Ciliary proteins Bbs8 and Ift20 promote planar cell polarity in the cochlea

Primary cilia were originally thought to be vestigial organelles without specific function. However, recent research has demonstrated that defects in primary cilia cause a range of developmental defects and human disorders collectively termed ‘ciliopathies’ (Lee and Gleeson, 2011; Waters and Beales, 2011), with Bardet–Biedl syndrome (BBS) considered to be an archetype for these disorders (Forsythe and Beales, 2013). Cilia are microtubule-based appendages continuous with the cell membrane but extending away from the cell surface (Fisch and Dupuis-Williams, 2011). At the core of each cilium is the ciliary axoneme, a microtubule-based structure containing either 9+2 or 9+0 microtubule doublets surrounded by soluble matrix and ciliary membrane. The proximal ends of these doublets are anchored to the basal body, a structure derived from the mother centriole following mitosis (Kobayashi and Dynlacht, 2011). The ciliary basal body is also a microtubule-organizing center that regulates ciliary and vesicular trafficking at the luminal surface (May-Simera and Kelley, 2012b; Moser et al., 2010). Finally, a transition zone located at the base of the ciliary axoneme and overlapping with the basal body plays a key regulatory role. All these components are considered to be part of the cilium, with disruption leading to ciliary defects. Many ciliary proteins have been identified based upon tight association with cilia; however, emerging evidence suggests that these proteins have additional, non-cilia-related functions. For example, intraflagellar transport (IFT) proteins have been reported at non-ciliary locations, including the membranous Golgi and dendrites of retinal neurons (Finetti et al., 2009; Sedmak and Wolfrum, 2010; Yuan and Sun, 2013). Similarly, Bbs proteins are implicated in actin cytoskeleton regulation (Hernandez-Hernandez et al., 2013; Tobin et al., 2008) and are localized at actin-rich structures in cultured cells and mouse cochleae (Hernandez-Hernandez et al., 2013; May-Simera et al., 2009).

Cilia also contribute to intercellular signaling pathways, including the planar cell polarity (PCP) branch of the Wnt signaling pathway, but the current understanding of how cilia are associated with PCP signaling is unclear (Wallingford and Mitchell, 2011). Although intricately linked, the phenomenon of PCP is not completely synonymous with PCP signaling. PCP refers to the uniform orientation of cells within an epithelium (Simons and Mlodzik, 2008; Vladar et al., 2009). PCP signaling is the information flow that is required to achieve this orientation; it is also more narrowly defined as the system of signaling that produces asymmetric subcellular localization of core PCP proteins. Cilia were first implicated in PCP signaling after PCP-like phenotypes were identified in Bbs mutants (Ross et al., 2005). One theory is that cilia regulate a switch between PCP and canonical Wnt signaling via sequestration of signaling molecules near the basal body (Lienkamp et al., 2012; Simons et al., 2005; Veland et al., 2013). However, the causal relationship between the cilium and PCP signaling has not been elucidated (Wallingford and Mitchell, 2011). One complication is that the morphological response to PCP signaling is usually the localization of the primary cilium; therefore, mutations in ciliary proteins that affect ciliary location will affect PCP but not necessarily PCP signaling.

In vertebrates, a striking example of PCP is the uniform orientation of stereociliary bundles on mechanosensory hair cells of the inner ear (Ezan and Montcouquiol, 2013). Stereociliary bundles are composed of a specialized cilium, called the kinocilium, positioned adjacent to elongated actin-rich microvilli called stereocilia, based on historical convention. The stereociliary bundles are polarized and the appropriate positioning of kinocilia relative to stereocilia is required for normal hair cell function. Although the precise cellular processes that mediate bundle orientation are still being elucidated, a key step is thought to be the directed migration of the kinocilium to a lateral position on the apical hair cell surface (Cotanche and Corwin, 1991; Denman-Johnson and Forge, 1999). Six ‘core’ PCP proteins are essential regulators of bundle orientation that, when mutated, lead to varying degrees of mis-orientation owing to disrupted intercellular signaling (Montcouquiol et al., 2003; Wang et al., 2006). Cilia-related proteins, including Ift88 and Kif3a, have also been shown to be required for appropriate bundle orientation (Jones et al., 2008; Sipe and Lu, 2011). Interestingly, many of these mutants also demonstrate defects in outgrowth of the cochlear duct, a process that is believed to require convergent extension, a conserved morphogenetic process that is also regulated by the PCP pathway.

These studies implicate cilia and cilia-related genes in two PCP-dependent processes: stereociliary bundle orientation and cochlear outgrowth. However, whether the phenotypes observed are a result of disruptions in the PCP process, the formation of cilia or both is unclear. To address these issues, we determined changes in stereociliary bundle PCP and cochlear outgrowth in mice with mutations in several, different classes of cilia-related genes. Our results suggest that a subset of cilia-related genes is required for trafficking of PCP molecules to the cell membrane in a cilia-independent manner and, as such, might play a more global role in protein trafficking within the cell.

Cochlear outgrowth (Fig. 1A) and stereociliary bundle orientation (Fig. 1B), two aspects of inner ear development known to be mediated through the PCP pathway, were analyzed using mouse lines with mutations in cilia-related genes. We obtained early postnatal or late embryonic cochlear tissue from Ift20, Ift25 (Hspb11 – Mouse Genome Informatics), Ift27, Gmap210 (Trip11^{Gt(AJ0490)Wtsi} – Mouse Genome Informatics) and Bbs8 (Ttc8 – Mouse Genome Informatics) mutant mice. Ift20, Ift25 and Ift27 are IFT complex B proteins required for both anterograde and retrograde IFT (Fig. 1C) (Follit et al., 2009; Lucker et al., 2005). Ift20 has additional roles related to Golgi-based sorting and vesicle trafficking of ciliary cargo (Follit et al., 2006), whereas Gmap210 anchors Ift20 to the Golgi complex (Follit et al., 2008). Bbs8 is thought to function as an adaptor protein for cargo undergoing IFT (Blacque et al., 2004; Tadenev et al., 2011). Despite a high degree of functional conservation between these molecules in other contexts, phenotypic variation in cochlear extension and bundle morphology was observed (Table 1). Cochleae from Ift27^−/− mutants were significantly shorter than in controls but displayed only mild bundle disruption. By contrast, Ift25^−/− and GMAP210^−/− cochleae were comparable to control littermates (supplementary material Fig. S1). Bbs8^−/− and Ift20^cko/cko mutants displayed more extreme PCP phenotypes and are described below.

Analysis of cochleae from P0 Bbs8^−/− mice revealed stereociliary bundle-orientation defects and flattened or misshapen bundles (Fig. 2A,B), but cochlear length was unchanged (supplementary material Fig. S2A). Consistent with other PCP mutants, stereociliary bundles were rotated and kinocilia were misplaced or occasionally absent. Kinocilia were often separated from stereociliary bundles, suggesting a loss of coupling between the structures. To confirm these changes, samples were examined by scanning electron microscopy (SEM) (Fig. 2C-I). At higher magnification, detached kinocilia and flattened bundle morphologies were visible (compare Fig. 2E with Fig. 2F,G). To quantify overall changes in kinocilia position and bundle orientation, both features were charted in wild-type (WT) and Bbs8^−/− cochleae (Fig. 2J,K). Both were mildly disrupted in Bbs8^−/− inner hair cells (IHCs), with most kinocilia and bundles still restricted to the lateral quadrant of the lumenal surface of hair cells. A more severe disruption was seen in outer hair cells (OHCs), where kinocilia and bundles were observed throughout the lumenal surface (Fig. 2J,K). Previous analyses of cochlear phenotypes in PCP mutants demonstrated variations in severity of bundle defects between each of the three rows of OHCs (Montcouquiol et al., 2003). However, a similar analysis in Bbs8^−/− cochleae indicated similar levels of defects in each row of OHCs. The flattened bundle morphology was further characterized by measuring the area between the vertex and ends of the two arms of each bundle, and the extent of bundle convexity (supplementary material Fig. S2B,C). Although the mean values for these metrics were unchanged, significantly greater variation in bundle convexity was observed in the absence of Bbs8. This is consistent with a role for Bbs8 in the specification of the shape, but not the overall size, of the stereociliary bundle.

As uniform orientation of stereociliary bundles is thought to be required for normal hearing, we sought to determine whether the bundle and kinociliary defects observed in Bbs8^−/− mice lead to deficits in auditory function. Hearing was assessed by measuring auditory brainstem response (ABR) thresholds between 4 and 24 kHz in 2- to 3-month-old mice. Surprisingly, no significant threshold elevations were observed (supplementary material Fig. S2D). As high frequency hearing often shows a greater susceptibility to systemic perturbations, we also examined hearing thresholds up to 45 kHz on a subset of the Bbs8^−/− mutants. Even at these higher frequencies, Bbs8^−/− mice did not have significantly elevated threshold shifts compared with controls (Fig. 3A). Measurable distortion-product otoacoustic emissions, a measurement of OHC function, also did not differ between Bbs8^−/− mutants and controls (Fig. 3B). The basis for the lack of an auditory phenotype is unclear. One possibility, corrective reorientation of bundles, as has been reported for Vangl2 CKOs mutants (Copley et al., 2013), was examined in 8-month-old mice. However, some bundle abnormalities were still present (data not shown).

These results are consistent with a role for Bbs8 in cochlear stereociliary bundle orientation, but not in the convergent extension movements that have been proposed to underlie cochlear extension. As bundle polarity and cochlear extension are often both disrupted in PCP mutants, we sought to determine whether Bbs8 similarly regulated PCP in two populations of polarized cells that do not undergo convergent extension: vestibular hair cells of the utricular maculae and ependymal cells lining the ventricular system of the brain.

The utricular sensory epithelium comprises a relatively uniform distribution of hair cells arranged in a fan-like shape with stereociliary bundles uniformly oriented along its radial axes (spines). In addition, a line of reversal is present near the mid-point along the medial-to-lateral axis such that bundles on either side of the line are oriented towards the center. At P0, the uniform orientation of bundles can be visualized based on labeling of stereocilia with phalloidin (Fig. 4A). By contrast, bundle orientation in Bbs8^−/− utricles appears irregular, resulting in whorls and disheveled-looking patches (Fig. 4B, arrows). Although these whorls are frequently seen in Bbs8 mutants and rarely in the controls, the orientation of individual hair bundles was examined to rule out possible disruptions in bundle structure during dissection or tissue preparation. To quantify this finding, the orientation of stereociliary bundles relative to neighboring hair cells was measured based upon β2-spectrin expression (Fig. 4C). β2-spectrin is a component of the cuticular plate that anchors the stereocilia and is absent from the fonticulus region at the base of the kinocilium. As a result, β2-spectrin reveals the location of the kinocilium on each cell. To determine whether absence of Bbs8 leads to changes in polarization, the average difference in the angle of polarization between one hair cell and its neighbors was determined for both control and Bbs8^−/− utricles (see Materials and Methods for details). Whereas the average angular difference was unchanged between mutants and controls, a statistically significant difference in standard deviation was measured between the two groups (Fig. 4D,E). The line of reversal was also more difficult to identify in Bbs8^−/− mutant utricles. These results are consistent with a disruption in bundle polarization.

PCP signaling has also been shown to play a role in the development of ependymal cilia, with defects leading to hydrocephalus (Tissir et al., 2010). Polarization of ependymal cells is tightly correlated with maturation and differentiation of multi-ciliated cellular morphologies. Consistent with this, Bbs8^−/− mice are born at Mendelian ratios yet are underrepresented at weaning and display hydrocephaly (data not shown). SEM of ependymal cells at P16 indicated severely stunted cilia in Bbs8^−/− ventricles (Fig. 4F,G). This phenotype was reminiscent of other PCP mutants in which ependymal cilia became basally embedded, instead of presenting normally on the apical surface (Tissir et al., 2010). Immunohistochemical analysis of the lumenal surface (length) at earlier time points (P9) suggests that ependymal cell maturation is compromised upon loss of Bbs8 (Fig. 4H,I, double arrow). There was also a wider variation in cell length in Bbs8^−/− tissue (Fig. 4J), which might have been caused by a failure of cells to polarize. Finally, undifferentiated cells with only a single spot of rootletin (suggesting a single primary cilium) were more commonly observed in mutant tissue (Fig. 4I, white arrow).

PCP phenotypes were also observed in cochleae from animals in which Ift20 was deleted from the inner ear by crossing Ift20^flox/flox mice with FoxG1^cre (referred to as Ift20^cko/cko; see Materials and Methods for details). Conditional mutants are not viable, and analyses were therefore performed at E18.5. Cochlear ducts from Ift20^cko/cko mice were significantly shorter than in littermate controls (Fig. 5A,B; supplementary material Fig. S3A) and showed a marked broadening of the sensory epithelia in the apex (supplementary material Fig. S3B,C). Ift20 is required for ciliogenesis, thus, as expected, kinocilia were absent in Ift20^cko/cko hair cells (Fig. 5C-G), which also served to confirm deletion of Ift20. Hair cells exhibited flattened and rotated bundles, similar to those observed in Bbs8^−/− mutant cochleae and other ciliary mutants that lack kinocilia (Fig. 5C-G) (Jones et al., 2008; Sipe and Lu, 2011). SEM confirmed these findings (Fig. 5E-G). As observed in other mutants lacking kinocilia, hair cells with circular stereociliary bundles could be identified, albeit infrequently (Fig. 5G, white arrow). As ciliary axonemes were missing in Ift20^cko/cko hair cells, changes in polarization were quantified based on the position of the basal body (Fig. 5H-J). In hair cells from control cochleae, the majority of basal bodies were located in the lateral quadrant (Fig. 5H, white arrow). By contrast, OHCs from Ift20^cko/cko cochleae often lacked basal bodies, and those that could be identified were more widely distributed across the lumenal surface (Fig. 5J). Analysis of basal body locations by row suggested that the first and third row OHCs show greater defects in the absence of Ift20 (Fig. 5J). Absent basal bodies might be a result of deeper positioning within the hair cells but could also reflect technical difficulties with the antibody reaction. By contrast, basal body locations on IHCs appeared comparable between control and Ift20^cko/cko (Fig. 5H-J).

Previous studies have suggested that Bbs and Ift proteins are not strictly associated with cilia and have shown localization to actin-rich cellular regions. Localization to actin-based structures might be significant in hair cells because the actin-rich stereocilia develop adjacent to the tubulin-based kinocilium, and molecules functioning in either structure could act to polarize the bundle. To determine where Bbs and Ift proteins localize within cochlear hair cells, immunogold labeling/TEM analysis was performed on cochleae at P0. Existing antibodies directed against Bbs8 proved unsuitable for TEM. However, as Bbs8 and Bbs2 are both components of the BBSome complex required for cilia biogenesis (Nachury et al., 2007), an antibody against Bbs2 was used as a proxy for Bbs8. Bbs2 localized to actin-rich structures in developing hair cells with a particular concentration along stereocilia and microvilli (Fig. 6A-C). Of particular interest was clustering of Bbs2 on vertical tracts of microfilaments contacting the apical surface (Fig. 6B). These microfilaments were similar in structure to filaments within microvilli, and are therefore presumably actin based. Similar actin-associated localization patterns for Bbs2 were seen in additional ciliated epithelia in the ventricular zone and choroid plexus (supplementary material Fig. S4). Ift20 labeling could be seen in the kinocilium (Fig. 6D, red arrows) and was also abundant in actin-rich microvilli (data not shown) and stereocilia of OHCs and IHCs (Fig. 6D-E′). Labeling was also observed near basal bodies (Fig. 6F) and associated centrioles (Fig. 6G).

These results demonstrate roles for Bbs8 and Ift20 in bundle polarization; however, the specific effects of these molecules are unclear. Previous studies have placed other cilia-related proteins, such as Ift88, downstream of the core PCP factors, by demonstrating that asymmetric localization of core PCP proteins occurs normally in these mutants. To determine whether a similar localization of core PCP proteins occurs in Bbs8 or Ift20 mutants, membrane localization of Vangl2 was determined by immunocytochemistry at P0 and E18.5, respectively (Fig. 7A-D). In control cochleae (Fig. 7A,C), Vangl2 was localized along both medial and lateral surfaces of pillar cells (Fig. 7A,C, white asterisk), and asymmetrically at junctions between medial hair cell and lateral support cell surfaces (Fig. 7A,C, white arrow). By contrast, in Bbs8 mutants, although Vangl2 was still detected along pillar cells, membrane accumulation at hair cell-support cell junctions appeared reduced (Fig. 7A,B). Similar changes were observed in Ift20^cko/cko cochleae, although overall membrane localization appeared somewhat more reduced (Fig. 7C,D). This result contrasts with localization in other cilia mutants, in which Vangl2 was shown to be unaltered, including Mkks, Ift27, Ift25, Gmap210 and Ift88 (supplementary material Fig. S5) (Jones et al., 2008). In order to determine whether the decreased labeling of Vangl2 in Bbs8 mutants was a result of decreased protein abundance or changes in membrane targeting, cochlear tissue was separated into cytoplasmic and membrane fractions and then probed for expression of Vangl2 by western blot. No significant difference between control and mutants was observed for either fraction (supplementary material Fig. S6), suggesting that Bbs8/Ift20 play a role in targeting Vangl2 to specific regions of the cell rather than in regulation or overall membrane localization.

Recent studies identified a G-protein-dependent signaling pathway acting in a cell-autonomous manner, independent of the core PCP proteins, that regulates the migration of the kinocilium and patterns the apical hair cell surface (Ezan et al., 2013). Moreover, localization of GTP-binding protein alpha-i subunit 3 (Gαi3), a key factor in this migration, is disrupted in Mkks (Bbs6) mutants. To determine whether Bbs8 plays a similar role, localization of Gαi3 was determined in Bbs8 mutants (Fig. 7E-G). Overall, there is an expansion of the Gαi3 domain in the absence of Bbs8, consistent with abnormal hair bundle morphology and kinocilia mislocalization. There also seems to be a larger variation in Gαi3 expansion over the surface of the hair cells, as opposed to a more uniform distribution in controls.

The results presented above are consistent with previous studies demonstrating an interaction between Bbs8 and Vangl2 in vitro (May-Simera et al., 2010). To determine whether a similar association exists in vivo, we performed co-immunoprecipitation (co-IP) pull-down assays from endogenous neuronal tissue using antibodies specific for Vangl2 or Bbs8. Neuronal tissue was used because of the greater relative abundance of both proteins. After IP, using an anti-Vangl2 antibody, Bbs8 was detected by western blot. This band was absent from control IPs (Fig. 7H, black arrow). Next, we examined possible associations of Ift20 with Vangl2 via co-IP, using tagged constructs transfected into HEK293 cells. We were able to identify Ift20-Flag after IP with Vangl2-GFP (Fig. 7I, black arrow). Ift88-GFP was used as a positive control, which also precipitated Ift20-Flag, whereas negative controls showed no bands. Together, these findings suggest that the PCP protein Vangl2 physically interacts with Bbs8 and Ift20, and that this interaction mediates the asymmetric accumulation of Vangl2 in polarized cells.

To identify pathways that might be altered as a result of the deletion of Bbs8, we performed a microarray analysis using cochlear tissue harvested from P0 Bbs8^+/+ and Bbs8^−/− mice. After normalization using the RMA algorithm, significantly differentially expressed genes with a twofold change or more (P<0.05) were selected, based on an ANOVA using the Partek Genomics Suite software. The resulting genes were analyzed using the Partek Pathway software, which identifies KEGG Pathways enriched with these genes. Three of the top four pathways with enrichment scores above 9.0 and highly significant P-values (P-values <8.6E-05) were related to cell adhesion, cell migration and cytoskeletal rearrangements (Table 2; supplementary material Table S1). Changes in all of these pathways can be a consequence of disruption in PCP signaling (Carreira-Barbosa et al., 2009; Cui et al., 2013; Ezan and Montcouquiol, 2013; Narimatsu et al., 2009; Saburi and McNeill, 2005), suggesting that the observed phenotypic defects in the Bbs8^−/− mice are attributable to its dysregulation. In addition, as previous work has suggested that multiple Wnt pathways can interact in a mutually antagonist fashion, we examined changes in genes directly involved in Wnt signaling. Here, we found downregulation of transcripts encoding many Wnt ligand and Wnt receptor molecules (Table 2).

There is growing evidence supporting an involvement of cilia-related proteins in regulation of PCP. However, understanding the potential role of cilia in this pathway is complicated due to the redundancy of ciliary and PCP genes. In addition, cell polarity is sensitive to generalized cellular abnormalities, making it difficult in some circumstances to directly link a mutation to PCP-specific deficits. To better understand this, we analyzed polarity defects within the cochleae of mice with targeted mutations in several cilia-related genes. It is remarkable that, although these proteins have similar functions in other systems, we observed distinct cochlear phenotypes for the different mutants. For example, deletion of Gmap210, Ift27 or Ift25 showed only mild cochlear disruptions. By contrast, more severe PCP defects were observed when Bbs8 or Ift20 were deleted, including changes in stereociliary bundle orientation and morphology, and mislocalization of the core PCP transmembrane protein Vangl2. The collective phenotype was greatest in Ift20 mutants that were missing ciliary axonemes altogether and also showed changes in cochlear duct extension. This phenotype is consistent with other ciliary mutants, such as Ift88 and Kif3a, in which complete loss of the ciliary axoneme results in shortened cochlear ducts and mis-positioned stereociliary bundles. However, in Ift88 and Kif3a mutants, the membrane localization of Vangl2 remains intact (Jones et al., 2008; Sipe and Lu, 2011). This is in contrast with the results presented here for Bbs8 and Ift20 mutants, which suggest that a subset of cilia proteins does affect PCP signaling at the level of PCP protein localization. These results also indicate that the PCP phenotype observed in a subset of ciliary mutants is related to the disruption of PCP signaling. We propose that those ciliary proteins contribute to protein trafficking, thereby regulating the asymmetric localization of PCP molecules.

Many proteins associated with cilia function occur in complexes localized to cilia-related subdomains, such as the basal body, transition zone or ciliary axoneme (van Dam et al., 2013). More recently, some of these proteins have also been observed in other cellular regions, such as dendrites of retinal neurons and epithelial focal adhesions (Finetti et al., 2009; Sedmak and Wolfrum, 2010; Yuan and Sun, 2013). These observations suggest that cilia-related proteins have additional cellular functions away from the cilium, although these functions could still be orchestrated via the basal body in its capacity as a microtubule-organizing center. Given that a primary role for ciliary proteins is the movement of cargo along microtubules, it is highly likely that these proteins might also regulate aspects of intracellular trafficking along the cytoskeleton, such as vesicular transport (Delaval et al., 2011; Kim and Tsiokas, 2011; Robert et al., 2007; Follit et al., 2008,, 2006; Pedersen et al., 2008; Sedmak and Wolfrum, 2010). For instance, Ift20 has been shown to contribute to vesicular trafficking from the Golgi to the base of the cilium (Follit et al., 2006). These observations, together with the disrupted membrane accumulation of Vangl2 in Ift20 mutants, indicate that Ift20 is required for the delivery of Vangl2-containing vesicles to the cell surface. These observations also suggest that other members of the IftB complex contribute to Vangl2 targeting. However, at least two members of that complex, Ift25 and Ift27, do not appear to play crucial roles, as deletion of either gene leads to only mild cochlear PCP defects. This suggests a unique requirement for Ift20 compared with other Ifts, as well as potential novel roles for this protein that might be independent of its association with Ift complexes. A similar result was observed in Bbs8 mutants, a finding that is consistent with the direct physical interactions between Bbs8 and Vangl2 [May-Simera et al. (2010) and this study]. Like Ift20, protein-localization studies have expanded the potential functions for Bbs proteins beyond the cilia. In the cochlea, Bbs2, 4 and 6 display a range of non-cilia-related cellular distributions, including in the vicinity of cellular membranes and actin-rich regions (May-Simera et al., 2009). In cultured cells, Bbs proteins are localized to actin-rich focal adhesions, where they negatively modulate the actin cytoskeleton (Hernandez-Hernandez et al., 2013). Similarly, we found Bbs2 enrichment in the actin-based stereocilia of hair cells. Together, these observations suggest that Ift20 and Bbs8 can function independently of the ciliary axoneme, and might traffic proteins along microtubules or filamentous actin to cellular locations other than cilia.

The basis for the lack of an auditory phenotype in Bbs8 mutants is unclear. One possibility is that there is a corrective reorientation of bundles, as has been reported for Vangl2 CKO mutants (Copley et al., 2013); however, our preliminary evidence suggests that this is not the case. Other cilia mutants, such as Mkks (Bbs6) and Alms1, have milder bundle disruptions than Bbs8 mutants, yet display more severe auditory phenotypes (Jagger et al., 2011; Ross et al., 2005). This suggests that auditory dysfunction in cilia mutants is not necessarily directly linked to alterations in stereociliary bundle morphology or mechanotransduction, further arguing that these genes encode additional intracellular functions independent of cilia. Based on these results, we propose that a subset of proteins that had originally been identified by their association with cilia might in fact function in broader roles related to the intracellular trafficking of membrane-bound proteins throughout the cell, although these functions might still be organized via the basal body. Bbs8 and Ift20 appear to be members of this group in that they act upstream of cilia localization by targeting Vangl2 (and other) PCP proteins to the membrane. Although this does not rule out a role for Bbs8 and Ift20 in ciliary migration, it seems clear that other cilia proteins, such as Ift88, Mkks and Kif3a, have more restricted functions and only contribute to ciliary migration.

One important function attributed to the primary cilium is regulating a transition between the canonical (β-catenin dependent) and non-canonical (PCP) Wnt signaling pathways (Simons et al., 2005). As canonical Wnt signaling regulates transcription (Dickinson and McMahon, 1992), we tested whether loss of Bbs8 might also affect gene expression. Indeed, microarray data from Bbs8^−/− cochleae show changes in gene expression. Many of these genes encode proteins involved in Wnt signaling and downstream effectors, suggesting additional consequences of Bbs8 deletion besides PCP signaling. As a result, loss of ciliary function might upset the delicate balance between canonical and non-canonical Wnt signaling. By contrast, other studies have suggested no connection between cilia and Wnt signaling. For example, Ift88 zebrafish mutants lack all cilia, but have normal canonical and non-canonical Wnt signaling (Borovina and Ciruna, 2013; Huang and Schier, 2009), as do Ift88, Ift172 and Kif3a mutant mouse embryos (Ocbina et al., 2009). These differences might reflect species-, tissue- or time-dependent differences in ciliary contributions towards Wnt signaling. Moreover, normal Wnt responsiveness might be retained if the basal bodies, and therefore trafficking to the basal body, remain intact in these mutants.

Our results, in combination with previous reports on the effects of deletion of Bbs8 (Tadenev et al., 2011), suggest a greater role for Bbs8 in PCP by comparison to other Bbs proteins (Ross et al., 2005). If this is the case, then patients with mutations in BBS8 should display an increased prevalence of ‘PCP-like’ phenotypes. Indeed, BBS8 was first identified in a family with Bardet–Biedl syndrome, in which a homozygous null mutation results in randomization of left-right body axis symmetry, another phenotype possibly related to PCP defects (Ansley et al., 2003; Aw and Levin, 2009). Several BBS patients with deleterious mutations in BBS8 have been identified, who in addition to the typical disease phenotypes (obesity, retinopathy, polydactyly) harbor additional clinical manifestations including renal cystic disease, shortened limbs, hearing impairment, dilated ventricles and situs inverses (P. Beales, personal communication), features that could be suggestive of perturbed Wnt signaling culminating in aberrant PCP. Altogether, our results and these observations are consistent with the role of ciliary proteins extending beyond the cilia and basal body. Clearly, a more complete understanding of the role of ‘ciliary’ proteins in cellular signaling pathways and other biological phenomena is crucial for our understanding of cellular and developmental biology, as well as for the development of targeted treatment strategies.

Animal care and use was in accordance with NIH guidelines and conformed to institutional ACUC regulations. Generation and genotyping of Bbs8^−/− and Ift20^flox/flox mutants have been described (Jonassen et al., 2008; Tadenev et al., 2011). Foxg1^Cre (Hebert and McConnell, 2000) animals were obtained from Jackson Laboratories and were crossed with Ift20^flox/flox mice to inactivate Ift20 in the developing inner ear. Gmap210 (Follit et al., 2008), Ift25 (Keady et al., 2012), Ift27 (Eguether et al., 2014) and Mkks (a gift from Philip Beales, UCL, London, UK) tissue was harvested at embryonic day 18.5 (E18.5) or postnatal day 0 (P0). The morning after mating was considered E0.5 and up to 24 h after birth was considered P0.

Temporal bones were isolated from embryonic and postnatal mice and fixed in 2-4% paraformaldehyde at 4°C for 2 h. Following microdissection of the cochlea and utricle, immunohistochemistry was performed as described (May-Simera and Kelley, 2012a). Images were obtained using a Zeiss 510 laser scanning confocal microscope. All images were captured from the basal turn of the cochlea unless otherwise stated. Methods used for quantification and additional antibody information are available in the supplementary material methods.

For SEM, temporal bones were dissected and fixed in EM fixative [2.5% glutaraldehyde (Sigma-Aldrich), 4% paraformaldehyde (Electron Microscopy Sciences) and 10 mM CaCl₂ in HEPES buffer] for 2 h at room temperature. Organs of Corti were micro-dissected and prepared for SEM using an S-4800 Hitachi electron microscope as described (May-Simera and Kelley, 2012a). For post-embedding immunogold, cochleae from P0 mice were prepared as described for light microscopy and SEM, and tissue was fixed in 4% paraformaldehyde and 0.5% glutaraldehyde in 0.1 M phosphate buffer, and further processed as described (Petralia et al., 2010; Petralia and Wenthold, 1999). Additional details are available in the supplementary material methods.

Endogenous co-IP of Bbs8 and Ift20 protein was performed using P0 brain lysate harvested in RIPA buffer (Tris-HCl, 50 mM; NaCl, 150 mM; 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, pH 7.6) containing protease and phosphatase inhibitors (Roche). Anti-Bbs8 and anti-Vangl2 antibodies were used for pull-down or western blot, respectively. For in vitro co-IPs, HEK293 cells were transiently co-transfected with Vangl2 GFP-pCLIg and IFT20-Flag plasmids using lipofectamine (Invitrogen). Cell lysates were harvested 48 h post transfection. Rabbit polyclonal anti-EGFP (Clontech) was used for pull-down, and mouse monoclonal anti-flag was used for western blot. Protein G dynabeads (Invitrogen) were used for co-IP, following the manufacturer's instructions. Post precipitation, protein was harvested in SDS-containing sample loading buffer and detected by western blot. Proteins were run on 4-12% SDS-PAGE (Invitrogen) using conventional protocols and Pico substrate (Cell Signaling) for chemiluminescent detection.

RNA was extracted from two Bbs8^−/− and two littermate control cochleae at P0 using the RNAqueous-Microkit (Ambion). Two collections were performed on separate days to produce two biological replicates. Total RNA was further purified on an RNAeasy column (Qiagen) and the RNA quality was assessed by an Agilent Bioanalyzer (Agilent Technologies). Target labeling and hybridization to GeneChips were carried out in the NIDDK Microarray Core facility using the GeneChip Mouse 430_2 Array purchased from Affymetrix. The analysis is described in detail in the supplementary materials and methods. The Affymetrix array data have been deposited at ArrayExpress under the accession number E-MTAB-3165.

We would like to acknowledge the editorial assistance of the NIH Fellows Editorial Board for carefully editing the manuscript. We are grateful to Tracy Fitzgerald, Tiansen Li, Chandrakala Puligilla, Kapil Bharti, Tiziana Cogliati, Julien Debbache and Christopher Brinson for technical assistance, scientific reading and commenting on the manuscript.