Targeted mutation of the talpid3 gene in zebrafish reveals its conserved requirement for ciliogenesis and Hedgehog signalling across the vertebrates

Hedgehog (Hh) signalling plays a crucial role in animal development, controlling cell type specification, proliferation and survival in a variety of contexts through a pathway that has been highly conserved during evolution (reviewed by Ingham et al., 2011; Ingham and Placzek, 2006). Although originally identified through genetic analysis in Drosophila, many of the components that modulate and transduce Hh signals have subsequently been functionally annotated in vertebrates, both through ES cell-mediated targeted mutagenesis and via forward genetic screens in zebrafish and mouse. Unexpectedly, several of the loci identified in the latter screens were found to encode intraflagellar transport (IFT), uncovering a hitherto unsuspected role of the primary cilium in mammalian Hedgehog signalling (Eggenschwiler and Anderson, 2007) that more recent analysis of the zebrafish oval (that encodes the IFT88 protein) and iguana (igu) mutants has shown to be a conserved feature of Hh signalling across the vertebrates (Glazer et al., 2010; Huang and Schier, 2009; Kim et al., 2010; Tay et al., 2010). Consistent with this conservation, the chicken talpid3 (ta3) mutation, originally identified on the basis of its eponymous limb defects (Ede and Kelly, 1964) and later found to disrupt transduction of Sonic hedgehog (Shh) activity in the limb as well as in the neural tube and somites (Davey et al., 2006; Lewis et al., 1999), was recently shown to identify a novel coiled-coil domain protein essential for primary ciliogenesis (Davey et al., 2006; Yin et al., 2009). Surprisingly, however, although the ta3 gene has been highly conserved from Nematostella to human (Yin et al., 2009), no mutant alleles of ta3 have been described in any other organism, despite the extensive screening efforts in mouse and zebrafish. One possible explanation for the failure to identify a zebrafish ta3 mutant could be that maternal expression of the gene masks the effects of its zygotic inactivation. Indeed, although a number of zebrafish IFT mutants have been isolated on the basis of their ciliogenesis defects (Drummond et al., 1998; Lunt et al., 2009; Sun et al., 2004), none of these displayed the severe phenotypes seen in their mouse counterparts owing to the disruption of the Hh pathway. Only when the maternal expression of IFT88 was removed by germline replacement was the full spectrum of its effects on ciliogenesis and Hh signalling revealed (Huang and Schier, 2009).

Here, we describe the identification of the zebrafish ta3 locus and its functional annotation via zinc-finger nuclease-mediated targeted mutagenesis. We show that maternally supplied ta3 product is indeed sufficient to support embryogenesis in the zebrafish but that complete elimination of ta3 function results in a phenotype strikingly similar to that of its chick counterpart. Our study provides further confirmation of the conserved role of the primary cilium in vertebrate Hh signalling and establishes a new and highly tractable model for the post-embryonic analyses of ciliopathies.

Adult fish were maintained on a 14 hour light/10 hour dark cycle at 28°C in the AVA (Singapore) certificated IMCB Zebrafish Facility. Previously described zebrafish strains used were: ptc1^hu1602, ptc2^tj222 (Koudijs et al., 2008); igu^ts294 (Wolff et al., 2004); Tg(eng2a:eGFP)ⁱ²³³ (Maurya et al., 2011); and Tg(fli::GFP) (Lawson and Weinstein, 2002).

A 2220 bp 3′ expressed sequence tag (EST) clone containing the putative zebrafish ta3 cDNA sequence (GenBank reference number XM_001338591.1) was identified by BLAST (Basic Local Alignment Search Tool) query of GenBank using the chick Ta3 cDNA (GenBank reference number NM_001040707.1). Based on the sequence of XM_001338591.1, primers were designed for RACE experiments to identify additional 5′ and 3′ cDNA sequences. Both the 5′ and 3′ RACE experiments were performed using the SMART RACE cDNA amplification kit (BD Biosciences Clontech). The reverse primers complementary to the 5′ region of the EST sequence (5TAr1: 5′-AGTGAGAGGTGTAGACTGGGTCGGCACT-3′) were used to amplify a 3403 bp 5′RACE. Similarly, a forward primer complementary to the 3′ region of the 5′RACE sequence (07Dec3Rf1, 5′-CAGGCTTCTCCCACACCAGTGATTCCAGC-3′) was used to amplify a 2242 bp 3′ RACE. The 4776 bp full-length cDNA (GenBank reference number JN088213) was amplified with JumpStart AccuTag LA DNA polymerase (Sigma) using the forward primer 5′-CCTGAACTACTACTGGACTAGTTACATGTTACTG-3′ (5UTRf5) and the reverse primer 5′-GCTATGAACGTCTCCAGGCACTTAGCAG-3′ (FL-07Dec3Rr2). Besides the full-length cDNA, the forward primers complementary to XM_001338591.1 EST (3TAf3, 5′-TACATCAGTGCTGTCACATGTGCCTGCTG-3′; 2189-FP1, 5′-ACAGAGCGGCGATCCAAACTCAGATG-3′) amplified another two different 3′RACE products, of 389 bp (GenBank reference number JN088214) and 1083 bp (GenBank reference number JN088215), respectively. The three 3′RACE products and the EST (XM_001338591.1) are non-conserved at their 3′ end, indicating the occurrence of alternative splicing.

Zinc-finger nucleases (ZFN) specific for the zebrafish ta3 gene were generated using Oligomerized Pool Engineering (OPEN) (Maeder et al., 2008) with selection using the bacterial one hybrid (B1H) system (Noyes et al., 2008). For the left subunit of the ZFN, OPEN pools 37 and 4 (GGCt), 108 (GGC) and 54 (GTG) were recombined by PCR and assembled in vector 1352omegaUV2 to create a phagemid library. Likewise for the right subunit; OPEN pools 3 and 41 (GGAt), 116 (TGA) and 31 (GTT) were assembled. The libraries each had ∼10⁷ independent clones. Bait plasmids with each required zinc-finger recognition half site sequence were constructed in a version of the pH3U3 vector modified to be chloramphenicol resistant rather than kanamycin resistant. Phage transduction was used to introduce each zinc finger protein (ZFP) library to US0 cells harbouring the respective bait plasmid and ∼2×10⁷ transductants were exposed to selection on 14 cm His selective (NM) plates with either 15 mM or 20 mM 3-AT. The largest looking colonies were re-streaked on NM plates with 20 mM 3-amino-1,2,4-triazole (3-AT). Sequencing of these selected clones and random clones picked prior to selection revealed that selection had induced a marked bias in the zinc-finger recognition helix residues. For each subunit, a clone with a prevalent sequence was chosen for making the ZFP. As a precaution, these chosen plasmids were retransformed into US0 cells with the respective bait plasmids to confirm growth on NM plates with 20 mM 3-AT. These selected zinc-finger domains were sub-cloned into the pCS2FlagFok1DD and pCS2HAFok1RR plasmids (Meng et al., 2008). Capped RNA from each plasmid was produced by in vitro transcription and a range of doses was injected into one-cell stage zebrafish embryos. Embryos injected with ∼100 pg had an appropriate 30% rate of deformity at 24 hpf (hours post-fertilisation). Genomic DNA prepared from non-deformed 24 hpf embryos injected with this dose was used as a PCR template to analyse potential somatic mutations. The site of the ZFN target overlaps an NlaIII endonuclease site. Amplicons from injected embryos were partially resistant to NlaIII cleavage while the amplicons from uninjected embryos were fully cleaved. Roche Titanium 454 amplicon sequencing showed that 108 out of 693 (16%) of the amplicon molecules had insertions or deletions at the target site. G0 adults derived from embryos injected with ZFN capped RNA were in-crossed and their progenies (G1) individually genotyped by PCR using the forward primer (5′-TGGCATGTCAGCAGAAAAACA-3′) and reverse primer (5′-TGGACTTTCTGCACTCCCTTGC-3′) followed by Sanger sequencing.

Donor embryos from a ta3ⁱ²⁶⁵/+ × ta3ⁱ²⁶²/+ intercross were labelled by injecting GFPnanos3′ UTR RNA (Koprunner et al., 2001). Wild-type host embryos were injected with morpholino antisense oligonucleotide against dead end (5′-GCTGGGCATCCATGTCTCCGACCAT-3′) to block germ cell development (Weidinger et al., 2003). Cells were transplanted from donors (1000 cell stage) into the hosts (dome stage). Donor embryos were genotyped by PCR using the primers described above. Transplanted host embryos were screened at 24 hpf for successful transfer of donor germ cells as indicated by GFP expression (Ciruna et al., 2002).

Standard in situ hybridization was performed with anti-Dig alkaline phosphatase and chromogenic substrate NBT/BCIP as previously described (Oxtoby and Jowett, 1993). RNA probe used for in situ hybridization of ta3 was synthesized from the 4779 full-length template and sheared into 500 bp fragments with 60 mM Na₂CO₃/40 mM NaHCO₃. RNA probes for other genes were prepared from templates as previously described: ptc2 (formerly ptc1) (Barth and Wilson, 1995; Concordet et al., 1996), nkx2.2 (Barth and Wilson, 1995; Concordet et al., 1996), L-fabp and trypsin (Lo et al., 2003).

Whole-mount antibody staining was performed as previously described at the following dilutions: mAb 4D9 (anti-Engrailed; DHSB) at 1:50-1:200; rabbit anti-Prox1 at 1:5000 (Elworthy et al., 2008); rabbit anti-γ-tubulin at 1:500 (Sigma); mouse anti-γ-tubulin at 1:500 (Sigma); and mouse anti-acetylated α-tubulin at 1:800 (Sigma). The secondary antibodies were: Alexa488-conjugated goat anti-mouse or anti-rabbit, Alexa546-conjugated goat anti-mouse and Alexa568-conjugated goat anti-rabbit, Alexa633-conjugated goat anti-secondary antibodies (1:1000, Invitrogen). Bright-field microscopy images were acquired with an AxioCam HRc mounted on a Zeiss AXIO Imager M2, Olympus DP70 on MVX10 or Leica DFC300 FX mounted on MZ16FA. Fluorescent specimens were imaged using the 60× or 100× oil immersion objective on an Olympus Fluoview 1000 confocal microscope. Images were acquired using Olympus FV10-ASW software.

The full-length coding region of the zebrafish ta3 cDNA was cloned as a ClaI/XhoI fragment immediately 3′ to the coding region of EGFP inserted into the BamHI/ClaI sites in the polylinker of pCS2+ (Turner and Weintraub, 1994) to generate a C-terminal in-frame fusion with EGFP. The plasmid was linearized with SacII. pCS2-GFP-Kif7 (Tay et al., 2005) and pCS2-GFP-Dzip1 (Kim et al., 2010) were linearized with NotI. Capped synthetic mRNAs were transcribed in vitro with SP6 mMessage mMachine Kit (Ambion) and injected into fertilized eggs. BAC Gli2a-GFP was used as previously reported (Kim et al., 2010).

Cyclopamine treatment followed a standard method with immersion in 40 μM cyclopamine (Toronto Research Chemicals) from the bud stage as previously described (Wolff et al., 2003).

Embryos were dechorionated, deyolked and homogenized manually in ice-cold PBS without Ca²⁺ and Mg²⁺ in the presence of complete protease inhibitor cocktail (Roche). The embryo pellet was lysed in RIPA buffer [50 mM Tris HCl (pH 8.0), 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, protease inhibitor cocktail and 1 mM PMSF]. Samples were microcentrifuged for 10 minutes at 4°C, loading buffer [62.6 mM Tris HCl (pH 6.8), 2% SDS, 0.01% bromophenol blue, 10% glycerol and 100 mM DTT] was added to the supernatant and the equivalent of 30 embryos run on each lane of a 7.5% acrylamide denaturing gel at 30 mA for 120 minutes, and electroblotted onto Immobilon-P polyvinylidene fluoride (PVDF) membrane (Millipore). PVDF strips were blocked in 5% milk powder PBS 0.1% Tween20 for 1 hour, and incubated with rabbit anti-zebrafish Gli2a (1:5000) (Maurya et al., 2011) for 1 hour at room temperature. After washing, primary antibody was detected with ECL HRP-conjugated anti-rabbit lgG (1:50,000). Chemiluminescent Substrate was SuperSignal West Femto (Pierce). The loading amount of protein extract among specimens was evaluated by β-actin level with rabbit anti-β-actin (1:5000; Cell Signalling). Signal quantification was performed using Adobe Photoshop software.

We amplified a 4.7 kb full-length cDNA (see Materials and methods) predicted to encode a 1554 amino acid protein (supplementary material Fig. S1) with 33/50%, 31/45%, 33/49% and 35/51% sequence identity/similarity to the chicken, human, mouse and Xenopus Ta3 proteins, respectively. The coding sequence of the zebrafish ta3 gene consists of 31 exons, conceptual translation of which predicts an encoded protein with five coiled-coiled motifs (Fig. 1B; supplementary material Fig. S2). The genomic sequence is located on linkage group 17 (LG17) in a syntenic region conserved between zebrafish, Xenopus, chick, mouse and human (Fig. 1A).

In situ hybridization revealed that newly fertilized eggs contain abundant maternally derived ta3 mRNA. The ta3 transcript is ubiquitously distributed in embryos from early cleavage stages through to 2 dpf (Fig. 1C,D). To investigate the subcellular distribution of the Ta3 protein in zebrafish embryos, we generated a construct encoding the Ta3 protein fused at its N terminus to EGFP (see Materials and methods) and injected mRNA transcribed from this construct into newly fertilized eggs. The EGFP-Ta3 fusion protein localized to the basal body and to its daughter centriole in somitic cells of wild-type embryos (Fig. 1E). Localization to the basal body was unaffected by the loss of wild-type Dzip1 protein in iguana mutants (Fig. 1F).

To analyse the function of talpid3 in the zebrafish embryo, we initially used the well-established approach of morpholino oligonucelotide mediated gene knockdown (Eisen and Smith, 2008). Morpholinos designed to block either translation or splicing of the ta3 transcript (supplementary material Fig. S3) were injected into newly fertilized embryos following standard protocols and the injected embryos were fixed and assayed for expression of Prox1 and Engrailed proteins, two sensitive read-outs of Hh pathway activity in the zebrafish myotome (Wolff et al., 2003), and for acetylated α-tubulin, a marker for cilia. Using three different morpholinos at varying concentrations, we failed to observe any reproducible changes in Prox1 or Eng expression or any cilium defects (data not shown).

Accordingly, we adopted a different strategy to inactivate ta3 function, namely the generation of stable germ line transmissible mutant alleles using zinc-finger nuclease (ZFN)-mediated targeted mutagenesis (Doyon et al., 2008; Foley et al., 2009; Meng et al., 2008). The online ZiFiT software was used to identify sequences in the ta3-coding region that are potentially amenable to targeted mutagenesis using the OPEN ZFN method (Maeder et al., 2008). Targeting this sequence (aGCCGCCCACgtccatGTTTGAGGAt) in the eleventh coding exon had the potential to create mutations at Thr 553 (Fig. 2A) in a region conserved with the chick Ta3 protein. This site was chosen in preference to others (including some closer to the N terminus) based on the notion that OPEN zinc-finger nucleases with first and third fingers directed against GNN recognition sites and middle fingers directed against either G/TNN recognition sites might be more effective (Ramirez et al., 2008). Embryos injected with mRNA encoding a ZFN pair selected for recognition of this sequence were found to carry a variety of deletions or insertions at the target site (data not shown) confirming the efficacy of this approach. Accordingly, we grew up adults from injected embryos and screened their progeny for similar lesions in ta3. In this way, we recovered a number of individuals transmitting insertion or deletion mutations, several of which cause frameshifts resulting in premature termination codons (Table 1) predicted to yield proteins truncated N terminal to the fourth coiled-coil domain.

Animals trans-heterozygous for ta3 mutant alleles predicted to encode truncated proteins completed embryogenesis and showed no defects in Eng or Prox1 expression in the myotome, nor any other manifestations of aberrant Hh pathway activity (data not shown). Consistent with this, the growth and morphology of the primary cilia appeared normal. By 4 dpf, however, some mutant larvae displayed dilation of the pronephric tubules and edema (Fig. 3B). Confocal analysis revealed that cilia became sparsely distributed within the pronephric tubules, which displayed an apparent overproliferation of epithelial cells (Fig. 3D). There was no evidence of body curvature typical of other zebrafish cilia mutants (Drummond et al., 1998; Lunt et al., 2009; Pathak et al., 2007; Sun et al., 2004). Most ta3 trans heterozygotes died before 14 dpf, although some could survive for up to 4-5 weeks. By contrast, animals homozygous for the ta3ⁱ²⁶⁷ allele, which causes a small 81 nucleotide in-frame deletion, were almost fully adult viable (Fig. 3E).

We surmised that the lack of a mutant phenotype in trans heterozygous ta3 embryos could be due to the significant levels of maternally derived ta3 mRNA present in newly fertilized eggs. To test this inference, we used the germ cell replacement technique (Ciruna et al., 2002) to generate chimeric females with ta3 mutant germ lines (see Materials and methods for details) and crossed these to ta3ⁱ²⁶⁴/+ heterozygous males. The resulting MZ (maternal + zygotic) ta3 mutant embryos exhibited a curled body and ectopic lens phenotype at 1 dpf (Fig. 4A,C). The domain of Eng expression was significantly expanded among the fast muscle fibres by 28 hpf, indicative of an increased number of Hedgehog-dependent MFFs (Fig. 4E,G). At the same stage, transcription of ptc2 was upregulated in the somites but downregulated in the neural tube, whereas expression of nkx2.2 was almost eliminated from the ventral neural tube (Fig. 4I,M). Treatment with the Smo antagonist cyclopamine had no effect on ptc2 or nkx2.2 expression in MZta3 embryos (Fig. 4K and data not shown), indicating that transduction of the Hh signal is disrupted downstream of Smo. By 2 dpf, the embryos exhibited mild cyclopia (Fig. 5A,B). Expression of the fli::GFP reporter revealed mis-branching and merging of the cerebellar central artery, which is associated with hindbrain haemorrhage, in over 70% of embryos (Fig. 5F,G). The pectoral fins of 4 dpf larvae were symmetric (Fig. 5D), reminiscent of the polydactylous limbs typical of the chick ta3 mutant. This phenotypic spectrum is very similar to that of iguana homozygotes (Sekimizu et al., 2004; Wolff et al., 2004) and of MZift88 mutant embryos (Huang and Schier, 2009). The MZta3 mutants died at 4 dpf. By contrast, ta3/+ heterozygotes derived from ta3 homozygous mutant oocytes (designated Mta3) showed no evidence of aberrant Hh pathway activity, their patterns of Hh-target gene expression being indistinguishable from wild-type embryos.

Confocal analysis of MZta3 embryos stained with anti-acetylated tubulin revealed a complete absence of both motile and primary cilia (Fig. 6, Fig. 7D). The loss of primary cilia is consistent with the evidence of aberrant Hh pathway activity in MZta3 embryos described in the preceding section. In wild-type embryos, GFP-tagged Gli2a has been shown to accumulate at the tip of primary cilia in response to Hh pathway activation, whereas similarly tagged DZIP1 protein encoded by igu accumulated at the basal body (Kim et al., 2010) (Fig. 6G,M). Using the same GFP fusion constructs, we investigated the localization of both proteins in the absence of primary cilia in MZta3 mutants. The basal body localization of DZIP1-GFP was unchanged (Fig. 6J), whereas Gli2a-GFP now also localized to the basal body (Fig. 6H). Similarly, GFP-Kif7, which, in wild-type embryos, was enriched along the axoneme or at the tip of primary cilium (Fig. 6K), also accumulated at the basal body in the absence of the cilium in MZta3 mutants (Fig. 6L).

In zygotically rescued Mta3 embryos, primary cilia and motile cilia in the pronephric duct developed relatively normally; however, the number and length of motile cilia present in Kupffer’s vesicle varied significantly compared with wild type (Fig. 7B,C). To assess the effects of these abnormalities, we examined laterality by the orientation of two asymmetrically located endodermal organs, the liver and exocrine pancreas (Chen et al., 2001), using in situ hybridization for l-fabp and trypsin respectively. Compared with wild type in which just 2.7% of larvae showed laterality reversal, such reversals were observed in 24.7% of Mta3 embryos. Among MZta3 embryos, the endodermal organs were absent in 51.7% of individuals, but when present showed randomized orientation (Fig. 7E-J). The absence of endodermal organs is consistent with a disruption of Hh pathway activity.

Several lines of evidence have implicated the primary cilium as the processing hub for Gli proteins. To explore this role further, we analysed Gli2a processing by western blot analysis of protein extracts from wild-type and MZta3 mutant embryos at 28 hpf. We used a polyclonal antibody raised against a region N-terminal to the cleavage site of the zebrafish Gli2a protein (Maurya et al., 2011), which therefore can detect both the full-length (activator) form (Gli2aFL) and the truncated repressor form (Gli2aR). Extracts from wild-type embryos showed two prominent bands when probed with this antibody, one of around 190 kDa, corresponding to Gli2aFL, and one of around 48 kDa, corresponding to Gli2aR (Fig. 8, lane 3). The levels of the larger band were significantly diminished whereas those of the smaller band were significantly increased in embryos treated with cyclopamine (Fig. 8, lane 1) such that the FL:R ratio was reduced from 0.45 typical of wild type to 0.05. Conversely, the presumed Gli2aR band was significantly diminished, whereas the intensity of the Gli2aFL band increased slightly in ptc1^hu1602; ptc2^tj222 double mutant embryos (Fig. 8, lane 2) the FL:R ratio increasing fourfold relative to wild type. These findings are consistent with a major role for HH in the modulation of Gli2a processing. In MZta3 extracts, the level of Gli2aR was slightly reduced compared with that of wild-type extracts, whereas the level of the Gl2aFL form was slightly enhanced (Fig. 8, lane 4), such that the FL:R ratio increased to 0.6; in addition, multiple forms of intermediate size could now be detected. In contrast to wild-type embryos, cyclopamine treatment of MZta3 embryos caused only a slight reduction in the FL:R ratio (Fig. 8, lane 6); interestingly, the FL:R in Mta3 embryos was also only slightly reduced after cyclopamine treatment (Fig. 8, lane 7).

Based on sequence comparison and synteny analysis, we have identified the zebrafish orthologue of the chicken ta3 gene, previously shown to encode a centrosomal protein that is essential for primary ciliogenesis (Yin et al., 2009). Consistent with these properties, we found that a zebrafish EGFP-Ta3 fusion protein localizes both to the basal body and to the daughter centriole in zebrafish embryonic cells. Our initial attempts to analyse ta3 function in the zebrafish using morpholino antisense oligonucleotides proved unsuccessful, highlighting the limitations of this approach (Eisen and Smith, 2008). To circumvent this problem, we used a previously described targeted mutagenesis approach (Doyon et al., 2008; Foley et al., 2009; Meng et al., 2008) to generate stable germline transmissible alleles of the zebrafish ta3 gene. Zinc-finger nucleases that recognise sequences close to the 5′ end of the gene were selected using the OPEN protocol. These reagents proved highly mutagenic when injected into zebrafish embryos, generating deletion and insertion mutations at high frequency. In contrast to their chicken counterpart, however, these zebrafish ta3 mutations proved to be embryonic viable, perhaps explaining why the ta3 locus had not previously been identified in the large-scale forward genetic screens performed in this species (Haffter and Nusslein-Volhard, 1996). By contrast, elimination of both zygotic and maternal ta3 activity achieved by germline replacement, resulted in a complex phenotype very similar to that of the chicken ta3 mutant, (Ede et al., 1974; Ede and Kelly, 1964; Head et al., 1992; Hinchliffe and Thorogood, 1974; Lee and Ede, 1989; Lewis et al., 1999), including defects in the neural tube, somites, limbs and craniofacial structures. In the neural tube, the loss of expression of the ventral marker nkx2.2 precisely mirrors the chick mutant phenotype and is indicative of a loss of Hh activity, as is the absence of liver and pancreas (Roy et al., 2001), that was seen in a significant proportion of mutant embryos; conversely, in the myotome, expansion of the Eng expression domain is consistent with a gain of Hh signalling (Wolff et al., 2003) and in line with this, we also found a significant expansion of the ptc2 expression domain in the somites. This latter effect mirrors the expansion of ptc2 expression throughout the somites of chick ta3 embryos, although we note that these mutants also exhibit a significant downregulation of ptc2 in the medial somite (Davey et al., 2006), an effect indicative of a reduction in Hh pathway activity.

The MZta3 phenotype is strikingly similar to that of mutant alleles of the iguana (igu) gene, the product of which, DZIP1, is another coiled-coil domain protein (Sekimizu et al., 2004; Wolff et al., 2004) also required for primary ciliogenesis (Glazer et al., 2010; Kim et al., 2010; Tay et al., 2010). In contrast to Ta3, DZIP1 protein localizes uniquely to the basal body and not to the daughter centriole (Kim et al., 2010; Tay et al., 2010); we found this localization to be unaffected in MZta3 mutant embryos, arguing against a direct interaction between the two proteins. Likewise, and less surprisingly, the basal body localization of Ta3 appears independent of DZIP1. Although our analysis has focused on the effects of elimination of maternally derived ta3 gene function on embryonic development, we note that it also demonstrates that ta3 activity in the germ line is dispensable for oogenesis. We also generated males carrying ta3 mutant germ lines; although these displayed reduced fertility, they were able to give rise to viable progeny at low frequency, in contrast to the reported sterility of chimeric ovl^{tz288b/tz288b} (ift88) males (Huang and Schier, 2009).

The response of cells to Hh signalling is mediated by the Gli2 and Gli3 proteins, both of which are characterized by an N-terminal repressor domain and a C-terminal activator domain, giving them the potential to act as bi-functional transcription factors. Various lines of evidence suggest that in mammals, Gli2 acts principally as an activator (Bai and Joyner, 2001), whereas Gli3 functions both as a repressor and activator (Wang et al., 2000; Motoyama et al., 2003). In line with this, Hh abrogates the efficient processing of Gli3 to its truncated repressor form that occurs in unstimulated cells (Litingtung et al., 2002; Wang et al., 2000), while Gli2 protein is only inefficiently processed even in the absence of Hh activity (Pan et al., 2006). Previous studies of chicken ta3 function investigated the processing of Gli3 in mutant embryos and found it to be significantly abrogated, resulting in a marked increase in the levels of the full-length form of the protein relative to its truncated repressor form (Davey et al., 2006). These findings are consistent with the Hh gain-of-function aspects of the ta3 mutant phenotype, though are less easy to reconcile with the loss of Hh-dependent gene expression seen in the neural tube.

Zebrafish have a single Gli3 protein but two Gli2 proteins, designated Gli2a and Gli2b, the former appearing to be the orthologue of mammalian and avian Gli2 based on sequence comparisons and conserved synteny (Karlstrom et al., 1999). Notably, the expression of Gli2b and Gli3 appears largely confined to the neurectoderm and its derivatives (Tyurina et al., 2005; Ke et al., 2008), whereas Gli2a is expressed more widely in mesodermal and ectodermal derivatives (Karlstrom et al., 2003). Moreover, functional analyses suggest that, in zebrafish, both Gli3 and Gli2 act as repressors and activators; although no reagents are available to study zebrafish Gli3 protein processing, we found a significant proportion of the Gli2a protein to be in the truncated ‘repressor’ form, a proportion that was markedly increased upon treatment of embryos with the Smo inhibitor cyclopamine and markedly decreased in the absence of Ptc activity. These findings suggest that, in zebrafish, modulation of Gli2a processing is an important mode of control of its activity by Hh signalling. In this context it is interesting that the so-called processing determinant domain (PDD) of Gli2 identified by Pan and Wang (Pan and Wang, 2007) shows 84% sequence identity between mouse and human but only 52% identity between zebrafish and mouse (supplementary material Fig. S4). Whether or not this divergence in sequence is sufficient to confer the differing Gli2 processing profiles observed between mouse and fish remains to be determined.

In contrast to the major abrogation of Gli3 processing seen in chick ta3 mutants (Davey et al., 2006), we found the reduction in Gli2aR levels in zebrafish MZta3 to be less dramatic. At the same time, we observed a slight increase in the levels of full-length Gli2a, as well as the appearance of bands of intermediate size, that are not detected at significant levels in wild-type embryos or in the absence of Smo or Ptc activity. One limitation of our analysis is that it was performed on whole-embryos extracts, making it impossible to relate the changes in Gli2a species to the contrasting effects on Hh target gene expression in different tissues. Moreover, in order to identify mutant embryos unequivocally on the basis of their phenotype, we were limited to analyzing Gli2a at 28 hpf, which is several hours after the crucial period for muscle cell type specification (Wolff et al., 2003). Nevertheless, the modest increase in Gli2a FL;R ratio is consistent with the apparent gain-of-function Hh phenotypes observed in MZta3 mutants. The fact that the pattern of processing was largely unaltered in response to cyclopamine treatment is in line with the failure of cyclopamine to modify the MZta3 phenotype and underlines the importance of the primary cilium as a hub for the modulation of Gli processing by Hh. Interestingly, we found that a GFP-tagged form of Gli2a that normally shuttles to the tip of the primary cilium in response to Hh activity (Kim et al., 2010) localizes to the basal body in MZta3 mutants. Similarly, we found that GFP-tagged Kif7 also localizes to the basal body in the absence of the primary cilium, consistent with the notion that Kif7 sequesters Gli proteins at the basal body in the absence of Hh activity. We note, however, that the FL form of Gli2a also appeared resistant to cyclopamine treatment in Mta3 embryos, despite their having normal primary cilia and showing sensitivity to cyclopamine treatment at the level of Hh gene expression.

Disruption of left-right (LR) asymmetry is one aspect of the zebrafish MZta3 phenotype that is not shared with the chick mutant (Davey et al., 2006), though a similar effect has recently been described in mouse ta3 homozygotes (Bangs et al., 2011). Previous studies have shown that ectopic Shh activity can cause reversals of L-R patterning, as indicated by changes in laterality of the heart and endodermal organs (Levin et al., 1995; Schilling et al., 1999). Thus, the effects seen in MZta3 embryos could reflect the partial de-repression of Hh pathway activity due to the loss of primary cilia. Significantly, however, we observed LR abnormalities in zygotically rescued embryos derived from ta3 mutant oocytes (Mta3) in which there is no evidence of aberrant Hh pathway activity but there is depletion of the motile cilia in Kupffer’s vesicle. Given the implication of nodal flow in establishing LR asymmetry (Hirokawa et al., 2006), we consider it most likely to be the disruption of motile ciliogenesis that underlies this particular aspect of the ta3 phenotype. With the exception of this requirement, which probably reflects the precocious differentiation of motile cilia in Kupffer’s vesicle, maternally derived ta3 product appears dispensable for normal development; at the same time, the zygotic ta3 phenotype indicates that maternally derived Ta3 is sufficient to support embryogenesis, as ta3 homozygotes first manifest a mutant phenotype, cystic kidneys, only during early larval stages. A similar phenotype has also been reported to develop in chick ta3 mutants at 7 days (Yin et al., 2009) and has also been described in a number of different zebrafish mutants, most of which have been found to disrupt genes encoding either centrosomal or IFT proteins (Drummond et al., 1998; Lunt et al., 2009; Pathak et al., 2007; Sun et al., 2004). In contrast to ta3, however, these zebrafish cystic kidney mutants have other phenotypic effects that manifest in the embryo leading to early lethality. In this respect, the late larval lethality of the ta3 mutant makes it an especially useful model for PKD and potentially for other ciliopathies.

We thank Iain Drummond for helpful suggestions, Sudipto Roy for the Kif7-GFP construct, Ashish Maurya for providing the eng2a:GFP reporter line, Cheryll Tickle for communicating data ahead of publication, and colleagues at ICMB and the MRC CDBG for their interest and advice.