The conserved oligomeric Golgi complex acts in organ morphogenesis via glycosylation of an ADAM protease in C. elegans

Secreted proteins are required to guide the migration of gonadal distal tip cells (DTCs) and shape the gonad in C. elegans(Kimble and Hirsh, 1979; Hedgecock et al., 1987). One of the best-characterized guidance molecules is UNC-6/netrin, which is secreted from neurons on the ventral midline and is required for dorsally oriented DTC migration (Hedgecock et al.,1990; Ishii et al.,1992; Serafini et al.,1996; Wadsworth et al.,1996). Gonad shape and DTC directional migration are also regulated by two different secreted ADAM family proteases, GON-1 and MIG-17(Blelloch and Kimble, 1999; Blelloch et al., 1999; Nishiwaki et al., 2000). GON-1 is required for both DTC motility and the expansion of gonadal arms. MIG-17 is required for the control of the direction of DTC migration, but not motility per se, and is secreted from body wall muscle cells and accumulates in the gonadal basement membrane. Remodeling of basement membranes by these ADAM proteases appears to be requisite for DTC migration and gonad morphogenesis.

The DTCs in mig-29 and mig-30 mutants show misdirected migration similar to that seen in mig-17 mutants. We have molecularly cloned mig-29 and mig-30 and found that they encode homologs of COG-3/Sec34 and COG-1/ldlBp, respectively, which are components of the conserved oligomeric Golgi (COG) complex - a multisubunit protein complex associated with the periphery of the Golgi in mammalian cells(Ungar et al., 2002). Therefore, mig-29 and mig-30 were renamed cogc-3and cogc-1, respectively, where cogc stands for COG component. The COG complex, also called the ldlCp complex, Sec34/Sec35p complex and GTC complex, has been identified independently by several groups working in both yeast and mammalian systems. The COG complex functions in multiple aspects of intracellular vesicle trafficking, including that between the endoplasmic reticulum (ER) and Golgi, between Golgi cisternae and from endosomes to Golgi (Krieger et al.,1981; Wuestehube et al.,1996; Walter et al.,1998; Spelbrink and Nothwehr,1999; Whyte and Munro,2001; Suvorova et al.,2002). Biochemical and electron microscopic analyses of the mammalian COG complex have identified eight protein components: COG-1-4 form lobe A of the complex, whereas COG-5-8 form lobe B(Ungar et al., 2002; Loh and Hong, 2004). Genetic experiments in S. cerevisiae have implicated each of these components in vesicular transport from the ER to Golgi. Disruption of any one of the lobe A components causes severe growth defects in yeast, whereas disruption of lobe B components results in only mild growth defects(Whyte and Munro, 2001; Ram et al., 2002). Genetic analysis in mutant Chinese hamster ovary (CHO) cells showed that COG-1/ldlBp and COG-2/ldlCp are required for proper glycosylation of low-density lipoprotein receptor (LDLR) (Kingsley et al., 1986). The Drosophila gene, four way stop,encodes a homolog of COG-5 that is essential for spermatocyte cytokinesis,elongation of spermatids and maintenance of Golgi morphology in males(Farkas et al., 2003). Recently, COG-7 was shown to be a member of a novel class of proteins involved in human congenital disorder of glycosylation (CDG), which causes multisystemic developmental abnormalities, often leading to death. In individuals with CDG who have a COG-7 defect, protein glycosylation and glycosyltransferase trafficking are affected because of the destabilization of the COG complex (Wu et al.,2004). These genetic and biochemical analyses suggest that the components of the COG complex are functionally distinct and have important cellular functions; however, the roles that these components play in organogenesis remain largely unexplored.

We now report that the C. elegans COG complex is crucial for organ morphogenesis. The C. elegans genome encodes homologs for all eight components of the mammalian COG complex. We found that two of the C. elegans lobe A components, COGC-3 and COGC-1, are required for proper glycosylation and gonadal localization of the MIG-17 protease, which controls gonadal DTC migration. In addition, we show that the intracellular localization and expression of a membrane-bound Golgi enzyme, nucleoside diphosphatase (NDPase), MIG-23, is abnormal in cogc-3 and cogc-1 mutants. We have previously shown that MIG-23 is required to achieve proper glycosylation of MIG-17 in muscle cells(Nishiwaki et al., 2004). These results suggest that the COG complex acts in gonad morphogenesis by regulating Golgi enzymes such as MIG-23, thereby affecting the glycosylation and function of the MIG-17 ADAM protease.

Worms were grown at 20°C (Brenner,1974). The following mutations were used: cogc-3(k181),cogc-1(k179) (this work), unc-11(e47, n2954)(Brenner, 1974; Nonet et al., 1999), lin-17(rh95, n3090) (Sawa et al.,1996), unc-119(e2498)(Maduro and Pilgrim, 1995) and mig-17(k174) (Nishiwaki et al.,2000).

Genetic mapping experiments assigned both cogc-3(k181) and cogc-1(k179) mutations between lin-17 and unc-11 on linkage group I. The cogc-3 and cogc-1 genes were cloned by SNP mapping (Wicks et al.,2001) followed by transposon tagging. Tc4 was inserted between nucleotide positions 8913/8914 of the genomic clone, Y71F9AM in cogc-3. Tc1 was inserted between nucleotide positions 80618/80619 of Y54E10A in cogc-1.

Nomarski and fluorescence microscopy were performed using a Zeiss Axioplan 2 microscope. To score the gonadal DTC migration defects, the trajectories of DTCs were deduced from the final shape of the gonad arms. The localization of MIG-23-GFP, YFP-TRAM, mRFP-SP12 and COGC-3 in body wall muscle cells was analyzed using a confocal laser scanning microscope (Radiance 2100 Rainbow,BioRad) equipped with a C-apochromat 63× (water immersion; NA 1.2) lens and controlled by Lasersharp 2000 software (BioRad).

The following primers were used to amplify the cogc-3 genomic fragment: 5′-GATGCCGTCAAGAACAAGGAGAAGAGATCG-3′ and 5′-TTCTTGTTGGAATGTGGGAAAAGGAAGCAG-3′. The following primers were used to amplify the cogc-1 genomic fragment:5′-GGCACGGACAAGTCGTCGCGTTGAAATATG-3′ and 5′-CTGCAAGGCTACCTATGCCTGCCTTGTGTTGG-3′. mec-7p::mig-17::GFP,unc-54p::mig-23::GFP and mig-17::GFP were constructed previously(Nishiwaki et al., 2000; Nishiwaki et al., 2004). mig-17::Venus was constructed by replacing the coding region of GFP in the mig-17::GFP plasmid with that of the Venus gene(encoding a yellow fluorescent protein variant)(Nagai et al., 2002). The myo-2p::YFP::TRAM plasmid was kindly provided by Tom A. Rapoport(Rolls et al., 2002). The unc-54p::mRFP fusion gene was constructed by joining the fragments containing the unc-54 promoter region with mRFP, a coding region for monomeric red fluorescent protein, kindly provided by Roger Y. Tsien (Campbell et al., 2002). The unc-54p::mRFP::SP12 gene was constructed by inserting the unc-54p::mRFP fragment to the NotI site of plasmid SP12,which was kindly provided by Anne Spang (Poteyaev et al., 2004). arIs37[pmyo-3::ssGFP] I was used to analyze endocytosis and secretion (Fares and Greenwald,2001a; Fares and Greenwald,2001b). mec-7p::cogc-3 and unc-54p::cogc-3 were constructed by joining the fragments containing the unc-54 or mec-7 promoter regions with the cogc-3 cDNA fragment and a 1.8 kb cogc-3 terminator fragment amplified by PCR. A full-length cogc-3 cDNA, yk863b10, was used.

DNA mixtures were first injected into unc-119(e2498) animals(Mello et al., 1991), and the transgenic arrays were transferred to cogc-3(k181);unc-119(e2498) or cogc-1(k179);unc-119(e2498) animals by mating. The genomic cogc-3 PCR fragment was injected at 30 μg/ml with 50 μg/ml of the marker plasmid sur-5::GFP, 10 μg/ml of unc-119⁺ plasmid pDP#MM016B(Maduro and Pilgrim, 1995) and 60 μg/ml pBR322. The genomic cogc-1 PCR fragment was injected at 10 μg/ml with 80 μg/ml sur-5::GFP, 10 μg/ml unc-119⁺ plasmid and 40 μg/ml pBR322. unc-54p::cogc-3 was injected at 20 μg/ml with 70 μg/ml sur-5::GFP, 10 μg/ml unc-119⁺ plasmid and 50μg/ml pBR322. mig-17::GFP was injected at 20 μg/ml with 10μg/ml unc-119⁺ plasmid and 120 μg/ml pBR322. mig-17::Venus was injected at 20 μg/ml with 10 μg/ml unc-119⁺ plasmid and 120 μg/ml pBSII KS(-). To construct animals for touch cell expression experiments, mec-7p::mig-17::GFP was injected at 100 μg/ml with 10 μg/ml unc-119⁺ plasmid and 40 μg/ml pBR322. mec-7p::cogc-3 was injected at 20 μg/ml with 10 μg/ml unc-119⁺ plasmid and 120 μg/ml pBR322. To construct animals carrying both mec-7p::mig-17::GFP and mec-7p::cogc-3, 100 μg/ml mec-7p::mig-17::GFP and 20μg/ml mec-7p::cogc-3 were co-injected with 10 μg/ml unc-119⁺ plasmid and 20 μg/ml pBR322. unc-54p::mig-23::GFP was injected at 10 μg/ml with 10 μg/ml unc-119⁺ plasmid and 130 μg/ml pBSIIKS(-). unc-54p::mRFP::SP12 was injected at 50 μg/ml with 70 μg/ml unc-54p::mig-23::GFP, 10 μg/ml unc-119⁺plasmid and 20 μg/ml pBSIIKS(-). myo-2p::YFP::TRAM was injected at 5 μg/ml with 10 μg/ml unc-119⁺ plasmid and 135μg/ml pBSIIKS(-).

The amino acid identities between the COG components of C. elegansand human are 23% (COGC-1), 26% (COGC-2), 30% (COGC-3), 33% (COGC-4), 22%(COGC-5), 23% (COGC-6), 19 % (COGC-7) and 20% (COGC-8) using the program GENETYX (Software Development). The expected protein from W01B6.9(cogc-7) has relatively weak overall homology with human COG-7(Fig. 2C), although the similarity is greater in the coiled-coil region; by contrast, it shows no significant similarity to yeast Cog7p. RNAi experiments were carried out as described (Maeda et al.,2001).

The N-terminal residues 34-260 of COGC-3 tagged with GST were produced in Escherichia coli. A synthetic peptide (EKEMEGKRESLR-EMVGRR)corresponding to a region of the COGC-1 N terminus was coupled to keyhole limpet hemocyanin to use for an antigen. Rabbit antiserum against COGC-3 was affinity purified on columns fixed with histidine-tagged COGC-3 polypeptide. The rabbit antiserum against the COGC-1 peptide was also affinity purified using the synthetic peptide.

Worms were disrupted by glass beads in a buffer containing 50 mM Tris-HCl(pH 7.4), 200 mM NaCl, protease inhibitor cocktail (Roche) and 1% (v/v) Triton X-100. The samples were immunoblotted with rabbit anti-GFP IgG (1:1000,Molecular Probes), mouse anti-α-tubulin IgG 12G10 (1:1000, J. Frankel and M. Nelson, Developmental Studies Hybridoma Bank, University of Iowa),anti-COGC-3 (1 μg/ml), anti-COGC-1 (1 μg/ml) or anti-MIG-23(Nishiwaki et al., 2004)(1:1000) at room temperature for 1 hour followed by incubation with peroxidase-conjugated anti-rabbit IgG or anti-mouse IgG (1:5000, Amersham Pharmacia Biotech) at room temperature for 1 hour.

The worm lysates were incubated with rabbit anti-GFP (1:1000, Molecular Probes) or anti-COGC-3 (1 μg/ml) at 4°C for 3 hours. Samples were precipitated with protein A-Sepharose beads (Amersham Biosciences). Normal rabbit IgG (Genzyme TECHNE) was used as a control antibody. PNGase F treatment(New England BioLabs) was performed according to the manufacturer's instructions.

For whole-mount immunohistochemistry, samples were processed as described by Yamaguchi et al. (Yamaguchi et al.,1983), except that they were immersed in methanol at -20°C for 5 minutes followed by acetone at -20°C for 5 minutes. After blocking in 2%bovine serum albumin (BSA) in phosphate-buffered saline (PBS), samples were incubated with anti-COGC-3 (1 μg/ml) or mouse monoclonal anti-GFP IgG (3E6,1:1000; Molecular Probes) in PBS containing 1% BSA at 4°C for 12 hours. Samples were then incubated with secondary antibodies TRITC donkey anti-rabbit IgG (Jackson), Texas Red donkey anti-rabbit IgG (Jackson) or fluorescein donkey anti-mouse IgG (Jackson) diluted 1:200 in PBS containing 1% BSA at room temperature for 1 hour and with DAPI (2 μg/ml; Wako) at room temperature for 10 minutes. For frozen sections, animals were fixed with 4%paraformaldehyde in PBS for 12 hours on ice. After washing in PBS, they were resuspended sequentially in 10, 20 and 30% sucrose in PBS at 4°C. After the animals sank to the bottom of the Tissue-Tek Cryomold (Miles), the overlying sucrose solution was removed and OCT compound (Miles) was added. Frozen sections (10 μm) were prepared using a cryostat (MICROM). After blocking the sections with 2% BSA in PBS, samples were incubated with rabbit anti-COGC-3 (1 μg/ml) or rabbit anti-GFP (1:1000, Molecular Probes) in PBS containing 1% BSA for 12 hours at 4°C, with TRITC donkey anti-rabbit IgG(1:100; Jackson) in PBS containing 1% BSA for 1 hour, with fluorescein phalloidin (10 U/ml; Molecular Probes) for 1 hour, or with DAPI (2 μg/ml;Wako) for 10 minutes at room temperature.

The C. elegans hermaphrodite gonad is bi-lobed, one lobe each in the anterior-right and posterior-left areas of the body cavity(Fig. 1A). Wild-type animals have U-shaped gonadal arms (Fig. 1B), in contrast to the variously abnormal shapes of the arms in mig-17 mutants resulting from meandering DTC migration(Fig. 1C). To study the mechanism of MIG-17-dependent regulation of DTC migration, we analyzed two mutants, cogc-3(k181) and cogc-1(k179), which were isolated spontaneously from the mutator strain mut-7(pk204). These mutants have mig-17-like meandering DTC phenotypes(Fig. 1D-G). As in mig-17 mutants, gonadal morphology in cogc-3(k181) and cogc-1(k179) mutants was highly variable, owing to the abnormal DTC migration. We classified the DTC migration phenotypes among mig-17(k174),cogc-3(k181) and cogc-1(k179) mutants with respect to the phase during which abnormal migration first became apparent(Fig. 1A). All three mutants had essentially the same DTC phenotypes: migration was normal during phase 1,the initial migration on the ventral body wall muscle, but migration became abnormal during phases 2 and 3 when the MIG-17 localizes to the gonadal basement membrane (Nishiwaki et al.,2000). DTC migration defects in cogc-3(k181) and cogc-1(k179) mutants were more severe and penetrant than those in the putative null mutant, mig-17(k174)(Fig. 1H). Both cogc-3(k181) and cogc-1(k179) mutants exhibited a protruding vulva phenotype.

The phenotypic similarity between cogc-3, cogc-1 and mig-17 mutants suggests that COGC-3, COGC-1 and MIG-17 function in a common pathway controlling DTC migration. To test this possibility, we examined the genetic interactions between these genes using putative null alleles. Both cogc-3(k181) and cogc-1(k179) appeared to be null alleles, as discussed in a later section. When cogc-3(k181) or cogc-1(k179) was combined with mig-17(k174), the defective DTC migration phenotypes were not enhanced and were similar to those seen in the single mutants cogc-3(k181) or cogc-1(k179)(Fig. 1I). These results,together with the fact that the defects in cogc-3(k181) and cogc-1(k179) mutants are more severe compared with mig-17(k174), support the idea that mig-17 acts in a common pathway with cogc-3 and cogc-1, and that mig-17function is totally compromised in cogc-3 and cogc-1mutants.

cogc-3 and cogc-1 genes were cloned by genetic mapping and injection rescue experiments, and were mapped between lin-17 and unc-11 in linkage group I. Genetic mapping using single nucleotide polymorphisms (SNPs) placed cogc-3 in the region between genomic clones W03D8 and Y54E10BL. Similarly, cogc-1 was localized between Y54E10A and W05F2. Sequence analysis of the cogc-3(k181)genome identified a Tc4 transposon insertion in exon 5 of the predicted gene Y71F9AM.4a, and the analysis of the cogc-1(k179) genome identified a Tc1 transposon insertion in exon 8 of the predicted gene Y54E10A.2, which constituted a single gene with part of Y54E10A.1 as determined by cDNA analysis (see Fig. S1 in the supplementary material). Both mutant phenotypes were rescued by injecting PCR products for the corresponding genomic regions(Fig. 2A). A series of RNA interference (RNAi) analyses suggested that the products from the longest mRNAs were responsible for cogc-3 and cogc-1 defects (see Fig. S1 in the supplementary material). Sequence analysis of cogc-3and cogc-1 cDNAs revealed that the encoded proteins are homologs of human COG-3/Sec34 and COG-1/ldlBp, respectively. These proteins correspond to two of the four lobe A subunits of the bi-lobed hetero-octameric COG complex required for intracellular vesicle trafficking and Golgi function(Fig. 2B,C; Fig. 3A,B)(Chatterton et al., 1999; VanRheenen et al., 1999; Suvorova et al., 2001; Ungar et al., 2002). cogc-3 comprises 14 exons encoding a predicted protein of 794 amino acids. The N terminus contains a coiled-coil domain that potentially acts in protein-protein interactions, whereas the C terminus has an EEA1-like domain that was originally identified in the EEA-1 protein, which tethers vesicles to endosomes (Christforidis et al.,1999; Loh and Hong,2002). cogc-1 comprises 15 exons encoding a predicted protein of 787 amino acids, and it also has an N-terminal coiled-coil domain(Fig. 2A,D).

Seven of the eight putative components of the C. elegans COG complex, including those corresponding to COGC-3 and COGC-1, have been identified based on their homology with mammalian and yeast components(Podos et al., 1994; Chatterton et al., 1999; Whyte and Munro, 2001). By searching sequence databases, we identified the remaining component, COGC-7,which is also encoded by the C. elegans genome(Fig. 3B). To test the role of each COG subunit in gonadal DTC migration, soaking RNAi experiments were performed in wild-type hermaphrodites at the first larval stage. The RNAi knockdown of any of the four components of lobe A, including both cogc-3 and cogc-1, resulted in meandering DTC phenotypes similar to those seen in cogc-3(k181) and cogc-1(k179)mutants (Fig. 1D-G and Fig. 3C-F). We also observed similar DTC migration defects in RNAi knockdowns of lobe B components. Interestingly, RNAi of lobe A components caused a much stronger effect than that of lobe B components (Fig. 3G), consistent with the observations in yeast that mutations in lobe A components result in more severe growth defects than those in lobe B(Whyte and Munro, 2001). These results suggest that all eight COG components participate in a common pathway that regulates DTC migration and that each component is functionally distinct during this process.

We performed western blot analysis using antibodies against the N termini of COGC-3 and COGC-1. The COGC-3 antibody detected an 87 kDa protein in wild-type animals that was not seen in the cogc-3(k181) mutant. In addition, the level of COGC-3 was lower in the cogc-1(k179) mutant(Fig. 4A). Likewise, the COGC-1 antibody detected an 83 kDa protein in wild-type animals that was not detected in the cogc-1(k179) mutant. As with the COGC-3 antibody data, the levels of COGC-1 were reduced in the cogc-3(k181) mutant(Fig. 4B). These results indicate that cogc-3(k181) and cogc-1(k179) are protein-null mutants and that COGC-3 and COGC-1 stabilize one another, although we cannot rule out the possibility that smaller isoforms of these proteins, undetectable by the antibodies, are still produced.

To examine whether COGC-3 and COGC-1 are incorporated in the same protein complex, we performed co-immunoprecipitation using wild-type worm lysate and anti-COGC-3. COGC-1 co-precipitated with COGC-3 in the presence of anti-COGC-3 but not with a control IgG (Fig. 4C,D), supporting the idea that COGC-3 and COGC-1 belong to the same protein complex.

We performed immunohistochemistry to detect COGC-3 localization in situ. COGC-3 was detected in most tissues, including muscle, intestine and hypodermis. COGC-3 expression in the germline was very low, or nil. As observed in mammalian cells (Suvorova et al., 2001; Unger et al., 2002), COGC-3 staining was often prominent in the perinuclear region, which may correspond to the ER or Golgi apparatus (Fig. 5A-C). No signal was detected in cogc-3(k181) mutants(Fig. 5D), indicating that anti-COGC-3 specifically detected COGC-3 in situ. In whole-mount staining,COGC-3 was strongly expressed in cells of the pharynx, vulva and seam hypodermis (Fig. 5E-J), and of the intestine (data not shown). We examined the intracellular distribution of COGC-3 in the body wall muscle cells using a Golgi enzyme, MIG-23/NDPase (a type II membrane protein), as a marker. Some of the anti-COGC-3 signal colocalized with MIG-23-GFP, in particular, at the perinuclear region(Fig. 6A-C). We analyzed the colocalization of the COGC-3 and YFP-TRAM(Rolls et al., 2002), an ER-marker in the body wall muscle cells. COGC-3 was widely distributed in the cytoplasm and partially colocalized with YFP-TRAM(Fig. 6D-F).

It has been reported that Golgi type II membrane proteins are mislocalized or destabilized in culture cells harboring mutations in COG components(Bruinsma et al., 2004; Oka et al., 2004). Thus, we examined whether the pattern of MIG-23-GFP expression was affected in cogc-3 and cogc-1 mutants. Although the punctate cytoplasmic expression of MIG-23-GFP seen in wild-type worms was still observed, the intensity and number of the GFP signals were significantly reduced in cogc-3 mutants (Fig. 6G,H). Furthermore, western blot analysis revealed that the amount of endogenous MIG-23 was also reduced in cogc-3 and cogc-1mutants (Fig. 6O). The distribution of the ER marker mRFP-SP12 was not affected in cogc-3mutants (Fig. 6I-N). These results suggest that MIG-23 is destabilized in cogc-3 and cogc-1 mutants and that either the intracellular sorting of MIG-23 is severely altered or the structure of the MIG-23-associated Golgi apparatus is disorganized.

MIG-17 is produced in and secreted from the body wall muscle cells(Nishiwaki et al., 2000). In the present study, we confirmed COGC-3 expression in these cells. If COGC-3 is required for MIG-17-mediated control of DTC migration, it is possible that this process requires COGC-3 expression in body wall muscle cells. The expression of cogc-3, which is controlled by the muscle-specific unc-54 promoter, partially rescued the cogc-3(k181)phenotype; however, this phenotype was not rescued upon expression of cogc-3 controlled by the touch neuron-specific mec-7promoter (Fig. 7A), suggesting that normal gonadal DTC migration requires COGC-3 activity in muscle cells,where MIG-17 is also expressed. One possible explanation of the partial rescue is that COGC-3 is also required in tissues besides muscle.

cogc-3 and cogc-1 encode components of the COG complex,which is required for intracellular vesicle trafficking and glycosylation. We therefore analyzed glycosylation and gonadal localization of MIG-17 in these mutant backgrounds. We compared the molecular sizes of MIG-17-GFP in wild type, cogc-3(k181) and cogc-1(k179) mutants using western blotting. The band at ∼100 kDa, which corresponds to the MIG-17-GFP proform in wild-type worms, was broader and contained lower molecular weight species in the cogc-3 and cogc-1 mutants. The broadening was not due to the protein degradation in the mutants because treatment of the samples with PNGase F, which removes all N-glycans, yielded sharp bands of ∼85 kDa both in wild type and mutants(Fig. 7B). These results indicate that MIG-17 glycosylation is incomplete in cogc-3 and cogc-1 mutants. We then examined the gonadal localization of MIG-17-GFP in both cogc-3(k181) and cogc-1(k179) mutants by immunostaining of worm cross-sections with anti-GFP. The MIG-17-GFP signal seen at the gonad surface in the wild type was almost undetectable in cogc-3(k181) and cogc-1(k179) mutants(Fig. 7C,D), indicating that the localization of MIG-17 is severely affected in these mutants.

The localization defect of MIG-17 could be due either to the failure of secretion of MIG-17 from the muscle cells or to the loss of MIG-17 affinity for the gonadal basement membrane. To examine whether the secretory pathway functions normally in the muscle cells in cogc-3 and cogc-1mutants, endocytotic uptake of GFP and MIG-17-Venus by coelomocytes was assessed according to the method of Fares and Greenwald(Fares and Greenwald, 2001a; Fares and Greenwald, 2001b). When we expressed either GFP containing the N-terminal secretion signal(ssGFP) or the MIG-17-Venus fusion protein in body wall muscle cells, both of these proteins localized to coelomocytes, which are scavenger cells that actively endocytose fluid from the pseudocoelom(Fig. 8A,D). Coelomocytes in cogc-3 and cogc-1 mutants showed normal accumulation of GFP and MIG-17-Venus (Fig. 8B,C,E,F). Moreover, MIG-17-Venus did not accumulate in the body wall muscle cells in these mutants (data not shown). These results indicate that MIG-17 is secreted normally from the muscle cells in cogc-3 and cogc-1 mutants.

We asked whether the function of cogc-3 in DTC migration is dependent on MIG-17. We addressed this problem by ectopically expressing MIG-17. MIG-17 is normally expressed in muscle cells, but ectopic expression in touch neurons under the control of the touch neuron-specific mec-7promoter also rescues DTC migration defects in mig-17 mutants(Nishiwaki et al., 2004). Conceivably, MIG-17 secreted from touch neurons acts on the gonad to achieve normal DTC migration. However, MIG-17 expression in touch neurons of cogc-3 mutants failed to rescue the DTC phenotypes. COGC-3 expression in touch neurons also failed to rescue cogc-3. By contrast, COGC-3 and MIG-17 co-expression in touch neurons substantially ameliorated the phase 2 DTC migration defects of cogc-3 mutants(Fig. 9). The rescue of phase 3 defects was weak (data not shown). These results indicate that MIG-17 is functionally impaired in cogc-3(k181) mutants, implying that COGC-3 normally requires MIG-17 for its action in DTC migration. Thus, it is probable that COG components other than COGC-3 are also normally expressed in touch neurons.

The COG complex is well conserved across species. All eight genes encoding the components of the COG complex, including COGC-3 and COGC-1, were identified in C. elegans. The cogc-3(k181) and cogc-1(k179) mutations in COG lobe A are probably null because the transposon insertions identified in these mutants are expected to produce termination codons. In fact, immunoblotting confirmed that these mutants are indeed null. However, worms carrying these mutations are viable and can proliferate as homozygotes, albeit very slowly. cogc-1 RNAi in cogc-3 mutants exacerbated the DTC migration defects (data not shown), suggesting that COGC-1 has other functions that are unrelated to COGC-3 function in gonadal DTC migration. These scenarios are very similar to those in the yeast lobe A mutants cog1, cog2 or cog3, each of which grows very slowly and exhibit synthetic lethality when combined with each other (VanRheenen et al.,1999; Ram et al.,2002).

In mammalian cultured cells, the level of COG-1 is reduced when COG-3 is depleted (Zolov and Lupashin,2005), but the level of COG-3 is not affected by depletion of COG-1 (Ungar et al., 2002; Oka et al., 2004). However, we observed in C. elegans that COGC-3 and COGC-1 stabilize each other. These results suggest that molecular interactions among the lobe A components in C. elegans may not be identical to those in mammals.

COGC-3 is widely expressed in tissues such as the pharynx, vulva, seam hypodermis and body wall muscle. Using the maker protein MIG-23, a membrane-bound Golgi NDPase expressed in these muscle cells, we found that COGC-3 also localizes to the Golgi. Interestingly, colocalization of COGC-3 and MIG-23 was more clearly seen around nuclei than in the rest of the muscle cell cytoplasm. This could reflect differential distribution of these proteins in the Golgi compartments. The human Golgi NDPase is expressed widely in the Golgi (Wang and Guidotti,1998), whereas the COG complex localizes primarily to the cis-Golgi, although it is also detected in other regions of the Golgi, including the late Golgi (Spelbrink et al., 1999; Kim et al., 2001; Suvorova et al., 2001). Thus,COG complex localization may follow a gradient that decreases from the early to late Golgi.

The punctate MIG-23-GFP expression was reduced in cogc-3 and cogc-1 mutants, suggesting that MIG-23 sorting or MIG-23-associated Golgi architecture may be perturbed in cogc-3 and cogc-1mutants. When examined throughout larval development, this phenotype was not apparent during early stages and manifested during L4 and adult stages (see Fig. S2 in the supplementary material). As Golgi-resident enzymes such as NDPase are thought to be continuously sorted and recycled to the Golgi, the observation that this phenotype manifests only later in development might be due to abnormal development of Golgi structure. We also detected lower steady state levels of MIG-23-GFP in cogc-3 and cogc-1 mutants. A similar reduction in type II Golgi membrane proteins, such as mannosidase II,GOS-28, GS15, GPP130, CASP, giantin and golgin-84, have been reported in the ldlB and ldlC CHO cells (Oka et al.,2004).

MIG-17 was underglycosylated in cogc-3 and cogc-1mutants. A number of studies have described functions for COG components in protein glycosylation. For example, glycosylation of invertase is defective in yeast mutants of cog1, cog3, cog5, cog6, cog7 and cog8(Whyte and Munro, 2001), and the glycosylation and stability of the low-density lipoprotein receptor (LDLR)is affected in the CHO mutant cells ldlB and ldlC(Kingsley et al., 1986; Reddy and Krieger, 1989). In addition, multiple glycosylation pathways are disrupted in a type of human CDG caused by a mutation in COG-7 (Wu et al.,2004). Although we have previously reported that MIG-17 is underglycosylated and defective in gonadal localization in mig-23mutants (Nishiwaki et al.,2004), by comparison the defects in gonadal localization of MIG-17 in cogc-3 and cogc-1 mutants are more pronounced. These results raise the possibility that multiple glycosylation enzymes, including MIG-23/NDPase, are affected in cogc-3 and cogc-1mutants.

We found that MIG-17 is secreted from the muscle cells of cogc-3and cogc-1 mutants. As for protein secretion, the function of the COG complex appears to differ depending on the context. For example, in the yeast mutants cog1 and cog2, secretion of many proteins is blocked(Kim et al., 2001), whereas secretion of invertase is delayed yet the extracellular secretion of the protein occurs at near wild-type levels(Bruinsma et al., 2004). Secretion of LDLR to the cell surface is not blocked in ldlB and ldlC cells(Kingsley et al., 1986; Reddy and Krieger, 1989). Thus, glycosylation and secretion of certain groups of proteins may be differentially dependent on the COG complex, implying separable functions for the complex. Alternatively, for many proteins, the COG complex may affect only the efficiency of secretion, and thus these proteins may be secreted slowly even in COG mutants.

The ectopic co-expression of both COGC-3 and MIG-17, but not the individual proteins, in touch receptor neurons rescued the phase 2 migration defects of cogc-3 mutants. Therefore, we conclude that COGC-3 regulates DTC migration through the action of MIG-17. Presumably, COGC-3 expressed in touch neurons facilitates the appropriate glycosylation of MIG-17, which is then secreted from the neurons and diffuses to the gonad where it controls DTC migration. Although we have not yet examined the functional link between the other COG components and MIG-17, the phenotypic similarity among the mutants and RNAi knockdowns suggests that all the COG components act together in MIG-17 glycosylation, which is essential for correct DTC migration. Although the defective phase 2 migration of cogc-3 was strongly rescued by MIG-17 expression in touch neurons, the rescue of phase 3 defects was weak,suggesting that MIG-17 function is more important for phase 2 than for phase 3 migration. In addition, the defects in both phase 2 and 3 are more severe in cogc-3 and cogc-1 compared with mig-17. These results suggest that proteins other than MIG-17 are also affected by the COG complex. The protruding vulval phenotype observed in cogc-3 and cogc-1 mutants suggests this possibility. We performed lectin blot analysis using HRP-conjugated peanut agglutinin (PNA), a lectin with affinity for immature glycans (Wu et al.,2004). We found that PNA-binding activities were elevated in both cogc-3 and cogc-1 mutants (see Fig. S3 in the supplementary material), indicating that these mutations affect the glycosylation of multiple proteins.

In conclusion, we propose that the COG complex regulates the localization and stability of the Golgi enzymes involving MIG-23 and thereby affects the glycosylation and function of the MIG-17 ADAM protease, which is crucial for directed DTC migration. Our findings provide a model for the function of the COG complex in organ morphogenesis. Human CDG causes multisystemic abnormalities that include mental defects and malformation of organs(Marquardt and Denecke, 2003);however, although a mutation in COG-7 was found to affect the trafficking and function of the glycosylation machinery, the molecular mechanisms linking glycosylation defects to clinical presentations remain to be elucidated(Wu et al., 2004). Nonetheless, the functional genetic and molecular studies of the COG complex in C. elegans provide the molecular basis for understanding CDG and other glycosylation-dependent diseases.

We are grateful to Toshihiko Oka for suggestions for experiments, and to Toshihiko Oka, Kunihiro Matsumoto, Naoki Hisamoto and Shinji Ihara for critical reading of the manuscript. We thank Alan Couslon, Andy Fire, Yuji Kohara, Takeshi Ishihara, Hitoshi Sawa, Roger Y. Tsien, Hiroshi Kagoshima,Melissa M. Rolls, Tom A. Rapoport, Dmitry Poteryaev, Anne Spang, Theresa Stiernagle and the Caenorhabditis Genetics Center for materials; Asako Sugimoto for the soaking RNAi protocol. We also thank the members of our laboratory for discussions.