Hedgehog regulation of superficial slow muscle fibres in Xenopusand the evolution of tetrapod trunk myogenesis

During vertebrate skeletal muscle development, several populations of myogenic cells arise within each somite due, at least in part, to signals impinging on the somite from neighbouring tissues(Borycki and Emerson, 2000; Pownall et al., 2002). In amniotes, many myogenic cells arise from the dermomyotome, an epithelial sheet covering the lateral somitic surface, although small numbers of early myogenic cells arise in chick prior to dermomyotome formation. A large part of the somite forms sclerotome. In zebrafish, the earliest populations of myogenic cells arise from several zones of the early somite, yet no dermomyotome has been described. Instead, the outer surface of the developing somite becomes covered in a layer of specialised slow muscle fibres and sclerotome formation is late and limited (Stickney et al.,2000). To shed light on the origin of the dermomyotome and the relationship of myogenic cells populations in zebrafish and amniotes, we examined myogenesis and dermomyotome formation in Xenopus laevis, an anuran species in which a `dermatome' has been described but is poorly characterised (Keller, 2000; Nieuwkoop and Faber,1967).

In zebrafish, the medial adaxial myogenic cell population arises early in presomitic mesoderm (PSM) next to the notochord. Adaxial cells have a distinct cuboidal morphology and express the myogenic basic helix-loop-helix transcription factors myf5 and myod, followed by slow myosin heavy chain (MyHC) (Chen et al.,2001; Coutelle et al.,2001; Devoto et al.,1996; Weinberg et al.,1996). Subsequently, as somite borders form, myf5 and myod mark a distinct population of cells in the lateral somite(Chen et al., 2001; Coutelle et al., 2001; Weinberg et al., 1996). The fate of the medial and lateral myogenic cells is known: medial adaxial cells form slow muscle and most migrate laterally to generate the superficial slow muscle fibres that lie within the somite under the epidermis(Blagden et al., 1997; Devoto et al., 1996). At a slightly later stage of development, lateral somitic cells give rise to the fast muscle that makes up the bulk of the myotome(Devoto et al., 1996). Subsequently, the myotome grows by the addition of further cell populations at its dorsal and ventral extremes, a situation reminiscent of the dorsomedial and ventrolateral dermomyotomal lips of amniotes(Barresi et al., 2001; van Raamsdonk et al., 1982; Veggetti et al., 1990).

Secreted signalling molecules encoded by Hedgehog (Hh) genes, which are expressed in ventral midline tissues, are required for appropriate medial slow muscle development (Barresi et al.,2000; Blagden et al.,1997; Coutelle et al.,2001; Currie and Ingham,1996; Du et al.,1997; Lewis et al.,1999; Norris et al.,2000). Hh is not required for myogenesis of lateral fast muscle or a second wave of slow fibres formed from the dorsoventral tips of the myotome(Barresi et al., 2001; Blagden et al., 1997). Thus, in fish, the distinct contractile fibre type of successive waves of fibres has permitted elucidation of several modes of fibre formation.

In amniotes, independent regulation of several somitic muscle precursor populations has also been described although no clear-cut distinction, e.g. on the basis of fibre type, between the products of different myogenic inductions has been reported (Hadchouel et al.,2003; Kahane et al.,2001; Sacks et al.,2003). In birds and rodents, as in fish, early populations of fibres express slow MyHC, whereas some later fibres do not. As in fish,prevention of ventral midline signalling or blocking sonic hedgehog(shh) in birds eliminates markers of the earliest myogenic cells in the somite (Borycki et al.,1998; Pownall et al.,1996), while shh overexpression can enhance expression of myod and terminal muscle differentiation(Johnson et al., 1994; Kahane et al., 2001). Mice with ablated Shh function show reduced epaxial myogenesis at the dorsomedial lip of the dermomyotome(Borycki et al., 1999; Chiang et al., 1996). Elimination of the Shh and indian hedgehog (Ihh) genes or their receptor Smoothened in the mouse leads to more severe loss of medial myogenesis (Zhang et al.,2001), just as occurs in zebrafish. Lateral myogenesis within the somite is relatively unaffected by loss of Hh signalling. However, the parallel role of Hh in induction of sclerotome(Fan et al., 1995), which constitutes a large part of the early somite in amniotes, and conflicting views on the role of Hh in dermomyotome development makes interpretation difficult (Borycki et al.,1999; Cann et al.,1999; Huang et al.,2003; Kruger et al.,2001; Teillet et al.,1998). Actinopterygian teleost species diverged from sarcopterygian amniote ancestors over 400 Mya. Nevertheless, the similarities between amniote and fish myogenesis raise the possibility that a common system has been co-opted to different ends during evolution. To address this issue,we turned to amphibia, which, although sarcopterygian derived, diverged from amniotes over 350 Mya.

In Xenopus laevis, the axial musculature begins to differentiate early (Cary and Klymkowsky,1994; Chanoine and Hardy,2003; Hopwood and Gurdon,1990; Hopwood et al.,1991). As in zebrafish, axial muscles are functional by 24-26 hours of development (stage 22-24, when the frog has 12-15 somites) and differentiation progresses in an anterior to posterior direction. Thus, over a number of developmental stages, anterior myotomes contain muscle cells that are more advanced than posterior myotomes. There is a major difference in fate of anterior and posterior somites in Xenopus: anterior (trunk)somites ultimately generate the complex vertebral and muscular structure of the metamorphosed adult, whereas posterior (tail) somites are fated to cell death (Nieuwkoop and Faber,1967). Although the differentiation of muscle fibres starts at a very early stage, differences between muscle fibre types have not been reported until stage 35 or later (Hughes et al., 1998; Kordylewski,1986; Schwartz and Kay,1988). As in fish, the bulk of the late Xenopus embryo somite is built of large fibres, whereas, on the lateral surface of the myotome, there is a monolayer of distinct fibres. In many anurans a cell layer designated dermatome exists in this lateral location(Radice et al., 1989). We describe the pattern of differentiation of slow and fast muscle fibre types at early stages and its dependence on anteroposterior position. We show that manipulation of Hh signalling can affect the decision between fast and slow muscle formation in a manner similar to that observed in zebrafish. We go on to characterise later somite myogenesis and suggest a model for evolutionary adaptation of myogenesis in the transition from fish to tetrapod bodyplan.

Xenopus laevis embryos were staged according to Nieuwkoop and Faber (Nieuwkoop and Faber,1967). Embryos at the two- to four-cell stage were injected with RNA prepared from pSP64T plasmid containing the full-length coding region of zebrafish shh. Whole-mount mRNA in situ hybridisation was performed using antisense RNA probes to Col1a1(Goto et al., 2000), Xptc2 (Takabatake et al.,2000), Pax3 (Martin and Harland, 2001), XMyf5(Hopwood et al., 1991), actin and XMyoD (Hopwood and Gurdon, 1990). Devitellinized embryos were exposed to 100μM cyclopamine or solanidine or ethanol vehicle control.

Antibodies used in this paper are available from Alexis, ATCC, the DSHB,Iowa and/or DSG Braunschweig, Germany. Antibody A4.1025 (IgG2a) recognises many, probably all, sarcomeric MyHCs in species from Drosophila to human (Dan-Goor et al., 1990). BA-F8 (IgG2b) was raised against human MyHC and reported to react with slow and cardiac MyHC in humans and mice, and BA-D5 (IgG2b) was raised against human MyHC and reported to detect slow MyHCs in rodent, chicken and zebrafish(Blagden et al., 1997; Schiaffino et al., 1989). Antibody EB165 (IgG1), raised against chicken fast MyHC, was a gift from Dr Everett Bandman (Cerny and Bandman,1986). The muscle marker 12/101 was a gift from Dr Jeremy Brockes. Cryosections of staged embryos fixed in Dent or paraformaldehyde fixative were stained according to (Blagden et al.,1997), with the use of class or subclass-specific secondary and fluorescent tertiary reagents (Jackson ImmunoResearch). After whole-mount in situ mRNA hybridisation, embryos were cryosectioned and reacted for all MyHC with A4.1025 or anti-proliferating cell nuclear antigen (PCNA, Sigma). Analysis and photomicrography was on Zeiss Axiophot or Axiocam. For plastic sections, devitellinized embryos were fixed in 2% glutaraldehyde in amphibian Ringer solution, embedded in Epon-Araldite 812 mixture and semi-thin sections stained with crystal violet and basic fuchsin.

We screened over 20 monoclonal antibodies known to recognise specific isoforms of mammalian or avian MyHC on sections of developing Xenopusand found several that distinguished two populations of muscle fibres in stage 35 embryos (Fig. 1). EB165 and BA-F8 react strongly with a single layer of superficial muscle fibres(Fig. 1B,C). BA-D5 reacts with the vast majority of the muscle fibres at this stage, as revealed by all MyHC antibody A4.1025 and by 12/101, a frequently-used Xenopus muscle marker (Fig. 1A,D; data not shown). Thus, in Xenopus, the 12/101;A4.1025;BA-D5 and EB165;BA-F8 antibodies distinguish two muscle fibre populations.

Slow muscle fibres are usually oxidative with abundant mitochondria. NADH tetrazolium reductase histochemistry reveals oxidative cells in the superficial location of EB165;BA-F8-reactive cells(Fig. 1E), as in zebrafish(van Raamsdonk et al., 1978). The presence of the BA-F8 slow MyHC epitope and abundant mitochondria in superficial cells suggests they are slow fibres. This expression persists through at least st48 (Fig. 1F,G). EB165 and BA-F8 also detect oxidative slow fibres weakly but preferentially in adult Xenopus leg muscle(Fig. 1H-J). Whole-mount staining revealed that EB165 also detects a subset of fibres in head muscles and that EB165⁺ fibres are present near the leading edge of the fibres forming from migratory muscle precursors on the abdomen(Fig. 1K,L)(Martin and Harland, 2001). So, EB165 reactivity may be transient during fibre type maturation in some fibre populations. Nevertheless, we hereafter employ BA-D5, A4.1025 and 12/101 as markers of all fibres, and BA-F8 and EB165 as markers of the superficial slow fibres. For simplicity, and by analogy with zebrafish, we hereafter refer to the fibres that do not react with BA-F8 and EB165 as `fast'.

We investigated the timing of appearance of the differentiated slow and fast muscle fibre populations in Xenopus(Fig. 2). The earliest-formed muscle is fast, appearing in anterior somites before stage 22(Fig. 2A,B). As in zebrafish,fast cells form the majority of medial/deep muscle in the older animal(Fig. 1F; Fig. 2K-N). By contrast, the slow fibres arise later in development, being first detected weakly in the tail tip of stage 27/28 embryos where a group of slow cells appear to span the somite from adjacent to the notochord towards the lateral somite surface(Fig. 2F, inset). At this stage, embryos have about 20 somites, and yet the older trunk somites do not have slow fibres (data not shown). By stage 35 the posterior half of the embryo, including somites 18-20, contains superficial slow cells outlining the lateral border of each somite (Fig. 2I, see also Fig. 5A). In somite 36 at the tail tip, the slow fibre markers span the somite transversely, as occurred earlier in the 20th somite (compare Fig. 2F, inset, 2J). As the functional studies below reveal, slow fibres formed prior to about stage 35 have a similar origin and we designate them `first wave slow fibres'. At later stages, slow superficial fibres appear in successively more anterior somites such that, by stage 48, all somites contain an outer layer of slow cells(Fig. 2K-N). This anterior extension of slow fibres roughly parallels the retraction of the gut, so that somites without underlying endodermal tissue contain slow fibres. The superficial slow fibres are detected preferentially at the dorsal and ventral extremes of the somite in stage 48 embryos, which is not the case at stage 35(Fig. 2I-N). Functional studies below indicate that these later-formed slow fibres have a distinct embryological origin and we designate them `second wave slow fibres'. Thus,because anterior (trunk) somites develop first, deep medial `fast' fibres arise first during development. In more posterior (tail) somites, by contrast,first wave slow fibres arise first, initially in the medial somite and then become located more superficially as the bulk of the somite differentiates into fast muscle. Second wave slow fibres arise later in all somites.

Superficial slow fibres in zebrafish arise next to the notochord and depend on axial signals carried by Hh proteins(Blagden et al., 1997; Du et al., 1997). To test the homology between the superficial slow muscle fibres of Xenopus and zebrafish with respect to their induction by Hh proteins, we overexpressed Shh in Xenopus embryos and analysed muscle fibre pattern at stage 35. Shh-injected, but not control uninjected or lacZ-injected,embryos show a mis-positioning and an increased incidence of slow muscle fibres (Fig. 3). Shh-injected embryos contain significantly more slow fibres/section(33.0±1.7 profiles, P<0.01) than controls(21.6±0.6). Both the extent and location of ectopic slow markers are patchy, with the location of ectopic slow fibres correlating with regions of co-injected lacZ activity (Fig. 3). However, ectopic slow fibres never exceeded a threefold increase in fibre number, nor did we detect loss of fast fibres. Strikingly,we did not observe increased slow fibres in anterior trunk somites. Thus,extra Hh signalling can augment first wave slow fibre formation, suggesting that Hh promotes formation of first wave slow fibres in Xenopus, as in zebrafish.

In zebrafish, notochord-derived Hh is required for adaxial cell myf5 and myod expression and slow fibre formation, but not for lateral myf5 and myod expression and fast muscle differentiation (Blagden et al.,1997; Coutelle et al.,2001). To test for a role of midline signals in Xenopusmyogenesis we ablated notochord, but not neural plate which is required for lateral myogenesis (Mariani et al.,2001). Extirpation of notochord causes significant reduction of XMyf5 expression. Yet there is no noticeable loss of XMyoDor actin expression, which confirms that neural plate remains intact. XMyf5 mRNA decline is particularly noticeable in adaxial cells flanking the anterior notochord (Fig. 4A). This result raises the possibility that a midline-derived signal promotes adaxial, but not more lateral, myogenesis in Xenopus,just as in zebrafish.

As notochord expresses Shh, we tested the hypothesis that Hh is required to generate slow muscle. Xenopus embryos were exposed to the Hh-signalling antagonist cyclopamine during the period of slow myogenesis. Cyclopamine inhibits Smoothened, a required component of the Hh signalling pathway (Chen et al., 2002; Frank-Kamenetsky et al.,2002). First, we demonstrated that cyclopamine inhibits Hh signalling in Xenopus embryos by examining the expression of the patched gene, Xptc2 (Takabatake et al., 2000). Expression of patched genes, which encode Hh receptors, is upregulated by Hh signalling(Ingham and McMahon, 2001). At stage 28, Xptc2 is highly expressed in tailbud regions where slow muscle cells are forming next to Xshh expression in the notochord,and also in somite chevrons close to banded hedgehog expression(Ekker et al., 1995). Treatment with cyclopamine blocks Xptc2 expression throughout the embryo at stage 13, 20, 28 and 36, whereas treatment with the related alkaloid solanidine has no effect (Fig. 4B and data not shown). Therefore, cyclopamine treatment is a specific Hh signalling inhibitor in Xenopus.

At early stages, cyclopamine has no effect upon medial XMyf5expression in trunk regions, where first wave slow muscle does not form, even though Xptc2 expression is suppressed (data not shown). However,cyclopamine downregulates XMyf5 expression in the PSM and nascent somites in the tail creating a `gap' in tailbud expression just where untreated embryos initiate first wave slow myogenesis in the tail(Fig. 4C). Reduction of XMyf5 is specific to the gap region at this stage, as expression in dorsal and ventral myotome of maturing somites is not affected. Strikingly,the missing XMyf5 expression domain is the region where Xptc2 and XMyf5 mRNAs are high in medial cells flanking notochord (see Fig. 4B, Fig. 7G). Moreover, expression of XMyoD is reduced in the tail, though less markedly than XMyf5, but is less affected anteriorly(Fig. 4C). Thus, Hh signalling is required for normal myogenesis in the region of the first slow fibre formation in the Xenopus tail.

We next determined whether the cyclopamine-induced reduction in MRF expression in the medial cells of tail PSM is paralleled by later changes in slow muscle. Cyclopamine greatly reduces slow muscle formation in tail somites of stage 36 and younger embryos (Fig. 5). The related alkaloid solanidine has no effect on slow myogenesis, paralleling its inability to block Hh signalling(Fig. 4B; data not shown). No effect of cyclopamine is detected before slow muscle formation begins: medial fast muscle appears normal in trunk somites up to stage 28 (data not shown). From stage 28 onwards, however, several defects are observed in posterior tail somites, where slow myogenesis is initiated in control embryos. Live cyclopamine-treated stage 36 embryos exhibit increased spontaneous movement and a slightly thinner tail tip when viewed from the dorsal surface (data not shown). Paralleling the loss of tail slow fibres, general muscle markers such as sarcomeric MyHC and 12/101, are also lost from tail tip, although tailbud outgrowth has not ceased (Fig. 5A). This result (1) shows that blockade of Hh signalling prevents terminal differentiation and/or survival of the slow fibres, rather than changing their character; and (2) confirms that slow fibres are the first to form in tail somites. Cyclopamine-treated embryos have a small but consistent reduction in the dorsoventral extent of musculature and a loss of chevron shape in tail somites (Fig. 5A). Defects are particularly marked in the medial somite(Table 1). Posterior muscle-containing somites beyond about somite 13 show an unusual segregation of a ventral cluster of fibres at stage 36 that parallels the altered XMyf5 expression pattern (Fig. 4C; Fig. 5A,arrowheads). In other words, zones of remaining muscle formation in cyclopamine-treated embryos still express XMyf5 mRNA. Slow muscle fails to form just where XMyf5 mRNA is missing. At stage 48, treated embryos have around 10 fewer somites containing muscle than controls, and the most posterior muscle-containing somites show a medial loss of muscle(Fig. 5C, Table 1). Thus,cyclopamine-treatment blocks initiation of slow muscle formation but has little if any effect on fast muscle formation. The Hh dependence of both XMyf5 mRNA and then slow MyHC expression in tail somites defines the first wave slow fibres.

In trunk somites, superficial slow muscle arises later than in tail somites, between stage 35-48 (Fig. 2). Examination of cyclopamine-treated embryos at stage 41 reveals that trunk slow muscle is formed, even though Hh signalling is inhibited(Fig. 5B,D). Thus, trunk slow fibre formation is Hh independent and this defines the second wave slow fibres. Despite the early lack of first wave slow fibres in tail somites,second wave slow fibre formation does occur in tail somites of cyclopamine-treated embryos (Fig. 5B,C). After cyclopamine treatment, anterior tail somites 15-30 lacked the first wave of slow muscle cells at stage 36, but by stage 41 slow fibres are present in these somites (compare Fig. 5A with 5B). First wave slow myogenesis initiates in the medial somite at the tail tip of untreated embryos and spreads outward dorsally and ventrally(Fig. 5A,B). In cyclopamine-treated embryos, by contrast, slow MyHC is initiated in the dorsal and ventral lips of the superficial myotome and appears to spread rapidly inwards to cover the surface of more anterior tail somites(Fig. 5B, arrows). Therefore,the late appearance of slow fibres is unlikely to arise from a delay in differentiation of first wave slow cells. A continuing defect in slow fibres is observed in the newly formed posterior tail somites at stage 41 and even more markedly at stage 48, indicating that effects of cyclopamine persist throughout tailbud outgrowth (Fig. 5B,C; Table 1). Our results suggest that cyclopamine ablates the first wave of slow fibres in the tail, but that a second wave of (Hh independent) slow fibre formation occurs in all somites, similar to that reported in zebrafish(Barresi et al., 2001). In trunk somites, this second wave generates the first slow fibres of the myotome, whereas in tail somites the second wave probably augments the slow fibres already formed around the time of somitogenesis by the first wave.

Trunk somites are not completely insensitive to cyclopamine. Although second wave slow fibres form normally, cyclopamine treatment reduces XMyoD expression in dorsal and ventral edges of trunk and anterior tail somites, whereas XMyf5 expression is normal at these locations(Fig. 4C). Subsequently, the somite is reduced dorsoventrally and there is a lack of what appears to be a transient population of slow fibres at the dorsomedial myotomal edge(Fig. 5C,D). In addition,ventral muscle fibres over the belly are aberrant, indicating disruption of ventral lip myogenesis. Section analysis of unmanipulated embryos reveals that strong XMyoD expression in the dorsal and ventral somite edges is associated with the dermomyotomal lips(Fig. 7E,F). Thus, Hh signalling is required for some aspects of trunk myogenesis.

To investigate the sources of second wave slow fibres formed in Xenopus somites, we prepared serial plastic sections of embryos at stages 22, 28 and 35 (Fig. 6A-F). Throughout the period, two epidermal cell layers surround the embryo as described (Nieuwkoop and Faber, 1967). At stage 22 and 28, the anterior 18 somites contain one striking specialisation, a layer of thin cells covering the lateral somite surface in some regions (Fig. 6A-C) (Blackshaw and Warner,1976; Hamilton,1969). As no slow fibres have yet formed in trunk somites, this layer is reminiscent of amniote dermomyotome. At stage 35, a single superficial monolayer of distinct cells is discerned both in trunk somites that lack first wave slow fibres and in tail somites that contain superficial first wave slow fibres (Fig. 6D-H). This distinct superficial cell layer is not slow muscle because it lacks MyHC and sarcomeric myofibrils(Fig. 6G,H; Fig. 7C-F). Pax3, a dermomyotome marker in amniotes (Bober et al., 1994; Goulding et al.,1994), is expressed in cells superficial to the differentiated muscle (Fig. 6I-K). Strikingly,however, expression is greater in trunk than in tail regions beyond about somite 12 from stage 29-34 (Fig. 6I,J). Later, at stage 37/38, Pax3 expression increases in posterior tail somites, possibly in parallel with dermomyotome formation(Fig. 6E,H,K). Like amniote Engrailed1 (Davidson et al.,1988; Davis et al.,1991), Xenopus En1 is also expressed at stage 33, level with the notochord in the outermost layer of trunk but not tail somites(Fig. 6L). We conclude that in trunk and anterior tail somites a cell layer, which hereafter we call dermomyotome, covers the superficial surface of the somite.

To examine the differentiation status of dermomyotome further, stage 35 embryos were double stained for MRFs and MyHC. XMyf5 and XMyoD are differentially expressed in Xenopus somites(Hopwood and Gurdon, 1990; Hopwood et al., 1991; Martin and Harland, 2001). Whereas XMyf5 is most highly expressed at dorsomedial and ventrolateral lips of the somite, XMyoD labels a dorsoventral chevron across each somite (Fig. 7A,B). As in amniotes, XMyf5 is strongly expressed at dorsomedial and ventrolateral lips of the dermomyotome(Fig. 7C,D). In addition, small groups of cells within the dermomyotome also express XMyf5,suggesting that myogenesis might be occurring at locations other than the lips(Fig. 7C, arrows). At a similar anteroposterior level, XMyoD mRNA accumulates only in the cells of the dermomyotome close to the dorsomedial lip and, unlike XMyf5, is also significantly expressed in MyHC-containing cells of the myotome proper(Fig. 7E,F). This expression suggests that, at the dorsomedial lip, myogenic cells in the surface layer initially express XMyf5, then accumulate XMyoD and finally terminally differentiate delaminating into the myotome proper, as occurs in amniotes. At the ventrolateral lip the structure usually appears distinct,with a ventral intense patch of XMyf5 expression flanked medially and laterally by more weakly stained cells that are still undifferentiated. At least in trunk somites, these regions probably correspond to migratory precursors (Martin and Harland,2001). Taken together, these data suggest the Xenopusdermomyotome contains several separate myogenic foci.

To further examine the fate of dermomyotomal cells in Xenopus, we analysed Col1a1 expression, which has been used to mark dermis(Goto et al., 2000). Xenopus Col1a1 is expressed in the somitic regions from stage 25 and widely throughout the dorsal body at later stages when dermomyotome is mature(Fig. 7K,L). Transverse sections reveal that the dermomyotomal layer expresses Col1a1, as does overlying tissue of the epidermis(Fig. 7M). Thus, the stage 35 Xenopus trunk dermomyotome shares many characteristics with amniote dermomyotome.

Comparison of MRF and MyHC expression in the tail bud provides insight into the early events in posterior muscle patterning when first wave slow and fast fibres types are generated. In Xenopus, the XMyf5 expression pattern is similar to that in fish (Fig. 4C, Fig. 7A)(Coutelle et al., 2001). The most posterior XMyf5 mRNA is abundant in a deep layer of cells adjacent to the notochord but is not detected in more superficial cells of the pre-somitic mesoderm. These XMyf5-expressing cells do not express MyHC (Fig. 7G). In anterior PSM, XMyf5-expressing cells no longer exclusively cluster around the notochord: some appear to span the somite, but XMyf5 is still not detected in the outermost layer of cells(Fig. 7H). Some XMyf5-expressing cells with mediolaterally elongated nuclei express PCNA, which often marks proliferating cells(Fig. 7I). Cyclopamine prevents this anterior PSM XMyf5 expression, but not tailbud expression(Fig. 4C). These data suggest that Hh signalling is required to maintain XMyf5 expression in medial cells, which then become orientated mediolaterally, simultaneously losing XMyf5.

As XMyf5 expression declines, XMyoD mRNA appears. The most posterior XMyoD expression is detected weakly in medial cells but within a few serial sections more anterior, is found exclusively superficially, consistent with loss of XMyf5 and accumulation of XMyoD during lateral migration(Fig. 7J; data not shown). Reduction of this most posterior XMyoD expression is seen in cyclopamine-treated embryos (Fig. 4C). In plastic sections of this region, cells can be seen elongated mediolaterally across the somite, reminiscent of XMyf5-expressing cells and the slow fibres in these somites [compare Fig. 2F (inset) 2J with Fig. 6C,F and Fig. 7H]. Further anterior, XMyoD mRNA is present in newly differentiated superficial muscle beneath the dermomyotome (e.g. Fig. 7F). It seems likely, therefore, that the most superficial layer of cells in the most posterior tail somites contains nascent differentiating slow fibres that express XMyoD, but that shortly thereafter a dermomyotome arises to overlie the slow muscle.

Our analysis of Xenopus myogenesis has revealed that (1) Xenopus early somites differentiate several distinct muscle fibre populations; (2) the common ancestor of tetrapods and teleosts developed first and second wave superficial slow and medial fast muscle fibre types in the manner first described in zebrafish – a pattern still extant in the Xenopus tail; (3) trunk myogenesis has undergone a striking modification during tetrapod evolution, involving a block on Hh-driven first wave slow myogenesis; (4) migration of early medial cells to the lateral somite surface may contribute to the apparent rotation of Xenopussomites and formation of dermomyotome; and (5) a dermomyotome similar to that of amniotes provides a source of myogenic cells during larval growth.

We have used antibody reagents to distinguish several fibre types in Xenopus. Abundant evidence shows that the superficial fibres are slow, including oxidative metabolism, early Hh dependence and labelling with BA-F8, a slow antibody in mammals. However, there is potential for confusion because BA-D5 and EB165 epitopes show a different spatial pattern in Xenopus to that observed in zebrafish embryos, where these antibodies also distinguish slow and fast fibres. In zebrafish, BA-D5 is first expressed in first wave slow fibres, located medially within the somite(Blagden et al., 1997). Subsequently, most of these BA-D5-reactive slow fibres migrate to a lateral and superficial position (Blagden et al.,1997; Devoto et al.,1996). In Xenopus, by contrast, the BA-D5 epitope is present in all/most larval muscle. Thus, the medial muscle that we have designated `fast' to fit with zebrafish nomenclature, may actually have a slow contractile rate. In amniotes, all early fibres express some slow MyHC gene,regardless of their later fate (Page et al., 1992). In zebrafish, the medial myotome differentiates into EB165-reactive fast muscle (Blagden et al.,1997). In Xenopus, by contrast, EB165 marks slow fibres. The simplest explanation is that in Xenopus two epitopes characteristic of MyHC isoforms have come to be expressed in different cell populations. These observations emphasise that evolution can rapidly alter MyHC fibre type, but that MyHC markers are, nevertheless, useful in conjunction with other functional and molecular data to distinguish cell types of different developmental origin within one species.

In Xenopus tail somites, slow fibres arise initially in contact with the medial surface adjacent to the notochord and then become located superficially, as the bulk of the somite differentiates into fast muscle. This is what happens throughout the body axis in zebrafish(Blagden et al., 1997; Devoto et al., 1996). Strong evidence for homology derives from (1) the mediolateral migration of slow precursors (see below); (2) the Hh-dependence of tailbud XMyf5 and XMyoD expression and slow fibre formation; and (3) the generation of extra slow fibres when Hh signalling is increased. Hh dependence is also a feature of the first wave of slow myogenesis in zebrafish(Barresi et al., 2000; Blagden et al., 1997; Coutelle et al., 2001; Du et al., 1997; Lewis et al., 1999; Norris et al., 2000). Similarly, in both species, formation of deep fast fibres and a second wave of superficial slow fibres is Hh independent(Fig. 8A, tail series)(Barresi et al., 2001; Blagden et al., 1997; Du et al., 1997). The striking similarities between zebrafish and Xenopus in the formation of slow and fast muscle at both cellular and molecular levels suggests that the common ancestor of Xenopus and zebrafish (i.e. of sarcopterygian and actinopterygian fish) developed muscle in the manner observed in Xenopus tails and throughout zebrafish. The widespread presence of a superficial slow muscle layer in agnathans and primitive jawed fish strongly suggests that primitive vertebrates had this organisation(Flood et al., 1977). So the direct ancestor of amniotes probably generated at least three waves of somite muscle fibres: early Hh-dependent first wave superficial slow, medial fast and later Hh-independent second wave slow.

Muscle cell diversity in Xenopus myotome may be even greater. Our data shows that cyclopamine blocks later somite growth throughout the axis accompanied by reduction in XMyoD and slow MyHC expression at the dorsal somite edge (Fig 4C, Fig. 5D). This raises the possibility that additional fibre population(s) may exist that are also Hh dependent, reminiscent of the Shh-dependent epaxial myogenesis in mice(Borycki et al., 1999). These cells are indicated as third wave slow cells in our model(Fig. 8A). However, in zebrafish the first slow wave is itself composed of two cell populations:engrailed-expressing medial muscle pioneer cells and migratory superficial slow fibres (Devoto et al.,1996). In Xenopus, we do not find medially located pioneer-like slow fibres. Although widespread Engrailed 1-like immunoreactivity was reported in tail somites(Davis et al., 1991), we observed no En1 mRNA in tail somites at similar stages. Although En2 mRNA was present in brain and head, it was absent from somites(B.L.M., unpublished). Thus, we found no evidence for sub-populations of first wave cells.

In Xenopus, first wave (i.e. early Hh dependent) slow muscle is only formed in the tail. Why is first wave slow muscle not formed in the trunk? Lack of Hh expression, secretion or responsiveness is unlikely because expression of Xshh and Xbhh are detected throughout the axis of stage 22-35 embryos (Ekker et al.,1995; Mariani et al.,2001; Stolow and Shi,1995). Moreover, Xptc genes, markers of Hh response, are upregulated flanking the notochord in trunk PSM dependent on Hh signalling(Koebernick et al., 2001; Takabatake et al., 2000)(Fig. 4B). Another possible reason for the missing first wave is that precocious fast muscle differentiation in Xenopus mesoderm occurs prior to Hh exposure,ensuring that the cells exposed to Hh are already committed to fast differentiation. However, we found that early Hh overexpression did not induce slow muscle in trunk somites, either at stage 22 or stage 35(Fig. 3) (A.G., unpublished). Nor did Shh overexpression ever completely converted tail somites to slow myogenesis, as can happen in zebrafish(Blagden et al., 1997). Although generation of fast and second wave slow fibres is similar in trunk and tail somites, it seems there is a block on Hh response that prevents Xshh and Xbhh from promoting differentiation or survival of first wave slow fibres in trunk somites(Fig. 8A, trunk series).

In both fish and Xenopus trunk, myf5 expression is highest close to the notochord, whereas myod expression persists further laterally (Chen et al.,2001; Coutelle et al.,2001; Polli and Amaya,2002; Pownall et al.,2002; Weinberg et al.,1996). Notochord ablation prevents medial trunk XMyf5expression but has little effect on lateral expression of MRFs. This finding suggests the presence of two distinct XMyf5-expressing myogenic cell populations in trunk, with the adaxial ones being dependent on signals from notochord. However, we found that cyclopamine does not modify adaxial trunk XMyf5 expression (M.E.P., unpublished), even though Hh is active anteriorly because Xptc genes are upregulated in the medial somite(Koebernick et al., 2001; Takabatake et al., 2000), and this Xptc expression is blocked by cyclopamine. So Hh signalling,although occurring, is not required for initial XMyf5 expression. In mouse and zebrafish, initiation of myf5 expression is also less sensitive to Hh removal than is myf5 maintenance(Asakura and Tapscott, 1998; Borycki et al., 1999; Coutelle et al., 2001; Kruger et al., 2001; Teboul et al., 2003; Zhang et al., 2001). In avian trunk, MRF regulation by Hh depends on other midline/ectoderm signals(Borycki et al., 2000). We suggest, therefore, that the normal function of Hh to maintain MRF expression in adaxial cells is blocked in Xenopus trunk somites, paralleling loss of first wave slow fibres. It is unclear whether differences in myogenesis between amniote trunk and tail are homologous to those in Xenopus.

Several pieces of evidence suggest that first wave slow cells migrate from adjacent to the notochord to the superficial somite surface as they terminally differentiate into slow fibres. First, MRF expression patterns suggest that medial presomitic mesoderm cells gain XMyoD as they lose XMyf5, move laterally and undergo terminal differentiation. Second, in tail tip regions slow MyHC and mediolaterally elongated nuclei are detected in single cells that span the somite from notochord to lateral surface. More anteriorly, slow fibres are located on the superficial somite surface and orientated anteroposteriorly, as are their nuclei. Third, blockade of Hh signalling leads to gap in tailbud XMyf5 expression followed by decreased XMyoD expression and absence of first wave slow muscle cells. Based on Xptc1 expression,Hh signalling acts medially (Koebernick et al., 2001). Taken together with the known migration of Hh-dependent first wave slow fibres in zebrafish(Blagden et al., 1997; Devoto et al., 1996), these data indicate that first wave slow fibres in Xenopus migrate to the somite surface around the time of somite formation. However, cell tracking in vivo would be required to prove this view.

The mediolaterally elongated nascent slow cells are reminiscent of the cells orientated perpendicular to the notochord in somite –I, the next somite to form (Hamilton,1969; Keller,2000; Youn and Malacinski,1981a). First wave slow cells re-orientate parallel to the notochord as each somite forms. Synchronously, underlying medial somite cells differentiate into fast muscle fibres and elongate in the same direction. Thus, the terminal differentiation of two waves of myogenic cells leads to a change from cells elongated perpendicular to the notochord to fibres orientated parallel to the notochord (Fig. 8B). Re-orientation of cells in successively older somites has been interpreted as demonstrating rotation of nascent Xenopus somites(Hamilton, 1969). Yet all cells do not rotate synchronously as a block(Youn and Malacinski, 1981b). Indeed, the morphology of `rotating' Xenopus somite cells and migrating zebrafish slow muscle is remarkably similar(Cortes et al., 2003; Youn and Malacinski, 1981b). Our data, therefore, raise the possibility that a fish-like cell migration coincident with somite formation accounts for many of the morphological changes, rather than rotation of the entire somite. Other anuran species do not show such a somite rotation (Keller,2000; Youn and Malacinski,1981a). In the light of our finding, we suggest that the previous interpretation of a wholesale rotation of Xenopus somites should be regarded with caution until in vivo cell tracking has demonstrated which cells move where in developing somites.

Much has been written concerning the evolution of paired appendages in the transition from fish to tetrapods, but less attention has been paid to evolution of the dramatic somitic modifications required for the move to land. Our findings focus attention on the evolution of the tetrapod trunk,particularly dermomyotome. Our data show that a superficial layer of slow muscle is probably the ancestral condition of the common ancestor of teleosts and anurans. Dermomyotome has not been described in teleosts, instead their myotomes grow by polarized hyperplasia, a process by which extra muscle fibres are generated in discrete superficial somitic zones, often at dorsal and ventral extremes. In zebrafish, these zones give rise to Hh-independent slow fibres (Barresi et al., 2001)and pectoral fin musculature (Neyt et al.,2000). In this paper, we show that Xenopus trunk somites do not form the first wave of slow fibres but develop a dermomyotomal layer shortly after their formation.

First wave slow fibre migration and re-orientation occurs in tail somites. If we are right that this cell re-orientation accounts for the seeming`rotation' of tail somites, then similar cell migrations probably explain the`rotation' of trunk somites (Hamilton,1969; Youn and Malacinski,1981b). Such movements may carry the notochord-dependent adaxial XMyf5-expressing cells to the lateral somite surface. Loss of contact with midline-derived signals may explain failure of maintenance of XMyf5 expression, as occurs in zebrafish notochord mutants(Coutelle et al., 2001). Whereas in the tail Hh drives XMyoD upregulation and slow fibre formation, in trunk such migratory cells may adopt another fate. Possible fates include fast muscle, myoblasts or dermomyotome. In chicken, most trunk myotomal cells arise from the medial half of nascent somites(Ordahl and Le Douarin, 1992),suggesting a medial origin of dermomyotome. If notochord-dependent adaxial cells form dermomyotome, they may not be committed to myogenesis despite XMyf5 mRNA expression. In mice, myf5-expressing cells in brain do not make muscle (Daubas et al.,2000). We suggest that at all axial somite levels medial XMyf5-expressing cells may migrate to lie on the superficial somite surface, where they may become exposed to other signals influencing their fate. Perhaps an altered response to Hh was a key evolutionary innovation permitting these migratory cells to pursue fates other than first wave slow muscle.

Regardless of its origin, a morphologically distinct cell layer appears on the superficial surface of Xenopus somites that we believe constitutes the dermomyotome. This MyHC-negative layer arises shortly after somitogenesis and expresses Pax3, Col1a1 and, in specific zones, XMyf5 and XMyoD. These zones correspond well with the reported sites of myogenesis in the amniote dermomyotome(Ordahl et al., 2001) and with the sites of polarized hyperplasia in fish(Fig. 8A). It is important to distinguish superficial slow fibres from dermomyotome. Indeed, histology and electron microscopy (Fig. 6)suggest that the outer layer of tail somites is initially slow muscle fibres. These rapidly become covered by a layer of small `spindly' cells, that coalesce into the dermomyotome overlying the slow fibre layer by stage 35. At the stages we examined, we found no evidence to support the earlier views that the dermomyotome is separated from the myotomal portion of the somite, nor that it forms a dermatome `curtain' draped over the myotomes of more than one somite (Blackshaw and Warner,1976; Hamilton,1969). Instead, we concur with the idea that dermomyotome is also segmented (Youn and Malacinski,1981b). As in the mouse(Davidson et al., 1988), in Xenopus trunk somites En1 mRNA is restricted to the medial region of the dermomyotome. We propose this layer is the evolutionary homologue of amniote dermomyotome, generating cells for myotome growth.

It is likely that proliferative cells at the dorsal and ventral somitic lips contribute cells to the dermomyotome, as occurs in birds(Ordahl et al., 2001)(Fig. 8A). In addition, the MRF-expressing lips probably yield the second wave of Hh-independent slow muscle fibres. In trunk somites, this `second wave' generates the first slow fibres. During metamorphosis the somites of the trunk increase in size and form numerous muscles of the back (Ryke,1953). Perhaps failure of generation of the earliest slow fibres is related to the evolution of a pool of cells within the dermomyotome adapted to building later muscle in tetrapods(Shimizu-Nishikawa et al.,2002).

We thank E. Bandman, J. Gurdon and J. Brockes for reagents; M. Walmsley and G. Fletcher for providing Xenopus embryos; D. Osborn for technical help; and C. Houart and P. Salinas for input. Funding was from MRC and EU(QLRT-2000-530, S.M.H.), and Italian Government travelling scholarships (A.G. and G.T.). B.L.M. is supported by an NIH grant GM42341 to Richard Harland.