Regulation of PCDH15 function in mechanosensory hair cells by alternative splicing of the cytoplasmic domain

Hair cells of the inner ear are mechanosensors for the perception of sound, head movement and gravity. The mechanically sensitive organelle of a hair cell is the hair bundle, which consists of rows of stereocilia of graded heights polarized across the apical cell surface. Bundle polarity is essential for normal hair cell function, as only deflections of the bundle in the direction of the tallest stereocilia increase the open probability of mechanotransduction channels (Shotwell et al., 1981). Hair bundle polarity is established through a complex series of morphogenetic events. At the onset of hair bundle development, microvilli cover the apical hair cell surface and surround a centrally positioned kinocilium. The kinocilium moves to the edge of the apical hair cell surface. Subsequently, the microvilli next to the kinocilium elongate. This is followed by elongation of the more distant rows, generating a hair bundle with a staircase pattern that subsequently matures to attain its final shape (Tilney et al., 1992; Schwander et al., 2010).

The kinocilium is important for establishing hair bundle polarity as bundles are misoriented in mice carrying mutation in orthologs of genes linked to the ciliopathy Bardet-Biedl syndrome (Ross et al., 2005), and in mice with a conditional mutation in the gene encoding the intraflagellar and intraciliary transport protein Ift88 (Jones et al., 2008). Hair bundle polarization is also affected in mice with mutations in genes that regulate planar cell polarity (PCP) (Curtin et al., 2003; Montcouquiol et al., 2003; Lu et al., 2004; Wang et al., 2005; Montcouquiol et al., 2006; Wang et al., 2006). However, little is known about the molecules that connect PCP signaling to the kinocilium and bundle polarization. Candidate molecules are proteins that form the extracellular filaments connecting the stereocilia of a hair cell to each other and to the kinocilium because they are appropriately positioned to coordinate bundle polarization across different rows of stereocilia and in relation to the kinocilium (Fig. 1A). In the mouse cochlea, developing hair bundles contain ankle links, transient lateral links, top connectors and tip links; mature cochlear hair bundles maintain top connectors and tip links. Developing cochlear hair bundles also contain one kinocilium that is connected to the longest stereocilia by kinociliary links (Goodyear et al., 2005). Genes linked to hearing loss encode several of the proteins forming these linkages (Müller, 2008; Gillespie and Müller, 2009; Petit and Richardson, 2009). Relevant to the current study are protocadherin 15 (PCDH15) and cadherin 23 (CDH23), which are components of kinociliary links, transient lateral links and tip links (Siemens et al., 2004; Lagziel et al., 2005; Michel et al., 2005; Rzadzinska et al., 2005; Ahmed et al., 2006; Senften et al., 2006; Kazmierczak et al., 2007). Tip links are heterophilic adhesion complexes consisting of PCDH15 homodimers interacting with CDH23 homodimers (Kazmierczak et al., 2007) (Fig. 1A). Kinociliary links show a similar asymmetry (Goodyear et al., 2010) (Fig. 1A). Functional null mutations in PCDH15 and CDH23 lead to defects in hair bundle structure and polarity, indicating the importance of linkages containing these proteins for bundle morphogenesis (Alagramam et al., 2001a; Di Palma et al., 2001; Wilson et al., 2001; Pawlowski et al., 2006; Senften et al., 2006; Lefevre et al., 2008). Tip links are thought to gate mechanotransduction channels in hair cells (Gillespie and Müller, 2009). However, the specific aspects of bundle development regulated by kinociliary links, transient lateral links and tip links are not known.

One way a single gene could contribute to the formation of functionally distinct linkages in hair bundles is by alternative splicing of the primary transcript. Hair cells express three prominent PCDH15 splice variants (PCDH15-CD1, PCDH15-CD2 and PCDH15-CD3) differing in their cytoplasmic domains (Ahmed et al., 2001; Alagramam et al., 2001a; Alagramam et al., 2001b; Ahmed et al., 2006; Senften et al., 2006). According to published data, PCDH15-CD1 and PCDH15-CD3 are abundantly detectable in hair bundles as they mature, whereas PCDH15-CD2 is abundant only in developing bundles (Ahmed et al., 2006). It has been hypothesized that PCDH15-CD1 and PCDH15-CD2 regulate hair bundle development, while PCDH15-CD3 has been proposed to be a tip-link component (Ahmed et al., 2006). To test this model and to define the function of PCDH15 isoforms in hair cells, we generated and characterized isoform-specific knockout mice.

The strategy for generating knockout mice (see Fig. S1 in the supplementary material) followed published procedures (Grillet et al., 2009b). Homology arms were amplified from genomic DNA by PCR (Phusion High-Fidelity DNA Polymerase, New England Biolabs). Exons 35, 38 and 39 were replaced with a pGK-neomycin cassette in PCDH15-ΔCD1, PCDH15-ΔCD2 and PCDH15-ΔCD3 targeting constructs, respectively. The linearized targeting vectors were electroporated into 129P2/OlaHsd embryonic stem cells. Targeted clones were used to generate germline-transmitting chimera and crossed to FLP deleter mice to remove the selection cassette. Mice were maintained on a mixed C57BL6×129SvEv background (for genotyping primers, see Table S1 in the supplementary material).

The PCDH15-CD2 rabbit antiserum was raised by Covance (Denver, PA) against a peptide specific for exon 38 (CSEGEKARKNIVLARRRP). Antibodies were affinity purified against the peptide coupled to agarose beads. Other antibodies used were anti-acetylated-α-tubulin (mouse, Sigma, 6-11B-1), pericentrin (rabbit, Covance), Vangl2 (rabbit, Santa Cruz, H-55) and Fzd6 (mouse, R&D Systems). Additional reagents were Alexa Fluor 488- and 568-phalloidin, Alexa Fluor 488 and 594 goat anti-rabbit, and Alexa Fluor 647 goat anti-mouse.

RNA was isolated using Trizol (Invitrogen, Carlsbad, CA). cDNA was synthesized from 500 ng of RNA with Superscript III reverse transcriptase (Invitrogen) and oligo(dT) primers. RT-PCR analysis and qPCR was performed as described (Belvindrah et al., 2007) (primers are listed in Table S1 in the supplementary material).

Whole-mount staining was carried out as described previously (Grillet et al., 2009b). For scanning electron microscopy (SEM), inner ears were dissected and fixed by local perfusion with 2.5% glutaraldehyde, 4% formaldehyde, 50 mM HEPES buffer (pH 7.2), 2 mM CaCl₂, 1 mM MgCl₂ and 140 mM NaCl for 2 hours at room temperature. Inner ear sensory organs were fine dissected and processed by a modified OTOTO method (Waguespack et al., 2007). To increase structural stability and image contrast, we substituted thiocarbohydrazide by 1% tannic acid, resulting in alternate baths of 1% osmium tetroxide and 1% tannic acid for 1 hour each. Samples were dehydrated, critical-point dried (Bal-Tec CPD 030), coated with a 4 nm platinum layer (Balzers BAF 300) and observed in a SEM Hitachi S-4800, operated at 5 kV. Transmission electron microscopy was carried out as described previously (Grillet et al., 2009a).

PCDH15-CD1, PCDH15-CD2 and PCDH15-CD3 were amplified from murine cochlear mRNA and cloned into pcDNA-SK+ (Invitrogen). In situ probes were generated towards a common sequence in the 5′ region and toward unique regions in each of the isoforms and used for in situ hybridization as described (Schwander et al., 2007; Grillet et al., 2009a) (primers used to generate probes listed in Table S1 in the supplementary material). The 5′ probe was generating by cloning a PstI fragment of PCDH15 (bp 1456-2196 of mRNA, Accession Number NM_023115) into pBS-SK+.

ABR and DPOAE measurements were carried out as described previously (Schwander et al., 2007; Grillet et al., 2009b). Mechanotransduction currents from cochlear outer hair cells (OHCs) were recorded as described previously (Grillet et al., 2009b).

We measured eye movements in response to whole-body head rotations in alert mice. A head bolt was placed for animal restraint. The post was oriented so that when the animal was placed into a restraining device atop a servo-controlled rotating table, the plane tangent to the flat part of the dorsal skull was pitched 30° ‘nose-down’ from Earth horizontal. The axes of the animal's horizontal semicircular canals aligned to within 10° of the Earth-vertical axis (Calabrese and Hullar, 2006).

The video-oculography technique was used for eye movement recording (Migliaccio et al., 2005). Marker arrays fashioned from photo paper saturated with fluorescent yellow ink were opaque except for three fluorescent 200×200 μm windows separated by 200 μm and arranged in a 45° right triangle. Images were acquired at a rate of 180 frames/second using IEEE1394 (‘Firewire’) cameras [500×400 pixel frame size; Dragonfly Express, Point Grey Research, Canada; resolution <0.3°; camera/lens resolution (263 pixels/mm) at the eye surface aligned with the center of corneal curvature to within 72 μm (Migliaccio et al., 2005)]. Data were interpolated on a 1 kHz time base using a nonlinear filter based on a running spline (LabVIEW). We adjusted the spline smoothness parameter so that the correlation coefficient R² between the raw and spline-filtered data was greater than 0.80.

For rotational testing, the center of the animal's skull was aligned with the motor's Earth-vertical axis. Transient yaw whole body rotation stimuli were delivered at 3000°/s² constant acceleration for 100 ms to a peak/plateau velocity of 300°/s lasting 0.8-1.0 s, followed by a 3000°/s² deceleration for 100 ms to rest at 90° from the starting position. Eye rotation data were converted to rotation vectors in head coordinates and analyzed as described previously (Migliaccio and Todd, 1999; Migliaccio et al., 2004; Migliaccio et al., 2005). Yaw (horizontal) component eye velocity data were inverted prior to gain calculation. Quick phases and saccades were removed. The start and end of quick phases/saccades were defined as the points at which eye acceleration rose above or fell below manually estimated maximum slow phase eye acceleration. Data from at least 10 stimulus trials were averaged to obtain data for determining response parameters. For each trial, response latency was computed as the time difference between the zero-velocity-intercept times for lines fit in a least-mean-square sense to eye and head velocity during the constant-acceleration stimulus portion.

For comparison to other species (Migliaccio et al., 2004), we parameterized the responses by computing the ratio of eye and head acceleration during the constant-head-acceleration segment (G_a) and the ratio of eye and head velocity during the velocity plateau (G_v) when a plateau was evident. During calculation of G_a and G_v, the horizontal component of eye and head velocity was used. Results are mean±s.d.; each data point represents the mean for one animal over at least 10 stimulus trials. G_a,G_v and latency data were analyzed using a one-way analysis of variance (ANOVA) with significance set at P=0.05 (Diggle et al., 1994).

To define the function of PCDH15 isoforms in hair cells, we generated mice lacking specific isoforms (Fig. 1C; see Fig. S1 in the supplementary material). The three best-characterized PCDH15 isoforms in hair cells are PCDH15-CD1, PCDH15-CD2 and PCDH15-CD3, which differ in their cytoplasmic domains (Ahmed et al., 2006). Exon 35 is specific for CD1, exon 38 for CD2 and exon 39 for CD3. We replaced exons 35, 38 and 39 in ES cells by gene targeting with a neomycin cassette flanked by Frt sites (Fig. 1C; see Fig. S1 in the supplementary material). Genetically modified mice were generated and the neomycin cassette was removed by crossing the mice to a mouse line expressing FLP (Rodriguez et al., 2000). Heterozygous mice were intercrossed to generate mice homozygous for the deletion of exon 35, 38 and 39. Mice were genotyped by PCR (Fig. 1D), and will be referred to as PCDH15-ΔCD1, PCDH15-ΔCD2 and PCDH15-ΔCD3.

To confirm that the genetic modifications selectively affect the expression of specific PCDH15 isoforms, we analyzed RNA of P1 mice by PCR (see Fig. S1D,E in the supplementary material). The identity of the amplified DNA fragments was confirmed by DNA sequencing (data not shown). PCDH15-CD1, PCDH15-CD2 and PCDH15-CD3 were detectable in wild-type mice. PCDH15-ΔCD1 mice lacked PCDH15-CD1 expression, whereas PCDH15-CD2 and PCDH15-CD3 were maintained. PCDH15-ΔCD2 lacked PCDH15-CD2 expression but not PCDH15-CD1 or PCDH15-CD3; PCDH15-ΔCD3 mice lacked PCDH15-CD3 expression but not PCDH15-CD1 or PCDH15-CD2.

Next, we carried out in situ hybridization on cochlear sections using probes specific for PCDH15-CD1, PCDH15-CD2 and PCDH15-CD3. We also included a probe to the 5′ end of PCDH15 mRNA common to all three isoforms, and sense control probes (Fig. 2; see Fig. S2 in the supplementary material). Previous studies have shown that PCDH15-CD1, PCDH15-CD2 and PCDH15-CD3 are expressed in hair cells (Ahmed et al., 2006). We confirmed these findings and extended them. The probe to the 5′ region of PCDH15 revealed that in wild-type mice, PCDH15 was expressed not only in mechanosensory hair cells but also in Kölliker's organ (Fig. 2A). PCDH15-CD1 and PCDH15-CD3 were expressed in hair cells and Kölliker's organ, whereas PCDH15-CD2 appeared to be specific for hair cells (Fig. 2A). The analysis of PCDH15-ΔCD1, PCDH15-ΔCD2 and PCDH15-ΔCD3 mice revealed that in each mouse line, the two PCDH15 isoforms that were not mutated were still expressed (Fig. 2B). The expression pattern for the remaining isoforms appeared unaffected, with the exception of PCDH15-CD2, which was ectopically expressed in Kölliker's organ of PCDH15-ΔCD3 mice (Fig. 2B). Quantitative RT-PCR showed that expression of PCDH15-CD1 was not substantially changed in PCDH15-ΔCD2 and PCDH15-ΔCD3 mice. Likewise, PCDH15-CD3 expression was relatively normal in PCDH15-ΔCD1 and PCDH15-ΔCD3 mice. However, PCDH15-CD2 was upregulated in PCDH15-ΔCD3 and downregulated in PCDH15-ΔCD1 mice (see Fig. S1E in the supplementary material).

PCDH15-ΔCD1, PCDH15-ΔCD2 and PCDH15-ΔCD3 mice survived into adulthood. To measure auditory function, we recorded in 1-month- and 6-month-old mice the auditory brain stem (ABR) response to click-stimuli. Auditory thresholds were normal in PCDH15-ΔCD1 and PCDH15-ΔCD3 mice, but they were elevated to above 90 dB in PCDH15-ΔCD2 mice (Fig. 3A,B). Analysis of hearing function in response to pure tones confirmed that hearing function in PCDH15-ΔCD2 mice was affected across the entire analyzed frequency spectrum (Fig. 3C).

To study OHC function, we measured the distortion product otoacoustic emissions (DPOAEs). In wild-type, PCDH15-ΔCD1 and PCDH15-ΔCD3 mice, DPOAEs were dependent on the stimulus intensity at a given frequency, with no obvious difference between wild-type and mutants (Fig. 3D). By contrast, DPOAEs were not detectable in PCDH15-ΔCD2 mice (Fig. 3D). Similar observations were made at all frequencies analyzed (6-28 kHz) (Fig. 3D), indicating that in PCDH15-ΔCD2 the function of OHCs was impaired.

Ames waltzer^av3J mice, which carry a predicted functional null allele of PCDH15 that affects all isoforms, are deaf, circle and show defects in hair bundle morphogenesis (Alagramam et al., 2001a; Pawlowski et al., 2006; Senften et al., 2006). We wondered whether the auditory phenotype of PCDH15-ΔCD2 mice was caused by a similar mechanism and whether hair bundle morphology was normal in PCDH15-ΔCD1 and PCDH15-ΔCD3 mice. We stained P6 cochlear wholemounts with phalloidin to label F-actin in stereocilia (Fig. 4A-D). The sensory epithelium of the three mutant mouse lines was patterned into one row of inner hair cells (IHCs) and three rows of OHCs. Although hair bundle morphology appeared normal in PCDH15-ΔCD1 and PCDH15-ΔCD3 mice, hair bundles of OHCs were affected in PCDH15-ΔCD2 mice, showing defects in polarization (Fig. 4C,F,I,J). This observation was confirmed by SEM (see Fig. S3A in the supplementary material). Unlike in Ames waltzer^av3J mice, hair bundles were with few exceptions not fragmented (Fig. 4E-J; see Fig. S3A,D and Fig. S4A-C in the supplementary material).

To quantify polarity defects, we determined the deviation of the position of the bundle center, which contains the longest stereocilia, from their normal position at the tip of the V-shaped bundle (Fig. 4K-M). Four percent of the mutant bundles formed circles and were excluded from the quantification. IHCs and OHCs in wild-type mice at P1 and P6 showed little variation in orientation (Fig. 4L,M), but bundle orientation of OHCs was highly variable in PCDH15-ΔCD2 mice, deviating up to 90° from the normal position, sometimes more (Fig. 4L,M). No such defect was seen in IHCs (Fig. 4L,M).

To define the mechanism that caused the hair bundle defect in PCDH15-ΔCD2 mice, we raised antibodies against PCDH15-CD2 and analyzed its expression in hair cells at P1. As reported (Ahmed et al., 2006), PCDH15-CD2 was expressed throughout hair bundles of OHCs; we also observed expression in the short microvilli present at the hair cell surface at P1 (Fig. 5A,B,E-H). Although OHCs were strongly positive for PCDH15-CD2, it was expressed at low levels in IHCs (Fig. 5A,B). All staining was abolished in PCDH15-ΔCD2 mice (Fig. 5C,D), attesting to the specificity of the antibody. Co-staining with antibodies to acetylated-α-tubulin, to reveal the kinocilium, confirmed expression of PCDH15-CD2 in proximity to the kinocilium of OHCs (Fig. 5E-H, arrows). In some hair bundles, staining was concentrated at kinociliary tips (Fig. 5G,H, arrows). In vestibular hair cells, PCDH15-CD2 staining was strong throughout the bundles of immature hair cells and near the interface between stereocilia and the kinocilium (Fig. 6). We conclude that in OHCs, PCDH15-CD2 is targeted to the stereocilia and kinocilium, and is appropriately positioned to contribute to the formation of linkages between steoreocilia, and between the stereocilia and the kinocilium. A similar expression pattern is observed in vestibular hair cells, but there is little expression in IHCs.

We next analyzed hair bundles of cochlear hair cells at P1 and P6 by staining with fluorescence labeled phalloidin to reveal actin, and antibodies to acetylated-α-tubulin to reveal the kinocilium. In wild-type mice, the kinocilium was tightly associated to the longest stereocilia at the bundle apex (Fig. 7A,A′,C,C′). By contrast, the position of the kinocilium was more variable in PCDH15-ΔCD2 mice (Fig. 7B,B′,D,D′), suggesting that kinociliary links might be affected. This was confirmed by SEM analysis. Although kinocilia in wild-type mice at P1 were connected to the stereociliary bundle (Fig. 7E-G), they were separated from the bundle of OHCs and IHCs of PCDH15-ΔCD2 mice (Fig. 7H-J; see Fig. S4 in the supplementary material). A similar phenotype was observed at E15 (Fig. 10).

Previous studies have suggested that several PCDH15 isoforms might be present at transient lateral links (Ahmed et al., 2006). Unlike in Ames waltzer^av3J mice, hair bundles of IHCs and OHCs were not fragmented in PCDH15-ΔCD1, PCDH15-ΔCD2 and PCDH15-ΔCD3 mice (Fig. 4A-D,F,I,J; Fig. 7B,B′,D,D′ and see Fig. S4 in the supplementary material), suggesting that some of these isoforms have redundant functions at transient lateral links. This was confirmed by SEM analysis. In PCDH15-ΔCD2 mice, the kinocilium was separated from the stereocilia, but linkages along the length of stereocilia were visible (Fig. 7H-J; Fig. 8A-D). Such linkages were also observed in hair bundles from PCDH15-ΔCD1 and PCDH15-ΔCD3 mice (see Fig. S5 in the supplementary material).

At P96, there was widespread degeneration of hair bundles in the cochlea of PCDH15-ΔCD2 mice (see Fig. S3C in the supplementary material), although some hair bundles were maintained (see Fig. S3D in the supplementary material). The degenerative changes probably contribute to deafness of PCDH15-ΔCD2 mice. Stereocilia of OHCs that were maintained in the mutants formed connections to the tectorial membrane, as evident from their imprint in the tectorial membrane (see Fig. S3B in the supplementary material).

PCDH15-ΔCD1 and PCDH15-ΔCD3 mice showed normal hearing, suggesting that tip links were unaffected. Deafness in PCDH15-ΔCD2 mice is probably caused by defects in hair bundle morphology and maintenance, but tip-link function may also be impaired. We therefore analyzed tip links by SEM in P7 mice. Linkages running from the top of stereocilia to the side of the next taller stereocilia were observed in PCDH15-ΔCD1, PCDH15-ΔCD2 and PCDH15-ΔCD3 mice (Fig. 8C-G), indicating that tip links were preserved.

To determine whether hair cells from PCDH15-ΔCD2 mice had functional transduction, we measured transducer currents in P7-P8 OHCs from the apical/middle part of the cochlea using whole cell recordings. We focused our analysis on hair bundles with minimal polarity defects. In agreement with previous findings (Kennedy et al., 2003; Kros et al., 2002; Stauffer and Holt, 2007; Waguespack et al., 2007), control OHCs had rapidly activating transducer currents, which subsequently adapted (Fig. 8H, blue traces); similar current traces were obtained with OHCs from PCDH15-ΔCD2 mice (Fig. 8H, red traces). The amplitudes of saturated mechanotransduction currents at maximal deflection were similar in controls and mutants (Fig. 8H,I), suggesting that the total number of transducer channels was not altered in mutants. We also plotted the current/displacement relationships and channel open probability/displacement relationship and observed no significant difference between wild-type and mutants (Fig. 8I,J), indicating that the transduction machinery of hair cells in PCDH15-ΔCD2 mice functioned normally. Finally, as PCDH15-CD3 has previously been proposed as a component of tip links (Ahmed et al., 2006), we also measured transducer currents in PCDH15-ΔCD3 mice but observed no defects (see Fig. S6 in the supplementary material).

The kinocilium of hair cells is thought to be important for the development of hair bundle polarity. To determine the extent to which kinocilia position and bundle rotation in PCDH15-ΔCD2 mice correlates, we determined kinociliary position relative to stereociliary bundles (Fig. 9A). In IHCs and OHCs of wild-type mice, the position of the kinocilium predicts the orientation of the bundle (Fig. 9B,C). By contrast, kinocilia position did not predict bundle orientation in PCDH15-ΔCD2 mice and frequently deviated 90° or more from the normal position (Fig. 9B,C).

In wild-type hair cells, the centrioles of the basal body undergo a similar polarized movement to the kinocilium (Jones et al., 2008). Consistent with these data, staining for pericentrin revealed the polar localization of the centrioles in OHCs from wild-type mice at P1 (Fig. 9D-F′). In PCDH15-ΔCD2 mice, the two centrioles and kinocilium were coordinately mislocalized (Fig. 9G-I′) and the two centrioles were no longer aligned appropriately relative to each other along the axis of bundle polarity (Fig. 9G-I′). We conclude that kinociliary links are required to coordinate the polarization of the kinocilium and their basal bodies with the polarization of the stereociliary bundle.

Random polarization of hair bundles, similar to that observed in PCDH15-ΔCD2 mice, is a hallmark of mice with mutations in orthologs of genes linked to the Bardet-Biedl syndrome (Ross et al., 2005), and in mice with mutations that affect the intraciliary transport protein Ift88 (Jones et al., 2008). The ciliary genes are thought to act downstream of components of the core PCP pathway. We therefore asked whether core PCP proteins showed normal asymmetric localization in PCDH15-ΔCD2 mice. We stained cochlear wholemounts of P1 mice with phalloidin and antibodies to Fzd6 and Vangl2, which are normally localized at the medial edge of OHCs (Fig. 9J-M′; data not shown). The asymmetric distribution of Fzd6 (Fig. 9L-M′) and Vangl2 (data not shown) was not altered in PCDH15-ΔCD2 mice, suggesting that PCDH15-CD2 acts downstream of the PCP pathway.

It is thought that the kinocilium of vestibular hair cells is required for mechanical coupling between the hair bundle and the overlying gelatinous matrix (Roberts et al., 1988). We therefore wondered whether kinociliary links and vestibular function were affected in PCDH15-ΔCD2 mice. Analysis at E15.5, P1 and P6 revealed that although the kinocilium was tightly linked to hair bundles in wild-type mice (Fig. 10A-B′; data not shown), the kinocilium was separated from the stereociliary bundle in PCDH15-ΔCD2 mice in hair cells in the utricle, saccule and semicircular canals (Fig. 10C-D′; see Fig. S7 in the supplementary material; data not shown). Stereocilia were also in contact with the overlying gelatinous matrix (see Fig. S7C-E in the supplementary material).

We next analyzed vestibular function. Unlike in Ames waltzer^av3J mice, which lack all three PCDH15 isoforms, PCDH15-ΔCD2 mice (and PCDH15-ΔCD1 and PCDH15-ΔCD3 mice) did not circle (data not shown). To evaluate vestibular function quantitatively, we measured eye movements in alert mice mediated by the vestibulo-ocular reflex in response to whole-body head rotations (Fig. 10E-G). Mice were immobilized in a superstructure mounted atop a motor able to deliver 3000°/s constant accelerations to a plateau velocity of 300°/s about the dorsoventral axis of the animal through the stereotaxic origin of the skull. Eye movements were measured using a three-dimensional video-oculography system. We quantified responses to whole-body, passive, transient rotations in darkness using three parameters: Ga (‘constant-acceleration gain’) was computed as the ratio of eye to head velocity averaged over the constant acceleration segment of the stimulus; Gv (‘constant-velocity gain’) was computed as the ratio of eye to head velocity averaged over the constant velocity segment of the stimulus; latency is the delay from the onset of head movement to the onset of eye movement, computed through linear extrapolation of eye and head responses during the constant-acceleration segment of the stimulus. Measurements were carried out with PCDH15-ΔCD2 mice (n=6), C57BL6×129 wild-type littermate controls (n=8), C57BL/6 wild-type controls (n=7), Ames waltzer^av3J mice (n=3) and one dead mouse (to rule out technical artifacts). Ames waltzer^av3J mice and dead mice did not show any response. However, there was no statistically significant difference in any of the parameter between PCDH15-ΔCD2 mice and wild-type mice (Fig. 9E-G). In addition, we analyzed the response to head pitch tilts about the intreraural axis and observed normal responses in the mutants (data not shown). These data show that vestibular function was surprisingly intact in PCDH15-ΔCD2 mice.

Here, we show that alternative splicing of PCDH15 regulates its function in hair cells. Using isoform-specific knockout mice, we show that PCDH15-CD1 and PCDH15-CD3 are not essential for hair cell function, whereas PCDH15-CD2 is required for the formation of kinociliary links. In the absence of PCDH15-CD2, hair bundles consisting of rows of stereocilia of graded heights develop but bundle polarity is affected, demonstrating that kinociliary links are not essential for the development for the stereociliary staircase but for the morphological transformation that lead to bundle polarization. Components of the PCP pathway are normally distributed in the sensory epithelia of PCDH15-ΔCD2 mice, suggesting that PCDH15-CD2 acts downstream of the PCP pathway. Hair bundles in PCDH15-ΔCD2 mice are gradually lost postnatally, indicating that the morphological defects affect hair cell function causing deafness. In addition, we cannot exclude that cochlear hair cells do not function normally because their stereocilia are not connected to the kinocilium. Unlike in Ames waltzer^av3J mice, which carry a predicted PCDH15 null allele, stereociliary bundles are not fragmented in mice individually lacking PCDH15-CD1, PCDH15-CD2 and PCDH15-CD3, suggesting that several PCDH15 isoforms contribute to the formation of transient lateral links. We also observed that tip links are present in PCDH15-ΔCD1, PCDH15-ΔCD2 and PCDH15-ΔCD3 mice, providing evidence that none of these isoforms is uniquely required at tip links.

One of the central findings of our study is that PCDH15-CD2 is required for the development of kinociliary links and the normal polarization of hair bundles. Initial studies by the Tilney laboratory suggested that the kinocilium is important for bundle polarization (Tilney et al., 1992), but the mechanism by which the kinocilium determines polarity had remained unclear. Recent studies have shown that mutations in components of the core PCP pathway and mutations affecting the kinocilium cause polarity defects. When PCP signaling is defective, hair bundles maintain intrinsic polarity but are randomly polarized in the apical cell surface. When the kinocilium is affected, bundles show random polarization within the apical surface and loss of intrinsic polarity (Curtin et al., 2003; Montcouquiol et al., 2003; Lu et al., 2004; Ross et al., 2005; Wang et al., 2005; Montcouquiol et al., 2006; Wang et al., 2006; Jones et al., 2008). The asymmetric localization of PCP proteins at the border between hair and support cells is not affected in ciliary mutants (Jones et al., 2008), suggesting that ciliary genes act downstream of PCP components (Jones and Chen, 2008). In PCDH15-ΔCD2 mice, the PCP components Frz6 and Vangl2 are normally distributed, and hair bundle morphology resembles that of the ciliary mutants, suggesting that PCDH15-CD2 behaves like a ciliary mutant acting downstream of the PCP pathway. In the PCDH15-ΔCD2 mutants, the kinocilium and stereocilia bundle are mislocalized, but not always in same direction, suggesting that kinociliary links are required to coordinate movement of the kinocilium and stereocilia.

We have previously shown that tip links in hair cells have an intrinsic asymmetry in the planar polarity axis of hair cells, where PCDH15 forms the lower end of tip links and CDH23 the upper end (Kazmierczak et al., 2007). CDH23 and PCDH15 are also asymmetrically distributed at kinociliary links (Goodyear et al., 2010). However, polarity is reversed relative to tip links, with PCDH15 being present in the kinocilium and CDH23 in the longest stereocilia (Fig. 1A). Our findings indicate that PCDH15-CD2 is the PCDH15 isoform in kinocilia and suggest a mechanism by which it might regulate polarity. As the kinocilium, unlike the stereocilia, contains microtubules, it seems likely that the CD2 cytoplasmic domain contains specific sequence motifs for targeting to the kinocilia that are absent in CD1 and CD3. In this way, a polarity axis is established in kinociliary links. The fact that some PCDH15-CD2 is present in stereocilia could be explained because actin-based molecular motors might also transport PCDH15-CD2. Alternatively, as PCDH15 molecules form dimers (Kazmierczak et al., 2007), PCDH15-CD2 might enter stereocilia as a heterodimer with PCDH15-CD1 and/or PCDH15-CD3.

Our data show that several PCDH15 isoforms have redundant function in hair cells. The hair bundles of Ames waltzer^av3J mice, which lack all PCDH15 isoforms, show bundle fragmentation (Alagramam et al., 2001a; Pawlowski et al., 2006; Senften et al., 2006). As stereociliary bundles are not significantly fragmented in PCDH15-ΔCD1, PCDH15-ΔCD2 and PCDH15-ΔCD3 mice, it seems that several of these isoform contribute to transient lateral links. Consistent with this finding, we observed by SEM abundant links between the stereocilia in all three mutants. Similarly, it seems that neither PCDH15-CD1, PCDH15-CD2 and PCDH15-CD3 are uniquely required for tip-link formation. Previous immunolocalization studies have failed to clearly define the PCDH15 isoform at tip links. PCDH15-CD3 has been the best candidate, as one CD3-specific antibody bound to the region of the lower tip link end. However, the same antibody did not bind to adult murine hair cells and a second PCDH15-CD3 antibody bound to stereocilia more broadly (Ahmed et al., 2006). Using antibodies against PCDH15 isoforms, we have also been unable to define the PCDH15 isoform at tip links, as different antibodies to the same isoform revealed distinct expression patterns. This variation might be caused by differences in antibody affinity or epitope masking. Nevertheless, our genetic studies show that PCDH15-CD1, PCDH15-CD2 and PCDH15-CD3 individually are not essential for tip-link formation. Extracellular filaments with tip link features are present in PCDH15-ΔCD1, PCDH15-ΔCD2 and PCDH15-ΔCD3 mice. In addition, auditory and vestibular function is preserved in PCDH15-ΔCD1 and PCDH15-ΔCD3 mice, and vestibular function in PCDH15-ΔCD2 mice. OHCs in PCDH15-ΔCD2 and PCDH15-ΔCD3 mice also show normal responses to mechanical stimulation in vitro. An important implication of our findings is that the mechanotransduction channel in hair cells, which is localized in proximity to the lower tip-link end (Beurg et al., 2009) might bind to the extracellular and/or transmembrane domain of PCDH15 that is shared by several PCDH15 isoforms.

We were surprised to find that vestibular function was not noticeably altered in PCDH15-ΔCD2 mice. The kinocilium of vestibular hair cells is coupled to the kinocilium and the overlying gelatinous extracellular matrices. Unlike in the cochlea, adult vestibular hair cells do not loose their kinocilium. It has therefore been assumed that the kinocilium mechanically couples the hair bundle to the overlying gelatinous matrix (Roberts et al., 1988). Our findings now suggest that murine vestibular hair cells function normally, even when linkages between the kinocilium and stereocilia are disrupted. Direct coupling of the stereocilia to the gelatinous matrices is probably sufficient to transmit motion to stereocilia and activate transduction channels.

Our functional data were collected with mice of a mixed C57BL6×129SvEv background. Although we did not observe differences in PCDH15 isoform expression between inbred mouse strains, we cannot exclude the possibility that different phenotypes might be observed on distinct genetic backgrounds.

Our findings could have important ramifications for the analysis of the genetic causes of sensory impairment. Several PCDH15 mutations have been identified that cause Usher Syndrome and recessive forms of deafness (Ahmed et al., 2001; Alagramam et al., 2001b; Ben-Yosef et al., 2003; Ouyang et al., 2005; Roux et al., 2006; Le Guedard et al., 2007). These mutations commonly map to the extracellular PCDH15 domain. Our findings suggest that it would be important to determine whether mutations that specifically affect the PCDH15-CD2 isoform might lead to auditory impairment. Last, mutations specific for PCDH15-CD1 and PCDH15-CD3 might lead to retinal disease without auditory dysfunction.

This work was funded by NIDCD grants DC007704, DC005965, DC005965-S1 (U.M.), DC9255, DC2390 (C.D.S.) and NIDCD-DIR-Z01DC000002 (B.K.); by the Skaggs Institute for Chemical Biology (U.M.); by the Dorris Neuroscience Center (U.M.); by a Ruth Kirschstein Predoctoral Fellowship award (S.W.W.); and by a CIRM training grant (S.W.W.). Deposited in PMC for release after 12 months.