Crb3 is required to organize the apical domain of multiciliated cells

Epithelia line the external surface and body cavities in animals and ensure exchanges with the extracellular environment as well as protection of internal organs (Buckley and St Johnston, 2022; Rodriguez-Boulan and Macara, 2014). This dual function relies on the organization of epithelial sheets. Epithelial cells are highly polarized, exhibiting structurally and functionally distinct apical and basolateral domains (Román-Fernández and Bryant, 2016). In a majority of epithelia in vertebrates, the apical surface forms microtubule-based protrusions called cilia (Apodaca, 2018). Primary cilia are antenna-like organelles perceiving a variety of environmental cues, including crucial signaling morphogens (Anvarian et al., 2019; Bangs and Anderson, 2017). In the ciliated epithelia lining the airways, the brain ventricles and the reproductive tracts, the apical domain of multiciliated cells (MCCs) is covered by numerous cilia beating coordinately to propel biological fluids (Boutin and Kodjabachian, 2019; Spassky and Meunier, 2017). Thus, structural and/or functional alteration of cilia lead to rare diseases with pleiotropic symptoms called ciliopathies (Reiter and Leroux, 2017).

Xenopus is a powerful model for studying motile cilia biology (Walentek, 2021; Werner and Mitchell, 2012). The embryos are indeed enwrapped in a protective mucociliary epidermis composed of MCCs, bearing hundreds of motile cilia, alternating in a salt and pepper pattern with secretory cell types [goblet cells, ionocytes and small secretory cells (SSCs)] (Boutin and Kodjabachian, 2019; Walentek, 2021). The development of the Xenopus epidermis is a gradual process starting at the segmentation stage with the partitioning of the prospective non-neural ectoderm into an outer epithelial layer made of goblet cells, and an inner mesenchymal layer where MCCs, ionocytes, SSCs and p63-positive basal cells will be born (Deblandre et al., 1999; Dubaissi et al., 2014; Dubaissi and Papalopulu, 2011; Haas et al., 2019; Quigley et al., 2011; Stubbs et al., 2006; Walentek et al., 2014). From mid-neurula onwards (stages 14–31), intense morphogenetic events reshape the tissue, with the sequential insertion of MCCs, ionocytes and SSCs in the outer layer, where they complete differentiation (Haas et al., 2019). Successful intercalation of MCCs is powered by cell shape changes and forces generated by the actin and microtubule cytoskeletons (Boisvieux-Ulrich et al., 1987; Chuyen et al., 2021; Collins et al., 2021, 2020; Lemullois et al., 1988; Sedzinski et al., 2016; Werner et al., 2011). Prior to their intercalation, MCCs undergo random planar migration, constrained by homotypic repulsion, such that they eventually disperse and intercalate at regular intervals (Chuyen et al., 2021). MCC movement is confined to the grooves formed between overlying epithelial cells (Chuyen et al., 2021). Radial intercalation involves the emission by MCCs of filopodia that pull on outer layer junctions to probe their stiffness (Ventura et al., 2022). MCCs preferentially intercalate into stiff high-fold vertices, where they provoke remodeling of the junction in preparation for apical expansion (Ventura et al., 2022). Once inserted into the outer layer, MCCs expand their apical domain, essentially via the 2D pressure exerted by a highly dynamic medial actin network generated by Formin 1 (Sedzinski et al., 2017; 2016).

While inserting into the outer layer, MCCs multiply centrioles deep in the cytoplasm, which are then encased via CP110 and the ciliary adhesion complex (FAK, Paxillin, Vinculin and Talin proteins) in an internal actin network that drives their apical migration (Antoniades et al., 2014; Walentek et al., 2014, 2016). Once docked at the apical surface, centrioles are named basal bodies (BBs). The dynamic ascension of centrioles relies on actomyosin contractility and is regulated by actors of the PCP signaling pathway, the core protein Dishevelled, the ciliogenesis and planar polarity effector (CPLANE) protein Inturned and the PCP effector RhoA (Adler and Wallingford, 2017; Boisvieux-Ulrich et al., 1990; Park et al., 2006, 2008). BBs serve as scaffold to initiate ciliogenesis, and are equipped with a basal foot and striated rootlets, which interact with actin filaments and microtubules to establish a regular lattice able to sustain intense ciliary beating strokes (Herawati et al., 2016; Mahuzier et al., 2018; Spassky and Meunier, 2017; Werner et al., 2011). In particular, an extensive reconstruction of the cortical actin meshwork in two layers enables the planar polarized anchoring of BBs to permit directional beating (Boisvieux-Ulrich et al., 1990; Ioannou et al., 2013; Mahuzier et al., 2018; Mitchell et al., 2007; Park et al., 2006, 2008; Werner et al., 2011). At stage 25, the most apical actin layer reaches its final density, actin bundles progressively encircle each BB, and from stage 26 onwards eventually mature in actin-based protrusion called microridges (Werner et al., 2011; Yasunaga et al., 2022). At stage 29–30, the definitive sub-apical actin layer connects each BB to its immediate posterior neighbor via ciliary adhesion complexes and to microridges via Ezrin (Antoniades et al., 2014; Werner et al., 2011; Yasunaga et al., 2022). This sophisticated actin architecture allows coordinated cilia beating, thus ensuring efficient flow production to clear the surface of the embryo from surrounding microbes (Dubaissi et al., 2018; Nommick et al., 2022).

To allow cell shape changes, deformation of the actin cytocortex must be coupled to deformation of the plasma membrane (Clark et al., 2014). The plasma membrane is attached to the underlying cytoskeleton via specific regulated linker proteins such as Ezrin, belonging to the Ezrin-Radixin-Moesin (ERM) protein family (Clark et al., 2014; Pelaseyed and Bretscher, 2018). In Xenopus MCCs, overexpressed versions of Ezrin have been shown to accumulate in the newly expanding apical membrane of intercalating cells, and from there to be superimposed on the apical actin meshwork maturing into microridges (Yasunaga et al., 2022). Ezrin depletion in Xenopus MCCs induces a functional alteration of the actin cytoskeleton, impacting centriole/BB migration, apical microridge formation, and anchoring of BBs to microridges (Epting et al., 2015; Yasunaga et al., 2022). Thus, the fine-tuned localization of Ezrin correlates with the control of MCC terminal differentiation. However, the molecular mechanisms underlying the proper control of Ezrin subcellular localization in Xenopus MCCs are elusive.

The Crumbs (Crb) polarity proteins have pivotal functions in processes involving quick remodeling of the actin cytoskeleton in coordination with the adjacent membrane and are ERM biochemical and genetical interactors (Aguilar-Aragon et al., 2020; Bajur et al., 2019; Flores-Benitez and Knust, 2015; Gao et al., 2016; Kerman et al., 2008; Letizia et al., 2011; Médina et al., 2002; Salis et al., 2017; Schottenfeld-Roames et al., 2014; Sherrard and Fehon, 2015; Simões et al., 2022; Tilston-Lünel et al., 2016; Vernale et al., 2021; Wei et al., 2015; Whiteman et al., 2014). Biochemical studies have demonstrated that the Crb cytoplasmic tail possesses a FERM (4.1, ezrin, radixin, moesin) domain allowing ERM binding (Médina et al., 2002; Sherrard and Fehon, 2015; Tilston-Lünel et al., 2016; Wei et al., 2015; Whiteman et al., 2014). Functional studies essentially performed in Drosophila embryos showed that Crb stabilizes the ERM protein Moesin at specific sub-apical domains depending on the nature of the remodeling tissues (Letizia et al., 2011; Salis et al., 2017; Sherrard and Fehon, 2015). Crb and ERM both independently organize the apical cytocortex and interact genetically for dampening the actomyosin contractility during dorsal closure in Drosophila embryos (Bazellières et al., 2018; Fehon et al., 2010; Flores-Benitez and Knust, 2015). In vertebrates, the Crumbs family member Crb3 displays a quasi-ubiquitous epithelial expression (Bazellieres et al., 2009). Strikingly, Crb3- and ezrin-deficient mice display the same abnormal intestinal phenotype, with villi fusion and microvilli atrophy (Charrier et al., 2015; Saotome et al., 2004; Whiteman et al., 2014). Both Ezrin and Crb3 have been shown to be required for ciliogenesis (Bazellières et al., 2018; Bo et al., 2023; Epting et al., 2015; Fan et al., 2004, 2007; Hazime and Malicki, 2017).

The small Rab-GTPase Rab11 family proteins are central regulators of both apical trafficking and ciliogenesis as it initiates a Rab-GTPase cascade driving membrane delivery for primary cilia growth (Knödler et al., 2010). In several model systems, Crb colocalizes with Rab11 endosomes, which are required for Crb apical delivery (Aguilar-Aragon et al., 2020; Bo et al., 2023; Buck et al., 2023; Iioka et al., 2019; Schlüter et al., 2009). In addition, Rab11 is expressed in Xenopus MCCs and accumulates in their subapical domain at the time of their intercalation, and Rab11-depleted MCCs display reduced apical surface (Kim et al., 2012).

We therefore investigated how Crb3, ERM proteins and Rab11 could cooperate to organize the apical domain and build protrusions, such as cilia in Xenopus MCC.

In Xenopus laevis, there are two crb3 homeologs, crb3.S and crb3.L, with highly conserved DNA (85% identity) and protein (85% identity) sequences. As no tools were available to observe the precise endogenous localization of Crb3 proteins in Xenopus laevis, custom-made homeolog-specific antibodies were generated. To do so, we chose the extracellular domain of Crb3.S (QNVTTPAPGKLSESA) and Crb3.L (QNVTTSAPDRLSESAR) as immunogens, as this region is the most divergent between the two proteins (non-identical amino acids are underlined). To assess the specificity of the Crb3.L and Crb3.S antibodies, we performed competition assays with the corresponding immunogenic peptides, which caused the extinction of immunofluorescent (IF) signals in whole embryos (Figs S2, S4). By combining these new tools with labeling for Utrophin, which stains the actin meshwork, acetylated-Tubulin, which stains stable cytoplasmic microtubules and cilia, and Centrin, which stains centrioles and BBs, we were able to show that Crb3.L and Crb3.S have dynamic and exclusive patterns in the mucociliary epidermis of Xenopus embryos (Fig. 1; Fig. S1).

We focused on the expression of Crb3.L in intercalating MCCs (Fig. 1). As Crb proteins are transmembrane proteins, punctuate cytoplasmic staining likely corresponds to vesicles. In inserting MCCs, Crb3.L-positive vesicles start to accumulate deep in the cytoplasm close to microtubule arrays (Fig. 1A,A1), at the level of the internal actin meshwork (Fig. 1B,B′), in the vicinity the ascending centrioles (Fig. 1C,D). Next, as apical emergence and conjoint centriole ascension proceeds, Crb3.L is progressively redistributed to the expanding apical domain proximate to actin filaments (Fig. 1B,B′,C,D) and appears in the ciliary shaft of growing cilia (Fig. 1A2). In contrast, Crb3.S exhibits an earlier and wider expression, being detected in all cells of the outer layer during gastrulation (Fig. S1). From gastrulation onwards, Crb3.S expression level is prominent in goblet cells, albeit at variable levels, presumably absent from ionocytes and particularly high in SSCs (Figs S1, S4). Strikingly, it has very low levels or is absent from intercalating and mature MCCs (Figs S1, S4).

Based on this analysis and the central function of Crb proteins in cell shape changes empowered by quick actin remodeling, we decided to evaluate the role of Crb3.L in the morphogenesis of the apical domain of MCCs.

To address the function of crb3.L during ciliogenesis, we used an antisense morpholino knockdown strategy with two morpholinos targeting either the translation initiation site (ATG-crb3.L-mo) or the 5′ untranslated region (5′UTR-crb3.L-mo) of the crb3.L mRNA. The efficiency of 5′UTR-crb3.L-mo was evidenced by the extinction of Crb.3L IF staining in whole embryos, in both MCCs and non-MCCs (Fig. S3). To unravel a potential effect on ciliogenesis, we examined stage 28 embryos, when MCCs are mature, with a well-developed ciliary tuft (Fig. 2, control). Crb3.L depletion affected the aspect of the ciliary tuft of MCCs, which contained fewer cilia. Quantitative analysis revealed that ciliogenesis was significantly impaired upon injection of either of the crb3.L targeting morpholinos, with 30% (95% c.i. of 24–37%, ATG-crb3L-mo) to 39% (95% c.i. of 26–53%, 5′UTR-crb3.L-mo) of morphant MCCs forming abnormal ciliary tufts with a strong reduction in the number of cilia (Fig. 2). As the two crb3.L targeting morpholinos displayed comparable effects, 5′UTR-crb3.L-mo was used in the rest of the study.

As correct BB apical docking is a prerequisite for proper ciliogenesis, we compared the BB position in tracer-injected control and morphant MCCs. Two complementary approaches were used to observe the organization of BBs: immunolabeling and transmission electron microscopy (TEM). At stage 28, in the control situation, BBs covered most of the apical area, with quite a regular distribution (Fig. 3A,B,E,F,I). In contrast, in 5′UTR-crb3L-mo-injected MCCs, defects in BB positions and distributions were readily apparent in confocal and TEM images. BBs remained stuck into the cytoplasm (Fig. 3C,D,G–I) or unevenly scattered, forming clumps when they reached the apical surface (Fig. 3I). To quantify the penetrance of this phenotype, we analyzed two parameters: the number of BBs reaching the apical domain and the dispersion of BBs along the apico-basal axis (Fig. 3J,K). The number of BBs reaching the apical surface was drastically decreased in morphant compared to control MCCs (Fig. 3J). Furthermore, BBs were located significantly deeper into the cytoplasm of morphant compared to control MCCs (Fig. 3K). Importantly, both aspects of the BB phenotype were significantly rescued by the injection of a HA-tagged Crb3.L construction insensitive to the 5′UTR-crb3.L morpholino (Fig. 3J,K).

All these data demonstrate that Crb3.L is essential for an efficient apical migration and/or docking of centrioles/BBs.

Migration and docking of BBs is a stepwise process relying on proper actin meshwork dynamics and organization. In several models, Crb has been shown to regulate actin dynamics (Bazellieres et al., 2009; Flores-Benitez and Knust, 2015; Röper, 2012; Salis et al., 2017; Sherrard and Fehon, 2015; Simões et al., 2022). As Crb3.L-positive vesicles are detected at the time the actin meshwork remodels (Fig. 1) (Ioannou et al., 2013), we hypothesized that Crb3.L could be involved actin cytoskeleton organization in MCCs. We therefore assessed apical actin meshwork organization in stage 28 control and 5′UTR-crb3L-mo-injected MCCs. In both noninjected (yellow contour arrowheads) and injected (white contour arrowheads) control cells, the labeled actin meshwork was dense and structured, and found in close proximity to the cell–cell junctions (Fig. 4A–D,I,J). In 5′UTR-crb3L-mo-injected MCCs, the central actin meshwork appeared to be much dimer, fragmented and largely disconnected from the cell junctions (Fig. 4E–H,K,L). Measurement of apical medial actin staining intensity confirmed a drastic decrease of ∼40% in morphant compared to control MCCs (Fig. 4N). At this stage, the mean apical medial area of MCCs was not significantly different between the two groups, suggesting that in mature MCCs there is no strict correlation between the apical cell surface area and the actin meshwork density (Fig. 4M).

These results show that in mature MCCs, Crb3.L is required for proper organization of the cortical actin meshwork, but seems dispensable for the expansion of the apical domain to its final dimension.

We and others have shown that ERM and CRB proteins cooperate to enable apical domain construction (Aguilar-Aragon et al., 2020; Bajur et al., 2019; Flores-Benitez and Knust, 2015; Letizia et al., 2011; Médina et al., 2002; Salis et al., 2017; Sherrard and Fehon, 2015; Tilston-Lünel et al., 2016; Wei et al., 2015; Whiteman et al., 2014). Previous studies in Xenopus have demonstrated that Ezrin-depleted MCCs exhibit a very similar phenotype to the one we observed in Crb3.L-depleted MCCs (Epting et al., 2015).

Thus, we hypothesized that Crb3.L might affect the expression and/or the subcellular localization of activated Ezrin. As a first step to evaluate this possibility, we analyzed the localization of phospho-activated ERM during MCC differentiation (Epting et al., 2015; and see Materials and Methods). pERM expression was detectable in the protruding membrane of MCC starting their insertion into the outer layer (Fig. 5, stage 18; Fig. S5). At this stage, pERM staining was also detected in close proximity to the ascending centrioles/BBs (Fig. S5, stage 21). As apical expansion proceeded, pERM formed bright patches on the expanding apical surface (Fig. 5, stage 22, 25), then the patches coalesced in a more structured network superimposed on the mature actin meshwork (Fig. 5, stage 25, 30). Accumulation of pERM at cell-cell junctions was observed only in fully mature cells (Fig. 5, stage 30).

The similar apical distribution of Crb3L and pERM suggests that Crb3.L could control the localization of activated pERM to regulate the nascent actin network during apical expansion. We thus compared the distribution of pERM in control and Crb3.L-depleted immature MCCs undergoing apical expansion. Crb3.L depletion induced a 33.2±11 % (mean±s.d.) decrease of apical pERM signal intensity (Fig. 6A,B,D,F).

In correlation with this effect, Crb3.L depletion provoked a 30±10% (mean±s.d.) decrease in the average apical medial surface of newly emerged cells (Fig. 4C). These results are in contrast to the apparently preserved area of stage 28 morphant MCCs (Figs 4M, 6C), and suggest that non-Crb3.L-dependent processes might rescue the initial defect in apical expansion.

These data advocate for Crb3.L being required for regulating pERM on the emerging apical surface of MCCs, so as to bridge newly formed actin filaments to the apical plasma membrane and promote early apical expansion.

Besides its function at the apical cytocortex, the striking abundance of Crb3.L in the vicinity of the BBs, as well as the BB migration defect induced by Crb3.L depletion, suggests that vesicle trafficking of Crb3.L might mediate BB ascension. In several model systems, Crb colocalizes with Rab11 endosomes, which are pivotal mediators of apical trafficking and are required for Crb apical delivery (Aguilar-Aragon et al., 2020; Bo et al., 2023; Buck et al., 2023; Iioka et al., 2019; Schlüter et al., 2009). In addition, Rab11a accumulate in the subapical domain of MCCs at the time of their intercalation (Kim et al., 2012). To ascertain the hypothesis that the BB migration defect could be linked to Crb3.L defective apical vesicular trafficking, we checked if Crb3.L also label Rab11a positive endosomes. During MCCs insertion and early apical surface expansion, detection of endogenous Rab11a via immunofluorescence shows that Rab11a endosomes are closed to ascending centrioles/BBs (Fig. 7A–B′). Moreover, Crb3.L labeling, by means of endogenous or HA-tagged construct detection, partially colocalizes with Rab11a staining (Fig. 7A–B′). These data suggest that some aspects of Crb3.L function in MCC could be mediated by Rab11a-positive endosomes at centrioles/BBs.

To test whether Rab11a-positive endosome trafficking was required for centriole/BB ascension in MCCs we performed Rab11a loss of function experiments using a well-characterized morpholino (Kim et al., 2012) and examine MCC phenotype in stage 28 embryos when MCCs are normally fully mature. Rab11a depletion led to a strong centrioles/BBs migration defect: centrioles/BBs aggregated in the cytoplasm or were unevenly distributed at the apical surface (Fig. 7C,D,F,G). The apical meshwork formation was affected in Rab11a-depleted MCCs with a 26% decrease in actin density between the Rab11a morpholino-injected cell and noninjected cells in mosaic morphants for a morpholino dose of 22 ng (Fig. 7C,E,H–J). This phenotype was more drastic with 30 ng of Rab11a morpholino leading to a 52% diminution of actin density between Rab11a morpholino-injected and noninjected cells in mosaic morphants. The size of the apical medial domain was also decreased upon Rab11a depletion, with a reduction of ∼25% between Rab11a morpholino-injected and noninjected cells in mosaic morphants injected with a dose of 22 ng, and 40% between Rab11a morpholino-injected and noninjected cells in mosaic morphants injected with a dose of 30 ng (Fig.7C–E,H,I,K). Thus, the phenotypes of Rab11a- and Crb3.L-depleted MCCs are very similar with respect to the BB migration defect and apical actin meshwork formation. We found that the reduction of the size of the apical medial domain is more pronounced with Rab11a depletion as it persists in mature MCCs. All together, these data support the hypothesis that Crb3.L-Rab11a vesicle-mediated apical trafficking is required for BB ascension, apical expansion and apical actin meshwork formation.

In this study, thanks to newly custom-made polyclonal antibodies, we show that the crb3.L homeolog is highly expressed in MCCs, with a dynamic pattern of expression at the nexus of BBs and the intensively remodeling cytoskeleton of microtubules and actin. Crb3.L-depleted MCCs displayed a complex phenotype associated with BB migration defects, a reduction in the apical surface and disorganization of the apical actin meshwork, thus precluding normal ciliogenesis. We further unveil the transient association of endogenous activated ERM proteins with ascending centrioles/BBs and its recruitment to the growing apical surface during the last step of MCC intercalation. The loss of apical enrichment of pERM paralleled the initial reduction in apical domain expansion in Crb3.L-depleted MCCs. Our data advocate for Crb3.L-anchoring Ezrin in an open active conformation at the apical surface of MCCs, thus allowing coordinated growth of the apical membrane with extensive actin cytoskeleton remodeling.

So far, studies addressing the question of Crb3 tissue and subcellular localization in Xenopus laevis have relied on the expression of a C-terminally GFP-tagged version of Xenopus tropicalis Crb3. The amino acid sequence of Xenopus tropicalis Crb3 displays 83% and 81% identity with Crb3.L and Crb3.S respectively (Chalmers et al., 2005; Wang et al., 2013). This construct allowed the description of the dynamic changes in Crb3 subcellular localization during differentiation and morphogenesis of the non-neural ectoderm, but did not address the function of endogenous Crb3. To determine which homeolog was predominantly expressed in MCCs, and to describe their subcellular localization, antibodies targeting each one of the proteins were generated. We controlled their specificity by immunizing peptide competition assays and a loss of function experiment for Crb3.L. These new antibodies allowed us to unveil the distinct but complementary expression of Crb3.L and Crb3.S in the developing mucociliary epidermis of Xenopus laevis embryos, supporting a division of labor scenario following gene duplication.

We confirm that a large pool of Crb3, namely Crb3.S, is indeed located in numerous small intracellular vesicles (Wang et al., 2013). Moreover, the progressive restriction of the endogenous Crb3.S to the outer layer at the time of its mucociliary differentiation is in accordance with the published pattern based on the expression of Xenopus tropicalis Crb3–GFP (Wang et al., 2013). These data advocate that Crb3.S is a marker signing the beginning of the differentiation of the mucociliary epidermis. Later on, the expression of Crb3.S becomes very variable among the different cell types and surprisingly extremely high in the SSC.

To date, very few SSC markers have been described and they are either transcription factors, such as Foxa1, or secreted compounds (Otogelin-like, HNK1, IPTKb, serotonin and Xpod) (Dubaissi et al., 2014; Kurrle et al., 2020; Walentek et al., 2014). Thus Crb3.S is, so far, the sole transmembrane protein described to be highly expressed in SSCs. It extends the short list of bona fide SSC markers and offers the unique advantage of allowing visualization of cell shape via endogenous staining of vesicles and cytoplasmic membranes.

The SSCs are central to the epithelial anti-infective barrier function in Xenopus laevis embryos because they secrete anti-microbial substances (Dubaissi et al., 2014; Walentek et al., 2014). Therefore, it would be interesting to study the impact of Crb3.S depletion on the secretory function of the SCC as well as its functional consequences on innate immunity.

Until now, studies on Crb3 localization in ciliated cells have focused on its localization in the shaft of primary cilia (Fan et al., 2004, 2007; Hazime and Malicki, 2017; Sfakianos et al., 2007). Accordingly, functional studies on Crb3 revealed its function in the regulation of intraciliary transport (Fan et al., 2004; Hazime and Malicki, 2017; Sfakianos et al., 2007) and ciliary membrane composition (Hazime and Malicki, 2017). Here, we do observe a spotty staining of Crb3.L along the axoneme, suggesting that Crb3.L might also regulate ciliary transport and/or ciliary shaft composition in the motile cilia of MCCs.

Before being detected in cilia, Crb3.L accumulates in the sub-apical domain of the intercalating MCCs, in the vicinity of the ascending centrioles/BBs, as well as at the newly expanding apical membrane. These subcellular localization data are well in accordance with the compound phenotype we observed: the association of defective centriole/BB ascension, alteration of the apical actin meshwork and apical surface reduction. These data suggest that, in MCCs, Crb3 is not exclusively required for ciliary transport, as it has been described for primary cilia (Fan et al., 2004; Hazime and Malicki, 2017; Sfakianos et al., 2007). Our work points that Crb3.L is likely an iterative player of the stepwise process of multiciliogenesis, being subsequently required for centriole/BB ascension, BB anchoring or fusion to the apical membrane, apical actin meshwork construction, and finally cilia building/maintenance.

We observed a partial colocalization of Crb3 and Rab11a labeling in a subpopulation of vesicles that were very close to BBs in inserting MCCs or MCCs initiating expansion of their apical surface. Moreover, Rab11a depletion altered BB ascension, as in Crb3.L morphant MCCs. Thus Crb3.L might regulate centriole/BB ascension through its interaction with the Rab11a apical trafficking endosome. Crb3.L- and Rab11a-positive endosomes might recruit molecular motors required for the centriole/BB ascension process. Crb3.L- and Rab11a-positive endosomes could also deliver specific regulatory proteins or structural cargos to ascending centrioles/BBs. Accordingly, Bo et al. demonstrated very recently that Crb3.L- and Rab11a-positive positive endosomes were carrying components of the γ-Tubulin complex to the BB of cells bearing a primary cilium (Bo et al., 2023).

A widely accepted function of Crumbs is to control the size of the apical domain in various model systems in vivo. In Drosophila, Crb loss of function leads to smaller apical surfaces in tracheal and pupal wing cells (Salis et al., 2017; Skouloudaki et al., 2019). Crb might also selectively participate in the expansion of apical domain subdivisions, such as the stalk and rhabdomere regions of Drosophila photoreceptors, the inner segment of the photoreceptor and the apical microvilli-like protrusions of mechano-sensory neurons in zebrafish, and the microvilli of enterocytes in mice (Charrier et al., 2015; Desban et al., 2019; Omori and Malicki, 2006; Pellikka et al., 2002; Richard et al., 2009; Whiteman et al., 2014). In summary, Crb is required for apical growth in tissues undergoing intense morphogenetic events.

In Xenopus laevis, overexpression of Crb3 leads to expansion of the apical domain of epithelial cells in the early embryo (Chalmers et al., 2005). However, the question of the contribution of Crb3 to the very dynamic process of the MCC apical expansion had not been evaluated. Here, we show that prior to apical MCC expansion, Crb3.L labels vesicles accumulating in the subapical domain of inserting cells. Further on, Crb3.L relocates from this vesicular pool to the apical membrane at the time of apical expansion. Thus Crb3.L-positive vesicles are ideally positioned for the fast delivery of a ready-made stock of membrane dedicated to apical membrane expansion.

Crb3.L depletion causes an initial reduction of the apical surface in association with decreased levels of F-actin and pERM at the apical cell aspect. MCC apical surface expansion is autonomously driven by 2D pressure powered by the apical medial F-actin pool (Sedzinski et al., 2016). Ezrin depletion in the Xenopus multiciliated epithelia leads to profound actin meshwork disorganization (Epting et al., 2015; Yasunaga et al., 2022). The Crb juxta membrane FERM-binding domain (FBM domain) binds to the FERM domain of ERM (Médina et al., 2002; Wei et al., 2015; Whiteman et al., 2014). In the ovarian follicle epithelium of Drosophila, Crb with a nonfunctional FBM domain does not stabilize the activated ERM Moesin in the subapical region resulting in the loss of filamentous actin (Sherrard and Fehon, 2015). Thus, in MCCs, Crb3.L might regulate phosphorylated Ezrin in the growing apical domain, allowing progressive growth of the actin meshwork in coordination with apical membrane delivery.

We showed that, in immature MCCs, Crb3.L vesicles first accumulate in the vicinity of ascending centrioles/BBs, where they display a partial co-distribution with Rab11a endosomes. Consistent with this, both Crb3.L and Rab11a loss of function caused defective apical migration of centrioles/BBs. At a later stage, Crb3.L is re-localized from the internal vesicular pool to the apical membrane of the emerging MCC. Consistent with this, in Crb3.L-depleted cells the apical actin meshwork is defective and pERM staining is decreased. Strikingly, in Xenopus MCCs, apical cytoskeleton organization and centriole/BB migration appear to be tightly coordinated (Ioannou et al., 2013; Nommick et al., 2022). It is therefore difficult to ascertain whether Crb3.L function is restricted to centriole migration, to cytocortex organization or acts in both. Answering this question would require new tools to discriminate between the functions of the distinct Crb3.L pools in live embryos. As an example, light-induced clustering elegantly demonstrated that Rab11 endosomes contribute to left-right organizer lumen formation in zebrafish (Aljiboury et al., 2023; Rathbun et al., 2020). Deeper resolution of the molecular mechanisms at play might also be achieved by proximity labeling strategies. As an illustration, the APEX2 proximity ligation assay allowed the molecular mapping of a Crumbs3-specific subcellular domain in MDCK cells (Tan et al., 2020). Further work based on such new tools might help to elucidate the function of distinct Crb3.L pools in MCCs.

In conclusion, our work unveils a new aspect of Crb3 function in multiciliogenesis in vertebrates. At the tissue level, Crb3 was shown to be required for the differentiation of the proximal airway mucociliary epithelium, including MCC progenitors (Szymaniak et al., 2015). Consequently, MCCs, and hence cilia, are absent from the proximal airway of lung targeted Crb3-knockout mice (Szymaniak et al., 2015). This likely explains the respiratory distress syndrome of germline Crb3-knockout mice (Charrier et al., 2015; Szymaniak et al., 2015; Whiteman et al., 2014). In our model, MCCs are present, but display an abnormal ciliary tuft, suggesting that Crb3.L is dispensable for MCC specification in the Xenopus mucociliary epithelium.

At the cellular level, Crb3 appears to be involved in ciliary trafficking in MCCs as suggested by the cilia phenotype of crb3-null zebrafish mutants (Hazime and Malicki, 2017). In our model, a ciliary trafficking defect is not excluded, but it was not possible to test here as centrioles/BBs were not docked at the apical membrane. Thus, Crb3-dependent trafficking seems to be iteratively reemployed during multiciliogenesis first for centrioles/BB ascension possibly via cooperation with Rab11 apical trafficking endosomes, next in cilia elongation/maintenance via interactions with intraflagellar transport (Hazime and Malicki, 2017; Sfakianos et al., 2007).

All experiments were performed following the Directive 2010/63/EU of the European parliament and of the council of 22 September 2010 on the protection of animals used for scientific purposes and approved by the ‘Direction départementale de la Protection des Populations, Pôle Alimentation, Santé Animale, Environnement, d'Ille et Vilaine’ (agreement number 7257).

Wild-type or albino Xenopus laevis females were obtained from Biological Resources Center (CRB France or NASCO, USA). Eggs were fertilized in vitro, and embryos were de-jellied and reared as previously described (Nommick et al., 2022). For microinjections, embryos were placed in a 5% Ficoll-buffered solution (Hatte et al., 2018; Nommick et al., 2022). To specifically target the epidermis, eight-cell embryos were injected in a ventral blastomere of the animal pole. For depletion experiments, embryos were injected with 15 or 30 ng of 5′UTR-crb3.L-mo or ATG-crb3.L-mo respectively, together with 200 pg of gfp-gpi mRNA. For rescue experiments, 5′UTR-crb3.L-mo was co-injected with 200 pg of mrfp-caax mRNA and the HA tagged resistant crb3.L mRNA.

Predictive open reading frames (ORFs) of crb3.L and crb3.S of Xenopus laevis were identified by comparing genomic, RNA and protein sequences of Xenopus tropicalis and Xenopus laevis, as well as using the annotated versions of the genomes (Xenbase). cDNAs were prepared via reverse transcription (SuperSriptIII first strand synthesis, Invitrogen) from RNA extracted with silica membrane column purification (PureLink RNA mini kit Thermo Fisher Scientific). The following primer were used to amplify the coding sequences of each gene: crb3.L, forward primer 5′-CTCTGCGTCCCTACCCTG-3′ and reverse primer 5′-TGTAAGGCGCAGTTTTGACC-3′; crb3.S, forward primer 5′-CAGGCTGATCTCTGCATCAC-3′ and reverse primer crb3.S: 5′-AGGAATCTCTCCTGGAGTTCA-3′.

PCR fragments were purified and cloned into pGEM-T Easy+ (Promega), and subcloned in pCS2+ using the EcoRI cloning site to obtain the 29bp-crb3.L-pCS2+ vector, 29 bp of the 5′UTR were present in this first construct. The HA tagged crb3.L rescue construct resistant to 5′UTR morpholino (HA-crb3.L-pCS2+), was constructed as follows. First, the HA epitope sequence was inserted immediately 3′ to the signal peptide sequence by site directed mutagenesis using the Quikchange XL site-directed mutagenesis kit (Agilent) to obtain the following intermediate vector: 29bp-HA-crb3.L-pCS2+. To this end the following primers were used: HAmutcrb3.L, forward primer 5′-CCTGTCATTAGTGAAAGCTTACCCATACGATGTTCCAGATTACGCTCAGAATGTCACCACTTCAG-3′ and reverse primer 5′-CTGAAGTGGTGACATTCTGAGCGTAATCTGGAACATCGTATGGGTAAGCTTTCACTAATGACAGG-3′.

The insertion of an optimal Kozak sequence with simultaneous deletion of the 5′UTR remnant were performed by PCR using as a template the restriction fragment obtained by the digestion of 29bp-HA-crb3.L-pCS2+ by EcoRI.

The following primers were used: forward 5′-GAGAGAATTCCGCCACCATGCTCGATTATCTA-3′ and reverse 5′-GAGACTCGAGACTAGTGATTTGTAAGGCGC-3′.

PCR fragments were digested, purified and cloned into PCS2+ using the EcorI-XhoI cloning sites. The accuracy of the sequences was verified by sequencing.

The plasmids pCS2+, pCS2+-RFP-caax and pCS2+-GFP-GPI were from lab stocks. pCS107-GFP-CP110 was a kind gift of Peter Walentek (Center for Biological System Analysis, Universitätsklinikum Freiburg, Freiburg, Germany).

Linearized plasmids were used as template for in vitro synthesis of mRNA using the SP6 m message machine kit (Ambion) and purified using the Megaclear Kit (Life Technology Ambion).

Two independent morpholinos were designed targeting either the 5′UTR of crb3.L, named 5′UTR-crb3.L-mo, encoded by 5′-ACAGTAACAGGGTAGGGACGCA-3′ or the translational initiation codon named ATG-crb3.L-mo encoded by 5′-TTAGTAGATAATCGAGCATGTGGAC-3′. The sequence of morpholino targeting the translation initiation codon of the two Rab11a homologs was 5′-TACCCATCGTCGCGGCACTTCTGAC-3′ as in Kim et al. (2012).

Embryos were staged and arrested at appropriated stages according to Faber and Nieuwkoop (Faber and Nieuwkoop, 2020). For stage 21–22, embryos were incubated at 13°C until late afternoon on the day of injection and then placed at 16°C for 2 days. For stage 28, embryos were incubated at 13°C for 4 days. For embryos that did not hatch, the vitelline membrane was manually removed with tweezers before proceeding to fixation. Fixation conditions varied according to the antibodies and/or the probes (see Tables S1–S3). After fixation, whole embryos were immunostained or were further processed for cryosections. Cryopreservation was achieved by incubation in serial graded sucrose solutions (5 to 30%). Embryos were then embedded in Optimal Cut Temperature medium (OCT VWR chemicals), snap frozen in a dry ice ethanol bath, and stored at −80°C, cryosections that were 5 to 10 µm thick were obtained with a cryostat.

For whole-mount immunostaining, buffers and detergents varied with the antibodies or probes. For staining including pERM and SiR actin, embryos were washed in Tris-buffered saline (TBS) with 0.05% NP-40 and blocked in the same solution with 15% fetal calf serum (FCS). The pERM antibody detects endogenous levels of Ezrin only when phosphorylated at threonine 567 but also reacts with the other ERMs (Moesin at threonine 577 and Radixin at threonine 564). Ezrin mRNA is detected via in situ hybridization in Xenopus laevis MCC (Epting et al., 2015). Thus, pERM antibody is likely detecting pEzrin in Xenopus laevis MCC but we cannot exclude it could also react with pMoesin and pRadixin. For the other stainings, embryos were washed in maleic acid and 0.1% Triton X-100 (MABX) and blocked in MABX with 15% FCS.

Information about the antibodies, fixation conditions, specific changes regarding the following general immunofluorescence protocol are provided in Tables S1–S3.

Next, embryos were usually incubated overnight at 4°C with the first antibodies (see Table S1). Embryos were then extensively washed for 4 to 6 h, incubated for 30 min to 1 h in blocking buffer, then transferred in the solution containing the appropriate combination of goat secondary Alexa Fluor-coupled antibodies at 1:500 for 1.5–2 h (see Table S2). After incubation, embryos were briefly washed and mounted in antifading medium (mowiol, DABCO) between slide and coverslip. Exceptions to this protocol were the increased incubation time for anti-Crb3 antibodies (3–4 days, 4°C), addition of SiR-actin (1:1000) to the fixative solution (1 h) and with the secondary antibodies (1.5 h) (see Table S3). An alternative to SiR-actin was Acti Stain used at 1:150 dilution in the secondary antibody solution.

For Crb3.L immunostaining on cryosections, reagents and conditions were the same as described above. Peanut agglutinin (PNA) was incubated on the slide for 15 min after the secondary antibody incubation and wash (see Table S3).

Specific antibodies to Crb3 homeolog L and S were obtained by immunizing rabbits using the speedy protocol from Eurogentec (Seraing, Belgium). The synthetic peptides identical to the ectodomain of Crb3.L (H-QNVTTSAPDRLSESAR-C) or Crb3.S (H-QNVTTPAPGKLSESA-C) were coupled with the KLH (keyhole limpet hemocyanin) carrier protein on their C-terminal part. These peptides were used for affinity purification of the immunized rabbits. Validation of the antibodies was demonstrated via immunizing peptide competition assays for Crb3.L and Crb3.S and depletion through morpholino knockdown for Crb3.L (see Figs S2,S4).

Competition assays were performed using the three following conditions. The antibodies against Crb.3 were incubated overnight at 4°C with either the immunizing peptide, or an unrelated peptide (QTISDPGEEDPPVSKC present in Oopsacas minuta type IV collagen, negative control) or only MABX buffer (positive control). The molar ratio between antibodies and peptides was 1 mol for 25 mol, respectively, and antibodies were used at 1:200 (2.6 μg/ml). After centrifugation (21,000 g for 4 h at 4°C) to get rid of potential antibody–peptide complexes, supernatants were used to perform the immunostaining protocol. Incubation of antibodies with the immunizing peptide caused the loss of immunostaining unlike incubation with Oopsacas minuta type IV collagen.

Confocal images were acquired using ZEISS LSM 510 and 780, Leica SP5 and SP8 confocal microscopes with 63× oil objectives. Sequential or simultaneous laser excitations were applied depending on fluorophores, combinations, spectra, staining brightness and persistence.

Centrin staining was used for BB detection. Apical surfaces of individual MCCs marked by the tracer corresponding to regions of interest (ROI), which were segmented manually. Maximum intensity projection of the three planes framing the MCCs apical surface was applied to Centrin pictures and converted into 8-bit format. Automated detection of Centrin dots was performed in the ROI using the function find maxima>point selection of Fiji software. For this purpose, images were acquired with 63× oil objective applying a 2.5 zoom and Z-slice interval of 0.4 µm.

Z-series were set to frame the totality of the Centrin staining in MCCs. The number of Z-slices containing Centrin dots was used as a proxy for the apico-basal dispersion of BBs in the cytoplasm. These values were converted into distances from the cell apex. For this purpose, images were acquired with 63× oil objective applying a 2.5 zoom and Z-slice interval of 0.4 µm.

Fluorescence of pERM and F-actin signals was measured on eight-bit pictures using the mean gray value function of Fiji on maximal intensity projection of the three most apical cell planes. The apical medial area delimited by the apical cell–cell junctions but excluding them was used as a proxi for apical surface measurements (intensity, size) for both stains.

pERM signal quantification was performed in mosaic tracer-injected control and crb3.L-depleted embryos to circumvent non-biological interindividual variations of the staining intensity. Measures of mean pixel intensity of pERM staining per cell were performed on confocal plane including at least two tracer positive (cti) and negative MCCs (ctni) in control embryos, as well as two 5′UTR-crb3L-mo tracer positive (moi) and negative (moni) MCCs in morphant embryos. Then, the average intensity (mean pixel intensity) of each cell population was calculated for each plane. The absence of variation of the average intensity between cti and ctni was a prerequisite for interpreting a variation between moi and moni.

F-actin staining was quantified in two ways according to the associated staining. When F-actin staining was coupled with pERM staining, F-actin staining was quantified the same way as pERM (see above). When F actin staining was not associated with pERM, medial and junctional actin intensity were measured. The line delimitating the medial domain was taken as the internal limit of the junctional region, the outer limit was automatically drawn by the make band tool of Fiji software, fixing thickness of the band at 2 µm. As junctional actin does not vary between 5′UTR-Crb3.L morpholino-injected and control cells, in mosaic morphants, it can be used as reference point for medial actin intensity.

For pERM and F-actin staining, images were acquired with a 63× oil objective and Z-slice interval of 0,5 µm.

Stage 28 embryos were processed for electron microscopy as previously described (Revinski et al., 2018). 80 nm sections were made with a Leica Ultracut UC7 (Leica, Germany). Images were acquired using a Tecnai G2 (Thermo Fisher Scientific) microscope and a Veleta camera (Olympus Japan).

Statistical analyses were performed with RStudio (version 1.4.1717). Before comparing the mean of variables, normality and homoscedasticity were evaluated with Shapiro and Barlett tests, respectively. When data followed normal and homoscedastic distribution, an unpaired two-tailed Student's t-test was applied for comparing two groups. When data did not complete these two conditions, non-parametric tests were used. For comparison of two groups, a Wilcoxon test was used. For comparisons including more than two groups, Kruskal–Wallis tests were used, followed by a post hoc Wilcoxon test with Bonferroni correction to examine differences between two means. Distributions are depicted as box plots, where the box represents the interquartile range (50% of the distribution) with the median highlighted, and whiskers highlight 1.5× interquartile range from the 25th and 75th percentile. A Chi-squared test was applied for percentage comparison. For all tests, the significance threshold was set at P<0.05 and is displayed as *P<0.05, ** for P<0.01, *** for P<0.005, **** for P<0.001.

We are grateful to E. Bazellières and A. Pasini for critical reading of the manuscript. We wish to thank P. A. Bidaud-Meynard, A. Pacquelet, J. Pécréaux and J. P. Tassan for stimulating discussion. We thank P. Walentek for providing the CP110-GFP plasmid. We thank X. Pinson, S. Dutertre, R. Flores-Flores and E. Castellani for their technical assistance with confocal microscopy. We thank F. Roguet and J. Maurais for Xenopus care. We thank all members of the Gene Expression and Development team (GED), in particular A. Mereau for her help on Western Blot. We thank J. P. Tassan and V. Thomé for sharing Xenopus expertise and/or reagents. We are grateful to R. Gibeaux and C. Callens for offering a safe harbor for Xenopus injections during IGDR building renovation. Light imaging was performed at the optical imaging and electron microscopy (PiCSL-FBI core facility) platforms of the Institute for Developmental Biology of Marseille (IBDM, France) and at the Microscopy Rennes imaging Center (MRiC, Biosit, Rennes). Electron microscopy experiments were performed at PiCSL-FBI facility. The PiCSL-FBI and MRiC, Biosit facilities are France Bio-Imaging infrastructures, supported by the French National Research Agency (ANR-10-INSB-04).

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261046.reviewer-comments.pdf