Gli2 and Gli3 have redundant and context-dependent function in skeletal muscle formation

Progenitor cells for skeletal muscles of the head and the body arise from the head and prechordal mesoderm and the paraxial mesoderm, respectively(Christ and Ordahl, 1995; Noden, 1983). Specifically,the paraxial mesoderm is segmented into somites that form by budding off the anterior end of the presomitic mesoderm(Pourquie, 2001). Within the somite, precursor cells for the epaxial musculature (deep back muscles)originate from the medial edge of the dermamyotome, called the dorsal medial lip of the dermamyotome (DML), delaminate and involute inward to enter the epaxial myotome (Ben-Yair et al.,2003; Denetclaw et al.,2001; Denetclaw and Ordahl,2000; Kalcheim et al.,1999). At the lateral edge of the dermamyotome, called the ventral lateral lip of the dermamyotome (VLL), cells involute inward to enter the lateral hypaxial myotome and form the ventral body wall and intercostal muscles (Cinnamon et al.,1999; Ordahl and Le Douarin,1992). Ultimately, both epaxial and hypaxial myotomes merge and form a continuous layer of differentiated muscle cells. At the axial level of the limbs, lateral dermamyotomal cells delaminate and migrate to the limb buds to form appendage skeletal muscles. Finally, at the level of cervical/occipital somites, cells migrate anteriorly and form the hypoglossal chord, which will contribute to tongue and pharyngeal muscles(Mackenzie et al., 1998; Noden, 1983).

Genetic studies have established that the myogenic regulatory factor (MRF)Myf5 plays a key role in the specification of mesodermal cells to the different myogenic lineages, because Myf5-deficient mouse embryos display a significant delay in myogenesis until the onset of Myod1expression (Braun et al.,1994; Tajbakhsh et al.,1997) and in the absence of three of the four MRFs, Myf5,Myod1 and Mrf4 (Myf6 – Mouse Genome Informatics)no skeletal muscle progenitor cell is formed(Kassar-Duchossoy et al.,2004; Rudnicki et al.,1993). Furthermore, Myf5-deficient muscle progenitor cells at the DML fail to enter the epaxial myotome and migrate aberrantly in the dermatome and the sclerotome(Tajbakhsh et al., 1996). The temporal and spatial expression pattern of Myf5 is consistent with its primary role in myogenic specification. Myf5 is first found in the DML progenitor cells for epaxial muscles at E8.0 (embryonic day 8.0),followed by the VLL progenitor cells for hypaxial muscles at E9.75(Ott et al., 1991; Tajbakhsh and Buckingham,2000).

As expected for a protein acting upstream in the cascade of gene activation leading to skeletal muscle differentiation, Myf5 transcriptional regulation is particularly complex and has only been comprehensively described recently. Regulatory elements have been found over a region covering nearly 140 kb upstream and downstream of the Myf5 transcriptional start(Buchberger et al., 2003; Carvajal et al., 2001; Hadchouel et al., 2003; Hadchouel et al., 2000; Summerbell et al., 2000). Each element appears to function specifically in the control of Myf5expression at discrete sites in the embryo, indicating that Myf5regulation is controlled independently in individual skeletal muscle progenitor domains. Consistent with this observation, distinct signals produced by tissues surrounding the somite have been found to activate Myf5 expression in epaxial and hypaxial muscle progenitor cells(Borycki and Emerson, 2000). Notably, Wnt signals, secreted by surface ectoderm and neural tube cells,Sonic hedgehog (Shh), secreted by notochord and floor plate cells, and Bmp4,secreted by lateral plate cells, cooperate to induce and establish the somitic pattern of Myf5 expression(Borycki et al., 1999; Dietrich et al., 1998; Ikeya and Takada, 1998; Pourquie et al., 1996; Tajbakhsh et al., 1998). In previous work, we have shown that Shh is required for Myf5 activation in epaxial, but not in hypaxial muscle progenitor cells(Borycki et al., 1999; Kruger et al., 2001). Furthermore, Shh signalling appears to directly act on Myf5transcription in epaxial progenitor cells. Indeed, transgenic mice expressing a reporter gene under the control of the epaxial somite (ES) enhancer [also referred to as EEE, Early Epaxial Enhancer(Teboul et al., 2002)],located at –6.1 kb of the Myf5 transcriptional start,recapitulates endogenous epaxial Myf5 expression(Gustafsson et al., 2001; Teboul et al., 2002). In the context of a heterologous promoter, expression of this transgene is abolished in a Shh^–/– background or following mutation of an internal Gli-binding site(Gustafsson et al., 2001),suggesting that Gli proteins, which mediate Shh signalling(Ingham, 1998; Ruiz i Altaba, 1997), directly control Myf5 expression in epaxial muscle progenitor cells.

In the mouse, three Gli proteins, Gli1, Gli2, and Gli3 have been characterised (Hui et al.,1994). As is the case for Cubitus interruptus (Ci), the fly homologue of Gli, Gli2 and Gli3 can both activate and repress Shh target gene transcription in vitro (Dai et al.,1999; Sasaki et al.,1999). However, in vivo analyses of mouse mutants have shown that Gli2 acts primarily as a transcriptional activator(Ding et al., 1998; Matise et al., 1998), whereas Gli3 retains a bipotential activity, acting as transcriptional activator in motorneuron and ventral interneuron specification and as transcriptional repressor in dorsal interneuron specification(Bai et al., 2004; Litingtung and Chiang, 2000; Meyer and Roelink, 2003; Motoyama et al., 2003; Persson et al., 2002). Gli1 functions as a transcriptional activator of Shh target genes in vitro and in overexpression studies (Dai et al.,1999; Lee et al.,1997; Ruiz i Altaba,1999; Sasaki et al.,1999). However, mouse mutants show that Gli1 is dispensable(Park et al., 2000), although it can mediate Shh signalling in the absence of Gli2(Bai and Joyner, 2001; Park et al., 2000). Together,these observations indicate a complex interplay between Gli proteins in vivo.

Thus, a prediction from our previous studies of myogenesis in Shh^–/– mice and ES trangenic mice would be that Gli proteins are required during somite myogenesis to mediate Shh signalling. To address this question and to investigate whether Myf5activation and somite patterning require the activator or the repressor function of Gli proteins, we have examined the expression pattern and the regulation of Gli1, Gli2, and Gli3 during somite formation in the mouse embryo. We also carried out genetic studies using Gli and Shh mutant mice, as well as transgenic mice expressing lacZunder the control of the Myf5 epaxial enhancer (ES), to test the redundant function and the transcriptional activity of Gli proteins in Myf5 activation and somite patterning. We found that Gli2 or Gli3 is required for epaxial muscle progenitor cell specification and that Shh plays an essential role in this process to convert Gli3, but not Gli2 into a transcriptional activator. In addition, our study reveals an unexpected differential role for Gli2 and Gli3 along the antero-posterior axis in patterning the epaxial, hypaxial and myotomal compartments. Together, these data establish that Gli genes have essential specific and redundant functions during skeletal myogenesis.

Gli2 and Gli3XtJ mice were maintained as heterozygous stocks and crossed to generate different mutant combinations. Shhmice (kindly provided by M. Maconochie), also maintained as heterozygous, were crossed to Gli3XtJ heterozygous mice to generate Shh/Gli3double mutant embryos, and to Gli2 heterozygous mice to generate Shh/Gli2 double mutant embryos. Myf5 ES enhancer transgenic mice, which carry the reporter gene lacZ under the control of the tk viral promoter and the ES enhancer(Gustafsson et al., 2001),were crossed to Gli2^+/–Gli3^+/– mice to produce Gli2^+/–Gli3^+/–ESLacZmice. These mice were then bred to homozygocity. Embryonic day 0.5 was the day vaginal plugs were found. Embryos were harvested between E8.0 and E10.5 by Caesarian section, and genotyping was performed on yolk sac DNA by Polymerase Chain Reaction (PCR) with primers described previously(Chiang et al., 1996; Maynard et al., 2002; Mo et al., 1997).

Embryos were collected at various stages and fixed overnight at 4°C in 4% paraformaldehyde, washed in PTW (0.1% Tween 20 in PBS) and processed for in situ hybridisation according to the protocol described elsewhere(Henrique et al., 1995). To allow for quantitative comparisons, embryos were treated together in a single tube and staining allowed to develop for the same length of time. Digoxigenin(DIG)-labelled antisense riboprobes were generated from linearised plasmids containing inserts for Myf5 (Ott et al., 1991), Myod1(Sassoon et al., 1989), Pax1 (Deutsch et al.,1988), Pax3 (Goulding et al., 1991), Pax7(Jostes et al., 1991), Gli1 (Hui et al.,1994), Gli2 (Hui et al., 1994), Gli3 (Hui et al., 1994), myogenin(Sassoon et al., 1989),scleraxis (Cserjesi et al.,1995), paraxis (Burgess et al., 1995), Lbx1(Jagla et al., 1995), Sim1 (Fan et al.,1996) and Noggin(McMahon et al., 1998). Stained embryos were photographed under a LEICA MZ12 stereomicroscope using a Spot digital camera (Diagnostic Instruments). Embryos were then embedded in 2%agarose in PBS and 80 μm transverse sections were performed using a vibratome. Sections were mounted on slides using Glycergel (Dako), and photographed under Nomarski optics on a LEICA DM-R microscope using a LEICA digital camera.

Using whole-mount in situ hybridisation, we examined the expression pattern of Gli genes during somitogenesis and compared it to that of the myotome-,syndetome- and sclerotome-specific genes Myf5, scleraxis, and Pax1 (Brent et al.,2003; Deutsch et al.,1988; Ott et al.,1991). Expression of Gli genes in the paraxial mesoderm begins at E.8.0 with the formation of the first somite pair (data not shown). Gli1 is not detected in the pre-somitic mesoderm and its expression begins in somite I (Fig. 1A). In contrast, Gli2 and Gli3 are expressed in the anterior pre-somitic mesoderm (Fig. 1B,C). In newly-formed somites, Gli1 is strongly expressed in the ventral medial cells surrounding the notochord and more diffusely in the dorsal sclerotome (Fig. 1D). As somites mature, Gli1 expression spreads dorsally and laterally (Fig. 1G),overlapping with Pax1 expression(Fig. 1J), which is specifically detected within the sclerotome and is excluded from the dermamyotome (Deutsch et al.,1988). Interestingly, Gli1 expression is downregulated in cells surrounding the notochord in anterior somites of E10.5 embryos and in hindlimb somites of E11.5 embryos (data not shown), at the time condensation of mesenchymal cells occurs around the notochord(Hall and Miyake, 2000). At E9.5, Gli2 is expressed throughout the newly formed somite in a domain encompassing both the sclerotome and the DML(Fig. 1B,E). In anterior somites, Gli2 expression decreases in the sclerotome(Fig. 1H), although it is upregulated at later stages (E10.5 and E11.5) at the time Gli1expression is downregulated (data not shown). In anterior somites, Gli2 (Fig. 1H) is expressed in a domain overlapping that of scleraxis and Myf5, which label the syndetome (Brent and Tabin,2004) and the myotome (Ott et al., 1991), respectively (Fig. 1K,L). At E10.5, Gli2 remains expressed in the DML of posterior, but not anterior somites (data not shown). At E9.5, Gli3is expressed abundantly in the lateral domain of newly formed somites and is weakly expressed in the sclerotome (Fig. 1C,F). Gli3 is also expressed in the DML(Fig. 1F,I), although it is detected slightly later than Gli2. As somites mature, Gli3becomes restricted to the ventral dermamyotome and the myotome(Fig. 1I), in a domain that overlaps that of Myf5 expression(Fig. 1L). In E10.5 and E11.5 embryos, Gli3 expression remains high in the DML and in the VLL, but is downregulated in the central myotome (data not shown). These observations are consistent with Gli2 and Gli3 playing a role in Myf5 activation in the DML and controlling epaxial and hypaxial myotome formation.

We previously showed that in the avian somite Gli1 expression is dependent on Shh signalling, and others showed that Gli1 expression is downregulated in the neural tube of Gli2^–/–and Gli2^–/–Gli3^–/– mice(Bai et al., 2004; Borycki et al., 1998; Ding et al., 1998; Matise et al., 1998). To investigate whether somitic Gli1 activation is controlled by Gli2 and Gli3, we analysed Gli1 expression by in situ hybridisation of various Gli mutant embryos. We found that Gli1 expression is nearly unchanged in Gli3^–/– or Gli2^+/–Gli3^–/– embryos(Fig. 2B,H,N,D,J,P). In contrast, Gli1 is downregulated in anterior somites of Gli2^–/– embryos(Fig. 2C,O), and loss of one Gli3 allele in the Gli2 mutant background nearly abolishes Gli1 expression in somites, although CNS expression persists(Fig. 2E,K,Q). Noticeably, no Gli1 transcript is detected in somites of Gli2^–/–Gli3^–/– embryos(Fig. 2F,L,R), indicating that activation of Gli1 requires Gli2 or Gli3. These data also reveal that Gli2^–/–Gli3^–/– embryos are molecularly similar to Gli1^–/–Gli2^–/–Gli3^–/–embryos. Therefore, no Gli-mediated target gene activation can occur in Gli2^–/–Gli3^–/–somites.

To address the function of Gli proteins in Myf5 activation, we examined the expression of Myf5 in wild-type, Gli2, Gli3 and Gli2/Gli3 mutant embryos by whole-mount in situ hybridisation. Myf5 activation in the epaxial myogenic progenitor cells of the DML appears normal in E9.5 Gli2^–/– and Gli3^–/– embryos (data not shown), suggesting that Gli2 and Gli3 may be functionally redundant. We therefore examined compound Gli2/Gli3 mutant embryos at E8.5, a stage at which Myf5 expression is solely activated in the epaxial muscle progenitor cells of the DML, under the control of the epaxial somite enhancer (ES)(Gustafsson et al., 2001; Teboul et al., 2002). In wild-type embryos, Myf5 transcripts are detected in seven out of 10-11 somites with strong expression in the six occipital somites(Fig. 3A,E,I). In contrast, Myf5 activation is delayed in Gli2^–/–Gli3^–/– embryos and only weak expression is observed in the occipital somites(Fig. 3D,H,L). Transverse sections show that Gli2^–/–Gli3^–/– somites have an abnormal morphology and remain epithelial(Fig. 3H,L), and Myf5transcripts are observed along a ventral epithelial extension(Fig. 3L). Noticeably, somites retain their normal morphology in Gli2^–/–Gli3^+/– and Gli2^+/–Gli3^–/– embryos(Fig. 3G,K), and Myf5expression, although weaker, is activated on schedule in Gli2^–/–Gli3^+/– and Gli2^+/–Gli3^–/– embryos(Fig. 3B,C,F,G). Together,these data demonstrate that Gli2 or Gli3 is required for Myf5activation in the epaxial muscle progenitor cells, and that one copy of either Gli2 or Gli3 is sufficient to activate Myf5.

To address whether the delay in epaxial Myf5 activation observed at E8.5 persists at later stages, we examined E9.5 wild-type and mutant embryos. No epaxial Myf5 transcript is detected in the three most newly formed somites in Gli2^–/–Gli3^–/– mice(Fig. 3P), whereas Myf5 is detected in DML cells of all wild-type somites(Fig. 3M). In Gli2^–/–Gli3^+/– and Gli2^+/–Gli3^–/– posterior somites, epaxial Myf5 expression is reduced(Fig. 3N,O). Consistent with our observation on Myf5 expression, activation of myogenin expression is delayed in Gli2^+/–Gli3^–/–and Gli2^–/–Gli3^–/–embryos (Fig. 5A″,A″′). It was not possible to assess later stages as double mutant embryos die between E9.5 and E10.5. To further demonstrate the requirement for Gli2 or Gli3 in Myf5 activation and the determination of epaxial muscle progenitor cells, we crossed the ES/LacZ transgenic mice, which express the β-galactosidase reporter gene under the control of the Myf5 epaxial somite enhancer (ES)(Gustafsson et al., 2001),into Gli2/Gli3 mutant mice. Wild-type and Gli2^–/–Gli3^–/– mutant embryos carrying the transgene were analysed at E9.5 for their reporter gene expression (Fig. 4). Similar to endogenous Myf5 expression, no β-gal staining is observed in the posterior somites of Gli2^–/–Gli3^–/– embryos(Fig. 4B,D), whereas wild-type embryos exhibit β-gal staining in the DML of all somites(Fig. 4A,C,E,G). Consistent with our in situ hybridisation results, this observation demonstrates that Gli2 and Gli3 play essential role in the specification of epaxial muscle progenitor cells via the direct control of the Myf5 ES enhancer activity in newly formed somites. However, we observe weak β-gal staining in anterior somites, which is mainly ectopically located in the dorsal dermamyotome and in the myotome (Fig. 4F,H), indicating that in anterior somites additional factors may participate in Myf5 activation and cooperate with Gli proteins to pattern correctly this expression along the dorsoventral and mediolateral axes.

To further characterise the defects observed in Gli2/Gli3 mutant somites, we performed in situ hybridisation using dermamyotomal, myotomal and hypaxial markers.

In interlimb and anterior somites of E9.5 Gli2^–/–Gli3^–/–(Fig. 3T,X) and Gli2^–/–Gli3^+/–(Fig. 3R,V), but not Gli2^+/–Gli3^–/–(Fig. 3S,W) embryos, Myf5 expression is upregulated and shifted ventrally. Because Myf5 becomes activated in the myotome and in the hypaxial muscle progenitor cells at E9.5, it is most likely that this observation indicates a function for Gli2 and Gli3 in the formation and/or patterning of the myotome and the hypaxial somitic domain. Consistent with this possibility, Myf5 transcripts take a central position in Gli2^–/–Gli3^–/– somites compared to wild-type somites (compare Fig. 3P to 3M), indicating that anterior and posterior somitic compartments are more severely affected. Furthermore, myogenin, whose transcripts mark myotomal cells as they delaminate from the DML and enter the epaxial myotome (Fig. 5A)(Sassoon et al., 1989), is mispatterned and found in a more ventral domain in Gli2^–/–Gli3^+/– and Gli2^–/–Gli3^–/– somites(Fig. 5A′,A″′). Notably, myogenin transcripts are found in an overlapping domain with Myf5, along the ventral epithelial extension that characterizes the double mutant somite (compare Fig. 5A″′ and Fig. 3L). Finally, Myod1 expression is also precociously activated and misexpressed in a ventral domain of the epaxial myotome in Gli2^–/–Gli3^–/– embryos(Fig. 5E″′). These data are consistent with the idea that Gli2 and Gli3 are essential for the activation of Myf5 in the epaxial muscle progenitor cells and the subsequent formation of the epaxial myotome, and are required for the normal dorsoventral patterning of the myotome.

To clarify the identity of the epithelial cells extending from the dorsal medial lip along the neural tube, we used the dermamyotomal markers paraxis, Pax3 and Pax7 (Burgess et al., 1995; Jostes et al.,1991). Paraxis, Pax3 and Pax7 are all expressed in the dermamyotome of wild-type embryos(Fig. 5B-D) and their expression is unchanged in Gli2^–/–Gli3^+/– and Gli2^+/–Gli3^–/– embryos(Fig. 5B′,B″,C′,C″,D′,D″). In contrast, cells in the ventral epithelial extension of Gli2^–/–Gli3^–/– somites express paraxis Pax3 and Pax7, although at lower levels than dorsal dermamyotomal cells (Fig. 5B″′,C″′,D″′). To investigate whether these cells express DML markers, we performed an in situ hybridisation using Noggin, which specifically labels DML cells from E9.5 onward(Fig. 5G)(McMahon et al., 1998). Noggin transcripts remain localised to the dorso-medial domain of Gli2^–/–Gli3^–/– somites and are not found in the epithelial cells that extend ventrally(Fig. 5G″′),indicating that these cells do not have the molecular identity of DML cells,although they do express dermamyotomal markers. Noggin expression has been proposed to provide a permissive environment for Myf5 expression in DML cells (Hirsinger et al.,1997; Marcelle et al.,1997). However, we find here that Myf5 expression in Gli2^–/–Gli3^–/– mice occurs in a non-Noggin-expressing domain, indicating that antagonising Bmp4 signals is possibly necessary but not sufficient for Myf5 activation in the dorsomedial somite.

We then examined the specification, differentiation, and patterning of hypaxial muscle progenitor cells in Gli mutant embryos using Pax3, Sim1,Myod1 and Lbx1 expression. Hypaxial muscle progenitor cells originate from the Pax3-positive, Sim1-positive lateral dermamyotomal cells (Fig. 5D,H). As they delaminate from the dermamyotome to enter the hypaxial myotome or migrate to the limbs, they activate Myod1 or Lbx1, respectively (Fig. 5E,F). Thus, Myod1 is detected at E10.0 in the hypaxial muscle progenitor cells of interlimb somites(Fig. 5E) and this expression is immediately preceded by that of Myf5(Tajbakhsh and Buckingham,2000). By E10.5, Myod1 is detected in all hypaxial progenitor cells, including those of the body wall, limb and tongue muscles(Sassoon et al., 1989). In contrast, Lbx1 is activated at E9.5 at the level of occipital and limb somites in a subset of hypaxial muscle progenitor cells(Fig. 5F)(Jagla et al., 1995; Uchiyama et al., 2000), which are committed to migrate to the limb bud or to the head(Brohmann et al., 2000; Dietrich et al., 1998; Gross et al., 2000). As expected from our previous analysis of Shh mutant mice(Borycki et al., 1999), none of the mutant combinations studied shows a defect in the onset of hypaxial muscle progenitor cell specification (Fig. 5D-D″′,E-E″′,F-F″′,H-H″′). Likewise, no change in Myod1 and Lbx1 expression pattern is observed in Gli2^–/–,Gli3^–/– and Gli2^–/–Gli3^+/– embryos at E9.5 and E10.0 (Fig. 5E′,F′ and data not shown). However, Myod1 is upregulated in the hypaxial domain of Gli2^+/–Gli3^–/– and Gli2^–/–Gli3^–/– somites,and both Myod1 and Lbx1 are upregulated and expand into the ventromedial somitic domain of Gli2^–/–Gli3^–/– embryos(Fig. 5E″,E″′,F″,F″′). Noticeably,no such upregulation occurs for Pax3 and Sim1 dermamyotomal expression in Gli2^+/–Gli3^–/–and Gli2^–/–Gli3^–/–somites (Fig. 5D″,D″′,H″,H″′). This indicates that in the absence of Gli2 and Gli3, although hypaxial muscle progenitor cells form and are specified normally in the dermamyotome, they become unproperly patterned as they enter the hypaxial myotome or migrate into the limbs, as is the case for Shh mutant mouse embryos.

Gli2 and Gli3 proteins contain recognition sites for protein kinase A-dependent phosphorylation, a process that leads to the proteolytic cleavage of the full-length protein into a shorter N-terminal polypeptide containing a repressor domain in Drosophila(Chen et al., 1998; Epstein et al., 1996; Jia et al., 2002; Price and Kalderon, 2002). In vitro, both Gli2 and Gli3, but not Gli1 have been shown to act as transcriptional repressors (Dai et al.,1999; Sasaki et al.,1999). Shh signalling prevents this cleavage and converts Gli2 and Gli3 into transcriptional activators(Aza-Blanc et al., 2000). In vivo, only Gli3 was found to have repressor function and to require Shh signalling to be converted into a transcriptional activator(te Welscher et al., 2002; Wang et al., 2000). This implies that defects observed in dorsoventral patterning of the neural tube of Shh mutant mice can be partially rescued in Shh/Gli3compound mutant mice, as the result of the loss of a Gli3 repressor function(Litingtung and Chiang,2000).

To test whether some of the somitic defects observed in Shh mutant mice are due to a repressor activity of Gli2 or Gli3, we analysed by in situ hybridisation the expression of Myf5 and myogenin in Shh^–/–,Gli3^+/–Shh^–/– and Gli3^–/–Shh^–/– embryos,as well as in Gli2^+/–Shh^–/–and Gli2^–/–Shh^–/–embryos. As shown before (Borycki et al.,1999), Myf5 expression is not activated in the epaxial muscle progenitor cells of the DML in Shh^–/–embryos (Fig. 6F,J), although its activation comes on schedule in the hypaxial muscle progenitor cells(Fig. 6F,J,N,V,V′). In addition, hypaxial Myf5 expression expands into the ventromedial somite (compare Fig. 6M and 6N,V′), and Myf5 and myogenin expression is greatly reduced in occipital somites of Shh^–/– embryos(Fig. 6B,F,R). Removal of one Gli3 allele in the Shh^–/– background restores Myf5 expression in the DML epaxial muscle progenitor cells of interlimb somites, but not posterior somites(Fig. 6G,K). However, removal of both Gli3 alleles in the Shh^–/–background fully restores Myf5 expression in DML muscle progenitor cells along the entire anteroposterior axis(Fig. 6H,L,P,T). Unlike for Gli3, loss of Gli2 alleles in the Shh^–/– background does not restore epaxial Myf5 expression (Fig. 6U-X′). This indicates that Shh signalling is required to convert Gli3, but not Gli2 into a transcriptional activator for Myf5expression in DML cells. These data also suggest that DML cells of posterior and occipital somites are more sensitive to a ratio of Gli3 activator versus Gli3 repressor ≤ 1 than are DML cells of interlimb somites.

Interestingly, Gli3^–/–Shh^–/– interlimb somites show persisting ventromedial expansion of myotomal and hypaxial Myf5 expression (Fig. 6P), whereas Gli2^–/–Shh^–/– interlimb somites have a less pronounced expansion especially in cells surrounding the notochord (compare Fig. 6X′ to 6V′). These data suggest that Shh signalling functions in a Gli3-independent manner in the mediolateral patterning of interlimb somites,presumably acting via a repressor function of Gli2.

Through a detailed analysis of single and compound Gli mutant mice,together with the analysis of transgenic mice carrying a reporter gene under the control of the Myf5 epaxial enhancer (ES), we demonstrate that Gli2 or Gli3 is required to specify epaxial muscle progenitor cells and to maintain proper mediolateral and dorsoventral patterning of the dorsal somite(data are summarised in Table 1). Our data also reveal that the formation of distinct myogenic lineages within the somite requires a combination of activating and repressing Gli functions along the antero-posterior axis of the embryo.

Gli1, Gli2 and Gli3 expression is activated in coordination with somite formation in the mouse embryo, with Gli2 and Gli3 expression beginning just before somite formation. Thus, as in the avian embryo (Borycki et al.,1998), Gli-dependent Shh response is linked to somitogenesis in the mouse embryo. This suggests that similar mechanisms control Gli gene expression in the mouse and in the bird, where we showed that Shh signals control Gli1 activation, and Wnt signals control Gli2 and Gli3 activation(Borycki et al., 1998).

Gli genes are differentially expressed in the somite, with Gli1restricted to the ventral somite, and Gli2 and to a lesser extent Gli3, initially expressed throughout the somite and then predominantly found in the dorsal somite. This expression pattern is consistent with our findings that Gli2 and Gli3 are required for Gli1activation and that Gli3 can only compensate for the loss of Gli2 in newly formed somites. However, Gli3 acts as a weak transcriptional activator,because the presence of one Gli3 allele in a Gli2 mutant background is not sufficient to activate Gli1 expression in newly formed somites. The overlapping expression of Gli2 and Gli3in the dorsal somite also accounts for their functional redundancy during somite myogenesis, although their respective preponderance in the medial and lateral somite is consistent with independent functions in mediolateral patterning of the somite.

We provide several lines of evidence that Gli2 or Gli3 is required to mediate Shh signalling in the specification of epaxial muscle progenitor cells from the DML. First, as for Shh^–/– embryos, Myf5 expression is not activated in DML cells of Gli2^–/–Gli3^–/– embryos,whereas activation occurs, albeit at lower levels, in Gli2^–/–Gli3^+/– and Gli2^+/–Gli3^–/– embryos. Second, no β-Gal+ cell is observed in newly formed somites of Gli2^–/–Gli3^–/– embryos crossed into the Myf5 ES/LacZ transgenic mice, which express the reporter gene in the epaxial muscle progenitor cells(Gustafsson et al., 2001; Teboul et al., 2002). Third,Shh signalling is necessary to convert Gli3R into Gli3A for Myf5activation in DML cells. This establishes that Gli2 and Gli3 both act as transcriptional activators of Myf5 expression in DML cells and are required for the transcriptional activity of the Myf5 ES enhancer(Fig. 7). These data are consistent with our previous findings that Shh signalling is necessary to initiate epaxial Myf5 expression, and that the Gli binding site located in the Myf5 ES enhancer is essential for the enhancer activity in posterior somites (Gustafsson et al., 2001). In contrast, the present data do not support previous findings by Teboul et al., that the Gli binding site in the Myf5 ES enhancer is required for the maintenance of epaxial Myf5 expression but not for its initiation(Teboul et al., 2003). We believe this discrepancy reflects the different behaviour of the transgenic constructs used, which in the latter case, contains upstream of the Myf5 promoter both the branchial arch enhancer and the Myf5ES enhancer (Teboul et al.,2003). It is possible that cross-talk between enhancers, yet to be characterised, occurs and alters the activity of the ES enhancer.

At E8.5, weak Myf5 expression can still be detected in anterior somites of Gli2^–/–Gli3^–/– embryos,but in a more ventral position. This expression has also been observed in Shh^–/– and Smo^–/– embryos(Gustafsson et al., 2001; Kruger et al., 2001; Zhang et al., 2001). The origin of these Myf5 transcripts remains to be identified. We believe that Gli2 and Gli3 are required for the activation of Myf5 but in the absence of Shh signalling, Myf5 expression in the epaxial progenitor cells can be activated with delay in a Shh-independent manner(Fig. 7). Indeed, we find that Myf5 expression in E9.5 Gli2^–/–Gli3^–/– anterior somites maps to the ES enhancer, indicating that other factors interacting either with the Gli binding site or with another site can drive Myf5expression at later stages. Further characterisation of the Myf5 ES enhancer is required to elucidate the molecular mechanisms that control its activity. In particular, it will be of interest to investigate whether other Gli-related transcription factors can bind the consensus Gli site in the ES enhancer. Candidate proteins are members of the Zic family of transcription factors, which bind consensus Gli binding sites with a lower affinity than Gli proteins (Mizugishi et al.,2001), and can physically associate with Gli(Koyabu et al., 2001). Alternatively, the residual Myf5 expression observed at E8.5 in Gli2^–/–Gli3^–/– somites could result from the deregulated activity of a distinct enhancer and for instance, derive from the upregulation of the myotome enhancer(Buchberger et al., 2003; Hadchouel et al., 2003). In agreement with this hypothesis, we find that Myf5 and myogenin expression overlaps in Gli2^–/–Gli3^–/–occipital somites. The comparative analysis of E8.5 Shh^–/– or Smo^–/– and wild-type embryos carrying a transgene with the full Myf5 regulatory region, in which the ES enhancer has been deleted, will be critical to definitely rule out this possibility.

The ventral position of epaxial Myf5 transcripts in Gli2^–/–Gli3^–/–reveals that dermamyotomal and myotomal cells are mislocated in the ventromedial somite. This observation correlates with the reduction or loss of Pax1-expressing cells in the ventral somite of Gli2^–/–Gli3^–/–mice (Buttitta et al., 2003),indicating that normal dorsoventral patterning of the somite is disrupted in the absence of Gli proteins. This defect is reminiscent of the ventral shift of neuronal cells in the developing central nervous system of Gli2^–/–Gli3^–/–mice (Wijgerde et al., 2002). This suggests that, as was demonstrated in the neural tube(Jessell, 2000), somitic cells adopt a specific cell fate along the ventro-dorsal axis according to the level of Hedgehog signalling transduced (Fig. 7).

Biochemical and in vitro studies have shown that Gli2 and Gli3 can act as transcriptional activators and repressors(Ruiz i Altaba, 1999; Sasaki et al., 1999), but until recently it was thought that in vivo, Gli2 had only activator function and Gli3 had only repressor function (Bai et al., 2002; Ding et al.,1998; Litingtung and Chiang,2000; Matise et al.,1998; Wang et al.,2000). Our data show that Gli2 and Gli3 have both activator and repressor function in the dorsal somite depending on the genetic context. For instance, Gli2 represses hypaxial gene expression in the absence of Gli3, as illustrated by the gradual increase in hypaxial gene expression in Gli2^+/–Gli3^–/– and Gli2^–/–Gli3^–/– embryos(Fig. 7). Conversely, Gli3 activates epaxial gene expression in the absence of Gli2, as Myf5expression, although reduced, is still observed in posterior somites of Gli2^–/–Gli3^+/– embryos and is lost in Gli2^–/–Gli3^–/–embryos (Fig. 7). These results are in line with recent studies showing that Gli3 has activator function in the specification of floor plate and V3 interneurons in the neural tube(Bai et al., 2004), and in the specification of sclerotomal cells in the somite(Buttitta et al., 2003).

Our data clearly indicate that the failure to activate Myf5 in the epaxial somitic domain of Shh^–/– embryos is due to the repressor activity of Gli3, as we observe a progressive restoration of Myf5 expression in Gli3^+/–Shh^–/– and Gli3^–/–Shh^–/– embryos. Moreover, Gli3 repressor clearly acts in a concentration-dependent manner on Myf5 activation (Fig. 7), and accounts for the more severe phenotype observed in Shh^–/– mice compared to Gli2^–/–Gli3^–/– mice. Similar effects have been reported in the dorso-ventral patterning of the neural tube and in the anteroposterior patterning of the limb(Litingtung and Chiang, 2000; Litingtung et al., 2002; te Welscher et al., 2002),revealing that repression of target genes by Gli3 in the absence of Shh signals constitutes a fundamental mechanism in tissue patterning. This remarkable result also indicates that Myf5 activation can occur in the absence of both Shh and Gli signalling, as Gli1 and Gli2 proteins are thought to not be active in the absence of Shh. This observation is in line with previous studies showing that combinatorial Wnt and Shh signalling activate Myf5 expression(Munsterberg et al., 1995; Tajbakhsh et al., 1998), and suggests that in the absence of Gli activity, other transcription factors activate the Myf5 ES enhancer and Myf5 transcription. Crossing Gli3/Shh mutant mice into the ES/lacZ transgenic background would confirm that this is the case. Nevertheless, the fact that the Myf5 ES enhancer has a single Gli site suggests that the variable phenotype observed in our compound mutants can only be explained if activator and repressor forms of Gli, alone or in association with partners, bind the Gli binding site with different affinities.

In a previous report, we showed that in the absence of Shh, hypaxial Myf5 expression expanded into the ventromedial somite(Borycki et al., 1999). Here,we find that expression of Myf5 and myogenin in epaxial myotomal cells is upregulated and expands ventrally in Gli2^–/–Gli3^+/– embryos, and expression of Myf5, Myod1 and Lbx1 in hypaxial muscle cells is upregulated and expands medially in Gli2^+/–Gli3^–/– and Gli2^–/–Gli3^–/– embryos,indicating that in addition to their role in epaxial muscle cell determination, Gli2 and Gli3 function also in dorsoventral and mediolateral patterning of the somite. Moreover, this somite patterning defect persists in Gli3^–/–Shh^–/– interlimb somites, in agreement with the idea that Gli3 and possibly Gli2, act synergistically in the lateral somite to repress the hypaxial programme. The exact mechanism of Shh/Gli function in mediolateral patterning remains to be determined, and in particular, whether Gli repressors act directly or indirectly on the hypaxial/myotomal genes. Evidence for a direct role would require that similar studies to that presented here are performed using the hypaxial Myf5 enhancer. In the avian somite, evidence for an indirect role proposes that mediolateral patterning is established via counteracting activities between a lateral Bmp4 gradient originating from the lateral mesoderm and a medial Shh gradient originating from the axial mesoderm(Hirsinger et al., 1997; Marcelle et al., 1997). It is still not clear how Bmp4 and Shh counteract each other, but one possibility could be that Gli3R interacts with Smad proteins, which mediate Bmp4 signalling in the control of hypaxial gene expression(Fig. 7)(Liu et al., 1998). Finally,mispatterning could result from the deregulated cell division of specific muscle progenitor cell subpopulations (i.e. epaxial myotome and hypaxial cells). In this view, Shh signalling has previously been shown to control cell cycle progression via the regulation of cyclins(Bai et al., 2004; Barnes et al., 2001; Kenney and Rowitch, 2000), and is involved in the balance of proliferation/differentiation of limb muscle cells (Amthor et al., 1999; Duprez et al., 1998). Future experiments are required to investigate the molecular factors controlling the activity of the enhancers of hypaxial and myotomal genes and the possible cooperation of Smad and Gli proteins, as well as the relationship between Shh signalling and cell cycle regulators in skeletal muscle progenitor cells.

The authors would like to thank M. Buckingham, E. Olson, S. Dietrich, J. Epstein, A. McMahon, K. Jagla, and C. M. Fan for providing us with cDNAs. We also thank members of the Borycki lab and Dr J. Caamano for critical reading of the manuscript. We are grateful to J. Tiedeken, A. Milano and L. Williams for their excellent technical assistance in breeding and genotyping the mouse colonies. This work was supported by grants from the March of Dimes(1-FY00-665) and the NIH (HD7796) to C.P.E. and from The Royal Society, the EU Framework Program 6 and The Wellcome Trust to A.-G.B.