Characterization of α1(IV) Collagen Mutations in Caenorhabditis elegans and the Effects of α1 and α2(IV) Mutations on Type IV Collagen Distribution

C. elegans strains were maintained and handled as described (Brenner, 1974). The C. elegans Bristol strain N2 has been designated as wild type. Animals homozygous for the emb-9 alleles g23, g34, b117, b189, and hc70 (Miwa et al., 1980; Wood et al., 1980; Cassada et al., 1981) were maintained at 15°C. The emb-9 alleles st540, st545 (Williams and Waterston, 1994), cg45, cg46, cg56, and cg57 (see below) were maintained as heterozygotes with qC1, a balancer chromosome that prevents recombination in the region of emb-9 (Edgley et al., 1995). The following let-2 alleles were used: g25, g30 (Miwa et al., 1980; Cassada et al., 1981), b246 (Wood et al., 1980), mn109, mn129, and mn139 (Meneely and Herman, 1979).

The emb-9(cg56) allele was generated using the strain unc-32(e189)/qC1; cgEx1[emb-9(+)rol-6(su1006)]. The unc-32 marker is tightly linked to emb-9 and used to identify chromosome III. cgEx1 is an extrachromosomal array carrying the wild-type emb-9 gene and rol-6(su1006), which confers a right roller phenotype (Mello et al., 1991; Kramer and Johnson, 1993). The strain was mutagenized with ethyl methanesulfonate (EMS) and grown at 15°C. F1 Rol non-Unc animals were isolated onto separate plates and shifted to 25°C. Plates were screened for the presence of Rol Unc and the absence of non-Rol Unc F2 progeny. One allele, cg56, was isolated and backcrossed twice using the following scheme: The unc-32 emb-9(cg56)/qC1; cgEx1 animals were crossed with N2 males, and cross progeny males were individually crossed into unc-32/qC1 hermaphrodites. Offspring from this cross that produced approximately one-quarter lethal and one-quarter qC1 progeny, indicating they were unc-32 emb-9(cg56)/ qC1, were selected and maintained.

Putative null mutations were generated using the strain unc-36(e251) emb-9(g23); qDp3. qDp3 is a free duplication that carries unc-36(+) and emb-9(+). The strain does not survive at 25°C because of the semidominance of emb-9(g23). The strain was mutagenized with EMS and grown at 15°C for two generations. The F2s were shifted to 25°C and animals that survived were isolated onto separate plates. Three alleles, cg45, cg46, and cg57, were isolated and backcrossed four times as follows: unc-36 emb9(cg45, cg46, or cg57); qDp3 animals were mated with N2 males, and the resulting males were crossed individually into unc36/qC1 hermaphrodites. F1 cross progeny that produced approximately one-quarter lethal and one-quarter qC1 progeny, indicating that they were unc-36 emb-9/qC1, were selected.

emb-9 mutants were scored for viability at 15, 20, and 25°C. Animals heterozygous for emb-9 were generated by crossing the mutant strains with wild-type males. The heterozygotes were placed individually onto separate temperature-equilibrated plates. Eggs laid within the next 6–7 h (25°C), 8–10 h (20°C), or 14–16 h (15°C) were counted, and the parent was removed. 1 d later, plates were examined to determine the numbers of eggs that had hatched, or that failed to hatch (Emb). Plates were scored again after 3–4 d (25°C), 5–6 d (20°C), or 7–8 d (15°C) to determine the number of animals that had developed to adults (Ad). Animals that were severely abnormal and sterile as adults were noted (Ste). Larvae not accounted for as adults were considered to have arrested during larval development (Acc). The number of Acc animals could be reliably inferred since the N2, unc-32, and unc-36 controls gave very low percentages of Acc larvae at all temperatures. The total Emb and Acc for N2, unc-32, and unc-36 were from 0–3%. These background values have been subtracted from the results obtained from the relevant emb-9 alleles linked to unc-32 or unc-36. At least 200 embryos were scored for each strain at each temperature.

st545 and st540 were previously identified as alleles of emb-9 by testing for complementation with the emb-9(hc70) allele (Williams and Waterston, 1994). We retested for complementation using the putative null allele emb-9(g23cg45). unc-36 emb-9(g23cg45)/qC1 hermaphrodites were crossed with N2 males, and single cross progeny males were mated with st545/qC1 and st540/qC1. The production of one-quarter embryonic lethal progeny from these crosses indicated failure of st545 or st540 to complement emb9(g23cg45). The presence of viable cross progeny males indicated that the mating was successful. The absence of qC1 homozygous cross progeny indicated that the male was of the genotype unc-36 emb-9(g23cg45)/+ +.

Genomic DNAs were prepared from emb-9 mutant strains and the wildtype N2 strain as previously described (Kramer et al., 1988). g23, g34, hc70, b117, and b189 carrying strains were grown as homozygotes at 15°C, and st540, st545, and cg56 were maintained as heterozygotes balanced across from qC1. The cg45, cg46, and cg57 alleles were maintained as unc36 emb-9; qDp3 strains.

Nucleotide alterations were identified using either the chemical mismatch method or by directly sequencing the entire emb-9 gene. PCR fragments (1–2 kb) of wild-type and emb-9 mutant DNAs were prepared by incubating genomic DNA (500–700 ng) with 1× PCR buffer (GIBCO BRL, Gaithersburg, MD), 2 mM MgCl₂ , 2.5 ng/ml BSA, 200 μM dNTPs, and 2.5 U Taq polymerase (GIBCO BRL) in a total volume of 100 μl. PCR samples were incubated at 94°C for 1 min, 47–63°C (depending on T_m of primers) for 1 min, and 72°C for 1 min. This temperature cycle was repeated 30 times in an automated temperature cycler. PCR fragments were generated so as to overlap one another by 200–300 bp. Six 1–2-kb fragments were used to span the emb-9 gene, including 484 bp 5′ of the start codon and 111 bp 3′ of the stop codon. PCR fragments were purified using Geneclean (Bio 101 Inc., La Jolla, CA). Nucleotide alterations in these fragments were identified using the chemical mismatch detection procedure (Montandon et al., 1989). Point mutations were identified by direct sequencing of PCR fragments in the areas of mismatch, using Sequenase (United States Biochemical Corp., Cleveland, OH) and a modified annealing procedure (Kusukawa et al., 1990).

Several mutations were identified by directly sequencing overlapping PCR fragments that span the entire emb-9 gene. PCR fragments were cleaned using Centricon 100 columns (Amicon, Beverly, MA). 200–300 ng of template DNA was used for Taq DyeDeoxy terminator thermal cycle sequencing (Appiled Biosystems, Inc., Foster City, CA). Samples were spun through spin columns (model Centri-Sep; Princeton Separations Inc., Adelphia, NJ) and dried in a vacuum centrifuge. Each sample was resuspended in 4 μl of formamide-EDTA solution (83% deionized formamide, 8 mM EDTA, pH 8.0). Samples were denatured at 90°C for 2 min, transferred immediately onto ice, and loaded onto a 6% polyacrylamide gel on an automated sequencer (model 373A; Applied Biosystems, Inc.).

Animals carrying temperature-sensitive alleles, emb-9(g23, g34, hc70, b117, b189) and let-2(b246, g25, g30, mn109, mn129, mn139), were maintained as homozygotes at 15°C. To collect embryos raised at higher temperatures, plates containing a mixed population were shifted from 15 to 20 or 25°C and allowed to lay eggs overnight.

Embryos were isolated from animals by alkaline hypochlorite (5:3:1 water/4M NaOH/bleach) treatment for 7–10 min. The eggs were washed several times with PBS, fixed with 3% paraformaldehyde in PBS for 15 min at room temperature, rinsed with distilled water (dH₂O) and then stored at −80°C in 100% methanol for 30 min to 7 d.

Fixed embryos were rehydrated, treated with 6 M urea, 0.1 M glycine, pH 3.5, for 15 min at room temperature, rinsed with dH₂O, and blocked for 1 h in PBS containing 0.5% Tween-20 (PBS-T) and 10% normal donkey serum (NDS)¹. Primary antibodies against C. elegans type IV collagen and myosin heavy chain B were added in PBS-NDS for 1 h at 37°C. The embryos were then washed three times for 15 min each with PBS-T and incubated for 1 h at 37°C with secondary antibodies (Jackson ImmunoResearch, West Grove, PA) and 0.1 mg/ml diamidinophenolindole in PBST–NDS, washed 3× for 15–30 min with PBS-T, and then mounted in DABCO medium (20% glycerol, 10% polyvinyl alcohol, 2.5% DABCO in 40 mM Tris-HCL, pH 8.3) with phenylenediamine freshly added to 1 mg/ml.

The chain-specific anti–type IV collagen antisera used in this study have been described (Graham et al., 1997). Affinity-purified antiserum NW1910 was used to identify the α1(IV) chain EMB-9, and affinity-purified NW68 antiserum was used to identify the α2(IV) chain LET-2. The 5-8 mouse monoclonal antibody against C. elegans myosin heavy chain UNC-54 (Miller et al., 1983) was used to stain body wall muscle cells. The primary antibodies were detected using fluorescein-conjugated donkey anti–rabbit and/or rhodamine-conjugated donkey anti–mouse IgG secondary antibodies.

Eight alleles of emb-9 were characterized for their effects at 15, 20, and 25°C. Progeny from heterozygous, emb-9/+, mothers were scored for those that arrested during embryogenesis (Emb) or larval development (Acc), or became sick, sterile adults (Ste) (Fig. 1). For recessive alleles, 25% or fewer animals should show abnormal phenotypes, while semidominant alleles could result in up to 75% abnormal animals. These emb-9 alleles were grouped into three classes based on their general phenotypic characteristics.

The class I alleles, b117, g23, hc70, b189, and g34, are all strongly temperature sensitive. At 15°C, there is a very low level of lethality or sterility. At 20°C, there is a large increase in larval lethality in all except the b117 allele. The vast majority of these larvae arrest at the L1 or L2 stages. At 25°C, there is approximately 25% embryonic lethality for each of the alleles, indicating that essentially all homozygous mutant animals arrest as embryos. This result is consistent with the 100% embryonic lethality seen when homozygous embryos are shifted to 25°C (Cassada et al., 1981; Kramer, J., unpublished results). Larval lethality and sick, sterile adults increase the total percentages of abnormal animals to 37 to 48%, indicating that each of these alleles is semidominant.

The class II alleles, st545 and st540, show high levels of embryonic lethality and mild temperature sensitivity. st545 is essentially a recessive, embryonic lethal. It shows ∼25% embryonic arrest at all temperatures and a very low level of larval arrest at 20 and 25°C. st540 appears to be mildly cold sensitive, displaying decreased embryonic lethality at 25 versus 15°C and some semidominant larval arrest at the lower temperature.

The class III allele, cg56, is a recessive, non–temperature-sensitive larval lethal. All cg56 homozygotes arrest during the L1–L2 larval stages at all temperatures. There is no apparent semidominance of this allele.

Most emb-9 alleles show semidominant effects, which are likely to result because mutant protein interferes with the functions of other molecules. Complete absence of EMB-9 would eliminate any dominant interfering effects and allow determination of the phenotype that results when the α1(IV) collagen chain is missing. For this purpose, a genetic screen was designed to isolate putative null mutations of the emb-9 gene.

Three putative emb-9 null alleles were generated by screening for mutations that eliminate the temperaturesensitive dominant lethality of emb-9(g23). A strain carrying two copies of g23 and a wild-type copy of emb-9 on a duplication, unc-36 emb-9(g23); qDp3[emb-9(+)], is viable and fertile at 15°C but dies at 25°C. This strain was mutagenized and screened for F2 animals that could survive and reproduce at 25°C, and three putative null alleles, cg45, cg46, and cg57, were identified. All three of these alleles cause less severe phenotypes than g23. g23 homozygotes grown at 25°C arrest at approximately the twofold length of embryogenesis, when the head and tail of the elongating embryo have just reached each other (Fig. 2,A). Arrested embryos show extensive herniations of the body wall. In contrast, cg45, cg46, and cg57 embryos elongate further, to at least the threefold stage of embryogenesis (Fig. 2, B and C). The putative null mutant animals also display less severe herniations of the body wall.

Phenotypic characterization of the putative null mutations indicates that they are recessive, embryonic lethals and belong to a new phenotypic class, class IV (Fig. 1). Approximately 25% of the progeny from cg45, cg46, or cg57 heterozygous mothers arrest as embryos, at the three- to fourfold stage. A small number of embryos hatch, but they are small, herniated, and paralyzed (Fig. 1, abn. L1). These larvae do not develop further and appear to be homozygous mutant embryos whose egg shells have ruptured. There is no significant effect of temperature on the phenotype of these mutants, as would be expected for null mutants.

Nine mutations in the emb-9 gene have been identified, three previously (Guo et al., 1991) and six here, by comparison with the wild-type emb-9 gene sequence. The precise alterations in the mutants are given in Table I and are shown graphically in Fig. 3. All nine of the mutations are in the triple-helical (Gly-X-Y) repeat domain of the EMB-9 chain. Six of the mutations result in replacement of glycine with aspartate, glutamate, or arginine. Two of the putative null mutations, cg45 and cg57, result in nonsense codons, and one, cg46, is a deletion of 497 bp. The cg46 deletion removes the last 22 nucleotides of intron 9 and 475 nucleotides of exon 10. Failure to splice this intron would result in translation termination after incorporation of three amino acids from intron 9.

Nucleotide alterations were not identified in st545 or st540 by either chemical mismatch detection or direct sequencing of mutant genomic DNAs. These mutations were previously assigned to emb-9 based on failure to complement the temperature-sensitive allele emb-9(hc70) (Williams and Waterston, 1994). We retested complementation of these alleles using the putative null allele cg45 and found that they do fail to complement (see Materials and Methods), indicating that they are indeed emb-9 alleles. Since these alleles are maintained as heterozygotes across from the qC1 balancer chromosome, the wild-type sequence in the background may have interfered with the detection of these mutant alterations.

We have determined the effects of emb-9 and let-2 mutations on type IV collagen distribution by staining mutant embryos with chain-specific antisera. The distributions of EMB-9 and LET-2 in wild-type embryos have been described (Graham et al., 1997) and are outlined here for comparison to the patterns in mutants. Antisera against EMB-9 and LET-2 show identical staining patterns, suggesting that the two chains assemble into a single heterotrimeric collagen molecule. In lima to 1 1/2-fold embryos, strong type IV collagen stain is seen within body wall muscle cells (Fig. 4, A and B). Beginning in the comma stage, secreted type IV collagen accumulates beneath the body wall muscles and around the forming pharynx and intestine. From the twofold stage on, most type IV collagen stain is extracellular. Both EMB-9 and LET-2 are detected in the region of the hypodermal basement membrane underlying the body wall muscle quadrants and in the basement membranes surrounding the pharynx, intestine, and gonad (Fig. 4, C–F). No type IV collagen is detected in the hypodermal basement membrane between the body wall muscle quadrants or on the pseudocoelomic face of the muscles. This pattern is altered by mutations in emb-9 or let-2.

All three putative emb-9 null mutations have the same effect on type IV collagen localization. Null mutant embryos do not show any detectable staining with EMB-9– specific antisera (Fig. 5, A and B). They do stain with LET-2– specific antisera (Fig. 5, C and D); however, all of the stain is intracellular, and no LET-2 is detectable in basement membranes. The intracellular staining pattern is punctate and frequently appears as two strong spots located on opposite sides of each of the body wall muscle cell nuclei. No LET-2 is detectable within hypodermal, intestinal, or pharyngeal cells. This is consistent with previous evidence that type IV collagen gene expression occurs in body wall muscle cells but is undetectable in hypodermal, intestinal, and pharyngeal cells (Graham et al., 1997). The intracellular accumulation of LET-2 in putative emb-9 null mutants indicates that LET-2, the α2(IV) chain, is not secreted in the absence of full-length EMB-9, the α1(IV) chain.

Most emb-9 (6 of 11) and let-2 (15 of 17) mutations are substitutions for glycine residues in the Gly-X-Y repeat domain of the collagen. We stained the five emb-9 and twelve let-2 temperature-sensitive glycine substitution mutants with anti–EMB-9 and anti–LET-2 antisera. In all cases, both antisera gave very similar patterns, indicating that EMB-9 and LET-2 chains are colocalized in these emb-9 and let-2 mutants. The primary effect seen was loss of type IV stain from basement membranes and the appearance of strong intracellular stain in body wall muscle cells. The degree of this effect varied depending on the temperature and allele and correlated with the severity of the phenotype. In the most severe allele/temperature combinations, all type IV collagen was intracellular, while milder combinations showed progressively less intracellular stain and more basement membrane stain.

The emb-9(g23) mutant is shown as an example of the pattern seen in the five emb-9 mutants (Fig. 6). At 25°C, essentially all type IV collagen stain is found as bright spots within body wall muscle cells (Fig. 6, A and B). There is little or no detectable extracellular staining in basement membranes. At 20°C, strong intracellular stain is still seen, but weak extracellular stain in basement membranes can be seen as well (Fig. 6, C and D). At 15°C, the intensity of the intracellular stain decreases markedly, and strong extracellular stain of basement membranes is seen (Fig. 6, E and F). Note that at 15°C, g23 mutants develop normally, at 20°C most hatch but arrest as early larvae, while at 25°C they arrest at the twofold stage of embryogenesis (Fig. 1).

The let-2(b246) mutation causes a milder temperaturesensitive phenotype than emb-9(g23). At 15°C, b246 homozygotes develop normally, at 20°C 78% complete development to adult, while at 25°C they all arrest at the twofold stage of embryogenesis (Sibley et al., 1994). The milder phenotype of b246 is reflected in the distribution of type IV collagen at different temperatures. At 25°C, type IV collagen accumulates intracellularly, and little or no basement membrane stain is detectable (Fig. 7,A). At 20°C, type IV collagen distribution is nearly normal with only a small amount of intracellular accumulation (Fig. 7,B). At 15°C, the type IV collagen distribution is wild-type in appearance (Fig. 7 C).

Six temperature-sensitive glycine substitution mutants of let-2 were analyzed by staining embryos raised at 15, 20, and 25°C with LET-2 antiserum. The percentages of embryos that showed strong intracellular stain in body wall muscle cells and that showed strong pharyngeal basement membrane stain were determined (Table II). In general, intracellular body wall muscle stain increased and pharyngeal basement membrane stain decreased with increasing temperature. Thus, as the temperature increases, secretion of mutant type IV collagen is more strongly inhibited. These results indicate that temperature-sensitive glycine substitution mutations in either EMB-9 or LET-2 cause both chains to be retained intracellularly. The degree of intracellular retention and the severity of the phenotype increase with increasing temperature.

emb-9(cg56) is a glycine substitution mutation that results in a non–temperature-sensitive, recessive larval lethal phenotype (Fig. 1). Staining of this mutant with either EMB-9 or LET-2 antisera shows the same pattern at all temperatures (Fig. 8). Both EMB-9 and LET-2 show strong accumulation in body wall muscle cells, but there is also very weak staining of the basement membranes underlying the body wall muscles. The small amount of extracellular staining in cg56 mutants is similar to what is seen in emb-9(g23) mutants grown at 20°C and is consistent with the fact that cg56 animals survive to the L1 stage, as do g23 animals at 20°C. The small amount of type IV collagen present in the basement membranes of cg56 mutant animals is sufficient to allow completion of embryogenesis.

In both of the class II mutants, st545 and st540, EMB-9 and LET-2 chains accumulate intracellularly in body wall muscle cells, and there is no detectable extracellular basement membrane stain. However, the pattern of intracellular accumulation in st545 and st540 is different from that in the other emb-9 and let-2 mutants. The stain in st545 and st540 mutants generally appears as a ring surrounding the nucleus (Fig. 9,A), while in all other emb-9 and let-2 mutants, the intracellular stain generally appears as two spots on opposite sides of the nucleus (Fig. 9,B and Figs. 5, 6, and 7). Apparently, the st545 and st540 mutations cause accumulation of type IV collagen in a different intracellular compartment within muscle cells than do the other mutations.

emb-9 and let-2 mutant embryos were costained with an antibody against myosin heavy chain B, which is expressed in body wall muscle cells. The myosin staining served as a control for permeability of embryos and permitted analysis of the organization of body wall muscles. We had noted previously that the body wall muscle quadrants were disrupted in emb-9 and let-2 missense mutants. However, in these mutants it is not clear if disruption is due to the absence of type IV collagen or the dominant interfering effects of abnormal type IV collagen. Analyses of the three putative emb-9 null alleles indicate that they do cause disruption of the body wall muscle quadrants. In wild-type animals, the four body wall muscle quadrants appear as four stripes extending from head to tail of the animal (Fig. 10,A). In temperature-sensitive emb-9 or let-2 mutants at high temperature, gaps in the body wall muscle quadrants can first be seen at the 1 1/2-fold stage and become extensive by the twofold stage (Fig. 10,B). In emb-9 null mutant embryos, gaps also begin to appear at the 1 1/2-fold stage, although they are generally less numerous and less severe than in missense mutants. In two- to threefold emb-9 null mutant embryos, numerous gaps in the body wall muscle quadrants are apparent (Fig. 10 C). Since gaps in the body wall muscle quadrants are seen in the emb-9 null mutants, disruption of the body wall muscle quadrants can result from the absence of type IV collagen in the basement membrane. The first appearance of gaps in the body wall muscle quadrants corresponds to the beginning of muscle contractile activity, suggesting that the forces of contraction cause muscle cells to detach from the body wall.

In wild-type and emb-9 or let-2 missense mutants, the EMB-9 and LET-2 chains always colocalize. In emb-9 null mutants, the LET-2 chain is retained intracellularly, indicating that it requires EMB-9 to be assembled and/or secreted. All characterized type IV collagen molecules are heterotrimers containing two α1-like and one α2-like chains (Hudson et al., 1993). EMB-9 and LET-2 are α1 and α2(IV)-like chains and are likely to be the only type IV collagen chains in C. elegans (Guo and Kramer, 1989). If EMB-9 and LET-2 assemble into a single collagen molecule, then mutations in either would interfere with the folding of the triple-stranded collagen molecule, resulting in interference with the assembly and secretion of both chains. All of these considerations suggest that they assemble into a single type IV collagen molecule containing two EMB-9 and one LET-2 chains.

The mutations of the α1(IV) collagen gene emb-9 show a range of phenotypic severities, similar to that found for the α2(IV) gene let-2 (Sibley et al., 1994). Most mutations are substitutions for glycine residues in the Gly-X-Y repeat domain of the molecule. It was noted that the phenotypic severities of glycine substitution mutations in let-2 are inversely correlated with the calculated helical stability of type IV collagen at the position of the mutation (Sibley et al., 1994). The most severe mutations are in regions of low stability, and the mildest mutations are in regions of high stability. This correlation is also apparent for the emb-9 glycine substitution mutations. The mildest allele (b117) is in a region of above average stability, while the more severe alleles (g23, hc70, b189, cg56) are in regions of below average stability. For both emb-9 and let-2, more severe alleles are located in lower stability regions, and milder alleles are in higher stability regions.

The phenotypes of most emb-9 and let-2 mutations are temperature-sensitive. As the temperature is increased, the percentage of arrested animals increases and arrest occurs earlier in development. Staining of mutant animals with type IV collagen antisera showed that as the temperature was increased, more of the mutant collagen was retained within cells, and less was detected in basement membranes. Thus, the more severe phenotype at higher temperature likely results because less type IV collagen is present in basement membranes. Mutant type IV collagen that is able to assemble into the basement membrane provides function and results in a less severe phenotype. We found no evidence for mutant type IV collagen that was able to assemble into basement membranes but did not provide function.

Assembly of collagen molecules requires registration of three chains at or near the carboxyl terminus, nucleation of the triple-helix, and propagation of the triple-helical structure towards the amino end of the molecule. The collagen triple helix is stabilized by hydrogen bonding and is an inherently temperature-sensitive structure (Engel and Prockop, 1991; Bachinger et al., 1993). Glycine residues are required at every third position because they are the only residue that can be accommodated in the center of the triple helix without distortion of the structure and loss of stabilizing bonds. Glycine substitution mutations in the Gly-X-Y repeat domain of collagen interrupt the folding of the triple helix, requiring renucleation on the amino side of the substitution for triple helix assembly to progress.

C. elegans type IV collagen glycine substitution mutations can be viewed in terms of their effects on triple helix stability. Substitutions in regions of low helical stability cause more severe effects than those in high stability regions. Introduction of a glycine substitution in an already low stability region would produce such an unstable area that renucleation of triple helix assembly might be very inefficient or impossible. Milder alleles, located in more stable regions, would allow renucleation to occur more readily. The temperature sensitivity of most glycine substitutions could result because increased temperature would reduce the stability of the entire triple-helical domain. The mutations causing the most severe phenotypes are located in regions of low triple-helical stability, and a small increase in temperature would be sufficient to prevent assembly. The milder mutations are located in regions with higher stability, and larger increase in temperature would be required to completely block assembly.

The three putative null mutations generated in this study all have phenotypic characteristics distinct from other emb-9 mutations. While all other mutations in emb-9 and let-2, except cg56, cause arrest at approximately the twofold stage of embryogenesis, the null mutants arrest later, at the three- to fourfold stage. Since null mutant animals are capable of developing further than the missense mutants, the missense protein must interfere with the function of some other molecule(s). The missense chains act in a dominant negative fashion, interfering not only with the wild-type type IV collagen, but also with some other molecule(s). This interference could occur intracellularly, in the RER or Golgi network, and/or extracellularly, in basement membranes.

As shown here for emb-9 and previously for let-2 (Sibley et al., 1994), the dominant lethality of temperature-sensitive glycine substitution mutations increases with increasing temperature. Antibody staining shows that as the temperature increases, less type IV collagen is secreted. At the highest temperature tested, no type IV is detected in basement membranes. Since the dominance of these mutations becomes more severe as more type IV collagen is retained intracellularly and less is secreted, it is likely that dominant interference is occurring intracellularly, not in the basement membrane. Accumulation of large amounts of the abnormal collagen within cells could interfere with the synthesis, processing, and/or secretion of other proteins, thereby resulting in a more severe phenotype than results from simple loss of type IV collagen.

In emb-9 null mutants, the LET-2 chain accumulates intracellularly but does not cause detectable dominant effects. Since LET-2 is the α2-like chain, it is not likely to undergo any trimeric assembly in the absence of EMB-9. In contrast, when there is a glycine substitution in either EMB-9 or LET-2, trimeric assembly is likely to initiate normally but be blocked when it reaches the site of the substitution. Intracellular accumulation of these partially assembled trimeric molecules appears to be more detrimental than accumulation of individual, unassembled LET-2 chains.

The dominant interference caused by mutations in type IV collagen and fibrillar collagens may result from different mechanisms. Glycine substitutions in mammalian fibrillar type I collagen inhibit collagen triple helix assembly and result in intracellular degradation of much of the mutant and normal collagen chains (Prockop et al., 1989; Engel and Prockop, 1991). Secretion of small amounts of glycine substitution containing type I collagen into the matrix can dominantly disrupt collagen fiber formation. Glycine substitutions can cause bulges or kinks in the collagen molecule (Prockop et al., 1989; Engel and Prockop, 1991). Since fibrillar collagens assemble into highly ordered, closely packed fibers, the presence of an abnormal collagen molecule distorts the packing of other molecules in the fiber. In contrast, type IV collagen normally contains 20 or more interruptions of its Gly-X-Y domain and forms a looser, less ordered network (Yurchenco and Ruben, 1987). Incorporation of a type IV collagen molecule containing a glycine substitution is not likely to greatly disrupt the type IV network. Collagen molecules containing glycine substitutions are likely to be much less disruptive to the type IV network than to the highly ordered fibers formed by fibrillar collagens. This conclusion is consistent with our evidence that the dominance of the C. elegans type IV collagen mutations is not an extracellular effect, but occurs intracellularly.

The effects on type IV collagen distribution in animals with mutations in type IV collagen genes appear somewhat different in mammals and C. elegans. In the kidneys of humans, mice, or dogs with mutations in the α3–α5(IV) collagen genes, type IV collagen is not detected in basement membranes or within cells (Gubler et al., 1995; Peissel et al., 1995; Cosgrove et al., 1996; Miner and Sanes, 1996; Thorner et al., 1996). In C. elegans mutants, type IV collagen can be reduced or absent from basement membranes but is correspondingly seen to accumulate within the cells that synthesize it. This difference could result from a higher synthesis rate and/or a lower degradation rate for type IV collagen in C. elegans. The C. elegans type IV collagen mutants arrest during embryogenesis, when there is high transcriptional activity of the type IV collagen genes (Graham et al., 1997), and the mutant animals are fixed and stained within a few hours after the time of their arrest. There may not be enough intracellular degradative activity in C. elegans to remove the accumulated type IV collagen during this short period.

In emb-9 null mutants, we detected disruption of the body wall muscles beginning at the 1 1/2-fold stage, when contractions first begin. The disruption worsens as muscle contractions become more vigorous, and muscle cells appear to pull away from the hypodermis. It has been shown that body wall muscle function is required for elongation of the C. elegans embryo beyond the two-fold stage (Waterston, 1989; Barstead and Waterston, 1991; Williams and Waterston, 1994). The disruption of the body wall muscles that we see is sufficient to explain why the embryos arrest. At the time of arrest, the organization of other tissues, such as the pharynx and intestine, appears relatively normal. The type IV collagen that assembles into the basement membrane that separates body wall muscles from the hypodermis is critical for maintaining attachment of the body wall muscles during the force of contraction.

We wish to thank Dr. S. Wolinsky and members of his laboratory for help with automated sequencing, Dr. D. Miller for providing 5-8 monoclonal antibody, and Drs. B. Williams and R. Waterston for providing C. elegans strains. We also thank Drs. C. Witkowski, E. Goodwin, and J. Crew for helpful suggestions on the manuscript.

This work was supported by National Institutes of Health grant HD27211 (J.M. Kramer).