The osr1 and osr2 genes act in the pronephric anlage downstream of retinoic acid signaling and upstream of wnt2b to maintain pectoral fin development

The odd-skipped (Odd/Osr) family of genes comprises evolutionary conserved zinc-finger transcription factors that lie at the top of the genetic hierarchy required for renal development in vertebrates and probably also in Drosophila (James et al., 2006; Tena et al., 2007; Wang et al., 2005). Mammalian genomes contain two paralogs, Osr1 and Osr2 (Lan et al., 2001; So and Danielian, 1999). In the mouse, Osr1 expression starts early (E7.5) in the intermediate mesoderm (IM), from where renal structures derive (Mugford et al., 2008; So and Danielian, 1999), and is maintained until kidney organogenesis occurs. Osr2, by contrast, is activated at stage E9.25 in the mesonephros, and later (stage E14.5) in the mesenchyme that surrounds the ducts of the mesonephros and metanephros (Lan et al., 2001). Osr1 knockouts lack renal structures (James et al., 2006; Wang et al., 2005), whereas Osr2 mutants have apparently normal kidney development (Lan et al., 2004). Xenopus and zebrafish genomes also contain two Osr genes, and, in contrast to the mouse genes, both of them seem to contribute to some extent to the formation of the kidney (Tena et al., 2007). Indeed, knock down of both genes generates stronger kidney defects than single depletions, indicating partial redundancy between both genes. A partial redundancy of these two genes is further observed in gain-of-function assays in Xenopus and zebrafish (Tena et al., 2007), and can also be observed in knock-in experiments in mice (Gao et al., 2009).

Beside the kidney, vertebrate Osr genes are expressed in many other tissues. Analysis of mutant lines has indicated that these genes are required for proper formation and/or patterning of the endoderm, the heart, the teeth, the palate, the bones and the synovial joints in the limbs (Gao et al., 2011; Kawai et al., 2007; Lan et al., 2004; Mudumana et al., 2008; Wang et al., 2005; Zhang et al., 2009). Expression studies in mice and chicken indicate that, in the limb, Osr genes are expressed from very early stages in a highly dynamic pattern (Lan et al., 2001; So and Danielian, 1999; Stricker et al., 2006). These genes start to be expressed at E11.5 in mouse or HH22 stage in chick limb buds in largely overlapping domains in the bud mesenchyme, the tissue that will form, among other cell types, the bones. Slightly later, the expression of Osr1 and Osr2 becomes largely complementary, whereby Osr2 is expressed more proximally and Osr1 more distally. Finally, at later stages during limb development, both genes are again co-expressed in the developing joints. Despite this complex and dynamic expression patterns covering most of the developing limb, tissue-specific ablation using a lateral plate mesoderm (LPM)-specific Cre line (Prx1::Cre) has recently indicated that Osr1 and Osr2 genes are only required for joint development (Gao et al., 2011). To obtain insight about other possible functions of these genes during limb formation, and the degree of conservation of Osr function during the development of vertebrate appendages, we have examined the requirement of both osr1 and osr2 genes during development of the zebrafish pectoral fins, the structures that are equivalent to forelimbs in tetrapods.

During early stages of development, the pectoral fin/forelimb field is induced with the specification of a group of LPM cells on either side of the embryo's trunk at precise positions along the anterior/posterior (A/P) axis (Capdevila and Izpisua Belmonte, 2001; Johnson and Tabin, 1997; Mercader, 2007; Tickle, 1999; Duboc and Logan, 2011). Interactions between the mesenchyme and the overlying ectoderm trigger the outgrowth of the fin/limb bud. The earliest molecular marker described as an initiator of limb bud formation is the T-box transcription factor Tbx5, which is expressed in the limb mesenchymal precursors of all tetrapods as well as all fish species analyzed so far (Agarwal et al., 2003; Begemann and Ingham, 2000; Gibson-Brown et al., 1996; Isaac et al., 1998; Logan et al., 1998; Ruvinsky et al., 2000; Saito et al., 2006; Tamura et al., 1999). Indeed, Tbx5 is both necessary and sufficient for forelimb formation (Agarwal et al., 2003; Garrity et al., 2002; Minguillon et al., 2005; Ng et al., 2002; Rallis et al., 2003; Takeuchi et al., 2003; Ahn et al., 2002). In zebrafish, tbx5a mutant heartstrings (hst) embryos, as well as tbx5a morphants, show a complete loss of pectoral fins (Garrity et al., 2002; Ng et al., 2002; Ahn et al., 2002). The cells of the LPM expressing tbx5a fail to aggregate to form the compact circular structure of the wild-type fin bud. An upstream regulator of Tbx5 in the establishment of the limb field is the retinoic acid (RA) pathway (Begemann et al., 2001; Gibert et al., 2006; Grandel and Brand, 2011; Grandel et al., 2002; Mercader et al., 2006; Mic et al., 2004; Zhao et al., 2009). During development, the limiting step in the synthesis of RA is catalyzed mainly by aldehyde dehydrogenase 1a2. The medaka, zebrafish and mouse mutants for Raldh2/aldh1a2 display absence of Tbx5 expression in the limb mesenchyme, and fail to form fins/limbs (Begemann et al., 2001; Grandel et al., 2002; Mic et al., 2004; Niederreither et al., 1999; Negishi et al., 2010). RA signaling has been proposed to regulate fin outgrowth in a two-step process: during gastrulation it promotes initiation of tbx5 expression, whereas during early somitogenesis it is necessary to expand and maintain the tbx5 expression domain (Begemann et al., 2001; Grandel et al., 2002; Grandel and Brand, 2011). These phenotypes can be rescued by the exogenous application of RA or by transplantation of wild-type cells (Gibert et al., 2006; Linville et al., 2004; Mercader et al., 2006; Mic et al., 2004; Niederreither et al., 1999; Zhao et al., 2009; Negishi et al., 2010).

The Wnt signaling pathway has also been proposed to regulate fin/limb outgrowth. In the chick, Wnt2b is detected in the IM as well as in the LPM of the wing bud field (Kawakami et al., 2001). In the zebrafish, two wnt2b orthologs have been described, wnt2ba and wnt2bb. wnt2ba expression is restricted to the IM, whereas wnt2bb expression has been described in the LPM at the time of fin formation (Ng et al., 2002; Ober et al., 2006). Wnt2b gain-of-function in the chick leads to ectopic limb formation (Kawakami et al., 2001), whereas morpholino-mediated wnt2ba gene silencing in the zebrafish downregulates tbx5 expression and impedes proper fin bud outgrowth (Ng et al., 2002; Wakahara et al., 2007). wnt2ba expression is regulated by RA signaling (Mercader et al., 2006; Negishi et al., 2010), suggesting that RA might control limb development by controlling wnt2b expression.

The role of the IM, a kidney precursor tissue, in controlling limb outgrowth is still controversial. Classical experiments involving surgical ablation of the mesonephros in the chick led to impaired limb outgrowth (Geduspan and Solursh, 1992; Stephens and McNulty, 1981). On the contrary, physical block of IM and LMP did not interfere with limb development (Fernandez-Teran et al., 1997). In addition, genetic ablation of the kidney anlage in mouse appears to be compatible with the development of a normal limb (Bouchard et al., 2002). In opposition to this finding, a recently reported Xenopus mutant of the nephronectin gene, a small diffusible integrin ligand necessary for metanephros formation, display a complete lack of forelimbs (Abu-Daya et al., 2011).

Here, we report that in zebrafish Osr genes are required within the kidney anlage to maintain proper levels of tbx5a expression in the LPM during early stages of limb development. This dependence is mediated, at least in part, by Wnt2b signaling, which originates at the pronephros and requires Osr function. We also show that Osr genes are controlled by RA signaling. Therefore, our studies allow connecting the RA and the Wnt pathways during early limb formation through Osr function.

Antisense RNA probes were prepared from cDNAs using digoxigenin or fluorescein (Boehringer Mannheim) as labels. Zebrafish specimens were prepared, hybridized and stained as described (Harland, 1991; Jowett and Lettice, 1994). For wnt2ba in situ hybridization, the blocking solution used was 2% blocking powder (Roche) in maleic acid solution (0.1 M, pH 7.5) and the antibody was diluted in the same blocking solution. For in situ hybridization on sections, embryos were cut using a Leica VT100S vibratome at 30 μm after staining and examined under the microscope.

All DNAs were linearized and transcribed as described previously (Harland and Weintraub, 1985) with a GTP cap analog (New England Biolabs), using SP6, T3 or T7 RNA polymerases. After DNAse treatment, RNA was extracted with phenol-chloroform, column purified and precipitated with ethanol. mRNAs for overexpression studies were resuspended in water and injected at the desired concentration in the yolk at the one-to two-cell stage. For knockdown experiments, zebrafish embryos were injected in the yolk at the one-to two-cell stage with 7.5-20 ng of morpholinos. The translation specific MOosr1 and MOosr2 morpholinos and the Osr overexpression constructs have been described previously (Tena et al., 2007). The MOosr1sp morpholino was described by Mudumana et al. (Mudumana et al., 2008). The MOosr2sp morpholino used in this study targets the acceptor splicing site from exon 3 (5′ -ATATATCTGAGGAACAGGCG AGAGG-3′). To evaluate its efficiency on blocking osr2 mRNA splicing, we designed primers in exons 2 and 4 (5′ -GGGAGACACTTCACCA AATCC-3′ and 5′-CGTGATCGTGAGTGCTATGG-3′). For RT-PCR, total RNA was extracted at 48 hpf from 25 morphants and control embryos and amplification was carried out for 30 cycles. A band of 297 bp is produced only if the MOosr2sp cause the skipping of exon 2. The elimination of this exon of the osr2 mRNA introduces several precocious stop codons. The MOwnt2ba morpholino used targets acceptor splicing site from exon 2 (5′-CTGCAGAAACAA - ACAGACAATTAAG-3′) and was injected at 6 to 10 ng per embryo. Primers used for RT-PCR to test efficiency of gene silencing were wnt2ba_ex1-F (5′ -GCCAGAGTGTGATGGAGTTG-3′) and wnt2ba_ex5-R (5′ -CATGCCTCTGGAGGACTTGT-3′). A band of 982 bp spanning exon2 to exon5 disappears in MOwnt2ba-injected embryos.

The following primers were used to detect lef1 mRNA levels in control and injected embryos by qRT-PCR assays: lef1F (5′-CATCCAGCCA - TTGTCAACC-3′), lef1R (5′ -CAGCATGAAAGCGTTTAGAGG-3′). Amplification of the ef1a gene (McCurley and Callard, 2008) was used for normalization (primers ef1aF 5′ -CTTCTCAGGCTGACTGTGC-3′ and ef1aR 5′ -CCGCTAGCATTACCCTCC-3′).

Embryos were incubated in the dark at 28°C in 10^-8 M all-trans retinoic acid (Sigma), diluted in E3 embryo medium, from a 10^-5 M stock solution in DMSO. DEAB (4-diethylaminobenzaldehyde, Sigma Aldrich), a competitive reversible inhibitor of retinaldehyde dehydrogenases (Begemann et al., 2004), was applied at a concentration of 10^-4 M diluted from a 0.1 M stock in DMSO in E3 media. As controls, wild-type embryos were treated with similar dilutions of DMSO without drugs. Embryos injected with inducible domain of β-catenin (Afouda et al., 2008), were incubated with dexamethasone (4 μg/ml, Sigma Aldrich) in E3 embryo medium (Kolm and Sive, 1995).

We have previously reported two zebrafish morpholinos (MOosr1and MOosr2) that effectively impair osr1 and osr2 mRNA translation in zebrafish (Tena et al., 2007) (supplementary material Fig. S1A). Embryos injected with 20 ng of any of the individual MOs show reduction of fin buds at 5 dpf (see Table 1 for phenotypes in embryos injected with different MOs), although the proportion of affected embryos and the reduction of fin size was higher in osr1 morphant (Table 1, Fig. 1A-C). At these higher doses, we also observed craniofacial defects (Fig. 1C) similar to those described in mouse embryos deficient for Osr2 (Lan et al., 2004). Co-injection of both MOs at half doses caused a fin reduction similar or slightly higher to that observed with MOosr1 alone (Table 1, Fig. 1D,E). Accordingly, in the double knockdown there was a higher proportion of embryos with no fins, than in osr1 single morphants (Table 1, Fig. 1E). Similar effects (Table 1) were observed with a different set of MOs that block the correct splicing of osr1 and osr2 mRNAs (Mudumana et al., 2008) (supplementary material Fig. S1B). These results indicate that both genes influence the genetic cascade that operates during fin development. We then examined the expression patterns of these genes during stages of early fin bud formation and compared them with that of tbx5a. The expression of tbx5a in the LPM starts at the 7-somite stage (Begemann and Ingham, 2000) (Fig. 2C). At this stage, osr1 was expressed all along the IM, whereas osr2 was present at very low levels only in the posterior IM (Fig. 2A,B). At the 13-somite stage, the expression of tbx5a in the LPM was stronger and both osr1 and osr2 localized adjacent to this domain at the anterior IM in the pronephric anlage (Fig. 2D-F). At this stage, in the posterior IM, osr1 was found in a more lateral domain (Fig. 2D, asterisk), as previously shown (Mudumana et al., 2008). The same relative distribution of tbx5a, osr1 and osr2 was observed at 24 hours post fertilization (hpf), although at this stage osr1 also showed expression in endodermal cells (Mudumana et al., 2008; Tena et al., 2007) (Fig. 2G-I). At 48 hpf, although osr1 is very weakly expressed in the growing limb bud (Fig. 2K), osr2 is observed in two patches, one in the anterior and another in the posterior edges of this territory (Fig. 2L). A stronger and broader expression of osr2 was observed in the fins at 72 hpf, whereas osr1 is only moderately expressed in the fin at this stage (Fig. 2M-O).

The rather late expression of Osr genes in the fin fields let us examine at what stage pectoral fin development was affected in the Osr morphant embryos. In embryos injected with single or combinations of osr1 and osr2 MOs, the expression of tbx5a was not clearly affected at the 7- to 8-somite stage (Table 1, Fig. 3A-D). However, from the 12-somite stage onwards, we observed a reduction of tbx5a (Table 1, Fig. 3E-P). In these experiments, we used as an internal control tbx5a expression in the eye (Fig. 3Q-T), which was not affected in injected embryos. Importantly, tbx5a rarely disappeared completely in the morphant embryos, as we frequently observed for pax2a in the anterior pronephros (Fig. 3U,X). This indicates that Osr genes influence the maintenance, but not the triggering, of tbx5a expression. By analyzing in the same embryos the expression of the pronephric marker pax2a and tbx5a, we observed that there is a high correlation between kidney and fin defects in the Osr morphant embryos (Fig. 3U-X). Nevertheless, a fraction of the morphant embryos, which is higher in osr2-deficient ones (see Table 1), showed reduced tbx5a expression but no apparently affected kidneys (Fig. 3W). This indicates that the fin phenotype can occur even at Osr levels that are enough for kidney formation. As both genes seem to participate in fin formation, we also evaluated the ability of individual Osr genes to rescue tbx5a expression in double Osr-morphant embryos. For this experiment, we used the Myc-tagged Osr mRNAs (MTosr1 or MTosr2), which are insensitive to the translation-blocking MOs (Tena et al., 2007). Both Osr genes were similarly capable of partially rescuing the expression of tbx5a in the pectoral fin territory, as well as the expression of pax2a in the pronephros (Fig. 4A-D). Thus, whereas 93% of the embryos morphant for both Osr genes showed reduced tbx5a and pax2a expression, this proportion was reduced to 83% (n=138) or 76% (n=250) in embryos co-injected with the Osr MOs and MTosr1 or MTosr2 mRNAs, respectively. As shown previously (Tena et al., 2007), the overexpression of the Osr mRNAs caused strong gastrulation defects, which is likely to preclude a higher proportion of rescued embryos. We conclude that both Osr genes influence pectoral fin development at the time when they are not expressed in this territory but are co-expressed in the kidney anlage (Tena et al., 2007).

Both in zebrafish and chick embryos, it has been reported that wnt2ba, which is expressed in the IM, is essential for limb formation by controlling tbx5a expression (Kawakami et al., 2001; Ng et al., 2002). In order to analyze the epistatic relationship of Wnt2b and Osr genes in controlling limb development, we first compared the temporal expression profile of the Osr and the wnt2ba gene. At the 7-somite stage, when the expression of tbx5a initiates (Fig. 2A), wnt2ba is still not expressed (Fig. 5A). The expression of this gene became only detectable in the IM from the 10- to 12-somite stage onwards, in a pattern reminiscent to that of the Osr genes (Fig. 5B,C). Indeed, sections through 22 hpf embryos showed that Osr genes and wnt2ba are both co-expressed in the pronephric area (Fig. 5D-I). A similar pronephric expression was found for the related wnt2bb gene (Ober et al., 2006), which we detected expressed in the kidney territory only from 21-somite stage onwards (Fig. 5F, inset).

As Osr depletion downregulated tbx5a expression precisely at the time wnt2ba became activated, we next examined the expression of the Wnt2b genes in Osr morphant embryos. In osr1 morphant embryos, the expression of both wnt2ba and wnt2bb genes was downregulated (Table 1; Fig. 6A,B,E,F). The injection of MOosr2 did not affect wnt2ba expression (Fig. 6C) but caused the downregulation of wnt2bb (Fig. 6G). The co-injection of both Osr MOs at half doses caused similar defects to those observed in osr1 single morphants (Table 1; Fig. 6D,H). Our results therefore suggest that Wnt signaling may contribute to the effect of Osr depletion on tbx5a expression and fins development. To test this, we overexpressed this pathway in embryos co-injected with both Osr MOs. For that purpose, we injected an mRNA encoding an inducible form of β-catenin fused to the dexamethasone-regulated glucocorticoid receptor (GR) domain (Afouda et al., 2008). This allowed us, upon adding dexamethasone to the injected embryos at tailbud stage, to activate the pathway only after gastrulation, which prevented early defects associated with increased Wnt signaling during this crucial period of development. In embryos injected with 50 pg of this β-catenin-GR mRNA, the expression of the Wnt-target gene lef1 was effectively induced, as determined by qRT-PCR, but only in the presence of the hormone (supplementary material Fig. S2). At the 13-somite stage, the expression of tbx5a was not affected in embryos injected with β-catenin-GR mRNA in the absence of dexamethasone but was apparently slightly expanded upon dexamethasone addition (69%, n=52; Fig. 6I,J). In embryos co-injected with both splicing-blocking Osr MOs and β-catenin mRNA, tbx5a was downregulated in the absence of dexamethasone (96%, n=77; Fig. 6K). However, tbx5a downregulation was observed in significantly fewer embryos upon adding this synthetic hormone (75%, n=67; Fig. 6L).

Finally, we examined the expression of osr1 and tbx5a in embryos with depleted wnt2ba function. Gene expression was analyzed at the 13-somite stage, when tbx5a downregulation in Osr morphant embryos was first detected and wnt2bb was still not present. In this experiment, we used a MO that efficiently blocks the splicing of wnt2ba mRNA (supplementary material Fig. S1C). In 13-somite embryos injected with MOwnt2ba, tbx5a expression was reduced (75%, n=90) whereas osr1 expression was not affected (Fig. 7). Moreover, tbx5a reduction was similar to that found in Osr morphant embryos (compare Fig. 7D with Fig. 3F-H). These results strongly suggest that Wnt2b signaling is downstream of Osr factors and mediates, at least in part, the effect of Osr genes on tbx5a expression and limb development. Moreover, as the onset of tbx5a activation precedes that of the Wnt2b genes, this signaling pathway is probably required for maintaining, but not initiating, tbx5a expression, something we also found for the Osr genes.

Retinoic acid (RA) is essential for proper kidney development and also for limb formation through the regulation of tbx5a (Begemann et al., 2001; Cartry et al., 2006; Gibert et al., 2006; Grandel and Brand, 2010; Mic et al., 2004). As Osr genes are key regulators of both processes, we examined the mutual relationship between these genes and the RA pathway. We first determined whether the expression of aldh1a2, an enzyme required for RA production was affected in Osr morphant embryos. As expected for downstream factors of the RA pathway, the expression of aldh1a2 was not affected in embryos injected with any of the Osr MOs or the combination of both of them (MOosr1, 3%, n=53; MOosr2, 2%, n=41; MOosr1+ MOosr2, 1%, n=40; Fig. 8A-D). We then manipulated RA signaling and examined Osr expression. To increase or reduce the RA signaling pathway, we incubated the embryos from tailbud to the 21- to 22-somite stages either with all-trans RA or with DEAB (4-diethylaminobenzaldehyde), a competitive reversible inhibitor of retinaldehyde dehydrogenases. Increasing RA signaling rescued the downregulation of the tbx5a expression promoted by DEAB treatment (48%, n=128; Fig. 8E-G), but was unable to rescue tbx5a expression in Osr morphant embryos (95% of the embryos with reduced tbx5a expression, n=47; Fig. 8H). This indicates that Osr genes are downstream of the RA pathway.

We then examined the effect of increasing or reducing the RA signaling pathway on the expression of the Osr genes. In embryos incubated with DEAB or RA from tailbud onwards, the osr1 expression domain at the 21- to 22-somite stage was reduced (96%, n=25) or expanded (45%, n=31), respectively (Fig. 8I-K). By contrast, these treatments had, if any, minor effects on osr2 expression (Fig. 8M-O). Therefore, during somitogenesis, only osr1 depends on RA. Next, we determined whether the overexpression of osr1 could overcome the loss of tbx5a expression under conditions of reduced RA signaling. Indeed, this was the case as tbx5a expression was recovered in 30% of the embryos exposed to DEAB when injected with osr1 mRNA (Fig. 8L,P).

Finally, we compared the temporal requirement of RA for the expression of Osr, wnt2b and tbx5 genes. Embryos were incubated with DEAB from 50% epiboly, tailbud or 6-somite stages to the 21-somite stage, and subsequently fixed and processed for in situ hybridization. When RA was inhibited at 50% epiboly, the expression of all genes, including osr2, was strongly impaired, effects that became less pronounced when DEAB was added at tailbud or 6-somite stages (Fig. 9). These results indicate that, as has been shown for fin development (Gibert et al., 2006; Grandel and Brand, 2010), RA is required during gastrulation and early somitogenesis for Osr gene expression.

Our work demonstrates an essential function of zebrafish Osr1 and Osr2 transcription factors during pectoral fin formation. Osr genes are required at initial stages of fin outgrowth, for maintaining tbx5 expression, which is the earliest marker involved in limb bud formation (Agarwal et al., 2003; Ahn et al., 2002; Garrity et al., 2002; Rallis et al., 2003). Interestingly, this requirement occurs at a developmental stage in which the Osr genes are not expressed in the fin bud primordia, but in the adjacent IM. A role for the IM during limb development has been previously been reported by others (Geduspan and Solursh, 1992; Stephens and McNulty, 1981; Abu-Daya et al., 2011). Our data further support an important function of the kidney anlage for limb bud formation.

As it has been previously found during early kidney formation in zebrafish and Xenopus (Tena et al., 2007), both Osr genes seem to be partially redundantly required for pectoral fin formation. A similar redundancy has also been shown during mouse joint formation (Gao et al., 2011) and in other knock-in studies (Gao et al., 2009). Nevertheless, during pectoral fin formation, we systematically observed that loss of osr1 function seems to produce stronger defects than the reduction of osr2 activity. This may be due to broader and earlier expression of osr1 in the IM territory in most vertebrates, and correlates with a stronger inhibition of Wnt2b signaling by osr1 gene silencing (Lan et al., 2001; So and Danielian, 1999; Stricker et al., 2006; Tena et al., 2007). Therefore, as proposed recently (Gao et al., 2009), the distinct developmental requirement exhibited by both genes in this and other processes is probably the result of divergence of the cis-regulatory regions that control the spatiotemporal expression of these genes, although the functional potential of both gene products is likely to be very similar.

Our results demonstrate that Osr genes are required in the IM territory to promote pectoral fin development indirectly. This indirect action could be explained, at least in part, by the diffusible molecule Wnt2b, a factor essential for tbx5a expression (Ng et al., 2002). Accordingly, we demonstrate that Wnt2b and Osr genes show overlapping expression domains in the IM, and that Wnt2b expression depends on the function of Osr transcription factors. Moreover, Wnt signaling partially rescues the tbx5a downregulation observed in Osr morphant embryos and the timing of wnt2ba activation at 12 somites coincides with the developmental period in which tbx5a expression became affected by Osr depletion. All these data suggest that the Wnt2 signaling from the kidney anlage mediates the Osr requirement for maintaining tbx5a expression and proper pectoral fin development. It should be nevertheless mentioned that the requirement of Osr factors for Wnt2b expression is probably indirect as these factors have been shown to behave as repressors (Tena et al., 2007).

The earliest known player of the signaling cascade that leads to pectoral fin induction is the RA signaling pathway. RA signaling during gastrulation has been shown to be important for the establishment of a tbx5a-positive fin field (Grandel and Brand, 2010; Grandel et al., 2002). During somitogenesis, RA derived from the somites is involved in maintaining and expanding the tbx5-positive limb precursors (Begemann et al., 2001; Gibert et al., 2006; Linville et al., 2004; Mercader et al., 2006). RA is also required for kidney formation (Cartry et al., 2006) and has been shown to be able to activate the expression of Osr genes when Xenopus animal caps or mouse ES cells differentiated to IM or kidney identities (Drews et al., 2011; Mae et al., 2010). Here, we show that in zebrafish embryos RA signaling is also required for Osr expression. We therefore propose a model in which RA signaling is required to activate the expression of Osr genes in the IM. These transcription factors are then essential at the IM for renal organ formation but also for indirectly promoting pectoral fin development by controlling Wnt2b expression in the kidney anlage. Wnt2b, produced in the Osr-expressing domain, then might diffuse to the LPM to maintain tbx5a expression and promote pectoral fin formation (Fig. 10). In this genetic cascade, in which the RA signaling is relayed through three different tissues, the Osr genes in the IM play an essential linker function that responds to a signal from the somites and transmitting it to the LPM. Moreover, by activating the Osr genes, RA is capable of controlling in a coordinated way two different developmental processes, kidney formation and pectoral fin development. Nevertheless, the RA signaling pathway is probably also controlling limb development in an Osr-independent way as the initiation of fin bud development requires RA but seems not to depend on Osr genes (Begemann et al., 2001; Gibert et al., 2006; Grandel and Brand, 2010; Grandel et al., 2002).

Neither Osr1 nor Osr2 knockout mouse embryos have been reported to present major limb defects (James et al., 2006; Lan et al., 2004; Wang et al., 2005). This could be due to a partial redundancy of both genes in the activation of Wnt2b, or another similar Wnt ligand, at the kidney anlage. A very recent report has examined the consequences of removing both genes at the early limb bud (Gao et al., 2011). In this study the impairment of both genes produced only late developmental defects in joint formation but it did not impair early limb bud development (Gao et al., 2011). However, in these experiments Osr1 was still present in the IM. To examine Osr requirement precisely in the activation of limb bud formation, it would be necessary to eliminate both Osr genes from the early kidney anlage. Nevertheless, there are several reasons to believe that Osr genes are also likely to be necessary for limb bud formation in other vertebrates. First, in all vertebrates examined, proper kidney and limb formation require RA signaling (Begemann et al., 2001; Cartry et al., 2006; Gibert et al., 2006; Grandel and Brand, 2010; Grandel et al., 2002; Linville et al., 2004; Mercader et al., 2006; Mic et al., 2004; Negishi et al., 2010; Niederreither et al., 1999; Zhao et al., 2009). Second, as we report here in zebrafish, this signaling pathway also controls Osr expression in other vertebrates (Drews et al., 2011; Mae et al., 2010). Third, it has also been reported that the kidney anlage is required for limb formation in other vertebrates (Geduspan and Solursh, 1992; Stephens and McNulty, 1981), and kidney formation depends on Osr function in all vertebrates (James et al., 2006; Mudumana et al., 2008; Tena et al., 2007; Wang et al., 2005). Finally, Osr expression at the IM precedes and overlaps that of Wnt2b in all vertebrates examined (Kawakami et al., 2001; Mercader et al., 2006; Mudumana et al., 2008; Ng et al., 2002; Stricker et al., 2006; Tena et al., 2007). Therefore, in all vertebrates, Osr expression is in the right place (the kidney anlage) and at the right time to relay RA signaling from the somites to the LPM, at least partially, through Wnt2b. Functional experiments would be required to demonstrate whether Osr genes also participate in the genetic cascade required for limb bud formation in other vertebrates.

We thank S. Hoppler, J. C. Izpisua-Belmonte, K. Takeshima and M. Allende for reagents. We are especially grateful to F. Casares for critical reading of the manuscript.