Tmprss3, a Transmembrane Serine Protease Deficient in Human DFNB8/10 Deafness, Is Critical for Cochlear Hair Cell Survival at the Onset of Hearing

Mutations in the type II transmembrane serine protease 3 ( TMPRSS3 ) gene cause non-syndromic autosomal recessive deafness (DFNB8/10), characterized by congenital or childhood onset bilateral profound hearing loss. In order to explore the physiopathology of TMPRSS3 related deafness, we have generated an ethyl-nitrosourea-induced mutant mouse carrying a protein-truncating nonsense mutation in Tmprss3 (Y260X) and characterized the functional and histological consequences of Tmprss3 deficiency. Auditory brainstem response revealed that wild type and heterozygous mice have normal hearing thresholds up to 5 months of age, whereas Tmprss3 Y260X homozygous mutant mice exhibit severe deafness. Histological examination showed degeneration of the organ of Corti in adult mutant mice. Cochlear hair cell degeneration starts at the onset of hearing, postnatal day 12, in the basal turn and progresses very rapidly toward the apex, reaching completion within 2 days. Given that auditory and vestibular deficits often co-exist, we evaluated the balancing abilities of Tmprss3 Y260X mice by using rotating rod and vestibular behavioral tests. Tmprss3 Y260X mice effectively displayed mild vestibular syndrome that correlated histologi-cally with a slow degeneration of saccular hair cells. In situ hybridization in the developing inner ear showed that Tmprss3 is in sensory hair cells in the cochlea and vestibule. Our results show that Tmprss3 acts as a permissive factor for cochlear hair cells survival and activation at the onset of hearing and is required for saccular hair cell survival. This mouse model will certainly help to decipher the molecular mechanisms underlying DFNB8/10 deafness and cochlear function.

Mutations in the type II transmembrane serine protease 3 (TMPRSS3) gene cause non-syndromic autosomal recessive deafness (DFNB8/10), characterized by congenital or childhood onset bilateral profound hearing loss. In order to explore the physiopathology of TMPRSS3 related deafness, we have generated an ethyl-nitrosourea-induced mutant mouse carrying a protein-truncating nonsense mutation in Tmprss3 (Y260X) and characterized the functional and histological consequences of Tmprss3 deficiency. Auditory brainstem response revealed that wild type and heterozygous mice have normal hearing thresholds up to 5 months of age, whereas Tmprss3 Y260X homozygous mutant mice exhibit severe deafness. Histological examination showed degeneration of the organ of Corti in adult mutant mice. Cochlear hair cell degeneration starts at the onset of hearing, postnatal day 12, in the basal turn and progresses very rapidly toward the apex, reaching completion within 2 days. Given that auditory and vestibular deficits often co-exist, we evaluated the balancing abilities of Tmprss3 Y260X mice by using rotating rod and vestibular behavioral tests. Tmprss3 Y260X mice effectively displayed mild vestibular syndrome that correlated histologically with a slow degeneration of saccular hair cells. In situ hybridization in the developing inner ear showed that Tmprss3 mRNA is localized in sensory hair cells in the cochlea and the vestibule. Our results show that Tmprss3 acts as a permissive factor for cochlear hair cells survival and activation at the onset of hearing and is required for saccular hair cell survival. This mouse model will certainly help to decipher the molecular mechanisms underlying DFNB8/10 deafness and cochlear function.
Hearing impairment is the most common sensory defect in humans. Approximately 1 in 1000 newborns is born deaf, and ϳ50% of these cases of congenital deafness are genetic. Recently, mutations in the TMPRSS3 gene on chromosome 21q22.3 were shown to cause human autosomal recessive nonsyndromic hearing loss (DFNB8/10) (1). TMPRSS3-related deafness is characterized by bilateral, severe to profound hearing loss, with no described middle ear or vestibular deficits. Environmental and/or genetic modifiers may influence the expressivity of hearing loss caused by TMPRSS3 mutations. For instance, missense mutation P404L, altering a highly conserved amino acid of the TMPRSS3 serine protease domain, causes congenital hearing loss in a Tunisian family (2), whereas in a family from Turkey, it leads to childhood onset (6 -7 years of age) deafness (2,3). Clinical presentation of hearing loss resulting from mutations in TMPRSS3 is indistinguishable from the other forms of non-syndromic recessive non-syndromic hearing loss, and its diagnosis is therefore based on both clinical and molecular evaluation.
TMPRSS3 is a type II transmembrane serine protease, structurally defined by a transmembrane domain located near the N terminus; a low density lipoprotein receptor A domain, which binds calcium (4) and low density lipoprotein (5); a scavenger receptor cysteine-rich domain that is involved in protein-protein interaction (6); and a C-terminal serine protease domain from the S1 family of the SA clan of serine-type peptidases for which the prototype is chymotrypsin (7). Sixteen different TMPRSS3 mutations that lie in all functional domains have been described to date and were found to disrupt the proteolytic activity of TMPRSS3 (8).
RT-PCR and RNA in situ hybridization experiments revealed significant Tmprss3 expression in the thymus, stomach, testis, and embryonic day 19 embryos and in specific cochlear tissues, including the stria vascularis, spiral ganglion neurons, modiolus, organ of Corti, and some cells of the Kölliker's organ (9). It was hypothesized that Tmprss3 may participate in the regulation of sodium homeostasis through its ability to activate the inner ear-expressed sodium channel (ENaC) in vitro (9). However, this hypothesis has been challenged by the fact that pseudohypoaldosteronism patients mutated in the ENaC gene show normal hearing thresholds (10). Therefore, the genuine role of Tmprss3 in the auditory system is currently unknown, and alternative pathways should be explored in vivo. Here, we report on the cellular, molecular, and phenotypic characterization of a mouse model for Tmprss3-related deafness.

Animal Handling
The care and use of animals followed the animal welfare guidelines of INSERM, under the approval of the French "Ministère de l'Alimentation, de l'Agriculture et de la Pêche". All efforts were made to minimize the number of animals used and their suffering.

Generation of Tmprss3 Mutant Mice; the Tmprss3 Y260X Protein-truncating Mutation
Ingenium Pharmaceuticals AG has developed a large library of cryopreserved mutant mouse sperm derived from a chemical mutagenesis program, in which male C3HeB/FeJ mice were mutagenized with ethyl-nitrosourea. The screen for deleterious mutations in the Tmprss3 gene was performed in G1 animals that were derived from the mating of mutagenenized G0 males with untreated C3HeB/FeJ females. A mutation search was performed using heteroduplex formation analysis by temperature gradient capillary electrophoresis.

Preparation of Mouse Inner Ear Extracts
Mouse cochleae were isolated, ground in a mortar, and homogenized manually in a lysis buffer containing antiproteases (Roche Applied Science). The lysis buffer contains 150 mM NaCl, 1% SDS, 1% PMSF, 1 mM EDTA, 10 mM Tris-HCl, pH 7.4. After centrifugation at 10,000 ϫ g for 5 min, the supernatants were used for Western blotting.

In Situ Hybridization
Total RNA was isolated from the inner ear of wild type mice and reverse transcribed. The fragment was PCR-amplified and subcloned into the pCR4Blunt-TOPO (Invitrogen) with the following primers: Tmprss3his-S2 (5Ј-GCACAGAAATCAT-GGGCTCACG-3Ј) and Tmprss3his-AS (5Ј-TAGTCATCTC-TTACTGTGACTAG-3Ј). Digoxigenin-labeled probes were synthesized using the Roche Applied Science digoxigenin labeling kit according to the manufacturer's instructions. The probe for peripherin was kindly provided by Dr. M. M. Portier. In situ hybridization was carried out as described (11).

Assessment of Hearing Function
Animals were anesthetized by an intraperitoneal injection of a mixture of Rompun 2% (3 mg/kg) and Zoletyl 50 (40 mg/kg). Both pinna were severed to facilitate the insertion and placement of the distortion product otoacoustic emission (DPOAE) 5 probe.
Distortion Product Otoacoustic Emission Recordings-DPOAEs were recorded in the external auditory canal using an ER-10C S/N 2525 probe (Etymotic Research Inc., Elk Grove Village, IL) and consisted of two emitters and one microphone. The two primary tones were generated and the distortion was processed by the Cubdis system HID 40133DP (Mimosa Acoustics Inc., Champaign, IL). The probe was self-calibrated for the two stimulating tones before each recording. The two tones were presented simultaneously, sweeping f2 from 0.5 kHz to 20 kHz by quarter-octave steps and maintaining the f2/f1 ratio constant at 1.2. The primary intensities of f2 and f1 were set at 60 and 55 dB sound pressure level (SPL) (reference 2 ϫ 10 Ϫ5 pascals), respectively. For each frequency, the cubic distortion product (2f1 Ϫ f2) and the neighboring noise magnitudes were measured and expressed as a function of f2. Data are expressed as means Ϯ S.E.
Auditory Brainstem Response Recordings-The acoustical stimuli were generated by an arbitrary function generator (type 9100R, LeCroy Corp.), consisting of 9-ms tone bursts with a 1-ms rise and fall time delivered at a rate of 10/s. Tone bursts were passed through a programmable attenuator and delivered to the animal by a JBL 075 loudspeaker (James B. Lansing Sound, Inc., Stamford, CT) in a calibrated free field. The audi-tory brainstem responses (ABRs) were derived from two needle electrodes respectively placed at the vertex and on the mastoid of the stimulated ear. The signal was differentially amplified (ϫ10,000) by a Tektronix (TM 503) differential amplifier and digitized (50 kHz sampling rate, with a 12-bit dynamic range and 1024 samples/burst), averaged 512 times, and stored on a computer (Dell Dimensions, Dell Inc., Round Rock, TX). Intensity-amplitude functions of brainstem responses were obtained at each frequency tested (2,4,6,8,10,12,16,20,26,and 32 kHz) by varying the intensities of the tone bursts from 0 to 90 dB SPL, in 5-dB steps. The ABR thresholds were defined as the minimum sound intensity necessary to elicit a clearly distinguishable response.
Endocochlear Potential Measurement-After a ventrolateral approach to the cochlea, the bone over the basal turn scala media was shaved to create a small fenestra through the thinned bone. A glass microelectrode filled with 0.15 M KCl and connected to a direct current amplifier (model KS-700, World Precision Instruments) was passed through the fenestra and into the scala media to record the endocochlear potential. A silversilver chloride reference wire was placed in the animal's neck musculature.

Ultrastructural Evaluation of the Tmprss3 Y260X Mouse Cochlea
Animals were decapitated under deep anesthesia, and their cochleae were prepared using our scanning electron microscopy (SEM) and transmission electron microscopy (TEM) standard protocols (see Ref. 12). For SEM, a total number of 39 mice (1 ear/animal), including eight Tmprss3 WT and 31 Tmprss3 Y260X homozygous mice (see Table 1 for details), were used. The surface of the organs of Corti was observed, from the cochlear base to apex, using a Hitachi S4000 electron microscope. For TEM, a total number of 16 mice, including 14 Tmprss3 Y260X homozygous mice (1 ear/animal taken at 10 days (n ϭ 1), 11 days (n ϭ 5), 12 days (n ϭ 5), 14 days (n ϭ 1), 19 days (n ϭ 1), and 4 months (n ϭ 1)) and two Tmprss3 WT mice (11 days (n ϭ 1) and 19 days (n ϭ 1)) were used. Ultrathin radial sections of the organ of Corti, taken at different levels along the cochlear spiral, were observed using a Hitachi 7100 electron microscope. In addition, the cristae ampullares, utricules, and saccules from four 4-month-old Tmprss3 Y260X homozygous mice were also investigated using TEM. The samples were processed and studied under the same conditions as the cochleae.

Sensory Hair Cell Counting
Quantitative evaluation of sensory hair cell loss was performed using SEM. The mouse cochlea length has been found to average 6 mm (13), including the basal (3 mm), the middle (2 mm), and the apical (1 mm) coils. The number of missing hair cells was counted in six segments of 1 mm each normally containing a mean number of 110 inner hair cells (IHCs) and 125 ϫ 3 outer hair cells (OHCs) per mm. Hair cells were counted as absent if the stereociliary bundles and cuticular plates were missing. Cytocochleograms represent the percentage of missing hair cells in the three rows of OHCs and in the row of IHCs in the basal, middle, and apical regions.

Assessment of Balance Function
Rotarod-The balance and motor coordination of Tmprss3 WT and Tmprss3 Y260X homozygous mice were assessed on a rotarod apparatus (Bioseb, Chaville, France). Ten adult mice of each genotype were first given a pretraining trial on day 1 in order to familiarize them with the rotating rod. They were trained to stay on the rod first at the lowest speed (4 rpm) and progressively at higher speeds (up to 10 rpm). They returned to the rotating rod following each fall (total of 4 -5 trials on day 1). Testing of the mice occurred on days 2, 3, and 4 at fixed speed (10 rpm), and latency to fall was measured for a maximum of 180 s. Each mouse underwent 4 trials/day; the first one was considered as training trial and the second, third, and fourth ones were recorded. For each day, the data were averaged for each mouse (values from the third and fourth trials) and then averaged for each group (Tmprss3 WT and Tmprss3 Y260X homozygous mice).
Behavioral Experiments-The vestibular rating score was estimated as described previously (14,15). Tmprss3 WT and Tmprss3 Y260X mice (n ϭ 10 each) were each scored from 0 to 4, corresponding to normal behavior to maximal vestibular deficit. A score of 0 means that behavior is normal; a score of 1 means that the behavior is abnormal, but no specific vestibular deficit is effectively determined; a score of 2 corresponds to an identified but slight vestibular deficit; a score of 3 describes an identified and evident deficit; and a score of 4 means that vestibular deficit is maximal. Seven different tests were sequentially scored and totaled to rate the vestibular deficit as follows. 1) Head bobbing was tested, when abnormal intermittent backward extension of the neck was observed. 2) Circling stereotyped movement was tested, ranging from none to compulsive circles around the animal's hips. 3) Retropulsion, a typical backward walk reflecting vestibular disturbance, was tested. 4) The tail-hanging reflex, which normally induces a normal forelimb extension to reach the ground, results in the ventral bent of the body and grip of the tail when the vestibular deficit is maximal. 5) The contact-inhibition reflex normally leads the animal to hold on to a metal grid in a supine position to return when its back touches the ground. In the case of vestibular deficit with a lack in the body orientation referential, this reflex is abolished, and the animal continues gripping the grid in a supine position. 6) The air-righting reflex is necessary for animals to land on their feet when they fall from a supine position; vestibular dysfunction impairs this normal reversal, and a maximal deficit leads the animal to land on its back when dropped from a height of 40 cm onto a foam cushion. 7) Swimming was tested, ranging from a normal swimming behavior to drowning due to loss of all proprioceptive clues. Scores for all of these tests were totaled to obtain the final vestibular deficit score.

Tmprss3 Is Critical for Cochlear Hair Cell Survival
Morphometric Analysis-Hair bundles stained for actin were quantified in the stacks of saccules from newborn and young animals using Metamorph imaging software. The number of hair bundles was manually counted and normalized by the area monitored to obtain the density of hair bundles per 100 m 2 of saccular epithelium.

Statistical Analysis
For the assessment of hearing function, all data are expressed as means Ϯ S.E. To evaluate the significance of the functional difference between Tmprss3 WT and Tmprss3 Y260X mice, statistical tests were performed on Sigmaplot 2000 for Windows (version 6.1). All comparisons between means were performed using Student's paired two-tail t tests. The data values are expressed as means Ϯ S.E. Density of hair bundles between developing and adult animals was compared using analysis of variance followed by Tukey's test. For vestibular deficit, behavioral score were compared between Tmprss3 WT and Tmprss3 Y260X homozygous mice using the Mann-Whitney test. For the rotarod experiment, one sample t test was performed.

Cloning of Tmprss3 Long Isoform and Expression in HeLa Cells
Total RNA was isolated from the inner ear of wild type mice and reverse transcribed. The coding regions of the Tmprss3 long isoform were PCR-amplified and subcloned into the EYFP-N1 vector (Clontech, Mountain View, CA) using TMPRSS3lSal1-S (5Ј-TTAGTCGACATGGCCGCTTCAGA-AATGGTGGAG-3Ј) and TMPRSS3lBamH1-AS (5Ј-TTAGG-ATCCAAAGTCTTCAGATCTCTCTCCAACTG-3Ј). Plasmid containing the Tmprss3 short isoform tagged with V5 was described previously (8). Plasmids were transfected into HeLa cells using the Lipofectamine LTX Plus transfection reagent (Invitrogen). The endoplasmic reticulum was labeled with a rabbit anti-calreticulin polyclonal antibody (Stressgen, Ann Arbor, MI) at 1:800. V5 epitope was revealed with a mouse anti-V5 monoclonal antibody (Invitrogen) at 1:1000.

Spiral Ganglion Neuron Counting
Cryostat sections of 12 m were taken throughout the entire cochleae of Tmprss3 WT (n ϭ 10) and Tmprss3 Y260X homozygous (n ϭ 10) mice. To better visualize neurons in the spiral ganglion, the sections were immunolabeled using ␤ III tubulin (type I spiral ganglion neurons, Monoclonal antibody, 1:3000, Covance, Emeryville, CA) and peripherin (type II spiral ganglion neurons, rabbit anti-peripherin polyclonal antibody, 1:400, Chemicon International, Temecula, CA) and visualized using a Zeiss confocal microscope. To avoid counting the same neurons twice, only one section of entire cochlea was used for quantification using ImageJ software. Quantitative data were expressed as the mean number of neurons per 3600 m 2 Ϯ S.E. Student's t test was used for statistical analysis of the data (p Ͻ 0.01).

Generation of Tmprss3 Mutant Mice; the Tmprss3 Y260X Protein-truncating Mutation
The Tmprss3 gene, which maps on mouse chromosome 17, is composed of 12 coding exons. Mutation screening of the Tmprss3 gene in the Ingenium Pharmaceuticals G1 mouse DNA archive revealed a T to A substitution in exon 7 (at a position 780 bp from the start codon, NM_024022.2), resulting in a nonsense mutation at amino acid 260 (Y260X) (Fig. 1A). This mutation would result in the production of a 194-amino acid truncated protein, deleted of most of its protease domain (Fig. 1B).
To evaluate the effect of the Tmprss3 Y260X mutation on RNA processing, total RNA was extracted from various tissues, including inner ear, brain, intestine, skeletal muscle, liver, and kidney of three 4-week-old Tmprss3 Y260X heterozygous mice. Sequencing chromatograms obtained from RT-PCR products encompassing the Y260X mutation showed clear heterozygosity at the mutated position, suggesting that both Tmprss3 WT and Tmprss3 Y260X alleles were expressed in all tissues tested (data not shown). Expression of the Tmprss3 WT and Tmprss3 Y260X homozygous mutant protein was studied by Western blots on protein extracts from P60 (Fig. 1C) and P10 (Fig. 1D) Tmprss3 WT and Tmprss3 Y260X homozygous mutant cochleae and HeLa cells transfected with Tmprss3. Several Tmprss3 antibodies were tested, but none proved to specifically detect Tmprss3. The best results were obtained with an antibody raised against the C terminus of Tmprss3 protein and already described by Guipponi et al. (9). Although this antibody showed nonspecific binding, the additional bands did not interfere with the detection of Tmprss3. The Tmprss3-Cter antibody revealed a strong 50 kDa band in wild type cochleae and in transfected HeLa cells (star). As expected, the 50 kDa band was absent in Tmprss3 Y260X homozygous mutant cochleae because the epitope used for the C-terminal antibody generation is no longer present in Tmprss3 Y260X homozygous mutant mice. However, due to the lack of specific N-terminal antibody, this experiment does not allow us to determine if the truncated protein is present or absent. Moreover, because the Tmprss3-Cter antibody cross-reacts with other proteins, we were not able to unequivocally identify the inner ear cells expressing Tmprss3 by immunohistochemistry. Finally, Tmprss3 Y260X homozygous mutant were born at the expected Mendelian ratio in matings between heterozygous animals and resembled the wild type and heterozygous littermates in viability, growth rate, fertility, and gait.

Tmprss3 Y260X Mutant Mice Are Completely Deaf
Auditory function was tested in 60-day postnatal mice by recording sound-evoked ABRs. Wild type Tmprss3 WT (n ϭ 10) and Tmprss3 Y260X heterozygous mice (n ϭ 10) showed the classical ABR waveforms (I-IV) in responses to all sound stimula-tion from 2 to 32 kHz with thresholds between 20 and 45 dB SPL. In contrast, the Tmprss3 Y260X homozygous (n ϭ 10) littermates showed no visible ABRs, even for stimulus intensities above 90 dB SPL ( Fig. 2A). This analysis was completed by DPOAE recordings, which reflect the activity of OHCs. Consistent with ABR recordings, DPOAEs were not significantly different in wild type Tmprss3 WT (n ϭ 10) and Tmprss3 Y260X heterozygous mice (n ϭ 10), albeit it was absent in Tmprss3 Y260X homozygous (n ϭ 10) littermates at all stimulus frequencies and intensities studied (Fig. 2B).

Tmprss3 Y260X Mutant Mice Display Vestibular Deficits
Several tests aimed at assessing the vestibular functions were performed on Tmprss3 Y260X homozygous mice (4 months and n ϭ 10). Fixed speed rotarod performance, measured as latency to fall, was significantly reduced for mutant mice when compared with wild type littermates (Fig. 2C). To further refine the vestibular defects, Tmprss3 mutants were subjected to seven This results in a truncated protein without catalytic activity. Shown are Western blots using the Tmprss3-Cter (C and D) antibody on mouse Tmprss3 WT and Tmprss3 Y260X homozygous cochleae at P60 (C) and P10 (D) and on non-transfected (NT) and transfected (T) HeLa homogenates. The Tmprss3-Cter antibody detects a strong overexpressed 50 kDa band (star) in the transfected HeLa homogenate with a faint signal in non-transfected HeLa cells, suggesting that Tmprss3 might be faintly expressed in HeLa cells. This band, corresponding to Tmprss3, is also detected in Tmprss3 WT cochlea extracts and is absent in Tmprss3 Y260X homozygous cochlea at P60 and P10. In all homogenates, nonspecific bands are also observed. MAY 13, 2011 • VOLUME 286 • NUMBER 19 JOURNAL OF BIOLOGICAL CHEMISTRY 17387 behavioral tests, including the swimming, head bobbing, circling, retropulsion, tail-hang reflex, contact inhibition of the righting reflex, and air-righting reflex tests. These tests showed that Tmprss3 Y260X homozygous mice, unlike wild type littermates, have an increased latency to return on the contact inhibition reflex, a tendency to swim on the side, and a tendency to occipital landing when landing at the tail-hanging reflex (Fig. 2D). Collectively, these tests showed that Tmprss3 Y260X homozygous mice present a mild but still significant vestibular deficit when compared with their wild type littermates. Overall, these results indicate that the Tmprss3 Y260X mutation leads in mice to autosomal recessively inherited early onset syndromic deafness.

Deafness Is Due to a Rapid and Massive Degeneration of Hair Cells at the Onset of Hearing
SEM Analysis-The organ of Corti of Tmprss3 Y260X homozygous mice (Fig. 3, A-E) was investigated using SEM and compared with the organ of Corti of Tmprss3 WT mice (Fig. 3, F-J), from birth to 4 months of age. Both groups of mice essentially showed the same features until 11 days. Then hair cells rapidly disappear in the Tmprss3 Y260X homozygous mice, whereas they were preserved in the wild type animals.
In the Tmprss3 Y260X homozygous mice, at birth, the IHCs and the OHCs showed immature stereociliary bundles all along the cochlea (data not shown). At postnatal day 7 (P7), the stereociliary bundles were well developed on both types of hair cells (Fig. 3A). At P11 (Fig. 3B), the stereociliary bundles showed an adult-like appearance (almost linear on IHC and W-shaped on OHC). In addition, the thick kinocilium that accompanied the stereociliary bundles during the period of hair cell development had completely regressed on the OHCs. The unusual presence of one or two additional thin and long kinocilia on the OHCs and the IHCs was the only abnormal feature FIGURE 2. Assessment of hearing and balance impairments in Tmprss3 Y260X mice. A, mean ABR thresholds (mean Ϯ S.E.) for P60 Tmprss3 WT (blue circles; n ϭ 10), P60 Tmprss3 Y260X heterozygous (green circles; n ϭ 10) and P60 homozygous mutant (red circles; n ϭ 10) mice. All Tmprss3 Y260X homozygous mice displayed no ABR wave, even at the highest intensity tested (100 dB). This is symbolized by an upward arrow on top of the red open circle. Inset, representative ABR recordings from Tmprss3 WT (left) and Tmprss3 Y260X homozygous (right) P60 mice at 20 -100 dB SPL of pure sound at 8 kHz. No ABR waveform is visible in the Tmprss3 Y260X homozygous mouse recordings. B, recordings of distortion product otoacoustic emissions (2f1 Ϫ f2) reflect the cochlear nonlinearity as a result of outer hair cell function. P60 Tmprss3 WT (blue circles; n ϭ 10) and P60 Tmprss3 Y260X heterozygous mice (green circles; n ϭ 10) have normal otoacoustic emissions, whereas P60 Tmprss3 Y260X homozygous mice have none (red circles; n ϭ 10). The background noise level of the recording system in the absence of sound is also depicted (shadowed area; n ϭ 10). Values are mean Ϯ S.E., and n indicates the number of animals tested. C, abnormal balance ability in the Tmprss3 Y260X homozygous mice. Shown is a comparative analysis of the time spent on fixed speed rotating rod (10 rpm) before falling (latency to fall in seconds) for P120 Tmprss3 WT (blue; n ϭ 10) and P120 Tmprss3 Y260X homozygous (red; n ϭ 9) mice. Each mouse was subjected to 4 trials of 180-s duration. The latency to fall was measured in the third and fourth trials. Data are mean Ϯ S.E. (n ϭ 10; t test, p Ͻ 0.05 for the genotype variable). D, evaluation of vestibular deficits in P120 Tmprss3 WT mice (blue; n ϭ 10) and in P120 Tmprss3 Y260X homozygous mice (red; n ϭ 9). Seven different tests were evaluated (scored from 0 to 4) and totaled to obtain the vestibular deficit (in arbitrary units (AU)). The position of the median value is shown as the horizontal line in the boxes. The bottom and top of this box identify the first and third quartiles. The height of the boxes corresponds to the interquartile range, as the characteristics of the data variability. The error bars mark the minimal and maximal non-distant values. Tmprss3 Y260X homozygous mice display slight but statistically significant vestibular deficits in comparison with wild type mice (**, p ϭ 0.002; Mann-Whitney test).  MAY 13, 2011 • VOLUME 286 • NUMBER 19 JOURNAL OF BIOLOGICAL CHEMISTRY 17389 seen in some cochleae (one cochlea of two at P7 and two of six at P11) of Tmprss3 Y260X homozygous mice (Fig. 3, B and inset) from birth to 11 days. These atypical structures were generally located near the thick kinocilium (or its footprint), although they could be seen in other parts of the cuticular plates.

Tmprss3 Is Critical for Cochlear Hair Cell Survival
A prominent finding observed in the Tmprss3 Y260X homozygous mice was the abrupt occurrence of massive hair cell loss from day 12 onward (Table 1 and Fig. 3, C-E). At P12, four animals among six Tmprss3 Y260X homozygous mice actually suffered hair cell loss. In two of these mice, missing IHCs and OHCs (Fig. 3C, inset) were observed in limited segments of the basal cochlear region (0.6 -0.9 and 1-1.5 mm from the basal extremity, respectively), the more apical parts of these cochleae being preserved (data not shown). The two other animals showed complete loss of IHCs and OHCs all along the basal and middle cochlear regions (i.e. the 5 mm starting from the basal extremity; Table 1). Strikingly, in these four specimens, both types of hair cells seemed to have disappeared simultaneously. At P13, two of six Tmprss3 Y260X homozygous mice showed complete (100%) hair cell loss. Three mutant mice showed severe (80 -90%) loss of hair cells with only a few remaining hair cells in the last apical 150 -200 m (Fig. 3D). A single mutant mouse did not show any signs of cochlear defect. At P14 onwards (Fig. 3E), all Tmprss3 Y260X homozygous mice (n ϭ 3 at P14 and n ϭ 6 at p Ն 15) showed complete hair cell loss along the length of the cochlear duct.
All Tmprss3 WT mice investigated from birth to 4 months (n ϭ 24) showed a full contingent of hair cells throughout the cochleae (Table 1 and Fig. 3, F-J), although a few scattered hair cells (Ͻ1% total hair cells; Table 1) were occasionally missing (Fig. 3J) in the cochleae of the oldest animals (2 and 4 months). Additional thin kinocilia were not seen on hair cells in wild type mice.
Ultratructural Analysis Using TEM-In 11-day-old Tmprss3 Y260X homozygous mice, the organ of Corti was almost mature (Fig. 4A). The tunnel of Corti and the spaces of Nuel were widely opened, and both IHCs and OHCs showed well formed stereocilia, including the tip links involved in the transduction process (Fig. 4A and upper inset). Afferent synapses were evidenced at the basal pole of the IHCs (Fig. 4B). Medial efferent synapses were already formed at the basal pole of OHCs, and small afferent buttons were also contacting the hair cells (Fig. 4C). The spiral lamina contained densely packed myelinated nerve fibers (Fig. 4A, lower inset).
Consistent with our SEM data, the organ of Corti of Tmprss3 Y260X homozygous mice was profoundly altered in the basal and middle cochlear turns at postnatal day 12 (Fig. 4D). Both IHCs and OHCs were absent or disappearing, and the tunnel of Corti and the spaces of Nuel were collapsed. Pieces of hair cells extruded from the epithelium were seen in the endolymphatic compartment (Fig. 4E), whereas remnants of hair cells (Fig. 4F) were seen in the scarring epithelium. In 4.5month-old Tmprss3 Y260X homozygous mice, the organ of Corti has completely regressed in the basal cochlear turn, where a single layer of undifferentiated epithelial cells remained. The more apical regions showed a scarred epithelium in which pillar cells were still recognizable (not shown).

Qualitative and Quantitative Analysis of Spiral Ganglion Neurons in Tmprss3 Y260X
In 12-and 14-day old Tmprss3 Y260X homozygous mice, the ganglion neurons and their associated glial cells had a healthy appearance (supplemental Fig. S1A). In contrast, at 4 months of age, loss of ganglion neurons had occurred, and the remaining ones showed several abnormalities, including retraction of the cell body, deformation of their nucleus, and disorganization of the myelin envelope (supplemental Fig.  S1B). Determination of the time course of ganglion neuron loss revealed that type I ganglion neuron density is similar in the basal and medial turns of the cochlea in wild type and Tmprss3 mutant mice until P90. Neuronal loss was evident at P180 and P365, with mutant mice displaying a much reduced density of type I neurons when compared with wild type animals (supplemental Fig. S1C).

Deafness Is Not Due to a Defect in the Stria Vascularis Function
TEM analysis revealed that the stria vascularis of both control and mutant mice displayed a typical appearance, with numerous blood vessels and prominent basolateral infoldings of the marginal cells interlacing with intermediate cell digita- Hair cell counts were performed using scanning electron microscopy. The full length of the cochlea is considered to be 6 mm, containing an approximate number of 660 IHCs and 2250 OHCs.  tions and normal basal cells (supplemental Fig. S2, A and B).

Missing OHCs and IHCs (%) and number (n) of
The only change observed in Tmprss3 Y260X homozygous mice was a slight swelling of intercellular spaces between the marginal and intermediate cells (supplemental Fig. S2B, arrow). The proper functioning of the stria vascularis was assessed by recording the EP in adult (P60) Tmprss3 WT and Tmprss3 Y260X homozygous mice. Both wild type and mutant mice showed normal values, suggesting normal physiology of the stria vascularis (Tmprss3 WT ϭ ϩ101.6 Ϯ 5.9 mV and Tmprss3 Y260X homozygous ϭ ϩ106.3 Ϯ 4.7 mV; p ϭ 0.83) (supplemental Fig. S2C).

Saccular Hair Cells Are Selectively Lacking in Tmprss3 Y260X Mice
The general histology of Tmprss3 Y260X homozygous mouse cristae (Fig. 5, A and B) and utricles (Fig. 5, C and D) displayed no abnormality. In these epithelia, hair cells (stars) and sup-porting cells (arrowheads) were organized similarly to the wild type. Similarly to wild type, tight junctions were present at the apex of these two cell types, whereas hair cells were polarized with their nuclei located above the nuclei of supporting cells, which lined the basal part of the epithelia. Higher magnification observations (Fig. 5, B and D) confirmed the presence in these sensory epithelia of the two hair cell types that could be identified by characteristic feature: pear shape and calyx-type afferent for type I hair cells; cylindrical shape and bouton afferents for type II hair cells.
Conversely, saccular epithelia (Fig. 5, E and F) displayed abnormal cell composition and organization. Epithelia were thinner with most hair cells lacking, although supporting cells looked normal. At higher magnification (Fig. 5F), no calyx were observed, although some large afferent terminals were present  within the epithelium, close to unidentified sensory-like cells with no hair bundle. In adult Tmprss3 Y260X homozygous mice (more than 4 months old), immunostaining of vestibular hair cells with myosin VIIa and actin fluorescent labeling of their hair bundles confirmed the nor-mal organization of the sensory epithelia in both cristae and utricles (supplemental Fig. S3, A and B), whereas it was severely altered in saccules (supplemental Fig. S3C). The lack of saccular hair cells was common to all Tmprss3 Y260X homozygous mice analyzed (n ϭ 3 males and n ϭ 3 females). Fluorescent phalloidin that stains actin filaments was used to highlight and quantify the hair bundles in the sacculus of Tmprss3 Y260X homozygous mice at different postnatal days (P0, P7, P10, and P27) and in adults (supplemental Fig. S3D). Between P0 and P10, when vestibular reflexes involved in balance are becoming established (onset of the vestibular behavior), no difference in the number of hair bundles could be quantified over time or in comparison with the Tmprss3 WT mice (analysis of variance and t test) ( Table 2). No statistical difference in the density of hair bundles was seen in P27 saccular hair cells of Tmprss3 Y260X homozygous mice compared with control mice (Table 2). Conversely, the density of hair cells was drastically and significantly reduced in the saccule of adult (Ͼ4 months) Tmprss3 Y260X homozygous mice ( Table 2). The observed time course of the selective saccular hair cells loss indicates that it relies on a progressive degenerative process following the functional activation of the hair cells.

Tmprss3 Is Expressed in Hair Cells
By in situ hybridization, we analyzed Tmprss3 expression in the developing inner ear at P7 and P11. At P7, cochlear and vestibular hair cells (HC) were strongly positive for Tmprss3 (Fig. 6, A-D); no signal was observed with a control peripherin probe, except in the type II neurons in the cochlea (arrowheads) (Fig. 6I). In addition, labeling was observed in the supporting cells of the organ of Corti (OC) as well as epithelial cells of the inner (ISS) and outer (OSS) spiral sulcus. Finally, expression was observed in interdental cells (IC) and at a lesser level in the spiral ganglion cells (SG). By P11, staining was still observed in the same cochlear and vestibular cells (Fig. 6, E-H), whereas no signal was detected with peripherin control probe, except in the type II neurons in the cochlea (arrowheads) (Fig. 6J).

A Novel Tmprss3 Isoform Is Expressed in the Mouse Inner Ear, Tmprss3f
The TMPRSS3 gene has been shown to express several alternative transcripts. These transcripts are predicted to encode for proteins of different sizes that mainly vary in their N terminus (1,16). Recently, a novel Tmprss3 isoform has been deposited in the GenBank TM nucleotide sequence data base under accession number NM_001163776. This novel mouse Tmprss3 mRNA (named Tmprss3f in this paper) is predicted to produce a protein with 22 additional amino acids at the N terminus (Fig. 7) in comparison with the first described isoform, Tmprss3a (NM_024022.2). These additional residues do not seem to encode for any known functional protein domain, like signal peptides. Because Tmprss3a does not possess an N-terminal signal peptide, it is expected that these additional amino acids will not affect the expression of the protein. We showed that this novel Tmprss3 isoform may also be present in humans because nucleotide sequence alignment analysis using BLAT (see the UCSC Genome Bioinformatics Web site) revealed the existence of a putative novel human TMPRSS3 exon that shows significant homology to the novel mouse isoform (supplemental Fig. S4). Similarity between human and mice sequences was 94%, and homology was 89% (supplemental Fig. S5).

Tissue Expression and Subcellular Localization of Tmprss3f
To investigate whether Tmprss3f exerts different functions than Tmprss3a, we analyzed its expression profile and subcellular localization. RT-PCRs were carried out on mouse P17 tissues (liver, cochlea, retina, brain, cerebellum, spleen, kidney, lung, heart, and muscle). Interestingly, Tmprss3a and Tmprss3f showed distinct expression pattern. Tmprss3a was strongly expressed in liver, cochlea, brain, cerebellum, spleen, lung, and muscle and at a lower degree in retina, kidney, and heart (Fig.  8A, middle), whereas Tmprss3f was strongly expressed only in the cochlea with very faint expression in the cerebellum, spleen, and muscle (Fig. 8A, top).
Finally, to determine if these two isoforms are differentially localized within the cell, tagged and wild type Tmprss3f and Tmprss3a proteins were transfected in HeLa cells. Both the short and the long Tmprss3 isoforms were found to colocalize with the ER marker calreticulin (Fig. 8, B and C). To ensure that this ER expression is not due to an overexpression artifact, the experiment was replicated without permeabilizing the cells. No labeling was observed at the plasma membrane (data not shown), suggesting that Tmprss3 is really localized in the ER. This novel Tmprss3 isoform could play a unique role in the cochlea. Indeed, the long N-terminal tail of Tmprss3f might be necessary for its targeting to specific ER or plasma membrane microdomains, where it could interact with different signaling molecules in a cell-specific manner.

DISCUSSION
Mutations in the TMPRSS3 gene cause non-syndromic autosomal recessive deafness (DFNB8/10). In order to explore the physiopathology of TMPRSS3-related deafness, we have generated a mouse mutant carrying a nonsense mutation in Tmprss3 (Tmprss3 Y260X ) that results in the deletion of most of the catalytic domain. This Tmprss3 Y260X homozygous mutation leads to syndromic autosomal recessive profound hearing loss in

Tmprss3 Is Critical for Cochlear Hair Cell Survival
P12). However, these cochlear hair cells may have an apparent normal morphology but may not be functionally mature. The ensuing brutal loss of these hair cells at P12, suggests that Tmprss3 play an essential role in the functional maturation of cochlear hair cells. At this time actually, improvement of mechanical properties of the organ of Corti promotes hair cell sensitivity (20). In addition, the fact that both inner and outer hair cells degenerate following the same pattern suggests that at least some signaling pathways controlled by Tmprss3 proteolytic cleavage are common and equally important in both hair cell types.
Regarding the vestibule, selective lack of saccular hair cells in Tmprss3 Y260X homozygous mice causes only a mild behavioral deficit, suggesting that other normal sensory epithelia (cristae and utricule) and probably also the vision and the CNS, as suggested in another mouse mutant, were able to partially compensate for saccular dysfunction (21). A normal number of saccular hair cells was counted in P0 Tmprss3 Y260X homozygous mice, when mechanotransduction channels are opened in vestibular hair cells (22), and no signs of hair cells loss were evident during the first postnatal days at the setup of endolymphatic composition and onset of vestibular reflexes. The partial and  A, RT-PCR specific to the Tmprss3a and Tmprss3f transcripts was performed on cDNA from 10 mouse tissues, which included liver, cochlea, retina, brain, cerebellum, spleen, kidney, lung, heart, and muscle, as indicated. All tissues demonstrated expression of Tmprss3a (top). In contrast, only cochlea and to a much lower extent cerebellum, spleen, and muscle demonstrated Tmprss3f expression (middle). GAPDH was used as a positive control (bottom). The mouse Tmprss3 cDNAs were cloned into pcDNA3-V5 (Tmprss3a) or pYFP-N1 (Tmprss3f) expression vectors and transiently transfected in HeLa cells. B, Tmprss3a-V5 (red) co-localizes with calreticulin (green), a marker of endoplasmic reticulum. C, Tmprss3f-YFP long form (green) co-localizes with calreticulin (red). No labeling differences were seen between the short and the long isoforms. Scale bars, 50 m. progressive degeneration of saccular hair cells occurred during adulthood, when saccular function is well established. These results showed that Tmprss3 is not required for saccular hair cell development and activation but seems critical for maintenance of hair cells in adult mice. It is interesting to note that Tmprrs3 is not necessary for the proper development of both cochlear and saccular hair cells because these hair cells develop normally in mutant mice, and therefore we can postulate that Tmprss3 is not involved in signaling pathways that mediate developmental and specialization processes. Tmprss3 rather acts as a maintenance factor for cochlear and saccular hair cell function and/or survival.
Three recent in vitro studies have shown that pathogenic TMPRSS3 mutations could exert their deleterious effect at the molecular and cellular level through disruption of TMPRSS3 proteolytic activity (3,9,23). Therefore, identification of the substrate of TMPRSS3 could facilitate our understanding of the role of TMPRSS3 in hearing. Until now, the ENaC sodium channel has been the only candidate substrate of TMPRSS3 in the inner ear (9). In vitro studies using Xenopus oocytes have shown that TMPRSS3 proteolytic activity is associated with increased ENaC mediated currents. The ENaC channel is composed of three subunits (␣, ␤, ␥), which are all expressed in the cochlea (24,25) and could participate in the reabsorption of sodium from the endolymph and therefore in maintaining the low sodium concentration in the endolymph (26). However, the TMPRSS3/ENaC functional hypothesis has been challenged by the fact that pseudohypoaldosteronism type 1 patients, who are homozygous for mutations in the ENaC ␣ subunit, exhibit normal hearing (10). In the present study, we showed that the absence of TMPRSS3 proteolytic activity did not affect stria vascularis function, as evaluated by EP recordings. Put together, these data suggest that the mechanism of TMPRRS3 pathogenicity is unlikely to be mediated through ENaC misregulation and, in a more general aspect, via alteration of stria vascularis function.
Tmprss3 belongs to the hepsin/TMPRSS subfamily of type II transmembrane serine proteases which contains seven members. Besides Tmprss3, another member of this subfamily (Tmprss1) has been found to be important for normal hearing (27). Severe hearing impairment in Tmprss1 Ϫ/Ϫ mice is accompanied by modest structural defects. Apart from the abnormal tectorial membrane, the other structures of the cochlea appeared to develop properly. These differences in cochlear phenotype in these two mouse mutants indicate that both genes make a unique contribution to inner ear function and suggest interaction with different substrates.
The N-terminal and transmembrane domains, which are two defining features of type II transmembrane serine proteases, are thought to contribute to the targeting of the type II transmembrane serine proteases to the plasma membrane (28). However, Tmprss3 subcellular localization, previously determined using various transfected cell lines appears to be invariably restricted to the endoplasmic reticulum (8,9). In addition, when transfected cells were treated with the protein synthesis inhibitor, cyclohexamide, expression of Tmprss3 was still visible in the ER. This indicates that this protease is stably anchored in this cellular compartment (8). Interestingly, using a similar approach Tmprss5, Tmprss6, Matriptase-2, and DESC4 expression was clearly seen at the plasma membrane (8, 29 -31), suggesting that the ER expression of Tmprss3 is unlikely to be an artifact of the method used. Altogether, these data suggest that Tmprss3 is primarily located in the endoplasmic reticulum membranes, where it might exert its function. However, at this stage, the possibility that upon stimulation, a fraction of Tmprss3 might be translocated to the plasma membrane cannot be ruled out.
The Tmprss3 Y260X homozygous mice display unique features, including abrupt and concomitant degeneration of inner and outer hair cells by the time of hearing onset. These characteristics may reflect a crucial role for Tmprss3 in the final steps of hair cell morphological/functional maturation. Without any doubt, the characterization of the cochlear substrate repertoire of Tmprss3 will provide significant insights into its role in the hearing process.