sall4 acts downstream of tbx5 and is required for pectoral fin outgrowth

Mutations in the gene SALL4 result in Okihiro syndrome [OS, also called Duane radial ray syndrome (DRRS), OMIM number 607323](Al-Baradie et al., 2002; Kohlhase et al., 2002). OS is caused by SALL4 haploinsufficiency(Borozdin et al., 2004) and is characterised by forelimb defects associated with the eye defect, Duane anomaly. The forelimb defects of individuals with OS range from subtle thumb abnormalities to severely truncated limbs (phocomelia)(Al-Baradie et al., 2002; Kohlhase et al., 2002). The thumb, which is the most anterior digit, is most commonly affected in OS(Borozdin et al., 2004). In addition to these defining features of OS, a range of less common abnormalities has been reported, including atrial septal defects (hole in the heart), ear problems, Hirschsprung's disease and pigmentation defects(Al-Baradie et al., 2002; Kohlhase et al., 2002). Another Sall gene family member associated with developmental defects in humans is SALL1, which when mutated results in Townes-Brocks syndrome(TBS, OMIM number 107480) (Kohlhase et al., 1998). Individuals with TBS have limb defects, abnormal ears,imperforate anus and kidney abnormalities(Kohlhase et al., 1998). The limb defects of individuals with TBS include preaxial polydactyly and triphalangeal thumb in the forelimbs, and syndactyly and club foot in the hindlimbs (Kohlhase et al.,1999). Sall1-null mice do not phenocopy TBS(Nishinakamura et al., 2001);however, mice expressing a truncated form of Sall1 do have TBS-like defects (Kiefer et al., 2003). This suggests TBS results from mutations that produce a truncated,dominant-negative form of SALL1 and is not due to SALL1haploinsufficiency. Consistent with such a model, mutations in SALL1associated with TBS are predicted to form truncated SALL1 protein products.

Sall1 and Sall4 belong to a family of zinc finger transcription factors that share homology to the founding member of the gene family, the Drosophila spalt gene(Reuter et al., 1989). There are four known Sal-like (Sall) members in vertebrates (Sall1-4) that are defined by the presence of an N-terminal Cys2-His-Cys zinc finger. Sall4 has a further seven zinc fingers of the Cys2-His2 type that are arranged into three double zinc-finger domains. An additional zinc finger is found in close proximity to the second double zinc finger(Al-Baradie et al., 2002; Kohlhase et al., 2002). The double zinc finger domains are characteristic of Sall gene family members. In Drosophila, spalt acts downstream of the T-box gene optomotor-blind (omb) and is required for correct patterning of the wing imaginal discs (de Celis et al., 1996; Del Alamo Rodriguez et al., 2004).

Members of the T-box transcription factor gene family are characterised by the presence of a conserved DNA-binding motif known as the T-domain. Mutations in several different T-box genes are associated with developmental disorders(for a review, see Packham and Brook,2003), including TBX5, which, when mutated in humans,results in Holt-Oram syndrome (HOS, OMIM number 142900)(Basson et al., 1997; Li et al., 1997). HOS, which is caused by TBX5 haploinsufficiency, is defined by heart and forelimb abnormalities (Packham and Brook,2003). The limb deformities seen in individuals with HOS range from thumb defects to phocomelia (Basson et al., 1997; Li et al.,1997). There is an anterior bias to the limb defects of individuals with HOS such that the thumb and radius bones are predominantly affected (Packham and Brook,2003). Less common defects reported in individuals with HOS include absent pectoral muscles and eye problems, such as Duane anomaly(Newbury-Ecob et al., 1996). Although TBX5 mutations are associated with HOS it has been predicted that these mutations only account for ∼30% of individuals with HOS(Cross et al., 2000).

Previous loss-of-function experiments in zebrafish and mouse, and misexpression of dominant-negative Tbx5 constructs in chick, have demonstrated that Tbx5 is required for the initiation and outgrowth of the forelimb (for a review, see Logan,2003). Identifying genes that genetically interact with Tbx5 could uncover genes with essential roles in normal limb development and which, when mutated in humans, may result in HOS-like phenotypes. The forelimb defects in individuals with OS and HOS are very similar. In both conditions there is an anterior bias to the limb defects and the left limb is more severely affected than the right. In addition to this phenotypic similarity, several individuals previously diagnosed with HOS, but lacking TBX5 mutations, have subsequently been shown to have mutations in SALL4 (Brassington et al., 2003). Zebrafish are a useful model species with which to study forelimb/pectoral fin development(Fischer et al., 2003; Garrity et al., 2002; van Eeden et al., 1996). We have used zebrafish to investigate the function of sall4 during limb development and to explore the relationship between sall4 and tbx5. We demonstrate an essential role for sall4 during pectoral fin development and show redundant functions between sall gene family members. Our results offer explanations for the similar limb phenotypes of individuals with OS and HOS.

Whole-mount RNA in situ hybridisation was performed essentially as described (Thisse et al.,1993). The following probes have been described previously: dlx2, fgf8, fgf10, fgf24, erm(Fischer et al., 2003), fgfr2 (Poss et al.,2000), sp9 (Norton et al., 2005), tbx5(Begemann and Ingham, 2000) and sall1a (IMAGE consortium–accession number BI880033). A 1.6 kb fragment of sall4 was isolated, with the primers 5′-CTACTAGTGTCTTACTATTCGCCCCTGAT-3′ and 5′-CAGAAGAAATCGATGCACCAT-3′ using RT-PCR on 24 hpf whole embryo RNA, and used as an in situ probe template. Section in situ analysis was performed by wax embedding and sectioning whole-mount preparations. Skeletal preparations were performed as previously described(Grandel and Schulte-Merker,1998).

To overcome problems with morpholino (MO) design, we cloned a fragment of sall4 pre-mRNA that spans the boundary between exon 1 and intron 1,using RT-PCR with RNA from zebrafish lines that we intended to inject. Using the sequence of this clone and zebrafish genomic sequence, we designed a MO that is antisense to the boundary between exon 1 and intron 1 of sall4 pre-mRNA: 5′-CGCTCCAAACTCACCATTTTCTGTC-3′. We used a 5 bp mismatch of this MO as a control:5′-CGgTCgAAACTgACgATTTTCTgTC-3′ (lower case letters indicate altered bases).

To test the efficiency of the sall4 MO, RT-PCR was performed using whole-embryo RNA from ∼20 embryos at 24 hpf, using the primers 5′-TACAAAACTTCTCGAATTCAC-3′,5′-GACATGCGCATTTCTACTCGAGGG-3′ and 5′-AGAATTCCGCAAACCCTTGTCTCCTCCG-3′ to detect spliced and un-spliced sall4 mRNA transcripts.

The sequence of the sall1a MO, which is antisense to the 5′UTR, is 5′-GGCTCACGCATCAGCCACGAAAGAA-3′. The tbx5and fgf24 MOs 5′-GAAAGGTGTCTTCACTGTCCGCCAT-3′ and 5′-GACGGCAGAACAGACATCTTGGTCA-3′, respectively, have previously been described previously (Ahn et al.,2002; Garrity et al.,2002; Fischer et al.,2003). All MOs were obtained from Gene Tools.

Embryos were laid at 10 am and this time was taken as 0 hpf. Embryos were incubated at 28°C and were further staged using criteria previously established (Grandel and Schulte-Merker,1998; Kimmel et al.,1995).

To understand the role of sall4 during limb development, we cloned the zebrafish sall4 homologue and studied its expression during pectoral fin development. Using in situ hybridisation, sall4 mRNA transcripts are first detectable in the mesenchyme and not the overlying ectoderm of the pectoral fin primordia at 22 hours post fertilisation (hpf, Fig. 1A). During early pectoral fin bud stages (32 hpf), sall4 is expressed throughout the fin bud mesenchyme (Fig. 1B). As the pectoral fins mature, transcripts remain detectable throughout the mesenchyme,with highest levels at the distal tip of the fin(Fig. 1C).

To investigate the function of sall4 during pectoral fin development, we designed an antisense morpholino oligonucleotide (MO) that inhibits splicing of sall4 pre-mRNA and subsequently leads to knockdown of sall4 function. Using such a splice-blocking MO is advantageous as the efficiency of gene knockdown can be tested using RT-PCR(Draper et al., 2001). In wild-type embryos, we detect spliced and unspliced sall4 mRNA, using RT-PCR (Fig. 1D,E). Only unspliced sall4 transcripts are detectable in embryos injected with 5 ng of the sall4 MO, whereas there is no effect on splicing in embryos injected with 5 ng of a 5 bp mismatch control MO(Fig. 1E). This demonstrates the sall4 MO can efficiently block production of the mature sall4 spliced transcript.

We allowed sall4 morphant embryos to develop until 3 days post fertilisation (dpf) and compared their pectoral fins with those of wild-type embryos. sall4 morphants have a range of pectoral fin defects, from a complete absence of both pectoral fins to those that develop to approximately wild-type size but are positioned perpendicular to the body(Fig. 1G-J). Injection of higher concentrations of sall4 MO results in an increase in the severity of pectoral fin defects (Fig. 1). All embryos injected with 5 ng of the control MO are apparently wild type at 3 dpf (n=64, data not shown). Some sall4 morphant embryos have pectoral fins that turn rostrally,towards the head of the embryo (Fig. 1I). We, and others (Garrity et al., 2002), have observed a similar pectoral fin phenotype when embryos are injected with low concentrations of a tbx5 MO(Fig. 1K).

We stained 5 dpf sall4 morphant embryos with Alcian Blue to study the individual elements that comprise their pectoral fin defects. At 5 dpf,wild-type zebrafish pectoral fins consists of a scapulocoracoid, postcoracoid process, endoskeletal disc and actinotrichs(Fig. 1L)(Grandel and Schulte-Merker,1998). During the third week in development, these larval pectoral fins begin to be remodelled to form the adult pectoral fin. The scapulocoracoid and postcoracoid process will form the scapula, while the endoskeletal disc will form the proximal radials which articulate the lepidotrichia (fin rays), which form from the actinotrichs(Grandel and Schulte-Merker,1998; Sordino et al.,1995). The most severely affected sall4 morphant pectoral fins only possess proximal elements, the scapulocoracoid and postcoracoid process (Fig. 1M). Other sall4 morphant pectoral fins retain a severely truncated endoskeletal disc (Fig. 1N). We also observe sall4 morphant pectoral fins that have an endoskeletal disc that is decreased in size and truncated distally(Fig. 1O). Any actinotrichs that form in sall4 morphant pectoral fins are truncated and scattered compared with wild-type embryos (Fig. 1O). These loss-of-function experiments demonstrate that although sall4 is not required for the initiation of pectoral fin development,it is essential for outgrowth of the pectoral fins. These results also show that proximal pectoral fin elements, the scapulocoracoid and postcoracoid process, form independently of sall4 function.

Owing to the similar limb phenotypes of individuals with OS and HOS, and because tbx5 expression precedes sall4 in the pectoral fin primordia (Fig. 2A)(Begemann and Ingham, 2000; Ruvinsky et al., 2000), we investigated if tbx5 is required for sall4 expression during pectoral fin development. We used a tbx5 MO of identical sequence to one previously demonstrated to phenocopy the ENU-induced tbx5mutation heartstrings (Ahn et al.,2002; Garrity et al.,2002). We observe, as previously described, that embryos injected with 4 ng of tbx5 MO have no pectoral fins at 3 dpf (data not shown). In tbx5 morphant embryos, sall4 expression is never detectable in the pectoral fin primordia (red arrows, compare Fig. 2B with 2C), although expression in other regions of the embryo is normal (black arrowheads, compare Fig. 2B and 2C). At the same stages, the pectoral fin primordia continues to express tbx5 mRNA transcripts (Fig. 2D),demonstrating that the cells of the fin primordia are still present and that the loss of sall4 expression is not simply due to apoptosis of these cells following MO knockdown of tbx5.

In the mesenchyme of the pectoral fin primordia, tbx5 is required for fgf24 expression (Fischer et al., 2003). The pectoral fins fail to form in fgf24mutant embryos, although the expression of tbx5 is initiated normally(Fischer et al., 2003). fgf24 is initially expressed in the fin bud mesenchyme and is required for the induction of fgf10 expression at 24 hpf, also within the mesenchyme (Fischer et al.,2003). As sall4 expression begins after fgf24(Fig. 1A, Fig. 2A)(Fischer et al., 2003), we tested the possibility that fgf24 acts downstream of tbx5 to initiate sall4 expression in a linear fashion, using an fgf24 MO demonstrated to phenocopy the fgf24 mutant ikarus (Fischer et al.,2003). Embryos injected with 6-9ng of fgf24 MO have no pectoral fins at 3 dpf (data not shown), consistent with previously published results (Fischer et al.,2003). When fgf24 morphant embryos were analysed at earlier stages (26hpf), sall4 expression is maintained in the pectoral fin primordia but in a diffuse pattern(Fig. 2E). This expression pattern is similar to tbx5 in fgf24 morphant pectoral fin primordia (Fig. 2F), and is consistent with the reported disruption of cell migration following loss of fgf24 function (Fischer et al.,2003). These results demonstrate that, although induction of sall4 expression requires tbx5, it is independent of fgf24.

Zebrafish with mutations in fgf10 lack pectoral fins,demonstrating that, like sall4, fgf10 is required for pectoral fin development (Norton et al.,2005). As fgf10 expression is first detected 2 hours after sall4 in the pectoral fin primordia(Ng et al., 2002), we addressed the possibility that sall4 is required for fgf10expression in the developing pectoral fins. For simplicity, we will now refer to embryos injected with 10 ng of sall4 MO as sall4morphants. In sall4 morphant pectoral fins, fgf10 expression initiates but is reduced when compared with wild-type pectoral fins (compare Fig. 3A with 3B, reduced in 51%at 30 hpf, n=35). At later fin bud stages, fgf10 expression is downregulated in anterior regions of sall4 morphant pectoral fins(compare Fig. 3C with 3D,downregulated in 83%, n=29), demonstrating that sall4 is required for correct fgf10 expression during pectoral fin development.

During mouse and chick limb development, Fgf10, which is expressed in the mesenchyme, signals to the overlying ectoderm to activate the expression of Fgf8 in cells that will form the apical ectodermal ridge (AER). In turn, FGF8 positively regulates the expression of Fgf10 in the mesenchyme, thereby establishing a positive feedback loop in which ectodermal and mesenchymal FGFs maintain the expression of one another. This feedback loop is essential for limb outgrowth (for a review, see Martin, 1998). We therefore predicted that the downregulation of fgf10 in the mesenchyme of sall4 morphant pectoral fin buds would lead to the downregulation of ectodermal FGFs and a breakdown in FGF signalling in the fin bud. During normal pectoral fin development, fgf24 is expressed in the mesenchyme from 18 hpf until ∼28 hpf when it then becomes downregulated in the mesenchyme and begins to be expressed in the overlying ectoderm(Fig. 2A)(Fischer et al., 2003). In sall4 morphant embryos, expression of ectodermal fgf24 and fgf8 are downregulated (fgf24: 36%, n=28; fgf8: 22%, n=58) or absent (fgf24: 36%; fgf8: 74%) from the fin ectoderm at 40 hpf but remains normal in other regions of the embryo (Fig. 3E,F; data not shown). dlx2 and sp9 are also expressed in the fin bud ectoderm and their expression is positively regulated by FGF signalling from the fin bud mesenchyme(Fischer et al., 2003; Norton et al., 2005). At early time points in pectoral fin development (32 hpf) dlx2 and sp9 expression is present in the ectoderm of all sall4morphant fin buds (dlx2 n=24; sp9 n=10). However, in more mature sall4 morphant fin buds (40 hpf), dlx2 and sp9 expression is downregulated (dlx2: 75% n=16; sp9: 58% n=12) or absent (dlx2 25%; sp925%) (Fig. 3G-J). In those sall4 morphant fin buds in which dlx2 and sp9expression is downregulated, we observed that although transcripts remain detectable in the posterior fin bud ectoderm(Fig. 3H,J, red arrowheads)they are absent from the anterior ectoderm (black arrowheads), consistent with the loss of fgf10 expression in the anterior of sall4morphant fin buds. The transcription factor erm is expressed throughout the fin bud mesenchyme and its expression is positively regulated by FGF signalling (Fischer et al.,2003; Roehl and Nusslein-Volhard, 2001). In sall4 morphant pectoral fins, erm expression is initially unaffected but is downregulated at 40 hpf(compare Fig. 3K and 3L), while remaining normal in other regions of the embryo. These results show that sall4 is required for fgf10 expression in the developing pectoral fins and the downregulation of fgf10 expression in sall4 morphant pectoral fins results in a breakdown in FGF signalling in the fin bud.

During mouse limb development, Sall1 and Sall3 are expressed in overlapping domains and deletion of either gene individually does not produce a limb phenotype(Nishinakamura et al., 2001; Parrish et al., 2004). This suggests Sall genes have redundant functions during mouse limb development. In sall4 morphant embryos, fgf10 expression is downregulated only in the anterior fin bud (Fig. 3D), suggesting that another Sall gene family member may perform a similar function to sall4 in the posterior fin bud. The expression patterns of sall1a, sall1b and sall3 during zebrafish embryonic development have previously been described(Camp et al., 2003). Of these three genes, only sall1a is expressed during pectoral fin development(Camp et al., 2003). sall1a is weakly expressed in the pectoral fin primordia at 24 hpf and becomes more visible at 26 hpf (Fig. 4A). At later fin bud stages, sall1a expression is seen in both the mesenchyme and ectoderm, and at greatest levels in the distal fin bud (Fig. 4B). This expression pattern is comparable with that of mouse and chick Sall1 during limb development (Buck et al., 2001; Farrell and Munsterberg,2000).

To address whether sall1a plays a role in pectoral fin development, we used a MO to knockdown sall1a mRNA translation and compared the pectoral fins of 3dpf sall1a morphants with those of wild-type embryos. Embryos injected with the sall1a MO have truncated and often absent pectoral fins, demonstrating sall1a is required for pectoral fin outgrowth (Fig. 4C). sall1a morphant pectoral fin defects differ from those of sall4 morphants, as we never observe upturned pectoral fins in sall1a morphants (see table in Fig. 4). We stained 5 dpf embryos injected with 2 ng of sall1a MO with Alcian Blue to study the skeletal defects. Proximal skeletal elements such as the postcoracoid process always form in sall1a morphant embryos(Fig. 4D). We also observed sall1a morphant pectoral fins in which the endoskeletal disc and actinotrichs are severely abnormal (Fig. 4E). The sall1a morphant pectoral fin defects observed are comparable with those seen in embryos injected with the sall4 MO. To understand the regulation of sall1a during pectoral fin development, we studied sall1a expression in tbx5 and fgf24 morphant embryos. sall1a is not expressed in tbx5 morphant pectoral fin primordia(Fig. 4G), but is expressed in the pectoral fins of fgf24 morphant embryos(Fig. 4H). sall1a is expressed in a diffuse pattern in fgf24 morphant pectoral fin primordia when compared with sall1a expression in wild-type pectoral fins (Fig. 4I). This expression pattern is consistent with a disruption in cell migration following loss of fgf24 function (Fischer et al.,2003) and is comparable with sall4 and tbx5expression in fgf24 morphant embryos(Fig. 2E,F). These results demonstrate that like sall4, sall1a expression in the developing pectoral fins is dependant on tbx5 but independent of fgf24.

As sall1a could be responsible for the maintenance of the posterior domain of fgf10 expression in the pectoral fin bud of sall4 morphants, we studied fgf10 expression in sall1a morphant embryos (embryos injected with 5 ng of sall1a MO). fgf10 expression initiates in sall1amorphant pectoral fin primordia but at reduced levels compared with wild-type embryos (Fig. 4J,K). At later fin bud stages, fgf10 expression is downregulated but most profoundly in the posterior of sall1a morphant fin buds(Fig. 4F, compare with Fig. 3C). We also studied the expression of the ectodermal fin bud markers dlx2 and sp9 in sall1a morphant embryos. At 32 hpf sp9 expression is downregulated in the posterior of sall1a morphant fin buds(Fig. 4L, black arrowhead, 13%n=24), but continues to be expressed in the anterior (red arrowhead). dlx2 is also expressed in all sall1a morphant fin buds at 32 hpf (n=25), but at 40 hpf becomes downregulated (37%, n=16)or is absent (63%). In those embryos displaying a downregulation of dlx2 expression, transcripts are detectable in the anterior fin bud ectoderm but are absent in the posterior(Fig. 4M). The preferential downregulation of dlx2 and sp9 expression in the posterior fin bud ectoderm is consistent with the downregulation of fgf10 in the posterior mesenchyme of sall1a morphant pectoral fins.

As sall1a and sall4 appear to perform similar roles in positively regulating the expression of fgf10 during pectoral fin development, we studied the phenotype of sall1a/sall4 double morphant embryos. The pectoral fins fail to form in the majority of embryos injected with 4 ng of sall1a and 4 ng of sall4 MO(Fig. 5, table). Methylene Blue stained sections of 48 hpf sall1a/sall4 double morphant embryos shows that, similar to fgf10^–/– zebrafish(Norton et al., 2005), a fin bud initially forms in these embryos (Fig. 5A-B). At 26 hpf, fgf10 expression is lost in sall1a/sall4 double morphant pectoral fin primordia, although it is expressed normally in other regions of the embryo(Fig. 5G). At 32 hpf expression domains of both dlx2 and sp9 are absent in the fin bud ectoderm of sall1a/sall4 double morphant embryos(Fig. 5C-F; dlx2 90%n=52; sp9 81% n=27). These results demonstrate that sall1a and sall4 perform common, semi-redundant roles in initiating the expression of fgf10 in the pectoral fin primordia. Furthermore, in the absence of sall4 function, sall1a is able to maintain the posterior domain of fgf10 expression, while following knockdown of sall1a function, sall4 can maintain the anterior domain of fgf10 expression.

Our results, together with those of others, have demonstrated that fgf10 expression in the developing pectoral fins is dependant on sall1a, sall4 and fgf24(Fig. 5G)(Fischer et al., 2003),although sall1a and sall4 expression is not dependant on fgf24. For fgf24 to activate the expression of fgf10 it must signal via an FGF receptor. We therefore investigated if the expression of an FGF receptor is regulated by sall1a and sall4. As fgf10 expression initiates in the pectoral fins of both sall1a (Fig. 4K)and sall4 (Fig. 3B)morphant embryos, but is not expressed in embryos injected with both sall1 and sall4 MO (Fig. 5G), we predicted that expression of this receptor will not initiate in embryos injected with both sall1a and sall4 MO. Limb outgrowth fails to occur in mice lacking Fgfr2(De Moerlooze et al., 2000; Xu et al., 1998) and therefore we studied the expression of fgfr2 during zebrafish embryonic development. fgfr2 expression is first detectable in the pectoral fin primordia mesenchyme at 23 hpf and is not expressed in the overlying ectoderm(Fig. 6A). fgfr2expression therefore initiates after sall1a and sall4transcripts are first detected in the fin bud mesenchyme. At 24 hpf, fgfr2 is expressed in the pectoral fin primordia of wild-type(Fig. 6B) and fgf24morphant (Fig. 6D) embryos, but is absent from the pectoral fins of embryos injected with 4 ng of sall1a and 4 ng of sall4 MO(Fig. 6C). This demonstrates that fgfr2 expression in the pectoral fin primordia is dependant on sall1a/sall4, but not on fgf24.

During vertebrate limb development, a cascade of signals are required to initiate and maintain limb outgrowth (for a review, see Logan, 2003). Our studies of sall4, the gene mutated in Okihiro syndrome, add another factor to the series of events that control limb outgrowth. Expression of Sall4in the developing limb has been conserved in several species(Barembaum and Bronner-Fraser,2004; Neff et al.,2005; Kohlhase et al.,2002) (S.A.H. and M.P.O.L., this study and unpublished). We have shown that sall4 is required for outgrowth of the pectoral fins, but not the initiation of pectoral fin development, as tbx5 and fgf24 are induced normally and proximal skeletal elements form in sall4 morphants. The upturned fin phenotype found in some sall4 morphant embryos (Fig. 1I) and those injected with low concentrations of tbx5 MO(Fig. 1K) demonstrate that reduction of sall4 and tbx5 function produces similar fin defects in zebrafish. This is consistent with the similarity of limb phenotypes seen in individuals with OS and HOS, which are caused by haploinsufficiency of SALL4 and TBX5, respectively. From fish to mammals, Fgf10 has an evolutionary conserved function that is essential for limb outgrowth (Min et al.,1998; Sekine et al.,1999; Norton et al.,2005). Disruption of Fgf10 signalling is the common cause of the similar abnormalities that arise from fish to humans, following perturbation of either Tbx5 or Sall4 function.

In the pectoral fin primordia, sall1a, sall4 and fgf24expression is dependant on tbx5(Fig. 2)(Fischer et al., 2003);however, expression of either sall1a/sall4 or fgf24 can occur independently of one another (Figs 2 and 4). Therefore, tbx5activates the expression of two different sets of target genes, both of which are required for pectoral fin outgrowth(Fig. 6E)(Fischer et al., 2003). sall1a/sall4 and fgf24 are required for the initiation of fgf10 expression and we have addressed how this interaction occurs. fgf24 must signal via a receptor to activate the expression of fgf10 in the pectoral fin primordia. In zebrafish, fgfr2 is first expressed in the pectoral fin primordia at 23 hpf, just after sall1a/sall4 expression is first detected and just prior to the initiation fgf10. The temporal and spatial expression pattern of fgfr2 therefore makes it a good candidate receptor to mediate the activation of fgf10 expression by fgf24. As sall1aand sall4 are zinc-finger transcription factors they are good candidates to directly positively regulate the expression of fgfr2,although conflicting data exists regarding whether Sall genes act as transcriptional activators or repressors(Kiefer et al., 2002; Li et al., 2004; Netzer et al., 2001; Onai et al., 2004). Our results support a model (Fig. 6E) in which sall1a/sall4 act as transcriptional activators to positively regulate fgfr2 transcription, and that fgf24 signals via fgfr2 to initiate fgf10expression in the fin bud mesenchyme.

Collectively, these results show that tbx5 regulates the expression of fgf10 in the pectoral fin primordia using a feed-forward model of transcriptional regulation(Fig. 6E). Feedforward transcriptional motifs have been most comprehensively characterised in studies in E. coli (Shen-Orr et al.,2002) and S. cerevisiae(Lee et al., 2002). In one branch of the pathway tbx5 activates the expression of sall1a/sall4, which in turn induce fgfr2 expression, and in the other branch tbx5 activates the expression of fgf24(Fig. 6E). The delay between the initiation of tbx5 and sall1a/sall4 expression suggests that this regulation may be indirect or that tbx5 requires a co-factor to activate sall1a/sall4 expression. A third possibility is that higher threshold levels of tbx5 protein are required to activate different target genes. tbx5 is likely to directly activate the expression of fgf24 as expression of fgf24 is detected only 1 hour after tbx5 (Begemann and Ingham, 2000; Fischer et al.,2003). In the mouse, Tbx5 has been shown to regulate the expression of FGFs directly (Agarwal et al., 2003). The expression of fgf24 in the pectoral fin primordia begins at 18 hpf, ∼6 hours before fgf10 expression commences at 24 hpf. fgfr2 expression is detected at 23 hpf(Fig. 6E). During the interval between the initiation of fgf24 and fgfr2 expression, we predict that fgf24 protein levels accumulate in the absence of receptor. Presumably when fgfr2 expression initiates, fgfr2 proteins are rapidly occupied by ligand, owing to the presence of a reservoir of fgf24. Although our results do not provide an explanation for the apparent `priming'of FGF signalling, we predict that in the pectoral fin mesenchyme the dynamics of this regulation would favour a paracrine rather than an autocrine mode of signalling, and would produce rapid, robust and uniform signalling via the FGF receptor.

Studies in mouse and chick have shown that Fgf10, which is expressed in the mesenchyme of the developing limb buds, signals to the overlying ectoderm to induce Fgf8 expression. These ectodermal and mesenchymal FGFs form a positive-feedback loop that is essential for outgrowth of the developing limbs (for a review, see Martin, 1998). Although this positive-feedback loop has been described in mouse and chick limb buds, it has been studied less in zebrafish pectoral fins. In zebrafish, the situation is different from chick and mouse, owing to the presence of fgf24. In chick, mouse and humans, no fgf24 gene has been found, and it appears fgf24 has been lost in the terrestrial vertebrate lineage(Draper et al., 2003). In zebrafish pectoral fin primordia mesenchyme, fgf24 is required for the expression of fgf10 (Fischer et al., 2003). Ectodermal FGFs fail to be expressed in fgf24 mutant fin buds (Fischer et al., 2003), demonstrating that, similar to limb development in higher vertebrates, mesenchymal FGFs are required for the induction of ectodermal FGF gene expression. However, previously it has been unclear whether fgf24 or fgf10 were required for induction of ectodermal FGFs. The downregulation of fgf10 in sall4morphant fin buds, and the subsequent loss of ectodermal FGF expression,suggests that it is fgf10, rather than fgf24, that is required for the induction of FGF expression in the overlying ectoderm(Fig. 3). These results are also supported by observations of zebrafish fgf10 mutants(Norton et al., 2005). The induction of ectodermal FGFs only after the time point at which fgf10is first expressed, and long after the induction of fgf24 expression(Fig. 6E), also supports a model in which fgf10 is the crucial mesenchymal signal.

The scapulocoracoid, a proximal pectoral fin skeletal element that is equivalent to the scapula in higher vertebrates, is always present in sall4 morphant embryos (Fig. 1M) and therefore forms independently of sall4 function. This differs from tbx5 and fgf24 mutant embryos, which lack all pectoral fin structures, including the scapulocoracoid(Ahn et al., 2002; Garrity et al., 2002). These experiments suggest that specification of proximal pectoral fin structures is dependant on tbx5 and fgf24 function and may occur at stages prior to the initiation of sall4 expression. A parallel situation occurs during mouse limb development as Tbx5 conditional knockouts lack all forelimb structures including the scapula and clavicle(Rallis et al., 2003), while Fgf10-null mice possess a scapula rudiment(Min et al., 1998; Sekine et al., 1999). The formation of these proximal skeletal elements also suggest that fgf24performs functions other than just the induction of fgf10 expression. Although the limb defects of individuals with OS and HOS are very similar,there are some clear differences. Defects affecting the proximal forelimb,such as hypoplastic clavicles, have been reported in individuals with HOS(Newbury-Ecob et al., 1996)but never in individuals with OS. Our data suggest that these proximal forelimb defects are not observed in individuals with OS, as these structures are specified independently of SALL4 function. Defects affecting proximal limbs elements such as the clavicle should therefore be specific to HOS and not OS.

The preferential downregulation of fgf10 in the anterior of sall4 morphant fin buds (Fig. 3) led us to investigate whether a sall4-related gene is required to maintain the posterior domain of fgf10 expression. Although the expression of sall2 is yet to be described during zebrafish development, it appears that the only other Sall gene expressed in the developing pectoral fins is sall1a(Fig. 4A,B). Interestingly,although sall4 is required for the anterior domain of fgf10expression in the fin bud (Fig. 3D), sall1a is required for the posterior domain(Fig. 4F). In sall1a(Fig. 4K) or sall4(Fig. 3B) morphant pectoral fin primordia, fgf10 expression initiates; however, it fails to commence in sall1a/sall4 double morphant embryos(Fig. 5G). This suggests that the functions of sall1a and sall4 are partially redundant,such that fgf10 expression initiates in the primordia in the absence of either gene individually, but at later stages is absent in either the anterior or posterior fin bud.

The pectoral fin defects observed following loss of sall1afunction are different from other vertebrates, as Sall1-null mouse embryos do not have a limb phenotype(Nishinakamura et al., 2001). This difference in phenotype can be explained by the variation in expression of a related gene, Sall3, that is expressed in an almost identical pattern to Sall1 during mouse limb development(Nishinakamura et al., 2001; Ott et al., 2001) but is not expressed in the developing zebrafish pectoral fins(Camp et al., 2003). Sall1 is most closely related to Sall3, suggesting that Sall1-null mice do not have a limb phenotype because Sall3can compensate for the loss of Sall1. As sall3 is not expressed during zebrafish pectoral fin development, it cannot substitute for sall1a and as a result sall1a morphant embryos have truncated pectoral fins.

Our studies of tbx5 and sall4 function during zebrafish pectoral fin development offer explanations to the similar limb defects seen in individuals with HOS and OS. We have shown that sall4 is a target of tbx5 and that tbx5 and sall4 act in a pathway required to establish an FGF signalling loop that signals between the mesenchyme and ectoderm of the fin bud. During normal limb development, FGFs expressed in the AER are an essential component of a feedback loop between the ectoderm and underlying distal mesenchyme that is required to maintain FGF signalling (for a review, see Martin,1998). The result of disrupting this positive feedback loop is demonstrated in classical embryological experiments in the chick in which the AER is surgically removed. When anterior regions of the AER are removed, limbs develop that lack anterior skeletal elements(Saunders, 1948). Similarly,alteration of either Tbx5 or Sall4 function preferentially leads to a disruption of Fgf10 in the anterior of the limb bud(Rallis et al., 2003) (this study) and it is loss of FGF signalling in this region that ultimately causes the anterior bias of the deletion deformities characteristic of both HOS and OS. An unresolved issue that remains is why the anterior fin bud is sensitive to the loss of sall4 function and tbx5 haploinsufficiency,as both genes are expressed uniformly throughout the early fin bud. A contributing factor could be that partial redundancy of Sall-related genes leads to the maintenance of fgf10 expression in the posterior limb. Another, not mutually exclusive, explanation is that sall4 is more susceptible to tbx5 levels than other Sall-related genes expressed in the limbs.

We thank Sebastian Gerety and David Wilkinson for the sall1a MO and sall1a in situ probes, and Wendy Hatton of histology (NIMR) for sectioning. We are grateful to Carl Neumann, Philip Ingham and Elke Ober for providing in situ probes, and to Will Norton and Carl Neumann for very generously sharing unpublished data. We also thank past and present laboratory members for their critical input and support. S.A.H. and M.P.O.L. are funded by the Medical Research Council; M.P.O.L. has received funding from the EMBO young investigators program.