Basement membranes are thin extracellular matrices that have important roles in the development and functions of many tissues. The major constituents of basement membranes include type IV collagens, laminins, nidogen (entactin), and proteoglycans. Interactions between these constituents and their interactions with cell surface molecules are critical for the assembly and function of basement membranes (Yurchenco, 1994 ; Timpl, 1996 ). Collagen IV and laminin can independently polymerize in a concentration-dependent manner to form separate networks in the basement membrane (Yurchenco and Furthmayr, 1984 ; Yurchenco et al., 1992 ; Yurchenco and Cheng, 1993 ). Connection between these independent networks is thought to be mediated by nidogen, which can bind both collagen IV and laminin with high affinity (Fox et al., 1991 ).

Nidogen is a 150-kDa glycoprotein that consists of two amino (G1, G2) and one carboxyl (G3) terminal globular domains that are connected by a rod domain composed primarily of endothelial growth factor (EGF) repeats (Durkin et al., 1988 ; Mann et al., 1989 ; Fox et al., 1991 ). Mammals express two closely related forms of nidogen, nidogen-1 and nidogen-2, that are encoded by separate genes (Kohfeldt et al., 1998 ). Both nidogen-1 and nidogen-2 bind type IV collagen, perlecan, and laminin with high affinity (Aumailley et al., 1993 ; Chung et al., 1993 ; Kohfeldt et al., 1998 ). The G3 domain of nidogen-1 binds strongly to a single EGF module in the laminin γ1 chain (Mayer et al., 1993 ), whereas nidogen-2 shows strong interaction with LG domains of the laminin α2 chain (Kohfeldt et al., 1998 ; Talts et al., 1999 ). Binding of nidogen-1 to collagen IV and perlecan occurs through the nidogen G2 domain (Reinhardt et al., 1993 ). Using purified proteins, nidogen-1 was shown to form a ternary complex with type IV collagen and laminin-1 (Aumailley et al., 1993 ). The ability of nidogen to bind both the collagen IV and laminin networks with high affinity suggests that this linking function may be critical for basement membrane assembly.

Several lines of evidence suggest that nidogen-1 has important functions in vertebrate development and differentiation. Nidogen-1 promotes attachment of cells in culture, mediated through the single RGD sequence within its rod domain and by sequences in the G2 domain (Yelian et al., 1993 ; Dong et al., 1995 ; Wu et al., 1995 ). Addition of antibodies that interfere with nidogen-1 binding to laminin disrupts epithelial morphogenesis of cultured embryonic kidney, lung, or submandibular glands (Ekblom et al., 1994 ; Kadoya et al., 1997 ). Nidogen proteolysis has been correlated with loss of epithelial function during normal mammary gland involution and during gland regression induced by ectopic stromelysin-1 expression (Alexander et al., 1996 ). Nidogen-1 can cooperate with laminin-1 to stimulate β-casein synthesis by cultured mammary epithelial cells (Pujuguet et al., 2000 ). These studies suggest that nidogen may have critical developmental functions in vivo.

To examine the in vivo functions of nidogen, we initiated studies of its roles in Caenorhabditis elegans development. Several basement membrane components have been shown to be essential for embryonic development in C. elegans (Kramer, 1997 ). Loss or alteration of type IV collagen (Guo et al., 1991 ; Sibley et al., 1993 ; Gupta et al., 1997 ), perlecan (Rogalski et al., 1993 ), or SPARC/osteonectin (Fitzgerald and Schwarzbauer, 1998 ) result in arrest during the mid to late stages of embryogenesis. Mutations in other basement membrane components that are required for the proper assembly of collagen IV, perlecan, or SPARC might be expected to also result in defective development.

The mechanisms of assembly of these molecules into basement membranes may differ. Perlecan is expressed in muscle cells and assembles into the basement membranes immediately adjacent to the cells that synthesize it (Moerman et al., 1996 ). In contrast, type IV collagen (Graham et al., 1997 ) and SPARC (Fitzgerald and Schwarzbauer, 1998 ) assemble into basement membranes associated with tissues that do not express the proteins. Epitope-tagged type IV collagen synthesized in body wall muscle cells can assemble into basement membranes that surround the pharynx, intestine, and gonad. Expression only in muscle cells was also shown to be sufficient to rescue the lethality of a type IV collagen null mutant and to provide modest fertility (Graham et al., 1997 ). In addition, type IV collagen is localized to certain basement membranes in C. elegans and absent from others, even though all are open to the pseudocoelom, which is the source of free collagen IV molecules. These studies indicated that a mechanism for directing type IV collagen assembly to particular locations must exist.

Laminin appears earlier than type IV collagen during mouse (Leivo et al., 1980 ; Dziadek and Timpl, 1985 ) and Drosophila development (Kuschegullberg et al., 1992 ) and during formation of new basement membranes in angiogenesis (Form et al., 1986 ). In a coculture of epithelial and mesenchymal cells, antisense inhibition of laminin expression blocked deposition of type IV collagen and nidogen at the cellular interface, indicating the importance of laminin for basement membrane assembly (De Arcangelis et al., 1996 ). Because nidogen can serve as a link between type IV collagen and laminin, it is possible that collagen IV could be localized by binding to nidogen that is associated with laminin already present at tissue surfaces. Thus, the earlier localized laminin could provide a pericellular binding site to concentrate collagen IV and allow its assembly. Nidogen would be required to meditate this interaction because collagen IV and laminin have not been found to directly interact.

To determine what roles nidogen might play in C. elegans development and whether it is involved in localization of type IV collagen assembly, we generated mutations in the C. elegans nidogen gene. Here we present characterization of the C. elegans nidogen gene nid-1 and analysis of mutations in it.

Brenner, 1974

Plasterk, 1995

Graham et al., 1997

Dziadek and Timpl, 1985

Graham et al., 1997

Finney and Ruvkun, 1990

Graham et al., 1997

Francis and Waterston, 1985

Sambrook et al., 1989

Mohler et al., 1998

To determine fecundity, individual animals were picked at the L4 larval stage and were transferred daily to new plates until they stopped laying fertilized eggs. The total numbers of eggs laid each day was counted using a dissecting microscope. For the reported analysis animals were maintained at 20 ± 0.5°C, but similar results were also obtained at 24°C. To count spermatozoa, young adult animals were fixed with 4% paraformaldehyde in PBS for 4 h at 4°C, washed in PBS, postfixed with 0°C methanol for 5 min, and incubated with 100 mM DAPI for 1 h. A z-dimension stack of digital fluorescence microscopic images through the spermatheca was generated, and sperm nuclei were counted. Actin filaments were visualized using a modified rhodamine phalloidin staining procedure (Strome, 1986 ; Rose et al., 1997 ). Adult hermaphrodite gonads were extruded by decapitation in culture medium (80 mM NaCl, 20 mM KCl, 10 mM MgCl₂, 5 mM HEPES, pH 7.2, 10% FBS, 0.01% tetramisole), fixed with 3% paraformaldehyde in PBS containing 0.1% Tween-20 (PBS-TW) for 2 h at room temperature, washed in PBS-TW, and incubated with 0.33 μM rhodamine phalloidin (Molecular Probes, Inc., Eugene, OR) in PBS-TW.

Waterston and Sulston, 1995

Springer, 1998

Figure 1 Structure of the C. elegans nid-1 nidogen gene. (A) Exons are indicated as boxes with numbers below. The nid-1A-specific exon 11 is colored dark gray, and the nid-1A– and B–specific exons 12–14 are colored light gray. The extents (more ...)

Table 1 Amino acid sequence comparisons of C. elegans and mammalian nidogens

Figure 2 Comparisons of nidogen protein structures and sequences. (A) Domain structures of C. elegans NID-1, mouse nidogen-1 (Mann et al., 1989 ), human nidogen-2 (Kohfeldt et al., 1998 ), and Ascidian nidogen (Nakae et al., 1993 ). Domains are represented (more ...)

The predicted exon structure of nid-1 suggested that some of the rod domain EGF repeats could be removed by alternative splicing. We analyzed RT-PCR products and cDNA clones derived from nid-1 to demonstrate that three isoforms (nid-1A, B, C) are generated by alternative splicing (Figures 1 and 2A). All three isoforms contain the amino G1 and G2, and carboxyl G3 globular domains, but they differ in the number of EGF repeats in the rod domain. The longest form, NID-1A, contains all 12 EGF repeats in the rod domain, NID-1B is missing the single EGF repeat encoded by exon 11, and NID-1C lacks 7 EGF repeats encoded by exons 11–14. The NID-1C rod domain, containing five EGF repeats, is most similar in length to mammalian nidogen-1, which has four EGF and one thyroglobulin repeat (Figure 2A).

There is a single RGD (700–702) sequence in mouse nidogen-1, located near the end of the first rod domain EGF repeat. This RGD sequence has been shown to provide a major nidogen cell attachment activity in vitro (Mann et al., 1989 ; Dong et al., 1995 ). There is also a single RGD (714–716) present in NID-1, and it is located in the same region of the molecule, at the start of the second rod domain EGF repeat. This RGD sequence is present in all three NID-1 isoforms. There are no RGD sequences in nidogen-2 or the ascidian nidogen.

Figure 3 Northern and Western blot analyses of nid-1 expression in wild-type and mutant animals. (A) Detection of nid-1 transcripts in RNA extracted from wild-type (N2) and nid-1 mutant (cg118, cg119) animals. RNAs were extracted from embryos (Embryo) or (more ...)

Probing with exon 13, which would hybridize to nid-1A and nid-1B, did not reveal any transcripts beyond those detected when exon 11 was used as a probe to visualize the nid-1A transcripts. The predicted size of the nid-1B transcript (4.6 kb) is only 4% smaller than nid-1A, so it may be present in low abundance but masked by the nid-1A band. These experiments show that nid-1A is the major transcript expressed during embryogenesis and that detectable nid-1C expression does not begin until after the completion of embryogenesis.

Specific NID-1 antiserum was raised against a bacterially expressed fusion protein containing the carboxyl-terminal G3 domain that is common to all NID-1 isoforms. Western blots of embryonic C. elegans extracts show a strong immunoreactive band migrating at ~170 kDa and a weak band at 160 kDa (Figure 3B). Extracts of mixed stage animals show the 170- and 160-kDa bands as well as a strong band at 150 kDa. The 170- and 150-kDa bands are likely to be NID-1A and C isoforms, because they are similar to the predicted masses of 172 and 136 kDa, respectively, and they exhibit the same temporal characteristics as the nid-1A and nid-1C transcripts (Figure 3A). The 160-kDa band may be NID-1B, because it is close to the 166-kDa predicted mass for NID-1B. Two or more lower-molecular-weight bands are variably seen in mixed stage extracts and may represent degradation products of NID-1. These results are consistent with the RNA expression data in showing that NID-1A is the major embryonic form and NID-1C is not strongly expressed until after embryogenesis.

Figure 4 Localization of NID-1 in wild-type animals. Animals were stained with anti–NID-1 (A, C, E, and G–K) and anti-myosin A (B, D, F, and L). (A and B) A comma stage embryo shows strong accumulation of NID-1 around body wall muscle cells (more ...)

Under the body wall muscles of larval and adult stage animals NID-1 is organized as punctate lines (Figure 4, K and L). These lines follow the rows of dense bodies within the muscle cells (Figure 5,A–C). NID-1 also accumulates strongly at the outer edges of the muscle quadrants and more weakly at the boundaries between muscle cells within each quadrant. Less organized NID-1 staining is seen in the regions between the body wall muscle quadrants, presumably associated with the epidermal basement membranes in these regions.

Figure 5 NID-1 localization to sublateral nerves and muscle. Wild-type N2 (A–E) or nid-1(cg118) (F and G) animals were stained with anti–NID-1 (A, C–G) or anti–α-actinin (B and C) antibodies. (A–C) Part of a body (more ...)

NID-1 accumulates along the four sublateral nerves that run beneath the center of each muscle quadrant (Figure 5, D and E). The sublateral nerves extend along dorso- and ventrolateral tracts from the nerve ring in the anterior of the animal to near the middle of the animal where they turn further lateral to positions coincident with the lateral edges of the body wall muscle quadrants. NID-1 accumulation on these nerves is seen in larval and adult animals and is similar in intensity to the staining at the edges of the body wall muscle quadrants. Staining along the edges of the body wall muscle quadrants appears to be associated with the muscle edges rather than the nerves in these regions because the staining closely follows the edge of the muscles and it does not display the left/right asymmetry expected for the ventral and dorsal nerve cords. As noted above, less organized NID-1 staining is also present in the regions between body wall muscle quadrants.

The effects of both deletions on nid-1 expression were examined at both the RNA and protein levels. cg118 animals contain nid-1 transcripts of ~3.2 and 2.2 kb, ~1.8 kb shorter than in wild type (N2), as predicted by the extent of the deletion (Figure 3A). The truncated nid-1A and C transcripts are present at approximately wild-type levels. No nid-1 transcripts are detectable in RNA extracted from cg119 animals, although control ama-1 RNA is detected at normal levels in these same samples. Western blots of cg118 extracts reacted with anti–NID-1 antiserum show strong bands of ~113- and 72-kDa apparent masses and a weak band of 99 kDa (Figure 3B). These sizes are similar to the 103-, 67-, and 97-kDa masses predicted for the cg118 truncated NID-1A, C, and B products, respectively. The truncated NID-1A and NID-1C proteins are present at approximately wild-type levels. No NID-1 protein is detectable on Western blots of extracts from cg119 animals reacted with anti–NID-1 antiserum (Figure 3B), and NID-1 is not detectable in these animals by immunofluorescence (Figure 6,E and F). These results show that the cg119 deletion allele is a molecular null, whereas the cg118 deletion mutant produces approximately wild-type levels of truncated NID-1 that is missing all of the G2 and part of the G1 domains.

Figure 6 NID-1 localization in cg118 and cg119 mutant animals. Animals were stained with anti–NID-1 antiserum (A, C, E, G, H, and I) or anti-myosin A (B, D, F, and J). (A and B) Twofold stage cg118 mutant embryo shows NID-1 localization under body (more ...)

Reinhardt et al., 1993

Kohfeldt et al., 1998

Graham et al., 1997

Figure 7 Type IV collagen localization in wild-type and nid-1 mutant animals. Wild-type N2 (A–D, K, L, N, P, and Q), cg118 mutant (E–H, M, O, R, and S) or cg119 mutant (I, J, T, and U) animals were stained with anti–LET-2 type IV collagen (more ...)

Ekblom et al., 1994

Kadoya et al., 1997

Nelson et al., 1983

Kimble and Hirsh, 1979

Table 2 Effects of nid-1 mutations on fecundity

Nidogen appears as a major component of basement membranes in all organisms that have been characterized. The ability of nidogen to bind with high affinity to both type IV collagen and laminin suggested that it might be a critical molecule for assembly of basement membranes. In vitro studies in vertebrate systems have supported the potential importance of nidogen for cell adhesion and tissue morphogenesis. We have shown that loss of nidogen in C. elegans results in viable animals that are fertile and that display no overt abnormal phenotypes. Loss of other C. elegans basement membrane components, such as type IV collagen (Guo et al., 1991 ; Sibley et al., 1993 ; Gupta et al., 1997 ), perlecan (Rogalski et al., 1993 ), or SPARC (Fitzgerald and Schwarzbauer, 1998 ), results in embryonic lethality. If nidogen were critical for the assembly or function of these constituents of basement membranes, then its absence would also be expected to result in lethality. The fact that this is not the case provides strong evidence that nidogen is not required for these molecules to assemble into functional basement membranes.

nid-1 is the only gene in C. elegans that encodes a classical nidogen with G1, G2, rod, and G3 domains. There are, however, other genes that encode nidogen-related domains. Three predicted gene products (B0393.5, D1044.2, K03H1.5) contain G1-like domains. All three also encode predicted transmembrane domains near their carboxyl-termini, suggesting that they are cell surface transmembrane proteins. Six distinct predicted gene products contain YWTD motifs related to the nidogen G3 domain: five are predicted to be transmembrane, one is predicted to be secreted. All six also contain EGF repeats but no other similarities to nidogen. There are no other genes that encode nidogen G2–related domains. It is possible that the nidogen G1– and/or G3–related domains of these other gene products could replace NID-1 functions in nid-1 mutant animals.

Our immunolocalization studies show that NID-1 accumulates at all sites where type IV collagen was previously shown to assemble (Graham et al., 1997 ) but is additionally found at sites where little or no collagen IV is detected. In particular, NID-1 shows strong accumulation around the nerve ring, whereas collagen IV is only weakly detected in this region. NID-1 accumulates on the sublateral nerves, where collagen IV shows no preferential accumulation. Both collagen IV and nidogen show strong accumulation under body wall muscles during embryonic development, but the level of nidogen decreases during larval and adult stages, whereas collagen IV levels are maintained. Nidogen accumulates strongly at the edges of the muscle quadrants and only weakly at the junctions between adjacent muscle cells, whereas collagen IV shows the opposite pattern, being strongest at muscle cell junctions. NID-1 also shows diffuse accumulation on the regions of the epidermal basement membrane that lie between the body wall muscle quadrants, whereas collagen IV is absent from these regions. Thus, although collagen IV and nidogen are generally coincident, there are sites where substantial differences in their relative abundance are seen. These results indicate that there is not a one-to-one correspondence between collagen IV and nidogen in C. elegans basement membranes.

Type IV collagen assembles into all basement membranes of C. elegans except on the regions of the epidermis that are located between the body wall muscle quadrants and on the pseudocoelomic faces of the muscle quadrants (Graham et al., 1997 ). We had expected that nidogen would be involved in localization of collagen IV assembly, based on its ability to link collagen IV and laminin. However, we showed that type IV collagen assembles in a completely normal manner in the absence of nidogen. How is assembly of type IV collagen restricted to particular locations? Type IV collagen could be localized to cell surfaces by binding to integrins. However, in mutants for two C. elegans integrins, ina-1 α-integrin (Baum and Garriga, 1997 ) and pat-3 β-integrin (Gettner et al., 1995 ), type IV collagen localization appears normal (J. Kramer, unpublished results). It is possible that collagen IV assembly is dependent on other integrins or some other unidentified molecule. It is also possible that there are redundant mechanisms for collagen IV localization, such that loss of any one binding molecule will not reveal abnormalities in collagen assembly.

The mammalian nidogen-1 G2 domain binds type IV collagen and perlecan with high affinity (Reinhardt et al., 1993 ). The C. elegans and mammalian G2 domains show strong sequence conservation, suggesting that the C. elegans G2 domain is likely to also be capable of binding collagen IV and perlecan. The nid-1(cg118) deletion completely removes the nidogen G2 domain as well as parts of G1 and the first EGF repeat of the rod domain. The cg118-deleted NID-1 accumulates to approximately normal levels and shows normal tissue localization. Thus, the G2 domain has little or no role in nidogen stability or assembly into basement membranes. Nidogen localization therefore appears to occur independently of collagen IV and perlecan interactions.

If linking between collagen IV and laminin were a critical aspect of nidogen function, then the cg118 G2 domain deletion might be expected to act as a dominant interfering mutant. The truncated NID-1 could bind to laminin but would be unable to bind type IV collagen. Mammalian nidogen-1 was found to cooperate with laminin-1 to stimulate β-casein synthesis by cultured mammary epithelial cells (Pujuguet et al., 2000 ). However, addition of the nidogen-1 G3 domain inhibited laminin-1 stimulation of β-casein synthesis, suggesting that the presence of the laminin-binding G3 domain in the absence of other parts of nidogen dominantly interferes with laminin signaling. In coculture studies, antisense inhibition of laminin expression blocked the accumulation of type IV collagen and nidogen at the epithelial/mesenchymal interface (De Arcangelis et al., 1996 ), leading to the suggestion that linkage of collagen IV to laminin via nidogen was essential for its assembly. The fact that the cg118 G2 deletion nidogen does not cause dominant phenotypes indicates that its binding to laminin does not interfere with any essential processes. If nidogen was required to link type IV collagen to laminin, then the cg118 deletion NID-1 should dominantly interfere with this process.

We have identified three alternative splice isoforms of NID-1 that contain differing numbers of EGF repeats in their rod domains. NID-1A is the longest and is the predominant embryonic form, whereas the shortest, NID-1C, only appears after the completion of embryogenesis. The different numbers of EGF repeats would change the spacing between the collagen IV and laminin-binding domains of the NID-1 isoforms and could also change their binding repertoires. Alternative splice forms of other nidogen genes have not been reported. However, mammals have two nidogen genes that may provide additional diversity of nidogen function.

The alternative splice isoforms of nid-1 are reminiscent of let-2, the α2(IV) collagen gene of C. elegans. let-2 generates two alternative splice isoforms that dramatically change relative abundance immediately after the completion of embryogenesis (Sibley et al., 1993 ). The C. elegans perlecan gene unc-52 also produces alternatively spliced variants that differ in their expression during development (Mullen et al., 1999 ). The basement membranes of C. elegans appear to undergo significant changes in composition on completion of embryogenesis.

Perturbation of nidogen-1 function in cultured mammalian kidney, lung, or submandibular gland (Ekblom et al., 1994 ; Kadoya et al., 1997 ) and correlation of nidogen proteolysis with mammary gland involution (Alexander et al., 1996 ) have suggested that nidogen has important roles in epithelial function. We particularly looked in C. elegans nid-1 mutants at tissues that would reveal defects in epithelial morphogenesis. In nid-1 mutants no defects were seen in the excretory canals, which are analogous to the vertebrate kidney, the gonad, which forms by directed migration of a basement membrane enclosed tube, or in organization of the hypodermal cells, which form the epidermis of nematodes. We conclude that nidogen is not required for morphogenesis of these epithelia in C. elegans. It is possible that such functions for nidogen have arisen in vertebrates or that the in vitro studies are misleading as to the in vivo functions of nidogen.

We showed that nid-1 mutations cause reduced fecundity. The nid-1 null mutant showed a greater reduction (32%) than the G2 domain deletion mutant (11%). Thus, NID-1 lacking the G2 domain is able to provide some, but not all, of the nidogen function required for normal fecundity. The cause of the reduced fecundity of nid-1 mutants has not been elucidated. However, NID-1 shows strong accumulation around the developing spermatheca, uterus, and vulva. Although no defects were detected in these structures, it is possible that they do not function optimally in the absence of nidogen. The reduced fecundity that results from either of these nid-1 mutations would result in sufficient selective pressure to ensure maintenance of the gene, even in the absence of any other functions.

During preparation of this article a report appeared showing that a nid-1 mutation, ur41, can affect positioning of longitudinal nerves in C. elegans (Kim and Wadsworth, 2000 ). The nid-1(ur41) mutation results in no overt morphological or behavioral phenotypes, in agreement with our findings. The nerve positioning defects are only detectable using GFP markers to visualize the affected axons. The most penetrant defect reported was mispositioning of the dorsal sublateral nerves up toward the dorsal midline. Our demonstration that NID-1 accumulates on the sublateral nerves indicates that it directly associates with the affected axons. The strong accumulation of NID-1 at the edges of the body wall muscle quadrants that we demonstrated is consistent with the proposal that nidogen may influence the guidance decisions of dorsoventrally migrating axons at the interfaces between muscle and epidermis (Kim and Wadsworth, 2000 ).

We thank Dr. J. Culotti for identifying the nid-1(ev608) allele, Dr. J. Hardin for jcIs1, Dr. Y. Kohara for cDNA clones, and A. Coulson for cosmids. Some strains used in these studies were provided by the Caenorhabditis elegans Genetics Center, which is supported by the National Institutes of Health Center for Research Resources. This work was supported by National Institutes of Health grant HD-27211, to J.M.K.