Wnt/β-catenin signaling directly regulates Foxj1 expression and ciliogenesis in zebrafish Kupffer’s vesicle

Cilia are microtubule-based organelles that are assembled and maintained by intraflagellar transport (IFT) proteins (Pedersen and Rosenbaum, 2008). Cilia can be broadly divided into sensory and motile subtypes (Satir and Christensen, 2007). Sensory or primary cilia form on nearly all interphase and nondividing cells (Pedersen and Rosenbaum, 2008) and are the site of transduction for developmental signaling pathways such as the Hedgehog (Hh) pathway and possibly the Wnt pathways (Eggenschwiler and Anderson, 2007). By contrast, motile cilia almost exclusively form in tissues that produce fluid flow, such as the respiratory airways, spinal canal and embryonic node (Eley et al., 2005). Genetic defects in motile cilia lead to primary ciliary dyskinesia (PCD), a heterogeneous syndrome that is characterized by respiratory infections, hydrocephalus, situs inversus, male infertility and occasionally cystic kidney and retinal degeneration (Afzelius, 1976; Olbrich et al., 2002; Kramer-Zucker et al., 2005; Badano et al., 2006). Motile cilia biosynthesis specifically requires Foxj1, a winged-helix domain-containing transcriptional factor (Chen et al., 1998). Recently, Foxj1 has been demonstrated to not only be necessary, but also sufficient for motile cilia synthesis (Stubbs et al., 2008; Yu et al., 2008). However, little is known about the molecular mechanism that governs foxj1 expression.

In a developing zebrafish embryo, motile cilia are present in cells lining the Kupffer’s vesicle (KV), a temporary organ equivalent to the mouse node and to Xenopus gastrocoel roof plate (GRP). These nodal cilia rotate in a clockwise direction to generate a leftward flow of extracellular fluid that is essential for the establishment of left-right (LR) asymmetry (Nonaka et al., 1998; Nonaka et al., 2002; Essner et al., 2005; Kramer-Zucker et al., 2005; Hirokawa et al., 2006). Uniquely in zebrafish, fate-mapping studies have shown that KV originates from a group of dorsal forerunner cells (DFCs) (Cooper and D’Amico, 1996; Melby et al., 1996). These DFCs retain cytoplasmic bridges with the yolk cell after other cells are no longer connected to the yolk. Consequently, genetic manipulation can be achieved exclusively in DFCs of zebrafish (Amack and Yost, 2004).

Motile cilia have also been found in the developing zebrafish pronephric ducts (PDs), otic vesicles (OVs) and floor plate (FP) (Kramer-Zucker et al., 2005). PD motile cilia are required for normal kidney organogenesis and ciliary motility in OVs ensures proper formation of otoliths, which are important for the ear’s sensory function (Riley et al., 1997; Kramer-Zucker et al., 2005; Yu et al., 2008; Colantonio et al., 2009; Yu et al., 2011). Two distinct foxj1 genes, foxj1a and foxj1b, have been identified in zebrafish tissues. Their expression is mutually exclusive in KV (foxj1a) and OVs (foxj1b), but partially overlaps in PDs and FP (Aamar and Dawid, 2008; Yu et al., 2008). Depletion of foxj1a results in abnormal LR asymmetry and renal cyst formation, whereas depletion of foxj1b leads to otolith disorganization (Stubbs et al., 2008; Yu et al., 2008; Yu et al., 2011). Whether foxj1 expression and cilia synthesis in different tissues are regulated by a general developmental mechanism remains to be investigated.

The Wnt/β-catenin pathway is activated upon binding of Wnt ligands to the Frizzled–low-density lipoprotein receptor-related protein (Fzd-LRP5/6) receptor complex. This causes β-catenin stabilization and translocation to the nucleus where it binds to the lymphoid enhancer factor and T cell factor (Lef and Tcf) transcription factors to activate Wnt target gene expression (Logan and Nusse, 2004; Willert and Jones, 2006). Wnt/β-catenin signaling has been implicated in the establishment of LR asymmetry. Moderate upregulation of Wnt signaling that does not alter dorsoanterior structure leads to no-looping hearts as seen in the zebrafish apc and mbl mutants (Carl et al., 2007; Lin and Xu, 2009); however, it has no effect on left-sided spaw expression, suggesting that LR patterning function of KV is not affected (Carl et al., 2007; Lin and Xu, 2009). Downregulation of Wnt signaling, as seen in mouse Wnt3a mutants and zebrafish wnt3a and wnt8a morphants, causes randomized organ laterality and randomized side-specific gene expression, which is likely to be a combined effect of Wnt activity on midline formation and KV function (Nakaya et al., 2005; Lin and Xu, 2009). The precise function of Wnt signaling in KV has yet to be elucidated.

Here, using zebrafish KV development as a paradigm, we aim to elucidate molecular mechanisms of how Wnt/β-catenin signaling regulates LR asymmetry. Our data reveal a temporal and KV cell-autonomous Wnt function in LR axis determination and cilia formation. Importantly, we show that this line of Wnt function can be ascribed to its direct transcriptional regulation of foxj1a. We also expand our studies to the developing PDs and OVs, two epithelia that require Wnt activity for their normal development and function. Our results indicate that reduction of Wnt signaling impairs foxj1 expression and ciliogenesis in both of these tissues, suggesting a more general role for the Wnt/β-catenin pathway in the control of foxj1 expression and cilia formation.

Wild-type (TL), Tg(hsp:β-catenin-GFP) and Tg(hsp:dkk1-GFP) strains of zebrafish were used for this work. The two heat-inducible transgenic lines were generated by injection of hsp70 promoter-driven dkk1 or β-catenin1 expression plasmids into one-cell staged embryos. Heat shock was carried out at 40°C for 60 minutes, unless otherwise specified in the Results.

To generate the Tg(0.6foxj1a:gfp) transgenic reporter line, a 0.6 kb fragment that is located approximately –5.2 kb to –4.6 kb upstream of the ATG start codon of the zebrafish foxj1a gene was inserted into a tol2 GFP vector. The resulting plasmids were co-injected with tol2 transposase RNA into one-cell staged embryos. To generate the Tg(0.6Δfoxj1a:gfp) transgenic reporter strain, all three putative Lef1/Tcf binding sites (CTGTT, CCTTTGTT and CACAG) in the 0.6 kb fragment were deleted using the QuickChange II XL Site-Directed Mutagenesis Kit (Agilent).

Antisense morpholino oligonucleotides (MOs) against β-catenin1 (ctnnb1 – Zebrafish Information Network) (Lyman Gingerich et al., 2005), fgfr1 (Neugebauer et al., 2009) and fzd10 (targeting the AUG site: 5′-TGAGTCCGACACCGGCAGCAAACAT-3′) were obtained from Gene Tools, LLC. Unless specified otherwise in the Results, 4 ng of each morpholino was used. To target morpholinos specifically to DFCs, lissamine-tagged MOs were injected into the yolk cell at ∼512-cell stage as previously described (Amack and Yost, 2004). The lissamine tag allowed us to then select embryos in which the dye had diffused throughout the yolk cell and into DFCs.

Full-length zebrafish axin1, foxj1a and fgf8a cDNA were amplified by an Expand High-Fidelity PCR system (Roche), using zebrafish cDNA as a template. The resulting cDNA fragments were cloned into the pCS2+ plasmid.

Capped mRNAs were synthesized from the above pCS2+ plasmids and pCS2-wnt8a construct (Lin et al., 2007) using the SP6 mMESSAGE mMACHINE kit (Ambion). To achieve DFC-specific over-expression, 4 ng of axin1 RNA, 2 ng of foxj1a RNA, 0.2 ng of wnt8a RNA or 50 pg of fgf8a RNA together with dextran-rhodamine lineage tracing dye (Molecular Probes) or lissamine-tagged MO were injected into the yolk cell at ∼512-cell stage.

Single-color whole-mount in situ hybridization and two-color fluorescent hybridization were conducted as previously described (Lin and Xu, 2009). spaw and pitx2 proportions were analyzed using Fisher’s exact test. Results were considered to be significant when P<0.01.

SU5402 treatment was carried out as previously described (Neugebauer et al., 2009).

Cilia were visualized by immunostaining using monoclonal anti-acetylated tubulin antibody (Sigma) as previously described (Lin and Xu, 2009). Cilia length and number were measured using AxioVision software and analyzed using a two-tailed Student’s t-test. Results were considered to be significant when P<0.01.

Embryos at ∼24 hours post-fertilization (hpf) were deyolked for embryo lysate extraction as previously described (Link et al., 2006). Lysates were loaded onto 4-20% Tris-HCl mini gradient gels (BioRad) and transferred to PVDF membranes. Anti-Fzd10 antibody (ProteinTech Group) was used at 1:400, and anti-Actin antibody (Santa Cruz Biotechnology) at 1:1000. Blots were developed using ECL plus Western Blotting Detection System (Amersham).

Previously, we showed that knocking down Wnt3a or Wnt8a in zebrafish results in cardiac asymmetry defects (Lin and Xu, 2009). Because Wnt/β-catenin signaling plays stage-dependent roles during development (Ueno et al., 2007), we decided to determine the time window during which Wnt signaling is needed for LR axis determination. To temporally reduce Wnt/β-catenin signaling, we generated a transgenic Tg(hsp:dkk1-GFP) fish line in which the expression of Dkk1 (Dickkopf1), an antagonist of the Wnt/β-catenin pathway (Glinka et al., 1998), was driven by a heat-shock promoter (supplementary material Fig. S1A,B). We found that Dkk1 activation in embryos from 30% epiboly to bud stage all randomized cardiac jogging. By contrast, induction of Dkk1 from the 3-somite stage onwards did not elicit significant jogging defects (A.C., unpublished). At the molecular level, induction of Dkk1 prior to the 3-somite stage altered left-sided gene expression, including spaw in the lateral plate mesoderm (LPM) and pitx2 in the posterior LPM (Table 1; supplementary material Fig. S1C). As expected, non-transgenic siblings heat-shocked at all stages displayed normal expression of spaw and pitx2 (Table 1). To activate Wnt/β-catenin signaling temporally, we generated a transgenic Tg(hsp:β-catenin-GFP) fish line (supplementary material Fig. S1D-F). Induction of β-catenin1 at stages up to 10-somites did not alter either cardiac jogging or spaw expression (Table 1) (A.C., unpublished). Thus, we conclude that Wnt/β-catenin signaling must be kept above a certain level before segmentation to ensure the establishment of LR asymmetry.

To determine whether Wnt signaling affects LR axis determination directly or indirectly via its role in maintaining midline integrity (Takada et al., 1994; Nakaya et al., 2005; Thorpe et al., 2005; Lin and Xu, 2009), we performed DFC-targeted knockdown of genes in the Wnt/β-catenin pathway. We first tested wnt3a and wnt8a, which were previously shown to be expressed near or within KV (Lin and Xu, 2009). Two-color fluorescent in situ hybridization revealed that they had overlapping expression with charon (dand5 – Zebrafish Information Network), a novel member of the Cer/Dan family of Nodal antagonists that has KV-specific expression (Hashimoto et al., 2004), confirming their KV localization (supplementary material Fig. S2I-N). We found that DFC-targeted injection of wnt3a or wnt8a MOs resulted in LR asymmetry defects (A.C., unpublished). However, we cannot ascribe a KV-cell autonomous function to them because Wnts are secreted molecules. We then turned our attention to frizzled receptors. We screened 12 annotated zebrafish frizzled homologs and found the expression of fzd10, which encodes a putative receptor for Wnt3a and Wnt8a (Momoi et al., 2003; Kemp et al., 2007), near or within KV (Fig. 1B,C). fzd10 expression can be detected in DFCs at ∼80% epiboly (Fig. 1A, arrow), and in the brain, dorsal neural tube and tail bud/KV at the 10-somite stage (Fig. 1B,C) (see also Nasevicius et al., 2000). To deplete Fzd10, we designed a translation-blocking MO of fzd10. Injection of this MO into one-cell staged embryos greatly suppressed Fzd10 levels, as revealed by western blotting (Fig. 1D). Consistent with the notion that fzd10 transduces the Wnt/β-catenin pathway (Terasaki et al., 2002; Momoi et al., 2003; Wang et al., 2005), the expression of specific targets of this pathway were inhibited, as demonstrated by depletion of sp5l from DFCs in DFC^fzd10MO embryos (Fig. 1F, arrow) and downregulation of axin2 in the brain, neural tube and tail bud region in fzd10 morphants (Fig. 1H) (Jho et al., 2002; Huang and Schier, 2009). Conversely, a role of fzd10 in the Wnt/PCP pathway, as suggested by studies in synovial sarcomas and synchronous colorectal tumors (Fukukawa et al., 2009; Nagayama et al., 2009), was excluded because defects in convergent extension movement, such as widened somite and abnormal ntl, hgg1 (ctsl1b – Zebrafish Information Network) and/or dlx3 staining, were not detected in fzd10 morphants (supplementary material Fig. S2O-R). As expected, asymmetry defects were observed in fzd10 morphants (Fig. 1J) and, more importantly, in DFC^fzd10MO embryos. For example, injection of 1 ng of MO resulted in a significant defect in spaw expression (50% left, 20% right, 1% absence, 29% bilateral; n=119) compared with those containing fzd10 MO in the yolk alone (95% left, 2% right, 3% bilateral; n=78) (Fig. 1I,J; P<1×10^–10). Injection of 4 ng of MO caused more severe defects (35% left, 35% right, 30% bilateral; n=63). These data suggest a KV-cell autonomous Wnt function in LR axis determination.

To consolidate this notion, we expanded our search to other components of the Wnt/β-catenin pathway and detected near or within KV expression of intracellular Wnt signal transducers axin2, β-catenin1 and β-catenin2 (ctnnb2 – Zebrafish Information Network) as well as transcription factors lef1 and tcf7 (supplementary material Fig. S2D-H). DFC^axin1RNA embryos showed moderate, but statistically significant, alterations in spaw expression (Fig. 1J; P<2×10^–04 compared with Yolk^axin1RNA embryos). Moreover, DFC^{β-catenin1MO} embryos also exhibited significant randomization in spaw expression relative to the Yolk^{β-catenin1MO} embryos (Fig. 1J; P<4.37×10^–05). Collectively, our data indicate that Wnt signaling has an intrinsic function in LR axis determination.

To gain insight into how Wnt signaling regulates LR patterning, we examined ciliogenesis in KV. We found that induction of Dkk1 at stages prior to the 3-somite stage led to a significant reduction in cilia length. Specifically, moderate Dkk1 induction (Dkk+) by heating Tg(hsp:dkk1-GFP) embryos for 30 minutes resulted in shorter cilia (4.38±0.56 μm) compared with heat-shocked non-transgenic siblings (5.46±0.47 μm) (Fig. 2A,B,J; P<3×10^–04) and wild-type embryos (5.67±0.37 μm) (Fig. 2D,J; P<1.34×10^–06), both of which had similar cilia length (P<0.43). The number of cilia in Dkk+ embryos remained normal (36±13 relative to 42±12 in non-transgenic siblings) (Fig. 2A,B,K; P<0.32). However, if a higher level of Dkk1 (Dkk++) was induced by increasing the duration of heat shock to 60 minutes, not only was cilia length further shortened (3.04±0.34 μm; P<8.61×10^–15), but the number of cilia was significantly reduced as well (23±11; P<2.0×10^–4) compared with non-transgenic siblings (Fig. 2C,J,K). Similarly, injection of fzd10 MO into one-cell staged embryos inhibited cilia length/number in a dose-dependent manner (Fig. 2J,K). By contrast, transient upregulation of Wnt signaling by inducible β-catenin1 expression had no significant effect on cilia length (Fig. 2J; P<0.58) or cilia number (Fig. 2K; P<0.14) compared with controls. Thus, we conclude that Wnt signaling is required for ciliogenesis in KV, which temporally correlates with its stage-dependent function in regulating LR patterning.

DFC-targeted reduction of Wnt signaling also disrupted ciliogenesis. Cilia in DFC^axin1RNA embryos were shorter (Fig. 2J; P<8.95×10^–10), but the number remained unchanged (Fig. 2K; P<0.69) compared with controls. DFC^{β-catenin1MO} embryos had shorter (P<7.99×10^–14) and fewer (P<4.88×10^–05) cilia (Fig. 2J.K). Furthermore, injection of 1 ng of fzd10 MO significantly reduced cilia length only (P<5.00×10^–11), whereas injection of 4 ng of MO reduced both cilia length (P<1.94×10^–14) and cilia number (P<14.07×10^–05) compared with controls (Fig. 2E,F,J,K). These data indicate a KV cell-autonomous control of cilia formation by Wnt signaling. To explore whether cilia motility is affected, we assessed the expression of dnah9 (dynein axonemal heavy polypeptide 9, also known as lrdr1). dnah9 is the zebrafish homolog of the mammalian Iv (inversus viscerum; now known as Dnahc11) gene that encodes a conserved dynein heavy chain required for cilia motility (Supp et al., 1997; Essner et al., 2005). Dkk1 induction and DFC-targeted injection of MOs against β-catenin1 or fzd10 all suppressed dnah9 expression (Fig. 2H,I and Fig. 3F), suggesting an impairment in cilia motility.

To address whether the abnormal ciliogenesis is secondary to general effects of Wnt signaling on patterning, we assessed KV morphogenesis. First, the expression of sox17, a marker of DFCs/KV lineage (Alexander et al., 1999), was evaluated at the bud stage for DFC specification and at 6-somite stage for KV cell maintenance. Neither Dkk1 activation nor DFC-specific injection of fzd10 MO markedly altered sox17 expression (supplementary material Fig. S3A-F). Second, we noted that embryos overexpressing Dkk1 and embryos depleted of Fzd10 had mostly normal KV morphology, despite being smaller when a higher level of Dkk1 was induced or a higher dose of MO was used (supplementary material Fig. S3G-I) (A.C., unpublished). Third, we found that the tight junction marker ZO1 was expressed normally in Dkk1-expressing embryos and DFC^fzd10MO embryos (supplementary material Fig. S3I). Together, these results indicate that Wnt signaling regulates ciliogenesis without disturbing DFC specification, KV formation or apical-basal polarization of KV cells. However, we cannot rule out the possibility that other aspects of KV development might be affected.

Because Wnt/β-catenin signaling functions mainly through transcriptional regulation of its targets, we reasoned that it might modulate genes implicated in KV ciliogenesis. The expression of ciliogenic transcription factor foxj1a is initiated in DFCs, becomes robust towards the end of gastrulation (Aamar and Dawid, 2008; Yu et al., 2008; Tian et al., 2009) and is then gradually downregulated in KV (A.C., unpublished). The correlation between foxj1a expression in DFCs/KV lineage and the temporal requirement for Wnt signaling in KV ciliogenesis prompted us to examine foxj1a transcripts. foxj1a expression was reduced in embryos expressing a moderate level of Dkk1 (30 minutes heat shock) and embryos injected with 1 ng of fzd10 MO (A.C., unpublished), and was nearly abolished in embryos expressing a higher level of Dkk1 (60 minutes heat shock) (Fig. 3B) and embryos injected with 4 ng of fzd10 MO (Fig. 3C). By contrast, the level of foxj1a remained normal following brief upregulation of Wnt signaling via β-catenin1 activation (Fig. 3D).

To determine whether Wnt signaling controls ciliogenesis and LR asymmetry through foxj1a, we performed foxj1a RNA rescue experiments. Unlike whole-body over-expression (Tian et al., 2009), DFC-targeted over-expression of foxj1a RNA did not appear to significantly alter left-sided spaw expression (Fig. 3K; P<0.25 compared with controls). It also did not cause any apparent changes in cilia length (P<0.65) and number (P<0.73) relative to controls (Fig. 3I,J). However, it significantly restored cilia length in DFC^fzd10MO embryos (from 3.20±0.59 μm to 4.68±0.31 μm; P<8.16×10^–09), DFC^{β-catenin1 MO} embryos (P<3.00×10^–04) and embryos overexpressing Dkk1 (P<1.83×10^–10) (Fig. 3I), which is in agreement with the documented role of foxj1a in cilia length elongation (Cruz et al., 2010; Yu et al., 2011). To a lesser extent, ectopic expression of foxj1a RNA enhanced cilia number in DFC^fzd10MO embryos (from 18±7 to 25±8; P<0.02), DFC^{β-catenin1MO} embryos (P<0.04) and embryos with activated Dkk1 (P<0.15) (Fig. 3J). In addition to cilia length and number, ectopic expression of foxj1a RNA was able to restore dnah9 expression in DFC^fzd10MO embryos (Fig. 3E-G). Lastly, ectopic expression of foxj1a RNA partially rescued abnormal spaw expression in DFC^{β-catenin1MO} embryos (P<4.00×10^–03) and DFC^fzd10MO embryos (P<3.00×10^–06) (Fig. 3K). Together, our results demonstrate a role of foxj1a in Wnt-regulated ciliogenesis and LR asymmetry.

Because FGF signaling has also been implicated in the regulation of ciliary growth (Neugebauer et al., 2009), we attempted to understand how these developmentally important pathways converge on ciliogenesis. Consistent with a previous report (Neugebauer et al., 2009), downregulation of FGF signaling by incubation with 10 μm of SU5402, an FGF pathway inhibitor, or injection with 1 ng of fgfr1 MO resulted in reduced foxj1a expression and reduced cilia length, but the number of cilia was maintained (supplementary material Fig. S4). We also found that treatment with 20 μm of SU5402 or injection with 4 ng of fgfr1 MO led to a greater inhibition of foxj1a expression and reduction of cilia length as well as a reduction in cilia number without significantly affecting DFC specification (Fig. 4B,I,J; supplementary material Fig. S4). These phenotypes are strikingly similar to those induced by blocking of Wnt signaling. Next, the genetic relationship between the Wnt and FGF pathways was examined by epistatic analysis. Although DFC-targeted overexpression of wnt8a RNA alone did not have an apparent effect on wild-type embryos (Fig. 4D,I,J), it restored foxj1a expression (Fig. 4B,C) and significantly enhanced cilia length (from 2.95±0.58 μm to 4.18±0.46 μm; P<3.13×10^–08) (Fig. 4I), and cilia number remained unchanged (Fig. 4J; P<0.35) in DFC^fgfr1MO embryos. By contrast, DFC-targeted overexpression of fgf8a RNA failed to rescue foxj1a expression (Fig. 4F,G), cilia length (Fig. 4I; P<0.21) or cilia number (Fig. 4J; P<0.24) in DFC^fzd10MO embryos. The amount of fgf8a RNA used in our rescue experiment was the highest amount that alone did not significantly disrupt either foxj1a expression (Fig. 4H) or embryo morphology (A.C., unpublished). Together, our data suggest that Wnt signaling functions downstream of FGF signaling in the regulation of foxj1a expression and ciliogenesis in the zebrafish KV.

We then investigated how Wnt signaling regulates foxj1a expression. We first examined the timing between Wnt inhibition/activation and foxj1a expression. Tg(hsp:dkk1-GFP) embryos were heat shocked at 30% epiboly, the earliest stage showing detectable foxj1a expression (Aamar and Dawid, 2008), for 30 minutes and were then collected 30 minutes later for assessment of foxj1a expression. Dkk1 induction prevented initiation of foxj1a transcription (supplementary material Fig. S5). Conversely, though β-catenin1 activation did not significantly alter foxj1a expression 4 hours after heat shock (Fig. 3D), it strongly and transiently enhanced foxj1a expression in DFCs 1 hour after heat shock (supplementary material Fig. S5). These rapid responses indicate that Wnt signaling does not require the synthesis of an intermediate protein in order to control foxj1a transcription.

To determine whether Wnt signaling directly regulates foxj1a transcription, we performed enhancer dissection. Bioinformatics analysis identified a 150 bp stretch of sequence that is homologous with tetradon sequence, 50 bp of which is also conserved in humans (Fig. 5A). Because the sequence partially overlaps with the predicted first exon, where essential cis-acting elements are typically located in proximity, we generated a series of constructs containing the 150 bp sequence and/or its flanking sequences (Fig. 5A). Two fragments (0.6 kb and 1.6 kb, respectively) were identified to sufficiently direct the expression of GFP fluorescent reporter in KV, PDs and FP by transient injection assay (Fig. 5A) (A.C., unpublished). The smaller 0.6 kb sequence was chosen to generate a transgenic reporter strain. The stable transgenic Tg(0.6foxj1a:gfp) embryos expressed gfp transcripts in DFCs and GFP in KV, PDs and FP (Fig. 5B,D-F), thus recapitulating endogenous foxj1a expression pattern. Importantly, DFC-targeted injection of fzd10 MO into Tg(0.6foxj1a:gfp) embryos abolished gfp expression in DFCs (Fig. 5C), suggesting the existence of Wnt-responsive elements in the enhancer region. Upon inspection of the sequence, three putative Lef1/Tcf binding sites were identified (Fig. 5A, red line). We then constructed a GFP reporter plasmid driven by the enhancer with deletions in all three putative Lef1/Tcf binding sites (0.6Δ). In stable transgenic Tg(0.6Δfoxj1a:gfp) embryos, GFP reporter expression appeared to be absent from KV and PDs, but was maintained in the FP (Fig. 5G-I), indicating that the putative Lef1/Tcf binding sites are required for tissue-specific foxj1a expression. Taken together, we conclude that Wnt signaling directly controls foxj1a transcription in KV.

To address whether Wnt-Foxj1 signaling regulates ciliogenesis beyond KV, we examined cilia in the developing kidney and inner ear. In the kidney, the Wnt/β-catenin pathway has been implicated in normal development in mice, Xenopus and zebrafish (Park et al., 2007; Lyons et al., 2009), and impaired Wnt/β-catenin signaling has been shown to cause cystic kidney in mice (Marose et al., 2008; Lancaster et al., 2009). We found that induction of Dkk1 did not grossly alter the expression of PD precursor marker pax2 (supplementary material Fig. S6); however, it diminished foxj1a expression (Fig. 6B, arrow). Given that Tg(0.6Δfoxj1a:gfp) embryos did not exhibit GFP reporter expression (Fig. 5H,I), Wnt signaling might directly regulate foxj1a transcription in PDs. Furthermore, induction of Dkk1 resulted in fewer and shorter cilia (4.17±0.49 μm relative to 6.79±0.18 μm in controls; P<1.03×10^–09) (Fig. 6G,H,K), as well as a loss of dnah9 expression in PDs (Fig. 6F, white arrow). Consistent with the cilia defects, cystic distension of PDs was present in day 1 embryos (Fig. 6M, arrow). However, Dkk1 activation did not affect foxj1a expression in the FP (Fig. 6B, asterisk), which explains the presence of GFP fluorescent protein in the FP of Tg(0.6Δfoxj1a:gfp) embryos (Fig. 5H,I) and suggests a tissue-specific function of Wnt.

The zebrafish ear has two types of cilia: short cilia and tethering cilia. Despite the disagreement regarding which one is the motile type, ciliary motility ensures normal ear development (Riley et al., 1997; Colantonio et al., 2009; Yu et al., 2011). Consistent with our findings in KV and PDs, Dkk1 activation did not alter the expression of otic placode precursor pax2 (supplementary material Fig. S6), but abolished the expression of foxj1b (Fig. 6D, arrowhead) and dnah9 (Fig. 6F, white arrowhead) in the otic placode. Moreover, induction of Dkk1 resulted in shorter (2.35±0.36 μm relative to 4.09±0.19 μm in controls; P<1.74×10^–08) and fewer tethering cilia as well as fewer short cilia (Fig. 6I,J,L). In concordance with the cilia defects, otolith assembly was perturbed. In contrast to the two otoliths positioned at the anterior and posterior ends of the otic vesicle by 2 days post-fertilization (dpf) in normal embryos, only one otolith was formed (Fig. 6O, arrowhead). These results define an essential role of Wnt/β-catenin signaling in PD and OV ciliogenesis, which is possibly mediated by foxj1a and foxj1b, respectively. Based on these data, we propose that Wnt-regulated ciliogenesis is a general mechanism during development.

In this paper, we put forward the novel concept that Wnt/β-catenin signaling regulates ciliogenesis via direct regulation of foxj1a transcription. First, we showed that DFC specification and KV morphogenesis are not markedly affected by impaired Wnt signaling, thus arguing that the effect of Wnt signaling on foxj1a expression/cilia formation is not secondary to its other functions, such as posterior fate specification (Schier and Talbot, 2005). Second, we placed Wnt signaling genetically downstream of FGF signaling in the regulation of foxj1a expression. In the future, similar genetic studies can be extended to examine relationships between the Wnt and other signaling pathways that are also implicated in foxj1a expression and/or ciliogenesis, for example, the Notch pathway (Lopes et al., 2010). Third, our finding that the putative Lef1/Tcf binding sites in the foxj1a enhancer are indispensable for foxj1a expression in KV firmly defines a direct transcriptional regulation of foxj1a by Wnt. Our ongoing hypothesis predicts that multiple signaling pathways could regulate ciliogenesis, among which Wnt is responsible for downstream signaling, acting directly at the level of foxj1a transcription. Lastly, although Wnt signaling is essential for ciliogenesis, it is not sufficient to induce ectopic foxj1a expression. Together with the notion that Wnt signaling has a broader function domain than Foxj1a, we also predict that Wnt signaling needs other co-factors to initiate ciliogenesis. Further analysis of the foxj1a enhancer, including searching for crucial cis-acting elements and identification of transcription factors that bind to these elements, will elucidate the transcription circuit that confers Wnt-regulated foxj1a expression.

Our data from several motile cilia-forming tissues suggest a general function of Wnt-Foxj1 signaling in governing cilia formation. It remains to be determined whether this function of Wnt is conserved in mammals. We found that two of the three putative Lef1/Tcf binding sites in the zebrafish foxj1a enhancer are conserved in the human FOXJ1 enhancer (Rawlins et al., 2007; Wang and Ware, 2009), suggesting a conserved function. Wnt3a knockout mice showed normal ciliogenesis in the node (Nakaya et al., 2005), but this does not necessarily indicate a lack of conservation of the Wnt-Foxj1-ciliogenesis cascade in mice. The following issues need to be addressed before any conclusions can be made. Do other components of Wnt signaling have redundant functions in the node? Is Foxj1 downregulated in the node of Wnt3a-deficient mice or other mouse mutants harboring impaired Wnt signaling? Are there other transcription factors that can compensate for the loss of Foxj1? For example, normal ciliogenesis in the FP of Foxj1-null mice was attributed to redundant functions of Rfx3 (regulatory factor binding to the X box), a protein important for cilia biosynthesis and cilia motility (Bonnafe et al., 2004; El Zein et al., 2009; Cruz et al., 2010). It would be interesting to know whether Rfx3 has a similar redundant function in the node. Therefore, more experimentation is required to assess to what extent the Wnt-Foxj1-ciliogenesis cascade is conserved in mammals, and tissue- and species-specificity need to be considered.

The present work uncovers a role of Wnt/β-catenin signaling upstream of ciliogenesis, which is distinct from studies that have focused on the effects of primary cilia on the Wnt/β-catenin pathway. The effects of cilia on the Wnt pathways have been substantially investigated, but there is a lack in consensus (Eggenschwiler and Anderson, 2007; Wallingford and Mitchell, 2011). For example, some studies suggest that cilia execute an inhibitory role on the Wnt/β-catenin pathway (Gerdes et al., 2007; Corbit et al., 2008) and are required for switching between the Wnt/β-catenin and Wnt/PCP pathways (Simons et al., 2005); others suggest a dispensable function of cilia in the Wnt pathways (Huang and Schier, 2009; Ocbina et al., 2009). Our discoveries will open doors for future studies to elaborate on a reciprocal relationship between Wnt activity and cilia formation, which might help to reconcile discrepancies in the roles of cilia in Wnt signal transduction.

The Wnt/PCP pathway has also been implicated in cilia motility. In contrast to the Wnt/β-catenin pathway that transcriptionally regulates foxj1a, PCP effectors and core proteins regulate cell and basal body polarization (Oishi et al., 2006; Park, T. J. et al., 2006; Park et al., 2008; Gray et al., 2009; Mitchell et al., 2009; Borovina et al., 2010; Guirao et al., 2010; Song et al., 2010). In fact, ligands and receptors that participate in these two pathways are colocalized in many ciliated tissues. Using KV as an example, genes encoding components of the Wnt/β-catenin pathway, such as wnt3a, wnt8a and fzd10, and genes encoding components of the Wnt/PCP pathway, such as wnt5b and fzd2, are co-expressed near or within KV (Oishi et al., 2006; Lin and Xu, 2009; Freisinger et al., 2010). Therefore, it is possible that the Wnt/β-catenin pathway governs the expression of genes that are essential for cilia formation, whereas the PCP pathway ensures that these cilia-forming components are assembled at a correctly polarized subcellular location, such as the apical surface. Further investigations are warranted to examine potential interactions between the two pathways in programming polarized cilia beating and subsequent unidirectional fluid flow.

The Wnt/β-catenin pathway is an important developmental pathway that has multifaceted functions during embryogenesis. It is involved in KV, kidney and inner ear development and function (Ohyama et al., 2006; Park et al., 2007; Jayasena et al., 2008; Lancaster et al., 2009; Lin and Xu, 2009; Lyons et al., 2009). However, Wnt signaling does not appear to be essential for FP development despite the presence of Wnt ligands (Placzek and Briscoe, 2005). In keeping with these observations, we found that Wnt is required for foxj1 expression in KV, PDs and OVs but not in FP. Our observations suggest a hypothesis predicting that defective ciliogenesis is an important cellular event that accounts for certain defects in Wnt-related developmental processes and diseases. As foxj1 expression was associated with differentiation in tissues, such as respiratory epithelial cells, radial glia of the brain and sensory hair cells of the inner ear (You et al., 2004; Park, K. S. et al., 2006; Jacquet et al., 2009; Yu et al., 2011), it is tempting to speculate that the Wnt-Foxj1 axis represents a differentiation pathway that drives the committed epithelium towards more terminally differentiated cells with cilia. Whether and how the Wnt-Foxj1 axis interacts with other aspects of Wnt signaling, such as cell-fate stabilization and cell proliferation, remains to be determined. Nevertheless, our findings assigned a novel function to Wnt/β-catenin signaling, which could have a significant impact on subsequent research areas that bridge cilia biology and developmental signaling.

We thank Mayo Clinic Zebrafish Core Facility for the zebrafish husbandry.