 7 injections with retrograde tracers (4 gold–lectin and 3 cholera toxin b subunit (CTb)). The injection areas (which were ∼500 μm deep) are shown on a cross-section of the cerebellum. B, enlarged image of the right hemisphere. For each injection spot, we established in which sagittal zone(s) it was located on the basis of staining in the inferior olive. The grey lines indicate the most likely locations of the sagittal zones in crus 1 and crus 2. C and D, two representative images of the staining pattern in the inferior olive. The white lines delineate the contralateral inferior olive, the coloured lines indicate the stained area. Note that the colour coding corresponds to that of the injection sites in panels A and B. Gold–lectin staining can be recognized as bright spots, while the CTb staining has a reddish appearance. DAO, dorsal accessory olive; MAO, medial accessory olive; PO, principal olive. Microscopic images of the staining in the ipsilateral trigeminal nuclei: E shows a stained area in the ventral part of the principal trigeminal nucleus (Pr5). F shows staining in the dorsal part of Pr5. In addition, a small spot in the trigeminal motor nucleus (M5) was found. G depicts staining in the ventral part of the spinal trigeminal nucleus pars interpolaris (Sp5i). Staining in the dorsal part of Sp5i can be seen in H. I and J, microscopic images of the staining in the contralateral pontine nuclei (Pn) (RtTg, reticulo-tegmental nucleus). I comes from the same animal as E and G, J comes from the same animal as F and H. The scale bars represent 200 μm in C–J.

Retrograde labelling was also prominent in the ipsilateral trigeminal nuclei, mainly the principal nucleus (Pr5) and the spinal nucleus pars interpolaris (Sp5i) (Fig. 8E–H). In line with the results of Yatim et al. (1996), we found that the dorsal parts of Pr5 and Sp5i projected mainly to the central part of crus 1 and crus 2, while the ventral parts projected more to the medial part (bordering the vermis) and lateral parts of crus 1 and crus 2. In addition, heavy labelling was also noted within the contralateral pontine nuclei (Fig. 8I and J), indicating that the pontine nuclei also project to crus 1 and crus 2. The contralateral trigeminal nuclei and the ipsilateral pontine nuclei also contained retrogradely labelled neurons, but to a much less extent (data not shown).

Complex spike responses were particularly often recorded from the central part of crus 1, and less often from the part bordering the vermis (Fig. 7H, Supplemental Fig. S2). This distribution mirrors the trigemino-olivary projections: the spinal nucleus projects mainly to the rostro-medial part of the dorsal accessory olive and the dorso-medial group and ventral leaf of the principal olive (Uzman, 1960; Huerta et al. 1983; De Zeeuw et al. 1996; Yatim et al. 1996). These regions project to the C3, D1 and D0 zones (Voogd & Ruigrok, 2004; Sugihara & Quy, 2007). As some complex spike responses to mechanical whisker stimulation were also found in the medial part of crus 1, a contribution from the so-called A2 zone, supplied by medial aspect of the caudal part of the medial accessory olive, cannot be excluded (see Figs 7H and 8B) (Voogd & Ruigrok, 2004). Simple spike responses to whisker stimulation were found in Purkinje cells dispersed over crus 1 and crus 2 (Fig. 7I, Supplemental Fig. S2), corresponding to the pattern of the trigemino-cerebellar mossy fibres, that originate from both the principal and the spinal trigeminal nuclei and innervate virtually all of crus 1 and crus 2 (Yatim et al. 1996).

Coherent firing of Purkinje cell pairs

Purkinje cells can cooperate in anatomically and functionally defined ensembles. We investigated coherent complex spike firing in our recordings to further define the extent to which Purkinje cells are functionally related to each other in crus 1 and crus 2 (Fig. 9A and B). We made cross-correlograms of the complex spike times of each Purkinje cell pair for which we had at least 600 s of simultaneous recording (Fig. 9C and D). Of the 295 Purkinje cell pairs tested, 124 (42%) showed a non-uniform cross-correlogram, indicating that the complex spike firing of these Purkinje cell pairs was not random with respect to each other. Correlated firing does not necessarily imply synchronous firing. Therefore, we made cross-correlograms with different bin sizes and tested whether one or both of the two bins around 0 ms exceeded the threshold, which was defined as the average + 5 s.d. of the interval −3 to −2 s. This led to three categories of synchrony: where complex spikes had an increased chance to occur within 10, 5 or 2 ms of each other. In total, 72 (24%) Purkinje cell pairs showed ‘10 ms synchrony’, of which 46 (16% of all pairs) showed ‘5 ms synchrony’ and 27 (9% of all pairs) ‘2 ms synchrony’ (Fig. 9F).

Figure 9. Synchronous firing in anaesthetised mice.

A, raster plot showing the occurrence of complex spikes in 7 Purkinje cells recorded simultaneously. Several Purkinje cell pairs showed a marked correlation in complex spike firing. Two pairs have been highlighted here: PC1 vs. PC5 and PC4 vs. PC7. Complex spikes that occurred within 10 ms of each other are marked red or green, respectively. Complex spikes that occurred within 100 ms of each other are marked dark red and dark green, respectively. B, relative locations of the recording electrodes used for the recordings in A. The inter-electrode distance is 305 μm (heart-to-heart) in the x- and y-direction. C, cross-correlogram of the complex spikes of PC1 vs. those of PC5 (2 ms bins). Inset: enlargement of the middle part (from −500 ms to +500 ms). The black line indicates the average number of complex spikes during the interval from −3 to −2 s. The green line indicates the threshold (mean + 5 s.d.). D, as for C but now for the pair PC4 vs. PC7. E, the synchrony indices for the two pairs shown in C and D and for the pair PC1 vs. PC3 (that did not show any correlation in complex spike firing). F, in total, 295 pairs of Purkinje cells were compared. In 42% of these, the patterns of complex spike firing were significantly correlated. About half of these pairs showed synchronous firing within 10 ms, with a part showing complex spike synchrony within 5 or even within 2 ms. G, the timing of the maximal peak in the cross-correlogram (y-axis) increased with larger (Euclidean) distances between the two Purkinje cells of each pair. Black line: linear regression line. H, as for G but now for the difference in depth. Purkinje cells that are located at a deeper location tended to fire before more superficial Purkinje cells. *P < 0.05 (linear regression). I, synchronous simple spike firing in a Purkinje cell pair of which both Purkinje cells responded to stimulation of the same whisker (10 ms bins). J, the centre part of the histogram in I, but now with 2 ms bins (green line). The red line indicates the histogram of the same experiment, but without the 500 ms following each whisker stimulation. Note that the y-axis is normalized to correct for changes in absolute number of simple spikes between both conditions. K, of all 20 Purkinje cell pairs of which both cells responded with a simple spike response to stimulation of the same whisker(s), 9 pairs showed simple spike synchrony (right). Of the 45 other pairs tested, only 5 showed simple spike synchrony (left). P < 0.01 (Fisher's exact test)

Thus, although we did find well-timed (2 ms) complex spike synchrony in 9% of the Purkinje cell pairs, in 33% of the Purkinje cell pairs there was coherent complex spike firing that was not truly synchronous. We wondered whether we could find any evidence for a travelling wave of complex spike activity over the cerebellar cortex. To this end, we plotted the distance between the cell pairs against the time of the maximal correlation (10 ms bins). The longer the distance between two Purkinje cells, the longer the average time between complex spikes (P= 0.0004, linear regression, slope = 14.7 ms mm⁻¹, Fig. 9G). Next, we investigated whether we could find any difference between the distance in transversal (medio-lateral) or sagittal (rostro-caudal) direction or in depth. We found no correlation between either transversal or sagittal distance and latency time (P= 0.1143 and P= 0.0854, respectively, linear regression, data not shown). However, when we looked at differences in the depth, we found that Purkinje cells that were situated at a deeper location tended to fire prior to more superficially located Purkinje cells (P= 0.0040, slope =−12.3 ms mm⁻¹, linear regression, Fig. 9H). Hence, we conclude that the difference in depth is the most important factor in determining the relative timing of Purkinje cells.

Simple spike synchrony, while prominent in the paramedian lobule, has been shown to be virtually absent in crus 2 (Heck et al. 2007). In line with this, we found little evidence for synchronous firing of simple spikes, except in Purkinje cell pairs that responded to the same whisker input. Of the 20 Purkinje cell pairs showing simple spike co-modulation, i.e. both Purkinje cells had simple spike responses to stimulation of the same whisker(s), 9 (45%) showed simple spike synchrony. Of the 45 Purkinje cell pairs tested that did not show this simple spike co-modulation, only 5 pairs (11%) showed simple spike synchrony (P= 0.0066, Fisher's exact test, Fig. 9K). In line with Heck et al. (2007), the extent of simple spike synchrony in crus 1 and crus 2 was relatively small, even in synchronously firing Purkinje cell pairs (Fig. 9I). The relative power of the synchrony was similar during periods of whisker movement and rest (Fig. 9J).

Complex spike synchrony and sensory input

Next, we investigated the impact of sensory input on coherent complex spike firing. To this end, we selected all simultaneously recorded Purkinje cell pairs which showed a complex spike response to whisker stimulation (15 out of 295 pairs (5%)). Indeed, 12 of these 15 ‘co-modulating’ Purkinje cell pairs (80%) showed correlated complex spike firing, compared to only 112 out of 280 other Purkinje cell pairs (40%) (P= 0.0028, Fisher's exact test). In line with this, complex spike synchrony also occurred much more often in co-modulating than in other Purkinje cell pairs (9 out of 15 (60%) vs. 63 out of 280 (23%) pairs, respectively, P= 0.0027, Fisher's exact test, 10 ms synchrony, data not shown).

Interestingly, when discriminating between Purkinje cell pairs of which both cells showed a strong complex spike response to whisker stimulation (>5 s.d.) and the other co-modulating pairs (8 and 7 pairs, respectively), it became immediately apparent that the strongly co-modulating Purkinje cell pairs had a much higher incidence of synchronous firing than the weaker co-modulating Purkinje cell pairs (Fig. 10D). Indeed, complex spike synchrony in weakly co-modulating Purkinje cell pairs did not differ from that in non-co-modulating Purkinje cell pairs. At the same time, strongly co-modulating Purkinje cell pairs showed a strong increase in synchrony (7 out of 8 pairs (88%) for 10 ms synchrony, compared to 63 out of 280 non-co-modulating cell pairs (29%), P= 0.0002; 5 out of 8 pairs (63%) for 5 ms synchrony, compared to 40/280 non-co-modulating pairs (17%), P= 0.0029; and 4 out of 8 pairs (50%) for 2 ms synchrony, compared to 23/280 non-co-modulating pairs (10%), P= 0.0033, Fisher's exact test). Thus, we conclude that strongly co-modulating Purkinje cell pairs belong to functional ensembles of synchronously active Purkinje cells.

Figure 10. Functional ensembles have an increased tendency to fire complex spikes in synchrony in anaesthetised mice.

A, joined peri-stimulus histogram (JPSTH) of two Purkinje cells that both showed a strong complex spike response upon whisker stimulation. The whisker movement is shown in green. The JPSTH of the complex spikes of the two simultaneously recorded Purkinje cells (upper right) is composed of a trial-by-trial correlation of the raster plots of the two Purkinje cells. The peri-stimulus histograms of the two Purkinje cells are also shown. The histogram in the lower left corner is the histogram over the bins on the 45 deg line in the JPSTH, normalized for the firing rate per bin. It can be seen that the degree of synchrony does not vary to a large extent over the period, which indicates that the complex spike synchrony is not significantly larger during whisker movement than during periods of rest. B, cross-correlograms of the Purkinje cell pair depicted in A. The green line is the cross-correlogram of the whole recording, the black line is constructed with the omission of the 500 ms following the onset of stimulation (see the black and blue lines in the histogram in the lower left corner in A). The y-axis has been corrected for the different absolute numbers of complex spikes for the two conditions. C, synchrony indices for the whole trace (green line) and for the trace without the 500 ms following movement onset (black line). For comparison, the brown line shows the synchrony indices after a random shuffling of the inter-spike intervals (ISIs). The data in the panels A, B and C originate from the same Purkinje cell pair. D, Purkinje cells that both show a strong modulation in complex spike firing (>5 s.d. of baseline rate, ‘strong co-modulation’) have a strongly increased chance to fire in a correlated or even a synchronous manner. Purkinje cell pairs of which both Purkinje cells showed a significant complex spike response to stimulation of the same whisker(s) (>3 s.d., but not both >5 s.d., ‘weak co-modulation’) did show an increase in correlated firing, but hardly of truly synchronous firing. *P < 0.01 (as compared to fraction of non-co-modulating Purkinje cell pairs, Fisher's exact test). E, Purkinje cell pairs showing strong complex spike co-modulation were located in a sagittal band. Each line connects the surface-projected locations of two Purkinje cells showing strong complex spike co-modulation. F, as for E but for the simple spikes. Simple spike co-modulation is predominantly found along the transversal axis.

When two Purkinje cells showed a complex spike response to stimulation of the same whisker(s), they were most often found within a sagittal zone (Fig. 10E). Remarkably, when two Purkinje cells both showed a simple spike response to stimulation of the same whisker(s), they were preferably oriented transversally (Fig. 10F). On average, the transversal distance between two Purkinje cells showing complex spike co-modulation was 0.23 ± 0.15 mm, while that of two simple spike co-modulating cells was 0.68 ± 0.37 mm (P= 0.0002, Student's t test). Thus, while the complex spike co-modulation is organized in line with the sagittal projection zones of the inferior olive (Ruigrok, 2010), that of the simple spike co-modulation follows, at least in the anaesthetised state, roughly the trajectories of the parallel fibres (Braitenberg & Atwood, 1958).

To compare complex spike synchrony during rest and during whisker movements, we constructed joint peri-stimulus histograms (Fig. 10A). The dots on the 45 deg line reflect true synchrony of the two Purkinje cells. The histogram of this 45 deg line, after correction for changes in the complex spike frequency, shows changes in synchrony over time (Fig. 10A, bottom left). It turned out that the complex spike synchrony is constant throughout the recording, irrespective of the whisker movement. Consistently, cross-correlograms of traces including and excluding the periods of whisker movement were nearly identical (Fig. 10B). Yet, we found in all eight strongly co-modulating Purkinje cell pairs that the synchrony index was slightly lower in the parts of recording excluding the whisker movements (Fig. 10C). Taken together, we conclude that Purkinje cells ‘listening’ to the same sensory input have a large chance to belong to the same functional ensemble of Purkinje cells, the location of which is not necessarily confined to a narrow anatomical band. However, the sensory input itself does not, or only to a limited amount, acutely contribute to the complex spike synchrony.

Discussion

Although the rodent whisker system is widely used as a model system to study sensory input, functional data on whisker encoding by the cerebellum were virtually absent. We show here that mechanical stimulation of a single whisker elicits both complex spike and simple spike responses in cerebellar Purkinje cells. The complex spike responses differ from the simple spike responses with respect to their receptive fields. In addition, Purkinje cell pairs showing complex spike responses to stimulation of the same whisker(s) are predominantly oriented along the sagittal plane, whereas Purkinje cell pairs showing simple spike responses to stimulation of the same whisker(s) are mainly oriented along the transversal plane. In either case, synchrony is more prominent among Purkinje cell pairs responding to the same whisker than between other Purkinje cell pairs, indicating the existence of functional ensembles of Purkinje cells cooperating in sensory encoding.

The impact of anaesthesia

The use of anaesthesia can have profound effects on the physiology of Purkinje cells (e.g. see Schonewille et al. (2006a)). Indeed, the average firing rate in our hands was about 20–25% lower in the Purkinje cells recorded in anaesthetised mice compared to those recorded in awake mice. Yet, qualitatively, the Purkinje cell responses to mechanical stimulation of the whiskers were rather similar in both states. In total, whisker responses were more often found in Purkinje cells of awake mice than in those of anaesthetised mice, and the ‘awake’ responses also tended to be larger and longer lasting. However, these effects can probably largely be explained by the larger number of whiskers stimulated with the air puff stimulation in awake mice than with the piezo-stimulation in anaesthetised mice. It should be noted that, in anaesthetised mice, stimulating three whiskers instead of one also increased the prevalence of Purkinje cell responses considerably (Table 1). The use of anaesthesia enabled us to make longer and more stable recordings, allowed a more accurate stimulation of single whiskers and, last but not least, avoided the interference with spontaneous whisker movements. For future studies, it will be very interesting to investigate the integration of the sensory input from the whiskers and the modulation of the whisker movements by the cerebellar Purkinje cells. In our present study, however, we aim to give a comprehensive description of the physiologically relevant parameters of whisker encoding by Purkinje cells.

Complex spike population coding

Whisker input can increase complex spike firing dramatically – in some cells by more than 2500%. Nevertheless, we found that on average only around 10% of the trials evoked a complex spike with a stimulus of mediocre strength. The incidence of complex spike responses could be modulated by direction, amplitude and speed of the whisker movement. Thus, although the complex spike response is in itself an all-or-none response, the chance of complex spike firing depends on the properties of the actual whisker movement. Interestingly, the number of whiskers stimulated hardly affected the complex spike response in a given Purkinje cell. In line with this, we found that most Purkinje cells received climbing fibre input from a single whisker only, confirming a preliminary study by Axelrad & Crepel (1977). Thus, the more whiskers stimulated, the larger the number of Purkinje cells showing a complex spike response, but the percentage of trials with a complex spike per Purkinje cell response remained constant.

The relatively low incidence of complex spike responses to whisker stimulation, even in responsive Purkinje cells, may call the physiological relevance of these responses into question. Our finding that the complex spike responses of individual Purkinje cells form poor representations of the whisker input confirms earlier studies using two-photon imaging (Ozden et al. 2009; Schultz et al. 2009). These latter studies describe that, in contrast to individual Purkinje cells, ensembles of coherently firing Purkinje cells do provide a reliable encoding of the sensory input. In line with this, we also found widespread complex spike synchrony. Remarkably, pairs of Purkinje cells of which both cells showed a complex spike response to stimulation of the same whisker(s), had a particularly high degree of complex spike synchrony and were mostly located in the sagittal plane. Thus, Purkinje cells are clustered in functional ensembles based on their complex spike firing pattern. Forming such functional ensembles greatly increases the reliability of sensory encoding (Ozden et al. 2009; Schultz et al. 2009).

Complex spike synchrony is controlled by olivary activity (Llinás & Sasaki, 1989; Blenkinsop & Lang, 2006; Marshall et al. 2007). The orientation of the climbing fibre projections in the sagittal plane (Voogd & Glickstein, 1998; Ruigrok, 2010) fits with our observation that functional ensembles of Purkinje cells are also largely oriented sagittally. Since also the output of Purkinje cells to the cerebellar nuclei is organized in sagittal zones, this complex synchrony may enhance the Purkinje cell impact on the neurons of the cerebellar nuclei (Gauck & Jaeger, 2000; Hoebeek et al. 2010).

Complex spike synchrony has been shown to be enhanced during motor activity, but not during sensory processing (Welsh et al. 1995; Welsh, 2002). In line with this, we found that the complex spike synchrony was not instantaneously enhanced following whisker stimulation, but rather reflected a continuous functional binding.

Fractured somatotopy and sagittal zones

Local field potential recordings in the granule cell layer have shown that the lobules crus 1 and crus 2 have a fractured somatotopy, implying that there are patches of granule cells responding to tactile and/or nociceptive stimuli to specific parts of the body (Eccles et al. 1972; Shambes et al. 1978; Bower & Woolston, 1983; Bower, 1997). In rats, the largest whisker projection area in the cerebellar cortex is located in crus 1, occupying the largest part of that folium. In addition, several smaller patches are situated in other folia, including crus 2 (Joseph et al. 1978; Shambes et al. 1978). Such a detailed somatotopic mapping has not been made at the climbing fibre level, but there is a large, although not complete, similarity in the receptive fields of the complex spikes of Purkinje cells and those of the underlying granule cells (Brown & Bower, 2001). This is in line with the finding that the projection areas of the mossy fibre and the climbing fibre are largely, but not completely, similar (Thach, 1967; Eccles et al. 1972; Garwicz et al. 1998; Voogd et al. 2003). In general, while the climbing fibres terminate in well-defined sagittal zones (Armstrong et al. 1973a,b; Groenewegen & Voogd, 1977; Ruigrok, 2010), the mossy fibre terminal zones are broader and have less-defined borders (Wu et al. 1999; Apps & Hawkes, 2009).

Tactile stimulation induces climbing fibre activity in a longitudinal microzone (Garwicz, 1997; Garwicz et al. 1998; Ozden et al. 2009; Schultz et al. 2009). At first sight, such longitudinal zones of both the climbing and the mossy fibre projections contradict the fractured somatotopy of the cerebellum (see Shambes et al. 1978). It has been suggested, however, that each patch is composed of one or several of these microzones (Garwicz et al. 1998; Apps & Hawkes, 2009). Our data on the Purkinje cell responses to well-defined whisker stimulations support both theories. The locations of the Purkinje cells responding to whisker stimulation (Fig. 7H and I) were rather similar to the whisker patches described in rats (Shambes et al. 1978). At the same time, Purkinje cell pairs showing complex spike responses to stimulation of the same whisker(s) were largely oriented in the sagittal plane. These findings are in support of the ‘one-map hypothesis’ (see Apps & Hawkes, 2009): the fractured somatotopy and the longitudinal zones might actually be different descriptions of the same underlying map.

Receptive field topography

Ekerot & Jörntell (2001) applied ‘natural’ stimulations to the cat forepaw and measured the receptive field in the C3 zone of the cerebellar cortex. They found that the receptive fields of the complex spike responses in Purkinje cells were largely similar to those of the mossy fibres terminating in the granular layer below these Purkinje cells. Thus, they concluded that the receptive fields of the mossy fibres and the climbing fibres were largely similar, confirming earlier reports (Thach, 1967; Eccles et al. 1972; Bower & Woolston, 1983; Garwicz et al. 1998). A substantial fraction of their Purkinje cells (36%) showed a decrease in simple spike firing following stimulation of the (centre of the) climbing fibre receptive field. This inhibition can probably be explained by the action of the molecular layer interneurons, which have a very similar receptive field as the climbing fibres (Ekerot & Jörntell, 2001). The importance of molecular layer interneurons is supported by the finding that cerebellar granule cells are largely silent during rest, but can be activated by high-frequency bursts of mossy fibre inputs, which in turn can be triggered by tactile input (Chadderton et al. 2004; Rancz et al. 2007). Because of the low firing rate of the granule cells during rest, it is unlikely that their activity can be further decreased by diminished mossy fibre input. Hence, the decrease in simple spike firing must have its origin within the cerebellar cortex, and the molecular layer interneurons seem to be the most likely cause of sensory-induced simple spike inhibition. Molecular layer interneurons receive similar inputs as the Purkinje cells in the same sagittal zone (Ekerot & Jörntell, 2001) and readily inhibit simple spike firing (Eccles et al. 1966; Häusser & Clark, 1997).

Remarkably, while the receptive fields of the mossy fibres and of the climbing fibres largely overlap with that of inhibited simple spike firing, increased simple spike firing could also be evoked by skin stimulation, but had a different receptive field (Ekerot & Jörntell, 2001). Finally, the receptive field size of the simple spike responses is subject to bidirectional plasticity (Jörntell & Ekerot, 2002). The topography of the Purkinje cell receptive fields on the scale of individual whiskers partially confirms that of cutaneous stimulations of the cat forepaw (see Ekerot & Jörntell, 2001). Whisker stimulation evoked both inhibitory and excitatory simple spike responses to tactile stimulation. As with forepaw stimulation, the receptive fields of the simple spike and complex spike responses of a given Purkinje cell could be different.

While the simple spike responses of a given Purkinje cell to forepaw stimulation were either inhibitory or excitatory, depending on the stimulation area, simple spike responses to whisker stimulation were often bi-phasic: an increase in simple spike firing was directly followed by a period of simple spike inhibition. The temporal sequence, first excitation, then inhibition, is in line with the putative role of molecular layer interneurons in causing the inhibition of simple spike firing, as discussed above. Another possible source of simple spike inhibition, the complex spike pause, was found to have only a limited effect (Fig. 3), making the involvement of molecular layer interneurons more likely. Because of the bi-phasic nature of simple spike responses in individual Purkinje cells to single whisker stimulation, the receptive fields of inhibitory and excitatory simple spike responses were closely related, and not seperated as following forepaw stimulation. The receptive fields of the complex spike responses were independent of those of the simple spike responses, but the receptive fields of inhibitory and excitatory simple spike responses to single-whisker stimulation were highly related, suggesting that molecular layer interneurons indeed have a local action.

Simple spike responses occurred with different latencies. We discriminated between early (latency <20 ms) and late responses (latency >40 ms). In both cases, an excitatory response could be followed by an inhibitory one. A similar division between early and late responses to sensory stimulation has been reported previously, and it has been shown that late responses involve the cerebral cortex (Kennedy et al. 1966; Woolston et al. 1981; Morisette & Bower, 1996). Both the somatosensory cortex and the motor cortex may therefore be involved in the late response (Fig. 1A). Early responses, therefore, most probably reflect activity of the (direct) trigemino-cerebellar and the trigemino-ponto-cerebellar mossy fibre pathways (Fig. 1A). We found that early responses were primarily associated with single-whisker stimulation, and late responses by multiple-whisker stimulation, indicating that both pathways convey different information.

Radial vs. beam hypothesis

The hypothesis of fractured somatotopy has been linked to the radial hypothesis of granule cell action (Bower, 1997). Granule cell axons ascend to the molecular layer, where they bifurcate and form long, transverse parallel fibres (Ramón y Cajal, 1888; Braitenberg & Atwood, 1958; Brand et al. 1976). The radial hypothesis poses that the ascending part of the axon is functionally (much) more relevant than the transverse parallel fibre beams (Llinás, 1982; Bower, 1997; Cohen & Yarom, 1998). This hypothesis is supported by the finding that the receptive fields of the complex spikes in Purkinje cells and those of the underlying granule cells are largely overlapping (Brown & Bower, 2001). Interestingly, we found that Purkinje cells of animals in the anaesthetised state show simple spike responses to stimulation of the same whisker(s) and simple spike synchrony, were predominantly oriented in the transversal plane (Fig. 10F), perpendicular to the Purkinje cells responding with complex spikes (Fig. 10E). These results remain to be confirmed in awake animals, but they raise the possiblity that there are functional ensembles of Purkinje cells oriented along the parallel fibre beams.

Our data strongly suggest that the input from a single whisker is not projected to a single spot in the cerebellar cortex, but spread over a larger area. This could facilitate the rapid integration of tactile input from specific areas into a larger sensory context, as suggested by Bower (1997). Thus, the cerebellar cortex acts differently to the cerebral cortex, where sensory input is first processed in a specific region, and integration with other sensory inputs occurs only at a later stage (Engel et al. 2001). In line with this, the receptive fields of the whiskers were dispersed over crus 1 and crus 2 and not grouped as in the barrel cortex (Woolsey & Van der Loos, 1970; Petersen, 2007).

Conclusion

To our knowledge, this is the first study on the encoding of somatosensory input to Purkinje cells using such well-defined minimal receptive fields as single whiskers. From our study, it has become clear that complex spike responses have a very small receptive field, consisting of only one or a few whiskers, while the simple spike firing pattern of a given Purkinje cell is modulated by several whiskers. The simple spike receptive field lacks a clear ordering. Furthermore, the receptive fields of nearby Purkinje cells are apparently completely intermingled. We propose that this ordering, together with the previously described fractured somatotopy (Shambes et al. 1978; Bower, 1997), facilitates the integration of sensory information from various sources. In addition, we provide evidence that, at least in the anaesthetised state, Purkinje cells cooperate in different functional ensembles, based on shared receptive fields and synchronous firing. In line with the cerebellar anatomy, complex spike responses are oriented in the sagittal plane and simple spike responses in the transversal plane. This implies that individual Purkinje cells are members of at least two functional ensembles, further facilitating the rapid integration of diverse sensory inputs. In view of the impact of the cerebellum on motor output, we expect that such rapid integration of sensory input will contribute strongly to the integration of sensory and motor commands and thus to the motor output of the animals.

Glossary

Abbreviations

C57BL/6 mice

C57 black 6 mice

CS

complex spike

CTb

cholera toxin b subunit

DAO

dorsal accessory olive

ISI

inter-spike interval

JPSTH

joined peri-stimulus histogram

LED

light-emitting diode

M5

trigeminal motor nuclei

MAO

medial accessory olive

PC

Purkinje cell

Pn

pontine nuclei

PO

principle olive

Pr5

trigeminal principal nucleus

RtTg

reticulo-tegmental nucleus

Sp5i

trigeminal spinal nucleus pars interpolaris

SS

simple spike

Author contributions

All experiments were performed at the Erasmus MC. L.W.J.B., S.K.E.K., J.S., T.J.H.R. and C.I.D.Z. conceived and designed experiments, L.W.J.B., S.K.E.K., J.S., B.F.M.R., F.Z., B.V.D.E., C.B.O., J.W.P., J.R.D.G., T.J.H.R. and C.I.D.Z. performed and analysed the experiments, L.W.J.B., S.K.E.K., T.J.H.R. and C.I.D.Z. wrote the manuscript. All authors approved the final version.