Type IV Collagen Is Detectable in Most, but Not All, Basement Membranes of Caenorhabditis elegans and Assembles on Tissues That Do Not Express It

Animals were maintained as described by Brenner (6). The C. elegans variety Bristol isolate N2 is designated as wild-type. The mutant strain used in this study, unc-36(e251) emb-9(g23cg45)/qC1[unc-36(+)emb-9(+)], was generated by reverting the temperature-sensitive dominant lethality of emb-9(g23) (20a). The cg45 mutation converts Gln1388 of EMB-9 to a termination codon.

cDNA fragments encoding the entire NC1 domains of EMB-9 and LET-2 were generated by reverse transcriptase–PCR amplification and cloned into pGEX fusion protein vectors (Pharmacia LKB Biotechnology, Inc., Piscataway, NJ). Fusion proteins were isolated in inclusion bodies after overexpression in Escherichia coli. The inclusion bodies were solubilized in SDS/β-mercaptoethanol sample buffer (30) and electrophoresed on SDS–polyacrylamide gels, and the fusion protein bands were excised from the gels and used to immunize rabbits. Synthetic peptides derived from the NC1 domain of EMB-9 (CVDQDKQFRKPMSQ) or from exon 9 of LET-2 (CREFTGSGSIVGPR) were coupled to BSA using m-maleinidobenzoyl-N-hydroxysuccinimide ester (21).

Rabbit antisera against fusion protein and peptide antigens were generated commercially (Hazelton Research Products, Inc., Denver, PA). Sera against synthetic peptides were affinity purified using Actigel (Serogene Bioseparations, Arcadia, CA) columns coupled with the same peptide. Fusion protein antisera were first affinity purified on columns coupled with the fusion protein to which they were generated. Cross-reacting antibodies were then removed by passing the affinity-purified anti–EMB-9 sera over a column containing the LET-2 fusion protein, and vice versa. Purified antisera were stored at −20°C in PBS with 50% glycerol, 1% BSA.

C. elegans extracts were prepared by washing mixed-stage populations extensively with large volumes of distilled water (dH₂O).¹ After pelleting, they were mixed immediately in an equal volume of 2× Laemmli sample buffer (30), containing 5% freshly added β-mercaptoethanol, and boiled for 10 min. Extracts were then incubated for 12–24 h at room temperature, boiled 10 min, and stored at −20°C.

Western blots were performed by standard procedures (21). C. elegans protein extracts were electrophoresed on either 7.5 or 4% polyacrylamide gels and transferred to nitrocellulose using 25 mM Tris, 192 mM glycine, 0.1% SDS, 10% methanol. Nitrocellulose filters were stained with Ponceau S to determine the position of molecular mass markers and then blocked for at least 1 h in BB (BB: 10% dry milk, 0.3% Tween 20 in PBS). Filters were incubated with anti–type IV collagen antibodies diluted 1:500–1:1,000 in BB for 1 h at room temperature and washed with 0.3% Tween 20 in PBS. They were then incubated with alkaline phosphatase– conjugated goat anti–rabbit antibodies (Sigma Chemical Co., St. Louis, MO) diluted 1:1,000 in BB, washed, and developed with BCIP/NPT (21).

For competition experiments, anti–type IV collagen antibodies were preincubated for 1 h at room temperature with 200–400 μg/ml of the fusion protein or 100 μg/ml of the peptide used as the original antigen. The antibody/antigen mixture was then used to stain blots as described above, or to stain animals as described below.

Animals were prepared for immunofluorescence analysis by one of three methods. In the first, eggs were collected from alkaline hypochlorite– treated animals (55), rinsed twice with dH₂O, fixed with 3% paraformaldehyde in PBS for 15 min at room temperature, rinsed twice with dH₂O, and stored at −80°C in 100% methanol for 30 min to 7 d. Alternatively, animals in dH₂O were pipetted onto slides subbed with 0.2% gelatin, 0.1% polylysine, 0.01% chrome alum, covered with a coverslip, and frozen on dry ice. The coverslip was cracked off with a razor blade, and the slide was immersed in −20°C methanol for 5 min, followed by −20°C acetone for 3 min, and air dried. Lastly, larvae and adults were prepared as described by Finney and Ruvkun (11).

After fixation, animals were rehydrated and either blocked immediately for 1 h in PBS containing 0.1% Triton X-100 (PBS-T) and 10% normal donkey serum (NDS), or treated with 6 M urea, 0.1 M glycine, pH 3.5, for 15 min (69) and rinsed 3× with PBS before blocking. Samples were incubated with antibodies against type IV collagen (EMB-9 or LET-2), myosin heavy chain B (UNC-54), or mouse anti-hemagglutinin (HA) monoclonal antibody (Boehringer-Mannheim Corp., Indianapolis, IN) in PBS-T–NDS overnight at 4°C for 1 h at 37°C, washed with PBS-T, incubated 1 h at 37°C with FITC donkey anti–rabbit and LRSC donkey anti– mouse secondary antibodies (Jackson ImmunoResearch, West Grove, PA) in PBS-T–NDS, washed with PBS-T containing 1 μg/ml diamidinophenolindole, followed by PBS alone.

To prepare frozen sections, animals were fixed with 3% paraformaldehyde in 0.1 M sodium phosphate buffer, pH 7.0, for 20 min at room temperature. After washing in water, they were resuspended in 30% sucrose in PBS at 4°C. After the animals sank to the bottom of the tube, the overlying sucrose solution was removed and an equal volume of OCT compound (Miles, Inc., Elkhart, IN) was added. Frozen sections were collected on polylysine-coated slides and dried for 30 min before being blocked and stained as described above.

Reporter constructs for emb-9 were generated from the plasmid pJJ359, which contains the complete emb-9 gene flanked by 4.3 kb of 5′ upstream and 0.5 kb of 3′ sequence (Fig. 1 A). pJJ359 is capable of transgenic rescue of emb-9 mutant animals. To generate the emb-9:lacZ reporter construct pJJ444, the 10-kb XmaI-SaII fragment from pJJ359 was inserted into XmaI-XhoI cut pPD34.110. PPD34.110 contains the E. coli lacZ gene with a membrane stop transfer sequence and 3′ untranslated sequences from unc-54 (13.). An emb-9:green fluorescent protein (GFP) reporter, pJJ446, was made by removing the lacZ region of pJJ444 by KpnI-SpeI digestion and inserting the KpnI-SpeI GFP-containing fragment from pPD95.75 (provided by A. Fire, S. Xu, J. Ahnn, and G. Seydoux).

Another emb-9:lacZ reporter construct, pJJ318, was made by inserting a HindIII-PvuI (−1672 to −18 relative to emb-9 start codon) fragment from the lambda phage clone CH#2 into HindIII-SmaI–digested pPD21.28. CH#2 contains the complete emb-9 gene and is capable of transgenic rescue of emb-9 mutant animals (20). pPD21.28 contains the E. coli lacZ gene with a nuclear localization signal and 3′ untranslated sequences from unc-54 (13).

Two let-2 reporter constructs were generated from the plasmid pJJ353, which contains the complete let-2 gene with 2.2 kb of 5′ upstream and 430 bp of 3′ untranslated sequence (Fig. 1 B). pJJ353 is capable of transgenic rescue of let-2 mutant animals (50). pJJ360 was made by ligating the 6-kb SphIEcoRV fragment from pJJ353 into SphI-SmaI–digested pPD34.110. pJJ361 was made by ligating the 4-kb SmaI-ApaI fragment from pPD34.110 with the 11-kb EcoRV-ApaI fragment from pJJ353. pPD34.110 contains the E. coli lacZ gene with a membrane stop transfer sequence and 3′ untranslated sequences from unc-54 (13). An additional let-2 reporter, pJJ325, was generated by cloning the 3.2-kb AvaI fragment from cosmid C12F7 (−3.2 to +20 kb relative to let-2 initiation codon) into SmaI-digested pPD22.04 (13).

DNAs were injected into hermaphrodite gonads as described (36) using pRF4[rol-6(su1006)] (28) as a transformation marker. The emb-9 reporter construct pJJ318 was integrated into the genome by γ-irradiation and selection of a line that segregates 100% roller progeny. β-galactosidase activity was visualized by staining as described (12), after fixation for 15 min with 3% paraformaldehyde in PBS and washing with PBS.

The pJJ400 construct was made by cloning the 3.6-kb HindIII-XbaI fragment from pJJ359 (Fig. 1,A) into the pCRII vector (InVitrogen, San Diego, CA), which is useful because it has no AatII sites. PCR was performed on first strand cDNA derived from mixed-stage C. elegans RNA using primers emb9ectA (−44 to −9, relative to translation start) and emb9ectB2 (+3268 to +3250). Emb9ectA has a single nucleotide change, T-38G, relative to wild-type to create a HindIII site. Emb9ectB2 has a synthetic HindIII site appended. PCR products were HindIII digested and cloned into HindIII-digested pJJ400 to generate pJJ401. The 1.6-kb AatII fragment was removed from pJJ401 and replaced with the 1.6-kb AatII fragment from pJJ359 to generate pJJ402. PCR was performed on first strand cDNA derived from mixed-stage C. elegans RNA using primers emb9ectD (+5725 to +5750) and embtwoHA, which is the complement of +7217 to +7231 with two HA epitope tags (67), a termination codon, and an Xba site appended. PCR products were SaII-XbaI digested and cloned into SaII-XbaI–digested pJJ402 to generate pJJ404. The insert in pJJ404 was isolated after NsiI-ApaI digestion and ligated into NheI-KpnI–digested pPD30.38, after both fragments had been blunt ended with T4 DNA polymerase. pPD30.38 contains the unc-54 body wall muscle myosin promoter and enhancer (35). The resulting clone, pJJ414 (Fig. 1 A), has the unc-54 promoter and enhancer, the emb-9 gene with introns 1–6, 10, and 11 removed and two copies of HA inserted immediately preceding the termination codon, and 3′ untranslated sequences from pPD30.38.

Plasmid pJJ427 was constructed by ligating the 3.5-kb ApaI-KpnI fragment from pPD30.38 with the 7.8-kb PvuI-PstI fragment from pJJ359 after blunt ending. This construct has the wild-type emb-9 gene (−15 to +7750) under control of the unc-54 promoter and enhancer. pJJ427 was used to generate transgenic strains in unc-36(e251) emb-9(g23cg45)/qC1 animals. The constructs pJJ445 and pJJ429 were generated by ligating the emb-9 gene and reporter containing NsiI-SpeI fragments from pJJ444 or pJJ446 with the unc-54 promoter containing NsiI-XbaI fragment from pJJ427.

We have used four antisera to examine the distribution of type IV collagen chains in C. elegans. Two antisera were generated against fusion proteins containing the NC1 domains of EMB-9 (NW155) or LET-2 (NW153), one against a peptide from the NC1 domain of EMB-9 (NW1910), and one against a peptide derived from an interruption of the Gly-X-Y domain of LET-2 (NW68). Each of the sera was affinity purified using the original antigen. In addition, the fusion protein sera were cross adsorbed on columns carrying the fusion protein from the other type IV collagen chain. The purified anti–EMB-9 antisera do not cross-react with the LET-2 fusion protein, and the purified anti–LET-2 antisera do not cross-react with the EMB-9 fusion protein (data not shown).

On Western blots of total C. elegans extracts, both EMB-9 antisera react with a major band of ∼240 kD, and both LET-2 antisera react with a major band of ∼210 kD (Fig. 2). The predicted molecular masses of EMB-9 and LET-2, after cleavage of signal peptides, are 168 and 165 kD, respectively. Their higher apparent molecular masses are at least partly due to the abnormally slow migration of collagens relative to globular protein molecular mass standards on SDS–polyacrylamide gels (17). The apparent molecular masses of the reacting bands differ depending on the percentage acrylamide in the gels, another characteristic of collagens. The fact that the EMB-9 and LET-2 antisera specifically react with proteins of different mobilities indicates that they are specific for the chains against which they were made. Activity on Western blots is eliminated by competition with the original antigen, but not by competition with antigen from the other type IV collagen chain (Fig. 2 B). Similar results were obtained in competition experiments on embryos prepared for immunofluorescence. The specificity of the antisera is further supported by the fact that the anti–EMB-9 antisera do not react with embryos homozygous for either a stop codon or an internal deletion of emb-9 (20a).

Ultrastructural analyses of C. elegans (1, 64–66) have identified basement membranes on the pseudocoelomic face of the hypodermis and surrounding the body wall muscles, pharynx, intestine, and gonad (Fig. 3). We have determined the distributions of the α1 and α2(IV) collagen chains EMB-9 and LET-2 in whole mount animals using specific anti–type IV collagen antisera. All four antisera generate essentially identical immunofluorescence patterns, so below we refer to type IV collagen localization rather than to the separate chains. The colocalization of EMB-9 and LET-2 chains further supports the specificity of the antisera and suggests that EMB-9 and LET-2 may assemble into the same heterotrimeric collagen molecule.

At 20°C, C. elegans embryogenesis takes ∼800 min from first division to hatching (55). Most cell divisions are completed during the first 350 min, after which the embryo begins elongation from an ovoid ball into a tube approximately four times the length of the egg. We refer to the stages of morphogenesis, from the start of elongation to hatching, by the length of the embryo relative to the egg shell, progressively as lima, comma, 1 1/2-fold, twofold, threefold, and fourfold.

Type IV collagen is first detectable in body wall muscle cells at the onset of morphogenesis, just preceding the lima stage (data not shown). At this early stage, the type IV collagen stain appears to be completely intracellular. By the 1 1/2-fold stage, strong type IV collagen stain is seen in body wall muscle cells and at the interface between them and the hypodermis (Fig. 4, A and B). Weak stain is seen surrounding the pharyngeal and intestinal primordia beginning at the comma stage, and this stain becomes intense by the 1 3/4-fold stage (Fig. 4, C and D). Intracellular stain is also seen in the head mesodermal cell and its lineal homologue MS.pppaaa by this stage (Fig. 4, E and F).

From about the comma stage on, stain for type IV collagen accumulates in four stripes at the interfaces between the hypodermis and the four body wall muscle quadrants (Fig. 4, G–J). There is no stain detectable in the hypodermal basement membrane that is located between the muscle quadrants. There is also no detectable stain on the pseudocoelomic face of body wall muscle cells. This same staining pattern is seen with all four type IV antisera and is not altered after treatment with acid urea to unmask potentially hidden epitopes (69).

After the twofold stage, six cells in the nerve ring frequently exhibit stain for type IV collagen (Fig. 4,K). Based on their positions, these cells appear to be the GLR cells, which are considered to have glial-like characteristics (66). The GLR cells have broad thin processes that separate the muscle arms that project under the nerve ring from the pharynx. At the time of hatching, strong type IV collagen stain is seen at the interface between the body wall muscle cells and the hypodermis and surrounding the pharynx, intestine, and primordial gonad (Fig. 4 L).

In larval and adult animals, the distribution of type IV collagen generally remains the same as that seen in late embryos; underlying the body wall muscles and surrounding the pharynx, intestine, and gonad (Fig. 5,C). As in embryos, type IV collagen is not detectable in the hypodermal basement membranes between body wall muscle quadrants or on the pseudocoelomic face of body wall muscles (Fig. 5, A and B). In larvae and adults, coelomocytes can be seen to stain with the type IV antisera (Fig. 5,D). The basement membrane surrounding the gonad stains for type IV collagen throughout larval development and in adults. However, in many L4 and adult animals the gonadal stain is not uniform, being more intense around the spermatheca (Fig. 5,E) and around the distal ends of the gonad arms. Although the basement membranes surrounding the pharynx and intestine stain with the type IV antisera, intracellular stain in these tissues was never seen in embryos, larvae, or adults. A schematic summary of the observed type IV collagen staining pattern is shown in Fig. 3.

The sites of emb-9 and let-2 expression have been determined using lacZ and GFP reporter constructs in transgenic animals. For both genes, the reporter constructs were derived from clones that are capable of rescuing mutant animals when present in transgenic arrays. For emb-9, constructs in which lacZ or GFP are inserted into exon 10 and that have 4.3 kb of 5′ upstream sequences, pJJ444/446 (Fig. 1 A), show strong activity in the cells that exhibit intracellular stain with anti–EMB-9 antisera. A construct with 1.7 kb of 5′ upstream sequences in a transcriptional fusion with lacZ, pJJ318, has also been analyzed, both as an extrachromosomal array and integrated into the genome. This construct has an expression pattern very similar to pJJ444 but often shows activity in two to four of the most posterior intestinal cells in early larvae and has lower activity in body wall muscles of late larvae and adults.

For let-2, 3.2 kb of 5′ upstream sequences, which is more than is required for transgenic rescue of mutant animals, shows very little activity when used in a transcriptional fusion with lacZ, pJJ325 (Fig. 1 B). However, insertion of lacZ into exon 13 of let-2 generates a translational fusion, pJJ360, that shows strong activity in the same cells that stain intracellularly with type IV collagen antisera. The presence of further 3′ exons and flanking sequences, pJJ361, have no effect on the observed expression pattern. The let-2 expression reporters, pJJ360/361, show the same spatial and temporal expression patterns as the emb-9 pJJ444/446 reporter constructs.

Weak β-galactosidase and GFP activity can first be detected in body wall muscle cells of lima stage embryos (data not shown). In embryos of all later stages, β-galactosidase and GFP activity is strong in body wall muscle cells (Fig. 6, A and B), and no activity is seen in pharyngeal, intestinal, or hypodermal cells. In situ hybridization of C. elegans embryos with a probe derived from emb-9 also shows localization in the body wall muscles (56). In L1 larvae (Fig. 6,D), body wall muscle cells continue to show activity, and activity is also seen in the ring ganglion (presumptive GLR cells), head mesodermal cell, coelomocytes, and intestinal and anal muscles. Beginning at the L2 larval stage, the distal tip cells of the developing gonad show strong activity (Fig. 6,C). In L3–L4 animals, the level of activity in the distal tip cells often appears to be greater than that seen in the body wall muscles (Fig. 6,E). In L4 animals, spermathecal cells begin to show β-galactosidase activity, and this activity can still be seen in adults (Fig. 6,F). In adult animals, type IV collagen reporter activity is seen in vulval and uterine muscles (Fig. 6,F), body wall muscles (Fig. 6,G), as well as distal tip cells (Fig. 6 H).

The emb-9 and let-2 reporters showed activity in the same cells that stained with the anti–EMB-9 and –LET-2 antisera, with the exception of some accessory muscles. Although the intestinal, anal, vulval, and uterine muscles showed activity with type IV expression reporters, intracellular staining of these cells with anti–type IV collagen antisera was not apparent. It is possible that antibody staining of these cells was weak and/or obscured by the intense stain of adjacent intestinal, gonadal, and body wall muscle basement membranes.

It is notable that no emb-9 or let-2 reporter activity was seen in pharyngeal, intestinal, or hypodermal cells of embryos, larvae, or adults. This result is consistent with the fact that anti–type IV collagen antisera did not show intracellular stain in these tissues. However, it is surprising because the pharynx and intestine are covered by basement membranes that do stain with the anti–type IV collagen antisera. Although the pharyngeal and intestinal basement membranes contain EMB-9 and LET-2, the genes encoding them are not expressed in these tissues.

The strong expression of emb-9 and let-2 in body wall muscle cells suggests that they could be the source of the type IV collagen found in the pharyngeal and intestinal basement membranes. To determine if type IV collagen produced in body wall muscle cells can assemble on other tissues, we constructed an epitope-tagged emb-9 gene that is transcriptionally controlled by the body wall muscle–specific unc-54 promoter and enhancer, pJJ414 (Fig. 1 A). The unc-54 gene encodes the body wall muscle myosin heavy chain MHC-B (33), and the specificity of its promoter has been well documented (44). The epitope tag consists of two tandem copies of the nine–amino acid HA peptide (67). Transgenic lines carrying the construct were generated in the wild-type N2 background and stained with anti-HA and –type IV collagen antibodies.

In lima through 1 1/2-fold embryos, the anti-HA stain is seen in body wall muscle cells (Fig. 7,A), coincident with stain for the LET-2 chain (Fig. 7,D). In later embryos, when wild-type type IV collagen is localized largely in basement membranes (Fig. 7,E), the HA-tagged molecules are mostly still intracellular (Fig. 7,B). This intracellular retention appears similar to that seen in animals with mutations in either of the type IV collagen genes (20a). In very late embryos that are close to hatching, weak HA staining is detectable on the surface of the pharynx and intestine (Fig. 7, C and F). Thus, the epitope-tagged EMB-9 chain is expressed in body wall muscle as expected; however, the presence of the HA epitope tag at the carboxyl end of EMB-9 appears to interfere with its proper assembly and/or secretion.

The inhibition of HA-tagged EMB-9 secretion is not complete, however, since in larvae and adults carrying pJJ414, strong staining with anti-HA antibodies is seen covering the pharynx and intestine (Fig. 7, G and H). This staining is coincident with that seen with anti–type IV collagen antisera (Fig. 7, I and J). Staining with HA antibodies beneath body wall muscles and over the gonad can also be detected. Animals in the same preparation that do not carry the transgenic array show no staining, demonstrating the specificity of the HA antibody (Fig. 7, G–J). These results show that type IV collagen produced in body wall muscle cells can assemble into the basement membranes covering other tissue.

As a test of how much type IV collagen function can be provided by expression in body wall muscle cells, we asked how far animals could develop if they produced functional type IV collagen only in body wall muscle cells. The body wall muscle-specific unc-54 promoter and enhancer were placed immediately upstream of the initiation codon of the complete wild-type emb-9 gene to create pJJ414 (Fig. 1 A). This construct was used to generate transgenic arrays in animals carrying an emb-9 nonsense mutation linked to an Unc marker maintained over a balancer chromosome, unc-36 (e251) emb-9(g23cg45)/qC1. Normally, this stain segregates no Unc progeny because of the strict embryonic lethality of the emb-9 nonsense mutation.

We expected that, at best, sterile adults could develop from emb-9(g23cg45) animals carrying pJJ427, since the normal gonadal type IV collagen expression would be absent. To our surprise, viable and fertile transgenic Unc animals were generated by injecting the unc-54:emb-9 construct. We have established several lines that are homozygous unc-36(e251) emb-9(g23cg45) and are rescued by arrays carrying the pJJ427 construct. The rescued animals grow slowly; many arrest during larval development, and those that reach adulthood have small brood sizes. However, the strains can be maintained. These results indicate that expression of functional type IV collagen only in body wall muscles is sufficient for C. elegans to develop to adulthood and to reproduce.

It is possible that the presence of the complete emb-9 coding region in the pJJ427 construct could alter the expression pattern of the unc-54 promoter and enhancer. To test this possibility, GFP was inserted into exon 10 of the construct to create pJJ429 (Fig. 1,A). Transgenic animals carrying pJJ429 show strong GFP activity only in body wall muscle cells (Fig. 8 A). No activity was seen in the pharynx, intestine, or gonad. Therefore, rescue of emb-9 (g23cg45) mutant animals is possible by expression of functional emb-9 solely in body wall muscle cells.

Embryos homozygous for emb-9(g23cg45) do not stain with anti–EMB-9 antisera and show only intracellular accumulation, no basement membrane stain, with anti– LET-2 antisera (20a). We stained pJJ427-rescued emb-9 (g23cg45) embryos with anti–EMB-9 antisera and found that EMB-9 is present in pharyngeal, intestinal, and gonadal basement membranes and under body wall muscles (Fig. 8, B–D). This result further demonstrates that type IV collagen produced in body wall muscle cells can assemble into basement membranes covering other tissues, and that it can do so in the absence of any other source of type IV. It also shows that type IV collagen derived from body wall muscles is sufficient for development of C. elegans to the adult stage and for at least partial fertility.

We have used specific antisera to localize the α1 and α2(IV) collagen chains of C. elegans, EMB-9 and LET-2. Type IV collagen was first detectable within body wall muscle cells of C. elegans embryos just beginning morphogenesis. Consistent with this result, the temperature-sensitive periods for both type IV collagen genes, emb-9 and let-2, also begin at this time (23,40; Kramer, J., unpublished results). Soon after, at the comma stage, secreted type IV collagen begins to form stripes under the body wall muscle quadrants. At approximately the same time, type IV collagen first appears around the primordial pharynx and intestine. By the end of embryogenesis, type IV collagen underlies the body wall muscle cells and surrounds the pharynx, intestine, and gonad primordium. The same general pattern of type IV collagen localization continues in larval and adult animals.

Antisera against EMB-9 and LET-2 give the same spatial and temporal staining patterns in wild-type animals, indicating that these chains are colocalized. Additionally, in animals with missense mutations of either chain, the distributions of the chains are altered but still colocalized, and in animals with nonsense mutations in emb-9, the LET-2 chain is retained intracellularly and not secreted (20a). The colocalization of EMB-9 and LET-2 and the inability of LET-2 to be secreted in the absence of EMB-9 indicates that the two chains do not function independently. Since EMB-9 and LET-2 are α1 and α2-like type IV collagen chains, respectively (19), they may assemble into a single heterotrimeric collagen molecule as do the vertebrate α1 and α2 chains. These results are consistent with the notion that EMB-9 and LET-2 function in a single heterotrimeric type IV collagen molecule.

Basement membranes in two regions of C. elegans did not stain with anti–type IV collagen antisera at any stage, those on the hypodermis between the body wall muscle quadrants and those on the pseudocoelomic face of body wall muscles (Fig. 3). The basement membranes in these regions appear ultrastructurally similar to those that cover the intestine and gonad (66) and do stain for type IV collagen. There are three possible explanations for the failure to detect type IV collagen in these regions of basement membrane.

First, it is possible that EMB-9 and LET-2 are present but their epitopes are masked in these regions. Acid-urea extraction, which can uncover masked epitopes in vertebrate basement membranes (69), of whole animals or frozen sections of C. elegans had no effect on the type IV collagen staining pattern. All four type IV antisera fail to stain these regions in animals that are fixed with paraformaldehyde or methanol/acetone and in frozen sections. These results make the possibility of epitope masking unlikely and support the idea that EMB-9 and LET-2 are absent from these regions.

Second, it is possible that there is type IV collagen present that is produced by an unidentified gene(s). Southern blot analyses indicated that emb-9 and let-2 are likely to be the only type IV collagen gene in C. elegans (19). At this time, with approximately one-half of the C. elegans genome sequence complete (61), no other type IV collagen genes have been identified. Multiple cDNA clones derived from emb-9 and let-2 have been identified by C. elegans EST database projects (62; Kohara, Y., personal communication), but no other type IV collagen transcripts have been seen. Although it is not yet possible to exclude the existence of another type IV collagen gene(s) in C. elegans, the probability is small.

Finally, it is possible that there is no type IV collagen in these regions of basement membrane. Although basement membranes can vary in the collagen isoforms they contain (37, 48), they all appear to contain some type IV collagen. There is evidence that during angiogenesis, newly forming basement membranes are laminin rich and lack type IV collagen (15), and there are cell lines that secrete matrix lacking type IV collagen (5, 16). However, there are no reports of fully formed in vivo basement membranes that do not contain type IV collagen. The regions in C. elegans on the surface of the hypodermis between body wall muscle quadrants and on the pseudocoelomic face of body wall muscle cells may be basement membrane–like structures that lack type IV collagen.

Whichever of the above possibilities is true, it is clear that the composition and/or modifications of the basement membranes between body wall muscle quadrants and on the pseudocoelomic face of muscles must be different from the other basement membranes in C. elegans. Another basement membrane component, the UNC-52 perlecan homologue of C. elegans, is also not detected in these same regions of basement membrane (42).

We have identified the cells that express the emb-9 and let-2 type IV collagen genes using reporter constructs in transgenic animals. With minor exception, the type IV collagen–expressing cells in C. elegans are mesodermal. The mesoderm of C. elegans is derived from four founder cells, MS, D, Cap, and Cpp (54, 55). The D, Cap, and Cpp founders generate only body wall muscle cells, all of which express type IV collagen. The MS cell generates body wall muscles and several other type IV collagen–expressing cell types (Fig. 9). Notably, all of the embryonic MS-derived cells express type IV collagen, except the pharyngeal cells. Type IV collagen expression is also confined to mesodermal cells in Drosophila (25, 32, 39) and sea urchin (9, 63). In vertebrates, type IV collagen expression has been reported in both mesenchymal and epithelial cells of several tissues (34, 49, 57, 58), although it is restricted to the mesenchyme in mouse intestine (52).

The combination of antibody localization and gene reporter results indicated that the pharynx and intestine do not express the α1 or α2(IV) genes, emb-9 or let-2, even though these tissues are covered by basement membranes that contain their products. Also, type IV collagen antisera stain the entire surface of the gonadal basement membrane, even though only a few cells located at the distal and proximal ends of the gonad express the type IV collagen genes. We demonstrated that α1(IV) chains synthesized in body wall muscle can assemble into basement membranes on these other tissues. Additionally, we found that expression of functional type IV collagen only in body wall muscles is sufficient for C. elegans to complete development and be at least partially fertile.

These results raise the question of how the sites of type IV collagen assembly into basement membranes are determined. Presumably, body wall muscle cells secrete type IV collagen from both their hypodermal and pseudocoelomic faces. Body wall muscle cells are closely apposed to the hypodermis, and collagen secreted into the small intervening space could assemble into the stripes seen beneath the body wall muscle quadrants. The type IV collagen that assembles into the pharyngeal, intestinal, and gonadal basement membranes is likely to be secreted from the pseudocoelomic face of the muscle cells. Notably, the basement membranes that cover the pseudocoelomic face of body wall muscle cells do not contain detectable type IV collagen. Since the body wall muscle quadrants extend from the head to the tail of the animal, secretion into and diffusion within the pseudocoelomic cavity could provide the necessary type IV collagen for all other tissues.

The remaining question is why all basement membranes that are exposed to the pseudocoelomic cavity do not contain type IV collagen? As noted above, type IV collagen is not detectable in the basement membranes on the pseudocoelomic face of muscle cells or on the hypodermis between body and wall muscle quadrants. Either something is present in these regions that blocks type IV collagen assembly, or something that is required to promote assembly is absent from these regions. We favor the latter possibility, since assembly could be promoted by simply increasing the local concentration of type IV collagen, while no inhibitors of assembly are known.

Type IV collagen can assemble into a polygonal lattice in vitro in a concentration-dependent manner (70). However, the in vivo concentration of type IV collagen may be too low to allow such unassisted assembly. Higher local concentrations of type IV collagen could be achieved if it bound to cell surface receptors or other cell-associated basement membrane components. Likely candidates for this function are integrins and/or nidogen, both of which have been identified in C. elegans (18; Kramer, J., unpublished results). Binding to these molecules could produce local concentrations of type IV collagen high enough for assembly to occur. Thus, the localization of type IV collagen may be determined by the localization of cell-associated molecules to which it binds.

The sources and final locations of type IV collagen have been determined in some vertebrate systems. In cocultures of mouse epithelium with chick endothelium, the developing kidney glomerular basement membrane was shown to have type IV collagen contributed by both species (49). In cocultures of fetal intestinal chick mesenchyme with rat endoderm, the resulting subepithelial basement membrane was found to have type IV collagen derived only from the mesenchyme (52, 53). In cocultures of bovine keratinocytes and human fibroblasts, both cell types contributed type IV collagen to the dermal–epidermal basement membrane (34), while nidogen was derived only from fibroblasts (14). So, the type IV collagen in a basement membrane can be derived from cells on either or both sides of it. In these cases, the two groups of cells contributing to basement membrane formation are generally present along a continuous interface. In C. elegans, the situation is different because the body wall muscle cells that produce type IV collagen are not immediately adjacent to all sites at which it assembles (Fig. 3). However, the localized assembly of type IV collagen in both the vertebrate and C. elegans studies suggest that it must interact with other molecules to direct its assembly.

The assembly properties of the perlecan homologue of C. elegans, UNC-52, are different from those of type IV collagen. UNC-52 is deposited in the basement membrane close to its sites of synthesis (42). The tissues that have UNC-52 in their basement membranes all express the unc-52 gene. Laser ablation of body wall muscle cells results in a hole in the UNC-52 antibody staining pattern where the cells are missing, indicating that UNC-52 does not move far from its site of synthesis. In C. elegans, two basement membrane components assemble in different ways; perlecan is deposited locally, while type IV collagen can assemble at sites distant from its site of synthesis.

We wish to thank David Miller (Vanderbilt University, Nashville, TN) for providing antibodies and Andrew Fire (Carnegie Institute, Baltimore, MD) for providing expression vectors.

This work was supported by National Institutes of Health grant HD22027 (J.M. Kramer) and NRSA GM15268 (P.L. Graham).