Positioning and maintenance of embryonic body wall muscle attachments in C. elegans requires the mup-1 gene

No genes or molecules that determine how muscle cells select their appropriate attachment points during development have as yet been identified. To identify such genes, we chose a genetic approach for the following reasons: the frequency at which any gene can be identified by mutational analysis depends solely on the function of the gene and its.mutant phenotype, and is independent of whether the gene is expressed at low levels in a few cells or at high levels in many cells. Other approaches to identify molecules involved in muscle pattern formation would require assumptions on their nature, distribution and abundance, as well as a suitable assay for their function in muscle development. A mutational analysis in a simple model system provides such an assay and tools to study the structure, function and regulation of genes of interest.

The nematode C. elegans (Brenner, 1974) is well suited for the study of muscle pattern formation. The body wall musculature in C. elegans consists of a small number of precisely patterned muscles arranged in a monolayer and individual muscle cells can be observed in vivo throughout development (Sulston and Horvitz, 1977; Sulston et al. 1983; Waterston, 1988). Unlike insect or vertebrate muscles which are syncytial, each body wall muscle in C. elegans is a single mononucleate cell. In insects, the primary muscle pattern is established by muscle pioneer cells which then grow through fusion with other myocytes (Ho et al. 1983; Ball et al. 1985). Muscle pattern formation in C. elegans is thus reduced to the problem of the attachment of single cells to appropriate attachment sites.

The 81 embryonic body wall muscles in C. elegans are spindle-shaped cells arranged in four longitudinal rows (Sulston et al. 1983; Waterston, 1988). At hatching each muscle cell consists of two sarcomeres, each of which is made up of two longitudinal bands of thick filaments and three longitudinal bands of thin filaments (Priess and Thomson, 1986). The myofilaments of body wall muscles attach to the hypodermis through multiple dense bodies distributed along the boundary between the muscle cells and hypodermis. Additional attachment structures resembling half-desmosomes attach the muscle cell to the hypodermis all along their interface (reviewed in Waterston, 1988).

The reproducible positioning of the embryonic body wall muscles, including their anteroposterior order, spacing and their segregation to the dorsal or ventral epidermis could be an accident of their birth positions or could also reflect intrinsic differences in epidermal preference (Hedgecock et al. 1987). There are four cases in the embryonic lineage where sister muscle cells attach to opposite dorsal and ventral quadrants apparently by active movements (Sulston et al. 1983; Hedgecock et al. 1987). This suggests that muscle positioning may involve interactions between muscle and hypodermis.

Vertebrate and invertebrate muscles have very similar structural components in the myofibrils (e.g. myosin and titin/twitchin) and muscle attachments (e.g. vinculin and α-actinin) (Macleod et al. 1979; Francis and Waterston, 1985; Barstead and Waterston, 1989; Benian et al. 1989). Cell adhesion and neuronal guidance molecules (e.g. integrin and laminin) are also conserved between vertebrates and invertebrates (Bogaert et al. 1987; reviewed in Harrelson and Goodman, 1988; reviewed in Dodd and Jessel, 1988; Hedgecock et al. 1990). The genes involved in positioning muscle attachments may therefore be similarly conserved.

We screened for larval lethal mutants with a disrupted body wall muscle pattern but otherwise wildtype anatomy (pharyngeal muscles, intestine, hypodermis and cuticle). These specimens can be recognized under the dissecting microscope because they have permanent bends or kinks in the body causing them to curl up in a characteristic manner. This phenotype was first observed by Ed Hedgecock (E. Hedgecock, personal communication), who named this class of mutants Mup (muscle attachment position). We selected, for further analysis, mutants with body wall muscles that are able to contract, but have ectopic processes that make morphologically wild-type attachments in inappropriate positions on the body wall. We report here the genetic and phenotypic characterization of mup-1, a gene that may have a role in both the positioning and the maintenance of body wall muscle attachments.

Worm strains were maintained and standard crosses performed as described in Brenner (1974). The wild-type parent of all strains used was C. elegans variety Bristol strain N2 (Brenner, 1974). All experiments were done at 20°C unless otherwise stated. Alleles used were as follows. LGI: dpy-5(e61); LGII: mup-l(e2347, e2430ts, e2434, e2436, e2438, e2439, e2485ts, e2486, e2487 and e2488) (this work), tra-2(ql22gf), lin-31(n301), unc-85 (el414am), unc-4(el20), dpy-10(el28 and m481::Tcl), dpy-2(e8), dpy-25(e817sd) unc-52(e444 and e669am); LGIII: lon-l(el85), daf-2(el368ts)’, LGIV: unc-5(e53), unc-31(e928), him-8(el489)’, LGV: dpy-ll(e224) \ LGX: mup-2(e2346ts). Rearrangement used: nDf3, (Greenwald and Horvitz, 1980).

L4 Daf-2(el368ts) hermaphrodites were mutagenized by standard EMS mutagenesis. Fj L4 hermaphrodites were incubated individually at 25°C for 2 to 3 days. All F₂ offspring which grow beyond the L2 stage arrest as dauer larvae. This allows slow growing worms and larval lethals with a Mup phenotype to be easily noticed on these plates. The mutations were recovered from the siblings by shifting them to 20°C. The use of daf-2(el368) for this screen was suggested by L. Avery.

were isolated in separate screens of 600 and 1300 F, clones, respectively.

Alleles e2438, e2439 and e2485ts were isolated in noncomplementation screens for mup-2 alleles. 600 F, animals from mutagenized unc-31(e928) parents were allowed to lay some eggs before they were mated to mup-2(e2346ts)/0 males which were grown at 15 °C. Mutations were recovered from heterozygous siblings from plates that have Mups on them. Allele e2438 was isolated in this screen. Allele e2439 was isolated in a similar screen (1000 clones) but daf-2(el368ts) parents were used instead of unc-31(e928) parents. Allele e2485ts was isolated from a similar screen of 1000 clones using daf-2(el368ts) parents and him-8(e!489); mup-2(e2346ts)/0 males instead of mup-2(e2346ts)/0 males.

The phenotype of nDf3/e2430ts ranges from Mups to viable Dpys with no Mup phenotype. New alleles of mup-1, including complete loss of function alleles, can thus be detected in this screen and recovered by outcrossing viable mup-l(new)/mup-1 (e2430ts) to N2. N2 L4 males and hermaphrodites were mutagenized and mated to generate males in which both chromosomes II were mutagenized. Individual F, males were mated with two or three e2430ts/tra-2(ql22gf) females on a 2.5 cm Petri dish. Tra-2(ql22gf) is a dominant feminizing mutation. All F, offspring were therefore produced from outcrossing. We screened 1080 matings (2160 chromosomes) for Dpys and Mups and isolated five mup-1 alleles, e2434, e2436, e2486, e2487 and e2488.

This scheme, suggested by T. Schedl, proved useful for non-complementation screens for autosomal embryonic or larval lethals and for testing complementation of such lethals.

The forward mutation rate was based only on the noncomplementation screens because, in the earlier screens, some alleles may have escaped our attention.

mup-1 alleles e2347 and e2430ts were mapped to chromosome II using the standard mapping strains, dpy-5 (e61); unc-4(el20); lon-l(el85) and unc-5(e53); dpy-11 (e224); lon-2(e678).

mup-1 (e2347) was positioned on the left arm of chromosome II by cft-two-factor crosses with unc-4(e120) and unc-85(el414amf. complete broods of mup-l(e2347) unc-4(el20)/ + + gave 1960 wild-type and 125 Unc-non-Mup offspring. Complete broods of mup-1 (e2347) unc-85 (el 414am)/ + + gave 1021 wild-type and 31 Unc-non-Mup offspring. The combined data placed mup-1 2.5 m.u. left of unc-85. mup-l(e2430ts) was placed close to or left of lin-31 since all 8 Unc-non-Lin recombinants from mup-1 (e2430ts) + + / + lin-31(n301) unc-85(el414am) picked up mup-1 (e2430ts) and none of the 7 Lin-non-Unc recombinants picked up mup-l(e2430ts). mup-l(e2347) complements unc-85(el414am), lin-31(n301) and dpy 25(e817sd), the only dpy gene in the region. The position of the mup-1 gene relative to lin-8 and lin-31 was not determined.

All mup-1 alleles were mapped close to or left of unc-85 in three-factor crosses with unc-85(el414am) dpy-10(el28). All Dpy-non-Unc recombinants from unc-85(el414am)+dpy-10(el28)/4-mup-1+hermaphrodites segregated Mups and none of the Unc-non-Dpy recombinants segregated Mups. All mup-1 alleles were complemented to mup-1 (e2430ts) by mating mup-1 (e2430ts) males to mup-1 +/+ tra-2(ql22gf) females and scoring for the presence of Mups.

Genetic map of the left arm of chromosome 2 indicating the map position of mup-1 gene.

L4 e2430ts and e2485ts hermaphrodites with wild-type muscle phenotype were singled and grown at 15 °C, 20 °C or 25 °C. Each hermaphrodite was transferred onto a fresh plate after each day of egg-laying. The complete brood size of each hermaphrodite was scored by counting the offspring of each category: (1) unhatched threefold Mups, (2) Mups lethal at the LI larval stage, (3) larval Mups that grow beyond the LI stage, (4) viable adults with bends and kinks in the body and (5) animals with no muscle phenotype (Table 1). Wild-type L3–L4s were counted and removed, while the remaining hatched and unhatched Mups were allowed to develop for 1 or 2 days before scoring. This allows the slow growing Mups to hatch or reach their terminal larval/adult stage.

L4 mup-1/ + hermaphrodites were singled and the complete brood of each hermaphrodite was placed into categories as above.

Offspring of genotypes nDf3/mup-1 (e2434) and nDf3/ mup-l(e2436) were obtained by crossing a single mup-1(2434)/+ male or mup-1 (e2436)/ + male to a single nDf3 / tra-1 (ql22gf) female, mup-1 (e2434)/mup-1 (e2436) offspring were similarly obtained by mating a single mup-l(e2434)/ + male to a single mup-l(e2436)/tra-2(ql22gf) female. All offspring were generated through outcrossing and were scored as above.

nDf3/+ males were not used for the cross because these are not as healthy as wild-type males. nDf3/ + hermaphrodites also produce more than 25 % embryonic and larval lethals. This sickness of nDf3/+ animals is not an unusual effect of large deletions and has made the use of nDf3 to show that the strong mup-1 alleles are nulls less straightforward.

mup-1 (e2430ts, e2436 and e2347); dpy-ll(e224) doubles were constructed while mapping mup-1 to a chromosome, mup-l(e2347) dpy-2(e8) was constructed by standard two-factor crosses between mup-l(e2347) +/+ dpy-2(e8).

mup-1 dpy-10(el28) double homozygotes were obtained from standard three-factor crosses between mup-l + +/ + unc-85(el414am) dpy-10(el28). The offspring of Dpy-non-Unc recombinants (genotype: mup-1+dpy-10/ + unc-85 dpy-10) were checked for suppressed Mups. mup-1 (e2430ts, e2485ts, e2436 and e2347) were suppressed by dpy-10 but mup-l(e2488, e2486, e2487, e2438, e2434 and e2439) were not.

The Mups completely suppressed by the dpy genes were fertile piggy Dpys with a wild-type muscle pattern. The genotype of these were checked by outcrossing them to N2 males. Viability of the double mutants was determined by scoring the percentage of Dpy-Mup offspring with no bends and kinks in the body from viable double mutants.

mup-1 +/+ unc-52 were constructed by mating mup-1 (e2347, e2436 or e2434)/+ males into unc-52(e444 or e669am) homozygotes. Non-Unc outcrossed offspring were picked and allowed to self. The progeny were screened for weak Dpys. 20 mup-l(e2436)/unc-52(e444) hermaphrodites produced 362 suppressed Mups amongst a total of about 1500 homozygous Mup offspring. This gives 24 % suppressed Mups compared to 2.2% from e2436/+ hermaphrodites (see Table 1). mup-l(e2347) +/+ unc-52(e444) produced homozygous viable Mups at twice the frequency compared to e2347/+ hermaphrodites (Table 1). unc-52 (e669am) does not rescue the Mup phenotype of mup-1 (e2436 or e2347) animals, mup-l(e2434) Mups were not suppressed by unc-52 (e444).

Animals were anesthetized with 0.5–1 % l-phenoxy-2-propanol and mounted on 5 % agar pads for observations with polarized and Nomarski light microscopy as described in Sulston and Horvitz (1977), Sulston et al. (1980) and Sulston et al. (1983). We used a Zeiss Axiophot with a 20x Neofluar or a 40 x Planapochromate Oil objective for polarized light microscopy and a 63 x Planapo Oil objective or a 100 x Plan Neofluar Pol objective for Nomarski and immunofluorescence microscopy. A Brace-Kohler or Senarmont compensator was used as described in Sulston and White (1980) to enhance contrast of the biréfringent muscles. Photographs were taken with Kodak Technical Pan film (developed with Kodak HC110).

To obtain consistent amounts of homozygous mup-1 embryos, L4 larvae of mup-1/+ were singled on 2.5 cm NGM plates seeded with OP50. After 3 days, those that did not produce mup-1 offspring were discarded, mup-1 (e2430ts) was grown as a homozygous strain. When the F_t were at their maximum egg production, embryos and adults were harvested by flooding the plates with M9 medium and rubbing the eggs off the agar with a plastic disposable inoculation loop. Approximately 1/6 of the eggs on the plate would be mup-1 homozygotes. We generally used 10-40 plates per preparation.

The eggs, larvae and hermaphrodites were spun down in a glass 30 ml corex tube and resuspended in 25 ml bleach solution (7 ml 13% (industrial) bleach and 1.25 ml 4 N NaOH made up to 25 ml with distilled water). After 2 min, the tube was shaken by hand for a few seconds to enhance the breakup of the hermaphrodites. When most hermaphrodites were broken up (6-8min), the solution was divided into two aliquots in 30 ml corex tubes. To each tube, 15 ml ice-cold 2 M sucrose/lOOmM NaCl was added. After gentle mixing, a layer of 1-2 ml of distilled water was placed on the sucrose cushion. Intact embryos were floated on the sucrose cushion by a 5-10 min centrifugation at 6000 revs min^-1 (bench Sorvall centrifuge). This sucrose step was essential to remove bacteria and decaying embryos, larvae and adults, which form a pellet at the bottom of the tube.

The embryos were treated gently in the following steps. They were collected in a 5 ml glass centrifugation tube, rinsed twice in 100 mM phosphate buffer, pH 7.0 (by gentle centrifugation). They were then fixed for 5-10 min in 3% freshly dissolved paraformaldehyde in 100 mM phosphate buffer, pH 7.0 and rinsed twice in distilled water (by centrifugation). As much supernatant as possible was removed and the glass tube was then chilled on ice. 3 ml of methanol (precooled to −20°C) was quickly pipetted on the pellet of embryos. The embryos were kept at −20°C for 10min to several days before use. With some but minimal loss of quality, the embryos can be stored in methanol (or acetone) at −70°C for several weeks. Depending on the antibodies, acetone was used instead of methanol or the fixation with formaldehyde was omitted. This procedure permeabilized more than 80% of the embryos for antibody staining and yielded, contrary to the freeze-fracture methods, intact embryos. This protocol was developed in consultation with R. Barstead. The larva in Fig. 3E,E’ was fixed and prepared for staining as described in Priess and Hirsh (1986).

The following steps were performed in 1.5 or 0.6 ml Eppendorf tubes. The embryos were rehydrated through a 80%, 50%, 20%, 0% methanol (or acetone) series and concentrated through low-speed centrifugation. They were resuspended in 0.05 % Tween 20+2 % milk powder (Brand Marvel) –PBS pH 7.2 and incubated with primary antibodies for 2h at room temperature or preferably overnight at 4 °C. They were washed 5 times in Tween–Marvel–PBS and incubated for 2h at room temperature with the secondary antibodies in Tween-Marvel-PBS. Secondary antibodies used were FTTC-labeled goat anti-mouse IgG (Sigma) and rhodamine-labeled goat anti-rabbit IgG (Sigma). After three 20 min washes in Tween–Marvel–PBS followed by three 5min washes in PBS, pH7.2, the embryos were equilibrated through a 20 %, 50 %, 70 % glycerol series in PBS, pH 8.0 and mounted on microscope slides in 90% glycerol-PBS, pH8.0 with Imgml^-1 paraphenylene diamine (for fluorescein-conjugated second antibodies) and/or 1mg ml⁻¹ propylgallate (for rhodamine-conjugated second antibodies). The slides were sealed with clear nail polish.

mAb MH4 binds intermediate filament-like structures on the hypodermal side of the muscle attachment (Waterston, 1988). mAb MH25 stains the intramembranous or extracellular regions underlying dense bodies, M lines and cell margins where thin filaments terminate (Francis and Waterston, 1985). mAb Ne2/lb4.14 stains a filamentous antigen in the seam cells in late threefold stage embryos (H. and R. Schnabel, personal communication). mAb MH27 stains the cell boundaries of all hypodermal cells and all desmosomes in the pharynx and intestine (Priess and Hirsh, 1986; Waterston, 1988). mAb Ne8/4c6.3 (from the MRC-LMB collection) binds a muscle antigen with a similar distribution to paramyosin in body wall muscles and a filamentous nuclear antigen in cleaving embryos. Because of the exceptionally low level of background binding of Ne8/4c6.3 compared to the other myofilament antibodies tested, mAb Ne8/4c6.2 allows the visualisation of the shape of the muscle cells (muscle arms and ectopic processes) as well as the development and organization of the myofilaments in early embryos. It does not bind to the pharynx and is therefore unlikely to be a paramyosin antibody.

The monoclonal antibodies were used alone or in double and triple combinations with polyclonal rabbit antisera raised to C. elegans myosin (R6-2; Waterston, 1989) or paramyosin (R224; Kagawa et al. 1989).

Worms were washed from NGM plates in PBS, pH 7.2 fixed for 2–4 h in 3 % paraformaldehyde in 100 mM phosphate buffer, pH 7.0 and rinsed in distilled water. The worms were permeabilized through a 20 %, 40 %, 60 % and 70 % ethanol series followed by rehydration through an ethanol series in the reverse order. They were then incubated for 2–4 h at room temperature (or overnight at 4°C) in 0.1% Tween–20–PBS, pH7.2 with 1–4 μg ml⁻¹ FITC-phalloidin (Sigma), rinsed2–3 times in PBS, pH 8.0, and then taken through a 20%; 50%; 75% glycerol-PBS, pH 8.0 series before mounting in 90% glycerol-PBS, pH8.0 with 1mg ml⁻¹ paraphenylene-diamine.

Animals were prepared for electron microscopy as described in White et al. (1986).

The muscle cells become arranged in four longitudinal rows roughly two cells abreast by the end of gastrulation (at the bean to early comma stage). Fig. 1 illustrates wild-type embryonic muscle and hypodermal development. In wild-type late bean/early comma stage embryos, paramyosin filaments criss-cross throughout the muscle cells (Fig. 1A’,B’)-Some muscle cells have a patch of paramyosin staining underlying a patch of intermediate filament (MH4) staining in alternate hypodermal cells. This is the earliest stage at which we observed co-localization of components of the myofilament lattice with the MH4 pattern in the hypodermis (390min, Fig. 1A,A′). The hypodermal cells in which the patches of MH4 staining first appear are, depending on the side of the embryo, alternating even- and odd-numbered cells (counting from the deirid) (Fig. 1A). Such a pattern of MH4 staining at 390 min of development could come about through a much earlier interaction between the muscles and hypodermis before the dorsal hypodermal cells interdigitate (i.e. before 310min; Sulston et al. 1983). This may help establish, before elongation takes place, the register between the hypodermal cells and the underlying muscle cells. The paramyosin filaments are located on either side of the longitudinal cell boundaries between neighbouring muscle cells (which at this stage are two cells abreast over most of their length in each muscle row). The shapes of the muscle and hypodermal cells change and their relative lengths increase during elongation (see Fig. 1). Some adhesion between the muscle cells and the hypodermis may therefore be required to maintain their register as the embryo elongates.

During elongation, as the dorsal hypodermal cells fuse to form hyp-7 (Singh and Sulston, 1978; Priess and Hirsh, 1986), the MH4 pattern becomes two thin lateral bands in the hypodermis (Fig. 1C–F). The randomly oriented paramyosin filaments in the muscle cells disappear early in elongation and most become concentrated in two bands underneath the MH4 bands (Fig. IC). Fig. 1A–F shows that both the bands of MH4 staining in the hypodermal cells and the underlying bands of paramyosin filaments remain narrower than the cells that produce it. This reflects a localized cell adhesion between the muscle cell and the hypodermis.

As the embryo elongates from the comma to the -fold stage, the myofilaments become attached to the hypodermis. By the U-fold stage the myofilaments are organized in sarcomeres parallel to the anterior-posterior axis. This staining pattern does not change as the embryos elongate to the threefold stage (Fig. 1D-E). The establishment of the pattern of muscle cells and their attachment to the hypodermis thus takes place early in morphogenesis before and during elongation. During elongation, the cells become progressively ‘flattened’ against the hypodermis, increasing the demarcation of the dorsal and ventral bands of muscle cells. In the head region, muscle arms project into the nerve ring from the comma stage onwards. Processes extending across the lateral seam into neighbouring muscle rows are rarely observed in wild type after the comma stage.

We have isolated and characterized ten alleles of the mup-1 gene. Five alleles, e2430ts, e2485ts, e2347, e2438 and e2439, were recovered in various screens for mutants with a Mup phenotype, mup-1 larvae have a smooth cuticle, a functional wild-type pharynx and a general wild-type anatomy but have inappropriately attached (Figs 2A,B, 3) or detached muscles (Figs 2B–D, 4). These ectopic body wall muscles can contract and make functional attachments across quadrants, mup-1 mutants can be easily distinguished from most other Mup mutants because they are dpy (for dumpy, i.e. shorter and wider than wild type) (Fig. 2).

mup-1 (e2430ts) is a weak allele and can be grown as a homozygous viable strain (see below). Larvae with the genotype e2430ts/nDf3 (nDf3 is a deficiency deleting the mup-1 gene; see Methods) range from Mups to viable Dpys. This shows that a screen for mutations based on their failure to complement e2430ts (noncomplementation screen) for Mups or Dpys can yield null alleles, irrespective of whether the loss of function phenotype of the locus is wild type, Mup or embryonic lethal. Five more mup-1 alleles e2434, e2436, e2486, e2487 and e2488 were isolated in screens of 2160 mutagenized chromosomes (see Methods).

The ten mup-1 alleles have a similar Mup phenotype, do not complement e2430ts and were all mapped left of unc-85 on chromosome II (see Methods), mup-1 (e2347) maps 2.5 map units left of unc-85 and complements known genes in the region (see Methods), mup-1 is therefore a new genetic locus.

A certain proportion of mup-1 embryos arrest in the late threefold stage and do not hatch. They are identical in phenotype to hatched LI Mups. These mup-1 embryos may fail to hatch because the head is held bent by inappropriately attached muscles (Fig. 2), restricting pharyngeal secretory activity that is required for hatching (Wood, 1988). Hatched animals may die early because the pharynx cannot pump well while those that can pump well survive longer, but may succumb later to problems of feeding, moulting or constipation. In the weaker alleles, more embryos hatch and grow through the different larval stages. The body wall musculature of larval or adult Mups does not deteriorate, suggesting that there may be little or no requirement for mup-1 function in muscle attachment and growth during larval development.

All the ten mup-1 alleles are recessive (25 % of the offspring of mup-1 /+ parents have a Mup or Dpy phenotype; Table 1) and represent an allelic series of reduction of mup-1 function:

Mup-1 allele e2430ts is the weakest. Its expressivity ranges broadly from very weak Dpys with wild-type muscles to larval lethal Mups with muscles that are detached from the hypodermis, or are attached in inappropriate places. The most severely affected e2430ts Mups are similar in phenotype to the strongest mup-1 alleles. The weakest e2430ts animals have an average broodsize (247, n=ll;s.D.=39at 20°C) similar to that of wild type (280 at 20°C; Byerly et al. 1976). mup-1 (e2485ts) is a stronger allele than e2430ts. Like e2430ts, e2485ts has a wide expressivity. The two alleles,

mup-1 (e2430ts) and mup-1 (e2485ts), are both cold and heat sensitive, with a more severe phenotype at 15°C and 25 °C than at 20 °C (Table 1). The other eight alleles are stronger and non-conditional.

e2430ts, e2485ts, e2436 and e2347 are progressively stronger alleles with a decreasing proportion of homozygous animals that grow beyond the LI stage or are viable adult Mups and Dpys. Most of the homozygous mup-1 animals of the strong alleles (see Table 1) arrest as unhatched threefold Mups or LI Mups and none of the e2434 and e2439 embryos develop beyond the LI larval stage. Since the unhatched threefold Mups have a similar phenotype to the LI mup-1 lethal larvae, the strong alleles are almost identical in phenotypic strength.

The phenotype of the weak alleles over the stronger ones is intermediate between the two alleles. e2436/ e2434, e2436/nDf3 (Table 1) and e2430ts in trans to the other mup-1 alleles all have a phenotypic strength that is intermediate between the two alleles (data not shown).

We use a combination of genetic criteria to determine the loss of function phenotype of the mup-1 gene:

1. The strong allele e2434 behaves like a deficiency deleting the mup-1 locus (nDf3) when placed in trans to a weak mup-1 allele, e2436. Furthermore, the phenotype of the strong allele e2434 is identical to that of e2434/nDf3 (Table 1).

2. All the six strong mup-1 alleles are recessive. They were isolated at a forward mutation rate of 5/2160 mutagenized haploid genome (see Methods), five times higher than that of most C. elegans genes of 1/2000 mutagenized haploid genome by EMS mutagenesis (Brenner, 1974; Meneely and Herman, 1979; Greenwald and Horvitz, 1980). Since gain-of-function alleles are rare, it is likely that these strong alleles are strong hypomorphs or amorphs.

3. From non-complementation screens, which allowed the isolation of null alleles irrespective of their phenotype, we isolated five alleles, of which four were strong.

4. The ten alleles form an allelic series with the strong alleles being identical in phenotype.

5. The weak alleles of mup-1 are suppressed by dpy-10 while the strong ones are not (see below). Hypomorphic alleles of glp-1 are similarly suppressed by dpy-10 and dpy-11 while complete loss-of-function alleles are not.

Based on this combined genetic evidence, we propose that the strong alleles are strong hypomorphic or amorphic mup-1 alleles while the remaining ones are weak hypomorphic alleles.

Homozygous e2430ts offspring from e2430ts/ + hermaphrodites are more viable than the offspring of homozygous e2430ts hermaphrodites. This is also the case with e2485ts (Table 1). This weaker phenotype is most probably due to wild-type mup-1 protein or mRNA contributed by e2430ts/ + hermaphrodites to e2430ts oocytes. Temperature-sensitive period analysis of another Mup mutant mup-3(e2431ts) also shows a maternal effect (Goh and Bogaert, unpublished data). All stronger mup-1 alleles are maintained as unbalanced heterozygous strains (because there are no balancers in the region). Since each heterozygous parent carries one wild-type copy of the mup-1 gene, homozygous mup-1 embryos from these parents acquire an equal amount of mup-l(+) product from their parent, independent of the strength of the mup-1 allele. The observed Mup phenotype of the putative null alleles may therefore reflect the consequence of the lack of zygotic mup-1 function and not the loss of both maternal and zygotic mup-1 function. Mosaic analysis of hermaphrodites homozygous for the mup-1 (null) in the germ line or a temperature-sensitive null mup-1 allele would be required to establish the null phenotype of the mup-1 gene.

mup-1 gene function appears not to be required for postembryonic development (attachment during muscle growth and moulting): e2347/ + hermaphrodites occasionally produce dumpyish worms, which are homozygous e2347 (checked by outcrossing to wild type). The body wall muscles of these viable mup-l(e2347) homozygotes are wild type and do not degenerate. From selfing they segregated 100% Mup lethals with a strong Mup phenotype. The muscles in the viable mup-1 larvae of the weaker alleles (e2430ts, e2485ts and e2436) which are not detached also retain wild-type muscle attachment as the muscles grow. As the volume of muscle contractile filaments grow, the contractile force increases much more rapidly than the increase in the area of attachment. Progressively stronger adhesion is thus expected to be required for maintaining muscle attachment. If mup-1 encodes a major component required for muscle attachment, we would expect the muscles to detach or degenerate progressively as the viable mup-1 larvae grow. The mup-1 gene is not likely to be essential for muscle function or myofilament assembly since mup-1 embryos undergo wild-type morphogenesis and postembryonic muscle growth. C. elegans embryos carrying mutations in genes that are essential for muscle function and assembly, such as myosin heavy chain A mutants, fail to elongate (Waterston, 1989).

All heterozygous male offspring from outcrossing homozygous mup-l(e2347) to wild-type males were wild-type and no weak male Mups were observed. This indicates that zygotic expression of mup-1 (+) is sufficient to rescue the Mup phenotype of mup-l(e2347). Since e2347is a strong but not a null allele, we cannot be sure that a total elimination of maternal mup-1 function can be complemented by a wild-type zygotic mup-1 activity.

The pattern of muscle and hypodermal cells and the colocalization of muscle filaments and the hypodermal antigens were studied by immunofluorescence staining with antibodies to various muscle and hypodermal components (see Methods). We distinguish an early muscle position defect (Fig. 3) and a later muscle detachment phenotype (Fig. 4) in mup-1 embryos:

In comma to twofold stage embryos, one or more ectopic processes containing myofilaments extend dorsalwards or ventralwards from the muscle cell body (Fig. 3A,B). These resemble muscle arms except that they are often forked. It is important to note that in the comma to twofold stage embryos, the muscle cell shape, myofilament organization (Fig. 3B) and the colocalization of muscle filaments and the hypodermal intermediate filament staining are wild type (as in Fig. 1, data not shown), except for this muscle position defect. We have occasionally observed mup-l(e2434 and e2430ts) embryos at the comma to the li-fold stage in which a large number of muscles have ectopic processes (Fig. 3C,C′). Threefold stage mup-1 embryos can exhibit the muscle position and detachment (see below) phenotypes simultaneously (Fig. 3D). Fig. 3E,E′ shows a mup-1 (e2434) larva with extensive ectopic muscle processes which contain myofilament. Individual body wall muscle cells can have simultaneously wild-type as well as ectopic myofilaments and muscle attachments to the hypodermis (Fig. 3F,F′). The attachments of ectopic body wall muscle processes can be compared to that of a male diagonal sex muscle (Fig. 6). The morphology at the electron microscope (EM) level of ectopic mup-1 (e2347) body wall muscle attachments (Fig. 6A,B) and that of male diagonal sex muscles (Fig. 6C) are identical. Half desmosomes linked by intermediate filaments anchor the muscle across the hypodermis to the cuticle. Dense plaques separated by a basement membrane are observed at the boundary between the muscle and hypodermis. Myofilaments are inserted into these dense plaques. This shows that mup-1 (e2347) body wall muscles make morphologically wild-type but ectopically positioned muscle attachments. Ectopic processes contract extensively, sometimes causing temporary indentations in the cuticle where these muscles are attached.

We observe ectopic processes in about 5 % of mup-1 embryos. This is likely to be an underestimate: in our preparations, only 1/6 of the embryos are homozygous mup-1 and only a fraction of these are oriented in such a way that the presence of ectopic muscle processes or the organization of the myofilament lattice can be examined (see Methods). The frequency of Mup embryos with ectopic processes appears to be higher in mup-l(e2430ts) embryos derived from homozygous mup-l(e2430ts) mothers than in Mups derived from heterozygous mup-l(e2434)/ + mothers (data not shown).

Until the twofold stage, most mup-1 embryos have a wild-type myofilament organization. In some muscle cells of mup-1 embryos beyond the twofold stage, the myofilament organization becomes progressively disorganized (Fig. 4A–C). Muscles may become detached in the late threefold stage. This second phenotype is distinct from the muscle position defect and may occur separately (Figs 2C,D, 4A–C). In the Mups with the strongest muscle detachment phenotype, all the muscle cells of one row, usually the ventral (60% of the embryos), are detached and some have a rounded up cell body some distance away from the hypodermis (Fig. 4). In Mups with a weaker terminal phenotype, the muscles from the dorsal or ventral quadrants detach from the hypodermis as a sheet (Fig. 2C,D), preventing unfolding at hatching.

The ectopically attached body wall muscles in the early stage embryos may be due to a defect in positioning muscle attachments and the later detachment phenotype may be due to insufficient adhesion of muscle cells to the hypodermis. Mups thus remain curled up because of muscle processes attached across quadrants (Figs2A,A′, 3E) or because of the detachment of muscles as a sheet (Fig. 2C–D).

mup-1 animals have a wild-type pharynx and intestine and, apart from the muscle defects, appear to be healthy dumpyish worms with a smooth, non-blistered cuticle (Fig. 2). Staining with various mAbs to muscle and hypodermal components (see Methods), compared to those of wild-type embryos (as shown in Fig. 1), shows that mup-1 embryos also have a wild-type arrangement of hypodermal cells and seam cells during gastrulation (data not shown). Elongation and hypodermal fusion is also wild-type, except that, in some specimens, the seam is wider, or the two lateral seams are fused dorsally in a region of the body where muscles are absent or disorganized (data not shown). We have observed this phenotype also in other Mup mutants (Bogaert and Goh, unpublished data) and consider it a secondary consequence of the absence of a patch of muscles on the hypodermis. Contact between the muscles and the hypodermis may be important in morphogenesis to squeeze the hypodermal cytoplasm into ridges and determine the shape of the hypodermal cells. EM sections through mup-l(e2347) embryos have shown disorganization in the nervous system around the pharynx and in the nerve bundles (data not shown). We have not examined this in detail. Pleiotropic effects in the nervous system of mup-1 embryos would not be surprising because the disruption of the muscle pattern would affect the general body plan. It remains to be established whether mup-1 also has a primary requirement in the nervous system (i.e. occurring before the muscle phenotype).

Mutations in three dpy genes suppress the mup-1 muscle phenotype: double mutants of mup-1 (e2347) with dpy-10(el28 or m481::Tcl), dpy-2(e8) or dpy-U(e224) are viable strains, mup-1 (e2347)-dpy doubles are stronger Dpys than the Dpys alone, dpy-2 completely rescues the muscle phenotype of 25% of the mup-1 (e2347); dpy-2(e8) offspring, while dpy-11 rescues 10% of the mup-l(e2347); dpy-ll(e224) animals. dpy-10(el28) is a weaker suppressor than dpy-2 and dpy-11. dpy-5(e61), which is a stronger Dpy than these other Dpys, does not suppress the Mup phenotype. Suppression is therefore not a consequence of shortening the length of the animal.

Not all mup-1 alleles are suppressed by dpy-10. The weak alleles, mup-1 (e2430ts, e2485ts, e2436 and e2347), are suppressed but the strong alleles, mup-1 (e2488, e2486, e2487, e2438, e2434 and e2439), are not. We conclude that the suppression of the Mup phenotype does not overcome the requirement for mup-1 function since it requires a partially functional mup-1 product.

The phenotype of Mups suppressed by Dpys are shown in Fig. 5. The best suppressed Dpy-Mups are fertile ‘piggy’ Dpys with a wild-type pattern of body wall muscles and a wild-type myofilament lattice structure (Fig. 5E).-Different partially suppressed double mutants of mup-1 and dpy-2, dpy-10 or dpy-ll can be (1) Dpys with a wild-type arrangement of muscle cells but with disorganized sarcomeres (indicated in Fig. 5C), (2) Mups in which muscle cells have extensive attachments in another quadrant or only a few myofilaments extending to another quadrant (Fig. 5B–D), (3) Mups with contracting myofilaments criss-crossing through the animal and attaching across quadrants (Fig. 5A), (4) Mups with muscles detached from a quadrant as a continuous sheet (not shown).

The two mup-1 alleles of intermediate strength, e2436 and e2347, are suppressed by one copy of maternal unc-52(e444) product to give weak Dpys with a wild-type muscle pattern. mup-l(e2436) +/+ unc-52(e444) hermaphrodites segregate a higher proportion of suppressed mup-1 animals compared to mup-l(e2436)/+ hermaphrodites: 24% versus 2.2%, respectively. A similar but less pronounced interaction was observed between mup-l(e2347) and unc-52(e444). The genotype of the suppressed weak Dpys, as tested by outcrossing to N2 males, are homozygous for mup-1 (e2436 or e2347) and unc-52(+), unc-52(e444)I + or unc-52(e444).

Two copies of mutant unc-52(e444) gene in the mother does not improve the viability of these two mup-1 alleles. Unc-52(e444) is therefore a dominant and maternal but not a recessive maternal or recessive zygotic suppressor of mup-l(e2347 and e2436). unc-52 (e669am) has no effect on the Mup phenotype of mup-l(e2347 or e2436) although both unc-52 (e444 and e669am) have a very similar phenotype. The putative mup-1 null allele, e2434 is not suppressed by unc-52 (e444). This interaction therefore appears to require the presence of 50% of maternal mutant unc-52(e444) product and a partial mup-1 function.

We describe the first member of a new class of late embryonic or larval lethal mutants (Mup mutants) in C. elegans in which the muscle attachment pattern is disrupted. Mup mutants are characteristically folded-up larvae with permanent bends or kinks in the body. These are caused by the detachment of body wall muscles or by muscle processes that make morphologically wild-type attachments in inappropriate positions on the body wall, resulting in muscles that criss-cross throughout the body (E. Hedgecock, personal communication; Bogaert and Goh, unpublished data).

The forward mutation rate at which mup-1 alleles were isolated in non-complementation screens is five times higher than that for the average C. elegans gene. This indicates that the mup-1 gene may encode a large protein product or a protein that is intolerant to minor amino acid changes.

The ten mup-1 alleles form an allelic series. The four weak and intermediate alleles fit the criteria of weak hypomorphic mutations and the six strong ones are strong hypomorphic or amorphic mutations. The variable expressivity of the mup-1 phenotype (i.e. from strong Mup to wild-type) in case of e2430ts, e2485ts and e2436, suggests that the mup-1 function is dosage dependent. There may be a threshhold requirement for mup-1 function in embryonic development, above which the mup-1 embryos of the weak and intermediate alleles can develop into weak Dpys with no muscle abnormalities.

mup-1 has both a maternal and a zygotic contribution. The rescue of the Mup phenotype of the strong hypomorphic allele, mup-1 (e2347), by a zygotic wildtype product, suggests that the maternal mup-1 function is not essential, mup-1 thus appears to be genetically similar to Drosophila mutants in l(l)myospheroid (Wright, 1960; Newman and Wright, 1981) where a complete zygotic rescue is also seen despite a maternal contribution of the mys product. The removal of both the zygotic and maternal mys(+) product results in a stronger phenotype than the removal of only the zygotic product (Leptin et al. 1989) as is also the case for mup-1 (Table 1). Because all maternal mup-1 (+) contribution cannot be removed from embryos of the strong alleles, the strongest mup-1 phenotype may reflect only the loss of zygotic mup-1 activity and may not be the complete loss of function phenotype.

The mup-1 gene product interacts with or is a component of the cell surface/extracellular matrix The requirement for mup-1 function is reduced but not eliminated in the background of dpy-2, dpy-10 and dpy-11. dpy-10 have recently been shown to encode a collagen gene and dpy-2 is likely to be a collagen gene (J. Kramer, personal communication).

The interaction of mup-1 with the dpy genes is analogous to that of the C. elegans glp-1 gene with dpy genes. As is the case for mup-1, hypomorphic mutations in glp-1 but not putative null mutations are suppressed by mutations in different dpy genes, glp-1 (Yochem and Greenwald, 1989) encodes a transmembrane molecule with epidermal growth factor (EGF) repeats in the extracellular domain and has strong homology to lin-12 (Greenwald et al. 1983; Yochem et al. 1988) and Notch (Wharton et al. 1985; Kidd et al. 1986). It appears to be required as a receptor for certain cell-cell interactions in C. elegans development (Austin and Kimble, 1987; Priess et al. 1987). A general disruption of the extracellular matrix could allow a partially active glp-1 molecule to function (e.g. by facilitating the diffusion of a partially active signal molecule) but does not completely remove the requirement for glp-1 function (Maine and Kimble, 1989). The suppression of mup-1 by dpy-2, dpy-10 and dpy-11 may work through an analogous mechanism.

The interaction of mup-l(e2436 and e2347) with unc-52 (e444) but not with unc-52(e669am) may reflect a sensitivity in the stoichiometry between the mup-1 and unc-52 gene products. Although both unc-52 alleles e444 and e669am have an indistinguishable paralysed phenotype, the phenotype of mup-1 (e2436 and e2347) is only suppressed by e444 but not by e669am. This suppression is unusual because it is maternal and dominant, unc-52(e444) affects larval and adult muscle growth and attachment, and has no embryonic muscle phenotype. The suppression by unc-52(e444), like that by dpy genes, also requires a partial mup-1 function. The sequence of the unc-52 gene indicates it encodes a putative cell adhesion molecule (T. Rogalski and D. Moerman, personal communication). The presence of 50% mutant unc-52(e444) product in the mother may lead to changes in the extracellular matrix in the embryo, which allow a partially active maternal and/or zygotic mup-1 product to function during embryogenesis.

We infer from the suppression of mup-1 hypomorphic alleles by the dpy genes and unc-52(e444) that the mup-1 gene product interacts with or is a component of the cell surface/extracellular matrix.

The ectopic processes in mup-1 mutants develop early in embryogenesis when the muscle cells and the hypodermis are interacting to establish a pattern of muscle attachment. They make morphologically wildtype attachments, which anchor the muscle processes orthogonally onto the hypodermis (Fig. 6B). We think therefore that this ‘muscle attachment position’ phenotype is a distinct phenotype and not a consequence of a secondary displacement of muscle cells with abnormal myofilaments that are unattached, as has been observed in some other muscle mutants (Waterston, 1989; Bogaert and Goh, unpublished data).

The second mup-1 phenotype, the detachment of the muscle cells as single rounded up cells or in sheets suggests the role of mup-1 in adhering the muscles to the hypodermis. This phenotype is due to the detachment of muscle cells but not their failure to attach because the muscle pattern in twofold-stage embryos are wild type and a progressively higher frequency of detached muscles are observed in the later stage embryos. Second, the detached muscle cells are frequently peeled away from the hypodermis as a contractile sheet of cells, which indicates that they had previously been attached in rows (Figs2C,D, 3E,E ′, 4C,C ′).

We have two not mutually exclusive explanations for the observed incomplete penetrance of the muscle attachment position phenotype in mup-1 embryos:

The first explanation follows from the observation that incomplete penetrance of phenotypes is not uncommon in loss of function mutants in genes involved in cell guidance and migration, such as unc-5, unc-6, unc-40 (Hedgecock et al. 1990) and unc-53 (Bogaert and Goh, unpublished data). These gene products may function as parallel distributed processors in a network of molecules with partially redundant functions (e.g. cell surface/extracellular matrix) analogous to what has been suggested for the organization of the cytoskeleton in yeast (Snyder, 1989; Novick et al. 1989) and Dictyostelium (Delozanne and Spudich, 1987; Knecht and Loomis, 1987). Damage to a parallel processor would not result in a completely penetrant phenotype or complete loss of function, but in an increased ratio of errors, which is what is observed in mup-1 and in cellguidance mutants.

The second explanation follows from the observation that mup-1 function has both a maternal and zygotic contribution. The maternal mup-1 (+) product in most comma stage mup-1 embryos may be sufficient for the formation of a wild-type muscle pattern and myofilament organization (paramyosin, MH4 and MH25 staining patterns (data not shown for MH25)), but some embryos or cells may receive less maternal mup-1 product leading to ectopic processes. The comma to H- fold stage mup-1 (e2434) and mup-1 (e2430ts) embryos in which most muscles have ectopic muscle processes (Fig. 3C,C’) could be such embryos. Such variation in phenotype due to varying amount of maternally contributed product amongst siblings have been observed in Drosophila and C. elegans. The higher frequency of ectopic processes in mup-l(e2430ts) embryos derived from homozygous mup-1 (e2430ts) hermaphrodites compared to those derived from mup-l(e2434)/+ hermaphrodites also supports the case.

We have shown that a reduction in or loss of mup-1 function leads to ectopic body wall processes that make morphologically wild-type attachments in inappropriate positions on the body wall, mup-1 may be a localized adhesion molecule required in early embryogenesis for attracting the formation of muscle attachments in the four rows of body wall muscles, corresponding to the narrow bands defined by MH4 and MH25 (Francis and Waterston, unpublished data; Waterston, 1989). Insufficient mup-1 levels or the absence of mup-1 product (as discussed above) may lead to the persistence of randomly oriented myofilaments in the body wall muscles, which develop into ectopic processes.

mup-1 embryos may have enough maternal and zygotic mup-1 function for the initial muscle attachment but, in the threefold stage when stronger attachment is required, the reduced mup-1 function may be insufficient for the muscles to adhere to the hypodermis, thus leading to their detachment from the hypodermis. Requirement for mup-1 function appears to end when the muscles are firmly anchored across the hypodermis in the cuticle, mup-1 is, in this respect, different from other genes such as unc-23 (Waterston et al. 1980) and unc-52 (Mackenzie et al. Y418) in C. elegans and l(l)myospheroid in Drosophila (Wright, 1960; Newman and Wright, 1981; Leptin et al. 1989) which are required for muscle attachment later in or throughout muscle development and growth.

We have identified the first gene involved in positioning muscle attachments during embryogenesis. The molecular cloning of the mup-1 gene may lead to further understanding of how myoblasts attach to their appropriate attachment points during development to form a precise muscle pattern.

We are indebted to J. N. Thomson for the electron microscopy and to Francis Leong for expert printing of the photographs. We thank R. Waterston and H. Kagawa for antibodies. Some nematode strains used in this work were provided by A. Chisholm in LMB and the Caenorhabditis Genetics Centre, which is funded by the NIH National Center for Research Resources (NCRR). We are especially grateful to R. Barstead, E. Hedgecock, H. and R. Schnabel and R. Waterston for advice and discussion throughout this work. We also thank many other colleagues including L. Avery, D. Moerman, C. Kenyon and colleagues, J. Kramer, J. Priess, M. Stem, T. Schedl, B. Williams and colleagues in LMB for advice, comments on the manuscript, and/or sharing unpublished results. Initial work on mup-1 was carried out in the Medical Research Council, Laboratory of Molecular Biology, Cambridge for which T.B. was supported by a Research Fellowship from Peterhouse College, Cambridge. We thank L. Lim and S. Brenner for their generous support.