Muscle arm development in Caenorhabditis elegans

The ability of a cell to extend membrane is essential for numerous processes throughout animal development. For example, neurons extend axons and dendrites to establish contact and exchange information between distantly located cells. Recently it has been shown that postmitotic striated muscle cells also extend membrane in a regulated fashion. During the development of the neuromuscular junction, muscles use specialized filopodia to contact incoming motor axons. This was first observed in Drosophila, in which actin-filled muscle membrane extensions, called myopodia, cluster at sites of motor axon innervation during embryonic development(Ritzenthaler et al., 2000). The term `myopodia' has since been adopted by others to describe the membrane extensions that emanate from embryonic mouse skeletal muscle myotubes and make contact with incoming motor axons (Misgeld et al., 2002). Muscle membrane extension may be a common feature of mammalian skeletal muscle cells, as cultured rat myotubes also extend`micro-spikes' at sites of motoneuron apposition(Uhm et al., 2001). In all three cases, muscle membrane extensions disappear after the neuromuscular junction has formed, suggesting that these extensions have specific roles in guiding or stabilizing incoming motor axons during development. However, it is unknown if the development of muscle membrane extensions in Drosophila, mice and rats is controlled by similar mechanisms. Nevertheless, these observations suggest that membrane extension is developmentally regulated in muscle cells. To better understand muscle membrane extensions we have begun to study an analogous process in the nematode worm Caenorhabditis elegans.

Adult C. elegans have 95 body wall muscles (BWMs) that are required for locomotion and movement of the head. These muscle cells are arranged in four quadrants along the length of the animal(Fig. 1). The two dorsal quadrants flank the dorsal cord, while the two ventral quadrants flank the ventral cord. Within each quadrant there are two longitudinal rows of muscle cells, a proximal row close to the nerve cord, and a distal row further from the cord that we refer to as distal BWMs. The BWMs of nematodes establish contact with motor axons using specialized membrane extensions called muscle arms (Stretton, 1976; Sulston and Horvitz, 1977; Sulston et al., 1983; White et al., 1986). Muscle arms have a simple morphology consisting of a thin stalk that emanates from the cell body and a bifurcated terminus that contacts the nerve cord. The 16 anteriormost BWMs in the head extend muscle arms exclusively to motor axons in the nerve ring (Fig. 1). Immediately posterior to the head muscles, 16 neck BWMs extend arms both to the nerve ring and to the nearest nerve cord. The remaining 63 BWMs extend muscle arms exclusively to motoneurons located in the nearest nerve cord. By contrast to the muscle membrane extensions observed during neuromuscular junction formation in Drosophila, rat and mouse, C. elegansmuscle arms are maintained throughout the life of the animal.

How muscle arms reach the nerve cord is not clear. However, two lines of evidence suggest that a chemoattractant guides the migration of muscle arms to the nerve cords. First, muscle arms extend to motor axons irrespective of their physical position. For example, commissural motor axons in unc-6/netrin or unc-5/RCM mutants do not complete their ventral-to-dorsal migrations and instead migrate along the longitudinal axis at sublateral positions. In these mutants, the dorsal BWM arms that normally extend to the motoneurons of the dorsal cord instead extend to the errant lateral motor axons (Hedgecock et al.,1990). Second, muscle arms extend to locations of dense core vesicle accumulation in the cell bodies of motoneurons in unc-104mutants (Hall and Hedgecock,1991; Zhou et al.,2001). This observation suggests that the dense core vesicles contain a muscle arm chemoattractant. unc-104 encodes a kinesin motor protein that is required for anterograde transport of vesicles within axons(Hall and Hedgecock, 1991; Zhou et al., 2001). These two lines of evidence suggest that muscle arm migration to the nerve cords may be analogous to axon extension toward the source of a chemoattractive cue.

Membrane extension is a conserved process that has been studied extensively in vitro and to a more limited extent in vivo. Membrane extension is normally inhibited by capping actin filaments at the barbed end, which prevents filament elongation (Pollard and Borisy,2003). However, local activation of ADF/cofilin at the leading edge can sever actin filaments and generate exposed barbed ends(DesMarais et al., 2004; Ghosh et al., 2004). F-actin elongation can then be initiated from these newly exposed barbed ends and provide the force necessary for membrane protrusion. Furthermore, the incorporation of G-actin-ATP monomers on to newly exposed barbed ends results in ATP-rich actin filaments that are the preferred sites of Arp2/3-mediated actin side-branching (Ichetovkin et al.,2002). Extensive Arp2/3-mediated filament side-branching results in a `dendritic array' actin network at the leading edge that supports the protrusion of a broad lamellae (Svitkina et al., 1997). In regions posterior to the leading edge,tropomyosin binds actin filaments, suppressing Arp2/3-mediated side branching and protecting actin filaments from ADF/cofilin-mediated severing and disassembly (DesMarais et al.,2002; Ono and Ono,2002). Many of these proteins are required for the development of the sarcomere (Ono, 2003) and before our work it was unclear if they could also be used to regulate membrane extension within muscle cells.

Here we present a detailed characterization of muscle arm development and have found it to be stereotypical, suggesting a high degree of genetic regulation. We also show that perturbation of various components of the actin and myosin machinery, including actin, tropomyosin, ADF/cofilin and muscle myosin, impair muscle arm extension and disrupt arm morphology. Our work has resulted in new insights about the mechanics of membrane extension in muscle cells and establishes C. elegans muscle arms as a useful in vivo model for membrane extension.

Unless otherwise indicated, nematodes were cultured at 20°C. RP247 trIs30 I was constructed by microinjecting pPRRF138.2(him-4p::MB::YFP), pPRZL44(hmr-1b::DsRed2) and pPR2.1(unc-129neural-specific promoter::DsRed2) together at 10, 80 and 40 ng/μl, respectively, into N2 (wild-type) adults using standard techniques (Mello et al.,1991). The membrane anchor sequence (Mb) used to localize YFP was obtained from Dr A. Fire's 1999 vector kit and is from the pat-3membrane-localization domain (Gettner et al., 1995). An extra-chromosomal array of trEx[pPRRF138.2;pPRZL44; pPR2.1] was integrated using a previously described protocol(Mitani, 1995). Resulting integrants were backcrossed to N2 four times. RP127 trIs25 was constructed similarly by injecting and integrating pPRRF138.2(him-4p::MB::YFP), pPRZL47(F25B3.3p::DsRed2) and pRF4(rol-6(su1006) into N2, at 20, 50 and 50 ng/μl, respectively. The pPRRF207(him-4p::gfp::act-1) transgene was generated by ligating a 5.6 kb EcoRI-Asp718 fragment from pPRZL138.1(him-4p::YFP) to the 2.4 kb Asp718-EcoRI fragment from pJH4.64(pei-1p::GFP::act-1), which was a generous gift from Geraldine Seydoux. We used this transgene instead of a phalloidin conjugate that binds to all actin because the actin within BWMs obscures visualization of the actin within the relatively long distal BWM arms (not shown).

To investigate larval muscle arm extension in a post-embryonic cell division mutant we made lin-6(e1466) dpy-5(e61)/hT2[qIs48](Wang and Kimble, 2001)animals transgenic for him-4p::Mb::YFP. lin-6(e1466) homozygotes are easily recognized, as the mutation is linked to dpy-5(e61), which has a distinct phenotype. Post-embryonic divisions occur in lin-6(e1466)mutants, but most daughter cells eventually die(Sulston and Horvitz, 1981). The molecular identity of lin-6 is not published.

For microscopy, worms were anaesthetized in 2-10 mmol/l Levamisole (Sigma)in M9 solution (Lewis, 1995) and mounted on a 2% agarose pad. We used a Leica DMRA2 HC microscope with standard Leica filter sets for GFP, YFP, CGFP and GFP Red epi-fluorescence to take all pictures. A 20× dry objective was used to take all pictures of adult worms, while a 63× oil objective was used to take pictures of young larvae and GFP::actin-expressing muscles. All muscle arm counts were done from photographs, which are available upon request. To study populations of worms at defined developmental stages, worms were first synchronized (Lewis, 1995) and then staged by the extent of germline and gonad cell proliferation (Kimble and Hirsh,1979).

The temperature-shift experiments were done by incubating worms at either 15 or 25°C for at least 24 hours. Synchronized L1s (Lewis, 1995) were deposited on plates with food and placed at the desired temperature. The developmental stage of the worms was monitored and pictures of the dorsal right muscle quadrants were taken when the animals reached adulthood.

We did not assume Gaussian distributions of the number of counted arms and therefore used a more stringent two-tailed Mann-Whitney rank sum test to assess the statistical significance of the observed differences.

All RNAi experiments were done by feeding worms dsRNA-producing bacteria as previously described (Timmons and Fire,1998). The pPRRF215 unc-54(RNAi) construct was made by isolating a 1700 bp PstI-BglII unc-54 genomic fragment that contains exons 4, 5 and the first 286 bp of exon 6 from the pPD5.41 unc-54 genomic clone (a gift from Dr A. Fire) and sub-cloning it into the RNAi feeding vector pPD129.36(Timmons and Fire, 1998) cut with PstI and BglII. pPRRF215 and the negative control (empty pPD129.36) vector were transformed into HT115 bacteria(Timmons et al., 2001) using standard protocols.

The width of muscle arms from dorsal right BWM number 11 was measured for ten young adult animals of each genotype using Openlab 3.1.5 software(Improvision Inc.). Arm width was measured at the midpoint of each arm,halfway between the body of the muscle and the nerve cord. Because the length of worms, and therefore the length or size of BWMs, can vary substantially between adults of different ages and of different genetic backgrounds, the arm widths were normalized to the total muscle length along the anterior-posterior axis to control for differences in animal size.

To better understand muscle arm biology in C. elegans, we performed a detailed analysis of muscle arm development. We examined muscle arms in populations of live hermaphrodite animals expressing membrane-anchored yellow-fluorescent protein (Mb::YFP) from the muscle-specific him-4promoter (him-4p) (Vogel and Hedgecock, 2001) (Fig. 1). This marker was used because it is expressed in only a subset of distal BWMs and therefore enables the clear visualization of individual muscle arms.

Expression of Mb::YFP from him-4p in first larval-stage (L1)larvae was limited to four distal BMWs posterior to the neck muscles in each quadrant (Fig. 2A). As development continued, two more distal BWMs posterior to the initial four in each quadrant also expressed Mb::YFP, albeit at lower levels(Fig. 1D,E). We were able to assign unambiguous identities to the 16 BWMs that expressed Mb::YFP brightly from him-4p because of the largely invariant cell lineage and the stereotypical pattern of BWM shapes and arrangements(Sulston and Horvitz, 1977; Sulston et al., 1983). For example, the dorsal right BWM Cppapap is the most anterior muscle that is always overlapped by neighbouring BWMs on either side and can therefore serve as a useful position marker (Fig. 1F). There are 14 BWMs born post-embryonically from the M-mesoblast cell that intercalate into the existing quadrants at variable positions (Sulston and Horvitz,1977). The general integration space is posterior to the gonad primordium, which is located just posterior to the four brightly fluorescing BWMs. The identities we assigned to the posterior muscles that fluoresced lightly are therefore based on position only and not lineage. For convenience we have renamed the BWMs with numbers according to their position along the anterior-posterior axis in each quadrant. For example, Cppapap is referred to as dorsal right muscle 15 (Fig. 1F,G).

To control for the possibility that the Mb::YFP muscle arm reporter may interfere with muscle arm outgrowth, migration or morphology, we examined BWMs expressing DsRed2 driven from the muscle-specific C26G2.1 promoter(P.J.R. and S. Quaggin, unpublished). There were no differences in arm morphology or the number of arms observed with the two different reporters(not shown and Table 2, line 35). We conclude that Mb::YFP driven from him-4p does not interfere with muscle arm number or morphology and is therefore a useful tool to study muscle membrane extension.

Previous work has shown that L1 larvae have one or two muscle arms per BWM,whereas adults have three to five muscle arms per BWM(Hall and Hedgecock, 1991; Hedgecock et al., 1990). The period of muscle arm extension and migration during larval development was not known at the initiation of our studies. To determine when larval muscle arm extension occurs we therefore counted muscle arms at three developmental stages (Table 1). Newly hatched animals had on average 1.7 (±0.8; n=480) muscle arms per BWM(Fig. 2A). We infer that these arms develop during embryogenesis as previously suggested(Hedgecock et al., 1990). By the second larval stage, distal BWMs extended an average of 3.4 (±1.0; n=320) muscle arms (Fig. 2B). This indicates that there is a burst of muscle arm extension after hatching and before the completion of L2 development. The number of muscle arms in adults (μ=4.0±1.0; n=480) suggests that additional arm extension occurs after the second larval stage(P<0.001).

The major period of larval muscle arm outgrowth is coincidental with the only period of post-embryonic motoneuron proliferation in hermaphrodites(Sulston and Horvitz, 1977). This coincidence raised the possibility that post-embryonic motoneurons induce larval muscle arm outgrowth. We reasoned that if larval muscle arm outgrowth is dependant on the birth of post-embryonic motoneurons, lin-6 mutant animals, which are defective for post-embryonic DNA replication(Sulston and Horvitz, 1981),should have fewer muscle arms than controls. lin-6(e1466) dpy-5(e61)adults had an average of 2.3 (±0.8; n=160) muscle arms per BWM, which is significantly less than wild type (μ=4.0±1.0; n=480) and dpy-5(e61) controls (μ=3.6±0.7; n=235) (P<0.001) (Fig. 2, Table 1). These results are consistent with the hypothesis that muscle arm outgrowth during larval development requires post-embryonic motoneuron proliferation.

We next examined the number of muscle arms extended by BWMs 9 to 19 in each quadrant in detail. The number of arms extended by different BWMs within the same quadrant was significantly different (P<0.001)(Table 1). In the ventral left quadrant, for example, muscle 11 extended an average of 4.6 arms, muscle 17 extended an average of 3.4 arms, and muscle 19 extended an average of 5.8 arms. A visual inspection of the position of muscle arm outgrowth from BWMs,the paths taken to the nerve cord, and the morphology of the arms suggest that muscle arm outgrowth and migration is also stereotypical(Fig. 2E-P). These results suggest the existence of extensive genetic regulation that controls several aspects of muscle arm development.

The UNC-104 kinesin motor protein is thought to transport the putative muscle arm chemoattractant along motor axons(Hall and Hedgecock, 1991). We therefore counted the number of muscle arms extended in the background of three unc-104 loss-of-function mutations, including the null mutation rh142 (Hall and Hedgecock,1991). In this and all further mutant analysis we considered the muscle arms only from the dorsal right quadrant BWMs, as all quadrants show similar defects and muscle arms are most clearly visible in this quadrant. The unc-104 mutants displayed significantly fewer muscle arms than did wild-type adult animals (P<0.001)(Table 2, lines 31-33; Fig. 3). Intriguingly, unc-104 mutant adults extended similar numbers of muscle arms to wild-type hatchlings. This suggests that UNC-104 is required for larval muscle arm outgrowth, but is dispensable for embryonic muscle arm development.

Membrane extension is dependent upon remodelling of the actin cytoskeleton(Pollard and Borisy, 2003; Ritzenthaler et al., 2000),prompting us to determine if the actin cytoskeleton is required for muscle arm extension. To visualize actin in muscle arms we used the him-4promoter to drive the expression of a GFP-C. elegans actin (ACT-1)fusion protein within the distal BWMs (Fig. 4). Similar functional GFP::ACTIN fusion proteins have been used in many other systems to investigate the subcellular distribution of actin(Fischer et al., 1998; Jacinto et al., 2002; Uchida and Yumura, 2004). In several lines made transgenic for him-4p::GFP::ACT-1 we observed GFP fluorescence in the isotropic bands of the BWM sarcomeres in a characteristic pattern that surrounds dense bodies (Fig. 4C) (Francis and Waterston,1985). This demonstrates that the GFP::ACT-1 fusion protein is properly localized. In the muscle arms of these animals, actin bundles were typically observed along the shaft and occasionally at the branched termini(Fig. 4A).

Actin loss-of-function mutations have no obvious phenotype, probably because of genetic redundancy. We therefore used an actin gain-of-function mutation (st15) to examine the consequences of actin perturbation on muscle arm development. The act-1,2,3(st15) allele is a mutation within the actin gene cluster act-1,2,3, although it is unknown which of the three actin genes is mutated(Landel et al., 1984; Waterston et al., 1984). We observed three muscle arm phenotypes in the act-1,2,3(st15)background. First, act-1,2,3(st15) adults had significantly fewer arms than controls, but more than L1 hatchlings(Table 2, line 10). Second, 91%of the mutant BWMs extended membrane in the general direction of the nearest nerve cord but did not make contact with it, compared with 10% of controls(Fig. 3, Table 3). We refer to this phenotype as errant membrane projections. Third, there was a dramatic arborization of muscle arm termini at the nerve cords(Fig. 3, Table 3). In addition to these arm phenotypes, 73% of act-1,2,3(st15) BWMs had flowing lateral ends(Fig. 3B), which differed significantly in morphology from the pointed lateral ends of wild-type controls (Table 3). We also examined the distribution of actin in the st15 mutant background using the GFP::ACT-1 reporter. We found that muscle arm termini in act-1,2,3(st15) mutants were enriched with actin compared with controls (Fig. 4). Moreover, we were unable to resolve individual actin filaments or actin filament bundles within the arm termini. These results demonstrate a crucial role for actin regulation in muscle arm development.

We investigated the consequences of disrupting actin dynamics on muscle arm extension. We first examined the role of ADF/cofilin, which has both actin-severing and actin depolymerization activities and is required for membrane extension in many cell types(Maciver and Hussey, 2002). The C. elegans genome encodes a single ADF/cofilin gene, unc-60. The unc-60 mRNA is spliced in two ways from a common first exon encoding only the initiator methionine to give two distinct ADF/cofilin orthologues: UNC-60A and UNC-60B(McKim et al., 1994). While UNC-60A is widely expressed, UNC-60B is specifically expressed in BWMs, vulva and spermatheca (Ono et al.,2003). Although deletions of unc-60A are lethal, all alleles of unc-60B are viable and result in defective BWM sarcomeres(Ono et al., 2003; Ono et al., 1999).

We observed three muscle arm phenotypes in unc-60B(su158) animals. First, unc-60B(su158) animals have as few muscle arms as wild-type L1 hatchlings (Fig. 3)(Table 2, line 11). This suggests that larval muscle arm extension is abrogated in animals lacking UNC-60B. Second, 70% of unc-60B(su158) BWMs have errant membrane projections (Fig. 3)(Table 3). Third, 29% of unc-60B(su158) BWMs have arborized arm termini, compared to 2% of controls (Table 3). In addition, 100% of the BWMs in an unc-60B(su158) background have flowing lateral ends, compared to none in controls(Fig. 3)(Table 3). The F-actin bundles in the muscle arms of unc-60B(su158) mutants appear disorganized in the shaft and there is an accumulation of actin fibres within the muscle arm termini, similar to act-1,2,3(st15) mutants(Fig. 4). These results suggest an important role for UNC-60B-mediated actin regulation in muscle arm extension.

Wild-type UNC-60B can both depolymerize and sever F-actin. The unc-60B(e723) mutation, however, severely reduces both these activities, while s1307 enhances both activities, and r398enhances depolymerization but eliminates severing(Ono et al., 1999). To investigate which activities are required for muscle arm extension, we examined muscle arms in these three unc-60B mutants(Table 2, lines 12-14). As expected, e723 adults had similar numbers of muscle arms to unc-60B(su158) null mutants and wild-type hatchlings(P>0.05). Conversely, unc-60B(s1307) mutants had similar numbers of muscle arms to wild-type adults (P>0.05). Interestingly, unc-60B(r398) mutants also had wild-type numbers of muscle arms (P>0.05). Because r398 retains F-actin depolymerization activity but has no severing activity, we conclude that UNC-60B-mediated severing is dispensable for muscle arm extension. One interpretation of this result is that the depolymerization activity of UNC-60B ensures that a pool of free G-actin is available for new arm extension.

Actin filament disassembly by UNC-60B is strongly enhanced by UNC-78, an actin-interacting protein 1 (AIP1) orthologue expressed in BWMs(Mohri and Ono, 2003; Ono et al., 2004). We therefore hypothesized that unc-78 mutants might also have muscle arm extension defects. Although the BWMs of unc-78(gk27) null mutants exhibit a similar number of lateral flowing ends to unc-60B(su158)(Fig. 3, Table 3), the number of muscle arms extended in unc-78(gk27) mutants did not differ significantly from controls (Table 2, line 15). This result suggests that the muscle arm defects observed in unc-60B(su158) and unc-60B(e723) mutants are not a secondary consequence of defects within the sarcomere, because sarcomere organization is similarly disrupted in unc-60B and unc-78 nulls(Ono et al., 2001).

ADF/cofilin-dependent actin disassembly is antagonized by tropomyosin,which structurally reinforces actin filaments(Bernstein and Bamburg, 1982; Blanchoin et al., 2000; DesMarais et al., 2002; Ono and Ono, 2002). We investigated whether C. elegans tropomyosin, called LEV-11 or TMY-1,might also be required for muscle arm extension. We examined muscle arms in two different lev-11(RNAi) backgrounds. lev-11(TM1-RNAi)targets the two lev-11 isoforms that are thought to be muscle-specific, while lev-11(TM2-RNAi) targets isoforms that are expressed in BWMs, the pharynx and the intestine(Kagawa et al., 1995; Ono and Ono, 2002). Both lev-11(TM1-RNAi) and lev-11(TM2-RNAi) resulted in a significant reduction in the number of muscle arms in adults(Table 2, lines 16 and 17). We therefore restricted further lev-11(RNAi) analysis to the TM1 isoforms. As the number of muscle arms in lev-11(RNAi)-treated animals is wild-type hatchlings (P>0.05), we conclude that tropomyosin is essential for larval arm extension.

Tropomyosin stabilizes actin filaments(DesMarais et al., 2002; Ono and Ono, 2002). We therefore tested whether lev-11(RNAi) could suppress the membrane extension defects observed in unc-60(su158) and act-1,2,3(st15) mutants, which probably result from excessive actin filament extension or stability. We observed that unc-60(su158);lev-11(RNAi) worms had fewer BWMs with errant membrane projections (19%vs 70%), flowing lateral muscle ends (83% vs 100%), and fewer muscle arms with arborized arm termini (17% vs 29%) than unc-60(su158) mutants alone(Table 3). lev-11(RNAi) similarly reduced the membrane-associated defects observed in an act-1,2,3(st15) background(Table 3). These results show that unc-60 and lev-11 act antagonistically in BWMs and are consistent with previous in vitro studies(Ono and Ono, 2002).

Intriguingly, the muscle arms of lev-11(RNAi)-treated animals were dramatically wider than wild-type controls (Figs 3, 4, 5, 6). These wide arms are unlikely to be an amalgamation of wild-type muscle arms, as the combined width of the remaining arms in lev-11(RNAi)-treated animals greatly exceeds the combined width of arms in wild-type worms(Fig. 6B). A model explaining the wide arms of lev-11(RNAi)-treated animals is presented in the discussion. In summary, these data indicate that lev-11 is required for both larval muscle arm outgrowth and proper muscle arm morphology.

Many of the mutants and RNAi-treated animals that we have examined are not only predicted to have disrupted actin dynamics, but are also severely uncoordinated. To test if muscle arm development is dependent on locomotion,we examined muscle arms in the background of two mutations that result in severely uncoordinated movement but were not known to disrupt actin regulation. First, we examined muscle arms in animals mutant for egl-30, which encodes an α-subunit of heterotrimeric G protein(Brundage et al., 1996). egl-30(n715) mutants were paralyzed, but had no muscle arm phenotypes, demonstrating that paralysis alone does not disrupt muscle arm development (Table 2, line 34; Fig. 6). We then examined unc-54, which encodes muscle myosin heavy chain B (MHC-B)(MacLeod et al., 1977). Surprisingly, unc-54 loss of function resulted in significantly fewer arms and greater arm widths than wild-type controls(Table 2, line 20; Fig. 6).

MHC-B is expressed in all muscles in C. elegans except for the pharynx (Ardizzi and Epstein,1987; Okkema et al.,1993) and is one of three components of thick filaments in BWMs,together with MHC-A, encoded by myo-3, and paramyosin, which is encoded by unc-15 (Kagawa et al.,1989; Miller et al.,1986; Schachat et al.,1977; Waterston et al.,1977). Because MHC-B is an integral component of thick filaments,we investigated if disruption of this structure is likely to be the primary cause of the muscle arm defects observed in unc-54 nulls. First, we examined muscle arms in unc-15 mutants, which contain MHC-B aggregates in the BWMs and do not form thick filaments(Waterston et al., 1977). Table 2 and Fig. 6 show that unc-15 loss-of-function mutants exhibited only a small decrease in the number of muscle arms and no increase in muscle arm width. This suggests that wild-type thick filaments are dispensable for muscle arm development. Next, we examined muscle arms in reduction-of-function backgrounds of myo-3. As the two available myo-3 mutants are lethal, we first examined myo-3 reduction-of-function by RNAi and found no muscle arm defects in either a wild-type background or the RNAi-sensitive background, rrf-3(pk1426) (data not shown). We then examined myo-3(st386) animals that were mosaic for a myo-3p::MHC-A::GFP rescuing extra-chromosomal array stEx30(Campagnola et al., 2002). The number of muscle arms extended by muscles expressing MHC-A::GFP was not significantly different from muscles without MHC-A (P>0.01; Fig. 7). In addition, there is no significant difference in the arm widths of muscles with or without MHC-A(P>0.01). Together, these results strongly suggest that the muscle arm defects observed in unc-54 mutants are not a secondary consequence of disrupted thick filaments. We conclude that muscle myosin heavy chain B is specifically required for muscle arm development.

We next determined when MHC-B is required for muscle arm development. The number of muscle arms in unc-54(e190) hatchlings is indistinguishable from wild-type hatchlings (Table 2, line 21), suggesting that the role of MHC-B in muscle arm extension is restricted to post-embryonic development. We then used a temperature-sensitive allele of unc-54(e1157) to determine when MHC-B is required for larval arm outgrowth. We found that MHC-B is required for muscle arm extension during early larval development and is coincidental with the period of muscle arm extension in wild-type larvae(Fig. 8).

unc-54(e190) mutants resembled lev-11(RNAi)-treated animals: the number of muscle arms was significantly reduced, the width of the remaining arms was significantly increased, the arm termini were infrequently arborized, and the BWMs did not display lateral flowing ends or membrane extension defects (Table 3). If MHC-B and LEV-11 regulate the same process, then lev-11(RNAi)-treated unc-54(e190) nulls should not have fewer muscle arms than lev-11(RNAi)-treated worms alone. Consistent with this hypothesis we found that unc-54(e190); lev-11(RNAi) animals had equivalent numbers of muscle arms to lev-11(RNAi) animals(Table 2, line 28). Furthermore, muscle arm widths of unc-54(e190); lev-11(RNAi) animals were not significantly different from those of lev-11(RNAi) animals and were not additive (P>0.05). This suggests that MHC-B and LEV-11 function together to regulate muscle arm extension and morphology.

Strikingly, unc-54(RNAi) enhanced the arm termini arborization phenotype in both act-1,2,3(st15) and unc-60(su158) mutants(Fig. 5, Table 3). This enhancement was both quantitative and qualitative, as the arborization was continuous from muscle to muscle in almost all act-1,2,3(st15); unc-54(RNAi) animals and in some unc-60(su158); unc-54(RNAi) animals. It is intriguing that these animals also had defasciculated nerve cords. Because unc-54 and unc-60B are specifically expressed in muscle and not neurons (Ardizzi and Epstein,1987; Okkema et al.,1993; Ono et al.,1999), our results suggest that the arborization of muscle arms can lead to nerve cord defasciculation. In addition, the enhancement of arm arborization by unc-54(RNAi) suggests that MHC-B normally restricts actin-based membrane extension in the muscle arm termini.

Our study provides a detailed description of muscle arm development in C. elegans and extends the pioneering work by others(Hall and Hedgecock, 1991; Hedgecock et al., 1990). At hatching, each BWM is connected to the nearest nerve cord by one or two muscle arms. The number of muscle arms is approximately doubled during the first larval stage, coincident with the birth of post-embryonic motoneurons. Each BWM is observed to extend a characteristic number of arms, and in many cases these arms emerge from the same location on the BWM and adopt a similar shape. Our work establishes muscle arm extension as a useful in vivo model of membrane extension because we show that muscle arm development relies on many of the same proteins that remodel the actin cytoskeleton in other systems.

Adults that have null mutations in unc-60B/ADF/cofilin or unc-54/MHC-B, or have been treated with lev-11/tropomyosin(RNAi), have similar numbers of arms to wild-type hatchlings. It is unlikely that these genes have redundant functions in arm extension because various double mutant combinations fail to further reduce the number of muscle arms. Rather, we propose a `two-phase' model of muscle arm development in which larval muscle arms develop by a fundamentally different mechanism from embryonic muscle arms(Fig. 9). Previous studies have shown that myoblast progenitors of the head and neck BWMs are born juxtaposed to the nerve ring. As embryonic development proceeds, these myoblasts move away from the nerve ring and leave muscle arms behind (C. R. Norris, I. A. Bazykina, E. M. Hedgecock, D. H. Hall, personal communication). As none of the cytoskeletal mutants we examined had fewer arms than hatchlings, we infer that all BWMs passively leave membrane connections behind as they move away from the nerve cords during embryogenesis (Phase I). We propose that BWMs actively extend muscle arms only during larval development (Phase II). Our model is supported by two additional observations. First, unc-54/MHC-B is required only for the extension of larval muscle arms, as unc-54mutant hatchlings have the same number of arms as wild-type hatchling controls. Second, the number of muscle arms in unc-104 mutant adults is similar to wild-type hatchlings. Because UNC-104 is required to transport vesicles that probably carry the putative chemotropic cue, the remaining muscle arms in the unc-104 mutants must not rely this cue for their development.

We have discovered unexpected roles for myosin heavy chain B(unc-54) in both phases of muscle arm development. First, MHC-B is required to regulate embryonic muscle arm morphology, as unc-54mutations result in a wide arm phenotype. The same phenotype is observed in lev-11(RNAi)-treated animals. Furthermore, lev-11(RNAi)-treated unc-54 nulls do not have wider arms than either single gene disruption alone. These results suggest that MHC-B and tropomyosin act together to regulate muscle arm morphology and are consistent with the known role of tropomyosin in mediating myosin-actin interactions in both muscle and non-muscle cells (Bryce et al., 2003; Gordon et al.,2000).

In our two-phase model of arm development, embryonic muscle arms are the trailing edge remnants of myoblast movement away from the motor axons. Intriguingly, inhibition of non-muscle myosin II activity in migratory neutrophils results in a failure to retract the trailing edge, which consequently expands in size and accumulates actin filaments(Eddy et al., 2000). It has been proposed that non-muscle myosin II generates tension within the cytoskeleton of the trailing edge and promotes dissociation of adhesion complexes from the substrate, allowing forward cell translocation(Eddy et al., 2000). We similarly propose that MHC-B, which is a muscle myosin II family member(Hodge and Cope, 2000; Sellers, 2000), also generates actin-filament tension within the developing embryonic muscle arms to restrict arm width. Our model is supported by the observations that disruption of unc-54 results in the expansion of embryonic muscle arms, the accumulation of actin within the arm terminus, and the dramatic enhancement of the actin-based arborization of termini in unc-60B and act-1,2,3 mutants.

MHC-B is also required for the active extension of muscle arms during the second phase of muscle arm development. The number of muscle arms in unc-54 null hatchlings is identical to wild-type hatchlings. Similarly, adult animals in which unc-54 was compromised during early larval development using a temperature-sensitive allele have a similar number of muscle arms as wild-type hatchlings. These results demonstrate that MHC-B is essential for larval muscle arm extension. This is the first evidence that the muscle myosin II MHC-B can play an active role in membrane extension. However, non-muscle myosin II is well known to be required for membrane extension in many cell types (Diefenbach et al., 2002; Wylie et al.,1998; Lo et al.,2004; Pollard and Borisy,2003). We speculate that unc-54 mutants fail to extend larval arms because MHC-B is needed to generate the actin-based tension required for membrane extension at the leading edge as previously described in other systems (Diefenbach et al.,2002). Our observation that unc-54(RNAi) and lev-11(RNAi)-treatments can suppress the lateral flowing BWM ends and membrane extension phenotypes seen in the actin gain-of-function mutant further supports this model.

UNC-60B/ADF/cofilin has two important biochemical activities required for the proper assembly and maintenance of the C. elegans sarcomere:F-actin depolymerization and severing (Ono et al., 1999). Although all the unc-60B alleles that we examined disrupt sarcomere development, only those alleles that inhibit the depolymerizing activity of UNC-60B have larval muscle arm extension defects. This suggests that it is the F-actin depolymerizing activity of UNC-60B/ADF/cofilin, and not the severing activity, that is crucial to membrane extension in muscle. These observations contradict a previous report that only the severing activity of ADF/cofilin is required for membrane extension in transgenic chick neurons(Endo et al., 2003). Although both studies rely on previous biochemical analysis of mutant ADF/cofilin activity, it is possible that the mutations behave differently in vivo, or have neomorphic activities that were not considered. Alternatively, chick cofilin and C. elegans UNC-60B may have either divergent properties or are regulated differently in neurons or muscle cells. In any case, our results suggest that enzymatic modulation of the actin cytoskeleton is crucial for muscle arm extension.

In summary, we have demonstrated that muscle arms are a useful model system to study the mechanics of membrane extension because they are readily observed in living animals and are amenable to genetic disruption. Indeed, this work has led to novel insights into the functions of conserved cytoskeletal components and regulators. We anticipate that the study of muscle arms in C. elegans will be valuable in uncovering new genes required for membrane extension and guided migration.

We thank Zhibin Lu, Raynah Fernandes, and Nicole Ricker for technical assistance; Jean-Louis Bessereau for bringing the him-4 promoter to our attention; David Hall for sharing unpublished results; Shoichiro Ono for the lev-11 RNAi clones and unc-60 alleles; Geraldine Seydoux for the construct encoding the GFP::ACT-1 fusion protein; Andrew Fire for vectors; and Andrew Spence for use of his equipment and many helpful discussions. We thank Brent Derry, Joe Culotti and James Craigie for comments on the manuscript. We would also like to express our gratitude to Theresa Stiernagle and the C. elegans Genetic Center, which is funded by the NIH National Center for Research Resources (NCRR), for sending us many worm strains. S.J.D. is supported by a Natural Sciences and Engineering Council of Canada doctoral fellowship. P.J.R. is a Canadian Research Chair in Molecular Neurobiology. This work was supported by a Terry Fox Foundation National Cancer Institute of Canada operating grant, a Terry Fox Foundation equipment grant, and infrastructure awards from the Canadian Foundation for Innovation and Ontario Innovation Trust.