Integrin α8 and Pcdh15 act as a complex to regulate cilia biogenesis in sensory cells

Non-motile or primary cilia are centriole-derived, microtubule-based projections present in most metazoan cell types that are involved in sensory processes such as mechanosensation, chemosensation and photosensation (Leroux, 2007; Falk et al., 2015). The formation of cilia includes the assembly of the axoneme and the directional transport of ciliary proteins by membranous and non-membranous trafficking (Nachury et al., 2007; Westlake et al., 2011). In recent years significant progress has been made in the identification of components involved in vesicular trafficking necessary for primary ciliogenesis (Leroux, 2007; Pazour and Bloodgood, 2008). Among the proteins identified, Rab8a, a member of the Rab family of small GTPases, plays a critical function (Deretic et al., 1995; Nachury et al., 2007). To promote cilia biogenesis, Rab8 needs to be activated by its guanine nucleotide exchange factor, Rabin8 (also known as RAB3IP), which is targeted to the base of the cilia by GTP-Rab11-positive vesicles (Nachury et al., 2007; Westlake et al., 2011; Hehnly et al., 2012). In vertebrates, primary cilia are implicated in several developmental pathways and act as signaling centers mediating intercellular communications (Ezratty et al., 2011; Breunig et al., 2008; Delling et al., 2016). Dysfunction of cilia, due to mutations in ciliary proteins or proteins involved in vesicular transport, is associated with a broad spectrum of human disorders that affect sensory neuron physiology (hearing/balance, olfaction and vision), as well as organ development and function (Delmaghani et al., 2016; Grati et al., 2015; Jagger et al., 2011; Rachel et al., 2012).

Hair cells, the neurosensory cells of the auditory and vestibular systems, are the mechanosensors for the perception of sound and head movements. Projecting from their apical surface is the hair bundle, which consists of rows of ascending height actin-filled stereocilia tethered to a single primary cilium, the kinocilium (Cosgrove and Zallocchi, 2014). During hair bundle development the kinocilium is physically connected via extracellular linkages formed between cadherin-23 (Cdh23) and protocadherin-15 (Pcdh15), which are essential for proper hair bundle integrity and function (Kazmierczak et al., 2007; Webb et al., 2011). Mutations in CDH23 or PCDH15 are associated with Usher syndrome type I and non-syndromic hearing loss in humans (Cosgrove and Zallocchi, 2014). Although the kinocilium is important for the establishment of hair bundle polarity (Jones and Chen, 2008,), it may also play additional roles as a modulator of mechanotransduction activity in immature hair cells as well as a linkage coupling the hair cell bundle to components of the extracellular matrix (ECM) (Roberts et al., 1988; Kindt et al., 2012). Recently, it has been shown that mutations in two kinociliary proteins, Cdc14a and Dcdc2, are associated with human recessive deafness (Delmaghani et al., 2016; Grati et al., 2015). Zebrafish morphants (MOs) for dcdc2 and cdc14a showed kinocilium abnormalities with the concomitant defects in hair cell morphology and function, reinforcing the notion of a direct involvement of primary cilia in hair cell function (Delmaghani et al., 2016; Grati et al., 2015).

Integrins are heterodimeric cell surface receptors composed of α and β subunits that function as adhesion molecules by binding extracellular matrix (ECM) proteins and as receptors by mediating signal transduction (Müller et al., 1997). In particular, integrin α8 (Itga8) has an obligatory association with the β1 subunit (Itgb1; Müller et al., 1997) and is selectively incorporated into the apical membrane of hair cells during development where it is thought to initiate the assembly of transmembrane complexes necessary for the maturation of apical structures (Littlewood-Evans and Müller, 2000). Itga8-deficient mice have balance problems, abnormal stereocilia and fusions between stereocilia and kinocilium, with perturbations in the distribution of ECM components. At the molecular level, Itga8 modulates actin stress fiber assembly via a Rhoa-dependent mechanism (Zargham et al., 2007a,b; Benoit et al., 2009). Mutations in Itga8 have been associated with bilateral renal agenesis and Fraser syndrome (Humbert et al., 2014; Talbot et al., 2016).

The present work focuses on the kinocilium of neuromast hair cells. Using zebrafish as the experimental model, we demonstrated ciliary localization and an association between Itga8 and Pcdh15. Loss of Itga8 or Pcdh15 function leads to a common phenotype, including kinociliary length dysregulation, impairment of endocytosis, and Rab8 and centrin mislocalization. These defects can be explained by a reduction in Rhoa activation, since constitutively active Rhoa is able to rescue these defects in Itga8 and Pcdh15a knockdown and mutant zebrafish.

Preliminary results from our laboratory performed in mouse auditory hair cells suggested the existence of a functional Itga8–Pcdh15 complex. To extend these findings in a more suitable model, we decided to analyze whether defects in Itga8 or Pcdh15a proteins resulted in zebrafish hair cell abnormalities. Knockdown zebrafish for both of these proteins were generated for by the injection of sub-optimal doses of morpholino suspensions into one-cell stage eggs (hereafter denoted MOs), and analyzed at 3 days post fertilization (dpf). When studying their gross morphology, ∼30% of the MOs (Itga8 or Pcdh15a) showed pericardial edema and slight body curvature (Fig. S1A–E′ and Table S1). Since these defects were not observed in the pcdh15a zebrafish mutant lines (orbiters, Seiler et al., 2005) (Fig. S1F–H), they were considered morpholino off-target effects and excluded from our experiments. Apical hair cell morphology was analyzed in control, MOs and orbiter mutants by co-staining for phalloidin (a hair cell bundle marker) and acetylated tubulin (an axoneme marker) (Fig. 1A–O). Super-resolution structured illumination microscopy (SR-SIM) analysis showed a significant reduction of the kinociliary length in the itga8- and pcdh15a-deficient animals (Fig. 1B,C,I–L) compared to that in the corresponding controls (Fig. 1A,H,K,L). The orbiter mutations not only resulted in an average reduction in the kinociliary length (Fig. 1L) but also in a dysregulation of ciliogenesis in general, as determined by the broader variation of the individual ciliary lengths (Fig. 1M) and by a shift in the distribution of the kinociliary length frequencies towards shorter kinocilia (Fig. 1N). The fact that pcdh15a MOs and mutants showed similar kinociliary defects demonstrates the specificity of the MO phenotype. Since Pcdh15 is required for proper hair cell mechanotransduction channel activity (Kazmierczak et al., 2007) and to exclude the possibility that the shortening of the kinocilium may be the result of mechanotransduction channel impairment, we analyzed kinociliary length in myosin VIIA mutants (mariner^tc320b; Ernest et al., 2000). The results obtained from these animals (Fig. 1L) demonstrated that although mechanotransduction activity is impaired (Ernest et al., 2000), kinociliary lengthening is not, suggesting an independency between both processes.

Kinociliary length was normal when the MOs were co-injected with the corresponding cRNAs for Itga8 or Pcdh15a (Fig. 1D,E,K,L), demonstrating the specificity of the morpholino effect. One of the immediate downstream effectors in Itga8-signaling cascade is the small GTPase Rhoa (Zargham et al., 2007a,b; Benoit et al., 2009). Since Rhoa ciliary localization and involvement in primary ciliogenesis has been demonstrated (Liu et al., 2007; Pan et al., 2007; Hernandez-Hernandez et al., 2013), we decided to assess whether constitutively active (CA) Rhoa (Schaefer et al., 2014; Farina et al., 2016) was able to rescue the Itga8 MO phenotype. The zebrafish genome possesses five rhoa genes, but only rhoab and rhoad are expressed at early developmental stages (Zhu et al., 2006, 2008). cRNAs coding for CA rhoab or rhoad (Schaefer et al., 2014; Farina et al., 2016) were co-injected with the Itga8 morpholino suspension and the kinociliary length analyzed in these animals. We observed rescue of the morphant phenotype when animals were co-injected with CA rhoad only; suggesting that Itga8 effect on cilia elongation is mediated by this GTPase (Fig. 1F,G,K).

The number of cells harboring a kinocilium was quantified in MOs and orbiter mutants (Fig. S2A). No significant differences were observed at 3 dpf. However, at 5 dpf orbiters showed a significant reduction (∼30–40%) in the fraction of cells carrying a cilium compared to that in wild-type (WT) animals. Since the number of mature hair cells per neuromast was similar across treatments, and because we did not observe any increase in hair cell death as judged with a TUNEL assay (Fig. S2B–E), these results suggest that Pcdh15a and Itga8 proteins are not only required for efficient cilia elongation in neurosensory cells but also for proper cilia maintenance.

To confirm the activation state of Rhoa, rhotekin pulldown and immunofluorescence experiments were performed in MOs and mutants (Fig. 2; Fig. S3). Total lysates from 1–2 dpf larvae were prepared from controls, Itga8 MOs, Pcdh15a MOs and the corresponding rescued MOs (+cRNA), and Rhoa activation analyzed by pulldown assay (Fig. 2A). We observed a significant decrease in active Rhoa abundance (GTP-Rhoa, framed area in Fig. 2A–C) when Itga8 or Pcdh15a were knocked down compared to controls. This deficiency in Rhoa activation was rescued when MOs were co-injected with the full-length cRNAs for itga8 or pcdh15a, confirming the direct involvement of these proteins in Rhoa regulation. The Rhoa activation state was also evaluated in the orbiter lines at 5 dpf (Fig. 2B). Again, we observed a significant decrease in the amount of GTP-Rhoa (Fig. 2B, framed area) when comparing the pull down results from mutant and WT animals (Fig. 2F), directly implicating Pcdh15a in Rhoa activation and at the same time confirming the specificity of the Pcdh15a MO phenotype. Finally, total lysates from 1–2 dpf larvae injected with the itga8 MO suspension alone or in combination with the CA rhoa cRNAs were employed to validate the activation state of Rhoa (Fig. 2C,G). As expected, co-injection of CA Rhoab (Itga8 MO+b) or CA Rhoad (Itga8 MO+d) resulted in a significant increase of active Rhoa (GTP-Rhoa, Fig. 2C, framed area) compared to Itga8 MOs. However, when both CA rhoa cRNAs were injected together (Itga8 MO+b&d) we observed a modest, although not significant, recovery of GTP-Rhoa, more likely due to the strategy employed to generate these animals (see Materials and Methods). To confirm the downregulation of Itga8 in these MOs and in the CA rhoa cRNAs co-injected MOs, western blot experiments were run in parallel with the pulldowns. The results presented in Fig. 2C (Itga8 immunoblot) demonstrate that GTP-Rhoa was increased despite a reduction in Itga8 protein abundance. Overall the data point to the existence of an interrelationship between Itga8 and Pcdh15a protein expression/function and Rhoa activation.

Rhoa localization was assessed in neuromast hair cells through immunofluorescence (Fig. S3) with a mouse monoclonal antibody (anti-Rhoa-1) generated and qualified by our laboratory (Fig. S4I–K). A significant reduction in the ciliary fluorescent intensity was observed in the Itga8 MOs compared to that in controls (Fig. S3A,B,K). When Itga8 MOs were co-injected with the full-length itga8 or CA rhoa cRNAs we observed a modest (although not significant) recovery of the fluorescence (Fig. S3C–E,K) compared to MOs values. On the other hand, Pcdh15a MOs showed a significant increase of Rhoa-positive fluorescence that returned to normal values when the MOs were co-injected with the corresponding cRNA (Fig. S3F,G,L). Similar results were observed with the orbiter strong animals, that is an increase in Rhoa-associated fluorescence, while orbiter weak animals did not show any significant changes compared to WT animals (Fig. S3H–J,L). The differences between the orbiter lines may, very likely, be the result of the different mutant alleles expressed in these animals (see Materials and Methods). The findings suggest that the absence of Itga8 or Pcdh15a protein function results in Rhoa inactivation and its mislocalization from or to the cilia. Again, the fact that we observed the same phenotype in pcdh15a MOs and mutants, in both pulldown and immunofluorescence experiments, confirms the specificity of the MO effect.

Since itga8- and pcdh15a-deficient animals showed kinocilia abnormalities (Fig. 1), we investigated Itga8 and Pcdh15a localization in hair cells. Kinocilia localization of Pcdh15 has already been reported for murine hair cells as well as its role in kinociliary linkage formation and planar cell polarity (Webb et al., 2011; Lelli et al., 2010). Likewise, apical localization of integrin α8β1 in immature hair cells has already been described, with Itga8 mutant mice showing abnormalities in hair bundle and kinocilia morphologies (Littlewood-Evans and Müller, 2000). Antibodies against Itga8 and Pcdh15a (CD1 variants, Maeda et al., 2017) were developed and qualified by our laboratory (Fig. S4A–H′). At 1 dpf in the inner ear (Fig. 3A–B′), we observed Pcdh15a and Itga8 colocalization in the hair cell precursors (Tanimoto et al., 2011), with Pcdh15a present at the hair cell bundle (asterisk), while Itga8 localized in both the hair cell bundle (asterisk) at and the cilium (arrowhead). Pcdh15a and Itga8 localization was assessed at up to 5 dpf in hair cells (Fig. 3C–D′). SR-SIM showed colocalization of both proteins in the hair cell bundle and kinocilia from WT neuromasts, with an accumulation towards the tip of the cilia (Fig. 3D,D′,I). Apical colocalization of both proteins was also detected in the orbiter lines (Fig. 3E,H′,J). However, we observed a mislocalization or re-distribution in those lines, with few hair cells showing a positive hair bundle (3G–H′, asterisks) or cilia immunostaining (Fig. 3E,E′,G,G′). Instead, we observed a positive signal around the hair cell bundle of what seems to be the cuticular necklace, a region known to be highly enriched with transport vesicles (Kachar et al., 1997). These differences in the distribution of Itga8 and Pcdh15a in the orbiter lines were frequently observed within the same animal, demonstrating heterogeneity within the mutant phenotype. However, one observation was consistent, pcdh15a mutant neuromasts with short kinocilia lacked Itga8 and Pcdh15a ciliary localization (Fig. 3F,F′,H,H′).

Itga8 and Pcdh15 ciliary localization and regulation was also analyzed in hTERT-RPE transfected cells. We not only observed ciliary localization for these two proteins (Fig. S5C–E), but also an increase in ciliary length when co-transfected (Fig. S5F), corroborating the direct involvement of Itga8 and Pcdh15a in ciliogenesis.

Since defects in Itga8 and Pcdh15a proteins resulted in similar ciliary phenotypes and because both proteins colocalize in hair cells, we decided to analyze whether there was an interdependency in their expression and/or localization. Immunoblots from 3 dpf morpholino controls (Fig. 4A, top blots) showed several bands for Itga8 between 100–150 kDa corresponding to the predicted molecular mass for the full-length Itga8 (∼118 kDa, GenBank: AEU12477.1) and likely representing different degrees of glycosylation (hereafter denoted the ‘multiplex’, red asterisk). Smaller bands (∼90 kDa, ∼75 kDa and ∼50 kDa, black asterisks) were also detected. In the case of Pcdh15, several variants have been described in mouse and zebrafish (Ahmed et al., 2006; Maeda et al., 2014; Reiners et al., 2005; Zallocchi et al., 2012a,b). Pcdh15a immunoblots from controls (Fig. 4A, bottom blots) showed expression of the full-length protein (Pcdh15a-CD1, ∼200 kDa, red asterisk) and three additional bands (100 kDa, 50 kDa and 40 kDa, black asterisks). Itga8 MOs had a reduction of all the Itga8 variants, while Pcdh15a MOs only showed a decrease of the full-length and the 100 kDa Pcdh15a variants (Fig. 4A; quantification in Fig. S6). Surprisingly, rescue with the corresponding full length cRNA (+cRNA), not only restored expression of the full-length protein but of the small variants (Fig. 4A; Fig. S6), suggesting they might arise from post-translational modifications. We note that the 50 kDa and 40 kDa Pcdh15a variants were not reduced in the specific MOs (Fig. 4A, bottom right blot; Fig. S6), suggesting that these variants may lack the morpholino target sequence or, alternatively, that they may represent non-specific bands. Finally, when looking at the expression of the reciprocal protein, we observed a reduction of the full-length Itga8 (multiplex, red asterisk), and 90 kDa and 50 kDa variants in the Pcdh15a MOs (Fig. 4A, top right blot) and a reduction of the Pcdh15a full-length and 100 kDa variant in the Itga8 MOs (Fig. 4A, bottom left blot). Re-expression of Pcdh15a resulted in recovery of the Itga8 small variants but not the full-length protein, while we observed a modest recovery of Pcdh15a variants (full length and 100 kDa) in the Itga8 rescue MOs (Fig. S6). In the case of the orbiter lines, full-length Pcdh15a was reduced (Fig. 4B, bottom blot, red asterisk), while an additional prominent band (more likely a truncated form, blue asterisk) was observed. In contrast to what we found for the Pcdh15a MOs, we did not observe a reduction of the 100 kDa Pcdh15a band (Fig. 4B, bottom blot, black asterisk) nor of the Itga8 variants (Fig. 4B, top blot) in the orbiter lines.

Ciliary localization of Itga8 and Pcdh15a was addressed in the MOs by confocal microscopy analysis (Fig. 4C–G′). Lack of Itga8 in the Itga8 MOs resulted in loss of Itga8 and Pcdh15a ciliary distribution (Fig. 4D,D′). Likewise, both proteins were absent from the kinocilium in the Pcdh15a MOs (Fig. 4F,F′). The lack of Pcdh15a ciliary distribution in the specific MOs that at the same time showed a decrease in ciliary length suggests that the full-length and/or the 100 kDa variants are the forms involved in kinocilium elongation. Finally, co-injection with the corresponding full-length cRNAs (Fig. 4E,E′,G,G′) restored ciliary localization of both proteins (arrowheads in merged images).

Collectively, the results demonstrate interdependency between Itga8 and Pcdh15a in terms of protein abundance and targeting, pointing to the existence of a putative protein complex in hair cells.

The Pcdh15a–Itga8 interaction was assessed by co-immunoprecipitation (co-IP) and by proximity ligation assay (PLA, Wang and Deretic, 2015) (Fig. 5). Reciprocal co-IPs showed an interaction: Itga8 immunoprecipitated the full-length and the 100 kDa variant of Pcdh15a (Fig. 5A, asterisks), while Pcdh15a immunoprecipitated the multiplex corresponding to the full-length Itga8 protein (Fig. 5D, red asterisk) and the ∼90 kDa variant (black asterisk). Co-IP with irrelevant antibodies, normal guinea pig serum (IP NGpS) and normal rabbit serum (IP NRS), confirmed the specificity of the interaction. The direct immunoprecipitations (Fig. 5B,C) showed that we were able to immunoprecipitate some of the corresponding variants for both Itga8 and Pcdh15a, but not all of them; this might be due in part to the inaccessibility of the antibody to the recognition site. PLA experiments (Fig. 5E–I) performed in 5 dpf WT fish (Fig. 5E) confirmed the existence of an Itga8–Pcdh15a complex (red dots) in hair cells. We observed the presence of the complex in the cell body and the apical aspect of hair cells (Fig. 5E, high magnification inset), suggesting that these proteins are being transported together. PLA results employing the orbiter lines (Fig. 5F,G) showed that although some Pcdh15a and Itga8 variants are still present (Figs 3–4) they are not forming a complex (no red dots, Fig. 5F,G). Fig. 5H shows a representative images of a PLA-negative control in which one of the primary antibodies was omitted.

These findings strongly point to an in vivo interdependency between Pcdh15a and Itga8 in hair cells, where reduced expression of one of the proteins affects stability and/or targeting of the other. Although there were some apparent discrepancies between Pcdh15a MOs and mutants (i.e. differences in the Pcdh15a variants that are being knocked down and presence/absence of Itga8 variants) that may be the result of the different strategies used to reduce Pcdh15a function, the final outcome is the same, an obligatory interaction between Itga8 and Pcdh15a for efficient protein complex localization to the apical aspect of hair cells. When comparing the pattern of protein variants expressed by Pcdh15a mutants and MOs, we observed that the abundance of the small variants was unchanged (50 kDa and 40 kDa variants) in both types of animals (Fig. 4A,B), correlating directly with the presence of positive immunofluorescence at the apical aspect of the neuromast hair cells (Figs 3–4). Since Pcdh15 apical localization has been addressed by different groups (Kazmierczak et al., 2007; Lelli et al., 2010; Webb et al., 2011; Zallocchi et al., 2012a,b; Maeda et al., 2014, 2017; Ogun and Zallocchi, 2014), it is very likely that the 50 kDa and 40 kDa bands are specific Pcdh15a variants still present in the mutants and MOs and not non-specific associations. Most importantly, Pcdh15a mutants and MOs showed a reduction of the full-length Pcdh15a (and 100 kDa variant in the case of the MOs) and a decrease in Pcdh15a ciliary distribution, which suggests that this is the variant(s) relevant for kinocilia elongation. As for Itga8, although no experiments were performed in Itga8 mutant lines, the fact that we observed the same phenotype in the absence of Itga8–Pcdh15a protein complex and that we were able to rescue those defects with the corresponding cRNAs, strongly points to the specificity of the Itga8 MO phenotype.

A functional role for the Usher proteins in the regulation of vesicular trafficking has been reported by different groups (Tian et al., 2014; Prosser et al., 2008; Yan et al., 2001; Sorusch et al., 2017). More importantly, primary cilia biogenesis has been linked to vesicular trafficking and to actin-based cytoskeleton rearrangements (Molla-Herman et al., 2010; Luo et al., 2012; Ghossoub et al., 2011). Given the reduction in kinociliary length and the decrease in active Rhoa when the Itga8–Pcdh15a complex is not present, we next addressed whether there is a relationship between these defects and the endocytic/recycling pathway. For this purpose, long-term FM1-43 incorporation (Fig. 6A–O) was used as a proxy for hair cell endocytosis (Ogun and Zallocchi, 2014). Downregulation of Itga8 (Fig. 6A,B) or Pcdh15a (Fig. 6H–I) resulted in a significant decrease in FM1-43 incorporation by the hair cells (Fig. 6G,N). Similar to the Pcdh15 MOs, 5 dpf orbiter mutants also showed a significant reduction in dye uptake (orbiter strong ∼85% reduction, orbiter weak ∼60-75% reduction) compared to WTs (Fig. 6K–N). After FM1-43 incubations of up to 90 min, we did not observe any increase in the fluorescence incorporated by the hair cells in MOs and mutants, suggesting that absence of the protein complex results in complete blockage of the endocytic/recycling pathway (Fig. S7). While co-injection with the itga8 cRNA rescued the phenotype (Fig. 6C,G), pcdh15a cRNA co-injected animals did not show any differences in the fluorescence incorporated compared to the specific MOs (Fig. 6J,N). Furthermore, no recovery was observed in these animals after 90 min (Fig. S7B,C,H), indicating that the modest re-expression of Pcdh15a (Fig. 4A, bottom right blot) may not be sufficient to rescue the endocytic defect. Analysis of dye uptake in Itga8 MOs co-injected with the CA Rhoa proteins resulted in partial but significant recovery (∼50%) of the fluorescence incorporated compared to MOs. The fluorescent intensity reached normal values (∼100%) when both CA proteins were co-expressed, suggesting that they are both involved in the regulation of the endocytic pathway in hair cells (Fig. 6D–F,G). The defects in FM1-43 uptake were specific to neuromast hair cells since different cell types present in the fish skin were able to incorporate the dye in the MO and mutant animals (Fig. S7J–O).

Given that previous works demonstrated a role for Rhoa in basal body and vesicular trafficking regulation (Pan et al., 2007; Bershteyn et al., 2010; Hernandez-Hernandez et al., 2013) and a relationship between Rhoa activity and Rab8a (Braun et al., 2015), we further characterized the ciliary defect in MOs and mutants by analyzing the status of vesicular (Rab8a, Rabin8 and Rab11a/b) and basal body/centrosomal proteins (γ-tubulin and centrin). As described previously (Westlake et al., 2011), we observed Rab8a ciliary localization in control animals (Fig. 7A,G,I). Lack of Itga8 or Pcdh15a proteins resulted in a significant reduction of ciliary Rab8a (Fig. 7B,E,H,J,AA) while there was an accumulation of Rab8a at the base of the cilia. Co-injection of the MOs with the corresponding cRNAs showed a significant recovery from the MO phenotype (Fig. 7C,F, AA). When Itga8 MOs were co-injected with the cRNAs coding for the CA Rhoa molecules, we only observed full recovery with CA Rhoad (Fig. 7D,AA) suggesting that this signaling molecule is the one involved in normal Rab8a ciliary transport. No correlation between Rab8a apical localization and protein abundance was observed between the different treatments (Fig. S8A), suggesting that the lack of ciliary Rab8 might be independent of the levels of total Rab8 expression. Because Rab8a ciliary localization depends on its centrosomal activation by Rabin8 (Nachury et al., 2007; Westlake et al., 2011), we assessed Rabin8 distribution in MOs and mutants. Similar to previous reports (Nachury et al., 2007; Westlake et al., 2011), we observed punctuate (vesicular) distribution of Rabin8 at the apical aspect of hair cells (Fig. 7K–P). Quantification of the fluorescent intensity (Fig. 7BB) did not reveal any significant differences, suggesting that Rabin8 is properly targeted to the periciliary region. Consistent with these results, the Rab11a/b distribution did not show any differences between controls and treated animals (Fig. S8C–G). Since centriolar satellites are also involved in the trafficking of components necessary for ciliation, we examined the possible contribution of the Itga8–Pcdh15a complex in centrin targeting to the centrosome (Nachury et al., 2007). While hair cells from controls showed the characteristic centrin punctate distribution corresponding to basal body/centriole localization (Laoukili et al., 2000; Trojan et al., 2008; Bachmann-Gagescu et al., 2015) (Fig. 7Q,W,Y), Itga8- and Pcdh15a-deficient animals showed defects in the transport and/or distribution of centrin (Fig. 7R,U,X,Z,CC). When Itga8 MOs were co-injected with the full-length cRNA (Fig. 7S) or CA rhoab cRNA (Fig. 7T), centrin localization to the centriole/basal body was restored compared to MOs (Fig. 7CC). On the other hand, pcdh15a rescued MOs did not show any recovery (Fig. 7U versus V, and Fig. 7CC), more likely due to the modest re-expression of the Pcdh15a variants (Fig. 4A). However, since we observed centrin defects in both pcdh15a MOs and mutants, this argues in favor of a specific morpholino effect. Centrin immunoblots for the different treatments did not show any clear correlation between its distribution at the apical aspect of hair cells and its abundance (Fig. S8B). Similar to the Rab8a results, this may suggest that proper localization of centrin by the Itga8–Pcdh15a complex may be independent of its abundance. Alternatively, since our immunoblots were performed employing whole-larva lysates, it is also possible that specific differences related to hair cells may have been masked by total centrin (and Rab8a) expression. The abnormalities in centrin distribution were not the result of altered basal body maturation or docking. Immunostaining with the basal body marker γ-tubulin (Fig. S8H–L) did not show any qualitative difference between controls and Itga8- or Pcdh15a-deficient animals, suggesting that in the absence of the Itga8–Pcdh15a complex, the cortical actin network involved in basal body docking is still intact (Pan et al., 2007) and that the defects in centrin localization are not due to abnormal basal body positioning but are more likely due to centrin transport.

Finally, to address whether the defects observed in Rab8a and centrin distribution were the result of Itga8–Pcdh15a protein complex impairment and not an indirect effect due to defective ciliogenesis, we analyzed the distribution of additional cilia-associated proteins. Fig. S8M–V shows that the axoneme-associated protein IFT54 (also known as TRAF3IP1; Bizet et al., 2015) and the transition zone-associated protein Cc2d2a (Bachmann-Gagescu et al., 2011) have a normal distribution in Itga8 MOs and Pcdh15a mutant animals, suggesting that the Itga8–Pcdh15a complex specifically regulates Rab8a and centrin distribution via a Rhoa signaling pathway.

The way an organism perceives its surrounding environment relies on sensory functions. The cilia, as the sensory organelles that project from the apical surface of neurosensory cells, are involved in key biological processes such as hearing and balance, vision and olfaction (Jagger et al., 2011; Jones and Chen, 2008; Tian et al., 2014; Jansen et al., 2015; Sorusch et al., 2014). In this work, we used zebrafish as an experimental model to demonstrate the existence of an Itga8–Pcdh15a protein complex that is involved in kinocilium biogenesis and maintenance in hair cells. Defects in Itga8 or Pcdh15a prevent the formation of the complex, leading to a decrease in Rhoa-mediated signaling with a concomitant reduction in kinociliary length as well as in the number of kinocilia per neuromast. Mechanistically, activated Rhoa promotes the vesicular transport of Rab8a into the cilium and the targeting of centrin to the centrosome.

Although a role in ciliogenesis for Pcdh15 was suggested more than 15 years ago (Murcia and Woychik, 2001), this is the first work presenting direct evidence of Pcdh15 involvement in ciliogenesis in neurosensory cells. In zebrafish hair cells, Pcdh15a (and Itga8) showed apical distribution with an accumulation towards the ciliary tip, far from the hair cell bundle, suggesting additional roles for Pcdh15a not related to the formation of the kinociliary link (Kazmierczak et al., 2007). Moreover, since the role of actin polymerization in ectosome release from the ciliary tip has been confirmed (Nager et al., 2017), it is appealing to speculate that the Itga8–Pcdh15a complex may be involved in this process through the activation of Rhoa.

Since Itga8 has an obligatory association with the integrin β1 subunit (Itgb1) (Müller et al., 1997), and because evidence for interactions between protocadherins and Itgb1 exists (reviewed in Yagi, 2008), our findings suggest the there is a ternary complex between Itga8 and Pcdh15a that is mediated through Itgb1. The presence of ciliary integrins (i.e. Itgb1) and their role in mechanosensation via ECM interactions has been previously documented, with the mechanical stimuli resulting in extensive ciliary signaling and ECM deposition (McGlashan et al., 2006; Praetorius et al., 2004; Seeger-Nukpezah and Golemis, 2012). In the case of neuromasts, they possess a gelatinous matrix structure, the cupula, anchored to the hair cells by the kinocilia (McHenry and van Netten, 2007). The presence of the Itga8–Pcdh15a complex along the ciliary membrane in intimate contact with the cupular ECM poses the question of its putative role as both a modulator of ECM protein deposition and as a mechanosensor receptor involved in the detection of external cues that, ultimately, will influence fish behavior.

While evidence exists for the activation of Rhoa by Itga8 (Zargham et al., 2007a,b; Benoit et al., 2009), the role of Rhoa in ciliogenesis is still controversial (Pan et al., 2007; Hernandez-Hernandez et al., 2013). We found that active Rhoa regulates the transport of ciliary cargo necessary for kinocilia formation and elongation in neurosensory cells. This is suggested because centrin and Rab8a are mislocalized upon Rhoa inactivation due to Itga8–Pcdh15a complex disruption. Rabin8 interacts and activates Rab8a (Nachury et al., 2007; Westlake et al., 2011; Lai et al., 2015) granting GTP-Rab8a access to the cilia compartment. If Rab8a is phosphorylated on Ser111, the Rabin8 interaction does not occur and, thus, Rab8a remains inactive (Lai et al., 2015). Since impairment of Itga8–Pcdh15a complex formation did not affect Rabin8 targeting to the pericentriolar recycling endosome (Nachury et al., 2007; Westlake et al., 2011), this suggests that the Rabin8–Rab8a interaction might be disrupted when the complex is absent, pointing to a possible role for Rhoa in the regulation of Rab8a phosphorylation in hair cells.

Centriolar satellites are cytoplasmic non-membranous granules involved in the recruitment of cargo to the base of the cilia (Nachury et al., 2007). The observation that centrin distribution is also affected, suggests that activation of Rhoa by the Itga8–Pcdh15a complex plays a more general role in the transport of ciliary components (i.e. it is acting as regulator of vesicular and non-vesicular trafficking).

The link between the cilium and endocytosis appears to be conserved in many ciliated organisms (Ghossoub et al., 2011; Mehta et al., 2014), with components regulating endocytosis also involved in ciliogenesis and ciliary signaling cascades (Luo et al., 2012; Rbaibi et al., 2012; Coon et al., 2012; Clement et al., 2013). Based on the above information, our data suggest that lack of Itga8–Pcdh15a complex formation leads to defects in cilia elongation and, as a result, in an impairment in endocytosis. Thus, by properly modulating ciliary lengthening and maintenance, the Itga8–Pcdh15a complex is indirectly regulating the endocytic pathway.

Although the effects observed in the pcdh15a MOs were specific (we observed similar defects with the pcdh15a mutants), we were unable to rescue endocytosis activity and centrin distribution in the MOs. This may be the result of the modest re-expression of Pcdh15a that was observed upon cRNA co-injection.

Based on our findings and in combination with multiple lines of evidence (Pan et al., 2007; Bershteyn et al., 2010; Hernandez-Hernandez et al., 2013; Nager et al., 2017; Lai et al., 2015; Farina et al., 2016), we propose a model (Fig. 8) in which the Itga8b1–Pcdh15a complex is activated by cupular ECM ligands. Receptor activation leads to an increased in GTP-Rhoa with its corresponding targeting to the ciliary or plasma membrane. Once activated, membrane bound-Rhoa modulates actin dynamics and, ultimately, the transport of cargo necessary for proper cilia growth and function.

In summary, our study lays the foundation for future work to uncover the identity of molecules downstream of Rhoa that lead to Rab8a and centrin ciliary localization, as well as the molecular requirements for Itga8–Pcdh15a complex formation and activation. Ciliary defects result in a diverse array of developmental abnormalities in humans collectively known as ciliopathies (Leroux, 2007; Rachel et al., 2012; Sorusch et al., 2014). The formation of this novel protein complex in the ciliary membrane of hair cells suggests the presence of a unique mechanism for the regulation of ciliogenesis in neurosensory cells and poses some major questions related to the possible function of similar complex(es) in other sensory systems, with potential implications for human pathologies.

Zebrafish (Danio rerio), AB and TL WT strains were grown at 28.5°C under standard conditions. Animal care and husbandry were overseen by the Institutional Animal Care and Use Committee at Boys Town National Research Hospital. Experimental larvae (1–5 dpf, both sexes) were grown in embryo medium (14 mM NaCl, 0.5 mM KCl, 0.25 mM Na₂HPO₄, 0.04 mM KH₂PO₄, 1.3 mM CaCl₂, 1 mM MgSO₄ and 4 mM NaHCO₃, pH 7.4) at 28.5°C with a 10-h-light–14-h-dark cycle. Animals were cryo-anaesthetized before the initiation of the experiments. The mutant lines orbiter^th263b, orbiter^tc256e and mariner^tc320b were provided by Teresa Nicolson (Oregon Health and Science University, Portland, OR) and have been previously described (Seiler et al., 2005; Ernest et al., 2000). The orbiter^th263b line (orbiter strong) harbors a non-sense mutation that results in acoustic and vestibular defects, while the orbiter^tc256e line (orbiter weak) harbors a missense mutation that results in a weak acoustic response. The mariner^tc320b line harbors a nonsense mutation in the myosin VIIA gene that leads to mechanotransduction channel activity impairment. Because homozygous animals for the orbiter and mariner mutation can be phenotypically screened when they start swimming, we were only able to use them between 4–5 dpf. Otic (O1 and O2), supraorbital (SO3), mandibular (M1 and M2), Middle (MI2) and Opercular (OP1) neuromasts were analyzed in this work.

Mismatch controls, scrambled and specific MO were designed by and purchased from GeneTools, OR. For the generation of MOs, suboptimal doses were injected in one-cell stage eggs. Itga8 MOs were generated by injection with 1 ng translation-blocking MO (5′-GGACAAACAAGAGTGCATGGCTTCA-3′) and 5 ng splice-blocking MO (5′-TAGTGTTTATGTGTTTCTGTAGGCC-3′). Pcdh15a MOs were generated by injection with 10 ng of translation-blocking MO (5′-CCTCCGCATCTTCACTTAATGCCTA-3′). For the generation of rhoab MOs, a translation-blocking MO (5′-CTTCTTGCGAATTGCTGCCATTTTG-3′) was used at 5.7 ng per injection (Zhu et al., 2008). In the case of rhoad MOs, a translation-blocking MO (5′-AGCTTCTTACGGATAGCTGCCAT-3′) was used at 8 ng per injection.

For the rescue experiments, morpholino-resistant cRNAs were generated as described previously (Ogun and Zallocchi, 2014). Rescue of Itga8 MOs was performed with 160 pg of itga8 cRNA while 180 pg of pcdh15a cRNA was used for Pcdh15a MOs. For the rescue of Itga8 MOs with CA rhoa cRNAs, suboptimal doses of 11.25 pg were used (Zhu et al., 2008). When co-injected together, each CA rhoa cRNA was at a dose of 8.4 pg.

Embryos or larvae were maintained in embryo media and life screened at room temperature using a stereomicroscope (MZ10F; Leica) with a Plan Apochromat 1.0× objective and a magnification of 6.3×. Bright field images were captured with a camera (DFC310 FX; Leica) and the Application Suite V4.0.0 acquisition software (Leica).

Total RNA was obtained by TRizol extraction of 3 dpf larvae followed by production of cDNA (SuperScript III, First-Strand kit, Thermo Scientific). PCRs for itga8 and pcdh15a were performed with the Expand Long Template PCR System (Roche), according to the manufacturer's instructions. PCRs for rhoab and rhoad were performed with the FastStart High Fidelity PCR System (Roche), according to manufacturer's instructions.

For the generation of a His-tagged itga8 construct containing the T7 promoter, two rounds of PCRs were performed. The first PCR used forward primer 5′-CTATAGGGCCCTTCAGCTAGCATGGACTATACCCGCACTCAC-3′ and reverse primer 5′-TGGTGGTGCGCCCTTGTGGATTTTTCT-3′. The second PCR used forward primer: 5′-ATTAATACGACTCACTATAGGGCCCTTCAGCTAGC-3′ and reverse primer 5′-TTTTATTTCAGTGGTGGTGGTGGTGGTGCGC-3′. The final fragment was cloned in a TOPO^® TA Cloning^® Kit (Thermo Scientific) and colonies screened for the construct with the right orientation. itga8 was also cloned in the CT-GFP Fusion TOPO^® Expression system (Thermo Scientific) with the following pair of primers: forward primer, 5′-ATGGATTACACTCGGACACACAGGGCTGAAG-3′; reverse primer, 5′-CCGCGTCTGTGGATTTTTCTGCGGTC-3′.

The pEF6/V5-His TOPO^® TA Expression system (Thermo Scientific) was used to clone pcdh15a for both generation of cRNA and expression in cell cultures. Primers designed for this purpose were as follows: forward primer, 5′-TGGGTTGGCAGTAGGCATTAAGTGAAGATG-3′; reverse primer, 5′-TACATCGTTCTTGTTGTCATATTTAACATCAGGGC-3′. Colonies were screened for the construct with the right orientation.

For the generation of constitutively active Rhoab and Rhoad, PCR products were obtained by using the following pair of primers: forward 5′-ATGGCAGCAATTCGCAAG-3′ and reverse 5′-TCACAGCAGACAGCATTTG-3′ for rhoab; forward 5′-ATGGCAGCTATCCGTAAGAAG-3′ and reverse 5′-TCATAACAGCAGGCAGC-3′ for rhoad. PCR fragments were cloned with the TOPO^® TA Cloning^® Kit (Thermo Scientific) and used as a template for site-directed mutagenesis (GeneArt^® Site-Directed Mutagenesis System, Thermo Scientific) of the glycine residue in position 14 to a valine residue. Primers were designed according to manufacturer's instructions. CA rhoab and rhoad were also cloned in the expression vector pAcGFP1-N1 (Clontech) employing the following pair of primers: forward NheI, 5′-ATGCTAGCATGGCAGCAATTCGCAAG-3′ and reverse XhoI, 5′-TACTCGAGCAGCAGACAGCATTTGTTGC-3′ for rhoab; forward NheI, 5′-GTGCTAGCATGGCAGCTATCCGTAAGAAGCTG-3′ and reverse XhoI, 5′-TACTCGAGTAACAGCAGGCAGCCGCT-3′ for rhoad.

All constructs were sequence verified by the University of Nebraska Medical Center DNA Sequencing Core Facility (Omaha, NE).

Guinea pig antibodies against an extracellular peptide sequence (REFESKPREVGRVYLY) of zebrafish Itga8 were developed under contract with Thermo Scientific. Rabbit antibodies against a C-terminal peptide (KNSDRFGCSPDVKYDNKNDV) of Pcdh15a-CD1 (homologous to human and mouse PCDH15; Seiler et al., 2005; Maeda et al., 2017) were developed under contract with Proteintech Group. Mouse monoclonal antibodies against a peptide sequence (VADIEVDSKQVELAC) common to both Rhoa proteins were developed under contract with GenScript (anti-Rhoa-1).

Other antibodies used in this study were: mouse monoclonal anti-actin, clone C4 (cat. #MAB1501, Millipore), mouse monoclonal anti-pan-centrin, clone 20H5 (cat. #04-1624, Millipore), mouse monoclonal anti-acetylated tubulin 6-11B-1 (cat. #T6793, Sigma-Aldrich), mouse monoclonal anti-γ-tubulin clone GTU-88 (cat. #T6557, Sigma-Aldrich), chicken anti-GFP (cat. #NB100-1614, Novus Biologicals), mouse polyclonal anti-Rabin8 (RAB3IP) (cat. #H00117177-B01P, Novus Biologicals), mouse monoclonal anti-Rab8a clone 3G1 (cat. #H00004218-M02, Novus Biologicals), mouse monoclonal anti-6x-His epitope tag HIS.H8 (cat. #MA1-21315, Thermo Scientific), rabbit anti-Rab11a/b (cat. #GTX128847, Gene Tex), mouse monoclonal anti-Rhoa 26C4 (cat. #sc-418, Santa Cruz Biotechnology), mouse monoclonal anti-V5 epitope (cat. #R960-25, Thermo Scientific), rabbit anti-IFT54 (cat. #ARP55316_P050, Aviva Systems Biology) and mouse anti-Cc2d2a (Bachmann-Gagescu et al., 2011).

HeLa and HEK293 cells (obtained from Dominic Cosgrove, Boys Town National Research Hospital, Omaha, NE) were cultured under standard conditions. Subconfluent cells were dissociated and electroporated with 5–10 µg of the corresponding vector and plated onto fibronectin-coated microscope slides at a ratio of 1:10. After 48 h, cells were fixed and prepared for immunofluorescence analysis.

The human telomerase reverse transcriptase retinal pigmented epithelial (hTERT-RPE) cell line was purchased from the ATCC and grow in DMEM/F12 containing 10% fetal bovine serum (FBS) and 0.01 mg/ml hygromycin B. Lipofectamine 2000 (Thermo Scientific) was used to transfect the cells with 5–10 µg of Itga8 or Pcdh15a constructs, and after 24 h switched to OptiMEM (Thermo Scientific) for 24-48 h to induce ciliogenesis. Cell were fixed and prepared for immunofluorescence analysis.

For co-immunostaining of Itga8, Pcdh15a and acetylated tubulin, 3–5 dpf zebrafish were fixed with 4% paraformaldehyde (PFA) in 1% PBSTw (PBS plus 1% Tween 20) overnight at 4°C. Samples were rinsed several times with PBSTw and blocked for 1 h at room temperature (RT) with PBSDT (0.25% Tween-20 and 1% DMSO in PBS) containing 3% bovine serum albumin (BSA) and 6% FBS. Incubation of primary antibodies (1:500) was done overnight in PBSTw 0.1%. After several washes, samples were incubated with the corresponding Alexa-Fluor-conjugated secondary antibodies and Alexa-Fluor 488-conjugated phalloidin (Thermo Scientific). Samples were rinsed thoroughly with PBSTw (0.1% Tween 20) and incubated for 30 min at RT with 4% PFA in PBS. After several washes, larvae were mounted for confocal or SR-SIM analysis.

For acetylated tubulin, Rab8 and Rab11a/b immunostaining, samples were processed as described previously (Ogun and Zallocchi, 2014).

For centrin, γ-tubulin and Rabin8 immunostaining, fish were fixed at 4°C for 4 h with 4% PFA, 4% sucrose and 0.01% Tween 20 in PBS and then rinsed twice with PBSDT. In the case of γ-tubulin immunostaining, samples were incubated with ice-cold acetone for 6 min at −20°C, washed for 5 min with distilled water (no rocking) and then washed two more times with PBSDT (rocking). Primary antibodies were diluted in PBSDT containing 3% BSA and 6% FBS at a 1:200 dilution, and samples incubated overnight with rocking at 4°C. After several washes, larvae were incubated with Alexa Fluor 594-conjugated anti-mouse-IgG antibody and Alexa-Fluor-488–phalloidin.

For Rhoa and IFT54 immunostaining, larvae were fixed with 4% PFA in PBS overnight at 4°C and rinsed several times with PBSTx (0.1% Triton X-100 in PBS). Samples were incubated for 1 h at 37°C in PBS without rocking and then with the primary antibodies against Rhoa-1 (1:200) or IFT54 (1:100) in PBSTx 0.1% containing 2.5% BSA and 2.5% FBS. Secondary antibodies were Alexa Fluor 594-conjugated anti-mouse-IgG or AlexaFluor 488-conjugated anti-rabbit-IgG.

Cc2d2a immunostaining was performed according to Bachmann-Gagescu et al. (2015).

HeLa and HEK293 cells were fixed with 4% PFA for 30 min at processed as described previously (Ogun and Zallocchi, 2014) but without permeabilization. Primary and secondary antibodies were used at 1:500 in blocking solution. Untransfected and transfected hTERT-RPE cells were fixed with 4% PFA, permeabilized for 7 min with PBS plus 0.1% Triton X-100 and prepared for immunofluorescence studies.

In all the cases, samples were mounted under coverslips using ProLong Gold antifade reagent with or without DAPI.

Zebrafish confocal images were captured at RT using a Zeiss LSM 800 confocal microscope employing the Airyscan function, except for in the dye-incorporation studies in which the regular confocal function was used. Z-stack images were acquired using the 63×, NA 1.4 oil objective at a 2×zoom, and with a sectioning set automatically to optimal. Z-stack cell confocal images were acquired using a 40×, NA 1.3 oil objective.

Zebrafish SR-SIM images were captured at RT using a Zeiss ELYRA PS.1 super resolution microscope. Z-stack images were acquired using a 63×, NA 1.4 oil objective, zoom 2×. Sectioning was set automatically to optimal.

Images were acquired and processed with ZEN 2 black edition software (Carl Zeiss). SR-SIM images were acquired and processed with the ZEN 2 blue edition software (Carl Zeiss). Z-stack images are presented as flat Z-projections. Only linear adjustments were made to brightness and contrast, and the final figures were assembled using Photoshop and Illustrator software (Adobe).

Long-term dye incorporation experiments were performed as described previously (Ogun and Zallocchi, 2014). Briefly, zebrafish larvae (3 dpf MOs and 5 dpf mutants) were pre-incubated for 10 min in the presence of BAPTA (Thermo Scientific) to disrupt the tip links and thus block FM1-43 entrance through the mechanotransduction channels, and then for 30 to 90 min with 3 µM of FM1-43FX (Thermo Scientific) in the presence of low BAPTA concentrations. After several washes, animals were fixed with PFA, counterstained with phalloidin and processed for fluorescence analyses.

At 5 days post-fertilization larvae were fixed with 4% PFA overnight, rinsed several times with PBSTx 0.1% and mounted in Optical Cutting Temperature compound for cryo-sectioning at 14 µm. PLA was performed by employing the Duolink^® In Situ Red Starter Kit Rabbit PLUS (Sigma-Aldrich) and according to manufacturer's instructions with the following modifications: coronal sections of 5dpf larvae were blocked with PBSTx containing 2.5% BSA and 2.5% FBS. Primary antibodies (anti-Itga8, anti-Pcdh15a and anti-acetylated tubulin) were diluted 1:200 in the same blocking solution. Secondary anti-guinea pig antibody (cat. #A18777, Life Technologies) was conjugated to the Probemaker MINUS probe as follows: antibody was dialyzed against PBS and then conjugated by employing the Duolink in situ Probemaker MINUS (Sigma-Aldrich) according to the manufacturer's instructions. Sections were incubated with anti-rabbit PLUS (1:100), anti-guinea pig MINUS (1:30), Alexa-Fluor 405-conjugated anti-mouse-IgG (1:300) and Alexa Fluor 488-conjugated phalloidin (1:300) in blocking solution. Samples were rinsed several times and mounted using using ProLong Diamond antifade reagent for confocal analyses. Negative controls were produced by omitting one of the primary antibodies (anti-Itga8 or anti-Pcdh15a).

The TUNEL assay was performed employing the in situ cell Death detection kit, TMR red (cat. #12156792910, Roche) and according to the manufacturer's instructions. During the final washes, larvae were incubated with Alexa Fluor 488-conjugated phalloidin. As positive control, 5 dpf WT larvae were incubated for 2 h with cisplatin 500 µM to induce apoptosis.

Co-IP experiments were performed under conditions that allow detection of primary and secondary associations according to Ogun and Zallocchi (2014) with some modifications. In brief, 30 μl of a 50% protein A–Sepharose CL-4B slurry (Sigma-Aldrich) was incubated overnight with 7.5 μl of rabbit anti-Pcdh15a, 7.5 μl of guinea pig anti-Itga8, or the corresponding pre-immune normal sera (normal rabbit serum was used as a specificity control for anti-Pcdh15a antibody, and normal guinea pig serum as a specificity control for anti-Itga8 antibody). Three days post-fertilization larvae were homogenized in co-IP buffer (10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.5 mM MgCl2, 0.5 mM CaCl2, and 1% Brij 97) containing protease and phosphatase inhibitors (Thermo Fisher Scientific), and 1-2 mg protein used for each co-IP.

Pulldown experiments (RhoA activation assay, cat. #BK036, Cytoskeleton, Inc.) were performed according to the manufacturer’s instructions with the following modification adapted for zebrafish. A total of 20–30 larvae (1–5 dpf) were homogenized in 170 µl cell lysis buffer. The pulldown was performed with 15 µl (25 µg) of Rhotekin-RBD protein beads and 200–500 µg of protein. Samples were run on a 15% acrylamide gel, transferred for 1 h and membranes were blocked for 1 h at RT with 3% milk. Membranes were incubated overnight with the primary antibody anti-Rhoa 26C4 (cat. #sc-418, Santa Cruz Biotechnology) at a dilution 1:250 in TBSTw (0.05% Tween 20). After several washes, the secondary horseradish peroxidase (HRP)-conjugated anti-mouse-IgG antibody was added at a dilution 1:3000 for 1 h at RT in TBSTw. Immunoblots were developed using Pierce ECL Plus Western blotting substrate. Total Rhoa immunoblotting was performed with 30–40 µg of protein. The primary antibody dilution was 1:500 and immunoblots were revealed using Pierce SuperSignal West Femto. His-Rhoa (5–10 ng, Cytoskeleton, Inc.) was included in each immunoblot as a control for specificity.

Embryos or larvae were deyolked and lysates processed as previously described (Ogun and Zallocchi, 2014). Protein samples were loaded at 15–35 µg per lane and transferred overnight. For Itga8 and Pcdh15a western blots, membranes were incubated in 5% milk blocking solution (0.1% Tween 20, 75 mM NaCl, 5% glycerol, and 5% non-fat dried milk in PBS) overnight. Primary antibody (1:1000) was added in blocking solution overnight with rocking. Membranes were rinsed and incubated with goat HRP-conjugated anti-rabbit-IgG (1:20,000) or goat HRP-conjugated anti-guinea-pig-IgG (1:10,000) secondary antibodies for regular western blots or with HRP-conjugated protein A (1:10,000; GE Healthcare) for co-IP experiments. Membranes were developed using Pierce ECL Western blotting substrate (Thermo Fisher).

For Rhoa immunoblots, employing the antibody developed by our laboratory (anit-Rhoa-1), membranes were incubated for 1 h at RT in 3% milk blocking solution and the primary antibody overnight at a dilution 1:200 in TBSTw 0.05%. Secondary antibody was used at 1:3000 in TBSTw 0.05%. For Rab8a and centrin-1, immunoblots membranes were blocked for 1 h at RT with 3% milk blocking solution and primary antibody diluted to the concentration suggested by the manufacturer. For pan-actin immunoblotting, membranes were stripped with western blot stripping buffer (Restore PLUS; Thermo Scientific) and probed as previously described (Ogun and Zallocchi, 2014).

hTERT-RPE cells were processed as described previously for other cell types (Zallocchi et al., 2012a,b), and 30 µg of protein used for western blot studies. Membranes were stripped and re-probed for actin as a loading control.

In all the cases, membranes were exposed to films and developed using a film processor.

Ciliary lengths were measured by employing ZEN 2 black edition software (Carl Zeiss), averaged for each treatment and presented as bar graphs. Individual kinociliary lengths are presented as scattered plot graphs. At least five independent experiments were performed; each containing at least ten fish and at least two neuromasts per fish were inspected. For hTERT-RPE cells, three independent experiments were performed. Ciliary length was measured in control cells and in transfected cells positive for ciliary Pcdh15a, Itga8 or both.

Three independent experiments were performed for the pulldown assays, co-IPs and the MOs/mutant immunoblots. Each lane in the blots corresponds to 15–60 larvae. Protein bands were quantified by using ImageJ software, version 1.51 g (NIH), and normalized to the total amount of Rhoa (tRhoa) (for pull downs) or actin (for regular immunoblots) and the numerical data expressed as a percentage of that in controls.

Two independent experiments were performed in the case of the controls, MOs, and rescued MOs immunostained for Itga8 and Pcdh15a. The number of neuromast hair cells showing Itga8 or Pcdh15a apical localization was expressed versus the total number of neuromasts inspected.

Two PLA independent experiments were performed with ten animals each, and one neuromast per animal was analyzed for the presence or absence of PLA staining.

Fluorescence intensities were calculated as described in Ogun and Zallocchi (2014) employing ImageJ software, and expressed as percentages of that in controls. For Rhoa and Rabin8 immunostaining, a region of interest (ROI) corresponding to the apical region of neuromasts hair cells was selected and the fluorescence calculated in at least three independent experiments. At least ten animals were analyzed per treatment and at least two neuromasts per animal. In the case of endocytosis experiments, the ROI corresponding to the whole neuromast was selected and the total fluorescence intensity calculated and plotted as a percentage of that in control. Five independent experiments were performed, with five to ten larvae per treatment and one or two neuromasts inspected per animal.

In the case of Rab8a immunostaining, its ciliary localization was qualitatively assessed, and the data expressed as a percentage of that in controls. For centrin immunostaining experiments, the punctate distribution that correlates with the point of insertion of the kinocilium (basal body/centriole) was qualitatively assessed and the final results plotted as percentages of that in controls. For both immunostainings, at least five independent experiments were performed with five to ten animals per treatment and at least one neuromast analyzed per animal.

To minimize bias during data processing, animal treatments were unknown to the person conducting the analyses.

Numerical data were graphically represented as scattered plots (individual values) or as bar graphs (means±s.e.m.). Statistical analysis was performed using Prism 5 (version 6.07). A non-parametric Kruskal–Wallis test followed by Dunn's multiple comparisons test was used to compare the proportion of ciliated hair cells. A two-tailed Student's t-test was used for the long-term endocytosis experiments (30–90 min) to compare Itga8 MOs versus controls and in all other experiments to compared Pcdh15a MO+cRNA versus Pcdh15a MOs. One-way ANOVA followed by Dunnett's multiple comparisons test were used to analyze the rest of the numerical data. Treatments were compared to control (statistics in black) and versus the corresponding MO (statistics in red).

We thank J. Taylor and J. Talaska for providing assistance with microscopy. Support for the UNMC Advanced Microscopy Core Facility was provided by the Nebraska Research Initiative, the Fred and Pamela Buffett Cancer Center Support Grant (P30CA036727), an Institutional Development Award (IDeA) from the NIGMS of the NIH (P30GM106397), the Nebraska Research Initiative Grant and Nebraska Center for Cellular Signaling CoBRE (NIH P30GM106397). We thank Dr T. Nicolson (Oregon Health and Science University, Portland, OR) for the mutant zebrafish lines, Dr S. Rocha-Sanchez (Creighton University) for cisplatin, R. Bachmann-Gagescu (University of Zurich, Switzerland) for the Cc2d2a antibody, O. Ogun for fish husbandry, D. Delimont for genotyping and S. Kennedy for figure preparation and art work. We also want to thank Dr D. Cosgrove (BTNRH) for critically reviewing the manuscript.