Delta 1-activated Notch inhibits muscle differentiation without affecting Myf5 and Pax3 expression in chick limb myogenesis

In vertebrates, all the myogenic cells that form the striated skeletal muscles of the limb and trunk originate from the somites. The medial halves of the somites give rise to back and intercostal muscles or the epaxial musculature (Ordahl and Le Douarin, 1992). Cells derived from lateral dermomyotomes migrate lateroventrally to produce the muscles of the body wall and the limbs, forming the hypaxial musculature (Ordahl and Le Douarin, 1992; Christ and Ordahl, 1995). Formation of limb skeletal muscle during vertebrate embryogenesis involves cellular commitment, migration, proliferation, growth arrest and differentiation. Each step involves the expression and activity of a specific panel of factors (Olson, 1992). The myogenic bHLH transcription factors, Myf5, MyoD (also known as Myod1), Mrf4 (also known as Myf6) and myogenin, which are also called myogenic regulatory factors (MRFs), have been shown to initiate the skeletal muscle differentiation program.

Ectopic expression of these MRFs is able to convert several non-muscle cell types into skeletal muscle in tissue culture (Weintraub et al., 1991), in transgenic mice (Miner et al., 1992; Santerre et al., 1993) and in Xenopus (Ludolph et al., 1994). Conversely, knockout of these genes leads to various muscle defects (Rudnicki et al., 1992, 1993; Braun et al., 1992a,b; Hasty et al., 1993; Nabeshima et al., 1993). Moreover, cells deprived of Myf5 or MyoD assume a non-muscle fate (Tajbakhsh et al., 1996; Kablar et al., 1999). These properties have led to the notion that MRFs trigger the successive events leading to skeletal muscle formation. Gene targeting has clearly defined a hierarchy among the MRF family members. Mice lacking Myf5 and MyoD (Rudnicki et al., 1993) do not form myoblasts or skeletal muscle. In contrast, in myogenin-null mice, myoblasts do form, as assayed by Myf5 and MyoD expression, but do not differentiate into muscle fibres (Hasty et al., 1993; Nabeshima et al., 1993). Myf5 and MyoD therefore appear to lie in a genetic pathway upstream of myogenin, the latter having a role in activating muscle cell terminal differentiation (Hasty et al., 1993; Nabeshima et al., 1993). In mice, there are numerous and consistent studies concerning the sequential expression of myogenic factors during somite and limb bud development, Myf5 being detected before MyoD in somites and limbs (Ontell et al., 1995; Tajbakhsh and Buckingham, 1999). In birds, studies of the timing of the expression of these factors in axial regions have led to the conclusion that, in contrast to the situation in mice, MyoD expression occurs before that of Myf5 by a few hours (Pownwall and Emerson, 1992; Boricky et al., 1997; Denetclaw and Ordahl, 2000). However, Hacker and Guthrie (1998) found that Myf5 transcripts were expressed first, followed by those of MyoD. In the limb, studies using in situ hybridisation of tissue sections (Williams and Ordahl, 1994), RT-PCR (Lin-Jones and Hauschka, 1996) and whole-mount in situ hybridisation (Hacker and Guthrie, 1998) have led to conflicting results concerning the timing of appearance of MyoD and Myf5 transcripts.

The progression through discrete developmental steps has been studied in muscle cell lines. The presence of Myf5 and MyoD is not itself sufficient to trigger differentiation in cell culture, since myoblasts exposed to growth factors continue to proliferate and to express Myf5 and/or MyoD (Yutzey et al., 1990). One well-described general mechanism influencing differentiation events during development is the Notch signalling pathway (reviewed in Artavanis-Tsakonas et al., 1999). The Notch pathway has been shown to operate at different steps during Drosophila myogenesis (Baylies et al., 1998). However, although mice bearing null mutations in the different Notch signalling components exhibit defects of somite formation, they do not display any muscle defects (Swiatek et al., 1994; Conlon et al., 1995; de Angelis et al., 1997). The lack of effect probably reflects the functional overlap among the Notch family members. In mammals, the only direct evidence of Notch involvement in myogenesis comes from in vitro studies where activated Notch or ligand-induced Notch signalling suppresses muscle differentiation in various mouse cell lines (Kopan et al., 1994; Nye et al., 1994; Lindsell et al., 1995; Shawber et al., 1996; Jarriault et al., 1998; Nofziger et al., 1999; Kuroda et al., 1999). No such evidence has been obtained in vivo.

Knockouts and studies on cell lines have proved very powerful in determining the genetic hierarchy of MRFs and giving clues about their functions. However, the exact functions of the proteins coded by Myf5 and MyoD during development are still not fully understood. Based on in vitro studies, Myf5 and MyoD are widely considered to have overlapping function associated with myoblast proliferation (Lassar et al., 1994; Molkentin and Olson, 1996). In order to gain insight into the respective roles of Myf5 and MyoD in avian limb myogenesis, we have characterised the cellular expression patterns of these genes. We found that in the chick limb, Myf5 can be detected at stage 20 with an expression domain similar to that of Pax3, MyoD being detected a few hours later (stage 22) in a more restricted domain. Activation of the Notch pathway in vivo led to a downregulation of MyoD expression, without affecting Pax3 and Myf5 expression, followed by an inhibition of terminal differentiation. Together, these results suggest that Myf5 acts upstream of MyoD, and that the Notch pathway is involved in the progression from the Myf5-expressing stage to the MyoD-expressing stage, during chick limb bud myogenesis.

Fertilised White Leghorn eggs (HAAS, Strasbourg, France) were incubated at 37°C. All grafting experiments were performed in ovo. Young embryos were staged according to Hamburger and Hamilton (HH) (1951), while old embryos were staged according to embryonic days in ovo. To facilitate comparisons, we report both staging for young embryos.

Infectious Delta/RCAS (Henrique et al., 1997) and control/RCAS viruses were produced in Chick Embryo Fibroblasts (CEF) as described by Duprez et al. (1998). Briefly, CEF were isolated from E10 O-line embryos (BBSRC, Institute for Animal Health, Compton, Berkshire, UK) and grown in DMEM (Gibco, BRL) containing 8% (v/v) fetal calf serum and 2% (v/v) chick serum supplemented with antibiotics. CEF were transfected transiently with retroviral recombinant DNA using Transfectam (Gibco, BRL) according to the manufacturer’s instructions.

Retrovirus-expressing cells were prepared for grafting as described by Duprez et al. (1998). Pellets of approximately 50 to 100 μm in diameter were grafted into the limb field of White Leghorn embryos around stage HH 16 of development (E2.5). Embryos were harvested at different times after grafting and processed for in situ hybridisation of whole mounts or tissue sections. Embryos grafted with control/RCAS-expressing cells did not exhibit any change in morphology (see also Duprez et al., 1996) or gene expression (data not shown). The numbers of embryos processed for in situ hybridisation of whole mounts are given in the text or in Table 1. In each experiment, two to six specimens were used for in situ hybridisation of tissue sections.

E3 embryos were injected in the amnios (near the heart and wing) with 200μl of 10mM BrdU (Amersham, Life Science), and were re-incubated for another 15 minutes. 1 μl of 10 mM BrdU was directly injected in the circulation of E7 embryos and fixed 1 hour after. The embryos were then fixed and processed for in situ hybridisation of sections.

In situ hybridisation of whole mounts and tissue sections

Embryos were fixed in 4% (v/v) formaldehyde and processed as previously described for in situ hybridisation of whole mounts and paraffin sections (Duprez et al., 1998, 1999). Antisense digoxigenin- and fluorescein-labelled RNA probes were prepared as follows: Myf5 (Saitoh et al., 1993); Pax3 and MyoD (Duprez et al., 1998), Delta1 (Henrique et al., 1997), Serrate2 and Notch1 (kind gift from Dominguos Henrique). For double in situ hybridisation, the fluorescein probe was revealed with NBT/BCIP reagents (Roche) first, then the digoxigenin probe with INT/BCIP (Roche).

Differentiated muscle cells were detected on sections and in cultures using a monoclonal antibody against sarcomeric myosin heavy chain, MF20 (Developmental Hybridoma Bank, University of Iowa, Iowa City). Proliferating cells were detected using a monoclonal antibody against BrdU (Amersham). In situ followed by immunohistochemistry was performed using successively the probes (Notch1, Serrate2, Delta1, Pax3, MyoD and Myf5) and the monoclonal antibody against BrdU or MF20 antibody.

In order to clear up the controversy concerning the timing of appearance of Myf5 and MyoD, we performed in situ hybridisation on serial transverse limb sections (Fig. 1). Myf5 and MyoD transcripts were not detected during the migration of the muscle progenitors from the somites to the limb bud (data not shown). This is in agreement with chick (Williams and Ordahl, 1994; Lin-Jones and Hauschka, 1996) and mouse data (Tajbakhsh and Buckingham, 1994; Ontell et al., 1995). In contrast, as soon as the Pax3-expressing myoblasts had reached their destination (Fig. 1A), defined as stage 20 (Chevalier et al., 1977; Christ et al., 1977), we were able to detect Myf5 transcripts (Fig. 1B), but no MyoD mRNA was observed (Fig. 1C). At that stage the ventral and dorsal muscle masses had not yet separated, as visualised by Pax3 expression (Fig. 1A). Using this in situ hybridisation technique, MyoD expression was first detected unambiguously at stage 22/23 (see Fig. 2F and Duprez et al., 1998).

Myf5 and MyoD show different expression domains during limb development

Whole-mount in situ hybridisation at stage 22/23 (E4) showed the Myf5 expression domain was larger than that of MyoD and closely resembled that of Pax3 (Fig. 2A-C). The Myf5 expression domain matched that of Pax3 throughout development (Fig. 1, Fig. 2 and data not shown). MyoD transcripts (Fig. 2F,G) were located in a subregion of the ventral and dorsal muscle masses expressing Myf5 mRNA (Fig. 2D,E). The Myf5 expression domain extended to near the ectoderm, while MyoD mRNA was located more centrally within the limb. In order to understand whether Myf5 and MyoD transcripts were located in the same cells where their expression domains overlapped, we performed double in situ hybridisation at stage 23. These experiments on transverse wing sections confirmed that the expression domain of MyoD mRNA (orange) was more restricted than and contained within that of Myf5 (purple) (Fig. 2H,I). All three possible expression combinations were observed (Fig. 2I): (1) Myf5⁺/MyoD⁻ cells (black arrows) were preferentially located near the ectoderm, (2) Myf5⁻/MyoD⁺ cells were preferentially found near the centre of the limb (black arrowheads) and (3) cells expressing both genes were found at the interface between these regions (white arrowhead). After stage 23, the MyoD expression domain spread to include all muscle masses, while Myf5 and Pax3 transcripts were progressively downregulated (see Figs 6, 8 and Duprez et al., 1998).

The Myf5 mRNA expression domain appeared identical to that of Pax3 (Fig. 1A,B; Fig. 2A,B). The similarity in the expression patterns of Pax3 and Myf5 transcripts suggested that their expression might be linked to the proliferative state of the cell. Since Pax3 expression has already been linked with proliferation (Epstein et al., 1995; Amthor et al., 1998), we studied the proliferative/differentiation state of Myf5-positive cells by performing BrdU incorporation experiments at stage 23 (see Materials and Methods). These experiments showed that some Myf5-expressing cells had indeed incorporated BrdU, similar to Pax3-expressing cells (Fig. 3A-D, arrows). In contrast, most of the cells expressing MyoD transcripts did not incorporate BrdU (Fig. 3E,F, arrowheads).

One known mechanism involved in differentiation processes in many systems and species is the Notch pathway (Artavanis-Tsakonas et al., 1999). We set out to investigate whether this signalling pathway is involved in myogenesis. We first analysed the endogenous cellular expression pattern of the Notch pathway components. From the literature it appeared that the Notch receptor might be ubiquitously expressed, the specificity of its action being determined by its ligands (Artavanis-Tsakonas et al., 1999). Indeed, whole-mount in situ hybridisation showed that the ligands Delta1 (Fig. 4A) and Serrate2 (Fig. 4B) were located in the muscle areas of the chick limbs at E5, while Notch1 transcripts were more uniformly distributed (Fig. 4C; Vargesson et al., 1998; Beckers et al., 1999). BrdU incorporation experiments indicated that high levels of the ligands Delta1 (Fig. 4G, arrows) and Serrate2 (Fig. 4D,E, arrows) were expressed in scattered cells that did not incorporate BrdU within the muscle masses. At E7, the ligand Serrate2 was only detected in MF20-positive cells (Fig. 4H). Delta1 was weakly detected at that stage (data not shown). The expression domain of the receptor Notch1 was larger than that of the ligand Serrate2 (compare the adjacent sections of Fig. 4E and 4F), suggesting that Notch1 was also expressed in non-myogenic cells. We could detect Notch1 mRNA in BrdU-positive cells (Fig. 4F, arrowheads), although most of the Notch1-positive cells were BrdU negative (Fig. 4F). In addition, Notch1 transcripts were clearly detected in mononucleated cells around the muscle fibres (Fig. 4I). These results indicate that the receptor Notch1 is expressed in immature myoblasts, while the cells expressing the ligands Delta1/Serrate2 are more advanced in myogenesis (postmitotic myoblasts and muscle fibres).

In order to understand the role of Notch signalling during the different steps of myogenesis, we activated the Notch pathway by over-expressing Delta1 using the RCAS retrovirus. The Delta/RCAS construct has been shown to be effective in retinal (Henrique et al., 1997), cartilage (Crowe et al., 1999), feather bud (Crowe et al., 1998; Viallet et al., 1998) and scale (Crowe and Niswander, 1998) formation. Based on described functions of Notch signalling in different systems and species, we hypothesised that constitutive activation of Notch signalling in muscle cells would lead to an inhibition of terminal muscle differentiation. Aggregates of Delta/RCAS-transfected cells (see methods) were grafted to stage 16 (E2.5) wing buds. In situ hybridisation of Delta1 transcripts in whole mounts showed the degree of virus spread 48 hours (Fig. 5A,B; n=5 out of 6) and 72 hours (Fig. 5C,D; n=2 out of 2) after grafting. In order to visualise the activation of Notch signalling, we looked for Notch1 expression after grafting, since it has been shown that activation of Notch signalling enhances Notch expression (Lewis, 1996). Overexpression of Delta1 in the limb bud led to an extension of the Notch1 expression domain (Fig. 5I,J; n=4 out of 5), reflecting an activation of Notch signalling (Micchelli et al., 1997; Franklin et al., 1999). In such grafted embryos, the MyoD expression domain appeared reduced in the region where ectopic Delta1 was detected (Fig. 5B,D), compared with the control wing, after 48 hours (Fig. 5E,F; Table 1) and 72 hours (Fig. 5G,H; Table 1). In contrast, the Myf5 (Fig. 5K,L; Table 1) and Pax3 (Table 1) expression domains were unchanged compared to the control wing. This demonstrates that the Notch pathway operates between the Myf5/Pax3-expressing and the MyoD-expressing stages.

Seventy-two hours after the Delta/RCAS-expressing cells were grafted into the wing bud, the control and operated wings were cut transversely through the forelimb region and hybridised with the Delta1 probe, revealing the extent of the spread of the virus (Fig. 6A,B,I,J). Adjacent sections hybridised with MyoD (Fig. 6C,D,K,L), Myf5 (Fig. 6E,F,M,N) and Pax3 (Fig. 6G,H,O,P) probes showed the normal and modified muscle pattern in the control and operated wings, respectively. The Delta/RCAS-infected right wing exhibited a downregulation of MyoD transcripts (compare Fig. 6C,D with 6K,L), while Myf5 and Pax3 expression appeared unaffected (compare Fig. 6E-H with 6M-P). Analysis of myosin expression using the MF20 antibody showed that there was a clear diminution in the number of differentiated muscle cells in the treated wing (Fig. 6O,P), compared with the control limb (Fig. 6G,H, in brown). This demonstrated that terminal differentiation is affected despite the presence of Myf5 and Pax3 transcripts.

We then examined the phenotype of the Delta/RCAS-infected limbs at E9.5 (7 days after grafting; n=4). Transverse sections were cut at the same level along the proximo-distal axis from the control (Fig. 7A,B) and manipulated (Fig. 7C,D) wings. In the limb shown in Fig. 7, only the posterior muscles were affected (Fig. 7D), although all muscles could be affected (data not shown). Examination of MyoD and myosin expression showed that the posterior muscles were reduced in size and disorganised (Fig. 7D), compared with the normal pattern (Fig. 7B). Moreover, the FDP (flexor digitorum profondus) muscle was absent in the Delta1-infected wing (Fig. 7D) but was clearly present in the control wing (Fig. 7B). Only certain muscles were affected (Fig. 7D) in this way by Delta1/RCAS infection despite the broad distribution of ectopic Delta1 transcripts (Fig. 7C). We interpreted this result as Delta-activated Notch acting only within a specific time window between the Myf5 and MyoD expression steps (see above). This time window is probably brief, so ectopic Delta1 would have to infect myoblasts at a very specific time in order to affect their further differentiation. High magnifications of the Delta1-infected muscle FCU (flexor carpi unlaris) and control FCU (Fig. 8) confirmed the absence of myosin in the MyoD-negative region (Fig. 8C,D) of the infected FCU, while myosin expression could be detected in the MyoD-positive region (Fig. 8D; arrow). In the control muscles, we could detect the MyoD mRNAs (purple) and myosin (brown) in most of the cells (Fig. 8A,B).

In situ studies of Myf5 and MyoD expression during development in the chick have been apparently contradictory (see Introduction). We have shown that Myf5 transcripts can be unambiguously detected before those of MyoD in the chick limb bud. A similar situation occurs in the chick somite (Hirsinger et al., 2001). These results are consistent with those obtained in mice. The expression domains of Myf5 and Pax3 essentially overlap during development. In contrast, MyoD mRNAs initially show a more restricted pattern, in the centre of the limb. During subsequent limb development, the MyoD expression domain spreads to include all muscle masses, while Pax3 and Myf5 are downregulated, although a low level of Pax3 transcripts can still be detected quite late (E10) (Duprez et al., 1998). Despite the absence of clear segregation between proliferative and postmitotic myoblasts in the limb, we observe a gradient of maturation from the ectoderm, where Myf5- and Pax3-expressing myoblasts proliferate (less differentiated), to the centre of the muscle masses, where myoblasts express MyoD and become postmitotic (more differentiated). There is probably an intermediate phase where the cells are Myf5⁺/MyoD⁺. This sequence of myogenic factor expression is shown in Fig. 9A.

In mice, Myf5 and MyoD have been described as being activated in a mutually exclusive manner in the musculature, Myf5 transcripts being first detected in the epaxial myotome and MyoD in the hypaxial myotome (Braun and Arnold, 1996; Cossu et al., 1996a, Tajbakhsh and Buckingham, 1999). Nevertheless, both genes are later co-expressed in the majority of cells with myogenic potential both in vivo and in vitro (Cossu et al., 1996b; Tajbakhsh and Buckingham, 1999). Moreover, Myf5-deficient embryos exhibit a 2-day delay in development of axial muscles, but normal formation of the limb musculature. Conversely, in MyoD mutant embryos, there is delayed limb muscle development and normal axial musculature formation (Kablar et al., 1997). This has been interpreted, in mice, as showing that Myf5 has a primary function in the regulation of axial muscles whereas MyoD is involved in limb muscle formation. Our chick expression data (see Results) provide no evidence for this dichotomy of function (Myf5/axial muscles versus MyoD/limb muscles). Instead, our results suggest involvement at different steps during myogenesis, Myf5 acting before MyoD. This has been already suggested by gene targeting analysis: in the absence of Myf5 and Pax3, mice do not express MyoD and fail to develop body skeletal muscles (Tajbakhsh et al., 1997). Moreover, it appears that Myf5 is activated first in both epaxial and hypaxial domains of mouse somites (Tajbakhsh and Buckingham, 1999). In addition, mouse Myf5 expression matches the main sources of myotomal precursors (Venters et al., 1999). These findings indicate that Myf5 initiates the body skeletal muscle differentiation program in both chick and mouse. The absence of muscle phenotype in the Myf5 knockout mice (Braun et al., 1992a,b) could be explained by Pax3 replacing the absent Myf5 and activating MyoD. Indeed, Pax3 appears to be sufficient, in some cellular contexts, to activate MyoD expression and thus initiate the myogenic program in vitro (Maroto et al., 1997) and in vivo (Tajbakhsh et al., 1997; Bendall et al., 1999). Alternatively, paraxis could be another candidate to assume the role of Myf5, since the double mutation paraxis⁻/Myf5⁻ shows muscle losses not observed in the single mutations (Wilson-Rawls et al., 1999a). The presence of muscle in the absence of MyoD (Rudnicki et al., 1992) has been interpreted as showing that it can be replaced by Myf5 (Rudnicki et al., 1993; Tajbakhsh and Cossu, 1997). However, Myf5 alone is insufficient to activate the myogenic program in the absence of the other three myogenic factors (Valdez et al., 2000), suggesting rather an overlap in the functions of myogenin, MyoD and Mrf4. It has already been shown that Mrf4 and MyoD can compensate for each other’s absence in muscle differentiation in mice, since the Mrf4⁻/MyoD⁻ double mutant displays a severe muscle deficiency, whereas mice lacking either Mrf4 or MyoD do not show defects in muscle development (Rawls et al., 1998). The absence of axial muscle defects in MyoD⁻ mice could be explained if Mrf4 were able to support muscle development. The transient expression of Mrf4 in myotome before its expression in late embryogenesis and postnatal muscles (Ontell et al., 1995) is consistent with this notion.

We have shown that Delta-activated Notch signalling in vivo downregulates MyoD expression and then inhibits terminal differentiation in the chick limb bud. This is the first demonstration in vivo of the involvement of Notch signalling in chick limb myogenesis. The cell-surface receptor Notch mediates communication between cells expressing Notch and cells expressing membrane-bound ligands such as Delta1 and Serrate2. Our examination of Notch1 and Delta1/Serrate2 expression shows that high levels of the ligands are detected in postmitotic cells and muscle fibres but that Notch1 is associated with mononucleated cells surrounding the fibres. These results are illustrated in Fig. 9B. Our overexpression data coupled with the in situ analysis can be interpreted as Notch signalling playing a role in maintaining the myoblasts in an undifferentiated state until myoblasts are correctly positioned to pursue their differentiation. This result is consistent with the known functions of Notch in other systems, e.g. retina, where progenitor retinal cells exposed to Delta1 are prevented from undergoing neuronal differentiation (Henrique et al., 1997; Dorsky et al., 1995); cartilage, where misexpression of Delta1 blocks chondrocyte maturation (Crowe et al., 1999); and feather, where overexpression of Delta1 inhibits feather development (Crowe et al., 1998; Viallet et al., 1998). Moreover, during development of adult indirect flight muscles in Drosophila, Notch activation causes failure of differentiation (Anant et al., 1998).

The use of the dominant-negative form of Delta1 that blocks Notch signalling (Henrique et al., 1997) failed to give any muscle phenotype when grafted into the limb (data not shown). This means that definitive proof of a physiological role for Notch in limb myogenesis is still lacking. However, successful block of Notch signalling using the dominant-negative form of Delta1 has only been reported for one system, the retina (Henrique et al., 1997). An alternative explanation is that Serrate2, another Notch ligand that is expressed in differentiated myotubes (see Results), compensates for the lack of Delta1 activity. Serrate has been indeed shown to compensate for the loss of function of Delta in Drosophila (Gu et al., 1995). Moreover, overexpression of Serrate2 can inhibit the differentiation of C2C12 myoblast cells (Lindsell et al., 1995). Serrate2 is expressed at the right time and place to play a role in limb myogenesis. However, it is not clear whether Delta1 and Serrate2 have identical functions or whether they use the same receptors during limb myogenesis.

Our results show that neither Myf5 nor Pax3 is affected by ectopic activation of Notch signalling, indicating that Notch signalling acts after the Pax3/Myf5 step (Fig. 9A). In addition, our results show that Myf5 and Pax3 are insufficient to allow further muscle differentiation in the absence of MyoD. This contrasts with the normal muscle phenotype in MyoD knockout mice (Rudnicki et al., 1992; Kablar et al., 1997). Our misexpression experiments do not allow us to exclude the hypothesis that Delta1-activated Notch acts on the MyoD-expressing lineage in the chick limb, leaving intact the Myf5-expressing pathway. But in that case we would have expected to observe normal terminal differentiation in our experimental limbs, as in the MyoD knockout mice (Rudnicki et al., 1992; Kablar et al., 1997). The Myf5 (and Pax3) pathway is insufficient, in our experimental context, to rescue terminal muscle differentiation in the absence of MyoD, since we observe fewer myosin-positive cells in Delta1-infected limbs. This reinforces the idea that myogenic factor(s) other than Myf5 compensate for the absence of MyoD in MyoD^−/− mice (Rawls et al., 1998; Wilson-Rawls et al., 1999b; Valdez et al., 2000). Alternatively, there might be a genuine difference between chick and mouse.

It is not clear whether activated-Notch acts on the transition (1) between Myf5-proliferative and MyoD-postmitotic cells or (2) between Myf5-postmitotic to MyoD-postmitotic cells. By analogy with the situation in the retina, where forced expression of Delta1 maintains proliferating neuroepithelial precursors (Henrique et al., 1997), we would have expected an extension of the Pax3 and Myf5 domains concomitant with an increase of the BrdU incorporation. However, the fact that we do not observe any change in the Pax3 and Myf5 domains (Figs 5, 6) or of BrdU labelling (data not shown) after Delta misexpression favours an action of Notch on Myf5 postmitotic cells. Whatever the situation, it is clear that Delta-activated Notch blocks further differentiation of Myf5-expressing cells.

We found a decrease of MyoD transcripts after ectopic activation of Notch signalling. We cannot conclude from our experiments whether the downregulation of MyoD transcripts is the result of an inhibition of gene activation or a defect in the maintenance of MyoD expression. However, studies on transfected cell lines revealed that activated Notch is able to inhibit MyoD transcription (Kuroda et al., 1999). Activated Notch also interferes with the muscle-inducing activity of the MyoD protein (Kopan et al., 1994). This interference has recently been shown to occur through a direct protein interaction between the ankyrin repeat region of Notch and MEF2C, an essential cofactor of MyoD, that blocks DNA binding (Wilson-Rawls et al., 1999b). Since MyoD and MEF2 participate in regulatory circuits involving positive transcriptional feedback loops (Thayer et al., 1989; Braun et al., 1989; Molkentin et al., 1995), the downregulation of MyoD expression we observed could also be the consequence of the inhibition of MEF2 activity.

It has been shown recently that MyoD is a direct, positive regulator of Xenopus Delta1 (Wittenberger et al., 1999). The detection of MyoD transcripts before those of Delta1 in the Xenopus gastrula indicates that MyoD triggers Notch signalling in this species (Wittenberger et al., 1999). From our results it is not clear whether MyoD induces Delta1, which would then trigger Notch signalling in adjacent myoblasts, or if Notch signalling is activated before the onset of MyoD expression. Chick MyoD and Delta1 expression seem to occur together in the limb (data not shown). However, Serrate2 transcripts are detected before those of MyoD in the limb (data not shown). Thus, we favour the hypothesis that high levels of ligand (Delta1, Serrate2) expression in a few cells would activate Notch signalling in adjacent cells. The ligand-positive cells would then differentiate by activating MyoD. The existence of a positive feedback loop (MyoD towards Delta1) would enhance and lock the differentiation process.

In vitro studies have generally concluded that MyoD is expressed in proliferative myoblasts (Lassar et al., 1994; Molkentin and Olson 1996), although this does not fit with the endogenous expression of MyoD in postmitotic cells of mouse and chick myotomes (Ontell et al., 1995; Brand-Saberi et al., 1996; Amthor et al., 1998; Hirsinger et al., 2001). In the chick limb, where the distinction between proliferative and postmitotic myoblasts is not obvious (in contrast to the situation in somites), we have shown that the majority of MyoD-expressing cells do not incorporate BrdU, indicating that they have withdrawn from the cell cycle and are postmitotic myoblasts. Moreover, experimental evidence from in vitro studies shows that MyoD induces withdrawal from the cell cycle independently of muscle differentiation (Davies et al., 1987; Crescenzi et al., 1990; Sorrentino et al., 1990; Trouch et al., 1993). No such effects have been reported for Myf5. In addition, Myf5 and MyoD show different expression profiles during the cell cycle in C2 cells. MyoD expression is maximal upon cell cycle exit (Kitzmann et al., 1998). A recent study of satellite cells reached the same conclusion that MyoD could be necessary for the transition from proliferation to differentiation (Yablonka-Reuveni et al., 1999). Interestingly, MyoD is able to activate myogenin (Hollenberg et al., 1993) and a cyclin inhibitor, p21, without any new protein synthesis (Otten et al., 1997; Cenciarelli et al., 1999). p21 activation is linked to cell-cycle exit (reviewed in Walsh and Perlman, 1997).

In conclusion, in vitro studies and genetic analysis, coupled with our in situ and in vivo experiments on chick limb, indicate that Myf5 initiates the myogenic program and that MyoD expression is the manifestation of subsequent differentiation. The in vivo signal regulating the transition from the Myf5 step to the MyoD step in the chick limb may involve Notch signalling.

We thank Marie-Aimée Teillet and Heather Etchevers for critical reading of the manuscript. We are grateful to Francis Beaujean, Michel Fromaget and Sophie Gournet for the illustrations. This work was supported by the Association Française contre les Myopathies (AFM) and the Centre National de la Recherche Scientifique (CNRS).