A Novel Role for Subunit C in Mediating Binding of the H+-V-ATPase to the Actin Cytoskeleton*

Primary proton transport by V-ATPases is regulated via the reversible dissociation of the V1V0 holoenzyme into its V1and V0 subcomplexes. Laser scanning microscopy of different tissues from the tobacco hornworm revealed co-localization of the holoenzyme and F-actin close to the apical membranes of the epithelial cells. In midgut goblet cells, no co-localization was observed under conditions where the V1 complex detaches from the apical membrane. Binding studies, however, demonstrated that both the V1 complex and the holoenzyme interact with F-actin, the latter with an apparently higher affinity. To identify F-actin binding subunits, we performed overlay blots that revealed two V1subunits as binding partners, namely subunit B, resembling the situation in the osteoclast V-ATPase (Holliday, L. S., Lu, M., Lee, B. S., Nelson, R. D., Solivan, S., Zhang, L., and Gluck, S. L. (2000) J. Biol. Chem. 275, 32331–32337), but, in addition, subunit C, which gets released during reversible dissociation of the holoenzyme. Overlay blots and co-pelleting assays showed that the recombinant subunit C also binds to F-actin. When the V1 complex was reconstituted with recombinant subunit C, enhanced binding to F-actin was observed. Thus, subunit C may function as an anchor protein regulating the linkage between V-ATPase and the actin-based cytoskeleton.

V-ATPases are ubiquitous proton pumps that are found in the endomembranes of all and the plasma membranes of many specialized eucaryotic cells (1,2). They comprise at least 12 subunits that are part of two different subcomplexes, i.e. a peripheral, catalytic V 1 complex with subunits A 3 B 3 CDE X FG Y H and a membrane-bound, proton-conducting V 0 complex with subunits ac 6 de. In the midgut of the tobacco hornworm (Manduca sexta) they are localized in the apical membrane of goblet cells where they exclusively energize all secondary active transport processes across the epithelium (3). During starvation or molt, pump activity abolishes due to the dissociation of the V 1 complex from the membrane (4). In the tobacco hornworm as well as in yeast, where reversible dissociation occurs upon the withdrawal of glucose, subunit C seems to be released into the cytoplasm because the purified V 1 complex contains not more than substoichiometric amounts of this polypeptide (5)(6)(7). The regulatory mechanisms responsible for reversible V-ATPase dissociation are not understood, although it seems likely that interactions of V-ATPase subunits with other cellular proteins control this process. In fact, several proteins have already been identified that bind to subunits of the V-ATPase and thus may link this enzyme to a comprehensive cellular network. Among recently discovered candidates is the yeast RAVE complex (8), which associates with the cytosolic V 1 complex via its rav1p component and is apparently essential for stable assembly of the V-ATPase holoenzyme (9). Another example is the plasma membrane V-ATPase in osteoclasts, which may be linked to glycolysis via the interaction of its subunit E with aldolase, an enzyme of the glycolytic pathway (10).
The involvement of proteins interacting with V-ATPases may also be relevant to sorting processes within the cell. In clathrin-coated vesicles, V-ATPase seems to be associated with the 50-kDa polypeptide of the AP-2 complex (11), which is involved in the internalization of proteins from the plasma membrane (12). Even more, this interaction may be necessary for activity and in vitro reassembly of the V-ATPase (13). Binding of V-ATPase to AP-2, however, may also be linked to the mechanisms of pathogenicity in human immunodeficiency virus (HIV) 1 infection. Direct interaction between the HIV accessory protein Nef and the V-ATPase subunit H was correlated with the ability of Nef to internalize CD4, the primary receptor for HIV (14). More detailed investigations revealed that Nef contacts the endocytotic machinery and therefore suggested an important role for subunit H of the V-ATPase in viral infectivity (15). Most recently, it was demonstrated that subunit H, which shows structural and functional similarities to ␤-adaptins (16), binds in vitro and in vivo to the C-terminal flexible loop of Nef as well as to AP-2 (17). Thus, subunit H may function as an adaptor for interactions between Nef and AP-2.
Binding of cellular proteins may also link V-ATPases to the cytoskeleton. In vivo experiments indirectly suggested that microtubules might be involved in glucose-dependent dissociation of the yeast V-ATPase (18). However, V-ATPases are also associated with the actin-based cytoskeleton. During osteoclast activation ruffled membrane V-ATPase directly binds to Factin, evidently linking V-ATPase transport to the reorganization of the actin cytoskeleton (19). Binding to F-actin is mediated by the amino-terminal domain of the V 1 subunit B (20). The interaction of actin and V-ATPase may occur only in systems where the V-ATPase is sorted into the plasma membrane of a cell; latrunculin, which disrupts actin filaments, had no effect on glucose-dependent dissociation of the V-ATPase from the endomembranes (18).
To investigate whether the interaction between actin and plasma membrane V-ATPases reflects a general phenomenon, we performed co-localization studies, co-pelleting assays, and overlay blots using the plasma membrane V-ATPase of Manduca midgut goblet cells. We show that, in vitro, both the V 1 complex and the V 1 V 0 holoenzyme bind directly to F-actin. In vivo, however, this interaction seems to be restricted to the holoenzyme. The interaction is mediated by subunits B and C; the latter subunit may function as a regulatory linker protein between the V-ATPase and the actin-based cytoskeleton.

EXPERIMENTAL PROCEDURES
Insects-Larvae of M. sexta (Lepidoptera, Sphingidae) were reared under long day conditions (16 h of light) at 27°C using a synthetic diet modified according to Bell and Joachim (21).
Immunohistochemistry-Midguts of the larvae were dissected, and the gut content was removed. After excision of longitudinal muscles, the tissue was stretched, cut into small pieces of about 5 mm 2 and fixed for 90 min at room temperature in PLP fixative (0.1 M sodium metaperiodate, 75 mM L-lysine, 2% (w/v) paraformaldehyde in 0.1 M Sørensen phosphate buffer, pH 7.4). Tissue embedding, cryosectioning, and immunostaining were performed as described previously (22). For labeling of the V 1 complex, the cryosections were treated with a 1:10 dilution of monoclonal antibody 221-9 to subunit A of the Manduca V-ATPase (22). Soluble subunit C was labeled with a 1:2000 dilution of monospecific, polyclonal antibodies to the recombinant protein (6). Visualization of the primary antibodies was performed with a 1:1000 dilution of Cy3-or TRITC-conjugated anti-mouse and anti-guinea pig F(abЈ) 2 -fragments (Sigma). To test for unspecific binding of the secondary antibody, control reactions were carried out without primary antibodies. For labeling of the actin filaments, cryosections were treated with FITC-conjugated phalloidin (5 g/ml in 20 mM phosphate-buffered saline (pH7.4), 0.1% Triton X-100, and 2% bovine serum albumin) for 40 min at room temperature. After rinsing three times for 5 min with phosphate-buffered saline, the sections were covered by Mowiol (Aventis, Frankfurt, Germany) and viewed with either an Olympus IX70 fluorescence microscope or a confocal laser-scanning microscope (Zeiss, LSM 410). Double-labeling with FITC-phalloidin and monoclonal antibody 221-9 was performed for co-localization studies; antibody 221-9 was detected with TRITC-conjugated anti-mouse F(abЈ) 2 -fragments (Sigma). To exclude mutual co-excitation, detection of fluorescence signals was performed by using two lasers with excitation wavelengths of ϭ 488 nm for the FITC-signal and ϭ 568 nm for the TRITC-signal and long pass emission filters of 515 and 595 nm, respectively.
Actin Binding Assay-For in vitro binding studies, rabbit skeletal muscle actin (Cytoskeleton, Inc.) was dissolved in water and depolymerized at 0°C for 30 min. After centrifugation for 30 min at 200,000 ϫ g and 4°C, G-actin (70 M) was polymerized for 1 h at 30°C in buffer P (5 mM Tris-HCl, pH 8.0, 40 mM KCl, 2 mM MgCl 2 , 0.2 mM CaCl 2 , and 1.2 mM ATP). Binding studies and determination of stoichiometries were performed at monomer concentrations below the critical concentration for actin polymerization. For this purpose, F-actin was stabilized with 10 M phalloidin and then diluted to 200 nM in buffer P containing 10 M phalloidin, 2 mM dithiothreitol, and varying concentrations of V 1 or V 1 V 0 ATPase, respectively. Samples were incubated for 2 h at room temperature and then centrifuged for 1 h at 200,000 ϫ g and 20°C. Protein amounts of the supernatant and the dissolved pellets were determined, and the proteins were analyzed by SDS-PAGE and subsequent silver staining.
Overlay Blots-Five micrograms of the purified V 1 complex or the V 1 V 0 holoenzyme, respectively, or 1 g of aldolase or the recombinant subunit C, respectively, were separated by SDS-PAGE and transferred onto a nitrocellulose membrane. Blocking was performed for 1 h in a buffer containing 3% gelatin in 20 mM Tris-HCl (pH 7.5), 500 mM NaCl, and 0.02% NaN 3 . The membrane was equilibrated three times for 5 min in buffer F (5 mM Tris-HCl, 0.2 mM CaCl 2 , 5 mM MgCl 2 , 0.2 mM ATP, and 200 mM KCl, pH 8.0) and then incubated for 1 h at room temperature in buffer F containing 1% gelatin and F-actin at a concentration of 0.1 mg/ml. After washing the membrane three times for 5 min with F buffer containing 0.05% Tween 20, bound actin filaments were detected by incubating the membrane in buffer F containing 1% gelatin and rabbit polyclonal anti-actin antibodies (Sigma) at a dilution of 1:200. Afterward, the membrane was rinsed three times for 5 min in buffer F containing 0.05% Tween. Subsequently, the membrane was incubated for 1 h in buffer F containing 1% gelatin and anti-rabbit IgG (whole molecule) conjugated with alkaline phosphatase at a dilution of 1:30,000. Following a third washing step that was performed as above, decorated protein bands were detected after rinsing the membrane in 0.34‰ nitro blue tetrazolium (NBT)/0.18‰ 5-bromo-4-chloro-3-indolyl phosphate (BCIP) in a reaction buffer consisting of 50 mM Tris-HCl (pH 9.5), 100 mM NaCl, and 50 mM MgCl 2 .
Coupling of C rec to Activated Sepharose Beads and Binding Studies with Cytosolic Midgut Extracts-Two hundred microliters of Affi-Gel (Bio-Rad) were washed three times with 4 ml of H 2 0 and then three times with 4 ml of coupling buffer consisting of 0.1 M MOPS-NaOH (pH 7.5), 9.6 mM ␤-mercaptoethanol, and 0.01% C 12 E 10 . Afterward the Affi-Gel was incubated for 12 h at 4°C in coupling buffer containing 0.1 mg/ml recombinant subunit C. Following three further washing steps with coupling buffer, the beads were incubated at 4°C for 1 h in 4 ml of 20 mM ethanolamine-HCl (pH 8.0) to block any remaining active esters. Ethanolamine-HCl was removed by washing the beads three times with 4 ml of 0.1 M MOPS-NaOH (pH 7.5) containing 9.6 mM ␤-mercaptoethanol. Cytoplasmic extracts from the midgut of fifth instar larvae were prepared as follows. Midgut tissue isolated from ten 5th instar larvae was washed three times in TEK buffer (16 mM Tris-HCl, 0.32 mM EDTA, 200 mM KCl, pH 8.1) and carefully broken up on ice in 4 ml of TEK buffer containing 5 mM Pefabloc (Biomol) using a glass/Teflon homogenizer. Insoluble cell compounds were removed by centrifugation for 20 min at 20,000 ϫ g and 4°C. The remaining supernatant containing the cytosolic extract was supplemented with 9.6 mM ␤-mercaptoethanol. The beads were incubated with 4 ml of the cytosolic extract by rotating the suspension slowly for 2 h at 4°C. Unspecifically bound proteins were removed by washing the beads three times with 4 ml of 0.1 M MOPS-NaOH (pH 7.5) containing 9.6 mM ␤-mercaptoethanol. As a positive control, rabbit muscle actin (Cytoskeleton) was dissolved in 50 mM MOPS-NaOH (pH 7.5) to a final concentration of 24 M and tested for binding to the beads as described above.
For analyzing bound proteins, the beads were spun down in a microfuge, resuspended in Laemmli buffer (125 mM Tris-HCl, 5% sucrose, 2% SDS, 2% ␤-mercaptoethanol, and 0.05% bromphenol blue) and boiled for 1 min. Identification of actin was performed by Western blotting after SDS-PAGE using polyclonal antibodies directed to a highly conserved actin domain (Sigma, A-2066).
Depletion of Residual Subunits C from the V 1 Complex and Reassociation of Recombinant Subunit C-The V 1 complex was purified according to Grä f et al. (23) except for the presence of 25% methanol during the last purification step on a Superdex 200 column (Amersham Bioscience), which led to complete removal of residual C subunits (24). Binding of recombinant subunit C to the C-depleted V 1 complex was performed by incubating 100 g of the V 1 complex with a 10-fold molar excess of recombinant subunit C for 16 h at 4°C in a buffer consisting of 150 mM NaCl, 20 mM Tris-HCl (pH 8.1), and 9.6 mM ␤-mercaptoethanol. Subsequently, the V 1 complex containing recombinant subunit C (V 1C ) was isolated by gel permeation chromatography on a Superdex 200 column. To analyze actin-binding properties, the V 1 complex lacking subunit C as well as V 1C were incubated for 1 h at room temperature with 200 nM of phalloidin-stabilized F-actin in a buffer consisting of 10 mM Tris-HCl (pH 7.5), 50 mM KCl, 1 mM MgCl 2 , 1 mM ATP, and 2 mM dithiothreitol. After centrifugation at 200,000 ϫ g for 1 h, the proteins in the pellet were quantified and analyzed by SDS-PAGE and silver staining.
Other Methods-The V 1 complex and the V 1 V 0 holoenzyme were purified from M. sexta larval midgut according to published protocols (23,25). Recombinant subunit C was expressed in Escherichia coli BL21 cells and purified by nickel-nitriloacetic acid (Ni-NTA) affinity chromatography as described previously (6). Protein concentrations were determined by the Amido Black method (26). Absolute protein concentrations were calculated based on the molar extinction coefficients of the proteins at 280 nm; respective absorption measurements were preceded by treatment of the proteins with 6 M guanidine hydrochloride (27). SDS-PAGE, Western blotting, and immunostaining were performed as described previously (25, 26,28).

V-ATPase and F-Actin Co-localize at the Apical Membranes of Epithelial Cells from Midgut, Malpighian Tubules, and Salivary Glands-
The midgut epithelium of the tobacco hornworm consists of columnar cells forming an extensive microvillar brush border, goblet cells with large cavities, and pluripotent stem cells from which both cell types derive (29). The microvilli of columnar cells are supported by actin filaments that are arranged along their longitudinal axes. Because the goblet cell apical membrane is also organized in a microvillus-like fashion (30), we looked for actin filaments using FITC-labeled phalloidin that exclusively binds to F-actin. As shown in Fig. 1A (left  panel), a strongly developed fluorescence signal was observed in the brush border membrane of columnar cells as had been observed previously (31). In addition, a previously unrecognized labeling in the region of goblet cell apical membranes was observed, although it was not as intense as the signal of the columnar cell brush border. Because it is known that V-ATPase resides in the goblet cell apical membranes, we also examined the subcellular V-ATPase distribution using the monoclonal antibody 221-9 to V-ATPase subunit A (22). As indicated by the yellow color of the merged images in the middle panel of Fig. 1A and the right panel of Fig. 1B, V-ATPase co-localized with actin filaments at the goblet cell apical membrane. Apical brush border membranes of epithelial cells from Malpighian tubules and salivary glands, which are known to contain V-ATPase as well (1), also exhibited co-localization of V-ATPase and F-actin (Fig. 1, C  and D). Thus, in all three cases where V-ATPase is present in the apical plasma membrane, it seems to be in a close neighborhood with bundled actin filaments of microvilli or with the loose network of actin filaments underneath the membrane.
Both the V 1 Complex and the Holoenzyme Bind to F-actin-Co-localization of V-ATPase and F-actin suggested that the plasma membrane V-ATPase may be associated with the actin cytoskeleton. To examine this possibility, we performed in vitro co-pelleting assays using the purified V 1 V 0 ATPase and F-actin from rabbit muscle. For this purpose, we used phalloidin-stabilized F-actin at a concentration which was below the critical concentration for actin polymerization. The critical concentration in our assays was about 0.4 M, a value that is in good agreement with previously published concentrations ranging from 0.1 to 1 M (32). After centrifugation at 200,000 ϫ g for 1 h, the proteins in the pellet were separated by SDS-PAGE and silver stained. In contrast to controls without actin, a significant portion of the holoenzyme was found in the pellet together with F-actin ( Fig. 2A). The same result was obtained when we used the purified V 1 complex instead of the holoenzyme (Fig. 2B). In both cases, binding to F-actin was a saturable process. Our findings imply that both the holoenzyme and the V 1 complex bind directly to F-actin.
To evaluate the stoichiometries of the interaction with actin, we determined molar protein concentrations after co-pelleting, taking into account the known actin amounts of the pellet. Although actin appeared to exhibit a slightly higher (2-3-fold) affinity to the V 1 V 0 holoenzyme as compared with the V 1 complex, binding reached its saturation at concentrations of about 50 nM for both the V 1 V 0 holoenzyme and the V 1 complex if the concentration of F-actin monomers was 200 nM (Fig. 2C). Thus, the molar ratio was estimated to be about 1 V-ATPase to 4 -5 F-actin monomers. Because the V 1 V 0 holoenzyme as well as the V 1 complex have more than 10-fold molecular masses as compared with the actin monomer, these results suggest that maximum binding capacity is limited, which is likely due to steric hindrance of F-actin binding to the V-ATPase. The determined stoichiometry was in the same range as those reported previously for the bovine V-ATPase, which was estimated to be 1:8 (19).
Subunits B and C Mediate Binding to F-actin-Recently, Holliday et al. (20) demonstrated that the bovine V-ATPase subunit B contains an F-actin binding site. To determine whether subunit B is responsible also for the binding of F-actin to the Manduca V-ATPase, we performed overlay blots. For that purpose, we separated the subunits of the V 1 V 0 holoenzyme and the V 1 complex by SDS-PAGE and stained them, after transfer onto nitrocellulose, with either Ponceau S or F-actin (Fig. 3, A and B, respectively). Immunodetection of bound F-actin revealed a protein band of 56 kDa for both enzyme preparations, confirming that subunit B interacted with actin filaments. To our surprise, we found an additional polypeptide of ϳ40 kDa in the V 1 V 0 holoenzyme preparation that was intensively labeled by anti-actin antibodies. Control reactions performed in the absence of F-actin showed no labeled protein bands (Fig. 3C). Based on the observed molecular mass, we assumed that the labeled polypeptide was the V 1 subunit C, which we had identified previously to be part of the Manduca V-ATPase (6). Because this subunit is present in preparations of the Manduca V 1 complex in substoichiometric amounts at the most (7), we were not astonished to observe labeling only in the V 1 V 0 holoenzyme. To confirm our assumption and to discriminate the labeled band from the V 0 subunit d, which has a similar molecular mass (33), we also tested the recombinant subunit C (C rec ) for F-actin binding. As expected, C rec reacted with actin filaments in the overlay blots (Fig. 3B,  lane 4). Moreover, it co-pelleted together with F-actin in a saturable manner (Fig. 4A). Estimation of the binding stoichiometry revealed a molar C rec /F-actin-monomer ratio of ϳ1:1 (Fig. 4B).
To date, we have demonstrated that subunit C of the purified V 1 V 0 ATPase as well as the recombinant subunit C bind to rabbit muscle actin. However, binding between V-ATPase and actin may depend on the source of the microfilaments, although rabbit skeletal muscle actin shares more than 93% identical amino acids with Manduca non-muscle actin, whose sequence (submitted to the EMBL/GenBank TM data base under accession no. AJ519536) we have recently deduced from a -Zap clone isolated from a Manduca midgut cDNA library (34).
To confirm that subunit C also binds to Manduca non-muscle actin, we covalently coupled the recombinant subunit C to beads of Affi-Gel 10 and incubated them with a cytoplasmic extract from the larval midgut. SDS-PAGE of proteins attached to subunit C revealed one major band that was identified immunologically in a Western blot as actin (Fig. 5). Thus, we could show that C rec not only binds to rabbit muscle actin but also to non-muscle actin from the midgut of the tobacco hornworm. This finding also suggests that the described interaction between C rec and actin may occur in vivo in the Manduca midgut.
Recombinant Subunit C Increases Binding of F-actin to the V 1 Complex-The isolated V 1 complex does not contain subunit C in significant amounts (see also Fig. 2B) and thus lacks one  4) were separated by SDS-PAGE and blotted onto a nitrocellulose membrane. A, proteins on the membranes were stained with Ponceau S. B, the blots were blocked with bovine serum albumin and then incubated with F-actin for 1 h at room temperature. After rinsing, bound F-actin was detected by its immunoreaction with an antiserum to actin. C, control blot treated as described above but in the absence of F-actin. For evaluation of the overlay blots we performed controls with aldolase and a low molecular weight marker containing phosphorylase b (94 kDa), albumin (67 kDa), ovalbumin (44 kDa), carbonic anhydrase (30 kDa), trypsin inhibitor (20 kDa), and ␣-lactalbumin (14 kDa). Aldolase and phosphorylase b had been shown previously to bind to F-actin and therefore acted as positive controls for the binding assay (41,42). Interestingly, we found that carbonic anhydrase was also decorated with F-actin, a finding that, to our knowledge, was not reported previously.

FIG. 4. Recombinant subunit C binds to F-actin.
A, phalloidinstabilized F-actin (200 nM) was incubated with the indicated concentrations of recombinant subunit C (C rec ) for 2 h at room temperature. The mixtures were centrifuged at 200,000 ϫ g for 1 h, and the pellets were analyzed by SDS-PAGE and silver staining. Arrowheads indicate the positions of actin and C rec . B, estimation of binding stoichiometry of actin monomers and C subunits. Phalloidin-stabilized F-actin (200 nM) was incubated in solutions containing recombinant C subunits at indicated concentrations. Following incubation and centrifugation, the molar concentrations were estimated densitometrically after normalizing protein amounts. Data were collected from three independent experiments and depicted as means Ϯ S.E. of the F-actin binding subunits. Therefore we hypothesized that the evidently lower affinity of the V 1 complex to F-actin as compared with that of the V 1 V 0 holoenzyme (see Fig. 2C) might be due to the lack of subunit C. Because the V 1 complex can be supplemented by incubation with C rec resulting in a V 1C complex containing significant amounts of subunit C (Ref. 24; see also Fig. 6A), we tested the capability of V 1C to bind F-actin. For this purpose, we performed F-actin co-pelleting assays with either the V 1C complex or the V 1 complex lacking subunit C (see "Experimental Procedures"). Reconstitution of the V 1 complex with C rec led to a significant increase in the amount of V 1 -ATPase subunits found in the pellet as compared with the V 1 complex lacking subunit C (Fig. 6, B and C). This result indicated an increase in binding capacity due to the presence of an additional actin binding subunit. V 1 Complexes That Have Detached from the Membrane Do Not Co-localize with F-actin-The above-mentioned results raised the question of whether binding of F-actin mediated solely by subunit B is sufficient to allow binding to V 1 complexes in vivo. To answer this question, we labeled cryosections of the posterior midgut from starving tobacco hornworms with either a monoclonal antibody to the V 1 subunit A to detect V 1 complexes, a monospecific antiserum to subunit C, or FITC phalloidin to visualize F-actin, respectively. In accordance with previous biochemical studies, the V 1 complexes became detectable in the cytoplasm of goblet cells due to dissociation of the V 1 V 0 holoenzyme during starvation (Fig. 7A). As expected, subunit C was found in the cytoplasm, too (Fig. 7B). The cytoplasmic fluorescence signals resulting from immunoreactions with subunit A and C were not distinguishable so that both, free V 1 complexes and C subunits, appeared to be evenly distributed throughout the cytoplasm of the goblet cells. By contrast, actin filaments were localized almost exclusively near the brush border membrane or close to the basal membrane (Fig. 7C). Thus, neither the detached V 1 complexes nor the C subunits appeared to co-localize with F-actin, suggesting that binding only to subunit B may not be sufficient for the attachment of actin filaments to the V-ATPase in vivo.

DISCUSSION
In the larval midgut of M. sexta, high amounts of non-muscle actin filaments are found in the microvilli of the columnar cell-derived brush border (Ref. 31; see also Fig. 1). Here we show that actin filaments also reside at the goblet cell apical membranes, which are organized in a microvillus-like fashion (29). Because goblet cell apical membranes contain V-ATPase at high densities, we asked whether this enzyme is linked to the actin-based cytoskeleton. Co-localization studies and different binding assays performed in this work clearly indicated FIG. 5. Binding studies with crude larval midgut extracts. Recombinant subunit C was coupled to Affi-Gel (Affi-C rec ) and incubated with a cytoplasmic extract from the midgut of M. sexta 5th instar larvae. Unspecifically bound proteins were removed, and the beads were loaded directly onto an SDS-polyacrylamide gel (lanes 2 and 4). As a positive control, Affi-C rec was incubated with F-actin at a final concentration of 24 M (lanes 1 and 3). The proteins were either stained with Coomassie Blue (lanes 1 and 2) or transferred onto a nitrocellulose membrane and stained with an anti-actin antiserum (lanes 3 and 4). An arrowhead marks actin. Control experiments showed that F-actin did not bind to non-coupled Affi-Gel. that the Manduca V-ATPase is able to interact with F-actin. Our results are in good agreement with those from previous studies demonstrating interaction between F-actin and V-ATPase during osteoclast activation (19). Like the Manduca midgut V-ATPase, the osteoclast V-ATPase resides in the apical plasma membrane. Thus, the interaction of a plasma membrane V-ATPase with cellular actin may reflect a general phenomenon that could be of functional relevance.
Overlay blots performed with rabbit muscle actin and either the V 1 V 0 holoenzyme or the V 1 complex confirmed the results of Holliday et al. (20) demonstrating that subunit B is involved in actin binding. To our surprise, however, we found that not only subunit B but also subunit C, which is a constituent part of the V 1 V 0 holoenzyme but not of the V 1 complex (5, 6, 7), binds to F-actin. In contrast to our results, previous studies with V-ATPase immunoprecipitated from metabolically labeled mouse marrow cultures or immunopurified from bovine kidney supplied no indications for subunit C to be involved in actin binding (19). Consequently, either binding properties of C subunits differ between insects and mammalian V-ATPases, or subunit C did not bind to actin in the mammalian system because it was not present in the V-ATPase preparations used for these binding studies. A reasonable explanation for the absence of subunit C may be based on the purification procedure using anti-V 1 antibodies for immunopurification. Because cytoplasmic V 1 complexes and membrane-bound holoenzymes are in dynamic equilibrium, immunopurification via anti V 1 antibodies might have favored enrichment of V 1 complexes rather than enrichment of the holoenzyme. Consistently, none of the reported V-ATPase preparations from mouse marrow and bovine kidney show the presence of either subunit C or any V 0 subunit (19). Thus, it seems possible that mainly the cytoplasmic V 1 complex instead of the V 1 V 0 holoenzyme had been investigated for actin binding and, therefore, only subunit B could be detected as actin binding protein.
In contrast to some other V-ATPase subunits, the function of the V 1 subunit C is still elusive. Subunit C is believed to be a peripheral stalk component that may not be essential for enzyme activity in a reconstituted system (35) but is important for assembly of the V 1 complex (36,37). Genetic disruption of the gene VMA5 encoding subunit C in yeast leads to a typical VMA Ϫ phenotype with cells that are unable to grow at neutral pH and in the presence of higher calcium concentrations. Vacuolar vesicles isolated from VMA5⌬ yeast strains showed no V-ATPase activity and only poor assembly of the peripheral V 1 portion onto the vacuolar membrane (37). Because the deletion of either subunit C or H disrupts the ability for assembly of the remaining V 1 subunits, it was suggested that subunits C and H might play a role in bridging the V 1 and V 0 complexes. Upon reversible disassembly of the V 1 V 0 complex, subunit C is the only protein that dissociates from the resulting subcomplexes (5). Both, the peripheral localization as well as its ability to dissociate from cytoplasmic V 1 complexes and reattach to the holoenzyme make subunit C an ideal candidate for regulation of the enzyme by the interaction with other cellular proteins. In contrast to other V-ATPase subunits, to date no cellular protein has been reported that would directly interact with subunit C. Thus, actin is the first cellular protein known hitherto that binds to subunit C.
Currently, we can only speculate on the meaning of the interaction between actin and V-ATPase. Our results indicate that the V 1 V 0 holoenzyme is bound to the filamentous actin cytoskeleton lining goblet cell apical membranes. Reduced Factin binding capacity of V 1 complexes in comparison to V 1C complexes and missing co-localization of cytoplasmic V 1 com-plexes and actin filaments suggest that, upon enzyme dissociation, V 1 complexes do not only detach from the membrane but also from actin filaments due to the loss of subunit C. Consequently, subunit C may support subunit B to retain the holoenzyme at the actin cytoskeleton and assist in releasing V 1 complexes into the cytoplasm. In conclusion we hypothesize that subunit C plays a crucial role not only in supporting V-ATPase assembly but also in controlling its linkage to the actin-based cytoskeleton in those cases where V-ATPase resides in the plasma membrane. By contrast, V-ATPase residing in the endomembrane system appears to be linked exclusively to microtubules because it was found that nocodazole, which disrupts microtubules, partially blocks dissociation of the V-ATPase in response to glucose depletion in yeast, whereas latrunculin, which disrupts actin filaments, had no effect (18). Independent of the type of filament, however, the interaction between V 1 subunits of the active holoenzyme and cytoskeletal proteins could assist the stator subunits of the V-ATPase, such as V 0 subunit a (2), in compensating the torque resulting from the rotation of central subunits.
The goblet cell apical membrane contains V-ATPase molecules at a very high density, which can be estimated by electron microscopy of isolated vesicles to be about 5,000 complexes/ m 2 . 2 The high density may lead to a lack of space for other transmembrane proteins. Indeed, SDS-PAGE of highly purified goblet cell apical membranes only shows the well defined protein pattern for V-ATPases without any indications for the presence of additional proteins at higher concentrations (28). Therefore, transmembrane proteins that are able to link the plasma membrane to the spectrin-based actin cytoskeleton via ankyrin, such as different antiporters, P-type ATPases, or ion channels (38), might be absent as well. For instance, the Manduca midgut completely lacks a Na ϩ /K ϩ ATPase (39), which had been presumed to bind to ankyrin in the Drosophila midgut (40). Other putative anchor proteins in the goblet cell apical membrane may be present at densities that are too low for anchoring the apical plasma membrane at the actin cytoskeleton. Based on these considerations, it appears plausible that the V-ATPase takes over this anchor function. For Drosophila, variability between different epithelia with respect to the organization of the membrane skeleton and its transmembrane anchors has been suggested (40). However, if there is variability regarding the choice of the anchor protein, then it would not be surprising if the linker protein connecting the anchor to the actin filaments is variable also. Subunit C may act as such a linker between the goblet cell apical membrane and the cytoskeleton, bypassing the necessity of further proteins because it directly binds to F-actin. Consistently, immunolocalization of spectrin in the Manduca midgut showed no signal at the goblet cell apical region, whereas spectrin was clearly localized in the region of the terminal web of the columnar cell brush border (31).