Prominins control ciliary length throughout the animal kingdom: New lessons from human prominin-1 and zebrafish prominin-3

Prominins (proms) are transmembrane glycoproteins conserved throughout the animal kingdom. They are associated with plasma membrane protrusions, such as primary cilia, as well as extracellular vesicles derived thereof. Primary cilia host numerous signaling pathways affected in diseases known as ciliopathies. Human PROM1 (CD133) is detected in both somatic and cancer stem cells and is also expressed in terminally differentiated epithelial and photoreceptor cells. Genetic mutations in the PROM1 gene result in retinal degeneration by impairing the proper formation of the outer segment of photoreceptors, a modified cilium. Here, we investigated the impact of proms on two distinct examples of ciliogenesis. First, we demonstrate that the overexpression of a dominant-negative mutant variant of human PROM1 (i.e. mutation Y819F/Y828F) significantly decreases ciliary length in Madin–Darby canine kidney cells. These results contrast strongly to the previously observed enhancing effect of WT PROM1 on ciliary length. Mechanistically, the mutation impeded the interaction of PROM1 with ADP-ribosylation factor–like protein 13B, a key regulator of ciliary length. Second, we observed that in vivo knockdown of prom3 in zebrafish alters the number and length of monocilia in the Kupffer's vesicle, resulting in molecular and anatomical defects in the left-right asymmetry. These distinct loss-of-function approaches in two biological systems reveal that prom proteins are critical for the integrity and function of cilia. Our data provide new insights into ciliogenesis and might be of particular interest for investigations of the etiologies of ciliopathies.

Prominins (proms) 6 are cholesterol-binding, pentaspan membrane glycoproteins conserved throughout the animal kingdom and expressed in numerous embryonic and adult epithelial cells (1)(2)(3)(4). In mammals, two prom proteins were identified (2). The first member of the family, PROM1 (CD133), has become a widely used surface antigenic marker for the identification and isolation of a broad range of somatic stem cell populations, and potentially cancer-initiating cells (reviewed in Ref. 5). In the visual system, PROM1 gene mutations are associated with various forms of retinal degeneration (6,7), which can be phenocopied in genetically modified mice carrying either a null allele or a mutated dominant form of human PROM1 (8,9). Murine models deficient in prom1 show a severe disorganization of photoreceptor outer segments, which are specialized sensory cilia (reviewed in Ref. 10). In zebrafish, the corresponding mammalian PROM1 gene is duplicated, and the co-orthologues are referred to as prom1a and prom1b (2,(11)(12)(13)(14). The knockdown of prom1b, but not prom1a, also led to photoreceptor degeneration (15). This observation was surprising, given that both prom1a/b are co-expressed in fish retina (13,16,17). Prom1 is found in rods and cones, where it is concentrated in the outer segments, particularly at the edges of open disks, irrespective of the species investigated (6,13,18) (reviewed in Refs. 7 and 19). These morphological and pathological data together with the presence of prom1 (or its paralogue prom2) in nonmotile and motile cilia found in various mammalian organ systems lay evidence of a conserved involvement of proms in the maintenance of ciliary structure and function (20 -22). Besides cilia, prom1 and prom2 are associated with various types of cellular protrusions, such as microvilli, filopodia, and lamellipodia, indicating a strong preference for highly curved membrane domains (1,2,23,24). The association of proms with small membrane vesicles that are budding from microvilli and primary cilia is in line with this preference (20,21,25).
Mutations in the ganglioside-binding site of PROM1 or variations in the level of membrane cholesterol influence its specific subcellular localization as well as the organization of membrane protrusions and/or the dynamics of the vesicles budding thereof (21,(26)(27)(28). The direct interaction of prom1 (or prom2) with cholesterol (21,26,29) and potentially other membrane lipids, such as monosialotetrahexosylganglioside (GM 1 ) (30, 31), might account for these peculiarities. These prom/lipid complexes might stimulate the interaction of prom1 with various membrane and/or cytoplasmic protein partners and hence modulate the architecture of the given protrusion. For instance, the binding of prom1 to protocadherin 21 organizes the nascent precursor of photoreceptive membranes at the base of the outer segment (9). The interaction of prom1 with the actin-related protein 2/3 (Arp2/3) complex, which mediates branching of actin networks, favors the formation of clusters of microvilli in epithelial cells and filopodia in nonepithelial cells (28). The latter observation might be associated with the implication of prom1 in cancer cell migration (32). The selective binding of prom1 to ADP-ribosylation factor-like protein 13B (Arl13b) or histone deacetylase 6 (HDAC6) can orchestrate the functionality and dynamics of primary cilia and consequently the activation of stem cells (33). The interactions of Prom1 with Arl13b/ HDAC6 appear to be dependent on cytoplasmic lysine 138 (numbered according to prom1 splice variant s2) (33,34), whereas binding to Arp2/3 complex is stimulated by phosphorylation of a C-terminal tyrosine residue located at position 819/ 828 (s1/s2 variant) (28,35). This post-translational modification also promotes the interaction of prom1 with the p85 subunit of phosphoinositide 3-kinase (PI3K), thereby activating the transformation of phosphatidylinositol 4,5-bisphosphate (PIP 2 ) into phosphatidylinositol 3,4,5-trisphosphate (PIP 3 ) at the inner leaflet of the plasma membrane (36), which plays a pivotal role in the organization of plasma membrane protrusions (28). For example, the expression of a mutant form of human PROM1 containing a single amino acid substitution in s1 and s2 splice variants (i.e. Tyr 819/828 3 Phe; Y819F or Y828F mutant, respectively) impaired the elongation of microvilli (28). Altogether, alterations in prom1, as observed in vivo in retinal pathologies or in vitro with various cell types, may influence the organization of its surrounding membranous microenvironment and its cross-talk with cytoskeleton regulators. As a consequence, the general structure of the plasmalemma and, hence, dynamic protrusions such as microvilli and cilia would be affected.
Here, we further dissect the role of proms in ciliary architecture by determining the impact of Y819/828F mutation in PROM1 splice variants s1 and s2 on the primary cilium of mammalian cells and the down-regulation of one of three prom paralogues in zebrafish (12,13). Zebrafish prom3 has attracted our attention due to its expression in the Kupffer's vesicle, an embryonic organ implicated in the left-right (L-R) axis determination via a cilium-based nodal flow mechanism (37). Being composed of monociliated cells, Kupffer's vesicle represents an excellent model for disentangling cilium-related questions (38). Our data revealed a direct implication of mammalian and nonmammalian proms for two separate paradigms of ciliogenesis.

Mutation of tyrosine 819/828 in human PROM1 reduces ciliary length
We recently demonstrated that the mutation Y819F or Y828F in human PROM1.s1 and PROM1.s2 splice variants, respectively, impairs tyrosine phosphorylation and the elongation of microvilli in epithelial Madin-Darby canine kidney (MDCK) cells (28,39). The prom1 splice variants differ solely by the inclusion of a 9-amino-acid-encoding exon in the N-terminal domain of prom1.s2 (40, 41). By using scanning EM (SEM), we evaluated whether the same mutation would affect the primary cilium. We observed that cilia were indeed shorter in cells expressing the Y819/828F mutant proteins ( Fig. 1A and Fig. S1), and some cells did not harbor a cilium (Fig. 1A, asterisk). In contrast, very long primary cilia with more than 10 m in length were regularly observed on WT PROM1-transfected cells ( Fig. 1B and Fig. S1). These observations prompted us to quantify ciliary length. WT cells and those expressing the mutant proteins were subjected to double immunofluorescence using anti-PROM1 and anti-acetylated ␣-tubulin (AcTub) antibodies. The latter visualizes the ciliary axoneme, indicating ciliary length. Samples were analyzed by confocal laser-scanning microscopy (CLSM). In agreement with SEM data, primary cilia were significantly shorter in cells expressing Y819/828F mutants compared with WT cells at 7 days postconfluence (dpc) (1.7 Ϯ 0.81 m (n ϭ 203) and 1.7 Ϯ 0.68 m (n ϭ 236) for PROM1.s1 Y819F and PROM1.s2 Y828F, respectively, versus 2.5 Ϯ 1.17 m (n ϭ 224) in control cells; p Ͻ 0.001, Mann-Whitney test) (Fig. 1C, left). Similar data were obtained at 14 dpc (Fig. 1C, right). These data differ from those previously obtained with the expression of WT PROM1, where an increase in ciliary length was observed for s1 (3.1 Ϯ 1.4 m (n ϭ 569)) and s2 (3.1 Ϯ 1.4 m (n ϭ 953)) variants ( Fig. 1C; 7 dpc; see also Ref. 33). The number of cells harboring no cilium was also increased upon expression of PROM1 Y819/828F mutants (Fig. 1D) (33).
Because the assembly and disassembly of primary cilia are tightly coordinated with the cell cycle, we evaluated the number of cells after 7 and 14 dpc. Surprisingly, the number of cells expressing PROM1 Y819/828F mutants was significantly increased compared with control cells or cells expressing WT PROM1.s1 or PROM1.s2 variant, suggesting an extensive proliferation at an early stage of polarization (Fig. S2). At 14 dpc, cell numbers were, however, similar in all cell lines ( Fig. S2) (data not shown), suggesting that cell division might not be the only cause accounting for the appearance of short cilia. These data suggest that mammalian prom1 positively influences the length of primary cilia. ). Arrow, primary cilium; dashed line, origin of cilium; *, cell without primary cilium; #, artifact of SEM preparation. For the immunofluorescence analysis (C and D), cells were double-immunolabeled for PROM1 and AcTub. The length of AcTub-labeled cilia was evaluated by CLSM and classified into three categories: Ͻ3, 3-5, and Ͼ5 m (C). The number of analyzed primary cilia is displayed in parentheses. Data (purple) obtained from cells expressing PROM1.s2 were taken from our recent study (33) and used for comparison. The percentage of total (blue) or PROM1 ϩ (black) cells with no cilium is indicated (D). The mean and S.D. (error bars) are presented. Scale bars, 500 nm (top panels) and 2 m (bottom panels) (A) and 1 m (B).

Mutation of tyrosine 819/828 in human PROM1 reduces its interaction with Arl13b and impacts Arl13b subcellular localization
We recently reported that the K138Q mutation impedes the interaction of PROM1 with Arl13b, a member of the Arl (ADPribosylation factor (Arf)-related) subfamily of the Ras superfamily of small GTPases that regulates ciliary length (42). This mutation results in the formation of shorter cilia (33). To find out whether the interaction of PROM1 with Arl13b also mediates the negative impact of the Y819/828F mutation on ciliary length, we immunoisolated PROM1 (WT and mutants) after cell solubilization using paramagnetic bead-conjugated AC133 mAb and determined its interaction with Arl13b by immunoblotting. We observed that the Y819/828F mutation interfered with PROM1-Arl13b interaction for both s1 and s2 splice variants (Fig. 2, A and B). Similar data were obtained with the reciprocal experiment (i.e. co-immunoisolation of PROM1 using Arl13b as bait) (data not shown). These data suggest that the impact of the Y819/828F mutation on primary cilium length involves PROM1-Arl13b interaction. To further dissect this issue, we investigated the subcellular localization of Arl13b within the ciliary compartment. As expected, the CLSM analyses of MDCK cells expressing PROM1.s2 revealed the presence of Arl13b along the ciliary axoneme (Fig. 2C). A partial co-localization with PROM1 was observed. The latter was nonuniformly distributed along the cilium as reported previously (20,33), which might indicate its association with specific membrane microdomains (30) (see "Discussion"). In contrast, the corresponding PROM1 mutant seemed to be uniformly expressed, like Arl13b, in short cilia (Fig. 2C). Unexpectedly, we observed an accumulation of Arl13b at the base of cilia (Fig. 2, C and D) together with a reduction at the level of axonemal microtubules in cells expressing PROM1 mutant (0.69 Ϯ 0.29 (n ϭ 6)) compared with those harboring WT PROM1 (1.55 Ϯ 0.44 (n ϭ 9); p Ͻ 0.01, Mann-Whitney test) as indicated by the Arl13b/AcTub immunoreactivity ratio (see also Fig. 2C, top right panels). Overall, these observations suggest that the mutation of the cytoplasmic tyrosine 819/828 in PROM1 causes some defects in the intraflagellar transport of Arl13b and its reduction in the ciliary axoneme and hence influences the ciliary length.

Expression and distribution of proms in zebrafish
Until now most of the experimental evidence linking proms with the structure and dynamics of cilia was derived from analysis of mammalian proteins. As prom paralogues are expressed redundantly in many tissues, very few models permit documenting their individual roles (2). Developing nonmammalian vertebrates offer effective in vivo functional assay systems for testing loss-of-function scenarios with well-defined readouts. We used zebrafish (Danio rerio) as an established vertebrate model to gain insights into the physiological function of a given nonredundantly expressed prom (2,13). Indeed, besides prom1a and prom1b, the zebrafish genome contains a third prom gene, named prom3 (Gene ID: 556596; chromosome 13). As is the case for other proms, several splice forms of prom3 are transcribed. We observed at least two distinct splice variants that differ by the inclusion of a facultative exon (exon B) in the 5Ј-UTR (Table S1).
Distribution analysis of the three zebrafish proms by wholemount in situ hybridization (ISH) revealed spatially and temporally restricted, but dynamically changing, expression patterns during embryogenesis. These differences are already .s1 or PROM1.s2 splice variants or the corresponding Y819/ 828F mutants were subjected either to immunoisolation (IS) using paramagnetic beads conjugated to mAb against PROM1 (AC133) or anti-fluorescein isothiocyanate (FITC) antibody as negative control (A and B) or processed for indirect immunofluorescence followed by CLSM (C and D). For immunoisolation, the input (PROM1, 1:10; Arl13b, 1:40) and immunoisolated materials were probed for PROM1 and Arl13b by immunoblotting (A). The arrowhead and asterisks indicate plasma membrane and endoplasmic reticulumassociated species of PROM1, respectively, whereas arrows show Arl13b. Molecular mass markers (kDa) are indicated (left). The ratio of Arl13b/PROM1 (plasma membrane species only) immunoreactivities was quantified (B). 3-6 independent experiments were performed. Data are presented as mean and S.D. (error bars). Individual values are shown. **, p Ͻ 0.01; ***, p Ͻ 0.001 (twotailed unpaired Student's t test). For immunofluorescence, cells were tripleimmunolabeled for PROM1 (green), Arl13b (red), and AcTub (white). Nuclei (nu) were counterstained with DAPI (blue). The three-dimensional view (x-z orientation) was built from 38 -45 x-y sections throughout the ciliary compartment (C, left panels) or x-y section (right top panels) and a composite of 5-10 x-y sections (right bottom panels) taken at the axoneme level (Ax, green bar) or at the base (red bar) of the primary cilium (C), respectively. The top view of three-dimensional reconstructions highlights Arl13 at the base of the cilium (D, dashed red line). Cells expressing either WT (left) or mutant (right) PROM1 are displayed. Green arrow, punctate PROM1 staining at the ciliary membrane of WT PROM1 ϩ cells. Asterisk, PROM1 associated with microvillar membrane. Scale bars, 1 m. Impact of prominins on the ciliary structure obvious at very early stages of development as, for instance, maternal deposition of prom1a transcripts could be detected as described previously (12). A detailed description of zebrafish prom expression domains starting at early blastula stage through somitogenesis stages (6-, 8-, and 16-somite stages) as well as in pharyngulas (at 22 h post-fertilization (hpf)) is presented in the supporting Results (Figs. S3 and S4).
The unique expression of prom3 in the transient embryonic structure called the Kupffer's vesicle (Fig. S3) caught our attention owing to its implication in the L-R axis determination via a cilium-based nodal flow mechanism. The fact that knocking down zebrafish Arl13b elicited phenotypes that are related to ciliary structures within Kupffer's vesicle (44) further prompted us to concentrate on zebrafish prom3, while using prom1a as a negative control, given that both prom1 paralogues are absent from this anatomical structure.

Depletion of zebrafish prom3 leads to left-right asymmetry defects
To address the biological role of prom3, an antisense depletion approach was undertaken using three independent morpholino-oligonucleotides (MOs). Two distinct strategies were employed. First, a translation blockage was utilized using two independent MOs (MO1 and MO2). To interfere with mRNA translation, the oligonucleotides were made complementary either to the translation initiation codon region of the message or the 5Ј-UTR (in a sequence stretch found in both splice variants; Fig. S5A). Second, the blockage of pre-mRNA splicing was applied using a third MO (MO3). It was designed to block pre-mRNA splicing allowing also for the direct evaluation of specificity by monitoring the impairment of the pre-mRNA processing (Fig. S5B). The impact of MO3 on mRNA splicing was verified by RT-PCR (Fig. S5, C and D). Following oligonucleotide injection, a gross morphological observation at 4 days postfertilization (dpf), revealed severe impairments in morphants. Indeed, the prom3 morphants induced by either MO1 (5 ng injected) or MO2 (10 ng) suffered from pericardial edema, and their body axis was frequently curved ventrally downward with a kink in the tail (Fig. 3, A and B). The body size was reduced (Fig. 3, A and B versus C). Similar observations were made with MO3 (10 ng) (Fig. S6). In line with the localization of prom3 transcripts in the intermediate mesoderm and in its successor structure, the pronephric duct (Fig. S4C), the rostral and middle portion of the pronephric duct appeared to be dilated compared with the control (standard control MO, 10 ng) (Fig. 3, E and G), yet no typical glomerular cyst formation was observed. In contrast, neither the control nor the antisense knockdown of prom1a resulted in any obvious gross morphological defects (Fig. 3, H-JЈ).
These observations prompted us to investigate the consequence of the depletion of prom3 at earlier stages of development (i.e. 30 and 40 hpf). The injection of MO3 resulted in an alteration of normal laterality (handedness). In contrast to control embryos where the normal leftward looping of the heart tube (situs solitus) was observed, the prom3 morphants had a randomized heart position as detected by ISH of cardiac myosin light chain 2 (cmlc2) (Fig. 3, K-P). They displayed either dextrocardia (situs inversus) (i.e. inversion of L-R asymmetry with respect to the median-sagittal plane as judged by left-right displacement of the developing heart tube) (Fig. 3, M and P; 8.5 and 12% of cases at 30 and 40 hpf, respectively; n ϭ 35 and 59) or an absence of laterality resulting in a median heart tube position (Fig. 3, L and O; 48.5 and 46% of cases at 30 and 40 hpf, respectively). Interestingly, the simultaneous application of one-fourth of two independent MOs blocking translation (i.e. MO1 (1.25 ng) and MO2 (2.5 ng)) resulted in a synergistic effect on fish morphology at 58 hpf leading to a severe phenotype similar to what has been observed upon injection of MO2 alone at an amount of 10 ng (Fig. 4, A-D). The combined injection of one-eighth of the single effective dose of each individual MO (MO1 ϩ MO2) elicited a laterality defect as observed with cmlc2 ISH (Fig. 4, E and F).
Next, we examined whether the randomized anatomical laterality of the morphants was evidenced by altered molecular asymmetries earlier in development. The expression of lefty-2, a member of the transforming growth factor-␤ superfamily, was analyzed as a typical example of asymmetrically expressed genes during the segmentation period (45,46). In contrast to control embryos, prom3 morphants displayed a randomized gene expression at 25 hpf (Fig. 5, A-D). Four configurations of lefty-2 localization were detected with different frequencies, depending on the nature and the amount of the MO used (Fig.  5E). Either right-sided or bilateral or even no expression was observed in addition to the normal left-sided appearance of this molecule. Thus, the antisense depletion of prom3 resulted in a L-R randomization of gene expression well before the manifestation of anatomical signs of handedness, suggesting that prom3 acts upstream of lefty-2 in the sequence of events leading to L-R determination in zebrafish.

Zebrafish prom3 regulates ciliary length in Kupffer's vesicle
What could explain the molecular and anatomical laterality defects of prom3 morphants? As indicated above, prom3 is, contrarily to prom1a and prom1b, selectively expressed in Kupffer's vesicle (Fig. S4, E-G). This transient embryonic structure of teleosts plays a pivotal role in the L-R axis determination. It is composed of around 50 -60 monociliated cells (47). The cilia, 3-5 m in length, have a major impact on the generation of a local microflow required for the establishment of correct laterality. Thus, Kupffer's vesicle serves as an integration focus between the early upstream epigenetic L-R information generated during cleavage stages by maternally deposited proton pumps and the well-conserved asymmetric Nodal/ Lefty/Pitx2 signaling cascade in the left lateral plate mesoderm (45). To gain insight into the role of prom3 in this developing structure, we applied prom3 MO2 and MO3. The samples were analyzed at the 8-somite stage by double immunolabeling using anti-AcTub antibody and anti-atypical protein kinase C (aPKC) antibody. (Despite the fact that the generation of antisera against prom3 was attempted, no specific antibody could be raised that would work on immunohistochemical staining. This might be explained by a high level of N-glycosylation in prom3 (see below).) Although the formation of Kupffer's vesicle itself was not impaired, we noticed that AcTub-labeled ciliary structures were severely affected (Fig. 6, A and B). Both the number of cilia per Kupffer's vesicle (Fig. 6C) and their length Impact of prominins on the ciliary structure ( Fig. 6D) were significantly reduced upon prom3 MO application. The antisense depletion altered neither the polarity of epithelial cells in Kupffer's vesicle nor their nuclear distribution as observed by aPKC and 4Ј,6-diamidino-2-phenylindole (DAPI) staining, respectively (Fig. 6, A and B). This strongly argues for a direct implication of prom3 in ciliogenesis rather than an indirect effect on cell polarity and proliferation in general. Altogether, these data suggest that compromised ciliary morphology was responsible for the observed laterality defects. The exact mechanisms by which zebrafish prom3 impacts on cilia will need to be further clarified once stable loss-of-function mutants and specific antibody tools against this protein are available. However, its analysis, through the molecular and morphological/anatomical readouts of the effective bioassay system applied, delivers strong evidence for the involvement of proms in ciliary biology in metazoan organisms, irrespective of their phylogenetic position during evolution.

The localization of zebrafish prom3 mimics that of mammalian prom2 and co-localizes with Arl13b within the ciliary compartment
Mammalian prom1 is selectively associated with the apical membrane of polarized epithelial cells, whereas its paralogue prom2 is distributed in a nonpolarized fashion (21,48). We therefore investigated the localization of zebrafish prom3 upon ectopic expression in MDCK cells. To that end, we generated expression constructs of zebrafish prom-3 either tagged with hemagglutinin (HA) epitope or as a fusion protein consisting of

Impact of prominins on the ciliary structure
the enhanced GFP (EGFP) linked to the cytoplasmic C-terminal domain of prom3. In previous studies, we showed that GFP does not interfere with the proper intracellular trafficking and targeting of mammalian prom1 and prom2 to plasma membrane protrusions (20,21). Upon high-speed centrifugation of transfected MDCK cell homogenates, prom3-HA/GFP was solely recovered in the membrane fraction, indicating that the fusion proteins were, as expected, tightly associated with the membrane (Fig. S7, A and B). Peptide-N-glycosidase F treatment led to the shift of their apparent molecular mass from ϳ160 to 90 kDa in the case of prom3-HA or from ϳ180 to 110 kDa for prom3-GFP, consistent with the presence of 20 potential N-glycosylation sites (Fig. S7, A and B) (data not shown).
Interestingly, CLSM analysis of cells growing as a polarized cell monolayer revealed that prom3-HA was expressed at both apical and basolateral membranes (Fig. 7A). Similar data were observed with prom3-GFP (Fig. S7C). Within the apical domain, prom3-HA-and prom3-GFP-tagged versions were observed at the primary cilium highlighted by AcTub staining as well as in microvilli (Fig. 7A and Fig. S7C). The subcellular localization of prom3 is reminiscent of that of mammalian prom2 (21). We also observed a co-localization of zebrafish prom3-HA with Arl13b in the ciliary compartment (Fig. 7B, arrow) as demonstrated for mammalian prom1, suggesting that prom-Arl13b interaction may be conserved among prom molecules and across species. As with PROM1 (Fig. 2C), the stain-ing of prom3-HA is often asymmetrically distributed in the ciliary membrane. Prom3-HA can be either concentrated on one side of the ciliary membrane throughout the entire axoneme ( Fig. 7B and Video S1) or appear in a patched fashion, including its concentration at the ciliary tip (Video S2). The ciliary length (i.e. 3.1 Ϯ 1.3 m; n ϭ 85) of cells expressing prom3-HA is similar to that of cells positive for WT PROM1 (see above).
Although the information concerning the molecular trafficking of membrane proteins in and out the ciliary compartment is still incomplete and sometimes conflicting (49,50), it has been shown that the proper targeting of some cilium-associated transmembrane proteins (e.g. seven-transmembrane receptor-1, rhodopsin, and smoothened) depends on two consecutive conserved hydrophobic and basic residues (51). Such residues (tyrosine and arginine) are present in the cytoplasmic C-terminal domain of mammalian prom1 and zebrafish prom3 in the vicinity of the last transmembrane segment (Fig. S7D). Mutation of these two residues to alanine (YR3 AA) in prom3-HA, however, did not affect its proper targeting. Just like the WT version, the mutated prom3 was found associated with primary cilia (Fig. S7E). The ciliary targeting motif of proms thus remains to be discovered.

Discussion
PROM1 has received widespread attention due to its clinical translation relevance as a surface marker of stem cells (52) and

Impact of prominins on the ciliary structure
cancer-initiating cells, thus representing a target for cancer eradication (5). Besides, various retinal degenerative diseases are associated with mutations in the PROM1 gene, qualifying it as a candidate for gene replacement therapy. Despite that, the function of prom1 and other members of the prom's family remained hypothetical for a long time (24). Recently, two studies have clarified the role of prom1 in the architecture of plasma membrane protrusions, such as microvilli, filopodia, and primary cilia (28,33). Prom1 can modulate their structures and functions by interacting with regulators engaged in their biogenesis and/or dynamics.
The dual interaction of prom1 with Arl13b or HDAC6 could influence the dynamics of ciliary length either in a positive or negative manner, respectively (53)(54)(55)(56)(57). As demonstrated with dental epithelial stem cells, such interactions can regulate temporal and spatial activation of stem cells and perhaps cancerinitiating cells (33). The implication of prom1 in ciliogenesis might have consequences in numerous cilium-based diseases known as ciliopathies (58). In the retina, its ablation prevents the proper formation of photoreceptor outer segments that represent modified sensory cilia. The deletion of Arl13b in adult mice via tamoxifen-induced Cre/loxP recombination caused similar retinal phenotypes (59,60). Some of the phenotypes of zebrafish prom3 morphants are strikingly reminiscent of those of a group of mutants with compromised pronephric function (61). Mutations in genes responsible for the assembly/ motility of cilia frequently result in edematic phenotypes, occasionally with cystic distension of the glomerulus or the pronephric tubule/duct (44,53,62). The link between proms and Arl13b regarding the regulation of cilia is strengthened by the decreased length of monocilia associated with the Kupffer's vesicle in prom3 knockdown. As a result, molecular and anatomical L-R asymmetry defects, a phenotype similar to that obtained with both zebrafish Arl13b mutants and morphants (44), were observed. These data suggest that across species,

Impact of prominins on the ciliary structure
prom-Arl13b interaction could play a key role in many cellular systems, which involve motile or nonmotile cilia, such as primary cilia. The latter host various signal transducers (e.g. sonic hedgehog and Wnt pathways) responsible for proper organ development (63). Our data may also explain the limited phenotypic and physiological defects after the ablation of a given prom paralogue in tissues expressing two of them simultaneously because a compensatory mechanism could occur (2,8,43).
Mechanistically, the mutation of the residue Tyr-819/828 located at the C terminus of PROM1 is particularly instructive because the reduced Arl13b binding suggests that tyrosine phosphorylation may be involved in regulating such interaction, similarly to what has been reported for PROM1 binding to PI3K-p85 subunit and Arp2/3 complex (28,36). Whether this modification would interfere with HDAC6 interaction that is mediated, as for Arl13b, by a lysine residue located in the first small intracellular loop of prom1 remains to be evaluated (33,34). Similarly, it remains to be determined whether the interaction of prom1 with the Arp2/3 complex, which is also dependent on its tyrosine 819/828 phosphorylation (28), could influence the ciliary structure (64). Investigating this interaction at the base of the cilium and/or its tip during the shedding of prom1 ϩ membrane vesicles (20, 28, 33, 65) will be of particular interest because it has been shown that actin polymerization in these subcompartments depends on the Arp2/3 complex (66 -68). In photoreceptors, the prom1-Arp2/3 complex interaction could stimulate the growth of membrane evaginations from the connecting cilium at the base of the outer segment, thus generating a new photoreceptor disc (69).
A further exciting question is how proms can influence microvilli and cilia, thus examining the connection of cholesterol-binding membrane proteins with actin-and tubulin-based cytoskeleton structures. Other constituents indirectly implicated in such interaction might be gangliosides, whose presence in the ciliary membrane was demonstrated (30). A recent study on the structural impact of mutations in the ganglioside-binding site of PROM1 revealed their influence on the interaction of prom1 with PI3K in a tyrosine phosphorylation-dependent manner (28). Therefore, the morphological disturbance of primary cilia observed upon overexpression of mutants Y819/828F may also reflect a disrupted interaction of proms with cholesterol-and/or ganglioside-rich membrane microdomains (21,26,30,31), as suggested for the altered microvillar structure (28). The local activity of PI3K that converts PIP 2 into PIP 3 in the inner leaflet of the plasma membrane can regulate the interaction of the lipid bilayer with the underlying structural core of cellular protrusions and/or the trans- Impact of prominins on the ciliary structure port of membrane components therein (28,(70)(71)(72)(73)(74). The accumulation of Arl13b at the ciliary base of cells expressing the PROM1 mutant may result from a defect in the aforementioned mechanism. It is known that Arl13b acts as a guanine-nucleotide-exchange factor for Arl3, and both proteins coordinate the intraflagellar transport (75)(76)(77)(78). Thus, prom-containing membrane microdomains may act as building units that elaborate and/or maintain the proper composition of microvillar and ciliary membranes and thus influence interactions with cytoplasmic regulatory proteins. They could also affect membrane fluidity by creating distinct liquid phases and, hence, influence the diffusion of their constituents (74). The uneven distribution of PROM1 (or prom3) observed in cilia of WT PROM1 (or prom3-HA) ϩ cells compared with its uniform spread in mutated PROM1 ϩ cells suggests the disruption of these membrane microdomains (Fig. 2C). A more in-depth study should dissect the impact of proms on each subdomain of the cilium, including the basal body and the transition zone as well as the molecular players involved in the intraflagellar transport (79). The interaction of proms with other membrane (e.g. protocadherin 21) or secreted (spacemaker) proteins can also organize the extracellular space between cellular protrusions (9,80,81). Altogether, these distinct interactions could explain the implication of proms in various cellular processes such as proliferation, differentiation, autophagy, migration, and carcinogenesis.
Interestingly, prom1 also plays a role in autophagy, where a phosphorylation-dependent interaction with ␥-aminobutyric acid receptor-associated protein (GABARAP), an initiator of autophagy, leads to the suppression of GABARAP-mediated ULK1 kinase activation and the subsequent initiation of autophagy (82,83). The prom1 phosphorylation could link ciliogenesis and autophagy (84, 85), which in turn may regulate the cell proliferation versus differentiation (86). This model is thrilling because cell differentiation has been linked to a reduction or loss of prom1 (52,87,88). A relationship between cell proliferation, cilia, and prom3 was suggested in zebrafish hair cells (89).
The apico-basolateral distribution of zebrafish prom3 is reminiscent of that of mammalian prom2 (21,48), which contrasts with the exclusive apical localization of prom1 (90). However, the cytoplasmic C-terminal domain of prom3 is distinct from that of prom2, which is known to be almost perfectly conserved among mammals (2). Also, genomic analysis revealed that zebrafish prom3 is not a true orthologue of the mammalian prom2 gene. Although both genes display a higher degree of compactness compared with prom1 (Table  S1), they do not show a conserved synteny (11). Still, prom3 might have evolved from the same ancestral locus as prom2 generated during the first round of vertebrate genome duplication (11,12,91).
In conclusion, our data revealed a participation of mammalian and nonmammalian prom paralogues in ciliary architecture and function beyond photoreceptors. With that, we contribute to the ongoing dissection of molecular and cellular networks underlying not only ciliopathies, but also developmental pathways linked to the integrity and dynamics of cilia.

Animals
Animal experiments were performed in strict accordance with German Animal Welfare legislation, and all procedures were performed in compliance with the UK Animals Act (Scientific Procedures, 1986) approved by the Institutional Animal Welfare Officer (Tierschutzbeauftragter). All efforts were made to minimize the suffering of animals, and those used for organ removal were sacrificed according to the Schedule 1 methods of humane killing. Experiments were carried out at the Max-Planck-Institute of Molecular Cell Biology and Genetics (MPI-CBG, Dresden, Germany), which holds licenses (Az.:74-9165.40-9-2000-1 and 74-9168.25-9-2001-1) for keeping and breeding of laboratory animals (Erteilung und Erlaubnis für das Züchten und Halten von Wirbeltieren zu wissenschaftlichen Zwecken) and collecting organs and tissues (Anzeigepflichtige Tierversuche), both issued by Regierungspräsidium Dresden, Saxony. No genetically modified animals were used or generated in the current study.

Fish embryos
Fish were raised and kept under standard conditions at 27°C at the fish facility of the MPI-CBG, Dresden. Freshly laid fertilized eggs were harvested into E3 embryo medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl 2 , 0.33 mM MgSO 4 , 0.00001% (w/v) methylene blue). Embryos were raised at 28.5°C and staged according to Kimmel et al. (92).

Cloning of zebrafish prom3
By searching the fourth assembly of the zebrafish genome of the Wellcome Trust Sanger Institute using the murine prom1 amino acid sequence as a query, a genomic contig (Zv4_scaffold1778) encoding for a previously unannotated, partial prom molecule was identified. With the sequence information gained, a cDNA fragment of this prom homologue, designated as prom3, was amplified by PCR from a 72-hpf whole-larvae cDNA pool using the oligonucleotides 5Ј-GACT-GATCCTGAGCTGTCTG-3Ј (jj-499) and 5Ј-ACAATGACT-TACTCATACACATCTG-3Ј (jj-500) as forward and reverse primers, respectively. Total RNA was used to prepare a first strand cDNA template with Superscript II reverse transcriptase (Invitrogen). As a result, a 1.2-kb cDNA fragment was obtained, subcloned into pCRII-TOPO TA vector (Invitrogen), and sequenced. The missing extremities were amplified by 5Ј-and 3Ј-RACEs using a SMART RACE cDNA synthesis kit (catalog no. 634914, Clontech) according to the manufacturer's instructions. The gene-specific oligonucleotides used for the first and second round of amplification of 5Ј-RACE were 5Ј-GATGAA-GTGTAGTCCACAAGGGTGCGTTCTG-3Ј (jj-593) and 5Ј-ACTAACATGAAGATCCAGGAGATCAG-3Ј (jj-594). Those for 3Ј-RACE were 5Ј-GACATCACATCTAACATAATGCCT-TTGCTGCTG-3Ј (jj-590) and 5Ј-GTGAGAGCTGAGAGCA-AGGTGTTCGTAGACTG-3Ј (jj-591). PCR products were extracted from the agarose gel (Qiagen GmBH, Helden, Germany) and subcloned into pCRII-TOPO TA vector. The identity of the inserts was determined by sequencing. Finally, the entire zebrafish prom3 coding sequence was amplified with Impact of prominins on the ciliary structure Accuprime Taq High Fidelity DNA polymerase (Invitrogen) using the oligonucleotides 5Ј-CAGAGGGAGACATTGGAG-CGGAC-3Ј (jj-616) and 5Ј-GTGATTTCGGATACAGCCAA-GTG-3Ј (jj-621) as forward and reverse primers, respectively. PCR products were subcloned into pCRII-TOPO Dual Promoter vector (Invitrogen). Under these conditions, we have identified two distinct splice variants of zebrafish prom3 that differ in their 5Ј-UTR. Their sequences were deposited in the GenBank TM database (accession numbers DQ341400 and HQ698925). Sequencing reactions were performed using an Applied Biosystems 3730 Genetic Analyzer at the MPI-CBG sequencing facility.
The nucleotide and protein sequences were analyzed by the HUSAR compilation of sequence analysis tools accessible online at the German Cancer Research Center (Deutsches Krebsforschungszentrum, Heidelberg, Germany) or European Molecular Biology Laboratory-European Bioinformatics Institutes (EMBL-EBI). The genomic organization of the Prom3 gene is presented in Table S1.

Plasmid construction
Eukaryotic expression plasmids Zf PB2-1 and Zf PB1-3 containing full-length coding sequence of zebrafish prom3 (DQ341400) fused in-frame at its C terminus to either EGFP or HA epitope tag nucleotide sequence, respectively, were constructed by selective PCR amplification. The 5Ј primer was 5Ј-GTCAGATCCGCTAGCAACATGAAGTCCAGAACGAAA-GCAG-3Ј (jj-685N), whereas the 3Ј ones were 5Ј-GACTGC-ACTCGAGTTGTTTCGCATTTTCCACGAACTC-3Ј (jj-687) and 5Ј-GACTGCACTCGAGTCAGGCGTAGTCGGGG-ACGTCGTAGGGGTATTGTTTCGCATTTTCCACGAA-CTC-3Ј (jj-686) for the EGFP and HA tag, respectively. In the latter oligonucleotide, TGA stop codon (boldface type) and coding sequence of HA tag (27 nucleotides double underlined) are indicated. The 5Ј primer contained an NheI restriction site, whereas the 3Ј primers carried an XhoI recognition site (underlined), allowing for directional subcloning into pEGFP-N1 vector (BD Bioscience Clontech) digested with the corresponding restriction enzymes. In all cases, prom sequences within the expression plasmids were sequenced to confirm the absence of any reading mistakes introduced by the DNA polymerase and/or subcloning steps.

Cell culture and transfection
MDCK cells (strain II, American Type Culture Collection, catalog no. ATCC CCL-34) were cultured as described previously (28). They were maintained in a humidified incubator at 37°C under a 5% CO 2 atmosphere. MDCK cells were stably transfected with either Zf PB2-1 (prom3-EGFP), Zf PB1-3 (prom3-HA), or Zf PD2-3 (prom3 YR-AA -HA) plasmids using Amaxa Nucleofector technology (Amaxa, Cologne, Germany). Two days after transfection, cells were selected by introducing 500 -600 g/ml Geneticin TM (G418 Sulfate, Thermo Fisher Scientific) into the culture medium and further expanded. After selection, the concentration of the selective agent was reduced to 250 g/ml. MDCK cells expressing human PROM1.s1 and PROM1.s2 and the corresponding Y819/828F mutants as well as those containing solely control vector were recently described (28). Note that under these conditions, Geneticin did not affect the ciliary length, as demonstrated with transfected cells containing expression vector alone (data not shown), ruling out a direct effect of the drug on polyphosphoinositides (93,94). To enhance the expression of the recombinant protein, 10 mM sodium butyrate was added to the culture medium for 16 h before certain experiments as described (90). Cells were used 7 and 14 dpc.

Probes for in situ hybridization
Antisense complementary DIG-labeled RNA (cRNA) probes were generated using a DIG labeling mix (Roche Molecular Biochemicals). To synthesize cRNA probes, the entire cDNA inserts containing the coding sequence of either zebrafish prom1a (HQ386793) or prom3 (DQ341400) were used. For zebrafish prom1b (AF373869), a 1.365-kb cDNA fragment (nucleotides 234 -1,599) was used. The average homology of the zebrafish prom1 co-paralogues along the stretch used for ISH is ϳ60% (13). Combined with the stringent hybridization and post-hybridization conditions, this precludes any crosshybridization. To generate templates for zebrafish lefty2 (BC097092) and cmlc2 (BC076232) probes, cDNA clones were obtained from the German Resource Center and Primary Database (German Science Center for Genome Research, RZPD, Berlin).

Microinjection of morpholino-oligonucleotides
Standard control and custom MOs were purchased from GeneTools. MOs made complementary to the AUG translational initiation site, the 5Ј-UTR, and the splice site of zebrafish prom3 were 5Ј-CTGCTTTGCTTCTGGACTTCATGTT-3Ј (MO1), 5Ј-GAGCGTGTTCCAGTTAATTCAGGTC-3Ј (MO2), and 5-TGGTGCTCACAGAATAAGTACAGTG-3Ј (MO3), respectively. MOs directed against the AUG initiation coding site Impact of prominins on the ciliary structure and 5Ј-UTR of zebrafish prom1a were 5Ј-AGCATGATTA-GTTCTAGTGTTGAAG-3Ј (MO1) and 5Ј-GTTCAGTAAG-ACACAATCACTGACA-3Ј (MO2), respectively. Control MO was 5Ј-CCTCTTACCTCAGTTACAATTTATA-3Ј. MOs were dissolved in H 2 O at a concentration of 40 g/l (stock solution) and further diluted (5 ng/nl) for injection in Danieau buffer containing 0.2% phenol red as tracking dye. They were injected at the one-cell stage into the yolk using a PV830 pneumatic PicoPump (WPI) in a maximum volume of 2 nl. At a given period of the developmental stage, zebrafish larvae were either observed in a native, unfixed state or further processed as described below.

PCR detection of nuclear pre-mRNA processing
Control MO and prom3-MO3-injected embryos were manually dechorionated with fine-tip watchmaker's forcipes and snap-frozen at the 9-somite stage. After total RNA isolation and reverse transcription using Superscript II reverse transcriptase (Invitrogen), PCR amplification was performed using oligonucleotides 5Ј-GATCCTCTACATTGTCCTGATG-3Ј (jj-719) and 5Ј-GACATCATCTAAGATGCCTG-3Ј (jj-720) as forward and reverse primers, respectively. Cycling parameters (denaturation at 94°C for 30 s, annealing at 50°C for 45 s, extension at 72°C for 1 min) were used for 40 cycles. To exclude amplification of genomic DNA that might be present in total RNA preparations, DNase digestion of the RNA sample was performed (10 min, room temperature) followed by a clean-up step (RNeasy kit, Qiagen) before cDNA synthesis.

Histological sections
Embryos were fixed in 4% paraformaldehyde (PFA), washed three times with PBS (pH 7.4), dehydrated, and embedded in paraffin. Serial sections were cut at 4 m with a rotary microtome. Slides were deparaffinized, stained with hematoxylin-eosin, and coverslipped with xylene-based mounting medium.

Whole-mount in situ
Embryos were PFA-fixed at the required stage of development. After dechorionation, embryos were dehydrated in methanol and stored at Ϫ20°C until use. Whole-mount ISH was performed as described (95). Stained embryos were postfixed in 4% PFA and observed using either an Olympus SZX12 microscope or a Zeiss Axioscope. Images were captured without glycerol clearing.

Whole-mount immunohistochemistry
Embryos were fixed in 4% PFA at the required stage of development. After dechorionation, embryos were washed three times for 10 min in PBST (PBS ϩ 0.1% Tween 20) and subsequently incubated with 0.005% SDS in 0.2% gelatin-PBS for 30 min. Embryos were then rinsed once with 0.2% gelatin-PBS followed by three periods of incubation in 0.2% saponin-PBS for 30 min each and one period of incubation in blocking buffer I (5% normal goat serum and 2% BSA diluted with 0.2% saponin-PBS) for 1 h at room temperature. Afterward, the samples were incubated simultaneously with anti-AcTub mouse mAb (1:800; clone 6-11B-1, catalog no. T6793, Sigma-Aldrich, Darmstadt, Germany) and anti-aPKC rabbit polyclonal antibody (1:200; catalog no. sc-216, Santa Cruz Biotechnology, Inc.) diluted in blocking buffer I overnight at 4°C. Primary antibodies were detected using goat anti-mouse Alexa Fluor546 -or anti-rabbit Alexa Fluor488 -conjugated secondary antibodies, respectively (1:750; Thermo Fisher Scientific) diluted in blocking buffer I, respectively. Embryos were counterstained with DAPI (1 g/ml; Thermo Fisher Scientific) to facilitate the identification of the Kupffer's vesicle according to its typical nuclear configuration. After washing with PBS, the caudal halves of the embryos were dissected and freed from yolk with fine tungsten needles. Samples were subsequently flat-mounted with Mowiol 4.88 (Calbiochem). Images were captured using a Zeiss 780 confocal laser-scanning microscope (Jena, Germany). Composite images were prepared from the digital data files using Fiji (96) and Adobe Photoshop and Illustrator software (San Jose, CA).

Immunoisolation
WT MDCK cells or those expressing human PROM1 variants were lysed in solubilization buffer (0.2% Triton X-100, 150 mM NaCl, 20 mM Tris/HCl, pH 7.5) containing complete TM protease inhibitor mixture (Roche Diagnostics GmbH, Manheim, Germany), 2 mM sodium orthovanadate and 10 mM sodium fluoride (both from New England Biolabs Inc., Ipswich, MA), and 10 mM ␤-glycerophosphate (Sigma-Aldrich) on ice for 30 min. After centrifugation (10 min, 16,000 ϫ g, 4°C), an aliquot of detergent lysates was removed as an input sample, and remaining materials were divided into two fractions. They were incubated with either 65 l of mAb AC133-MicroBeads, or 130 l of anti-FITC-MicroBeads as negative control, for 1.5 h at 4°C prior to magnetic separation using MS columns (all from Miltenyi Biotec, Bergisch Gladbach, Germany). Materials retained in columns were washed five times with 1 ml of icecold PBS containing 2 mM EDTA and 0.1% Triton X-100 and then eluted with preheated (95°C) Laemmli sample buffer according to the manufacturer's instructions. Input and bound fractions were analyzed by immunoblotting for PROM1 and Arl13b (see below).
Alternatively, Arl13b was immunoisolated. After precleaning the detergent cell lysates with 100 l of MACS Protein A MicroBeads (Miltenyi Biotec) for 30 min at 4°C, followed by the magnetic separation, lysates were divided into two fractions. One was incubated with 4.5 l of rabbit antiserum against Arl13b (catalog no. 17711-1-AP, Proteintech, Manchester, UK), and the other was incubated without as a negative control, for 20 min at 4°C. Afterward, 100 l of MACS Protein A MicroBeads was added for 1 h at 4°C. Bound fractions were isolated with a magnetic column and processed as described above.

Membrane preparation
MDCK cells expressing either zebrafish prom3-HA, prom3 YR-AA -HA, or prom3-EGFP were scraped with a rubber policeman in ice-cold PBS and collected by centrifugation at 1,200 ϫ g for 5 min at 4°C. Cell pellets were homogenized with a glass-Teflon pestle in sucrose buffer (300 mM sucrose, 5 mM EDTA, 10 mM Hepes-KOH, pH 7.5, containing Complete Protease Inhibitor (Roche Molecular Biochemicals) at 4°C and Impact of prominins on the ciliary structure subsequently incubated for 30 min on ice. Lysates were centrifuged at 1,000 ϫ g for 10 min at 4°C, and the resulting supernatants were subjected to ultracentrifugation at 60,000 ϫ g for 60 min at 4°C. The pellets containing membrane fractions were resuspended in 1% SDS, 20 mM sodium phosphate and boiled for 2 min at 95°C, whereas cytosol fractions were recovered in supernatants. Membrane fractions were subjected to endoglycosidase treatment (see below) prior to their analysis by SDS-PAGE and immunoblotting.

Endoglycosidase digestion and immunoblotting
MDCK membrane extracts were incubated in the absence or presence of 1 unit of peptide-N-glycosidase F overnight at 37°C according to the manufacturer's instructions (Roche Diagnostics GmbH).

MDCK cells expressing human PROM1
Cells were rinsed with Ca/Mg buffer (PBS containing 1 mM CaCl 2 and 0.5 mM MgCl 2 ) and incubated in blocking buffer III (Ca/Mg buffer containing 0.2% gelatin) for 30 min at 4°C. For cell surface staining of PROM1, cells were incubated with mouse mAb CD133/1 (1:50; Miltenyi Biotec) diluted in blocking buffer III for 1 h at 4°C. After washing with PBS, cells were fixed in 4% PFA in PBS for 30 min at room temperature and then incubated for 10 min in PBS containing 50 mM NH 4 Cl. Fixed cells were permeabilized and blocked with 0.2% saponin in 0.2% gelatin/PBS for 30 min at room temperature. To visualize the primary cilium, cells were labeled with mouse monoclonal AcTub antibody (1:1000; clone 6-11B-1, Sigma-Aldrich) for 30 min in permeabilization buffer, followed by Alexa Fluor546 (or 647)-conjugated goat anti-mouse IgG 2b (1:600; Thermo Fisher Scientific), whereas Alexa Fluor488-conjugated goat anti-mouse IgG 1 antibody (1:500; Thermo Fisher Scientific) was used to detect anti-PROM1 antibody as described (33). In some experiments, Arl13b was detected with the rabbit antiserum against Arl13b and Alexa Fluor555conjugated donkey anti-rabbit IgG (HϩL) (1:500; Thermo Fisher Scientific). Nuclei were visualized with DAPI (1 g/ml). Finally, cells were washed three times in PBS containing 0.2% gelatin, three times in PBS, and twice in H 2 O and mounted in Mowiol 4.88. For the measurements of the ciliary length, cells were not mounted but were directly observed with a Zeiss LSM 700 or LSM 780 confocal laser-scanning microscope (Jena, Germany).

MDCK cells expressing zebrafish prom3
Cells were washed with PBS and then fixed and permeabilized as described above. To evaluate the distribution of the transgenes, zebrafish prom3-HA-or prom3 YR-AA -HAexpressing cells were labeled with a rat monoclonal HA tag antibody (1 g/ml) followed by the goat anti-rat IgG (HϩL) conjugated either to Cy2 TM (1:600; Jackson ImmunoResearch Inc.) or Alexa Fluor488 (1:500; Thermo Fisher Scientific), whereas in prom3-EGFP-transfected cells, the green fluorescence was monitored. The detection of AcTub and Arl13b was performed as described above. After mounting, cells were observed with a Zeiss LSM 780 or 510 Meta confocal laserscanning microscope. The microscope settings were such that multipliers were within their linear range. The images shown were prepared from data files of LSM 5 LIVE (Göttingen, Germany) using Volocity 3D Analysis software (PerkinElmer Life Sciences), Fiji, and Arivis Vision 4D software (Arivis AG) as well as Adobe Photoshop and Illustrator software.

Measurements
Primary cilia of MDCK cells were quantified using Huygens Professional software (Scientific Volume Imaging, Hilversum, The Netherlands) as described (33). After deconvolution of Z-stack images, the advanced object analyzer was used to obtain ciliary length. Fluorescence intensities of Arl13b and AcTub in the axoneme were quantified with Fiji. The number and length of cilia in Kupffer's vesicles were evaluated using the segmented line tool of the Fiji software on maximum projection of Z-stack images captured from flat-mounted tissues with a Zeiss LSM 780 confocal laser-scanning microscope.

Scanning EM
MDCK cells were prepared for SEM analysis as described (21). Briefly, cells grown on fibronectin-coated coverslips were fixed in 2% glutaraldehyde overnight at 4°C. After dehydration in an acetone gradient (25-100%), cells were critical pointdried in a CO 2 system (Critical Point Dryer CPD 300, Leica). Samples were mounted on aluminum stubs, coated with gold in a sputter coater (SCD 050, BAL-TEC GmbH), and examined with a scanning electron microscope (JSM-7500F, JEOL).

Statistical analysis
Statistical analyses were performed using GraphPad Prism version 6.00 for Mac (GraphPad Software, La Jolla, CA) or

Impact of prominins on the ciliary structure
Microsoft Excel software. Two-tailed unpaired Student's t test was used for the analysis of the nuclei numbers observed in MDCK monolayer for a given surface area, Arl13b/PROM1 ratio, cilium number, and their length in Kupffer's vesicle, whereas a Mann-Whitney test was applied for the length of cilium in MDCK cells and Arl13b/AcTub ratio. Observed differences were regarded as significant if the calculated p values were Յ0.05. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.

Data availability
All data are contained within the article and supporting materials. The zebrafish prom3 sequences were deposited in the public GenBank TM database under accession numbers DQ341400 and HQ698925.