Origin of muscle satellite cells in the Xenopus embryo

Muscle satellite cells are the stem cells of skeletal muscle (Hawke and Garry, 2001; Morgan and Partridge, 2003; Chargé and Rudnicki, 2004). They proliferate and contribute to new muscle formation during the growth of the animal, and then become quiescent in the adult. They are small cells lying beneath the basement membrane surrounding the multinucleate muscle fibers. Skeletal muscle does not show continuous cell turnover in the adult organism but following muscle damage satellite cells become reactivated. They are then released from the muscle fibers, divide, start to express myogenic transcription factors and become myoblasts. These may develop into myofibers or fuse with pre-existing fibers. Following reactivation, the satellite cells also undergo renewal divisions to maintain a population of undifferentiated satellite cells, although the cell number tends to fall with age (Kuang et al., 2008).

The biology of satellite cells in the adult has been well studied, but their origin in the embryo remains unclear. Previous studies in the mouse and chick embryo indicated that satellite cells originate from the somites and in particular from a Pax3/7-positive population of cells located in the central region of the dermomyotome of the somite (Armand et al., 1983; Gros et al., 2005; Relaix et al., 2005). But these studies did not indicate the origin of the satellite cells prior to somitogenesis, nor whether they come from the same or different region as the myoblasts that form the primary and secondary myofibers. Studies in the zebrafish indicate that at least some satellite cells come from anterior border cells within the somites, that later form a layer on the outer surface of the myotomes (Hollway et al., 2007). Although an origin from the somites is now thought most likely, some other studies had previously suggested an origin from quite distinct embryonic structures such as the dorsal aorta (De Angelis et al., 1999).

The Xenopus embryo offers an opportunity to investigate this problem. Unlike the mouse or chick, the neurula stage embryo is easily accessible and is well suited to microsurgical experiments. Muscle satellite cells certainly exist in Xenopus. Indeed they were first discovered in another frog species (Mauro, 1961). Our previous work on regeneration of tail muscle indicated that the Xenopus tail muscle contains abundant Pax7-positive cells and that these cells are able to regenerate the tail muscle following tail amputation (Chen et al., 2006). Like mammalian satellite cells, these Pax7-positive cells are small, with limited cytoplasm, and lie beneath the basement membrane of the muscle fibers. In the tadpole, which is growing in size, these cells are frequently in cycle, as shown by labeling with BrdU, whereas the myofiber nuclei are not. Although the Xenopus cells do not display all the molecular markers that have been associated with satellite cells in the mouse, by virtue of their Pax7 expression, morphology, location and behavior during muscle regeneration, they may be considered to be the exact equivalent of satellite cells in mammals. Satellite cells have also been examined in urodele amphibians and seem similar except that they are completely surrounded by the myofiber basement membrane (Morrison et al., 2006).

Our previous work on the cell lineage of the regenerating tail suggested that the paraxial mesoderm of the open neural plate neurula may contain few if any satellite cell precursors (Gargioli and Slack, 2004). The present study was aimed at investigating where the satellite cells do come from in the embryo and what factors lead to their formation.

We show here that a satellite cell precursor zone may be identified before the formation of the somites. It does not lie in the paraxial mesoderm, normally considered to be equivalent to the presomite plate, but in the adjacent dorsal part of the lateral plate. We show that the formation of satellite cells is promoted by BMP signaling. A ventral-to-dorsal gradient of BMP activity is a general feature of early vertebrate development. In Xenopus, it is set up at the beginning of gastrulation and is based on a widespread expression of BMP genes in the ventrolateral sector of the embryo, coupled with a localized secretion of BMP inhibitors (Chordin and Noggin) from the organizer region. It is well recognized that the BMP gradient controls mesoderm patterning during gastrulation (Hemmati-Brivanlou and Thomsen, 1995; Dosch et al., 1997; De Robertis, 2009) and we show here, at least with regard to satellite cell formation, that this gradient remains active into the neurula stages. We further show that the transcription factor Pax7 itself, although it is needed for satellite cell function in regeneration (Chen et al., 2006), is not involved in the initial formation of satellite cells.

For fate mapping of the mesoderm, transgenics were prepared and grafts were performed, as described previously (Gargioli and Slack, 2004). Hosts were cultured until stage 48/49.

Explants from neurulae were microdissected from stage 13-14 embryos corresponding to paraxial mesoderm or dorsolateral mesoderm, but including all three germ layers. Explants were individually cultured in 24-well plates in 0.5×NAM for 4 days. BMP treatment was 100 ng/ml rBMP4 (R&D Systems) for 24 hours followed by 3 days in 0.5×NAM.

For inhibition of BMP by Noggin, explants were dissected from HGEM-Noggin transgenic embryos (Beck et al., 2006), placed in 0.5×NAM plus or minus rBMP4 (100 ng/ml) and 4 hours later subjected to a heat shock (34°C for 15 minutes). Explants were cultured at room temperature for 24 hours, subjected to a second heat shock, and cultured for an additional 3 days in 0.5×NAM at room temperature.

Embryo culture and bFGF (50 ng/ml, R&D Systems) treatment of animal cap explants were carried out as described previously (Isaacs et al., 1994). Before dissection of animal caps, synthesized mRNA of candidate factors were individually injected in a volume of 9.2 nl into the animal pole region of both blastomeres at the two-cell stage.

To detect gene expression in animal cap explants, RT-PCR was performed on RNA extracted from five animal caps at day 5. One microgram of RNA was reverse transcribed and 1/20th of the cDNA reaction volume (1 μl) was amplified 30 cycles by PCR. Primers are listed in Table S1 in the supplementary material.

Whole-mount in situ hybridization and immunostaining were performed as described previously (Gargioli and Slack, 2004; Chen et al., 2006). For whole-mount immunofluorescence of explants and tadpole tails, tissue was fixed using a variable pH paraformaldehyde fixative procedure modified from Daughters et al. (Daughters et al., 2001). Briefly, tissue was incubated at room temperature for 5 minutes in 2% PFA (pH 6.5), followed by a 1 hour incubation in 2% PFA (pH 8.5) at room temperature. Primary antibodies used in the study were Pax7 (DSHB), NCAM (Chemicon), GFP (Abcam), pSmad1 (Cell Signaling) and Myosin (made in the J.M.W.S. laboratory).

Stained tissue was visualized on an inverted confocal microscope and analyzed using FV1000 software (Olympus). Photomicrographs are composite images of 5 μm optical slices through the tissue compressed along the z-axis and represent approximately 150 μm depth. Images were further processed using Photoshop (Adobe Systems).

To identify the origin of muscle satellite cells in the neurula, we carried out a fate-mapping study of the trunk/tail mesoderm of stage 13/14 Xenopus embryos. This was done by making orthotopic grafts of small pieces of tissue from CMV-GFP transgenic donors to wild-type host embryos, allowing them to develop to a stage at which satellite cells have differentiated (stage 48), and then counting the satellite cells within the area of the graft. Open neural plate neurulae (stage 13-14) were used for transplantation because this is before somite formation and at this stage the mesoderm is easily isolated free from the other germ layers. Furthermore, the position of the newly formed open neural folds provides an anatomical reference for the underlying tissue regions that were grafted. We carried out orthotopic grafts of three regions of mesoderm in the prospective trunk as shown in Fig. 1A,B: paraxial, dorsolateral and ventrolateral. The paraxial tissue is the thick mesoderm lying beneath the neural plate, while the dorsolateral and ventrolateral regions are of approximately the same breadth (100 μm) but lesser thickness. The dorsolateral region expresses many lateral plate markers, although expression of some somite markers also overlaps into this region (Pollet et al., 2005). All three regions populate much of the tail, as well as the trunk (Tucker and Slack, 1995). We raised the host embryos to stage 48 tadpoles, and checked the position of the graft by GFP expression in the whole tadpole before preparing transverse sections. The GFP-positive domain resulting from each type of graft was sectioned and stained with antibodies to GFP and to Pax7. The number of labeled myofibers and satellite cells was counted within a standard volume (10×10 μm sections, area 120 μm²). Pax7 staining was used, together with cell morphology, to identify satellite cells. There are also some Pax7-positive cells in the central nervous system at this stage (Chen et al., 2006) but these cannot be confused with muscle satellite cells.

The ventrolateral grafts never contributed any cells to the myotomes, but the paraxial and dorsolateral grafts both did so (Table 1). The gross distribution of labeled cells is shown for some typical closed neurulae (whole mounts showing GFP, stage 20-25, Fig. 1C-E), and for cross sections of typical tailbud stages at trunk level (drawings, stage 32, Fig. 1F-K). Cells from the paraxial region populate the central part of the mature somites. Cells from the dorsolateral region accounts for about one third of cells in the somites. They are always found in the ventral region and in 67% cases also in the dorsal region. This behavior is not surprising in view of the known dorsal convergence of mesoderm in the neurula and post-neurula stages (Hamilton, 1969). The gross picture remains the same at the later stage used for the analysis of satellite cell formation (tadpole stage 48). Among the three regions of mesoderm grafted, we found that only cells from paraxial and dorsolateral grafts end up within the myotomal skeletal muscle. The ventrolateral grafts populate entirely the connective tissue of the body wall or the gut wall (100%, n=15). For this reason, we excluded the ventrolateral region from further consideration.

The GFP construct used includes a nuclear localization sequence and the protein is enriched in nuclei. However, it is also present in the cytoplasm and so it is easy to identify labeled myofibers in transverse sections not containing a myonucleus. Although the transgenic embryos should theoretically express GFP in all cells, when sections were examined we found that on average only 92% of myofibers and 63% of satellite cells were positive for GFP under the conditions of our analysis (n=10, see Fig. S1 in the supplementary material). The percentage of labeled myofibers may be expected to be higher than other cell types because cell fusion means that not all nuclei in a fiber need be transgenic for the fiber to be green. Because each transgenic tadpole is a founder animal, we took the precaution of keeping all the donor embryos and raised them to the same stage as the corresponding hosts. Donors tend to have some defects but can usually survive to stage 48 and so each one was sectioned and the percentage of GFP-positive cells recorded for the myofibers and the satellite cells. In this subset of donors (n=19), the figures were 95% for myofibers and 49% for satellite cells, and there was no difference between the groups of donors used to provide the paraxial and dorsolateral grafts. These donor counts were used to correct the numbers of GFP-positive cells found in the corresponding hosts, and convert them to numbers of graft-derived cells. A total of 19 host tadpoles were counted and the counts were of a 120 μm² area across 10 sections spanning the region containing GFP-labeled cells. The results are expressed in terms of the number of cells of each of the two types formed from the graft within the standard volume that was counted. The collected data are shown in Table 1 and typical specimens in Fig. 2B,C. The results show that the paraxial mesoderm gave rise to a high proportion (71%) of myofibers but few (17%) satellite cells derived from the graft, whereas the dorsolateral region gave rise to many satellite cells (51%) but fewer myofibers (38%) derived from the graft. The precision of this type of study is necessarily limited by small dissection errors and small variations in healing of grafts into position. So the spatial resolution cannot be very high. However, the result clearly indicates that in normal development the satellite cell precursors originate from a more lateral level than the myofibers, most coming from outside the thickened region of paraxial mesoderm.

The fate-mapping result shows only the origin of the satellite cells in normal development and does not give any information about whether their progenitors are already specified at the time of grafting. This was investigated by culturing explants of tissue from early neurulae (stage 13-14) for 5 days until sibling controls had reached stage 48. These explants were full thickness, containing the overlying ectoderm as well as the mesoderm of the appropriate level, and results are shown in Fig. 3.

This showed two things very clearly. First, Pax7-positive cells do develop in the dorsolateral explants (mean 39.3 cells/explant). In some cases (four out of 12) they did so in the complete absence of myofibers, indicating that they can differentiate autonomously and do not require any interaction with developing myofibers. Second, virtually no Pax7-positive cells differentiated in the paraxial explants (mean 0.56 cells/explant). The specification study shows a sharper difference between the paraxial and dorsolateral levels than the fate map study. This may be because the fate map study has a larger variance due to the grafting procedure, or because the environment in the two types of explants reinforce the initial difference in progenitor cell representation, or both. Because the position of satellite cells under the muscle fiber basement membrane cannot be defined if they are not associated with fibers, for this and other explant experiments, we refer to ‘Pax7-positive cells’ rather than ‘satellite cells’. However, we believe that they are in fact satellite cells.

Because of the well established importance of Pax7 for satellite cell function, we investigated whether Pax7 itself is involved in the initial formation of satellite cells by overexpressing either Pax7 or a dominant-negative construct, Pax7EnR. The activity of Pax7EnR as an inhibitor of Pax7 was demonstrated in our previous publication (Chen et al., 2006). To limit damage to other parts of the embryo, these overexpression experiments were carried out by injection of mRNA into single blastomeres at the 32-cell stage. According to the fate map (Dale and Slack, 1987), blastomere C3 projects to the paraxial region of the trunk and blastomere C4 to the lateral plate of the trunk. The domains filled by these blastomeres at trunk level thus approximate to the regions studied in our microsurgical experiments, with C3 roughly corresponding to the paraxial mesoderm and C4 to the dorsolateral and ventrolateral mesoderm (see Fig. S2 and Table S2 in the supplementary material).

Following injection of Pax7 mRNA, there was an induction of ectopic Pax3 (C3 or C4 injection) and Myf5 (C3 injection only) (Fig. 4; see Table S3 in the supplementary material), indicating that the injected RNA is biologically active. But the number and position of muscle fibers remained normal.

To examine the effects on satellite cell formation, we examined sections of the C3-injected tadpoles at stage 45, double immunostained with Pax7 and NCAM antibodies. The reason for using NCAM as well as Pax7 is because there might be residual Pax7 protein present resulting from the Pax7 RNA injection. Ten specimens of each kind of injection were used, with five sections counted from each. This showed no increase in the number of satellite cells on the injected side (Fig. 4J). The same was true of the C4-injected embryos. There was also no effect on the number of satellite cells following injection of 100 pg Pax7EnR mRNA into the C4 blastomere (Fig. 4J).

The overexpression experiment shows that Pax7, either alone or together with the induced Pax3, is not sufficient to specify either myogenic or satellite lineages. The dominant-negative experiment shows, in addition, that Pax7 is not required for satellite cell formation. There is no doubt that Pax7 is required for satellite cell function at later stages, but by then the Pax7EnR RNA will have degraded and will no longer be active.

In order to investigate what factors do affect satellite cell formation, we adapted the animal cap assay. Isolated animal caps do not form either myofibers or satellite cells. However, significant muscle masses are induced following treatment with fibroblast growth factors (Slack et al., 1987). Interestingly, we found that such induced caps do not generate Pax7-positive cells. Although the intact embryo contains some Pax7-positive neurons by stage 48, our animal cap explants also did not produce any neural tissue in these experiments.

For testing factors, two-cell stage embryos were injected with either control (GFP) or candidate factor mRNA. Animal caps were removed at blastula stage 8-9, induced to form mesoderm with FGF2 and maintained in culture medium for 5 days to allow time for the development of Pax7-positive cells. In response to mRNA encoding early myogenic regulatory factors, including Pax3, Pax7 and Myf5, there was no expression of endogenous Pax7 or formation of Pax7-positive cells (see Table S4 in the supplementary material). This confirmed the results from the whole embryo experiments presented above. Similar results were observed for factors affecting the Notch and Wnt pathways, which are known to regulate myogenesis and somite formation (Bryson-Richardson and Currie, 2008; Dequéant and Pourquié, 2008). However, when we overexpressed Bmp4, we observed a significant increase in Pax7 expression both by RT-PCR and by cell counts after Pax7 immunostaining (mean 17 cells/cap) compared with control (mean 0.22 cells/cap) (see Fig. S3 and Table S4 in the supplementary material). A dose-response study with Bmp4 mRNA showed that the highest number of Pax7+ cells was achieved at 50 pg of mRNA (Fig. 5). Similar results were achieved by treatment with BMP4 protein (see Fig. S4 in the supplementary material).

Increases in Pax7 expression and Pax7-positive cells were also observed after overexpression of Msx1 (mean 27 cells/cap), a known downstream target of BMP signaling (Suzuki et al., 1997), and Fgf8 (mean 21 cells/cap), which is expressed in the posterior mesoderm surrounding the blastopore in the pre-somite stage embryo (Christen and Slack, 1997).

We next wanted to determine whether we could trigger the formation of Pax7-positive cells in neurula stage paraxial explants, which form virtually no Pax7-positive cells on isolation (Fig. 3), by induction with BMP4. Paraxial explants cultured in normal medium did not develop Pax7-positive cells, whereas with 100 ng/ml BMP4 for 24 hours followed by an additional 3 days in culture they formed a large number (mean=31; Fig. 6). Dorsolateral explants formed similar numbers of Pax7-positive cells with or without the BMP treatment. When paraxial explants were treated with BMP starting at different stages, the increase was only seen when treatment was commenced at stage 13, and not at stage 18 or 22 (Table 2). This shows that competence to form Pax7-positive cells is only present early on and this is consistent with the data on Noggin-induced inhibition (see below), indicating that the BMP-induced specification of satellite cell progenitors occurs during the neurula stages. In this experiment, the ‘paraxial’ explants from later stages than 13, because they are the same width as at stage 13, contain former dorsolateral tissue which has converged in the intervening period. For this reason, the larger number of Pax7-positive cells in the controls is to be expected.

We next wanted to determine if the satellite cell inducing factors identified by means of our in vitro assays also regulated their formation in vivo. We co-injected either Bmp4 (10 pg), Msx1 (100 pg) or Fgf8 (100 pg) with Gfp (80 pg) mRNA as a lineage tracer into either C3 or C4 blastomeres of a 32-cell stage embryo to fill paraxial or lateral mesoderm of the trunk. Expression domains were verified by observation of GFP expression the next day, and embryos were cultured in 0.1×NAM for 5 days. The tails of tadpoles were then fixed and analyzed for changes in satellite cell number by whole-mount immunofluorescence for Pax7 and muscle Myosin. A standard square was counted for each tadpole, so the figures represent the numbers of Pax7-positive cells within a certain volume of tail muscle, not in the domains filled by C3 or C4 blastomeres. This is why the control figure is similar for both.

In C3-injected embryos we found that Bmp4 (mean=303), Msx1 (mean=287) and Fgf8 (mean=291) all showed significant increases in the number of Pax7-positive cells compared with control (mean=142; Fig. 7). Similarly, overexpression of these factors in lateral plate (C4) increased the number of Pax7-positive cells formed by Bmp4 (mean=211), Msx1 (mean=265) and Fgf8 (mean=223) compared with control (mean=127; Fig. 7). Although overexpression in lateral plate did increase satellite cell formation the percent increase was greater when these factors were overexpressed in paraxial mesoderm. In these cases, the tail muscle morphology was disrupted to varying degrees but the increase in number of satellite cells did not seem to lead to enhanced muscle formation.

If satellite cell progenitor formation is controlled by BMP4 in the neurula stage mesoderm, then we should be able to reverse the increase of Pax7-positive cells in paraxial explants by overexpression of Noggin. We overexpressed Noggin by using a heat shock-inducible Noggin transgenic line (HGEM-Noggin) previously generated in the laboratory (Beck et al., 2006). First, we needed to establish that the induced Noggin is sufficiently active to inhibit concentrations of BMP that will induce the formation of Pax7-positive cells. To do this, we subjected paraxial explants from stage 13/14 HGEM-Noggin transgenic embryos to a 15-minute heat shock to induce expression of Noggin. Explants were then maintained in media containing BMP4 for 20 hours and subjected to a second heat shock. They were then transferred to normal medium without BMP4 and cultured for an additional 3 days. As expected from the experiments described above, paraxial explants treated with BMP4 form many Pax7-positive cells (mean=30; Fig. 8C), but following induction of Noggin the number of Pax7-positive cells is returned to the control level (mean=0.23; Fig. 8D). This demonstrates that the heat shock induction of Noggin in this transgenic line is able to inhibit levels of BMP4 that are required to induce satellite cell formation.

The most critical experiment was then to dissect out dorsolateral explants from neurulae, heat shock to induce Noggin, and examine the number of Pax7-positive cells formed (Fig. 8B). This showed a dramatic reduction of numbers of Pax7-positive cells compared to controls (Fig. 8A). This loss of function experiment enables us to conclude that the formation of Pax7-positive cells from the dorsolateral mesoderm is indeed dependent on endogenous BMP signaling. Furthermore, the timing of this experiment indicates that the satellite cell progenitors cannot be determined, i.e. irreversibly committed, before the neurula stage, because the Noggin induction can prevent their formation when administered to explants that are at a stage equivalent to the commencement of neural tube closure. It also confirms that the effects of BMP signaling that we observe are exerted during the period of satellite cell specification, within 24 hours of the open neurula, and not at later stages of myogenesis (3-5 days post-fertilization).

BMP signaling is mediated by phosphorylation of Smad1 and Smad5, so it is to be predicted that such phosphorylation would be detectable in the dorsolateral mesoderm of the neurula. Accordingly, we immunostained sections of open neural plate embryos with anti-Smad1-phosphate. Some positive nuclei are found in the dorsolateral mesoderm, but not in the paraxial mesoderm (mean 8.8 positive cells/3 sections, n=5 embryos; Fig. 9A). This is consistent with an active response to BMP signaling in this region at this time.

We also carried out in situ hybridization for Msx1. In common with previously published results, there is clear expression in the neural crest (Fig. 9B). However, on sections there is no visible expression in any part of the mesoderm (Fig. 9C). So although Msx1 has a potent effect on Pax7-positive cell formation following overexpression, we cannot be sure that it is really an intermediate step in the process in vivo.

In the past, the majority view has been that satellite cells originate from the same organ rudiment as produces the skeletal muscle (Armand et al., 1983; Gros et al., 2005; Relaix et al., 2005). A minority has considered that they derive from a completely different site such as the dorsal aorta, other blood vessels or the bone marrow (De Angelis et al., 1999; Gussoni et al., 1999; Seale and Rudnicki, 2000). Our results indicate that neither view is entirely correct. At least in Xenopus embryos, the satellite cells arise from a region that is morphologically part of the lateral plate mesoderm at the neurula stage, adjacent to, but distinct from, the main paraxial plate. Some of our experiments show that Pax7-positive cells can arise without accompanying myofibers, indicating that the progenitor cell populations for satellite cells and myofibers must be different.

In the Xenopus neurula, the dorsolateral mesoderm is distinct from the paraxial mesoderm in that it is thinner and is located outside the neural folds (Fig. 1). However, it does converge to become incorporated into the somites in the course of dorsal convergence movements in the late neurula and early tailbud stages [see Fig. S2 in the supplementary material and Hamilton (Hamilton, 1969)]. The ‘presomite plate’ is often considered to be identical to the thick paraxial mesoderm, but if the presomite plate is defined as all cells that later contribute to the somite, it must obviously include some part of the lateral plate. Once they become visible, satellite cells are found throughout the myotomes, so the progenitor cells must migrate as individuals into the central myotome region, which is formed from the paraxial mesoderm. We do not know whether all the cells from a specific dorsolateral level of mesoderm become satellite cells, or whether only some do so with the others becoming muscle fibers or fibroblasts; however, the sparse staining for Smad1-phosphate shown in Fig. 9 suggests that only some cells are responding to BMP at this time. The mouse and chick experiments, indicating an origin from the central dermomyotome, were carried out after somite formation and therefore corresponded to a much later stage in Xenopus than our own experiments, perhaps around the equivalent of stage 35. For this reason, we consider that the Xenopus data presented here are not necessarily inconsistent with the data from the amniote species. We predict that if it is possible to locate a population of satellite cell precursors in amniotes before somitogenesis then they too are likely to lie in the dorsal lateral plate, and experiments to test this hypothesis are currently in progress.

There has been much recent work on the zebrafish devoted to establishing how various types of muscle cell arise from the somites (Hollway et al., 2007; Stellabotte and Devoto, 2007; Stellabotte et al., 2007). This has shown the existence of a population of anterior border cells within each newly formed somite, not expressing myogenic regulatory factors, that migrate to the outer surface of the myotome and form a layer of thin cells that are Pax7 positive. Some of these subsequently enter the myotomes and become satellite cells. Others contribute to myofibers of myotomes, body wall or pectoral fin, or to fibroblast-like cells in the dermis and dorsal fin. This illustrates the principle that satellite cells may arise separately from the main myotomal fiber mass and migrate into it. However, the course of somitogenesis in Xenopus is not the same as the zebrafish. Concurrent with segmentation of the somites there is an early rotation of all the prospective myotome cells to an anteroposterior orientation such that each myotome is one cell long (Hamilton, 1969). There is also a layer described as the dermatome, or the dermomyotome, lying outside the myotomes. This is visible in Plate 38 of Hausen's atlas (Hausen and Riebesell, 1991) and has been described by various authors previously (Hamilton, 1969; Blackshaw and Warner, 1976). According to a study by Grimaldi et al. (Grimaldi et al., 2004) this layer is uniformly positive for Pax3. Owing to the difference of cell movements during segmentation, we consider it very unlikely that this layer comes from the anterior of each somite, as seen in the zebrafish. We believe that it arises from the dorsolateral mesoderm as a result of the dorsal convergence movement described by Hamilton. If so, it is likely that the prospective satellite cells are located in this layer before entering the myotomes. However, in Xenopus this layer of cells is not uniformly Pax7 positive. In fact there is very little Pax7 expression at all in the tailbud stages (Chen et al., 2006).

Pax7 is a transcription factor expressed in muscle satellite cells but not in mature myofibers (Seale et al., 2000). It was initially reported that Pax7⁻ mice lacked satellite cells, but was subsequently shown that they did develop satellite cells which were not capable of self-renewal (Oustanina et al., 2004). In Xenopus, we showed that Pax7 was required for regeneration of muscle following tail amputation (Chen et al., 2006), although in mice there remains some dispute about the need for Pax7 for muscle regeneration (Lepper et al., 2009).

In Xenopus, Pax7 is expressed in the neural crest during the neurula stages. Subsequently, there is a low level expression in somites and myotomes, but no Pax7-positive outer layer as is seen in the zebrafish (Chen et al., 2006). There is extensive expression in neurons forming in the dorsal half of the neural tube and spinal cord. The picture in the mouse and chick is generally similar, although in these amniote species there is a higher and more general expression in the dermomyotome (Lacosta et al., 2005; Horst et al., 2006). In mammals, the related transcription factor Pax3 is expressed in the whole early somite and satellite cell precursors express both Pax3 and Pax7 (Relaix et al., 2005; Buckingham and Relaix, 2007).

As Pax7⁻ mutant mice do form satellite cells, Pax7 must be dispensable for their initial formation in development, even if it required later for self-renewal. But because of the close association with Pax3, it is possible that Pax3 rescues the formation of satellite cells in the absence of Pax7. Our own evidence clearly confirms that Pax7 is not required for satellite cell formation. We show that overexpression of Pax7 in the presomite plate or in FGF-treated animal caps does not increase satellite cell formation, and that antagonism of Pax7 function with Pax7EnR does not reduce satellite cell formation from the dorsolateral mesoderm. We also show that Pax3 does not drive satellite cell formation, or Pax7 expression, in FGF-treated animal caps (see Table S4 in the supplementary material). These collected data make it very unlikely that Pax7 or Pax3 drives the initial specification of satellite cells in development.

It is legitimate to ask whether it is sufficient to use the expression of Pax7 to identify satellite cells. In our previous publication (Chen et al., 2006), we showed that they corresponded to morphological satellite cells and that they were active in DNA synthesis during tail growth. Many, but not all, of the Pax7-positive cells in muscle also express NCAM, a recognized marker for satellite cells in mouse. In many of the explant experiments reported here, the Pax7-positive cells are separated from muscle fibers. In such circumstances the satellite cells obviously cannot be localized under the basement membrane of fibers, which is a major criterion for identification. This is why we refer to ‘Pax7-positive cells’ rather than ‘muscle satellite cells’ in the relevant sections. However, we believe that they are in fact satellite cells. The only other Pax7-positive cells in the body during the relevant developmental stages are various classes of neuron in the central nervous system, but none of our experiments generated ectopic neural tissue, and the Pax7-positive cells do not resemble neurons morphologically.

The ventral-to-dorsal gradient of BMP was discovered in Xenopus and later shown to be a general feature of early vertebrate development (Hemmati-Brivanlou and Thomsen, 1995; Yamagishi et al., 1995; Jones et al., 1996; Dosch et al., 1997; Kurata et al., 2001; De Robertis, 2009). BMPs become activated autonomously due to maternal BMP activity activating their own genes' transcription. The expression domains are later refined by negative interactions that ensure that cells cannot at the same time express BMP genes and organizer genes (Reversade and De Robertis, 2005). The gradient depends on diffusion of BMP protein from the producing zone, coupled with inhibition due to binding of Chordin and Noggin from the organizer. The activity of signaling is converted into a nuclear level of phosphorylated Smad-1 and -5 (Simeonia and Gurdon, 2007). Many genes are activated by BMP signaling, and an obvious candidate to be in the satellite cell formation pathway was Msx1 (Suzuki et al., 1997; Maeda et al., 1997). We find that that Msx1 behaves very similarly to BMP in our assays of Pax7-positive cell formation. However, we do not find that it is expressed in the dorsolateral mesoderm of the neurula, so unless it is active at sub-detection levels, we conclude that it is not involved in satellite cell formation in normal development. A similar consideration may apply to FGF8. This is expressed in the posterior of the forming axis and has been considered a posteriorizing factor in other vertebrates as well as in Xenopus (Christen and Slack, 1997; Dubrulle and Pourquie, 2004). The role of FGF8 has been investigated in zebrafish myogenesis and has a negative effect on the formation of Pax7-positive cells, albeit these are cells of the dermomyotome layer rather than the definitive satellite cells (Hammond et al., 2007).

We believe that satellite cell specification is occurring during neurula stages rather than the tadpole stages for two reasons. First, the effect of BMP treatment on paraxial explants is most pronounced when the BMP is added from stage 13, rather than from stage 18 or later (Table 2). This means that competence to respond to BMP is present in the open neurula and falls away by neural tube closure. Second, satellite cell formation in dorsolateral explants from HGEM-Noggin neurulae is suppressed by heat shock The Noggin mRNA induced by heat shock decays after about 24 hours (Beck et al., 2006) so this inhibitory effect cannot still be persisting during the tadpole stages when the Pax7 expression actually becomes visible.

The BMP gradient is normally supposed to be a phenomenon of the gastrula, as this is when neural induction and mesoderm dorsalization take place (Lettice and Slack, 1993). But we also know of studies indicating some lability of mesoderm explanted from the neurula of urodele embryos (Yamada, 1937; Forman and Slack, 1980). So it does seem likely that the BMP gradient can remain active through the neurula stages and participate in additional dorsoventral decisions, such as the formation of the satellite cell precursor zone.

Very recently, two other studies have investigated the effects of BMP signaling on satellite cell numbers, although in rather different contexts from our own work. This effect is in the same direction as that which we have described, although at a much later stage of development, when satellite cells are already present in the muscles. In the first study it was shown that additional BMP signaling in the limbs of late chick embryos can stimulate satellite cell number, while Noggin will reduce it (Wang et al., 2010). In the second study, focusing on the myotomes of zebrafish embryos, it was shown that upregulation of BMP delays the differentiation of the Pax7-positive dermomyotome cells to myofibers leading to a temporary increase in Pax7-positive cell numbers (Patterson et al., 2010). Although these works describe processes leading in the same direction as ours, we feel that the events concerned are later in development that the satellite cell precursor specification that we describe. In our experiments, the effects of BMP signaling are exerted at neurula stage when there are no Pax7-positive cells present, and the satellite cells themselves differentiate several days later.

In summary, our work demonstrates an origin for muscle satellite cells from the dorsolateral mesoderm and a dependence of their formation on the BMP gradient in the early embryo. We consider that this mechanism is quite likely also to exist in higher vertebrates.

This work was supported by a NIH (NIAMS) Training Grant #T32-AR050938, NIH (NIGMS) R01 grant #R01GM088500 and Wellcome Trust Programme Grant #60082. Deposited in PMC for release after 6 months.