Gene activation in the amphibian mesoderm

Amphibian embryos are big, develop in large numbers outside the mother, and can readily survive surgery. These properties have made them favoured experimental material for embryologists for over a century. Removing pieces of embryonic tissue for culture in isolation, or transplantation to another region of an embryo, was used to show that cell potency is progressively restricted in early development by a series of cellular interactions called inductions (Spemann, 1938; Holtfreter and Hamburger, 1955; Gurdon, 1987). The most celebrated induction is the direction by the mesoderm of neural differentiation from ectodermal cells that would otherwise have formed skin (Spemann and Mangold, 1924). Only comparatively recently was it discovered that the mesoderm itself develops in response to a previous induction (Nieuwkoop, 1969; Sudarwati and Nieuwkoop, 1971). Following the work of Nieuwkoop, we have gained in amphibia over the last two decades unparalleled knowledge of the origin and early development of the mesoderm (reviewed by Gerhart and Keller, 1986; Smith, 1989; Hopwood, 1990).

The mesoderm arises in the blastula, the hollow ball of cells that results from rapid cleavage of the egg cytoplasm following fertilization. When isolated and cultured, tissue from around the equator of the blastula differentiates into mesodermal derivatives in accordance with its normal fate. Expiants of tissue from the pigmented animal cap form (atypical) epidermis, and expiants of the yolky vegetal tissue from the opposite pole of the embryo form (poorly differentiated) endoderm. However, when animal and vegetal pieces are cultured in combination, mesoderm forms from the animal tissue (Nieuwkoop, 1969; Sudarwati and Nieuwkoop, 1971). This suggests that at least part of the mesoderm normally develops from animal cells at the equator of the blastula in response to induction by vegetal cells (Fig. 1).

A major obstacle to the study of amphibian development is that the long generation time (at least several months) makes impractical the extensive screens for mutants that have been so fruitfully applied to Drosophila melanogaster (Nüsslein-Volhard and Wieschaus, 1980; Nüsslein-Volhard et al. 1987; Anderson, 1987). However, since the discovery of the conservation of the homeobox between flies and frogs (Carrasco et al. 1984), the hybridization of Drosophila probes to Xenopus libraries under conditions of reduced stringency has allowed vertebrate embryologists into the well-stocked larder of the Drosophila geneticists. Several close and distant relatives of genes that regulate development in Drosophila have now been cloned from the genome of the frog, Xenopus laevis. Such genes can provide useful molecular markers for the study of a vertebrate embryo that is particularly accessible to analysis and consequently well understood at the cellular and tissue levels of organization. Xenopus embryos also allow relatively straightforward testing of the function of cloned genes. Examples of this approach are the cloning and tests of function of the homeobox-containing genes, Xhox-1A and Xhox3 (Harvey and Melton, 1988; Ruiz i Altaba and Melton, 1989).

We have used low stringency library screening to isolate two Xenopus genes, belonging to the helix-loop-helix family of DNA-binding proteins (Murre et al. 1989), that are expressed early in mesodermal differentiation. One is a frog homologue of MyoD (Davis et al. 1987). We find that the early embryonic expression of Xenopus MyoD (XMyoD) is consistent with a role in initiating muscle differentiation in normal development, and go on to show that it can indeed activate muscle-specific gene expression in embryonic cells. The other is a relative of the twist gene of Drosophila, which is required for mesodermal differentiation in flies (Simpson, 1983; Niisslein-Volhard et al. 1984). We show that this gene (Xtwi) is expressed widely in the early frog mesoderm, but not in the myotomes, where XMyoD is expressed.

MyoD is a mouse cDNA that, when transfected into cultured fibroblasts, can convert them into myoblasts (Davis et al. 1987). It encodes a helix-loop-helix DNA-binding protein that can interact with promoter elements required for the transcription of several muscle genes (Lassar et al. 1989). A Xenopus cDNA was cloned which encodes a protein that shares 73 % of its amino acid residues with MyoD (Hopwood et al. 1989a; Harvey, 1990; Scales el al. 1990). We have used probes made from this XMyoD cDNA to study the pattern of transcription of XMyoD in early embryos. This has been compared to that of the cardiac actin gene, until now the earliest marker of muscle differentiation in Xenopus (Mohun et al. 1984). We have asked, in particular, if XMyoD shows a pattern of expression consistent with a role in the normal activation of such genes as the cardiac actin gene.

The location of XMyoD transcripts has been determined by in situ hybridization, and by Northern blotting of RNA from dissected embryo parts. A cardiac actin probe labels the myotomal cells of the somites, from which the axial musculature will develop, and a XMyoD probe labels the same region (Fig. 2A,B). A stronger case for the specificity of XMyoD expression was made by Northern analysis of tissues dissected from a post-neurula embryo: XMyoD transcripts, like cardiac actin RNA, are restricted to the somite fraction (Fig. 2C). Later in development the cardiac actin gene is also expressed in the developing heart, but we have not detected XMyoD transcripts in this tissue (Hopwood et al. 1989a). (The eventual specificities of cardiac and skeletal isoforms of cr-actin for heart and skeletal muscle, respectively, are not established until adulthood [Mohun et al. 1984].) Thus, whilst the location of XMyoD transcripts is consistent with a role in muscle gene activation in the somites, another factor(s) must activate the same contractile protein genes in the early heart. All of the mammalian members of the family of myogenic factors related to MyoD are also skeletal muscle-specific (Sassoon et al. 1989; Braun et al. 1989; Rhodes and Konieczny, 1989; Braun et al. 1990).

Cardiac actin transcripts first accumulate in mid-gastrulae (stll of Nieuwkoop and Faber, 1967) (Cascio and Gurdon, 1986). XMyoD RNA is detectable at a very low concentration in eggs and early embryos, the number of transcripts first rising in early (st 10) gastrulaeto reach a peak in late neurulae, before declining in later development. By measuring absolute numbers of transcripts from the two genes, we found that the rise in the XMyoD transcript level genuinely preceded the appearance of cardiac actin RNA by one to two hours (Fig. 3). This would be enough time for XMyoD protein to be made, enter nuclei, and activate cardiac actin transcription, and for mature cardiac actin transcripts to accumulate. In contrast, in mouse embryos, myogenin and cardiac actin RNAs were first detected by in situ hybridization at 8.5 days post coitum, but MyoD transcripts were not detectable until two days later (Sassoon et al. 1989). It therefore seems unlikely that MyoD could activate the cardiac actin gene in the somites of mouse embryos. However, transcripts encoding a quail relative of MyoD, qmfl, first appear at about the same time as cardiac actin RNA (de la Brousse and Emerson, 1990). In the amniotes, therefore, the necessary temporal sequence of myogenic factor expression followed by the appearance of contractile protein mRNAs has yet to be established.

The foregoing description of XMyoD expression showed that XMyoD RNA is present in the right place at the right time to play a role in activating musclespecific gene expression in the somites. MyoD can activate muscle genes in a wide variety of cells in culture, although even in the successful cases the frequency of activation has often been low (Davis et a!. 1987; Weintraub et al. 1989; Schafer et al. 1990). We have asked if XMyoD can activate muscle genes in embryonic cells at the time when they are normally being allocated to the muscle lineage or to other pathways of differentiation. We expressed XMyoD ectopically by microinjection of synthetic mRNA into early embryos (Gurdon et al. 1974; Woodland et al. 1974; Andrews and Brown, 1987), a technique that has been used to create widespread high levels of several mRNAs which are normally spatially restricted (Harvey and Melton, 1988; Kintner, 1988; McMahon and Moon, 1989; Ruiz i Altaba and Melton, 1989; Detrick et al. 1990). A major concern in interpreting the results of experiments of this kind is to distinguish between specific and non-specific effects, the latter often including reproducible syndromes such as abnormal gastrulation leading to microcephaly. A stringent negative control is the injection of transcripts identical to the test RNA but containing a single point mutation that alters an amino acid known or predicted to be essential for function of the protein. A control XMyoD RNA was made by mutating a G residue to a C, which changes alanine 114 of the XMyoD protein to a proline (XMyoD114P RNA) (Fig. 4A). While minimally changing the primary structure of the RNA, this would be expected to disrupt the first helix of the helix-loop-helix domain, preventing dimerization and thus DNA binding (Davis et al. 1990).

We tested the effect of XMyoD RNA injection in animal cap cells (Hopwood and Gurdon, 1990), which normally give rise only to ectodermal derivatives. XMyoD or XMyoD114P RNA was injected into twocell embryos; the animal caps were separated at the late blastula stage (st9) and then cultured until after the time of normal activation of the cardiac actin gene in the somites. Northern blotting showed that this gene was activated in the animal caps from XMyoD-injected embryos, but not in controls injected with the mutant RNA (Fig. 4B). XMyoD could also activate the skeletal actin gene (data not shown) and desmin synthesis (N. D. Hopwood, A. Pluck and J. B. Gurdon, in preparation). It can thus bypass mesoderm-inducing factors and activate muscle-specific gene expression in animal cap cells. XMyoD is likely to act directly on the cardiac actin promoter. It has been shown that MyoD can bind to sites in the promoters of muscle genes (Lassar et al. 1989), and we note that, in addition to a CArG sequence that is essential for transcription (Mohun et al. 1989), there are potential XMyoD-binding sites located further upstream in the Xenopus cardiac actin promoter (M. V. Taylor, N. D. Hopwood, and T. J. Mohun, unpublished data).

We have shown that the cardiac actin gene was activated in animal caps of XMyoD-injected embryos to an extent similar to its normal activation in the myotomes of early neurulae (Hopwood and Gurdon, 1990). However, transcription of the cardiac actin gene was not sustained, the 12/101 muscle-specific antigen that is normally activated later in myogenesis (Kintner and Brockes, 1984) did not accumulate, the animal caps did not form muscle, and whole XMyoD-injected embryos also developed substantially normally. This could be because a high enough concentration of XMyoD protein was not maintained, as the injected RNA was progressively degraded. However, we did create a concentration of XMyoD RNA in the XMyoD-injected animal caps several-fold greater than that in normal myotomal cells of late neurulae, and note that the XMyoD RNA concentration normally falls significantly once somites have formed (Hopwood et al. 1989a).

A more likely explanation for the failure to differentiate muscle is that other specific factors are required to co-operate with XMyoD to activate the full myogenic programme. Other members of the MyoD family are potential co-operators, but since these myogenic factors can activate each other’s synthesis, it is possible that their expression would in fact be activated in XMyoD-injected animal caps. If so, they are clearly insufficient to cause muscle differentiation. Two other factors have been shown to be able to convert cultured fibroblasts into myoblasts, but it is not yet known if they are involved in normal myogenesis: the uncharacterized genomic locus, myd (Pinney et al. 1988) and the oncogene, v-ski (Colmenares and Stavnezer, 1989). A possible role for another specific factor(s) has been suggested by the recent discovery of Id, a helix-loop-helix protein which lacks the basic region that is believed to mediate DNA-binding in other members of the family (Benezra et al. 1990). A protein such as Id could sequester in non-functional complexes either XMyoD itself, or a necessary partner of XMyoD, and so be a buffer against the effects of XMyoD in animal cap cells. Myogenesis might require the inactivation of negative regulators like Id.

We conclude that XMyoD is sufficient, together with components present in animal cap cells, to activate expression of several muscle genes, but not to cause complete myogenesis. The significance of this experiment is that it shows the ability of XMyoD to activate muscle genes in normal embryonic cells, at the time when muscle differentiation normally begins. The expression of cDNA clones in animal cap cells by microinjection of synthetic mRNAs, alone or in combination with XMyoD, should provide an effective assay for additional factors involved in myogenesis.

The twist gene of Drosophila is required for mesodermal differentiation: twist mutant embryos lack mesoderm and internal organs (Simpson, 1983; Niisslein-Volhard et al. 1984). Twist gene products accumulate in mesodermal (and some endodermal) cells as an early zygotic response to the activity of the dorsal group of maternal effect genes (Thisse et al. 1987, 1988). The twist protein is a member, like MyoD, of the helix-loop-helix family of DNA-binding proteins (Thisse et al. 1988; Murre et al. 1989).

We have used a probe made from the twist gene to isolate a related Xenopus cDNA, called Xtwi (Hopwood et al. 19896). It encodes a protein of predicted M_T 18000, about one-third the size of twist, and related to its C-terminal region. The twist and Xtwi proteins are very similar in the putative DNA binding and dimerization domain and also share 11 out of a stretch of 14 amino acids near their C-termini; the rest of the proteins are substantially different. In view of the specific expression of Xtwi in the early mesoderm (see below), we believe that twist and Xtwi are probably homologues, although the possibilities cannot be excluded either that the two genes evolved independently, or that there is another Xenopus gene more similar to twist than Xtwi.

The expression of a gene related to twist in the mesoderm of a vertebrate encourages us to seek further similarities in the mesodermal differentiation of the two groups. It is already known that the dorsal protein, which is believed to activate the twist gene, shares a large domain with the avian proto-oncogene, c-rel (Steward, 1987). Transcripts that encode proteins related to dorsal are present in Xenopus embryos at the time of Xtwi gene activation (K. R. Kao and N. D. Hopwood, unpublished data). A putative Xenopus homologue of the snail gene (Boulay et al. 1987), which, like twist, is activated in the Drosophila zygote by dorsal and required for development of ventral structures, is also expressed in early frog embryos (Sargent and Bennett, 1990).

Xtwi transcripts first accumulate in early gastrulae about an hour later than zygotically transcribed XMyoD RNA appears (Hopwood et al. 1989b). We have used in situ hybridization and Northern blotting of dissected tissues to show that in early embryos Xtwi transcripts are present only in the mesoderm. Later in development, the Xtwi gene is also expressed in the neural crest, a transient structure in vertebrate embryos that gives rise to a wide variety of tissues, including the peripheral nervous system, adrenal medulla, pigment cells, and much of the muscle and connective tissue of the face and neck (reviewed by Le Douarin, 1982). Xtwi transcripts are restricted to the anterior mesoderm of early neurulae, presumably reflecting the earlier differentiation of anterior structures. Only by the tailbud stage are they found in the most posterior regions.

The three major divisions of the early mesoderm are the notochord, or embryonic backbone, the somites, and the lateral plate, the last of which gives rise, among a variety of other structures, to the heart, and smooth muscle of the gut. The somites consist of three parts: the myotomes, from which axial muscles develop; the sclerotome, which gives rise to skeleton; and the dermatome, which contributes the connective tissue of the skin. The Xtwi gene is expressed in the notochord (anterior half of the embryo) and lateral plate mesoderm (anterior quarter) of early (stl4) neurulae (Fig. 5). There is also Xtwi RNA anterior to the somites in what may be prechordal mesoderm. However, we have detected no Xtwi transcripts in the somitic cells, which express muscle markers such as cardiac actin and XMyoD. This lack of Xtwi expression in the myotomes contrasts to its otherwise widespread expression in the differentiating mesoderm and is the most striking feature of its pattern of transcript accumulation. Although we do not yet know the location of Xtwi transcripts in gastrulae, it is unlikely that they are present in the prospective somites of gastrulating embryos. This is because the anterior-posterior progression in the onset of Xtwi expression allows us to examine the more posterior parts of the embryo as they begin to express Xtwi; in doing so we have never seen any Xtwi transcripts in the myotomes. Indeed, myotomal differentiation, as indicated by transcription of the cardiac actin gene, has been established throughout the somites of the trunk many hours before Xtwi RNA is first detectable in the posterior half of the embryo (Hopwood et al. 1989(7). This suggests that Xtwi is unlikely to function primarily to define mesodermal cells as non-myotomal, since myotomal differentiation appears, as judged by the expression of myogenic factors and contractile proteins, to be specified positively at an earlier stage in the development of the mesoderm.

By the late tailbud stage (st28-30), Xtwi transcripts are still detectable in the notochord, particularly in posterior regions, and in anterior and posterior lateral plate mesoderm (Fig. 7). We have not seen, at the stages examined (up to st32), any Xtwi expression in lateral plate cells in the middle of the embryo. Instead, there is strong expression around the notochord, which probably corresponds to mesenchymal cells derived from the sclerotomes. Expression becomes more ventral and lateral posteriorly. Since Xtwi is itself the most general marker of non-muscle mesoderm in Xenopus, and no specific, early markers are available for mesoderm outside the myotomes and notochord, it is difficult to assess either the degree of generality or the meaning of the specific and evolving pattern of Xtwi expression. However, it is possible that all non-myotomal mesodermal cells express the gene at some time, the pattern of expression reflecting the order of mesodermal cell differentiation.

We have previously described Xtwi expression in the anterior (cephalic) neural crest from mid-neurula stages onwards (Hopwood et al. 1989b). Fig. 6 shows strong labelling of anterior neural crest cells in a post-neurula embryo (st21). Having now examined slightly later embryos, we can report that it is also expressed in the neural crest along the length of the trunk (Fig. 7). It is difficult, in the absence of other early neural crest markers in the frog, to know what derivatives these Xtwi-expressing cells will form. However, we have not seen Xtwi transcripts in the cephalic ganglia, some of which are derived from the neural crest.

As expected from the mesodermal location of XMyoD and Xtwi transcripts, these genes are activated following mesoderm induction. We carried out an experiment based on Nieuwkoop’s (see Introduction), in which blastulae were dissected into animal, equatorial, and vegetal parts, which were cultured separately until sibling embryos became late (stl2i) gastrulae. Animal and vegetal pieces were also cultured in combination for the same period. No XMyoD or Xtwi transcripts were present in the animal or vegetal pieces cultured alone (the maternal contribution of XMyoD RNA having been degraded), but transcription of XMyoD and Xtwi genes was activated in the animal:vegetal conjugates (Hopwood et al. 1989a; Fig. 8A). Taken together with knowledge of the location of XMyoD and Xtwi transcripts in normal development, this shows that the genes are activated following mesoderm induction.

The neural crest arises at the lateral margins of the neural plate as a consequence of its induction by the mesoderm. In order to test the responsiveness of the Xtwi gene to neural induction, tissue recombinates were made between inducing gastrula mesoderm and responsive gastrula ectoderm. After culture until sibling embryos reached the tailbud stage, the mesoderm was removed from the conjugates. The induced ectoderm, but not control uninduced ectoderm, contained Xtwi transcripts (Fig. 8B), which indicates that Xtwi expression is activated in the neural crest following neural induction. First commencing in mid-neurulae (stl4–15), transcription of the Xtwi gene is the earliest specific activity of the amphibian neural crest so far described. We conclude that Xtwi is activated in response to two successive early inductions.

We thank Anne Pluck for in situ hybridizations, Juanita Roche for comments on the manuscript, and the Cancer Research Campaign and the Royal Society for support.