Fine-tuning of Hh signaling by the RNA-binding protein Quaking to control muscle development

In zebrafish, motility is initially induced by the contraction of segmental muscles that are located in repeated units – the somites – along each side of the body through the action of motoneurons. Each somite is V-shaped and contains predominantly two differentiated muscle cell types: the superficial slow-twitch fibers and the deep fast-twitch fibers (Devoto et al., 1996; Roy et al., 2001). Slow muscle precursors initially abut the notochord and derive from the so-called adaxial cells. Slow muscle precursors differentiate and elongate along the main body axis to span the entire length of the somite. They then undertake a lateral migration through the fast muscle precursors. During their migration, the slow muscle precursors are thought to trigger maturation of the fast muscle precursors (Henry and Amacher, 2004). These latter precursors elongate to reach the somite boundary and, once this process is completed, the fast muscle cells fuse to create syncytial myotubes (Snow et al., 2008) by an as yet uncharacterized mechanism (Krauss, 2007; Moore et al., 2007; Srinivas et al., 2007).

Within each somite, Hedgehog (Hh) activity emanating from the notochord induces the differentiation of the medially adjacent paraxial mesodermal cells, called adaxial cells, into slow-twitch muscle fibers (Barresi et al., 2000; Blagden et al., 1997; Coutelle et al., 2001; Currie and Ingham, 1996; Currie and Ingham, 1998; Du et al., 1997; Hirsinger et al., 2004; Lewis et al., 1999a; Lewis et al., 1999b; Roy et al., 2001). Overexpression of Hh leads to an excess of slow muscle cells, whereas inhibition of the pathway in the smo mutant, which lacks the Hh receptor Smoothened, prevents the differentiation of adaxial cells into their derivatives (Barresi et al., 2000; Hirsinger et al., 2004). Depending on the level of Hh signaling, adaxial cells give rise to two subpopulations: slow muscle cells (low level) and a specialized group of Engrailed-positive muscle cells termed the muscle pioneers (MPs; high level). The MPs populate the medial part of the somite known as the horizontal myoseptum, which separates the ventral from the dorsal half of the somite, and their presence is tightly correlated with the V-shape of the somite (Ekker et al., 1992; Hatta et al., 1991; Roy et al., 2001; Wolff et al., 2003); U-shaped somites often lack MPs. Later, an intermediate level of Hh activity induces the medial fast fibers (MFFs) (Roy et al., 2001), a subpopulation of paraxial mesodermal cells that surround the MPs and also express Engrailed (Wolff et al., 2003).

The activity of the secreted Hh proteins is mediated by their binding to the large seven-multipass transmembrane protein Patched 1 (Ptc1, or Ptch1) (Concordet et al., 1996). Once bound to Hh, Ptc1 ceases to inhibit the twelve-transmembrane receptor Smo (Chen et al., 2001), inducing Hh signaling transduction. The outcome of this pathway is the absence of cleavage and the activation of the Gli family of zinc-finger transcriptional effectors. Gli enters the nucleus to activate the transcription of various target genes, including ptc1, and to specify the different muscle cell types (Concordet et al., 1996). As the level and the timing of Hh signaling are critical for the specification of the different muscle cell types, somite morphogenesis is a complex process that requires fine-tuning. Recent studies have revealed a role for miRNA in the control of the Hh pathway, suggesting a level of post-transcriptional regulation, but this area remains largely unexplored (Flynt et al., 2007).

The post-transcriptional regulation of gene expression plays a fundamental role in the temporal and spatial modulation of developmental processes. Various molecules are involved in this fine-level regulation in addition to miRNAs, including RNA-binding proteins. Within this latter category, Quaking proteins belong to the signal transduction and activation RNA (STAR) family of RNA-binding proteins (Vernet and Artzt, 1997). Their action is mediated by their KH domain, a prevalent RNA-binding domain that recognizes and binds a specific sequence in the 3′UTR of the target mRNAs. STAR proteins function in pre-RNA splicing (Arning et al., 1996; Chawla et al., 2009; Matter et al., 2002; Paronetto et al., 2007; Stoss et al., 2004), mRNA export (Coyle et al., 2003; Larocque et al., 2002; Reddy et al., 2002), mRNA stability (Larocque et al., 2005; Nabel-Rosen et al., 1999) and protein translation (Jan et al., 1999; Lee and Schedl, 2001; Lee and Schedl, 2004; Saccomanno et al., 1999; Schumacher et al., 2005). STAR family members have been identified throughout the evolution of the metazoan lineage and regulate various processes of cell differentiation, such as myelination in mice or tendon differentiation in flies (Ebersole et al., 1996; Nabel-Rosen et al., 1999; Nabel-Rosen et al., 2005; Nabel-Rosen et al., 2002; Sidman et al., 1964; Volk et al., 2008).

In this study, we report the isolation of a new zebrafish mutation, t31954, that affects motility. Within the somite, this mutation leads to delayed maturation of fast muscle and specification problems for the Hh-induced muscle cells. Molecular analysis shows that allele t31954 carries a loss-of-function mutation in quaking A (qkA), a zebrafish STAR family member that is highly related to vertebrate Quaking. We provide evidence that Hh signaling is downregulated in the somitic territory when qkA function is lost. qkA^t31954 homozygous mutants are still responsive to overactivation of Hh signaling but the pattern of muscle cell type induction is modified, suggesting that qkA acts downstream of the Hh signaling pathway to regulate muscle cell type specification. By in silico analysis, we find that gli2a mRNA carries a Quaking response element (QRE) within its 3′UTR. Using a sensor assay, we show that this QRE is essential for gli2a translation and mRNA stability. Consistent with this notion, the loss of qkA function rescues slow muscle fibers in yot mutant embryos, which express a dominant-negative Gli2a isoform. Our results thus uncover a novel role for Quaking as a fine-level post-transcriptional regulator of the Hh signaling pathway that is essential for the proper specification of the various muscle cell types.

The qkA^t31954 allele was isolated in a large-scale ENU mutagenesis screen organized by the ZF-MODELS EU project. Embryos were obtained from natural crosses of either wild-type (WT) or heterozygous yot^ty119/+ or qkA^t31954/+ mutant fish and staged as previously described (Kimmel et al., 1995; Westerfield, 1993).

Heterozygous qkA^t31954/+ mutant fish were outcrossed to the polymorphic WIK strain for mapping. Approximate mapping was performed on 48 mutants and siblings from a single F1 cross. The mutation was first mapped to LG17 by bulk segregant analysis. Primer sequences for simple sequence polymorphisms (SSLPs) were obtained from the Boston Massachusetts General Hospital (MGH) zebrafish and ZFIN databases (Fig. 2A). Embryonic genomic DNA from individual zebrafish larvae (n=442 embryos) was prepared as described (Nüsslein-Volhard and Dahm, 2002). PCR products were electrophoresed on 2% MetaPhor plus 1% ultra pure agarose (TEBU-bio) gels.

RNA was extracted from 24-hpf embryos with TRIzol (Molecular Probes), then reverse transcribed with the AccuScript High Fidelity First-Strand cDNA Synthesis Kit (Stratagene). The cDNA was PCR amplified using the qkA_forward 5′-AAAGAGCACGCCAGACAGAGAC-3′ (5′UTR) and qkA_reverse 5′-GGCGGCTCATCATTGGGTGAAG-3′ (3′UTR) primers. cDNA products were cloned into pSPE3-Rfa. prdm1a was amplified with prdm1a_Fwd 5′-ctactcgagACTAGCAGTGGGACAAGACAGG-3′ and prdm1a_Rev 5′-ctatctagaCGGCCTACATTAAAAATCAACC-3′ and cloned into pBluescript SK⁻ (Stratagene).

Genomic DNA from single qkA^t embryos was amplified by PCR using primers 5′-CAGAGTTTCCACACTGTTAGTGG-3′ and 5′-ACATCAAAGTACAAGCCTCAAGC-3′. PCR products were digested with BclI and resolved on a 3% agarose gel. yot^ty119 genotyping was performed as described (Vanderlaan et al., 2005).

Capped mRNAs were synthesized from pSPE3-Rfa and pCS2⁺ linearized with KpnI and NotI, respectively, and transcribed with the T7 (qkA) and SP6 (GFP and kaede) RNA polymerases (Ambion) then injected at 50 ng/μl (GFP), 10 ng/μl (kaede) or 100 ng/μl (qkA). Morpholinos (MOs) were obtained from Gene Tools and included: qkA exon 1-intron 1 splicing site MO (qkA^gt-MO, 5′-TCAGATTTACACACACATACCTTCG-3′); qkA translation-blocking MO (qkA^ATG-MO, 5′-CCTTCACCTCCATCTCCCCGACCAT-3′); and gli2a translation-blocking MO (gli2a^ATG-MO, 5′-GAGGTGGGACTTGTGGTCTCCATGA-3′). To ensure local MO delivery (Fig. 3), MO was injected together with GFP RNA into one cell of the 8-cell embryo and sorted at gastrulation, selecting embryos with labeled clones restricted to the margin.

To target muscle cells, ~20-50 labeled cells were removed from WT donors (sphere stage) and transplanted to the margin of unlabeled hosts according to Hirsinger et al. (Hirsinger et al., 2004). Genotypes of donor and/or host embryos derived from heterozygous intercrosses were determined by PCR and/or by fast startle escape-touch response.

Probe synthesis and in situ hybridization were performed following standard protocols (Hauptmann and Gerster, 1994).

Immunohistochemistry was performed as follows on at least 50 mutant and WT sibling embryos. In brief, embryos were fixed in 4% paraformaldehyde (PFA) in PBS and stored in methanol at −20°C. After progressive rehydration, 24- or 48 hpf-embryos were incubated in proteinase K (10 μg/ml) for 15 or 60 minutes, respectively, washed three times with PBS containing 0.1% Tween 20 and post-fixed with 4% PFA for 20 minutes at room temperature. After four washes, blocking with 1% DMSO, 1% BSA, 10% normal goat serum in PBS containing 1% Triton X-100 for 1 hour was followed by incubation with primary antibodies overnight at 4°C, then incubation with secondary antibodies at room temperature for 2.5 hours. Primary antibodies were: F59 (1/100), F310 (1/100), MF20 (1/100), Pax7 (1/100) and 4D9 (1/50) (all DSHB); Prox1 (Reliatech, 1/200); and GFP (Torrey-Pines, 1/1000). Secondary antibodies included Alexa Fluor 488- or 594-conjugated anti-mouse IgG1 or anti-mouse IgG2b and Alexa Fluor 488- or 568-conjugated anti-rabbit (Invitrogen) at 1/1000. Specimens were visualized by laser-scanning inverted confocal microscopy.

The full 3′UTRs of wnt11 and gli2a were fused to the 3′ end of d2egfp using XhoI and XbaI (Giraldez et al., 2005). Sequences were amplified from genomic DNA with the following primers and cloned into double-digested pCS2+d2egfp: 3′UTR-wnt11_Fwd 5′-CTACTCGAGCAGACCGTCTTCACCAATAGACCT-3′ and 3′UTR-wnt11_Rev 5′-CTATCTAGACAGGCTGAAATGCATGGCTTTCAT-3′; 3′UTR-gli2a_Fwd 5′-CTACTCGAGTAATTGAAGGCATTACTCTA-3′ and 3′UTR-gli2a_Rev 5′-CTATCTAGAGGTCATACAGTGAAAGTCAA-3′. The resulting constructs were linearized by NotI and then RNAs transcribed using SP6 RNA polymerase (Ambion). RNAs were co-injected at the one-cell stage with kaede mRNA. Individual embryos were prepared for western blot analysis (trunk) and genotyping (head).

Each sample was incubated with RIPA solution (Flynt et al., 2007) containing protease inhibitors and then an equal volume of Laemmli 2× loading buffer containing β-mercaptoethanol was added. After genotyping, three embryos were pooled and resolved on a 10% SDS-polyacrylamide gel, blotted onto a Hybond ECL membrane (GE Healthcare) and blocked with 5% non-fat dried milk in TBS (Hauptmann and Gerster, 1994) containing 0.1% Tween 20. Immunodetection was carried out using rabbit anti-GFP antibody (Torrey Pines, 1/5000-1/10,000), anti-Kaede (MBL, 1/1000) and anti-Actin (Sigma, 1/1000) as primary antibodies and HRP-conjugated anti-rabbit secondary antibody (Roche, 1/2000). Detection was carried out using the Lumi-Light chemiluminescence reagent (Roche) and ChemiCapt apparatus (Vilber Lourmat). The intensities of all bands were normalized to Kaede signal and Actin levels (three independent experiments).

The recessive lethal embryonic mutation t31954 exhibits impaired motility at 24 hours post-fertilization (hpf). This feature was used to distinguish mutant embryos from WT siblings and allowed us to map the t31954 mutation to the locus referred to as qkA (see below). From now on, we refer to this mutation as qkA^t31954 or, in short, qkA^t.

Consistent with the impaired motility, qkA^t homozygous embryos exhibited somite defects. First, the border of the somite segments was not V-shaped but U-shaped (Fig. 1A-D) and slightly irregular. However, early and late boundary markers, including Fibronectin, Laminin and Dystrophin, did not show marked discontinuity indicating that the border is present (data not shown). In addition, somite muscles were affected. Mononucleated slow-twitch fibers are aligned in a parallel array of ~20 fibers along the dorsoventral axis in WT 27-hpf embryos and encompass the entire length of the somite (Fig. 1A,A′, white arrow). In qkA^t homozygotes, they were reduced in number, the arrays were disorganized and exhibited gaps (Fig. 1B,B′, white asterisk), branching and abnormal fiber direction (Fig. 1B, white arrowhead), resulting in the invasion of the fast muscle domain (Fig. 1B′). Consistent with the U-shaped somite, the MPs, a subpopulation of medial slow muscle cells in the immediate vicinity of the notochord that specifically co-expresses Prox1 and the homeoprotein Engrailed (Ekker et al., 1992; Glasgow and Tomarev, 1998; Halpern et al., 1993; Hatta et al., 1991; Roy et al., 2001), were also reduced (Fig. 1E,F, white arrows; Table 1). In addition, the MPs were mislocated and situated away from the notochord in some cases, suggesting that they were pseudo-MPs (see Fig. 5D-F; data not shown) (Wolff et al., 2003).

At 27 hpf, WT fast muscle cells, labeled by F310 (Crow and Stockdale, 1986)], are elongated and arranged in three diagonal arrays of multinucleated myotubes that are anchored to the anterior and posterior somite boundary and exhibit cortical labeling (Henry et al., 2005; Roy et al., 2001; Snow et al., 2008; Snow and Henry, 2009). By contrast, qkA^t mutant F310-positive cells were not structured like fibers, appeared immature, with predominantly cytosolic fMLC labeling, and were not organized into recognizable arrays or adopted an array disposition parallel to the main axis (Fig. 1C,D). Moreover, cortical labeling with β-catenin antibody showed that a large portion of the mutant fast muscle fibers do not span the entire somite length and are mononucleated, suggesting that cell fusion failed to occur (compare Fig. 1G with 1H, arrows). Similar to MPs, MFFs were also reduced in number in qkA^t mutants (Fig. 1F; Table 1).

The requirement for qkA could be first traced during early to mid-somitogenesis. At the 10-somite stage, qkA^t mutant adaxial cells were not all juxtaposed to the notochord, in contrast to WT (Fig. 1I,J). At early stages, myod (myod1 – Zebrafish Information Network) expression appeared unaffected, indicating that the specification of the myogenic lineage was properly initiated (see Fig. S1 in the supplementary material).

Thus, qkA^t homozygous embryos exhibit a severe and persistent muscle phenotype that affects the morphogenesis of slow and fast fibers as well as maturation of the latter (data not shown). In addition, slow muscle cells, including the MPs and MFFs, are reduced. Together, our results indicate that qkA is required to control both slow and fast muscle cell fates as well as their maturation (fast) and morphogenesis (fast and slow).

qkA^t31954 was meiotically mapped to linkage group (LG) 17 using MGH polymorphic markers, at 1.5 cM from z9633 and 0.11 cM from z21435 (Fig. 2A). Closest to the latter marker, the gene qkA was found in the Zv7 Ensembl assembly of the zebrafish genome. qkA cDNA was cloned and sequenced from WT, heterozygous and qkA^t homozygous mutant embryos. qkA contains eight exons that span 84.2 kb of genomic DNA (Fig. 2B). Sequencing of the cDNA obtained from the qkA^t homozygote showed a T-to-A transversion in position +467 relative to the ATG start codon, which was confirmed on genomic DNA (Fig. 2B,C). This mutation leads to the loss of the BclI site at +464, which was confirmed on PCR-amplified genomic DNA (Fig. 2D), allowing us to easily genotype qkA^t mutant embryos.

RT-PCR of zebrafish qkA cDNA generated two splice variants: qkA1 and qkA2. Zebrafish qkA1 includes all the exons, whereas qkA2 lacks the 3′ part of exon 6 (Fig. 2B). qkA1 and qkA2 mRNAs encode highly conserved 383 and 342 amino acid RNA-binding proteins, respectively (Fig. 2E), that belong to the STAR family (Hardy et al., 1996). Proteins of this family include a KH domain surrounded by two sites, QUA1/NK and QUA2/CK, which together form the STAR domain. The protein can homo/heterodimerize through the NK/CK domains, whereas the KH domain mediates direct protein-mRNA interaction in order to upregulate or downregulate mRNA expression (Cox et al., 1999; Larocque et al., 2005; Larocque et al., 2002; Larocque and Richard, 2005; Nabel-Rosen et al., 2002; Noveroske et al., 2002; Wu et al., 2002). Moreover, one of the mouse quaking (Qk) transcripts, QkI-5 (Ebersole et al., 1996), contains a nuclear localization signal (NLS) sequence at the C-terminus, which is also detected in zebrafish QkA1/2 [366 to 372 (QkA1) and 325 to 331 (QkA2)]. Additionally, there is a significant conserved synteny between regions containing zebrafish qkA and mouse Qk. We therefore conclude that zebrafish qkA is the orthologue of mouse Qk (see Fig. S2 in the supplementary material). The T-to-A transversion in qkA^t leads to a change from a highly conserved isoleucine to asparagine (I156N) in the KH domain, which probably results in a misfolding of this domain that is likely to alter QkA function (Fig. 2E; see red box in Fig. S2 in the supplementary material).

The qkA^t mutation was phenocopied by morpholino antisense oligonucleotide (MO) (Nasevicius and Ekker, 2000) injections, either splice blocking (qkA^gt-MO; Fig. 2F-H,J-L) or translation blocking (qkA^ATG-MO; data not shown). qkA^gt-MO (Fig. 2B) induced a truncated QkA isoform devoid of the STAR domain (Fig. 2E; see Fig. S3 in the supplementary material). The phenotypes of the qkA^t mutant or morphants were efficiently rescued by qkA1 or qkA2 mRNA overexpression, leading, at 27 hpf, to a striking recovery of both slow and fast muscle fiber morphology (Fig. 2I,M; data not shown). However, overexpression of qkA1 or qkA2 in WT embryos did not noticeably affect either slow or fast fiber development (data not shown). Thus, our data show that the muscle phenotypes observed in qkA^t homozygous embryos result from the loss of qkA zygotic function and qkA therefore appears to be a permissive factor for muscle development.

qkA is expressed throughout paraxial mesoderm (Baxendale et al., 2009; Tanaka et al., 1997) (data not shown) and could be required within fast muscle progenitor cells or in their environment. To address this issue, qkA function was blocked locally with qkA^ATG-MO (Fig. 3A). The presence of the MO within a large group of fast muscle progenitors did not affect their capacity to fuse, to elongate, nor to adopt a diagonal arrangement (Fig. 3E-G, n=9/10). A similar result was obtained when qkA^ATG-MO was restricted to the notochord/floor plate domain (Fig. 3B-D, n=8/8). In sharp contrast, presence of the qkA^ATG-MO in a domain that included slow muscle cell progenitors but excluded fast muscle progenitors led to the typical qkA^t immature fast muscle phenotype (Fig. 3H-J, n=9/11). Hence, QkA function is not required in fast muscle cells, but in their environment. Consistent with this result, WT fast muscle cell progenitors grafted into a qkA^t mutant fast muscle domain, or qkA^t mutant fast muscle cell progenitors grafted into WT fast muscle domain, developed as host tissue, confirming the non-autonomous requirement for QkA in fast muscle cells (Fig. 3K-M, n=22/22).

qkA is required for the morphogenesis and/or fate of Hh-induced slow muscle cells and MPs, suggesting that QkA could control these processes by regulating the Hh pathway. Consistent with this notion, somitic expression of the Hh pathway read-outs ptc1 and prdm1a (Baxendale et al., 2004; Concordet et al., 1996) was dramatically reduced in non-axial qkA^t tissue during somitogenesis (Fig. 4A,B,D,E, arrows; see Fig. S4A,B,D,E in the supplementary material), but not in the notochord nor presomitic mesoderm (PSM) (Fig. 4C, arrowhead; data not shown). Likewise, the number of Pax7-positive dermomyotomal myogenic precursors (Feng et al., 2006) was higher in qkA^t than in WT (Fig. 4G,H; n=6 embryos, four somites/embryo), similar to the situation in which Hh signaling is impaired. By contrast, the expression of the Hh-regulated gene nkx2.2a was not altered in the central nervous system (data not shown). Thus, qkA is required to specifically regulate the Hh pathway within trunk mesoderm.

If QkA acts within the Hh signaling pathway to control slow muscle/MP morphogenesis, maturation and/or induction, inhibition of the pathway should lead to phenotypes similar to those induced by qkA loss of function. The transcription factor Gli2a is required for Hh signal transduction in muscle development. Similar to the effects of qkA loss of function, in gli2a morphants the expression profile of ptc1 and prdm1a was attenuated during somitogenesis (Fig. 4C,F; see Fig. S4C,F in the supplementary material) and some slow muscle cells were found medial to the bulk of fast muscle cells [compare Fig. 4I-K with 4L-N (qkA^t) and 4O-Q (gli2a^MO)] (Wolff et al., 2003). Finally, the organization of the fast muscle fibers was impaired in gli2a morphants (see Fig. S5 in the supplementary material). Together with the fact that gli2a is required for MP induction, our results strongly suggest that qkA acts on slow muscle and MP morphogenesis and fate by regulating the Hh signaling pathway.

To directly test the idea that QkA controls the fate of muscle cells by acting on Hh signaling, high levels of Hh protein were induced in the progeny of qkA^t heterozygous carriers by injection of RNA. Slow muscle fibers express the marker Prox1, whereas MPs express both Prox1 and Engrailed (Ekker et al., 1992; Glasgow and Tomarev, 1998; Halpern et al., 1993; Patel et al., 1989; Roy et al., 2001; Wolff et al., 2003), and both can be distinguished from MFFs, which express Engrailed only (Roy et al., 2001; Wolff et al., 2003). In WT or qkA^t heterozygotes (Fig. 5G-I,M; Table 1), shh mRNA overexpression led to a twofold increase in both slow muscle fibers and MPs, and to a marked decrease in MFFs. By contrast, shh overexpression in qkA^t homozygotes (Fig. 5J-M; Table 1) led to an increase in MPs and slow fibers but to a strong increase in the size of the MFF population. These results are consistent with the idea that Hh signaling is still partially active when QkA activity is downregulated (owing to the persistence of Hh-dependent muscle cells). However, the response to shh overexpression in terms of the generation of different muscle cell types is altered in QkA-deficient embryos as the MFF fate is promoted rather than inhibited. Altogether, combined with the downregulation of ptc1 and prdm1a expression and the upregulation of Pax7⁺ dermomyotomal cells in qkA^t homozygotes, these results show that qkA plays a major role in fine-tuning the Hh signaling pathway in the mesoderm.

How does QkA act on the Hh signaling pathway? In vertebrates, Quaking regulates different mRNA targets by interacting with their pre-messenger 3′UTR (Larocque et al., 2002; Li et al., 2000; Nabel-Rosen et al., 2002; Schachner and Bartsch, 2000) via a QRE (Galarneau and Richard, 2005). A Quaking-binding site has been defined by two motifs, UACUMAY and UAAY, separated by 1-20 nucleotides. This structure has been highly conserved across evolution. The high degree of sequence conservation between mouse Qk and zebrafish QkA KH domains suggested that the QRE is probably conserved in zebrafish as well. We screened in silico, using the Ensembl database, the 3′UTRs of genes encoding the different components of the Hh signaling pathway for the presence of a QRE. gli2a, which encodes a transcription factor that promotes ptc1 transcription and is required for MP, slow muscle and MFF specification (Du and Dienhart, 2001; Roy et al., 2001), includes a sequence that is highly related to the canonical QRE in its 3′UTR (Fig. 6A). QkA could thus regulate the Hh signaling pathway through gli2a mRNA binding.

As no antibodies are available for Gli2a, we performed a sensor assay (Leucht et al., 2008) using a destabilized GFP (d2eGFP) as a reporter fused upstream to the 3′UTR of gli2a mRNA (d2eGFP-gli2a) or to the 3′UTR of wnt11 mRNA (d2eGFP-wnt11), which does not carry any obvious QRE, as a control. These constructs were injected at the one-cell stage into the progeny of qkA^t heterozygotes and d2eGFP expression was assayed by western blotting on mid-somitogenesis embryos. Whereas d2eGFP levels were similar in WT and qkA^t homozygous d2eGFP-wnt11-injected embryos, d2eGFP expression derived from d2eGFP-gli2a injection was significantly decreased in qkA^t homozygous versus WT sibling embryos (Fig. 6B, compare left with right panel; Fig. 6C; Table 2). Although this latter effect might seem modest, the assay was performed on whole trunk extracts. Gli2a regulation by QkA might be restricted to adaxial cells and thus swamped by the expression of gli2a in surrounding tissues. Thus, this result is consistent with the fact that endogenous gli2a mRNA is stabilized, or its translation facilitated, by QkA protein.

To address the role of QkA in gli2a mRNA stability more directly, we took advantage of the yot mutant allele, which carries a point mutation in the coding region of the gli2a transcript that results in the expression of a truncated repressor form of the Gli2a protein (DR-Gli2a). Embryos homozygous for this mutant allele display a complete loss of response to Hh signaling in the myotome, including a loss of all slow muscle fibers as well as MPs (Karlstrom et al., 2003). Consistent with this being due to constitutive Gli2a repressor activity, the phenotype can be suppressed by a translation-blocking gli2a MO (Wolff et al., 2003). We reasoned that if QkA stabilizes the gli2a transcript, inactivation of QkA should similarly suppress the yot^ty119 phenotype. We injected yot^ty119 homozygous embryos with qkA^MO and, as predicted, observed significant rescue of slow muscle fiber differentiation (Fig. 6D-F).

Altogether, our data reveal a novel and essential role for QkA proteins in the regulation of the Hh signaling pathway and in the specification of the different muscle cell types in the zebrafish embryo.

We have identified and characterized qkA, which encodes a highly conserved member of a large family of RNA-binding proteins, as a new locus required for somitic muscle development. We also present the first direct evidence that the post-transcriptional regulator QkA acts on muscle development in zebrafish by regulating the Hh signaling pathway via the transcription factor Gli2a.

Both slow and fast muscle fibers require qkA function. qkA loss of function leads to both a reduction in the number of slow muscle cells, including MPs (see below), and in altered morphogenesis for the general population of slow muscle cells. In particular, their parallel horizontal alignment is disrupted at 24 hpf and slow cells invade the fast muscle domain. Although it is difficult at this stage to trace the origin of the altered alignment, the invasion of the fast muscle domain suggests that the distal migration of slow muscle cells could have been hampered. This process is known to rely on the complementary expression of n- and m-Cadherins (Cortes et al., 2003), so qkA might interact with these latter pathways. Consistently, some adaxial cells do not contact notochord in qkA^t mutants, suggesting that adhesion is altered. Alternatively, qkA could be involved in the regulation of early movements controlled by the PCP components knypek and trilobite (glypican 4 and vang-like 2 – Zebrafish Information Network). A detailed analysis of adaxial cell movements during somitogenesis is required to clarify this issue.

Fast fibers express the appropriate differentiation markers (F310, MF20) at 27 hpf in qkA mutants/morphants and many of them exhibit the classical multinucleated morphology. However, unlike their WT counterparts, MF20 and F310 stainings appear diffuse, progenitors often fail to undergo fusion and, although they are aligned along the main body axis, fast muscle cells do not adopt a tense fiber-like aspect. As myogenic markers are expressed as in the WT, cell type specification seems to occur properly. However, at 48 hpf, fast fibers recover partially (see Fig. S6 in the supplementary material), suggesting that QkA is required for the on-time maturation of the fast muscle fibers. How is qkA required for fast muscle maturation? Both local inhibition of qkA within fast muscle progenitors and transplantation experiments demonstrate that qkA is not required within fast muscle cells but in their environment. Fast muscle cells or their precursors are known to require interaction with their environment/neighbors to properly develop. Notochord is required for slow muscle development as a source of the Hh morphogen (Devoto et al., 1996; Halpern et al., 1993; Krauss et al., 1993). However, creating a large clone of qkA-negative cells in notochord does not delay fast muscle maturation. Similarly, muscle cells require an interaction with somite/myotome boundaries for their development. However, although the boundary might appear morphologically altered, the expression of segment boundary matrix components appears, in general, normal, making it unlikely that QkA acts on fast muscle cells via the segment boundary. Finally, fast muscle progenitors are known to require slow muscle cells for their maturation (Henry and Amacher, 2004). Slow muscle cells do differentiate and most of them migrate through the field of fast muscle progenitors when qkA function is inhibited. So one intriguing possibility is that qkA is required for the action of slow muscle cells on the maturation of fast muscle cells. Alternatively, other cell/tissue types, such as the dermomyotome (Feng et al., 2006), could be involved. Restoration of qkA activity in specific regions of qkA-deficient embryos and analysis of the resulting fast muscle phenotype is required to answer this point.

We provide several lines of evidence that qkA is responsible for the different muscle cell type phenotypes observed in qkA^t mutants. First, the mutation maps to the qkA linkage group. Second, qkA harbors a missense mutation in the qkA KH domain, at a highly conserved residue. Third, the mutation can be rescued by qkA mRNA injections. Fourth, qkA^MO injections mimic the t31954 phenotype. Altogether, these results allow us to unambiguously assign the phenotype to a loss of qkA function. Consistent with this notion, the t31954 allele is recessive. Thus, qkA is required for somitic muscle development. How can it fulfill its function? STAR family members act as RNA-binding proteins that are able to regulate different biological processes by either preventing translation (Saccomanno et al., 1999) or interfering with RNA transport/splicing (Larocque et al., 2002). In this process, RNA binding requires the KH domain, as well as a consensus sequence UACUAAY-N(1-20)-UAAY (Galarneau and Richard, 2005). In this study, we showed that the mutation in the qkA^t RNA-binding KH domain involves a change from the hydrophobic Ile to the polar Asn. In mice, a mutation involving a similar change in a neighboring position within the Qk KH domain has been analyzed at the molecular level. The mouse Qk^k2 allele causes a transformation from Val to Glu, which leads to a loss of RNA-binding activity but not hetero/homodimerization (Cox et al., 1999; Larocque et al., 2002; Noveroske et al., 2000), most likely by perturbing the hydrophobic core of the Qk KH domain. The t31954 mutation should therefore also affect the QkA RNA-binding activity.

We present several lines of evidence that QkA interacts with the Hh signaling pathway. First, inhibition of qkA function leads to a reduction in the size of cell populations that are derivatives of the adaxial cells, which are classically induced by Hh signaling. Second, the number of Pax7⁺ dermomyotomal cells is increased. Third, slow muscle cells invade the fast muscle domain in both qkA^t mutants and when Hh signaling is inhibited (gli2a^MO). Fourth, expression of ptc1 and prdm1a, two major read-outs for Hh activity, is markedly reduced in the trunk of qkA^t. Interestingly, this reduction is observed in paraxial but not in axial mesoderm, nor in PSM, indicating that qkA is required only locally and temporally for full activation of the Hh pathway. Thus, qkA appears to be required for the proper regulation of the Hh signaling pathway within paraxial mesoderm. When the Hh pathway is overactivated in qkA^t, adaxial derivatives are induced to a similar level as in WT, suggesting that either the t31954 allele is not null or that qkA acts redundantly with other genes to regulate muscle fates. The slow muscle phenotype is stronger in qkA morphants than in qkA^t mutants, suggesting that qkA^t does not completely abolish qkA function, although the redundancy hypothesis cannot be discounted.

Hyperactivation of the Hh pathway, which normally leads to a reduction in the MFF population, results in a dramatic increase in the MFF population when qkA is attenuated. Two mechanisms could explain such an increase. First, Hh activity represses the fast muscle (and thus MFF) fates and QkA could act by preventing this repression. Alternatively, qkA could act on the timing of MFF induction by lengthening the competence period of fast muscle progenitors to Hh signals. Further experiments are required to resolve these issues.

How does QkA act on the pathway? Among the different Hh signaling pathway members available in the zebrafish Ensembl database, only gli2a exhibits a canonical QRE. Moreover, gli2a appears to be a good candidate target mRNA for QkA as it is required for slow muscle induction. Furthermore, we show that the gli2a 3′UTR confers sensitivity to QkA in vivo (sensor assay) and provide evidence for in vivo functional interaction between QkA and gli2a (the analysis of the yot mutant). Interestingly, the capacity of mouse Qk and the related C. elegans GLD-1 to regulate Gli1 (or C. elegans tra-1) by binding to their QREs and repressing translation has been described (Lakiza et al., 2005). Our data are consistent with an action at the translational level, although they support an activation rather than a repression, potentially stemming from a difference in the QRE-encompassing region or structural differences between zebrafish QkA and GLD-1. Furthermore, our data suggest that the regulation of the Hh pathway by Qk-gli interaction is highly conserved throughout evolution and provide direct evidence for the in vivo implication of such an interaction in a developmental process. However, the muscle phenotype resulting from qkA loss of function cannot be entirely explained by an action through gli2a mRNA. In particular, the MFF population is reduced in qkA^t but increased in gli2a morphants (Wolff et al., 2003). Although the mechanism of action of Gli2a is rather complex because of the existence of activator and repressor states, a reduction in qkA translation would be expected to lead to a decreased capacity of activation. Thus, other QkA-regulated genes need to be evoked to explain its action. It is conceivable, for instance, that Gli1 is an indirect target of QkA, considering the fact that combined loss of function for gli1 and gli2a leads to a complete loss of Engrailed-positive cells in the somite (Wolff et al., 2003). Alternatively, other QRE-negative genes might be involved.

We propose the following model (Fig. 7). In WT embryos, gli2a is transcribed and the pre-messenger is stabilized by QkA activity. Then, gli2a mRNA is exported to the cytoplasm and translated. In response to Shh, the activated form of Gli2a transduces Hh signaling by entering into the nucleus and inducing the transcription of target genes, including ptc1 (Fig. 7A). When qkA function is inhibited, the stabilization of gli2a mRNA is not ensured, leading to reductions in the gli2a transcript and Gli2a protein concentration (Fig. 7B). Alongside the negative regulation of the Hh transduction component Sufu by miRNAs (Flynt et al., 2007), our results provide a novel regulator of this pathway that allows fine-tuning to generate the different muscle cell populations of the somite.

We are very grateful to Philip Ingham (IMCB Singapore and University of Sheffield, UK) for suggesting the yot qkA^MO experiment and for allowing Rosemary Kim to assist in its execution. We thank P. Lemaire and L. Bally-Cuif for the gifts of plasmids pSPE3-Rfa and d2egfp; F. Bouallague for excellent animal care; and E. Hirsinger and members of the F.M.R. lab for comments on manuscript. R.L. was supported by a fellowship from the Ministère de l'Enseignement Supérieur et de la Recherche. This work was supported by the EU FP6-ZF MODELS and FP7-ZF-HEALTH programs, the ANR grant ANR06-BLAN-0321, the Association Française contre les Myopathies grant #F12172A/S01818 and the Association pour la Recherche sur le Cancer grant #5020.