Sp8 and Sp9, two closely related buttonhead-like transcription factors, regulate Fgf8expression and limb outgrowth in vertebrate embryos

Understanding the molecular mechanisms that control how outgrowth of different tissues and organs of the embryo are established and maintained is one of the major questions in the study of embryonic development. The advent of molecular and genetic techniques have provided powerful tools with which to dissect the ways in which transcription factors and signaling molecules are used to allow neighboring cells to communicate and adopt positional information, which are provided by the action of embryonic signaling centers. The signaling centers direct the primordia of tissues and organs towards outgrowth and pattern formation for proper morphogenesis. Formation and maintenance of signaling centers should be tightly regulated, spatially and temporally, in order for them to control tissue development properly. Although there are many different model systems in which, by the action of signaling centers, molecular and genetic mechanisms of outgrowth have been studied, the vertebrate limb is one of the best-understood models.

Limb outgrowth requires the formation and maintenance of three different signaling centers: the apical ectodermal ridge (AER) controls proximodistal growth; the zone of polarizing activity, which is located in the posterior mesenchyme is responsible for anteroposterior pattern formation; and the non-ridge ectoderm directs formation of the dorsoventral axis. Their coordinated action constructs the three-dimensional morphology of the limb(reviewed by Capdevila and Izpisua Belmonte, 2001; Niswander,2003; Tickle,2002a). Among these signaling centers in the limb bud, the AER,which is a thickened epithelial structure positioned at the distal edge of the limb bud at the dorsoventral boundary, is pivotal for maintaining limb outgrowth. Surgical removal of the AER results in cell death in the mesenchyme and abrogates limb outgrowth (Dudley et al., 2002; Rowe et al.,1982; Sun et al.,2002). The importance and necessity of the AER in limb outgrowth is a conserved feature of vertebrate development as illustrated in mice, chick and zebrafish (Grandel and Schulte-Merker,1998; Tickle,2002b). Despite recent extensive studies, the molecular and genetic mechanisms that control initiation and maintenance of the AER in these model organisms is far from being understood.

The morphogenesis of the AER can be divided into two processes.

(1) The induction of AER precursor cells in the surface ectoderm that will migrate toward the dorsoventral boundary and form the AER. These cells start to express fibroblast growth factor 8 (Fgf8), a member of the Fgf superfamily that acts as an essential signaling molecule involved in vertebrate limb outgrowth (Crossley et al., 1996b; Lewandoski et al.,2000; Moon and Capecchi,2000; Vogel et al.,1996).

(2) Maturation of the AER that results in formation of the characteristic,thickened structure (Loomis et al.,1998). The initial induction of AER precursor cells depends on the activity of Fgf10 emanating from the lateral plate mesoderm(Min et al., 1998; Ohuchi et al., 1997; Sekine et al., 1999). In the absence of Fgf10, no Fgf8 expression is detected, indicating that induction of the AER precursor cells does not take place, and the AER is not formed.

In addition, ectodermal Wnt/β-catenin signaling and Bmp signaling are essential for induction of Fgf8 expression in AER precursors(Ahn et al., 2001; Barrow et al., 2003; Kengaku et al., 1998; Pizette et al., 2001; Soshnikova et al., 2003). The AER precursors migrate to the dorsoventral boundary, and form a structurally distinguishable AER by apical compaction(Loomis et al., 1998). Secreted from the AER, Fgf8 signals to the underlying mesenchyme, and Fgf10 produced by the distal mesenchyme in turn signals to the AER, resulting in establishment of the reciprocal positive feedback loop that maintains the expression of each one (Ohuchi et al.,1997; Xu et al.,1998). This positive-feedback loop, mediated by different splicing isoforms of Fgfr2 (Arman et al.,1999; Ornitz et al.,1996), maintains the AER and limb outgrowth. Further studies have shown that Wnt/β-catenin activity is also required for maintaining Fgf8 expression in the AER (Barrow et al., 2003; Kawakami et al.,2001; Soshnikova et al.,2003). Our current understanding of AER development implies that the concerted signaling of these and possibly other growth factors leads to the activation of different transcription factors that in turn elicit the instructions that permit proper limb development.

Among the different transcription factors involved in these processes, it was recently shown, by gene targeting analysis, that Sp8, a buttonhead (btd)-like zinc (Zn) finger transcription factor is required for maintaining, but not for initial induction of Fgf8expression in AER precursor cells (Bell et al., 2003; Treichel et al.,2003). The Sp family of transcription factors is united by a particular combination of a C2H2 type Zn-finger DNA-binding domain and btd domain (Bouwman and Philipsen, 2002; Philipsen and Suske, 1999). Eight Sp members have been identified in mouse and human. Although the necessity of Sp8 for mouse limb outgrowth has been reported, it is largely unknown how Sp8 interacts with the previously mentioned signaling pathways to maintain limb outgrowth. For example, although Fgf10, Wnt/β-catenin and Sp8 are factors that are required for AER maintenance and subsequent limb outgrowth, it is not clear whether they are sufficient or if other factors are also required.

To gain further insights into the role of Sp genes during vertebrate limb development, we have performed several experiments. They include the isolation of a novel btd-like Zn-finger transcription factor, Sp9. Sp9contains a btd domain and Zn-finger domain that are highly homologous to those of Sp8. Using mouse, chick and zebrafish embryos, we have studied the embryonic expression pattern, regulation and role of Sp8and Sp9 during limb development. Both Sp8 and Sp9are expressed in the AER, but regulated differently by Wnt and Fgf signaling. Loss- and gain-of-function approaches revealed that both Sp8 and Sp9 positively regulate Fgf8 expression in the AER and contribute to limb outgrowth in vertebrate embryos.

Chick Sp8 and Sp9 were obtained by screening cDNA libraries. The full-length mouse Sp8 was obtained by combining an EST clone and a PCR clone, and mouse Sp9 was obtained as EST clones. Zebrafish sp8 and sp9 were obtained by screening a 24-hours post fertilization (hpf) cDNA library (Stratagene) and 5′RACE. The screening of libraries and PCR were carried out by following standard procedures. The nucleotide sequences of chick Sp8, chick Sp9, mouse Sp9, zebrafish sp8 and zebrafish sp9 are deposited in GenBank with the Accession Numbers AY591906,AY591907, AY591908, AY591904 and AY591905, respectively.

Chick Sp8 and Sp9 probes were derived from a partial open reading frame covering 3′ to the Zn-finger domain and 3′UTR in order to avoid cross hybridization. The mouse Sp8 probe contains 1 kb of the 3′UTR and the Sp9 probe contains 750 bp covering the 5′UTR and partial ORF, 5′ to the btd domain. Zebrafish sp8 and sp9 probes are derived from 3′UTR sequences. Chick Fgf8 and zebrafish fgf8 probes have been described previously (Ng et al., 2002; Vogel et al., 1996). Zebrafish prx1 was cloned by RT-PCR and confirmed by nucleotide sequencing(Accession Number BC053228).

Chick and mouse embryos were examined by whole-mount in situ hybridization and Alcian green cartilage staining as described(Vogel et al., 1996). Zebrafish embryos were examined by whole-mount in situ hybridization as described (Ng et al.,2002).

Mouse embryos deficient for Fgf10(Min et al., 1998; Sekine et al., 1999), Dkk1 (Mukhopadhyay et al.,2001) and Lrp6(Pinson et al., 2000) were used. Zebrafish mutants, heartstrings (hst)(Garrity et al., 2002; Ng et al., 2002), neckless (nkl) (Begemann et al., 2001) and dackel (dak)(Grandel et al., 2000) have been described previously.

The full-length mouse Sp8 and Sp9 were subcloned into RCAS BP(A) vector. In order to construct dominant-active and-negative forms of Sp8 and Sp9, a part of the open reading frame (from the btd domain to the termination codon; ΔN-Sp8 andΔ N-Sp9) was fused to a VP16 activation domain (VP16-) or an Engrailed-repressor domain (EnR-), and subcloned into the RCAS BP(A)vector. The dominant-active β-catenin clone has been described previously(Capdevila et al., 1998). Retroviral production was performed as previously described(Vogel et al., 1996). Staging of chick embryos was according to Hamburger and Hamilton (HH; Hamburger and Hamilton, 1951). Prospective limb fields of chick embryos at HH stage 9-11 were infected with the viruses. An RCAS BP(A)-alkaline phosphatase virus was used as a control and no phenotypic changes in gene expression or limb morphology were observed. The injected embryos were developed until desired stages and fixed for analysis.

Heparin beads were soaked in Fgf10 (1 mg/ml) or Fgf8 (1 mg/ml). AG-X beads were soaked in the Fgf receptor kinase inhibitor SU5402 (Calbiochem), at 2 mg/ml in DMSO. The beads were implanted into stage 19-21 developing chick limb buds as described previously (Kawakami et al., 2003). Chick embryonic fibroblasts were infected with RCAS BP(A)-Wnt3a(Kengaku et al., 1998), and implanted into limb buds as described previously(Wada et al., 1999). Control beads soaked in PBS or DMSO and chick embryonic fibroblasts with RCASBP(A)-alkaline phosphatase were used at the same stage and no change in gene expression was observed. The manipulated embryos were incubated for desired periods and processed for in situ hybridization analysis.

Morpholino oligo nucleotides were designed by and obtained from GeneTools LLC (Eugene, OR). The zebrafish sp8 morpholino lies from nucleotide position –1 to +24, relative to the translation start site:5′-TTTGTTACACGTCGCAGCCAACATG-3′.

The zebrafish sp9 morpholino sequence lies from nucleotide position –14 to +11, relative to the translation start site:5′-CTATAAAACATAGCTGGCTTGTGTG-3′.

The standard control oligonucleotide available from GeneTools was used. The morpholinos were solubilized in 1×Danieu's solution and injected into one-cell stage zebrafish embryos at a range of 5-15 ng/embryo(Ng et al., 2002).

In the course of our experiments to characterize clones involved in chick limb development, we isolated a clone expressed in the AER, which contained a btd-like Zn-finger domain. Screening cDNA libraries resulted in isolation of clones encoding three different btd-like genes, Sp8,Sp5 and a novel Zn finger-containing clone. We termed this new clone chick Sp9, based on the fact the clone contains a btd domain and C2H2-type Zn-finger domain, characteristic domains conserved in Sp family members, and that its deduced amino acid sequence of the C-terminal region was different from that of Sp8 (Fig. 1). Previous analysis, including a large-scale systematic characterization of mouse transcriptome, identified Sp8(Bouwman and Philipsen, 2002; Ravasi et al., 2003). The analyses, however, did not identify Sp9. To clarify whether Sp9 exists in other vertebrates, we searched databases and found a mouse EST clone (Accession Number AW494427 in NCBI database) and a human sequence (hCT1831218 in Celera database). Based upon the deduced amino acid sequence, we concluded that these sequences encoded Sp9(Fig. 1). Subsequently, we screened a zebrafish cDNA library and obtained two closely related Zn finger-containing clones. Sequence analysis revealed that these clones show a high degree of similarity to the mouse Sp8 and Sp9 genes(Fig. 1C). Our analyses indicate that Sp9 is a novel Zn finger containing btd-like factor expressed in human, mouse, chick and zebrafish.

We found that Sp9 and Sp3 are arranged in close proximity on mouse chromosome 2 (293 kb away) with their promoters facing each other,similar to Sp8-Sp4 (chromosome 12), Sp1-Sp7 (chromosome 15) and Sp2-Sp6(chromosome 11) (Bell et al.,2003). We also found that Sp5 is located close to Sp3, but on the opposite side to where Sp9 is located and also further away from Sp3 than Sp9 is (2564 kb). A similar gene arrangement was also observed with human Sp1-Sp9 genes. The arrangement of Sp genes revealed that eight genes are arranged in a pair-wise manner on four different chromosomes, and Sp5 appears to be in a position that is independent of the other Sp genes. This arrangement suggests that Sp genes arose from a common ancestor by tandem duplication and chromosomal duplication, as proposed for other gene families(Agulnik et al., 1996).

The highly homologous Zn finger domain made it difficult to compare amino acid sequences of Sp8 and Sp9 with other Sp family members within the domain(Fig. 1B). Phylogenetic analysis of deduced amino acid sequences of mouse Sp1-Sp9 together with two Drosophila Sp family members, btd and Sp1revealed that mouse Sp8 and Sp9 are closely related to Drosophila Sp1 (Fig. 1A). Sp8 contains characteristic Ser-rich, Ala-rich and Gly-rich stretches in its N-terminal domain. The zebrafish Sp8 is slightly different from the other Sp8 sequences in its N-terminal region, and is shorter than the others (Fig. 1C). The Sp9 sequence contains Ser-rich and Ala-rich domains in its N-terminal region;however, it does not contain the Gly-rich domain found in Sp8. Both Sp8 and Sp9 contain a Gly-rich sequence in the region 3′ to the Zn-finger domain. In all of the Sp8 and Sp9 sequences analyzed in this study, the amino acid sequences from the btd domain to the Zn-finger domain are identical except for one amino acid (Fig. 1B). The high degree of conservation of amino acid sequences among different species hints at their importance in vertebrate evolution.

In order to understand the roles of Sp8 and Sp9, we first decided to analyze their embryonic expression pattern in chick, mouse and zebrafish embryos by in situ hybridization.

Sp8 expression was detected in early stages of chick embryos as an oval and 2 stripes at HH stage 5 (Fig. 2A), which became two lateral stripes at HH stage 7, running along the anteroposterior body axis from the head ectoderm to the anterior region of the primitive streak (Fig. 2B). At HH stage 8, strong expression in the anterior neuroectoderm, which forms the central nervous system, was observed(Fig. 2C), and at HH stage 9, Sp8 is expressed in the most anterior region of the forebrain,midbrain, neural groove and Hensen's node(Fig. 2D). The expression in the midbrain was later confined to the midbrain/hindbrain boundary, a signaling center that controls midbrain development(Fig. 2G)(Chi et al., 2003; Crossley et al., 1996a; Lee et al., 1997). At HH stage 15-16, the expression of Sp8 was broadly observed in the surface ectoderm of the limb-forming fields (Fig. 2E,F, arrowheads), in addition to the neural tube. The signal in the neural tube marks proliferating neural cells, and is excluded from the dorsal and ventral region (Fig. 2F). The expression in the limb field became confined to the distal region of the limb at HH stage 16 and 17, with still scattered signal visible in the surface ectoderm of the limb bud(Fig. 2H,I). Sp8 is strongly expressed in the AER and weakly in the ectoderm in the developing limb bud at HH stage 21 (Fig. 2J,K). At this stage, the neural tube expression was restricted to proliferating interneurons with a reversed triangle shape, and excluded from the dorsal-most and ventral-most regions(Fig. 2L). The expression in the AER was detected throughout later stages of limb development(Fig. 2M).

A similar expression pattern was observed in mouse and zebrafish embryos. Mouse Sp8 expression was detected in the forebrain,midbrain/hindbrain boundary and neural tube(Fig. 2N,O; Bell et al., 2003; Treichel et al., 2003). At the stage of limb bud outgrowth, Sp8 is expressed in a scattered manner in the ventral ectoderm (Fig. 2P,Q), and is later confined to the AER(Fig. 2R). In zebrafish, sp8 was detected in the forebrain, prospective midbrain/hindbrain boundary, tail bud, and neural keel and neural tube in the somitogenesis stages (Fig. 2S-U). sp8 is expressed in the pectoral fin bud, and the signal is restricted to the apical fold, a structure corresponding to the AER in chick and mouse embryos (Fig. 2V-Y).

Overall, the expression pattern of Sp8 is conserved among the different vertebrate species studied. In particular, the expression in the limb buds, forebrain, midbrain/hindbrain boundary and neural tube suggests that Sp8 has a conserved role in the development of these structures during mouse, chick and zebrafish embryogenesis.

Sp9 expression was not detected in early stages of chick development (HH stage 3-7). A strong signal was detected at HH stage 8 in the neural groove and anterior part of the regressing Hensen's node(Fig. 3A). Sp9expression showed a restricted pattern in the nervous system, like Sp8, but is excluded from the forebrain(Fig. 3B-D). The expression was detected in the anterior hindbrain (Fig. 3B,C), which will be confined to the midbrain/hindbrain boundary at HH stage 13 (Fig. 3D). In the developing limb, it is expressed in the AER and weakly in the distal surface ectoderm (Fig. 3E-G). Unlike Sp8, Sp9 is expressed in a small area in the anterior border at HH stage 27 and later (Fig. 3H). After HH stage 28, expression in the AER disappears, and the transcripts start to be detected in the anterior and posterior edges of the autopod, but are excluded from the distal edge(Fig. 3I).

In mouse embryos, Sp9 was detected in the AER during limb development as well as in the distal region of the ectoderm, in a similar manner to that of the chick (Fig. 3J-L). In zebrafish, sp9 is expressed in the forebrain,unlike in the chick and mouse, and in the prospective midbrain/hindbrain boundary during early somitogenesis (Fig. 3M,N), which is narrower than that of sp8. As development proceeds, sp9 is also detected in the hindbrain, as well as in the apical fold of the developing pectoral fin, and this expression extends proximally, as compared with that of sp8(Fig. 3O-R). A section of sp9 hybridized embryos shows its strong expression in the basal stratum of the pectoral fin ectoderm (Fig. 3S) (Grandel and Schulte-Merker, 1998). Like Sp8, the conserved expression of Sp9 in the AER in mouse, chick and zebrafish embryos alludes to the fact that Sp9 may also have a key role during vertebrate limb development.

It has recently been reported that Sp8 has a role in maintaining Fgf8 expression and limb outgrowth in mice(Bell et al., 2003; Treichel et al., 2003). In order to examine whether Sp9 also has a role in limb development, as well as to examine possible molecular and genetic interactions of Sp8and Sp9 with known signaling pathways involved in limb development,we made use of zebrafish pectoral fin mutants. We initially analyzed hst mutants, animals that bear a point mutation in the tbx5-coding sequence that results in a loss of function mutation(Garrity et al., 2002; Ng et al., 2002). Tbx5 is a mesenchymal factor required for limb bud initiation and outgrowth as an upstream regulator of fgf10 in the pectoral fin field(Ahn et al., 2002; Ng et al., 2002). We observed significant downregulation of sp8 and sp9 expression in pectoral fin buds of hst mutants at 40 hpf(Fig. 4B,E), when compared with wild-type embryos (Fig. 4A,D). This result places sp8 and sp9 downstream of tbx5in limb development, and suggests that sp9 as well as sp8must play a role in normal limb outgrowth. Similar results were obtained by using nkl mutant embryos (data not shown), which carry a mutation in the raldh2 gene (Begemann et al.,2001). Given that nkl mutation lies upstream of tbx5 function in the pectoral fin formation, it further confirms that sp8and sp9 are ectodermal factors downstream of mesenchymal signals.

We also studied dak mutant embryos, where fin buds start to grow but their outgrowth fails to be maintained(Grandel et al., 2000). We observed significant downregulation of sp8 and sp9 at 40 hpf when compared with wild-type embryos (Fig. 4C,F). This result further supports the involvement of sp8 and sp9 in fin outgrowth.

These analyses show that expression of sp8 and sp9correlates with fin/limb development in zebrafish, and defects in fin/limb development are associated with their downregulation. The specific mutations in hst and nkl indicates that both sp8 and sp9 are ectodermal factors downstream of tbx5, a mesenchymal factor required for proper limb outgrowth.

The Wnt and Fgf signaling pathways are two of the major pathways that positively control Fgf8 expression in the AER (reviewed by Kato and Sekine, 1999; Martin, 1998; Tickle and Munsterberg, 2001; Yang, 2003). We therefore wanted to investigate whether the expression of Sp8 and Sp9is regulated by these two signaling pathways.

Fgf10 is essential for inducing and maintaining Fgf8expression and AER formation during limb development. In order to analyze whether Fgf signaling also regulates expression of Sp8 and Sp9, we first analyzed their expression in Fgf10^–/– mouse embryos. We observed downregulation of Sp8 and Sp9 in embryos lacking Fgf10 (Fig. 5A-D). This result provides genetic evidence that initial induction of Sp8and Sp9 in the ectoderm depends on Fgf10 signaling from the lateral plate mesoderm, and further confirms the result obtained by zebrafish pectoral fin mutant analysis (Fig. 4).

In order to examine the role of Fgf signaling in a spatially and temporally controlled manner, we carried out an experiment in which Fgf protein and the Fgf receptor inhibitor were applied in the developing chick limb. Implantation of beads soaked with Fgf10 in the anterior distal region of the limb resulted in upregulation of both Sp8 and Sp9 at 6 hours post-implantation (n=18/23 and 15/20 for Sp8 and Sp9, respectively). The expression domain of Sp8 and Sp9 in the AER was extended in the Fgf10-bead-implanted limb, when compared with the contralateral control limb(Fig. 5E,F,I,J). In addition,both Sp8 and Sp9 were upregulated in the broad region exclusively in the limb ectoderm. We also examined the effect of Fgf8-soaked beads in the expression of Sp8 and Sp9 in the developing chick limb. However, we did not observe significant changes in their expression at 6 and 12 hours post-implantation (data not shown). These results indicate that Fgf10 signaling emanating from the mesenchyme is sufficient for expression of Sp8 and Sp9 in the ectoderm. To further examine the requirement of Fgf signaling in their expression in the AER, we implanted beads soaked with a Fgf receptor tyrosine kinase inhibitor, SU5402. Twenty-four hours post implantation, expression of Sp8 and Sp9 are significantly downregulated in close proximity to the beads(Fig. 5G,H,K,L, n=7/9 and n=6/8 for Sp8 and Sp9, respectively). These results demonstrate that Fgf signaling is necessary and sufficient for expression of both Sp8 and Sp9 in the ectoderm.

Wnt3a, which signals through the β-catenin pathway, is expressed in the AER precursors and the established AER in the developing chick limb, and has been demonstrated to regulate Fgf8 expression(Kawakami et al., 2001; Kengaku et al., 1998). In order to examine the role of the Wnt/β-catenin signaling pathway on Sp8 and Sp9 expression, we implanted cells expressing Wnt3a into the anterior distal tip of the chick limb bud at HH stage 19-21. This manipulation resulted in induction of ectopic expression of Sp8 in the limb ectoderm in close proximity to the implanted cell pellet, at 6 and 12 hours post-implantation (n=2/6 at 6 hours, and n=5/6 at 12 hours; Fig. 6A,B). Moreover, retrovirus-mediated overexpression of dominant-active β-catenin resulted in ectopic induction of Sp8in a broad region of the chick limb ectoderm (n=5/5; Fig. 6C), which was associated with the formation of ectopic ridge-like structures(Fig. 6D). Contrary to what was observed with Sp8, we did not observe significant changes in the expression of Sp9 in the limb ectoderm after the same experiments were performed (data not shown).

In mouse embryos, although Wnt3a is not expressed during limb development, another Wnt gene, Wnt3 is expressed in the ectoderm and regulates limb development (Barrow et al.,2003). To further gain insights into the regulation of Sp8 and Sp9 expression by the Wnt/β-catenin signaling pathway, we performed the following experiments in mouse embryos. First, we used Lrp6^–/– embryos for a loss-of-function approach. Lrp5/6 is a component of a Wnt-receptor complex that is required for canonical β-catenin-dependent signaling(He et al., 2004). We observed a variety of phenotypes in Lrp6^–/– embryos,including total absence of limb bud outgrowth, severe truncation of limbs and truncations of the posterior part of the body. We chose embryos with no gross morphological defects, in order to avoid possible misinterpretation because of secondary effects. In E10.5 Lrp6^–/– embryos,we observed downregulation of Sp8 expression in the AER(Fig. 6E,F). Particularly, the domain of expression was shorter when compared with that of the wild-type limb. Second, we used Dkk1^–/– embryos as a gain of function approach. Dkk1 is an extracellular Wnt antagonist, andβ-catenin-dependent signaling is upregulated in Dkk1^–/– embryos(Mukhopadhyay et al., 2001). We observed a stronger Sp8 expression in the AER and ectopic Sp8 expression outside the AER in 12.5 dpc Dkk1^–/– embryos(Fig. 6G,H). However, in both Lrp6^–/– and Dkk1^–/– embryos, we did not observe significant changes in the expression of Sp9 (data not shown). These results provide genetic evidence that Wnt/β-catenin signaling positively regulates Sp8, but not Sp9. Furthermore Wnt regulation of Sp8 appears to be a conserved feature in vertebrate limb development.

To further investigate the roles of Sp8 and Sp9 during limb development, we used gain- and loss-of-function approaches by using viral-mediated expression in chick embryos. First, we overexpressed full-length Sp8 and Sp9 throughout the entire limb bud, and observed expansion of the AER marked by an expanded Fgf8-expressing domain (Fig. 7A,B). This phenotype was observed after overexpressing both Sp8 and Sp9(n=3/17 and 10/60, for Sp8 and Sp9, respectively). This result indicates that Sp8 and Sp9 can positively regulate Fgf8 expression and AER morphology.

The role of Sp8 and Sp9 was also tested by using dominant-active and dominant-negative constructs in which Sp8 and Sp9 were fused to a VP16 or an EnR. We used a N-terminal-deleted form of Sp8 (ΔN-Sp8) and Sp9(ΔN-Sp9) because of the transgene size limitation of RCAS system (∼2 kb) and to avoid possible interferences resulting from the repetitive amino acid sequence in the N-terminal domain (which might interfere with the activity of these constructs). Injection of RCAS-VP16-ΔN-Sp8 and RCAS-VP16-ΔN-Sp9 resulted in a phenotype similar to that of full-length Sp8 and Sp9. The distal region of the limb bud was enlarged and the AER was elongated (n=4/29 and n=4/42 for VP16-ΔN-Sp8 and VP16-ΔN-Sp9, respectively; Fig. 7C,D). Correlating with the expansion of the distal mesenchyme phenotype, we observed ectopic digit formation in some of the affected embryos at later stages (n=3/25; Fig. 7J). These results show that Sp8 and Sp9 act as activators of Fgf8expression during limb development.

Next, we tried to examine the roles of Sp8 and Sp9 by a loss-of-function approach. Injection of RCAS-EnR-ΔN-Sp8 and RCAS-EnR-ΔN-Sp9 produced similar phenotypes, resulting in significant downregulation of Fgf8 expression in the AER(n=3/15 and n=11/28 for EnR-ΔN-Sp8and EnR-ΔN-Sp9, respectively; Fig. 7E,F). This downregulation of Fgf8 could lead to a partial loss of the AER, which was associated with an indentation phenotype in RCAS-EnR-ΔN-Sp8-injected embryos(Fig. 7G,H). Correlating with hypoplasia of the AER, we observed a variety of skeletal defects at later stages in RCAS-EnR-ΔN-Sp8 or RCAS-EnR-ΔN-Sp9 injected embryos. In the most severe cases, we observed a small projection (Fig. 7K). Where limbs should have formed, small cartilaginous rudiments were observed. The shoulder griddle, which is not formed from the limb bud,appears normal. In the milder phenotypes, even though the limb was formed, it lacked several skeletal elements. Consistent with an indentation phenotype(Fig. 7H), Fig. 7L shows a phenotype lacking the radius and digit II. These results strongly support that both Sp8 and Sp9 may act to maintain the expression of key signaling molecules such as Fgf8, and the AER during limb development.

Because the frequency of phenotype is not very high (10-39%), we carefully examined virus infection using an RCAS probe in embryos injected with the aforementioned viruses. In 54-64% of embryos analyzed (n=28-44),we observed widely spread infection of RCAS viruses in the ectoderm, and widely spread strong mesenchymal infection was observed in 25-36% of the embryos. Some embryos showed a phenotype with only infection to the ectoderm(data not shown). This analysis, together with lack of phenotype produced by control RCAS virus (e.g. RCAS-alkaline phosphatase, n=60),indicates that the phenotype observed was specific for each virus injection experiment.

The high degree of amino acid sequence conservation between Sp8and Sp9 raised the possibility that the VP16- and EnR-fusion constructs used, may act redundantly, and the phenotype observed could be not specific for Sp8 or Sp9. To address this possibility, and also to avoid redundancy between Sp8 and Sp9, we performed knock-down experiments in the zebrafish using morpholinos. The effects of sp8 and sp9 morpholinos on the development of the pectoral fin were analyzed using the fgf8 and prx1 markers. prx1 is known to be expressed broadly in the chick limb bud mesenchyme in an AER-independent manner(Nohno et al., 1993), and was used to visualize the morphology of the developing pectoral fin in this study. Injection of either sp8 morpholino or sp9 morpholino resulted in downregulation of fgf8 expression in the apical fold(Fig. 7M-O). As summarized in Table 1, we observed complete loss of fgf8 expression in 10% of sp9 morphants and very faint fgf8 signal in 55% of morphants, resulting in fgf8downregulation in a total of 65% of sp9 morphants at 36 hpf. sp8 morphants showed milder phenotypes, and fgf8 expression was downregulated in 33% of the sp8 morphants. Fig. 7 shows the typical `faint signal' phenotype, in which the signal is nearly invisible(Fig. 7M-O). Consistent with downregulation of fgf8 expression, injection of either sp8morpholino or sp9 morpholino resulted in interfering with pectoral fin outgrowth as visualized by expression of prx1(Fig. 7P-R). Moreover, we observed a synergistic effect of sp8 morpholino and sp9morpholino on the expression of fgf8. Co-injection of sp8morpholino and sp9 morpholino at a lower dose (5 ng each) resulted in downregulation of fgf8 expression in ∼80% of morphants. Increasing the amount of morpholinos (10 ng each) produced higher efficiency of downregulating fgf8 expression, and fgf8 expression was completely abolished in more than half of morphants.

These results clearly indicate that downregulation of function of either sp8 or sp9 is sufficient to downregulate fgf8expression and fin outgrowth. It also confirms the results observed in chick experiments with dominant-negative constructs, and further indicates that the coordinated actions of sp8 and sp9 might be required for fgf8 expression in the zebrafish fin.

Large scale bioinformatic analyses have characterized Zn-finger transcription factors in the mouse transcriptome, including Sp8(Bouwman and Philipsen, 2002; Ravasi et al., 2003). Our experimental approach has identified an additional novel Zn finger transcription factor, Sp9, in human, mouse, chick and zebrafish, as well as Sp8 in chick and zebrafish. The Sp family members are characterized by highly conserved btd domain and Zn-finger domain in their C-terminal region, which binds to the GC box on DNA. There is only one amino acid difference between Sp8 and Sp9 in these regions. Moreover, the amino acid sequence of this region is completely conserved among vertebrate species analyzed in this study. This high degree of sequence conservation during vertebrate evolution suggests not only that Sp8 and Sp9 may have essential roles, but also that they may have redundant activities. The existence of Sp9 in invertebrates has not yet been reported. However, the identification of Sp8 in the beetle as an essential factor for limb outgrowth(Beermann et al., 2004), and the fact that Sp8 can functionally replace btd in Drosophila (Treichel et al.,2003) indicates that Sp8/btd has a common role in appendage development in vertebrates and invertebrates.

As previously demonstrated, the release of the mouse and human genome sequences revealed that Sp genes are arranged in a paired manner on chromosomes; Sp1-Sp7, Sp2-Sp6 and Sp4-Sp8. Our analysis identified Sp9 in close proximity to and in an opposite direction to Sp3. Based upon the arrangement of Sp genes on chromosomes, it is likely that a single primordial gene underwent a tandem duplication event and produced progenitor genes for the Sp1-Sp4 subfamily and Sp6-Sp9 subfamily. Then whole-cluster duplication(s) might have taken place to generate four Sp clusters. This is also supported by the proposal that the diversity of the amino acid sequences outside the Zn-finger domain were created by gene duplication (Kolell and Crawford,2002). A similar scenario to gene evolution has been proposed for the Tbx2 subfamily genes, Tbx2-Tbx5(Agulnik et al., 1996). Our finding that Sp5 is not linked to other Sp genes supports a previously proposed evolutionary mechanism of Sp genes, in which Sp5might be an evolutionary link between the Sp family and KLF family,another Zn finger factor family whose primary structure of the Zn finger domain is related to that of Sp family but lacks the btd domain(Ravasi et al., 2003; Treichel et al., 2001).

While necessity of Sp8 for maintaining Fgf8 expression in mice has recently been demonstrated, its placement in the genetic cascade that permits normal limb outgrowth and possible additional roles remained unknown. Furthermore, our identification of Sp9, a novel btd-like gene generated a new question: does Sp9 have a similar or distinct role from Sp8? In order to try to address these issues and to gain insights into the mechanisms of Sp8 and Sp9 action, we analyzed their possible regulation by different signaling pathways known to play a key role during vertebrate limb development.

Genetic and embryological analyses have revealed that Fgf10-Fgfr2b in tandem is pivotal for inducing the expression of Fgf8 in the AER precursor cells, migration of AER precursors from the surface ectoderm to the dorsoventral boundary, and formation of the AER(Gorivodsky and Lonai, 2003; Min et al., 1998; Ohuchi et al., 1997; Sekine et al., 1999). This also maintains expression of Fgf8 in the AER in the established limb(Arman et al., 1999; Xu et al., 1998). Our analysis using Fgf10^–/– embryos, as well as the manipulation of chick embryos, revealed that both Sp8 and Sp9 are ectodermal targets of Fgf10 signaling emanating from the mesenchyme during initiation and outgrowth of the limb bud(Fig. 5). This is supported by zebrafish mutant analysis, where retinoic acid and tbx5 lie upstream of the fgf10 signaling cascade in the mesenchyme(Begemann et al., 2001; Garrity et al., 2002; Ng et al., 2002). In these mutant embryos, sp8 and sp9 are significantly downregulated(Fig. 4), which further supports the fact that the regulation of sp8 and sp9 by mesenchymal signals is a conserved feature during vertebrate evolution.

In the ectoderm, Wnt/β-catenin signaling is known to be a crucial factor for induction and maintenance of Fgf8 (reviewed by Yang, 2003). Unlike Fgf10,however, we observed differential regulation of Sp8 and Sp9by Wnt/β-catenin signaling in both chick and mouse embryos(Fig. 6). Although we did not observe alteration of Sp9 expression, Sp8 expression was positively regulated by Wnt/β-catenin signaling. The fact that Fgf10 signaling and Wnt/β-catenin signaling can induce Fgf8 in the ectoderm raised the possibility that activation of Sp8 might be mediated through Fgf8 protein (Barrow et al., 2003; Kawakami et al.,2001; Kengaku et al.,1998; Ohuchi et al.,1997). This, however, does not seem to be the case, as exogenously applied Fgf8 could not induce Sp8 and Sp9, while Fgf10 could. Our molecular and genetic analysis positions Sp8 as a downstream factor of Fgf10 and Wnt/β-catenin, while Sp9 is placed downstream of Fgf10, but independent of Wnt/β-catenin.

Our gain- and loss-of-function analyses have unveiled a role of Sp9 and a new role of Sp8 as positive regulators for Fgf8 expression and AER formation. Our results with viral constructs of full length and VP16-fused forms of Sp8 and Sp9 strongly suggest that both genes are able to activate Fgf8 expression as transcriptional activators (Fig. 7). This is also supported by the complementary results obtained using EnR fusion constructs and morpholinos. The fact that the proximal region of the Fgf8 gene is GC-rich and includes several copies of consensus Sp1-recognition sequences (Brondani et al., 2002) further suggests that Sp8 and Sp9directly regulate Fgf8 expression in the limb.

During mouse embryogenesis, the precursors of the AER originate in the ventral ectoderm (Kimmel et al.,2000; Loomis et al.,1998). In the chick, these cells are distributed in the wide range of the surface ectoderm, both dorsally and ventrally(Altabef et al., 1997; Michaud et al., 1997). Interestingly, the distribution of Sp8 transcripts at the time of Fgf8 induction correlates with the appearance of the AER precursors. Expression of Sp8 is detected in the AER precursors with a ventrally biased manner in mouse embryos (Fig. 2P,Q) (Bell et al.,2003; Treichel et al.,2003), and in a wide region of the surface ectoderm in chick(Fig. 2E,F). It has been demonstrated that Fgfr2b, the high-affinity receptor for Fgf10, is expressed widely in the surface ectoderm(Arman et al., 1999). The expression pattern of Sp8, together with the ability of Sp8and Sp9 to induce Fgf8 expression suggests that Sp8contributes to the initial induction of Fgf8 in mouse and chick. Consistent with this, lower levels of Fgf8 expression in the limb-forming area were observed in Sp8^–/–embryos at E9.5, when Fgf8 expression becomes evident in the limb-forming field (Treichel et al.,2003). The initial expression of Fgf8 was not abolished completely in Sp8^–/– embryos; however, this could be due to the redundant activity of Sp9. It will therefore be interesting to analyze double mutants of Sp8 and Sp9.

Our data not only support the previously shown requirement of Sp8in the expression of Fgf8 and limb outgrowth, but also demonstrate that Sp9 is required for Fgf8 expression and limb outgrowth. Our experiment with sequence-specific morpholinos excludes the possibility of redundant activity of the EnR fusion constructs(Fig. 7M-R). Therefore, the fact that downregulation of either Sp8 or Sp9 is sufficient to downregulate Fgf8 expression and limb outgrowth, as well as the synergistic effects produced by co-injection of sp8 morpholino and sp9 morpholino on fgf8 expression in zebrafish, strongly support the argument that these Sp factors may cooperate during normal limb outgrowth. A similar scenario could take place in other regions of the embryo where both genes are co-expressed, such as the midbrain/hindbrain boundary.

Vertebrate limb outgrowth requires proper activity of the AER. As such,induction and maintenance of the AER is a central issue for limb outgrowth and morphogenesis. Molecular and genetic approaches have revealed that Fgf10,Wnt/β-catenin and Sp8 are crucial factors for these processes. Our studies have identified a novel btd-like transcription factor Sp9. Both Sp8 and Sp9 are regulated by Fgf10, and Sp8 is additionally regulated by Wnt/β-catenin signaling. Furthermore, our results indicate that Sp8 and Sp9 mediate the induction and maintenance of Fgf8 expression in the AER precursors and in the established AER, allowing proper limb/fin outgrowth in mouse, chick and zebrafish.

We thank May Schwarz for technical and editorial assistance. We also thank Carles Callol, Ilir Dubova, Hiroko Kawakami, Harley Pineda, Marina Raya and Gabriel Sternik for experimental assistance. We are most grateful to Dr Juan Hurle for electron microscopy images of RCAS-dominant-activeβ-catenin-injected embryos, to Dr Keisuke Sekine for his assistance and to Drs Naoyuki Wada, Setsuko Sahara and Ryuichi Shirasaki for the discussion. We express our appreciation to Dr Thomas Schilling for the zebrafish mutant neckless, and to Dr William C. Skarnes for the Lrp6^–/– mouse. T.M. is partially supported by a fellowship from the Uehara Memorial Foundation and J.R.L. is supported by a fellowship from Fundação para a Ciencia e a Tecnologia. This work was supported by grants from FEDER to J.R.L.; and from the G. Harold and Leila Y. Mathers Charitable Foundation, the National Science Foundation and the National Institutes of Health to J.C.I.B.