The V0-ATPase mediates apical secretion of exosomes containing Hedgehog-related proteins in Caenorhabditis elegans

All developmental and physiological functions performed by epithelia depend on the polarized targeting of the plasma membrane and secreted proteins to either the apical or basolateral plasma membranes (Rodriguez-Boulan et al., 2005). Cargo proteins sorted in the Golgi apparatus and the endosomal system through sets of basolateral- and apical-specific sorting determinants are transported to the plasma membrane following partially different routes (Hoekstra et al., 2004; Rodriguez-Boulan et al., 2005). Although basolateral secretion has been fairly well characterized, the mechanisms involved in apical trafficking remain poorly defined (Rodriguez-Boulan et al., 2005).

Basolateral sorting signals usually correspond to tyrosine or dileucine residues found in the COOH terminus of proteins. They are recognized by basolateral-specific adaptor complexes (Bonifacino and Lippincott-Schwartz, 2003; Rodriguez-Boulan et al., 2005), such as AP-1B in epithelial cells (Folsch et al., 1999). Before membrane fusion and SNARE action, vesicles are thought to be tethered to the basolateral membrane by the exocyst complex (Whyte and Munro, 2002), which was initially identified in yeast (TerBush and Novick, 1995). In metazoans, the exocyst is required for basolateral delivery of the LDL receptor in MDCK cells (Grindstaff et al., 1998; Yeaman et al., 2001), of E-cadherin in the Drosophila melanogaster notum (Langevin et al., 2005), and for Rhodopsin1 transport in D. melanogaster photoreceptor cells (Beronja et al., 2005). Recent results suggest that AP-1B and the exocyst act primarily in recycling endosomes (Ang et al., 2004; Beronja et al., 2005; Langevin et al., 2005; Lock and Stow, 2005; Satoh et al., 2005), which underlines the central role of this organelle in sorting processes. Indeed, recycling endosomes may be compartmentalized into apical- and basolateral-related domains, or even divided into distinct organelles, suggesting that they could also play a critical role in apical trafficking (Hoekstra et al., 2004; Rodriguez-Boulan et al., 2005).

Aside from this possible role of recycling endosomes, all other aspects of sorting along the basolateral and apical routes seem to differ. Apical signals are more diverse and often correspond to posttranslational adducts, such as lipids or glycans (Schuck and Simons, 2004; Rodriguez-Boulan et al., 2005). For instance, the Hedgehog morphogen is secreted apically upon cholesterol addition, but basolaterally otherwise (Gallet et al., 2003). No specific apical cytosolic complex, akin to AP-1B or the exocyst, has been identified so far. Instead, protein clustering, possibly through lipid rafts, is thought to mediate the sorting and transport of apical cargoes (Schuck and Simons, 2004; Rodriguez-Boulan et al., 2005). In particular, glycosyl phosphatidylinositol–linked proteins appear to form oligomers that are directly targeted to the apical membrane (Paladino et al., 2004, 2006; Hua et al., 2006). Several proteins have been proposed to play an active role in apical protein clustering, raft formation, and/or apical delivery, such as caveolins (Kurzchalia et al., 1992), annexin 13b (Fiedler et al., 1995), and the tetraspan protein VIP17/MAL (Cheong et al., 1999; Puertollano et al., 1999). However, their mechanistic roles have not been fully elucidated, or their implication has been questioned (Manninen et al., 2005). In addition to the limited understanding of apical secretion at the molecular level, it is not clear whether the terminal fusion process involves small vesicles, such as those defined at synapses, or larger organelles, such as secretory lysosomes (Blott and Griffiths, 2002).

Hence, despite the many critical findings originating from tissue culture cells (Rodriguez-Boulan et al., 2005), investigations with other systems and other cargo proteins could help to elucidate the mechanisms involved in apical exocytosis. Caenorhabditis elegans, which has contributed to decipher the mechanisms controlling vesicular trafficking (Nurrish, 2002), provides such an in vivo model. We have chosen to analyze apical secretion of cuticle proteins by the epidermis. The cuticle includes glycosylated collagens, glycosyl phosphatidylinositol–linked cuticlins, and lipid-modified Hedgehog-related peptides (McMahon et al., 2003; Sapio et al., 2005; Zugasti et al., 2005). We previously suggested that the gene che-14 is required for cuticle secretion (Michaux et al., 2000). The CHE-14 protein is the C. elegans orthologue of Drosophila Dispatched, which participates in apical targeting of cholesterol-modified Hedgehog (Burke et al., 1999; Gallet et al., 2003).

While searching for che-14 alleles (Michaux et al., 2000), we uncovered several additional mutations inducing che-14–like phenotypes and reasoned that they might identify new components of the apical trafficking pathway. Two such mutations, mc37 and mc38, proved to be small deletions behaving as genetic-null alleles of the gene vha-5 (unpublished data). The gene vha-5 encodes one of the four C. elegans “a” subunits of the V0 sector of the vacuolar H⁺-ATPase (V-ATPase), and is required for development beyond the L2 larval stage (Oka et al., 2001; Pujol et al., 2001). The V-ATPase is a multisubunit protein complex consisting of two subcomplexes called the V0 and V1 sectors (Fig. 1 A). The cytosolic V1 sector hydrolyses ATP and provides the energy to pump protons through the transmembrane proteolipid pore formed by the V0 sector (Nishi and Forgac, 2002). The V-ATPase is the main proton pump establishing a pH gradient in the secretory and endocytic pathways. It generates a proton-motive force that is essential to load synaptic vesicles with neurotransmitters before secretion (Amara and Kuhar, 1993). The V-ATPase is also found at the apical plasma membrane of polarized cells, where it is essential for osmoregulation in animal excretory systems (Nishi and Forgac, 2002). More recently, biochemical and genetic data suggested that the V0 sector can play a role independently from the V1 sector. In Saccharomyces cerevisiae, vacuoles deficient for the “a” subunit Vph1p do not fuse efficiently (Peters et al., 2001; Bayer et al., 2003). In D. melanogaster, neurons lacking the “a” subunit Vha100 accumulate vesicles in synaptic terminals (Hiesinger et al., 2005). In both cases, the defects were independent of the proton gradient and placed downstream of SNARE function (Peters et al., 2001; Bayer et al., 2003; Hiesinger et al., 2005).

By further dissecting the role of vha-5 using targeted mutagenesis, and by comparing phenotypes resulting from the inactivation of V1 or V0 subunits, we uncover a specific role for the V0 sector in mediating secretion to the apical membrane. In particular, we show that the V0 sector is required for apical secretion of Hedgehog-related peptides through a multivesicular compartment able to release exosomes.

To determine the distribution and subcellular localization of VHA-5, we raised polyclonal antibodies against its cytoplasmic NH₂ terminus. In addition, we generated a COOH-terminal VHA-5∷GFP fusion, which rescued the larval lethality caused by the vha-5(mc38) deletion (Fig. 1 B). The VHA-5 antiserum recognized a 105-kD protein in wild-type extracts (Fig. 1 C, lane a). To prove its specificity, we examined extracts from vha-5(mc38) homozygous animals carrying the rescuing VHA-5∷GFP construct. The VHA-5 antiserum failed to detect the ∼105-kD band in these extracts, but detected an ∼135-kD band (Fig. 1 C, lane b). These results are consistent with vha-5(mc38) being a small deletion associated with a frameshift (Fig. 1 B and not depicted) and with the presence of 257 additional residues in the GFP-fusion protein. We conclude that the VHA-5 antiserum is specific and that vha-5(mc38) is a molecular null mutation.

In agreement with previously published observations (Oka et al., 2001; Pujol et al., 2001), we found that VHA-5 is expressed in the H-shaped excretory cell corresponding to the C. elegans kidney-like organ (Fig. 1, D and E). It is also expressed in the main epidermal syncytium (Fig. 1, D–F), which had previously been overlooked. The excretory cell extends long processes called excretory canals where osmoregulation takes place (Nelson and Riddle, 1984), whereas the epidermis controls body length and apical cuticle secretion (White, 1988). VHA-5 colocalized apically with the V1 subunit VHA-8 in both tissues (Fig. 1 E; note that VHA-5 is not expressed in the lateral epidermis). VHA-5 was localized at the level of apical membrane stacks by immunogold staining (Fig. 1 G). Consistent with VHA-5 distribution and a role of the V-ATPase in osmoregulation (Nishi and Forgac, 2002), vha-5(mc38) larvae filled with fluid and died at the L1 stage (unpublished data), which corresponds to the phenotype observed after laser ablation of the excretory cell (Nelson and Riddle, 1984). In addition, vha-5(mc38) L1 larvae had a severe malformation of the lateral cuticular specializations known as alae (Fig. 1 D and Fig. 2 A), which are primarily synthesized by the lateral seam cells. Although VHA-5 is not expressed in these cells, the main epidermal syncytium also contributes to their morphogenesis (Sapio et al., 2005). Because VHA-5 is transmembraneous and not cuticular, the simplest interpretation for this phenotype is that vha-5 mutations compromise the secretion of proteins needed for alae formation.

As outlined in the previous section, the V0 sector may fulfill two distinct functions; either working together with the V1 sector to mediate proton pumping or working alone, as in yeast and D. melanogaster neurons, to mediate membrane fusion. To determine which of these functions could account for the cuticle secretion defect observed in vha-5(mc38) larvae, we examined the role of other V-ATPase subunits in cuticle formation using the RNAi approach. If improper proton pumping is responsible for the aforementioned vha-5 cuticle defects, RNAi knockdown of either V0 or V1 subunits should result in similar cuticular defects. Conversely if the loss of a V0-specific function accounts for the vha-5 cuticular phenotype, only RNAi knockdown of V0 subunits should phenocopy vha-5 cuticle defects. We chose two V1 subunits (VHA-8 and -13) and one V0 subunit (VHA-4) encoded by single-copy genes, which were, thus, expected to be ubiquitously expressed and to display RNAi phenotypes of comparable severity. In addition, we tested RNAi against the three genes encoding the V0 “c” subunit (vha-1, -2, and -3), which are >78% identical at the nucleotide level. We found that the RNAi phenotype of vha-1 was the strongest and was directly comparable to that of vha-4, -8, and -13 (Fig. 2 B′), presumably because it reflects knockdown of all three paralogs.

RNAi against these V1 or V0 subunit genes led to 100% lethality in the progeny of treated animals (Fig. 2 B‘, bottom bars). It is likely that most embryos died because of a defect in yolk endocytosis, which is known to be sensitive to proton pumping (Choi et al., 2003). In agreement, we found that yolk vitellogenin-GFP accumulated in the pseudocoelom of RNAi-treated animals rather than in oocytes and embryos (Fig. S1). Despite the strong lethality induced by loss of V-ATPase function, a few L1 larvae hatched and, invariably, died filled with fluid before the L2 larval stage (Fig. 2 B’‘), as observed for vha-5(mc38) larvae. Strikingly, we observed before their death that L1 hatchlings displayed severe alae defects when RNAi targeted the V0 subunits vha-1 or -4, but had wild-type alae after knockdown of the V1 subunits vha-8 or -13 (Fig. 2, B and B’). Consistent with the phenotype of vha-5–null mutants, RNAi against vha-5 also affected alae formation (Fig. 2, B and B‘–C), although lethality was weaker because VHA-5 is not ubiquitously expressed like VHA-4. One trivial explanation for the persistence of normal alae after V1 subunit knockdown could be that RNAi was less efficient than for V0 subunits. It is unlikely, as the lethality rates and the larval osmoregulation defects observed after V1 and V0 subunit knockdown were comparable (Fig. 2, B’ and B‘’), hinting that both RNAi were equally effective. To support this idea, we submitted a vha-8∷gfp transgenic strain to vha-8 RNAi and verified that it induced a drastic decrease of VHA-8∷GFP fluorescence (Fig. 2 D). We conclude that the V0 sector is required independently from the V1 sector for apical secretion of some cuticle components.

If the V0 sector has two distinct functions, it should be possible to recover vha-5 alleles that affect either its V0-specific secretion function or its V0+V1 proton-pump function. The vha-5 distribution and the aforementioned phenotypes indicate that reducing V0-specific function should affect cuticle secretion, whereas impairing proton pumping should affect the excretory canal responsible for osmoregulation. To identify such mutations, we used a plasmid rescue strategy, whereby we generated mutations by using PCR on a rescuing vha-5∷gfp construct, introducing them into vha-5(mc38)/+ animals, and recovering live homozygous vha-5(mc38) animals whenever possible (Fig. 3 A). We modified charged or large hydrophobic residues, as well as residues previously mutated in the yeast Vph1p (Leng et al., 1996, 1998). We generated 56 mutations (Fig. 3, B and C; and Fig. S2); 42 had no obvious phenotype by differential interference contrast (DIC) microscopy (Fig. 3 B, stars), and eight failed to rescue, indicating that those residues are essential for VHA-5 function (Fig. 3 B, white boxes). More interestingly, six substitutions rescued the vha-5(mc38)–induced lethality and affected the cuticle, the excretory canal, or both.

These six mutations defined three classes, which we will call “cuticle mutations” (L786S, E830Q, and V844F), “canal mutations” (W190A and R191A), or “mixed mutations” (W327A). First, animals carrying cuticle or mixed mutations were significantly shorter and dumpier than wild-type animals or animals carrying canal mutations (Fig. 4 A). This phenotype is frequently observed for mutations affecting cuticle components (McMahon et al., 2003). Western blot analysis using the VHA-5 antiserum detected similar amounts of mutant VHA-5∷GFP proteins (Fig. 4 B), implying that expression level differences do not explain phenotypic differences. Second, scanning electron microscopy (SEM) showed that adult alae were strongly affected in the former, but not the latter, animals (Fig. 4 C and Fig. S3 A). Third, confocal microscopy using the mutant VHA-5∷GFP as a marker and transmission electron microscopy (TEM) revealed that the excretory canal of animals with canal or mixed mutations, but not with cuticle mutations, was abnormal (Fig. 4, D and E; and Fig. S3, A and B). Their excretory canals had an increased section, often with multiple lumens, and 3–10 abnormal whorls per canal (see mc38; Ex[W190A] in Fig. 4 [D and E]). Strikingly, we observed similar phenotypes by knocking down V1 or V0 subunits by RNAi from L3 larval stage until adulthood (Fig. 5). Thus, we infer that the defects induced by canal mutations reflect an impairment of V0+V1 proton pumping, and that they are caused by loss-of-function rather than by gain-of-function mutations. Expansion of the excretory canal in these mutants might help to compensate for the decrease in proton-pumping efficiency. Surprisingly, animals with canal mutations did not show any proton-pumping defect in the epidermis. Possibly, proton pumping is preserved in this tissue because the “a” subunit VHA-7, which is expressed in the epidermis but not in the excretory cell (Oka et al., 2001; Pujol et al., 2001), compensates for the mutated VHA-5. In contrast, we infer that the cuticle defects induced by cuticle mutations reflect an impairment of the V0-specific secretion function, which would not be compensated by other “a” subunits, probably because they are not endowed with this specific function. The existence of various “a” subunits with possibly different functions in the epidermis is reminiscent of the difference observed in yeast between Vph1p and Stv1p (Kawasaki-Nishi et al., 2001). Lastly, we suggest that the mixed mutation W327A affects both V0+V1 and V0-specific functions. We note that cuticle mutations are located in the last transmembrane domain or in the COOH-terminal luminal tail, whereas canal mutations are in the NH₂-terminal cytoplasmic part, which is more likely to interact with the V1 sector (Nishi and Forgac, 2002). We conclude that the V0-specific and V0+V1 functions of VHA-5 are genetically separable.

If indeed the V0 sector is involved in secretion, vha-5 cuticle mutants should accumulate secretory organelles. At low magnification, TEM through the epidermis showed that animals carrying cuticle or mixed mutations contained significantly more and larger dense organelles than wild-type adults or animals carrying canal mutations (Fig. 6 A and Fig. S3 C). At higher magnification these organelles appeared as multivesicular bodies (MVBs; Fig. 6, B and C). MVBs are endosome-derived organelles containing 30–90 nm vesicles, which grow from early and recycling endosomes or from the trans-Golgi network and evolve into lysosomes or into secretory organelles (Raiborg et al., 2003). Hence, MVB accumulation may reflect either an endocytic/degradation pathway or a secretory defect. To distinguish between these two possibilities, we examined whether vha-5 mutants had normal lysosomes. In addition, we compared vha-5 defects to those induced by strong mutations in vps-27, rme-8, and cup-5, which are three essential genes acting at different steps along the endocytic route (Zhang et al., 2001; Treusch et al., 2004; Roudier et al., 2005). The rationale for this comparison is that if vha-5 cuticle mutations affect endocytosis, then vps-27, rme-8, or cup-5 mutations should induce cuticle phenotypes comparable to those of vha-5 mutants. As found in other systems (Luzio et al., 2001), we could observe in all vha-5 mutants intermediate late endosome–lysosome compartments corresponding to enlarged MVBs with multilamellar structures (called hybrid-MVBs in the next section) and normal lysosomes, suggesting that the endocytic/degradation pathway was not qualitatively affected (Fig. 6 C). In contrast, cup-5(ar465) mutants accumulated large MVBs in their epidermis that did not evolve into lamellar structures (Fig. 7 A). Furthermore, vps-27(ok579), rme-8(b1023), and cup-5(ar465) mutants had normal alae, unlike vha-5 cuticle mutants (Fig. 7, B and C). We conclude that cuticle mutations are unlikely to affect degradation, and, rather, they affect a secretory pathway.

To understand the relevance of MVB accumulation in vha-5 mutants, we reinvestigated secretion in the wild-type C. elegans epidermis. In hematopoietic cells, some MVBs release their vesicle content into the extracellular space and, thus, play a role in exocytosis, in addition to their well defined role in the endosomal pathway (de Gassart et al., 2004). The vesicles released by fusion of MVBs with the plasma membrane were originally called exosomes in antigen-presenting cells. In support of the notion that secretion in the C. elegans epidermis involves exosomes, we observed small light MVBs containing 50–100-nm vesicles just beneath the apical plasma membrane. Moreover, we occasionally saw vesicles immediately above the plasma membrane in the inner cuticular layer, strongly suggesting that a MVB had released its intralumenal vesicles (Fig. 6 B). These MVBs were always found in the vicinity of epidermal apical membrane stacks, a structure whose role is so far unknown (White, 1988). In contrast, we rarely observed similar MVBs adjacent to the plasma membrane in cuticle mutants, or they were darker (Fig. 6 C). These data suggest that the MVB-limiting membrane can fuse with the apical membrane to release exosomes in the cuticle, and show that this process is impaired in vha-5 cuticle mutants. It raises the possibility that the V0 sector is critical for MVB fusion with the apical membrane during exosome release.

An important expectation of the cuticle defects described so far is that we should be able to identify cuticular proteins whose secretion depends on VHA-5 activity. Cuticle proteins include collagens and Hedgehog-related peptides (McMahon et al., 2003; Zugasti et al., 2005). We found that the collagen DPY-7 was efficiently secreted in vha-5(mc38)–null animals, in animals carrying cuticle mutations, as well as in che-14(mc35) mutants (Fig. S4). We turned our attention onto Hedgehog-related peptides, which appeared as good candidates for three reasons. First, vha-5 alae defects partially resemble those observed in che-14 mutants. Second, CHE-14 is homologous to Dispatched, which is a protein required for Hedgehog release (Burke et al., 1999; Michaux et al., 2000). Third, despite the absence of a Hedgehog homologue in C. elegans, its genome contains several Hedgehog-related peptides required to generate a normal cuticle, although their precise roles remain unknown (Aspock et al., 1999; Zugasti et al., 2005).

We tagged with GFP the secreted domain of the Hedgehog-related peptides WRT-2 and -8 (Fig. 8 A and Fig. S3 E), which are expressed in the epidermis (Aspock et al., 1999). We found that animals bearing cuticle mutations, but not canal mutations, accumulated VHA-5∷RFP and WRT-2∷GFP or -8∷GFP in discrete entities in their epidermis (Fig. 8, B and C; and Fig. S3, D and F). These entities most likely correspond to the dense and hybrid MVBs (Fig. 6, C and D) because VHA-5∷RFP also colocalized (Fig. 8 C) with the MVB marker VPS-27∷GFP (Roudier et al., 2005). Moreover, both VHA-5 antiserum and a GFP antiserum targeting WRT-2∷GFP decorated the MVBs of cuticle mutants (Fig. 8 D and Fig. S5 B). Last, in wild-type nontransgenic animals, in addition to membrane stacks (Fig. 1 G), VHA-5 was found at the MVB-limiting membrane, in intralumenal vesicles, and in the cuticle (Fig. 8 E and Fig. S5 A), suggesting that it could act at different steps in the secretion of vesicle (see Discussion). Importantly, the VHA-5 protein with the substitutions L786S (Fig. 8 B) or E830Q (not depicted) could reach the plasma membrane in heterozygous vha-5(mc38)/+ animals, which strongly suggests that their intracellular retention in homozygous mc38 animals is caused by the loss of a trafficking function rather than by misfolding. Consistently, the WRT-2/8 proteins were not retained intracellularly either in heterozygous vha-5(mc38)/+ animals, despite the presence of the L786S (Fig. 8 B), or in E830Q transgenes (not depicted). These results indicate that the V0 sector plays a key role in a specific apical secretion pathway that is taken on by Hedgehog-related proteins, but not by collagens.

Whereas basolateral secretion is known to depend on the activity of specific complexes (AP-1B and the exocyst), no such complex has been implicated in apical secretion. In addition, although recycling endosomes appear to play a central role in basolateral secretion, their importance in apical secretion is still under active investigation. Our characterization of mutations affecting the V-ATPase “a” subunit VHA-5 sheds new light on the apical biosynthetic secretory pathway. We could observe the fusion of MVBs with the apical plasma membrane in wild-type animals, and the subsequent release of exosomes. In contrast, we found that some VHA-5 mutations accumulate MVBs in their epidermis and prevent the secretion of Hedgehog-related proteins. Thus, we propose a model whereby apical secretion of Hedgehog-related proteins involves their incorporation into the intralumenal vesicles of MVBs, and their subsequent release when MVBs fuse with the apical plasma membrane (Fig. 9). Furthermore, we suggest that the V0 sector of the V-ATPase plays a key role in this process.

We can envision two scenarios for the role of the V0 sector. First, vha-5 mutations affecting cuticle formation could decrease V-ATPase proton pumping along the biosynthetic secretory route to indirectly impair secretion. Consistent with this possibility, mutations in S. cerevisiae Vph1p (L746S, E789Q, and V803F) corresponding to the cuticle mutations (L786S, E830Q, and V844F, respectively) strongly reduce, but do not eliminate, proton pumping (Leng et al., 1996, 1998). We think, however, that this possibility is unlikely because in yeast proton uptake is not limiting for fusion of Vph1p-defective vacuoles (Bayer et al., 2003). Alternatively, these mutations could uncover a direct role of the V0 sector in apical exocytosis, independently of the V1 sector. Four observations strongly support this notion. First, inhibition of V1 subunits did not affect cuticle formation, although it strongly impaired osmoregulation and endocytosis. Second, we obtained specific V0 mutations inducing a strong cuticular phenotype without any apparent pumping-related defect in the excretory cell. Third, these mutations selectively affected the secretion of Hedgehog-related proteins, but not collagens. Last, strong mutations in well characterized genes blocking different steps of fluid-phase endocytosis did not affect cuticle structure. The specific accumulation of dense and hybrid MVBs, but not of light MVBs, in cuticle mutants, as well as the detection of significant amounts of WRT-2∷GFP within these MVBs, suggests that VHA-5 is neither required to form MVBs nor to load them with cargo proteins. Instead, the presence of VHA-5 at the MVB-limiting and apical membranes suggests that the V0 sector could drive the fusion of MVBs with the apical membrane via the formation of V0 sectors transcomplexes between both membranes (Fig. 9), as suggested in yeast vacuole fusion and at the D. melanogaster synapse (Bayer et al., 2003; Hiesinger et al., 2005; Peters et al., 2001). In vertebrates, 300-nm procollagen-I rod bundles assemble in the ER and travel through the Golgi lumen (Bonfanti et al., 1998). Assuming that worm collagens are secreted this way, they would not fit into exosomes, suggesting that there are at least two distinct secretion pathways in the epidermis, one involving the V0 sector and another followed by collagens.

What could explain a common requirement for the V0 sector during the C. elegans apical exocytosis, yeast vacuole fusion, and D. melanogaster synaptic transmission? The prevailing view is that a SNAREpin complex initiates membrane fusion once a vesicle has been docked to a proper membrane (Chernomordik and Kozlov, 2003; Jahn et al., 2003). Although the V0 sector is thought to act downstream of SNAREs in yeast and D. melanogaster (Peters et al., 2001; Bayer et al., 2003; Hiesinger et al., 2005), we cannot exclude that it also acts in parallel to SNAREs, at least in C. elegans, to dock MVBs. Another possibility is that V0 transcomplexes initiate the formation of a protein pore, as initially suggested in yeast (Peters et al., 2001). On the other hand, expansion of the fusion pore is considered as the limiting step in membrane fusion, and might require additional catalysts in vivo (Chernomordik and Kozlov, 2003). Such a role could be fulfilled by the V0 sector, either to overcome constraints caused by the big size and/or the specific lipoproteic content of C. elegans epidermal MVBs and yeast vacuoles, or to allow rapid synaptic transmission in D. melanogaster neurons (Hiesinger et al., 2005).

Irrespective of the precise role of the V0 sector in membrane fusion, our findings bear potentially important implications. First, morphogens such as Wingless and Hedgehog in D. melanogaster, or Sonic-Hedgehog at the mouse node, might be secreted through a similar pathway because their secretion involves particles possibly related to exosomes (Greco et al., 2001; Gallet et al., 2003; Panakova et al., 2005; Tanaka et al., 2005). A major objective will be to determine whether CHE-14 and Dispatched act in the aforementioned secretory pathway, and, if so, at which step. Second, several other cell types, such as antigen-presenting cells, reticulocytes, and some epithelial cells, can release exosomes (de Gassart et al., 2004), which might thus also require the V0 sector for their secretion. In particular, the V0 sector might be directly associated with the transmission of some infectious diseases because viruses, such as HIV and the prion protein, can be disseminated through MVBs and the exosome-releasing machinery (de Gassart et al., 2004; Fevrier et al., 2005). Likewise, the aforementioned secretory pathway could be involved in innate immunity because expression of the Hedgehog-related peptide GRD-3 is induced in C. elegans upon bacterial infection (Couillault et al., 2004). Third, our findings raise the issue of the origin of the MVBs. Interestingly, the apical recycling endosomes have been recognized to play an important role in biosynthetic secretory pathways (Hoekstra et al., 2004). Future studies should reveal whether the secretory MVBs that we described could originate from this compartment.

In conclusion, our work shows that trafficking to the apical membrane of at least some lipid-modified proteins involves specific protein complexes (the V-ATPase V0 sector), much as trafficking to the basolateral membrane, and predicts a key role for MVBs in apical exocytosis.

Worms were grown at 20°C (unless noted otherwise; Brenner, 1974). The identification of mc37 and mc38 as vha-5/F35H10.4-null mutations will be described elsewhere (the gene affected by mc37 and mc38 was initially named rdy-1; see www.wormbase.org). Marker and alleles used were as follows: cup-5(ar465) (Treusch et al., 2004), rme-8(b1023ts), bIs1[vit-2∷gfp; rol-6(su1006)] (Zhang et al., 2001), and vps-27(ok579)/unc-24(e138) dpy-20(e1282) (Roudier et al., 2005). The L1 alae phenotype of rme-8(b1023ts) was scored by allowing adults to lay eggs for 1 h at 15°C, and then transferring embryos to 25°C after egg laying. RNAi was performed using the following bacterial clones from the Wellcome–Medical Research Council library (Kamath et al., 2003): vha-1, III-5A20; vha-4, II-5J16; vha-5, IV-4O13; vha-8, IV-3I08; and vha-13, V-9O06 (www.wormbase.org). To score L1 larvae, RNAi was induced in L4 larvae; to score adults, RNAi was induced in larvae at the L2–L3 molt. Feeding plasmids were retransformed into fresh HT115 (DE3) bacteria, selecting for tetracycline and ampicillin resistance.

Cloning of the vha-5 coding sequence with a 2.8-kb promoter upstream of the GFP coding sequence in the pPD95.75 vector (Fire kit) generated a rescuing vha-5∷gfp construct. A vha-5∷mrfp construct was obtained by replacing the GFP with the monomeric red fluorescent protein (mRFP) coding sequence in the vha-5∷gfp construct. A vha-8∷yfp construct was obtained by cloning the vha-8 coding sequence and a 3-kb promoter upstream of the YFP coding sequence in the pPD136.64 vector (Fire kit). To generate wrt-2∷gfp and wrt-8∷gfp constructs, we cloned wrt-2 and -8 genomic DNA with their 5′ and 3′ regulatory sequences into pBSKII-derived plasmids. The GFP coding sequence was inserted in a nonconserved region of the predicted secreted peptide (Fig. 8 A and Fig. S3 D).

The vha-5∷gfp construct was mutated using the QuikChange Site-Directed Mutagenesis kit (Stratagene). Each desired mutation, and the entire vha-5 coding sequence of most important plasmids, was verified. Mutant plasmids were microinjected in heterozygous vha-5(mc38)/unc-5(e53) at 3 ng/μl, along with the marker pRF4 [rol-6(su1006)] at 100 ng/μl, wrt-2∷gfp or wrt-8∷gfp constructs at 30 ng/μl (when relevant), and pBSKII plasmid at up to 200 ng/μl. Absence of unc-5 animals in the progeny was used as a criterion for rescue. mRFP versions for the mutations W190A, R191A, W327A, L786S, E830Q, and V844F were obtained from GFP derivatives without PCR amplification and resulted in similar phenotypes. At least two independent extrachromosomal lines were initially examined for each mutation. More detailed analysis was performed on a representative line.

VHA-5 polyclonal antibodies were raised in rabbits injected with a purified GST fusion protein containing VHA-5 residues I29–M302, which was obtained by cloning a fragment amplified from the cDNA yk458f4 (a gift from Y. Kohara, National Institute of Genetics, Mishima, Japan) into the vector pGEX-2T. Total worm extracts were solubilized in 8 M urea/2% SDS by sonication, before 8% acrylamide gel electrophoresis and Western blotting. VHA-5 antiserum was used at 1:2,000, the actin monoclonal antibody (act-2D7; Institut de Génétique et de Biologie Moléculaire et Cellulaire collection) at 1:4,000; primary antibodies were revealed with a SuperSignal kit (Pierce Chemical Co.). Immunofluorescence was performed using the VHA-5 antiserum at 1:1,000 dilution and the DPY-7 monoclonal antibody (gift from I. Johnstone, Wellcome Centre for Molecular Parasitology, Glasgow, Scotland) at a 1:50 dilution.

Animals were mounted on 4% agarose pads in M9, anaesthetized with 0.2% tricaine/0.02% tetramisole in M9. For DIC imaging, we used a microscope (Axioplan; Carl Zeiss MicroImaging, Inc.) coupled to a camera (CoolSNAP; Roper Scientific) under a 100× objective (PlanApo; Leica). For Fig. 4 A, we took at least 40 pictures of adult worms per strain and used ImageJ (National Institutes of Health) to measure the distance between the rectum and the grinder. Confocal images were captured on a confocal microscope (SP2-AOBS; Leica), scanning every 122 nm for XZ sections through a 100× objective with a 2.15× zoom (Fig. 8, B and C; and Fig. S3 D) or a 4× zoom (Fig. 4 E and Fig. 5). Images were processed with the Tcstk software (McMahon et al., 2001) and edited using Photoshop 7.0 (Adobe). Microscopes were in an air-conditioned room (20–21°C).

L4 larvae were transferred onto fresh plates for 24 ± 2 h at 20°C before fixation. For TEM, but not for SEM, animals were sectioned and fixed for at least 24 h in 2.5% glutaraldehyde/2% paraformaldehyde/0.1 M sodium cacodylate, pH 7.2, at 4°C, and then postfixed for 5 h with 2% osmium tetroxide in the same buffer at 4°C, dehydrated in graded alcohol/water mixes, and embedded in Epon. Ultrathin 70-nm sections were contrasted with uranyl acetate and lead citrate. Sections were observed with a microscope (CM12; Philips) operating at 60 kV. Quantification of the excretory canal section area was obtained using the Metamorph software after scanning images were captured at a 17,000× magnification. Quantification of MVBs was performed on 3,600×-magnified images. Quantification of the mean area occupied by organelles (Fig. 6 D) was obtained using ImageJ and dividing the total surface of each organelle subtype by the cytoplasmic surface of the hyp7 epidermis section. At least four animals per mutant strain were examined, and more than nine pictures per animal from different ultrathin sections were analyzed. For SEM, animals were postfixed for 1 h with 2% osmium tetroxide at 4°C, dehydrated, and critical point dried in hexamethyldisilazane. Fixed animals were mounted on stubs, coated with palladium, and observed through a microscope (XL20; Philips). At least 20 animals per strain were analyzed.

Adult worms were frozen with a high pressure freezing apparatus (EMPACT-2; Leica) in 20% BSA/M9 medium. Cryosubstitution was conducted as in Muller-Reichert et al. (2003). Ultrathin sections were collected on formvar-carbon–coated copper grids and processed for immunogold labeling. Blocking was performed in PBS/glycine 150 mM, and then in PBS/1% BSA/0.1% Cold Water Fish Skin Gelatin (CWFSG; Aurion) for 30 min. Rabbit anti–VHA-5 at 1:1,000 and rabbit anti-GFP at 1:500 (ab6556; AbCam) were incubated for 1 h in PBS/0.1% CWFSG. 10 nm protein A–coupled gold beads (1:50; University Medical Center, Utrecht, Netherlands) were incubated for 1 h in PBS/0.1% CWFSG. Postfixation was achieved in 2.5% glutaraldehyde, contrasted by uranyl acetate/lead citrate. Images were acquired at 60 kV on a microscope (Morgagni; FEI) with a charge-coupled device camera (Megaview III; Soft Imaging System).

Fig. S1 provides a control for Fig. 2 B, showing that RNAi against vha-5 was efficient. In addition, it presents the yolk endocytosis defects induced by the loss of V-ATPase activity (yolk proteins are produced by the intestine and are essential for embryonic development); it suggests that RNAi treatment against V0 and V1 subunits was equally effective, and contributes to establish that alae differences described in Fig. 2 are meaningful. Fig. S2 summarizes the main phenotypes observed in vha-5(mc38) animals carrying transgenes with the mutations shown in Fig. 3. Fig. S3 presents the excretory canal, cuticle and MVB phenotypes induced by the mutations R191A, W327A and V844A, which are discussed but not illustrated in the main text, and shows that WRT-8∷GFP accumulates in mc38; Ex[vha-5(E830Q)∷rfp] animals; it should be viewed along with Figs. 4, 6 and 8. Fig. S4 shows that secretion of the collagen DPY-7 is not affected by vha-5 or che-14 mutations. Fig. S5 provides larger pictures and controls for the immunogold experiments shown in Fig. 8.

We are grateful to Guillaume Belliard, Grégoire Michaux, and Emmanuel Sotirakis for their help in isolating vha-5 mutations, and to Satis Sookhareea for expert assistance in producing VHA-5 antiserum. We thank the Institut de Génétique et de Biologie Moléculaire et Cellulaire Imaging Facility for help with confocal and EM, and Patrick Schultz for access to the EMPACT2. We thank Andy Fire, Yuji Kohara, Johnny Fares, Barth Grant, Iain Johnstone, Renaud Legouis, and the CGC for reagents. We thank Jean-Louis Bessereau, Thierry Galli, Pierre Gönczy, Nathalie Pujol, and Christelle Gally for critical reading of the manuscript.

S. Liégeois and A. Benedetto were supported by fellowships from the Ministère de la Recherche, and from the FRM. This work was supported by funds from the Centre National de la Recherche Scientifique, Institut National de la Santé et de la Recherche Médicale, and by a grant from the Ministère de la Recherche (programme ACI).