Role of granule-cell transmission in memory trace of cerebellum-dependent optokinetic motor learning

Significance

Purkinje cells and vestibular nuclei receive optokinetic and retinal slip signals that induce optokinetic motor learning. The role of multiple cerebellar circuits in the adaptive optokinetic response (OKR) was investigated by selective and reversible blockade of granule-cell transmission to Purkinje cells. This blockade abrogated both short-term and long-term OKR adaptation, but adaptive OKRs were immediately induced when the granule-cell transmission was recovered in pretrained, adaptive OKR-negative mice. Eye movement by electrical stimulation of the flocculus was elevated by long-term but not by short-term OKR training. Simultaneous flocculus and OKR stimulation produced normal OKR adaptation in adaptive OKR-negative mice. This study demonstrates that vestibular nuclei serve as a critical circuit responsible for formation and storage of OKR adaptation.

Abstract

Adaptation of the optokinetic response (OKR) is an eye movement enhanced by repeated motion of a surrounding visual field and represents a prototype of cerebellum-dependent motor learning. Purkinje cells and vestibular nuclei (VN) receive optokinetic and retinal slip signals via the mossy fiber-granule cell pathway and climbing-fiber projections, respectively. To explore the neural circuits and mechanisms responsible for OKR adaptation, we adopted the reversible neurotransmission-blocking (RNB) technique, in which granule-cell transmission to Purkinje cells was selectively and reversibly blocked by doxycycline-dependent expression of transmission-blocking tetanus toxin in granule cells. Blockade of granule-cell inputs abolished both short-term and long-term OKR adaptation induced by repeated OKR training, but normal levels of both responses were immediately evoked in the pretrained RNB mice by OKR retraining once granule-cell transmission had recovered. Importantly, eye movement elicited by electrical stimulation of the cerebellar focculus was elevated by long-term but not by short-term OKR training in adaptive OKR-negative RNB mice. Furthermore, when the flocculus of adaptive OKR-negative RNB mice was electrically excited in-phase with OKR stimulation, these mice exhibited long-term adaptive OKR. These results indicate that convergent information to the VN was critical for acquisition and storage of long-term OKR adaptation with conjunctive action of Purkinje cells for OKR expression. Interestingly, in contrast to conditioned eyeblink memory, the expression of once acquired adaptive long-term OKR was not abrogated by blockade of granule-cell transmission, suggesting that distinct forms of neural plasticity would operate in different forms of cerebellum-dependent motor learning.

The cerebellum is the critical site for motor learning, and adaptation of the optokinetic response (OKR) is a prototype of cerebellum-dependent motor learning (1–4). Continuous oscillatory rotation of a screen for a few hours strengthens the gain of the OKR, which goes back to basal levels in the light after training (short-term OKR) (2, 3). Repeated training by screen rotation for a few days progressively increases the OKR and long-lastingly maintains the enhanced OKR (long-term OKR), when animals are kept in the dark after training (2, 3, 5). The basic eye movement (basic OKR) observed in untrained naïve animals is mediated by a closed-loop circuit consisting of nucleus reticularis tegmenti pontis (NRTP), vestibular nuclei (VN), and oculomotor nuclei (OMN) (6, 7) (Fig. S1). In OKR adaptation, optokinetic signals are transmitted to the VN directly through mossy fibers derived from NRTP and indirectly to Purkinje cells via the mossy fiber-granule cell pathway whereas retinal slip signals are transmitted to Purkinje cells via the nuclei of the accessory optic system (AOSn)-inferior olive-climbing fiber pathway (6, 7) ( Fig. S1). A number of studies using lesion analysis (3, 8) and electrophysiological (9–11), pharmacological (3, 12, 13), and gene knock-out techniques (5, 14–16) demonstrated that the circuits of both Purkinje cells and VN are indispensable for memory trace of the adaptive OKR. However, OKR adaptation involves multiple learning processes: at least acquisition, expression, and storage of motor learning. The neural circuits and underlining mechanisms involved in OKR adaptation still largely remain to be elucidated.

To address this issue, we adopted a gene-manipulating technique termed reversible neurotransmission-blocking (RNB) (17, 18). In the RNB technique, tetanus toxin is restrictedly expressed in granule cells under the control of a tetracycline-controlled reverse transactivator (rtTA). Tetanus toxin specifically cleaves synaptic vesicle VAMP2 (19), resulting in blockade of transmitter release from synaptic vesicles (19) (Fig. S1). In RNB mice, granule-cell inputs to Purkinje cells can be thus selectively blocked and reversibly recovered by administration and omission, respectively, of doxycycline (DOX). As a consequence, when granule-cell transmission is blocked, the optokinetic signal is not transmitted to Purkinje cells but is still conveyed to the VN via the direct mossy-fiber pathway. This reversible blocking technique thus allows us to delineate distinct roles of Purkinje cells and VN in the memory trace of OKR learning. The present study demonstrates that VN are critical for acquisition and storage of long-term OKR memory in intimate association with the Purkinje-cell circuit for expression of OKR adaptation.

Results

Effects of Reversible Blockade of Granule-Cell Transmission on OKR Adaptation.

OKR was evoked by rotating a striped screen with sinusoidal oscillation (0.2 Hz, 18 deg/s) and recorded with a computer-assisted infrared video camera (Fig. S2). Animals were trained by screen rotation for 1 h every day and kept in darkness after each training. The gain of eye movement was calculated as the ratio of amplitudes of eye velocity relative to the intensity of the stimulus. The basic OKR was determined by averaging the gains of the initial 12 cycles of eye velocity of untrained animals on day 1. The gain of short-term OKR in the trained animals was assessed by averaging the gains of the first 12 cycles (1 min) of eye movements every 6 min during 60-min training in each day whereas long-term OKR was determined by pursuing changes in the average gains of the initial 12 cycles of 1-h training from day 1 to day 5.

The WT and RNB mice were pretreated with DOX for 14 d. The naïve WT and RNB mice showed comparable levels of basic OKR on day 1 (WT vs. RNB, 0.34 ± 0.03 vs. 0.29 ± 0.03, P = 0.63, Student t test, n = 7 each) (Fig. 1A), indicating that blockade of granule-cell transmission had no effect on the basic OKR. The WT mice then explicitly displayed both short-term OKR during 1-h training throughout 5 d (for detailed statistical analysis, see Table S1; n = 7) and long-term OKR by the successive 5-d training [F(1,6) = 8.25, P < 0.0001, one-way repeated-measures ANOVA, n = 7] (Fig. 1A). Thus, the gain of short-term OKR of WT mice was almost saturated on day 5. In marked contrast to the WT mice, the DOX-treated RNB mice (On-RNB) failed to evoke either short-term OKR or long-term OKR (Fig. 1A) (for detailed statistical analysis, see Table S1). These results indicate that transmission from granule cells to Purkinje cells is indispensable for evoking both short-term and long-term adaptive OKRs.

Fig. 1.

Short-term and long-term OKR adaptation of WT and RNB mice under the DOX-On or DOX-Off condition. DOX was administered 2 wk before behavioral analysis, and the administration was continued during 5-d OKR training (DOX-On). DOX was then omitted after behavioral analysis from day 6 (DOX-Off). (A) The gain of short-term OKR was determined by averaging the gains of the first 12 cycles (1 min) of eye movements every 6 min during 60 min from day 1 to day 5 and on days 19 and 20 (n = 7), and the mean and SEM (a vertical solid line) of the averaged gains are indicated every 6 min. The gain of long-term OKR was determined by averaging the gains of the first 12 cycles of eye movements each day and is indicated by the dashed line. Animals were kept in darkness between tasks as indicated by the black bars. In the Inset, representative traces of eye movements of WT (black) and RNB mice (red) are indicated at each time point. (B and C) OKR dynamics of WT (B) and RNB (C) were determined by rotating a screen with a maximum velocity of 18°/s at 0.1, 0.2, 0.4, and 0.8 Hz before and after OKR training. Ten cycles of eye movements were used to calculate gains of OKR dynamics.

We then addressed whether learning processes of OKR could change during blockade of granule-cell transmission to Purkinje cells. The DOX-treated WT and RNB mice were trained by 1-h OKR stimulation from day 1 to day 5 and then kept in darkness from day 6 to day 18 in the absence of DOX treatment. These animals were then tested by measuring OKR by 1-h training on days 19 and 20. The WT mice retained their OKR memory induced by 5-d training and showed high levels of long-term OKR comparable with those evoked on day 5 (day 5 vs. day 19 and day 20, 0.63 ± 0.05 vs. 0.69 ± 0.04 and 0.70 ± 0.04, respectively; for day 19, F(1,6) = 0.42, P = 0.842; one-way repeated-measures ANOVA, n = 7) (Fig. 1A). Remarkably, the DOX-withdrawn RNB mice (On-Off RNB) showed comparable, high levels of long-term OKR on days 19 and 20 [day 19, 0.74 ± 0.03; day 20, 0.74 ± 0.04; WT vs. RNB on day 19, F(1,60) = 0.0012, P = 0.973, two-way repeated-measures ANOVA, genotype effect, n = 7 each]. The gains of the long-term OKR of the On-Off RNB mice were thus markedly different between day 5 and days 19 and 20 [day 5 vs. day 19 and 20, 0.31 ± 0.05 vs. 0.74 ± 0.03 and 0.74 ± 0.04; day 5 vs. day 19, F(1,6) = 17.91, P < 0.001, one-way repeated-measures ANOVA, n = 7]. The results indicate that OKR learning is acquired and stored despite the absence of granule-cell transmission to Purkinje cells.

The gain of long-term OKR is known to increase independently of the frequency of OKR stimulation during consecutive training (3, 20). We addressed whether OKR training with 0.2 Hz changed the dynamic property of OKR (OKR dynamics) by measuring gains of OKR at different frequencies of OKR stimulation (Fig. 1 B and C). The basic OKRs of DOX-treated naïve WT and RNB mice showed a virtually identical pattern of OKR dynamics on day 1, with a peak gain at 0.1 Hz and decreasing gain as the frequency of the stimulus was increased (pre-day 1 in Fig. 1 B and C). Upon 5-d training with stimulation of 0.2 Hz, the WT mice showed an overall increase in the gain of OKR in response to different frequencies of the stimulus and a shift of OKR dynamics to a higher range of frequencies of stimuli (Fig. 1B) (for statistical analysis, see Table S2). Consistent with the retention of long-term OKR following a 2-wk interval of OKR training, the change in OKR dynamics was maintained in WT mice on day 19 (Fig. 1B and Table S2). Markedly, however, the DOX-On RNB mice showed no such change in OKR dynamics with 5-d training and kept the OKR dynamic pattern identical to that of the pretrained naïve mice (Fig. 1C and Table S2). Importantly, when granule-cell transmission was recovered by omission of DOX, the 5 d-trained DOX–On-Off RNB mice showed the OKR dynamics characteristic of long-term OKR and exhibited no difference in the dynamic pattern from that of the WT mice (Fig. 1C and Table S2). These results further support the view that the OKR adaptation is induced and stored without granule-cell transmission to Purkinje cells during time points of OKR training.

Effects of Blockade of Granule-Cell Transmission on Acquired OKR Adaptation.

We next addressed whether acquired adaptive OKR could be maintained or abrogated by blockade of granule-cell transmission to Purkinje cells. As expected, the WT and RNB mice, when DOX was omitted, showed both short-term and long-term OKRs during 5-d training (Fig. 2A) (for statistical analysis, see Table S3). Then, these mice were treated with DOX from day 6 to day 20 and subjected to 1-h OKR stimulation on days 19 and 20. Strikingly, DOX–Off-On RNB mice retained their OKR memory and exhibited long-term OKR comparable with that of WT mice [WT vs. RNB, 0.62 ± 0.07 vs. 0.78 ± 0.03 on day 19, 0.73 ± 0.15 vs. 0.77 ± 0.02 on day 20; WT vs. RNB, F(1,6) = 0.5488, P = 0.595, two-way repeated-measures ANOVA, genotype effect, n = 7 each] (Fig. 2A) (for further statistical analysis, see Table S3). Furthermore, the DOX–Off-On RNB mice showed OKR dynamics characteristic of long-term OKR on day 19 (Fig. 2 B and C and Table S2). Earlier studies indicated that DOX treatment of RNB mice for 2 wk is sufficient to block granule-cell transmission to Purkinje cells (17, 18). Furthermore, when RNB mice were treated with DOX not only during 5-d training but also after training up to day 20, these DOX-treated RNB mice showed no long-term OKR by OKR stimulation on days 19 and 20. These results indicate that, when OKR memory is once acquired and expressed, granule-cell transmission to Purkinje cells is no longer required for long-term OKR memory.

Fig. 2.

Long-term OKR memory is retained even when granule-cell transmission is blocked. WT and RNB mice were free from DOX up to day 5, and DOX was then administered from day 6 to day 20. Short-term and long-term OKRs (A) and OKR dynamics of WT (B) and RNB mice (C) were determined as described in the legend of Fig. 1.

Induction of Neural Plasticity for OKR Adaptation.

It is plausible that neural plasticity is induced in the VN by repeated training and changes the output that reflects adaptive eye movements (3, 5, 21). Because Purkinje cells are the sole output neurons in the cerebellar cortex (2), suprathreshold electrical stimulation of the flocculus could have similar impacts as the synchronized firings of Purkinje cells on the downstream VN (Fig. 3A). We first examined possible adaptive changes occurring at the VN after OKR training by quantifying eye movements elicited by electrical stimulation of the flocculus (Fig. 3), which mimics Purkinje-cell firings (10). An electrode was embedded and maintained for 5 d in the right flocculus (H zone) responsible for horizontal eye movements toward an ear (Fig. 3A). Suprathreshold electrical stimulation was then delivered to the cerebellar cortex at the different stages of OKR training. The flocculus and paraflocculus are located adjacent to each other, but eye movements induced by electrical stimulation were previously found to be different between these two loci (22). When the position of an embedded electrode was histologically examined after electrical experiments, unilateral stimulation of the flocculus confirmed induced horizontal eye movements toward the ear (Fig. 3B) whereas the stimulation of the paraflocculus evoked torsional eye movements. Upon electrical stimulation of the flocculus, WT mice showed no change in gain of eye movements before or after training on day 1 (Fig. 3 C and D) [pre-day 1 vs. post-day 1, F(1,30) = 0.06, P = 0.81, two-way repeated-measures ANOVA, training effect, n = 4]. Similarly, no alteration of eye movements was observed on day 5 before or after successive 5-d training (Fig. 3D) [pre-day 5 vs. post-day 5, F(1,30) = 1.79, P = 0.19, two-way repeated-measures ANOVA, training effect, n = 4]. Remarkably, when eye movements were compared in the same mouse between day 1 training and day 5 training, significantly larger eye movements were evoked on day 5 than on day 1 when long-term OKR adaptation was established (Fig. 3D) [pre-day 1 vs. post-day 5, F(1,30) = 30.17, P < 0.001, two-way repeated-measures ANOVA, training effect, n = 4]. The control experiment confirmed that embedding an electrode and keeping it in place for several days per se had no enhancing effect on eye movement unless the animals were successively trained with OKR stimulation. These results indicate that neural plasticity in the VN occurs during the emergence of long-term OKR by 5-d training but not during induction of short-term OKR by 1-d training. Remarkably, the DOX-On RNB mice exhibited eye movements identical to those of electrically stimulated eye movements of the WT mice in both short-term and long-term OKR adaptation: no increase in electrically stimulated eye movements was observed before and after 1-d training on either day 1 or day 5, but prominent facilitation of eye movements was noted after OKR training from day 1 to day 5 (Fig. 3E) [pre-day 5 vs. pre-day 1, F(1,40) = 35.37, P < 0.001, two-way repeated-measures ANOVA, training effect, n = 5], even though long-term OKR never occurred in the DOX-On RNB mice. These results indicate that mice with blocked granule-cell transmission still retain the ability to undergo neural plasticity at the VN by repeated OKR stimulation although this plasticity failed to express adaptive OKR.

Fig. 3.

Enhancement of eye movements of trained animals by flocculus stimulation. (A) Scheme of blockade of granule-cell transmission (red cross) and electrical stimulation of the flocculus (marked by yellow). (B) Ten traces of eye movements of WT mice were superimposed when the right flocculus was electrically stimulated with currents of 20–60 μA. (C) WT and RNB mice were stimulated at the time points indicated by the colored arrows: black and blue, respectively, 5 min before and 5 min after OKR training on day 1; red and green, respectively, 5 min before and 5 min after OKR training on day 5. The right flocculus was stimulated with a pulse of 100 Hz for 0.5 s. (D and E) Horizontal eye positions of WT mice (D) and DOX-On RNB mice (E) at the end of electrical stimulation are plotted against different current amplitudes. Each color indicates time points of electrical stimulation shown in C. Error bars indicate SEM. Insets above the graphs indicate representative traces of horizontal eye movements at pre-day 1 and pre-day 5 in response to electrical stimulation (blue bars) with a normalized current amplitude of 3.

Induction of OKR Adaptation by In-Phase Electrical Stimulation of the Cerebellar Cortex.

Because repeated OKR training was found to be capable of inducing neural plasticity in the VN of DOX-On RNB mice, we addressed the possibility that adaptive OKR could be expressed by appropriate signals from the cerebellar cortex. To examine this possibility, we subjected DOX-On RNB mice to 1-h training for 5 d. We then applied electrical stimulation to the flocculus of the DOX-On RNB mice together with rotation of a screen for 15 min after OKR training on day 5 (Fig. 4A). The pulse rate of electrical stimulation was sinusoidally modulated by increasing the modulatory depth every 1 min (baseline ± amplitude, 50 ± 20 pulses per s), and the frequency of the electrical stimulation was increased (in-phase) or decreased (out-of-phase) when the screen was rotated (0.2 Hz) toward the ear for 15 min (Fig. 4A). Then, 1 min after this simultaneous stimulation, we measured the gain of OKR to screen rotation in the absence of electrical stimulation (Fig. 4B). The OKR gain increased when the stimulation was applied in-phase with screen rotation [before vs. after in-phase electrical assistance on day 5, F(1,48) = 30.49, P < 0.001, two-way ANOVA, assistance effect, n = 6] (Fig. 4B). In contrast, the gain remained unchanged or suppressed when the stimulus was applied out-of-phase (Fig. 4B). The assistance by in-phase but not by out-of-phase electrical stimulation also shifted the OKR dynamic pattern to that of long-term OKR (Fig. 4C). The same set of experiments was performed by electrical stimulation of the paraflocculus, which is not related to induction of horizontal OKR (23) (Fig. 4 B and D). The assistance by in-phase electrical stimulation to the paraflocculus neither enhanced the gain of OKR nor changed OKR dynamics of DOX-On RNB mice [gain of OKR before vs. after assistance by electrical stimulation on day 5, F(1,40) = 0.11, P = 0.73, two-way ANOVA, assistance effect, n = 5]. These results indicate that long-term OKR adaptation is stored in the VN of DOX-On RNB mice and could be expressed by the appropriate signals from the cerebellar cortex.

Fig. 4.

OKR of DOX-On RNB mice after assistance by electrical stimulation. (A) Scheme of experimental procedures to examine effects of electrical stimulation on OKR. After OKR training of DOX-On RNB mice for 5 d, the animals received an additional screen rotation for 15 min together with in-phase (red) or out-of-phase (blue) sinusoidal electrical stimulation at either the ipsilateral flocculus or paraflocculus. Gains of OKR were determined from day 1 to day 5. On day 5, OKR was further measured in the absence of electrical stimulation 15 min after simultaneous stimulation with screen rotation and electrical stimulation. (B) Gains of OKR were measured according to the experimental procedures described in A. Representative traces of OKRs after simultaneous in-phase or out-of-phase electrical stimulation are indicated in the Inset. (C and D) OKR dynamics were determined at the indicated time points described in A. In C, OKR dynamics are also shown for DOX-On mice treated with out-of-phase electrical stimulation. Error bars indicate SEM.

Discussion

The present investigation revealed that selective blockade of granule-cell transmission to Purkinje cells abolished both short-term and long-term OKR adaptations in cerebellum-dependent motor learning. However, when granule-cell transmission recovered, normal levels of long-term OKR were immediately evoked in the pretrained RNB mice from the beginning of the second OKR stimulation. Importantly, the adaptive OKR-negative On-RNB mice showed a dynamic pattern of basic OKR characteristic of untrained mice even after 5-d training, but the OKR dynamics shifted into a pattern of long-term OKR when granule-cell transmission was recovered. These results demonstrate that input transmission of granule cells to Purkinje cells is indispensable for expression of short-term and long-term OKRs but is not required for formation and storage of adaptive OKR.

Electrophysiological examination indicated that 5-d OKR training, but not 1-h OKR training, enhanced eye movement elicited by electrical stimulation of the cerebellar flocculus. This finding is consistent with the study by Shutoh et al. (3), who reported that long-term, but not short-term, OKR training enhanced neural excitability of the medial VN. Importantly, our study further indicated that 5-d OKR training potentiated electrically stimulated eye movement of On-RNB mice, even though these mice never showed long-term OKR. These findings support the notion that VN is a critical circuit that is responsible for the formation and storage of OKR adaptation.

The optokinetic signals that trigger adaptive OKR are transmitted to the VN directly through mossy fibers from the NRTP and indirectly to Purkinje cells through the mossy fiber–granule cell pathway (6, 7). The retinal slip signals are mediated by climbing fibers from the inferior olive and transmitted to Purkinje cells and then to the VN (6, 7). The optokinetic and retinal slip signals are thus conveyed to both Purkinje cells and the VN. Conjunctive stimulation of parallel fibers and climbing fibers induces long-term depression (LTD) at the parallel fiber–Purkinje cell synapses and suppresses the tonic inhibition of Purkinje cells onto the VN (2, 4). It has been proposed that LTD causes increased sensory input transmission and induces adaptive response for downstream motor movement (1, 2). Shutoh et al. (3) reported that reversible flocculus shutdown with lidocaine extinguishes the memory trace of short-term OKR but not that of once-acquired long-term OKR and proposed that short-term OKR memory is formed initially within the cerebellar cortex and then transferred to the VN to be consolidated to a long-term memory (3, 5). Injection of lidocaine into the flocculus, however, inhibits transmission from both granule cells and climbing fibers to Purkinje cells. In fact, they observed that chronic lesioning of both the cerebellar flocculus and inferior olive abolished OKR adaptation (3). In contrast to these pharmacological studies, the present study selectively blocked granule-cell transmission, and this blockade has been shown to produce no obvious effect on climbing-fiber transmission to Purkinje cells (18). Thus, direct inputs from mossy fibers and those from climbing fiber–Purkinje cells onto the VN remained intact in On-RNB mice and would induce neural plasticity in the VN circuit. This neural plasticity could then allow prompt induction of adaptive OKRs, once the granule-cell transmission had been recovered. Thus, our study supports the view that convergent information to induce neural plasticity in the VN is critical for acquisition and storage of long-term OKR adaptation with conjunctive action of Purkinje cells for OKR expression.

When the long-term OKR was once established, blockade of granule-cell transmission did not abrogate the acquired adaptive long-term OKR. Furthermore, the OKR gain increased, by coupling of electrical excitation of the flocculus with optokinetic stimulation under the transmission blockade. In addition, injection of lidocaine into the flocculus has been shown not to inhibit the established long-term OKR (3). These results indicated that, once adaptive OKR is established, granule-cell transmission to Purkinje cells is no longer required for long-term OKR. These findings suggest that the neural plasticity induced in the VN by repeated OKR training increases OKR gain by enhancing the involvement of the closed-loop circuit responsible for basal OKR. Interestingly, a previous study using the RNB technique indicated that blockade of granule-cell transmission prevents not only expression of adaptive eyeblink response during conditioning of näive mice but also that of once-acquired eyeblink memory (18). The memory trace of OKR adaptation thus differs from that of conditioned eyeblink motor learning. Notably, eyeblink motor learning requires the well-matched coincidence of conditioned stimuli (tone or light) and unconditioned stimuli (air puff or an electric shock), and this coincidence is supposed to rarely occur in an animal’s life. Thus, the tight linkage between conditioned and unconditioned stimuli at the cerebellar cortex is necessary for this form of cerebellum-dependent motor learning (24, 25). In contrast, the OKR adaptation is induced by enhancing naturally occurring eye movement to adjust it to motion of the visual field. Accumulated evidence indicates that gene knockout of several cerebellar cellular components differentially affects different prototypes of cerebellum-dependent motor learning (4, 21, 26, 27). The different forms of cerebellum-dependent motor learning could thus be regulated by the involvement of different cellular components, as well as by distinct forms of neural plasticity in different cerebellar circuits.

Materials and Methods

Animals.

RNB mice and their WT littermates, obtained by mating TeNT and Tet transgenic mice (18), were used for all experiments. DOX was administered in pellets containing 6 mg/g DOX and in drinking water containing 2 mg/mL DOX and 10% (wt/vol) sucrose. After electrophysiological and behavioral analyses, the genotypes of mice were determined by using tail biopsies and PCR analysis. The analysis was performed by operators who were blind to the genotype of the mice. All animal experiments were approved by the Animal Committee under the guidelines of Osaka Bioscience Institute.

Eye-Movement Experiments.

Head holding and surgery of anesthetized animals were performed as described previously (20). The apparatus and measurement of OKR are described in Fig. S2. OKR was tested by sinusoidal oscillation of a striped screen (striped size 5°) by 18°/s (maximum screen velocity) at 0.2 Hz, unless otherwise stated. Over 10 cycles of the evoked eye movements, free from artifacts due to blinks and saccades, were selected for averaging. Correction programs (Matlab) were used to delete artifacts. The gain of eye movement was defined as the ratio of the peak-to-peak amplitude of eye movements to that of the screen oscillation. Short-term OKR was examined by exposing a mouse to 1-h sustained screen oscillations at 0.2 Hz (maximum screen velocity, 18°/s). Long-term OKR was examined by successively giving 1–h trainings over 5 d or for 2 d after a 2-wk interval, during which the trained animals were kept in darkness and free from OKR training. During short-term and long-term trainings, the mice were kept in the dark except for training sessions. OKR dynamics was determined by 10 cycles of screen rotation at 0.1, 0.2, 0.4, and 0.8 Hz (maximum velocity, 18°/s).

Electrophysiology.

Under ketamine (80 mg/kg) and xylazine (15 mg/kg) anesthesia, a head holder was attached to the skull, and a small hole (<1 mm in diameter) was drilled in the skull above the right flocculus (3 mm lateral and 6 mm caudal from bregma) and covered with 2% agarose gel and dental adhesive. After the surgery, the mice were allowed to recover for at least 3 d. Horizontal eye movements were induced by electrical stimulation of the flocculus with current amplitudes of 20–60 μA for 0.5 s (duration, 1 ms; interval, 9 ms; 50 pulses) through an insulated stainless wire (100 μm). Eye movements were measured in darkness after dropping pilocarpine hydrochloride (Santen) to decrease and stabilize pupil size. Amplitudes of electrical stimulation were normalized by assigning amplitudes that induced maximal eye movements as a value of 5. Then, eye movements were measured by applying electrical stimulation with normalized amplitudes from 1 to 4. The effect of electrical stimulation of the flocculus or the paraflocculus on OKR adaptation was examined as follows: An electrode was embedded and maintained in either the flocculus or paraflocculus of On-RNB mice. These mice were subjected to 1-h OKR training for 5 d. After OKR training on day 5, they received an additional 15-min screen rotation at 0.2 Hz (maximum velocity, 18°/s) together with sinusoidally modulated electrical stimulation at 0.2 Hz (10). This modulation was made by increasing modulation depths (peak to trough) at 10 pulses per s every 1 min from the baseline ± amplitude of 50 ± 20 pulses per s up to the peak of 120 pulses per s (19). The phase of electrical pulse-rate modulation relative to the screen velocity was adjusted by increasing or decreasing the frequency of the electrical stimulation (in-phase or out-of-phase, respectively), when the screen was rotated toward the ear. Immediately after this manipulation, the gain of OKR was measured in the absence of electrical stimulation. After behavioral analysis, anesthetized mice were perfused with PBS followed by 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.3), and the position of the electrode in either the flocculus or paraflocculus was determined.

Statistical Analysis.

Statistical analysis was performed by using Student t test, one-way or two-way repeated-measures ANOVA, and two-way ANOVA. The data of statistical analyses are presented in Tables S1–S3.