Aberrant cochlear hair cell attachments caused by Nectin-3 deficiency result in hair bundle abnormalities

The tissues and organs in mammals are composed of multiple heterogeneous cell types, which are arranged in complex patterns, including the checkerboard-like cell arrangement. In the organ of Corti, the auditory epithelium of the snail-shaped cochlea in the inner ear, sensory hair cells (HCs) are interdigitated with nonsensory supporting cells (SCs), forming a checkerboard-like cellular pattern (Barald and Kelley, 2004; Klein and Mlodzik, 2005) (Fig. 1A-D). In addition to this unique cell arrangement, HCs are equipped with a uniform orientation of hair bundles on their apical surfaces (Fig. 1C,D). This polarised pattern of the hair bundles is a typical example of planar cell polarity (PCP), which refers to the polarisation of a field of cells within the plane of a cell layer. The checkerboard-like cell arrangement and the polarised structures of the hair bundles are essential for appropriate perception of sound (Yoshida and Liberman, 1999).

The hair bundles on HCs are actin-based structures uniformly aligned on the apices of HCs with V-shapes pointing unidirectionally and abneurally towards the outer (lateral) border of the cochlear duct. In mice, the development and maturation of the hair bundles occur from two perpendicular directions, from the basal to the apical turn and from the medial to the lateral side of the cochlea, during a period between the late embryonic stage and postnatal day (P) 14 (Denman-Johnson and Forge, 1999). During this period, a single tubulin-based kinocilium, a specialised primary cilium extending from the basal body, emerges in the centre of the HC surface (Fig. 1E), and then undergoes directed migration towards the lateral edge. Subsequently, stereocilia become organised around the kinocilium to form the V-shaped bundles (Fig. 1F). The kinocilium retracts at ∼P10 (Fig. 1G).

The mechanism underlying the polarity of the kinocilium and stereocilia on HCs (hereafter referred to as HC polarity) is not fully understood, although many PCP molecules have been identified as regulators of HC polarity. These include frizzled (Fz), dishevelled, Celsr, vang-like (Vangl) and prickle (Goodrich and Strutt, 2011; Gray et al., 2011). Ciliary proteins, such as Kif3a and Lis1, also contribute to proper HC polarity with appropriate organisation of microtubules (Sipe et al., 2013; Sipe and Lu, 2011).

The mechanism underlying the checkerboard-like cellular patterning of HCs and SCs had not been revealed, but we recently found that nectins, immunoglobulin-like cell-cell adhesion molecules (CAMs), which comprise a family of four members (Nectin-1, -2, -3 and -4; now known as Pvrl1-4 - Mouse Genome Informatics), regulate this mechanism in the mouse auditory epithelium (Togashi et al., 2011). The characteristic features of nectins are that they promote homophilic and heterophilic trans-interactions between members (heterophilic trans-interactions are Nectin 1-3, Nectin 1-4 and Nectin 2-3), and that their heterophilic interactions are stronger than their homophilic interactions in the following order: Nectin 1-3>Nectin 2-3>Nectin 1-1, 2-2 and 3-3 (Harrison et al., 2012; Ikeda et al., 2003). In mice, Nectin-1, -2 and -3 are expressed in the auditory epithelium, of which Nectin-1 and -3 are differentially expressed in HCs and SCs, respectively. Genetic deletion of Nectin-3 leads to disruption of the checkerboard-like cellular pattern with aberrant attachments between HCs by a homophilic interaction of Nectin-1. Thus, the heterophilic trans-interaction between Nectin-1 in HCs and Nectin-3 in SCs mediates heterotypic adhesion between these two cell types and contributes to the checkerboard-like cellular patterning.

Nectins are involved in the formation of adherens junctions (AJs) cooperatively with cadherins in various cell types, including epithelial and endothelial cells and fibroblasts (Takai et al., 2008a; Takai et al., 2008b). In addition, nectins regulate the formation of tight junctions (TJs), of which major CAMs are junctional adhesion molecules (JAMs), occludin and claudins, which act cooperatively with cadherins (Takai et al., 2008a; Takai et al., 2008b).

During the course of a study on the roles of nectins in the checkerboard-like cellular patterning of HCs and SCs in the mouse auditory epithelium, we noticed that the hair bundles were morphologically abnormal in Nectin-3-deficient mice compared with those in wild-type mice. Therefore, in the present study, we investigated these abnormal phenotypes in detail.

There were no apparently different phenotypes in the auditory epithelium between wild-type and heterozygous Nectin-3 (Pvrl3) knockout (Nectin-3^+/-) mice. In the Nectin-3^+/- auditory epithelium, HCs were interdigitated with SCs, forming a checkerboard-like cellular pattern (Fig. 2A). The hair bundles formed a V-shape and were symmetrically arranged about the kinocilium as estimated by the immunofluorescence signals for F-actin and the basal body marker γ-tubulin, and by scanning electron microscopy (Fig. 2A,B). However, a number of HCs were aberrantly attached to each other directly, which was confirmed by transmission electron microscopy, and the orientation and morphology of the hair bundles on aberrantly attached HCs were disordered in Nectin-3-deficient (Nectin-3^-/-) mice in the postnatal stages (Fig. 2A-C; supplementary material Fig. S1A,B and Fig. S2).

In the Nectin-3^-/- auditory epithelium, the aberrant attachments between HCs were grouped into two categories: (1) HCs attached in the same row and (2) HCs attached in different rows (Fig. 2). Aberrant attachments between HCs in the same row were more frequently observed than were aberrant attachments between HCs in different rows [64% (17, 28 and 19% from the lateral to medial attached HC pairs), versus 36% (26 and 10% for the lateral and medial HC pairs, respectively); n=4 mice]. In HCs attached in the same row, the hair bundles were orientated towards the site of attachment and asymmetrically arranged, and their morphologies were deformed (Fig. 2E,H). They were disorganised into a straight line as if they were connected (flat bundles) or split into several clumps (split bundles) (Fig. 2A; Fig. 3A,B). Flat bundles were more frequently observed than split bundles (Fig. 3C,F). The kinocilium was aberrantly located near the boundary between attached HCs (Fig. 4A,C). In HCs attached in different rows, the hair bundles were orientated towards the attached sites and asymmetrically arranged (Fig. 2F-H), and formed a straight line down the centre of HCs (Fig. 3A,B). However, the orientation and morphology of the hair bundles were less disordered in the lateral HCs (Fig. 2F,G; Fig. 3C). The kinocilium was aberrantly located near the attached sites in the medial HCs (Fig. 4A,D,E). The basal body at the base of the kinocilium was mislocalised towards the attached sites in aberrantly attached HCs (Fig. 4F). However, these abnormalities of the hair bundles and the kinocilium were not observed in unattached HCs (Figs 2, 4). These results indicate that the orientation and morphology of the hair bundles and the positioning of the kinocilium are disturbed by the aberrant attachments between Nectin-3^-/- HCs.

In the vestibular epithelium of the saccule, sensory HCs and nonsensory SCs form a mosaic pattern and HCs are equipped with a uniform orientation of the hair bundles on their apical surfaces, similar to the cochlear epithelial cells (Deans et al., 2007). In contrast to the cochlear epithelium, HCs were not attached to each other and HC polarity did not change in Nectin-3^-/- mice: the hair bundles face away from each other across the striola in both Nectin-3^+/- and Nectin-3^-/- mice (supplementary material Fig. S3B). Consistently, differential expression patterns of nectins similar to those observed in the cochlear epithelium were not observed in the surface view of the Nectin-3^+/- vestibular epithelium: the immunofluorescence signals for Nectin-1, -2 and -3 were observed at the boundaries between HCs and SCs and between neighbouring SCs (supplementary material Fig. S3A). These results support the conclusion that the orientation and morphology of the hair bundles and the positioning of the kinocilium in the auditory epithelium are disturbed by the aberrant attachments between Nectin-3^-/- HCs, and also support the previous conclusion (Togashi et al., 2011) that the differential expression of Nectin-1 in HCs and Nectin-3 in SCs contributes to the checkerboard-like cellular patterning in the cochlear epithelium.

To determine when the abnormal phenotypes of the hair bundles and the kinocilium were observed in aberrantly attached Nectin-3^-/- HCs, we analysed the localisation of γ-tubulin in E16.5 mouse HCs. The maturation of the organ of Corti starts from the basal turn and proceeds to the apical turn of the cochlea; therefore, HCs in the apical turn are less mature than those in the middle turn (Lim and Anniko, 1985). In both the apical and middle turns of the Nectin-3^-/- cochlea, HCs were aberrantly attached to each other and the immunofluorescence signals for Nectin-1 and afadin (Mllt4 - Mouse Genome Informatics) were concentrated at the boundary between attached HCs (Fig. 5A,B). In the apical turn of the Nectin-3^+/- cochlea, the signal for γ-tubulin was observed in the centre of the apical surface of HCs, whereas it was positioned at the lateral side in the middle turn (Fig. 5C; supplementary material Fig. S4A). In the Nectin-3^+/- cochlea, aberrantly attached HCs were fewer than in the Nectin-3^-/- cochlea and were decreased during maturation (Fig. 5B). These results indicate the correct establishment of HC polarity at E16.5, which is consistent with the previous observations that the terminal differentiation of the organ of Corti starts between E14.5 and E18.5, and that HC polarity is established by E18.5 in the mouse auditory epithelium (Kelly and Chen, 2007). By contrast, in the Nectin-3^-/- cochlea at E16.5, the signal for γ-tubulin was observed in the centre of the apical surface of HCs in the apical turn even though HCs were aberrantly attached (Fig. 5C; supplementary material Fig. S4A). However, the localisation of γ-tubulin was rotated towards the attached sites in the middle turn (Fig. 5C; supplementary material Fig. S4A,B). These abnormalities were not observed in unattached HCs (Fig. 5A,C). These results indicate that the localisation of γ-tubulin is disturbed in the late stage, but not in the early stage, of the establishment of HC polarity, and suggest that the cellular machinery governing HC polarity is disturbed by the aberrant attachments between HCs during the establishment of HC polarity in the Nectin-3^-/- auditory epithelium.

Some PCP components are asymmetrically localised at the cell membranes and loss of function of PCP components causes the misorientation of the hair bundles, abnormal positioning of the kinocilium and shortening of the cochlear duct, indicating that PCP components are implicated not only in HC polarity, but also in the cell arrangement called convergent extension (Montcouquiol et al., 2006; Wang et al., 2006a; Wang et al., 2006b). Genetic ablation of PCP components causes shortening of the cochlear duct and accumulation of HCs at the apex (Montcouquiol et al., 2003; Wang et al., 2006a). To assess convergent extension in the Nectin-3^-/- cochlea, we measured the length of the cochlear duct and examined whether HCs accumulated at the apex by immunostaining for myosin VIIa. The length of the cochlear duct was not distinguishable between Nectin-3^+/- and Nectin-3^-/- mouse cochleae (5.58±0.23 mm versus 5.81±0.18 mm, respectively, mean ± s.d., n=3, P>0.05 by Student’s t-test) (Fig. 6A). HCs were not accumulated at the apices of Nectin-3^+/- and Nectin-3^-/- mouse cochleae (Fig. 6B). To determine whether PCP components are involved in the abnormal phenotypes of the Nectin-3^-/- cochlea, we compared the distributions of Vangl1 and Fz6 (Fzd6 - Mouse Genome Informatics) in Nectin-3^+/- and Nectin-3^-/- cochleae. In the Nectin-3^+/- auditory epithelium, the immunofluorescence signals for Vangl1 and Fz6 were concentrated along the boundary between the medial edges of HCs and the lateral edges of SCs (Fig. 6C,D). The localisation of the signals for Vangl1 and Fz6 was essentially unchanged in the Nectin-3^-/- auditory epithelium (Fig. 6C,D). Taken together, these results suggest that the essential function of PCP components is maintained in the Nectin-3^-/- auditory epithelium.

Because changes in the hair bundle phenotypes of Nectin-3^-/- HCs are unlikely to be related to the PCP pathway, they appear to depend on a non-autonomous effect, which probably requires Nectin-1. To verify this idea, we analysed the hair bundle phenotypes of Nectin-1^-/- HCs. In Nectin-1^-/- mice, HCs were aberrantly attached to each other; however, the number of these cells was much lower than that in Nectin-3^-/- mice (Togashi et al., 2011). Similar to aberrantly attached Nectin-3^-/- HCs, aberrantly attached Nectin-1^-/- HCs also showed disturbances of the orientation and morphology of the hair bundles (supplementary material Fig. S5A), although neither Nectin-2 nor Nectin-3 was concentrated at the boundary between aberrantly attached HCs (supplementary material Fig. S5B). Thus, common abnormal phenotypes in aberrantly attached HCs strongly support the idea that the non-autonomous effect of the heterophilic interaction between Nectin-1 and -3 is essential for the correct formation of the orientation and morphology of the hair bundles.

Nectins are involved in the formation of the epithelial apical junctional complex (AJC), which includes AJs and TJs (Takai et al., 2008a; Takai et al., 2008b). We therefore assessed the distributions of components of AJs and TJs. In the surface view of the Nectin-3^+/- auditory epithelium, the immunofluorescence signal for Nectin-1 was concentrated at the boundary between HCs and SCs, but not at the boundary between neighbouring SCs (Fig. 7A). Other AJ components, including Nectin-2, afadin, E-cadherin (cadherin 1 - Mouse Genome Informatics) and β-catenin, and the TJ component ZO-1 were concentrated at the boundaries between HCs and SCs and between neighbouring SCs (Fig. 7B-F). In the Nectin-3^-/- auditory epithelium, the signal for Nectin-1 was markedly concentrated at the contact sites between aberrantly attached HCs, whereas it was not or was hardly observed at the boundary between HCs and SCs, presumably because Nectin-1 molecules can trans-interact with each other but not with Nectin-2 molecules (Fig. 7A). Furthermore, the signals for other AJ and TJ components were also markedly concentrated at the contact sites between aberrantly attached HCs, whereas they were weakly observed at the boundaries between HCs and SCs and between neighbouring SCs (Fig. 7B-F). The same results were observed irrespective of whether HCs were attached in the same or different rows.

In the lateral view of the Nectin-3^+/- auditory epithelium, the signals for Nectin-1 were concentrated at the AJC area of the boundary between HCs and SCs, and those for Nectin-2, afadin and ZO-1 were concentrated at the AJC area of the boundaries between HCs and SCs and between neighbouring SCs (Fig. 7A-C,F; supplementary material Fig. S6A). However, the signals for E-cadherin and β-catenin extended further along the apicobasal axis than did those for Nectin-1, -2, afadin and ZO-1 (Fig. 7D,E; supplementary material Fig. S6A). In the Nectin-3^-/- auditory epithelium, the signal for Nectin-1 was concentrated at the boundary between aberrantly attached HCs. Its distribution was extended towards the basal side and coincided with that for F-actin, whereas it was hardly observed at the boundaries between HCs and SCs and between neighbouring SCs (Fig. 7A; supplementary material Fig. S6A). The signals for other AJ and TJ components were also concentrated and their distributions were extended towards the basal side and coincided with that for F-actin at the boundary between aberrantly attached HCs, like the signal for Nectin-1, whereas they showed only weak signals at the boundaries between HCs and SCs and between neighbouring SCs (Fig. 7B-F; supplementary material Fig. S6A). The localisation of the AJ and TJ components is schematically shown in Fig. 7G. Collectively, these results indicate that the Nectin-3-mediated adhesion between HCs and SCs plays a role in the proper localisation of AJC components along the planar and apicobasal axes.

Within the AJC, nectins are physically associated with the apicobasal polarity protein (ABP) components Patj (Inadl - Mouse Genome Informatics) and Par-3 (Pard3 - Mouse Genome Informatics) (Adachi et al., 2009; Takekuni et al., 2003). We finally compared the distribution of ABP components in Nectin-3^+/- and Nectin-3^-/- auditory epithelia. Three major ABP complexes, namely, the apical Crumbs complex consisting of Crumbs3 (Crb3), Pals1 (Mpp5 - Mouse Genome Informatics) and Pals1-associated tight junction protein (Patj), the Par complex consisting of Par-3, Par-6 (Pard6 - Mouse Genome Informatics) and atypical protein kinase C (aPKC), and the lateral Scribble complex consisting of Scribble, discs large (Dlg1) and Lgl, are essential for epithelial ABP formation (Pieczynski and Margolis, 2011). In the surface view of the Nectin-3^+/- auditory epithelium, the immunofluorescence signal for Pals1 was concentrated at the phalloidin-negative apical surface of HCs, which was polarised towards the lateral side (Fig. 8A). The signal for Par-3 was uniformly distributed along the boundaries between HCs and SCs and between neighbouring SCs (Fig. 8B). In the Nectin-3^-/- auditory epithelium, the area positive for Pals1 signal was rotated towards the contact sites between aberrantly attached HCs (Fig. 8A), and the signal for Par-3 was concentrated at the contact sites between aberrantly attached HCs (Fig. 8B). These results indicate that the localisation of Pals1 and Par-3 along the planar axis changes in aberrantly attached Nectin-3^-/- HCs.

In the lateral view of the Nectin-3^+/- auditory epithelium, the signal for Pals1 was observed in the apical area without overlapping that for Nectin-1 at the boundary between HCs and SCs (Fig. 8A; supplementary material Fig. S6B and Fig. S7A). The signal for Par-3 was concentrated at the apical area partly overlapping that for Nectin-1 (Fig. 8B; supplementary material Fig. S6B and Fig. S7B). In the Nectin-3^-/- auditory epithelium, the distribution of the signal for Pals1 was not extended towards the basal area of the boundary between aberrantly attached HCs (Fig. 8A; supplementary material Fig. S6B and Fig. S7A). The distribution of the signal for Par-3 extended into the basal area, partly overlapping that for Nectin-1 at the boundary between aberrantly attached HCs, compared with that at the boundaries between HCs and SCs and between neighbouring SCs (Fig. 8B; supplementary material Fig. S6B and Fig. S7B). These results indicate that the localisation of Par-3, but not that of Pals1, along the apicobasal axis, changes in aberrantly attached HCs. The localisation of the ABP components is schematically shown in Fig. 8C. Taken together, these results indicate that the Nectin-3-mediated adhesion between HCs and SCs is crucial for the correct localisation of Pals1 and Par-3.

We previously showed that Nectin-1 and -3 were differentially expressed in HCs and SCs, respectively, and that, in the Nectin-3^-/- auditory epithelium, HCs were aberrantly attached to each other, thereby leading to a disturbance of the checkerboard-like cellular pattern of HCs and SCs. On the basis of these observations, we proposed that the heterophilic trans-interaction between Nectin-1 in HCs and Nectin-3 in SCs plays a key role in the checkerboard-like cellular patterning (Togashi et al., 2011). Nectins are involved in the formation of AJs and TJs in epithelial cells (Takai et al., 2008a; Takai et al., 2008b). Here, we showed that in the Nectin-3^-/- auditory epithelium, junctions between HCs and SCs were disrupted with aberrant localisation of AJ and TJ components as well as ABP components. Moreover, the absence of either Nectin-1 or -3 caused disturbances of the orientation and morphology of the hair bundles and/or the positioning of the kinocilium; however, these abnormal phenotypes were observed only in aberrantly attached HCs, but not in unattached ones. Thus, the abnormal phenotypes of HCs are likely to be due to a non-autonomous effect depending on the heterophilic interaction between Nectin-1 and -3. In addition, in Nectin-3^+/- mice, the aberrant attachments between HCs were seen at E16.5, but not in the postnatal stages, suggesting that these attachments are released during maturation. However, in Nectin-3^-/- mice, the abnormal phenotypes began during the migration of the hair bundles in the early developmental stages and continued until the postnatal stages. These results indicate that the changes in junction formation caused by the aberrant attachments between HCs result in the disruption of HC maturation associated with the abnormal phenotypes of the hair bundles in the Nectin-3^-/- auditory epithelium.

Similar phenotypes of the hair bundles were reported in homeobox transcription factor Emx2-deficient mice and in mice mutant for the Notch ligands Dll1 and Jag2 (Holley et al., 2010; Kiernan et al., 2005). In these mice, HCs were attached to each other and lost their polarity. However, these mice showed additional abnormalities, such as defects in HC development or changes in the numbers of HCs, suggesting that the abnormal signals govern all HCs. Lack of these abnormalities suggests that Emx2 and the Notch pathway function normally in Nectin-3^-/- mice.

PCP components regulate HC polarity and convergent extension (Goodrich and Strutt, 2011; Gray et al., 2011). Genetic deletion of PCP components causes hair bundle misorientation, shortening of the cochlear duct and abnormal accumulation of HCs at the apex of the duct (Montcouquiol et al., 2003; Wang et al., 2006a). Similarly, genetic deletion of ciliary proteins, such as Ift88, Kif3a and Lis1 (Pafah1b1 - Mouse Genome Informatics), causes hair bundle misorientation and shortening of the cochlear duct without affecting the asymmetric distribution of PCP components (Jones et al., 2008; Sipe et al., 2013; Sipe and Lu, 2011). In addition, there is a genetic interaction between Ift88 and Vangl2 in regulating HC polarity and convergent extension (Jones et al., 2008). By contrast, ciliary protein mutant mice, but not PCP mutant mice, show morphological changes in the hair bundles. From these findings, ciliary proteins may function either downstream of or parallel to the PCP pathway (Ezan and Montcouquiol, 2013). In Nectin-3^-/- mice, the establishment of convergent extension, the localisation of Vangl1 and Fz6, and the organisation of the striolar reversal zone in the vestibular epithelium were all normal; however, the orientation and morphology of the hair bundles were disrupted only in aberrantly attached HCs. These results suggest that PCP components do not significantly contribute to the abnormal phenotypes of Nectin-3^-/- HCs, but rather that some signal(s) initiated from ciliary proteins is altered by the abnormal attachments in Nectin-3^-/- HCs.

How are the signals regulating HC polarity disturbed by the aberrant attachments between HCs? Are there some common downstream signals between PCP components and ciliary proteins? One of the candidates for a common signalling pathway is the Rac1/p21-activated kinase (PAK) signalling pathway. This signalling pathway serves as an effector of the Wnt pathway in PCP (Habas et al., 2003) and ciliary proteins (Sipe et al., 2013; Sipe and Lu, 2011), and the trans-interactions of nectins induce the activation of Rac1 (Kawakatsu et al., 2002). In Madin-Darby canine kidney cells, Rac1 is transiently activated by nectins during the initiation of their trans-interactions and then inactivated once the trans-interactions are established (Kitt and Nelson, 2011). Rac1/PAK signalling regulates the organisation of the actin and microtubule cytoskeletons (Bokoch, 2003), and Rac1 knockout mice showed defects in HC morphogenesis similar to those observed in Nectin-3^-/- HCs (Grimsley-Myers et al., 2009). To evaluate the mechanism underlying HC polarity, we attempted to examine the activation of Rac1/PAK signalling in the Nectin-3^-/- auditory epithelium, but were unable to determine whether this pathway was activated (data not shown). Therefore, it remains to be elucidated whether Rac1 is activated only at crucial areas of HCs and in the crucial stages of development of the auditory epithelium when the orientation and morphology of the hair bundles and the positioning of the kinocilium are determined, although it can be speculated that the misorientation and dysmorphology of the hair bundles in the Nectin-3^-/- auditory epithelium may be caused by the ectopic activation of Rac1/PAK signalling.

The ABP component Patj binds to nectins (Adachi et al., 2009) and forms a ternary complex with Pals1 and Crb3 (Makarova et al., 2003). This ternary complex is upstream of the apical ABP complex Par-3-Par-6-aPKC in epithelial cells (Hurd et al., 2003). Par-3 binds to nectins and regulates the formation of TJs cooperatively with afadin in epithelial cells (Ooshio et al., 2007; Takekuni et al., 2003). The ABP component Scribble is genetically associated with Vangl (Montcouquiol et al., 2003). However, the molecular links between AJC and PCP components remain unclear. Here, we showed that Pals1 and Par-3 were aberrantly localised in aberrantly attached Nectin-3^-/- HCs and that AJC formation was disturbed at the boundary between aberrantly attached HCs. At the boundary between aberrantly attached HCs, the distributions of the AJC components Nectin-1, Nectin-2, afadin, E-cadherin, β-catenin and ZO-1, and the ABP component Par-3, all of which were concentrated there, extended along the apicobasal axis. Pals1 localised to the apical surface of HCs was rotated towards the attached sites. Par-3 regulates microtubule dynamics and centrosome orientation through dynein (Schmoranzer et al., 2009) and the cadherin-catenin complex regulates microtubule dynamics (Harris and Tepass, 2010). Together with these previous findings, it can be speculated that the homophilic interaction of Nectin-1 at the boundary between aberrantly attached Nectin-3^-/- HCs alters microtubule dynamics through positional changes in AJC and ABP components. This is consistent with the previous findings that genetic deletion of microtubule-associated ciliary proteins such as Ift88, Kif3a and Lis1 resulted in a disruption of HC polarity and morphological changes in the hair bundles (Sipe et al., 2013; Sipe and Lu, 2011). Therefore, heterophilic interaction of nectins probably leads to an elaborate checkerboard-like cellular pattern by arranging the localisation of AJC and ABP components and by controlling microtubule dynamics and, thus, contributing to the formation of HC polarity and the morphology of the hair bundles.

The detailed molecular mechanisms underlying the orientation and morphology of the hair bundles, the positioning of the kinocilium and the localisation of PCP components remain unknown, but we propose here a model for the abnormal positioning of the basal body and the kinocilium in aberrantly attached Nectin-3^-/- HCs, as schematically shown in Fig. 9. In the Nectin-3^+/- auditory epithelium, the PCP pathway is activated along the medial-lateral side axis by extracellular cues from PCP signalling during the maturation of the cochlea. Because Nectin-1 and -3 are uniformly distributed along the boundary between HCs and SCs, signals such as Rac1 and PAK should be appropriately activated at the lateral edge to propagate the PCP pathway. As a result, the basal body and the kinocilium (grey dot) move towards the lateral side of HCs (green arrows). In the Nectin-3^-/- auditory epithelium, however, at the contact sites between aberrantly attached HCs, Nectin-1 accumulates presumably because Nectin-1 molecules can trans-interact with each other but not with Nectin-2, and the signals may be activated by the Nectin-1-afadin complex; therefore, the basal body and the kinocilium move towards the sum (brown arrows) of the signals induced by the extracellular cues of PCP signalling (green arrows) and the signals induced by aberrant attachments between HCs (blue arrows), thus causing the aberrant positioning of the basal body and the kinocilium and the subsequent misorientation and dysmorphology of the hair bundles. If this is the case, these mechanisms may provide insights into why HCs are interdigitated with SCs to form a checkerboard-like cellular pattern in the auditory epithelium. The separation of HCs by SCs, which leads to the checkerboard-like cellular patterning of these two types of cells, is at least required for HC polarity.

Nectin-1 (Pvrl1) knockout (Nectin-1^-/-) and Nectin-3 (Pvrl3) knockout (Nectin-3^+/- and Nectin-3^-/-) mice were generated as described (Inagaki et al., 2005). The animal experiments were approved by the Institutional Animal Care and Use Committee and carried out according to the Kobe University Animal Experimental Regulations.

The following antibodies (Abs) were used: mouse anti-acetylated tubulin monoclonal Ab (mAb) (T6793, 1:500, Sigma-Aldrich); rat anti-mouse Nectin-1 mAb (D146-3, 1:200, MBL); rat anti-mouse Nectin-2 mAb (D083-3, 1:200, MBL); rabbit anti-l-afadin polyclonal Ab (pAb) (A0349, 1:200, Sigma-Aldrich); rabbit anti-γ-tubulin pAb (T3195, 1:200, Sigma-Aldrich); rabbit anti-myosin VIIa pAb (25-6790, 1:400, Proteus BioSciences); rabbit anti-Vangl1 mAb (HPA025235, 1:2000, Sigma-Aldrich); mouse anti-Fz6 mAb (1:5000, a gift from Dr J. Nathans, Johns Hopkins University, MD, USA); rat anti-E-cadherin mAb (1:500, a gift from Dr M. Takeichi, RIKEN, Kobe, Japan); mouse anti-β-catenin mAb (1:400, a gift from Dr M. J. Wheelock, University of Nebraska, NE, USA); mouse anti-ZO-1 mAb (339100, 1:100, Invitrogen); rabbit anti-protein associated with Lin-7 (Pals1) pAb (sc-33831, 1:100, Santa Cruz Biotechnology); and rabbit anti-Par-3 pAb (07-330, 1:400, Millipore). The primary Abs were visualised using donkey fluorochrome-conjugated secondary Abs (1:400). The fluorochromes used were Cy3 and Cy5 (Millipore). F-Actin was visualised using Alexa 488-conjugated phalloidin (1:100, Invitrogen).

For whole-mount immunofluorescence microscopy, dissected organs were fixed with 2% paraformaldehyde (PFA) in Hanks’ balanced salt solution at room temperature (RT) for 30 minutes, washed in PBS, and then permeabilised with 0.2% Triton X-100 in PBS for 15 minutes. The samples were blocked with a blocking solution containing 10% normal donkey serum and 1% bovine serum albumin in PBS at RT for 30 minutes, followed by incubation with primary Abs in the blocking solution at 4°C overnight. The samples were washed with PBS, then incubated in secondary Abs in Can Get Signal Solution A (Toyobo Life Science) at RT for 75 minutes. The samples were flat-mounted on glass slides with glycerol gelatine (Sigma-Aldrich). A z-stack of whole-mount cochleae was imaged using a confocal microscope (LSM700; Carl Zeiss). Fluorescence intensity was measured using ImageJ software (NIH).

Mouse cochlear samples were fixed with 2.5% glutaraldehyde (GA) and 2% PFA in 0.1 M phosphate buffer (PB) at RT for 30 minutes, and then prepared using the osmium-thiocarbohydrazide-osmium-thiocarbohydrazide-osmium method as previously described (Hunter-Duvar, 1978). The samples were viewed by scanning electron microscopy.

Mouse cochlear samples were fixed with 2.5% GA and 2% PFA in 0.1 M PB at RT for 1 hour. They were washed in PBS, transferred to a 25% sucrose solution, then embedded in OCT compound and frozen on dry ice. The samples were viewed with transmission electron microscopy.

For quantification of the orientation of the hair bundles, HCs at the basal turns of Nectin-3^-/- and littermate control cochleae were analysed at P1. To evaluate the angle of the orientation of the hair bundles, the angle formed by the intersection of a line connecting the free ends of the stereocilia and a line parallel to the pillar cells was measured using ImageJ software.

We thank Drs J. Nathans (Johns Hopkins University), M. Takeichi (RIKEN) and M. J. Wheelock (University of Nebraska) for the generous gifts of reagents; Drs T. Uemura (Kyoto University) and M. Furuse (Kobe University) for helpful discussions and critical reading of the manuscript; Dr H. Sakaguchi (Kyoto Prefectural University of Medicine) for performing the osmium-thiocarbohydrazide-osmium-thiocarbohydrazide-osmium method; and Dr S. Kitajiri (Kyoto University) for analysis of the vestibular system.