FGF negatively regulates muscle membrane extension in Caenorhabditis elegans

Plasma membrane extension is central to both development and disease. Epiboly, ingression, axon extension and cell migration are all processes in animal development that require actin-based membrane extension at the leading edge of the cell (Pollard and Borisy,2003). This activity is often hijacked in cancerous cells to promote neo-vascularization of tumors(Folkman, 1971; Ogawa et al., 2000; Plate et al., 1992) and the metastatic colonization of healthy tissue(Keilhack et al., 2001). Many of these migrations and cell extensions are regulated by well characterized guidance cues, cell-surface receptors and downstream signaling elements that function together to control cytoskeletal reorganization(Dickson, 2002; Pollard and Borisy, 2003). However, the components that guide the migration and cell extension of other cell types remain a mystery (Rorth,2003).

Intriguingly, developing muscles in a variety of animals also extend plasma membrane in a regulated fashion. At the sites of incoming motor axons, muscle membrane extensions called microspikes or myopodia have been observed in Drosophila (Bate,1990; Ritzenthaler et al.,2000), Xenopus(Kullberg et al., 1977), rat(Uhm et al., 2001) and mouse(Misgeld et al., 2002), and probably facilitate the development of the neuromuscular junction (NMJ). Mouse myopodia are enriched with clusters of acetylcholine receptors (AChRs)(Misgeld et al., 2002),whereas in rat, agrin can elicit the microspikes and mimic the effects of incoming axons (Uhm et al.,2001). Although agrin is well known to activate the MuSK receptor tyrosine kinase and thereby promote clustering of AChRs at the postsynaptic membrane (DeChiara et al.,1996; Gautam et al.,1996), it is unknown if agrin elicits muscle membrane extension through a similar signaling pathway.

The muscles of C. elegans also extend specialized plasma membrane to motor axons (Stretton,1976; Sulston and Horvitz,1977; White et al.,1986). In wild-type adult animals, there are 95 body wall muscles(BWMs) that are distributed along the length of the worm in four rows(Sulston and Horvitz, 1977; Sulston et al., 1983)(Fig. 1). There are two dorsal and two ventral BWM quadrants. The dorsal muscles extend membrane called muscle arms to the dorsal nerve cord, where muscle arm termini harbor the postsynaptic element of the NMJ (White et al., 1976). Similarly, ventral muscles extend arms to the ventral nerve cord and form NMJs (Fig. 1). Unlike microspikes or myopodia, muscle arms are retained after the formation of the NMJ and therefore provide a convenient model system for the genetic analysis of muscle membrane extension.

Although several lines of evidence suggest that muscle arms are guided to the nearest nerve cord (Hall and Hedgecock, 1991; Hedgecock et al., 1990), the nature of the guidance system is unknown. We and others suggest that there are two phases of muscle arm development(Dixon and Roy, 2005) (C. R. Norris, I. A. Bazykina, E. M. Hedgecock and D. H. Hall, personal communication). During embryogenesis, muscle arms probably arise through a passive process that is initiated by juxtaposition between the myoblast and the nascent motor axon. As the myoblast moves from the nerve cord, muscle membrane remains connected to the nerve cord, resulting in an embryonic muscle arm. By contrast, muscle arm extension during larval development is an active process that requires regulators of the actin cytoskeleton, including ADF/Cofilin (unc-60B) (Dixon and Roy, 2005). It is unknown if microspikes, myopodia, and muscle arms share common regulatory mechanisms.

Here, we present results from two RNAi screens for genes required to regulate muscle membrane extension in C. elegans. In a small screen of 24 genes that guide other cell extensions and migrations in the worm, we found that disruptions in an FGF pathway result in ectopic membrane extensions(EMEs) from the BWMs. To find additional genes that regulate EMEs, we also screened 847 genes that result in an uncoordinated (Unc) or paralyzed (Prz)phenotype when targeted by RNAi (Kamath et al., 2003; Simmer et al.,2003). We reasoned that some of these behavioral phenotypes may result from defects in muscle membrane extension. Indeed, we found that several integin and laminin components not only have EMEs, but also have muscle arm extension defects (MADs) in which the larval muscle arms fail to extend. Our results are consistent with a model whereby integrins and laminins are needed for directed muscle arm extension to the nerve cords, while FGF signaling provides a general mechanism to regulate membrane extension.

All constructs used in this work are described in Table S2 in the supplementary material. Details of plasmid construction are available upon request. All strains used in this work are described in Table S3 in the supplementary material. Transgenic animals were generated using standard injection and integration techniques(Mello et al., 1991) and resulting integrants were back-crossed to N2 (wild-type) three times. For microscopy, worms were anesthetized in 2-10 mM Levamisole (Sigma) in M9 solution and mounted on a 2% agarose pad. All images were acquired with a Leica DMRA2 compound microscope equipped with a Retiga 1300 digital camera (Q imaging) using Open Lab (Improvision) software. Images were processed using Volocity (Improvison) and Adobe Photoshop 7.0 (Adobe Systems). All muscle arm counts were of the dorsal right-hand nerve cord. Unless otherwise indicated,all counts were of muscles 9-21 and performed in the background of RP247[trIs30], as described(Dixon and Roy, 2005).

Twenty-four RNAi by feeding vectors targeting known cell migration guidance genes were constructed by inserting 0.8-1.5 kb fragments of genomic DNA isolated by PCR from exon-rich regions of each gene into the EcoRI site of the L4440 plasmid (Timmons and Fire, 1998). These vectors were transformed into HT115 bacteria(Timmons et al., 2001) using standard protocols. RNAi by feeding vectors targeting let-60, ptp-2,soc-2, egl-15 and 14 other RTK genes (Table S1 in the supplementary material) and 847 Unc or Prz genes were from the Ahringer library(Kamath et al., 2003). A similar procedure was used for all RNAi-by-feeding experiments. L4-stage worms(the P_os) were first plated onto 6 cm MYOB-agar plates lacking food and allowed to run for ∼0.5 hour to remove any contaminating OP50 bacteria. Then, for the screens of guidance and RTK genes, four worms were re-plated into a 6 cm NGM-RNAi-inducing agar dish containing a single RNAi-inducing bacterial clone. For the screen of 847 Unc or Prz genes two P_os were re-plated into separate RNAi-seeded wells of a 12-well dish. In both cases, after three days growth at 20°C, late L4 or young adult F1 progeny were scored for the EME phenotype (+/–) at the dissection microscope. Positive hits were re-tested and the EME phenotype was quantified by counting the number of individual membrane processes extending from the body wall muscles into the lateral hypodermal space. In each case,the side of the animal was noted. Equal numbers of the left and right hand sides were counted for each strain and averaged. All statistical analyses of muscle arm counts and EMEs were performed using Student's t-test for independent samples.

The ability of sem-5 to rescue the EME phenotype was tested by injecting trIs10; sem-5(n1779) worms with a construct directing the expression of sem-5 genomic DNA (sem-5g) either in BWMs(myo3p::sem-5g) or pan-neuronally (F25B3.3p::sem-5g). The presence of the rescuing arrays were followed in the BWMs and nervous system with the co-injection reporters myo-3p::cfp and F23B3.3::dsred2, respectively. Control strains contain the co-injection markers plus pBluescript (Stratagene) DNA. The ability of egl-15 to rescue the EME phenotype in muscle was tested by injecting trIs10; egl-15(n484) worms with a construct directing the expression of egl-15(5A) cDNA (courtesy of O. Hobert) from the myo-3promoter. The presence of the rescuing arrays was followed with the co-injection reporter myo-3p::cfp. A control line was also generated using the same concentration of pBluescipt instead of myo3p::egl-15(5A)^cDNA. We also generated worms with an Ex array directing the production of Mb::YFP in the BWM in the background of oxIs14; egl-15(n484); otEx1270 and oxIs14;egl-15(n484); otEx1269 (Bulow et al.,2004). These otEx arrays drive expression of EGL-15A in the hypodermis from the dpy-7 promoter and rescue the axon defects of egl-15(n484) (Bulow et al.,2004). As a negative control, cousins lacking the otExarrays were also counted. Before counting EMEs, we selected animals that brightly fluoresced in the co-injection fluorescent channels to ensure that animals carried the extra-chromosomal array in many of their cells.

Constructs directing the BWM-specific expression of dsRNA against gfp,unc-115, sem-5 and egl-15 gene products were generated by subcloning a coding-rich genomic fragment for each gene into the pPRSD102 plasmid, a derivative of pPD95.86 that contains an expanded MCS 3′ to the myo-3 promoter, and isolating clones for gfp, unc-115, sem-5 and egl-15 fragments in both orientations with respect to the myo-3 promoter. To target mb::yfp mRNA for degradation, for example, two plasmids directing the production of RNA in opposite directions are co-injected along with the plasmid directing expression of a DsRed2 in BWMs to mark the array. All cell specific RNAi experiments were performed in a trIs10; sid-1(qt9)background. Multiple transgenic lines were isolated from each injection and from these lines, those worms that displayed robust expression of the co-injection marker in the body wall muscles were scored for EMEs.

Embryos from trIs10; egl-15(n484) and trIs10; sos-1(cs41)worms were raised at either 20°C (for up-shift experiments) or 25°C(for downshift experiments) for 12 hours, then harvested at room temperature using the hypochlorite method and plated onto MYOB agar plates lacking food(Lewis and Fleming, 1995). The eggs were allowed to hatch overnight at either 20°C or 25°C without food, causing them to arrest as L1 hatchlings. The next day, the synchronized L1 population was divided and transferred to seven different plates and stored at either 20°C or 25°C (the 0 hour time point). Individual plates were transferred between 20°C and 25°C incubators when appropriate. For example, in trIs10; egl-15(n484) up-shift experiments, one plate containing L1s harvested and synchronized at 20°C was transferred to 25°C after 6 hours (the 6 hour time point), another plate transferred to 25°C after 12 hours (the 12-hour time point), etc. All strains were scored for the EME phenotype upon reaching young adulthood.

let-756 rescue was tested using various promoters that drive let-756 expression in different cell types. RP175 and RP176 are two independent lines of the same genotype Ex[pPRRF186(let-756p::YFP);(pPRRF187(let-756p::LET-756)]; let-756(s2887) unc-32(e189)III that were used in a mosaic analysis to determine where let-756 expression was required to rescue the lethality conferred by let-756(s2887). Both RP175 and RP176 depend on the Ex array for viability, as no non-fluorescent animals are ever observed in these two strains(n>1000). We collected 47 RP175 and 95 RP176 animals that were dramatically mosaic for the Ex array, being present in only a few let-756-expressing cells. Of these, 10 RP175 animals had YFP expression exclusively in the BWMs, five had YFP expression exclusively in the pharynx and one had expression in only one CAN neuron. For RP176, 23 animals had YFP expression exclusively in the BWMs, seven had YFP expression exclusively in the pharynx and five had expression in only one or two CAN neurons. We found no RP175 or RP176 animals with YFP expression in only the G1 and G2 glandular cells or only in the intestine.

To identify genes required to regulate muscle plasma membrane extension, we first built a strain called RP1 that expresses a membrane-anchored YFP(Mb::YFP) (Dixon and Roy,2005) in all BWMs from a transgenic array called trIs10(Fig. 1). Because BWMs of the dorsal quadrants extend muscle arms to the dorsal cord, and the ventral BWMs extend muscle arms to the ventral cord(Dixon and Roy, 2005; Hedgecock et al., 1990; White et al., 1986), the left and right lateral hypodermal ridges are devoid of BWM membrane extensions in wild-type animals (Fig. 1). We therefore expected that any EMEs projecting into the lateral space as a result of disrupted plasma membrane extension would be readily visible in living RP1 animals. To test this, we examined the plasma membrane of BWMs in unc-5 mutants. In unc-5 mutants, commissural motor axons fail to extend to the dorsal cord and instead extend longitudinally along the lateral and ventral hypodermal ridges(Hedgecock et al., 1990). Muscle arms from dorsal BWMs extend to these errant lateral and ventral motor axons as a secondary consequence(Hedgecock et al., 1990). As expected, muscle arms that extend to the misguided axons are readily visible in trIs10; unc-5(e53) animals(Fig. 1), demonstrating that trIs10 enables the visualization of BWM membrane in the lateral hypodermal space.

We next investigated if genes required for the guidance of cell extensions and migrations might also play a role in regulating BWM membrane extension. Twenty-four genes required for guidance were targeted in an RNAi pilot screen using RP1 worms (see Materials and methods and Table S1 in the supplementary material). sem-5, which encodes a small adapter protein that is orthologous to Grb2 and functions downstream of receptor tyrosine kinases(Borland et al., 2001; Clark et al., 1992; Moghal and Sternberg, 2003; Stern et al., 1993), was the only RNAi target that resulted in EMEs(Fig. 1). Two sem-5hypomorphs, n1779 and n2019, both have significantly more EMEs than the reporter strain alone (P<0.001, Fig. 1C), confirming that disruption of sem-5 results in the EME phenotype. We also observe EMEs in sem-5(RNAi) animals in a reporter strain that expresses DsRed2 throughout the BWMs, demonstrating that the EME phenotype is independent of the Mb::YFP reporter (Fig. 1C). Furthermore, sem-5(RNAi) does not induce EMEs in all cells, as the plasma membrane of the seam cells and DA and DB motoneurons in animals treated with sem-5(RNAi) do not exhibit EMEs (n=100 animals, see Fig. S1 in the supplementary material). These results suggest that sem-5 is required to prevent ectopic membrane extensions from the BWMs.

The EMEs revealed by sem-5 hypomorphic activity often resemble the misguided muscle arms in mutants such as unc-5(e53). However, three lines of evidence suggest that EMEs are not a secondary consequence of misguided axons as in unc-5 mutants. First, the morphology and position of the DA, DB, DD and VD commissural motor axons in sem-5(RNAi)-treated RP1 animals is wild type(Fig. 1H,I). Second, disruption of sem-5 by RNAi in worms expressing a fluorescent reporter in all neurons (see Table S3 in the supplementary material) does not produce lateral axon guidance defects (n=10 animals). Third, EMEs do not extend to any particular lateral axon (data not shown), making it unlikely that EMEs extend into the lateral space because of altered lateral neuron fate. Together these results suggest that EMEs are not a secondary consequence of defects in neuronal development.

To determine where sem-5 is required to regulate membrane extension from the BWMs, we first examined the cells that express a YFP reporter from ∼3.0 kb of promoter and enhancer sequences of sem-5. Animals carrying this transgene express YFP in many cells throughout development and in the adult, including the vulval precursor cells,hypodermis, intestine, neurons and the BWMs (see Fig. S2 in the supplementary material). To test if sem-5 expression in muscles is sufficient to prevent the EME phenotype, we built a construct designed to express SEM-5 specifically in muscle cells from the myo-3 promoter(myo-3p::SEM-5) (Okkema et al.,1993). Three independent transgenic lines containing myo-3p::SEM-5 were isolated in the background of the sem-5(n1779) hypomorphic mutation. Each line displayed significantly fewer EMEs than the negative control (P<0.001, Table 1). By contrast,pan-neuronal expression of SEM-5 does not rescue the EME phenotype(Table 1). These results show that sem-5 expression in muscle is sufficient to rescue the EME phenotype.

To investigate if sem-5 expression in BWMs is necessary to prevent EMEs, we developed an approach to knock down gene expression in a BWM-specific manner. We built constructs designed to express double-stranded RNA sequence(dsRNA) corresponding to the coding sequence of the targeted gene specifically in the muscles. To prevent the systemic distribution of the muscle-expressed dsRNA (Fire et al., 1998) and limit the RNAi effect to the muscles, we expressed the dsRNA in the background of a sid-1(qt9)-null mutant (a kind gift from Craig Hunter). SID-1 is required for dsRNA uptake into cells(Feinberg and Hunter, 2003; Winston et al., 2002). Therefore the RNAi effect should be limited to the BWMs that express the dsRNA. To test this approach, we injected trIs10; sid-1(qt9) animals with three transgenes: two that direct the production of dsRNA against gfp and a third that expresses nuclear-localized DsRed2 in muscle to mark the presence of the extra-chromosomal (Ex) array that contains the transgenes. We expected that YFP expression would be knocked down in muscles that have the Ex array because gfp-coding sequence is 99% identical to yfp, while neighboring muscles that do not have the Ex array should maintain YFP expression. In trIs10; sid-1(qt9) animals mosaic for the Ex array, 94% of DsRed2(+) BWMs were YFP(–) (n=143),while 99% of DsRed2(–) BWMs were YFP(+) (n=79)(Fig. 2A). Conversely, in trIs10; sid-1(qt9) lines expressing dsRNA in BWMs targeting unc-115, which is not expressed in muscles(Lundquist et al., 1998) and the sequence of which is unrelated to gfp, 100% of DsRed2(+) BWMs were YFP(+) (n=201) (Fig. 2B). These results demonstrate that the RNAi effect can be restricted to a cell of interest through the cell-specific expression of dsRNA in the background of a sid-1 null mutation.

To determine if sem-5 expression in the BWMs was necessary to prevent EMEs, we isolated four independent lines of trIs10;sid-1(qt9) animals expressing sem-5(dsRNA) specifically in BWMs and found that each had significantly more EMEs than negative controls(Fig. 2). Together, our results suggest that sem-5 expression in BWMs is both necessary and sufficient to negatively regulate ectopic membrane extension from the BWMs.

SEM-5 and its mammalian ortholog Grb2 can function as adaptor proteins linking ligand-bound receptor tyrosine kinases (RTKs) to the guanine nucleotide exchange factor SOS that activates Ras within the cell(Lowenstein et al., 1992; Moghal and Sternberg, 2003; Stern et al., 1993). To determine if sem-5 regulates muscle membrane protrusions downstream of an RTK, we screened 14 C. elegans RTKs genes for the EME phenotype using RNAi (see Table S1 in the supplementary material). egl-15,which encodes the lone FGF receptor (FGFR) in C. elegans(DeVore et al., 1995), was the only RTK that resulted in the EME phenotype(Table 1). Three egl-15 hypomorphic alleles that we examined also have EMEs(Fig. 3A,B, Table 1). Although BWM spatial expression of egl-15 has not yet been reported, both SAGE analysis and Affymetrix microarray analysis of cultured C. elegans myoblasts reveals low, but significant, expression of egl-15 mRNAs (Don Moerman and David Miller, III, personal communication). Knocking down egl-15activity specifically in BWMs using the same RNAi approach outlined above for sem-5 results in significantly more EMEs than control strains(Fig. 2D). Moreover, a construct directing the expression of an egl-15 cDNA from the myo-3 promoter in the BWM rescues the EME phenotype of trIs10;egl-15(n484) worms (Table 1). Conversely, expression of egl-15 in the hypodermis does not rescue the EME phenotype (Table 1, lines 32-35). These results suggest that FGFR activity within the BWMs is required to prevent the EME phenotype.

We next tested if other known FGF pathway genes play a role in regulating membrane extension from the BWMs. There are two putative FGF ligands upstream of egl-15 encoded by egl-17(Burdine et al., 1997) and let-756 (Roubin et al.,1999). The number of EMEs in egl-17(n1377) null animals did not differ from negative controls (P>0.05). By contrast, let-756(s2613) hypomorphs had significant numbers of EMEs(P<0.001) (Fig. 3C, Table 1), suggesting that LET-756 is the ligand necessary to prevent EME formation. Numerous genes function downstream of egl-15, including the adapter proteins sem-5(Grb2), soc-1(Gab1) and soc-2(Shoc-2/Sur-8),the non-receptor tyrosine phosphatase ptp-2(Shp2), the Ras guanine nucleotide exchange factor sos-1 (Son of Sevenless), and the Ras-family GTPase let-60 (Ras)(Borland et al., 2001). We determined that soc-2(RNAi), ptp-2(RNAi) and let-60(RNAi)each induced EMEs (Fig. 3D, Table 1). EMEs were also observed in strains containing soc-1, sos-1 and let-60hypomorphic alleles (Table 1). Together, these data suggest that FGF signaling negatively regulates ectopic membrane extensions from the BWMs.

To determine when FGF signaling is required to prevent EMEs, we performed temperature shift experiments using temperature-sensitive alleles of egl-15(n484) (Goodman et al.,2003) and sos-1(cs41)(Rocheleau et al., 2002) (see Fig. 4 and Materials and methods for details). Both egl-15(n484) and sos-1(cs41)worms shifted to 25°C before the third larval stage had more EMEs than the negative control animals kept at 20°C, while those shifted after the third larval stage did not (Fig. 4A,B). Conversely, mutant L1 hatchlings raised at 25°C and then immediately shifted to 20°C for the remainder of larval development show no more EMEs than animals raised continuously at 20°C(Fig. 4C,D), indicating that the FGF pathway is not required during embryogenesis to regulate EMEs. The number of EMEs in animals left in the non-permissive temperature significantly increased at each successive time point until the middle of the third larval stage at the 24 hour time point, when the number of EMEs was not different than animals raised constitutively at 25°C. These results show that FGFR signaling is required specifically during early to mid larval development to prevent EME formation. Larval muscle arm extension occurs exclusively during this same period (Dixon and Roy,2005), suggesting that the FGF pathway functions to prevent widespread ectopic extension of the muscle membrane during this crucial period.

To investigate the effects of upregulation of the FGF pathway on membrane extension, we examined BWM plasma membrane in a clr-1loss-of-function mutant. clr-1 encodes a BWM-expressed receptor tyrosine phosphatase that negatively regulates EGL-15 activity(Kokel et al., 1998). The clr-1(e1745) temperature-sensitive allele can suppress the EMEs induced by egl-15 and sem-5 hypomorphic activity(Fig. 3E,F, Table 1), consistent with the role of CLR-1 in antagonizing ELG-15-mediated signaling. Similarly, the let-60(n1046) hypermorph is able to partially suppress the EME phenotype induced by sem-5(RNAi) and egl-15(RNAi)(Fig. 3G,H; Table 1). These data are consistent with previous work showing that let-60 functions in the FGF pathway, but does not mediate all aspects of EGL-15 function(DeVore et al., 1995; Sundaram et al., 1996).

Interestingly, clr-1(e1745) mutants have a dramatic reduction in the number of muscle arms compared with controls(Fig. 3I,J, Table 2), suggesting that larval muscle arms fail to extend in these mutants(Dixon and Roy, 2005). The MAD phenotype of clr-1 mutants is partially suppressed by sem-5(RNAi) and completely suppressed in an egl-15(n1783)background (Fig. 3L,M, Table 2). These results suggest that the level of EGL-15 activity is a key determinant of the ability of the BWM to extend plasma membrane: decreased EGL-15 activity results in ectopic membrane extensions, while increased EGL-15 activity abrogates larval muscle arm extension.

To understand better how let-756 restricts ectopic muscle membrane extensions, we first examined the let-756 expression pattern using let-756 promoter and enhancer sequence (let-756p) to drive YFP expression. Worms transgenic for let-756p::YFP express YFP in the BWMs, the pharynx, the CAN neuron and two unidentified neurons of the head as previously described (Bulow et al.,2004; Popovici et al.,2004) (see Fig. S2 in the supplementary material). Although the same cells expressed a LET-756::YFP fusion protein driven from the let-756p, we also observed nuclear localization of LET-756::YFP,along with pericellular expression around the BWMs that is very similar to that of CLR-1 (Kokel et al.,1998) (Fig. 5A-D).

We rescued the lethal phenotype of let-756(s2887) null mutants with an extra-chromosomal array that contained both the let-756p::YFP reporter and a second construct that used the same let-756p to drive the expression of let-756 genomic coding sequence. Using the let-756p::YFP reporter to mark the cells that expressed LET-756 from the rescuing array, we determined that LET-756 expression in a single CAN neuron, a single BWM or cells in the pharynx was sufficient to rescue the lethality of let-756(s2887) nulls. Importantly, we could find no let-756(s2887) homozygotes that did not express YFP(n>1000), indicating that physiologically relevant LET-756 is not expressed in cells that do not also express YFP from the array. Moreover, when only a small number of BWMs or pharynx cells had the rescuing array in let-756(s2887) homozygotes, the animal had a pronounced bulge near the site of YFP expression, suggesting local rescue of the scrawny phenotype(Fig. 5E,F). Together, our observations demonstrate that transgenic expression of let-756 from any one of a number of distinct cell types is sufficient to rescue the lethal phenotype of let-756 nulls.

As let-756 expression from distinct cell types rescued the lethal phenotype of let-756 nulls, we tested if the same was true for the EME phenotype. We observed complete rescue of both lethality and EMEs in let-756(s2887) nulls expressing LET-756 from the endogenous let-756p, in neurons from the pan-neuronal F25B3.3 promoter or in the hypodermis from the dpy-7 promoter(Fig. 5G). This suggests that LET-756 functions non-autonomously to negatively regulate EMEs from the body wall muscles.

Postsynaptic termini of the neuromuscular junction reside on the distal ends of muscle arms (Dixon and Roy,2005; Stretton,1976; Sulston and Horvitz,1977; White et al.,1986). We reasoned that disruption of muscle arm extension might therefore result in locomotory defects. In an effort to identify additional genes that regulate muscle membrane extension, we screened 847 genes that result in an uncoordinated or paralyzed phenotype when targeted by RNAi(Kamath and Ahringer, 2003; Simmer et al., 2003), 23 of which resulted in a strong EME or MAD phenotype (M.A., R.F., S.J.D. and P.J.R., unpublished). One of these genes, lam-1, encodes the only laminin β-subunit in C. elegans(Huang et al., 2003; Hutter et al., 2000) and results in significant EMEs when disrupted by RNAi or mutation(Fig. 6, Table 3). pat-4, which encodes an integrin-linked kinase ortholog(Mackinnon et al., 2002), and pat-6, which is required for the assembly of integrin complexes(Lin et al., 2003), were also identified in our screen (Table 3). These results prompted us to investigate other integrin and laminin genes (Cox and Hardin,2004; Hutter et al.,2000). epi-1 (α_B-laminin), lam-1 (β-laminin), lam-2 (γ-laminin) and pat-2 (α-integrin) each induced EMEs when targeted by RNAi and had significant reductions in the number of muscle arms reaching the nerve cord in either RNAi or hypomorphic mutant backgrounds(Fig. 6, Table 2). By contrast, other cell-adhesion genes that we examined by RNAi or mutations, such as all 13 predicted cadherin genes (Cox et al.,2004), conferred no EMEs (not shown). Similarly, disruption of ina-1 (α-integrin) by RNAi or mutation did not result in EMEs(Table 2). This was a surprising result, as the INA-1/PAT-3 integrin heterodimer probably binds laminin, while the PAT-2/PAT-3 complex probably binds UNC-52/perlecan(Baum and Garriga, 1997; Rogalski et al., 1993). However, this prompted us to examine perlecan mutants. We found that unc-52(e998) mutants phenocopy the MAD and EME phenotypes of pat-2(RNAi) (Fig. 6, Table 2). These data suggest that the PAT-2/PAT-3 integrin complex functions with laminins and perlecan to regulate muscle membrane extension.

Given the phenotypic overlap between disruptions in pat-2, the laminins and the FGF pathway, we asked if these components interact genetically. We tested if the clr-1(e1745) and let-60(n1046)mutants that upregulate FGF pathway activity could suppress the EMEs induced by epi-1(RNAi), lam-1(RNAi), lam-2(RNAi) and pat-2(RNAi). In all cases, clr-1(e1745) significantly reduced the number of EMEs induced by these RNAi treatments as did let-60(n1046), albeit less effectively (Fig. 6, Table 3). These results show that the FGF pathway interacts genetically with both integrin and laminins.

We have discovered that an FGF pathway is a general negative regulator of plasma membrane extension from the body wall muscles of C. elegans. Too little FGFR signaling results in ectopic membrane extensions and too much signaling prevents membrane extension. Using temperature-sensitive alleles of two genes in the FGF pathway, we demonstrate that FGF signaling is required during early larval development to prevent EMEs. Previous work has shown that the level of let-756 FGF mRNA peaks during the second larval stage and declines thereafter (Roubin et al.,1999), providing additional evidence that the LET-756 FGF pathway is required during this period to inhibit EMEs. Although FGFR signaling acts in muscles to regulate membrane extension, let-756 can be expressed from a variety of spatial locations to prevent the formation of the ectopic membrane extensions without compromising muscle arm extension. This argues against LET-756 acting as either an attractant or repellant of muscle membrane extension. Instead, we speculate that the interaction of LET-756 with EGL-15 acts as a switch to negatively regulate membrane extension by unknown molecular mechanisms.

Unlike the FGF pathway, we found that disruptions in laminin, integrin and perlecan result in both ectopic membrane extensions and fewer muscle arms reaching the nerve cord. These phenotypes are consistent with wayward muscle arm extensions that fail to reach their target. We speculate that the laminins, integrins and perlecan, which together constitute a significant part of the extracellular matrix and muscle adhesion complex(Cox and Hardin, 2004, Hutter et al., 2000), could be required to regulate the spatial distribution of factors that guide muscle arms to the nerve cord. Alternatively, disruption of laminin, integrin and perlecan may interfere with proper localization of receptors on the plasma membrane of muscles or with adhesion of the muscle membrane extensions to the substratum (Huang et al.,2003). Under any model, the ectopic membrane extensions resulting from compromised laminin and integrin function depend on FGF pathway activity as increased FGFR activity suppresses their extension.

Our work demonstrates a new role for FGF signaling in C. elegansin which FGFR activity regulates muscle membrane extension specifically during early larval development. During this same period, EGL-15 plays at least four other roles in the animal. First, EGL-15 acts as a receptor for EGL-17, which guides the migration of the sex myoblasts to the incipient vulva(Burdine et al., 1998; Burdine et al., 1997). The egl-15 gene encodes two isoforms by splicing between two alternative exons called 5A and 5B (Goodman et al.,2003). Null mutations that are specific to the 5A exon, including egl-15(n484), reveal that only EGL-15(5A) is necessary for the guidance of the sex myoblasts and is therefore likely to be the only isoform on the sex myoblasts required to bind EGL-17(Goodman et al., 2003). Although egl-15(n484) and let-756(s2613) mutations result in EMEs from the BWMs, the egl-17(n1377) null allele does not. Our observations therefore suggest that EGL-15(5A) is capable of receiving either EGL-17 or LET-756 ligands, depending on the cellular context. This conclusion is further supported by the fact that expression of LET-756 under the control of egl-17 regulatory elements can partially rescue the sex-myoblast migration defects (Goodman et al.,2003).

The second well-established role for FGFR signaling is in fluid homeostasis. Hyperactivity of EGL-15 results in fluid filled worms, called a clear or Clr phenotype (Kokel et al.,1998). By contrast, hypomorphic let-756 or egl-15 alleles that do not specifically disrupt EGL-15A result in a scrawny phenotype that is consistent with the worms lacking sufficient fluid(DeVore et al., 1995; Roubin et al., 1999). The null phenotype of let-756 or egl-15 is early larval lethality,which probably results from improper fluid regulation(Birnbaum et al., 2005; Huang and Stern, 2005). Recent work has demonstrated that EGL-15 activity in the hypodermis is essential for fluid regulation and that LET-756 probably signals to EGL-15 in the hypodermis from the BWMs (Huang and Stern,2004). These results are seemingly contradictory to our observations that LET-756 expression from a transgenic array can rescue the essential let-756 function from any number of cells, including a single BWM, a single CAN neuron or a limited number of cells in the pharynx. However, this difference can be reconciled by noting that the specific requirement of LET-756 in BWMs was determined by Huang and Stern(Huang and Stern, 2004) using a free chromosomal duplication, while we used a transgenic array that typically results in transgene overexpression(Mello et al., 1991). We therefore suggest that overexpression of LET-756 from few cells of any type might be sufficient to regulate fluid balance and that the ligand can travel some distance within the animal to trigger EGL-15 activity.

A third role for EGL-15 is in axon extension, guidance and the positional maintenance of two interneuron axons along the ventral midline(Bulow et al., 2004). We believe that EMEs are not a secondary consequence of compromising these roles of EGL-15 for many reasons. First, the egl-15(n484) allele has no axonal guidance defects (Bulow et al.,2004), but does result in severe EMEs. Second, the cellular focus of EGL-15 in axon extension and guidance is in the hypodermis(Bulow et al., 2004), whereas we have demonstrated that egl-15 and sem-5 are both autonomously required within the BWMs to antagonize EMEs. Third, EGL-15 is required embryonically for axon extension and guidance(Bulow et al., 2004), whereas our temperature-shift time-course experiments using temperature-sensitive alleles demonstrated that egl-15 and sos-1 are dispensable during embryonic development to prevent EMEs. Fourth, the motor axon extension and guidance defects of egl-15 nulls are relatively minor, being restricted to a single anterior commissure(Bulow et al., 2004), whereas we observe EMEs all along the entire length of the animal. Fifth, the role of EGL-15 in the maintenance of axon position is restricted to two ventral cord interneurons and does not require the intracellular domains of EGL-15(Bulow et al., 2004). These observations are inconsistent with EMEs being induced by defects in the maintenance of axon position because the egl-15(n1783) mutation that specifically disrupts the EGL-15 kinase domain(Goodman et al., 2003) results in EMEs, along with other intracellular transducers of the EGL-15 signal,including sem-5, sos-1 and let-60. Last, we did not observe EMEs projecting to any particular axon in the lateral hypodermal ridge in FGFR mutants. We therefore conclude that the aberrant membrane projections that result from compromised FGFR signaling are not a secondary consequence of defects in axon extension, guidance or positional maintenance.

Finally, an FGF pathway is required for acetylcholine receptor (AChR)expression at the neuromuscular junction in C. elegans(Gottschalk et al., 2005). egl-15 acts in parallel with cam-1, which encodes a ROR-family receptor tyrosine kinase ortholog, to regulate the presence of AChRs at the NMJ, without altering the transcription levels of AChR subunits(Gottschalk et al., 2005). We have examined several mutants involved in neurotransmission at the neuromuscular junction, including cam-1/ROR(Francis et al., 2005), unc-18/UNC-18 (Weimer et al.,2003), cha-1/choline acetyltransferase(Alfonso et al., 1994), unc-29/nicotinic AChR subunit(Fleming et al., 1997) and unc-25/glutamic acid decarboxylase(Jin et al., 1999), none of which have muscle membrane extension defects (S.J.D. and P.J.R., unpublished). This suggests that neurotransmission does not regulate muscle membrane extension. However, the work of Gottschalk and colleagues raises the exciting possibility that the FGF pathway may regulate the localization or membrane insertion of other receptors that regulate BWM membrane extension.

We gratefully acknowledge Andrew Spence for many helpful comments and use of his equipment, Michael Stern for strains and helpful comments, Oliver Hobert for strains and the egl-15(5A) cDNA, Craig Hunter for the sid-1(qt9) null mutant, and Zhibin Lu and Jason Moffat for the construction of several transgenes. We also thank 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 predoctoral scholarship from the Natural Sciences and Engineering Research Council of Canada. P.J.R. is a Canadian Research Chair in Molecular Neurobiology. This work was supported by a National Cancer Institute of Canada operating grant, a Terry Fox equipment grant, and infrastructure awards from the Canadian Foundation for Innovation and Ontario Innovation Trust.