Suppression of Tumor-related Glycosylation of Cell Surface Receptors by the 16-kDa Membrane Subunit of Vacuolar H+-ATPase*

The glycosylation of integrins and other cell surface receptors is altered in many transformed cells. Notably, an increase in the number of β1,6-branched N-linked oligosaccharides correlates strongly with invasive growth of cells. An ectopic expression of the Golgi enzymeN-acetylglucosaminyltransferase V (GlcNAc-TV), which forms β1,6 linkages, promotes metastasis of a number of cell types. It is shown here that the 16-kDa transmembrane subunit (16K) of vacuolar H+-ATPase suppresses β1,6 branching of β1integrin and the epidermal growth factor receptor. Overexpression of 16K inhibits cell adhesion and invasion. 16K contains four hydrophobic membrane-spanning α-helices, and its ability to influence glycosylation is localized primarily within the second and fourth membrane-spanning α-helices. 16K also interacts directly with the transmembrane domain of β1 integrin, but its effects on glycosylation were independent of its binding to β1integrin. These data link cell surface tumor-related glycosylation to a component of the enzyme responsible for acidification of the exocytic pathway.

Integrins and other cell surface receptors are extensively N-glycosylated by enzymes that reside in the endoplasmic reticulum (ER) 1 and Golgi complex. Upon the insertion of newly synthesized polypeptides into the ER, oligosaccharyltransferase adds Glc 3 Man 9 GlcNAc 2 -P-P-dolichol to asparagine residues. The three glucoses and all but the last five mannose residues on the oligosaccharide core are progressively trimmed off as the protein moves from the ER to the cis-Golgi by a series of glucosidases and mannosidases. Upon entering the medial Golgi, the N-acetylglucosaminyltransferases GlcNAc-TI and GlcNAc-TII begin the process of rebuilding the oligossacharides with GlcNAc residues. In various cells and tissues, GlcNAcs-TIII, -TIV, and/or -TV make subsequent GlcNAc additions, and this rebuilt backbone serves as a structure to which other complex sugars are added.
Many cancer cells exhibit altered glycosylation patterns on surface receptors, but the complexity and differences in oligosaccharides in normal cells have made it difficult to establish causal relationships between sugar modifications and abnormal cell growth properties. Nevertheless, there is a growing body of convincing evidence that hyperactivity of GlcNAc-TV promotes a metastatic phenotype (1)(2)(3)(4)(5)(6)(7)(8)(9). ␤ 1 integrin and the epidermal growth factor receptor (EGF-R) are both targeted by GlcNAc-TV (10 -13), which adds GlcNAc to the oligosaccharide backbone via a ␤1,6 linkage. In addition, both receptors can have significant involvement in cancer cells. ␤ 1 integrin is implicated in the invasive processes of many tumor cells (14 -19), and an elevated expression of the EGF receptor is an indicator of poor prognosis for many cancers including breast, ovarian, and uterine (20 -23).
We showed recently that the transmembrane domain of ␤ 1 integrin interacts with the 16K subunit of vacuolar H ϩ -ATPase (V-ATPase) (24), the enzyme that acidifies the Golgi and exocytic and endocytic compartments (25,26). 16K is a membranespanning protein that folds as a four-helix bundle and assembles into a hexamer, forming the membrane proton channel of the enzyme. In addition to its role in the V-ATPase, 16K has been reported to form gap junctions and neurotransmitter release channels (27)(28)(29). It is believed to play a role in the function of the beta form of platelet-derived growth factor receptor (PDGF-␤R) with which it interacts and participates in trimeric complexes with the E5 protein of papillomaviruses (30). Deletion of the fourth transmembrane helix of 16K leads to reduced interactions with E5, the PDGF-␤R, and ␤ 1 integrin (31,24), and this truncated protein induces anchorage-independent growth in 3T3 cells and enhances their growth as tumors in nude mice (31). Even less is known about how 16K might regulate ␤ 1 integrin functions. A role for 16K was suggested by experiments showing that fibronectin induces a redistribution of 16K-containing vesicles into proximity of the points of contact between the cell and the extracellular matrix (24).
Considering that integrin functions are dependent upon glycosylation and because all of the molecules of interest including E5, 16K, the PDGF-␤R, the EGF-R, and ␤ 1 integrin pass through the ER and Golgi, we conducted a study to assess whether 16K regulates receptor processing. Using lectins specific for different processing intermediates, we found that 16K inhibits the addition of ␤1,6-branched oligosaccharides to both ␤ 1 integrin and the EGF-R. This result correlated with a loss of invasive growth. Each individual transmembrane helix of 16K was expressed, and both the second and fourth helices were able to inhibit ␤1,6 branching. This inhibition occurred independently of direct binding to the receptors. These and other data highlight a critical role for the ␤ 1 integrin transmembrane domain in the glycosylation of the protein and reveal that 16K can regulate processing events that are implicated in cancer.

EXPERIMENTAL PROCEDURES
Cell Culture, Transfections, Antibodies, and Lectins-Human embryonal kidney cells (HEK293) constitutively expressing T7 polymerase (a gift from Dr. M. A. Billeter, Institut fur Molekularbiologie, Abteilung, Switzerland) were grown in ␣-minimum Eagle's medium in 10% fetal bovine serum at 37°C in 5% CO 2 . All transfections were done using a standard calcium phosphate procedure. Unless otherwise described, HEK293 cells were incubated with precipitate for 3 h, and lysates were made after 24 h of expression. The HSV and T7 antibodies were obtained from Novagen, and the anti-PDGF-␤R was from Santa Cruz Biotechnology Inc. Alexa 488-cojugated goat anti-mouse was obtained from Molecular Probes, Inc. Agarose-conjugated Phaseolus vulgaris leucoagglutinin (L-PHA) and concanavalin A (ConA) and alkaline phosphatase-conjugated rabbit anti-mouse IgG were purchased from Sigma. Alkaline phospatase-conjugated goat anti-rabbit was from Chemicon.
Assembly of T7-tagged Wild Type and W728G Mutant ␤ 1 Integrin-Oligos encoding the 22 amino acid rat ␤ 1 integrin signal sequence (32) followed by the 11 amino acid T7 epitope (33) were generated, annealed, and ligated into the BamHI and EcoRI sites of pXJ41 to make XJ41-T7. These oligos also added an XbaI site upstream of the BamHI site. The assembled full-length bovine ␤ 1 integrin cDNA (34) was inserted into this vector by directionally cloning PCR-amplified products using the upstream primer 5Ј-GCTCTAGAGAAAATAGATGTTTG-3Ј and downstream primer 5Ј-CCGCTCGAGTCACTCATACTTCGGATT-3Ј into the XbaI and XhoI sites of XJ41-T7 (pXJ41-T7-␤ 1 ). The mutation of Trp 728 of ␤ 1 integrin to glycine was done following the procedure of Kirsch and Joly (35) using the upstream primer 5Ј-GCTGAGCATAAAGAATGT-3Ј and the downstream primer 5Ј-TAAAAGCTTCCCAATCAGCA-3Ј.
Cloning of Human 16K cDNA and Generation of 16K Mutants-The human 16K cDNA was cloned by reverse transcription-PCR from the human pancreatic tumor cell line CRL-80 using the primers (H16K-Up) 5Ј-ACATGTCCGAGTCCACG-3Ј and (H16K-DN) 5Ј-CTACTTTGTG-GAGAGGATG-3Ј. The HSV-tagged 16K cDNA was made by directionally cloning a PCR product into the EcoRI and BamHI sites of the plasmid XJ40-KKO (36), which adds the HSV tag to the carboxyl terminus of the protein. The 16K PCR fragment was generated using the primers (16K3-UP) 5Ј-CGCGAATTCATGTCCGAGTCCAAGA-3Ј and (16K-3DN) 5Ј-CGGGATCCCTTTGTGGAGAGGATG-3Ј. 16K mutants were generated by PCR amplification using primers with appropriate restriction enzymes spanning amino acids identified in Fig. 4. PCR fragments were HSV epitope-tagged at the carboxyl terminus by cloning into pXJ40-KKO.
Assembly of T7-tagged EGF Receptor-The EGF-R was cloned into the pXJ41 expression vector, and oligos encoding the 71 amino acid signal sequence (37) followed by the T7 epitope tag were annealed and ligated upstream of the coding sequence.
Western Blot Analysis-RIPA lysates made from HEK293 cells transiently transfected with pXJ41-T7-␤ 1 were resolved on 8% SDS-PAGE, transferred to nitrocellulose, blocked with 5% skim milk, and probed with anti-T7 antibody. Detection was with nitro blue tetrazolium/5bromo-4-chloro-3-indolyl phosphate following incubation with alkaline phosphatase-conjugated rabbit anti-mouse IgG. Alternatively, lysates were treated with endoglycosidase H (2 units) or glycopeptidase F and incubated at 37°C overnight in recommended buffers by the manufacturer (Calbiochem). Agarose-conjugated L-PHA and ConA were used to isolate ␤1,6-branched and high mannose forms of ␤ 1 integrin, respectively. Lysates made from cells co-transfected with HSV-tagged 16K were treated with anti-HSV antibody and agarose-conjugated protein A, and recovered complexes were analyzed by Western blot as described above. For experiments in which 16K was co-transfected with ␤ 1 integrin or the EGF-R, total amounts of DNA were made equal by the addition of parental pXJ41 vector.
Migration/Invasion Assays-Invasion assays (38) were performed in Costar transwell chambers in which 10 g of bovine plasma fibronectin or laminin (Sigma) was adsorbed overnight. Before adding the cells, the protein matrix was rehydrated for 2 h with ␣-minimum Eagle's medium without serum. 2 ϫ 10 4 HEK293 cells transiently transfected with 16K were added to the top chamber and were allowed to penetrate the matrix for 22 h at 37°C. The chambers were then washed three times with PBS, and any remaining cells were removed from the top surface using a cotton swab. Cells that had penetrated the membrane and reached the lower surface were detected by Giemsa staining and counted. Control cells transfected with the empty pXJ41 vector (mock transfected) were able to invade all matrices tested.
Adhesion Assays and Immunofluorescence-HEK293 cells were transfected with HSV-tagged 16K for 12 h, split, and aliquoted onto untreated coverslips or coverslips coated with 1.8 g of fibronectin or laminin. After an overnight incubation at 37°C, cells were fixed in 4% formaldehyde, permeabilized with 0.1% Triton X-100, blocked in 10% goat serum, and treated with anti-HSV antibody for 1 h at room temperature. Following three washes in PBS, cells were incubated in 2% goat serum containing Alexa-488-conjugated rabbit anti-mouse secondary antibody, washed, mounted, and viewed using a Zeiss Axiovert 200.
Pulse-Chase Experiments-HEK293 cells transiently transfected with ␤ 1 integrin or co-transfected with 16K were split 12 h following transfection. After 24 h, cells were preincubated for 45 min in methionine-free Dulbecco's modified Eagle's medium containing 2% fetal bovine serum and 2 mM glutamine, washed twice with PBS, and then incubated in fresh methionine-free medium containing 2% fetal bovine serum, 2 mM glutamine, and 0.3 mCi/ml [ 35 S]methionine for 5 min at 37°C. Cells were washed twice with PBS and either lysed immediately in RIPA buffer or incubated in fresh Dulbecco's modified Eagle's medium with excess methionine and 10% fetal bovine serum for 10 or 20 min or 1, 2, 6, or 18 h, respectively, followed by RIPA buffer lysis. Immunoprecipitations using anti-T7 antibody were recovered with protein A-Sepharose and resolved on 8% SDS-polyacrylamide gels, which were then subjected to fluorography (Amplify, Amersham Biosciences, Inc.), dried, and exposed to film. Nontransfected HEK293 cells labeled for 4 h and treated with anti-T7 antibody were used as a negative control.
Flow Cytometry-Cells transfected with wild type T7-tagged ␤ 1 integrin or with W728G mutant-tagged integrin were fixed in 2% formaldehyde and transferred to a microfuge tube. Cells were washed three times in PBS and recovered by low speed microcentrifugation. Cells were then incubated with anti-HSV antibody for 1 h at 4°C, washed, and incubated with Alexa-488-conjugated secondary antibody for 30 min at 4°C. Cells were washed three times with PBS and immediately analyzed using a Coulter Elite flow cytometer.

Identification of Processing
Intermediates of ␤ 1 Integrin-16K and ␤ 1 integrin were tagged with HSV and T7 epitopes, respectively, to permit specific detection following transfection into HEK293 cells. ConA was used to detect molecules with terminal mannose residues that are found predominantly in the ER, and the lectin L-PHA was used to detect the ␤1,6branched GlcNAc residues added by GlcNAc-TV. The T7 epitope-tagged ␤ 1 integrin produced three main products (Fig.  1, lane 1) that were identical in size to endogenous forms of ␤ 1 integrin (data not shown). The middle band (ϳ120 -125 kDa) was reactive with ConA (lane 3) and sensitive to endoH glycosidase (lane 5), identifying it as the high mannose form present in the ER. Its size was reduced to that of the ϳ110 -115 kDa lower form by both endoH and glycopeptidase F (lane 4). The lower form did not react with either ConA or L-PHA, identifying it as unglycosylated core protein. The upper form (ϳ125-135 kDa) contained L-PHA-reactive molecules (lane 2), identifying it as the most mature product. Treatment with brefeldin A, which blocks the transport from the ER to the Golgi, inhibited the appearance of only the uppermost band, confirming that the intermediate and low molecular weight forms were ER intermediates (data not shown).
The previously reported interaction of endogenous 16K and ␤ 1 integrin in rat L6 myoblasts (24) was confirmed in HEK293 cells by co-immunoprecipitation experiments using the epitopetagged proteins. Lysates from cells co-transfected with T7tagged ␤ 1 integrin and HSV-tagged 16K were immunoprecipitated with anti-HSV antibody, and complexes were analyzed by Western blots using anti-T7 antibody. The ϳ110 kDa core protein was predominantly found in association with ␤ 1 integrin (lane 6). Immunoprecipitations using anti-HSV from lysates of cells transfected only with T7-tagged ␤ 1 integrin served as a negative control (lane 7).
Expression of 16K Inhibits ␤1,6 Branching of ␤ 1 Integrin-We next assessed whether the repertoire of ␤ 1 integrin forms was dependent on the levels of 16K in the cell (Fig. 1, B-E). Increasing amounts of the vector encoding HSV-tagged 16K were co-transfected with a constant amount of T7-tagged ␤ 1 integrin. As more 16K was made (Fig. 1D), the largest form of the integrin disappeared (Fig. 1B). This corresponded with a loss of L-PHA-reactive ␤ 1 molecules (Fig. 1C). There was not a simultaneous build-up of ER-resident forms of the integrin, suggesting that the decrease was because of an alteration in ␤ 1 integrin processing rather than retention of the integrin in the ER. Western blot analysis of the same lysates for endogenous PDGF-␤R (Fig. 1E) showed that 16K does not affect overall expression levels of other glycosylated cell surface receptors.
Inhibition of ␤1,6 Branching by 16K Correlates with a Loss of Migratory Abilities of HEK293 Cells-In transwell invasion assays (Fig. 2), the expression of 16K abrogated migration through fibronectin and laminin (LN) matrices, demonstrating a correlation between the loss of ␤1,6 branching and the invasive abilities of cells. HEK293 cells transiently transfected with 16K were split, and aliquots were plated on plastic (Fig. 2, A  and D), fibronectin (B and E), or LN (C and F). Cells expressing 16K, identified by immunofluorescence using anti-HSV antibody, were abundant on plastic but nearly absent from the population that attached to fibronectin or LN. This finding suggests that the inhibition of migration results from an inhibition of attachment.
The Effect of 16K on ␤ 1 Integrin Processing Is Independent of V-ATPase Activity-One possible explanation for the effect of 16K on glycosylation is that GlcNAc-TV activity is highly sensitive to the pH of the Golgi, and that by overexpressing 16K, the activity of the V-ATPase is altered. If glycosylation was in fact sensitive to pH, it would be predicted that a major disruption of V-ATPase function would lead to altered processing. Dicyclohexylcarbodiimide (DCCD) is a potent inhibitor of V-ATPase (39), and it was used to treat cells that had been transfected with T7-tagged ␤ 1 integrin. After 2 h in DCCD, no changes in the forms of integrin or in the amount of L-PHAreactive molecules were seen (Fig. 3A). Acridine orange was used to confirm that the DCCD had dramatically inhibited acidification of cellular compartments (Fig. 3B). Pulse-chase labeling of transfected cells demonstrated that within 1 h newly synthesized ␤ 1 integrin polypeptides are processed into all three prominent molecular mass forms (Fig. 3C). Therefore, we concluded that the lack of any effect on glycosylation of a 2-h treatment with DCCD was strong evidence that 16K is not altering glycosylation by altering V-ATPase activity. The pulsechase experiments also showed that over an 18-h period the ␤1,6-branched form of ␤ 1 integrin is never present when16K is expressed.
The Inhibition of Glycosylation by 16K Is Not Dependent upon a Direct Association with ␤ 1 Integrin-Although 16K is primarily a V-ATPase component, its interaction with ␤ 1 integrin raised the possibility that it also functions as a shuttling molecule that can direct the integrin to bypass GlcNAc-TV. There are many ER/Golgi components implicated in the trafficking of molecules, such as ERGIC-53, Rab6, the ADP-ribosylation factor family of proteins, COP1, COP2, and SNARES (39 -47). ERGIC-53, for example, transports glycoproteins through compartments, presumably via direct association with glycan intermediates (48). Because glycosylation occurs sequentially through different compartments, we examined whether the inhibitory effect of 16K on glycosylation of ␤ 1 integrin was dependent on their association. A series of HSV epitope-tagged truncated 16K molecules (Fig. 4, upper panel) was made, and their ability to bind ␤ 1 integrin in co-immunoprecipitation experiments was compared with their ability to promote the loss of ␤1,6 branching.
The interaction of 16K with ␤ 1 integrin required the region of the protein spanning helices 2-4 with helix 4 contributing the most to the interaction (Fig. 4C, lanes 8 -10). This finding confirmed our earlier results using the yeast two-hybrid assay and direct protein interaction studies, showing the importance of helix 4 for the interaction (24). Although no individual helices formed stable interactions with ␤ 1 integrin, helices 2 and 4 suppressed ␤1,6 branching as effectively as full-length 16K antibody. Three forms of T7-tagged ␤ 1 (ϳ110, 125, and 135 kDa) were seen (arrows). The interaction of 16K with ␤ 1 integrin was confirmed by analysis of lysates made from cells co-transfected with T7-tagged ␤ 1 integrin and with HSV-tagged 16K. Complexes were retrieved by immunoprecipitating with anti-HSV antibody (lane 6) followed by agarose-conjugated protein A. Lysates of cells transfected with T7-tagged ␤ 1 integrin immunoprecipitated with anti-HSV served as a negative control (lane 7). B-E, HEK293 cells were co-transfected with a constant amount of ␤ 1 -T7 integrin and increasing amounts (in g) of HSV-tagged 16K. Proteins with ␤1,6-linked oligosaccharides were isolated using agarose-conjugated L-PHA and identified by probing with anti-T7 antibody (panel C). Aliquots of lysates probed with anti-T7 (panel B) showed that overall levels of ␤ 1 integrin were similar in all samples, and anti-HSV antibody verified that 16K was expressed incrementally (panel D). Increasing amounts of 16K resulted in the reduction of the largest L-PHA-reactive form of ␤ 1 integrin (see arrow in panel B, lanes 2-6, and panel C, lanes 2-6) but did not alter the expression of the PDGF-␤R (panel E).
( Fig. 4, A and B, lanes 3, 5, and 10), whereas helices 1 and 3 had only a small effect (lanes 2 and 4). Helix 2 was less active when present with helix 1 (lane 6), but otherwise all variants tested that had either helix 2 or 4 could suppress L-PHA reactivity (lanes 7-9). Helices 1 and 3 resulted in less of the ConAreactive smaller form of ␤ 1 integrin. These data indicate that specific regions of 16K can affect processing independently of direct association with ␤ 1 integrin. 16K is believed to have arisen from a duplication of an 8-kDa progenitor protein with two hydrophobic helices (49 -51), which may underlie the similarity in functions of alternate helices with respect to the glycosylation of ␤ 1 integrin.
Amino Acids within the Transmembrane Domain of ␤ 1 Integrin Are Required for Control of ␤1,6 Branching-These results, which show that membrane-spanning hydrophobic fragments of 16K could influence ␤ 1 integrin processing, led us to predict that the transmembrane domain of ␤ 1 integrin played a role in processing. We created mutations spanning the integrin transmembrane domain, expressed them as T7-tagged pro-teins, 2 and assessed their effect on glycosylation. Whereas most mutations had little effect, the conversion of Trp 728 , which is buried in the membrane near the cytoplasmic surface (Fig. 5A), to glycine caused processing to become strikingly similar to that seen when 16K was co-expressed with wild type ␤ 1 integrin. That is, there was a significant reduction of ␤1,6 branching of this mutant (lanes 1-6). At the same time, the core protein form of this mutant retained its ability to interact with 16K (lanes 7-10). The mutation of Trp 728 to glycine does not affect the ability of the integrin to get to the cell surface (Fig.  5B) nor its ability to bind to the integrin matrices fibronectin, laminin, or vitronectin (data not shown) and, therefore, is probably not altering the conformation of the integrin. These data further demonstrate that glycosylation can be orchestrated from within the membrane and in addition suggest that 16K might target an intermediary factor whose interaction with ␤ 1 integrin is abrogated by this tryptophan mutation.

FIG. 2. 16K inhibits cell invasion and adhesion to the integrin ligands laminin and fibronectin. HEK293 cells transfected with 16K
were examined for their ability to penetrate a Costar transwell apparatus coated with 10 g of laminin (LN) or fibronectin (FN). Control cells transfected with empty pXJ41 vector (mock transfected) were able to invade all matrices tested, and this ability was lost upon the expression of 16K. HEK293 cells transfected with HSV-tagged 16K were split and plated overnight on plastic (A and D), fibronectin (B and E), and laminin (C and F). A-C show phase-contrast views of the cells, whereas D-F show immunofluorescence with anti-HSV followed by Alexa 488-conjugated secondary antibody. Cells expressing 16K were able to bind to tissue culture plastic but lost the ability to attach to either laminin or fibronectin. Only the cells that were nontransfected or expressing low levels of 16K were able to attach. Scale bar ϭ 250 M.
FIG. 3. Inhibition of V-ATPase activity does not affect ␤1,6 branching of ␤ 1 integrin. A, treatment of HEK293 cells transfected with ␤ 1 integrin (lanes 2, 3, 5, and 6) with the V-ATPase inhibitor DCCD (lanes 3 and 6) did not inhibit ␤1,6 branching as detected using agarose-conjugated L-PHA followed by Western blot analysis with anti-T7 antibody (lanes 5 and 6). B, acridine orange staining showed that the concentration of DCCD used in panel A reduced the pH of organelles acidified by the V-ATPase. C, shown are pulse-chase experiments in which HEK293 cells were transfected with either ␤ 1 integrin alone (lanes 2-8) or co-transfected with ␤ 1 integrin and 16K (lanes 9 -15). Cells were pulsed with [ 35 S]methionine for 5 min, and complexes were retrieved from lysates using anti-T7 antibody and agarose-conjugated protein A at 2, 5, 10, and 20 min and 1, 6, and 18 h, respectively. The larger, L-PHA-reactive form of ␤ 1 integrin (arrow) was seen after 1 h (lane 6) in cells transfected with ␤ 1 integrin alone and was not detected when 16K was present. Scale bar ϭ 150 M.
Glycosylation of the EGF-R Is also Inhibited by 16K-The receptor for the EGF-R also has ␤1,6-branched oligosaccharides, and it too was T7 epitope-tagged and expressed in HEK293 cells along with T7-tagged ␤ 1 integrin and increasing amounts of HSV-tagged 16K (Fig. 6). ␤1,6-Branched proteins were isolated using L-PHA-conjugated agarose and analyzed by Western blots probed with anti-T7 antibody. As with ␤ 1 integrin, the L-PHA-reactive form of the EGF-R was suppressed by 16K (lanes 1-4). Co-immunoprecipitation experiments (data not shown) failed to detect any interaction between 16K and the EGF-R, again confirming that modulation of glycosylation by 16K occurs without direct interaction with the glycosylation substrate. DISCUSSION Integrins and other surface receptors undergo extensive post-translational modification throughout the exocytic pathway with glycosylation being a major determinant of receptor functions. Considering that the acidic environment of this pathway is regulated by V-ATPase, it is intriguing that one of FIG. 4. Inhibition of glycosylation does not depend on a direct association between ␤ 1 integrin and 16K. Upper panel, 16K derivatives comprised of specific hydrophobic helices were generated with carboxyl-terminal HSV tags. Lower panel, HEK293 cells were transfected with ␤ 1 -T7 only (lane 1) or co-transfected with ␤ 1 -T7 and the 16K derivative (lanes 2-10). Integrin expression was assessed by Western blot analysis of RIPA lysates probed with anti-T7 antibody (panel A), and glycosylation by Glc-NAc-TV was assessed by incubating extracts with agarose-conjugated L-PHA and analyzing recovered proteins with anti-T7 antibody (panel B). Proteins interacting with 16K or 16K mutants were isolated by immunoprecipitation using anti-HSV antibody followed by agaroseconjugated protein A, and recovery of ␤ 1 -T7 integrin was monitored by Western blot analysis using anti-T7 antibody (panel C).
the subunits of this enzyme, 16K, can play a role in processing. As we have shown, this involvement appears to be independent of V-ATPase activity, indicating that 16K in fact has multiple roles. There is little known about the cellular regulators of glycosylation, and our data show that for GlcNAc-TV, a specific subdomain of 16K, can have a major impact on ␤ 1 integrin processing. Furthermore, to add to the complexity of these data, 16K can form a stable complex with ␤ 1 integrin, but the effect of 16K on glycosylation is not dependent upon this interaction.
There are several events that might explain these data. On one hand, 16K may act as a trafficking molecule to assist in the movement of ␤ 1 integrin through the Golgi, and 16K overexpression could cause the integrin to be diverted prematurely to proteasomes or the cell surface without full processing occurring. On the other hand, 16K may somehow regulate Glc-NAc-TV or the mechanisms, which determine how the transferase finds its targets.
For various reasons, the first of these explanations is not satisfactory. The fact that an ␣-helical subdomain of 16K, which has no ability to bind ␤ 1 integrin, can regulate processing and that the EGF-R is also affected even though it does not bind to 16K makes a cargo receptor role for 16K seem unlikely. ER to Golgi cargo receptors, such as ERGIC-53, are also structurally unrelated to 16K, functioning as mannose-specific lectins (40). In addition, the overall levels of ␤ 1 integrin or the EGF-R are not significantly reduced by the addition of 16K, indicating that proteolysis is not being enhanced.
Nevertheless, it should be noted that the V 0 subunit of the yeast V-ATPase, which is comprised of a hexamer of 16K, Vma6, Vph1, Vma11, and Vma16, has recently been found to catalyze membrane fusion of vacuoles in a calmodulin/calciumdependent and V-ATPase-independent manner (53). The authors propose that V 0 subunits from opposing membranes form a pore spanning both membranes, providing a mechanism for membrane fusion. Thus, the possible involvement of 16K in the trafficking of molecules within compartments should not be excluded.
The possibilities that 16K regulates GlcNAc-TV or that it plays a role in the identification of glycosylation targets of the transferase must be considered in light of the fact that 16K is a highly hydrophobic membrane protein. GlcNAc-TV has a single membrane-spanning hydrophobic domain compared with the four within a 16K monomer or the twenty-four within the 16K hexamer that constitutes the proton channel of V-ATPase. Although there is as yet no supporting evidence, 16K could act as a docking site in the membrane to cluster molecules with complementing functions. The transmembrane 4 superfamily family of tetraspan proteins is believed to carry out such a role in plasma membranes (54,55). It is relevant in this regard that the transmembrane domain of ␤ 1 integrin was found in our study to play a key role in the ability of that molecule to become ␤1,6-branched. Notably, the mutation of Trp 728 resulted in levels of ␤1,6 branching that were identical to those seen when 16K was overexpressed. The most direct interpretation is that Trp 728 is required for a direct interaction between GlcNAc-TV and ␤ 1 integrin or for the interaction with another molecule(s) that is needed for GlcNAc-TV to have access to ␤ 1 integrin. Do et al. (56) showed that ␤1,6 branching is in fact dependent upon GlcNAc-TV having access to suitable oligosaccharide acceptors. Increasing amounts of 16K could compete for GlcNAc-TV or for this other molecule, thereby altering glycosylation of ␤ 1 integrin. It is predicted that other ␤1,6-branched receptors such as the EGF-R also participate in similar complexes with which 16K can compete. Clearly, it will now be necessary to search for other putative transmembrane partners of ␤ 1 integrin, the EGF-R, GlcNAc-TV, and 16K. We recently identified in a yeast two-hybrid screen two additional uncharacterized binding partners for ␤ 1 integrin (24).
The reduction in ␤1,6 branching in response to increasing amounts of 16K may also be the result of the inhibition of GlcNAc-TV expression. The GlcNAc-TV gene is itself sensitive to signaling pathways with ets-dependent activation by src kinase and by her-2/neu stimulation of the Ras-Raf signaling  1, 4, 7, and 9) ␤ 1 integrin. Four of the wild type T7-␤ 1 samples (lanes 2, 5, 8, and 10) and two of the W728G mutant samples (lanes 7 and 9) had HSV-16K included in the transfection. Western blots of RIPA lysates probed with anti-T7 as shown in lanes 1-3 and lanes 9 and 10 confirm that the mutant and wild type are expressed at similar levels. ␤1,6-Branched proteins, recovered using L-PHA-conjugated agarose, are reduced similarly by co-transfection with 16K or by the mutant (lanes 4 -6). Immunoprecipitations with anti-HSV antibody from co-transfected cells (lanes 7 and 8) show that the W728G mutation does not interfere with binding of the core integrin protein to16K. B, flow cytometry experiments revealed that the integrin mutant W728G is expressed on the surface of the cell at levels comparable with those of the wild type-tagged ␤ 1 integrin.
FIG. 6. 16K inhibits ␤1,6 branching of the EGF-R. HEK293 cells were co-transfected with T7-tagged ␤ 1 integrin and the EGF-R. Lanes 1-4 contain equal amounts of T7-tagged EGF-R and ␤ 1 integrin with increasing amounts of HSV-tagged 16K. Lanes 5 and 6 show lysates transfected with only ␤ 1 integrin or EGF-R, respectively. A, Western blot analysis of RIPA lysates treated with anti-T7 antibody. B, increasing amounts of 16K suppress ␤1,6 branching of both ␤ 1 integrin (lower band) and the EGF-R (upper band), as detected using L-PHA-conjugated agarose and Western blot analysis with anti-T7 antibody. pathway playing a role (57,58). In addition, the processing of ␤ 1 integrin can be accelerated by TGF-␤ 1 through the stimulation of Ras (59), suggesting that any effects of 16K on signaling pathways using Ras could affect glycosylation. It is also a reasonable expectation that the involvement of 16K in cell surface receptor glycosylation will in turn perturb gene expression pathways by altering receptor function. In support of this finding, it has been suggested that a role of N-linked glycosylation is to prevent surface receptors from dimerizing in the absence of an appropriate ligand (60). Additionally, the inhibition of ␤1,6 branching in GlcNAc-TV knockout mice enhances the clustering of T cells and lowers T cell activation (61).
There are very few reports in the literature of cellular regulators of GlcNAc transferases. Guo et al. (52) reported a modest (Ͻ50%) decrease in GlcNAc-TV activity in response to the metastasis-suppressor gene nm23-H1 and the tumor suppressor gene p16, but they did not examine specific receptors such as ␤ 1 integrin or the EGF-R. In our experiments, we observed a Ͼ90% loss of ␤1,6 branching of these targets and a concurrent dramatic reduction in migration through extracellular matrix components. 16K may provide unique opportunities for the intervention in ␤1,6 branching and reduction of invasive cell growth.