Cerebrocortical activation following unilateral labyrinthectomy in mice characterized by whole-brain clearing: implications for sensory reweighting

Posture and gait are maintained by sensory inputs from the vestibular, visual, and somatosensory systems and motor outputs. Upon vestibular damage, the visual and/or somatosensory systems functionally substitute by cortical mechanisms called “sensory reweighting”. We investigated the cerebrocortical mechanisms underlying sensory reweighting after unilateral labyrinthectomy (UL) in mice. Arc-dVenus transgenic mice, in which the gene encoding the fluorescent protein dVenus is transcribed under the control of the promoter of the immediate early gene Arc, were used in combination with whole-brain three-dimensional (3D) imaging. Performance on the rotarod was measured as a behavioral correlate of sensory reweighting. Following left UL, all mice showed the head roll-tilt until UL10, indicating the vestibular periphery damage. The rotarod performance worsened in the UL mice from UL1 to UL3, which rapidly recovered. Whole-brain 3D imaging revealed that the number of activated neurons in S1, but not in V1, in UL7 was higher than that in sham-treated mice. At UL7, medial prefrontal cortex (mPFC) and agranular insular cortex (AIC) activation was also observed. Therefore, sensory reweighting to the somatosensory system could compensate for vestibular dysfunction following UL; further, mPFC and AIC contribute to the integration of sensory and motor functions to restore balance.

Following unilateral damage to the vestibular inner ear, oculomotor and postural symptoms such as nystagmus and postural instability occur, but gradually recover over time in a process known as vestibular compensation 1-3 . Brain plasticity is thought to be responsible for vestibular compensation, because functional recovery occurs without regeneration of the vestibular periphery. Vestibular compensation includes changes in the patterns of floccular inhibition of the vestibular nucleus complex (VNC) and the recovery of spontaneous firing and responsiveness in ipsilesional VNC neurons 2,4,5 . In addition to vestibular compensation, appropriate postural control after vestibular damage also depends on dynamic changes in multimodal sensorimotor control systems known as "sensory reweighting" 6 . Therefore, when studying the recovery after vestibular damage, it is important to consider not only vestibular compensation, in which the cerebellum-brainstem pathways are involved, but also the cerebrocortical mechanisms of sensory reweighting.
Immediately after acute vestibular insults, postural control shifts to visual and/or somatosensory dependence, which can substitute impaired vestibular function in humans 7,8 . Functional brain imaging studies have www.nature.com/scientificreports/ level as that in the sham-treated group by day 4 after UL, which was faster than the recovery observed on the head roll-tilt (Fig. 1C).
Whole-brain images. Figure 2 shows whole-brain images of sham-treated (Sham2 and Sham7) and UL individuals (UL2 and UL7) on day 2 and day 7 after surgery. Neural activation of the whole brain can be comprehensively visualized using a panoramic view, from which 3D images can also be obtained (Supplementary Movie 1). Hippocampal neurons were activated non-specifically in both groups.
Visual cortex (V1) and somatosensory cortex (S1). On day 2 after surgery, there were no significant bilateral differences in the number of activated neurons in V1 and S1 between the UL and the sham-treated group. On day 7 after surgery, the number of activated neurons in the bilateral S1 was significantly higher in the UL than in the sham-treated mice, whereas no significant differences between these groups were found in V1 (Figs. 3,4,5,6).

Posterior granular insular cortex (pGIC) and auditory cortex (A1).
There was no significant difference in the number of activated neurons in the pGIC between the UL and the sham-treated group at both day 2 and day 7 after surgery (Figs. 7, 8A). On day 2 after surgery, there were no significant differences in the number of activated neurons in A1 between the UL and the sham-treated group. However, on day 7 after surgery, the number of activated neurons in the bilateral A1 was significantly higher in the UL group than in the shamtreated group (Figs. 7, 8B).

Medial prefrontal cortex (mPFC) and agranular insular cortex (AIC). Whole-brain analysis under
comprehensive and panoramic views demonstrated that activated neurons were found bilaterally in the medialfrontal cortex ( Fig. 9A; arrowhead) and the parieto-temporal region ( Fig. 9A; arrows) on UL2. A survey of arbitrary cross-sections ( Fig. 9B-E) revealed that the activated neurons were located in the mPFC (Fig. 9C)  www.nature.com/scientificreports/ and the caudate putamen (Fig. 9E). On UL7, activated neurons could also be found bilaterally in the anteriortemporal region ( Fig. 10A; arrowheads) and the temporal region ( Fig. 10A; arrows). A survey of arbitrary cross sections from UL7 ( Fig. 10B-E) revealed that the activated neurons were located in the agranular insular cortex (AIC) (Fig. 10C) and in S1 (Fig. 10D), which is a target area for sensory reweighting. There were a few activated neurons in these areas in individuals Sham2 and Sham7 (data not shown). Regarding the cell count of the four subregions in the mPFC (Figs. 11,12), the number of activated neurons was significantly higher in the UL group than in the sham-treated group on day 2 after surgery in the medial precentral cortex (mPC), dorsal anterior  www.nature.com/scientificreports/ cingulate cortex (dAC), and prelimbic cortex (PL). This number gradually decreased, but it was still significantly higher in the UL group on day 7 (Fig. 12). A few activated neurons were observed in the infralimbic cortex (IL).
In the cell count of the three subregions in AIC (Figs. 11,13), there were no significant differences between the groups on day 2, but the number of activated neurons in all the subregions was higher in mice from the UL group on day 7 compared to sham-treated mice.  There were no significant differences in V1 between the UL and the sham-treated group. (B) Activated neurons in the somatosensory cortex (S1). There were significantly more activated neurons in the UL group than in the shamtreated group on day 7 after surgery. Error bars indicate the standard error of the mean.

Discussion
All the mice that underwent chemical UL showed ipsilesional head roll-tilt and impaired rotarod performance. Although the present study did not assess direct vestibular damages, these behavioral results were similar to those from a previous study using the same concentration of arsanilate in which peripheral vestibular organ damages were confirmed histologically 16 . The head roll-tilt peaked on day 2 and lasted up to 10 days after the UL (Fig. 1B). Ito et al. reported that the head roll-tilt peaked on day 2 and recovered by day 4 after UL in mice that had been administered arsanilate using a post-auricular approach. The effects of intratympanic injection of arsanilate were persistent in this study compared to the administration of the same arsanilate dose using a post-auricular approach 16 . This difference may therefore be explained by the different routes of administration. Intratympanic injection without skin incision may cause the arsanilate to be retained for longer in the tympanic cavity, resulting in stronger effects on the vestibular periphery. The rotarod test has been widely used to investigate vestibular imbalance after labyrinthectomy 17 and motor deficits due to brain injury 18,19 . In our results, rotarod performance was disrupted from day 1 to day 3, and had recovered by day 4 after UL (Fig. 1C). The recovery observed on the rotarod performance was much faster than that for the head roll-tilt. Since the head roll-tilt is mainly controlled by otolith inputs, the recovery of the head roll-tilt may be directly proportional to vestibular compensation 16 . On the other hand, the rotarod performance seems to be related not only to vestibular function, but also to motor control and coordination by other sensory modalities, including sensory reweighting. Taken together, these results suggest that vestibular compensation had not yet been completed between days 2 to 10 after UL, and that sensory reweighting occurred on day 4 after UL. Our investigation of neural activation in the cerebral cortices revealed no significant differences in V1 and S1 between the UL and the sham-treated group on day 2, but significantly higher neural activity at S1 in the UL group on day 7. The results of the neural activity at S1 in UL animals are in agreement with the results of the rotarod behavioral tests.  There were no significant differences in A1 between the UL and the sham-treated group on day 2 after surgery; however, there were significantly more activated neurons in A1 of the UL group than in the sham-treated group on day 7 after surgery. Error bars indicate the standard error of the mean. www.nature.com/scientificreports/ Our study showed that on day 7 after UL, the number of activated neurons in both the right and left somatosensory cortices (S1) was significantly higher in UL than in sham-treated animals, while no significant differences were found on day 2 ( Fig. 6B). Neural activation in S1 was not only noticed when the region was pre-selected as a potential area for sensory reweighting (Figs. 3B, 5), but also when a series of arbitrary crosssections where scanned for signs of neural activation after UL (Fig. 10D). According to the rotarod results, sensory reweighting started not earlier, but later than day 4 after UL, and it was already present on day 7 after UL ( Fig. 1C). Taking these results together, activation of the somatosensory cortex (S1) (Figs. 5, 6B) may play a role in sensory reweighting after UL in mice. In contrast, there were no significant differences in the number of activated neurons in either the right or left visual cortex (V1) between the UL and sham-treated groups on day 2 or on day 7 after UL (Figs. 4, 6A), demonstrating that the visual cortex (V1) was not activated after UL. This result suggests that sensory reweighting after vestibular loss in mice may not depend on visual input. According to motor function tests, including rotarod tests performed in normal and visually impaired mice with and without somatosensory inputs from the whiskers, the somatosensory inputs from whiskers are more essential for maintaining motor function in mice 20 . These results strongly support our findings that somatosensory (whisker) inputs play a key role in sensory reweighting after UL in mice, unlike the mechanism of sensory reweighting in humans. Although there is no evidence of direct neural connections from the peripheral vestibule or vestibular nuclei to S1 and V1, multisensory cortical areas in the mouse homologous to areas 2v, 3a, and PIVC in monkeys would convey vestibular, somatosensory, and visual information to S1 and V1 21 .
It has been reported for rodents that the parieto-temporal (PT) region receives most of the projections from the vestibular nuclei 22 . However, there have been no reports on the vestibular cortex in mice. In this study, we assumed that the vestibular cortex in mice might be located in the pGIC, which is anatomically homologous to the vestibular cortex of other species (PIVC, ASSS, and PT). We set a 3D ROI for the pGIC as a potential vestibular cortex in mice. However, activated neurons were not found in the pGIC after the UL (Fig. 9A). It is suggested that the mouse vestibular cortex could not be identified in the pGIC using a vestibular deafferentation model, and that alternative research models, such as vestibular stimulation models, may be required to identify the vestibular cortex in mice.
Our study showed a significant increase in the number of activated neurons in the auditory cortex (A1) in mice on day 7 after UL, but not on day 2 (Fig. 9B). Because both UL and sham-treated animals were kept in the same conditions in terms of auditory input 2 days before decapitation, this difference was not due to reactivation www.nature.com/scientificreports/ by cochlear inputs from the healthy side. It may be considered that this is not the result of sensory reweighting but of disinhibition in A1, induced by the loss of peripheral sensory inputs due to cochlear damage by the UL.
In the survey on arbitrary cross sections from whole-brain imaging, the activated neurons were found in the mPFC on day 2 and day 7 after UL (Fig. 12). In rodents, the mPFC is divided into four subdivisions from dorsal to ventral: mPC, dAC, PL, and IL [23][24][25] (Fig. 11A). The mPC and dAC are known to be linked to various  www.nature.com/scientificreports/ motor behaviors, while PL and IL are associated with diverse emotional, cognitive, and mnemonic processes 23,26 . Stimulation of the mPC and dAC has been reported to induce movement of the eye, vibrissa, head, and hind limbs [27][28][29][30] . Therefore, the mPC and dAC in rodents are homologous to the frontal eye field, which acts as a supplementary motor and premotor cortex in primates [31][32][33] . The mPC overlaps with the vibrissa motor cortex (vM1) and receives numerous cortical inputs from the visual, somatosensory, auditory, parietal, retrosplenial, and orbital areas [34][35][36][37] . The mPC also projects to the superior colliculus 38 and subcortical nuclei, involving oculomotor control 39,40 . Therefore, it is thought that the mPC plays a role in receiving multiple sensory inputs, in processing and in transforming them into motor functions 41 . In contrast to the motor-associated functions of the mPC and dAC, PL and IL are linked to the limbic system 25,31,42 and profoundly influence visceral and autonomic activities, such as gastrointestinal motility and blood pressure control 43,44 . In our study, activated neurons were found in the mPC, dAC, and PL of the mPFC subdivision after UL (Fig. 12). Based on these neural connectivity and functions of the mPFC, the activation of the mPC and dAC at the early stage after UL may be related to changes in oculomotor control and integrative motor activity. PL activation may also be related to visceral and autonomic changes after UL. The basal ganglia (including the caudate putamen) control muscle tonus for posture, locomotion, and cognition through the basal ganglia-cortical and the basal ganglia-brainstem loops 45,46 . In our study, the activated neurons in the caudate putamen were observed in UL2 but decreased in UL7 (Figs. 9E, 10E). The results may be driven by the activation-deactivation of direct motor and postural control systems rather than sensory reweighting mechanisms. The insular cortex (IC) comprises three different subareas, which differ in their cytoarchitecture of layer 4: the agranular (AIC), dysgranular (DIC), and granular (GIC) cortices (from ventral to dorsal). In our results, activated neurons were found in the AIC on day 7, but not on day 2 after UL. The IC has been suggested to function as a hub linking large-scale brain systems. It receives multisensory inputs from the outside environment, such as auditory, somatosensory, visual, olfactory, and gustatory information, as well as from inside the body, such as blood pressure, heartbeat, oxygenation, and input from the digestive system 47    www.nature.com/scientificreports/ cortices and the amygdala. In contrast, the anterior subregion shows more connections to the motor cortex, prefrontal cortex, striatum, and sensory cortex 48 . In our results, activated neurons were found in all subregions (middle, posterior, and anterior) of the AIC on day 7 after UL. Neuronal activation of the somatosensory cortex (S1) (Figs. 5, 6B) may substitute for the loss of vestibular inputs following UL (that is, in sensory reweighting) leading to a recovery in the rotarod performance (Fig. 1C). Activation of the AIC (Fig. 13) may also play a role in sensory reweighting by integrating somatosensory and motor functions. It has been reported that visual and somatosensory information are integrated at the early levels of the central nervous system such as the vestibular nuclei [49][50][51][52] , suggesting that sensory reweighting might occur at the brainstem level. Meanwhile, studies in humans have demonstrated changes of neural activities in the visual, somatosensory, and insular cortex after vestibular damage 9 , which is in partial agreement with our results: Only the somatosensory cortex, but not the visual cortex, was activated after UL, probably because rodents depend more on the sensory input from the whiskers than on visual inputs. Since how higher cortices such as V1, S1, mPFC, and IC influence the vestibular system is unknown, further research is needed.

Materials and methods
This study was carried out in compliance with the ARRIVE guidelines, and the protocols were approved by the Institutional Animal Care and Use Committee of Niigata University (IACUC; approval codes #SA00673 and #SD00787). All experiments were performed in accordance with the guidelines and regulations provided by IACUC.
Animals. Twenty C57BL/6J (CLEA Japan, Tokyo, Japan) and 16 Arc-dVenus male mice 53 , 8-10 weeks of age at the time of the experiments, were used for behavioral studies and whole-brain imaging studies, respectively. Because behavioral measurement could have affected the neuronal activities and CUBIC data, naive mice were used in behavioral studies. Arc-dVenus mice are transgenic heterozygotes mice that express fluorescent destabilized Venus protein under the control of the Arc immediate-early gene. The mice were housed in plastic cages with a 12-h light/dark cycle and under temperature-controlled conditions. Food and water were provided ad libitum. Behavioral studies. Twenty C57BL/6J naïve mice (10 in each group, UL and sham-treated) were used for behavioral observation. All of the mice that received intratympanic arsanilate showed behavioral abnormalities such as spontaneous nystagmus (SN) directed to the contralesional side and head roll-tilt directed to the ipsilesional side, indicating that intratympanic injection of arsanilate induced damage to the ipsilateral vestibular periphery. We did not use SN as a behavioral correlate of recovery after UL because of technical difficulties in the quantitative measurements of SN. To measure behavioral changes after left-side UL, the experimental animals were subjected to the head roll-tilt and rotarod tests. These two tests were used to estimate vestibular compensation and sensory reweighting, respectively, occurring in response to UL.

Procedure of chemical unilateral labyrinthectomy.
Head roll-tilt. Head roll-tilt, defined as the angle of the sagittal plane from the vertical plane, has been previously used to evaluate vestibular imbalance 16 . We measured the head roll-tilt on photographs once a day and used it as an indicator of vestibular imbalance following UL (Fig. 1A).

Rotarod test.
The rotarod test has been widely applied to monitor motor performance in naïve mice 54,55 , in addition to changes in balance after labyrinthectomy 17 and brain injury 18,19 . Mice were placed on the rod at a slow rotational speed (4 rpm). The speed was then gradually increased to 40 rpm over 120 s. Each session lasted 180 s. The session ceased if the mouse fell from the rod, at which point the time to fall (TTF) was recorded. The experiment consisted of three trials per day and the best performance was recorded. These measurements were performed during 14 consecutive days following UL.
Whole-brain imaging study. Sixteen Arc-dVenus mice were used for 3D visualization using CUBIC analysis. To minimize the effects on other sensory modalities, including external visual and auditory inputs, all mice used for imaging studies were placed in a light-and sound-deprived room for two days before decapitation. Animals were decapitated on days 2 or 7 after UL or sham surgery (4 groups of 4 mice each).
Tissue clearing. Arc-dVenus mice were anesthetized with a mixture of medetomidine hydrochloride (0.75 mg/kg body weight [BW]), midazolam (4 mg/kg BW), and butorphanol tartrate (5 mg/kg BW) on day 2 or day 7 after UL. The mice were then perfused with phosphate buffered saline (PBS pH 7.4) and 4% paraformaldehyde in PBS through the left ventricle. The brains were removed, post-fixed overnight in 4% paraformaldehyde in PBS at 4 °C and stored in PBS. The brains were rendered transparent for imaging according to the updated CUBIC method 14,56 . For 4-5 days, the brains were immersed in CUBIC-L [10/10 wt% chemical cocktail of N-butyldiethanolamine (B0725, Tokyo Chemical Industry) and Triton X-100 ( To remove autofluorescence signals in the finalized dVenus images, autofluorescence images were subtracted from the dVenus images using Fiji software. Furthermore, considering that the brains swell in response to the clearing process and that the size of the swollen brain differed across the experimental groups, the brain size was calculated and normalized using the ratio of the size. Acquired brain images were overlapped using an automatic transformation algorithm 56 for quantitative comparison. Reconstitution and analysis of volume-rendered images were conducted using the 3D visualization and image processing software Imaris (version 8.1.2, Bitplane, Zürich, Switzerland). The dVenus signal was overlaid with the RedDot™2 signal, and the cortical layers were identified by the arrangement of neuronal nuclei. A whole-brain image for each experimental group was thus created, as shown in Fig. 2.
Since both cortices are located in the posterior part of the insular cortex and are surrounded by the somatosensory and auditory cortices, the pGIC was set as a possible cortical area corresponding to the human/primate PIVC and cat ASSS. Since UL could influence not only the vestibular organ but also the cochlea, the 3D ROIs were set on the pGIC as a possible "vestibular cortex" in mice, and the primary auditory cortex (A1) to investigate the effects of loss of sensory inputs from the inner ear. In addition, potentially significant areas where the number of activated neurons could change following UL were surveyed by arbitrary cross sections in Imaris. Box-shaped 3D ROIs were set up on these target areas for sensory reweighting (V1 and S1) (Fig. 3), sensory deprivation (pGIC and A1) (Fig. 6), and newly identified cortical areas (mPFC and AIC) (Fig. 7) according to the mouse brain atlas 57 . The number of activated neurons in the 3D ROIs was automatically counted after the signal threshold was set, and compared between the UL and sham-treated groups. The mPFC and the agranular insular cortex (AIC) were divided into subregions. Because the mPFC in rodents is anatomically and functionally divided into four subregions, namely, the medial precentral cortex (mPC), dorsal anterior cingulate cortex (dAC), prelimbic cortex (PL), and infralimbic cortex (IL), from dorsal to ventral [23][24][25] , the cell count in the mPFC was performed separately for these four subregions. Since the AIC is divided into three subregions from rostral to caudal (anterior [aAIC], middle [mAIC], and posterior [pAIC]) based on their connectivity with other cortical and subcortical regions 48 , the cell count in the AIC was performed separately for these three subregions.

Statistical analysis.
The results for the head roll-tilt and rotarod tests were analyzed with repeated measures two-way analysis of variance (ANOVA) followed by post-hoc Tukey`s tests to identify the postoperative days at which there were significant difference between the UL and sham-treated subjects. Differences in the total cell count were analyzed by the Mann-Whitney test using SPSS Statistics software (version 26.0 , IBM, Armonk, NY, USA). Statistical significance was set at P < 0.05.
Ethics approval. All experiments were approved by the ethics committee of animal experiments from Niigata University and were carried out in accordance with the approved guidelines.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. www.nature.com/scientificreports/