Abstract
Disordered protein ubiquitination has been linked to neurodegenerative disease, yet its role in inner ear homeostasis and hearing loss is essentially unknown. Here we show that progressive hearing loss in the ethylnitrosourea-generated mambo mouse line is caused by a mutation in Usp53, a member of the deubiquitinating enzyme family. USP53 contains a catalytically inactive ubiquitin-specific protease domain and is expressed in cochlear hair cells and a subset of supporting cells. Although hair cell differentiation is unaffected in mambo mice, outer hair cells degenerate rapidly after the first postnatal week. USP53 colocalizes and interacts with the tight junction scaffolding proteins TJP1 and TJP2 in polarized epithelial cells, suggesting that USP53 is part of the tight junction complex. The barrier properties of tight junctions of the stria vascularis appeared intact in a biotin tracer assay, but the endocochlear potential is reduced in adult mambo mice. Hair cell degeneration in mambo mice precedes endocochlear potential decline and is rescued in cochlear organotypic cultures in low potassium milieu, indicating that hair cell loss is triggered by extracellular factors. Remarkably, heterozygous mambo mice show increased susceptibility to noise injury at high frequencies. We conclude that USP53 is a novel tight junction-associated protein that is essential for the survival of auditory hair cells and normal hearing in mice, possibly by modulating the barrier properties and mechanical stability of tight junctions.
SIGNIFICANCE STATEMENT Hereditary hearing loss is extremely prevalent in the human population, but many genes linked to hearing loss remain to be discovered. Forward genetics screens in mice have facilitated the identification of genes involved in sensory perception and provided valuable animal models for hearing loss in humans. This involves introducing random mutations in mice, screening the mice for hearing defects, and mapping the causative mutation. Here, we have identified a mutation in the Usp53 gene that causes progressive hearing loss in the mambo mouse line. We demonstrate that USP53 is a catalytically inactive deubiquitinating enzyme and a novel component of tight junctions that is necessary for sensory hair cell survival and inner ear homeostasis.
Introduction
Hair cells in the mammalian cochlea are highly specialized epithelial cells that convert sound-induced vibrations into electrical signals through K+-mediated depolarization (Fig. 1). The apical surface of hair cells is surrounded by the endolymph, a K+-rich extracellular fluid produced by the stria vascularis, whereas the basal end is surrounded by the low K+ perilymph (Wangemann, 2006). Tight junctions (TJs) connect the apices of hair cells and supporting cells in the reticular lamina and separate the basal and apical surfaces of these cells, preventing paracellular leakage of solutes (Gulley and Reese, 1976). Likewise, TJ barriers in the stria vascularis maintain the endolymph at a high positive resting potential, the endocochlear potential (EP), which provides the driving force for hair cell transduction.
TJs are composed of integral, peripheral membrane and intracellular proteins that form a network of sealing strands near the apical cell borders. Major integral membrane TJ proteins are the claudins and occludin that join the membranes of adjacent cells via trans-homophilic interactions while connecting intracellularly to the actin cytoskeleton through adaptor proteins, such as zonula occludens (ZO)-1/tight junction protein (TJP) 1, ZO-2/TJP2, and ZO-3/TJP3 (Furuse et al., 1993; Steed et al., 2010; Tamura and Tsukita, 2014).
Disturbance of cochlear ion homeostasis attributable to impaired TJs is the cause of many forms of hearing loss. Mutations in CLDN14, encoding claudin-14, underlie human hereditary deafness DFNB29 (Wilcox et al., 2001). Claudin-11 null mice lack the EP and develop hearing loss as a result of defects in basal cell TJs of the stria vascularis (Gow et al., 2004; Kitajiri et al., 2004). In contrast, hearing loss in claudin-9, claudin-14, occludin, tricellulin, and ILDR1-deficient mice has been ascribed to altered leakiness of TJs in the organ of Corti and an increased concentration of K+ around the basolateral surfaces of the outer hair cells (OHCs), resulting in hair cell degeneration (Ben-Yosef et al., 2003; Nakano et al., 2009; Nayak et al., 2013; Kitajiri et al., 2014; Morozko et al., 2015). Finally, genomic duplication and overexpression of TJP2 has been linked to nonsyndromic deafness DFNA51 (Walsh et al., 2010).
TJs are highly dynamic structures, and their barrier function is modulated by different signaling pathways and posttranslational protein modifications, including phosphorylation and palmitoylation (Antonetti et al., 1999; Van Itallie et al., 2005). The role of protein ubiquitination in this process is less well understood. Ubiquitination regulates protein stability, function, and/or localization and is critical for a multitude of cellular functions (Hershko and Ciechanover, 1998). Ubiquitin is conjugated to lysine residues in target proteins by E3 ubiquitin ligases either as a single unit (monoubiquitination) or as a branched chain (polyubiquitination) (Komander and Rape, 2012). In particular, lysine 48-linked polyubiquitin chains target proteins for proteasomal degradation. Protein ubiquitination is counter-regulated by deubiquitinating enzymes (DUBs), which cluster in five protein families: ubiquitin C-terminal hydrolases (UCHs), ubiquitin-specific proteases (USPs), ovarian tumour proteases (OTUs), Josephins, and JAB1/MPN/MOV34 metalloenzymes (JAMMs) (Komander et al., 2009). The USP family constitutes the largest family. Despite extensive studies on E3 ligases in inner ear homeostasis (Zenker et al., 2005; Nelson et al., 2007), the physiological role of USPs and other DUBs is poorly defined.
Here, we report the phenotypic characterization of a mutant Usp53 allele termed mambo, which was isolated from a forward genetics screen in mice. mambo mice carry a point mutation in the predicted catalytic domain of Usp53 and exhibit a rapidly progressive hearing loss. Usp53 is expressed broadly in the inner ear, and OHC survival is affected selectively in the cochlea. USP53 and the adaptor proteins TJP1 and TJP2 bind to each other and colocalize at TJs in epithelial cells, suggesting that they form a complex critical for TJ function. Remarkably, OHCs evade degeneration in organ culture, indicating that unfavorable extracellular conditions promote OHC degeneration in mambo mice. Together with previous studies, our findings support the idea that cochlear fluid homeostasis is tightly regulated by the actions of ubiquitin ligases and DUBs and suggest that molecular components of the ubiquitin pathway are potential risk factors for progressive hearing loss.
Materials and Methods
All procedures were performed in accordance with research guidelines of the institutional animal care and use committee of Rutgers University. Mice of either sex were studied.
Ethylnitrosourea mutagenesis screen, auditory brainstem response, and distortion product otoacoustic emission measurement.
The ethylnitrosourea (ENU) mutagenesis protocol and primary phenotypic screen has previously been described (Reijmers et al., 2006; Schwander et al., 2007). The measurement of auditory brainstem responses (ABRs), distortion product otoacoustic emissions (DPOAEs), as well as the assessment of vestibular function in open-field and swim tests followed our published procedures (Schwander et al., 2007).
Linkage analysis and DNA sequencing.
Genome-wide single nucleotide polymorphism (SNP) genotyping for linkage analysis was performed as described previously (Schwander et al., 2007). Affected mambo mice were bred with BALB/cByJ mice. To identify the mambo mutation, a list of annotated and predicted genes in the affected interval was established using the University of California Santa Cruz (UCSC) genome browser. Total RNA was isolated from inner ear tissue of postnatal day 5 (P5) wild-type and mambo mice. cDNA was generated using oligo-dT primers and SuperScript II reverse transcriptase (Life Technologies). Annotated and predicted genes in the genomic interval were amplified using Phusion DNA polymerase (New England Biolabs) and sequenced with primers specific to the full-length gene. In addition, exon-based sequencing of candidate genes was performed with primers designed via the Exon Primer function in the UCSC Genome browser (http://genome.ucsc.edu/).
Mouse genotyping.
Genotyping was performed by PCR using a set of primers that flank the mambo mutation in the Usp53 gene: forward primer 796f, 5′- CTTCAGATACACTTTGATTTTCATTG-3′, and reverse primer 796r, 5′-GGTTCAGATGAACAAAACTAAGACC-3′. PCR fragments were purified and digested with TspRI for 4 h at 65°C to give a 310 bp product in wild-type mice, 110 and 200 bp products in homozygous mutants, and all three products in heterozygous littermates.
DNA constructs.
The mouse Usp53 gene contains 16 exons, of which 15 are coding, distributed over 51 kb of genomic DNA. Several partial and two complete sequences of Usp53 transcripts have been described. A short transcript (GenBank AK045953.1) encodes a putative protein fragment with an incomplete N-terminal catalytic domain (amino acids 1–274) and nine additional amino acids. The apparent full-length cDNA (NCBI Reference Sequence: NM_133857.3) encoding mouse USP53 (1069 aa) was amplified from cochlear RNA by RT-PCR and inserted in-frame into XhoI and SacII sites of pEGFP-C1 (Clontech). The mambo mutation (p.C228S) was introduced into USP53 by overlap PCR using XhoI/SacII as flanking restriction sites. Mutation of the predicted Src homology 3 (SH3) binding domain (amino acids 1008–1013; PPPLPP to AAALAA) was performed by overlap PCR using SalI/SacII sites. The catalytic domain (amino acids 1–348) and the tail domain (amino acids 393–1069) of USP53 were amplified by PCR and subcloned into pEGFP–C1 using XhoI/SacII sites. Additional C-terminal truncation mutants C1 (amino acids 1–1017), C2 (amino acids 1–900), and C3 (amino acids 1–676) were amplified by PCR and inserted in-frame into SalI/SacII (C1, C2) and KpnI/SacII (C3) sites of pEGFP-C1. Full-length human TJP2 in p2xFlag CMV2 was obtained from Addgene (plasmid 27415; Oka et al., 2010). All constructs were verified by DNA sequencing.
Histology, electron microscopy, and immunolocalization studies.
Whole-mount staining, transmission electron microscopy (TEM), and scanning electron microscopy (SEM) of cochlear sensory epithelia were performed as described previously (Senften et al., 2006; Schwander et al., 2007). To generate anti-USP53 antisera, rabbits were coinjected with two synthetic peptides derived from the sequence of mouse USP53 (LNRSQMQTSGRRAPVKLSHD, STASAPRLERVGLSPDVGV). The peptide sequences show no homology with any other protein in publically available databases. The antibodies were purified from the IgG fraction of the antisera by affinity chromatography against each of the two peptides. In addition, two rabbit polyclonal anti-USP53 antibodies, HPA035844 and HPA035845, were obtained from Sigma-Aldrich. All three anti-USP53 antibodies specifically detected green fluorescent protein (GFP)–USP53 fusion protein when expressed in HEK293T cells. Protein expression was evaluated by immunofluorescence analysis with USP53 antibodies as described previously (Senften et al., 2006), except that cochlear tissues were fixed in 2% PFA for 40 min at room temperature and incubated with primary antibodies overnight at room temperature on a rocking table. Additional antibodies were as follows: mouse anti-Sox2 (Santa Cruz Biotechnology), rabbit anti-myosin VIIa (Proteus Biosciences), mouse anti-calbindin (Sigma), chicken anti-GFP (Rockland), rabbit anti-GFP (Sigma), rabbit anti-GFP (Abcam), rabbit anti-ZO-1 and anti-ZO-2 (Life Technologies), rabbit anti-occludin (Abcam), rabbit anti-UCHL3 (Cell Signaling Technology), anti-ZO-1 Alexa Fluor 488 conjugate (Life Technologies), anti-FLAG horseradish peroxidase (HRP) conjugate (Sigma), Alexa Fluor 488 and 568 anti-rabbit and anti-mouse (Life Technologies), Alexa Fluor 488 anti-chicken (Life Technologies), Alexa Fluor 350 and 488 phalloidin (Life Technologies), and HRP-conjugated anti-rabbit (GE Healthcare).
Quantitative RT-PCR.
RNA was isolated from finely dissected cochleae using the ARCTURUS PicoPure RNA Isolation kit (Thermo Fisher Scientific) and treated with DNase I (Qiagen). cDNA was synthesized from 72 ng of RNA using the SuperScript IV First-Strand Synthesis System (Thermo Fisher Scientific) and random hexamers. Quantitative PCR was performed using Usp53 isoform-specific primers and iTaq Universal SYBR Green Supermix (Bio-Rad) in a Rotor-Gene 3000 thermal cycler/real-time DNA detection system (Corbett Research). Expression data were normalized using phosphoglycerate kinase 1 (Pgk1) as a reference gene. Primers included the following: short isoform of Usp53, forward, 5′-GAAGTGTCCTAGTAACTGTGGCC-3′, and reverse, 5′-GAATGAAAGCAACTGTGATACCCC-3′; long isoform of Usp53, forward, 5′-CGACACAGGGATTTGGTTGATG-3′, and reverse, 5′-CAGAGGTGTAGCTCTCATGGG-3′; and Pgk1, forward, 5′-CAACAACATGGAGATTGGCACA-3′, and reverse, 5′-ACAGTAGCTTGGCCAGTCTTG-3′. No difference in the efficiency of primers against the two Usp53 isoforms was detected on serial dilutions of plasmid DNA. The efficiency was close to 100% for both sets of primers. For quantification purposes, a common threshold fluorescence value for all primer pairs was chosen within the early exponential growth phase. The fractional quantification cycle numbers Cq for which the threshold was achieved were extracted from the curve for each primer pair. The fold difference in the abundance of each Usp53 isoform relative to Pgk1 was calculated as 2(CqPgk1 − CqUsp53), where CqPgk1 and CqUsp53 are quantification cycle numbers for Pgk1 and Usp53, respectively. The levels of nonspecific amplification by each primer pair were assessed with no-RT controls (cDNA prepared without reverse transcriptase). The nonspecific amplification levels were >20-fold lower than the specific signal for each pair of primers.
In situ hybridization.
In situ hybridization (ISH) was performed on 12-μm-thick cryosections, as described previously (Schwander et al., 2007; Grillet et al., 2009). The RNA probes are complementary to cDNA segments corresponding to amino acid residues 118–474 (probe 1) and 473–797 (probe 2) of mouse USP53.
Immunoprecipitation and Western blot analysis.
HEK293T cells were transfected with X-tremeGENE 9 DNA transfection reagent (Roche). After 20 h, cell extracts were prepared in 20 mm potassium phosphate, pH 7.4, 500 mm NaCl, 1% NP-40, 0.5% sodium deoxycholate, 50 μm ZnCl2, and MS-SAFE protease inhibitor (Sigma). Lysates were diluted 1:4 in 20 mm potassium phosphate buffer containing 50 μm ZnCl2 and precleared with 30 μl of protein A-Sepharose (GE Healthcare) for 1 h at 4°C. Lysates were incubated with anti-GFP antibody (ab290; Abcam) for 1.5 h at 4°C, followed by overnight incubation with 30 μl of protein A-Sepharose at 4°C. Immunocomplexes were washed three times with immunoprecipitation buffer containing 20 mm potassium phosphate, 125 mm NaCl, 0.25% NP-40, 0.125% sodium deoxycholate, and 50 μm ZnCl2 before addition of Laemmli's sample buffer and SDS-PAGE. Proteins were transferred to PVDF membrane with an iBlot system (Life Technologies), and the membrane was blocked with 2% dry milk (Amresco) for 30 min. The blots were probed with the following antibodies: rabbit anti-GFP (Sigma), rabbit anti-UCHL3 (Cell Signaling Technology), mouse anti-FLAG (Sigma), rabbit anti-ZO-1 and anti-ZO-2 (Life Technologies), and rabbit anti-occludin (Abcam). Primary antibodies and proteins were visualized with HRP-conjugated anti-rabbit antibody (1:20,000; GE Healthcare) using the ECL2 detection system (Thermo Fisher Scientific).
DUB inhibitor assay.
HEK293T cells were transfected with X-tremeGENE 9 DNA transfection reagent (Roche). After 24 h, cells were solubilized in 50 mm Tris, pH 7.5, 150 mm NaCl, 1% NP-40, 50 μm ZnCl2, 130 μm bestatin (Sigma), 1 μm pepstatin A (Sigma), and 1 mm DTT for 20 min on ice. Lysates from transfected and nontransfected (mock) cells were treated with each of the following DUB inhibitors at a concentration of 50 μm: ubiquitin-vinyl methyl ester (Ub-VME), ubiquitin-propargylamide (Ub-PA), or ubiquitin-vinyl sulfone (Ub-VS; Boston Biochem) for 1 h at 37°C, boiled in Laemmli's sample buffer for 10 min at 95°C, and analyzed by SDS-PAGE and Western blotting.
Culture and injectoporation of cochlear explants.
Injectoporation of cochlear explants was performed following recently established protocols (Xiong et al., 2014). Organs of Corti were dissected from P3–P4 wild-type mice, cut into three pieces, and cultured on an uncoated glass coverslip in DMEM/F-12 with 10 ng/μl ampicillin for 4 h at 37°C. Subsequently, adherent cochlear explants were placed between two platinum wire electrodes (Surepure Chemetals) and injected with plasmid (1 μg/μl) between the second and third row of OHCs using a microinjection pipette (2–3 μm diameter). Pipette and electrodes were positioned using a BX51 upright microscope with a 60× objective (Olympus), and two micromanipulators (MPC-200; Sutter Instruments). A series of three to five square pulses (15 ms length, 1 s intervals) with a magnitude of 60 V was applied using an ECM 830 electroporator (Harvard Apparatus). Organs of Corti were cultured for 8–12 h in DMEM/F-12, fixed with 4% PFA, and stained with anti-TJP2 antibody (Life Technologies) and phalloidin to visualize TJs and F-actin in stereocilia, respectively. Samples were imaged using a BX63 fluorescence microscope and CellSens software (Olympus).
Biotin tracer assay.
Temporal bones were dissected from mice at 1 month of age, and the round and oval windows were opened in PBS containing 1 mm CaCl2 (PBS/C). Cochleae were perfused through the round and oval windows with 100 μl of 10 mg/ml EZ-Link Sulfo-NHS-LC-Biotin (Pierce Chemical) in PBS/C for 5 min and then flushed with PBS/C five times. The temporal bones were then fixed by perilymphatic perfusion with 10% TCA for 1 h, rinsed once in PBS, and decalcified in 0.12 m EDTA in PBS for 2 d at room temperature. The tissue samples were rinsed in PBS, incubated in 30% sucrose in PBS overnight at 4°C, flash frozen in OCT compound in liquid nitrogen, and sectioned at 15 μm. Frozen sections were incubated with streptavidin Alexa Fluor 568 conjugate (Life Technologies) for 1 h to visualize the distribution of the biotin tracer.
Endocochlear potential.
EP recordings followed published protocols with some modifications because of the size and fragility of the mice at P10–P11 (Ohlemiller et al., 2006). Adult mice were anesthetized (60 mg/kg sodium pentobarbital, i.p.) and positioned ventrally in a custom head holder. Core temperature was maintained at 37.5 ± 1.0°C using a thermostatically controlled heating pad in conjunction with a probe (model 73A; Yellow Springs Instruments). An incision was made along the midline of the neck, and soft tissues were blunt dissected laterally to expose the trachea and the animal's left bulla. A tracheostomy was then made, and the musculature over the bulla was cut posteriorly to expose the bone. Using a hand drill, a small hole was made in the ventral surface of the bulla, and the opening was expanded to expose the cochlear base. Using a tri-corner pick, a small hole was made in the cochlear capsule directly over the scala media of the lower basal turn. Glass capillary pipettes (40–80 MΩ) filled with 0.15 m KCl were mounted on a hydraulic microdrive (FHC) and advanced until a stable positive potential was observed that did not change with increased electrode depth. The signal from the recording electrode was led to an AM Systems model 1600 intracellular amplifier. A silver/silver chloride ball electrode inserted into the neck muscles served as ground. Mice tested at P10–P11 were unstable under sodium pentobarbital, so that a ketamine–xylazine (80 and 15 mg/kg, respectively) mouse mixture was used in these mice.
Noise exposure.
Mice were exposed fully awake to 8–16 kHz octave band noise for 2 h at 98 dB SPL. Noise exposure was performed in an Industrial Acoustics double-walled sound booth. Up to two 30 × 19 × 13 cm reverberant plastic cages containing one mouse each were suspended 50 cm below an exponential horn (Selenium Corneta HM4750-SLF). Noise was generated digitally using custom Labview routines and presented using a Tucker-Davis Technologies RZ6 in combination with a Crown Audio power amplifier. Sound level was monitored in real time by Brüel and Kjær Type 2203 sound-level meter and tracked by custom software. Mice were 7 weeks of age at the time of exposure. For the ABR measurement, animals were anesthetized with a mixture of ketamine–xylazine (80 and 15 mg/kg, i.p., respectively) and placed in a prone position on a platform with an ES-1 free-field speaker (Tucker-Davis Technologies) 7 cm directly lateral from the right ear. Subdermal platinum electrodes (Grass Instruments) were placed behind the right pinna (reference), on the vertex (active), and under the skin of the back (ground). A rectal probe was used to monitor temperature that was maintained near 38°C using a direct current-based isothermal pad (FHC). Tone bursts 5 ms in duration (0.5 ms cos2 rise/fall) were presented 500–1000 times at 20/s in descending intensity using a 5 dB minimum step size until wave I of the ABR could no longer be discerned visually. Recording used Biosig32 and Tucker-Davis Technologies hardware. Baseline ABR tests were obtained at 5, 10, 20, and 28.3 kHz 2 d before exposure (preexposed) and 14 d after (exposed).
Results
A missense mutation in the Usp53 gene underlies hearing loss in mambo mice
The mambo mouse mutant was generated on a C57BL/6/6J background in a large-scale mouse ENU-mutagenesis screen for novel recessive deafness traits (Schwander et al., 2007). mambo mice presented with a low acoustic startle response (ASR), a rapid and involuntary reflex involving the contraction of muscles evoked by a loud sound (Schwander et al., 2007). Because reduced ASR can be caused by deficits in emotional processing or motor or auditory function, the mice were tested for ABR, which is a measure of sound-evoked neuronal activity (Zheng et al., 1999). ABR thresholds >90 dB SPL were measured for a broadband click stimulus, indicating severe hearing impairment (Schwander et al., 2007). Heritability testing of mambo mice on the C57BL/6J background revealed that they inherit their deafness phenotype recessively (Schwander et al., 2007).
To map the affected genomic locus, we outcrossed affected mambo mice to the BALB/cByJ mouse strain. The resulting F1 hybrids were intercrossed to generate F2 mice for ABR phenotyping and linkage analysis with SNP markers as described previously (Wiltshire et al., 2003; Schwander et al., 2007). Consistent with a fully penetrant recessive trait, we found ∼24% (13 of 55) of the F2 animals to be affected in ABR tests. SNP mapping of the F2 progeny identified a 1.6 Mb interval (122.3–123.9 Mb) on chromosome 3 that segregated with the mambo phenotype (Fig. 2A). We next designed primers for all annotated (n = 12) and predicted (n = 3) genes in the interval, amplified mRNA from the organ of Corti of affected mambo and C57BL/6J mice by RT-PCR and sequenced the amplicons (Fig. 2B; and data not shown). For some genes, we also screened genomic DNA for mutations by sequencing exons, intron/exon boundaries, and most of the 5′ and 3′ untranslated regions. We identified only one mutation: a T>A transversion (c.682T>A) located in coding exon 7 of the Usp53 gene (p.C228S; Fig. 2C). The mambo mutation introduces a TspRI restriction site that enabled us to genotype mice by PCR-restriction fragment length polymorphism (PCR-RFLP) assay (Fig. 2D). The point mutation was uniquely homozygous in mice that displayed the deafness phenotype (data not shown). Based on these findings, we conclude that the point mutation in the Usp53 gene underlies the deafness phenotype in mambo mice.
Progressive hearing loss and OHC dysfunction in mambo mice
To investigate the nature of hearing impairment in mambo mice, we determined ABR thresholds in 2.5-, 4-, and 8-week-old mice in response to broadband click stimuli starting at 90 dB SPL and then decreasing in intensity. The recorded potentials were plotted against the response time at any given sound intensity (Fig. 3A). The first peak (I) represents the activity of the auditory nerve; the subsequent peaks (II–IV) arise in neuronal populations of the auditory brainstem. At P18, ABR thresholds were elevated in homozygous mambo mice (62 ± 4 dB SPL, mean ± SEM, p < 0.005, two-tailed t test) when compared with heterozygous littermates (39 ± 4 dB) or age-matched wild-type mice (42 ± 5 dB; Fig. 3B). By P30, homozygous mambo mice were profoundly hearing impaired (ABR thresholds >80 dB), and, by P60, most of them were deaf (ABR thresholds >90 dB; Fig. 3B). Analysis of ABRs elicited by pure tones between 4 and 32 kHz revealed that mambo mice were equally affected across all frequencies (Fig. 3C). We conclude that the hearing loss in mambo mice is early onset and rapidly progressive in nature and affects all tested frequencies.
Disruption of the first peak of the ABR waveform (Fig. 3A) could be caused by defects in hair cells and/or spiral ganglion neurons. One possibility is that the decreased auditory sensitivity in mambo mice is attributable to compromised OHC function, which leads to a loss of cochlear amplification. Therefore, we recorded DPOAEs, which depend on the mechanical activity of OHCs. DPOAEs were present in 2-month-old wild-type and heterozygous mambo mice but were absent in homozygous mutant animals at all of the frequencies analyzed (6–28 kHz; Fig. 3D,E). The results indicate that the functional impairment of OHCs contributes to hearing loss in mambo mice, although OHC dysfunction alone does not cause complete deafness. Aside from hearing loss, no differences between mambo and wild-type mice were observed in gross neurologic function and reflexes. In particular, vestibular function was intact (Schwander et al., 2007).
Defects in hair cell morphology in mambo mice
The abnormal DPOAE recordings observed in mambo mice suggested that the integrity of cochlear hair cells might be compromised. Therefore, we analyzed the morphology of hair cells in cochlear sensory epithelia from homozygous mambo mice and age-matched wild-type mice by SEM (Fig. 4A). At P5, the cochlear sensory epithelium was normally patterned into one row of inner hair cells (IHCs) and three rows of OHCs, and hair bundles exhibited the characteristic staircase-like arrangement of stereocilia. By P11, clear changes in the morphology and number of OHCs became evident in mambo mice (Fig. 4A, arrow). OHCs showed signs of degeneration, including ruffling of the apical cell membrane and fused stereocilia (Fig. 4A, arrowhead). In addition, OHCs occasionally displayed subtle defects in bundle polarity (data not shown). TEM analysis of the cochlear duct at P10 revealed incomplete and lumpy chromatin condensation in OHC nuclei, indicative of early apoptotic changes (Fig. 4B, arrowheads). However, no signal for cleaved caspase-3 was observed in hair cells of mutant cochlea by immunohistochemistry (data not shown), suggesting that the lack of USP53 triggers OHC death via a caspase-independent pathway.
We next analyzed hair cell morphology by staining inner ear sections from 1- and 3-month-old mice with hematoxylin–eosin (Fig. 4C). Consistent with the SEM results, we observed a significant loss of OHCs in the cochlea of mambo mice at 1 month of age, whereas IHCs and spiral ganglion neurons appeared unaffected (Fig. 4C, top). To evaluate survival of hair cells and supporting cells, we counted cell bodies in serial sections through the cochlear duct from wild-type and mambo mice (n = 4 and 4). In addition to the loss of OHCs (p < 0.005, two-tailed t test), there was a small but statistically significant decrease in the number of Deiters' cells (p < 0.05) but not IHCs (p = 0.8; Fig. 4D). By 3 months of age, most cochlear hair cells were missing and degeneration of the entire organ of Corti was evident, indicating that OHC loss is followed by the degeneration of IHCs in adult mice (Fig. 4C, row 2, arrowheads). Degeneration of peripheral axons and somata of spiral ganglion neurons was widespread (Fig. 4C, rows 2 and 3, arrows). Because there were no obvious morphological abnormalities in sensory neurons of the early postnatal inner ear, initial hearing loss in mambo mice is likely caused by defective hair cell function. Finally, the gross anatomy of the vestibular end organs appeared normal, which is consistent with the nonsyndromic hearing loss phenotype of mambo mice (Fig. 4C, row 4; Schwander et al., 2007).
Immunofluorescence staining of cochlear whole mounts revealed that hair cells in mambo mice differentiate normally and display specific hair cell markers, such as myosin VIIa, but start to degenerate at approximately P8 (Fig. 4E, top panels). At P10, OHCs were found frequently in aberrant positions among Sox2-positive nuclei of Deiters' cells, indicating disruption of cochlear microarchitecture (Fig. 4E, bottom panels). OHCs are lost progressively along the cochlea in a mosaic pattern, rather than following a basal-to-apical gradient (data not shown). Altogether, these data support the hypothesis that the degeneration of OHCs is the initial trigger of pathological changes in the cochlea, which is reflected in the reduced DPOAE responses in mambo mice.
The mambo mutation affects the putative catalytic domain of USP53
Several mRNA isoforms of Usp53 have been described. The longest isoform that we could amplify from mouse cochlea is identical to NM_133857.3 and comprises 15 coding exons, encoding a protein of 1069 aa (Fig. 5A). The USP53 protein has a predicted N-terminal USP catalytic domain (amino acids 1–348) and belongs to the USP superfamily of DUBs (Quesada et al., 2004; Reyes-Turcu et al., 2009). The mambo mutation (c.682T>A) introduces a missense mutation (p.C228S) in coding exon 7 of the Usp53 gene, which encodes part of the USP catalytic domain of USP53 (Fig. 5A). A shorter cloned cDNA for Usp53 (AK045953) contains coding exons 1–7, which encode a putative protein fragment that consists of an incomplete N-terminal catalytic domain (amino acids 1–274) and nine extra amino acids resulting from usage of an alternative 3′ splice site in exon 7 (Fig. 5B). These are the only known Usp53 isoforms that contain exon 7, which is affected by the mambo mutation. We assessed the relative abundance of the short and long (full-length) Usp53 isoforms with respect to the reference gene Pgk1 in cochlea of wild-type mice at P6 by RT-qPCR (Fig. 5C). Levels of full-length Usp53 mRNA were at least 10-fold higher compared with the short isoform. A 10-fold difference in the abundance of the two isoforms was also detected in heterozygous and homozygous mambo mice in the age range P4–P8 (data not shown). The data indicate that full-length Usp53 is the predominant isoform in the cochlea and likely most relevant to inner ear function.
We found USP53 orthologs solely in vertebrate genomes (human, mouse, rat, chicken, Xenopus laevis, and zebrafish) but not in arthropods (Drosophila melanogaster) or nematode (Caenorhabditis elegans). Multiple sequence alignments showed that the affected residue, C228, is conserved among all USP53 orthologs (Fig. 5D). BLAST (basic local alignment search tool) database searches showed that the catalytic domain of USP53 has the highest sequence similarity (65% identical and 78% similar residues) to the catalytic domain of USP54, a protein of unknown function (data not shown). In addition, USP53 and USP54 show significant sequence similarity to echinus, a Drosophila protein required for ommatidial rotation and programmed cell death in the fly eye, within their respective catalytic domains (Copeland et al., 2007; Montrasio et al., 2007). The sequence similarity to echinus extends to the C-terminal tail region in USP54 but not in USP53, indicating that the two DUBs may have acquired distinct functions. A functionally essential histidine residue of the His-box, which contributes to the recognition of the C terminus of ubiquitin, is absent in both USP53 and USP54 (Fig. 5A), suggesting that these proteins may lack catalytic activity (Hu et al., 2002; Quesada et al., 2004).
USP catalytic domains share a common conserved fold, resembling an opened right hand with thumb, palm, and fingers subdomains (Hu et al., 2002). The catalytic triad is formed by the thumb (Cys) and palm subdomains (His, Asp/Asn; Fig. 5A, arrows), whereas the fingers subdomain serves as the primary ubiquitin interaction site. Two adjacent double cysteine motifs (CXXC) in the fingers subdomain coordinate one zinc ion in many of the USPs (Avvakumov et al., 2006; Renatus et al., 2006). USP53 contains two double cysteine motifs in a region that overlaps with the fingers subdomain of other USP proteins, although the second double cysteine motif diverges from CXXC to CXXXC (Ye et al., 2009). Strikingly, the mambo mutation converts the second cysteine in the CXXXC motif into a serine (Fig. 5D). Thus, a similar zinc binding site may be present in USP53, and ubiquitin binding rather than catalysis may be affected by the mambo mutation.
To test the catalytic activity of USP53, we performed a DUB inhibitor assay. DUB inhibitors contain ubiquitin conjugated to a reactive group, such as vinyl methyl ester (Ub-VME). During ubiquitin binding to the DUB, the active-site cysteine will react with the reactive group and form a covalent complex between the DUB and the ubiquitin molecule, which can be detected as a shift in molecular weight by Western blotting. Therefore, we treated lysates from HEK293T cells overexpressing wild-type GFP–USP53 with different DUB inhibitors, followed by Western blotting with antibodies against USP53 or UCHL3, another DUB, which served as a positive control. In samples treated with DUB inhibitors, a shift in molecular size was observed for UCHL3 but not for USP53 (Fig. 5E), suggesting that USP53 is a catalytically inactive DUB. These results are consistent with previous enzymatic activity studies showing that USP53 is devoid of catalytic activity against ubiquitin–β-galactosidase fusion protein (Quesada et al., 2004).
mambo is expressed broadly in the inner ear
To examine the expression pattern of Usp53 in the mouse inner ear, we performed ISH assays on frozen sagittal sections from wild-type mice at P4. Identical results were obtained using two independent RNA probes that hybridize to non-overlapping regions of Usp53 mRNA (Fig. 6A). Expression of Usp53 mRNA was detected at intermediate levels in hair cells and a particular subset of cochlear supporting cells, including pillar cells, Deiters' cells, and Hensen's cells (Fig. 6B–I). Usp53 was also expressed in the cochlear nerve and fibrocytes of the spiral ligament, but no expression was observed in stria vascularis and spiral limbus (Fig. 6B–D,F–H). Nonspecific hybridization signal in the tectorial membrane and Reissner's membrane was also observed in sections incubated with a Usp53 sense probe, which served as a negative control (Fig. 6J–M). An example of ISH for a hair cell-specific gene, Pejvakin, is given for comparison (Fig. 6N–Q). A schematic representation of the Usp53 expression pattern in the mouse organ of Corti at P4 is given in Figure 6, R and S.
USP53 localizes to TJs in cochlear epithelial cells
To examine the subcellular distribution of USP53, we expressed a cDNA encoding GFP–USP53 fusion protein in hair cells and supporting cells by injectoporation of cochlear explants from P4 wild-type mice following a recently described protocol (Fig. 7A,B; Xiong et al., 2014). We found that GFP–USP53 accumulated at the apical cell borders in transfected hair cells and supporting cells, and its localization overlapped precisely with endogenously expressed TJP2 (Fig. 7C). A z-axis analysis of the vertical distribution of GFP–USP53 showed that the fusion protein specifically localized to the apical-most region (∼1 μm) of hair cells (Fig. 7D) and supporting cells, in which the apical processes of these cells are joined by TJs. To verify whether the association of USP53 with TJs is tissue or cell-type specific, we analyzed its subcellular localization in heterologous cells. We chose Madin-Darby canine kidney (MDCK) cells for this analysis because they form a polarized epithelium with TJs when grown to confluence. In transfected MDCK cells, GFP–USP53, but not GFP alone, prominently colocalized with TJP2 at TJs (Fig. 7E, arrows) and sporadically at intracellular puncta (Fig. 7E, arrowheads), indicating that USP53 and TJP2 might be able to form protein complexes at these subcellular sites in different types of epithelial cells.
To verify whether heterologously expressed USP53 (GFP–USP53) reflects endogenous USP53 distribution in the organ of Corti, we generated rabbit polyclonal antisera (Rb111) against two peptides from the C-terminal tail domain of mouse USP53 (Fig. 5A). The antisera were affinity purified against each peptide separately. In addition, we tested two commercially available antibodies (HPA035844 and HPA035845; Sigma) that recognize different epitopes in the C-terminal tail of the protein (Fig. 5A). The affinity-purified antibodies specifically detected USP53 in extracts from transfected HEK293T cells (Fig. 8A; and data not shown). The antibodies were also effective in immunohistochemistry as they strongly labeled cell borders of cells overexpressing GFP–USP53 in injectoporated cochlear explants (Fig. 8B; and data not shown). Immunolabeling of acutely dissected wild-type organs of Corti revealed a circumferential staining pattern for endogenous USP53 in areas of cell–cell contacts between hair cells and Deiters' cells (Fig. 8C, left, D). No alterations in the distribution of USP53 were observed in heterozygous and homozygous mambo mice (Fig. 8C). Furthermore, no signal was observed with preimmune serum or secondary antibody alone (Fig. 8D; and data not shown). Together, these findings suggest that USP53 can be recruited to TJs in epithelial cells of the inner ear and in heterologous epithelial cells.
Several TJPs have been localized to the inner ear, and a number of them have been implicated in hearing, including members of the claudin family (claudin-9, claudin-11, and claudin-14), occludin, tricellulin, ILDR1, and TJP2 (Fig. 1). Mutations in the respective genes lead to deafness in mice and humans (Wilcox et al., 2001; Ben-Yosef et al., 2003; Gow et al., 2004; Kitajiri et al., 2004, 2014; Nakano et al., 2009; Walsh et al., 2010; Borck et al., 2011; Morozko et al., 2015). To verify whether USP53 acts in a common pathway with any of these proteins, we searched for biochemical interactions. First, we sought to map the domain in USP53 that mediates its association with TJs by expressing various USP53 fragments as GFP fusions in the cochlear sensory epithelium (Fig. 9A). Although full-length USP53 and its C-terminal tail domain localized to cellular junctions, the catalytic domain and GFP alone were distributed evenly in the cytoplasm, indicating that the tail domain controls subcellular localization of USP53. Consistent with these findings, the mambo mutation, which affects the catalytic domain, did not perturb USP53 localization at the junctions. Analysis of the USP53 primary sequence using Scansite 3 (Obenauer et al., 2003) with high-stringency criteria identified a putative binding site for SH3 domains (Fig. 5A). This finding seemed intriguing because of the fact that the major junctional scaffolding proteins TJP1 and TJP2 contain SH3 domains (Jesaitis and Goodenough, 1994). However, disruption of the SH3 binding site did not alter the subcellular localization of USP53 (Fig. 9A), indicating that other elements in the tail domain promote its recruitment to TJs.
USP53 binds TJP1 and TJP2
Based on the expression and distribution of USP53 in cochlear epithelial cells, we next asked whether the DUB is part of the multiprotein complex of TJs. We overexpressed full-length USP53, catalytic domain, C-terminal tail, USP53–mambo, and USP53 lacking the predicted SH3 binding site as GFP fusions in HEK293T cells and performed coimmunoprecipitation assays for endogenously expressed TJPs. Both native TJP1 and TJP2 coprecipitated with wild-type and full-length mutant USP53 proteins, as well as with the C-terminal tail, but not with the catalytic domain or GFP alone. In contrast, the transmembrane TJ protein occludin failed to associate with USP53 (Fig. 9B,C). Coimmunoprecipitation of overexpressed FLAG-tagged TJP2, rather than native TJP2, with GFP–USP53 confirmed the specific interaction between USP53 and TJP2 (Fig. 9D). No interactions were observed between USP53 and occludin, claudin-9, or claudin-14 by coimmunoprecipitation (data not shown), suggesting that recruitment of USP53 to TJs likely occurs through TJP1/2 binding. Serial C-terminal deletion mutants of mouse USP53 were generated to map the TJP2 interaction site to an ∼300 aa region adjacent to the catalytic domain (Fig. 9E). The findings are consistent with the results from injectoporation studies and suggest that USP53 associates with TJ adaptor proteins via its C-terminal domain, possibly regulating their levels or activity.
TJ properties and the EP in mambo mice
The apices of cochlear hair cells and supporting cells in the reticular lamina are joined by a band of TJs that separate the endolymph from the perilymph (Fig. 1) (Gulley and Reese, 1976). Given the junctional localization of USP53, we hypothesized that impaired TJs in the organ of Corti of mambo mice may allow endolymphatic K+ to leak across the reticular lamina, thus reducing the K+ gradient and EP. To explore this possibility, we measured the EP at two time points: (1) during the development of the EP (P10); and (2) after it reached mature levels (P30) (Fig. 10A). At P10, the EP in mambo mice (39 ± 2 mV) was not significantly different from control mice (36 ± 2 mV). However, at P30, the EP increased to >100 mV in control mice, although it remained ∼17% lower in the mutants (p < 0.005, two-tailed t test).
Claudin-11-deficient mice lack the EP because of impaired TJs in the basal cell layer of the stria vascularis, the structure responsible for generating the EP (Gow et al., 2004). By ISH, we did not detect Usp53 expression in the stria vascularis (Fig. 6D,H). However, given the importance of this structure for the generation and maintenance of the EP, we next examined its morphology and function. Hematoxylin–eosin staining of inner ear sections revealed no degenerative changes in the stria vascularis of 3-month-old mambo mice (Fig. 10B,C). To analyze the paracellular permeability properties of the stria vascularis, a rhodamine-labeled biotin tracer was injected into the round and oval windows of heterozygous and homozygous mambo mice at P30, a time point at which a reduced EP was observed. The biotin-based tracer was not detected in the intrastrial space of homozygous mambo mice and heterozygous siblings, indicating that the tracer did not penetrate beyond the TJ barrier in the basal cell layer of the stria vacularis (Fig. 10D–G, arrows). Consistent with the lack of Usp53 expression in the stria vascularis, these data suggest that the paracellular permeability properties of the stria vascularis are largely unaffected in mambo mice.
Therefore, we analyzed the structure of TJs in the reticular lamina of the organ of Corti using thin-section TEM. At P10, the morphology of TJs between OHCs and Deiters' cells appeared normal at the ultrastructural level (Fig. 11A–D). Kissing points, in which TJ strands between adjacent cells cause occlusion of plasma membrane, could be observed in both mambo and control mice (Fig. 11C,D, arrows). Furthermore, immunostaining of cochlear whole mounts revealed no alterations in the distribution of the TJ proteins TJP1 and TJP2 in P6 mambo mice (Fig. 11E–H). These findings suggest that the apical membranes of hair cells and supporting cells are properly fused together by continuous TJ strands.
An impairment of TJ barrier function and increased paracellular flux of solutes, particularly K+, has been associated with OHC loss, although no morphological alterations in TJs have been reported (Ben-Yosef et al., 2003; Nakano et al., 2009; Kitajiri et al., 2014). To examine whether changes in the extracellular environment contribute to OHC degeneration in mambo mice, we assessed the degree of OHC loss in explant cultures of the organ of Corti in a low K+ milieu (4.16 mm K+). Organs of Corti from P7 homozygous mambo mice and heterozygous littermates were cultured for 3 d ex vivo, followed by immunostaining with TJP1 and calbindin antibodies to label TJs and hair cells, respectively (Fig. 11I–L). For comparison, acutely dissected organs of Corti from 10-d-old heterozygous and homozygous mambo mice were stained with the same antibodies (Fig. 11M–P). Significant loss of OHCs was observed in vivo in homozygous mutants (Fig. 11N,P) but not in heterozygous littermates (Fig. 11M,O). In addition, signs of OHC degeneration, including fusion of stereocilia (Fig. 11N, arrow), submerging of OHC bodies under the reticular lamina, and expansion of the phalangeal processes of Deiters' cells, were evident (Fig. 11N,P, arrowheads). In contrast, OHCs in mutant organs of Corti survived when cultured ex vivo, suggesting that hair cell loss in mambo mice is triggered by changes in their microenvironment rather than intrinsic factors (Fig. 11I–L).
Heterozygous mambo mice show increased susceptibility to noise-induced hearing loss
Most epithelial cell sheets are sealed by the apical junctional complexes composed of TJs that form a permeability barrier for solutes and adjacent adherens junctions (AJs) that provide mechanical stability to tissues (Farquhar and Palade, 1963; Aijaz et al., 2006). The junctions between OHCs and Deiters' cells are unique in that they combine ultrastructural and molecular features of TJs and AJs and are referred to as tight adherens junctions (TAJs; Nunes et al., 2006). These combined features may allow TAJs to serve as an effective barrier to solute diffusion while withstanding the mechanical forces imposed by sound-induced vibrations and OHC electromotility. In support of such a dual role for TAJs, targeted deletion of the TAJ protein vezatin in OHCs renders mice more susceptible to noise-induced hearing loss (NIHL; Bahloul et al., 2009). Therefore, we reasoned that a similar impairment of TAJs may be present in Usp53 mutant mice.
The early-onset hearing loss in homozygous mambo mice precludes a meaningful study of noise-induced ABR threshold changes. Therefore, we challenged heterozygous animals, which do not show any hearing impairment, with an acoustic trauma. We exposed 7-week-old heterozygous mambo mice and age-matched wild-type mice to continuous broadband noise (8–16 kHz) at 98 dB SPL for 2 h and determined ABR thresholds 2 d before and 2 weeks after noise exposure (Fig. 12A). ABR thresholds were increased in both wild-type and heterozygous mambo mice in the post-exposure group, predominantly at high frequencies (Fig. 12B). However, threshold shifts were significantly larger in heterozygous mutants (50 ± 6 dB at 28.3 kHz) than in wild-type controls (23 ± 6 dB at 28.3 kHz; Fig. 12C). This indicates that one mutant allele of Usp53 is sufficient to undermine the structural and functional integrity of the cochlear sensory epithelium, possibly by affecting the mechanical properties of TAJs.
Discussion
Over the past few years, several ubiquitin ligases have been implicated in inner ear function and cochlear homeostasis. Fbxo2, a ubiquitin ligase F-box protein, interacts with connexin 26, a gap junction protein involved in potassium recycling, and regulates its levels in the cochlea (Nelson et al., 2007). Fbxo2 null mice develop age-related hearing loss. Other E3 ubiquitin ligases, such as Nedd4–2, Rma1, and UBR1, have been implicated in the pathogenesis of syndromic deafness, including Bartter's syndrome, Pendred syndrome, and Johanson–Blizzard syndrome, respectively (Embark et al., 2004; Zenker et al., 2005; Lee et al., 2012). However, with the exception of a single study describing an association between variants in the Usp31 gene and hearing loss in dogs (Yokoyama et al., 2012), there have been no reports regarding the involvement of DUBs in auditory functions. Here we provide evidence that USP53 is a novel component of TJs in cochlear epithelial cells and affects the survival of OHCs at the onset of hearing. First, progressive hearing loss in USP53-deficient mambo mice is preceded by degenerative changes in OHCs, starting at approximately P8. Although Usp53 mRNA is distributed broadly throughout the cochlea, a feature shared with other deafness-related TJ genes (Borck et al., 2011), OHCs are selectively affected. Second, USP53 is localized to TJs in a broad range of cochlear epithelial cells, including hair cells, and binds to the TJ adaptor proteins TJP1 and TJP2. Third, OHCs evade degeneration in cochlear organotypic culture in a low K+ milieu. Finally, the resistance of the organ of Corti to noise-induced damage is compromised in Usp53 heterozygous mutant mice. We conclude that USP53 associates with TJ scaffolding proteins and is likely critical for maintaining the functional integrity of TJs, which protect OHCs from toxic extracellular factors and mechanical damage.
The onset of OHC defects in mambo mice coincides with the initial generation and rise of the EP. We show that mambo mice are initially capable of generating a normal EP, which is consistent with the lack of Usp53 expression in the stria vascularis. Although the EP is reduced by 20% in adult mambo mice, the paracellular permeability of the stria vascularis is unaltered. This raises the possibility that the barrier function of epithelial cells lining the scala media is compromised. The EP decline is unlikely to reflect focal fluid leakage through lesions created in the reticular lamina by degenerating OHCs because the formation of gaps is typically prevented by concomitant expansion of the phalangeal processes of Deiters' cells (Forge, 1985; Raphael and Altschuler, 1991a,b; Raphael, 1993; Leonova and Raphael, 1997). Accordingly, our SEM analysis did not reveal any perforations of the reticular lamina in mambo mice (data not shown). On balance, the evidence suggests that subtle aberrations in TJs of the organ of Corti may lead to EP decline.
A reduction of the EP and hearing loss have also been reported in claudin-11 null mice, a phenotype that has been attributed to defective TJs in the stria vascularis (Gow et al., 2004; Kitajiri et al., 2004). Normal EPs have been reported in tricellulin, claudin-9, and claudin-14 mutant mice (Ben-Yosef et al., 2003; Nakano et al., 2009; Nayak et al., 2013), although OHC loss in these mouse models was attributed to the leakage of endolymph across the reticular lamina. It has been suggested that spatially restricted TJ leakage and a compensatory increase in K+ secretion by the stria vascularis may account for these findings (Ben-Yosef et al., 2003; Nakano et al., 2009). The small EP decline in mambo mice could be explained by the broader expression pattern of Usp53 or a more deleterious effect of USP53 deficiency on TJ permeability. An accumulation of K+ in the space of Nuel has been shown to cause prolonged depolarization and longitudinal contractions of OHCs (Zenner, 1986), which may lead to their degeneration. Our observation that OHCs of mambo mice do not degenerate ex vivo in low K+ medium is consistent with this possible mechanism.
Another principal finding is the increased susceptibility of heterozygous mambo mice to NIHL. This result can perhaps be best explained by a dual mechanism, in which USP53 influences both the ion barrier function of TAJs and their resistance to mechanical stress. Although one functional allele of Usp53 is sufficient to protect the epithelium under limited noise conditions, TAJs might be weakened and more prone to noise-induced damage. To our knowledge, this is the first report of a gene encoding a TJP that promotes increased susceptibility to NIHL in heterozygous mice. Targeted homozygous deletion of another TAJ component, vezatin, in OHCs increases the susceptibility of mice to acoustic trauma. Vezatin may bridge the junctional complex to the F-actin-based cell cortex via its interaction with radixin (Bahloul et al., 2009). In analogy, USP53 binds to TJP1 and TJP2, which form a link between transmembrane TJ proteins and the cytoskeleton (Fanning et al., 1998, 2002; Wittchen et al., 1999), suggesting that a similar mechanism may render mambo mice more susceptible to noise injury. It is tempting to speculate that the increased permeability and decreased mechanical stability of TAJs may be directly related: because of the extremely low endolymphatic Ca2+ concentration (20–100 μm; Ikeda and Morizono, 1988), an increase in TJ permeability may expose AJs to the low endolymphatic Ca2+ environment. This, in turn, might affect trans-interactions between E-cadherins, which require Ca2+ concentrations as high as 1 mm (Pertz et al., 1999), and alter the mechanical properties of AJs.
Our data show that USP53 binds to the TJ adaptor proteins TJP1 and TJP2 but not the TJ transmembrane proteins occludin, claudin-9, or claudin-14. The association of USP53 with TJPs poses an interpretational challenge for the observed OHC phenotype in mambo mice. Aside from their role in regulating TJ assembly and stability, TJPs translocate to the nucleus to regulate gene expression (Bauer et al., 2010). Indeed, overexpression of TJP2 and altered expression of apoptosis-related genes may contribute to hearing loss in DFNA51 patients (Walsh et al., 2010). Although we cannot exclude the involvement of intrinsic cellular mechanisms in the pathophysiology of mambo mice, the mitochondrial apoptotic pathway is unlikely to underlie OHC death because no caspase-3 activity was detected in degenerating OHCs. Moreover, OHC degeneration is rescued in low K+ medium, suggesting that extracellular factors trigger hair cell death. In addition, the decreased EP suggests that increased TJ permeability may allow the diffusion of K+ and other extracellular factors across the reticular lamina leading to the described toxic effects of high extracellular K+ on OHCs.
USP53 shows no catalytic activity in DUB inhibitor assays, which is consistent with the lack of an essential histidine residue in its active site. To our knowledge, this is the first report of an animal model of a human disorder caused by a mutation in an inactive DUB. Only 7 of the 83 members of the DUB superfamily are predicted to be inactive, including six of the USP family (Komander et al., 2009). Some inactive DUBs, such as Sad1, a yeast homolog of USP39, USPL1, and USP52, do not bind ubiquitin and have been implicated in diverse physiological processes (van Leuken et al., 2008; Schulz et al., 2012; Hadjivassiliou et al., 2014; Jonas et al., 2014). The physiological roles and ubiquitin-binding properties of USP50 and USP54 are unknown. It is possible that the ubiquitin-binding capability of USP53 has been preserved.
Three TJ proteins necessary for hearing, occludin, tricellulin, and ZO-2, are known targets of ubiquitination (Wagner et al., 2011), and ubiquitination of occludin regulates TJ permeability (Murakami et al., 2009; Rao, 2009). Furthermore, levels of TJP2 are controlled by the proteasome (Quiros et al., 2013) and augmented in DFNA51 patients (Walsh et al., 2010). It is tempting to speculate that binding of USP53 to TJP2 may interfere with the function of active DUBs in a dominant-negative manner by blocking the removal of ubiquitin from TJ proteins. In mambo mice, the mutant USP53 catalytic domain may fail to bind ubiquitin on target proteins and shift the dynamic balance between ubiquitination and deubiquitination, thus influencing stability or activity of TJ proteins. Alternatively, USP53 may target ubiquitin-like modifiers, such as NEDD8 or ISG15, as reported for other USPs (Gong et al., 2000; Catic et al., 2007), or it may have acquired a protein–protein interaction function, in analogy to the inactive catalytic domain of PRPF8 (Pena et al., 2007). It will be important to analyze the specificity of USP53 for ubiquitin and ubiquitin-like modifiers and identify their target TJ proteins.
Although the precise mechanism by which USP53 regulates cochlear homeostasis remains to be elucidated, the results of our study establish USP53 as a novel TJ component necessary for OHC survival and for the integrity of the auditory system. Our findings also highlight the efficiency of forward genetics approaches in identifying genes that are likely relevant to progressive hearing loss in humans.
Footnotes
This work was supported in part by National Institute on Deafness and Other Communication Disorders (NICDC) Grant DC013331 (M.S.). The generation of the mambo mice and the identification of the affected gene was performed with funding from NIDCD Grant DC014713 and DC007704 to Ulrich Mueller (The Scripps Research Institute, La Jolla, CA). We thank Rajvi Shah and other members of the Schwander laboratory for helpful discussions, Wei Xiong for technical assistance with injectoporation assays, and Bechara Kachar for providing us with expression constructs for occludin and claudins.
The authors declare no competing financial interests.
- Correspondence should be addressed to Martin Schwander, Rutgers, The State University of New Jersey, 604 Allison Road, Piscataway, NJ 08854. schwander{at}biology.rutgers.edu