SF/HGF is a mediator between limb patterning and muscle development

In the vertebrate limb, signalling centres regulating pattern formation in all three dimensions are well established. The apical ectodermal ridge (AER) is known to be crucial for proximodistal limb bud outgrowth and skeletal patterning (Saunders, 1948). AER signalling has been demonstrated to be mainly mediated by various fibroblast growth factors (FGFs) expressed in the ridge (reviewed in Martin, 1998). The zone of polarizing activity (ZPA), which is located in the posterior limb bud mesenchyme, has been shown to regulate the development of anteroposterior pattern in the limb (Saunders and Gasseling, 1968). ZPA activity has later been shown to be mediated by Sonic Hedgehog (SHH) signalling (Riddle et al., 1993). The dorsoventral axis in the limb is thought to be established by signals from the ectoderm, for example, Wnt7a (Parr and McMahon, 1995).

It has been shown that the precursor cells of skeletal limb muscles originate from the ventrolateral edges of the dermomyotomes (Christ et al., 1974, 1977; Chevallier et al., 1977). During limb bud outgrowth up to stage 25, they migrate actively from the base of the limb towards its tip in a directed fashion (Wachtler et al., 1982; Brand-Saberi and Christ, 1992). Therefore, some investigators favoured the AER as a candidate for direction control in myogenic cell migration (Gumpel-Pinot et al., 1984). However, other authors found evidence that there is no immediate signalling influence from the AER, but rather that properties of the stationary mesenchyme direct the migrating cells towards the tip of the limb bud (Brand-Saberi et al., 1989). Bladt et al. (1995) found that, in mice lacking the c-met receptor or its ligand, scatter factor/hepatocyte growth factor (SF/HGF), no muscle precursor cells enter the limb, and the musculature of limbs, distal tongue and diaphragma is absent. The protooncogene c-met (Cooper et al., 1984), which encodes a transmembrane tyrosine kinase, is expressed in epithelia that de-epithelialize after binding of the ligand, SF/HGF (Stoker et al., 1987), a secreted 90 kDa glycoprotein heterodimer structurally related to plasminogen (Nakamura et al., 1989). Moreover, c-met has been shown to be expressed in migrating myogenic precursor cells invading the lateral plate and the limb mesenchyme (Bladt et al., 1995; Yang et al., 1996). In the chick, SF/HGF is expressed in the limb fields of the lateral plate adjacent to the lateral dermomyotomes from which myogenic precursor cells detach, and also later in the developing limb bud mesenchyme (Myokai et al., 1995; Thery et al., 1995).

In this study, we demonstrate for the first time that SF/HGF expression in the chick limb is controlled by signals involved in limb bud patterning. In the absence of a functional AER, no expression of SF/HGF in the limb bud is observed. Excision of the ZPA results in ectopic and enhanced SF/HGF expression in the posterior limb bud mesenchyme. We further demonstrate that this latter result is paralleled by a shift of Pax-3-expressing cells towards the posterior limb bud mesenchyme, indicating a role of the ZPA in positioning of the premuscle masses. Moreover, we present evidence that, in the limb bud mesenchyme, SF/HGF increases the motility of myogenic precursor cells and has a role in maintaining their undifferentiated state during migration. We present a model for a crucial role of SF/HGF during migration and early patterning of muscle precursor cells in the vertebrate limb.

Fertilized eggs of Gallus gallus (White Leghorn) and Coturnix c. japonica were obtained from a local breeder and incubated at 38°C and 80% relative humidity for the time required. The stages of the embryos were determined according to Hamburger and Hamilton (1951). Limbless embryos were obtained from Drs Ursula Abbott and Jacqueline Pisenti, Davis CA.

Eggs were windowed and the vitelline membrane and the amnion slit open in the area of operation. The AER was excised at stages 18–22 using ophthalmologic scissors. For ZPA removal, the posterior third of stage 19–22 limb buds was excised also by using ophthalmologic scissors.

For application of FGF-2, BMP-2 and BMP-4, heparin-coated acrylic beads of approximately 80 μm in diameter (Sigma, Germany) were rinsed in PBS and individually transferred into protein solution (FGF-2: 50 μg/ml; BMP-2: 100 μg/ml; BMP-4: 2 or 20 μg/ml). Factors were diluted in PBS+0.1% BSA and applied on stage 19-21 embryos. FGF-2 was obtained from Pepro Tech, Rocky Hill NJ, BMP-2 and BMP-4 were obtained from Genetics Institute, Cambridge MA. Recombinant SF/HGF was produced in Sf9 insect using the baculovirus expression system followed by a one-step purification on heparin sepharose (Weidner et al., 1993). Beads were soaked in factor at 4°C for at least 1 hour prior to implantation. To implant a bead into a limb bud, a small slit was cut through the ectoderm into the dorsal mesenchyme using tungsten needles and the bead was pushed through the slit into the mesoderm. N-SHH was prepared as described (Cann et al., 1999) and applied using Ni-NTA-Agarose beads (Qiagen), which were washed in Tris-NaCl buffer (pH 8.5), individually transferred into factor (0.65 mg/ml), and soaked in factor at 4°C for at least 1 hour prior to implantation. Implantation procedure was carried out as described above.

The operated eggs were sealed with medical tape and reincubated for 18-24 hours. Embryos were inspected, killed and fixed in 4% paraformaldehyde overnight at 4°C, dehydrated in a graded methanol series and stored at −20°C.

For quail-chick transplantations, limb bud mesenchyme was taken from quail donors at stages 19–22 and implanted into limb buds of stage 21–24 chicken hosts, the host being always older than the donor. SF/HGF was injected into the host limb mesenchyme proximal and distal to the implant at a concentration of 100 μg/ml using a hand-drawn glass pipette. Control injections were carried out with Locke saline solution or 100 μg/ml BSA in PBS. Embryos were reincubated for 4 to 6 days, killed and treated for immunohistochemistry as described below. Migration of myogenic quail cells was diagnosed when a substantial number of quail nuclei was detected in myotubes at a distance of more than 300 μm to the transplant, without being continuous with stationary graft tissue. Cells were considered not to have migrated when they were continuous with the graft or less than 300 μm away from stationary graft tissue.

Embryos destined for MyoD expression analysis were operated at stages 20 or 21. SF/HGF was injected into the limb bud mesenchyme at a concentration of 100 μg/ml, PBS was injected as control. Specimens were reincubated for 24 hours, fixed and processed for in situ hybridization as described below.

Embryos were rehydrated in a graded methanol series and processed as described by Nieto et al. (1996). Visualization of the hybridization product was achieved by use of the digoxigenin RNA labelling and detection kit by Boehringer (Germany) according to the recommendations of the supplier. The following probes were used in this study: avian SF/HGF (kindly provided by Dr Claudio Stern, see Thery et al., 1995), a 2.3 kb insert cloned into pBluescriptSK⁻, quail Pax-3 (kindly provided by Dr Christophe Marcelle and Dr Michael Stark, San Diego CA), a 1543 bp insert cloned into pBluescriptSK⁺, and chick MyoD (kindly provided by Dr Bruce M. Paterson, Bethesda MYL), a 1518 bp insert cloned into pBluescriptKS⁺.

Specimens destined for anti-quail labelling in sections were processed as described (Zhi et al., 1996).

Although in earlier studies (Myokai et al., 1995; Thery et al., 1995) the expression pattern of SF/HGF in the chick embryo has been described in some detail, we reexamined SF/HGF expression in the limb bud by whole-mount in situ hybridization from stage 18 to 26 to elucidate the possible role of SF/HGF in the migration control of myogenic precursor cells (Fig. 1).

At stage 18, SF/HGF is expressed throughout the limb bud mesenchyme. At stage 19, expression is slightly decreased in the posteriormost limb bud mesenchyme (Fig. 1A). The anterior bias of SF/HGF expression becomes more conspicuous at stages 20-22 (Fig. 1B,C). At stage 22, expression decreases further in the proximal limb mesenchyme. At stage 23, SF/HGF transcripts are restricted to the anterodistal limb mesenchyme and a narrow band of fainter expression extending from the proximal to the distal margin of the limb in the central limb mesenchyme (Fig. 1D). At stage 24, expression could only be detected in the anterodistal tip of the limb (Fig. 1E), where it persists up to stage 25 at low level. From stage 26 on expression was no longer detected.

SF/HGF stimulates directed myogenic cell migration Myogenic precursor cells in the limb bud mesenchyme are known to migrate in proximodistal direction towards the tip of the limb bud (Wachtler et al., 1982). SF/HGF is expressed in the limb bud mesenchyme up to stage 25, and is known to have mitogenic, motogenic and morphogenic properties in vitro and in vivo, thereby acting in a paracrine fashion (reviewed in Matsumoto and Nakamura, 1996). To test for a possible influence of SF/HGF on the migratory behaviour of myogenic cells in the limb bud, we transplanted blocks of stage 19-22 quail wing bud mesenchyme containing myogenic cells into chick host wing buds being at least one stage more advanced. Proximally and distally to the transplant, we also injected SF/HGF into the adjacent host mesenchyme and determined the location of quail cells in the host mesenchyme after 4 to 6 days of reincubation (Fig. 2). In 86% (n=21) of the cases examined, quail nuclei were found in myotubes of host muscles at considerable distance from the transplant, from 300 μm up to more than 2 mm (Fig. 2B,C). In controls after injection of BSA or Locke solution, quail cells were largely confined to the area of transplantation (75% not migrated, n=12), i.e. those counted as not migrated were either in direct contact with stationary graft tissue (cartilage, fibroblasts) or emigrated less than 300 μm from this tissue. In similar transplantation experiments without application of factors, it was shown that donor cells never migrate into older host tissue (Brand-Saberi et al., 1989). Here, we could confirm this as in all controls without injection no migration was observed (100% not migrated, n=4). At fixation, in SF/HGF-injected embryos as in controls, the transplant was situated in most cases in the cubital or distal humeral region of the wing. After SF/HGF application quail cells were detected distal to the transplant in the muscles of the zeugopod and autopod (Fig. 2A), as well as in the proximal stylopod (Fig. 2D) and even in the pectoral muscle of the trunk (not shown).

These data suggest that after SF/HGF application, myogenic cells in the limb bud are motilized and migrate in a retrograd or prolonged fashion, depending on the presence of SF/HGF in the surrounding stationary mesenchyme. However, it should be noted that, after reincubation beyond stage 25, which is necessary for myotubes to form, the long-range translocation of myogenic quail cells in our experiments does not result from active migration but from interstitial growth in the developing wing. Nevertheless, our results indicate that the localized presence of SF/HGF in the stationary mesenchyme of the limb bud provides guiding cues to the migratory muscle precursor cells in a paracrine fashion.

In vitro studies showed that, in fibroblasts, motile and invasive cell behaviour can be induced by SF/HGF (Giordano et al., 1993; Dugina et al., 1995). To test if SF/HGF is involved in the maintenance of the undifferentiated, motile state of myogenic precursor cells in the chick limb, we monitored the expression of a marker gene for differentiating muscle cells, MyoD (Davis et al., 1987; Weintraub et al., 1989), after application of SF/HGF to the limb bud. We injected SF/HGF into the dorsal mesenchyme of stage 20 or 21 limb buds, where the dorsal premuscle mass is about to form. After 24 hours reincubation, MyoD expression in the injected limb buds was severely reduced when compared to the untreated, contralateral limb bud (94%, n=16) (Fig. 3). This indicates that, in the presence of SF/HGF, myogenic cell differentiation is impeded. Thus, we conclude that SF/HGF is able to prevent premature differentiation of myogenic cells in the limb bud, which is a prerequisite for continuing migration.

To reexamine the question whether the AER has a role in myogenic cell migration in the developing limb, we tested for a relationship between AER signalling and SF/HGF expression by monitoring SF/HGF expression after microsurgical AER excision (Fig. 4A-E). At all stages examined, SF/HGF expression in AER excised limb buds was severely reduced or absent (n=14; Fig. 4B). To verify these results in a different approach, we looked at SF/HGF expression in the chick mutant, limbless (Prahlad et al., 1979). limbless embryos stop limb bud outgrowth shortly after its initiation because the ectoderm is not competent to form an AER in response to inducing signals from the mesenchyme (Fallon et al., 1983; Carrington and Fallon, 1988). Limbless mutants are therefore an apt system to investigate processes in the absence of a functional AER, without invasive manipulations (Noramly et al., 1996). As expected, we could not find SF/HGF expression in the rudimentary limb buds of limbless embryos from stage 20 onwards (n=12; Fig. 4C). However, weak SF/HGF expression was detected in early stages up to stage 18/19, before the newly formed AER starts to promote limb bud outgrowth (not shown). Thus, the AER is necessary for maintenance of SF/HGF expression in the limb bud mesenchyme, but not for early SF/HGF expression in the limb field. It has been shown recently that FGF can ectopically induce SF/HGF expression in the lateral plate mesoderm (Heymann et al., 1996). As the AER is the major source of FGF in the growing limb bud (reviewed in Martin, 1998), we speculated that the AER can induce SF/HGF expression in the limb bud mesenchyme via FGF signalling. To test this hypothesis, we implanted FGF-2-soaked beads into the distal mesenchyme of limb buds after AER excision and tested the embryos for SF/HGF expression. We observed a clear rescue of SF/HGF expression in the vicinity of the FGF beads (n=10; Fig. 4D), but not of control beads soaked in PBS (n=4; Fig. 4E). Thus, we conclude that, in the chick limb bud, the AER induces SF/HGF expression via FGF signalling.

We next examined the influence of the AER on the distribution of myogenic cells in the limb bud. Therefore, we hybridized specimens with a probe specific for Pax-3, which is a marker for undifferentiated muscle precursors in the limb (Bober et al., 1994; Goulding et al., 1994). After AER removal, Pax-3 was still detectable in the limb mesenchyme, though in a slightly aberrant pattern (n=3; Fig. 4H). This is not contradictory to the above results, as myogenic cells may well have entered the limb bud up to the time of operation, which was in this series stage 19 or stage 21. Moreover, the short reincubation time of 24 hours might allow for some remaining SF/HGF activity in the mesenchyme in the absence of an AER. In the limbless mutants, however, the rudimentary forelimb buds were devoid of any Pax-3 expression (stages 20–24) while, in the hindlimb buds, weak Pax-3 expression was in some cases still detectable (n=9; Fig. 4F). This might be due to the delayed onset of AER degeneration in hindlimb buds (Noramly et al., 1996). However, in the somites adjacent to the limb buds, Pax-3 expression in the lateral tips of the dermomyotomes had vanished, similar to the wild-type situation (Fig. 4G). In wild-type embryos, these Pax-3-expressing cells detach from the dermomyotomes and invade the limb buds. As, in limbless embryos, no Pax-3-positive cells were seen in the forelimb buds, it is likely that these cells had been eliminated by apoptosis after having detached from the somites. Apoptosis is known to stop further limb outgrowth in mutant limb buds from stage 19 onwards (Fallon et al., 1983). Taken together, we observe persistent Pax-3 expression in AER-excised limb buds even though SF/HGF expression is absent whereas, in limbless embryos, from stage 20 onwards, neither Pax-3 nor SF/HGF transcripts can be detected in the limb rudiments.

To investigate a potential role of the ZPA in myogenic cell migration and muscle pattern formation, we analyzed SF/HGF expression after microsurgical ZPA removal by in situ hybridization using a probe specific for SF/HGF (Fig. 5). At all stages examined, after ZPA removal, we observed SF/HGF expression not only in the anterodistal domain corresponding to the normal expression pattern, but also in an additional domain along the posterior margin of the operated limb (n=30; Fig. 5A), which clearly exceeded the central band of fainter expression. Moreover, in situ hybridizations gave an overall more intense signal in the operated limb compared to the contralateral side, indicating that ZPA removal allows for enhanced SF/HGF expression in the limb bud mesenchyme. To test whether the altered SF/HGF expression pattern after ZPA removal is paralleled by an altered distribution of myogenic precursor cells in the limb, we hybridized ZPA-excised embryos with a probe specific for Pax-3. Due to the loss of tissue by ZPA removal, the Pax-3 expression domain in operated limbs was generally smaller than in control limbs. Interestingly, however, we observed a shift of the Pax-3 expression domain, which represents the premuscle masses, towards the posterior margin of the operated limbs (n=9; Fig. 5B). The anterior portion, which was devoid of Pax-3-positive cells, was larger in ZPA-excised embryos. Taken together, after ZPA removal, we observe ectopic SF/HGF expression in the posterior limb mesenchyme as well as a posterior shift of muscle precursor cells. From this we conclude that the SF/HGF expression domain is regulated by signals from the ZPA and is involved in directing myogenic cell migration in the chick limb bud, leading to the establishment of the positions of the dorsal and ventral premuscle masses.

The finding that, in the absence of the ZPA, SF/HGF is expressed in the posterior limb bud mesenchyme, in contrast to its anterodistally restricted expression in the normal limb, suggests that, in the normal limb, the ZPA represses SF/HGF expression in the posterior mesenchyme. Therefore, we tested proteins known to be synthesized by cells of the ZPA for a potential inhibitory influence on SF/HGF expression. As SHH is considered to be the polarizing signal in vivo (Riddle et al., 1993), we were interested in whether SHH can repress SF/HGF expression in the limb bud. To test this, we implanted beads soaked in SHH into the anterior limb bud mesenchyme and tested the embryos for SF/HGF expression after 24 hours of reincubation. Surprisingly, we did not find downregulation of SF/HGF transcription in the vicinity of the beads but, on the contrary, upregulation of SF/HGF transcription in the anterior marginal mesenchyme of the operated limb, some distance from the bead (n=14; Fig. 5C). The activity of the protein was evident from the occurrence of ectopic outgrowth in the anterior limb bud mesenchyme and from in vitro control experiments (not shown). This leads us to assume that SHH is not a negative regulator of SF/HGF in the posterior limb mesenchyme. In a second approach, we repeated the above experiment using beads soaked in BMP-2, which is also expressed in the ZPA (Francis et al., 1994). Here, we observed downregulation of SF/HGF expression in the vicinity of the bead, indicating that BMP-2 can inhibit SF/HGF expression (n=6; Fig. 5D). In contrast, BMP-4-soaked beads were incapable of repressing SF/HGF expression in the limb bud, but resulted in an overall decrease of limb bud size probably due to apoptosis (n=5; Fig. 5E). Therefore, BMP-2 is a good candidate for the putative repressor of SF/HGF expression in the posterior limb bud mesenchyme.

In this study, we analyzed the influence of SF/HGF on the behaviour of myogenic precursor cells in the chick limb bud and monitored the effect of signals from the main signalling centres in the limb, AER and ZPA, on SF/HGF expression. We found that SF/HGF keeps the migrating myogenic precursor cells in an undifferentiated state and increases their motility. Moreover, our data indicate that SF/HGF has an influence on the direction of myogenic cell migration within the limb bud mesenchyme. We present evidence that SF/HGF expression in the limb bud depends on FGF signalling from the AER, and that inhibitory signals from the ZPA restrict SF/HGF expression to the anterodistal mesenchyme of growing chick limb buds (Fig. 6).

SF/HGF has long been known to de-epithelialize and motilize epithelial cells in vitro (Stoker et al., 1987) and, thereby, to be involved in invasive carcinoma growth (Weidner et al., 1990). Nevertheless, SF/HGF is also involved in various developmental processes implying epitheliomesenchymal transitions, and has trophic functions for regenerating cells in liver, kidney and lung (reviewed in Matsumoto and Nakamura, 1996). SF/HGF is expressed in mesenchymal cells and is secreted into the extracellular matrix, where it functions in a paracrine fashion to activate the c-met transmembrane tyrosine kinase, which is expressed in epithelial cells (reviewed in Birchmeier and Birchmeier, 1993). In vitro studies showed that SF/HGF-mediated phosphorylation of c-met and focal adhesion kinase triggers the dynamic formation of focal adhesion points, cell spreading, subsequent disruption of cell-cell contacts and cell crawling (Matsumoto et al., 1994).

Recent work suggests an important role of SF/HGF in the recruitment and emigration of myogenic precursor cells from the lateral dermomyotomes into the limb buds (Bladt et al., 1995; Brand-Saberi et al., 1996; Dietrich et al., 1999). Embryological experiments indicate that these directed cell movements require signals from the proximal limb bud mesenchyme (Christ et al., 1978).

Several lines of evidence indicate that SF/HGF is this postulated inducing signal. In the mouse, c-met is strongly expressed in the ventrolateral edges of the dermomyotomes, and also in cells that migrate from the somites to the limbs. Concomitantly, around the time of emigration of myogenic precursor cells, SF/HGF is expressed in the medial limb field mesenchyme directly abutting the lateral dermomyotomes (Bladt et al., 1995). Similar expression data are known from the chick (Thery et al., 1995; Myokai et al., 1995, own data). Interestingly, in SF/HGF homozygous mutant mice (Schmidt et al., 1995), as well as in c-met homozygous mutants (Bladt et al., 1995), the lateral dermomyotomes fail to de-epithelialize and no emigration of myogenic cells occurs, leaving the limb buds devoid of skeletal muscle (Dietrich et al., 1999). In previous experiments, we could demonstrate that injection of SF/HGF protein into the somatopleure of chick embryos at interlimb level induces detachment and emigration of myogenic cells from the dermomyotomes (Brand-Saberi et al., 1996).

Thus, the de-epithelializing and motogenic properties of SF/HGF in myogenic cell migration are well established. Yet, it has been unclear what might be the function of the persisting expression of SF/HGF in the limb bud mesenchyme during later stages of limb bud outgrowth up to stage 25.

In this study, we tested for a possible role of SF/HGF in directing myogenic cells during their later migration in the mesenchyme of the outgrowing limb bud. We showed that SF/HGF application can strongly increase myogenic cell migration and that ectopic SF/HGF is able to reinitiate and reverse migration in the limb bud. We conclude that SF/HGF mobilizes myogenic cells also in normal development, and is involved in directing them towards their target tissue.

However, Heymann et al. (1996) did not observe a preferential migration of myogenic cells towards beads soaked in SF/HGF and implanted into the flank of chick embryos. From this, they excluded a chemotactic influence of SF/HGF on myogenic cells, which was suggested by Takayama et al. (1996) who found muscle cells in the central nervous system of transgenic mice after forced expression of SF/HGF in the neural tube. Our results do not imply a direct guiding property of a SF/HGF source, but rather suggest that SF/HGF secreted by the stationary mesenchymal cells allows for the maintenance of motility in myogenic cells. We postulate that the dynamic expression domain of SF/HGF in the limb bud, moving from the proximal mesenchyme in younger buds to the distal mesenchyme in older buds, reflects the domains of active myogenic cell migration as maintained by the SF/HGF-mediated c-met signalling pathway. Thus, the expression control of SF/HGF represents an actual directing influence on myogenic precursor cells in the limb bud. This is in line with the observation that migrating myogenic precursor cells express c-met and Pax-3, the latter being considered to be involved in c-met expression control (Bladt et al., 1995; Epstein et al., 1996; Daston et al., 1996). Moreover, our results are supported by the findings of Dietrich et al. (1999) who found evidence that, in the mouse, SF/HGF expression coincides with the routes of migrating muscle precursor cells, as identified by Lbx 1 expression (Dietrich et al., 1998), and by the observation of Brand-Saberi and Christ (1992) that myogenic cell migration ceases after stage 25, which is the latest stage showing SF/HGF expression. Interestingly, Zhi et al. (1996) found that muscle precursor cells detaching from the most caudal somite adjacent to the wing region (somite 21) sometimes do not take part in wing muscle formation, or they remain in the proximal mesenchyme and are restricted to proximal muscle blastemas. This may reflect the absence of SF/HGF from the posterior limb bud mesenchyme from stage 19 onwards. Consistent with our model, we found that after SF/HGF injection into the limb bud mesenchyme, the expression of MyoD, which is a marker for differentiating and stationary myoblasts, is downregulated, suggesting that SF/HGF keeps myogenic cells in an undifferentiated, motile state.

Taken together, we conclude from these results that the expression of SF/HGF in the limb bud mesenchyme is necessary for myogenic cell migration. We hypothesize that the dynamic SF/HGF expression domain shifting from proximal-to-distal positions during limb bud outgrowth indirectly guides the myogenic cells migrating in its wake towards their distally located destinations.

In the limb bud mesenchyme, SF/HGF is expressed in a spatially restricted and temporally dynamic pattern. Therefore it is likely that SF/HGF expression in the limb bud is controlled by a precise regulatory mechanism, coordinating the expression pattern in time and space. So far, however, the postulated regulatory mechanism is essentially unknown.

Considering that SF/HGF expression in the limb bud is, at all stages examined, restricted to the mesenchyme subjacent to the AER, we monitored SF/HGF expression after AER removal. We observed that, in the absence of the AER, SF/HGF transcripts cannot be detected in the operated limb. Consistent with this, in limbless mutant chick embryos, which lack a functional AER (Prahlad et al., 1979; Fallon et al., 1983), the limb rudiments are also devoid of SF/HGF expression, except in early stages before the ridge has formed. These results demonstrate that SF/HGF expression in the limb bud mesenchyme requires signals from the AER. We show that FGF-2 bead application to AER excised limb buds rescues SF/HGF expression in the vicinity of the bead. This is in line with the findings of Heymann et al. (1996), who observed ectopic SF/HGF expression in the lateral plate mesoderm after FGF-2 bead application to the flank of chick embryos. Thus, our data suggest that the AER maintains SF/HGF expression in the limb bud mesenchyme via FGF signalling. Although we could induce SF/HGF expression in the limb bud mesenchyme by application of FGF-2 protein, we cannot rule out the possibility that other FGFs, namely FGF-8 which is expressed, as is FGF-2, throughout the AER, are the natural inducers of SF/HGF in the limb bud. This is not easy to determine because FGF functions are highly redundant (reviewed in Martin, 1998).

We next examined whether the anterior bias of SF/HGF expression in the limb bud from stage 19 onwards is due to signals from the ZPA. We found that, after excision of the ZPA, SF/HGF expression in the limb bud is generally more intense and that its expression domain is extended towards the posterior margin of the limb bud mesenchyme. This indicates that, in the normal limb bud, SF/HGF expression is restricted to the anterior portion by inhibitory signals from the posterior limb bud mesenchyme. We next tested whether factors expressed in the posterior limb bud mesenchyme can influence SF/HGF expression. Surprisingly, we found that anterior application of SHH enhanced SF/HGF expression at the anterior margin of the limb bud mesenchyme, whereas BMP-2 was observed to downregulate SF/HGF expression in the vicinity of the bead. This indicates that BMP-2, but not SHH, inhibits SF/HGF expression in the posterior limb bud mesenchyme. These results seem to be hard to reconcile, as SHH is thought to induce BMP-2 expression in the limb mesenchyme, BMP-2 thus being a putative downstream target of the SHH signalling pathway. It should be noted that, after SHH bead application, SF/HGF was upregulated exclusively in the anterior marginal limb bud mesenchyme, but not in the immediate vicinity of the more proximomedially located bead. Laufer et al. (1994) observed that SHH can induce BMP-2 expression only when applied close to the AER, and not when applied to proximomedial positions in the limb bud. They found, moreover, that FGF-4 from the AER is required to induce the competence of the mesoderm to respond to SHH, and that shh and FGF-4 expression are linked via a positive feedback loop. Therefore, an explanation for the above results could be that SHH released from the implanted bead induced FGF-4 expression in the anterior marginal ectoderm, which in turn induced upregulation of SF/HGF in the underlying mesenchyme. BMP-2 application, however, could exert its inhibitory function without being counteracted against by FGF signalling from the ectoderm. Yet, this model still does not explain the conflicting functions of SHH and BMP-2 which are, in normal limb buds, coexpressed in the posterior limb bud mesenchyme. Thus, these data remain ambiguous and will need further investigation.

Taken together, our results suggest that the AER maintains SF/HGF expression in the distal limb bud mesenchyme via FGF signalling, and that the anterior bias of the SF/HGF expression domain is due to an inhibitory influence in the posterior mesenchyme, probably mediated by BMP-2.

We showed that SF/HGF is required for motilizing and directing myogenic cell migration in the limb bud, and that its expression is controlled by the AER and the ZPA.

In limbless embryos, which never form a functional AER (Fallon et al., 1983), we found that muscle precursor cells detach from the lateral tips of the dermomyotomes at limb level, similar to the wild-type situation. However, in the wing bud rudiments of limbless specimens at stage 20 or later stages, no Pax-3-positive cells were found. Thus, obviously, the limbless gene product is not involved in the recruitment of myogenic precursor cells from the somite.

From our results, we speculate that myogenic cell migration is a two-step process. In a first step, myogenic precursor cells de-epithelialize, detach from the lateral dermomyotomes and populate the limb field mesenchyme. In a second step, during limb bud outgrowth, these cells migrate actively towards their distal destinations in the limb bud mesenchyme. SF/HGF is involved in both steps, but with different roles. In the early phase, it is involved in the epitheliomesenchymal transition of the myogenic precursor cells and their subsequent motilization. In the later phase, it is needed to keep the cells undifferentiated and motile and guide them during their migration within the limb bud mesenchyme. In a different context, this latter property of SF/HGF may also have an implication regarding carcinoma metastasis formation, since we show that SF/HGF acts not only on the detachment of epithelial and epithelia-derived cells, but also on the motile phase of individually migrating cells.

After ZPA removal, we observed a shift of Pax-3-expressing cells towards the posterior margin of the operated limb bud. This is consistent with the idea that the presence of SF/HGF in the mesenchyme is needed to make way for myogenic cell migration, as the SF/HGF expression pattern is, in a similar fashion, extended to the posterior margin of the ZPA excised limb bud.

Our data therefore suggest a SF/HGF-mediated role of the ZPA in patterning of premuscle masses. This strongly supports the hypothesis that early stages of limb muscle pattern formation are regulated by the same signalling mechanisms that pattern the limb skeleton.

We appreciate the help of Professor Ursula Abbott and Dr Jacky Pisenti, in providing us with limbless-mutant embryos. We thank Ulrike Pein and Lidia Koschny for excellent technical assistance, Professor Frank Stockdale for critical comments on the manuscript, and Drs Moises Mallo and Benoit Kanzler for helpful support. This work was supported by the Deutsche Forschungsgemeinschaft, DFG-grant Br 957/5-1.