C. elegans mig-6 encodes papilin isoforms that affect distinct aspects of DTC migration, and interacts genetically with mig-17 and collagen IV

Little is known about the dynamic extracellular matrix (ECM) processes that are evoked to ensure the orchestrated cell movements and proliferation required to produce organs of the proper shape, size and organization. However, during the past decade, several proteinases, including matrix metalloproteinases (MMPs) and ADAM family metalloproteinases, have been found to be necessary for normal morphogenesis and for cancer metastasis by remodeling the ECM (Chang and Werb,2001; McFarlane,2003).

We are using genetics to understand the molecular mechanisms that regulate a readily monitored cell migration in vivo - the migration of the two C. elegans hermaphrodite distal tip cells (DTCs)(Nishiwaki, 1999; Su et al., 2000). The sequential three-phase migration pattern of these two cells determines the final shape of each of the two U-shaped hermaphrodite gonad arms(Fig. 1A). The DTCs are born post-embryonically near one another in the ventral mid-body during the first larval stage. Their phase 1 migration comprises a longitudinal migration in opposite directions away from the mid-body using the ventral body muscle basement membrane as a substratum for migration. During phase 2, the DTCs turn and migrate across the lateral epidermal basement membrane towards the dorsal body wall muscles. During phase 3, the DTCs reorient again and migrate centripetally on the dorsal body muscles - back towards the mid-body - where they normally stop (Fig. 1A).

Among the first genes known to affect DTC migration is unc-6,which encodes a secreted repulsive cue (UNC-6/netrin) for axon and DTC migrations that occur along the dorsoventral (DV) axis, as well as unc-5 and unc-40/DCC, which encode receptors that mediate the repulsive effects of UNC-6/netrin (Chan et al., 1996; Hedgecock et al., 1987; Leung-Hagesteijn et al., 1992). In mutants of these genes, the DTCs frequently fail to initiate the phase 2 migration, but if they do initiate it, this migration appears normal. This raises the question of what molecular mechanisms are required to guide the phase 2 DTC migration after it is initiated by UNC-6,UNC-5 and UNC-40 activities?

We have found two major phenotypic classes of mig-6 mutant alleles. Class-l mutations [aka mig-6(l)] hinder the rate of DTC movement during all phases of its migration, whereas class-smutations [aka mig-6(s)] alter DTC guidance during the second(ventral to dorsal) phase of its migration. The nature of mutational lesions,RNAi, and rescue by endogenous and cell specific expression all show the mig-6 class-s and class-l mutations affect the function of the two alternatively spliced mRNAs, mig-6S and mig-6L, which encode MIG-6S and MIG-6L proteins, respectively.

We cloned mig-6 and found that it is the previously reported c-ppn gene (registered as ppn-1 in Wormbase) of C. elegans (Kramerova et al.,2000), which is highly related to genes encoding the secreted multi-component ECM proteins Drosophila papilin and Manduca sexta lacunin (Kramerova et al.,2000; Nardi et al.,1999). Previous histological studies have shown that papilin and lacunin are constituents of basement membranes and suggest that they have roles in the morphogenesis of epithelial tissues. Single papilin orthologues are found in C. elegans and Drosophila genomes. In this report, we use the name mig-6, the first published name for this gene(Hedgecock et al., 1987) and now the official gene name in Wormbase.

DTC migration defects and consequent gonad phenotypes of gon-1 and mig-17 mutants, which both encode secreted ADAMTS metalloproteinases(Blelloch and Kimble, 1999; Nishiwaki et al., 2000), are similar or identical to mig-6(l) and mig-6(s) alleles,respectively. Intergenic non-complementation suggests that mig-6S and mig-17 act in the same mechanistic pathway to ensure normal phase 2 DTC guidance: consistent with immunostaining results that suggest MIG-6S regulates MIG-17 basement membrane localization. Additional genetic evidence indicates that MIG-6S is antagonistic to non-fibrillar network collagen IV in guiding phase 2 DTC migrations. These data suggest that MIG-6 regulates distinct aspects of DTC migration by dynamically regulating the abilities of specific proteinases to remodel different basement membranes encountered during sequential phases of DTC movements.

mig-6 class-l alleles: oz alleles were provided by Dr T. Schedl, Washington University; mig-6(e1931) was provided by the Caenorhabditis Genetics Center.

mig-6 class-s alleles: mig-6(ev701) was obtained from an EMS-induced (Brenner,1974) screen for dominant DTC mutants. Other mig-6(s)alleles were obtained in F2 screens. A heterozygote of mig-6(ev788)was isolated by sib-selection from a formaldehyde-induced deletion library(Johnsen and Baillie, 1988). ev788 deletes nucleotides 490 in exon 4 to 1543 in exon 6(gaatctggaaacttctacta.......tgagcaagagaagttcgaca). Class-s and null alleles were out-crossed four times and dpy-11(e224) mig-6 doubles were made and balanced by translocation eT1 (III, V) or nT1[qIs51] (IV, V) (Edgley et al.,2006). dpy-11 mig-6(ev788)/nT1[qIs51] segregates non-Dpy heterozygous worms with DTC defects and arrested embryos. All other strains were provided by the Caenorhabditis Genetics Center.

mig-6 was mapped between two deficiencies, sDf35 and sDf20, and 30 cosmid clones (kind gifts from the Sanger Centre) in this region were tested for their ability to rescue the mig-6(s) and mig-6(l)mutants by injecting 10 μg/ml of each DNA with 50 μg/ml of sur-5::gfp co-transformation marker.

In rescue experiments designed to identify the mig-6 gene,class-s mutant DTC defects were scored as clear patches in the body -typically caused by altered DTC migration patterns, which rarely occur in wild-type animals (Hedgecock et al.,1987). In all other rescue experiments, three classes of DTC defects (see Results) were scored using Nomarski optics to examine the shape of gonad arms.

To score rescue of lethality and sterility, lines heterozygous for the transgene and for dpy-11(e224) mig-6 were established[dpy-11(e224) was used in initial rescue experiments and later found to further reduce viability]. Homozygous dpy-11 mig-6(s) segregants were scored as rescued for lethality (designated `+') if broods of transgenic animals were greater than 100 (broods of non-transgenic siblings were always less than 100). Homozygous dpy-11 mig-6(l) segregants are like mig-6(l) homozygotes in that they normally have no progeny(n=500 individuals); therefore, when a strain segregated dpy-11 mig-6(l) homozygotes that in the presence of the transgene produced any progeny at all (typically 20% of transgenic animals did this), the transgene was scored as rescuing (designated `+').

Animals doubly heterozygous for emb-9 and mig-6(ev701)were made by crossing dpy-11 mig-6(ev701)/nT1[qIs51](VI, V) to unc-36(e251)emb-9(g23cg46)/qC1 dpy-19 (e1259) glp-1(q339) III. F1 non-GFP-expressing male progeny [unc-36(e251) emb-9(g23cg46)/++; dpy-11 mig-6(ev701)/++] were crossed into dpy-18(e364)/eT1 III; unc-46(e177)sDf20/eT1[let-500(s2165)]V (Edgley et al., 2006). F1 Unc L4 larvae {unc-36(e251) emb-9(g23cg46)/eT1 III; dpy-11 mig-6 (ev701)/eT1 [let-500(e2165)]V} were cloned and the genotype was confirmed by outcrossing to N2 males and finding that F1s segregate Dpy and non-Dpy non-Unc animals with phase 2 DTC migration defects. Control strain unc-36(e251)/eT1 III; dpy-11 mig-6(ev701)/eT1[let-500]V was made by the same protocol. dpy-5 unc-13/++ I; dpy-11 mig-6(e1931)/++ V males were crossed into smg-1(r861) I to produce dpy-5 unc-13/smg-1 I; dpy-11 mig-6(e1931)/++ V. smg-1self-progeny were homozygosed and their Dpy progeny were checked for sterility and DTC defects. Ten out of 200 Dpy progeny were non-sterile and confirmed dpy-11 mig-6(e1931)/dpy-11 + recombinants.

Worms were observed and photographed using a Leica DMRA2 microscope equipped with Hamamatsu ORCA-ER digital camera. Lengths along the core of gonadal arms from the tip of the anterior gonad to the tip of posterior gonad were measured using OpenLab software (Improvision).

To make a full-length cDNA of mig-6S, the 5′ upstream region was amplified by PCR from a wild-type cDNA pool using an SL1 primer and a reverse primer in exon 5, then spliced to the partial cDNA yk5c8 (from the National Institute of Genetics, Japan). For mig-6L, the above fragment was then combined with partial cDNAs yk5c8 and yk257g8. A mig-6S genomic/cDNA minigene construct (pZH125) was made by splicing yk5c8 (exons 5 to 11a) and a KpnI-BsrGI genomic fragment spanning a 5 kb region 5′ upstream to exon 5. The mig-6L genomic/cDNA minigene (pZH117) was made by splicing yk3e8 (exons 8 to 11),yk257g8 (exon 11 to exon 18) and a KpnI-BsaBI genomic fragment spanning a 5 kb region 5′ upstream to exon 8. To produce heterologous promoter driven minigenes, 3.0 kb of the lag-2 promoter(for pZH116 and pZH118) from pJK590 (gift from Dr J. Kimble) and 2.2 kb of the myo-3 promoter (for pZH135 and pZH145) from pPD96.52 (gift from Dr A. Fire) were substituted for the endogenous 5′ regulatory regions of mig-6S and mig-6L genomic/cDNA minigenes.

A 210 bp genomic PCR fragment 5′ to and including the start codon was linked to the 5′ adjacent 5 kb KpnI-MunI fragment of mig-6(+).This was sub-cloned into the KpnI site of gfp expression vector pPD95.79 (from Dr A. Fire). Protocols for a large scale in situ hybridization of C. elegans larvae can be found at http://nematode.lab.nig.ac.jp/method/insitu_larvae.html.

Standard procedures were used. Mutational lesions were determined by direct sequencing of PCR amplified cDNA fragments.

cDNA fragments were PCR amplified using T7-tagged gene-specific primers and the M13-20 universal primer. Primer sequences are available on request. dsRNA preparations (0.2-0.5 mg/ml) were injected into intestine or pseudocoelom of wild-type adults. The progeny of the injected animals were examined for lethality and DTC defects.

For RNAi of (1) mig-6S, (2) mig-6L or (3) both, the following fragments were subcloned into the pPD129.36 RNAi vector (kind gift of Dr A. Fire): (1) an AvaI fragment of yk3e8 corresponding to the 3′ UTR of mig-6S (within intron 11 of mig-6L)(Fire et al., 1998), (2) a HindIII-SpeI fragment of yk257g8 (exons 14-15) and (3) a BsaBI-HindIII fragment of yk3e8 (exon 8). For RNAi by feeding, these were transfected into HT115 (DE3) bacteria(Timmons et al., 2001). emb-9 RNAi utilized a clone from the RNAi bacterial library (gift from Dr J. Ahringer).

Temperature-sensitive emb-9(g34ts) collagen mutant animals were cultured at 16°C until the late L2 stage, then cultured at 25°C for 8 hours. After an additional 20 hours of culture at 16°C, L4 larvae were examined by Nomarski optics for gonad defects.

A mig-17 promoter-driven translational fusion between mig-17 and gfp(Nishiwaki et al., 2000) and an unc-119(+) co-transformation marker (from Dr D. Pilgrim) were introduced into unc-119(e2498) by microparticle bombardment(Praitis et al., 2001). By cloning non-Unc dauer worms, several mig-17::gfp integrated transgenic lines were established and passed into appropriate strains using standard genetic techniques.

For whole-mount immunostaining, worms were fixed in Bouin's fixative(Duerr, 2006). Frozen sections of synchronized early L4 worms were prepared according to Kubota et al.(Kubota et al., 2006). Embryos were permeabilized with alkaline hypochlorite and fixed with 3%paraformaldehyde (Miller and Shakes,1995). For blocking and antibody dilution, PBS containing 0.5%Triton X-100 and 1% bovine serum albumin (Sigma) was used.

Rabbit polyclonal antibodies were raised against MIG-6 peptide SGQKETGNWGPWVPE(C) (amino acids 70-85) and affinity purified. Goat polyclonal anti-SAD-1 (Hung et al., 2007)and rabbit anti-EMB-9/collagen IV antibodies were kind gifts from Dr M. Zhen(Mt. Sinai Hospital, Toronto) and Dr J. M. Kramer (Northwestern U.),respectively. Rabbit anti-GFP polyclonal antibody (A11122; used at 1:200 dilution) and Alexa Fluor 568 phalloidin (used at 20 U/ml) were purchased from Invitrogen. Secondary antibodies (Molecular Probes) were goat anti-rabbit IgG-Alexa fluor 488 (used at 1:2000 dilution), goat anti-mouse IgG-Alexa Fluor 546 (1:1000) or donkey anti-goat IgG-Alexa Fluor 594 (1:400).

In class-l mig-6 mutants (e1931, oz113, oz159 and oz90) originally described by Hedgecock et al.(Hedgecock et al., 1987), DTC migration is extremely slow relative to the wild type, resulting in foreshortened gonad arms (Fig. 1B,C). The rate of DTC migration found in the mig-6(l)mutants is about one-third of the wild-type rate (14 versus 38 micrometers per hour) (Fig. 1C). DTCs in mig-6(l) mutants nevertheless initiate phase 2 migrations toward the dorsal muscles. However, the ability of the DTCs to reach the dorsal muscle with normal timing is impeded in these mutants. The distal region of the gonad does not grow in diameter and looks like a small appendage to the bulbous proximal arm (Fig. 1B). mig-6(l) mutant animals are also completely sterile. These phenotypes are fully penetrant and recessive for all four mig-6(l) alleles.

We also identified several semi-dominant, highly penetrant class-salleles of mig-6 in which the DTCs migrate at an approximately normal rate but have specific defects in phase 2 migrations. In mig-6(s)mutants, the DTCs appear to detach from ventral muscle bands at the onset of their phase 2 migration, but their subsequent migration usually follows one of three patterns in addition to the wild-type pattern(Table 1). For the phase 2D pattern, DTCs migrate diagonally until they reach the dorsal body muscles,which they then follow back to the mid-body region(Fig. 2A). For the phase 2V pattern, DTCs start the phase 2 ventral to dorsal migration onto the lateral epidermis normally, but quickly return to the ventral body muscles and complete phase 3 by migrating centripetally along these muscles back towards the mid-body (Fig. 2B). This causes gonad arm defects roughly reminiscent of those observed in mutants of unc-5, unc-6 and unc-40(Hedgecock et al., 1990). Finally, for the phase 2M category, the DTCs, after a diagonal phase 2 migration as in phase 2D defects, migrate towards mid-body but meander along the DV axis of the lateral epidermis as they do(Fig. 2C). We suggest that this phenotype results from loss of precise DTC guidance along the DV axis of the lateral epidermis during the phase 2 migration.

mig-6(s) mutations are semi-dominant and, as heterozygotes, show the same spectrum of DTC migration patterns as homozygotes. In class-s heterozygotes, phase 2V and 2D patterns are most prevalent,whereas meandering gonads (phase 2M) are less frequent (especially anterior arms) than in homozygotes (Table 1). Thus, we consider the meandering gonad phenotype to be caused by a greater loss of mig-6 function. This is supported by mig-6S-specifc RNAi experiments (see below).

We mapped mig-6 to a short region of LGV and rescued the DTC phenotypes with two overlapping cosmids (C32A1 and C37C3)then with gene C37C3.6 (Fig. 3A). C37C3.6 encodes at least two mRNAs, a short(C37C3.6a) and a long (C37C3.6b) form, herein designated mig-6S and mig-6L, respectively(Fig. 3B). The predicted 3′ ends of these transcripts are available as EST cDNA clones (Y.K.,National Institute of Genetics, Mishima, Japan).

The two predicted transcripts share the same exon structure and open reading frame from exon 1 into exon 11, but differ in their 3′extensions. mig-6S includes the whole of exon 11, which encodes a stop codon and about 320 base pairs of 3′ untranslated mRNA, whereas mig-6L includes only part of exon 11 as this exon is spliced to exon 12 using an alternative donor site.

Two transcripts of approximately the predicted sizes are detected by Northern blot analysis (Fig. 3C). RNAase protection experiments show that both transcripts are abundant in embryos and less abundant during all larval stages (see Fig. S1 in the supplementary material).

As shown in Fig. 3D, the predicted MIG-6 proteins bear a common signal sequence for secretion at their N terminus, but no predicted transmembrane domain, suggesting they are secreted. The signal peptide is followed by thrombospondin type 1 (TSP1)repeats (also known as TSRs), the first two of which are separated by a cysteine rich spacer region (CR1), then a series of cysteine-rich lagrin repeats (CR2) and six predicted Kunitz-type serine proteinase inhibitor (KU)domains. The C terminus of MIG-6L contains five additional Kunitz domains and a single predicted immunoglobulin (Ig) C2-type domain(Fig. 3D). Secreted basement membrane proteins with similar domain organization as MIG-6 are a lacunin from the moth Manduca sexta (Nardi et al., 1999), and papilins from Drosophila(Kramerova et al., 2000) and the nematode Haemonchus contortus(Skuce et al., 2001). Genes encoding papilin related proteins with similar overall domain organization are also found in human and mouse genomes. These proteins differ in the number of repeats of a given type of domain, having, for example, only a single KU domain in the mammalian version.

In Drosophila, alternative papilin splice forms are known that differ only in the number of KU and IgC2 domains they encode(Kramerova et al., 2003). Whether mammalian papilin-like genes encode multiple mRNAs and proteins is unknown. The N-terminal TSP1 repeats and the intervening cysteine rich CR1 domain, together termed the `papilin cassette', have highest evolutionary conservation among related proteins (38% identity and 54% similarity between MIG-6 and Drosophila papilin, and 36% identity and 50% similarity between MIG-6 and human or mouse papilin)(Fig. 3D). Comparisons among these related molecules have been described previously(Kramerova et al., 2000; Nardi et al., 1999).

To explore stage-specific requirements for the two mig-6transcripts, we compared RNAi of specific transcripts by injection (affecting embryonic and post-embryonic expression in the subsequent generation), and by feeding starting as L1 larvae (affecting post-embryonic expression in the same generation) (Table 2). RNAi of mig-6L by either method phenocopies the mig-6(l) mutant DTC defects (Fig. 1B). This suggests that mig-6L alone normally encodes class-lfunctions that are required post-embryonically for a wild-type rate of DTC migration.

Although ∼60% of self-fertilization progeny of mig-6(s)homozygotes survive to adulthood, RNAi of mig-6S by injection causes severe embryonic and early larval lethality allowing only about 1-2% of animals to survive to adulthood. Thus, RNAi is more potent at reducing embryonic MIG-6S activity than are any of the mig-6(s) mutations,probably because they all cause amino acid substitutions that primarily affect postembryonic DTC migrations (see below). This mig-6S RNAi effect is also indistinguishable from the effect of injecting dsRNA that targets both transcripts (Table 2),suggesting that the null phenotype is lethal. To further examine this possibility, we isolated a predicted null allele(Fig. 3D) by formaldehyde-induced direct deletion (Materials and methods). mig-6(ev788) is deleted for a genomic segment ranging from nucleotide 490 in exon 4 to 1543 in exon 6. All possible splicing of the mutant transcript is predicted to produce an early out-of-frame protein. As predicted from the RNAi experiments, all ev788 homozygotes are embryonic or early larval lethals.

RNAi of mig-6S by feeding, which should affect post-embryonic functions of MIG-6S in the same generation, phenocopies the phase 2D and 2V DTC defects predominant in class-s heterozygotes but at a lower penetrance and with rare meandering DTC defects(Table 2). The reduced penetrance and severity (i.e. rarity of phase 2M defects) may occur because of incomplete disruption of the mig-6S message by this method. In fact,predicted enhancement of RNAi efficacy by feeding in an rrf-3 mutant background (Simmer et al.,2002) enhances the frequency of meandering DTCs(Table 2). These results suggest that mig-6S alone encodes class-s functions required post-embryonically for properly executing the second phase DTC migration and verify that meandering DTC defects reflect a greater loss of mig-6S protein.

Because ∼40% of the fertilized eggs laid by homozygous class-shermaphrodites die in embryogenesis and as L1 larvae (see Table S1 in the supplementary material), it was surprising to find that 132 of 526 progeny(25%) from class-s mig-6(ev701) heterozygotes were viable mutant homozygotes that became adults, the majority of which had meandering DTC defects. This suggests that development of these mutants is largely normal but is defective in a way that specifically affects post-embryonic DTC migrations.

We determined molecular lesions for eight mig-6 alleles. Three class-l mutations introduce a premature stop codon predicted to delete the MIG-6L-specific KU repeats and the C-terminal IgC2 motif(Fig. 3D). Combined with the finding that RNAi silencing of mig-6L produces a class-lmutant phenotype and that a clone including only mig-6S (pZH96) is not sufficient to rescue mig-6(e1931) class-l mutants(Fig. 3B), we conclude that MIG-6L is solely responsible for the DTC migration functions that are defective in mig-6(l) mutants.

In C. elegans early stop codons, like those observed for the class-l alleles, may trigger nonsense-mediated decay of mRNAs. However, the smg-1(r861) mutation, which is known to suppress this decay (Blelloch and Kimble,1999), did not suppress the mig-6(e1931) class-lDTC migration phenotype, as smg-1(r861); dpy-11 mig-6(e1931) was 100%sterile (n=190) and had class-l mutant DTC defects. The tested allele (e1931) is a nonsense mutation that eliminates the last three Kunitz domains and the IgC2 domain(Fig. 3D), suggesting that at least one of these Kunitz domains and/or IgC2 is essential for MIG-6L function.

In contrast to class-l alleles, all class-s alleles are missense mutations that affect a TSP-1 repeat or a CR1 domain in the papilin cassette (Kramerova et al.,2000), or a cysteine-rich lagrin repeat(Fig. 3). Previous studies found that the `papilin cassette' and the CR1 region within the cassette are required for tight binding to ECM (Kuno and Matsushima, 1998).

The DTC migration defects in most of the class-s mutant heterozygotes are more penetrant than in heterozygotes of the putative null allele ev788 but exhibit a similar spectrum of phenotypes(Table 1). If ev788 is a true null (see below), the higher penetrance of defects in class-s heterozygotes (e.g. ev701/+) must result from a greater than 50%(half dose) loss of mig-6(+) activity and indicates that these mutants are anti-morphs for DTC migration (i.e. they interfere with wild-type activity).

In principle, it is possible that the differential expression of mig-6S and mig-6L, as opposed to different inherent abilities of the proteins they encode, are responsible for the differences in their function. In situ hybridization with a mig-6L-specific probe shows that mig-6L is restricted to the DTCs(Fig. 4A) and what are likely to be coelomocytes (not shown, but see Fig. 4E); however, a probe that detects both transcripts revealed additional expression in body wall muscles(Fig. 4B). Transcriptional reporters for mig-6 revealed expression in DTCs, body wall muscles,CAN neurons, head mesodermal cells, GLR cells and coelomocytes(Fig. 4C-E). This expression pattern is consistent with the in situ hybridization data; however, it does not distinguish whether mig-6S is normally expressed in the DTCs, as we know to be the case for mig-6L.

MIG-6S localizes to embryonic muscles(Fig. 4F). In double immunostaining experiments with anti-myosin and anti-MIG-6 antisera, MIG-6 appears to be localized near the muscle surface (not shown). In the mig-6(ev788) putative null allele there is no expression by muscle(Fig. 4G) even though the epitope is near the N terminus and therefore little else is predicted to be encoded by this allele (Fig. 3D) (and Materials and methods).

MIG-6 in larvae also localizes to intestine, pharynx and gonad basement membranes (Fig. 4H-J). There is no obvious abnormal localization (either intracellular or extracellular) of MIG-6 proteins in mig-6(s) mutants(Fig. 4K,L); however,functionally relevant concentration differences could have been missed by this technique.

A genomic DNA construct (pZH95), including regulatory sequence and sequence encoding both mig-6 transcripts rescued the lethality and DTC migration defects of the null allele to near wild-type levels (see Table S2 in the supplementary material). A genomic/cDNA minigene construct (pZH125)encoding MIG-6S could not rescue class-l mutant sterility (not shown)or DTC defects (Table 3), but could partially rescue viability and DTC defects of mig-6(s)homozygotes (slightly) and mig-6(s) heterozygotes (substantially)(Table 3). A genomic/cDNA minigene encoding MIG-6L (pZH117) partially rescued the DTC defects of the class-l homozygotes, but not homozygous class-s DTCs or lethality (Table 3); however,substantial rescue of class-s heterozygote DTC defects by mig-6L did occur (Table 3). These data indicate that when expressed under the control of nearly identical endogenous regulatory sequences, MIG-6S cannot substitute for MIG-6L but MIG-6L can partially substitute for MIG-6S in guiding phase 2 DTC migrations.

To explore the significance of mig-6 expression in DTCs and in muscles, we engineered mig-6S or mig-6L minigenes under the control of either of two heterologous 5′ regulatory regions. The lag-2 promoter (lag-2p) drives expression primarily in the hermaphrodite DTCs (Blelloch and Kimble,1999; Nishiwaki et al.,2000), whereas the myo-3 promoter (myo-3p)drives expression in all of the body-wall muscles(Okkema et al., 1993). For each transgene, rescuing activity was examined in class-s and class-l mutants. We found that only pZH118[lag-2p::mig-6L] rescues the mig-6(l) mutant DTC migration defects to roughly the same extent as the mig-6L minigene driven by endogenous promoter (pZH117 in Table 3). These results indicate that MIG-6L functions cell autonomously in the DTCs to regulate the rate of their migration and are consistent with the in situ hybridizations,which show that mig-6L expression in larvae is limited largely to DTCs (Fig. 4A). Also as predicted, MIG-6S cannot substitute for MIG-6L when expressed in the DTCs(pZH116 in Table 3).

Although mig-6S is normally expressed in body wall muscles,neither the myo-3p::mig-6S nor the myo-3p::mig-6L hybrid minigenes rescued the homozygous class-s or class-l DTC phenotypes. Transgenic animals that carry both myo-3p::mig-6S and lag-2p::mig-6S also failed to rescue the homozygotes (data not shown).

Mutants of mig-17 display DTC defects that mimic DTC defects of mig-6(s) homozygotes, but at lower penetrance. Although the mig-17 mutant defects are recessive, animals that are doubly heterozygous for mig-6 class-s alleles ev701 or k177, and the mig-17 putative null allele, k174, show higher penetrance (Fig. 5A) and expressivity (higher frequency of type 2M DTC migrations) of defects than respective mig-6(s) heterozygotes. This intergenic non-complementation suggests that MIG-6S functions in the same pathway (but see Yook et al., 2001) as MIG-17, a known ADAMTS metalloproteinase that is required (like MIG-6S) to guide phase 2 DTC migrations.

As shown previously using anti-GFP antibodies(Nishiwaki et al., 2000),MIG-17::GFP is distributed evenly with moderate intensity along the gonad basement membrane in wild-type animals(Fig. 5B,C). We found that in many mig-6(ev700) class-s mutant animals, MIG-17::GFP is distributed unevenly along the surface of the gonad, with high expression of the protein concentrated at variable places along the gonad arms(Fig. 5D). In some but not all cross-sections of mig-6(s) mutant animals, regions of abnormally high interspersed with regions of abnormally low expression are observed in spaces between gonad and body muscle bands or hypodermis(Fig. 5E). In other sections,MIG-17::GFP appears within the gonad arm (not shown). Western blots (see Fig. S3 in the supplementary material) show that the amount of MIG-17 protein is not dramatically changed in mig-6(ev700), suggesting that MIG-6S does not significantly affect the overall level of MIG-17, but does prevent over-accumulation in variable regions of the gonad arms. Whether this is a direct or an indirect effect of mutant MIG-6S proteins on MIG-17 is unknown.

Previous anecdotal reports of the effects of collagen mutations on DTC migration (Cassada et al.,1981) prompted us to examine collagens as possible targets of MIG-6 activity. Collagen type IV is one of the major components of specific basement membranes in C. elegans and other animals. Two EMB-9/collagen IV alpha 1 chains and one LET-2/collagen IV alpha 2 chain normally form a single heterotrimeric collagen. Glycine substitution mutants in Gly-X-Y repeats of the EMB-9 or LET-2 exhibit temperature-sensitive embryonic lethality, caused by reduced secretion and assembly of collagen type IV into basement membranes (Gupta et al.,1997). We exposed the emb-9(g34ts) mutants to the non-permissive temperature for 8 hours of L2 to L3 larval development at the time when the DTCs initiate their phase 1 migration. emb-9(g34ts)hermaphrodite animals so treated have a normal morphology, but their gonads phenocopy the mig-6(l) mutant gonad morphology, suggesting a possible functional/regulatory connection between MIG-6L and EMB-9(Fig. 6A,B).

Given the above data, we attempted to examine the relationship of MIG-6S with collagen IV function. Suppression of the phase 2 DTC defect was partial when RNAi was used to reduce EMB-9/collagen IV in mig-6(ev701)heterozygotes, but effectively complete when the wild-type dose of emb-9 was halved in heterozygotes for the putative null(Gupta et al., 1997) allele emb-9(g23cg46) (Fig. 6C). A lesser but substantial suppression was also caused by let-2 temperature-sensitive mutations(Fig. 6C). Thus, a reduction in collagen IV protein (by RNAi or a heterozygous null) or a reduction in collagen IV folding and secretion (caused by point mutation) can suppress mig-6 class-s DTC defects. These results indicate that collagen IV acts antagonistically to MIG-6S in an extracellular mechanism that affects phase 2 DTC migration. We did not observe any difference in EMB-9 distribution between wild type and mig-6 or mig-17 mutants using anti-EMB-9 antibodies for immunostaining; however, functionally relevant concentration differences could have been missed by this technique.

Papilin is a basement membrane protein that is essential for embryonic development, as shown by RNAi knockdown experiments in Drosophila and C. elegans (Kramerova et al.,2000). In the present study, we identified two different classes of MIG-6/papilin proteins and analyzed their functions by class-specific mutations and RNAi. Our results indicate that both of the MIG-6/papilin isoforms, MIG-6S and MIG-6L, function specifically to regulate DTC migration and MIG-6S also affects embryogenesis. There are several reasons to believe that the DTC and embryonic functions of the MIG-6 proteins are separate. First, the class-specific RNAi by feeding, which affects MIG-6 post-embryonic function in the same generation, can produce either class-s or class-l mutant DTC phenotypes. Second, all DTC phenotypes can be rescued to some degree by post-embryonic expression of one or the other of the two isoforms. Finally, the DTC function of MIG-6S proteins is genetically separable from the observed function of MIG-6S in embryogenesis as class-s mutant heterozygotes have normal sized broods with one-quarter mutant homozygous progeny that manifest DTC defects. Taken together, these results indicate that mutations that affect either MIG-6 isoform can cause specific defects in post-embryonically derived basement membrane structures required for DTC migration without affecting basement membrane functions required for embryogenesis.

We have found that the two isoforms of MIG-6/papilin are made by different cell types in C. elegans and normally have distinctly different functions in DTC migration. Class-l mutations, which eliminate the three C-terminal KU and the single IgC2 C-terminal domains (predicted to only affect MIG-6L), appear to specifically reduce the rate of DTC migration to about one-third that of the wild type. Other more subtle gonad defects are likely in class-l mutants as they are sterile and their distal gonad arms have a greatly reduced cross-sectional diameter compared with the proximal arm.

There are several reasons to believe that mig-6(l) mutations primarily affect the gonad basement membrane. Most importantly, MIG-6L is expressed by DTCs as they migrate and can act cell autonomously in larvae to promote DTC migration even though it is predicted to be a secreted protein. It is possible that the MIG-6L specific C-terminal region may include a retention signal or a motif (e.g. one or more KU domains and/or the IgC2 domain) that help localize it to the gonad basement membrane.

Because emb-9/collagen IV alpha 1 have been suggested to affect DTC migration (Cassada et al.,1981), we further characterized the DTC migration defects caused by inactivating emb-9 during DTC migration. The phenotypic similarity of post-embryonically induced DTC defects in emb-9 and in mig-6(l) mutants is striking. As both MIG-6L and EMB-9 are produced by the DTC, it is tempting to speculate that MIG-6L promotes post-embryonic collagen IV function, and that it could do so by downregulating the activity of proteinases that degrade collagen IV as is known for ADAM proteinases in other animals (Roy et al.,2004).

Because the phenotypes of gon-1 mutants, mig-6(l) mutants and post-embryonically induced DTC defects in emb-9 mutants are all highly penetrant, we could not test genetic interactions between them. Furthermore, gon-1/+; mig-6(e1931)/+ double heterozygotes are normal;therefore, additional studies will be required to elucidate the molecular mechanisms of cell surface remodelling required for a normal rate of DTC migration.

The class-s point mutations affect both long and short isoform amino acid sequences. However, RNAi experiments indicate that it is alteration of the short isoform that causes a defect in phase 2 DTC migration. Unlike MIG-6L, expression of MIG-6S in the DTCs did not significantly rescue class-s homozygote DTC defects [possibly owing to adverse effects of anti-morphic mig-6(s) alleles], but did significantly rescue heterozygotes. This suggests that MIG-6S, like MIG-17(Nishiwaki et al., 2000), may affect DTC migration cell non-autonomously; however, the tissue of origin for the most prominent effect of MIG-6S on DTC migration is still unknown.

Genetic interactions between mig-6(s) and mig-17/metalloproteinase mutants suggest that they act in the same mechanistic pathway used to guide phase 2 DTC migrations. MIG-17 is an ADAMTS metalloproteinase secreted as a pro-form from body wall muscle. As reported previously (Ihara and Nishiwaki,2007), MIG-17::GFP localizes to the entire gonad and intestinal surface after the initiation of the phase 2 DTC migration. We found that in mig-6(s) mutant animals MIG-17::GFP did not consistently show uniform distribution along gonad and intestine surfaces but instead it often showed discontinuous overaccumulations along these organs (see Fig. 5 legend). These results suggest that mutant MIG-6S does not efficiently retain MIG-17 in the gonad basement membrane. The lack of MIG-17 in this basement membrane could then cause phase 2 DTC migration defects similar to mig-17 mutants(Kubota et al., 2006; Nishiwaki et al., 2004). The abnormal distribution of MIG-17 in many mig-6(s) mutant animals suggests that MIG-6S plays a role in regulating MIG-17 basement membrane localization; however, whether this involves a physical interaction (direct or indirect) between these proteins or results from an unrelated abnormality in DTC function caused by mig-6(s) mutations is unknown.

MIG-6S and MIG-17 could modify interactions between the DTC and the lateral epidermal basement membrane, which is normally used as a substratum to support and guide phase 2 DTC migrations. Consistent with a functional or regulatory relationship between MIG-6 proteins and basement membrane proteins is the finding that emb-9 RNAi, as well as emb-9/collagen IV alpha 1 and let-2/collagen IV alpha 2 mutations, clearly suppresses mig-6(s) mutant DTC defects. Thus, genetic reduction of mutant or wild-type forms of collagen IV are both able to suppress mig-6(s)defects, suggesting that losing mutant forms of collagen IV [which are abnormally retained in the endoplasmic reticulum(Gupta et al., 1997) and might have caused co-retention of MIG-6 mutant protein] does not contribute to the suppression. The suppression observed is therefore reminiscent of the ability of fbl-1 (fibulin) mutations to suppress gon-1/adamts proteinase and mig-17 mutant DTC defects(Hesselson et al., 2004; Kubota et al., 2004) and suggests that secreted forms of collagen IV and MIG-6S act antagonistically,either in parallel mechanistic pathways (like fibulin and GON-1) or in the same pathway.

The genetic and phenotypic characterization of mig-6, mig-17 and collagen IV mutants reported here shows that MIG-6S and MIG-17 act antagonistically with secreted collagen IV to guide phase 2 DTC migrations and suggests that MIG-6L and collagen IV act together to promote a normal rate of DTC migration. We speculate that the abnormal distribution of MIG-17 in mig-6 class-s mutants causes an overabundance of normally or abnormally localized collagen IV, which hinders phase 2 DTC guidance. This model predicts that reducing the collagen IV level or altering its supramolecular structure could also rescue the DTC migration defects of mig-17 mutants. In fact, one of the suppressors of mig-17 is a let-2 allele that has a mutation in the non-collagenous (NC) domain of the collagen IV alpha 2 chain (Ohkura and Nishiwaki, personal communication). NC domains are involved in forming a supramolecular network among triple helical protomers and this allele may have lost the ability to form this network. According to this model, EMB-9/collagen IV could be a target for inactivation by MIG-17/ADAMTS acting with MIG-6S. The possible pairing of the MIG-6S isoform with the MIG-17 metalloproteinase to regulate collagen IV extracellular activity could be a theme shared with MIG-6L and the GON-1 ADAMTS, except the latter pair would promote collagen IV function and the former pair would negatively regulate it or antagonize it in some other way. We therefore speculate further that collagen IV levels in gonad basement membrane may need to be regulated oppositely in phase 1 and phase 2 DTC migration, and that this is accomplished by the two isoforms of MIG-6.

We thank Drs T. Schedl, A. Coulson, J. Kimble, A. Spence, A. Fire, J. Ahringer, D. Pilgrim, W. Hung and R. Steven for materials; the Caenorhabditis Genetics Center for strains; Drs H. Qadota, R. Ikegami, G. Dalpe, B. Nash and N. Levy-Strumpf for their help and suggestions;and Dr S. Cordes, J. Yu, L. Zhang and L. Brown for invaluable technical support. We also thank Drs M. Zhen and J. M. Kramer for anti-SAD-1 and anti-EMB-9 antibodies, respectively.