Profiling of the Tetraspanin Web of Human Colon Cancer Cells *S

Tetraspanins are integral membrane proteins involved in a variety of physiological and pathological processes. In cancer, clinical and experimental studies have reported a link between tetraspanin expression levels and metastasis. Tetraspanins play a role as organizers of multimolecular complexes in the plasma membrane. Indeed each tetraspanin associates specifically with one or a few other membrane proteins forming primary complexes. Thus, tetraspanin-tetraspanin associations lead to a molecular network of interactions, the “tetraspanin web.” We performed a proteomic characterization of the tetraspanin web using a model of human colon cancer consisting of three cell lines derived from the primary tumor and two metastases (hepatic and peritoneal) from the same patient. The tetraspanin complexes were isolated after immunoaffinity purification using monoclonal antibodies directed against the tetraspanin CD9, and the associated proteins were separated by SDS-PAGE and identified by mass spectrometry using LC-MS/MS. This allowed the identification of 32 proteins including adhesion molecules (integrins, proteins with Ig domains, CD44, and epithelial cell adhesion molecule) (EpCAM), membrane proteases (ADAM10, TADG-15, and CD26/dipeptidyl peptidase IV), and signaling proteins (heterotrimeric G proteins). Importantly some components were differentially detected in the tetraspanin web of the three cell lines: the laminin receptor Lutheran/B-cell adhesion molecule (Lu/B-CAM) was expressed only on the primary tumor cells, whereas CD26/dipeptidyl peptidase IV and tetraspanin Co-029 were observed only on metastatic cells. Concerning Co-029, immunohistofluorescence showed a high expression of Co-029 on epithelial cells in normal colon and a lower expression in tumors, whereas heterogeneity in terms of expression level was observed on metastasis. Finally we demonstrated that epithelial cell adhesion molecule and CD9 form a new primary complex in the tetraspanin web.

Tetraspanins are integral membrane proteins characterized by the presence of four transmembrane domains delimiting three short intracellular domains and two extracellular regions of unequal size. They exhibit significant sequence identity as well as specific structural features in the larger of the two extracellular domains (1)(2)(3). All cell types studied so far express several tetraspanins, often to a high level. These molecules have been implicated in a large variety of physiological processes such as immune cell activation, cell migration, cell-cell fusion (including fertilization), and various aspects of cellular differentiation. These molecules have also been shown to play a role in infectious diseases (e.g. malaria, hepatitis C virus, and human immunodeficiency virus), and several genetic diseases are linked to mutations in certain of these molecules (e.g. X-linked mental retardation, retinal degeneration, and incorrect assembly of human basement membranes in kidney and skin) (1)(2)(3)(4)(5)(6)(7)(8).
In cancer, clinical studies have reported a link between tetraspanin expression levels and prognosis and/or metastasis. Indeed a high level of the tetraspanins CD9 and CD82/ KAI-1 on tumor cells is associated with a favorable prognosis in breast, lung, colon, prostate, and pancreas cancers. Additionally a decreased expression level of these molecules is correlated with metastasis in these cancers (for a review, see Refs. 1 and 4). In contrast, overexpression of CD151 in lung, colon, and prostate cancers was correlated with poor prognosis (9 -11). Furthermore using in vitro and in vivo experimental models, CD9 and CD82 have been shown to act as "metastasis suppressors," whereas CD151 was shown to increase the metastatic potential (1,4,(12)(13)(14).
The function of tetraspanins is still not precisely known. We have demonstrated that several tetraspanins associate with one or a few specific molecular partners, forming small primary complexes. Thus the tetraspanin CD151 associates directly with the integrins ␣3␤1 and ␣6␤1 (3,15), whereas CD9 and CD81 have been shown to associate with two molecules with immunoglobulin domains, CD9P-1 and EWI-2 (3, 16 -18). Furthermore tetraspanins have been shown to associate with each other by a mechanism involving palmitoylation and cholesterol (3, 19 -22). Under lysis conditions preserving tetraspanin to tetraspanin interactions, these molecules have been shown to partition into the low density fractions of a sucrose gradient, indicating association with detergent-resistant domains in the membrane (3,22). Thus, tetraspanins may organize particular microdomains on the plasma membrane to which they target their partner proteins. These microdomains appear to be different from the so-called lipid rafts. The entire set of interactions involving tetraspanins is called the "tetraspanin web" (1,3,23).
A better description of the composition and the organization of the tetraspanin web appears essential to understand the function of these molecules. Moreover a comparison of the tetraspanin web in primary and metastatic cancer cells may provide new clues for understanding the role of tetraspanins in metastasis. In this study, we performed a proteomic analysis of CD9-containing complexes by mass spectrometry using several cancer cell lines from the same patient.

EXPERIMENTAL PROCEDURES
Cell Culture-The human Isreco1, Isreco2 and Isreco3 cell lines were described previously (24). HeLa and colon carcinoma cell lines Caco-2, SW48, HT29, SW480, Lovo, Colo205, and SW620 were obtained from ATCC. All cell lines were cultured in Dulbecco's modified Eagle's medium supplemented with 10% FCS, 2 mM glutamine, and antibiotics (all from Invitrogen). The cells were maintained in a 37°C humidified incubator in the presence of 5% CO 2 .
Flow Cytometry Analysis-Cells were detached using a non-enzymatic solution (Invitrogen), washed, and stained with saturating concentrations of primary mAb. After washing three times with medium, cells were incubated with 10 g/ml FITC-labeled goat anti-mouse antibody. After washing, cells were fixed with 1% formaldehyde in PBS. All incubations were performed for 30 min at 4°C. The analysis of cell surface staining was performed using a FACSCalibur (BD Biosciences).
Immunoisolation of CD9-containing Complexes and In-gel Tryptic Digestion-For identification of CD9-associated molecules, 5 ϫ 10 8 cells were lysed in 40 ml of lysis buffer containing 10 mM Tris, pH 7.4, 150 mM NaCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 1% Brij97 in the presence of protease inhibitors. Insoluble material was removed by centrifugation at 12,000 ϫ g for 15 min, and the lysates were precleared three times successively with Sepharose 4B beads (Amersham Biosciences) coupled to BSA, to goat serum (Sigma), and then to an isotype-matched mAb. Isolation of CD9-containing complexes was performed using beads coupled to mAb ALB-6. The beads were washed five times with lysis buffer, and the proteins were eluted using 1% Triton X-100 and then acetone-precipitated. The proteins were separated by 5-15% SDS-polyacrylamide gel electrophoresis under non-reducing conditions. For profiling of the tetraspanin complexes, gels were silver-stained as described previously (18). For mass spectrometry analysis, the gels were stained with colloidal Coomassie Blue (Bio-Rad). The proteins were excised and destained. The gel pieces were incubated in 100% acetonitrile for 10 min, dried, and incubated in 100 mM ammonium bicarbonate containing 10 mM DTT for 30 min at 56°C. After cooling to room temperature, the DTT solution was replaced with 55 mM iodoacetamide in 100 mM ammonium bicarbonate for 20 min at room temperature in the dark. The gel pieces were washed in 100 mM ammonium bicarbonate for 20 min, dehydrated in 100% acetonitrile, and dried. The gel pieces were swollen in a digestion buffer containing 25 mM ammonium bicarbonate and 100 ng of trypsin (Roche Applied Science). Following enzymatic digestion overnight at 37°C, the resulting peptides were extracted with 50 l of 5% formic acid for 15 min at 37°C followed by addition of 100 l of 100% acetonitrile for another 15 min at 37°C. The peptides were then dried and rehydrated in 1% formic acid.
LC-ESI-MS/MS on Ion Trap and Data Analysis-LC-MS/MS analyses were performed using an ESI ion trap mass spectrometer (LCQ Deca XP, ThermoElectron, San Jose, CA) coupled on line with a capillary nano-HPLC system (LC Packings) (Dionex, Amsterdam, ND) for liquid chromatography. The capillary column used in this study was a PepMap C 18 reverse phase (75-m inner diameter, 15 cm) (LC Packings). A linear 20-min gradient (flow rate, 170 nl/min) from 5 to 50% acetonitrile in 0.1% (v/v) aqueous formic acid was performed. All data were collected in centroid mode using data-dependent acquisition mode. After the acquisition of a full MS scan (m/z 400 -2000 Da) in the first scan event, the three most intense ions present above a threshold of 10 5 counts were subsequently isolated for fragmentation (MS/MS scan). The collision energy for the MS/MS scan events was preset at a value of 35%. The sequences of the MS/MS spectra were identified by correlation with the peptide sequences from human proteins present in the non-redundant protein sequence database (nr from the National Center for Biotechnology Information (NCBI)) using the SEQUEST algorithm incorporated into the Finnigan BIOWORKS 3.1 software. The SEQUEST search results were initially assessed by examination of the Xcorr (cross-correlation) and the ⌬Cn (delta normalized correlation) scores. As a general rule, an Xcorr value of greater than 1.5, 2.0, and 2.5, respectively, for 1ϩ, 2ϩ, and 3ϩ charged peptides and a ⌬Cn greater than 0.1 were accepted as a positive identification (29).
Cell Surface Biotinylation, Chemical Cross-linking, Immunoprecipitation, and Western Blot-Surface labeling of cells with EZ-Link sulfo-NHS-LC-biotin (Pierce) was performed as described previously (16,18). Briefly cells were washed three times in Hank's buffered saline and incubated for 30 min in 10 mM Hepes, pH 7.3, 150 mM NaCl, 0.2 mM CaCl 2 , 0.2 mM MgCl 2 containing 0.5 mg/ml EZ-Link sulfo-NHS-LC-biotin. For cross-linking, the cells were incubated for 30 min at 4°C, in the culture flask, with 100 or 500 M DSP (Pierce) in the same buffer. After cell surface biotinylation or chemical cross-linking, cells were washed three times in 20 mM Tris, pH 7.4, 150 mM NaCl, 0.2 mM CaCl 2 , 0.2 mM MgCl 2 before lysis.
Cells were lysed directly in the flask in the lysis buffer (10 mM Tris, pH 7.4, 150 mM NaCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 0.02% NaN 3 ) containing 1% of the appropriate detergent (Brij97, Triton X-100, or digitonin) (Sigma; Calbiochem) and protease inhibitors. Digitonin was first dissolved in methanol at the concentration of 10% (w/v) and then diluted in lysis buffer without CaCl 2 and MgCl 2 as described previously (15). After 30 min at 4°C, insoluble material was removed by centrifugation at 10,000 ϫ g, and cell lysate was precleared for 2 h by addition of 1 ⁄1000 volume of heat-inactivated goat serum and 20 l of protein G-Sepharose beads (Amersham Biosciences). Proteins were then immunoprecipitated by adding 2 g of specific antibody and 10 l of protein G-Sepharose beads to 200 -400 l of lysate. After 2 h of incubation at 4°C under constant agitation, beads were washed five 1 The abbreviations used are: mAb, monoclonal antibody; NHS, N-hydroxysuccinimide; DSP, dithiobis(succinimidyl)propionate; Lu/B-CAM, Lutheran/B-cell adhesion molecule; EpCAM, epithelial cell adhesion molecule; ADAM, a disintegrin and metalloprotease; TADG-15, tumor-associated differentially expressed gene 15; GPCR, G proteincoupled receptor; DAPI, 4Ј,6-diamidino-2-phenylindole. times in lysis buffer containing 1% of the appropriate detergent. The immunoprecipitates were separated by 5-15% SDS-polyacrylamide gel electrophoresis under non-reducing conditions and transferred to a PVDF membrane (Amersham Biosciences). Western blotting on immunoprecipitates was performed using specific mAbs. Proteins were revealed by enhanced chemiluminescence (PerkinElmer Life Sciences) after incubation with a streptavidin-biotinylated horseradish peroxidase complex (Amersham Biosciences) when the first antibody was coupled with biotin; otherwise a secondary goat anti-mouse antibody coupled with horseradish peroxidase (Amersham Biosciences) was used.
Immunofluorescence Staining and Confocal Microscopy of Frozen Sections-Four micrometer-thick serial sections of frozen human colon were fixed for 20 min in acetone at Ϫ20°C. After drying at room temperature, the sections were incubated for 10 min in PBS containing 10% heat-inactivated goat serum and then with 5 g/ml mAb in the same buffer for 20 min at room temperature in a moist chamber, washed in PBS, and further incubated for 20 min with goat antimouse-FITC and DAPI. After three washes the sections were mounted in Mowiol and examined with a Leica DMR fluorescence microscope. For confocal microscopy, the sections were incubated with 5 g/ml TS9b (CD9, IgG2b) and HEA125 (EpCAM, IgG1) mAbs and then with a combination of goat anti-mouse IgG2b and IgG1 antibodies labeled, respectively, with Alexa Fluor 488 and Alexa Fluor 568 (Molecular Probes). Appropriate controls were performed to assess the specificity of the labeling. Analysis was performed with a TCS SP2 confocal microscope (Leica, Wetzlar, Germany).

Proteomic Analysis of CD9-containing Complexes of Isreco
Cell Lines Using LC-MS/MS-To better define the role of tetraspanins, we investigated the composition of CD9-containing complexes in three colon carcinoma cell lines. The Isreco1 cell line was established from a primary colon cancer, whereas Isreco2 and Isreco3 were established, respectively, from liver and peritoneal metastases from the same patient (24). The two metastatic cell lines were shown to express a lower level of CD9 than did the primary tumor using PCR display and Western blot analysis (24). However, flow cytometric analysis indicated that the metastatic cell lines still expressed a high level of CD9 as compared with other surface antigens (see below).
The profiling of CD9 complexes of the three cell lines was performed by immunoprecipitation experiments. After biotin labeling of cell surface proteins, cells were lysed with the mild detergent Brij97. We have shown previously that under these conditions, tetraspanin-tetraspanin interactions are preserved, and tetraspanin complexes can be immunoprecipitated by mAbs directed against any tetraspanin (23). CD9containing complexes were isolated by immunoprecipitation, and the CD9-associated proteins were eluted using the more stringent detergent Triton X-100, which dissociates tetraspanin-tetraspanin associations (15,23). After SDS-PAGE, the proteins were either transferred to a PVDF membrane and revealed by enhanced chemiluminescence or silver-stained to visualize the pattern of tetraspanin-associated molecules. The pattern of proteins co-immunoprecipitated with CD9 from the primary tumor cell line exhibited several differences compared with the pattern obtained from the two metastatic cell lines (Fig. 1). One of the major differences was an intense 30-kDa band observed in the pattern of the metastatic cell lines and absent in the immunoprecipitates collected from the primary tumor cell line.
The components of CD9 complexes in the Isreco cell lines were identified by mass spectrometry. The CD9-associated proteins were isolated from ϳ5 ϫ 10 8 cells using anti-CD9coated beads, separated by SDS-PAGE, and stained using silver or colloidal Coomassie Blue. Each lane was systematically cut (in about 20 slices), and the proteins were in-gel FIG. 1. Profiling of CD9-containing complexes. The CD9-containing complexes were solubilized using the mild detergent Brij97 and isolated by immunoprecipitation experiments using specific CD9 mAbs. The associated proteins were eluted using the more stringent detergent Triton X-100 and separated by SDS-PAGE. A, the proteins were labeled with biotin before lysis and transferred to a PVDF membrane after SDS-PAGE. The proteins were revealed by streptavidin-peroxidase and chemiluminescence. B, the proteins were revealed by silver staining. IP, immunoprecipitation; cont., control.
digested with trypsin. To eliminate background proteins from analysis, the IgG1-coated beads that were used in the last preclearing step were treated identically to anti-CD9 beads.
Any protein identified in both samples was not considered as a tetraspanin-associated protein. The resulting peptides were analyzed using LC-MS/MS, which allows separation, isola- . At a precise time (e.g. 27.74 min, dotted straight line), the mass spectrum obtained is shown (middle panel) in which a parent ion can be selected (e.g. m/z ϭ 699.7, black arrow). The fragmentation of that parent ion led to MS/MS spectrum generation containing b and y ions and thus sequence information of the parent ion (lower panel). The amino acid sequence can be deduced after search in the NCBI database using the program SEQUEST. The putative sequence of the peptide is shown with associated Xcorr and ⌬Cn. THis peptide sequence led to the identification of the protein CTL2. tion, and fragmentation of peptides from a complex mixture and thus protein identification from a single peptide (Fig. 2). The proteins were identified with one to eight peptides (see supplemental data).
A comparative analysis of the three Isreco cell lines showed that some proteins were identified in all cells (such as EpCAM, ADAM10, CTL1/CDw92, and CTL2) whereas others were differentially detected (Table I). The proteins Lu/B-CAM, TADG-15, syntaxin-3, some G proteins, and most tetraspanins were detected only in the primary tumor cell line Isreco1. By contrast, the tetraspanin Co-029 was detected only in the metastatic cell lines, and CD26 was detected only in Isreco3.  a Integrin ␤4 exhibited a 125-kDa molecular mass because of a proteolytic cleavage as described previously (61).

Profiling of the Tetraspanin Web
Cell Surface Expression and Association with CD9 of Selected Proteins-The differences observed in the CD9 complexes collected from the three cell lines may be the consequence of different levels of expression of these molecules. We thus examined the expression levels of tetraspanins and some of their associated proteins at the cell surface by indirect immunofluorescence and flow cytometry when antibodies were available (Fig. 3).
There was a reasonable relationship between the expression of the different molecules tested and their association with CD9. Thus, CD9P-1, CDw92, and EpCAM, which are expressed by all three cell lines, were detected in the CD9 immunoprecipitates collected from the three cell lines. Additionally the differential detection of CD26 in Isreco3 and of Co-029 in Isreco2 and Isreco3 reflects differences in expression. On the other hand, Lu/B-CAM was expressed only by Isreco1, consistent with the detection of Lu/B-CAM peptides only in this cell line. It should be noted that EWI-2 and integrins ␣6 and ␤4 were differentially detected by mass spectrometry despite a similar expression at the surface of all three cell lines. This apparent discrepancy may suggest a differential targeting of these proteins to the tetraspanin web. According to this hypothesis, we observed a higher amount of integrin ␣6␤4 co-immunoprecipitated with CD9 in Isreco2 and Isreco3 as compared with Isreco1 (data not shown).
To further validate the association with CD9 of some novel identified molecules, we examined whether the interactions could be observed by reverse co-immunoprecipitation (Fig.  4). After cell surface biotinylation and lysis using Brij97, immunoprecipitation experiments were performed using mAbs directed against Co-029, EpCAM, CTL1/CDw92, Lu/B-CAM, and CD26. The pattern of proteins co-immunoprecipitated with Co-029 in the metastatic cell lines was identical to that of CD9 (Fig. 4A). This suggests that when the tetraspanin Co-029 is expressed it is included in the tetraspanin web, and it associates with the same proteins as CD9. Co-029 is a 30-kDa molecule, and immunoprecipitation after Triton X-100 lysis demonstrated that the intense 30-kDa molecule present in the CD9 or Co-029 immunoprecipitates collected from the metastatic cell lines is indeed Co-029 (data not shown). The major proteins immunoprecipitated with anti-CD26, -Lu/B-CAM, and -EpCAM mAbs exhibited a molecular mass of 110, 80, and 40 kDa, respectively. This is consistent with the previously described molecular masses for these molecules. CDw92 was poorly labeled with biotin and was not visible at the exposure shown in Fig. 4. Each of these molecules coimmunoprecipitated from Isreco1 cell lysates a band co-mi-

FIG. 4. Association of novel identified proteins with tetraspanins.
A, after cell surface biotinylation, cells were lysed using Brij97. Immunoprecipitation experiments were performed using mAbs directed against tetraspanins CD9 and Co-029 or novel identified proteins. After SDS-PAGE, the proteins were transferred to a PVDF membrane and revealed by streptavidin-peroxidase and chemiluminescence. B, cells were lysed using Brij97 followed by immunoprecipitation and Western blotting with biotin-labeled CD9 mAbs. IP, immunoprecipitation; Int., integrin. grating with CD9. This band, as well as a band co-migrating with Co-029, was present in EpCAM and CDw92 immunoprecipitates collected from Isreco2 and Isreco3 cell lysates (Fig.  4A). The identity of this band as CD9 was demonstrated by Western blotting using CD9 mAbs (Fig. 4B).
EpCAM Is a CD9 Molecular Partner-Inside the tetraspanin web, certain tetraspanins have been shown to specifically and directly interact with a limited number of proteins, their molecular partners (forming primary complexes). These specific associations can be observed using detergents that disrupt tetraspanin-tetraspanin interactions or precipitate tetraspanins that interact with each other as in the case of digitonin (1,15,16,18). To identify a tetraspanin partner for EpCAM, CDw92, and Lu/B-CAM, immunoprecipitation experiments were performed after cell lysis using digitonin. Under these conditions, EpCAM but not CDw92 clearly immunoprecipitated a molecule co-migrating with CD9 (Fig. 5A). This band was identified as CD9 by Western blotting (Fig. 5B). Importantly EpCAM co-immunoprecipitated a higher amount of CD9 than did a well established CD9 partner, CD9P-1. In addition, no CD81 was detected in the EpCAM immunoprecipitate. An additional band of ϳ20 kDa was co-immunoprecipitated with EpCAM (Fig. 5A). This protein might correspond to Claudin 7, which was identified recently as an EpCAMassociated protein (32). No band co-migrating with EpCAM was observed in the CD9 immunoprecipitate after digitonin lysis. This suggests that the CD9 mAb may dissociate the CD9⅐EpCAM complex or that EpCAM interaction prevents the binding of the CD9 mAb. This is not unprecedented as CD81 mAbs fail to co-immunoprecipitate CD19 (a well characterized partner) (33), and certain CD151 mAbs fail to co-immunoprecipitate the integrins ␣3␤1 and ␣6␤1 (15). It should be noted that a weak band co-migrating with CD9 was also observed in the Lu/B-CAM immunoprecipitate after digitonin lysis (Fig.  5A). However, we believe that it would be premature to consider Lu/B-CAM as a CD9 partner because we could not observe Lu/B-CAM⅐CD9 complexes after Western blotting or in cross-linking experiments.
Cross-linking experiments were performed to determine whether CD9 interacts directly with EpCAM. Intact Isreco1 cells were first pretreated with DSP as a cross-linking reagent. Then cells were lysed under stringent conditions to disrupt the non-covalent associations, and immunoprecipitation experiments were performed with a specific anti-EpCAM mAb. The immunoprecipitates were run under non-reducing conditions, and the complexes containing EpCAM were first visualized by Western blot using the anti-EpCAM mAb. This approach revealed the existence at the cell surface of two complexes containing EpCAM with molecular masses of ϳ60 and 80 kDa. The ϳ60-kDa complex was recognized by the CD9 mAb after Western blotting. Importantly the molecular mass of this complex is consistent with a complex containing only CD9 (24 kDa) and EpCAM (40 kDa). The ϳ80-kDa complex may correspond to EpCAM dimers (34) (Fig. 5C). To gain further information about the potential relevance of CD9⅐EpCAM complexes, the distributions of these molecules in normal and colon cancer were compared by confocal microscopy. There was a substantial colocalization of these two molecules in the normal colon and a lower level of colocalization in primary tumor and metastasis (Fig. 6).  (Fig. 7). Both molecules were expressed in primary colon cancer and adjacent "normal" colon tissue. A clear reduction of B-CAM expression was observed in all tumor biopsies, whereas a reduction of Co-029 expression was observed in about half of the eight samples tested (Fig. 7 and data not shown). In addition, the molecular weight of Co-029 was lower in tumor samples than in normal tissues. This suggests that modifications of the protein Co-029 that remain to be characterized may occur in certain tumors (Fig. 7). The molecular weight of Co-029 in the panel of cell lines described below was identical to that observed in normal adjacent tissues (data not shown).

Expression of Components of the Tetraspanin Web in Normal Tissue and Colon
We then examined the expression of these molecules in primary tumors, normal adjacent colon, and metastasis from the same patients by immunolabeling of frozen sections. On normal colon tissue, an intense labeling of the lateral surface of epithelial cells was observed with the CD9 and Co-029 mAbs. The expression of Co-029 was restricted to epithelial cells, whereas CD9 was also expressed by mesenchymal cells. The overall expression of CD9 and Co-029 was lower on primary and metastatic tumors with Co-029 labeling being heterogeneous (Fig. 8). Similar results were obtained with two additional patients.
The high expression of Co-029 observed on Isreco2 and Isreco3 contrasted with the low expression of this molecule observed in the metastasis of the three patients analyzed. This suggested a possible variability of Co-029 expression in metastasis. To challenge this hypothesis, we first tested the expression of Co-029 in a panel of cell lines derived from primary tumors and metastasis. As shown in Fig. 9A, the two cell lines derived from low grade colon cancers (HT29 and Caco-2) had a higher Co-029 expression than the other primary tumor-derived cell lines. The Co-029 expression level was highly variable in the metastatic cell lines as compared with the cells derived from primary tumors. Variability in the level of expression of Co-029 was also observed when analyzing a small series of liver metastasis by immunohistochemistry. Both Co-029-positive and -negative metastasis were observed, and this heterogeneity was confirmed by Western blot (Fig. 9B).

DISCUSSION
In this study, we investigated the composition of CD9containing complexes in colon carcinoma cells using proteomics. This analysis led to the identification of 32 proteins and showed that CD9 associates with a so far unsuspected variety of membrane proteins. Most of these molecules can be classified as adhesion molecules and/or molecules with Ig domains, membrane proteases, signaling molecules, and finally tetraspanins.
Several categories of adhesion molecules in the tetraspanin web were identified including integrins (␣3␤1, ␣6␤1, and ␣6␤4), molecules with Ig domains (CD9P-1, EWI-2, and Lu/ B-CAM), and EpCAM. The protein Lu/B-CAM exhibits five Ig-like domains and is a laminin receptor with a restrictive specificity to laminin ␣5 chain (35). Lu/B-CAM was detected at the surface of the Isreco1 cell line but not on the metastatic cell lines. Furthermore Lu/B-CAM expression was decreased in tumor samples compared with normal adjacent tissue as determined by Western blot. However, Lu/B-CAM was mainly detected in some stroma cells in tissue sections (data not shown), and thus further work will have to determine whether the loss of expression of this molecule on metastatic cells is a general feature of colon cancer. Another new molecule in the tetraspanin web is EpCAM. EpCAM is a molecule with two epidermal growth factor-like domains that functions as a homophilic cell-cell adhesion molecule. EpCAM is expressed in many human epithelial tissues (36) and is overexpressed in the majority of epithelial carcinomas (for a review, see Ref. 37). For this reason, it has attracted major attention as a target for mAb-based immunotherapy of carcinomas (38). There is also evidence that EpCAM regulates cell proliferation of carcinoma cells (39,40). During the course of this study, Claas et al. (31) demonstrated an association of EpCAM with CD9 and Co-029 in rat cells. This association could be observed only when using weak detergents. In this study we demonstrated FIG. 8. Expression of CD9 and Co-029 in normal and cancer colon. Cryostat sections of normal human colon, colon cancer, and colon metastasis in the liver (all from the same patient) were stained with mAb TS9 to CD9 (A) or AZM22 to Co-029 (B) and then with a FITC-labeled goat anti-mouse antibody. The acquisition time was 2 s except for the staining of Co-029 in the primary tumor and the liver metastasis for which there was a 4-s acquisition. On the right is shown a superimposition with the DAPI staining (blue) that gives an estimate of the number of cells. Bar, 100 m (identical in all panels). that the interaction of EpCAM with CD9 can be visualized using digitonin (and thus is observed under conditions where tetraspanin to tetraspanin interactions are not observed or are strongly diminished) and stabilized by chemical cross-linking. Therefore, CD9⅐EpCAM constitutes a new primary complex in the tetraspanin web. Recently knock-down of EpCAM by RNA interference was shown to strongly diminish migration and invasion of a breast cancer cell line in vitro (39). It will be of special interest to determine whether the effects of EpCAM and tetraspanins on cell migration and invasion are functionally linked.
An exciting finding of this study is the presence of several transmembrane proteases in the tetraspanin web. These components may shed new light on the function of tetraspanin complexes, which may regulate proteolytic activities at the cell surface. Indeed transmembrane proteases participate in extracellular proteolysis such as degradation of extracellular matrix components, regulation of chemokine activity, and release of membrane-anchored growth factors, receptors, and adhesion molecules that influence cell growth and motility (41). Our data suggest that ADAM10 is a component of the tetraspanin web in colon carcinoma cells. ADAM (a disintegrin and metalloprotease) proteins are membrane-anchored metalloproteases that process and shed ectodomains of membrane-anchored growth factors, cytokines, and receptors. ADAMs also have essential roles in cell-cell and cellmatrix interactions and therefore in a variety of physiological and pathological processes including angiogenesis and cancer (42). ADAM10 may play a role in cancer through its ability to cleave transmembrane precursors of epidermal growth factor receptor ligands, including heparin-binding epidermal growth factor (43), which has long been known to associate with tetraspanins (44,45). CD26/dipeptidyl peptidase IV is a 110-kDa glycoprotein that belongs to the prolyl-oligopeptidase family. It selectively removes the N-terminal dipeptide from peptides with proline or alanine in the second position. Thus CD26 truncates many bioactive molecules including growth factors, chemokines, neuropeptides, and vasoactive peptides and is responsible for the inactivation of many bioactive peptides (46,47). It is expressed on a variety of tissues including T lymphocytes and endothelial and epithelial cells. CD26 plays an important role in immune regulation, signal transduction, and apoptosis as well as in tumor progression (46,47). In colorectal cancer, although CD26 is not present on normal human colon epithelium, it is sometimes aberrantly expressed in colon tumors. TADG-15 (tumor-associated differentially expressed gene 15)/matriptase is a transmembrane trypsin-like serine protease. TADG-15 expression was found in all types of epithelia. TADG-15 has been shown to cleave and activate several proteins that may play a role in the growth and invasion of cancer cells during tumor progression such as hepatocyte growth factor/scatter factor, urokinase plasminogen activator, and protease-activated receptor-2 (48 -50). TADG-15 was identified in the CD9 complex with only one peptide. Further work will be necessary to confirm this interaction Other findings that may shed new light on the function of tetraspanins are the presence of signaling molecules in the complexes. Only a few signaling molecules were demonstrated to associate with tetraspanins. Indeed the association of tetraspanins with phosphatidylinositol 4-kinase and with several protein kinase isoforms has been reported previously (3). However, the interaction of phosphatidylinositol 4-kinase with tetraspanins was only observed using detergents milder than Brij97, and that of protein kinase C was only observed after activation with phorbol esters. Recently Little et al. (30) showed that a fraction of heterotrimeric G proteins, G␣ q and G␣ 11 subunits, specifically associates with CD9 and CD81 as well as an unidentified G␤ subunit. Our data suggest that additional G proteins could be associated with tetraspanins. Heterotrimeric G proteins are coupled with seven-transmembrane domain receptors also called G protein-coupled recep- FIG. 9. Co-029 is heterogeneously expressed on colon metastasis. A, the expression level of Co-029 at the cell surface of cell lines originated from colon tumor (T) or metastasis (M) was determined by flow cytometry. B, biopsies from colon metastasis in the liver were used as the source of protein extracts for Western blotting experiments. Expression of tetraspanins Co-029 and CD9 was checked. Extracts from HeLa (uterine cervical cancer) and Lovo (colon metastasis) cell lines as well as from a biopsy of normal liver were also used. tors (GPCRs) (51). However, no GPCR was identified in our analysis of CD9-containing complexes. This may be related to a difficulty in obtaining peptides from largely hydrophobic proteins. Alternatively tetraspanins may associate with G proteins in the absence of GPCR.
Mass spectrometry and flow cytometry analysis of CD9containing complexes allowed the detection of 10 tetraspanins in the Isreco1 cell line. Among them, Tspan14/DC-TM4F2, Tspan-1/NET-1, Tspan9/NET-5, and Tspan15/NET-7 were observed only by mass spectrometry as there are no available reagents to these molecules. Tspan1/NET-1, Tspan9/NET-5, and Tspan15/NET-7 were not observed on the metastatic cell lines. This may indicate a lower expression level of these tetraspanins. In this regard, two tetraspanins (CD9 and CD82) of five studied by flow cytometry were down-regulated at the surface of the metastatic cell lines. By contrast, the tetraspanin Co-029 was detected by mass spectrometry only in CD9 complexes collected from the metastatic cell lines Is-reco2 and Isreco3. Flow cytometry and immunoprecipitations showed the expression of Co-029 only in Isreco2 and Isreco3 cell lines. Co-029 was originally identified by the generation of mAbs against a human colon carcinoma cell line (52,53) and was reported to be expressed in a variety of carcinomas. It was later identified in the rat as a molecule expressed on a metastatic subclone of a pancreatic adenocarcinoma cell line but not on a low metastasizing subclone of the same cell line (54). Co-029 was shown to be overexpressed in cirrhosis (55) as well as in hepatocarcinoma in particular with intrahepatic spreading (56). Altogether these results prompted us to examine the expression of Co-029 in the colon in comparison with CD9 whose expression in the normal colon has been reported (57). Surprisingly a high expression of Co-029 was observed in the normal colon that was confirmed by Western blot analysis. In contrast to CD9, the expression of Co-029 was restricted to epithelial cells. Both molecules were only expressed at lateral surfaces. For three patients, we could compare the normal adjacent colon with the primary tumor and the liver metastasis. In these patients, the expression of Co-029 was low both in the primary tumor and in metastasis in apparent contradiction with the analysis of Isreco cell lines. This discrepancy led us to investigate a possible heterogeneity of Co-029 expression in colon metastatic cells. Such heterogeneity is shown through the analysis of different cell lines and different biopsies of colon metastases in the liver. It will be necessary to determine in further studies with larger series of patients whether the level of Co-029 expression in metastasizing cells could influence their behavior and the clinical outcome. In this regard, it has been suggested that Co-029 may promote metastasis using in vitro and in vivo experimental models (58 -60). Indeed transfection of a low metastasizing rat pancreatic cell line with Co-029 resulted in increased metastatic potential with massive bleeding around the metastases. In addition, transfection of Co-029 in HCC cells promoted development of intrahepatic metastatic lesions. Further work will have to determine whether Co-029 plays an active role in the metastatic process during colon cancer progression.
In conclusion, we largely extended the list of molecules associating with CD9 and therefore to the tetraspanin web. Clinical studies have previously shown a link between the expression level of tetraspanins and tumor progression and metastasis in many cancers. We made the observation using tumor cell lines and a few patient samples that several components of the tetraspanin web appear to be differentially expressed during tumor progression. That may change tumor cell properties and contribute to invasion and metastasis. The clinical and functional relevance of the association in the plasma membrane of the various components of the tetraspanin web remain to be addressed in further studies.