The Nc1/Endostatin Domain of Caenorhabditis elegans Type Xviii Collagen Affects Cell Migration and Axon Guidance

Basement membranes are specialized zones of extracellular matrix (ECM) that perform essential mechanical and signaling roles in development. Besides providing structural support, the ECM modulates the complex array of signals presented to cells during development and normal homeostasis. Components of the ECM—for example, netrin (Culotti and Merz 1998)—signal to cells directly, whereas others localize latent growth factor signaling molecules and/or act as cofactors in growth factor signaling (Dinbergs et al. 1996). In addition, basement membrane degradation produces proteolytic fragments that can reveal cryptic ligands that affect cellular processes (Stringa et al. 2000). The best characterized cell–ECM interactions are mediated through integrins (Jones and Walker 1999), but additional matrix receptors have been identified, such as dystroglycan (Hemler 1999) and discoidin domain receptors (Vogel 1999).

Genetic characterization of basement membrane components and their receptors in Caenorhabditis elegans has been used to understand the in vivo functions for some of these molecules. Mutations in genes encoding type IV collagen chains emb-9 and let-2 (Guo et al. 1991; Sibley et al. 1994), perlecan unc-52 (Rogalski et al. 1993), alpha integrins ina-1 (Baum and Garriga 1997) and pat-2 (Williams and Waterston 1994), and beta integrin pat-3 (Gettner et al. 1995) can result in lethality. Mutations in other C. elegans basement membrane components are nonlethal but do affect developmental functions, such as neurogenesis. The null phenotypes for the genes encoding nidogen nid-1 (Kang and Kramer 2000; Kim and Wadsworth 2000) and netrin unc-6 (Ishii et al. 1992) are cell migration and axon guidance defects in the nervous system. In addition, mutations that abrogate the function of specific domains of INA-1 result in cell migration defects (Baum and Garriga 1997). It is apparent that ECM molecules or their subdomains can provide distinct signaling and structural functions.

Vertebrate type XV and XVIII collagens are closely related basement membrane molecules of unknown function (Kivirikko et al. 1994; Muragaki et al. 1994; Oh et al. 1994a; Rehn et al. 1994). Both collagens are broadly expressed and have been localized to a wide variety of basement membranes, including those of the endothelium, epidermis, kidney, lung, liver, peripheral nerves, and brain (Oh et al. 1994b; Rehn and Pihlajaniemi 1994, Rehn and Pihlajaniemi 1995; Muragaki et al. 1995; Hagg et al. 1997; Musso et al. 1998; Saarela et al. 1998a,Saarela et al. 1998b; Sasaki et al. 1998). The COOH-terminal 20-kD fragment of type XVIII collagen, termed endostatin (ES), was isolated from tumor cell culture medium as an inhibitor of endothelial cell proliferation. ES has been shown to inhibit endothelial cell proliferation and migration and angiogenesis both in vivo and in vitro (O'Reilly et al. 1997; Yamaguchi et al. 1999). ES has also been reported to induce apoptosis of appropriately stimulated endothelial cells in vitro (Dhanabal et al. 1999). Treatment of mice bearing experimental tumors with ES can result in tumor regression and dormancy (Boehm et al. 1997; O'Reilly et al. 1997). The equivalent COOH-terminal fragment of type XV collagen, termed restin, has also been reported to inhibit angiogenesis and endothelial cell migration (Ramchandran et al. 1999).

The complete 38-kD COOH-terminal noncollagenous domain of type XVIII collagen, termed NC1, has been identified in tissue homogenates, suggesting that it exists as a physiologic cleavage product (Sasaki et al. 1998). At least three functional elements exist within the NC1 fragment: an association domain, a proteolytically sensitive hinge region, and the ES domain. NC1 domains oligomerize noncovalently into trimers through their association domain, whereas ES remains monomeric. NC1 oligomers also show a pattern of interactions with other matrix molecules that is distinct from ES, indicating that the NC1 and ES fragments of type XVIII collagen may have distinct functions (Sasaki et al. 1998, Sasaki et al. 2000).

We have identified a type XV/XVIII collagen homologue in C. elegans, designated cle-1. We show that deletion of the cle-1 NC1 domain results in cell migration and axon guidance defects. These defects can be rescued by ectopic expression of the NC1 domain, which is shown to trimerize in vitro but not by the monomeric ES domain. Ectopic ES expression dominantly inhibits cell motility, phenocopying the NC1 domain deletion and suggesting that ES inhibits NC1 activity. Similar results for vertebrate type XVIII collagen are reported by Kuo et al. 2001(this issue), and, together, these data suggest that monomeric ES can inhibit the promigratory activities of trimeric NC1.

C. elegans culture and manipulation were performed using standard methods (Brenner 1974). All strains were maintained at 20°C. The following strains were used: wild-type N2 var. Bristol, MT2180 nDf23/unc-13(e1091) lin-11(n566), and DP19 unc-119(ed3). The following integrated green fluorescent protein (GFP) neural marker strains were used: a pan neuronal marker (evIs111 [F25B3.3::GFP]), a DA/DB motor neuron marker (evIs82 [unc-129::GFP]), and a mechanosensory neuron marker (muIs32 [mec-7::GFP]).

The cle-1 gene structure was determined by analysis of cDNAs generated by the genome project and sequencing of reverse transcription (RT)–PCR products generated from total C. elegans RNA. The cle-1 isoforms were first reported as three genes (C36B1.2, C36B1.1, and F39H11.4, in 5′ to 3′ order) available from GenBank/EMBL/DDBJ under accession numbers Z80215 and Z81079. The following cDNAs were examined: cm11a3 (Z14325), cm10e6 (M88890), yk96a11 (D75069, D72240), yk297a9 (C26933, C53322), yk12b4 (D27520), and yk238e4 (C61305, C52129). Sequence analysis shows that yk238e4 is a full-length cle-1B cDNA and that cm11a3 is a full-length cle-1C cDNA containing an SL1 splice leader. All nucleotide numbering is based on cosmid F39H11 sequences annotated with cle-1 information (AF164959).

The pBA52 (cle-1A::GFP) transcriptional reporter was generated by inserting the SalI–BamHI (3703–8654) fragment from cosmid F39H11 into pJJ471. The pJJ471 vector contains the polylinker and synthetic transmembrane domain of pPD34.110 (Fire et al. 1990), followed by GFP from pPD95.75 and an unc-54 3′ end cassette. pBA36 (cle-1B::GFP) contains a PCR-generated fragment (10996–11969) inserted into the MscI site of pPD116.16. pBA35 (cle-1C::GFP) was made by fusing an Nsi1 fragment (17696–20385) to PstI-digested pPD114.35. Descriptions of pPD plasmids are available at www.ciwemb.edu/pages/firelab.html.

For ectopic expression of CLE-1 domains, full-length cle-1C cDNA was generated by PCR from cm11a3 using primers SL1 (GGTTTAATTACCCAAGTTTGAG) and cle-1FullR1 (TGGAGCTGACTCATTTTTTGAAGTCGGATG). The product was subcloned into MscI–NheI-digested pPD114.108 (mec-7::GFP::let-838 3′UTR), replacing the GFP coding sequence. The resulting plasmid was digested with EcoRV and SmaI and religated to make pBA43 (mec-7::CelNC1), or digested with Acc65I, blunted by Klenow treatment, and religated to make pBA46 (mec-7::CelES). To make mec-7::CelAD, pBA43 was digested with BglII and NheI and ligated to the BamHI–XbaI GFP coding fragment from pPD114.35. All clones were sequenced to confirm that they had the expected structure.

Transgenic strains were generated as previously described (Mello et al. 1991). The cle-1 promoter::GFP expression plasmids (50 μg/ml) were coinjected with the pDP number MM016B [unc-119+] plasmid at 100 μg/ml into unc-119(ed3) animals. Ectopic expression plasmids at 25 μg/ml were coinjected with pPD114.108 (mec-7::GFP) at 25 μg/ml and tph-1::GFP at 100 μg/ml (Sze et al. 2000) into N2 wild-type animals. Transformants were identified by GFP expression using a Leica MZ3 fluorescent dissecting microscope. Expression of ectopic NC1 or ES domains was checked by RT-PCR of RNA from transgenic animals using the SL1 and F1NestTGA primers (TCATTTTTTGAAGTCGGATGTG). Only lines showing products of the expected size, 1,340 and 815 nt, respectively, for ectopic CelNC1 and CelES transcripts were analyzed. Transgenic arrays were passed into the cg120 background by mating to cg120 males.

Deletion mutagenesis was performed as described (Barstead 1999) using UV/TMP as a mutagen. Primers used to identify cg120 were MUTF1 (AACAACAAGCTGCGACAACTCC; 21948–21969), MUTR1 (TCCGGCTACTATCCACAACGAC; 25867–25888), NESTF1 (TGCCCCACCTTCCAATCAATG; 21979–21999), and NESTR1 (CCTCAAATGGGACAACAGCAACC; 25620–25641). The cg120 deletion was shown to remove nucleotides 22756–24758 by sequencing of PCR fragments derived with primers flanking the deletion. The original mutant strain was out-crossed to wild type 10 times before further analysis.

Double-stranded RNA (dsRNA) was prepared as described (Fire et al. 1998) using the complete yk96a11 cDNA as a template. The resulting dsRNA includes sequences extending from nucleotides 21405–25636 on the genomic sequence. dsRNA was injected at 1 mg/ml into the evIs82 or the cle-1(cg120) evIs82 strains. Similar results were obtained using dsRNA at 4 mg/ml.

Microscopy of living animals was performed by mounting on a 2% agarose pad in a small drop of M9 containing 10 μg/ml tetramisole as an anesthetic. Immunohistochemistry of mixed staged cultures (Bettinger et al. 1996) and embryos (Graham et al. 1997) was performed as described. Fluorescent and Nomarski images were obtained using a ZEISS Axoiophot microscope equipped with a Photometrics Sensys charge-coupled device camera. Images were serially deconvolved using MicroTome™ software (VayTek).

Primers CE-3 (5′-ATAAGCTTATGGAGCAGCTGCTGATCCAGT-3′) and CE-4 (5′-ATGGTACCTGATGATGGTGTTAGTATGGAA-3′) were used to generate a 469-bp RT-PCR fragment that was cloned into KpnI–HindIII-digested vector pQE-41 (QIAGEN), which adds an NH₂-terminal His tag to amino acids 287–439 (CLE-1C). The resulting clone, cecolPQE-41, was transformed into Escherichia coli strain M15, and protein expression was performed according to manufacturer's instructions (QIAGEN). Purified fusion protein was used to immunize rabbits. The resulting antisera were affinity purified on columns carrying the purified fusion protein.

CLE-1 NC1 domain sequences were amplified by PCR from cDNA yk234e8 using NC1-5′ (GATCGGCCCAGCCGGCCCATCATCACCATCACCATGACGATGACGATAAGggaactgcagtaaatg-ttcatgcaaccacggtg) and NC1-3′ (GATCGGATCCTTACTTTTTTGAAGTCGGATGTGATTCGATGCAAGC). The CLE-1 ES domain was similarly amplified using ES-5′ (GATCGGCCCAGCCGGCCCATCATCACCATCACCATGACGATGACGATAAGGT-TCATAATAAAGATCGAGTTATTCATATGATTGC) and the NC1-3′ primer. The amplified sequences append NH₂-terminal 6xHis tag and enterokinase cleavage sites to the NC1 or ES domains. PCR products were ligated into SfiI–BamHI-digested pSecTag2A (Invitrogen) as an in-frame fusion with the Ig κ secretion signal and the resultant plasmids transfected by calcium phosphate into 293T cells. Conditioned media from transfected 293T cells were purified over Ni–agarose (QIAGEN) and dialyzed overnight into PBS. Purified CLE-1 NC1 and CLE-1 ES proteins were cross-linked with ethylene glycol bis-succinic acid (EGS) and analyzed by SDS-PAGE, followed by Western blotting using anti–CLE-1 antisera.

The cle-1 gene (collagen with ES domain) is the only homologue of vertebrate type XV/XVIII collagens in the C. elegans genome. cle-1 is a complex gene using three promoters and alternative splicing to generate at least three mRNAs designated cle-1A-C (Fig. 1). Each CLE-1 isoform has a unique first exon beginning with a predicted signal peptide. All isoforms share the collagenous Gly-X-Y repeat domain and COOH-terminal domain with 55% similarity to mouse ES (Fig. 1B and Fig. E). The CLE-1A and -B forms add a region with 35% similarity to vertebrate type XV/XVIII collagens that is related to the NH₂-terminal thrombospondin domain also found in collagens IX, XVI, and XIX (Fig. 1B and Fig. D). The CLE-1A–specific NH₂-terminal domain contains two fibronectin type III (FNIII) repeats with 40% similarity to the fourth and fifth repeats in the UNC-40/DCC family of netrin receptors (Fig. 1B and Fig. C).

4 introns are located in identical positions in cle-1 and the mammalian type XVIII and XV collagen genes, indicating that these genes are true homologues. At the amino acid sequence level, CLE-1 is equally similar to types XV and XVIII collagen. However, both cle-1 and vertebrate type XVIII collagen are expressed as three isoforms (Muragaki et al. 1995; Rehn and Pihlajaniemi 1995), whereas type XV has no identified isoforms. Also, both the CLE-1 and vertebrate type XVIII collagen ES domains have a four–amino acid loop (Fig. 1 E) that is not present in type XV (Sasaki et al. 2000). This loop contains Arg158 and has been shown to contribute strongly to the heparin binding property of ES (Sasaki et al. 1998). Based on these comparisons, we conclude that CLE-1 is most similar to type XVIII collagen. The most notable difference is that the type XVIII collagen NH₂-terminal domain contains a region related to the Wingless receptor frizzled, whereas the equivalent CLE-1 domain is related to the netrin receptor unc-40 (Chan et al. 1996).

cle-1 expression was analyzed in transgenic animals carrying GFP transcriptional reporters driven by sequences upstream of the unique first exon of each isoform. cle-1A::GFP expression was first detected in comma stage (∼390 min) embryos in cephalic neurons (Fig. 2 A). Expression in interneurons and rectal epithelial cells is seen by the twofold stage, ∼60 min later, and continues through larval and adult stages (Fig. 2 E). cle-1B::GFP expression was first observed at the threefold stage (∼520 min) of embryogenesis in four neurons of the lateral ganglia and in the DD dorsal motorneurons (Fig. 2 B). The postembryonic VD dorsal motorneurons also express cle-1B::GFP starting in the second larval stage, and this pattern persists through larval and adult stages (Fig. 2 F).

cle-1C::GFP expression is first seen in comma stage embryos in pharynx and body wall muscles (Fig. 2 C) and later appears in the canal-associated neurons (CAN), head mesodermal cell, accessory muscles, and intestine (Fig. 2 D). During larval development, expression is restricted to the ventral GLR glia–like cells, the canal-associated neurons, and the head mesodermal cell (Fig. 2 G). Beginning in the late fourth larval stage and continuing through adult, cle-1C::GFP expression appears in somatic sheath cells of the gonad and reappears in body wall muscles (Fig. 2 H).

Anti–CLE-1 antiserum was raised against a region common to all isoforms and used to localize the protein in whole animals. CLE-1 strongly accumulates in a punctate pattern associated with the nerve ring and the dorsal and ventral nerve cords. This pattern is first visible at the twofold stage of embryogenesis (∼460 min) and persists in all subsequent stages (Fig. 3). Beginning in the L4 larval stage, CLE-1 is also observed in basement membranes surrounding the gonad, pharynx, and intestine (Fig. 3 C), and under body wall muscles where it preferentially accumulates at junctions between muscle cells and along dense body lines (Fig. 3 D). The apparently more limited distribution of CLE-1 in early development may result from low sensitivity of detection with the antibody.

We isolated a deletion of cle-1, cg120, which removes exons 18–20 and places the final three exons out of frame (Fig. 1 A). This deletion results in CLE-1 proteins lacking the COOH-terminal 231 amino acids, which includes nearly the entire COOH noncollagenous NC1 domain. CLE-1 protein is detectable in cg120 animals by immunofluorescence (Fig. 3, E–H), demonstrating that stable truncated protein is produced.

CLE-1 localizes to the nerve cords, nerve ring, pharynx, intestine, and gonad of cg120 animals in a pattern similar to wild type (Fig. 3, E–H). However, the intensity of CLE-1 staining is reduced in cg120, presumably because 45% of the region used as immunogen is removed by the deletion, and/or the truncated protein is less stable. Also, the accumulation of CLE-1 at the junctions between body wall muscle cells is absent from cg120 animals (Fig. 3 H), suggesting that the NC1 domain is necessary for this localization.

To determine whether there is general disruption of basement membranes in cg120 animals, we examined the distribution of type IV collagen, a major basement membrane component. The collagen IV distribution in cg120 mutants was indistinguishable from wild type (Fig. 3I and Fig. J), and, when normalized to myosin staining, collagen IV levels in the mutants were equivalent to wild type. Type IV collagen assembly appears unaffected by the cg120 deletion, suggesting that general basement membrane structure is not disrupted.

30% of cg120 deletion homozygotes are egg-laying defective (Egl), 30% display distal tip cell migration defects resulting in mispositioned gonad arms, 10% have male tail defects, and they generally display slightly incoordinated movement (Unc). In addition, 1–5% arrest as late embryos or first stage larvae (L1). These phenotypes can be found in various combinations in single animals. The frequency of embryonic or larval arrest increases to ∼35% in animals with cg120 in trans to a deficiency of the region, cg120/nDf23, consistent with the antibody staining results in demonstrating that cg120 is not a null allele.

The Egl phenotype can arise from defects in migration or function of either the hermaphrodite-specific neurons (HSNs) that control egg laying or the sex myoblasts that generate the egg laying musculature. These cells normally migrate from the tail to the midbody region, near the presumptive vulva (Garriga et al. 1993; Burdine et al. 1997). Egl defects can be classified as neuronal or muscular by examining egg laying in response to serotonin or imipramine treatment (Desai and Horvitz 1989). Egg laying by cg120 animals was strongly stimulated by serotonin treatment but only variably stimulated by imipramine, consistent with impaired HSN neural function (data not shown).

HSN positions were visualized in adults using a tryptophan hydroxylase GFP marker that is expressed in these cells (Sze et al. 2000). In 27% (n = 250) of cg120 animals one or both HSNs were located significantly posterior and/or dorsal of their normal position, suggesting that they failed to complete migration from the tail (Fig. 4 B). HSN axon guidance defects were also observed in cg120 animals but may be secondary effects of cell mispositioning (Baum and Garriga 1997). Animals displaying the Egl phenotype generally had mispositioned HSNs, suggesting that the Egl phenotype results from the HSN migration defect.

In addition to HSNs, other cells that undergo long range migrations during C. elegans development include the anterior mechanosensory neurons (ALML/R and AVM) and the canal-associated neurons. We analyzed these cells for migration defects in cg120 mutants (Fig. 5; Table and Table). Interestingly, although the canal-associated neuron is the only one of these cells that expresses cle-1, it did not exhibit any migration defects (Fig. 5 J). In 37% (n = 200) of cg120 animals one or more of the mechanosensory neurons were mispositioned (Fig. 5 B; Table). The ALMs are most often displaced anteriorly, whereas the AVM is displaced posteriorly and/or dorsally. In most cases the cells are mispositioned on the anterior–posterior trajectory along which they normally migrate, although occasionally they are also displaced dorsally or ventrally.

Axon guidance defects are also detected in cg120 animals. In wild-type animals, sensory and motor axons migrate orthogonally to the longitudinal axis to innervate the ventral or dorsal nerve cords (Fig. 4 C). In cg120 mutants, these migrations are frequently defective (32%; n = 50), with axons migrating at oblique angles, crossing one another, or defasciculating from axon bundles (Fig. 4 D and 5 F). These defects may partially explain the mild uncoordination of cg120 animals.

Defects in axon guidance for the DA and DB motorneurons were specifically visualized using an unc-129::GFP reporter (Colavita et al. 1998). The cell bodies of these motorneurons are located in the ventral cord and extend axons along the body wall to the dorsal cord. In wild type, these axons exit directly from the cell body and do not extend along the ventral cord. In 35% (n = 100) of cg120 animals, axons abnormally exit the ventral cord at some distance from the cell body (Fig. 4 F; Table). Particular DA and DB axons exit the ventral nerve cord on the right or left side in a stereotypic pattern. In 15% (n = 100) of cg120 animals, one or more axons exit the ventral cord on the incorrect side of the animal.

cg120 animals also display defects in migrations of nonneuronal cells of the male tail and the hermaphrodite gonad. Formation of the male tail involves complex cell migrations and interactions (Sulston et al. 1980; Emmons 1992). 10% (n = 50) of cg120 males show fusions of sensory rays 1–2 or 7–8–9, and sensory ray 6 is often short and broad (Fig. 6 B). Despite these defects, most cg120 homozygous males can mate. In the hermaphrodite gonad, the most common defect (32%) is a premature dorsal turn of the anterior distal tip cell (Fig. 6 E). The posterior gonad shows similar defects with lower frequency (16%). Gonads are also often slightly shortened relative to wild type, suggesting that the distal tip cells do not migrate completely along the anterior–posterior axis.

As an independent means of assessing the effects of loss of cle-1 function, we performed RNA interference (RNAi) against the gene. RNAi has been shown to cause loss-of-function phenotypes for numerous C. elegans genes (Fire et al. 1998). We injected dsRNA generated from a region of cle-1 that is common to all three isoforms and analyzed the resulting phenotypes.

Significant levels of both embryonic and larval lethality were seen in offspring from hermaphrodites injected with cle-1-specific dsRNA (Fig. 7; Table). Embryonic arrest most commonly occurred during early morphogenesis, at approximately the 1.5-fold stage. Larval arrest occurred at the first larval stage, and arrested larvae were small and had misshapen pharynxes that failed to pump. Animals that survived to the adult stage had motorneuron positioning and axon guidance defects (Table), gonad migration, or male tail defects and were sterile (Table). These defects are similar to those observed in cg120 mutants. Injection of dsRNA into cg120 mutant animals resulted in higher levels of lethality and sterility (Table). The occurrence of similar phenotypes, but at higher penetrance in RNAi-treated versus cg120 mutant animals, suggests that RNAi results in greater cle-1 loss-of-function than the cg120 deletion.

The cg120 deletion removes the NC1 domain that consists of three subdomains, an association domain, a hinge region, and the ES domain. To further investigate the functions of these subdomains, we ectopically expressed the full NC1, the association domain, and hinge regions, or the ES domain in wild-type and cg120 backgrounds. Secreted CLE-1 domains were expressed in mechanosensory neurons using the mec-7 promoter, which has been previously used to analyze cell and axon guidance (Hamelin et al. 1993; Colavita and Culotti 1998). The mechanosensory neuron mispositioning caused by cg120 was rescued by ectopic NC1 expression, with the frequencies of specific cell position defects reduced 5–10-fold (Table). In contrast, expression of the ES domain or the association–hinge region did not rescue these migration defects. Other cg120 phenotypes, such as egg-laying and gonad migration defects, were not rescued by NC1 expression in mechanosensory neurons, possibly due to a limited range of action of the NC1 domain.

Ectopic expression of the ES domain in the wild-type background caused dominant cell position defects, similar to cg120 mutants (Table). The mechanosensory neurons were generally mispositioned along their normal trajectories but short of their normal final positions (Fig. 5 D; Table). Eight independent ectopic ES–expressing transgenic lines showed similar defects with variable penetrance, suggesting that the effect may be dosage sensitive. Ectopic NC1 or association–hinge domain expression in the wild-type background showed no effect on mechanosensory neuron positioning (Table), indicating that the complete NC1 domain and the ES domain are functionally distinct.

Ectopic ES also affected cells that did not express it. Expression of the ES but not the NC1 domain in mechanosensory neurons resulted in distal tip cell migration defects in 40% of hermaphrodite gonads and tail defects in 30% of males (n = 50) (Fig. 6 C). In both cases, the defects are similar to those seen in cg120 animals, again indicating that NC1 deletion and ectopic ES expression cause similar defects. The distal tip cells and cells of the male tail do not express the ectopic ES, indicating that it acts cell nonautonomously to affect their function.

A major functional difference between type XVIII collagen NC1 and ES domains is the ability of NC1 to oligomerize (Sasaki et al. 1998; Kuo et al. 2001, this issue). Recombinant CLE-1 NC1 domains expressed in human embryonic kidney 293 cells were found to form trimers (Fig. 8 A), whereas CLE-1 ES domains remained monomeric. Thus, the C. elegans domains behave similarly to the mammalian domains in regards to oligomerization, suggesting that the functional differences between ectopically expressed CLE-1 NC1 and ES domains may reflect differences in their ability to oligomerize via the association domain.

We have identified cle-1 as the homologue of vertebrate type XVIII collagen in the avascular nematode C. elegans. Although there are some 150 collagen genes in C. elegans and more than 20 in vertebrates, only types IV and XVIII are conserved between them (Kramer 1997; Hutter et al. 2000). We have demonstrated that deletion of the COOH-terminal NC1 domain from the C. elegans type XVIII collagen homologue cle-1 allows for assembly of the truncated protein in basement membranes and does not result in gross disruption of basement membrane structure. Loss of the NC1 domain causes migration defects of multiple cells and axons. Particular migration defects can be rescued by ectopic expression of the complete NC1 domain, but not by its ES subdomain. In fact, ectopic ES expression causes migration defects similar to those resulting from the NC1 domain deletion, suggesting that ES may interfere with NC1 domain function. Similar antagonistic roles of type XVIII collagen NC1 and ES domains on cell migration have been described for vertebrate cells in culture (Kuo et al. 2001, this issue). Conservation of type XVIII collagen across these distant phyla indicates that, like type IV collagen, it is a fundamental matrix component that has conserved functions.

The three isoforms of CLE-1 differ in their domain compositions and their temporal and spatial expression patterns. Notably, the CLE-1A and B isoforms are expressed only in neurons. CLE-1 is broadly distributed in C. elegans basement membranes but accumulates to highest levels in association with the nervous system. The elements of the C. elegans nervous system are associated with basal laminae (White et al. 1986), but, until recently, no matrix components had been localized to them. Antibody localizations of nidogen, NID-1 (Kang and Kramer 2000), and CLE-1 (this report) to neural elements indicates that they are components of these neural basement membranes. Mutations in nid-1 (Kim and Wadsworth 2000; Kang and Kramer 2000) and cle-1 disrupt normal neural patterning, indicating that these matrix molecules have important roles in development of the C. elegans nervous system.

The cg120 CLE-1 NC1 domain deletion results in multiple cell and axon migration defects, but very little (<5%) lethality. However, placing cg120 across from a larger deletion of the region or RNAi against cle-1 results in much greater levels of lethality (25–39%). The NC1 domain deletion is clearly only a partial loss-of-function mutation, and stronger loss-of-function appears to increase lethality. Although RNAi can generate apparent null phenotypes, some neuronally expressed genes have been shown to be refractory to RNAi (Tavernarakis et al. 2000). Since RNAi in the cg120 background results in more penetrant lethality than in the wild-type background, it is likely that cle-1 function is only partially inhibited by RNAi. Complete loss of cle-1 function may result in more highly penetrant or strict lethality. The embryonic lethal animals from cle-1 RNAi appear similar to mutants that disrupt epidermal (hypodermal) function (Williams-Masson et al. 1997; Costa et al. 1998), suggesting that CLE-1 may function in epidermal migration and/or adhesion.

In C. elegans, several lines of evidence indicate that cell migration and axon guidance require integration of a variety of signals (Wadsworth and Hedgecock 1996). The migration defects we have described in cg120 mutants are not fully penetrant, indicating that the CLE-1 NC1 domain is only one of the factors influencing these migrations. In fact, most mutations that affect cell and axon migrations in C. elegans are only partially penetrant (Wadsworth and Hedgecock 1996). Mutations in two genes—ina-1, an α-integrin (Baum and Garriga 1997), and mig-2, a Rac-like GTPase (Zipkin et al. 1997)—cause partially penetrant migration defects that are most similar to cg120. Interestingly, Kuo et al. 2001(this issue) demonstrate that the promigratory activity of the type XVIII collagen NC1 domain requires Rac and Cdc42 function. The relationship between cle-1, ina-1, and mig-2 is currently being investigated.

We have shown that the CLE-1 NC1/ES domain affects cell and axon migrations in C. elegans. The collagen XVIII ES domain has antiangiogenic activity and has been reported to inhibit migration of endothelial cells specifically (O'Reilly et al. 1997; Dhanabal et al. 1999). Nematodes lack endothelial cells, but we have demonstrated effects on neurons and other cells. When cultured on complex matrix such as Matrigel, the migrations of other mammalian cell types, including neurons, were recently shown to respond to collagen XVIII NC1/ES domains (Kuo et al. 2001, this issue). In complex natural environments, the migrations of multiple cell types can be affected by mammalian and nematode collagen XVIII NC1/ES domains, indicating a conservation of function from nematodes to vertebrates.

In cle-1(cg120) NC1 domain deletion mutants, the migrations of multiple cells are incomplete, suggesting that the C. elegans NC1 domain has promigratory activity. Ectopic cle-1 ES domain expression in the wild-type background causes similar cell migration defects, indicating that ES has antimigratory activity (Fig. 8 B). Ectopic ES does not exacerbate the migration defects caused by the NC1 deletion, suggesting that ES negatively regulates the NC1 promigratory activity. These findings are consistent with those of Kuo et al. 2001(this issue), showing that trimeric collagen XVIII NC1 or oligomerized ES domains have promigratory activity for several cell types and that monomeric ES inhibits these activities. The C. elegans NC1 and ES domains are, respectively, trimeric and monomeric in vitro, suggesting that differences in oligomerization may explain their different in vivo effects. Multivalency of ES domains could be important to cluster receptors or the matrix molecules to which they bind, as a means of activation (Giancotti and Ruoslahti 1999).

The type XVIII collagen NC1/ES domain has a conserved role in regulating cell motility from C. elegans to vertebrates. In C. elegans, it most notably affects neurogenesis, presumably because this process involves numerous long range cell and axon migrations. In vertebrates, the best documented effects are on angiogenesis, which also involves extensive cell migrations. Related mechanisms may control cell migrations in neurogenesis and angiogenesis, and indeed molecules such as neuropilin (Soker et al. 1998) have been shown to affect migrations of both neural and endothelial cells. The role of type XVIII collagen in C. elegans neurogenesis suggests that it may also affect vertebrate neuronal cells.

We thank Y. Kohara (National Institute for Genetics, Mishima, Japan) for cDNA clones, A. Coulson (Sander Centre, London, UK) for cosmids, and A. Fire (Carnegie Institution of Washington, Baltimore, MD) for expression vectors. We thank J. Culotti (Samuel Lunenfeld Research Institute, Toronto, Ontario, Canada), C. Kenyon (University of California, San Francisco, CA), and D. Pilgrim (University of Alberta, Edmonton, Alberta, Canada) for GFP marker strains. We are indebted to K. Javaherian and S. Park (Childrens Hospital, Harvard Medical School, Boston, MA) for purification and analysis of recombinant CLE-1 fragments. 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 grants from the Health Sciences Council of the Academy of Finland (T. Pihlajaniemi) and the National Institutes of Health (HD27211, J.M. Kramer).