A mouse model for nonsyndromic deafness (DFNB12) links hearing loss to defects in tip links of mechanosensory hair cells

Abstract

Deafness is the most common form of sensory impairment in humans and is frequently caused by single gene mutations. Interestingly, different mutations in a gene can cause syndromic and nonsyndromic forms of deafness, as well as progressive and age-related hearing loss. We provide here an explanation for the phenotypic variability associated with mutations in the cadherin 23 gene (CDH23). CDH23 null alleles cause deaf-blindness (Usher syndrome type 1D; USH1D), whereas missense mutations cause nonsyndromic deafness (DFNB12). In a forward genetic screen, we have identified salsa mice, which suffer from hearing loss due to a Cdh23 missense mutation modeling DFNB12. In contrast to waltzer mice, which carry a CDH23 null allele mimicking USH1D, hair cell development is unaffected in salsa mice. Instead, tip links, which are thought to gate mechanotransduction channels in hair cells, are progressively lost. Our findings suggest that DFNB12 belongs to a new class of disorder that is caused by defects in tip links. We propose that mutations in other genes that cause USH1 and nonsyndromic deafness may also have distinct effects on hair cell development and function.

Dramatic progress has been made in the identification of gene mutations that cause hearing loss, but we know comparatively little about the mechanisms by which the mutations lead to disease. Interestingly, different mutations in a gene can cause distinct disease outcomes. The cadherin 23 gene (CDH23) provides a striking example. Predicted CDH23 null mutations lead to deaf-blindness (USH1D), whereas missense mutations cause nonsyndromic deafness (DFNB12) (1–13). A polymorphism in Cdh23 is linked to age-related hearing loss (14). Similarly, mutations in the genes for myosin VIIa (MYO7A) and protocadherin 15 (PCDH15) cause USH1 and nonsyndromic deafness (http://webh01.ua.ac.be/hhh/).

Recent studies in mice suggest that USH1 is caused by defects in hair cell development. Each developing hair cell contains at the apical surface a single kinocilium and rows of stereocilia, which form the mechanically sensitive organelle of a hair cell (Fig. 1A). Extracellular filaments connect the stereocilia and kinocilium of a developing hair cell (15). CDH23 and PCDH15 are components of transient lateral links, kinociliary links, and tip links (Fig. 1A) (16–20), and their cytoplasmic domains bind to protein complexes containing harmonin, MYO7A, and sans (21–25). Predicted null mutations in murine USH1 genes cause defects in hair bundle development (24, 26–34), suggesting that USH1 proteins form transmembrane complexes that regulate hair bundle morphogenesis. Failure of these complexes likely causes USH1.

Fig. 1.

Analysis of auditory function. (A) Diagram of a developing hair bundle. Stereocilia are connected to each other and to the kinocilium by various linkages. Linkages that contain PCDH15 and CDH23 are indicated. (B) The diagram shows the localization of CDH23 and PCDH15 at tip links. (C) Click-evoked ABR for a 2-month-old wild-type (black lines) and a salsa mouse (red line) at different sound intensities (dB). ABR waves I–IV are indicated. (D) Average auditory thresholds for 3-week-old and 2-month-old mice [wild-type n = 4 for 3 weeks (white) and 2 months (gray); salsa n = 5 for 3 weeks (orange) and 2 months (red)]. The mean ± SD is indicated; a Student's t test was performed. (E) Auditory thresholds in 3-week-old (triangles) and 2-month-old (circles) mice as determined by pure tone ABR recordings. In contrast to wild-type (gray and black lines) salsa mutants (orange and red lines) showed progressive hearing loss. (F) Analysis of movement in the open field. salsa mice (red) showed normal numbers of small-diameter rotations (>2.75-cm radius) and were not hyperactive. As a positive control, the sirtaki mouse line (gray) that has vestibular defects (35) is shown. Values are mean ± SD. A Student t test was performed. **, P < 0.01; ***, P > 0.001.

So far, there are no animal models for nonsyndromic deafness caused by mutations in any USH1 gene, and the mechanism by which such mutations cause disease is unclear. In an N-ethyl-N-nitrosourea (ENU) mutagenesis screen (35), we have now identified salsa mice, which suffer from progressive hearing loss and carry a Cdh23 missense mutation that is predicted to affect Ca²⁺ binding by the extracellular CDH23 domain. Similar mutations in the human CDH23 gene cause DFNB12 (1, 5, 6, 10, 13, 36). Unlike in mice with predicted Cdh23 null alleles, hair bundle development appears unaffected in salsa mice. Instead, tip links are progressively lost, suggesting that similar mutations in DFNB12 patients lead to deafness by affecting tip links.

Results

Progressive Hearing Loss in salsa Mice.

In an ENU mutagenesis screen, we identified salsa mice, which show no auditory startle response (35). Recordings of the auditory brainstem response (ABR) revealed that salsa mice suffer from progressive hearing loss. Wild-type mice at 3 weeks and 2 months of age had ABR thresholds to click stimuli at 20 ± 5 dB. Thresholds in salsa mice were by 3 weeks at 78 ± 15 dB and by 2 months at 100 ± 5 dB (Fig. 1 C and D). salsa mice were hearing-impaired across all frequencies (Fig. 1E). No defect was observed in heterozygous salsa mice, demonstrating recessive mode of inheritance (35). Distortion product otoacoustic emissions were not detected in salsa mice (Fig. S1), indicating that outer hair cell function was perturbed. Movement and the ability to swim were unaffected (Fig. 1F) (35), indicating that the vestibular system of salsa mice was intact.

salsa Mice Carry a Cdh23 Point Mutation.

salsa mice were derived on a C57BL/6J background (35). To identify the affected gene, we crossed salsa mice to 129S1/SvImJ mice. Offspring were intercrossed to obtain F₂ mice for ABR measurements and DNA preparation. By using single-nucleotide polymorphisms (SNPs), the affected genomic locus was mapped to a 4-MB interval on chromosome 10 containing Cdh23 (Fig. S2A). Sequencing revealed a single point mutation, A2210T, in exon 22 of Cdh23 (Fig. 2A). Compound heterozygous mice carrying 1 salsa allele and 1 previously reported mutant Cdh23 allele (waltzer^v2J) (27) were deaf (Fig. S2B), confirming that the salsa mutation causes deafness. The salsa mutation leads to a Glu737Val substitution within an LDRE motif in the seventh cadherin repeat of CDH23 (Fig. 2B), which is conserved in CDH23 across species and in cadherin repeats of other cadherins (Fig. 2B), and is required for Ca²⁺ chelation (Fig. 2C) (37, 38). Several mutations that cause DFNB12 resemble the salsa mutation and affect Ca²⁺-binding motifs (LDRE, DXND, DXD; Fig. 2D and Table S1) (1, 5–7, 10, 13, 36). salsa mice are therefore a model for some forms of DFNB12. Mutations that lead to predicted CDH23 null alleles cause USH1D (Fig. 2D) (1–5, 7–12), which is modeled by mutations in waltzer mice (Fig. 2D) (27, 33, 34).

Fig. 2.

The mutation in salsa mice maps to a Ca²⁺-binding motif in CDH23. (A) Sequence chromatograph from wild-type and salsa mice reveals an A-to-T mutation in exon 22 of Cdh23. (B) The C-terminal part of EC7 of CDH23 from different species is shown. The Glu737Val substitution in salsa affects a conserved Ca²⁺-binding motif (yellow boxes). CDH1 and CDH2 are shown for comparison. (C) CDH23 EC7/8 wild-type (blue) and salsa mutant (red) sequences threaded onto the E-cadherin EC1/2 by using the automodel class and a sequence alignment produced by T-Coffee. Energy-minimized model shows that the Glu737Val mutation affects Ca²⁺ coordination. (D) Domain structure of CDH23 indicating the 27 extracellular cadherin repeats (blue). Missense mutations in Ca²⁺-binding motifs in the CDH23 extracellular domain cause DFNB12 in humans (purple shaded box). Nonsense and splice site mutations have been identified in waltzer mice and USH1D (gray shaded box). SS indicates signal sequence; TM, transmembrane domain.

Progressive Tip-Link Loss in salsa Mice.

Although hair bundles are disrupted in waltzer mice at early postnatal ages (Fig. 3F) (27, 34), hair bundles were unaffected in salsa mice at postnatal day 5 (P5) and P28 (Fig. 3). This suggests that CDH23 plays an important role in steps beyond hair bundle development. Staining with CDH23 antibodies revealed that CDH23 was targeted to the stereocilia of hair cells in salsa mice by P5 and P10 (Fig. 4 A–F and K), but expression was only occasionally detectable by P30 (Fig. 4 G–K). However, no defects in CDH23 cell surface transport were observed (Fig. S3). Because CDH23 is a tip-link component (16, 20), tip-link maintenance may be affected. Tips of stereocilia show characteristic tenting that is thought to be the consequence of tip-link-mediated tension (39–42). Tenting can therefore be used to quantify the presence of tip links. Tenting was observed in wild-type hair cells at P5, P21, and P60 (Fig. 5 A, C, and H and Figs. S4 and S5). In salsa mice, tenting was observed throughout the cochlear duct at P5 (Fig. 5 B and H and Fig. S4); by P21, tenting was no longer observed in the basal turn of the cochlea, and was absent in all hair cells by P60 (Fig. 5D and Fig. S5). salsa mice showed no signs of vestibular dysfunction (Fig. 1F), and tenting of stereociliary tips and tip links were maintained in vestibular hair cells (Fig. 5 F and G). Finally, by P90, the organ of Corti and spiral ganglion degenerated in salsa mice, whereas the vestibule was unaffected (Fig. S6). Staining for activated caspase 3 showed that cochlear hair cells and spiral ganglion neurons died by apoptosis (Fig. S6).

Fig. 3.

Preserved hair bundle morphology in salsa mice. (A–F) Scanning electron microscopy micrographs of cochlear whole mounts from P5 wild-type mice and salsa and waltzer mutants. (A and B) The organ of Corti in salsa mice was patterned normally in 3 rows of outer and 1 row of inner hair cells. (C–F) Hair bundles in salsa displayed the characteristic polarized morphology with a single kinocilium. Hair bundles in waltzer mice were fragmented. (G–L) At P28, hair bundle morphology was indistinguishable in wild-type and salsa mice. [Scale bars: A and B (5 μm), C–F (2 μm), G and H (10 μm), and I–L (2 μm).]

Fig. 4.

Progressive loss of CDH23 expression in salsa mice. (A–J) Cochlear whole mounts of wild-type and salsa mice were stained with an antibody against the CDH23 cytoplasmic domain (green) and with phalloidin (red). (A and B) Levels of CDH23 expression in homozygous salsa and wild-type mice at P5 were similar. (C and D) Higher-magnification view of CDH23 expression in hair bundles at P5. (E and F) At P10, CDH23 staining was maintained at stereociliary tips. (G–J) At P30, CDH23 staining was barely detectable in salsa. Arrowheads point to CDH23 in wild types and residual staining in salsa. (K) Quantification of CDH23 staining in hair cells at P10 and P30. salsa mice (black) showed reduced numbers of CDH23 puncta in hair bundles at P30. Values are mean ± SD. A Student's t test was performed. ***, P < 0.001. [Scale bar: A and B (8 μm), C–J (2 μm).]

Fig. 5.

Progressive loss of tenting at stereociliary tips. (A and B) Scanning electron microscopy analysis in P5 wild-type and salsa mice revealed tip tenting in stereocilia of cochlear hair cells (arrowheads). (C and D) Defects in tenting in salsa mice at P60 (arrowheads in C indicate tenting in wild-type). (E–G) Tips in vestibular hair cells display normal tenting. Tip links (arrowhead in G) were preserved. (H) Quantification of tip tenting at P5 and P21. Tenting was reduced in hair cells from the medial and basal cochlear turns at P21. *, P < 0.05. [Scale bars: A–D (300 nm), E and F (500 nm), and G (150 nm).]

Defects in Hair Cell Function by P7/8.

To test whether hair cell function was affected before detectable morphological changes, we recorded mechanotransduction currents from P7/8 cochlear hair cells in response to 5-ms bundle deflections ranging from −400 to 1000 nm. Current-displacement [I(X)] relationship plots did not reveal a difference between wild-type and salsa mice (Fig. 6 A and B). Peak currents at maximal deflection were similar in controls (419 ± 40 pA) and mutants (427 ± 35 pA). We next analyzed the adaptation kinetics of mechanotransduction currents in outer hair cells stimulated with 100-ms deflections of 100–800 nm. Normalization of each current trace to its peak current showed faster transduction current decline in salsa compared with wild type, which was only observed at deflections above 300 nm (Fig. 6 C and D). We conclude that hair cell function was mildly affected by P7/8.

Fig. 6.

Mechanotransduction currents. Data from wild type are represented in blue and from salsa in red. (A) Examples of transduction currents in cochlear outer hair cells at P7 in response to 5-ms mechanical stimulation. (B) Current-displacement [I(X)] relationships revealed no obvious difference between wild type and salsa. (C) Examples of transduction currents in cochlear outer hair cells at P7 in response to 100-ms mechanical stimulation. (D) Averaged transduction currents for deflections between 100 and 800 nm expressed as a percentage of peak currents at 800 nm. Current amplitude was lower in homozygous salsa mice between 300-nm and 800-nm deflection. Data are shown as mean ± SEM. Student's 2-tailed unpaired t test was performed (*, P < 0.05; **, P < 0.01).

The salsa Mutation Affects CDH23 Adhesive Function.

CDH23 interacts with PCDH15 to form tip links (16). To test the effects of the salsa mutation on CDH23 properties, we incubated the purified extracellular domains of CDH23 with or without the salsa mutation fused to Fc tags (referred to as CDH23wt-Fc and CDH23salsa-Fc, respectively) with the purified His-tagged extracellular domain of protocadherin 15 (PCDH15-His; Fig. 7A). Protein complexes were analyzed by Western blotting. As reported, PCDH15-His bound to CDH23wt-Fc in a Ca²⁺-dependent manner (Fig. 7B) (16). We also observed Ca²⁺-dependent interactions between PCDH15-His and CDH23salsa-Fc, but interactions were diminished (Fig. 7B). To determine whether mutations in CDH23 that cause DFNB12 reduce interactions with PCDH15, we engineered 2 human mutations into CDH23-Fc (Fig. 7A) (1). Both mutations reduced interactions of CDH23 with PCDH15 (Fig. 7B).

Fig. 7.

salsa and DFNB12 mutations affect interactions between CDH23 and PCDH15. (A) Diagram of CDH23 and PCDH15 constructs. The extracellular domains were fused to a His or Fc tag. (B) CDH23Fc and the mutant derivatives were incubated with PCDH15-His in the presence of 1 mM EDTA or in the presence of 10 μM and 100 μM Ca²⁺. Protein complexes were purified and analyzed by Western blotting. Complex formation was observed in the presence but not absence of Ca²⁺. The salsa and DFNB12 mutations weakened interactions between CDH23 and PCDH15. (Right) Controls for input amounts of CDH23Fc and mutant derivatives.

Discussion

Previous studies have shown that predicted Cdh23 null alleles perturb hair bundle development, likely as a consequence of defects in the filaments that connect the stereocilia and kinocilium of a developing hair cell. In contrast, we show here that a Cdh23 missense mutation leads to progressive tip-link loss and, ultimately, to hair cell death. In salsa mice, a small defect in mechanotransduction by cochlear hair cells was apparent at P7/P8, indicating that hair cell function was already affected. At subsequent ages, tip links were lost, as revealed by diminished levels of CDH23 protein in stereocilia and loss of tenting of stereociliary tips. Because predicted CDH23 null alleles are associated with USH1D, whereas missense mutations similar to the one observed in salsa mice cause DFNB12, we propose that USH1D is caused by developmental defects in hair bundles and DFNB12 by tip-link defects.

Our biochemical studies show that the salsa mutation affects interactions of CDH23 and PCDH15, even though the mutation affects an amino acid outside the ligand-binding domain in EC1 (16). Hair cell function was also maintained for some time in salsa mice, and tip links sustained substantial forces in electrophysiological experiments. These findings seem contradictory at first glance, but the study of classical cadherins provides clues that likely explain the results. Ca²⁺ binding rigidifies the cadherin extracellular domain. Disruption of Ca²⁺ binding influences protein structure, often perturbing adhesive function (37, 38). In analogy to classical cadherins, it is likely that defects in ligand binding caused by the salsa mutation report changes in CDH23 structure that are propagated over a considerable distance within the extracellular domain. Structural changes might predispose tip links to breakage, necessitating frequent regeneration that cannot be sustained. CDH23 and PCDH15 at tip links might also carry additional posttranslational modifications, which stabilize their interaction but were not present in the recombinant proteins. An interesting implication of our findings is that some CDH23 mutations may predispose individuals to hearing loss caused by mechanical insults. In fact, polymorphisms in CDH23 are associated with noise-induced hearing loss (43), and mice heterozygous for the Cdh23 waltzer allele show increased noise susceptibility (44). Finally, a polymorphism in Cdh23 is associated with age-related hearing loss (14), which may be related to instability of tip links.

All USH1 proteins control hair bundle development (15). CDH23, PCDH15, and MYO7A are also expressed in mature hair cells (the mature expression pattern for harmonin and sans has yet to be determined). As tip-link proteins, CDH23 and PCDH15 are thought to gate transduction channels, whereas MYO7A contributes to channel adaptation (45). Missense mutations in the genes for MYO7A and PCDH15, similar to the salsa mutation in CDH23, might also affect tip-link function without effects on hair cell development. The duality of USH1 protein function for hair cell development and mechanotransduction provides a plausible explanation for the different disease phenotypes that are associated with distinct mutations in USH1 genes (http://webh01.ua.ac.be/hhh/). Based on our findings, we predict that some missense mutations in USH1 genes, which cause nonsyndromic deafness, affect the mechanotransduction machinery of hair cells without effects on hair bundle development. Null mutations instead affect hair cell morphogenesis. It has been proposed that USH1 genes also participate in photoreceptor morphogenesis (46), a process that is likely affected by USH1 null alleles but not by missense mutations that affect mechanotransduction. Strikingly, tip links in vestibular hair cells and vestibular function are maintained in salsa mice and DFNB12 patients, which may be a consequence of the different mechanical properties of these low-frequency mechanoreceptors. Consistent with this model, loss of tip links in salsa mice was first observed in the basal part of the cochlea, which contains hair cells that respond to the highest frequencies.

Materials and Methods

An extended section is provided as SI Materials and Methods.

ENU Mutagenesis, Functional Studies, Positional Cloning, Histology, and Biochemistry.

ENU mutagenesis, analysis of ABRs and distortion product otoacoustic emissions, test of vestibular function, staining of sections and whole mounts, electron microscopy, positional cloning, the expression and purification of recombinant CDH23 and PCDH15, and protein interaction studies have been described previously (16, 35, 47, 48). Tip tenting was quantified by counting tented stereocilia in the medium row in inner hair cell bundles (in 2–6 hair cells per time point and genotype). For quantification of tip staining, the number of CDH23 fluorescent puncta in bundles was determined (>27 hair cells per time point and genotype).

Molecular Modeling.

Using Modeler version 9v4 (49), cadherin repeat 7 and 8 of CDH23 were threaded onto cadherin repeat 1 and 2 of CDH1 using the automodel class and the sequence alignment produced by T-Coffee (50). The model was energy-minimized to remove bad contacts. The minimized, threaded coordinates were used to mutate Glu-66 to Val to produce a mutant model. Minimization consisted of 20 cycles of conjugate gradient minimization followed by 50 cycles of molecular dynamics optimization using the Verlet algorithm and finished with 20 cycles of conjugate gradient minimization.

Mechanotransduction Currents.

Outer hair cells in the apical/middle turn were recorded. Cells were whole-cell-patched, and hair bundles were deflected with a stiff glass probe.