1. Myosin Heavy Chains

C. elegans expresses four distinct muscle myosin heavy-chain (MHC) isoforms, each encoded by a separate gene (Schachat et al. 1977; Waterston et al. 1982; Dibb et al. 1989). Pharyngeal muscles contain MHC C and D, whereas MHC A and B are present in all other muscles including both striated and single-sarcomere body muscles (Epstein et al. 1974; Garcea et al. 1978; MacKenzie et al. 1978a; Ardizzi and Epstein 1987). The gene encoding MHC B is designated unc-54 , as it was first identified mutationally in Brenner's screen for uncoordinated mutants. Miller et al. (1986) determined the correspondence between the remaining MHC isoforms and three other C. elegans MHC genes: myo-3 , myo-2 , and myo-1 encode MHC A, C, and D, respectively.

Although individual thick filaments in body-wall muscle contain both MHC A and MHC B, these myosins primarily form homodimers rather than heterodimers (Schachat et al. 1978). The two isoforms are differentially localized within the filament (Miller et al. 1983). In an adult thick filament 10 μm in length, MHC A forms the central 1.8 μm, and the remaining 4 μm at either end is composed predominantly of MHC B.

The Rod Portion of MHC A and MHC B

The first available MHC sequence for any organism was that of MHC B (MacLeod et al. 1981; Karn et al. 1983). The rod portion exhibits many features common to known α-helical coiled-coil domains in which two α-helices interact along a hydrophobic core (McLachlan and Karn 1982, 1983). The rod has a seven-amino-acid periodicity (abcdefg) with hydrophobic residues at positions a and d. Charge distribution reveals an additional 28-residue periodicity with a biphasic charge structure. This 28-residue motif is repeated approximately 38.5 times, resulting in alternating bands of charge along the length of the rod. This stabilizes dimer-dimer interactions at staggers that are odd multiples of 14. Cumulative experimental and modeling data suggest a 98-residue stagger (14 × 7) between MHC rods, which would correspond to a 14.6-nm spacing (McLachlan and Karn 1982). This is very close to the 14.3-nm period of myosin heads in vertebrates determined using X-ray analysis.

McLachlan and Karn noted that four “skip” residues would be needed in the MHC B rod to maintain the heptad and 28-residue repeats. These additional residues affect the hydrophobic seam between two α-helices. Skip residues are conserved in sarcomeric myosins from diverse organisms, although available sequences of nonmuscle myosins do not show regularly spaced skip residues (see Dibb et al. 1989). A likely possibility is that skip residues could be responsible for constraining rod-rod interactions during formation of highly ordered thick filaments.

Sequence comparisons between the four C. elegans myosins show only modest conservation of the rod portions (compared to strong conservation in the head portions) (Dibb et al. 1989). The rod portions of MHC A and MHC B are more similar to each other than to other isoforms; it has been suggested that this may reflect a functional constraint on rod divergence (Dibb et al. 1989). One possibility is that similarity between MHC A and B rods might reflect structural features needed to coassemble in regions of the thick filament where these two isoforms overlap.

The bipolar structure of thick filaments requires an antiparallel packing in the middle of the filament, with parallel packing in the two filament arms. The central antiparallel region is named the “bare zone” because this region lacks myosin heads. Antiparallel packing of myosin rods with complete overlap would predict a bare zone of approximately 160 nm, the measured size in several studies (see, e.g., Craig 1977). This region of the filament is occupied by MHC A (Miller et al. 1983). A direct implication of this observation is that MHC A dimers must be able to pack in an antiparallel manner and thus carry out the nucleating step in filament assembly (Fig. 3b).

Figure 3

Possible myosin and paramyosin interactions during assembly. (Arrowheads) Amino termini; (open bars) paramyosin molecules; (closed (more...)

Genetic studies support the idea that MHC A has a special role in filament assembly (Brenner 1974; Epstein et al. 1974; MacLeod et al. 1977; Waterston 1989). Animals lacking MHC A die as embryos with completely nonfunctional muscles and severely impaired thick-filament assembly (Waterston 1989). In contrast, mutants lacking MHC B are viable with weakly contractile muscles that progress to near paralysis during later larval stages (Epstein et al. 1974). Mutants lacking MHC B have thick filaments, some of wild-type length, which consist entirely of MHC A (MacKenzie and Epstein 1980; Epstein et al. 1986). This indicates that MHC A dimers are capable of standard parallel packing, not just the antiparallel packing seen at the center of each filament. Indeed, MHC A appears to be capable of replacing MHC B if expressed at sufficient levels: An increase in myo-3 (MHC A) gene copy number can rescue movement and thick-filament defects of animals lacking MHC B (Riddle and Brenner 1978; Maruyama et al. 1989). The converse is not true: Overexpression of MHC B cannot rescue mutants lacking MHC A (Waterston 1989; P. Hoppe, pers. comm.).

These genetic observations suggest that unique feature(s) in the structure of MHC A must permit assembly at the M-line. The unique features could involve the capacity for antiparallel packing or interactions with specific M-line constituents. By making a set of DNA constructs encoding MHC A/MHC B chimeras, it has been possible to map two critical regions within the MHC A rod, either of which is sufficient to confer on MHC B the ability to rescue an MHC A null (P. Hoppe and R.H. Waterston, pers. comm.). The regions affecting myosin mutant rescue are a 263-residue segment toward the middle of the rod and a segment of 169 residues near the carboxyl end (the rod has a total length of ˜1088 residues). In each of the critical segments, MHC A is more hydrophobic on the outer surface of the myosin dimer. This surface is thought to mediate dimer-dimer interactions for filament assembly; an exciting possibility is that hydrophobicity in this region may allow antiparallel packing of MHC A. Previous modeling has emphasized surface charge distribution along the rod surface in guiding filament assembly (McLachlan and Karn 1982; Kagawa et al. 1989); these results may indicate additional guidance based on hydrophobicity profile.

Functional Analysis of the Myosin Head

The sliding filament model (Huxley and Hanson 1954) provided an eloquent explanation of how sarcomeres and thus muscle could be shortened by sliding thick and thin filaments past one another. The process is ATP-dependent, requires direct interaction of the myosin head and actin, and requires a change in the conformation of the myosin head to transform chemical energy from ATP hydrolysis into mechanical energy of myosin head movement along an actin filament. The recently available crystal structure for the myosin head fragment of chicken skeletal muscle has been a major breakthrough in developing models of actomyosin function (Rayment et al. 1993a,b). The presence of an ATP-binding pocket was expected, as was the proximity of two conserved thiol residues (designated SH1 and SH2) to this pocket. The positions of the catalytic site and actin-binding region on opposite faces of the molecule were unexpected, as was the long narrow cleft that extends from underneath the catalytic site to the actin-binding face. These structural details have suggested that interactions between nucleotide and the active site not only affect the power-stroke determining myosin movement, but can also disrupt myosin-actin binding by opening the long cleft (Rayment et al. 1993b; Rayment and Holden 1994).

In light of recent structural information, it is of interest to consider functional data from mutations in the myosin head. Two broad classes of unc-54 missense mutations within the head region have been identified.

The first class apparently affects the contraction-relaxation cycle in a manner that increases the duration of tight binding between myosin and actin. This class of mutations was isolated in a screen for intergenic suppressors of specific unc-22 (twitchin) mutants (Moerman et al. 1982: Dibb et al. 1985). Although some of these suppressor mutations have global effects on motility and muscle structure, they are rare and often subtle; hence, their isolation was facilitated by the power of the reversion screen. Depending on the allele, homozygous mutant animals may exhibit little or no effect on motility, or they may have a slow rigid movement, comparable to a rigor state. Eight alleles of this class have been sequenced (Table 1) (Dibb et al. 1985; D.G. Moerman et al., unpubl.): Three affect amino acids flanking the myosin ATP-binding site (Walker et al. 1982), two alter amino acids near the conserved SH1 thiol residue (Wells and Yount 1979), two are in the actin-binding region (Mornet et al. 1981; Sutoh 1983), and one allele is a double mutation that may also affect the actin-binding region. Most of these mutations have near-normal muscle structure, but two mutations in the actin-binding region have some disorganization within the A-band (Moerman et al. 1982; D.G. Moerman, unpubl.). All substitutions are at residues conserved between nematode and vertebrate myosin.

Table 1

Contractile activity mutations affecting the head region of the unc-54 myosin heavy chain.

This first class of unc-54 missense mutations should be useful in defining different functional interactions of myosin. For example, the mutations located at the putative myosin-actin interface may permit tighter binding between actin and myosin to occur, whereas mutations near the ATP-binding site may affect ATP interactions with the pocket or consequent effects on the cleft. These mutant myosin forms should be particularly useful for biochemical analysis in combination with in vitro motility systems (for review, see Cooke 1995; Ruppel et al. 1995), including systems to analyze single myosin molecules (Finer 1994; Ishijima et al. 1994).

The second class of unc-54 myosin head mutants, identified as dominant-negative mutations, suggests a role for the myosin globular head in filament assembly (Bejsovec and Anderson 1988, 1990). Heterozygous unc-54(d) animals are slow, or paralyzed, and have disrupted thick-filament assembly and organization. Homozygous animals are very sick or inviable, depending on the allele, and accumulate only low levels of MHC B (Bejsovec and Anderson 1988). Since homozygous unc-54(0) mutants are paralyzed, not lethal, the unc-54(d)-induced lethality cannot be due to low levels of MHC B alone. Bejsovec and Anderson (1990) proposed that the remaining altered MHC B in these mutants acts as a poison for filament assembly. Considering this hypothesis, the location of these mutations came as a surprise: All of these dominant-negative mutations mapped to the head region of myosin (31 mutations spread over 15 different sites; Bejsovec and Anderson 1990), none mapped to the rod portion of the molecule. The mutations are located in two major clusters, one near the ATP-binding site and the other near a weak actin-binding site. There is precedence for the view that the MHC head portion is involved in assembly. ATP and actin have been shown to promote myosin aggregation in vitro (Mahajan et al. 1989), and phosphorylation of regulatory light chains has been shown to regulate assembly of smooth muscle and nonmuscle myosin filaments (Citi and Kendrick-Jones 1987). Although the results of Bejsovec and Anderson suggest a role for the head region in filament assembly, it remains possible that these mutations exert their effect in some other manner, for example, by facilitating turnover of myosin. How the head region of myosin might contribute to filament assembly or stability is not understood, and this constitutes a challenge for the future.

At least one dominant allele of unc-54 affecting the rod region (e1152) has also been identified (Dibb et al. 1985). Unlike the class identified by Bejsovec and Anderson (1988), e1152 animals accumulate normal amounts of MHC B (MacLeod et al. 1977). The MHC B in this mutant can form dimers but does not aggregate to properly organized thick filaments. Heterozygous animals are slow, whereas homozygous animals are severely paralyzed. Several other dominant alleles of unc-54 result in similar heterozygote phenotypes and myosin accumulation patterns, but they have not been sequenced. A number of these alleles are lethal when homozygous.

2. Myosin Light Chains

The globular head region of a single MHC has two associated smaller polypeptides important for regulation of MHC activity. These two associated proteins are called the regulatory and essential (or alkali) myosin light chains (MLCs). Calcium regulation of myosin ATPase occurs through the regulatory MLCs. Two electrophoretically separable bands of 16,000 daltons and 18,000 daltons have been identified that correspond to MLCs (Harris et al. 1977). The larger band represents the regulatory MLCs of which there are two almost identical isoforms (Cummins and Anderson 1988). The two regulatory MLC genes of the nematode, designated mlc-1 and mlc-2 , are closely linked on the X chromosome, being separated by only 2.6 kb (Cummins and Anderson 1988). To date, one “essential MLC” gene ( mlc-3 ) has been identified (S. Sprunger and P. Anderson; C. White and P. Anderson; both pers. comm.).

mlc-1 and mlc-2 perform redundant functions within body-wall muscle (Rushforth et al. 1993; A. Rushforth and P. Anderson; C. White and P. Anderson; both pers. comm.); animals that are deleted for either the mlc-1 or mlc-2 gene appear to have wild-type body-wall muscle structure and normal movement. However, the double-mutant mlc-1(0) mlc-2(0) is paralyzed, muscle is defective, and animals die as L1/L2 larvae. Within the pharynx, the redundancy appears to be incomplete: mlc-1(0) animals exhibit normal pharyngeal function, and mlc-2(0) animals often exhibit pumping defects and larval arrest.

3. Paramyosin

This protein is a major component of thick filaments in many invertebrates (Cohen et al. 1970, 1971; Levine et al. 1976). Paramyosin is encoded by a single gene in C. elegans, unc-15 (Kagawa et al. 1989). Loss-of-function unc-15 mutants are severely paralyzed and have disorganized body-wall muscle (Waterston et al. 1977). Paramyosin is also present in pharyngeal muscles, but it is apparently not essential in these cells (Waterston et al. 1974). Paramyosin is a major component of the core of thick filaments and as such interacts directly with both MHC A and MHC B (MacKenzie and Epstein 1980). Protein content estimates of late larval thick filaments show a ratio of 4.5 paramyosin:3.1 MHC B:1.0 MHC A (Honda and Epstein 1990). Thick filaments isolated from unc-15 mutants are fragile and easily shear during isolation (MacKenzie and Epstein 1980), perhaps a clue regarding the in vivo role of paramyosin.

C. elegans paramyosin consists of 872 amino acid residues, with a structure characteristic of a myosin rod, including features of an α-helical coiled-coil throughout most of its length (Kagawa et al. 1989; H. Kagawa, pers. comm.). Like the myosin rod, paramyosin has a heptad periodicity (abcdefg)_n with hydrophobic residues concentrated in the a and d positions and charged residues concentrated in the remaining positions. Paramyosin also has a 28-residue meta-repeat of alternating charge. Paramyosin and myosin also share skip residues in the same positions, although paramyosin has an additional skip residue between the positions of skip residues 3 and 4 of myosin.

While the myosin rod and paramyosin have an overall 40% sequence similarity, paramyosin has more hydrophobic residues and fewer glycine residues than do myosin rods. The increased hydrophobicity is consistent with the localization of paramyosin within the filament. Fewer glycine residues, as well as intradimer salt bridges, may lead to a paramyosin rod that is more rigid than the corresponding region of myosin (Cohen et al. 1987; Kagawa et al. 1989). This is perhaps required for a core structure within the thick filament. The carboxyl terminus of paramyosin is similar to the carboxyl termini of MHC A and B, but paramyosin has a unique 29-residue hydrophobic amino portion. This region contains seven serine residues; two of which are known to be phosphorylation targets in the nematode (Schriefer and Waterston 1989). In several systems, the phosphorylation of myosin can regulate filament assembly or enzymatic activity. Although there is no direct evidence as to the function of paramyosin phosphorylation, it is intriguing that only free (nonfilament-associated) paramyosin is phosphorylated (Dey et al. 1992). This suggests that phosphorylation of paramyosin could regulate its assembly in vivo.

Assembly of Paramyosin

The presence of alternating zones of negative and positive charges along the outer surface of a paramyosin coiled-coil helix suggests that, like myosin, paramyosin assembly may be largely mediated by intermolecular ionic interactions (McLachlan and Karn 1982; Kagawa et al. 1989). An intriguing picture of thick-filament structure has come from “best fit” estimates of molecular packing based on these interactions. As more data become available, it will be interesting to see this picture confirmed and/or modified. In particular, nonionic interactions could also affect molecular packing within the thick filament (e.g., paramyosin shares the hydrophobicity profile for the coat position residues implicated in antiparallel MHC A packing; P. Hoppe and R.H. Waterston, pers. comm.).

Unlike myosin rods, which have maximal interactions at a parallel stagger of 98 residues, paramyosin rods appear to interact optimally at parallel staggers of 493 residues (an overlap region of 330 residues) (Kagawa et al. 1989). This is an axial displacement of approximately 720 Å (nematode paramyosin is ˜1211 Å in length) and is in good agreement with observations on paramyosin paracrystals (Fig. 3a) (Cohen et al. 1971; Waterston et al. 1974). To form and extend thick filaments in vivo, parallel paramyosin must interact with myosin rods. Models for this interaction based on charge distributions suggest an alignment with the myosin head almost contacting the amino terminus of paramyosin (Fig. 3d) (Kagawa et al. 1989).

Paramyosin might also be expected to pack in an antiparallel manner in the central zone of each filament. The existence of such “antiparallel” paramyosin is still under debate. Ardizzi and Epstein (1987) and Epstein et al. (1993) did not see labeling in the central zone with antibodies to paramyosin; this could reflect absence of paramyosin from the region or inaccessibility of epitopes. Charge distribution analysis suggests that antiparallel-packed paramyosin would be most stable with a stagger of 358 residues (Fig. 3c) (Kagawa et al. 1989).

A large-scale criterion can be used to assess the filament packing models above. It is believed that paramyosin acts as a template upon which the myosin is assembled (Kagawa et al. 1989 and references therein). For this to occur, interactions between these molecules should allow adjacent packing with minimal requirement for gaps or bending. This higher-order constraint is well satisfied by the filament packing models described above (Fig. 3d).

Much of the above modeling is supported by analysis of mutations in unc-15 (Gengyo-Ando and Kagawa 1991). Mutants lacking paramyosin have aggregated deposits of myosin but do not form structures resembling native thick filaments (Waterston et al. 1977). Missense alleles of paramyosin have disorganized muscle; under polarized light, large birefringent needle-shaped paramyosin aggregates can be observed within the muscle cells. These animals can be either weakly (e.g., e1215) or severely (e.g., e73) uncoordinated. e73 is semidominant and has been used to isolate both intergenic and intragenic suppressor mutants (Riddle and Brenner 1978: Brown and Riddle 1985). e73 results in a charge reversal in the middle portion of the molecule (D-342 to K). Three other missense alterations, su228 (R-837 to C), e1215 (Q-809 to R), and e1402 (L-799 to F), are restricted to a small, unusually hydrophilic region near the carboxyl terminus.

The charge-based model for parallel paramyosin assembly described above would place the e73 site opposite the region where the other three mutations are located (see Fig. 4). Further support of this model comes from the analysis of the intragenic revertants of e73 (Brown and Riddle 1985; Gengyo-Ando and Kagawa 1991). Three revertant changes were sequenced: m193 (E-586 to K) results in limited restoration of movement, whereas m208 (R-826 to K) and m209 (E-835 to K) allow more effective restoration of movement. When superimposed on the parallel assembly model described above, the latter two revertants contact the e73 region. The e73 and m209 sites are directly opposite each other (Fig. 4). Gengyo-Ando and Kagawa (1991) propose that e73 may promote abnormal paramyosin aggregate (paracrystal) formation by increasing the self-affinity of the molecule. m209 could decrease this affinity by restoring the charge repulsion at this position.

Figure 4

Mutational effects on parallel packing by paramyosin. (A) Mutation sites are indicated by allele number. Paramyosin molecules are drawn as associated in the (more...)

The model of parallel paramyosin assembly was based on one-dimensional considerations of the ionic interactions between structures (McLachlan and Karn 1982; Kagawa et al. 1989). As noted by Gengyo-Ando and Kagawa (1991), the situation is clearly more complex in vivo. The single charge reversals caused by paramyosin missense mutations do not significantly alter the calculated interaction scores, yet they dramatically affect assembly. Perhaps the mutations pinpoint regions of the molecule crucial for the initial steps of parallel paramyosin assembly. Just how complex the dynamic interactions between paramyosin self-association and paramyosin-myosin association are in vivo is illustrated by interactions between e73 and its intergenic suppressor myo-3 /sup-3. (Riddle and Brenner 1978; Brown and Riddle 1985; Maruyama et al. 1989). Motility can be partially restored to e73 animals by increasing MHC A levels (Maruyama et al. 1989). How increasing myosin counters paramyosin self-assembly is not understood, but this in vivo result should give pause to those who rely solely on in vitro reconstruction experiments to determine essential components in complex molecular assemblages.

Paramyosin and the Thick-filament Core Structure

In experiments initially designed to map the relative positions of MHC A and B within a thick filament, Epstein et al. (1985) found a new substructure within the filament which they have called the “core.” Successively higher salt concentrations (0.1 M to 0.75 M) applied to purified thick filaments will gradually solubilize the filaments, progressing from the ends toward the center. Using antibodies to mark the myosins, it was shown that, first MHC B was solubilized, and then MHC A, finally leaving an insoluble region with MHC A, a result that confirmed the differential distribution of these two myosin isoforms within the filament. When this minifilament was examined using electron microscopy, a “core structure” 15 nm in diameter was observed extending beyond the MHC A region of the filament.

New methods have allowed analysis of the filament core in more detail (Fig. 5a) (Epstein et al. 1988; Deitiker and Epstein 1993). Although some paramyosin is solubilized along with myosin in the above treatments, 30% of paramyosin remains with the core. Thus, paramyosin is a major component of core fractions (Deitiker and Epstein 1993). Paramyosin-rich core structures have also been identified in other invertebrates (e.g., Limulus; Levine et al. 1982). In C. elegans, minor protein constituents associated with the core have been identified (Deitiker and Epstein 1993), and a model for the nematode core structure has been proposed (Fig. 5b,c,d,e) (Epstein et al. 1995).

Figure 5

Models of thick-filament substructures in C. elegans. (Reprinted, with permission, from Deitiker and Epstein 1993.) (a) Layers of the thick filament (in each case, (more...)

4. The Twitchin Protein Superfamily

The identification of the thick-filament protein twitchin in the nematode marked the first occurrence of a new muscle component being discovered through genetic analysis. Mutants in the unc-22 gene (Brenner 1974) exhibit disorganized muscle sarcomeres and constant “twitch” of body muscles (Moerman and Baillie 1979; Waterston et al. 1980). The twitching phenotype suggests a role in regulating the actomyosin contraction-relaxation cycle. Early support for this hypothesis came from genetic reversion studies and immunofluorescence: (1) Rare missense alleles located in the head region of MHC B can suppress unc-22 -induced twitching (Moerman et al. 1982) and (2) twitchin colocalizes with MHC B in muscle A-bands (Moerman et al. 1988).

Sequencing of unc-22 revealed a large protein (6839 amino acids) with a single protein kinase domain near the carboxyl terminus, multiple copies of a fibronectin type-III-like domain (motif I, 31 copies), and an immunoglobulin superfamily C2-like domain (motif II, 30 copies) (Benian et al. 1989, 1993). Each motif is approximately 90–100 amino acids in length, and these motifs are arranged in a I-I-II pattern (and rarely as I-I-I-II) throughout most of the protein. Several additional motif II repeats are present at the amino and carboxyl termini (Fig. 6).

Figure 6

Schematic of domains in the deduced amino acid sequence of twitchin. (Gray boxes) Motif I (fibronectin type-III-like domain); (white boxes) motif II (immunoglobulin (more...)

The discovery of an intracellular protein with fibronectin and immunoglobulin motifs was initially surprising, since previously described motifs of these types were all in extracellular proteins or domains. There are now several members of this protein family, including a set of vertebrate muscle components (titin/connectin [Labeit et al. 1990; Labeit and Kolmerer 1995], myosin light-chain kinase [Olson et al. 1990; Shoemaker et al. 1990], C protein [Einheber and Fischman 1990], 86-kD protein [Fischman et al. 1991], telokin [Gallagher and Herring 1991; Collinge et al. 1992], skelemin [Price and Gomer 1993], and M protein [Noguchi et al. 1992]) and several invertebrate variants of twitchin (the Drosophila twitchin homolog, projectin [Ayme-Southgate et al. 1991; Fyrberg et al. 1992] and a set of arthropod, annelid, and mollusc muscle components called “mini-titins” [Nave and Weber 1990; Lakey et al. 1990; Vibert et al. 1993; Probst et al. 1994; for review, see Ziegler 1994]). Titin, C protein, and twitchin have all been shown to associate with myosin (Labeit et al. 1992; Okagaki et al. 1993; cited in Deitiker and Epstein 1993). A possible unifying theme for members of this family might thus be association with myosin.

Although there are likely to be structural and other similarities, there is no reason a priori to assume that protein family members have identical functions. Despite similar structural motifs in titin and twitchin (immunoglobulin and fibronectin domain repeats, and a serine/threonine protein kinase near the carboxyl end), their distribution and proposed functions are different. Titin, at 3000 kD, is the largest member of this muscle protein family; at approximately 1 μm in length, it can extend over half a vertebrate sarcomere, from M-line to Z-line (Wang et al. 1984; Furst et al. 1988). Two distinct functions for titin have been proposed on the basis of distribution within the sarcomere. Association of titin with thick filaments has led to the suggestion that it acts as a template or “protein ruler” for thick-filament assembly (Whiting et al. 1989; Trinick 1994; Labeit and Kolmerer 1995). The extension of titin across the I-band from the ends of thick filaments to the Z-line and its elastic nature suggested that it might be responsible for the passive elasticity of muscle (Maruyama et al. 1976 1977; Horowits et al. 1986, 1989; for review, see Trinick 1994; Ziegler 1994; also see Labeit and Kolmerer 1995).

The phenotype of unc-22 mutants, the localization of the protein product to the A-band region of the sarcomere, and the presence of a kinase domain within the protein have led to the suggestion that twitchin is involved in the regulation of muscle contraction (Brenner 1974; Moerman et al. 1988; Benian et al. 1989). Genetic studies on unc-22 mutants demonstrate that twitchin does not function as a protein ruler to regulate thick-filament length. Lack of twitchin does not directly inhibit filament assembly, rather the constant twitching of the muscle cell destroys previously assembled filament integrity. Young animals have near-normal muscle structure, but this structure gets progressively more disorganized with increasing age (D.G. Moerman, unpubl.). Whereas lack of twitchin in a wild-type background leads to a phenotype of twitching and disorganized muscle, lack of twitchin in an unc-54(s75) background leads to relatively normal movement and muscle (Moerman et al. 1982; also see section on myosin head mutants). By altering the crossbridge cycle of myosin-actin interactions, one releases the animal from any requirement for twitchin. This would not be the expected result if twitchin were acting as a protein ruler determining thick-filament length and stability; this result is instead compatible with twitchin having a role in regulating contraction.

Drosophila projectin offers an instructive example of isoform diversity. Projectin in asynchronous indirect flight muscle is limited to the I-band region, whereas projectin in synchronous muscles is associated with the A-band region (Vigoreaux et al. 1991). Projectin isoforms vary in size but are the products of a single gene (Ayme-Southgate et al. 1991). The localization of different isoforms of projectin to different subcompartments of a sarcomere in different muscle types suggests that this protein may have multiple functions. Drosophila indirect flight muscles are stretch-activated; thus, the role of projectin in these muscles may be analogous to the elasticity function proposed for titin in vertebrate muscle. Conversely, the location of projectin to muscle A-bands in synchronous muscles has led to the suggestion that, similar to twitchin, it may be involved in the regulation of contraction (Ayme-Southgate et al. 1995).

Studies on the kinase domain of nematode twitchin have shed light on protein kinase structure and autoregulation in general. The kinase domain of twitchin has activity in vitro, with myosin light-chain peptides the preferred substrates (Lei et al. 1994). Similar to other myosin light-chain kinases, twitchin undergoes autophosphorylation: Thr-5910, a site just upstream of the catalytic core (residues 5940–6197) is the primary target (Lei et al. 1994). In addition, twitchin contains an autoinhibitory site just carboxy-terminal to the catalytic core which can bind to the catalytic site and inhibit its function (Hu et al. 1994; Lei et al. 1994). X-ray crystallographic studies of twitchin's kinase demonstrate the steric mechanism of this autoinhibition. The inhibitory fragment does not simply act as a “pseudosubstrate” but actually “mirrors” the active site making contacts with residues (Hu et al. 1994).

Whether myosin light chains are the normal phosphorylation target in vivo for twitchin is unknown (J. Heierhorst et al., in prep.). Since nematode muscle has both myosin-based and thin-filament-based calcium regulatory systems (Lehman and Szent-Gyorgi 1975; Harris et al. 1977), a role for twitchin as part of a myosin-based regulatory system would be plausible. Recent results obtained in studies on Aplysia californica add strong circumstantial support to this suggestion. Experiments designed to identify proteins mediating muscle relaxation induced by neuropeptide cotransmitters (cardioactive peptides and myomodulins) identified a 750-kD phosphoprotein as a major substrate in this pathway (Probst et al. 1994). This protein is the Aplysia twitchin homolog; its level of phosphorylation was directly related to the change in muscle relaxation rate (Heierhorst et al. 1994; Probst et al. 1994). These physiological studies in Aplysia, when combined with genetic and molecular studies from C. elegans, make a strong case for a dynamic role for twitchin in control of the actomyosin contraction-relaxation cycle.

Recent studies of the unc-89 gene provide a genetic handle on other functions for the twitchin/titin superfamily. unc-89 mutations affect the M-line at the center of the bare zone, the region of bipolar thick filaments free of myosin heads. The function of the M-line may be to maintain thick filaments in proper register. Examination of unc-89 mutant body-wall muscle reveals a normal thick-filament number, but improper alignment of thick filaments; in addition, there is no M-line matrix (Waterston et al. 1980). unc-89 mutants move well but are thinner and more transparent than wild-type animals. UNC-89 is a 732-kD member of the twitchin/titin intracellular immunoglobulin superfamily (Benian et al. 1996). The protein consists from amino to carboxyl end of a complex series of domains, including SH3, CDC24, and PH; seven immunoglobulin domains; a KSP-containing multiphosphorylation domain; and finally 46 immunoglobulin domains in tandem. The region of similarity to CDC24 suggests that UNC-89 may be coupled to a G-protein-mediated signal transduction pathway. UNC-89 has been localized to the middle of muscle A-bands and may be specifically associated with the M-line (Benian et al. 1996). In vertebrates, five nonmyosin proteins have been localized to the M-line: creatine kinase (Strehler et al. 1983), myomesin (Vinkemeier et al. 1993), skelemin (Price and Gomer 1993), M protein (Noguchi et al. 1992), and the carboxyl terminus of titin (Gautel et al. 1993). As with UNC-89, the latter four proteins are members of the twitchin family.

1. Actins

One of the most conserved proteins phylogenetically, actin forms the core component of thin filaments and binds and activates myosin. C. elegans has four known muscle actin genes (Files et al. 1983; Albertson 1985; Waterston et al. 1984; Landel et al. 1984; Avery 1993a; Stone and Shaw 1993), encoding nearly identical proteins (Krause et al. 1989). These are typical invertebrate actins in that they resemble the sequence of vertebrate cytoplasmic actins, not vertebrate muscle actins (Vandekerckhove and Weber 1978). A candidate for cytoplasmic actin, act-5 , has been recently identified in C. elegans (L. Schriefer and R.H. Waterston; J. Waddle and R.H. Waterston; both pers. comm.). This actin is somewhat divergent from the previously identified actins, and, unlike these other actins, it is expressed in the intestine.

The study of C. elegans actin genes offers an instructive example of in vivo genetic analysis of a multigene family. Actin mutants were first identified as semidominant slow to paralyzed animals designated unc-92 . Genetic mapping and reversion analysis suggested that these might be actin mutations (Landel et al. 1984; Waterston et al. 1984). These mutants have disorganized body-wall muscle and aggregates of thin filaments at the ends of muscle cells (Waterston, et al. 1984). Pharyngeal muscle is also disrupted. These early results suggested the possibility that the dominant mutations might be missense alleles of individual actin genes and that the frequent reversion could result from second-site loss-of-function mutations inactivating the aberrant gene. Molecular studies have confirmed these hypotheses (Landel et al. 1984; L.A. Schriefer and R.H. Waterston, pers. comm.). The crystal structure of actin is known (Holmes et al. 1990; Kabsch et al. 1990), and the sequencing of dominant alleles reveals that several are amino acid substitutions located on an interactive actin face or near the ATP-binding pocket (L.A. Schriefer and R.H. Waterston, pers. comm.) The implication is that these alterations disrupt actin-actin interactions during filament assembly. Some of the revertants analyzed have insertions or deletions in act-1 or act-3 , indicating that elimination of a particular actin isoform, not the repair of the lesion, is the mode of phenotypic reversion (Landel et al. 1984). This type of revertant is possible presumably because multiple actin genes are expressed in these muscle tissues and the level of actin required in these muscles is less than that produced when either act-1 or act-3 are deleted (no data yet for act-4 ).

2. The Troponin-Tropomyosin Complex

This complex regulates the calcium-sensitive interaction of actin with myosin. Genes for nematode tropomyosin, troponin C (TnC), and troponin T (TnT), have all been cloned and sequenced, and mutations have been identified (Myers et al. 1996; H. Kagawa and B. Williams, pers. comm.; H. Kagawa et al., in prep.). Tropomyosin and TnC are encoded by the lev-11 and pat-10 genes, respectively (Williams and Waterston 1994; H. Kagawa and B. Williams, pers. comm.); TnT is encoded by mup-2 (Myers et al. 1996). There is evidence that multiple tropomyosin isoforms are encoded by a single-copy gene with alternative splicing (H. Kagawa et al., in prep.), but the situation is not yet clear for TnC and TnT, which may have other family members.

Loss-of-function mutations in lev-11 , pat-10 , or mup-2 lead to late embryonic lethality. Interestingly, none of these genes affect early assembly of the sarcomere; all act after muscle contraction has begun (Williams and Waterston 1994; Myers et al. 1996). Mutations in lev-11 or pat-10 result in an identical phenotype of severe paralysis from the earliest stages of development, whereas mup-2 mutants are capable of muscle contraction and continue morphogenesis until the threefold stage. The primary defect noted in mup-2 animals is a displacement of the dorsal muscle quadrants, particularly at the bends in the embryo. This displacement appears to be due to a loss of muscle attachment to the underlying matrix, since most mup-2 animals at the twofold stage have properly organized and attached muscle quadrants.

The isolation of a temperature-sensitive (ts) allele for mup-2 has permitted analysis of TnT function in later stages of muscle activity (Myers et al. 1996). This mup-2 mutation allows growth but causes a contraction phenotype: The animals tremble in liquid medium with poor coordination between muscle cells along the animal's length (Myers et al. 1996). This phenotype is reminiscent of unc-22 mutants and some non-null alleles of lev-11 and unc-54 (Brenner 1974; Lewis et al. 1980a; Dibb et al. 1985). It is intriguing that similar phenotypes can arise from defective components of thick filaments ( unc-22 twitchin, unc-54 myosin) or thin filaments ( lev-11 tropomyosin, mup-2 troponin). The trembling/twitching phenotypes suggest regulatory defects in the muscle contractile cycle. Nematode muscles use both thick- and thin-filament-based calcium regulatory systems in controlling contraction (Harris et al. 1977). Analysis of mutations that perturb regulation in just one of the two filament types may illuminate the nature and purpose of this dual regulation.

A second phenotype of mup-2 (ts) mutants grown at a restrictive temperature is that they are sterile (Myers et al. 1996). Sterility appears to result from lack of fertilization of the oocytes; the oviduct of these animals is enlarged and full of oocytes, but these oocytes fail to pass through the spermatheca. Oocyte movement along the oviduct and through the spermatheca is dependent on the myoepithelial sheath, a contractile structure (Hirsh et al. 1976) containing actin (Strome 1986b), myosin (Ardizzi and Epstein 1987), and attachment structures (e.g., perlecan; G. Mullen and D.G. Moerman, unpubl.). Comparative video recordings show that mup-2 mutants are severely deficient in oviduct contractile function (Myers et al. 1996).

3. Other Thin-filament-associated Proteins

Two genes recently cloned and sequenced which have roles in thin-filament stability and function, unc-60 and unc-87 , take us beyond the more familiar constituents of sarcomere thin filaments. Depending on the allele, unc-60 mutant animals exhibit a limp paralysis, slow movement, or lethality (Waterston et al. 1980; McKim et al. 1994). Large aggregates of thin filaments are observed at the ends of cells in all viable alleles examined (Waterston et al. 1980). unc-60 is a complex locus encoding two transcripts that share an initiator methionine but have separate subsequent exons (McKim et al. 1994). The two encoded proteins are 165 and 153 amino acids in length, with 38% identity and show similarity to the cofilin/destrin family of actin-binding proteins, a group known to depolymerize actin (Yonezawa et al. 1985; Nishida et al. 1987; Abe et al. 1989, 1990; Moriyama et al. 1990). Whether UNC-60 has a function similar to these proteins is unknown, but the unc-60 mutant phenotype of actin aggregates is consistent with this protein having a regulatory role in filament assembly. McKim et al. (1994) proposed that UNC-60 may have a function similar to actophorin in Acanthamoeba castellani, where actophorin acts in conjunction with α-actinin to promote bundling of actin filaments in vitro (Maciver et al. 1991). In the absence of actophorin, Acanthamoeba actin filaments continue to grow and cannot be bundled by α-actinin.

The single lethal allele, unc-60(s1586), is a small deletion interrupting both open reading frames (ORFs) (McKim et al. 1994). Therefore, it may be that mutations affecting only one transcript give rise to the Unc class of alleles. It will be interesting to learn if these mutations are limited to one ORF or whether both are targets. The s1586-bearing animals arrest development at mid-larval stages, which is surprising given that thin-filament regulatory proteins (e.g., tropomyosin) lead to late embryonic lethality, as does the absence of other key sarcomere constituents (see below). Perhaps unc-60 is a member of a functionally redundant gene family.

Mutations in unc-87 result in a severe larval paralysis, whereas adults are somewhat less affected (they exhibit a limp paralysis). unc-87 animals have disorganized body-wall muscle with collections of both thick and thin filaments. As with unc-60 mutants, small clumps of thin filaments are visible at the ends of cells (Waterston et al. 1980; Goetinck and Waterston 1994a). The unc-87 gene has recently been sequenced and was found to encode a 357-amino-acid 40-kD protein with a portion showing sequence similarity to the actin-binding domain of vertebrate calponins (Goetinck and Waterston 1994b). Calponins were first identified in vertebrate smooth muscle as binding to F-actin, Ca⁺⁺/calmodulin, and tropomyosin (Takahashi et al. 1986; Vancompernolle et al. 1990). Although their role in smooth muscle is not fully understood, one suggestion is that calponins form part of the thin-filament regulatory system (Abe et al. 1990; Winder and Walsh 1990). It is not clear whether UNC-87 has a regulatory role in muscle. In particular, it lacks homology with the calponin regions identified as tropomyosin- or calmodulin-binding sites (Vancomopernolle et al. 1990; Mezgueldi et al. 1992; Goetinck and Waterston 1994b). However, it is clear that UNC-87 is closely associated with the thin filaments. In wild-type and tropomyosin mutant ( lev-11 ) animals, UNC-87 is found in muscle I-bands, and in act-3 mutants, UNC-87 colocalizes with the misplaced thin filaments (Goetinck and Waterston 1994b).

Since muscle contraction appears to exacerbate the unc-87 phenotype, Goetinck and Waterston (1994a) speculated that retarding muscle contraction might alleviate the muscle disorder in these mutants. Indeed, unc-54 missense alleles that retard the contraction/relaxation cycle can partially suppress unc-87 disorders. Although the role of UNC-87 in muscle is unknown, it appears that this protein is not involved in the early stages of sarcomere assembly, since threefold-stage embryos have wild-type muscle which then becomes progressively more disorganized as the animals grow. On the basis of its structure, localization, and interaction with other muscle-affecting genes, Goetinck and Waterston (1994a) suggested that UNC-87 serves as a structural component to maintain lattice integrity during contraction.

1. Genes Necessary for Early Sarcomere Organization

Two broad phenotypic classes of lethal mutations affecting early muscle have been described in C. elegans, the Mup class (muscle positioning; Hedgecock et al. 1987; Goh and Bogaert 1991), and the Pat class (paralyzed and arrested elongation at twofold; Williams and Waterston 1994). These mutants are primarily embryonic or early larval lethals. The gene and product for some of these mutants are known; these are either components of the attachment complex (i.e., dense body and M-line: deb-1 -vinculin, Barstead and Waterston 1991a; pat-3 -β-integrin, Gettner et al. 1995; pat-2 -α-integrin, Williams and Waterston 1994 and pers. comm.; unc-52 -perlecan, Rogalski et al. 1993) or essential components within the sarcomere ( myo-3 -MHC A, Waterston 1989; lev-11 -tropomyosin, Williams and Waterston 1994 and pers. comm., Kagawa 1995 and pers. comm.; mup-2 -troponin T, Myers et al. 1996; pat-10 -troponin C, B. Williams and H. Kagawa, pers. comm.). Several other Pat mutants have yet to be characterized on a molecular level (Williams and Waterston 1994; K. Norman and D.G. Moerman, unpubl.).

The Pat phenotype offers a ready and efficient screen for mutants with disruptions in sarcomere assembly and organization, since animals arrested at the twofold stage of elongation are quite easy to identify. Elongation is the process whereby the embryo changes from a ball of cells to a worm. A hypodermal role in elongation had been described (Sulston et al. 1983; Priess and Hirsh 1986). The demonstration that myo-3 mutants arrested at the twofold stage of embryogenesis was the first indication that intact functional muscle is necessary to facilitate elongation of the embryo (Waterston 1989). Mutations in myo-3 do not affect the early stages of elongation. myo-3 mutants appear to develop normally until the 1.5-fold stage, when contraction would normally begin. These animals fail to initiate muscle contractions; the paralyzed animals cease elongation shortly thereafter (twofold stage; see Fig. 9). Pharyngeal morphogenesis continues in these animals and they hatch as inviable larvae.

That the Pat phenotype might be a common feature of mutations affecting early events in myofilament assembly was suggested by the results of Barstead and Waterston (1991a), who demonstrated that mutants lacking vinculin ( deb-1 ) also have a Pat terminal phenotype. Lethal alleles of unc-45 , a gene involved in thick-filament assembly (Epstein and Thomson 1974; Waterston 1988), also die as twofold-arrested and paralyzed animals (Venolia and Waterston 1990). Williams and Waterston (1994) carried out a screen designed to sample the whole genome for recessive-lethal mutations leading to the Pat phenotype. They identified ten novel genes and also obtained lethal “Pat” alleles of three genes previously identified only by non-null viable alleles ( unc-52 , unc-112 , and lev-11 ). Although the screen has not yet been carried out to saturation, this new array of mutants has proven invaluable for deducing several of the essential events in myofilament formation.

2. Sarcomere Assembly and the Membrane

A physical linkage between the myofibril array and the membrane is critical in allowing myofilament contraction to move the animal. This linkage is provided by the dense body structure (see Fig. 1). Major intracellular constituents of this structure in C. elegans include α-actinin, found throughout the dense body (Francis and Waterston 1985; Barstead et al. 1991b), and vinculin and talin, which are found nearer the sarcolemma (Barstead and Waterston 1989; Moulder et al. 1996). A transmembrane complex containing β-1 and α-integrin components anchors the myofilament lattice (Gettner et al. 1995; B. Williams, pers. comm.). Perlecan is a basement membrane component associated with the extracellular surface of these anchoring regions (Francis and Waterston. 1991; Rogalski et al. 1993).

The nematode dense body structure is similar in many respects to the vertebrate focal adhesion site (Burridge et al. 1988). Vertebrate models of myofibrillar assembly emphasize the importance of submembranous adhesion plaques as assembly points for myofibrillar components (for review, see Epstein and Fischman 1991). In cultured cardiac myocytes, myofibrillar components are found adjacent to the sarcolemma, and myofibril formation occurs at the plasma membrane (Dlugosz et al. 1984; Schultheiss et al. 1990). There is now direct evidence from studies in C. elegans and Drosophila demonstrating the essential role of membrane-associated proteins as nucleation sites for myofibril assembly. β-integrin is the transmembrane anchor at the base of dense bodies and M-lines in the nematode (Francis and Waterston 1985; Gettner et al. 1995). Mutations in the structural gene for β-integrin, pat-3 , lead to a twofold arrest, and myosin and actin are not organized into sarcomeres in muscle cells (Williams and Waterston 1994). These observations on pat-3 mutants suggest that although this β-integrin is not necessary for muscle cell migration or the formation of muscle quadrants, it is essential for assembling the myofilament lattice within individual muscle cells in the nematode. Similar results have been obtained in Drosophila with the lethal(1) myospheroid locus which encodes a β-integrin subunit (Wright 1960; Newman and Wright 1981; MacKrell et al. 1988; Volk et al. 1990).

Components of muscle-associated basement membranes are also important contributors to myofibril assembly and organization and may act as anchors for the transmembrane integrin complex. For example, mutations in type IV collagen ( let-2 and emb-9 ), or perlecan ( unc-52 ), or overexpression of SPARC/osteonectin all lead to alterations in muscle during embryogenesis and thus result in a lethal phenotype (Guo et al. 1991; Sibley et al. 1993; Rogalski et al. 1993, 1995; Schwarzbauer and Spencer 1993; Williams and Waterston 1994; for review, see Kramer 1994b and this volume). Particularly striking is the similarity in the null phenotypes for unc-52 and pat-3 (Rogalski et al. 1993; Williams and Waterston 1994). Similar to pat-3 mutants, unc-52 Pat mutants fail to assemble a myofilament lattice, although all of the appropriate muscle constituents are present (Fig. 10); this appears to be the null phenotype of unc-52 (Rogalski et al. 1995).

Figure 10

Immunofluorescence antibody staining of wild-type and unc-52 embryos. Embryos are stained at approximately 420 min after the first cleavage. For reference, (more...)

The unc-52 gene encodes the nematode homolog of perlecan, the core protein of the mammalian basement membrane heparan sulfate proteoglycan (Fig. 11) (Noonan et al. 1991; Kallunki and Tryggvason 1992; Murdock et al. 1992; Rogalski et al. 1993; see Kramer, this volume). Nematode perlecan is found in the extracellular matrix (ECM) linking muscle to hypodermis and is concentrated under the dense bodies and M-lines where integrin anchoring occurs (Fig. 12) (Francis and Waterston 1991). Mammalian perlecan co-aligns with focal adhesion sites in fibroblasts (Singer et al. 1987). Recent evidence suggests that mammalian perlecan and integrin may directly interact with each other (Chacravarti et al. 1995).

Figure 11

Schematic diagram of unc-52 and possible protein products. unc-52 consists of 26 exons and has three alternative (more...)

Figure 12

High-resolution immunofluorescence staining of hypodermal, basement membrane, and muscle components, illustrating the relationship of these tissues and attachment structures. Staining of (more...)

An attractive interpretation of these results is that perlecan and integrin form a nucleating complex for sarcomere assembly and that several sets of components assemble independently onto such centers. This model incorporates the current data and emphasizes the compartmental structure of the sarcomere. The limited effects of myo-3 and deb-1 mutations exemplify the compartmental nature of myofilament assembly (Waterston 1989; Barstead and Waterston 1991a; Hresko et al. 1994; Williams and Waterston 1994). In deb-1 mutants lacking vinculin, actin is not well organized, but integrin and talin are present and properly organized within the sarcolemma (Hresko et al. 1994; Moulder et al. 1996). This observation implies that the base of the transmembrane complex is assembled at the membrane independent of other components of the dense body and thin filaments. This model of initiation of myofilament assembly does not address several key questions. For example, how are perlecan-integrin complexes recognized separately by talin/vinculin/ α-actinin assemblages and M-line constituent assemblages, and how is the spacing and alternating order between integrin transmembrane sites determined? These are clearly important parameters for initiating myofilament assembly and for determining the length and orientation of a myofilament unit.

One clear implication of studies to date is that spacing of transmembrane complexes is not dependent on intact ordered myofilaments, since sarcomere spacing is not perturbed in mutants with primary defects in either thick or thin filaments (Waterston et al. 1980; Zengel and Epstein 1980; Waterston et al. 1984; also see Waterston and Francis 1985; Hresko et al. 1994). The distance between dense bodies and the M-line does increase during growth (for review, see Waterston and Francis 1985). A body-wall muscle cell in a newly hatched L1 is two sarcomeres wide and has A-bands with approximately 100 filaments. As the animal increases its mass, muscle cells grow, so that as an adult each muscle cell is about 10 sarcomeres wide and each A-band has 600 or more filaments (MacKenzie et al. 1978b). To accommodate this growth, the number of A-bands and the size of individual A-bands both increase (L1: width = 0.5 μm, depth = 0.4 μm; Adult: width = 1.1 μm, depth = 1.4 μm). To maintain the stagger of the filaments when the width of the filaments is increased, individual filaments must grow longer, from 5 μm in an L1 to approximately 10 μm in an adult (see Waterston and Francis 1985). A direct implication of increased filament length is that spacing between integrin complexes must also increase during development; yet evidence to date suggests that the former does not determine the latter. This flexibility in the positioning of the integrin complex during growth stands as a major challenge that must be accommodated in any model of myofilament assembly.

Whether the “ruler” or “organizer” for determining the final position of integrin in the muscle membrane is internal or external to the cell is unknown. The identification of perlecan as important for sarcomere assembly and attachment in the nematode is the first evidence that forces outside a muscle cell can influence internal structure. At this time, only a few major proteins of the ECM are identified; we know little about their roles either in organizing the ECM or in cell attachment. One feature evident from this rudimentary dataset is that the ECM undergoes structural changes during development. In particular, unc-52 expresses tissue- and temporal-specific perlecan isoforms (G. Mullen and D.G. Moerman, unpubl.). Viable mutations affecting a specific unc-52 isoform lead to late larval defects, including muscle attachment defects and fracturing of dense bodies (MacKenzie et al. 1978b; Waterston et al. 1980; Rogalski et al. 1995). Precise analysis of internal muscle structure as these mutants develop may illuminate the connection(s) between basement membrane structure and internal muscle organization (i.e., dense body positioning).

3. Muscle-Hypodermal Interactions

For muscle contraction to be useful to the animal, the series of links anchoring the myofilament lattice must eventually lead to the hypodermis and overlying cuticle. Studies on Ascaris as well as C. elegans reveal that the hypodermis in regions adjacent to muscle contains tonofilaments (Bartnik et al. 1986; Francis and Waterston 1991). The tonofilament arrays are similar to intermediate filaments in size and morphology and react to antibodies specific for intermediate filaments (Fig. 12) (Bartnik et al. 1986; Francis and Waterston 1991). A monoclonal antibody specific to nematode intermediate filaments, MH4, stains bands of tonofilaments running circumferentially from one side of a muscle quadrant to the other (Francis and Waterston 1991). Individual bands are about 1 μm wide and can be resolved as a doublet with a narrow gap. Two other monoclonal antibodies, MH5 and MH46, identify other components of this attachment network. These antibodies show a pattern similar to that of MH4, although MH5 staining is punctate and MH46 staining is more uniform (Francis and Waterston 1991). MH5 staining is also more intense in regions overlying muscle-muscle contacts. MH5 may be reacting with a component of the hemidesmosome, the membrane attachment structure for the tonofilaments. The MH46 antigen is a large basement membrane component of hypodermal origin with several fibronectin type III repeats as well as novel sequences (M. Hresko and L. Shrieffer, pers. comm.).

The location of the tonofilament arrays and associated organelles in regions apposed to muscle suggests that this network may be the link between muscle and the cuticle. However, there is no direct or fixed relationship between the tonofilaments and hemidesmosomes of the hypodermis with the major attachment structures within muscle. In examining transverse and longitudinal sections of nematode cuticle and muscle, Francis and Waterston (1991) found tonofilaments and their associated organelles adjacent to all major components of muscle, including A-bands, I-bands, dense bodies, and M-lines. The periodicity of the tonofilaments and hemidesmosomes is not mediated by an interaction with muscle, but rather by their association with the annuli (a pleated ridge) of the cuticle. Hemidesmosomes are located directly beneath the annuli. An implication of these observations is that tension developed by muscle may be distributed to the hypodermis by the basement membrane, perhaps through perlecan and the MH46 antigen. Interestingly, perlecan, a product of muscle, is distributed over the whole of the basement membrane underlying muscle but is concentrated at dense bodies and M-lines, whereas the MH46 antigen, a product of the hypodermis, is distributed within the basement membrane underlying muscle in a pattern similar to the distribution of hemidesmosomes within the hypodermis. How these two proteins may be linked is presently unknown.

From their birth, muscle cells are intimately associated with the hypodermis. The migration of muscle cells from a lateral position adjacent to the seam cells to either a dorsal or ventral quadrant adjacent to hypodermis implies interaction between muscle and hypodermis (Sulston et al. 1983; Hedgecock et al. 1987; Goh and Bogaert 1991). Throughout the time when muscle cells are migrating from the midline to form the four muscle quadrants, a series of temporally related changes are occurring in the hypodermis. These will eventually culminate in the formation of tonofilaments and hemidesmosomes in the region of the hypodermis adjacent to muscle (see Fig. 7). Gradually (from ˜290 minutes after the first cleavage onward), antigens to MH4, MH5, and MH46 can be detected in dorsal and ventral, but not seam, hypodermal cells. At 310 minutes, presumptive hypodermal attachment structures have started to concentrate under muscle cells. At 350 minutes, the MH46 antigen is found colocalized with perlecan at the boundaries of adjacent muscle cells and the hypodermis, and by 390 minutes, the antigens to MH4 and MH5 are concentrated under the muscle contractile complex (Goh and Bogaert 1991; Hresko et al. 1994). At 420 minutes, these hypodermal antigens, perlecan, and the myofilament lattice of muscle are co-extensive. By 430–450 minutes, the MH4 and MH5 antigens appear organized into tonofilaments and associated hemidesmosomes similar to those of an adult (Francis and Waterston 1991; Goh and Bogaert 1991; Hresko et al. 1994).

The close association of muscle and hypodermis throughout much of development, and the identification of hypodermal attachment structures only adjacent to muscle cells, suggests that these tissues may communicate to coordinate their development. The recruitment of a hemidesmosomal complex within the hypodermis does appear to be the result of a signal received from the underlying muscle. It has been shown that hypodermal cells organize hemidesmosomes only in regions adjacent to muscle cells and not in regions adjacent to areas where muscle cells have been experimentally removed (P. Shrimankar and R. H. Waterston, cited in Hresko et al. 1994).

A few genes have been identified which may have a role in muscle-hypodermal interactions. Mutations in some of the mup genes disturb muscle positioning as well as attachment (Hedgecock et al. 1987; Goh and Bogaert 1991; E. Gatewood and E. Bucher, pers. comm.). Another interesting group are the muscle attachment (mua) mutants described by J. Plenefisch and E. Hedgecock (pers. comm.). Body-wall muscle is initially properly placed and well organized in Mua animals, but during larval growth, muscle progressively detaches. Several complementation groups distinct from known pat, mup, or unc genes have been identified (J. Plenefisch and E. Hedgecock, pers. comm.). Alleles of unc-23 and viable (Unc) alleles of unc-52 convey phenotypes similar to this group of dystrophic mutants.

The fibronectin-like gene identified using the monoclonal antibody MH46 (Francis and Waterston 1991; M. Hresko, pers. comm.) is an intriguing candidate for a hypodermal factor involved in muscle interactions; mutants with defects in this gene arrest at the twofold stage of elongation but are not paralyzed (M. Hresko, pers. comm.). Muscle twitching occurs, and the muscle detaches from the hypodermis when contraction begins. In these twofold-arrested animals, the hemidesmosomes are not restricted to the region adjacent to muscle, but are found throughout the dorsal and ventral hypodermis. Mutants with specific defects in the hypodermis may help dissect the structure and function of the hemidesmosomal complex in a manner similar to what has been achieved for the muscle dense body. For example, it will be interesting to see if this structure has components common to mammalian hemidesmosomes, in particular an α6-β4 integrin complex (for review, see Garrod 1993). Beyond merely dissecting the attachment complexes of muscle and hypodermis, we will need to resolve how these two tissues coordinate and facilitate their behavior during morphogenesis if we are to understand how a fully functional muscle quadrant is established.