ER stress inhibitor attenuates hearing loss and hair cell death in Cdh23erl/erl mutant mice

Hearing loss is one of the most common sensory impairments in humans. Mouse mutant models helped us to better understand the mechanisms of hearing loss. Recently, we have discovered that the erlong (erl) mutation of the cadherin23 (Cdh23) gene leads to hearing loss due to hair cell apoptosis. In this study, we aimed to reveal the molecular pathways upstream to apoptosis in hair cells to exploit more effective therapeutics than an anti-apoptosis strategy. Our results suggest that endoplasmic reticulum (ER) stress is the earliest molecular event leading to the apoptosis of hair cells and hearing loss in erl mice. We also report that the ER stress inhibitor, Salubrinal (Sal), could delay the progression of hearing loss and preserve hair cells. Our results provide evidence that therapies targeting signaling pathways in ER stress development prevent hair cell apoptosis at an early stage and lead to better outcomes than those targeting downstream factors, such as tip-link degeneration and apoptosis.

Hearing loss is one of the most common sensory impairments in humans. Mouse mutant models helped us to better understand the mechanisms of hearing loss. Recently, we have discovered that the erlong (erl) mutation of the cadherin23 (Cdh23) gene leads to hearing loss due to hair cell apoptosis. In this study, we aimed to reveal the molecular pathways upstream to apoptosis in hair cells to exploit more effective therapeutics than an anti-apoptosis strategy. Our results suggest that endoplasmic reticulum (ER) stress is the earliest molecular event leading to the apoptosis of hair cells and hearing loss in erl mice. We also report that the ER stress inhibitor, Salubrinal (Sal), could delay the progression of hearing loss and preserve hair cells. Our results provide evidence that therapies targeting signaling pathways in ER stress development prevent hair cell apoptosis at an early stage and lead to better outcomes than those targeting downstream factors, such as tip-link degeneration and apoptosis. The cadherin 23 (CDH23) protein is localized in the upper part of the tip link in hair cells, where it functions as a key component. [1][2][3] In humans, Cdh23 mutations cause nonsyndromic autosomal recessive deafness (DFNB12) and Usher syndrome type 1D, USH1D, characterized by deafness associated with retinitis pigmentosa and vestibular dysfunction. [4][5][6] In mice, Cdh23 mutations lead to hearing loss with or without vestibular dysfunction. [7][8][9] Recently, we identified a novel point mutation (T208C) of Cdh23 and named this mutation erlong (erl). 10 The Cdh23 erl/erl mice (erl mice) proved to be animal models of DFNB12. Previously, we showed that hair cell apoptosis is one of the pathological mechanisms leading to hearing loss in this mutant. 10 Here, we aimed to reveal that ER stress signaling is the upstream pathway leading to hair cell apoptosis in mice with the erl mutation, and we sought to find potential therapeutics.
The perturbation of endoplasmic reticulum (ER) homeostasis leads to the accumulation of unfolded or misfolded proteins in the ER lumen, resulting in ER stress. The unfolded protein response (UPR) is subsequently triggered to alleviate this stress and to restore ER homeostasis, promoting cell adaptation and survival. Conversely, if the stress is prolonged, or if the adaptive response fails, the apoptosis pathway will be initiated. [11][12][13] The UPR is mediated through three ER transmembrane receptors: protein kinase RNA-like ER kinase (PERK); activating transcription factor-6 (ATF6); and inositolrequiring enzyme 1 (IRE1). [11][12][13][14] Upstream of this network, the ER chaperone immunoglobulin-binding protein (BiP) functions as a main regulator by dissociating from the PERK, ATF6 and IRE1 luminal domains and thus activating these effectors' stress responses. As a downstream signal, the CCAAT/ enhancer-binding protein-homologous protein (CHOP) is considered to be a very sensitive indicator of ER stress conditions, and its induction mostly promotes cell death. 11,13 A recent study of zebrafish mutants revealed that the Usher proteins, including CDH23, formed a complex and then preassembled at the ER. 15 Defects in the Usher proteins disrupted the complex formation, induced the protein-trafficking defects, and finally triggered ER stress and apoptosis. Hence, blocking ER stress pro-apoptosis factors could reduce cell death in zebrafish mutants. 15 Because the erl mutation affects Cdh23, the same gene that causes USH1D in humans, it is reasonable to speculate that the ERL-CDH23 protein might endure defective trafficking, thus triggering ER stress and leading to hair cell apoptosis and hearing impairment in erl mutants. In the current study, we tested the expression of the ER chaperone BiP and the pro-apoptotic factor CHOP in the erl and control C57BL/6 J (B6) mouse cochleae. Our data show that ER stress and the UPR response were the underlying causes of apoptosis in the erl cochleae, with the PERK signaling being a contributor. More importantly, we report here that the small molecular compound of ER stress inhibitor Salubrinal (Sal) could slow down the progression of hearing loss and hair cell death in erl mice. These results offer a potential therapy for human DFNB12 and possibly protect the visual function of individuals with Usher syndrome.

Results
CDH23 partly failed to reach the top of hair bundles and were co-localized with BiP in the subapical regions of OHCs in erl mice. We measured the CDH23 distribution in cochlear outer hair cells (OHCs) in erl mice and B6 mice. At P4, the CDH23 protein was specifically localized at the top of the OHCs in B6 mice, as revealed by confocal 3D images, indicating the localization of the tip links in hair bundle stereocilia (Figure 1a). In contrast, the CDH23 protein localized from the stereocilia to the nuclei of the OHCs in erl mice. Portions of the CDH23 proteins failed to reach the top of the hair bundles and remained in the OHC cytoplasm. Immunostaining showed that the ER chaperone BiP was more readily detected in almost the same region and co-localized with cytoplasmic CDH23 in the erl mouse OHCs at P6 (Figure 1b), whereas no specific BiP signaling was detectable in the cytoplasm of the B6 mouse OHCs. Line charts confirmed the higher expression of BiP and its co-localization with CDH23 in the erl OHC cytoplasm The PERK arm of UPR was activated in the erl mouse cochleae. The ER stress marker BiP was upregulated in the erl mice inner ears, as compared with those of B6 mice. In the measurements of mRNA isolated from the erl mouse cochleae, BiP was upregulated at P12 but downregulated at P30 (Figure 2a). Immunostaining showed not only a greater expression of BiP in OHCs but also that BiP was expressed at higher levels in spiral ganglion (SG) cells and in stria vascularis (StV) in the erl cochleae at P6 and P12 (Figure 2b and c). Consistent with the confocal threedimensional (3D)_images, the BiP signals were mainly localized in the OHC cytoplasm and formed a hat-like pattern revealed the location of CDH23 in the B6 (voxel-size 512 × 512 × 0.045 μm 3 and total volume 23.2 × 23.2 × 14.9 μm 3 ) and erl (voxel-size 1024 × 1024 × 0.027 μm 3 and total volume 27.2 × 27.2 × 27.2 μm 3 ) OHCs at P4. In the B6 mice, CDH23 protein was specifically localized at the tip of stereocilia of OHCs. The erl mice OHCs showed a wide range of CDH23 expressions (from the stereocilia to the nucleus). Around the nuclei, no detectable CDH23 signal was found in B6 mice, while in erl mice, many spots of CDH23 signal was detected in this region (n = 3 mice per group). Scale bars, 5μm. (b) The location and co-localization of BiP and CDH23 in the B6 and erl cochleae at P6. The B6 mice cochleae showed faint BiP signals and clear CDH23 signals, particularly on the tops of the OHCs. BiP was highly detected in the OHCs, SG, and StV of erl mice, and it was co-localized with CDH23 in the subapical regions of the OHCs (n = 3 mice per group). Scale bars, 50 μm (left); 20 μm (right). (c) The line charts show two-dimensional graphs of the raw fluorescence intensities of the white arrows drawn in the panels to the left, as analyzed by Image J software. The X-axis represents distance (pixels) along the line, and the Y-axis is the pixel intensity. The BiP signals were highly expressed in erl mice and partially overlapped with the CDH23 signals (n = 3 mice per group) above the OHCs' nuclei (Figure 2b and c). ER stress effector PERK's kinase target, the phosphorylated alpha-subunit of the eukaryotic initiation factor (eIF2α), also showed strong localization in the perinuclear regions of the OHCs in the erl cochleae ( Figure 3).
Downstream of the PERK arm, the pro-apoptotic factor CHOP was also expressed at higher levels in the erl cochleae ( Figure 4). At the mRNA level, Chop maintained a higher expression in the erl cochleae compared with B6 cochleae at P12 and P30 (Figure 4a). Immunostaining performed at the same age showed that CHOP was highly detected in the OHCs, SG and StV of erl mice and was mainly detected in the perinuclear regions of OHCs (Figure 4b and c). However, a bare CHOP signal was present in B6 mouse cochleae.
Disruption of the Chop gene protected hearing and OHCs in erl mice. To further confirm the involvement of Chop in ER stress-induced apoptosis in erl mice, we used double-mutant mice containing the erl mutation in the Cdh23 gene and a disruption in the Chop gene. The inbred mice were obtained by crossing Chop -/mice with Cdh23 erl/erl mice. Nine genotypes of mice sharing the same genetic background (C57BL/6) were generated: Cdh23 +/+ Chop +/+ ; Cdh23 +/+ Chop +/-; Cdh23 +/+ Chop -/-; Cdh23 +/erl Chop +/+ ; Cdh23 +/erl Chop +/-; Cdh23 +/erl Chop -/-; Cdh23 erl/erl Chop +/+ ; Cdh23 erl/erl Chop +/-; and Cdh23 erl/erl Chop -/-. Supplementary Figure 1 shows the genotyping of the double-mutant Cdh23 erl/erl Chop -/mice. We measured the hearing and cytocochleograms of the Cdh23 erl/erl Chop -/-; erl (Cdh23 erl/erl Chop +/+ ); and Chop -/-(Cdh23 +/+ Chop -/-) mice. At 10 weeks, the auditory-evoked brainstem response (ABR) thresholds in the Cdh23 erl/erl Chop -/double-mutant mice were significantly better than those in the erl mice ( Figure 5a). The surface preparation performed at the same age revealed less loss of OHC in the Cdh23 erl/erl Chop -/double-mutant mice in comparison with Cdh23 erl/erl mice. In the erl mice, several spots of OHC loss were exhibited in the basal and middle turns, and a few OHC losses were shown in the apical turns of the cochleae. However, in the Cdh23 erl/erl Chop -/double-mutant mice, no OHC loss was found in the entire length of the cochleae (Figure 5b). The quantitative study showed that the mean percentage of OHC loss in the Cdh23 erl/erl Chop -/mice was significantly lower than that in the erl mice ( Figure 5c).
Sal prevented hearing loss in erl mice. After receiving corresponding treatments, the erl mice in the test, vehicle and control groups underwent ABR and distortion product otoacoustic emission (DPOAE) testing at the same ages. The ABR thresholds in the test group were significantly lower than those in the vehicle and control groups in response to 16 ( Figure 6a and b), 8 and 32 kHz tone-burst and click stimuli (Supplementary Figure 2). No significant difference was found between the vehicle and control groups. The otoprotective effect was further confirmed by the higher DPOAE amplitudes. At 8 weeks, the DPOAE amplitudes in the test group were significantly greater at high frequencies (Supplementary Figure 3a). At 12 weeks, the test group showed much higher DPOAE amplitudes at almost all frequencies (Figure 6c). At 16 weeks, mice in all three groups showed lower DPOAE amplitudes, whereas the test group's amplitudes remained higher at low frequencies (Supplementary Figure 3b).
Sal protected against OHC death in erl mice. A cochlear surface preparation was performed on erl mice in the test, Hearing loss and hair cell death in mutant mice by ER stress J Hu et al vehicle and control groups at 12 weeks. OHC impairment was obvious in the vehicle and control groups, while OHC loss was very minimal in the test group (Figure 7a). In the vehicle and control groups, several substantial contiguous spots of OHC loss were observed in the basal and middle turns, and some OHC loss was seen in the apex turns. In contrast, the mice in the test group showed only small amounts of OHC loss in the basal turn and rare OHC loss in the middle turn. The mean percentage of OHC loss in the test group was significantly lower than that in the vehicle and control groups (Figure 7b). No significant difference was found between the vehicle and control groups. A scanning electron microscope was used to analyze OHC morphology and subcellular structures at 12 weeks (Figure 7c). The images showed that untreated erl mice exhibited almost total OHC loss and no detectable subcellular structure of hair bundles in the basal turn. In contrast, the Sal-treated mice showed small amounts of OHC loss and approximate normal arrangements of hair bundles.
Sal suppressed ER stress-induced apoptosis in erl mice. ER stress and apoptosis-related genes and proteins were downregulated after the Sal treatments. The mRNA expression levels of BiP, Chop, caspase-3 and caspase-12 genes were tested at P30 and 12 weeks. The results showed that Bip and Chop decreased in the Sal group versus the vehicle group at P30, and caspase-3 was downregulated at 12 weeks (Figures 8a and b). However, no significant difference was found in the caspase-12 gene expression (Supplementary Figure 4). BiP and cleaved caspase-3 protein extracted from the

Discussion
As a novel mutation of the Cdh23 gene, the erl mutant mice exhibited the postnatal onset of hearing loss starting at P27, which progressed to total deafness at~P100. This is considered to be an animal model for human DFNB12. 10 Given their time window from initial hearing-loss initiation to total deafness, these mutant mice are ideal tools for testing new otoprotective drugs. We previously revealed that erl mice treated with erythropoietin and Z-VAD-FMK could be significantly protected against OHC death and rescued from hearing loss through the blocking of the apoptosis pathway in erl mice. 10,16 In the current study, the upstream apoptosis response in the erl cochleae was discovered to be relevant to ER stress and the UPR.
Under the quality-control surveillance of the ER, only correctly folded proteins can be transferred by the Golgi complex and sent to their final destinations, while unfolded or misfolded proteins are retained in the ER and ultimately degraded by the ubiquitin-dependent proteolytic pathway or autophagy. 17,18 The accumulation and aggregation of unfolded or misfolded proteins in the ER lumen increase ER loading and disrupt ER homeostasis, ultimately leading to ER stress. 19,20 Under mild-to-moderate stimulation, ER stress is an adaptive and restorative response that leads to cell adaptation. In contrast, if the stress is beyond the ER's adaptive capacity, the protective signaling will switch to pro-apoptotic responses. ER stress has been linked to the pathogenesis of neurodegenerative diseases, such as Alzheimer's disease and Parkinson's disease, as well as to the pathogenesis of cell death, such as renal tubule lesions in diabetes. [21][22][23] In zebrafish mutants, the defective CDH23 proteins in hair cells failed to be transferred and contributed to the formation of Usher protein complexes, therefore leading to ER stress. 15 Taken together, these results offered us new insight into the pathological mechanisms of and potential therapies for sensorineural deafness. Using an animal model for mammals, we questioned whether similar protein-folding defects and the activation of ER stress happened in mice with the Cdh23 mutation.
We provided evidence that part of the ERL-CDH23 protein exhibited abnormal localization in the OHC (from the stereocilia to the nucleus, Figure 1a), failing to reach the top of hair bundles and thus leading to reduced migration to the tip links. Moreover, the ERL-CDH23 protein in the subapical region of the OHC was co-localized with ER chaperone BiP (Figures 1b  and c). The BiP protein was found to be highly expressed in OHCs, SG and StV in the erl cochleae at P6 and P12, and the BiP mRNA was upregulated at P12 but downregulated at P30 (Figure 2). Considering the UPR pathway, these results suggest that BiP was dissociated with the ER transmembrane receptor and then combined with defective CDH23 at the early stage of this response. Downstream of this pathway, eIF2α phosphorylation was detected in the erl cochleae (Figure 3), and the Chop mRNA and protein were upregulated at P12 and P30 (Figure 4). These results indicate that the PERK arm of the UPR was activated. CHOP was a key mediator when the ER stress-induced UPR signaling progressed to apoptosis. 24,25 The results of the auditory test, surface preparation and OHC counting in our double-mutant Cdh23 erl/erl Chop − / − mice further confirm that the deletion of the Chop gene could partly protect hearing and preserve the OHC in mice with the Cdh23 erl/erl mutation ( Figure 5). Taking these data together with the previous study on the induction of apoptosis-related genes in the erl mouse cochleae, 10 we suggest that apoptosis is an event downstream of the induction of the UPR, with the effector PERK signaling pathway being a contributor. Thus, we selected the PERK arm of the UPR as a therapeutic target.
It is interesting to note that CHOP does not present nuclear localization in OHCs, as expected for a transcription factor (Figures 4 and 8). However, CHOP does not have a nuclear localization signal. Its interaction with other stress-induced transcription factors mediates its nuclear localization. 26 It is therefore possible that in OHCs, CHOP is excluded from the nucleus due to the absence of interacting transcription factor partners. It was also reported that cells with cytoplasmic and nuclear localized CHOP gave distinct gene expression profiles. It was shown that that cytoplasmic CHOP inhibited the migration, while nuclear CHOP caused a G1 cell cycle arrest. 27 In this erl model, particularly in OHCs, CHOP is localized in the perinuclear regions in between the nuclei and hair bundles of the OHCs. An OHC develops a polarized bundle of stereocilia that is important for mechanotransduction. 28 Many hair bundle proteins migrate toward the hair bundle; thus, we speculate that the cytoplasmic CHOP may regulate the expression of protein migration-related genes. Further study is warranted on CHOP's role(s) aside from inducing apoptosis on OHC degeneration and functionality.
Sal is a small molecular compound (480 Da) that prevents cell death from ER stress-induced apoptosis by selectively inhibiting the dephosphorylation of eIF2α. 29 Studies show that Sal resists ER stress-induced cell death, improves cell survival and delays disease processes. [30][31][32][33][34] Our hearing tests confirmed our hypothesis that Sal could protect hearing by inhibiting ER stress-induced hair cell apoptosis in the erl mice. In a previous work, we reported that progressive hearing loss started at P27 in erl mice. 10 In this study, as early as 4 weeks postnatal, the 32-kHz-stimulus-evoked ABR thresholds in the test group were 10 dB better than in the control and vehicle groups (Supplementary Figure 1b). At all subsequent time points, the Sal-treated mice showed better ABR thresholds under all stimuli. Referring to the correspondence between high and low frequencies and to the basal-apical gradient of the basilar membrane, 35,36 the auditory tests indicate that the OHC death specifically started in the basal turns and then spread through the entire cochleae, which could be remedied by Sal treatment. Our previous studies targeting anti-apoptosis treatments also showed otoprotective effects, 10,16 but Sal showed an extended duration (up to 16 weeks). In our histological examination, we observed substantial contiguous and isolated OHC losses in the vehicle and control groups; in contrast, the Sal-treated mice showed only small amounts of OHC loss (Figure 7). This result provides a good correlation between our anatomical observations and functional tests.
After the Sal treatment, the upregulated Bip and Chop mRNA were decreased at P30, and the caspase-3 was decreased at 12 weeks (Figure 8a), indicating that Bip and Chop were affected by Sal treatment at an early age and that the caspase-3 mRNA was influenced later. However, the cleaved caspase-3 proteins (detected by western blot) were found to be decreased by Sal at P30. Our cleaved caspase-3 antibody (Asp175) was able to recognize the large fragment (17 kDa) of activated caspase-3 resulting from the cleavage adjacent to Asp175, but it was incapable of detecting the fulllength caspase-3 protein. These results suggest that the caspase-3 cleavage was blocked at P30 by Sal treatment. The caspase-12 mRNA remained unchanged at P30 and 12 weeks (Supplementary Figure 4), indicating that caspase-12 might not be involved in the feedback of the Sal treatment. The BiP proteins (detected by Western blot) and CHOP proteins (detected by immunostaining) were also downregulated after Sal treatment (Figure 8). These results, together with the previous data, 10 suggest that the defective CDH23 accumulated in the ER and then increased the ER loading and destroyed ER homeostasis, leading to the dissociation of the BiP from PERK. The downstream upregulated CHOP led to caspase activation and apoptosis. If the adaptive UPR reduces the protein load in the ER, then the restoration of protein synthesis will promote cell survival. Otherwise, if protein synthesis overrides the restoration of proteostasis, cell death will be triggered. 37 Sal inhibits the dephosphorylation of eIF2α, which inhibits protein synthesis when it is phosphorylated by PERK. This increased phosphorylation of eIF2α limits protein synthesis (including that of CDH23 and CHOP), relieves ER loading, and promotes cell survival. Potential mechanisms of the pro-apoptotic function of uncontrolled protein synthesis during ER stress have recently been described. 37,38 As an animal model of DFNB12, the erl mice with Sal treatment began at an early age (starting from P7) and showed significant hearing protection for a long duration (up to 16 weeks). By considering the signaling pathway of ER stress-induced apoptosis, we suggest that therapies targeting this specific signaling pathway could prevent hair cell loss at an early stage and achieve more effective and sustained results. USH1D patients who carry different mutations of the same gene (Cdh23) as DFNB12 exhibit congenital deafness and postnatal-onset blindness. Thus, we assume that early treatment with Sal might prevent ophthalmic dysfunction in Usher syndrome, thus highlighting Sal's application as a therapeutic agent in the treatment of animal models for Usher syndrome.

Conclusion
In summary, this study is the first to consider ER stressinduced hair cell apoptosis as having a key role in the hearing loss and hair cell death of mice with the Cdh23 mutation. Protein synthesis and related protein-folding processes are early molecular events in ER stress and apoptosis. Thus, protein-folding errors triggered by genetic mutations should be considered the earliest therapeutic target. Many Food and Drug Admission (FDA)-approved anti-ER stress drugs are presently available on the market and thus can be repurposed for otoprotection.

Materials and Methods
Experimental design. This study aimed at assessing ER stress-induced hair cell apoptosis in the mechanism of hearing loss and hair cell death in Cdh23 erl/erl mutant mice. The measurements of ER stress indicators were performed between Cdh23 erl/erl mutant mice, control B6 mice, as well as the Chop gene knockout (C57BL/6 background) mice. The ER stress inhibitor Salubrinal (Sal) was systematic given in Cdh23 erl/erl mutant mice by intraperitoneal injection. The hearing, OHC reservation and ER stress indicators were tested between Sal-treated, vehicle-treated and untreated Cdh23 erl/erl mice.
Sal treatment. The erl mice received different treatments by intraperitoneal injection. In total, 70 erl mice were divided into three groups with either gender: a test group (treated with Sal, 0.5 mg kg − 1 , R&D ng Cat# 2347), a vehicle group (treated with an equal volume of dimethylsulfoxide, DMSO), and a control group (untreated). All of the treatments started from postnatal (P) day 7, with subsequent injections occurring every other day for the first 12 weeks, every 3 days for 2 additional weeks, and then once every week for the duration of the experiments. The Sal dosage was selected from two sets of preliminary experiments, which showed it to be safe and hearing protective. The starting age of treatment was selected to prevent apoptotic gene upregulation in erl mice detected in our previous work. 10 PCR. The cochleae were quickly isolated, and then TRIzol (Invitrogen, Eugene, OR, USA) was used for total RNA extraction. The Super Script III First-Strand Synthesis System (Invitrogen) was used to synthesize cDNA. Quantitative PCR was performed with the reagent of SYBR Green FAST Mastermix (QIAGEN, Germantown, MD, USA) on an ABI 7300 Sequence Detection System (Foster City, CA, USA) . Primer sequences are listed in Table. S1. The cycle reaction was performed as follows: 95°C for 10 min, 40 cycles of 95°C for 15 s and 60°C for 1 min, and finally a dissociation curve of 95°C for 15 s and 60°C for 15 s. The gene expression was calculated relative to the housekeeping gene GAPDH and then analyzed using the 2 − ΔΔCT method. Semi-quantitative PCR was performed using the DNA Engine (Bio-Rad, Hercules, CA, USA) with the cycle reaction described above.
ABR and DPOAE testing. A computer-aided evoked potential system (Intelligent Hearing Systems, the Smart-EP software 3.30, Miami, FL, USA) was used as previously published. 39 Briefly, the mice were anesthetized, and their body temperature was maintained at 37°C. Sub-dermal needle electrodes were used. The recording electrode was inserted at the vertex of the skull; the ground electrode was inserted in the apex of the nose; and the reference electrodes were fixed near each ear. Click and 8, 16 and 32 kHz tone bursts were channeled through an inserted earphone. The ABR threshold was identified as the lowest stimulus level at which clear and repeatable waveforms were recognized (Figure 2a). The DPOAE measurement was conducted for pure tones at frequencies ranging from 4.4 to 20.3 kHz by using the Intelligent Hearing System (Smart EP 3.30 Software). Frequencies were acquired with the F2:F1 ratio of 1.22 and with primary stimulus of 65/55 dB SPL.
Surface preparation and hair cell counting. As we previously described, 40 the temporal bones were fixed in 4% PFA, and the basilar membrane was then carefully dissected and cut into three separate fragments: apical, middle and basal turn. The fragments were permeabilized in 2% Triton X-100, stained for F-actin with phalloidin (Invitrogen), and observed with a fluorescence microscope (Leica DM4500 B). Hair cells were counted as being present if cell bodies and the V-shaped hair bundles were intact.
Scanning electron microscope. The mice were fixed with 2.5% glutaraldehyde, and then the whole cochleae were dissected. The bony capsule, spiral ligament, and Reissner's membrane were carefully removed to expose the whole organ of Corti. Afterwards, the specimens were fixed in 1% osmium tetroxide (OsO4) for 1 h three times and in 1% thiocarbohydrazide (TCH) for 1 h twice (the OTOTO technique). The specimens were dehydrated in a gradient ethanol series, critical-point dried using CO 2 , and finally coated in palladium. Then, the samples were viewed under a high-resolution scanning electron microscope (FEI Helios NanoLab 650, Germany).
Statistical analysis. The data are presented as mean ± S.E.M. The analysis was performed using the SPSS 18.0 software. The data were statistically analyzed using one-way ANOVA and Student's t-tests. P-valueso0.05 were considered significant.

Conflict of Interest
The authors declare no conflict of interest.