The Tetraspan Protein Epithelial Membrane Protein-2 Interacts with β1 Integrins and Regulates Adhesion*

The growth arrest-specific-3 (GAS3)/PMP22 proteins are members of the four-transmembrane (tetraspan) superfamily. Although the function of these proteins is poorly understood, GAS3/PMP22 proteins have been implicated in the control of growth and progression of certain cancers. Epithelial membrane protein-2 (EMP2), a GAS3/PMP22 family member, was recently identified as a putative tumor suppressor gene. Here, we addressed the normal function of EMP2 by testing the prediction that it influences integrin-related cell functions. We observed that EMP2 associates with the β1 integrin subunit. Co-immunoprecipitation and immunodepletion experiments indicated that ∼60% of β1integrins and EMP2 can be isolated in common protein complexes. Whereas this association between EMP2 and β1 integrin may be direct or indirect, it has features of integrin heterodimer selectivity. Thus, by laser confocal microscopy, EMP2 colocalized with α6β1 but not α5β1 integrin. Increased expression of EMP2 also influenced the integrin heterodimer repertoire present on the plasma membrane. EMP2 specifically increased the surface expression of the α6β1 integrin while decreasing that of the α5β1 protein. Reciprocally, reduction in EMP2 expression using a specific ribozyme decreased surface expression of α6β1 integrin. Accordingly, these EMP2-mediated changes resulted in a dramatic alteration in cellular adhesion to extracellular matrix proteins. This study demonstrates for the first time the interaction of a GAS3/PMP22 family member with an integrin protein and suggests that such interactions and their functional consequences are a physiologic role of GAS3/PMP22 proteins.

understood tetraspan proteins are connexins, which form the major structural elements of gap junctions. On the other hand, tetraspanins have recently gained prominence for their role as "molecular facilitators" in the assembly of mixed protein signaling complexes, including those involving integrins (1,2). Functionally, tetraspanins have been shown to influence integrin-mediated events such as cell proliferation, migration, and tumor cell invasion in a variety of cell types (3)(4)(5)(6)(7)(8).
Currently, there are at least six known members of the GAS3/PMP22 family: PMP22, EMP1 (or TMP), EMP2 (or XMP), EMP3 (or YMP), PERP, and brain cell membrane protein 1 (9 -12). These genes share ϳ30 -40% amino acid identity (11,13). All GAS3/PMP22 members are predicted to contain two extracellular loops and a small cytoplasmic tail, but beyond this structural information, little is known about the endogenous function(s) associated with most of the family members. What is known is that many of these proteins appear to be involved with cell proliferation, cell-cell, and cell-matrix interactions and/or myelin formation. Dysregulation of certain family members has been linked with disease (11,14). For example, PMP22 has gained prominence due to the fact that genetic alterations in this gene (e.g. mutations, deletions, or duplications) lead to peripheral neuropathies such as Charcot-Marie-Tooth type 1A and Dejerrine Sottas syndrome (14 -17).
Epithelial membrane protein-2 (EMP2) was first identified based on its homology to PMP22 (11). Our laboratory initially encountered EMP2 as part of a search for genes involved in the malignant progression of B cell lymphomas using suppression subtractive hybridization (13). Disruption of EMP2 yielded dramatic phenotypes. Specifically, in a model of B cell lymphoma progression, EMP2 appears to act as a tumor suppressor (13). Moreover, ectopic overexpression of in cultured cells promoted stress-induced apoptosis (13).
In this study, we begin to address the biochemical function of GAS3/PMP22 family members. We test the idea that they may share with tetraspanins the capacity to interact with integrins (1,18,19). Integrins are ␣␤ heterodimeric receptors that bind and organize cellular responses to ECM proteins and cellular receptors (21,22). There is exceptional combinatorial diversity in this receptor system, with 16 ␣ and 9 ␤ subunits and alternate splicing of individual subunits (11,23). Individual integrin ␣␤ heterodimers display distinct specificities for various ECM proteins and cell surface receptors, so changes in the expression of these various integrins have profound effects on the repertoire of cell proliferation, survival, adhesion, and migration (12,14,24).
The present study tests whether EMP2 phenotypic effects are mediated through its interaction with integrins. We show that EMP2 associates with the ␤ 1 integrin subunit and appears to have a selective effect on the surface expression of the ␣ 6 ␤ 1 integrin heterodimer. This change in integrin expression mediated by EMP2 confers unique binding properties onto the cell, namely adhesion to the ECM component laminin. These obser-vations reflect an important role for EMP2 in the regulation of an integrin-mediated cellular phenotype.
Incorporated into these oligonucleotides was the 22 bases of the hammerhead ribozyme conserved catalytic core (underlined) and two 10 nucleotide EMP2-specific recognition domains (GenBank TM accession number AF346627; nucleotides 204 -224 of the murine EMP2). To construct the ribozyme, the two oligonucleotides were phosphorylated, hybridized to one other, and ligated into the unique HpaI site located in the 3Ј-untranslated region of the green fluorescent protein gene in plasmid pEGFP-N3 (Clontech, Palo Alto, CA).
Northern Analysis-In vivo ribozyme cleavage of EMP2 transcripts was verified by Northern analysis. Total RNA was isolated from NIH3T3 or 3T3/RIBO cells using an RNA purification kit (Qiagen, Valencia, CA). RNA (5 g) was subjected to agarose electrophoresis, transferred to a nylon membrane (Amersham Biosciences) by capillary action, and cross-linked by UV irradiation (Stratalinker; Stratagene, San Diego, CA). A full-length EMP2 cDNA fragment (516 bp) was labeled using random primer synthesis (Amersham Biosciences) with [ 32 P]dCTP as previously described (13). Membranes were prehybridized with Rapid-Hyb buffer (Amersham Biosciences) for 1 h and hybridized with labeled probe for 2 h at 65°C. Blots were washed with a high stringency buffer (60°C, 0.1ϫ SSC, and 0.5% SDS) and exposed to x-ray film.
EMP2 contains multiple glycosylation sites (13). In order to detect EMP2 in coimmunoprecipitation experiments, N-linked glycans were cleaved using peptide:N-glycanase (New England Biolabs, Beverly, MA). Eluates were treated as per the manufacturer's instructions at 37°C for 2 h. Eluted proteins were separated by SDS-PAGE and transferred to a nitrocellulose membrane (Amersham Biosciences). Membranes were stained with Ponceau S (Sigma) to determine transfer efficiency. Membranes were blocked with 10% low fat milk in PBS containing 0.1% Tween 20 and probed with an anti-␤ 1 integrin monoclonal antibody or EMP2 antisera. Protein bands were visualized using a horseradish peroxidase-labeled secondary antibody (BD Biosciences; Southern Biotechnology Associates, Birmingham, AL) by chemiluminescence (ECL; Amersham Biosciences). Experiments were repeated at least four times.
Immunodepletion-NIH3T3 (1 ϫ 10 7 ), 3T3/V (1 ϫ 10 7 ), or 3T3/EMP2 (1 ϫ 10 7 ) cells were lysed and solubilized as described above. Each lysate was divided into three equal portions to be immunodepleted overnight using (i) protein-A/G agarose alone, (ii) protein-A/G agarose plus EMP2 antisera, or (iii) protein A/G-agarose plus an anti-␤ 1 integrin antibody (clone TS2/16). Samples were centrifuged, and the unbound supernatant was collected. This supernatant was diluted 1:1 in 2ϫ Laemmli buffer and boiled for 5 min. Samples were treated with peptide:N-glycanase as described above for EMP2 detection. Supernatants were then subjected to SDS-PAGE followed by Western analysis. Protein expression was quantitated using the Personal Densitometer SI and ImageQuanNT software (Amersham Biosciences). Stoichiometry was calculated by dividing the volume intensities before and after immunodepletion as described by Stipp et al. (28). Experiments were repeated three times.
Adhesion Assays-A standard static adhesion assay (15-20 min) was performed as previously described (29). Briefly, 96-well plates were precoated overnight with the ECM substrates laminin, fibronectin, poly-D-lysine (Roche Molecular Biochemicals; 5-10 g/ml), or 1% fatty acid-free bovine serum albumin (Sigma). For collagen I or collagen IV (Becton Dickinson Labware, Bedford, MA), plates were coated 2 h as per the manufacturer's instructions. Cells (7 ϫ 10 4 ) were plated onto the ECM in serum-free conditions and incubated at 37°C. Unbound cells were washed away. Bound cells were stained with toluidine blue and then lysed using 2% SDS (Biowhittaker, Walkersville, MD). The resultant soluble toluidine blue was quantitated by measuring the absorbance at 595 nm. Binding to each ECM was performed in triplicate. Each experiment was repeated at least three times. An unpaired Student's t test was used to confirm significance between 3T3/EMP2 and 3T3/V as well as between 3T3/V and 3T3/RIBO.
In antibody blocking experiments, cells were preincubated with various dilutions of anti-␣ 5 or anti-␣ 6 integrin monoclonal antibody for 60 min at 4°C. Cells (7 ϫ 10 4 ) were aliquoted into a 96-well plate precoated with laminin or poly-D-lysine and allowed to adhere for 30 min. Unbound cells were washed away, and bound cells were quantitated as described above. Each experiment was repeated five times.
Confocal Microscopy-3T3/V, 3T3/RIBO, or 3T3/EMP2 cells were plated overnight onto glass coverslips (Fisher). Cells were fixed and permeabilized in cold methanol for 30 min at Ϫ20°C and then blocked with 1% normal goat serum for 45 min. Cells were incubated 2 h at room temperature in a humidified chamber with the primary antibody and then washed 3-4 times with PBS plus 0.01% Triton X-100 (PBST). Cells were incubated overnight with a fluorescein isothiocyanate (FITC)conjugated donkey anti-rat IgG and Texas Red-conjugated donkey antirabbit IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) at 4°C in a humidified chamber. Cells were washed with PBST, rinsed briefly with double-distilled H 2 O, and mounted onto microscope slides using a 3.5% n-propyl gallate-glycerol solution (Sigma).
Laser-scanning confocal microscopy was used to assess the distribution and colocalization of proteins. Samples were analyzed using a Fluoview laser-scanning confocal microscope (Olympus America Inc., Melville, NY). To simultaneously detect FITC-labeled and Texas Redlabeled cells, samples were excited with argon and krypton lasers at 488 and 568 nm. Light emitted between 525 and 540 nm or above 630 nm was recorded for FITC or Texas Red, respectively. 20 -30 horizontal (x, y) sections were obtained at 0.5-m intervals. Colocalization experiments were studied in single x-y optical sections and merged using the Fluoview image analysis software (version 2.1.39). In all experiments, cells were observed using a 60ϫ oil immersion objective. Each experiment was repeated at least four times.
Flow Cytometry-The membrane expression of ␣ 5 , ␣ 6 , and ␤ 1 integrin subunits was assessed by flow cytometry. Cells were fixed in 2% paraformaldehyde (w/v) in PBS for 20 min on ice. Cells were incubated with primary antibody (1:200) for 30 min on ice in PBS plus 2% fetal calf serum, washed two times, and then incubated with red-phycoerythrinconjugated anti-rat Ig, light chain antibody (0.25 g/10 6 cells; BD Biosciences) for 30 min on ice. Negative control cells were incubated with the secondary antibody alone. After two consecutive washes, cells were resuspended in PBS and analyzed with a FACScan flow cytometer (BD Biosciences). Integrin expression levels were calculated as mean fluorescent intensity (MFI). Experiments were repeated three times.

Production of Cell Lines with Varying Levels of EMP2-In
order to elucidate the cellular function of EMP2, we first created a number of cell lines engineered to express different levels of the gene. We capitalized on the fact that wild type NIH3T3 cells express moderate levels of EMP2, so we could create variants that overexpressed (3) or underexpressed (described below) this gene compared with the wild type cells. The latter was accomplished through the stable introduction of an EMP2-specific hammerhead ribozyme vector (30), which downregulates EMP2 expression by specific cleavage of its transcript. Northern blot analysis was used to directly validate that the EMP2-specific ribozyme cleaved the EMP2 message (Fig.  1A). Wild type NIH3T3 cells yielded a single EMP2 transcript at 4.8 kb (Fig. 1A). However, NIH3T3 cells stably transfected with an EMP2 ribozyme (called 3T3/RIBO cells) retained only a minimal signal for the normal transcript and displayed an additional lower band representing the cleaved EMP2 transcript (arrow). The residual intact EMP2 transcript probably reflects variable ribozyme expression in this nonclonal transfectant population.
We used Western analysis to characterize the protein levels of EMP2 in empty vector control NIH3T3 cells (3T3/V), cells expressing a FLAG-tagged recombinant EMP2 (3T3/EMP2), and cells expressing the EMP2-specific ribozyme (3T3/RIBO) (Fig. 1B, upper panel). Compared with 3T3/V cells, EMP2 protein levels were higher in 3T3/EMP2 cells and undetectable in 3T3/RIBO. Western analysis was also performed to measure the levels of an independent tetraspan protein (the tetraspanin, CD9). Protein levels of CD9 were unaffected among the three recombinant 3T3 cell lines (Fig. 1B, lower panel). These observations indicated that our recombinant 3T3 cell lines were modified as intended for EMP2 expression and suggested that the modification was specific for EMP2.
EMP2 Associates with the ␤ 1 Integrin Subunit-Using these engineered NIH3T3 variants, we examined the normal cellular function of EMP2. Initially, we chose a candidate approach drawing on the observation that numerous members of the tetraspanins associated with integrin dimers containing the ␤ 1 subunit (12,13,22). Specifically, we sought to establish whetherEMP2associatedwithintegrinsand/ormodifiedintegrindependent function. In order to determine whether EMP2 and the ␤ 1 integrin subunit could associate, a coimmunoprecipitation approach was employed. Cellular lysates were produced from NIH3T3 (or 3T3/V) and 3T3/EMP2 cells, incubated with an anti-␤ 1 integrin antibody to immunoprecipitate this molecule, and subjected to Western analysis using EMP2 antisera. As shown in Fig. 2A, Western analysis demonstrated that EMP2 immunoprecipitated with the ␤ 1 integrin subunit. The reciprocal experiment showed similar results; lysates treated with an anti-EMP2 antibody immunoprecipitated EMP2 plus ␤ 1 integrin protein (Fig. 2B). Positive controls, in which EMP2 and ␤ 1 integrin immunoprecipitates were probed for the cognate protein (Fig. 2, C and D), suggested that a substantial amount of these proteins were in a common coimmunoprecipitable complex.
To estimate the fraction of the ␤ 1 integrin subunit that associates with EMP2, immunodepletion experiments were performed according to the procedure of Stipp et al. (28) (see "Materials and Methods"). Lysates of 3T3/V or 3T3/EMP2 cell lysates were incubated overnight with anti-␤ 1 integrin antibody or EMP2 antisera, and the antibody with bound proteins was then immunodepleted using protein A/G beads. The amounts of EMP2 and ␤ 1 integrin in the lysates before and after immunodepletion were quantitated by Western analysis and densitometry, and these values were used to calculate the percentage of protein in the coimmunoprecipitation (Fig. 2, E and F). As expected, anti-EMP2 and anti-␤ 1 integrin each depleted Ͼ90% of their cognate antigen (EMP2 and ␤ 1 integrin, respectively). Anti-EMP2 immunoprecipitation coimmunoprecipitated 56% of the ␤ 1 integrin subunit from 3T3/V cells and 70% from 3T3/EMP2 cells. Anti-␤ 1 integrin antibody coimmunoprecipitated ϳ65% of EMP2 in both cell types. Taken together, these data indicate that ϳ60% of ␤ 1 integrin and EMP2 pools (the averages of the depletion numbers above) are present in a common co-immunoprecipitable complex after Nonidet P-40 solubilization. This association between EMP2 and ␤ 1 integrin may be direct or indirect. However, it should be noted that the interaction between EMP2 and ␤ 1 integrin was preserved in the presence of 1% Nonidet P-40, which unlike certain other nonionic detergents (CHAPS, Brij-97) disrupts most tetraspan-tetraspan interactions observed after such detergent solubilization (7,8,32). It thus appears that the association was not due to this type of nonspecific hydrophobic interactions.
We further examined the association and colocalization of EMP2 and the ␤ 1 integrin subunit using laser-scanning confocal microscopy. NIH3T3, 3T3/V, or 3T3/EMP2 cells were stained with anti-␤ 1 integrin and anti-EMP2, and serial laser confocal images were captured and analyzed using Fluoview image analysis software (Fig. 3). In NIH3T3 (not shown) or 3T3/V cells (Fig. 3), EMP2 expression is predominantly perinuclear, with an additional component of dispersed cytoplasm granules and low levels of plasma membrane expression. The perinuclear staining is consistent with endoplasmic reticulum and Golgi apparatus localization. 2 Overexpression of EMP2 in 3T3/EMP2 cells resulted in increased EMP2 on the plasma membrane. When 3T3/V cells were double-labeled with anti-EMP2 and anti-␤ 1 integrin, there was a significant colocalization of staining both in the perinuclear region and cytoplasmic 2 M. Wadehra, L. Goodglick, and J. Braun, unpublished data. vesicles. In 3T3/EMP2 cells, ␤ 1 integrin colocalized with EMP2 on the plasma membrane. These results were specific for ␤ 1 integrin, since no colocalization was observed between EMP2 and another tetraspan protein, CD9, in either 3T3/V or 3T3/ EMP2 cells (data not shown).
EMP2 Expression Modulates Binding of Cells to the ECM-We next examined whether the apparent association of EMP2 with the ␤ 1 integrin subunit had functional consequences. Specifically, we tested whether modulation of EMP2 expression affected the binding of cells to the ECM. To evaluate this, 3T3/V, 3T3/EMP2, and 3T3/RIBO cells were incubated for 15-20 min in plates coated with various ECM proteins (laminin, fibronectin, collagen I, collagen IV). Poly-D-lysine and 1% fatty acid-free bovine serum albumin were used as positive and negative substrates, respectively. Unbound cells were rinsed away, and the attached cells were stained with toluidine blue and quantitated as described under "Materials and Methods." Interestingly, the cells that overexpressed EMP2 (3T3/EMP2 cells) exhibited an approximate 2-fold increase in laminin binding compared with 3T3/V cells (p Ͻ 0.05; Fig. 4A), and a 4-fold increase in laminin binding compared with 3T3/RIBO (p Ͻ 0.005; Fig. 4A).
In contrast to the results with laminin, EMP2 expression levels inversely correlated with fibronectin binding. Specifically, the cells with lower levels of EMP2 (3T3/RIBO cells) demonstrated significantly greater binding to fibronectin compared with either 3T3/V or 3T3/EMP2 cells (Fig. 4B). Within the 15-20-min incubation period, none of the cell variants displayed significant binding to collagen I or IV even in the presence of Mn 2ϩ , which is known to enhance integrin-mediated adhesion (34) (data not shown).
Laminin binding is primarily, although not exclusively, mediated by the ␣ 6 ␤ 1 integrin (35). To verify that laminin binding in 3T3/V and 3T3/EMP2 cells was mediated by ␣ 6 integrins, we tested whether an anti-␣ 6 integrin would successfully block this interaction. Cells were incubated with anti-␣ 6 integrin or with anti-␣ 5 integrin as a control and then allowed to adhere to a laminin-coated plate for 30 min. As anticipated, anti-␣ 6 integrin but not anti-␣ 5 integrin antibodies specifically reduced binding of 3T3/V and 3T3/EMP2 cells to laminin-coated plates (Fig. 4, C and D). cific protein using protein A/G-agarose beads plus antibodies specific for EMP2 or ␤ 1 integrin (clone TS2/16). Protein A/G-agarose beads alone without conjugated antibodies served as a negative control. The lysates were fractionated by SDS-PAGE and examined by Western analysis using anti-␤ 1 integrin (clone 18) or anti-EMP2 antisera. E and F show levels of EMP2 and ␤ 1 integrin after immunodepletion, respectively.

FIG. 2. Stoichiometric analysis of the binding of EMP2 and ␤ 1 integrin subunit.
A-D, cell lysates were prepared from 5 ϫ 10 6 3T3/V cells or 3T3/EMP2 cells in 1% Nonidet P-40 and immunoprecipitated using rabbit anti-EMP2 antisera, anti-␤1 integrin antibody (clone TS2/ 16), or the appropriate isotype control IgG antibodies. Precipitates were fractionated by SDS-PAGE and subjected to Western analysis using rabbit anti-EMP2 or anti-␤ 1 integrin (clone 18) as indicated in the panels. The ␤ 1 integrin (A) and EMP2 (B) immunoprecipitates were probed for coimmunoprecipitation with EMP2 antisera and anti-␤ 1 integrin, respectively. As positive controls, EMP2 (C) and ␤ 1 integrin (D) immunoprecipitates were probed for the same cognate proteins. Lanes 1 and 2 are lysates of 3T3/V cells; lane 3 is a lysate of 3T3/EMP2 cells. Heterogeneity in ␤ 1 integrin sizes relates to differences in glycosylation states (20). E and F, cell lysates from 1 ϫ 10 7 3T3/V or 3T3/EMP2 cells were divided into equal portions and depleted of spe-FIG. 3. EMP2 colocalizes with the ␤ 1 integrin subunit. 3T3/V and 3T3/EMP2 cells were fixed in methanol and stained with rabbit anti-EMP2 (FITC secondary) and anti-␤ 1 integrin (clone 9EG7; Texas Red secondary) as described under "Materials and Methods." Colocalization between EMP2 and the ␤ 1 integrin subunit appears as yellow and is visualized in the far right panels. Images were captured on a Fluoview laser-scanning confocal microscope and merged using the Fluoview image analysis software (version 2.1.39). Cells are magnified ϫ600.
EMP2 Colocalizes Specifically with the ␣ 6 ␤ 1 Integrin Heterodimer-We demonstrated above that EMP2 could coprecipitate with the ␤ 1 integrin subunit and that EMP2 levels could modulate binding to laminin presumably through altering the expression and/or activity of the ␣ 6 integrin subunit. Here we assessed whether EMP2 colocalized with the ␣ 6 integrin subunit and whether altered protein levels of EMP2 would modulate the cellular distribution of the ␣ 6 integrin protein. 3T3/V or 3T3/EMP2 cells were labeled with anti-EMP2 and anti-␣ 6 integrin antibodies, and analyzed for cellular distribution and colocalization using confocal microscopy. As shown in Fig. 5A, EMP2 colocalized with the ␣ 6 integrin subunit in both 3T3/V and 3T3/EMP2 cells. Similar to the ␤ 1 integrin chain, the distribution of the ␣ 6 integrin subunit appeared more abundant on the cell surface with increased expression of EMP2 in 3T3/EMP2 cells (Fig. 5A). In contrast to the ␣ 6 integrin subunit, EMP2 did not colocalize with the ␣ 5 integrin chain (Fig. 5B). In 3T3/V cells, the ␣ 5 integrin subunit was present on the plasma membrane; interestingly, overexpression of EMP2 (in 3T3/EMP2 cells) did not appear to increase surface expression of the ␣ 5 integrin chain but rather increased ␣ 5 integrin staining in the endoplasmic reticulum-Golgi and cytoplasmic vesicles (Fig. 5B).
EMP2 Alters the Surface-expressed Repertoire of Integrins-Flow cytometry was used to further assess the surface expression of the ␣ 6 , ␣ 5 , and ␤ 1 integrin subunits in 3T3/V and 3T3/EMP2 cells. Cells were fixed; stained with antibodies detecting ␣ 6 , ␣ 5 , or ␤ 1 integrins; and quantitated by MFI for levels of surface expression using a FACScan flow cytometer. As shown in Fig. 6, A-C, the level of the ␣ 6 integrin subunit on the plasma membrane directly correlated with the level of EMP2. 3T3/EMP2 cells expressed Ն50% more ␣ 6 integrin chain on their surface (MFI ϭ 64.3) compared with 3T3/V cells (MFI ϭ 43.2) or 3T3/RIBO cells (MFI ϭ 39.7) (Fig. 6, A-C). In contrast, the expression of ␣ 5 integrin protein on the cell surface was inversely correlated with the levels of EMP2, with 3T3/RIBO cells expressing Ն2-fold more ␣ 5 integrin chain on their surface (MFI ϭ 180.7) compared with 3T3/V cells (MFI ϭ 87.7) and 3T3/EMP2 cells (MFI ϭ 45.2) (Fig. 6). Whereas EMP2 levels directly or indirectly modulated the plasma membrane expression of ␣ 5 and ␣ 6 integrin subunits, the surface levels of the ␤ 1 integrin subunit remained constant in 3T3/V, 3T3/EMP2, and 3T3/RIBO cells (Fig. 6, G-I). Moreover, the level of EMP2 expression had no effect on the surface expression of another membrane protein, CD9 (data not shown). These findings suggest that EMP2 expression can directly or indirectly regulate the surface repertoire of ␣ 6 versus ␣ 5 integrin subunits. DISCUSSION EMP2 belongs to the enigmatic GAS3/PMP22 family of tetraspan proteins, of which to date little is known regarding function or protein-protein interactions. Here we show for the first time a heterologous binding partner for a GAS3/PMP22 family member. We demonstrated by coimmunoprecipitation and immunodepletion that the majority of EMP2 and the ␤ 1 integrin subunit exist in apparently common protein complexes. By laser confocal microscopy, EMP2 colocalized with FIG. 4. EMP2 expression selectively augments laminin binding by an ␣ 6 -dependent process. Cells (3T3/V, 3T3/EMP2, and 3T3/RIBO) were incubated in serum-free media at 37°C for 20 min in wells coated with laminin or fibronectin (A and B, respectively); bovine serum albumin (BSA) was used as a negative control protein. Wells were washed to remove unbound cells. Adherent cells were stained with toluidine blue and lysed with 2% SDS. Absorbance was measured at 595 nm. Poly-D-lysine binding was used to normalize for plating differences between the cell types. Values are represented as the mean Ϯ S.D. for triplicate wells (*, p ϭ 0.004; **, p ϭ 0.005; Student's unpaired t test). To determine which integrin isoforms accounted for this change in adhesion, 3T3/EMP2 or 3T3/V cells were incubated for 1 h with antibodies to ␣ 6 or ␣ 5 integrin (C and D, respectively). Cells were added to wells containing laminin or poly-D-lysine. Cells were assayed for binding as described above. A representative experiment is shown; five independent experiments yielded similar results. ␣ 6 ␤ 1 but not ␣ 5 ␤ 1 integrin, suggesting that EMP2 associates with a discrete subset of integrins. Moreover, modulating EMP2 expression levels reciprocally changed the surface expression of ␣ 6 ␤ 1 and ␣ 5 ␤ 1 integrin, with a corresponding alteration in cell adhesion to ECM proteins. These observations suggest that EMP2 selectively associates with certain integrin isoforms and that EMP2 levels in effect regulate integrin surface expression and function. The apparent association of EMP2 and ␤ 1 integrins raises several issues. Our data indicate that the majority of both the EMP2 and ␤ 1 integrin pools (ϳ60% each) can be isolated in reciprocally immunoprecipitable complexes. The association of EMP2 and ␤ 1 integrin probably is not a simple isolation artifact for two reasons. The interaction between EMP2 and ␤ 1 integrin was observed even in the presence of 1% Nonidet P-40, a detergent that overcomes some types of nonspecific hydrophobic interactions observed with other isolation conditions for tetraspan proteins (7,8,32,36). In addition, we did not observe an association between EMP2 and CD9, a distinct tetraspanin protein, using either biochemical or laser confocal assays.
Our experimental conditions do not enumerate the absolute molecular abundance of proteins in these pools, so we cannot ascertain the molecular stoichiometry of EMP2 and ␤ 1 integrin in these complexes. Also, it is not clear whether the coimmunoprecipitation of EMP2 and the ␤ 1 integrin chain reflects a direct association of these proteins or an indirect association involving additional components of a heterogeneous noncovalent complex. For example, various tetraspanins appear to form homologous and heterologous multimers as well as heterogeneous complexes containing integrins and other protein species (4, 6 -8). Thus, it is likely that the molecular complexes bearing EMP2 and ␤ 1 integrins include other tetraspan molecules and additional classes of membrane proteins. Definition of this heterogeneity and the process of complex formation is needed to understand the structural basis of the tetraspan-integrin association.
Mechanistically, an elegant biochemical study recently has provided direct evidence for binding of CD151 and ␣ 3 ␤ 1 , involving the second extracellular loop of CD151 and amino acids 570 -705 of the ␣ 3 subunit (32). ␤ 1 integrin association has been reported to be a feature distinguishing CD81 from CD9 (40), and segment-swapping experiments are beginning to define tetraspan regions involved in the interaction (40,43). These recombinant chimera and cross-linking approaches should be useful in resolving whether EMP2 directly interacts with ␣ 6 ␤ 1 integrins and whether it shares homologous regions with these proteins involved in ␣ 6 and/or ␤ 1 interaction and selectivity.
The present study demonstrates that EMP2 levels modulate the selectivity of cell adhesion to ECM proteins, and this appeared to result from EMP2-induced changes in ␣ 5 ␤ 1 and ␣ 6 ␤ 1 integrin surface expression. However, it should be noted that other integrin heterodimers also can contribute to laminin and fibronectin binding (29,37,38). For example, the ␣ 2 and ␣ 3 subunits are part of heterodimer pairs that are known to also bind fibronectin as well as collagen I and IV (38). In addition, ␣ 6 ␤ 4 integrin is known to contribute to laminin binding (35,37). Nevertheless, modulation by EMP2 of these integrin subunits seems unlikely. First, little binding to collagen I or IV was observed, suggesting that bioactive ␣ 2 and ␣ 3 integrin subunit expression is minimal in our experimental system. Second, ␤ 4 integrin protein could not be detected in these cells by Western analysis (data not shown). We are currently exploring whether EMP2 interacts with and/or influences the expression of other integrin subunits.
Tetraspanins are prominent in their association with integrins and modulation of integrin-mediated functions like cell proliferation, survival, migration, and tumor cell invasion (3)(4)(5)(6)(7)(8). The present study indicates that EMP2 also shares such properties but is distinct in its subcellular localization, integrin isoform selectivity, and functional consequences. Thus, it appears that tetraspanins and EMP2 (perhaps representative of other GAS3/PMP22 proteins) share the structural features permitting integrin association but diverge in integrin isoform specificity, subcellular localization, and/or trafficking properties. We propose that the tetraspanins and GAS3/PMP22 proteins serve complementary and/or competing roles in delivery and/or stabilization of integrin isoforms to the cell surface. Differential expression of the various tetraspanin and GAS3/ PMP22 proteins may thereby play an important role in shaping a cell's repertoire of integrin-mediated functions.
The mechanism by which EMP2 regulates the surface expression of ␣ 6 ␤ 1 integrin remains to be defined. Integrins are typically expressed in a large intracellular pool, which offers the potential for rapid responsiveness through pool mobilization. It is unknown how this pool is mobilized in a subunitspecific fashion upon activation through certain signaling pathways and cell cycle transitions (31,33,38). One possible explanation for our results may be that the association of EMP2 and ␣ 6 ␤ 1 prolongs the half-life of this integrin (26). Alternatively, EMP2 may selectively promote the heterodimer formation of the ␤ 1 integrin subunit with the ␣ 6 versus ␣ 5 integrin chain. Finally, EMP2 may function in the trafficking of specific membrane-bound proteins from the ER-Golgi compartment to the plasma membrane. EMP2 appears to modify the association of a diverse set of proteins and glycolipids with caveolin-independent lipid rafts. 3 The mechanism of these interactions may thus provide a fresh insight into the processes affecting surface receptor expression and the consequences for cellular functional phenotype. Clarification of such EMP2-dependent regulatory mechanisms may offer new perspectives on the function of GAS3/PMP22 proteins as well as on the cell biology of integrins.