The Cdc14B phosphatase contributes to ciliogenesis in zebrafish

Regulation of the cell cycle during embryonic development is crucial to the morphogenesis of tissues and organs of proper shape and size. Key players in this tight regulation are the Cyclin-dependent kinases (Cdks), the oscillating activity of which mediates an orderly progression through the cell cycle (Morgan, 1997). Cdc14 phosphatases reverse Cdk1-dependent phosphorylation events in diverse organisms and therefore contribute to proper mitotic exit and cytokinesis. Like many mitotic regulators, Cdc14 phosphatases localize dynamically through the cell cycle to diverse cellular structures, where they presumably act on various substrates (Trinkle-Mulcahy and Lamond, 2006).

Cdc14 was first discovered in Saccharomyces cerevisiae, where it is essential for mitotic exit (Stegmeier and Amon, 2004). Cdc14 has been conserved throughout evolution and performs orthologous functions in other organisms (Clifford et al., 2008; Stegmeier and Amon, 2004). However, Cdc14 enzymes are not essential for cell division in Schizosaccharomyces pombe (Cueille et al., 2001; Trautmann et al., 2001) or Caenorhabditis elegans (Saito et al., 2004), indicating that redundant phosphatases must be involved in mitotic exit in these organisms.

In vertebrates, Cdc14 is encoded by at least two genes, CDC14A and CDC14B. Studies in cell culture and Xenopus laevis and mouse oocytes have implicated CDC14A in mitosis, cytokinesis (Bembenek and Yu, 2001; Kaiser et al., 2004; Kaiser et al., 2002; Krasinska et al., 2007; Mailand et al., 2002), meiosis (Schindler and Schultz, 2009b) and DNA repair (Mocciaro et al., 2010), whereas CDC14B is reported to act in G1-phase length regulation (Rodier et al., 2008), centriole duplication (Wu et al., 2008), spindle stability (Cho et al., 2005), meiosis (Schindler and Schultz, 2009a), DNA damage response or repair (Bassermann et al., 2008; Mocciaro et al., 2010) and activation of the zygotic genome (Buffone et al., 2009). Interestingly, CDC14B is dispensable for cell proliferation in human cell culture (Berdougo et al., 2008). Taken together, our knowledge about CDC14 phosphatase function in vertebrates, particularly at the organismal level, remains limited.

To investigate CDC14 function during vertebrate development, we used zebrafish. The zebrafish genome contains three cdc14 genes that we found to be ubiquitously expressed throughout embryogenesis. Two of the genes encode Cdc14A paralogs, whereas the third encodes Cdc14B. Using an antisense morpholino oligonucleotide (MO) approach and an insertional mutation in the cdc14b locus to inhibit Cdc14B function during early development, we observed that cdc14b-deficient embryos developed through embryonic and early larval stages, but manifested a variety of abnormalities, including hydrocephaly, curved body, kidney cysts and defects in left-right (LR) asymmetry, which were reminiscent of phenotypes in zebrafish mutants with defective cilia (Bisgrove and Yost, 2006; Kramer-Zucker et al., 2005). Indeed, we found that Cdc14B is necessary for proper cilia formation in the Kupffer's vesicle (KV), the kidney duct and the sensory patches of the inner ear. The phosphatase activity of Cdc14B is required for its role in ciliogenesis as it is for its role in cell division control; these functions of Cdc14B can be dissociated from one another. Our study has uncovered a new role for Cdc14 phosphatases in vertebrates and suggests that the process of ciliogenesis and consequently a number of developmental processes are exquisitely sensitive to Cdc14B function.

Wild-type (WT) AB^*, cdc14b^vu426/vu426, tp53^zdf1/zdf1 (Berghmans et al., 2005), Tg[cmlc2:GFP], Tg[sox17:GFP] (Sakaguchi et al., 2006) and ace^x15/+ adult zebrafish, Danio rerio, were maintained as described previously (Solnica-Krezel et al., 1996). Embryos were staged according to morphology (Kimmel et al., 1995).

cdc14 cDNAs were amplified from a cDNA library obtained from 24 hours post fertilization (hpf) embryos (see Table S1 in the supplementary material), except for cdc14b_tv2, which was obtained from the zebrafish gene collection (ZGC:55844). cdc14b_tv1 was sub-cloned into the pCS2+ vector with or without sequences encoding a 5′GFP epitope tag.

Primers used for cdc14b_tv1 and cdc14b_tv2 transcript variants are listed in Table S1 in the supplementary material. The level of expression was normalized to that of actin.

Three antisense MOs (Gene Tools, Philomath, OR, USA) were designed to target cdc14b isoforms: MO1-cdc14b (MORPH1734, 5′-CACATGACTGTTGGAAAAGAACAGG-3′), MO2-cdc14b (MORPH1736, 5′-CTCGCTTCAACTAAAAACAAAGAGC-3′) and MO3-cdc14b (MORPH1735, 5′-TTTGCAGCAGAAGCGGTACAGCATG-3′) (see Fig. S1 in the supplementary material).

For the rescue and protein localization, synthetic mRNAs from cdc14b_tv1 and centrin (gift from Adrian Salic) were made by in vitro transcription using the mMessage mMachine System kit (Ambion, Austin, TX, USA).

Embryos fixed at 2.5 days post fertilization (dpf) in 4% paraformaldehyde were dehydrated, progressively infiltrated and embedded in JB-4 plastic resin (Polysciences, Warrington, PA, USA). Sections (5 μm) through the kidney were cut using a microtome (Leica RM2265) and stained with 0.5% Toluidine Blue.

Embryos and larvae were fixed in 4% paraformaldehyde overnight. Permeabilization was carried out using Proteinase K (10 μg/ml, Roche, Mannheim, Germany). For whole-mount in situ hybridization, embryos and larvae were hybridized overnight at 68°C with the following digoxigenin-labelled probes: cdc14b_tv1, cdc14b_tv2, nkx2.5 (Chen and Fishman, 1996), foxa3 (Odenthal and Nusslein-Volhard, 1998), southpaw (Long et al., 2003), insulin (Milewski et al., 1998), no tail (Schulte-Merker et al., 1994), myo7a (Ernest et al., 2000), erm1 (Munchberg et al., 1999), pea3 (Munchberg et al., 1999) and pax2a (Krauss et al., 1991). Detection and staining were carried out as previously described (Thisse and Thisse, 2008). For immunohistochemistry, embryos were incubated with the following primary antibodies: rabbit anti-GFP (1:500, Torrey Pines Biolabs, East Orange, NJ, USA), mouse anti-acetylated tubulin (1:250, Sigma-Aldrich, St Louis, MO, USA), mouse anti-TAT1 (1:250) (Woods et al., 1989), mouse anti-ZO-1 (1:250, Zymed laboratories, San Francisco, CA, USA) and mouse anti-γ-tubulin (1:500, Sigma-Aldrich, St Louis, MO, USA). Secondary antibodies were goat anti-rabbit or goat anti-mouse (Invitrogen, Carlsbad, CA, USA). F-actin staining was carried out using Alexa-fluor 594 phalloidin (2.5 μM, Invitrogen, Carlsbad, CA, USA) and nuclei were counterstained with TOPRO-3 (Molecular Probes, Eugene, OR, USA).

The ApopTag Peroxidase In Situ Apoptosis Detection Kit (Chemicon, Temecula, CA, USA) was used for the TUNEL assay.

Images were acquired using an Imager Z1 compound microscope and AxioVision software (Zeiss, Oberkochen, Germany) (in situ hybridization, sections) or Astro IIDC software (Aupperle Services and Contracting, Calgary, Canada) (movies) and with an LSM510 inverted confocal microscope (Zeiss, Oberkochen, Germany) (immunohistochemistry). Cilia length and pronephric duct diameter were measured following acquisition of z-stacks and reconstruction using the LSM Image acquisition software (Version 4.2.0.121, Zeiss, Oberkochen, Germany). Particle path tracking and average intensity per pixel were determined using the Metamorph software (Version 7.1.3.0, Molecular Devices, Silicon Valley, CA, USA).

We identified three zebrafish cdc14 paralogous genes, cdc14a1 (NM_201149.1, zgc:63654, chromosome 2), cdc14a2 (NM_001128538, chromosome 22) and cdc14b (NM_200179.1, zgc:55844, chromosome 8) (see Figs S1, S2 in the supplementary material), from a BLAST search of the Ensembl database using the sequence of S. pombe Cdc14 phosphatase, Clp1. Alignment of each zebrafish Cdc14 with CDC14 phosphatases from other organisms showed high sequence identity between the N-terminal catalytic domains and significant divergence among the C termini (see Fig. S3 in the supplementary material). Using RT-PCR from a 24 hpf cDNA library, at least two transcript variants were detected for cdc14a1, one for cdc14a2 and two for cdc14b (see Fig. S1 in the supplementary material). Additional transcripts might be expressed. As redundancy is likely to exist between the cdc14a genes, we focused our functional studies on cdc14b.

Quantitative RT-PCR and in situ hybridization revealed that cdc14b_tv1 and cdc14b_tv2 mRNAs were maternally provided and ubiquitously expressed throughout embryonic and early larval development (see Fig. S4 in the supplementary material; data not shown).

To investigate the function of Cdc14B during embryogenesis, we purchased a mutant line from Znomics (Portland, OR, USA), designated vu426, carrying a 7 kb retroviral insertion in the first intron of cdc14b sequence (see Fig. S1 in the supplementary material). This insertion was located upstream of both cdc14b isoform start codons. Quantitative RT-PCR at 10 hpf revealed that the expression of cdc14b was downregulated in the homozygous cdc14b^vu426/vu426 mutants when compared with wild-type embryos (Fig. 1A). cdc14b_tv1 expression was reduced by ∼50%, whereas that of cdc14b_tv2 was diminished by about 25%. However, this decrease in cdc14b expression did not lead to an overt morphological phenotype (Fig. 1B,C).

We therefore employed three MOs to inhibit the function of Cdc14B during embryogenesis (see Fig. S1 in the supplementary material): MO1-cdc14b was designed to interfere with splicing of both cdc14b_tv1 and cdc14b_tv2 transcripts. MO2-cdc14b and MO3-cdc14b were designed to prevent translation of cdc14b_tv1 and cdc14b_tv2, respectively. A dose series revealed that, whereas injections of low MO dose at the one-cell stage resulted in no overt morphological defects, injections of 4.5 ng, 3.5 ng and 8.5 ng of MO1-, MO2- and MO3-cdc14b, respectively, all produced similar phenotypes, which are described below. These doses were used throughout the experiments unless otherwise specified. The effect of MO1-cdc14b on cdc14b expression was monitored over time (Fig. 1F): wild-type transcripts were almost completely depleted at 10 hpf, but were detectable again from 1 dpf, indicating that from this stage, loss of Cdc14B function was incomplete. Sequencing of the aberrant transcripts from the embryos injected with MO1-cdc14b revealed that the fifth exon, predicted to encode a critical region of the catalytic domain (see Fig. S3 in the supplementary material) (Gray et al., 2003) was spliced out. This suggested that the aberrant transcripts encoded non-functional proteins.

Following injection of each cdc14b MO, no obvious morphological phenotype was observed prior to gastrulation, probably due to maternal contribution (see Fig. S4A,B in the supplementary material). By 24 hpf, all cdc14b-deficient embryos, referred to as cdc14b morphants, presented numerous morphological abnormalities, including hydrocephaly and dorsally or ventrally curved body (Fig. 1G-K). These defects persisted through later stages and from 3 dpf, kidney cysts also became apparent (Fig. 1L-R). Interestingly, the cdc14b morphant larvae did not display a strong body axis curvature as seen in zebrafish cilia mutants (Cao et al., 2010; Omori et al., 2008; Sullivan-Brown et al., 2008; Tsujikawa and Malicki, 2004; Zhao and Malicki, 2007). However, this weaker phenotype has previously been observed when MOs against genes encoding cilia proteins were injected (Ferrante et al., 2009; Hurd et al., 2010; Ko et al., 2010; Tsujikawa and Malicki, 2004). This difference in the phenotypes is probably due to the decreased efficiency of the MOs by this time in development. In addition, the size of the body, eyes and ears was reduced. Interestingly, low doses of MO1-cdc14b, which did not give rise to morphological defects when injected in wild-type embryos, led to a phenotype resembling that of cdc14b morphants when injected in cdc14b^vu426/vu426 embryos (Fig. 1D,E). This is consistent with the notion that the vu426 allele is hypomorphic and provides further support for the specificity of MO1.

To understand the mechanism underlying the reduced size of the embryos deficient for cdc14b, we looked at apoptotic cell death in the tail of MO1-cdc14b morphants at 27 hpf. TUNEL analysis revealed increased apoptosis in the morphants in comparison to non-injected control siblings (see Fig. S5A in the supplementary material). To assess whether cell death, due to either MO toxicity or Cdc14B loss of function, contributed to the morphological defects observed in cdc14b morphants, we used p53 zebrafish mutants, in which apoptosis is significantly suppressed (Berghmans et al., 2005). When each cdc14b MO was injected into embryos produced by tp53^zdf1/zdf1 parents, cdc14b morphants still displayed hydrocephaly, curved bodies, smaller eyes and ears and, although less pronounced, a shorter body (see Fig. S5B-K in the supplementary material). Therefore, excessive apoptotic cell death was unlikely to explain the morphological defects in cdc14b-deficient embryos, although it was in part responsible for the reduced morphant body size. We next analyzed the number of mitotic figures in the tail of MO1-cdc14b morphants at 27 hpf. Cell proliferation was diminished in cdc14b morphant embryos when compared to non-injected control siblings (see Fig. S5L in the supplementary material), and thus could also contribute to the reduced size of cdc14b-deficient embryos.

Hydrocephaly, curved body and kidney cysts are often associated with ciliogenesis defects in zebrafish (Wessely and Obara, 2008). Zebrafish cilia mutants also exhibit improper specification of LR asymmetry. Thus, we examined cdc14b morphants for defects in LR asymmetry using markers for asymmetrically positioned organs. Expression of heart markers nkx2.5 and cmlc2 (Chen and Fishman, 1996; Yelon et al., 1999) at 31 hpf showed that, as expected, the heart tube shifted to the left in most uninjected control embryos, whereas in the cdc14b morphants the shifting direction was random (Fig. 2A-D; see Fig. S6A-D in the supplementary material). Expression patterns of foxa3 and insulin, markers of the visceral organs (Milewski et al., 1998; Odenthal and Nusslein-Volhard, 1998), at 49 hpf, indicated that gut looping and localization of the liver and pancreas were also randomized in cdc14b morphants (Fig. 2E-H; see Fig. S6E-H in the supplementary material). We analyzed the expression of the earliest known marker asymmetrically expressed, southpaw (spaw), encoding a Nodal-related ligand (Long et al., 2003), at 18 hpf. Whereas spaw was expressed in the left lateral plate mesoderm (LPM) in the uninjected control embryos, in the cdc14b morphants its expression was seen either in the left LPM, the right LPM, both sides of the LPM or was absent (Fig. 2I-L). Injection of synthetic cdc14b RNA (up to 0.4 ng) did not alter expression of spaw (see Fig. S6I in the supplementary material).

As injection of low doses of MO1-cdc14b in cdc14b^vu426/vu426 mutant embryos could phenocopy the morphological defects manifested by cdc14b-depleted embryos, we wanted to know whether LR asymmetry defects could also be seen in the injected mutants. Analyzing heart location, the non-injected cdc14b^vu426/vu426 mutants did not display LR asymmetry defects. However, injection of low doses of MO1-cdc14b induced this defective phenotype in mutant, but not in wild-type embryos (Fig. 2M).

To verify the specificity of the LR asymmetry phenotype in the cdc14b morphants, we performed rescue experiments using synthetic cdc14b_tv1 mRNAs and spaw expression at 18 hpf as a read-out. MO1-cdc14b was injected at the one-cell stage alone or co-injected with the cdc14b_tv1 mRNAs. The expression of spaw exclusively in the left LPM was restored in a significant fraction of co-injected embryos in comparison with the embryos injected with MO1-cdc14b only (Fig. 2N).

Mechanisms leading to LR asymmetry have been characterized in many species (Basu and Brueckner, 2008). In zebrafish, the ciliated KV is the organ that establishes LR asymmetry (Essner et al., 2005). Within the KV, beating cilia generate a directional fluid flow that results in asymmetric gene expression. To determine whether Cdc14B was required for LR asymmetry in the KV, we performed a different type of rescue experiment. MO1-cdc14b was first injected at the one-cell stage. While some embryos were kept as controls, the rest were injected into the yolk with cdc14b_tv1 or cdc14b_tv1-gfp (which encodes a Cdc14B C-terminal GFP fusion) mRNAs at 3 hpf, after the yolk syncytial layer became separated from the embryo proper (Kimmel and Law, 1985). By injecting into the yolk syncytium at 3 hpf, we aimed to target the mRNAs specifically to the dorsal forerunner cells (DFCs) that give rise to the KV (Cooper and D'Amico, 1996; Melby et al., 1996). Indeed, between 3 and 4 hpf, only the DFCs share cytoplasmic bridges with the yolk cell where the mRNAs are injected. Aberrant spaw expression in the cdc14b morphants was partially suppressed after cdc14b_tv1 or cdc14b_tv1-gfp mRNAs were injected at 3 hpf (Fig. 2N). The rescue experiments, in addition to similar LR asymmetry defects produced by injection of three different MOs and by injection of low doses of MO1-cdc14b in cdc14b^vu426/vu426 mutant embryos, led us to conclude that the LR asymmetry phenotype in cdc14b morphants is specifically due to a loss of Cdc14B function rather than non-specific activity of the MOs.

As Cdc14 function in cell cycle control relies on its phosphatase activity (Trinkle-Mulcahy and Lamond, 2006), we investigated whether this activity was necessary for the role of Cdc14B in establishing LR asymmetry. We injected MO1-cdc14b at the one-cell stage and then, at 3 hpf, synthetic cdc14b_tv1 mRNA carrying a mutation in the catalytic domain such that it encoded a phosphatase dead protein (D253A) (Wolfe et al., 2006). Injection of the mutated mRNAs in the morphant embryos did not rescue the LR asymmetry defect (Fig. 2N), indicating that the phosphatase activity of Cdc14B was required for its role in establishing LR asymmetry. Injection of the mutated mRNA alone (up to 0.4 ng) did not affect LR asymmetry establishment (see Fig. S6I in the supplementary material), suggesting that at this level of expression the phosphatase-dead protein did not exert a dominant-negative activity.

To identify the cellular basis of the LR asymmetry defect in the cdc14b morphants, we first examined the integrity of the midline as disruption of this structure can lead to LR asymmetry defects via mis-localization of the cues defining the left and right sides (Bisgrove et al., 2000; Danos and Yost, 1996). Similar expression of no tail (Schulte-Merker et al., 1994) in the notochord of uninjected control siblings and cdc14b-deficient embryos at 8 hpf and 10 hpf (see Fig. S7A-D in the supplementary material) suggested that the midline was not affected.

We then examined KV morphogenesis. The KV had normal size and shape, and was properly inflated at 14 hpf (Fig. 3A,B; see Fig. S7E,F in the supplementary material) (Oteiza et al., 2008). Furthermore, sox17, a marker of the DFCs that give rise to the KV, as well as no tail (Melby et al., 1996; Schulte-Merker et al., 1994) were correctly expressed in the KV at 10 hpf and 8.5 hpf, respectively (Fig. 3A,B; see Fig. S7A,B in the supplementary material).

As Cdc14B is implicated in cell cycle control, we next investigated whether the number of cells that make up the KV was reduced in the Tg[sox17:GFP] embryos injected with cdc14b MOs at three developmental time points: (1) at 8 hpf, when the KV is not formed yet and the DFCs are about to divide, (2) at 10 hpf, when the KV is forming and cell division has occurred and (3) at 14 hpf, when the KV is fully formed and functional, as described by previous studies (Amack et al., 2007; Cooper and D'Amico, 1996). No significant differences were observed in the number of KV cells between uninjected control embryos and cdc14b morphants at any of these stages (Fig. 3C).

Finally, we examined cilia of the fully formed KV in the cdc14b morphants at 14 hpf. The correct number of KV cilia were present (Fig. 3D-F) and they could beat in the morphants, suggesting that their structure was not significantly affected (see Movies 1 and 2 in the supplementary material). However, our morphometric analyses indicated that the length of KV cilia was significantly reduced in the cdc14b morphants compared with that observed in uninjected control embryos (Fig. 3D,E,G; see Fig. S7G-J in the supplementary material). To show that the cdc14b MOs had a specific effect on KV ciliated cells, we injected MO1-cdc14b and MO2-cdc14b at 3 hpf to target the DFCs that will form the KV. We found that in these cdc14b-depleted embryos, the length of the KV cilia was also diminished (Fig. 3G). Moreover, using the same principle of rescue with sequential injection of MO1-cdc14b at the one-cell stage and cdc14b_tv1-gfp RNA at 3 hpf, we showed that the decrease in length of the KV cilia was specifically due to Cdc14B loss of function. Interestingly, in gain-of-function experiments, whereas 0.2 ng of cdc14b_tv1-gfp mRNAs did not produce a difference in the length of the KV cilia (Fig. 3G; see Fig. S7K in the supplementary material), 0.4 ng was sufficient to cause a significant increase in cilia length (Fig. 3G; see Fig. S7L in the supplementary material). However, this increase in cilia length was not associated with a LR asymmetry phenotype (see Fig. S6I in the supplementary material).

Recent studies indicate that cilia length is important for normal fluid flow in the KV (Neugebauer et al., 2009). Tracking of particle paths within the KV confirmed that fluid flow was disrupted in the KV of cdc14b-deficient embryos (Fig. 3H,I), presumably owing to their shorter length. Altogether, these results suggest that Cdc14B is involved in LR asymmetry establishment via a role in ciliogenesis.

The ubiquitous expression of cdc14b and the morphological phenotype of the cdc14b morphants suggested that other types of cilia could also be affected. Thus, we examined a structurally different type of motile cilia in the posterior kidney duct of uninjected control and MO1-cdc14b-injected embryos at 27 hpf. As in the KV, these cilia were present in the correct number (Fig. 4A-C). This observation also implied a correct number of ciliated cells, as cells are monociliated in the posterior kidney duct (Liu et al., 2007b). We confirmed that the number of cells was unaffected in the kidney duct of the cdc14b morphants at 2.5 dpf (Fig. 4F) (Sullivan-Brown et al., 2008). However, the kidney cilia were significantly shorter (Fig. 4D). Defects in cilia can result in fluid flow disruption causing enlargement of the duct (Kramer-Zucker et al., 2005; Sun et al., 2004). Indeed, the duct diameter was significantly enlarged in the cdc14b-deficient embryos compared with that of uninjected control siblings (Fig. 4A,B,E).

Next, we investigated primary cilia in the five sensory patches of the inner ear (anterior, lateral and posterior cristae, utricular and saccular maculae) (Whitfield et al., 2002) of 4 dpf larvae injected with MO1-cdc14b. Again, we found a significant decrease in the length of the kinocilia in the saccular macula and in the other sensory patches but also a decrease in their number (Fig. 5A-D; see Fig. S8A-C in the supplementary material). To clarify if the decrease in kinocilia number was due to a decrease in the number of hair cells from which kinocilia emanate, we analyzed the size of the sensory patches using myo7a, a marker of the hair cell terminal differentiation (Ernest et al., 2000). We observed a slight decrease in the size of the five sensory patches in the cdc14b morphants (Fig. 5E,F) and a diminution in the number of cells counted in the saccular macula (Fig. 5G). All five sensory patches consist of a double-layered epithelium that results from the asymmetric division of precursor cells into supporting cells and hair cells (Haddon et al., 1998). When we analyzed the structure of the maculae, we found that the cdc14b-deficient larvae displayed a double-layered epithelium similar to that of non-injected control siblings (Fig. 5H,I), and also that the ratio of the hair cell to supporting cell numbers was comparable in cdc14b morphant and control larvae (Fig. 5J). This indicated that a cell proliferation defect occurred prior to formation of the double-layered epithelium leading to a reduced number of precursor cells and thereby a diminished number of hair cells. Confirming this, we found that, as previously described (Platt, 1993; Sobkowicz et al., 1995), each kinocilium in the cdc14b morphant larvae was always associated with an actin-rich structure called stereocilia at the surface of the hair cells (Fig. 5K). Altogether, these results suggest that Cdc14B can affect ciliogenesis independently of its impact on cell proliferation.

To further examine the role of Cdc14B during ciliogenesis, we analyzed its subcellular localization. Owing to the lack of specific antibody, we injected in the one-cell stage embryos cdc14b_tv1-gfp mRNAs, as previously used for the rescue experiments, and monitored the dynamic localization of the protein in embryos at 8 hpf. Cdc14B was restricted to the nucleolus during interphase and was released in the cytoplasm during mitosis (Fig. 6A), consistent with the localization of Cdc14B from other vertebrate organisms (Berdougo et al., 2008; Cho et al., 2005; Kaiser et al., 2002; Mailand et al., 2002; Schindler and Schultz, 2009a; Wu et al., 2008). Also consistent with previous studies (Berdougo et al., 2008; Buffone et al., 2009; Schindler and Schultz, 2009a), we observed Cdc14B at the chromosomes and the mitotic spindle. However, we did not detect Cdc14B at centrosomes (Buffone et al., 2009; Cho et al., 2005; Wu et al., 2008) or at basal bodies, suggesting that Cdc14B affected ciliogenesis from another location.

In order to define the step of ciliogenesis regulated by Cdc14B, we first analyzed centriole migration to the membrane in the epithelial cells of the posterior kidney duct at 27 hpf. Although MO1-cdc14b morpholinos were not 100% efficient at 27 hpf, we reasoned that if cilia were shorter in the kidney, the reduction in cdc14b expression must be sufficient to disrupt ciliogenesis and therefore to impair migration of the centriole at the apical membrane if this process was regulated by Cdc14B. The centrioles appeared normally positioned at the apical side of the epithelial cells in distended kidney duct of both non-injected and MO1-cdc14b injected embryos (Fig. 6B,C). In addition, kinocilia emanated from the membrane rather than from the cytoplasm when viewed in cross-section of the sensory patches (data not shown) in morphant larvae. Altogether, we interpret these results to mean that Cdc14B is unlikely to regulate the first steps of ciliogenesis involving movement of the centrioles/basal bodies to the membrane.

The correct number of cilia in the KV as well as the posterior kidney duct suggested that nucleation of the cilia occurred properly in cdc14b-depleted embryos. Therefore, we reasoned that Cdc14B might influence cilia length control. The FGF pathway modulates cilia length via maintenance of a transcriptional network leading to the expression of intraflagellar transport (IFT) proteins required for cilia growth (Neugebauer et al., 2009). Therefore, we tested whether Cdc14B acted in the same or parallel pathway to FGF to regulate cilia length. Our analysis of cdc14b morphants revealed that two FGF target genes, erm1 and pea3 (Raible and Brand, 2001; Roehl and Nusslein-Volhard, 2001), were correctly expressed in the DFCs at 8.5 hpf (Fig. 7A-D) and that pax2a, another FGF target gene (Hans et al., 2004; Phillips et al., 2001; Reifers et al., 1998), was also normally expressed in the eyes, the midbrain-hindbrain boundary, the otic placode and the developing pronephros at 17.5 hpf (Fig. 7E,F; data not shown).

Next, we analyzed the effect of both Cdc14B and FGF8 loss of function on cilia length. To this end, we used a null allele of fgf8, x15 (Kwon and Riley, 2009), and we injected a low-dose of MO1-cdc14b in fgf8/acerebellar^x15/x15 (fgf8/ace^x15/x15) homozygous mutants and siblings. In contrast to what was previously described (Neugebauer et al., 2009), at 14 hpf, KV cilia were shorter in non-injected fgf8/ace^x15/x15 mutants in comparison with their siblings (Fig. 7G,H,K). This could be explained by the fact that we used a different fgf8/ace mutant allele. In embryos injected with low doses of MO1-cdc14b, the length of KV cilia was diminished in siblings and even further reduced in the fgf8/ace^x15/x15 homozygous mutant embryos (Fig. 7I-K). The fact that loss of Cdc14B function enhanced the effect of the loss of FGF signaling on KV cilia length, in addition to the normal expression of FGF targets in cdc14b-deficient embryos, suggested that Cdc14B and the FGF pathway act in parallel to regulate cilia length.

Microtubule acetylation is required to maintain cilia. Using similar conditions to acquire images of KV cilia stained with an anti-acetylated-tubulin antibody in morphant and non-injected control embryos at 14 hpf, we observed that the average intensity of microtubule acetylation was reduced in the morphants (Fig. 8A-C). These results provide a plausible mechanism whereby Cdc14B could regulate cilia microtubule stability. However, although statistically significant, the observed decrease in the acetylation level of the cilia microtubules was small and might be insufficient on its own to account for the shorter cilia in cdc14b morphant embryos.

Here, we examined the role of Cdc14B, one of two Cdc14 family members in vertebrates, during zebrafish embryonic development. Using an insertional mutant and three different antisense MOs to impair Cdc14B function, we found and confirmed by rescue and gain-of-function experiments, that Cdc14B plays an important role in ciliogenesis, and consequently in a number of developmental processes, including LR asymmetry establishment, and kidney and ear development.

The Cdc14 phosphatase family is best known for its roles in cell cycle control, especially in mitotic exit (Stegmeier and Amon, 2004). By analyzing concomitantly cilia length and cell number in the KV, in the kidney duct and in the inner ear sensory patches, we were able to show that, in these cases, the function of Cdc14B in ciliogenesis could be dissociated from any potential role in cell division. Thus, Cdc14B appears to be functioning in a new biological process in zebrafish when compared with its established functions in yeasts.

Several explanations can account for the absence of cell cycle defects, despite a significant ciliogenesis defect in some of the tissues we studied in cdc14b-deficient embryos: (1) sufficient Cdc14B activity could remain after MO injections for cell division to occur; (2) unknown and non-targeted isoforms of Cdc14B and/or Cdc14A enzymes could compensate for Cdc14B loss of function in mitosis; and/or (3) other phosphatases are responsible for dephosphorylating mitotic proteins and allowing the cell to proceed through M-phase. It is also possible that Cdc14B is required for cell division in a subset of cells but dispensable for a majority of cell divisions in zebrafish.

In the tail as well as in the population of hair cell precursors, we observed a decrease in the number of mitotic figures in cdc14b-depleted embryos. This could be due to less proliferation, as discussed above, or delayed cell cycle progression and development. It should also be noted that in S. pombe, cells lacking the Cdc14 ortholog Clp1/Flp1 display a reduced length owing to premature entry into mitosis (Cueille et al., 2001; Trautmann et al., 2001). Thus, the diminished size of cdc14b-deficient embryos might result from smaller cells. Further study will be required to distinguish between these possibilities and discover whether the role of Cdc14 phosphatases in yeast cell cycle control can be extended to vertebrates.

In yeast, Cdc14 phosphatases are controlled by their subcellular localization and the same is thought to be true of their vertebrate counterparts. During interphase, Cdc14 is sequestered and kept inactive in the nucleolus, whereas it is released and fully active during late stages of mitosis (Trinkle-Mulcahy and Lamond, 2006). Ciliogenesis is a cell cycle-dependent process. During G1, the centriole migrates from the cytoplasm to the apical surface of the cell where it docks to the membrane (Dawe et al., 2007). Elongation of the cilium then occurs. During G2, the cilium retracts and it is absent during the following M-phase. If zebrafish Cdc14B is active only in anaphase, it would probably influence the dynamics of ciliogenesis early in the process. Our analysis of the first steps of ciliogenesis indicates that Cdc14B probably does not influence centriole migration and docking at the membrane (Fig. 6B,C) nor is it required in cilia nucleation. Furthermore, cilia beat in embryos depleted of Cdc14B, suggesting that their structure is relatively intact (see Movies 1 and 2 in the supplementary material). Based on these observations, we propose that Cdc14B regulates cilia elongation and/or maintenance.

Cdc14 proteins function in cell division control via their phosphatase activity. Because overexpression of the wild-type but not phosphatase-dead Cdc14B protein suppressed LR asymmetry defects of the cdc14b morphants (Fig. 2N), we conclude that Cdc14B requires its phosphatase activity to establish LR asymmetry. Knowing that the LR asymmetry defect in cdc14b-deficient embryos is likely due to the reduced length of the KV cilia, it is expected that Cdc14B affects ciliogenesis via dephosphorylation of substrates yet to be identified. As phosphoregulation is already known to be a key mechanism regulating ciliogenesis (Gerdes et al., 2009), it is reasonable to think that Cdc14 phosphatases evolved to participate in this process.

Proteomic analyses of cilia and flagella in different organisms have revealed the presence of many kinases and phosphatases in these structures (Boesger et al., 2009; Liu et al., 2007a; Ostrowski et al., 2002; Pazour et al., 2005). Notably, phosphatase inhibitor 2 has been proposed to promote stability of the cilia via tubulin acetylation of the cilia axoneme due to indirect inhibition of histone/tubulin deacetylases such as HDAC-6 (Wang and Brautigan, 2008). On the contrary, the Aurora A kinase is involved in cilia disassembly via phosphorylation and activation of HDAC-6 (Pugacheva et al., 2007). Interestingly, mammalian CDC14B as well as S. cerevisiae Cdc14 have been previously shown to have a role in microtubule stabilization (Cho et al., 2005; Higuchi and Uhlmann, 2005). Although cdc14b morphant embryos display a decrease of the acetylation level of cilia microtubules (Fig. 8A-C), we cannot at this point conclude whether the role of Cdc14B in ciliogenesis is limited to this effect.

Several models have been proposed to explain cilia length regulation. One relies on a balance between assembly and disassembly rates of the cilia microtubules (Marshall and Rosenbaum, 2001). Numerous kinases have previously been involved in regulation of cilia/flagella length. In Chlamydomonas reinhardtii, Long Flagella 2 (LF2) and LF4 encode a CDK-related kinase and a MAP kinase, respectively, that are implicated in flagellar length control (Berman et al., 2003; Tam et al., 2007). In zebrafish, a homolog of LF2p, cell cycle-related kinase, was recently implicated in cilia assembly (Ko et al., 2010). Given the affiliation between Cdc14 phosphatases and CDKs, and the fact that LF2p and its homolog localize to the cytoplasm rather than to the cilia/flagella, it will be interesting to test genetic and molecular interactions between these components.

Cilia defects have not been described in C. elegans lacking Cdc14, possibly because the role of Cdc14 in ciliogenesis is redundant with other phosphatases in this organism. However, the role of Cdc14B in ciliogenesis might be conserved in other invertebrates or vertebrates. Certainly, the characterization of cdc14 mutants in Chlamydomonas reinhardtii would shed light on the conservation of this function in invertebrates.

We identified three zebrafish cdc14 genes, two homologs of cdc14a and a single cdc14b gene, with numerous splice variants. The diversity of cdc14 transcripts will probably be a complicating factor in the functional analysis of Cdc14 proteins in vertebrate species. Their variable domains might allow specific subcellular localization, substrate interaction or regulation, and, in turn, the evolution of specific functions. In our studies, we simultaneously targeted two major transcripts of cdc14b and future studies will be necessary to determine whether they contribute equally to ciliogenesis.

In human cells, Cdc14A localizes to centrosomes (Kaiser et al., 2002; Mailand et al., 2002). It will therefore be interesting to determine whether Cdc14A also participates in ciliogenesis and if its function is redundant with Cdc14B or is independent of Cdc14B. In addition, it will be important to determine whether an imbalance of Cdc14 phosphatase activity underlies or impacts human ciliopathies.

The authors thank members of the Gould and Solnica-Krezel labs for many valuable discussions and comments on the manuscript, Dr B. Blanco-Sanchez and Dr T. Whitfield for their advice on the sensory patches study, and the Westerfield lab for reagents. This work was supported by a Martha Rivers Ingram endowment to L.S.K. and the HHMI, of which K.L.G. is an investigator. Deposited in PMC for release after 6 months.