A. Cellular Roles in Early Myogenesis
To a first level of resolution, the questions of how muscles are formed in C. elegans can be addressed by examination of the cell lineage (Sulston and Horvitz 1977; Deppe et al. 1978; Kimble and Hirsh 1979; Sulston and White 1980; Sulston et al. 1983). The idea that all muscles might derive from clonal myogenic commitment of a few early blastomeres was appealing to earlier investigators, who had determined that muscle is predominantly derived from only two of the four-cell-stage blastomeres (Boveri 1888). However, a more detailed examination of the lineages makes a simple early-clonal-commitment model untenable. This is most dramatically seen by the presence of a single body-wall muscle derived from the AB lineage (Figs. 13 and 14) (Sulston et al. 1983). Although it is the only body-wall muscle to derive from this lineage, the differentiated cell exhibits no evident distinctions from its distant cousins that form body-wall muscles from the P1 lineages. Body-wall muscles likewise arise in a piecemeal way from the MS lineage, with clonal groups of one to four body-wall muscles interspersed in the lineage tree with branches yielding a variety of nonmuscle cells. Derivation of body-wall muscles from the C and D lineages exhibits a much less complex pattern: Two granddaughters of the C cell give rise to simple clones of muscle, while D gives rise precisely to 20 muscle cell descendants. The different muscle lineage patterns suggest that several different pathways might be responsible for the initiation of myogenesis in body-wall muscle (Fig. 14).
Pharyngeal muscles, derived from AB and MS, are produced by a piecemeal pattern in a manner similar to that of the body-wall muscle lineages derived from these two founder cells (Sulston et al. 1983). The pharyngeal muscles are, however, not closely related to body-wall muscle lineages but are instead interspersed with lineages producing nonmuscle pharyngeal cells. The remaining embryonic muscles (the four intestine-associated [nonstriated] body muscles) are also produced from AB and MS; these are closely related in lineage to striated body-wall muscles.
B. Autonomy and Nonautonomy in Muscle Cell Commitment
The invariance of the cell lineage does not necessarily imply a noninteractive commitment of early embryonic cells to specific patterns of descendants. A variety of cell ablation and isolation experiments have suggested that interactions of potential myogenic precursors with neighboring cells have a key role in determining the pattern of cells that eventually differentiate into muscle (Priess and Thomson 1987; Wood 1991; Hutter and Schnabel 1994; Schnabel 1994,1995). These studies are still in progress (see Schnabel and Priess, this volume). From the viewpoint of muscle patterning, the most striking feature of the characterized cell interactions is that they occur very early in the affected cell lineages, several cell divisions before any overt differentiation of muscle cells is observed in the embryo (Priess and Thomson 1987; Schnabel 1995). Thus, whatever information is imparted through the cell interactions must be stably maintained through several cell divisions before eventually becoming manifest in expression of muscle-specific differentiation products.
Given the large number of early cell interactions that influence later commitment to muscle cell fates, it is valid to ask if there is any point at which myogenic commitment has become an intrinsic feature of specific cells. In the case of the D blastomere, it has been possible to address this question with a conceptually simple isolation experiment. Starting from the first embryonic division, the cells not in the lineage leading to D can be ablated just as they are born (Schnabel 1995). This leaves an isolated D cell, surrounded by debris resulting from ablation of other embryonic cells. The resulting partial embryo yields approximately 20 cells that express components indicative of body-wall muscle differentiation. This experiment strongly suggests that some type of intrinsic commitment is indeed occurring in the early embryo.
The invariant cell lineage could be taken as suggesting that the myogenic differentiation program might be switched on by a mechanism that counts cell cycles. Mutations blocking cell division (Gossett et al. 1982) and pharmacological agents (Cowan and McIntosh 1985) have been used to examine embryos in which cell cleavage arrests while other developmental events (including DNA synthesis) proceed. Both sets of manipulations resulted in cleavage-arrested embryos (2–90 cells total) containing specific subsets of cells that had initiated the myogenic program, as assayed by production of muscle filament components. It thus appears that a myogenic commitment program can proceed in the absence of a fully operational cell cycle. These experiments do not rule out the possibility that some aspect of the cell cycle might continue in cleavage-blocked embryos and have a role in myogenic commitment.
Experiments by Edgar and McGhee (1988) have addressed the ability of muscle differentiation to occur in cases in which DNA replication (instead of cell cleavage) has been blocked. These experiments demonstrated that muscle differentiation can occur without the last several rounds of DNA synthesis. This argues against a model in which replication cycles are counted to precisely time the onset of muscle differentiation. Edgar and McGhee observed that earlier blockage of DNA replication (two to three divisions before final differentiation) could inhibit muscle differentiation. This was in contrast to their observations with gut differentiation markers and suggested that some feature(s) dependent on S phase of the cell cycle might have a role in executing the process of myogenic commitment.
C. Genes Responsible for Muscle Patterning
Genetic and biochemical analyses have given us some understanding of the on events at the two temporal extremes of the muscle-patterning process: the early events preceding myogenic commitment and terminal events that give rise to differentiated muscle. The middle stages in this process are still a mystery and are the focus of research using a variety of experimental approaches.
1. The Early Body Plan
The establishment of the early blastomere identities is in large part controlled by maternally encoded factors, which have initial roles in determining patterns for both muscle and nonmuscle tissues. For a more extensive discussion of these genes, see Schnabel and Priess and Kemphues and Strome (both this volume). The influences of these genes on muscle can be summarized as follows:
The earliest determinative events in the embryo may be the formation of the asymmetric anterior-posterior axis (see Kemphues and Strome, this volume). As muscle is predominantly derived from the posterior daughter at the first cleavage (P1), it is not surprising that genes responsible for early anterior-posterior axis formation (par genes) also affect the pattern of muscle cells produced.
Many of the cell-cell interactions in the early embryo (which both restrict and enable subsequent muscle-containing lineages) require the GLP-1 signaling pathway (see Schnabel and Priess, this volume). Thus, mutations in glp-1 lead to both missing muscles and ectopic muscle. Some of the other components required for this signaling pathway have also been identified, including one of several proposed GLP-1 ligands, APX-1, which functions in interactions between P2 and ABp (Mango et al. 1994b; Mello et al. 1994), and lag-1 , which may act downstream from glp-1 (Lambie and Kimble 1991). There is as yet no working hypothesis of how glp-1 function might result several divisions later in modifications to the pattern of differentiated muscle and nonmuscle cells.
The myogenic potential of the MS and P2 lineages depends on separate programs that appear to set the initial identities of these blastomeres (Mello et al. 1992). Genetic analysis has implicated the gene skn-1 in setting the identity of MS (Bowerman et al. 1992a, 1993; Mello et al. 1992), with pal-1 setting the identity of P2-derived muscle precursors (C.P. Hunter and C. Kenyon, pers. comm.). Both genes produce maternally encoded transcription factors.
2. The Terminal Differentiation Program
The structural proteins described in this chapter are the end products of the muscle differentiation pathway. As nucleic acid and antibody probes for these components are generated, it becomes possible to follow the onset of transcription and translation of the corresponding genes. The four muscle myosin heavy-chain genes were the first components for which such probes were available (MacLeod et al. 1981; Karn et al. 1983; Dibb et al. 1989), and thus they have been the most extensively analyzed. Myosin protein accumulation in body-wall muscle cells begins just before the terminal division of myogenic precursors (Epstein et al. 1993; Hresko et al. 1994). If this is indeed the earliest expression of muscle-specific differentiation markers, it would contrast with the clonal gut lineage, for which expression of differentiated products can begin up to three divisions before cessation of cell division. In situ hybridization experiments demonstrated that expression of myosin genes is controlled at the level of mRNA accumulation (Evans et al. 1994; Seydoux and Fire 1994). Quantitative assays for mRNA and protein levels were consistent with this conclusion and with primary control at the level of transcription; the latter assays also suggested a modest posttranscriptional regulation modulating relative levels of the MHC isoforms (Honda and Epstein 1990).
An intensive genetic analysis of the major MHC isoform (encoded by unc-54 ) was facilitated by the fact that null mutations are viable and have an easily identified paralyzed phenotype (Brenner 1974; Anderson and Brenner 1984). It was hoped initially that the large number of mutant alleles would provide both structural and regulatory mutations and that the regulatory mutations might point to specific modes of gene control. Surprisingly, none of the unc-54 mutations that have been characterized appear to affect the regulation of the gene. Of more than 75 sequenced mutations, all but three affect the coding region by either changing critical amino acids, introducing stop codons, or affecting splice junctions (Dibb et al. 1985; Eide and Anderson 1985a,b; Bejsovec and Anderson 1990; D. Moerman, J. Kiff, and R.H. Waterston, unpubl.). Two alleles appear to be 5′duplications which are intriguing but do not immediately shed light on unc-54 regulation (Eide and Anderson 1985c). The remaining allele is a deletion in the 3′UTR which leads to RNA degradation by the SMG system (Eide and Anderson 1985a; Pulak and Anderson 1993; see Anderson and Kimble, this volume). The deleted sequences do not appear to have an essential role in determining the pattern of unc-54 regulation, since the 3′deletion allele is fully functional and properly regulated in a Smg– genetic background (Okkema et al. 1993).
The analysis of cis-acting sequences was greatly bolstered by the development of a DNA transformation system which allowed reporter constructs (lacZ fusions) to be rapidly assayed for expression pattern (Fire et al. 1990). Using such assays, Okkema et al. (1993) found that each of the MHC genes contains several separated elements, each sufficient to direct muscle-type-specific expression. All of the elements identified were positively acting enhancer or promoter elements: There was no evidence for negative regulation mediated through specific promoter or enhancer elements. Given the multiplicity of positively acting regulatory sites for each gene, the failure in the original genetic screens to recover individual point mutations with large effects on expression is not surprising.
Several of the cis-acting elements responsible for MHC gene activation have been dissected in more detail, with the results suggesting combinatorial modes of action. Evidence for a combinatorial “AND” function comes from a strong enhancer within the unc-54 third intron (Jantsch-Plunger and Fire 1994). This enhancer contains a set of four separable subelements, at least one of which has a broader specificity (body-wall muscle plus body hypodermis) than the complete enhancer (body-wall muscle only). Dissection of the myo-2 upstream enhancer revealed both combinatorial “AND” and “OR” functions. This enhancer carries at least one element active throughout the pharynx (in muscle and nonmuscle tissue), as well as an element acting in only a subset of pharyngeal muscles (Okkema and Fire 1994).
It seems likely that the mechanisms regulating MHC gene promoter/ enhancer activities will be used for coordinated activation of a larger set of muscle filament components. Analysis of the tropomyosin gene tmy-1 (Kagawa et al. 1995) provides a similar example of type-specific regulation (body-wall vs. pharynx) by promoter activation while indicating that additional complexity in regulation can be generated by differential splicing. Studies of basement membrane collagen and perlecan expression have similarly revealed a variety of differentially spliced mRNAs with distinct regulation, including both muscle-specific and more generally expressed forms (see Kramer, this volume).
D. Diversity within and between Muscle Cell Classes
Although all body-wall muscles have the same overall shape and organization, these cells are distinguishable by several features. First, as noted above, the cells derive from several different branches of the lineage tree. Second, cells have distinctive and fixed anterior-posterior (as well as dorsal-ventral) positions. In particular, the cells at the front and rear margins of the muscle quadrants might need special attachment properties in order to maintain overall integrity of the animal. Third, body-wall muscle cells in different parts of the animal have different connectivity patterns with the nervous system (White et al. 1986). These similarities and differences highlight the question of the extent of diversity in the body-wall muscle gene expression program.
At one extreme, body-wall muscle cells from different lineages (or in different parts of the body) might use completely different sets of regulatory factors in carrying out differentiation. Alternatively, it was conceivable that all body-wall striated muscles express a completely uniform differentiation program and that each is equivalent in gene expression (with any differences due to cellular context and/or posttranscriptional events). These extreme models both appear unlikely. The analysis of cis-acting sequences regulating body-wall MHC expression suggested at least some common elements to gene expression in the different cells, since each of the MHC promoter and enhancer elements (both wild-type and mutant) acts uniformly in all body-wall muscles (Okkema et al. 1993; Jantsch-Plunger and Fire 1994). Although MHC regulation may be uniform, indications of nonhomogeneity in body-wall muscle gene expression come from a variety of analyses. In a pilot screen for enhancers that act nonhomogeneously in body-wall muscles, two such elements were found: one with preferential expression in anterior body-wall muscles and a second with preferential expression posterior (A. Fire and S. Xu, unpubl.). In addition, expression of the homeotic selector gene mab-5 has been observed in a subset of posterior muscles (D. Cowing and C. Kenyon, pers. comm.; also see Wang et al. 1993). Although the differential activity of the selector genes and the two characterized position-dependent body-wall muscle enhancers demonstrate a mechanism for expressing genes nonhomogeneously in this tissue, there is not yet any functional indication of how such a mechanism might be used by the animal.
The nonstriated single sarcomere muscles (pharyngeal, vulval, uterine, and intestine-associated) have wide variation in their cellular morphology. The nonpharyngeal single-sarcomere muscles ("minor muscles") express many of the same differentiated components as body-wall muscles, including the same set of MHC isoforms (Ardizzi and Epstein 1987). Nevertheless, differences between striated and nonstriated muscle cells are readily observed in the timing and eventual levels of structural protein accumulation. This is exemplified by preferential expression of different unc-52 splice variants in the different muscle classes (G. Mullen and D.G. Moerman, unpubl.) and by differential activity of promoter and enhancer elements from the unc-54 gene (Okkema et al. 1993). Pharyngeal muscles are likewise a diverse group (see Avery and Thomas, this volume), having in common perhaps only their pharyngeal location and expression of the MHC isoforms encoded by myo-1 and myo-2 . One functional subgrouping within pharyngeal muscle is suggested by two independent experiments: embryonic staining with monoclonal antibody 3NB12, and activity assays for the “B” subelement of the myo-2 enhancer. Both experiments define a subset of pharyngeal muscles that include cells m3, m4, m5, m7 but strikingly excludes the large muscle cell m6 (see Fig. 15) (Okamoto and Thomson 1985; Priess and Thomson 1987; Okkema and Fire 1994).
Several ongoing lines of research are suggesting that gene expression in nematode nonstriated muscles might have an even richer complexity than previously imagined. Lynch et al. (1995) have recently produced lacZ fusions to many genes defined solely from the genome sequencing project and have reported promoter regions active in a variety of subsets of muscle cells, including subsets of the vulval and pharyngeal muscles. M. Krause and L. Avery (pers. comm.) have examined a striking restriction in the distribution of the HLH-2 protein, which is observed in muscle nuclei of the m5 class, but only in a subset of these nuclei. Two distinct and very specific enhancer elements have been found upstream of the C. elegans gene ceh-24 . One is active only in the lone m8 pharyngeal muscle, and the other is active only in vulval muscles (B. Harfe and A. Fire, unpubl.).
E. A Mystery: Acquisition of Muscle Cell Fates in the Mid-stage Embryo
In principle, it should be possible to find the mechanistic links between the early events establishing blastomere identity and much later elucidation of the differentiated muscle pattern. A half dozen cell cleavages occur in the intervening period of embryogenesis, with the onset of extensive overt differentiation coinciding with the last cleavage. A detailed understanding of this “mystery” phase of embryogenesis will require the identification of genes and gene products that are differentially distributed or differentially active during late proliferation. This will potentially be informative (1) as markers for intermediate stages in cell-type-specific commitment and (2) as candidates for molecules with functional roles in the process.
Several types of screens are being undertaken to identify myogenic genes active in the mid-stage embryo. Working from control sequences known to be responsible for structural protein gene expression, it should be possible to isolate regulatory factors that provide successive steps backward through the regulatory pathway. This has been done successfully in the case of the pharyngeal-muscle-specific “B” element of the myo-2 enhancer. A homeodomain protein, CEH-22, was identified by its ability to bind to that element and was subsequently shown to be expressed in precisely the cells in which the “B” element is active (Okkema and Fire 1994). CEH-22 is a member of the “NK2” family of homeodomain factors (see McGhee and Krause, this volume). In several organisms so far examined, multiple members of this family have been found (Drosophila, Kim and Nierenberg 1990; Planarians, Garcia-Fernandez et al. 1991; vertebrates, Price et al. 1992; C. elegans, B. Harfe and A. Fire, unpubl.). These components have been implicated in mesodermal and ectodermal specification and differentiation, including most intriguingly the murine NKx2.5 factor, which has been implicated in cardiac differentiation (Lyons et al. 1995). The role of CEH-22 in pharyngeal development has been addressed by genetic analysis and misexpression studies. Loss-of-function mutations in the ceh-22 gene result in loss of “B” element activity, whereas ectopic expression of CEH-22 in body-wall muscles can result in ectopic activity of the endogenous myo-2 gene (P. Okkema and A. Fire, unpubl.). ceh-22 expression is activated somewhat earlier than myosin gene expression, but it is limited to pharyngeal muscle precursors (Okkema and Fire 1994). Hence, although a step has been taken back through the series of regulatory events, more such steps will be essential before the process is completely charted.
At the same time, a more classical genetic approach has been taken to look for genes that are required for muscle differentiation and commitment. Mutations that fully disrupt these processes with no detectable differentiation of the affected muscle class would be expected to have lethal phenotypes (Waterston 1989). If muscle development is controlled by a consecutive series of differentiation factors, each dependent on previous factors for its expression, then one might expect to find a large number of such mutations. Lethal screens have been carried out for mutations affecting body-wall muscle differentiation (see, e.g., Williams and Waterston 1994), but no point mutations that specifically eliminate expression of muscle-specific differentiation markers have been isolated. Because the precise arrest phenotype to be expected from such a mutation is not clear, these screens focused on mutations that arrest as paralyzed twofold-elongated embryos (characteristic of complete loss of function for the contractile apparatus; Waterston 1989). A more general screen was carried out using deficiency mutations, assaying only for the production of terminally differentiated muscle components (Ahnn and Fire 1994). The surprise of the deficiency screens has been that very few loci in the zygotic genome are required uniquely for muscle differentiation. One candidate locus (yet to be identified on a molecular level) was identified in screening approximately 80% of the genome. These studies suggest extensive maternal contribution to later embryonic development, widespread redundancy in the commitment/differentiation program, or both.
Muscles associated with the reproductive system develop post-embryonically. The specific role of vulval muscles in egg laying allows them to be the object of a straightforward type of genetic screen. Animals without vulval muscle function fail to lay eggs, leading to a “bag of worms” phenotype in which the young worms hatch inside of the parent. Many mutants that affect the functioning of body-wall muscles (e.g., unc-54 null mutations) have similar effects on the vulval muscles, and thus exhibit egg-laying defects (Trent et. al. 1983). Mutations affecting vulval muscle function can be enriched for by prescreening for animals that fail to lay eggs and thus retain their eggs internally. A secondary screen can then be carried out to identify specific defects in vulval muscle (M. Stern and H.R. Horvitz, pers. comm.). The resulting genes are designated “sem” for sex muscle abnormality (some are called “egl” for egg-laying-defective). From these screens (Stern and Horvitz 1991; Garriga and Stern 1994; M. Stern, pers. comm.), mutations have been isolated that affect the generation and cell fate of sex muscle precursors (sex myoblasts) during larval development ( egl-31 , sem-1 , sem-4 ), the divisions of sex myoblasts ( sem-1 ), sex myoblast migrations ( egl-15 , egl-17 ; Stern and Horvitz 1991), and the attachment orientation of the vulval muscles ( unc-53 ; T. Bogaert, pers. comm.). The sem-2 and sem-4 genes have been proposed to cause presumptive sex myoblasts to adopt characteristics of body-wall muscle fate (M. Stern, pers. comm.). sem-4 encodes a zinc-finger-containing factor that has roles in a variety of neural tissues in addition to muscle (Basson and Horvitz 1996). The sem-2 locus may likewise be more general than just sex myoblast differentiation, since lethal mutations have been isolated that exhibit defective anterior body-wall muscles (C. Colledge and R.H. Waterston, pers. comm.).
Genetic screens for components affecting pharyngeal muscle development have likewise begun (Schnabel and Schnabel 1990; Avery 1993a; Mango et al. 1994a). So far, two mutations from these screens are of particular interest from the perspective of commitment. Both of these define zygotically acting genes required for all tissues within the pharynx. pha-1 encodes a putative DNA-binding protein required for complete differentiation of both muscle and nonmuscle tissue (Granato et al. 1994). The mutant animals produce only a rudimentary pharynx. pha-1 is evidently not required for pharyngeal muscle cell commitment, since the mutant embryos can still express ceh-22 and low levels of pharyngeal myosin (P. Okkema, pers. comm.). pha-4 mutants have a much more dramatic effect on pharyngeal development. Mutants lack an evident pharynx and all known pharyngeal markers, both muscle and nonmuscle (Mango et al. 1994a). It will be of interest to determine the connection among PHA-1, PHA-4, and the pharynx-specific (muscle+nonmuscle) enhancer subelements such as that upstream of myo-2 (Okkema and Fire 1994).
A second type of functional screen seeks factors that can promote muscle differentiation when expressed ectopically. Transformation technology in C. elegans is inadequate to implement such a screen solely in C. elegans: As an alternative, Krause and colleagues have been isolating homologs of two classes of vertebrate myogenic factors that had been shown to promote muscle differentiation in tissue culture cells. The most extensively characterized family of vertebrate myogenic factors are the MyoD subfamily of helix-loop-helix (HLH) transcription factors. C. elegans apparently contains only a single member of this HLH subfamily (Krause et al. 1990; Chen et al. 1992). This factor (designated HLH-1) has a striking expression pattern, with expression activated in mid-stage embryonic cells whose clonal descendants will give rise only to striated muscles. This suggests that a myogenic program has indeed been activated in these cells and nominates HLH-1 as a candidate execution factor for such a program. Mutational analysis indicates that HLH-1 is not the only execution factor in this program. Striated muscle differentiation proceeds in mutant animals lacking HLH-1, with the number and overall grouping of the resulting muscle cells apparently normal (Chen et al. 1992, 1994). The resulting muscles contract only poorly, so that the animals are inviable.
The analysis of hlh-1 function in C. elegans contrasts somewhat with similar studies in mice, where a double knockout of two myoD family members results in loss of muscle tissue (Rudnicki et al. 1993). The apparent paradox between the two systems is magnified in that the C. elegans factor can function in vertebrate tissue culture cells to promote muscle differentiation (Krause et al. 1992), and at least one of the vertebrate family members (myogenin) can substitute for C. elegans hlh-1 (Chen et al. 1994). This paradox might reflect a fundamental difference in the primary function in the two species or alternatively could reflect distinct cellular consequences (cells absent vs. abnormal) resulting from a similar primary defect in the muscle differentiation program.
A second family of vertebrate myogenic factors, the “MEF2” family of MADS box factors (Yu et al. 1992), appears also to have a single C. elegans homolog (M. Park and M. Krause, pers. comm.; see McGhee and Krause, this volume). In Drosophila, zygotic expression of the single mef-2 homolog has been shown to be necessary for all muscle differentiation (Bour et al. 1995; Lilly et al. 1995). Embryos homozygous for deficiencies removing the known C. elegans mef-2 homolog can still initiate muscle differentiation (J. Ahnn and A. Fire, unpubl.; M. Krause, pers. comm.). These deficiencies also delete considerable numbers of flanking genes, complicating further interpretation of the arrest phenotype. As point mutations in the locus become available, it should become clear what role this locus might have in myogenesis.
Searches based solely on expression pattern offer the possibility of directly obtaining genes that act at early stages in the myogenic pathway. A “gene-trap” screen (Hope 1991), as well as more directed schemes for expression pattern screening (Lynch et al. 1995), should be useful in this regard. Monoclonal antibody screens provide a variant of this approach. Two components with intriguing expression patterns have been identified in these screens. The glyoxylate cycle enzyme complex isocitrate lyase/ maleate synthetase (Liu et al. 1995) appears in myogenic lineages (although not exclusively), coincident with the onset of differentiation, and an unidentified protein recognized by the 3NB12 antibody appears in postcleavage pharyngeal precursors (Okamoto and Thomson 1985). Although the functional significance of these molecules is unclear, they provide useful markers for assessment of early muscle commitment events. In the case of hlh-1 , it has been instructive to view the gene as a marker for body-wall muscle cell commitment, leading to an investigation of regulatory sequences generating the myogenic precursor expression pattern. These studies have suggested that distinct lineage-specific signals in the hlh-1 promoter are responsible for the complete expression pattern (Krause et al. 1994). The corresponding regulatory components are likely to be the more general mesoderm-specifying or blastomere-specifying factors that act in combination to generate the myogenic pattern.
Cold Spring Harbor Laboratory Press, Cold Spring Harbor (NY)
Riddle DL, Blumenthal T, Meyer BJ, et al., editors. C. elegans II. 2nd edition. Cold Spring Harbor (NY): Cold Spring Harbor Laboratory Press; 1997. Section III, Specification of Muscle Patterns: Programs for Muscle Determination and Differentiation.