Shh and Wnt signaling pathways converge to control Gli gene activation in avian somites

In vertebrate embryos, paraxial mesoderm gives rise to multiple tissues, including skeletal muscles, cartilage, bones and dermis (Christ and Ordahl, 1995). The specification of these cell types requires multiple regulatory processes that are precisely coordinated with segmentation of the paraxial mesoderm to form somites (Borycki et al., 1997; Borycki and Emerson, 2000). Regulatory genes that control segmentation are expressed periodically in the segmental plate mesoderm and at the boundary between segmental plate mesoderm and newly formed somites (Palmeirim et al., 1997; McGrew and Pourquie, 1998). When somite formation initiates, the most anterior segmental plate mesoderm undergoes a mesenchymal-epithelial transition that leads to the formation of epithelial somites. In contrast to the process of segmentation, which is cell-autonomous, somite formation is controlled by unidentified extrinsic signals from the overlying surface ectoderm (Borycki et al., 1997; Sosic et al., 1997; Palmeirim et al., 1998). Following somite formation, cells in the newly formed somite acquire competence to respond to signals from surrounding tissues (Borycki et al., 1997). These signals positively and negatively control the dorso-ventral and medio-lateral patterning of gene expression in the somite (Munsterberg and Lassar, 1995; Pourquie et al., 1995; Pownall et al., 1996), leading to the determination of the different somite cell lineages (Lassar and Munsterberg, 1996; Borycki and Emerson, 2000; Dietrich, 1999).

Sonic hedgehog (Shh) produced by the ventral notochord and floor plate is essential for the induction of the muscle determination factors, MyoD and Myf5, and the sclerotome determination factor, Pax1, in somites of avian embryos (Johnson et al., 1994; Ebensperger et al., 1995; Munsterberg et al., 1995; Borycki et al., 1998). Additional signaling molecules, however, interact cooperatively with Shh to achieve proper activation and spatial patterning of gene expression and cell determination in somites. Wnt signals from the dorsal neural tube are required to maintain the epithelial architecture of the dermomyotome in the medial somite (Ikeya and Takada, 1998), and neural tube Wnts participate with Shh to control MyoD expression (Munsterberg et al., 1995) and to induce Noggin, which represses BMP4 signaling from the lateral plate to establish the medio-lateral somite axis (Hirsinger et al., 1997; Marcelle et al., 1997; Borycki and Emerson, 2000). Signals from the surface ectoderm are required for the survival and for the maintenance of epithelial organization of dermomyotomal cells through the activation of Pax3 (Fan and Tessier-Lavigne, 1994; Dietrich et al., 1997; Borycki et al., 1999b), and through the activation of Paraxis (Sosic et al., 1997), respectively. Finally, together with lateral plate BMP signals, they induce hypaxial lineage determination (Pourquie et al., 1996; Dietrich et al., 1998).

In order to identify the mechanisms that control the initiation of Shh signaling during somite formation, we have investigated the expression of Gli genes in somites of quail embryos. Gli genes are a family of three zinc finger transcription factor effectors that transduce Shh signals in responding cells. In mouse, human and Xenopus, three Gli genes have been identified: Gli1, Gli2 and Gli3 (Kinzler et al., 1988; Ruppert et al., 1990; Hui et al., 1994; Marine et al., 1997). These vertebrate Gli genes are homologues of the Drosophila transcription factor, Cubitus interruptus (Ci) (Alexandre et al., 1996). In avians, Gli1 and Gli2 have been previously identified (Marigo et al., 1996b). Gli1 is preferentially activated in the ventral somite where Pax1 is activated and sclerotome forms, whereas Gli2 is first activated throughout the somite and then becomes restricted in its expression to the myotome in the dorsal somite (Borycki et al., 1998).

Here we report on the identification of an avian Gli3 cDNA and present an experimental analysis of Gli1, Gli2 and Gli3 regulation during somite formation. These studies show that Gli3 is co-activated with Gli2 and Gli1 at the onset of somite formation in quail embryos, providing a basis for the initiation of Shh signaling in newly formed somites (Borycki et al., 1998). Surprisingly, however, in vivo and in vitro experiments provide evidence that each of these Gli genes is controlled differently. Gli2 and Gli3 are regulated by Wnt signaling from the surface ectoderm and neural tube, and these signals are probably transduced through β-catenin. Surface ectoderm/ neural tube Wnt signaling has both negative and positive functions that switch at the time of somite formation. In segmental plate mesoderm, Wnt signaling represses Gli3, but in segmental plate undergoing somite formation, Wnt signaling promotes somite formation and the activation of Gli2 and Gli3. In contrast, Gli1 activation is regulated by Shh signaling, which initiates at the time of somite formation and which also functions to dorsalize the expression of Gli2 and Gli3 to myotome progenitors during somite maturation. These findings, therefore, establish a role for Wnt signaling in the control of the Shh signaling pathway through the regulation of Gli2 and Gli3 activation, and provide a mechanistic basis for the known synergistic action of surface ectoderm/neural tube and notochord signaling in somite cell specification (Buffinger and Stockdale, 1994; Fan and Tessier-Lavigne, 1994; Munsterberg et al., 1995; Stern et al., 1995; Borycki et al., 1997, 1998, 1999a; Fan et al., 1997; Tajbakhsh et al., 1998).

A Hamburger-Hamilton (HH) stage-12 quail cDNA library cloned into pBK-CMV vector (Stratagene) was used as template in a Polymerase Chain Reaction (PCR) with a T3 primer and a degenerate primer designed from the published human, mouse and Xenopus Gli3 sequences (Gli3-3′ZF: GCCTTCCCARTGGCAGTTTGTCTC; in these sequences, N=A,G,T,C; Y=C,T; R=A,G) to amplify a product of approximately 1300 bp. This PCR fragment was further purified by PCR amplification using the primer Gli3-3′ZF and a nested degenerate primer Gli3-5′ (Gli3-5′: CCNTACATYARYCCNTAYA-TGGAC), which produced a 900 bp cDNA fragment with 88% identity to human GLI3. This 900 bp quail Gli3 cDNA was then used as a probe to screen a stage-12 quail cDNA library. A full-length quail Gli3 cDNA was isolated and sequenced on both strands (GenBank accession numbers: AF231111 and AF231112).

RatB1a fibroblasts (Finney and Bishop, 1993) were infected with the replication-defective retrovirus vector MV7 or with equivalent retroviruses carrying Wnt-1 or Wnt-4 cDNAs, as described (Jue et al., 1992). The cells were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM) (Gibco-BRL) containing 10% Fetal Calf Serum (FCS) (Gibco-BRL) and 1× Penicillin/Streptomycin/Fungizone (Gibco-BRL). 3 days before coculture with segmental plate explants, cells were trypsinized and transferred into 60 mm Petri dishes (Falcon) where they grew in suspension forming cell aggregates, which were cultured adjacent to segmental plate explants in collagen gels, as described below.

HH stage 12-13 quail embryos were obtained by incubating Coturnix coturnix japonica eggs for 45-48 hours at 38°C in humidified atmosphere. Embryos were transferred onto paper rings (Whatmann). Surgeries, including notochord separation, axial tissue separation and surface ectoderm removal were performed as previously described (Borycki et al., 1998). Embryos were then cultured for 7 or 16 hours using a modified New culture system, in presence of a mix of 90% thin albumin and 10% PBS/streptomycin. Segmental plate explants were dissected from quail embryos at HH stage 12-13. When indicated, surface ectoderm was removed after incubation for 45 seconds in Dispase I (Boehringer Mannheim). Explants were then cultured for 16 hours at 37°C in a CO₂ incubator on two different substrates: (1) on a 12-well plate coated with 1% agarose in DMEM-10% FCS (1:1); LiCl (Sigma) was added to the culture medium and the agarose at the final concentration of 10 mM when indicated. A range of LiCl concentrations from 2 to 20 mM was initially tested, and 10 mM LiCl gave optimal effects on gene expression with no observed cytotoxicity; or (2) embedded in a collagen gel (Boehringer Mannheim) according to the manufacturer’s protocol, and cocultured adjacent to Wnt- or Mv7-expressing cell aggregates in the presence of DMEM-10% FCS. Embryos and explants were then transferred into an Eppendorf tube, rinsed with PBS, fixed at 4°C in 4% formaldehyde, 2 mM EGTA in PBS and analyzed by in situ hybridization, as described below.

Embryos and explants were fixed in 4% formaldehyde, 2 mM EGTA in PBS at 4°C for 16 hours and analyzed by in situ hybridization as previously described (Henrique et al., 1995) using DIG-labeled (Boehringer Mannheim) RNA probes. RNA probes were synthesized from chick Gli1 and Gli2 cDNA (Marigo et al., 1996b) and quail Gli3 cDNA. Embryos were photographed using a video camera (Leica) and embedded in 4% low-melting point agarose (Sigma), and 100 μm transverse sections were cut using a Vibrotome 1000. Sections were photographed on a DIC microscope (Leica).

Two chicken Gli proteins were previously identified (Marigo et al., 1996b) and classified as Gli1 and Gli2 (Borycki et al., 1998). As the mammalian Gli gene family comprises three members, we conducted a PCR-based cloning and library screen to isolate a 5025 bp full-length quail Gli3 cDNA clone. Sequence analysis reveals that this quail clone is homologous to vertebrate Gli3 cDNAs (Fig. 1). Quail Gli3 encodes a 1576 amino acid protein that has 66% identity with Xenopus Gli3, 69% identity with mouse Gli3 and 70% identity with human Gli3. Quail Gli3 also has 98-99% sequence homology with human, Xenopus and mouse Gli3 in the zinc finger domain and displays highly conserved sequences in the N- and C-terminal domains (Fig. 1), some of which have been described previously (Ruppert et al., 1990; Marine et al., 1997). The quail Gli3 cDNA we isolated has an additional 65 bp insertion in its 5′ region, probably resulting from alternative splicing (Fig. 1). RT-PCR analysis, using primers on either side of the insertion site, was performed on RNA recovered from whole embryos and from dissected somites at HH stage 12. A minor amplified PCR product identified includes the 65 bp insertion found in the Gli3 cDNA sequence, whereas the major amplified PCR product lacks this insertion (data not shown), supporting the view that this insertion is an alternatively spliced exon. The minor Gli3 transcripts with this insertion sequence are enriched in somite transcripts. The translation product encoded by this cDNA has three in-frame stop codons that would lead to the production of an N-terminal Gli3 fragment (Fig. 1). The 3′ region of this cDNA also has an in-frame methionine and Kozak sequence that could be utilized to produce the Gli3 C-terminal protein (Kozak, 1989) that would lack most of the N-terminal repressor domain (Dai et al., 1999). It remains to be investigated whether these variant Gli3 proteins are produced in somites.

Whole-mount in situ hybridization methods were used to compare the expression of Gli3 with Gli1 and Gli2 during somite formation in stage-12 quail embryos (Fig. 2). We observed that Gli3, like Gli1 and Gli2 (Borycki et al., 1998), is undetectable in the segmental plate mesoderm, but is activated to high levels in paraxial mesoderm initiating somite formation (Fig. 2A,D,G). Gli3, like Gli2, is expressed throughout the newly formed somite, with higher expression in the domain surrounding the somitocoele and lower expression in the lateral somite (Fig. 2E,H). In contrast, Gli1 expression is restricted to the ventromedial somite domain encompassing the domain of Pax1 expression, but including the more dorso-medial domain of MyoD activation (Fig. 2B,C). The spatial patterning of Gli1 expression remains unchanged during somite maturation, but Gli2 and Gli3 expression becomes dorsalized to the dorso-medial domain of MyoD expression and myotome formation (Fig. 2F,I). Significantly, Gli2 and Gli3 expression are patterned identically in posterior somites; however, as somites mature along the antero-posterior axis, Gli2 expression becomes predominant in the medial myotome, whereas Gli3 expression becomes predominant in the lateral myotome (Fig. 2F,I). In addition, Gli3, but not Gli2, is strongly expressed in the posterior mesoderm and tail bud (Fig. 5B,E) and thus could have a specific function in the early processes of mesoderm development. The tail bud expression of Gli3 in stage-13 embryos overlaps with the domain of expression of Bmp4 and Hoxd10, two target genes of Shh signaling that are implicated in gut formation, but is not expressed in the caudal intestinal portal domain where Shh is expressed (Roberts et al., 1995). In the neural tube, Gli3 expression extends more posteriorly than does Gli2 or Gli1 (Figs 2A,D,G, 5B,E), but expression of all three Gli genes in the neural tube extends more posteriorly in stage-12 embryos than their expression in paraxial mesoderm, which is delimited by the boundary of somite formation. The posterior expression of these Gli genes as well as Ptc1 in the neural tube provides evidence that the notochord is producing functional levels of Shh in the posterior embryo where somites have not yet formed (Borycki et al., 1998). Gli1 expression is predominant in the ventral neural tube, Gli2 expression is predominant in the central neural tube, and Gli3 expression is restricted to the dorsal-most neural tube (Fig. 2B,E,H), as reported in the mouse embryo (Hui et al., 1994; Lee et al., 1997; Sasaki et al., 1997).

Surgical and Shh bead implantation experiments were undertaken to investigate the role of the notochord and Shh in the control of Gli genes during somite formation. Previously, we reported that Shh produced by the notochord is required for the activation of Gli1 expression during somite formation (Borycki et al., 1998). To examine whether notochord Shh signals control Gli2 and Gli3 activation, we surgically separated the notochord and floor plate from the paraxial unsegmented mesoderm on one side of stage-12 quail embryos, leaving the contralateral side as a control. Embryos were then cultured for 16 hours, during which time somites formed around the site of surgery, and Gli gene expression was assayed by in situ hybridization (Fig. 3). In contrast to Gli1 activation, which is blocked by notochord separation (Fig. 3A,B), we found that Gli2 and Gli3 activation does not require notochord signals (Fig. 3E,F,I,J). However, we observed that the notochord is required for the dorsalization of Gli2 and Gli3 expression, which normally occurs after somite formation (Fig. 3F,J). To determine whether Shh signaling is responsible for the dorsalization of Gli2 and Gli3, we surgically separated the notochord from the segmental plate mesoderm on one side of the embryo and implanted a bead soaked in a solution of purified N-terminal fragment of Sonic hedgehog (N-Shh), which has been shown to convey Shh signaling activity (Riddle et al., 1993; Marti et al., 1995; Roelink et al., 1995). In situ hybridization analysis of embryos 16 hours after culture revealed that N-Shh restores Gli1 activation, as we have previously reported (Borycki et al., 1998) (Fig. 3C,D), and also restricts Gli2 and Gli3 expression to the dorsal, myotome-forming region of the somite (Fig. 3G,H,K,L). We conclude that Shh signaling is not required for Gli2 and Gli3 activation during somite formation, but is required for the dorsalization of Gli2 and Gli3 expression during somite maturation.

Additional surgical and explant studies were undertaken to investigate the roles of surface ectoderm and neural tube in Gli2 and Gli3 regulation during somite formation. Explants of segmental plate mesoderm from stage-12 quail embryos were cultured for 16 hours in presence or in the absence of overlying ectoderm (Fig. 4A), and Gli2 and Gli3 expression was assayed by in situ hybridization. As previously reported (Borycki et al., 1998), surface ectoderm induces both somite formation and Gli2 expression in segmental plate explants (Fig. 4B,C). In contrast, we found that Gli3 is activated to high levels in paraxial mesoderm explants after removal of surface ectoderm (Fig. 4D,E). To determine whether surface ectoderm signals actively repress Gli3 expression in the segmental plate mesoderm, we preformed in vivo surgical experiments. Surface ectoderm overlying the posterior segmental plate mesoderm and axial tissues were separated on one side of stage-12 embryos (Fig. 5A), as we found that the neural tube provides redundant signals for Gli2 and Gli3 regulation (Borycki et al., 1998). Embryos were cultured for 7 hours in order to allow the surgery site to remain at the level of the segmental plate mesoderm. When Gli2 and Gli3 expression was assayed by in situ hybridization, we observed that Gli3 is prematurely activated to high levels in the segmental plate mesoderm on the operated side, whereas Gli3 is not detected in the segmental plate mesoderm on the unoperated control side (Fig. 5B,C). Transverse sections reveal that, on the operated side, Gli3 is expressed throughout the segmental plate mesoderm (Fig. 5D). In contrast, Gli2 is not expressed in segmental plate mesoderm in absence of surface ectoderm and axial tissues (Fig. 5E-G), demonstrating that distinct mechanisms control Gli2 and Gli3 expression: surface ectoderm and neural tube signals positively control the activation of Gli2 expression and the initiation of somite formation, whereas Gli3 expression is a cell-autonomous property of the segmental plate mesoderm in the absence of surface ectoderm. Gli3 activation at the time of somite formation, therefore, involves a derepression mechanism that must be mediated by a change in surface ectoderm signaling or in the response of the newly forming somite to these signals.

To identify the surface ectoderm signaling molecules responsible for the regulation of Gli2 and Gli3 expression during somite formation, we undertook a candidate molecule approach. Wnts are known surface ectoderm and neural tube signaling molecules (Parr et al., 1993; Parr and McMahon, 1995; Cadigan and Nusse, 1997; Willert and Nusse, 1998; Eastman and Grosschedl, 1999). To test whether Wnt proteins control Gli gene expression in the segmental plate mesoderm, we cocultured quail segmental plate mesoderm explants without surface ectoderm in collagen gels, either with RatB1a fibroblast cells stably infected with a MV7 control vector or with this vector expressing mouse Wnt1 or Wnt4. Explants without surface ectoderm cocultured with MV7-expressing RatB1a fibroblasts activate Gli3 expression (Fig. 6D), and do not express Gli2 (Fig. 6A). When explants are cocultured with Wnt1-expressing RatB1a fibroblasts, Gli2 expression is induced to high levels throughout the unsegmented paraxial mesoderm in greater than 80% of explants (Fig. 6B). Wnt4-expressing RatB1a fibroblasts also induce Gli2 expression in segmental plate explants in absence of surface ectoderm, but to a lower level than Wnt1-expressing cells (Fig. 6C). Wnt1-expressing RatB1a cells maintain the repression of Gli3 in segmental plate explants (Fig. 6E), which would express high levels of Gli3 in the absence of surface ectoderm, whereas Wnt4-expressing cells are less potent repressor of Gli3 (Fig. 6F). These data, therefore, establish that Wnt1 and Wnt4 can provide both positive and negative signaling activities for Gli2 activation and Gli3 repression.

Wnt1 and Wnt4 transduce their signals through inhibition of Glycogen Synthase Kinase-3 beta (GSK-3β), which results in the stabilization of β-catenin in responding cells, leading to formation of nuclear β-catenin/LEF1/tcf transcription factor complexes (Willert and Nusse, 1998; Eastman and Grosschedl, 1999). β-catenin-dependent Wnt signaling can be mimicked by LiCl treatment, which also acts to increase β-catenin by inhibiting GSK-3β β (Klein and Melton, 1996; Hedgepeth et al., 1997). To test whether activation of Gli2 and Gli3 is controlled by a β-catenin-dependent signaling mechanism, as predicted by the activities of Wnt1 and Wnt4 on paraxial mesoderm explants, we cultured segmental plate mesoderm explants from stage-12 quail embryos, without surface ectoderm, in the presence or absence of 10 mM LiCl. Consistent with the explant assays described above, we found that, in absence of LiCl, explants activate Gli3 at high levels (Fig. 7E) but do not form somites, retain a characteristic rough morphology, and do not express Gli2 (Fig. 7A). However, treatment of explants with 10 mM LiCl induces somite structures, as attested by the epithelial shape and the condensation of explant cells, and by the segmented expression pattern of Gli2 and Gli3 in these explants (compare Fig. 7B′,F′ with 7A′,E′). The somite structures formed in LiCl-treated explants are not as well organized as observed in explants associated with surface ectoderm (compare Fig. 7B′,F′ with 7C′,G′), suggesting that additional surface ectoderm signals are required to complete the process of somite formation. LiCl also activates Gli2 and Gli3 expression in a segmental pattern (Fig. 7B,F), as well as maintains the repression of Gli3 in unsegmented regions of explants (Fig. 7F). Given the activity of LiCl as an inhibitor of GSK-3β, these data indicate that Gli2 and Gli3 activation during somite formation is controlled by β catenin-dependent Wnt signaling. In contrast to Wnt1 and Wnt4 activities, LiCl treatment activates Gli2 locally in somite segments, but not in the unsegmented regions, indicating that LiCl treatment can coordinate Gli2 and Gli3 activation with somite formation. Therefore, during the process of somite formation, Wnt/β-catenin signaling undergoes a negative to positive switch from repression of Gli3 in segmental plate mesoderm to activation of somite formation and Gli2 and Gli3 in newly formed somites.

To examine whether Gli2 and Gli3 expression is a downstream response to the induction of somite formation by LiCl, we compared the activity of LiCl on Gli2 and Gli3 expression in unsegmented paraxial mesoderm explants that were maintained in association with their overlying surface ectoderm, which induces robust somite formation. Explants associated with surface ectoderm were found to express much higher levels of Gli2 (Fig. 7D) and Gli3 (Fig. 7H) in response to LiCl than did untreated explants (Fig. 7C,G), even though somite formation was equivalent in treated and untreated explants (Fig. 7D,H versus Fig. 7C,G). The higher levels of Gli2 and Gli3 expression in LiCl-treated explants provides evidence that both Gli2 and Gli3 can respond to LiCl, independent of the initiation of somite formation. The enhanced expression of Gli3 in somites formed in LiCl-treated explants in contact with surface ectoderm indicates that Gli3 expression is also subject to positive regulation through β-catenin, following its derepression at the time of somite formation. β-catenin, therefore, has a central role coordinating the activation of both Gli2 and Gli3 with the initiation of somite formation.

In this study, we have investigated the regulation of Gli genes during somite formation in quail embryos. We report that Gli1, Gli2 and Gli3 are co-activated in temporal coordination with the process of somite formation. The activation of these Gli genes coincides with the initiation of Shh responsiveness in newly formed somites, as indicated by the activation of a set of Shh-dependent genes that includes Ptc1, which is a known Shh-responsive gene (Chen and Struhl, 1996; Goodrich et al., 1996; Marigo et al., 1996a), as well as Myf5, MyoD and Pax1, which control somite cell specification (Borycki et al., 1998, 1999a). Considering that molecular and genetic evidence indicate that Gli transcription factors are effectors of Shh signal transduction in vertebrates (Hynes et al., 1997; Lee et al., 1997; Platt et al., 1997; Sasaki et al., 1997, 1999; Ding et al., 1998; Matise et al., 1998), these findings support a model for the initiation of Shh responsiveness in newly formed somites through the coordinated activation of this set of Gli genes. Once these Gli genes are expressed, they can then respond to Shh, which is constitutively expressed by the underlying notochord along the axis of the embryo. Somites, therefore, provide the first example of an embryonic tissue in which Shh transduction is controlled at the level of Gli gene regulation. It remains to be determined whether the Shh-responsive genes such as MyoD and Pax1, which become activated following somite formation, are direct targets of Gli transcription factors or whether these genes are regulated by downstream effectors of Shh signaling. Investigations of the expression, processing and functional activities of Gli proteins and splicing variants, such as described in this study, will be required to define the gene targets of Gli transcription factors in newly formed somites.

Embryo surgery and segmental plate mesoderm explant studies reveal that distinct signaling mechanisms regulate each of the three Gli genes in somites. Gli2 and Gli3 are subject to positive and negative regulation, respectively, by Wnt signaling from the surface ectoderm and neural tube. LiCl studies provide supporting evidence that Wnt regulation of Gli2 and Gli3 is mediated by β-catenin, which coordinates the activation of these genes with the initiation of somite formation. Prior to this study, there has been no information on the mechanisms that regulate Gli2 and Gli3 transcription and on the role of Gli regulation in the control of Shh signaling. Therefore, future investigations of the functions of Wnts and β-catenin in the control of Gli2 and Gli3 in other Shh-responsive tissues, including the nervous system and limb (Hui et al., 1994; Marigo et al., 1996b; Lee et al., 1997; Sasaki et al., 1997), will be of interest to determine if these mechanisms for Gli regulation are widely utilized. In contrast to Gli2 and Gli3, which are controlled by surface ectoderm/neural tube Wnt signaling, Gli1 activation is controlled by Shh signaling from the notochord. The cooperative interactions of surface ectoderm/neural tube and notochord signaling in the activation of the three Gli genes provides a molecular explanation for earlier findings that these tissues and Shh and Wnt signaling function synergistically to regulate the expression of somite-specific genes (Buffinger and Stockdale, 1994; Fan and Tessier-Lavigne, 1994; Munsterberg et al., 1995; Munsterberg and Lassar, 1995; Stern et al., 1995; Borycki et al., 1997, 1998; Fan et al., 1997; Tajbakhsh et al., 1998).

The results of our in vivo and in vitro studies reveal that Gli3 is repressed by surface ectoderm/neural tube signals in segmental plate mesoderm, whereas somite formation and Gli2 and Gli3 activation are under the positive control of surface ectoderm/neural tube signals. Therefore, the surface ectoderm/neural tube signaling that regulates Gli2 and Gli3 activation undergoes a switch at the time of somite formation. Our Wnt coculture experiments provide evidence that Wnts are the surface ectoderm/neural tube signaling molecules that mediate Gli3 repression in segmental plate mesoderm as well as the activation of Gli2 and Gli3 to initiate Shh signaling at the time of somite formation. In support of the essential role played by Wnt signaling in somites, injection of Frzb-1 expressing cells into the placenta of a pregnant female mouse blocks the Shh-dependent activation of Myf5 activation in the somites of embryos (Borello et al., 1999). As both Wnt1 and Wnt4 transduce their signaling activity through β-catenin, by inhibition of GSK-3β (Willert and Nusse, 1998; Eastman and Grosschedl, 1999), these findings indicate that both the repressing and the activating functions of Wnt signaling in Gli gene regulation are mediated through β-catenin.

A molecular model for a switch in Wnt signaling at the time of somite formation is shown in Fig. 8. This model proposes that this switch is controlled by a regulated change in the association of β-catenin/LEF1/tcf transcription complexes with cofactors such as Groucho, NLK and CtBP (Fisher and Caudy, 1998; Brannon et al., 1999; Ishitani et al., 1999; Meneghini et al., 1999). These cofactors modulate whether β-catenin/LEF1/tcf transcription complexes have repressing or activating gene regulatory functions. Presumably, the segmentation genes, which control the initiation of somite formation (Palmeirim et al., 1998), could mediate this switch in β-catenin function.

Gli gene regulation by Wnt signaling from surface ectoderm and neural tube is coordinated with the initiation of somite formation. We observed that, although Wnt1 and Wnt4 can repress Gli3 and activate Gli2 in explant coculture assays, LiCl can induce the entire repertoire of signaling activities of surface ectoderm/neural tube on segmental mesoderm, including somite formation and the regulated activation of Gli2 and Gli3. A previous report has shown that β-catenin levels increase in segmental plate cells prior to somite formation in chick embryos, and that LiCl treatment induces ectopic somite formation and β-catenin activation in segmental mesoderm (Linask et al., 1998). A number of explanations can account for the observed differences in the activities of LiCl and Wnts in our assays. One possibility is that the Wnts produced in vivo by the surface ectoderm and neural tube have more potent signaling activity or are produced at higher levels than the Wnt1 and Wnt4 produced in our coculture assays, leading to higher accumulation of β-catenin to levels required for somite formation and Gli regulation. Consistent with this possibility, LiCl is a chemical inhibitor of GSK-3β that bypasses all upstream processes required for Wnt signaling, and in this study, its activity has been optimized to mediate a maximal response for somite formation and Gli activation. Alternatively, surface ectoderm and neural tube may provide signals, in addition to Wnts, that are required for somite formation and Gli regulation, and LiCl may be able to activate the entire array of responses directed by these signals. Finally, we do not exclude the possibility that matrix factors present in collagen gels used to assay Wnt activity in our Wnt coculture assays may selectively prevent somite epithelialization. Collagen is not used in LiCl treatment experiments, in which segmental mesoderm explants are maintained on agarose substrates. Consistent with this possibility, collagen has previously been shown to interfere with Pax1 and Sim1 expression in somite explant cultures (Fan and Tessier-Lavigne, 1994). In any case, our findings provide clear evidence that Wnts acting through β-catenin are critical signals that regulate both somite formation and Gli2 and Gli3 activation.

The accumulation of β-catenin in newly forming somite cells could be regulated in vivo by mechanisms that regulate the levels of Wnt synthesis or the activities of Wnts. At this time, we do not know which Wnts are produced by surface ectoderm and neural tube to regulate somite formation and Gli activation. Wnt1 and Wnt4, which have activities in our induction coculture assays, are reasonable candidates, based on their expression in surface ectoderm (Fan et al., 1997) and neural tube (Parr and McMahon, 1994). It is notable that several different Wnt genes, including Wnt1 and Wnt4, are activated in the neural tube at the level along the embryo axis where somites are forming (Hirsinger et al., 1997; Marcelle et al., 1997), providing a source of new Wnts and increased levels of Wnt signaling to neighboring segmental plate cells that are initiating somite formation. In addition, Wnt signaling to newly forming somites could also be increased by a regulated reduction in the expression of inhibitors of Wnt signaling such as Dickkopf-1 (dkk-1), WIF-1, Cerberus and members of the Frizzled related protein (FRP) family (Finch et al., 1997; Leyns et al., 1997; Wang et al., 1997; Glinka et al., 1998; Hsieh et al., 1999; Piccolo et al., 1999). These inhibitors bind to Wnts and block their interactions with receptors. In support of this possibility, such Wnt inhibitors have been shown to be expressed at high levels in segmental plate mesoderm and their expression markedly decreases at the boundary of somite formation (Hoang et al., 1998). The enhanced Wnt signaling to newly forming somites, as a result of decreased expression of Wnt inhibitors and increased Wnt expression in the adjacent neural tube, could contribute to the observed accumulation of β-catenin in newly forming somites (Linask et al., 1998). These increased levels of β-catenin, in turn, would promote the cell adhesion processes of somite formation through its interaction with N-cadherins (Bellairs, 1985; Duband et al., 1987; Hatta et al., 1987), as well as promote the transcriptional activation of Gli2 and Gli3 through the formation of β-catenin/LEF1/tcf transcription complexes. These findings, therefore, provide evidence that Wnt signal transduction, like Shh signal transduction, is highly regulated during somite formation, and this regulation plays a central role in coordinating the process of somite formation with the regulated activation of Gli2 and Gli3 to initiate Shh signaling.

In contrast to Gli2 and Gli3, which we show are regulated by Wnt/β-catenin signaling, Gli1 activation in somites is shown to be controlled by Shh signaling. Shh also controls Gli1 expression in other tissues that are responsive to Shh signaling, including the limb and neural tube (Marigo et al., 1996b; Lee et al., 1997; Sasaki et al., 1997). As Gli1 is activated in newly formed somites when Gli2 and Gli3 are activated, Gli1 may be regulated by Gli2 or Gli3, through Shh signaling. In support of this model, the mouse Gli1 promoter has Gli binding sites and is transactivated by Gli3 (Dai et al., 1999) and Gli2 mutant mice do not activate Gli1 expression in the floor plate of the neural tube (Ding et al., 1998; Matise et al., 1998). An alternative possibility is that other Shh transcription effectors control Gli1 activation in somites, for instance COUP-TFII (Krishnan et al., 1997). In either case, the regulatory elements that regulate Gli1 activation are subject to Shh-dependent regulation in coordination with somite formation, which places Gli1 expression in somites downstream in the Shh signaling pathway initiated at somite formation (Fig. 8). As Shh signaling also dorsalizes Gli2 and Gli3 expression following their activation in somites, the maintenance of Gli1 expression in the ventral somite is independent of Gli2 and Gli3, regardless of their potential roles in Gli1 activation. Gli1 could mediate the dorsalization of Gli2 and Gli3 expression directly, or through downstream Shh effectors. Notably, recent data report evidence for Shh-mediated activation of Sfrp2, a Wnt1- and Wnt4-specific antagonist, in the ventral somite (Lee et al., 2000), providing a mechanism by which Shh could mediate Gli2 and Gli3 dorsalization during somite maturation. The roles of Gli1 in sclerotome specification in the ventral somite, and Gli2 and Gli3 in myotome specification in the dorsal somite, remain to be discovered.

We thank Lori Mendham for her technical assistance, and the members of the Emerson laboratory for suggestions and support. We also thank Peter Klein for his suggestions and advice, Ivan Bergstein for providing Wnt4 cDNA, Andrew McMahon for a gift of purified recombinant N-Sonic hedgehog protein, and Clifford Tabin for sharing sequences before publication and for critical reading of this manuscript. This work was supported by grants from the NICHD and the Muscular Dystrophy Association to C. P. E., and from the NCI to A. M. C. B, who is a recipient of an Irma T. Hirschl career scientist award.