Epb41l5 interacts with Iqcb1 and regulates ciliary function in zebrafish embryos

Cilia are antenna-like structures that extend from the surface of the apical membrane of cells and play important roles in sensing a variety of extracellular signals (Carvalho-Santos et al., 2011; Ishikawa, 2017; Loreng and Smith, 2017; Pazour and Witman, 2003). Cilia-mediated signaling regulates cell proliferation, differentiation and function, and plays essential roles in embryonic development and tissue homeostasis in adults (D'Angelo and Franco, 2009; Goetz and Anderson, 2010; Tasouri and Tucker, 2011). Defects in ciliary structure and function lead to a number of human diseases called ciliopathies (Avasthi et al., 2017; Cao et al., 2010; Dell, 2015; Estrada-Cuzcano et al., 2012; Hildebrandt et al., 2009; Kagan et al., 2017; Klena et al., 2017; Oud et al., 2017).

Ciliogenesis is initiated by the formation of membrane vesicles at the distal end of the mother centriole (Mirvis et al., 2018; Nigg and Raff, 2009). The centrosome migrates to the apical cell cortex where the mother centriole transforms into the basal body and initiates assembly of the axoneme (Avidor-Reiss et al., 2017; Bernabé-Rubio and Alonso, 2017; Linck et al., 2016). Intraflagellar transport (IFT) mediates bi-directional transport of cargo proteins in and out of the cilia (Ishikawa and Marshall, 2017; Kim et al., 2010; Li and Hu, 2011; Pedersen and Rosenbaum, 2008). This cargo includes various structural components of the cilia and proteins crucial for signal transduction (Corbit et al., 2005, 2008; Madhivanan and Aguilar, 2014; Nachury, 2014; Wheway et al., 2018).

Cilia have a specialized domain at the base of the cilium, named the ciliary transition zone (TZ). The TZ is an ultrastructurally defined complex barrier that acts as a gate to control protein entry and exit from the cilia. The ciliary transport is highly selective, which enables the establishment of a unique composition of soluble and membrane proteins in the cilia. Previous studies have identified multiple TZ components, many of which are encoded by the genes responsible for Meckel–Gruber syndrome (MKS) (Barker et al., 2014; Dawe et al., 2007; Kagan et al., 2017; Williams et al., 2011; Zhao and Malicki, 2011), Joubert syndrome (Garcia-Gonzalo et al., 2011; Kagan et al., 2017; Lee et al., 2012; Mitchison and Valente, 2017; Romani et al., 2013) and nephronophthisis (NPHP) (Chih et al., 2012; Craige et al., 2010; Garcia-Gonzalo et al., 2011; Kee et al., 2012; Sang et al., 2011; Williams et al., 2011). NPHP and MKS proteins form modules and cooperate to establish the TZ and compartmentalize the ciliary domain (Avidor-Reiss et al., 2017; Chih et al., 2012; Garcia-Gonzalo et al., 2011; Gonçalves and Pelletier, 2017; Takao et al., 2017; Williams et al., 2011). Although the importance of the TZ for cilia has been established, it is largely unknown how the TZ is specified, organized and maintained.

Erythrocyte protein band 4.1 like 5 (EPB41L5) is an adaptor protein that contains the FERM (band 4.1, ezrin, radixin and moesin) domain (Baines, 2006; Moleirinho et al., 2013; Tepass, 2009) and regulates morphogenesis of epithelial and neuroepithelial cells. First, EPB41L5 functions in apico-basal polarity by modulating the distribution of Crumbs at the apical membrane (Christensen and Jensen, 2008; Gamblin et al., 2018; Gosens et al., 2007; Hoover and Bryant, 2002; Hsu et al., 2006; Jensen et al., 2001; Jensen and Westerfield, 2004; Laprise et al., 2006; Perez-Vale and Peifer, 2018). Second, EPB41L5 is required for disassembly of cadherin-based adherens junctions to promote epithelial-to-mesenchymal transition (Hirano et al., 2008; Lee et al., 2007) and apical detachment of differentiating neurons (Matsuda et al., 2016). Third, EPB41L5 is involved in the formation and maintenance of focal adhesion in podocytes in adult mice (Hirano et al., 2008; Schell et al., 2017). Furthermore, EPB41L5 has also been shown to induce apical constriction in cultured epithelial cells (Nakajima and Tanoue, 2010, 2011) and Xenopus ectoderm (Chu et al., 2013). In this manner, EPB41L5 plays diverse functions in epithelial morphogenesis and its functional diversity might be determined by its partner proteins and/or cell types.

Although EPB41L5 does not interact with actin directly, EPB41L5 appears to play roles in modulating organization of the actin cytoskeleton, in particular actin at the apical or sub-apical cortex. Indeed, EPB41L5 interacts with actin modifiers, including p114RhoGEF (Nakajima and Tanoue, 2010, 2011; Schell et al., 2017) and vimentin (Hirano et al., 2008). However, it remains incompletely understood how EPB41L5 modulates actin networks for epithelial morphogenesis.

In this study, we report that EPB41L5 interacts with IQ calmodulin-binding motif-containing 1 [IQCB1, also called nephrocystin 5 (NPHP5)]. IQCB1 has been implicated in nephronophthisis, an autosomal recessive cystic kidney disease (Barbelanne et al., 2013; Downs et al., 2016; Hildebrandt et al., 2009; Otto et al., 2005; Schäfer et al., 2008; Stone et al., 2011). We show that EPB41L5 suppresses IQCB1 accumulation at the ciliary base in ciliated hTERT-RPE1 cells and at the centrosome in non-ciliated cells. We also show that zebrafish embryos deficient in epb41l5 and embryos expressing the C-terminally modified Epb41l5 have ciliary dysfunction. Furthermore, we demonstrate that EPB41L5 reduces IQCB1 interaction with centrosomal protein 290 (CEP290, previously known as NPHP6) in HEK293 cells. Taken together, we propose that EPB41L5 controls the integrity of the ciliary base and centrosome through IQCB1 and CEP290.

To identify EPB41L5-interacting proteins, we conducted mass spectrometry analysis of proteins physically interacting with EPB41L5. IQCB1 was identified by immunoprecipitation of mouse EPB41L5 expressed in HEK293 cells. IQCB1 contains three IQ calmodulin-binding motifs, one coiled-coil domain and a CEP290-interacting domain at the C terminus (Fig. 1B) (Barbelanne et al., 2013). We confirmed interaction between EPB41L5 and IQCB1 by co-immunoprecipitation of myc-tagged zebrafish or mouse EPB41L5 and FLAG- or HA-tagged human IQCB1 (Fig. 1C,D).

We next identified protein domains involved in the interaction between EPB41L5 and IQCB1 (Fig. 1A,B). We found that the N-terminal FERM domain of zebrafish Epb41l5 (amino acids 1–239) was required and sufficient for Epb41l5 binding to IQCB1 (Fig. 1C). Removing the C-terminal conserved domain of Epb41l5 (Epb41l5ΔCTD) increased Epb41l5 binding to IQCB1 (Fig. 1C), suggesting that the CTD inhibits Epb41l5–IQCB1 interaction. Lack of a strong signal in co-immunoprecipitation of the full-length Epb41l5 protein might also be a result of its poor stability (Matsuda et al., 2016). On the other hand, the internal domain of IQCB1 containing three IQ motifs and a coiled-coil domain (amino acids 287–443) was required and sufficient for the Epb41l5–IQCB1 interaction (Fig. 1D). Taken together, these results suggest that EPB41L5 interacts with IQCB1 via the FERM domain of Epb41l5 and the IQ-coiled coil domain of IQCB1.

We next asked whether EPB41L5 and IQCB1modulate each other's subcellular localization. As previously reported (Hirano et al., 2008; Matsuda et al., 2016; Nakajima and Tanoue, 2010, 2011), FLAG-tagged Epb41l5 is associated with the plasma membrane in hTERT-RPE1 cells (Fig. 2A) and with the basolateral membrane of polarized Madin–Darby canine kidney (MDCK) epithelial cells (Fig. S1A,A″). When IQCB1 and Epb41l5 were coexpressed, IQCB1 did not alter Epb41l5 localization in either hTERT-RPE1 cells (Fig. 2C′) or MDCK cells (Fig. S1C,C″).

By contrast, Epb41l5 modified IQCB1 localization. We observed that HA-tagged IQCB1 accumulated at the base of cilia marked by the ciliary protein ARL13B and in cytoplasmic puncta (Fig. 2B; Fig. S1B–B″), which is consistent with previous observations (Barbelanne et al., 2015, 2013; Das et al., 2017). We note that the immunostaining of endogenous IQCB1 using a commercial anti-IQCB1 antibody was below the detection limit (data not shown). Coexpression of Epb41l5 reduced IQCB1 accumulation at the ciliary base (Fig. 2C,F). Conversely, epb41l5 knockdown by short hairpin RNA (shRNA; Fig. 2E) further enriched cilium-associated IQCB1 immunostaining (Fig. 2D,G). We confirmed that coexpression of exogenous Epb41l5 rescued the accumulation of IQCB1 at the ciliary base in epb41l5 knocked-down cells (Fig. 2G). Taken together, these results suggest that EPB41L5 suppresses IQCB1 localization at the ciliary base.

We sought to understand how EPB41L5 suppresses IQCB1 localization at the ciliary base. A previous study showed that the coiled-coil domain of IQCB1 was involved in IQCB1 localization at the centrosome (Barbelanne et al., 2013), which develops into the ciliary base. We showed that EPB41L5 binds to IQCB1 through the region containing the coiled-coil domain (Fig. 1D). This raised the possibility that EPB41L5 competitively inhibits IQCB1 association with the ciliary base. Consistent with this hypothesis, coexpression of the N-terminal FERM-FA domain of Epb41l5 [Epb41l5(FERM-FA)] was sufficient to reduce IQCB1 accumulation at the ciliary base (Fig. 3A). The efficacy of the suppression might be stronger than full-length Epb41l5 (Fig. 2F). Coexpression of the C-terminal fragment of Epb41l5 [Epb41l5(Cfrag)] did not change IQCB1 accumulation at the ciliary base (Fig. 3B). These results suggest that EPB41L5 binding to IQCB1 suppresses IQCB1 localization at the ciliary base.

Next, we investigated whether the domain between amino acids 287 and 443 was required for IQCB1 displacement from the ciliary base. As described in a previous study (Barbelanne et al., 2013), ciliogenesis was impaired in cells expressing IQCB1Δ(287–443) (Fig. 3C′). Therefore, we examined IQCB1 association with the centrosome. First, we confirmed that IQCB1Δ(287–443) showed colocalization with centrosomal proteins γ-tubulin (Fig. S2A) and NEDD1 (Fig. S2B), as well as ARL13B (Fig. 3C). For this analysis, we used ARL13B as a centrosomal marker to make this more comparable with analyses of other IQCB1 deletion mutant proteins. It should be noted that IQCB1Δ(287–443) association with the centrosome was weaker than that of full-length IQCB1 (Fig. 3C,C″). IQCB1(287–443) did not show accumulation at the centrosome (Fig. 3E,E″). These observations confirm the previous observation that the CEP290-binding domain at the C terminus is important for IQCB1 localization to the centrosome (Barbelanne et al., 2013; Stone et al., 2011). Nevertheless, IQCB1Δ(287–443) accumulation at the centrosome was not altered by EPB41L5 overexpression (Fig. 3D,F). Taken together, these results suggest that EPB41L5 binding to the IQ-coiled-coil domain suppresses centrosomal association of IQCB1.

We next looked at whether Epb41l5 has a regulatory role in cilia formation or function in vivo using zebrafish embryos. First, we analyzed cilia formation in epb41l5-deficient embryos. We used an epb41l5 translation blocking morpholino (epb41l5-MO^ATG) (Hsu et al., 2006; Jensen and Westerfield, 2004; Matsuda et al., 2016) for embryos younger than 18 h post fertilization (hpf), because homozygous moe^b476 mutants are not identifiable either genetically or morphologically at 18 hpf or earlier. Also, translation blocking morpholinos are effective at minimizing the effects of maternally loaded epb41l5 transcripts in embryos. For embryos older than 18 hpf, we used epb41l5 null mutants mosaic eyes (moe) (Hsu et al., 2006; Jensen et al., 2001; Jensen and Westerfield, 2004; Kramer-Zucker et al., 2005b). Cilia formation was assessed by immunostaining.

We found that cilia were formed in the pronephric duct in moe^b476 mutants (Fig. 4A–B′) and Kupffer's vesicle (KV) in epb41l5-MO^ATG morphants (Fig. 4C–F). The presence of cilia in epb41l5 deficient embryos was expected because epb41l5 overexpression or knockdown did not alter cilia formation in hTERT-RPE1 cells (Fig. 2A,D) or in MDCK cells (data not shown). This is also consistent with the presence of nodal cilia in mouse epb41l5 null mutants (Lee et al., 2007).

However, pronephric cilia in moe^b476 zebrafish mutants differed from those in wild-type siblings. In wild-type embryos, pronephric cilia form bundles (Liu et al., 2007), so that individual cilia are not visibly distinguishable (Fig. 4A,A′). In contrast, individual cilia were easier to distinguish in moe^b476 mutants (Fig. 4B,B′). This suggests that the cilia in epb41l5-deficient embryos have defects in their function. Consistent with this, the expression of spaw, the first gene asymmetrically expressed in the lateral plate mesoderm (LPM) (Long et al., 2003), was randomized in epb41l5-MO^ATG morphants (Fig. 4G,H). Cilia are primarily responsible for directional fluid flow generation in KV. Therefore, these results suggest that KV cilia have functional defects, leading to randomized left–right (LR) patterning in epb41l5-MO^ATG morphants. We also observed randomized cardiac jogging in epb41l5-MO^ATG morphants, which was rescued by co-injection of epb41l5 mRNA (Fig. S3). This confirms that LR patterning defects are specific effects of the epb41l5-MO^ATG morpholino.

Previous studies showed that epb41l5 deficiency led to severe loss of epithelial integrity in developing zebrafish and mouse embryos (Hsu et al., 2006; Jensen et al., 2001; Jensen and Westerfield, 2004; Lee et al., 2010, 2007; Matsuda et al., 2016). Because cilia form at the apical membrane in epithelial cells, loss of epithelial integrity in epb41l5-deficient embryos could alter the spatial distribution of cilia, resulting in cilia dysfunction. Alternatively, EPB41L5 may have a more direct role in cilia, independent of its function on epithelial morphogenesis. In that case, EPB41L5 interaction with IQCB1 might mediate this process.

To explore a more direct role of Epb41l5 in cilia in vivo, we analyzed cilia in a novel allele of zebrafish epb41l5 mutants. The epb41l5Δctd mutants were generated using the CRISPR/Cas9 genome engineering system (Chang et al., 2013; Hwang et al., 2013; Li et al., 2016). Three guide RNAs (gRNAs) were designed to target the splicing donor site of exon 25 (Fig. 5A). Indels were confirmed near the gRNA target sites in the genome (Fig. 5B). Exon 25 was spliced out in epb41l5Δctd transcripts (Fig. 5C), resulting in a frame-shift and a premature stop codon (Fig. 5D).

Although no epb41l5 transcripts were detectable in epb41l5 null mutants moe^b476, the epb41l5Δctd transcripts were present in epb41l5Δctd mutants (Fig. 5C), which are expected to produce Epb41l5ΔCTD protein (Fig. 5D). We confirmed that three independent alleles of epb41l5Δctd mutants (Fig. 5B) produced the same epb41l5Δctd transcripts (data not shown). Importantly, all epb41l5Δctd mutants (epb41l5Δctd1, epb41l5Δctd2 and epb41l5Δctd3) showed largely normal eye pigmentation, body curvature and brain ventricle inflation (Fig. 5F,I,L), compared with wild-type embryos (Fig. 5E,H,K) and moe^b476 null mutants (Fig. 5G,J,M). These results suggest that epithelial integrity was largely maintained in epb41l5Δctd mutants. Unexpectedly, heterozygous epb41l5Δctd mutants showed severe male infertility (data not shown). This made it challenging to obtain epb41l5Δctd homozygous mutant embryos for further analyses.

To overcome the limited availability of epb41l5Δctd mutants, we took two alternative approaches. First, we designed a morpholino antisense oligonucleotide (epb41l5-MO^SpD; Fig. 5A), which targeted the splicing donor site of exon 25. We validated that epb41l5-MO^SpD injection resulted in the same transcripts as in epb41l5Δctd mutants (Fig. 5C). As expected, epb41l5-MO^SpD morphants did not have significant defects in epithelial morphogenesis, with a largely normal hindbrain ventricle and apical localization of ZO1 at the ventricular surface (Fig. 5O). Second, we overexpressed exogenous Epb41l5 lacking the C-terminal 60 amino acids by mRNA injection (Epb41l5Δ60 in Fig. 5D). The combination of epb41l5Δctd mutants, epb41l5-MO^SpD morphants and embryos expressing exogenous Epb41l5Δ60 helped assess the direct role of Epb41l5 on cilia and reduced the concern for any ‘off-target’ effect in morphants (Gerety and Wilkinson, 2011; Joris et al., 2017; Kok et al., 2015; Law and Sargent, 2014; Robu et al., 2007; Schulte-Merker and Stainier, 2014; Stainier et al., 2017).

We found that both KV cilia and pronephric cilia formed in epb41l5-MO^SpD morphants (Fig. 6B,E,E′) and embryos expressing Epb41l5Δ60 (Fig. 6C,F,F′), as was observed in wild-type embryos (Fig. 6A,D,D′). This further confirmed that Epb41l5 is not required for cilia formation. However, cilia appeared to be abnormal in these embryos. In wild-type embryos, individual pronephric cilia were distinguishable because they formed bundles (Fig. 6D,D′). On the other hand, individual pronephric cilia were easier to distinguish in both epb41l5-MO^SpD morphants (Fig. 6E,E′) and embryos expressing Epb41l5Δ60 (Fig. 6F,F′). This suggests failure of cilia bundle formation in these embryos.

These embryos also showed randomized LR patterning. The direction of cardiac jogging is regulated by LR patterning signals from KV (Amack et al., 2007; Amack and Yost, 2004; Chen et al., 1997; Essner et al., 2005). We found that cardiac jogging was randomized in epb41l5-MO^SpD morphants (Fig. 6K), epb41l5Δctd mutants (Fig. 6L) and embryos expressing Epb41l5Δ60 (Fig. 6M). Randomized cardiac jogging in epb41l5-MO^SpD morphants was rescued by epb41l5 mRNA co-injection (Fig. S3), confirming the specificity. Expression of spaw was also randomized in the LPM in epb41l5-MO^SpD morphants (Fig. 4H). charon is considered the first asymmetric flow target gene in medaka, zebrafish, frog and mouse (coco in frogs and Cerl2 in mice) (Hashimoto et al., 2004; Hojo et al., 2007; Lopes et al., 2010; Nakamura et al., 2012; Sampaio et al., 2014; Schweickert et al., 2010). We found that charon was more symmetrically expressed in epb41l5-MO^SpD morphants (Fig. 6N,O). Because cilia are primarily responsible for directional fluid flow generation in KV, these results suggest that cilia have motility defects in embryos expressing Epb41l5ΔCTD.

To assess a direct effect of Epb41l5ΔCTD expression on cilia motility, we did live imaging of pronephric cilia in epb41l5-MO^SpD and epb41l5-MO^ATG morphants. For live imaging of cilia, tg[actb:arl13b-gfp] embryos (Borovina et al., 2010) were imaged on a confocal microscope at 28–30 hpf. Pronephric cilia have been shown to beat at a frequency of 20.0±3.2 Hz in wild-type embryos (Kramer-Zucker et al., 2005a), which is faster than the scanning speed of a conventional confocal microscope. Indeed, individual pronephric cilia were difficult to image in control wild-type embryos (arrowheads in Fig. 6G; Movie 1) as compared to primary cilia in muscle progenitor cells (arrows in Fig. 6G). Notably, pronephric cilia were more clearly captured in epb41l5-MO^ATG morphants (arrowheads in Fig. 6H; Movie 2) or epb41l5-MO^SpD morphants (arrowheads in Fig. 6I; Movie 3) than in wild-type embryos (Fig. 6G; Movie 1). This suggests that cilia had reduced motility in epb41l5-MO^ATG or epb41l5-MO^SpD morphants. Taken together, these results suggest that Epb41l5 is required for cilia motility. Without normal Epb41l5 function, the ciliary base may lose its integrity, resulting in abnormal cilia motility.

Next, we asked whether Epb41l5 regulates the integrity of the ciliary base together with Iqcb1. To test genetic synergy, we co-injected a small amount of epb41l5-MO^ATG and iqcb1 morpholino (Schäfer et al., 2008). First, we confirmed that single partial knockdown of either epb41l5 or iqcb1 had minimal impacts on embryogenesis in general (Fig. 6Q–S). With partial knockdowns of both epb41l5 and iqcb1, embryos showed LR patterning defects (Fig. 6P). In addition, these double knockdown embryos showed severe body curvature (Fig. 6T,U), previously associated with ciliary dysfunction (Austin-Tse et al., 2013; Becker-Heck et al., 2011; Schottenfeld et al., 2007; Sullivan-Brown et al., 2008). These results suggest that Epb41l5 and Iqcb1 regulate cilia via the same genetic pathway.

We next explored how EPB41L5 and IQCB1 modulate the integrity of the ciliary base and centrosome. Previous studies showed that IQCB1 interacts with the centrosomal protein CEP290 and that CEP290 is required for IQCB1 localization to the centrosome (Barbelanne et al., 2013; Sang et al., 2011).

We hypothesized that EPB41L5 inhibits IQCB1 interaction with CEP290, leading to IQCB1 dissociation from the centrosome. To test that, we coexpressed EPB41L5, IQCB1 and CEP290 in HEK293 cells for a competitive co-immunoprecipitation assay. First, we confirmed the CEP290–IQCB1 interaction by immunoprecipitation (Fig. 7A,B). EPB41L5 coexpression reduced the overall levels of CEP290 and IQCB1 (Fig. 7C,D), as well as the amount of immunoprecipitated CEP290 (Fig. 7A,D) and IQCB1 (Fig. 7B,D). Nevertheless, EPB41L5 coexpression further reduced the amount of co-immunoprecipitated CEP290 (Fig. 7B,D) and IQCB1 (Fig. 7A,D), supporting our hypothesis. On the other hand, CEP290 coexpression did not reduce the interaction between IQCB1 and EPB41L5 (Fig. 7B).

Additionally, we found that EPB41L5 was co-immunoprecipitated by CEP290 (Fig. 7A), suggesting that EPB41L5 might modulate CEP290 centrosomal localization. This possibility was confirmed in RPE-1 cells (Fig. 7E,F,J). The effect was mediated by the N-terminal FERM-FA domain, but not by the C-terminal fragment of Epb41l5 (Fig. 7G–J). These results show that EPB41L5 displaces both IQCB1 and CEP290 from the centrosome.

In this study, we demonstrated a previously uncharacterized role for EPB41L5 in regulating cilia function. Previous work investigated a potential link of EPB41L5 to cilia in mouse epb41l5 null (Lulu) mutants that exhibit LR patterning defects (Lee et al., 2010). Cilia on the left–right organizer are primarily responsible for directional fluid flow generation, and randomized LR patterning is one of the consequences of defects in nodal flow (Basu and Brueckner, 2008; Cartwright et al., 2008; Dasgupta and Amack, 2016). As cilia with normal length formed in the node of Lulu mutants, the authors suggested that cilia disfunction in Lulu mutant mice is secondary to abnormalities in epithelial morphogenesis of the node. However, their explanation did not rule out the possibility that Epb41l5 has an additional role in cilia.

Zebrafish embryos expressing Epb41l5ΔCTD allowed us to distinguish the roles of Epb41l5 in cilia and epithelial morphogenesis. These embryos maintain largely normal epithelia, although minor epithelial defects contributing to abnormal body curvature and pericardial edema cannot be excluded. We showed that cilia motility and LR patterning were similarly impaired in epb41l5-deficient embryos and the embryos expressing Epb41l5ΔCTD. This suggests that Epb41l5ΔCTD maintains its function on epithelial morphogenesis but fails to form or maintain functional cilia.

Our results suggest that EPB41L5 is not required for cilia assembly. Cilia were formed both in epb41l5-deficient embryos and embryos expressing Epb41l5ΔCTD. This is consistent with the presence of cilia in mouse Lulu mutants (Lee et al., 2010). Nevertheless, EPB41L5 regulates ciliary function. This function of EPB41L5 in cilia appears to be mediated by its interaction with IQCB1 and CEP290. We show that EPB41L5 suppresses the localization of both IQCB1 and CEP290 at the ciliary base and centrosome. We propose that this leads to reduced cilia function, such as reduced cilia motility (Fig. 7K–M). This is also consistent with previous reports on roles of IQCB1 and CEP290 in ciliopathies (Baala et al., 2007; Barbelanne et al., 2013; Coppieters et al., 2010; den Hollander et al., 2006; Helou et al., 2007; Leitch et al., 2008; Otto et al., 2005; Sayer et al., 2006; Valente et al., 2006).

Although we show that EPB41L5 suppresses IQCB1 association with the centrosome, EPB41L5 is probably not the only regulator determining IQCB1 localization at the centrosome. A previous study showed that CEP290 interaction via its C-terminal domain is required for IQCB1 localization to the centrosome (Barbelanne et al., 2013). However, CEP290 interaction appears to be insufficient to localize IQCB1 to the centrosome, because IQCB1 lacking the coiled-coil domain maintains CEP290 binding but failed to localize to the centrosome (Barbelanne et al., 2013). In our experiments, IQCB1 mutants missing the C-terminal CEP290-interacting domain still localized to the centrosome, whereas the localization was weaker than that of full-length IQCB1 (Fig. 3C). Our interpretation is that efficient centrosome localization of IQCB1 requires both the coiled-coil domain and the CEP290-interacting domain. EPB41L5 only regulates IQCB1 association with the centrosome mediated by the coiled-coil domain.

Our study highlights dynamic remodeling of the centrosome and ciliary base under physiological and pathological conditions. We demonstrate that EPB41L5 regulates ciliary function through IQCB1 and CEP290 at the ciliary base and centrosome; however, the details of the underlying mechanism remain unclear. Besides direct effects on IQCB1 and CEP290, EPB41L5 can affect cilia by associating with Mind bomb 1 (MIB1) (Dho et al., 2019; Matsuda et al., 2016), which ubiquitinates CEP290 (Villumsen et al., 2013; Wang et al., 2016). Also, EPB41L5 might regulate cilia through actin remodeling, because actin-targeting drugs restored cilia formation in IQCB1-depleted and CEP290-depleted RPE1 cells (Barbelanne et al., 2013). Additional studies are needed to determine how EPB41L5 affects cilia function.

Full-length or truncated forms of zebrafish Epb41l5 and mouse EPB41L5 were generated in a previous study (Matsuda et al., 2016). pCBF-FLAG-tagged human IQCB1 was kindly provided by Dr William Tsang (Montreal Clinical Research Institute, Montreal, Canada; Barbelanne et al., 2013). Full-length and truncated forms of IQCB1 were PCR amplified using Choice Taq DNA polymerase (Denville Scientific) and cloned into a pCS2 expression vector using restriction enzymes (New England Biolabs) and the DNA ligation kit, Mightly Mix (Takara). Primers used for PCR amplification are listed in Table S1. Full-length mouse CEP290 (plasmid #27381) was obtained from Addgene. NotI and StuI fragments were subcloned into NotI and SmaI sites of pCS107 expression vector; thereafter, a FLAG tag was inserted into the NotI site. Plasmid DNAs were purified using the Qiagen Plasmid Midi kit (Qiagen).

HEK293, hTERT-RPE1 and MDCK cells were purchased from ATCC. HEK293 and MDCK cells were maintained in DMEM (Corning) supplemented with 10% FBS. hTERT-RPE1 cells were maintained in DMEM:F-12 medium (Corning) supplemented with 10% FBS. HEK293, hTERT-RPE1 and MDCK cells were transfected using polyethylenimine (PEI; Polysciences), X-tremeGene HP (Roche) and TransIT-X2 (Mirus), respectively. For cilia formation, cells were grown in serum-free medium for 24 h prior to fixation.

Cell lysates were extracted from HEK293 cells using RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) with Protease Inhibitor Cocktail III (Calbiochem). Cell lysates were incubated with primary antibodies (Table S2) and then with protein A/G Sepharose (Santa Cruz Biotechnology) at 4°C for at least 6 h. Sepharose beads were washed in TBST containing 0.05% Triton X-100. Beads were heated at 95°C for 5 min in SDS sample buffer. After SDS–PAGE and transfer to a nitrocellulose membrane with 0.2 µm pore size (GE Life Sciences), western blots were performed. Antibodies used are listed in Table S2. Chemiluminescent signals were acquired using Clarity ECL Western Blotting Substrates (Bio-Rad) on the ChemiDoc MP Imaging System (Bio-Rad).

For immunostaining, hTERT-RPE1 cells and MDCK cells were plated on coverslips (Electron Microscopy Sciences) and Transwell (Corning), respectively. Cells were fixed in 4% PFA (Electron Microscopy Sciences) for 15 min at room temperature then permeabilized with 0.1% Triton in PBS for 2 min. After blocking in 1% BSA for 30 min, cells were incubated with primary antibodies (Table S2) for 1 h and then incubated with secondary antibodies conjugated with AlexaFluor 488, AlexaFluor 647 (Invitrogen) or Cy3 (Jackson ImmunoResearch) for 1 h. Immunostained cells were mounted in Mowiol 4-88 (Sigma). Images were acquired on a Nikon A1R confocal microscope or an Olympus BX51 light microscope equipped with an Olympus CCD camera DP73.

All zebrafish were maintained and handled according to recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. A protocol for animal use was approved by the Institutional Animal Care and Use Committee (IACUC) at Rutgers University and Ichan School of Medicine at Mount Sinai. AB wild-type fish were obtained from ZIRC. moe^b476 mutants were kindly provided by Dr Monte Westerfield (Institute of Neuroscience, University of Oregon, Eugene, OR; Jensen and Westerfield, 2004). Tg[bact:arl13b-gfp] transgenic zebrafish were kindly provided by Dr Brian Ciruna (The Hospital for Sick Children, Toronto, Canada; Borovina et al., 2010). Tg[dusp:ma-gfp]^pt21 zebrafish were kindly provided by Dr Michael Tsang (University of Pittsburgh, Pittsburgh, PA; Molina et al., 2007).

The CRISPR/Cas9 gene editing system was used to generate epb41l5Δctd mutants. Three gRNAs (Table S1) were designed to target the splicing donor site of exon 25 in epb41l5. gRNAs were generated by oligonucleotide assembly and PCR-based methods (Carrington et al., 2015). The HiScribe T7 Quick High Yield RNA Synthesis kit (New England Biolabs) was used to synthesize gRNAs, which were then purified using Microspin G-25 Columns (GE Healthcare). Cas9 protein was purchased from PNA Bio. gRNA (100 pg) and Cas9 protein (250 pg) were mixed prior to microinjection. The gRNA-Cas9 complex was microinjected into a cell of one-cell stage embryos. At 24 h, genome DNA was extracted to determine the efficacy of gene editing. Primers used for PCR are listed in Table S1.

Morpholino oligonucleotides (Table S1) were purchased from GeneTools (Oregon, USA). p53 morpholino was co-injected to prevent p53-dependent cell-death and associating off-target effects of morpholinos (Robu et al., 2007). The mMessage mMachine SP6 Transcription kit (Invitrogen) was used to synthesize mRNAs encoding Epb41l5Δ60. Synthesized mRNA was purified by LiCl precipitation. Morpholinos and mRNA were microinjected into the yolk of one- to two-cell stage embryos.

Ribo probes for in situ hybridization were labeled with digoxigenin-UTP (Roche) using SP6, T3 or T7 RNA polymerase (Roche). In situ hybridization was performed as described previously (Matsuda and Chitnis, 2009). The BCIP/NBT substrate kit (Vectra lab) was used for coloration. Images were taken on a Leica MZ10 stereomicroscope equipped with an Olympus DP73 CCD camera.

Embryos were fixed in 4% PFA in PBS overnight at 4°C. After permeabilization with 0.1% Triton X-100 for 10 min, embryos were incubated with primary antibodies (Table S2) in 1% BSA in PBS overnight at 4˚C. Embryos were then washed in PBS and incubated with AlexaFluor 488- (Invitrogen) or Cy3-conjugated secondary antibodies (Jackson laboratory) in 1% BSA in PBS overnight at 4°C. After washing with PBS, embryos were transferred into 25, 50 and 75% glycerol in PBS. An A1R confocal microscope system (Nikon) was used for imaging.

Morpholinos were injected into the yolk of one- to two-cell stage tg[actb:arl13b-gfp] embryos. At 26–28 hpf, manually dechorinated embryos were mounted in low-melting agarose (Lonza). Time-lapse images were taken on a Nikon A1R confocal microscope system at 5 s intervals for 3 min.

Quantification of fluorescence signals was performed by analyzing individual single plane images. Integrated fluorescence intensity of immunostaining was measured using an ImageJ plugin. The Student's t-test was used to test association of continuous variables. When the cell frequency was not equal to zero, the Chi-squared test was used to test categorical variables. When the cell frequency was equal to zero, the Freeman–Halton extension of the Fisher exact probability test was used to test categorical variables.

We would like to thank members of the M.M. laboratory for technical assistance and comments on the manuscript. We thank Dr Sergei Sokol for helpful suggestions and editing the manuscript.