Ectopic Myf5 or MyoD prevents the neuronal differentiation program in addition to inducing skeletal muscle differentiation, in the chick neural tube

Skeletal muscle differentiation is crucially dependent on four basic-helix-loop-helix (bHLH) transcription factors, Myf5, MyoD (also known as Myod1), Myogenin and Mrf4 (also known as Myf6), which are named the myogenic regulatory factors (MRFs) (Weintraub et al., 1991). Null mutations in mice have revealed hierarchical relationships and apparent functional overlap among the MRFs. The primary MRFs, Myf5 and MyoD, are required for the determination of skeletal myoblasts,whereas the secondary MRFs, Myogenin and MRF4, act later in the program as differentiation factors (reviewed by Tajbakhsh and Buckingham,2000; Arnold and Braun,2000; Bergstrom and Tapscott,2001). Although Myf5 and MyoD can compensate for the absence of each other (Braun et al.,1992; Rudnicki et al.,1992; Rudnicki et al.,1993), they probably have distinct functions in the embryo. Myf5 transcripts are consistently detected prior those of MyoD in somites and limbs of mouse and chick embryos(Ontell et al., 1995; Tajbakhsh and Buckingham,2000; Delfini et al.,2000; Hirsinger et al.,2001) and in the chick head(Hacker and Guthrie, 1998). More importantly, in the absence of Myf5(Tajbakhsh et al., 1996), but not in that of MyoD (Kablar et al., 1999), presumptive myogenic cells adopt non-muscle fates(Tajbakhsh et al., 1996),demonstrating the importance of Myf5 for specifying muscle cells. Beside the recognised master role of the MRFs in triggering myogenesis in vertebrates, there is emerging evidence that other transcription factors are important for muscle formation. MEF2 transcription factors, although not specific to muscle lineage, are necessary co-factors of the MRFs to initiate myogenesis (Molkentin and Olson,1996). Pax3(Tajbakhsh et al., 1997), Lbx1 (Schäfer and Braun,1999; Gross et al.,2000; Brohmann et al.,2000), Mox2 (Mankoo et al., 1999) and Six1(Laclef et al., 2003)transcripts are all detected in skeletal muscle lineages and the corresponding knockout mice display a clear muscle phenotype, demonstrating their involvement in the molecular network controlling muscle development. The respective misexpression of the Lbx1(Mennerich and Braun, 2001)and Six1 (Heanue et al.,1999) genes in mesodermal chick explants confirm their involvement in the muscle molecular network. However, although current knowledge suggests that these transcription factors act genetically upstream or in parallel to the MRFs, the precise transcriptional relationships between the MRFs and these transcription factors are not completely understood. Growth factor signalling pathways also interfere with the network of muscle transcriptional factors, as the block of FgfR4 signalling in the chick limb downregulates Myf5and MyoD expression, leading to an inhibition of muscle differentiation. This shows that FgfR4 plays a crucial role in the cascade of molecular events leading to terminal muscle differentiation(Marics et al., 2002).

MRFs were discovered by their ability to convert various cell types into differentiation-competent myogenic cells(Weintraub et al., 1991). The MRFs have been shown to activate directly the expression of muscle-specific genes through interaction with E-box DNA-binding sites located in the promotors of these genes. Some cell types are permissive for myogenic conversion, while others are not(Weintraub et al., 1989). In addition, in a subset of permissive cell types, forced expression of MyoD is able to inhibit the ongoing differentiation program, in addition to promoting the expression of the myofibrillar proteins. For example, converted myosin-positive chondroblasts and RPE (retinal pigmented epithelial) cells following retroviral infection by MyoD are negative for cartilage-specific molecules and for melanin granules, respectively(Choi et al., 1990). However,there is less information on the ability of the MRFs to initiate ectopic myogenesis in vivo at ectopic sites in the embryo. Experiments expressing Myf5 or MyoD at ectopic sites in Xenopus and mouse embryos have shown that both Myf5 and MyoD can initiate expression of early muscle differentiation markers (Hopwood and Gurdon, 1990; Hopwood et al., 1991; Frank and Harland,1991; Miner et al.,1992; Faerman et al.,1993; Santerre et al.,1993). According to these studies, forced expression of Myf5 or MyoD does not induce the full set of markers,reflecting muscle terminal differentiation, which consequently has led to the general idea that Myf5 and MyoD have incomplete myogenic capabilities in vivo(reviewed by Pownall et al.,2002). Nevertheless, there are two reports of muscle differentiation in brain tissues: in Xenopus hindbrain after injection of Xmyf5 or XmyoD into blastomeres of two- to 32-cell stage (Ludolph et al.,1994) and in the adult mouse brain of transgenic mice expressing bovine Myf5 under the control of a viral promoter(Santerre et al., 1993). However, in these studies, the ectodermal origin of the ectopic muscle cells was not completely proved.

In this report, we investigate the consequences of forced expression of the mouse Myf5 and MyoD (Myod1 - Mouse Genome Informatics) genes in the chick neural tube, after electroporation. We show that ectopic mouse Myf5 and MyoD in the neural tube induces skeletal muscle differentiation in addition to inhibiting the neuronal differentiation program. This system allowed us to analyse, in an in vivo context, the transcriptional relationships between several factors known to be involved in myogenesis.

Fertilised eggs from commercial sources (JA 57 strain, ISA, Lyon, France)were incubated at 38°C. All electroporation were performed in ovo. Embryos were classically staged according to Hamburger and Hamilton (HH)(Hamburger and Hamilton,1992).

The complete mouse MyoD- and Myf5-coding sequences (a gift from S. Tajbakhsh, Pasteur Institute, Paris) were inserted into the pCAβ expression vector, which contains the hybrid chicken β-actin promoter/CMV enhancer (Yaneza et al.,2002). This hybrid promotor has been shown to drive efficiently the expression of the inserted gene in chick neural tube(Momose et al., 1999; Yaneza et al., 2002; Dubreuil et al., 2002).

Embryos were co-electroporated with GFP/pCAβ, a vector encoding enhanced versions of green fluorescence protein that also served as control(Momose et al., 1999), and mouse Myf5/pCAβ or mouse MyoD/pCAβ. The recombinant expression vectors were used at 1 to 4 μg/μl. The DNA was microinjected into the lumen of the neural tube at the trunk level of HH13-14 (E2) chick embryos with micropipettes. Electrodes were placed on either side of the embryos adjacent to somites 10-20. A square-wave stimulator (AM-Systems) was used to deliver five pulses of current at 35 volts for 50 ms each. With unilateral pulses,only the right parts of neural tubes were transfected. Embryos were allowed to develop at 38°C for 1-4 days, and processed for in situ hybridisation to paraffin wax-embedded tissue sections.

Embryos were fixed and processed for in situ hybridisation to tissue sections as previously described (Delfini et al., 2000). The antisense digoxigenin-labelled mRNA probes,were prepared as described: mouse MyoD and mouse Myf5, Myf5,MyoD, Myogenin, Delta1, Serrate2, Notch1, Pax3(Delfini et al., 2000), Paraxis (Delfini and Duprez,2000), FgfR4(Edom-Vovard et al., 2001) and SCG10 (Stanke et al.,1999). The probes for Six1, Mox1, Mox2, Id2 originate from the UMIST EST library (Boardman et al., 2002).

Differentiated muscle cells and neural crest cells were detected on sections by using the monoclonal antibodies MF20 and HNK1, respectively(Developmental Hybridoma Bank, University of Iowa). Axonal projections were recognised using the monoclonal antibody 3A10 recognising a neurofilament-associated antigen (Developmental Hybridoma Bank, University of Iowa). Immunohistochemistry were performed after the in situ hybridisation.

The regulation of each gene was tested on three to six embryos electroporated with mouse MyoD and on a minimum of two embryos electroporated with mouse Myf5, except for Mef2c (which was not tested after mouse Myf5 electroporation). For all probes, the multiple experiments gave consistent results (no regulation, down- or upregulation), showing 100%efficiency. Moreover, the mouse MyoD and Myf5 electroporated embryos always gave the same results, in term of the direction of regulation observed. The chick Myf5 and MyoD probes were tested on mouse tissue sections of E12.5 embryos and did not give any signals, excluding the problem of probe crossreaction between species.

Mouse Myf5 or MyoD recombinant vectors were coelectroporated with a GFP/pCAβ in the right hand sides of E2 chick neural tubes, at the trunk level. The use of the GFP allowed us to select the embryos that were well electroporated along the AP axis of the embryos. However, in order to visualise the location of the electroporated genes, we performed in situ hybridisation experiments to transverse sections using the mouse-specific probes for Myf5 and MyoD. The ectopic expression of mouse Myf5 and MyoD genes can be detected throughout the right sides of the neural tube and in the emigrating neural crest cells 1 day after electroporation(Fig. 1A,E). We used this system to analyse the transcriptional regulation between the MRFs in vivo. Our data show that forced expression of mouse Myf5 and MyoDinduces the ectopic expression of MyoD, but not that of Myf5, in the neural tube and neural crest cells(Fig. 1A-C,E-G). Myogenin, which is known to act downstream of Myf5 and MyoD genes (Nabeshima et al.,1993; Hasty et al.,1993), is also clearly ectopically induced in the neural tube after electroporation of mouse Myf5(Fig. 1A,D) and MyoD(Fig. 1E,H). Four days after electroporation, chick MyoD and Myogenin transcripts are still ectopically detected in MF20-positive cells (data not shown).

In conclusion, in this system, Myf5 expression is not regulated by itself or mouse MyoD, while that of MyoD expression is regulated by itself and by mouse Myf5.

In order to determine whether either mouse Myf5 or MyoDwere able to induce marked myogenesis in neural tissues, we performed immunohistochemistry experiments with the MF20 antibody directed against sarcomeric myosin heavy chains (MHC) after in situ hybridisation using the mouse Myf5 or MyoD probes(Fig. 2). Four days after electroporation of either mouse Myf5(Fig. 2A-C) or MyoD(Fig. 2D-H), we can observe MF20-positive cells in the neural tube(Fig. 2D,G,H see also Figs 3, 4) and in the neural crest cell derivatives (Fig. 2A-F). These MF20-positive cells show the elongated aspect of skeletal muscle fibres. High magnifications of these MF20-positive cells show striations reminiscent of the classical sarcomeric organisation of normal muscle fibres, indicating that terminal muscle differentiation has occurred in ectodermal tissue(Fig. 2C,F,H). However, it is not clear whether these MF20-positive cells are multinucleated. One possibility (not exclusive with a multinucleated state) is that myosin expression spreads in the axons of the neurons. It is noticeable that the orientation of the MF20-positive cells can follow the path of the axons of sensory neurons in the DRG, dorsal root ganglia(Fig. 2D-F) or that of the motoneuron axons (Fig. 2G,H). MF20-positive cells can be observed transversally and longitudinally anywhere along the dorsoventral axis of the neural tube. However, we never observed MF20-positive cells in the proliferating ventricular zone, consistent with the absence of ectopic mouse Myf5 or MyoD in this part of the neural tube, 4 days after electroporation.

The activation of myosin heavy chains in neural tissues prompted us to ask whether ectopic expression of mouse Myf5 or MyoD simply activates myogenesis in the neural tube or whether it could also act negatively to exclude expression of the endogenous cellular phenotype in vivo. We first analysed the overall pattern of developing axonal projections, using the 3A10 antibody, which recognises an associated-neurofilament protein. We found that ectopic muscle differentiation, assayed by MF20 labelling(Fig. 3A), clearly disrupts the organisation of the developing axons (Fig. 3B, arrows) in the neural tube. A merged image showing both myosin and neurofilament expression reveals no double-labelled cells(Fig. 3C,D), indicating a mutual exclusion of muscle and neuronal differentiation programs. In addition,we observed a downregulation of HNK1 labelling in neural crest cells that expressed mouse Myf5 and MyoD (data not shown). We also investigated the effect of ectopic myogenesis on the transcription of neuronal markers. Of the multitude of specific neuronal differentiation markers(reflecting the various neuronal cell types), we chose the pan-neuronal differentiation marker SCG10 (Anderson and Axel, 1985). SCG10 encodes a membrane-bound protein that accumulates in the growth cones and perinuclear cytoplasm of developing neurons (Stein et al., 1988). The function of this protein is not completely elucidated but SCG10is expressed very specifically in the central and peripheral nervous systems(Wuenschell et al., 1990; Mori et al., 1990) and thus provides a very useful general marker of nervous systems. We also used the cell-adhesion molecule BEN(Pourquié et al., 1990)as a marker of the motoneurons and DRGs, although not homogeneously expressed in those structures (Fournier-Thibault et al., 1999; Fraboulet et al.,2000). BEN belongs to the immunoglobulin superfamily and is suspected of playing a role in the fasciculation of axons(Pourquié et al.,1992). Four days after electroporation, forced-expression of mouse MyoD (Fig. 4A,B) led to a clear downregulation of the expression of SCG10(Fig. 4C,D) and BEN(Fig. 4E,F) where the ectopic mouse MyoD is detected (Fig. 4A,B). MF20 labelling is detected (in brown) precisely in the area where the expression of SCG10 and BEN is downregulated by ectopic mouse MyoD expression(Fig. 4B,D,F). Based on the downregulation of SCG10 expression at stage HH20(Fig. 4G,H), we conclude that the inhibition of the neuronal differentiation program has already started, 1 day after electroporation. These results show that the forced expression of a MRF diverts cells from the neuronal differentiation program to a muscle program.

Notch signalling and the HLH transcription factor Id2 are thought to be negative regulators of myogenesis. Overexpression of Id protein inhibits the muscle differentiation program by association with E2A proteins in vivo(Jen et al., 1992). Activation of the Notch pathway inhibits myogenesis in the chick and Xenopusembryos (Delfini et al., 2000; Hirsinger et al., 2001; Kopan,1994). We found that ectopic mouse MyoD(Fig. 5A) or Myf5(data not shown) does not upor downregulate Delta1 expression(Fig. 5B). However, ectopic mouse MyoD activates the expression of another Notch ligand, Serrate2 (Fig. 5C). The expression of the receptor Notch1 is also activated by ectopic mouse MyoD (Fig. 5D),reflecting an activation of Notch signalling(Wilkinson et al., 1994; Lewis, 1996; Delfini et al., 2000; Hirsinger et al., 2001). Forced expression of either mouse Myf5 (data not shown) or MyoD also activates the expression of Id2(Fig. 5E,F).

One day after electroporation, using double-labelling in situ hybridisation(data not shown), we could not discriminate precisely at a cellular level the location of the induced Serrate2 and Notch1 versus the ectopic MRF. However, 4 days after electroporation, the receptor Notcth1 is not detected in MF20-positive cells(Fig. 6A,B,D). In addition,there is clearly an ectopic expression of Notch1 surrounding the MF20-positive cells (Fig. 6A,B,D), although Serrate2 transcripts can be detected in the MF20-positive cells (Fig. 6E). This expression is reminiscent of the endogenous location of Notch components during normal myogenesis(Delfini et al., 2000; Hirsinger et al., 2001). We conclude that ectopic mouse Myf5 or MyoD activates the Notch pathway via its ligand Serrate2.

We used this system to analyse the expression of several genes involved in myogenesis. Mef2c has been shown to be a direct transcriptional target of myogenic bHLH factors in transgenic mice(Wang et al., 2001). Consistent with this result, ectopic mouse MyoD induces the expression of Mef2c in the neural tube (Fig. 7A,B). Pax3 and Paraxis are two transcription factors expressed in myogenic precursors prior to MRF expression in limbs(Delfini and Duprez, 2000) and somites (Burgess et al.,1996). In contrast to Paraxis, Pax3 is endogenously expressed in the dorsal part of the neural tube. Ectopic expression of either mouse Myf5 (Fig. 7C)or MyoD (data not shown) does not seem to modify the endogenous expression of Pax3 (Fig. 7D) and does not activate any Paraxis expression in the neural tube (data not shown). The invalidation of the transcription factor Mox2 (Mankoo et al.,1999) and that of Six1(Laclef et al., 2003) in mice led to downregulation of the expression of Myf5 and MyoD,respectively, in the limb. However, the reciprocal relationship, i.e. regulation of Mox2 and Six1 by the MRFs has not been investigated. In our experiments, forced expression of either mouse Myf5 (data not shown) or MyoD(Fig. 7E-H) did not induce any ectopic expression of Mox1, Mox2 or Six1 in the neural tube,whereas they are normally detected in the somites.

It has been shown that FgfR4 expression is a necessary step for terminal muscle differentiation (Marics et al., 2002). However, the transcriptional relationships between FgfR4 and the myogenic factors are not completely understood. We found that either mouse Myf5 (data not shown) or MyoD(Fig. 8A) is able to induce ectopic expression of FgfR4 (Fig. 8B) in the neural tube and neural crest cells 1 day after electroporation. This last result indicates the existence of an unexpected regulatory loop between MyoD and FgfR4. We also investigated how fast this upregulation occurs after electroporation of mouse MyoD. Fluorescence of GFP is detected roughly 3 hours after electroporation. We observed an ectopic expression of FgfR4 in the neural tube, 3(n=1/1), 4 (n=1/1), 5 (1/1) and 6 (n=5/5) hours after electroporation of mouse MyoD (Fig. 8C,D). This result indicates that the induction of FgfR4expression by MyoD might be direct.

We have shown that in addition to triggering skeletal muscle differentiation in the neural tube, the ectopic expression of either mouse Myf5 or mouse MyoD also leads to an inhibition of the endogenous neuronal differentiation program. Transitions between two distinct cell types derived from the same embryonic sheet (ectoderm, mesoderm or endoderm) have been documented. Classical examples of such transdifferentiation in mammals are the following: (1) conversion of pigmented epithelial cells from the iris into lens, both derived from the ectoderm(Okada, 1980); (2) transition of pancreatic cells to hepatocytes, both derived from foregut endoderm(reviewed by Horb et al.,2003); (3) conversion of muscle cells to adipocytes, both derived from mesoderm; this conversion has been shown to be inhibited by Wnt signalling (Ross et al.,2000); and (4) transdifferentiation of oesophagal smooth to skeletal muscle during normal mouse development under the influence of Myf5(Kablar et al., 2000). Lineage restriction into ectodermal, mesodermal and endodermal germ layers during development was thought to be irreversible. However, the switch across embryonic germ layers takes places during tail regeneration in the axolotl,where cells originally derived from the ectoderm contribute to muscle and cartilage originally made from mesoderm(Echeverri and Tanaka, 2002). In addition, although the issues are still controversial, recent data indicate the existence of adult stem cells from various tissues, including brain, skin,liver and bone marrow that can form cell types of different lineages when exposed to novel or foreign environments (Anderson et al., 2001; Blau et al., 2001). Muscle differentiation has already been observed in neural tube explants under the influence of Myf5 and/or MyoD in specific conditions(Tajbakhsh et al., 1994; Maroto et al., 1997). Myosin expression has been observed in the few Myf5-lacZ-positive cells in culture of neural tubes from the Myf5-lacZ mice(Tajbakhsh et al., 1994). Induction of ectopic Myf5 and MyoD after the forced expression of Pax3/RCAS is also able to induce muscle differentiation assayed by the expression of myosin by RT-PCR in neural tube explants(Maroto et al., 1997). However, in this latter study, markers for the endogenous neuronal program were not examined.

The proliferating cells lining the lumen of the embryonic neural tube have stem cell characteristics (Anderson,2001). It is conceivable that the forced presence of mouse Myf5 or MyoD in those cells triggers the muscle differentiation program instead of the endogenous neuronal differentiation program. Consistent with this hypothesis, we never observed coexistence of the late neuronal differentiation markers and ectopic MRF.

We have found that forced expression of mouse MyoD is able to activate the expression of inhibitors of myogenesis, Id2 and Notch components. The activation of Notch components by MyoD is also observed in the Xenopus embryos, where XMyoD has been shown to activate the Notch pathway, via its ligand Delta1(Wittenberger et al., 1999). This stimulation of myogenesis inhibitor transcription by MyoD is also consistent with the large scale analysis of the genes regulated by MyoD in 10T1/2 cells, in which the mRNA levels of Notch signalling components and Id2 are increased (Bergstrom et al.,2002). This type of regulation reinforces the idea that the bHLH factor MyoD uses the classic scheme of lateral inhibition for generating muscle precursors. MyoD will trigger myogenic differentiation in some cells and drive the activation of the Notch pathway in neighbouring cells, which will in turn repress the expression and the activity of the bHLH gene,stopping muscle differentiation in those cells. Consistent with this, forced activation of the Notch signalling pathway (using the Delta1/RCAS virus)inhibits MyoD expression in the chick limb and somite(Delfini et al., 2000; Hirsinger et al., 2001) and interferes with MyoD activity (Kopan et al., 1994; Wilson-Rawls et al., 1999).

Notch components and Id proteins are general inhibitors of differentiation,including neuronal differentiation. Id proteins are associated with neural cell proliferation and inhibit differentiation in a variety of systems(Martisen and Broner-Fraser et al., 1998; Lyden et al., 1999; Norton, 2000). This activation of myogenesis inhibitors by the myogenic bHLH factor, MyoD is to be related to that of negative regulators of neurogenesis by the proneural bHLH genes, such as neurogenins (Ngns) (Bertrand et al.,2002). Ngn2 activates the expression of Id2 in the chick spinal cord (Dubreuil et al.,2002). The Ngns also activate the expression of Notch ligands,which will then trigger the activation of Notch pathway(Ma et al., 1996; Ma et al., 1998; Dubreuil et al., 2002). This property to induce differentiation and to generate supernumerary progenitors through the canonical lateral inhibition mechanism seems to be a general feature of bHLH factors. This general mechanism could provide an explanation of why, in the competition between the (ectopic) muscle and (endogenous)neuronal bHLH transcription factors, the myogenic factor is stronger than the proneural gene in imposing its downstream differentiation program in the neural tube context. First the expression MyoD, which is driven by an artificial promoter leads to high levels of MyoD protein. In addition, the activation of Notch signalling by the bHLH factors will be able to repress as a feedback action the endogenous expression of proneural genes(Ma et al., 1996) but not that of the transfected mouse MyoD driven by artificial promoters. Interestingly, in a mesodermal context, retroviral misexpression of proneural genes, such as Ngns, in the somites is sufficient to induce the ectopic expression of SCG10, in addition to sensory-neuron-specific markers,in dermomyotomes (Perez et al.,1999). However, this study did not analyse whether the endogenous myogenic program was inhibited or co-existed with the ectopic neurogenic program.

We have shown that both mouse Myf5 and MyoD induce ectopic expression of MyoD, but not that of Myf5, in the chick neural tube. Although we cannot rule out completely the possibility that chick Myf5 and MyoD (versus mouse) would activate the expression of endogenous avian Myf5, these results are completely in line with those obtained in Xenopus. Injection of XMyf5 and XMyoD activates the transcription of the endogenous MyoD gene but not that of the Myf5 gene in animal cap cells(Hopwood et al., 1991). The autoregulatory loop of MyoD is also observed in several mouse muscle cell lines (Thayer et al.,1989; Braun et al.,1989a; Bergstrom et al.,2002). This is also in agreement with the fact that the DRR enhancer of the MyoD gene contains conserved E-box binding sites and thus may be a direct target of regulation by MyoD itself(Kablar et al., 1997; Kablar et al., 1999). The absence of Myf5 activation either by itself or by MyoD has also been noted in vitro (Braun et al.,1989b; Bergstrom et al.,2002). Both mouse Myf5 and MyoD lead to ectopic expression of Myogenin in the chick neural tube, which is consistent with Myogenin activation by Myf5 and MyoD in vitro(Braun et al., 1989a; Braun et al., 1989b; Bergstrom et al., 2002) and in vivo (Miner et al., 1992; Santerre et al., 1993), and with mouse genetic studies that place Myogenin downstream of Myf5 and MyoD (Hasty et al., 1993; Nabeshima et al.,1993).

Our in vivo results have several implications concerning the involvement of Myf5 versus MyoD in myogenesis. First, they support the idea that Myf5 is the first MRF to specify the skeletal muscle program in vertebrates, MyoD having later function(Tajbakhsh et al., 1997; Tajbakhsh and Buckingham,2000; Delfini et al.,2000; Hirsinger et al.,2001). Another aspect of our results is that MyoD can induce muscle differentiation in the absence of Myf5, Paraxis, Six1,Mox1 and Mox2 in vivo, as the conversion of neural tissues into skeletal muscle cells can occur in the absence of the corresponding transcripts. This could be related to the block of muscle differentiation in the absence of MyoD, despite the presence of Myf5, Pax3 and Paraxis after Notch signalling activation in the chick somites and limbs (Delfini et al., 2000; Hirsinger et al., 2001). A last aspect is that we never found any differences in terms of gene regulation between Myf5 and MyoD. The fact that all the genes induced by mouse MyoD are also induced by mouse Myf5 is expected, as mouse Myf5 induces MyoD. However, the converse is not true. We might expect to find genes only regulated by mouse Myf5, but we have tested at least 20 genes without observing this cast.

In Xenopus, chick and mouse, Myf5 and MyoDexpression has been detected in non-muscle mesodermal tissues. In Xenopus, ventral (nonsomitic) mesoderm transiently expresses MyoD during gastrulation (Frank and Harland, 1991). Myf5 transcripts also have been detected at low levels in presegmental mesoderm in chick(Kiefer and Hauschka, 2001; Hirsinger et al., 2001) and in mouse (Gerhart et al., 2000)embryos, suggesting the existence of inhibitory mechanisms of protein translation or activity. In relation to this, it is important to realise that,even in vitro, the myogenic conversion property of the MRF can only occur after growth factor depletion (Weintraub et al., 1989; Yutzey et al.,1990), showing the importance of inhibitory mechanisms in muscle differentiation. Myf5 transcripts have also been detected in the neural tube (Tajbakhsh et al.,1994) and specific neurons of the brain(Tajbakhsh and Buckingham,1995; Daubas et al.,2000). However, in this latter instance, the Myf5 protein is not produced, providing evidence for post-transcriptional control mechanisms for Myf5 in the mouse brain (Daubas et al.,2000). If such post-transcriptional regulation exists in the neural tube, it is not strong enough to block the protein synthesis of the ectopic MRF or, alternatively, the recombinant constructions do not contain the regulatory sequences necessary to respond to such a signal.

The observation of FgfR4 induction in the neural tube after forced expression of MyoD indicates the existence of an unexpected positive feedback loop from MyoD to FgfR4 during myogenesis. This FgfR4 upregulation by MyoD is consistent with that observed in human cells (S. J. Tapscott, personal communication). The upregulation of FgfR4 observed 3 hours after electroporation supports the idea of a direct regulation. Although the FgfR4 mutant mice are not very informative (Weinstein et al.,1998), the fact that in the chick limb FgfR4 transcripts are detected before those of MyoD(Marcelle et al., 1995) and that inhibition of FgfR4 signalling leads to a downregulation of MyoDexpression (Marics et al.,2002), clearly places FgfR4 as acting upstream of MyoD. However, the precise role of FgfR4 in myogenesis is not completely clear. FgfR4 transcripts are unambiguously detected in mononucleated cells surrounding the muscle fibres (Marcelle et al., 1994; Edom-Vovard et al., 2001) and FgfR4 downregulation is concomitant with terminal differentiation in muscle cell lines (Halevy et al.,1994), indicating a role in myoblast proliferation of this signalling pathway (Marcelle et al.,1995). However, inhibition of FgfR4 signalling seems not to modify cell proliferation in the chick limb, suggesting instead a role in muscle differentiation (Marics et al.,2002). The existence of a feedback loop from MyoD to FgfR4 could highlight a dual action of FgfR4 signalling pathway in myogenesis. In the chick limb at stage HH22/23, FgfR4 transcripts present an expression domain similar to that of Myf5 compared with the restricted MyoD domain(Delfini et al., 2000) (data not shown). In the limb muscle masses we can distinguish a FgfR4- and Myf5-positive domain close to the ectoderm and a more central domain positive for FgfR4, Myf5 and MyoD. One hypothesis is that low levels of FgfR4 signal would allow myoblast proliferation in the Myf5 domain of the muscle masses, while high levels (induced by MyoD)could initiate muscle differentiation in the MyoD domain. This hypothesis could be related to the biphasic regulation of FgfR4expression by different concentrations of Fgfs in muscle cell lines; this differential expression of FgfR4 is correlated with the state of muscle differentiation of the cells(Halevy et al., 1994; Pizette et al., 1996). Interestingly, in the somites, FgfR4 transcripts display two distinct domains of expression. FgfR4 transcripts are located along the rostral and caudal edges of the dermomyotomes in mitotically active cells that do not express MyoD (Kahane et al., 2001). In a second domain, FgfR4 transcripts are also co-localised with those of Myf5 and MyoD in the sublipdomain, an area subjacent at the dermomyotomal lips, which exhibits intermediary properties between dermomyotome and myotome(Cinnamon et al., 2001).

In conclusion, we have shown that either mouse Myf5 or MyoD is able to induce muscle differentiation in neural tissues at the expense of the endogenous neuronal differentiation program. This provides an in vivo system, in which to study regulation between muscle factors in the absence of the classical muscle molecular context.

We thank Muriel Altabef, Christo Goridis, Chaya Kalcheim, Pascal Maire,Olivier Pourquié and Marie-Aimée Teillet for discussion and/or critical reading of the manuscript. We are grateful for providing reagents to Sharagim Tajbakhsh (coding sequences of mouse Myf5 and MyoD), Jonathan Gilthorpe (pCAβ vector), Hermann Rohrer (SCG10 probe) and Ketan Patel(Mef2c probe). This work was supported by the Association Française contre les Myopathies (AFM), the Association pour la Recherche contre le Cancer (ARC), the Ministère de la Recherche (ACI jeunes chercheurs) the Fondation pour la Recherche Médicale (FRM) and the Centre National de la Recherche Scientifique (CNRS). MCD is supported by the French Ministery and the Fondation Bettencourt Schueller.