β1 Integrin Binds the 16-kDa Subunit of Vacuolar H+-ATPase at a Site Important for Human Papillomavirus E5 and Platelet-derived Growth Factor Signaling*

Integrins mediate adhesive interactions between cells and the extracellular matrix, and play a role in cell migration, proliferation, differentiation, cytoskeletal organization, and signal transduction. We have identified an interaction between the β1 integrin and the 16-kDa subunit of vacuolar H+-ATPase (16K). This interaction was first isolated in a yeast two-hybrid screen and confirmed by coimmunoprecipitation and inin vitro binding assays using bacterially expressed proteins. Immunofluorescent studies performed in L6 myoblasts expressing both native and epitope-tagged 16K demonstrate co-localization with β1 integrin in focal adhesions. Deletion of the fourth of four transmembrane helices in 16K results in loss of interaction with β1 integrin in vitroand in the two-hybrid system, and less prominent staining in focal adhesions. This helix is also required for ligand-independent activation of platelet-derived growth factor-β receptor signaling by the human papillomavirus E5 oncoprotein. Overexpression of 16K or expression of 16K lacking this helix alters the morphology of myoblasts and fibroblasts, suggesting that the interaction of 16K with integrins could be important for cell growth control. We also discuss the possible role 16K might play in integrin movement.

Integrins comprise the major family of cell surface receptor proteins that interact with the extracellular matrix (ECM) 1 or with counter receptors on adjacent cells. Integrin-ECM interactions (reviewed in Refs. [1][2][3] control the adhesion and migration of cells, as well as the transduction of signals regulating such processes as cell growth and differentiation. Integrin receptors involved in ECM interactions cluster at the points where stable binding to the matrix occurs, producing "focal adhesions" to which cytoskeletal components and signaling molecules co-localize (4,5). Integrin ␣/␤ heterodimers can be divided into subfamilies on the basis of their common ␤ subunit. Those of the ␤ 1 subfamily include at least nine different heterodimers that function primarily as receptors for collagens, laminins, and fibronectin.
Much of the literature on integrins deals with their extracellular interactions with ECM ligands and how cytoplasmic domains tweak signaling pathways and influence cytoskeleton structure. However, little is known about the mechanisms whereby they redistribute in the cell membrane during focal adhesion formation; how they shuttle through the cell during synthesis, endocytosis, and transcytosis; and how they become targeted for degradation or recycling. It is believed that interactions within the plasma and organellar membranes are critical to these processes (6). Several types of integral plasma membrane proteins have been demonstrated by co-immunoprecipitation to interact with integrin heterodimers. These are the integrin-associated proteins, which interact with certain ␣ v integrin heterodimers (7), EMMPRIN (8), and members of the four transmembrane domain (TM4) protein family: CD9, CD53, CD63, CD81, CD82, and NAG-2, which have been shown to complex with the ␣ 3 ␤ 1 integrin (9 -11). Little is known about the functional consequences of interactions with these proteins, although integrin-mediated Ca 2ϩ influx has been shown to be regulated by integrin-associated proteins (12).
Signaling from integrins and from transmembrane receptor tyrosine kinases can converge in the focal adhesion (13). For example, binding of growth factor PDGF-B causes dimerization of the ␤ form of its receptor (PDGF-␤ receptor), leading to the activation of the receptor's kinase activity and autophosphorylation. PDGF-B stimulation mimics many signaling responses triggered by integrin adhesion, such as phosphorylation of focal adhesion kinase, paxillin, and mitogen-activated protein kinases ERK-1 and ERK-2 (14,15). The convergence of integrin and PDGF-␤ receptor functions is supported by data showing that fibronectin, collagen, or anti-integrin IgG binding to fibroblasts leads to phosphorylation of the receptor in the absence of PDGF (16).
The vacuolar H ϩ -ATPase (V-ATPase) is a large multisubunit enzyme present in intracellular membrane compartments such as endosomes, lysosomes, clathrin-coated vesicles, and the Golgi complex, where it plays a vital role in the maintenance of endocytic and exocytic pathways (17)(18)(19)(20)(21). One of its subunits, a 16-kDa protein (16K) also known as ductin, has been found in the plasma membrane of renal and kidney epithelial cells, macrophages, and some tumor cell lines (22)(23)(24)(25). 16K is composed of four hydrophobic regions thought to form transmembrane ␣-helices, and plays a key role in the assembly and function of the V-ATPase pump (26). In addition, it comprises structures independently of the V-ATPase, including gap junctional complexes and mediatophores, which are involved in the release of neurotransmitters (27)(28)(29)(30). 16K also forms a complex with the E5 oncoprotein of papillomaviruses (31,32) and has been proposed to mediate the ability of E5 to cause ligandindependent activation of the PDGF-␤ receptor (33).
We report here the identification of an interaction between the transmembrane domains of ␤ 1 integrin and 16K. The interaction was detected using a yeast two-hybrid screen, and confirmed by various in vitro and in vivo binding assays. We find the distribution of 16K within cells to be ECM-dependent, and show that perturbation of 16K expression can affect the morphology of cells. These and other data in the literature showing a requirement for transmembrane domains for proper integrin distribution (e.g. Ref. 34) led us to hypothesize that 16K utilizes transmembrane interactions with receptors to shuttle them through the cell and to cell membranes.
Western Blot Analysis of Bait Protein Expression in Yeast-Yeast carrying the bait plasmids were grown to A 600 ϭ 0.7, harvested by centrifugation, and cells from 1.5 ml of culture were resuspended in 50 l of Laemmli sample buffer, frozen at Ϫ80°C, rapidly lysed in a boiling water bath for 5 min, then spun briefly to remove debris. Protein lysates were fractionated on 15% SDS-polyacrylamide gels, transferred to nitrocellulose, and probed with rabbit anti-LexA polyclonal antibody (gift from Erica Golemis). Detection was with alkaline phosphatase-conjugated goat anti-rabbit secondary antibody.
Cell Lines and Culture-The rat myogenic cell line, L6 described by Yaffe (40), the rat embryo fibroblast cell line (Ref52) and the human pancreatic tumor cell line Capan-2 were obtained from the American Type Culture Collection. L6 and Ref52 cells were cultured at 37°C in a 5% CO 2 atmosphere in ␣-modified minimal essential medium (␣-MEM), supplemented with 10% fetal bovine serum. Capan-2 cells were grown in McCoy's 5a medium with 10% fetal bovine serum.
Antibodies-The rabbit polyclonal anti-HA antibody (Y-11) was purchased from Santa Cruz Biotechnology. Anti-16K-M and anti-16K-N are rabbit polyclonal antibodies raised against N-terminal peptides from rat 16K and were used in immunoprecipitations and immunofluorescence, respectively (41,42). JHSL32 is a rabbit polyclonal antibody raised against the 70-kDa subunit of V-ATPase (43). Monoclonal Armenian hamster anti-mouse ␤ 1 integrin (HM␤1-1) was obtained from PharMingen. For double labeling experiments, primary antibodies were visualized using fluorescein isothiocyanate-conjugated mouse antihamster IgG (PharMingen), followed by fluorescein isothiocyanate-conjugated goat anti-mouse IgG (Sigma), or Texas red-conjugated goat anti-rabbit IgG antibody (Molecular Probes). Alkaline phosphataseconjugated rabbit anti-mouse IgG secondary antibody was from Sigma.
In Vitro Binding Assays-Transmembrane and cytoplasmic integrin segments were synthesized as tripartite fusions with GST and a fragment of protein A by cloning appropriate PCR-amplified products into the BamHI and EcoRI sites of pAGEX-2T (44). Such an approach with protein A tagging has previously been used in in vitro binding assays (45). The 116-amino acid protein A tag (14 kDa) provides solubility and a uniformly sized larger carrier for the small protein domains that vary in size from 3 to 10 kDa and have different charges and hydrophobicities, thereby minimizing artifacts in binding assays. The TRIC fragment was amplified using 5Ј-ATCGGATCCGCCACTGTTGTAGAGA-C-3Ј and 5-CTAGAATTCCCTCATACTTCGGATT-3Ј, the transmembrane region using 5Ј-CGGATCCATTATTCCAATTGTAG-3Ј and 5Ј-CGGAATTCCCAAATCAGCAGCAAT-3Ј, and the cytoplasmic region using 5Ј-CGGGATCCAAGCTTTTAATGA-3Ј and 5Ј-CTAGAAT-TCCCTCATACTTCGGATT-3Ј. These fusions were coupled to glutathione-agarose beads, and the protein A-tagged ␤ 1 moiety released by cleavage with thrombin. EGTA was added to 5 mM, and the beads with GST removed by centrifugation. The protein A-tagged integrin fragments were incubated with glutathione-agarose beads to which were coupled GST fusion proteins of full-length (GST-16K) or truncated versions (GST-TM1 and GST-TM4) of 16K. All GST fusion proteins were expressed using pGEX-5X-3. GST-16K was made using primers (5Ј-A-CGCGGATCCACATGGCTGACATCAAG-3Ј and 5Ј-CCGGAATTCCTA-CTTTGTGGAGAG-3Ј; 16K upstream and downstream, respectively) tagged with BamHI and EcoRI sites to PCR-amplify the full-length cDNA, GST-TM1 was isolated as a BamHI-EcoRI fragment from the two-hybrid isolate in pVP16, and GST-TM4 was amplified from the full-length cDNA using the 16K upstream primer and an internal downstream primer (5Ј-CCGGAATTCCTACACGAACAGTCGAGGC-3Ј; TM4 downstream). For in vitro binding assays, 2 g of GST-16K, GST-TM1, or GST-TM4 protein coupled to beads were incubated at 4°C with 2 g of soluble protA-TRIC, protA-TR only, or protA-IC only, for 1 h with gentle shaking in 40 mM Hepes, pH 7.9, 100 mM KCl, 5 mM MgCl 2 , 1 mM dithiothreitol, 0.5% Nonidet P-40, 1% milk powder. The beads were recovered by centrifugation and washed five times in binding buffer without milk powder. Beads were then suspended in sample buffer, heated, and the samples loaded onto 10% SDS-polyacrylamide gels. The proteins were transferred to nitrocellulose and protein A fusions detected directly by standard Western blot analysis using the rabbit anti-mouse alkaline phosphatase-conjugated antibody.
Immunoprecipitations-Cells grown as monolayers were chilled to 4°C, washed with cold PBS and lysed in cold RIPA buffer (150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 50 mM Tris, pH 7.5) for 30 min. They were then scraped from the dishes with a rubber policeman, transferred to a microcentrifuge tube, and spun for 10 min at 4°C. The soluble protein was divided into 25-g aliquots to which anti-16K-M, rabbit control serum, or HM␤1-1 was added. After 1 h at 4°C, protein A-Sepharose (20 l of 50% slurry/ml) was added and the incubation continued for 30 min. Complexes were recovered by centrifugation, washed three times with PBS, and run on 10% SDS-polyacrylamide gel electrophoresis under reducing conditions. Western blot analysis was performed using HM␤1-1 primary antibody, mouse antihamster secondary antibody, and alkaline phosphatase-conjugated rabbit anti-mouse antibody for detection.
Metabolic Labeling of Cells-Cells grown in 35-mm culture dishes were starved of methionine for 20 min in ␣-MEM free of methionine and serum. Following this period, 100 Ci of [ 35 S]methionine was added and cells were labeled at 37°C for 8 h. The supernatant was removed, cells were washed with 1 ml of ice-cold PBS, and harvested with a rubber policeman. Cells were collected by gentle centrifugation at 3000 rpm for 5 min, lysed in 200 l of RIPA buffer on ice for 10 -15 min, and centrifuged at high speed for 2 min. The supernatants were collected into a clean tube, and 40 l was used in each immunoprecipitation. Immunoprecipitations were performed as described above, and complexes were run on a 10% SDS-polyacrylamide gel. Following electrophoresis, the gel was fixed in 10% acetic acid and 40% methanol for 30 min, transferred to Amplify solution (Amersham Pharmacia Biotech), and treated for 30 min. The gel was dried and exposed to X-AR film (Eastman Kodak Co.).
Epitope Tagging and Transfections-The HA-epitope (YPYDVPDYA) from the influenza virus hemagglutinin protein HA1 (46) was added, along with a SacII restriction site, to the N terminus of both 16K and TM4 by PCR amplification using an upstream primer encoding the epitope, 5Ј-TCCCCGCGGATGTACCCCTACGACGTGCCCGACTACG-CCGCTGACATCAAGAACAACCCCGAA-3Ј, and the downstream primers used to make GST-16K and GST-TM4. Initially, SacII-EcoRI fragments from these PCR products were cloned into pTRE and expressed with the tet-On system (CLONTECH). In later experiments we switched to using the pXJ40-KK0 vector, which enables consistent high levels of expression from an SV40 promoter, by amplifying the tagged full-length or TM4 fragments with terminal EcoRI and BamHI sites to allow insertion into this vector. The effects of 16K and TM4 on cell morphology were seen with both pTRE and pXJ40-KKO systems, but the latter proved to be a more reliable vector. Stable cell lines were made by cotransfecting vector plasmids with the hygromycin B resistance plasmid pTK-hyg (CLONTECH), using the calcium phosphate procedure, and selecting for colonies that grew in 200 g/ml hygromycin.
Cell Culture Immunohistochemistry-Cells were grown on glass coverslips in dishes, media was removed and cells were washed twice with PBS. Cells were fixed with 4% paraformaldehyde in PBS, washed three times with PBS, and permeabilized by incubation with 0.1% Triton X-100 for 30 min. They were again washed three times in PBS, then blocked in 10% rabbit serum in PBS. Primary antibody was diluted according to suppliers recommendations in 2% rabbit serum and incubated with cells for 1 h, followed by one wash with PBS, and incubation with secondary antibody for 30 min at room temperature. For doublelabeling experiments this procedure was then repeated for the other primary and secondary antibodies, after which the cells received three 10-min washes in PBS. They were then covered with mounting medium, affixed to glass slides, and viewed using a Bio-Rad MRC-600 laser scanning confocal microscope equipped with a krypton-argon mixed gas laser. Cells treated with preimmune rabbit serum instead of primary antibodies, or untransfected control cells treated with anti-HA antibody, served as negative controls. Single-labeling experiments in which the secondary antibodies were mismatched with the primary antibodies demonstrated that there was no cross-reactivity between antibodies in the double-labeling experiments.
Bead Binding Assays-Polystyrene latex beads (Sigma) were coated with fibronectin, polylysine, and bovine serum albumin as described by Grinnell and Geiger (47). Approximately 3 ϫ 10 6 polystyrene latex beads (11.9 m diameter) were incubated for 1 h at 22°C in 0.5 ml of ␣-MEM containing 25 g of human plasma fibronectin (Sigma), 50 g of polylysine (Sigma), or 50 g of bovine serum albumin. Bead binding assays were performed according to the procedure of Burbelo et al. (48). Cells were incubated with 25 g/ml cycloheximide (Sigma) in fibronectin-depleted serum. After 2 h cells were washed twice with PBS, then detached with 0.05% trypsin and 25 g/ml cycloheximide. Cells (1 ϫ 10 5 ) were subsequently plated on coverslips for 60 min at 37°C in ␣-MEM containing 25 g/ml cycloheximide. Each type of bead (approximately 2 ϫ 10 6 ) was incubated with cells at 37°C for 20 min. Cells were then fixed and stained with anti-16K antibody as described above. Controls in which primary antibody was replaced with preimmune serum showed no immunofluorescence.

RESULTS
␤ 1 Integrin Interacts with 16K-The TRIC bait for the two hybrid screen comprised the transmembrane domain of ␤ 1 integrin, flanked by 16 amino acids on the extracellular side and the full intracellular cytoplasmic domain (Fig. 1A). This was done to avoid using only a short highly hydrophobic bait. Nine his ϩ lacZ ϩ clones that did not react with the control lexA-lamin bait or with the empty BTM116 plasmid were obtained from screening 2 ϫ 10 6 transformants, all encoding proteins that interacted with this bait and the ␤ 1 -TR bait, but not the ␤ 1 -IC bait consisting of the intracellular domain flanked by only four transmembrane residues. In addition, none of these proteins interacted with a bait containing the transmembrane domain of the ␣ 5 integrin subunit. Expression of both the ␣ 5 and ␤ 1 transmembrane (TR) bait proteins in yeast was similar, as detected by an anti-LexA antibody (Fig. 1A), but none of the proteins interacted with the ␣ 5 bait. Thus, the lack of an interaction between 16K and ␣ 5 -TR was not simply due to poor expression of that bait protein, and the interaction with the ␤ 1 transmembrane domain was not due solely to a nonspecific interaction of hydrophobic helices. Three of the cDNA clones recovered in this two hybrid screen encoded overlapping segments of the 16-kDa (16K) subunit of V-ATPase (Fig. 1C), two contained identical regions of an uncharacterized human cDNA, KIAA0025, previously isolated from a human immature myeloid cell line (49), one encoded a fragment of a sugar transport protein (50), and three did not share homology with any data base entries. All contained hydrophobic regions.
Of the clones identified, 16K has several known functions that were intriguing with respect to potential involvement with integrins. The V-ATPases are responsible for acidification of vesicles, which determines whether receptor-ligand complexes are degraded or recycled (19,21). Additionally, the 16K subunit can independently of the ATPase form gap junctions (30).
All of the two hybrid isolates of 16K were missing the first transmembrane domain, proteins we refer to as TM1. A representative filter of the two hybrid assay showing the interaction of one of the three TM1 isolates with the ␤ 1 -TR and ␤ 1 -TRIC baits is shown in Fig. 1B (center). To confirm that the absence of the first transmembrane helix or some part of the 3Ј-nontranslated region in these clones was not creating an artifact in the screen, we expressed the full-length 16K coding region in the prey plasmid pVP16 and found it to also interact specifically with the ␤ 1 -TRIC and ␤ 1 -TR baits, and not the ␤ 1 -IC bait (Fig. 1B, left). Andresson et al. (33) reported that the last helix of 16K, TM4, was required for its interaction with the E5 protein of bovine papillomavirus. To determine whether the interaction with ␤ 1 integrin also required this helix, we constructed a TM4 deletion mutant of 16K in pVP16, and found that it was no longer able to interact with any of the baits in a two-hybrid assay (Fig. 1B, right). When we tested the fulllength 16K, TM1, and TM4 proteins with the ␣ 5 -TR bait, none were found to interact, and filters with no colonies were obtained, identical to that seen when TM4 was tested.
In Vitro Binding Assays-GST fusion proteins of full-length 16K, or the mutants lacking TM1 or TM4, were produced in bacteria, coupled to glutathione-agarose beads and assessed for their ability to bind fragments of ␤ 1 integrin. The original TRIC bait, the transmembrane domain alone (TR only), or the cytoplasmic domain alone (IC only) were synthesized as tripartite fusions with GST and a fragment of protein A, using the pAGEX2T vector (Fig. 2). Cleavage of these fusions with thrombin released soluble protein A-tagged proteins (protA-TRIC, protA-TR only, protA-IC only). The protein A moiety allows for direct detection with antibody conjugates, thus eliminating the need for different primary antibodies specific for each integrin fragment, and thereby enabling quantitative comparison of the abilities of the different fragments to bind to the protein on beads. The protein A-tagged molecules were incubated with the beads to which equivalent amounts (2 g), based on Coomassie staining (lanes 11-13), of GST-TM1, full-length GST-16K, or GST-TM4 were coupled, and complexes recovered by centrifugation. Western blot analysis of these complexes with an antibody conjugate showed that the TRIC fragment used in the initial two hybrid screen was able to bind full-length 16K(GST-16K) and 16K lacking TM1 (GST-TM1) with similar affinities (Fig. 2, lanes 3 and 4), in both cases retaining approximately 5% of the input protA-TRIC molecule (one quarter of the input is shown in lane 1), even after five rounds of washing the beads. The removal of TM4 prevented binding (lane 2). Various negative controls for the protA-TRIC binding were used, including beads carrying no fusion protein, beads loaded with only GST, and beads loaded with a DNA-binding protein, the transcription factor TEF-1. None of these bound the input molecule (data not shown). Since GST-TM1 bound as efficiently as GST-16K, and was easier to purify from bacterial cells, it was used for subsequent experiments. The ␤ 1 transmembrane region alone was similarly able to bind GST-TM1 coupled to beads (lane 7), whereas the intracellular region was not (lane 10). Neither "protA-TR only" nor "protA-IC only" bound to glutathione-agarose beads to which no fusion protein was coupled (lanes 6 and 9). The binding assays in Fig. 2 were reproduced with at least 10 independent preparations of the fusion proteins and negative controls.
Coimmunoprecipitation and Co-localization of ␤ 1 Integrin and 16K-Further evidence that ␤ 1 integrin and 16K interact in vivo was obtained from co-immunoprecipitation analyses, using extracts from rat L6 myoblasts. These cells express ␤ 1 integrin and produce visible focal adhesions (Ref. 51 and our own observations). Immunoprecipitation of RIPA lysates with anti-16K recovered a protein that was reactive to anti-␤ 1 antibody, and identical in size to immunoprecipitated ␤ 1 (Fig. 3,  lanes 1 and 3). We also metabolically labeled cells with [ 35 S]methionine, and immunoprecipitated with anti-16K an identically sized labeled product, which comigrated with one of the proteins present in complexes pulled down by anti-␤ 1 (Fig. 3, lanes  5 and 6). With immunoprecipitations of non-labeled extracts, we detected several molecular mass forms of ␤ 1 interacting with 16K (lane 3). The larger of these (ϳ160 -170 kDa) was less efficiently pulled down with our anti-␤ 1 antibody (lane 1), which recovered primarily the ϳ125-kDa ␤ 1 species also seen in the radiolabeled extracts.
Double fluorescence labeling was used to localize ␤ 1 integrin and 16K in L6 cells and in an unrelated human pancreatic tumor cell line (Capan 2), chosen because of reports that 16K is up-regulated in human metastatic pancreatic cancer (42). Con-sidering that 16K can be present both as part of the V-ATPase complex and on its own in gap junctions, it was anticipated as well that ␤ 1 would likely not co-localize to all sites where 16K is found. In agreement with this, in L6 cells 16K was found predominantly in the cytoplasm (Fig. 4A). However, a portion localized to focal adhesions, where much of the ␤ 1 integrin (Fig.  4B) was located. The pancreatic cell line Capan 2 also showed co-localization of ␤ 1 with 16K in the plasma membrane, both in isolated cells (Figs. 5, B and E) and in clusters of compacted cells that typify the growth form of this cell line in culture (Fig.  5, A and D). The co-localization of 16K with ␤ 1 was largely independent of the V-ATPase, since an antibody to the en-FIG. 1. Interacting 16K cDNA clones isolated in the yeast two-hybrid screen. A, the transmembrane-spanning regions of ␤ 1 and ␣ 5 integrins, from which the bait proteins were constructed, are shown at the top. The dotted lines below each sequence indicate the residues included in each of the baits. The inset shows a Western blot of yeast cell extracts prepared from cells expressing the ␤ 1 -TRIC, ␤ 1 -TR, and ␣ 5 -TR fragments. The blot was probed with anti-LexA antibody, and two independent transformants are shown for each of the three baits indicated. B, representative membrane filters on which yeast cells were screened for two-hybrid interactions between ␤ 1 fragments and several different 16K partners. The full-length 16K protein, an original isolate from the two-hybrid screen that lacks the first transmembrane helix (TM1), and a mutant lacking the fourth transmembrane helix (TM4) were made as fusions with the VP16 activation domain and screened with the various bait fusion proteins indicated at the periphery around each circle. Negative controls included transformation of these plasmids alone (no bait plasmid), with the bait plasmid (empty bait plasmid) or with an irrelevant protein (lamin). No growth on selective media was seen in these controls. The VP16 fusion proteins were also tested for their abilities to interact with the ␤ 1 -TRIC, ␤ 1 -TR, and ␤ 1 -IC fragments. C, the full-length reading frame of 16K (155 amino acids) is shown as an open box at the top. The graph is a hydropathy plot (Kyte and Doolittle), showing four hydrophobic membrane-spanning regions, TM1 to TM4. The three non-identical but overlapping regions of 16K that interacted uniquely with the TRIC and TR baits, but not the IC bait are shown as black bars above the graph, with their N-terminal amino acid indicated, and the number of base pairs downstream of the stop codon shown at the right.
zyme's 70-kDa subunit (70K) showed that it was largely absent from focal adhesions (Fig. 4, C and D). Capan 2 cells showed a striking dissimilarity in 70K staining versus that of either 16K or ␤ 1 . In clumps of cells (Fig. 5C) and single cells (Fig. 5F), the 70K staining was predominantly in the cytoplasm and absent from plasma membranes and surfaces where cell-cell contact was occurring.
Epitope Tagging of 16K-The inability of the TM4 mutant to bind ␤ 1 raised the possibility that this truncated protein's distribution in cells might be altered. To permit distinction of a TM4 mutant from native 16K, we generated stable L6 cell lines expressing either HA-epitope-tagged full-length or TM4 mutant proteins. Cells expressing the TM4 mutant showed variations from typical myoblast structure, and the HA-TM4 protein (Fig. 4G) had a more perinuclear distribution than that of epitope-tagged full-length 16K (Fig. 4E). The HA-tagged fulllength protein, but not tagged TM4 protein, showed a distribution that mirrored that of ␤ 1 integrin, similar to native untagged 16K (panel A). ␤ 1 staining revealed that focal adhesions in the TM4-expressing cells (panel H) were less punctate than in untransfected control cells (panels B and D). Nevertheless, these cells expressing TM4, when stained with anti-16K and anti-␤ 1 , had endogenous 16K that colocalized with ␤ 1 to these structures (data not shown). These experiments showed that FIG. 2. In vitro binding of ␤ 1 and 16K. At the top is shown the C-terminal TRIC segment of ␤ 1 integrin used in the two-hybrid screen, as well as two additional fragments consisting of only the transmembrane or intracellular regions. The lower part of the figure shows Western blot assays used to detect whether protein A-tagged integrin fragments were retained on glutathione beads to which 2 g of different GST fusions of 16K were coupled. Lanes 2-4 show binding of the 25-kDa protA-TRIC molecule to full-length 16K (GST-16K) and to 16K lacking TM1 (GST-TM1), but not to a mutant lacking TM4 (GST-TM4). Lanes 6 and 7 show binding of the 17-kDa "protA-TR only" molecule to beads carrying GST-TM1 but not to beads alone (-ve). Lanes 9 and 10 show that the 20-kDa "protA-IC only" molecule is not able to bind to GST-TM1. Lanes 1, 5, and 8 contain one-fourth (0.5 g) of the soluble protA-tagged molecule used in each of the three sets of experiments. Lanes 11, 12, and 13 show Coomassie-stained aliquots of the GST-TM1, GST-16K, and GST-TM4 proteins used in the assays, confirming that the amount of each protein that was used in the binding assays was equivalent.

FIG. 4. Co-localization of native and epitope-tagged full-length
16K with ␤ 1 integrin in focal adhesions in L6 myoblasts. Cells were grown on coverslips, fixed, and doubly labeled. All were stained for ␤ 1 integrin, using the HM␤1-1 hamster monoclonal antibody, mouse anti-hamster IgG secondary antibody, and a fluorescein isothiocyanateconjugated goat anti-mouse tertiary antibody. The cells were co-stained with either rabbit anti-16K, a rabbit antibody to the 70-kDa subunit of V-ATPase (70K), or rabbit anti-HA followed by Texas Red-conjugated goat anti-rabbit antibody. The right column (panels B, D, F, and H) shows ␤ 1 staining, while the left column (A, C, E, and G) shows the location of native 16K (A), native 70K (C), epitope-tagged 16K (E), and epitope-tagged TM4 mutant (G) within the same cells. Arrows point to focal adhesions. Scale bar ϭ 25 m. the fourth transmembrane helix of 16K is important for the cellular localization of the protein in vivo, and plays a role in defining focal adhesions.
Fibronectin-dependent Redistribution of 16K-Although the mechanisms whereby integrins become redistributed in the membrane are not known, the presence of integrin ligands can lead to formation of focal adhesions (52). We used fibronectincoated beads to cluster integrins in L6 cells, and looked for a concurrent redistribution of 16K. Cells were first trypsinized and incubated with cycloheximide in fibronectin-depleted serum to obtain an initial random dispersement of fibronectin receptors. The binding of cells to fibronectin-coated beads then can then be used to induce redistribution of receptors to the points of contact (53). When this experiment was done, 16K redistributed to a pool at those contact points (Fig. 6A). Beads coated with polylysine, a positively charged nonspecific adhesion promoting polypeptide, bound to cells, but 16K did not localize to the contact points (panel C). These results suggest that insoluble fibronectin initiates redistribution of 16K by clustering and immobilizing ␤ 1 integrin.
Altered Morphology of L6 Myoblasts and Ref52 Cells-The above experiments predicted that perturbation of 16K within cells could potentially lead to a change in cell-ECM interactions, and in this regard it had been previously reported that the expression of a TM4 mutant of 16K in 3T3 fibroblasts led to anchorage-independent growth as assessed by formation of colonies in soft agar (33). We used L6 and rat embryonic fibroblast (REF52) cells to inquire whether 16K might be involved in the regulation of cell morphology and growth (Fig. 7). Stable transfectants overexpressing full-length 16K, or expressing the TM4 mutant, showed a change in morphology in both cell types. L6 cells with 16K became more spindle-shaped, and divided in such a way as to remain lined up lengthwise. The TM4 mutant protein, in contrast, caused them to have less well defined edges, and to show reduced effects of contact with other cells. REF52 cells overexpressing full-length 16K became strikingly elongated and underwent a conversion to a flattened morphology, while maintaining long tenuous end-to-end connections. REF52 cells expressing the TM4 mutant became shorter and broader than control cells, with less well defined edges. The full-length REF-16K transfectants had a doubling time approximately twice that of control cells or REF-TM4 transfectants, a possible consequence of their highly extended shape. Thus, with both a myoblast and a fibroblast cell line, the overexpression of wild type 16K led to an elongation of the cells, while the TM4 mutant led to a smaller, more distorted appearance. The conclusions from these experiments were based on three independently isolated stable cell lines for both full-length and TM4 transfectants in both cell types, all of which exhibited the described phenotypes. DISCUSSION Integrins, like many membrane-bound receptors, are directed after synthesis to the cell membrane in vesicles that bud off from the endoplasmic reticulum and cycle through the Golgi complex (54). These vesicles emerge from the trans-Golgi and fuse with the cell membrane, extruding the amino termini of integrin chains to the extracellular environment. Integrins are not thought to be in this way directed to focal adhesions, but rather to redistribute within the membrane to those points in response to binding to an ECM ligand (4). This response has been shown, at least with ␤ 1 and ␤ 3 integrins, to involve the cytoplasmic domain, perhaps through interaction with a cytoskeletal component, with ligand binding overcoming a block to directed movement (55)(56)(57).
The finding of an interaction between the 16-kDa subunit of V-ATPase and the transmembrane domain of ␤ 1 integrin raises a number of new possibilities for how integrins move. Specifically, 16K could be an important mediator of integrin movement via vesicles that shuttle receptor-ligand complexes through the cell. These complexes are engulfed in coated and non-clathrin-coated vesicles en route to endosomes, where 16K is abundant and where the V-ATPases play a key role in regulating acidification that leads to complex dissociation (58 -61). Our observation that fibronectin-coated beads caused 16K, and as seen in Fig. 6B, associated vesicles, to pool at the site where contact between the cell and the bead was made is consistent with the hypothesis that at those points there is active engulfment and cycling of receptors. Experiments using fibronectincoated beads and anti-␤ 1 integrin-coated beads showed a strikingly similar accumulation of ␤ 1 integrin, ␣-actinin, actin, talin (53), and rho (48) in the cytoplasm adjacent to the site of attachment to beads. It is a reasonable hypothesis that 16K is helping to direct and stabilize ␤ 1 transmembrane interactions during the recycling process.
Vesicles derived when clathrin is removed can also shuttle complexes through the process of transcytosis, which ␤ 1 integrin has been observed to undergo in fibroblasts, hepatic epithelial cells, and Chinese hamster ovary cells (62,63). In light of this, it is possible that focal adhesions form when integrins that are being continuously endocytosed and recycled back to the surface of the cell become retained only at points where they contact the ECM. The redistribution need not involve movement through the membrane, but rather a repetitive recycling that "finds" the points where the ECM would allow for formation of a focal adhesion. Once integrins contact a ligand, their half-life at that point is extended (64), and subsequent cytoskeletal proteins of the focal adhesion aggregate. The ligand-dependent requirement of the cytoplasmic domain of integrins for their relocation to focal adhesions can be considered in light of this hypothesis of integrin cycling. The need for ligand binding could simply reflect the need to get integrins into mobile vesicles through internalization of receptor-ligand complexes. Once internalized only the cytoplasmic domain would protrude from these vesicles, and through its interactions with talin, vinculin, paxillin, ␣-actinin, and other cytoskeletal associated proteins direct vesicle movement along the cytoskeletal network that organizes in response to focal adhesions.
The importance of transcytosis and integrin recycling also has relevance to migratory cells, where it has been proposed that integrins holding the cell to the matrix move rearwards as the cell advances, thereby polarizing within the cell the formation and dissolution of integrin-ligand complexes (65). Further attachments at the cell's leading edge would require replenishments of matrix receptors to move the cell forward. Consistent with this, it has been shown that neutrophil migration involves endocytosis of integrins from the rear of the cell, followed by transport through the cell and exocytosis at the leading edge (66).
The perturbation of cell morphology observed in our fulllength and TM4 transfectants suggests a role for 16K in regulating cell-ECM interactions. The TM4 mutant, although reportedly defective in ATPase activity (67) and ␤ 1 binding (our data), is nevertheless able to interact with 16K (33). One interpretation is that the presence of the TM4 mutant proteins in the ATPase interferes with the normal acidification of endosomes and leads to greater recycling of intact receptors to the cell membrane, thereby promoting migration and leading to cells with shorter extensions. Bretscher (65) predicted, based on his analysis of fibronectin receptor cycling (62), that fibroblasts would be longer if receptor cycling were impaired. Overexpression of native 16K led to elongation of REF52 cells, possibly because the additional molecules caused ␤ 1 to be aberrantly retained in the cell membrane. An impairment of endosome acidification by the TM4 mutant is reminiscent of the observation that the E5 transmembrane oncoprotein of papillomaviruses, which binds 16K (32), also prevents this acidification, and leads to enhanced receptor signaling (68).
The involvement of 16K with integrins also provides an interesting link with the observation that adhesion of fibroblasts to fibronectin leads to alkalinization of the cytoplasm (69). The V-ATPases do not need a counter-ion when they pump protons across membranes (70), and they are one of the principle enzymes that can affect cellular pH. The movement of H ϩ ions from the cytoplasm into acidic vesicles could be enhanced when contact with a particular matrix component leads to reorganization of the repertoire of receptor-ligand complexes, a process that would necessarily involve endosome processing. Within the V-ATPase enzyme, the membrane bound hexamer of 16K molecules, known as the V 0 subunit of the enzyme, provides a docking site for the V 1 subunit that includes the 70K protein, which leads to ATPase activity (67, 71). Immunofluo- rescence experiments showed that the distribution of 70K does not extend into the cell membrane to the extent of 16K, raising the possibility that membrane 16K is playing a role in receptor internalization, and once internalized in vesicles could provide attachment sites for the remainder of the V-ATPase. It is unlikely, based on the 70K staining, that all of the membrane 16K we observe is part of a functional V-ATPase enzyme. Some of the 16K could be serving as gap junctions.
There is additional evidence that 16K might play a role in the membrane independently of its direct participation in V-ATPase or gap junctions. The papillomavirus E5 oncoprotein binds both 16K and the PDGF-␤ receptor (31,32). The TM4 mutant of 16K that fails to interact with ␤ 1 also cannot bind E5, and can cause anchorage-independent growth in 3T3 cells (33). There is strong evidence that signaling through both PDGF-␤ and ␤ 1 integrins converges. Both PDGF and fibronectin lead to phosphorylation of the PDGF-␤ receptor itself and of focal adhesion kinase FAK 125 , and to recruitment of paxillin to focal adhesions (14). 16K coimmunoprecipitates in a trimeric complex with E5 and PDGF-␤ receptor and is proposed, through its interaction with E5, to facilitate dimerization of the receptor and resulting signaling in the absence of PDGF (72,73). 16K is in this way implicated in transformation by E5. Although we can only speculate at present, it is possible that 16K mediates the reported PDGF-independent activation of the PDGF-␤ receptor following ␤ 1 integrinligand interaction.
Recently, another integrin-binding protein, integrin-linked kinase, was identified as binding to the cytoplasmic domain of ␤ 1 and ␤ 3 integrins, and it was reported that overexpression of integrin-linked kinase altered the morphology of intestinal epithelial cells and induced anchorage-independent growth (74,75). These effects are reminiscent of the effect of the 16K TM4 mutant on 3T3 fibroblasts (33) and are consistent with our observation that TM4 alters the morphology of adherent cells. It is possible that integrin-linked kinase overexpression leads to its accumulation onto ligated fibronectin receptors and disrupts their internalization and recycling.
The involvement of 16K with ␤ 1 integrin bridges cell membrane-ECM interactions with the internal processing and trafficking of molecules. It will be of interest to determine whether 16K is in fact involved with more than just ␤ 1 and PDGF-␤ receptor functions, and whether observations showing its overexpression in some cancer cells (42) and the inhibition of cell growth by an inhibitor of V-ATPases, bafilomycin (76,77), reflect a major role of 16K in cell growth control.