Wnt signaling and the activation of myogenesis in mammals

Giulio Cossu, Ugo Borello

Author Affiliations

  1. Giulio Cossu*,1 and
  2. Ugo Borello1
  1. 1 Istituto Pasteur‐Fondazione Cenci Bolognetti, Dipartimento di Istologia ed Embriologia Medica, Universitá di Roma ‘La Sapienza’, Via A.Scarpa 14, 00161, Roma, Italy
  1. *Corresponding author. E-mail: cossu{at}


In the amniote embryos, specification of skeletal myoblasts occurs in the paraxial mesoderm in response to a number of signaling molecules produced by neighboring tissues such as neural tube, notochord and dorsal ectoderm. Candidate molecules for this complex signaling activity include Sonic hedgehog, Wnts and Noggin as positive activators and BMP4 as a possible inhibitor. Recently, the receptors and the post‐receptor pathways for Sonic hedgehog and Wnts have been characterized, and this has opened up the possibility of linking these signaling events to the activation of myogenic regulatory factor genes such as Myf5 and MyoD and functionally related genes such as Pax3. Here we focus on the role of Wnts, their putative receptors Frizzled and the soluble antagonist Frzb1 in regulating mammalian myogenesis. Although it is becoming evident that the signaling downstream of Frizzled receptors is much more complex than anticipated, it is conceivable that it may lead to transcriptional activation of Myf5 and MyoD and to initiation of myogenesis. However, the fact that both Wnts and Sonic hedgehog have a strong effect on cell proliferation and survival suggests that they may contribute to the overall process of myogenesis by a combination of these different biological activities.

Neighboring tissues influence somite differentiation

All skeletal muscles of the vertebrate body, with the exception of most head muscles, are derived from somites, epithelial structures which form in the paraxial mesoderm in a cranio–caudal developmental sequence over an extended period of embryogenesis (Brand‐Saberi and Christ, 1999). Muscle progenitor cells become specified within the dermomyotome, which is located dorsally under the ectoderm. Cells located in the ventral sclerotome, adjacent to the notochord, will contribute to cartilage and bone. At the onset of somitogenesis, cell fate is not yet determined since rotation of the epithelial somite in a dorsal/ventral or medial/lateral direction does not perturb subsequent differentiation that takes place according to the final orientation of the somite. This strongly suggests that signals from the environment specify the identity of cells within the newly formed somite.

Work carried out in the 1980s (reviewed in Cossu et al., 1996a) established that in explants of paraxial mesoderm, axial structures (the neural tube–notochord complex) are required to promote myogenesis, but only in pre‐somitic mesoderm and newly formed somites: more mature somites do not need the presence of neighboring tissues. Further experiments showed that only precursors of epaxial (back) muscles, located in the dorso‐medial domain of newly formed somites, are dependent upon signals from axial structures; in contrast, precursors of hypaxial (limb and body wall) muscles, located in the lateral half of the paraxial mesoderm, do not need the neural tube–notochord complex but rather require a signal from dorsal ectoderm for myogenic commitment (reviewed in Cossu et al., 1996a). It was later shown that explants of murine paraxial mesoderm, when cultured in the presence of axial structures, activate Myf5. In contrast, they will activate MyoD when cultured with their own dorsal ectoderm (Cossu et al., 1996b). This suggests that in mammals, axial structures activate myogenesis through a Myf5‐dependent pathway, while dorsal ectoderm acts through a MyoD‐dependent pathway. It is important to stress here that this does not mean that medial progenitor cells express Myf5 and lateral cells MyoD. The situation is more complex: medial cells express Myf5 and are indeed absent in Myf5 null embryos (although only at an early stage; Braun et al., 1992). In lateral cells, Myf5 may activate MyoD and myogenesis (see arrow in Figure 2), but this is not the only possibility since dorsal ectoderm can still activate MyoD in the absence of Myf5 (possibly through Pax3; Tajbakhsh et al., 1997). Soon after these initial inductive events, the great majority of myogenic cells express both MyoD and Myf5 proteins, although at different levels. This correlates with observations on older heterozygote Myf5/nlacZ mice where different types of muscle fibers all express β‐galactosidase, as well as MyoD. It also explains the phenotype of the single MyoD or Myf5 knock‐out mice where Myf5 null embryos initially have epaxial muscle defects whereas MyoD null embryos have predominantly hypaxial muscle defects (Kablar et al., 1997; S.Tajbakhsh and M.Buckingham, unpublished results) but the remaining gene is sufficient to support almost normal skeletal muscle development throughout the body.

Figure 1.

A simplified scheme of signaling molecules in newly formed epithelial somite. Shh (ochre dots), produced by notochord (Nc) and floor plate, acts on the ventral domain of newly formed epithelial somites, inducing sclerotome, and also on the dorso‐medial domain, inducing medial dermomyotome. Wnt1 (red dots), produced by dorsal neural tube (NT), acts (with Shh) on the dorso‐medial domain of newly formed somites (Sm), where Myf5 expression is observed soon after and epaxial progenitors are specified. Wnt7a (blue dots), produced by dorsal ectoderm (DE), acts on the dorso‐lateral domain, where hypaxial progenitors are specified. BMP4 (brown polygons), produced by lateral mesoderm (LM), prevents MyoD activation and early differentiation in the lateral domain of somites. Its action is counteracted by direct binding of Noggin (green triangles) produced by dorsal neural tube.

In the lateral myogenic progenitor cells, MyoD expression and subsequent terminal differentiation are transiently repressed in order to allow migration to the limb and body wall where muscle formation will take place. Although irreversibly committed (as shown by classic transplantation experiments), these myogenic progenitors do not express any member of the MyoD family either in the somite or during migration (Tajbakhsh and Buckingham, 1994), when they can be identified by the expression of the transcription factor Pax3 (Bober et al., 1994) and the receptor tyrosine kinase c‐Met (Bladt et al., 1995). It is therefore likely that their differentiation is repressed by signals derived from adjacent tissues. Mechanical separation of the lateral plate mesoderm from the paraxial mesoderm in the chick embryo induced expression of MyoD in the lateral half of somites where it is not observed normally. Furthermore, it was demonstrated that cells expressing BMP4 (Pourquié et al., 1996) could replace this inhibitory activity.

Signaling molecules replace the activity of neighboring tissues

A ventralizing signal emanating from the notochord and subsequently from the floor plate of the neural tube is required to activate Pax1 and specify a sclerotomal fate: Sonic hedgehog (Shh), normally produced by the notochord and the floor plate, can mimic its action (Fan and Tessier‐Lavigne, 1994). However, it has also been shown that Shh is required to promote myogenesis, and indeed in Shh null embryos epaxial myogenesis is absent whereas progenitors of hypaxial myogenesis, initially located in the dorso‐lateral region of newly formed somites and thus more distant from the notochord, are specified normally (Boricky et al., 1999). In addition, members of the Wnt family were also shown to activate myogenesis in the dorsal part of the somite (Münsterberg et al., 1995; Stern et al., 1995).

Because the onset of MyoD and Myf5 expression is spatially and temporally regulated in mouse embryos, it was important to know whether the differential activation of Myf5 and MyoD is promoted by different members of the Wnt family that are expressed in neural tube or in the dorsal ectoderm (Parr et al., 1993). Indeed, we observed that the Myf5‐inducing activity of the neural tube can be replaced by cells expressing Wnt1 and, to a lesser extent, Wnt4, while MyoD activation by dorsal ectoderm can be replaced by Wnt7a‐expressing cells (Tajbakhsh et al., 1998). Wnt7a is expressed in the correct spatio‐temporal pattern to be a candidate molecule for this activity. Sonic hedgehog synergizes with both Wnt1 and Wnt7a in explants from paraxial mesoderm of early E8.5 but not older E9.5 embryos.

With current information at hand, a parsimonious model of signaling activity can be proposed, as shown in Figure 1. Shh, produced by notochord and floor plate, activates Pax1 in the future sclerotome and, in conjunction with BMP4 (Murtaugh et al., 1999), initiates chondrogenesis in this anlagen. Shh, in conjunction with Wnt1 (and possibly other Wnts), activates myogenesis in the future dermomyotome, via a Myf5‐dependent pathway. Different Wnts such as Wnt7a activate myogenesis in the lateral domain, probably through a MyoD‐dependent pathway. This activity is inhibited by BMP4 to prevent premature differentiation; the negative action of BMP4 is counteracted, probably through direct protein–protein interaction, by Noggin which is produced by the dorsal neural tube in a Wnt‐dependent manner (Hirsinger et al., 1997; Marcelle et al., 1997).

Figure 2.

A model of Wnt signaling in somites. Wnt1 binds to Fz1 and, through a β‐catenin pathway, leads to activation of Myf5. Gli also converges on Myf5 activation. In contrast, Wnt7a binds to Fz7 and, through a β‐catenin independent pathway (G protein/C kinase?), leads to activation of MyoD. Pax3 also converges on MyoD activation, but whatever is downstream of the signaling is not known. Furthermore, Myf5 can activate MyoD, while the reverse is uncertain. Frzb1 can interfere with this process by sequestering Wnts in the extracellular space. Note that no experimental evidence for direct Wnt1–Fz1 or Wnt7a–Fz7 preferential binding is available so far.

Although the model illustrated in Figure 1 accommodates our present understanding, the rapid increase in the number of known molecules potentially involved in signaling during embryogenesis suggests that the model we now have will become more elaborate, and thus formal experimental evidence will be required. This will probably include in situ inhibition by specific antibodies (when they become available) together with detailed analysis of mutant and possibly compound mutant embryos.

From signaling molecules to downstream genes

If Shh and Wnts are the main ‘instructive’ molecules required to activate myogenesis, as current evidence suggests, then it becomes important to identify and understand their receptors and the downstream post‐receptorial pathways that lead to activation of Myf5 and MyoD. Also the relative role and mechanism of cooperation between the two different molecules need to be understood. At present, most studies are based on gene expression analysis, explant cultures and analysis of mutant phenotypes.

Recent work by Boricky et al. (1998) showed that genes of the Shh signal response pathway (Murone et al., 1999), specifically Patched, the Shh receptor, and Gli, zinc finger transcription factors, are activated in coordination with somite formation. They also showed that the expression of Patched, Gli and Gli2/4 is patterned differentially in the somite. This may explain how the same signaling molecule can specify two mutually exclusive fates such as cartilage and skeletal muscle. This provides an alternative explanation to a model in which a concentration gradient of Shh discriminates between sclerotomal and myotomal induction, as supported by notochord transplantation studies (Pourquié et al., 1993). More recently, the same authors (Boricky et al., 1999) showed that in Shh null embryos, epaxial myogenesis is abolished, while development of hypaxial musculature proceeds quite normally. This is in keeping with a crucial role for Shh in initiating epaxial myogenesis, before or together with the action of Wnts (epaxial muscle cells are the first to differentiate in the myotome). However, it is puzzling that in mouse explant cultures Shh can also cooperate with Wnt7a to activate MyoD (lateral pathway), even though neither Shh nor other members of the family have been detected in dorsal ectoderm. It remains to be verified whether hypaxial (lateral) myogenesis is independent of this signaling molecule in vivo and requires other, as yet unknown molecules that cooperate with Wnts.

Vertebrate homologs of the products Drosophila Frizzled are putative Wnt receptors (Bhanot et al., 1996). Several members of this family have been cloned in different organisms (Wodarz and Nusse, 1998; Dierick and Bejsovec, 1999). In order to correlate the Wnt‐dependent activation of myogenesis with the expression of specific Frizzled candidate receptors, we studied the expression of eight murine Frizzled (1 and 3–9) genes during mouse somitogenesis. All the Frizzled genes studied have a complex and partially overlapping pattern of expression in different regions of the embryo, and many of them (Fz1, 3, 7, 8 and 9) have specific expression in the epithelial somites as well as in the newly formed myotomes (Borello et al., 1999b). Since different domains of somites are defined by differential and overlapping expression of the various basic helix–loop–helix (bHLH) transcription factors (Patapoutian et al., 1995), it is tempting to speculate a possible correlation between the subsets of bHLH and Frizzled expressed. However, this analysis awaits development of specific antibodies that may distinguish among different Frizzled genes at a single cell level. In newly formed somites, Fz1 is expressed along the medial border, consistent with a possible preferential interaction with Wnt1 which is produced from the adjacent dorsal neural tube. On the other hand, Fz7 is expressed in a pattern complementary to Fz1, i.e. along the lateral and caudal edge of newly formed somites, consistent with the possibility of a preferential interaction with Wnt7a.

Wnt1 acts through the classic Dishevelled→ GSK3→ β‐catenin→ Tcf pathway (Wodarz and Nusse, 1998; Eastman and Grosschedl, 1999); in contrast, Wnt7a appears to act via a β‐catenin‐independent pathway (Kengaku et al., 1998) and leads to activation of MyoD rather than Myf5 (Tajbakhsh et al., 1998). It is thus tempting to speculate that Fz1 and Fz7 may mediate the differential response to Wnt1 and Wnt7a and activate different intracellular pathways (via either β‐catenin or G proteins/C kinase; Sheldahl et al., 1999) leading to differential activation of Myf5 or MyoD in epaxial or hypaxial progenitors (Figure 2). In the last few months, understanding of the Wnt/Fz signaling pathway has grown dramatically in complexity, with new proteins appearing continuously. Some of these proteins may act as co‐receptors (proteoglycans and LDL receptor homologs) while others interact with the axin–GSK3–β‐catenin complex (kinases, phosphatases, etc.), suggesting the existence of several possible steps of regulation as well as branching of the pathway (Peifer, 1999; Peters et al., 1999). This suggests that, in addition or alternatively to differential expression of Wnt receptors, other downstream molecules may be responsible for differential activation of Myf5 and MyoD. In any case, in vitro functional experiments with dominant‐negative molecules and specific inhibitors as well as analysis of appropriate null embryos will be needed to test whether Wnt1/Fz1 and Wnt7a/Fz7 (for once at least the numbers coincide) are key activators of Myf5 and MyoD, respectively, and to begin to unravel the specificity of each transduction pathway.

Another possible level of regulation of Wnt signaling may be exerted by the sFRPs (soluble Frizzled‐related proteins), a new class of genes recently identified in several laboratories (Leyns et al., 1997; Wang et al., 1997). These are secreted molecules with a strong homology to the Frizzled extracellular domain. Among the genes examined, only Frzb1 was found to be expressed in the pre‐somitic mesoderm and newly formed somites and thus its possible role in regulating myogenesis was investigated in detail. Frzb1 totally inhibits myogenesis in cultures of pre‐somitic mesoderm or newly formed somites, but has no effect on more mature somites or on myogenic cell lines, and thus appears to act differently from intracellular myogenic inhibitors such as Id or Twist. In order to examine the effect of Frzb1 overexpression in vivo, we developed a method based on transient transfection of cells with a Frzb1 expression vector and injection of transfected cells into the placenta of pregnant females before the onset of materno‐fetal circulation. Frzb1, secreted by transfected cells, accumulated in the embryo and caused a marked reduction of caudal structures. Myogenesis was strongly reduced and, in the most severe cases, abolished. Genes downstream of the Wnt signaling pathway such as En1, Noggin and Myf5 were down‐regulated but Pax3 and Mox1 were not, thereby excluding a generalized toxic effect (Borello et al., 1999a). The results obtained with this new method are in keeping with the Wnt1/Wnt3a double knockout and corroborate the idea that Wnt signals may act by regulating both myogenic commitment and expansion of committed cells in the mouse mesoderm. Indeed, in mouse embryos lacking both Wnt1 and Wnt3a, the medial compartment of the dermomyotome is not formed and the expression of Myf5 is severely reduced (Ikeya and Takada, 1998) but not abolished, probably because partial activation by Shh has occurred.

From the data discussed above, it appears that, at least medially, Shh and Wnts cooperate to activate myogenesis; they may also instruct diversification between epaxial and hypaxial myogenesis. How this can be achieved in molecular terms is still far from clear. Gli zinc fingers, activated by Shh, may bind directly to regulatory regions of Myf5 and MyoD promoters, whose complexity has made these studies difficult. Similarly, Wnt1‐activated Tcf–β‐catenin complex may directly activate transcription of target genes (among which are possibly Myf5 and MyoD) but also contribute to changes in chromatin configuration at these loci, making them more easily accessible to other transcription factors. It should be remembered, however, that Shh has been reported to be an important survival factor for paraxial mesoderm (Teillet et al., 1998; Cann et al., 1999) and it also has mitogenic activity on myoblasts (Duprez et al., 1998). Several Wnts, including Wnt1 and Wnt7a, have strong mitogenic and often transforming activity and most probably also act as survival factors (Wodarz and Nusse, 1999). Thus both molecules have the potential both to activate, directly or indirectly, transcription of Myf5 and MyoD, and to promote survival and expansion of the committed population. In other words, if we assume that a myotome is formed of four differentiated cells, these may in one extreme case derive from immediate differentiation of four progenitors responding to an inducing molecule. In the opposite situation, only one cell is induced that divides twice and generates four differentiated cells (Figure 3). In reality, combined action of Shh/Wnt on transcriptional activation, proliferation and survival must account for the final number of differentiated cells in a given structure such as the myotome. This is also relevant to the last issue discussed in this review, namely how different fates are chosen within contiguous and probably equivalent cells of the epithelial dermomyotome (Tajbakhsh and Cossu, 1997).

Figure 3.

Alternative mechanisms for the generation of a myotome. In response to inducing factors, several committed cells may differentiate directly without cell division (left). Alternatively, fewer committed cells need to proliferate in order to attain the final number of differentiated myotomal cells (right). In both cases, inhibition of differentiation of surrounding dermomyotomal cells is required.

The generation of myoblast diversity

It is important to recognize that the signals from neighboring tissues do not lead directly to terminal myogenic differentiation, at least in the majority of somitic cells exposed to the signaling molecules. Only a fraction of the cells exposed to Shh and Wnts signals are induced to differentiate terminally into myotomal muscle, while the remaining cells are probably kept in a committed but undifferentiated state. In this state, cells will divide, proliferate and sometimes die until the correct number of myoblasts is attained in the right place and at the right time to produce primary and secondary fibers and satellite cells during later development (Cossu et al., 1996a).

Although not proved definitively (see Bianco and Cossu, 1999; Miller et al., 1999), it is widely accepted that commitment of all myogenic progenitors occurs in the epithelial dermomyotome. It is thus necessary to postulate a mechanism by which one committed cell may prevent neighboring cells from adopting the same fate. In Drosophila, lateral inhibition through Notch and Delta has been shown as the probable mechanism by which adult myogenic progenitors are selected in response to Wg signaling (Baylies et al., 1998). It thus appears likely that a similar mechanism may operate in the mammalian somite. Indeed, several Delta and Notch genes are expressed in the somites (McGrew and Pourquié, 1998) and Notch inhibits myogenesis, probably through different intracellular mechanisms (Nofziger et al., 1999; Wilson‐Rawls et al., 1999). However direct evidence for a role for Notch in diversifying cell fate in mammalian somites is still missing.

Receptors for growth factors may be pertinent targets for Notch signaling. It has been proposed that the dorsal portion of the neural tube produces growth factors such as fibroblast growth factors (FGFs) that prevent differentiation of committed progenitors (Buffinger and Stockdale, 1995). If this is the case, only progenitors that will form the myotome may express few or no receptors (possibly as a consequence of Notch expression) and thus undergo terminal differentiation. From this point of view, it is interesting to note that the neural tube produces various FGFs, and the first cells which form the myotome are the only myogenic cells which do not express the FGF receptor FREK (Marcelle et al., 1995). While a causal relationship between lack of growth factor receptors and early differentiation remains to be demonstrated, this mechanism may be operating at different phases of muscle histogenesis (Cusella De Angelis et al., 1994) to explain the occurrence of different developmental choices within an apparently homogeneous population of committed myogenic progenitors.


We thank M.Buckingham, S.Tajbakhsh and B.Wold for helpful discussions and critical reading of the manuscript. Work in the authors' laboratory was supported by grants from Telethon, Fondazione Istituto Pasteur‐Cenci Bolognetti, EC, HFSP and MURST.