Structure, function and regulation of the coated vesicle V-ATPase

Clathrin-coated vesicles play a critical role in both receptor-mediated endocytosis and intracellular membrane traffic (for a review, see Forgac, 1989). As illustrated in Fig. 1, during endocytosis ligand-receptor complexes become concentrated in clathrin-coated pits, which pinch off from the cell surface to form clathrin-coated vesicles. Among the ligands internalized by this pathway are low-density lipoprotein, transferrin, insulin and epidermal growth factor. Rapid uncoating of the vesicle and subsequent membrane fusion events lead to delivery of the ligand-receptor complexes to CURL (compartment of uncoupling of receptor and ligand, Geuze et al. 1983). Acidification of CURL by the V-ATPases causes dissociation of ligand-receptor complexes, allowing recycling of unoccupied receptors to the cell surface and transport of the dissociated ligands to lysosomes for degradation.

Clathrin-coated vesicles play a similar role in the intracellular targeting of newly synthesized lysosomal enzymes from the Golgi apparatus to lysosomes (Fig. 1). By virtue of the mannose 6-phosphate recognition marker, lysosomal enzymes bind to the mannose 6-phosphate receptor in the trans-Golgi, where they become concentrated in clathrin-coated pits that pinch off to form clathrin-coated vesicles. Vesicle uncoating and delivery to an acidic compartment results in dissociation and recycling of the receptors to the trans-Golgi and targeting of the lysosomal enzymes to lysosomes.

Clathrin-coated vesicles also play a role in membrane retrieval at synaptic terminals following synaptic vesicle fusion (Pfeffer and Kelly, 1985). Because synaptic vesicles contain a V-ATPase which provides the electrochemical driving force for neurotransmitter uptake, it is likely that coated vesicles retrieve this pump from the plasma membrane and deliver it to synaptic vesicles.

The coated vesicle V-ATPase transports protons from the cytoplasm across the vesicle membrane uncoupled to the countertransport of other cations (Forgac and Cantley, 1984). This pump is therefore electrogenic and requires the movement of another ion for proton transport to occur. As illustrated in Fig. 1, the membrane potential generated by the proton pump is dissipated by the activity of a parallel chloride channel (Arai et al. 1989; Xie et al. 1983). We have recently shown that this chloride channel is controlled by protein-kinase-A-dependent phosphorylation (Mulberg et al. 1991). Because acidification is dependent upon chloride conductance, modulation of channel activity represents an important mechanism for controlling vacuolar acidification (see below).

The coated vesicle H⁺-ATPase, like other V-ATPases, is sensitive to micromolar concentrations of N-ethylmaleimide (NEM) and 7-chloro-4-nitrobenz-2-oxa-l,3-diazole (NBD-CI) (Arai et al. 1987b). Protection of activity by ATP suggests that both reagents react with groups present at the catalytic site. We have recently identified the cysteine residue responsible for NEM-sensitivity of the coated vesicle V-ATPase (Feng and Forgac, 1992) (see below). In addition, protection of the NEM-reactive cysteine by NBD-CI indicates that both reagents inhibit activity by reacting with the same residue.

Unlike the P-ATPases (Pedersen and Carafoli, 1987), the coated vesicle V-ATPase is resistant to vanadate (Forgac et al. 1983) and does not form a phosphorylated intermediate during turnover (Forgac and Cantley, 1984). In addition, the coated vesicle V-ATPase is resistant to the F-ATPase inhibitors oligomycin and aurovertin (Forgac et al. 1983), but is sensitive to dicyclohexylcarbodiimide (DCCD), a carboxyl reagent which inhibits proton flux (Arai et al. 1987a). As with other V-ATPases (E. J. Bowman et al. 1988a), the coated vesicle H⁺-ATPase is sensitive to nanomolar concentrations of bafilomycin (M. Myers and M. Forgac, unpublished observations).

Our current model of the coated vesicle V-ATPase (Adachi et al. 1990b), based on the structural data described below, is shown in Fig. 2. The V-ATPase complex is divided into a peripheral V₁ domain containing subunits of relative molecular mass 73 (A), 58 (fi), 40, 34 and 33 ×10³ and an integral V_o domain composed of the 100, 38, 19 and 17 (c) × 10³M_r subunits. The entire complex, which has a relative molecular mass of 750000, contains three copies each of the A and B subunits, six copies of the c subunit and one copy of the remaining polypeptides. The catalytic nucleotide binding sites are located on the A subunits, whereas the B subunits are also believed to contribute to ATP binding. The nucleotide binding domain is connected to the membrane via the 40, 34 and 33 ×10³Mr ‘accessory’ subunits. The V_o domain is responsible for proton translocation whereas the 17×10³Mr c subunit is responsible for the DCCD-sensitivity of proton transport. The 100×10³Mr subunit is a transmembrane glycoprotein of as yet unknown function. The V-ATPases thus closely resemble the F-ATPases of mitochondria, chloroplasts and bacteria, with which they share sequence homology. Unlike the F₁ and F_o domains of the F-ATPases, however, the Vi and V_o domains do not function independently. The possible significance of this finding to the mechanisms regulating the activity of the V-ATPases is discussed below.

The coated vesicle V-ATPase is composed of nine subunits of relative molecular mass 100, 73 (A), 58 (B), 40, 38, 34, 33, 19 and 17 (c) ×10³, which are immunoprecipitated as a single macromolecular complex by monoclonal antibodies directed against the native enzyme (Arai et al. 1987b). Using quantitative amino acid analysis, we have shown that these subunits are present in a stoichiometry of 1OO₁A₃B340₁38₁34₁33₁ 19₁17₆ (Arai et al.(1988). A similar subunit composition is observed for the V-ATPases of kidney (Gluck and Caldwell, 1987), chromaffin granules (Moriyama and Nelson, 1989), Neurospora crassa (Bowman et al. 1989), yeast (Kane et al. 1989) and plants (Parry et al. 1989; Lai et al. 1988).

The V-ATPase complex is divided into two structural domains. The peripheral V₁ domain, which has a relative molecular mass of 500×10³, has the structure A₃B₃40₁34₁33₁ (Arai et al. 1988) and is removed from the membrane by chaotropic agents such as KI and KNO3 (Arai et al. 1989; Adachi et al. 1990b). Dissociation of V| is activated by submicromolar concentrations of ATP (Arai et al. 1989). The V₁ subunits are released from the membrane as monomers which are capable of reassembling into various subcomplexes (Puopolo and Forgac, 1990; Puopolo et al. 1992b), as discussed below. The integral V_o domain contains four subunits in a stoichiometry of 100₁38₁ 19_1C6 (Arai et al. 1988). These polypeptides remain assembled as a complex of 250×10³M_r following removal of V1 and detergent solubilization (Zhang et al. 1992).

Topographical studies indicate that the V₁ domain is oriented towards the cytoplasmic side of the membrane (Arai et al. 1988; Adachi el al. 1990a). Thus, all the V₁ subunits are labeled by membrane-impermeant reagents in intact coated vesicles where only the cytoplasmic surface is exposed (Arai et al. 1988). In addition, all the V₁ subunits are cleaved by trypsin in reconstituted vesicles containing the purified V-ATPase oriented with the cytoplasmic surface facing out (Adachi et al. 1990a). Of the V_o subunits, the 100 and 38×10³M_r subunits are exposed on the cytoplasmic side of the membrane based on their labeling by impermeant reagents and their sensitivity to trypsin added from the cytoplasmic surface (Arai et al. 1988; Adachi et al. 1990a). The 100, 19 and 17×10³Mr subunits also possess luminal domains since they show a significant increase in labeling by impermeant reagents upon detergent permeabilization of intact coated vesicles (Arai et al. 1988). Lectin binding studies indicate that the 100×10³M_r subunit possesses covalently bound carbohydrate terminating in sialic acid (Adachi et al. 1990b). Since sialic acid is found exclusively on the noncytoplasmic surface of the membrane, this result provides further evidence that the 100x10³M_r subunit is a transmembrane glycoprotein (Adachi et al. 1990b).

Both the c and the 19×10³M_r subunits are highly hydrophobic proteins extensively buried in the bilayer, based upon their extraction by organic solvents (Arai et al. 1987a) and their high proportion of nonpolar amino acids (Arai et al. 1988). Heavy labeling of the c subunit with the photoactivated hydrophobic reagent [¹²⁵I]TID confirms that a large proportion of this protein is in contact with the lipid bilayer (Arai et al. 1988). Interestingly, [¹²⁵I]TID does not label the 38 and 19×10³Mr subunits while a small amount of labeling of the 100x10³M_r subunit is observed (Arai et al. 1988). This result is consistent with the transmembrane orientation of the 100×10³M_r-polypeptide and suggests that the 19×10³M_r subunit is shielded from contact with the lipid bilayer through interaction with other membrane-spanning domains, probably contributed by the c and 100×10³Mr subunits. The absence of [¹²⁵]TID labeling of the 38×10³M_r polypeptide is also consistent with the sequence of the corresponding chromaffin granule protein, which shows no putative transmembrane helices (Wang et al. 1988). The 38×10³M_r subunit therefore appears to be a peripheral polypeptide, which remains tightly bound through protein-protein contacts to the integral V_o domain.

Crosslinking studies of the coated vesicle V-ATPase using the reversible crosslinking reagent 3,3′-dithiobis(sulfosuccinimidylpropionate) (DTSSP) have revealed extensive contact between the A and B subunits as well as between the c subunit and the 40, 34 and 33×10³M_r subunits (Adachi et al. 1990b). These results have led to the model shown in Fig. 2 where the A and B subunits form a hexameric complex attached to the V_o sector via a bridge formed by the accessory subunits.

As mentioned above, dissociation of the V₁ subunits from the membrane by treatment with KI and ATP results in their initial release as monomers. Removal of the chaotropic agents in the presence of membranes containing V_o results in the reassembly of functional V₁ V₀ complexes (Puopolo and Forgac, 1990). If KI and ATP are removed in the absence of V_o and reassembly is monitored by sedimentation on glycerol density gradients, the following results are obtained. A subcomplex of approximately 500×10³M_r is formed which has the structure A₃B₃34₁33₁ (Puopolo et al. 1992b). This subcomplex completely lacks the 40×10³M_r subunit, which appears in the ‘monomeric’ fraction, together with approximately 50% of the A, B and SS×10³M_r subunits. Interestingly, the monomeric fraction is almost completely devoid of the 34×10³M_r subunit (Puopolo et al. 1992b).

Evidence for interaction between the 40 and 33x10³M_r subunits derives from immunoprecipitation experiments using the monoclonal antibody 1C-11G (Puopolo et al. 1992b). This antibody immunoprecipitates both the 40 and 33x10³M_r subunits from the ‘monomeric’ fraction, suggesting that these two subunits are complexed. That the antibody is directed against an epitope expressed on the native 33×10³M_r subunit is indicated by its ability to immunoprecipitate the V₁ (−40×10³Mr) subcomplex (Puopolo et al. 1992b). Thus, during in vitro reassembly, the 33×10³M_r subunit assembles with either the 40×10³M_r polypeptide or with the remaining V1 subunits to form the V₁(−40×10³M_r) subcomplex.

The Vo subunits, unlike the Vi subunits, remain together as a complex of 250×10³M_r following removal of Vi with KI and ATP (Zhang et al. 1992). The V_o subunits are present in the same stoichiometry in the free V_o domain as in Vi V_o and possess many of the same properties. Thus, the c subunit in the free V_o domain is extracted by organic solvents and is labeled by [^l4C]DCCD. In addition, a monoclonal antibody has been isolated which recognizes the 100x10³M_r subunit in both complexes. Both the 100 and 38×10³M_r subunits show the same tryptic cleavage pattern in the free V_o complex as in V|V₀, although the sensitivity of these polypeptides to proteolysis is increased, suggesting that they are more exposed to the aqueous phase on removal of V₁ (Zhang et al. 1992).

The existence of multiple nucleotide binding sites on the coated vesicle V-ATPase is suggested by several lines of evidence. First, the dependence of ATP hydrolysis on the concentration of ATP indicates the existence of two ATP binding sites with Kd values of 80 and 800μmoll⁻¹ (Arai et al. 1989). Saturation of the higher-affinity site leads to increased proton transport, whereas occupation of the low-affinity site appears to decrease proton transport, resulting in a decreased H⁺/ATP stoichiometry. ATP binding to a still higher affinity site (Kd 200 nmol 1⁻¹) activates dissociation of the peripheral Vi domain from the integral V_o domain (Arai et al. 1989), suggesting a decrease in stability of the V-ATPase complex.

Further evidence for the existence of multiple nucleotide binding sites has come from studies of the nucleotide analog 2′,3′-O-(2,4,6-trinitrophenyl)adenosine 5′-triphosphate (TNP-ATP) (Adachi et al. 1990a). TNP-ATP inhibits activity of the purified V-ATPase with Kd values of 50 nmol I⁻¹ and 3 μmol 1⁻¹ and protects both the A and B subunits from tryptic cleavage with a Kd of l–2mmoll⁻¹ (Adachi et al. 1990a). Finally, the coated vesicle V-ATPase contains multiple copies of both the A and B subunits which have been shown to possess ATP binding sites either by direct chemical labeling or by sequence homology (see below).

Evidence that the catalytic nucleotide binding site is located on the A subunit comes from its labeling by [³H]NEM and [^I4C]NBD-C1 in an ATP-protectable manner (Arai et al. 1987b). Similar results have been obtained with V-ATPases from a variety of sources (Forgac, 1989). Labeling of the B subunit of the plant V-ATPase using an ATP analog has also been reported (Manolson et al. 1985). Moreover, as first demonstrated by Zimniak et al. (1988), E. J. Bowman et al. (1988,b) and B. J. Bowman et al. (1988), the A and B subunits from plant and Neurospora crassa vacuoles are evolutionary related both to each other and to the α andβ subunits of the F-ATPases. Extensive data from chemical modification studies and mutational analysis indicate that both the α and β subunits are involved in nucleotide binding, with the catalytic site located on the βsubunit (Penefsky and Cross, 1991; Ysern et al. 1988; Senior, 1988). The greatest sequence conservation between the V-ATPase and F-ATPase subunits occurs in regions which are critical for nucleotide binding (Zimniak et al. 1988; Penefsky and Cross, 1991). Thus, the nucleotide binding subunits of the V-ATPases and F-ATPases have been derived from a common evolutionary ancestor.

Like the V-ATPases from other sources, the A and B subunits of the coated vesicle H⁺-ATPase show similar sequence homology (Puopolo et al. 1991, 1992a; Sudhof et al.1989) The bovine A subunit appears to be encoded by a single gene, which gives rise to a single transcript in all tissues tested (Puopolo et al. 1991). The bovine B subunit, in contrast, is encoded by at least two and possibly three, genes, which give rise to multiple transcripts in a tissue-specific manner (Puopolo et al. 1992a). Expression of both a 3.2 and 2.0 kb mRNA can be detected in all tissues examined except brain, where expression of only the 3.2 kb message is detected. Southern analysis is also consistent with the existence of multiple genes encoding the B subunit (Puopolo et al. 1992a).

We have recently identified the cysteine residue responsible for the sensitivity of the V-ATPases to sulfhydryl reagents (Feng and Forgac, 1992). We demonstrated that, in addition to NEM, the coated vesicle V-ATPase is also inhibited by cystine in an ATP-protectable manner. Unlike NEM, however, inhibition by cystine is reversible upon treatment with reducing agents, such as dithiothreitol (DTT). That cystine inhibits activity by forming a disulfide bond with the same cysteine residue that reacts with NEM was demonstrated by the fact that extensive treatment of the cystine-reacted enzyme with NEM followed by treatment with DTT restores activity of the coated vesicle V-ATPase. We then took advantage of this protection of the NEM-reactive cysteine by cystine to label this residue selectively. The catalytic cysteine was first protected with cystine, the enzyme was reacted extensively with NEM, the disulfide bond was reduced with DTT and the catalytic cysteine residue was then labeled with fluorescein maleimide. Under these conditions, fluorescein maleimide labels only the 73×10³M_r A subunit. Proteolytic cleavage of the labeled A subunit gives rise to a single 3.9×10^3xM_r fragment, which is labeled with fluorescein maleimide on cyteine residue 254 (Feng and Forgac, 1992). This residue is conserved as cysteine in all the V-ATPase A subunit sequences obtained thus far (Puopolo et al. 1991) and is located in the Walker consensus ‘A’ sequence GXGKTV (Walker et al. 1985). Moreover, the corresponding residue is valine in the F-ATPases and serine in the archaebacterial H⁺-ATPases (Penefsky and Cross, 1991), consistent with the insensitivity of the latter two classes to sulfhydryl reagents. As discussed above, we have demonstrated that this same cysteine residue is responsible for the sensitivity of the V-ATPases to NBD-C1. Based upon identification of this cyteine residue at the catalytic site of the A subunit, the existence of highly conserved consensus sequences involved in nucleotide binding (Walker et al. 1985) and the structures proposed for the ATP binding site of F₁ (Penefsky and Cross, 1991; Garboczi et al. 1988; Duncan et al. 1986) and adenylate kinase (Kim et al. 1990), we propose the model for the structure of the catalytic site of the V-ATPases shown in Fig. 3. Much additional work will be required to determine how accurately this model describes the true structure.

The 17×10³M_rc subunit is responsible for the DCCD-sensitivity of proton transport by the coated vesicle V-ATPase (Arai et al. 1987a). Reaction of the c subunit with DCCD results in complete loss of proton transport activity, although ATP hydrolysis is only inhibited if the H⁺-ATPase is embedded in the membrane. Comparison of the stoichiometry of DCCD labeling with the degree of inhibition suggests that complete blockage of proton conduction occurs after reaction of only one-sixth of the c subunits with DCCD (Arai et al. 1987a). As with the c subunit of F_o, this result suggests that formation of a proton channel requires the cooperative interaction of six copies of the c subunit, in agreement with the measured subunit stoichiometry (Arai et al. 1988).

Cloning and sequence analysis of the c subunit of the chromaffin granule V-ATPase (Mandel et al. 1988) has confirmed that this polypeptide is the vacuolar counterpart of the Fo c subunit, which is present in a stoichiometry of 10–12 copies per F_o complex (Foster and Fillingame, 1982). Despite being twice the size of the F_o protein, the vacuolar c subunit has only a single buried aspartate (the likely site of reaction with DCCD), suggesting that the number of transmembrane helices contributed by the c subunit has been conserved while the number of buried carboxyl groups has been reduced by a factor of two. This observation has suggested a mechanism in which the H⁺/ATP stoichiometry of the V-type and F-type H⁺-ATPases was altered during the course of their evolution (Cross and Taiz, 1990).

One report has suggested that the c subunit alone is sufficient to form a proton channel (Sun et al. 1987). Based on the homology between the vacuolar and F_oc subunits (Mandel et al. 1988) and on the extensive genetic and biochemical evidence indicating that the c subunit of F_o is not sufficient to form a proton pore (Cain and Simoni, 1986; Schneider and Altendorf, 1985), it seems most likely that additional subunits are also required for proton conduction through V_o. One possible candidate for the vacuolar counterpart to the a subunit, which is critical for proton conduction through F_o, is the 19×10³Mr polypeptide. Like the a subunit, this protein is present as one copy per complex and is extensively buried in the bilayer (Arai et al. 1988).

We have investigated the proton conductance properties of the V_o domain of the coated vesicle V-ATPase (Zhang et al. 1992). Proton conductance was measured both in native membranes from which V₁ had been removed and in reconstituted vesicles containing the isolated V_o complex by uptake of the fluorescent dye 9-amino-6-chloro-2-methoxyacridine (ACMA) in response to a K⁺-and valinomycin-induced membrane potential. In neither case was any DCCD-inhibitable proton conduction detected, although proton conduction could readily be detected upon addition of the proton ionophore carbonyl cyanide p-chlorophenylhydrazone (CCCP) and although the V_o domain was still competent to reassemble with the V| subunits to give a functional V-ATPase. These results suggest that one or more of the V_o subunits may be suppressing proton conduction through V_o (Zhang et al. 1992). The possible role of such suppression in vivo is discussed below.

To address the role of the accessory subunits of the coated vesicle V-ATPase, we have developed a protocol for reassembly of the V-ATPase from the dissociated V₁ and V_o domains that restores both ATP hydrolysis and proton translocation (Puopolo and Forgac,1990). Reassembly involves attachment of the complete complement of V₁ subunits to the V_o sector, is time-dependent and protein-concentration-dependent and gives rise to a reassembled V-ATPase with inhibitor sensitivities identical to those of the native enzyme.

Recently, we have employed this reassembly system to address the role of the 40 and SA×10³M_r subunits (Puopolo et al. 1992b). As explained above, we have isolated a Vi (−40×10³M_r) subcomplex which contains the A, B, 34 and 33×10³Mr subunits. This subcomplex, although devoid of ATPase activity, is able reassemble onto the V_o domain to give a complex possessing approximately 50% of the proton transport activity obtained using the complete complement of Vi and V_o subunits (Puopolo et al. 1992b). Reassembly requires that the V₁ (−40×10³M_r) subcomplex first be dissociated with KI and MgATP, followed by reassembly in the presence of V_o. Thus, the assembled subcomplex appears to be unable to bind to the membrane sector. Similarly, addition of the isolated 40x10³M_r subunit to the reassembly mixture is without effect on activity unless the V₁ (−40×10³M_r) subcomplex is first dissociated, under which conditions addition of the 40×10³M_r subunit restores activity to maximal levels. The binding sites for both V_o and for the 40×10³M_r subunit thus appear to be inaccessible in the assembled V₁ (−40×10³ M^r) subcomplex.

The monomeric V₁ fraction lacking the 34x10³M_r subunit obtained by sedimentation (see above) is also competent to reassemble with V_o to give a partially active complex (Puopolo et al. 1992b). The absence of either the 40 or 34×10³M_r subunits, however, makes the reassembled complexes unstable to detergent solubilization and immunoprecipitation. These results, summarized in Table 1, suggest that, although not absolutely required for coupling of ATP hydrolysis and proton transport, both the 40 and 34×10³M_r subunits are required for stability and maximal activity of the coated vesicle V-ATPase. These subunits may therefore play some role in regulation of either assembly or activity of the V-ATPases in vivo.

It is important to note that, unlike F₁ and F_o, the V₁ and V_o domains of the coated vesicle V-ATPase do not appear to function independently. Thus the V₁ (−40x10³M_r) subcomplex lacks ATPase activity (Puopolo et al. 1992b) and no conditions have yet been identified that allow binding of the 40×10³M_r subunit to this subcomplex in vitro. Moreover, preliminary results suggest that a V₁ complex containing the 40×10³M_r subunit formed in vivo is still unable to hydrolyze ATP (M. Myers and M. Forgac, in preparation). Similarly, as discussed above, the V_o domain, unlike the F_o domain, is unable to carry out passive, DCCD-inhibitable proton conduction (Zhang et al. 1992).

Although it is clear that cells are able to modulate the pH of the various intracellular compartments differentially (Forgac, 1989), the mechanisms of modulation remain uncertain. For example, clathrin-coated vesicles have been demonstrated to contain a V-ATPase capable of acidifying the lumen of the vesicle (Forgac et al. 1983; Stone et al. 1983), yet numerous data suggest that ligand-receptor complexes do not become exposed to an acidic pH until delivery to CURL. Thus, ligand-receptor complexes remain associated in endocytic coated vesicles and small peripheral endosomes (Geuze et al. 1983) and endocytic coated vesicles appear to be neutral organelles within the cell, based upon their failure to accumulate the electron microscopic probe 3-(2,4-dinitroanilino)-3′-amino-N-methyldipropylamine (DAMP) (Anderson and Orci, 1988). Finally, rat liver endocytic coated vesicles loaded with the pH-sensitive fluorescence probe fluorescein-5′-isothiocyanate (FITC)-dextran appear to be incapable of ATP-dependent acidification (Fuchs et al. 1987).

To determine whether the failure of endocytic coated vesicles to acidify was due to the absence of the proton pump, we have carried out immunocytochemical studies of MDBK cells, a bovine epithelial cell line, using monoclonal antibodies directed against the coated vesicle V-ATPase (Marquez-Sterling et al. 1991). We have demonstrated that, in addition to staining endosomes, lysosomes and portions of the Golgi complex, these antibodies label the plasma membrane and vesicles just beneath the plasma membrane. Moreover, significant co-localization of the V-ATPase with clathrin was observed at both the cell surface and in peripheral vesicles, indicating that the vacuolar proton pump is present in endocytic coated vesicles (Marquez-Sterling et al. 1991). These results, together with those described above, suggest that endocytic coated vesicles are not acidified because the activity of the coated vesicle H⁺-ATPase is suppressed within the cell.

A number of possible mechanisms may be involved in controlling vacuolar acidification. Changes in subunit composition of the V-ATPase represent one important regulatory mechanism which may be employed. One candidate for such a modulatory subunit is the 100×10³Mr polypeptide, which is present in the coated vesicle V-ATPase and other V-ATPases but appears to be absent from the V-ATPase found in the plasma membrane of kidney cells (Gluck and Caldwell, 1987). Alternatively, regulation may be accomplished through substitution of organelle-specific isoforms of particular V-ATPase subunits. The existence of multiple isoforms of the human (Bernasconi et al. 1990) and bovine (Puopolo et al. 1992a) B subunit has recently been demonstrated. The degree of vacuolar acidification may in some cases be controlled by changes in the number of V-ATPases in a particular compartment. This mechanism operates in regulating acid secretion across the luminal membrane of the intercalated cells of the mammalian kidney (Gluck et al. 1982). In this case, the density of pumps in the luminal membrane is controlled by reversible fusion with the plasma membrane of intracellular vesicles containing a high density of vacuolar proton pumps.

A third possible mechanism for regulating vacuolar acidification involves control of coupling between ATP hydrolysis and proton translocation. As discussed above, there are several experimental conditions that have been shown to alter the tightness of this coupling for the coated vesicle V-ATPase, including high concentrations of ATP (Arai et al. 1989), partial proteolysis (Adachi et al. 1990a) and detergent solubilization (Arai et al. 1987a). These results suggest that the enzyme may be poised in a state in which the stoichiometry of proton transport can readily be altered in response to the appropriate intracellular signal. A fourth possible regulatory mechanism is suggested by the apparent inactivity of the separated V₁ and V_o domains with respect to ATP hydrolysis and proton conduction (Puopolo et al. 1992b; Zhang et al. 1992) (see above). This mechanism would involve control of ATP-driven proton transport by controlling attachment of the V₁and V_o domains. Such a mechanism would necessitate that at least the V_o domain was not functional as a passive proton channel, since otherwise, in any compartment containing multiple copies of V_o, no significant pH gradient could be achieved until all the V_o sites had been occupied. Further evidence consistent with this model comes from the observation that coated vesicles appear to contain an excess of V_o domains over those required to form functional V₁ V₀ complexes (Zhang et al. 1992) and from the observation that a soluble pool of V₁ domains exists free in the cytoplasm (M. Myers and M. Forgac, in preparation).

A final mechanism of controlling vacuolar pH involves control of the chloride channel required for acidification. We have recently demonstrated that both chloride conductance and ATP-dependent acidification in coated vesicles are modulated by a protein-kinase-A-dependent phosphorylation and that this effect is the result of alteration in the activity of the chloride channel (Mulberg et al. 1991). It is possible that several of the mechanisms described above may be employed in regulating vacuolar acidification in vivo.

This work was supported by National Institutes of Health Grant GM 34478 and GM 44828 and American Heart Grant-in-Aid 890772. M.F. is an American Heart Association Established Investigator.