Oligomeric State of the Colon Carcinoma-associated Glycoprotein GA733-2 (Ep-CAM/EGP40) and Its Role in GA733-mediated Homotypic Cell-Cell Adhesion*

The GA733-2 antigen (GA733) is a homotypic calcium-independent cell adhesion molecule (CAM) present in most normal human epithelial cells and gastrointestinal carcinomas. Because oligomerization of some CAMs reg-ulates cell adhesion and signal transduction, the corre-lation between GA733 oligomeric state and cell-cell adhesion was investigated. Sedimentation equilibrium studies showed that full-length (-FL) GA733 exists as dimers and tetramers in solution, whereas the GA733 extracellular domain (-EC) is a monomer. The K d of GA733-FL is less than 10 n M for the monomer-dimer association, whereas the dimer-tetramer association is about 1000-fold weaker ( K d ; 10 m M ). Chemical cross- linking of purified GA733-FL in solution resulted in a major product corresponding to GA733 dimers, and minor amounts of trimers and tetramers. However, GA733-EC cross-linked under the same conditions was consistently a monomer. Chemical cross-linking of dissociated colon carcinoma cells produced predominantly GA733 dimers, whereas cross-linking of cells in monolayers yielded some tetramers as well. GA733-FL re-tained its cell-cell adhesion function as shown by inhibition of cell aggregation, whereas monomeric GA733-EC was inactive. These data show that GA733

More recently, it was shown to function as a Ca 2ϩ -independent homotypic cell adhesion molecule (CAM) (3). However, it appears to belong to a novel class of CAMs distinct, by its structural features, from the four known superfamilies of CAMs (immunoglobulin-type, selectins, cadherins, and integrins) (for review see Refs. 4 and 5). The sequence of GA733-2 indicates no significant homology with other known proteins except with GA733-1 (Trop-2), previously cloned by Linnenbach et al. (6). The amino acid sequences of GA733-2 and GA733-1 are ϳ50% identical.
GA733-2 (referred to hereafter as GA733) is synthesized as a 32.8-kDa nonglycosylated precursor, but migrates in SDS-PAGE gels as a 40-kDa band. The extracellular portion of GA733 contains three potential N-linked glycosylation sites and consists of a cysteine-rich N-terminal region followed by a cysteine-poor region. GA733 has a short transmembrane domain and a cytoplasmic tail of 26 amino acids with potential sites of phosphorylation on tyrosine and serine residues.
GA733 is present in most epithelial cells and is overexpressed in the vast majority of gastrointestinal tumors (7)(8)(9)(10). Moreover, emerging evidence correlates the increase of GA733 expression with other types of cancer lesions, notably in cervical squamous epithelia (11). The increase of cell proliferation in cervical squamous epithelia, which are normally GA733-negative, correlates with the expression of GA733. An increase of the number of positive cells as well as in levels of GA733 protein was observed during progression of cervical lesions from cervical intraepithelial neoplasia grade I to grade III (11).
The significance of increased GA733 expression in cancer lesions remains mysterious. Surprisingly, a recent study from our group showed that GA733 inhibits invasion of tumor cells in vitro when transfected into CT-26 mouse colon carcinoma cells (12). In the same study, the mouse homologue of GA733, mEGP (a homotypic CAM with 82% amino acid sequence identity with GA733) expressed by transfected mouse CT-26 cells (that are mEGP-negative) significantly decreases cell growth in vitro and inhibits metastasis in vivo. Although the mechanism(s) responsible for these two phenomena remain unclear, the strong intercellular adhesion profile of CT-26-mEGP cells (clumping and clustering observed in vitro) may inhibit cell dispersion and limit their migration capacity (12). The same mechanism was proposed to explain decreased metastases in nude mice by human breast carcinoma cells overproducing another homotypic adhesion molecule, E-cadherin (13). Alternatively, the decrease of cell growth may be due to the activation of certain mEGP-associated signal transduction pathways that down-regulate cell proliferation, when activated by increased homotypic engagement of mEGP. Although very little is known about the biochemical properties of GA733 and its murine homologue, mEGP, other than their role as homotypic CAMs, it is likely that they participate in "inside-out" and "outside-in" cell signaling pathways controlling cell growth and adhesion. This hypothesis is supported by a recent report showing that GA733-1 (Trop-2) is a calcium signal transducer (14).
Several clinical trials targeting GA733 by anti-idiotypic antibodies and recombinant protein vaccines have shown induction of humoral and cellular immunity in colorectal cancer patients (15)(16)(17)(18)(19)(20). More importantly, CO17-1A mAb against GA733 enhances survival of patients with colon carcinoma in a phase II randomized control study (21,22). Furthermore, a recombinant adenovirus expressing GA733 significantly enhanced survival of mice bearing established CT-26-GA733 tumors (23). The inhibitory effect of GA733 on cancer cell growth and invasion, its increased expression in human cancer lesions, and the beneficial treatment with mAbs against GA733 in experimental animals and patients appear to be paradoxical. Clearly, further characterization of the biochemical and functional properties of this protein and its role in tumor progression is needed.
Recent studies have suggested that, in addition to mediating homotypic associations between opposing cell membranes, cis interactions between some types of receptors on the same cell surface can play an important role in regulating cell adhesion, signal transduction, and migration (24). For example, it has been shown that C-CAM (25), PECAM-1 (26), and ICAM-1 (27,28) exist as cis-dimers in the cell membrane. Furthermore, dimeric forms of recombinant soluble ICAM-1 have significantly higher affinities for its purified ligand, LFA-1 (CD11a/ CD18), whereas the monomeric form shows no binding (27,28), illustrating the central role CAM oligomerization can play in cell-cell adhesion.
The molecular organization of GA733 within normal epithelia and colon carcinoma cells remains unknown, despite its use with some success as a target for passive and active immunotherapy of colon cancer in human clinical trials (15)(16)(17)(18)(19)(20)(21)(22). Elucidating the cis-oligomerization of GA733 on the cell membrane and the mechanism of GA733-mediated cell adhesion should contribute to a better understanding of GA733 function and may lead to substantial improvements of GA733 targeting in immunotherapy of colon cancer.
In this study, we investigated the oligomeric state of membrane-associated and soluble forms of GA733 and their ability to block cell-cell adhesion. We show that GA733-FL, which inhibits cell aggregation, self-associates to form high affinity cis-dimers in solution and in intact colon carcinoma cell membranes, whereas the soluble recombinant extracellular domain (GA733-EC) is a monomer and does not inhibit cell aggregation.
Protein Purification-Recombinant GA733-FL and GA733-EC were produced in insect cells using the baculovirus system as described previously (30,31). GA733-FL was essentially produced as described for GA733-EC (31), except that a full-length GA733 cDNA was used (32). All purification steps were performed at 0 -4°C. For recombinant GA733-FL, typically 500 ml of baculovirus-transfected Sf9 insect cells (ϳ12 ϫ 10 8 cells) were harvested by centrifugation at 48-h post-infection, and the cell pellet was resuspended in lysis buffer (50 l/10 6 cells) containing 100 mM Tris-Cl (pH 7.6), 10 mM EDTA, 10 mM EGTA, 1 mM ␤-mercaptoethanol, 2% Triton X-100, 40 mM sodium chloride, 0.15 mM phenylmethylsulfonyl fluoride (PMSF), 2 mM leupeptin, and 10 g/ml pepstatin, pH 7.4. The lysate was sonicated, centrifuged for 20 min at 16,000 ϫ g, and the supernatant was filtered through a 0.22-m filter and loaded onto a column containing 30 mg of affinity-purified GA733 mAb coupled to Sepharose CL-4B by CNBr activation. The column was washed with a buffer containing 100 mM Tris, pH 7.4, with 0.05% Triton X-100, 10 mM EDTA, 10 mM EGTA, 0.15 mM PMSF, 2 mM leupeptin, and 10 g/ml pepstatin, washed with the same buffer without detergent and bound proteins were eluted with 50 mM glycine, pH 2.5. The pH of the peak fractions was immediately neutralized, followed by dialysis against PBS (10 mM sodium phosphate, 130 mM NaCl, pH 7.4) supplemented with 0.15 mM PMSF, 5 mM EDTA, 0.02% sodium azide. The quality of purifications was monitored by SDS-PAGE and Western blot analysis using the GA733 mAb.
During the purification process, some covalent cross-linking of GA733 occurs presumably due to the activity of insect cell transglutaminases, enzymes that catalyze the formation of ␥-glutamyl-⑀-lysine bonds. Oligomers of GA733 (up to tetramers, but primarily dimers) could be detected by Coomassie Blue-stained SDS gels and on Western blots. Maintaining 100 mM Tris, with 10 mM EDTA and 10 mM EGTA throughout the washes, substantially minimized this artifactual crosslinking and was used in most purifications. Alternatively, we were able to completely eliminate this artifactual cross-linking by using 0.5% Triton X-100 in the wash and elution buffers. However, such preparations were used only for chemical cross-linking experiments, because Triton X-100 could not be efficiently removed by dialysis and attempts to use ion-exchange chromatography to remove the detergent were unsuccessful, making these preparations unsuitable for either cell aggregation assays or analytical ultracentrifugation studies.
The secreted GA733-EC was purified from the culture supernatant of baculovirus-infected Hi Five insect cells using a GA733 mAb-Sepharose affinity column. The column was washed with PBS, and the bound proteins were eluted with 100 mM triethylamine, pH 11.5, purchased from Pierce.
Protein Assay-Protein samples were concentrated using Centriprep 10 concentrators (Amicon, Beverly, MA). GA733-EC was concentrated in PBS, whereas GA733-FL was concentrated in PBS containing either 0.5 mM octyloxyethylene dodecyl ether (C 12 E 8 ) or 10 mM pentaoxyethylene octyl ether (C 8 E 5 ) (Sigma). GA733-FL protein concentrations were determined using the bicinchoninic acid protein assay (Pierce). The concentration of GA733-EC was determined by spectrophotometry at 280 nm in a 1-cm path length using an extinction coefficient of 0.926 (mg/ml) Ϫ1 calculated from the GA733-EC sequence according to Pace et al. (33).
HPLC Gel Filtration-For analytical ultracentrifugation experiments, recombinant proteins were further purified by HPLC gel filtration to remove trace contaminants and aggregates. GA733-EC was concentrated as described previously, injected into two TSK columns, G3000 SWXL and G2000 SWXL (Toso-Haas, Japan), connected in series, and separated at 0.8 ml/min using PBS. GA733-FL was concentrated in PBS containing 0.5 mM C 12 E 8 and chromatographed on a Superose 12 column (Amersham Pharmacia Biotech) at 0.5 ml/min using PBS containing 0.15 mM PMSF and 0.5 mM C 12 E 8 .
Mass Spectrometry-Matrix-assisted laser desorption/ionizationtime of flight (MALDI-TOF) mass spectrometry was carried out on a Voyager DE-PRO mass spectrometer (PE Biosystems, Framingham, MA). Recombinant protein solutions (2 l) were mixed 1:1 with a saturated solution of sinapinic acid in 33% acetonitrile and 0.1% trifluoroacetic acid and allowed to air dry on the sample target before analysis. Several 2-fold serial dilutions of the protein samples were tested, and protein A was used as an external and internal standard.
Analytical Ultracentrifugation-Experiments were performed in an Optima XL-I analytical centrifuge. Sedimentation equilibrium runs were performed with three initial concentrations of the protein sample. Absorbance or fringe displacement data was collected every 4 -6 h until equilibrium was reached, as determined by comparison of successive scans using the MATCH program.
Sedimentation equilibrium experiments were performed using either the interference optics (for GA733-EC) or the absorbance optics (for GA733-FL) to measure the protein concentration gradient. For experiments with GA733-EC, three cells were assembled with double-sector 12-mm centerpieces and sapphire windows. The cells were loaded with 110 l of HPLC gel filtration buffer as the reference, and 110 l of sample at concentrations of 2.0, 1.0, or 0.5 mg/ml. A blank scan of distilled water was taken before the run, to correct for the effects of window distortion on the fringe displacement data (34). Experiments were performed at 4°C or 30°C, at 23,000 -30,000 rpm. Fringe displacement data was collected every 4 -6 h until equilibrium was reached.
Sedimentation equilibrium experiments on GA733-FL were performed in the presence of either C 8 E 5 or C 12 E 8 nonionic detergents. For experiments using C 8 E 5 , affinity-purified GA733-FL was concentrated in PBS containing 0.15 mM PMSF and 10 mM C 8 E 5 and dialyzed against the same buffer for 24 h at 4°C. For experiments in the presence of C 12 E 8 , GA733-FL was chromatographed using HPLC gel filtration as described above. Fractions containing purified GA733-FL were then concentrated to 0.1-0.4 mg/ml and dialyzed for at least 24 h at 4°C against PBS containing 0.15 mM PMSF and 0.5 mM C 12 E 8 in 22% D 2 O (v/v). In the presence of detergent, the protein is part of a proteindetergent complex that has a buoyant molecular mass, M c (1 Ϫ Ј), containing contributions from the protein and the bound detergent (35,36), where M P is the molecular weight of the protein, P and Det are the partial specific volumes of the protein and detergent respectively, ␦ Det is the number of grams of detergent bound per gram of protein, and is the density of the solvent. Because the partial specific volume of C 12 E 8 is known (0.973 cm 3 /g (37)), the density of the solvent can be adjusted to 1/ Det with D 2 O. In this case the second term of the equation equals zero, effectively removing the contribution of the bound detergent. The density of the dialysis buffer described above was calculated to be 1.0278 g/cm 3 at 20°C using the SEDNTERP program. For experiments on GA733-FL, three cells were assembled with double-sector 12-mm centerpieces and quartz windows, and loaded with 110 l of dialysis buffer as the reference, and 110 l of sample at three protein concentrations (typically 0.2, 0.1, and 0.05 mg/ml). Experiments were performed at 4 or 20°C at speeds between 14,000 and 23,000 rpm. Absorbance data at 280 or 230 nm was collected every 4 -6 h until equilibrium was reached. For each scan, data were acquired every 0.001 cm with up to nine replicates in continuous scan mode.
For all experiments, attainment of sedimentation equilibrium was determined by comparison of successive scans using the MATCH v.7 program, and the data were edited using the REEDIT v.9 program (both programs kindly provided by David Yphantis). Nonlinear regression fitting of the sedimentation equilibrium data with various models was performed using the NONLIN program (38). The reduced molecular weight, of a protein is defined as: ϭ M(1 Ϫ ) 2 /RT, where M and are the molecular weight and the partial specific volume of the protein, respectively, R is the gas contant T is temperature in Kelvin, and the density of the solvent (34). The program SEDNTERP was used to calculate of GA733-EC and GA733-FL, using the known amino acid sequence of the proteins and an estimate of the amount of carbohydrate determined from MALDI mass analysis, which was treated as mannose groups for calculation purposes of . At least three data sets from different loading concentrations and/or rotor speeds were fitted simultaneously. Goodness of fit was determined by examination of the residuals and minimization of the variance. The association constants returned by NONLIN from self-association models were converted to the molar scale using the calculated molar extinction coefficient of the protein.
Chemical Cross-linking-Prior to cross-linking, the cells were washed twice with PBS. Adherent or detached Caco-2 and Colo-205 cells were overlaid with PBS, and cross-linking was initiated by the addition of either DSG or BS 3 at different concentrations (0.1, 0.2, 0.5, 1, and 2 mM). After 5 min at room temperature, the reaction was quenched for 15 min with Tris-Cl, pH 7.6, at a final concentration of 50 mM. Cells were lysed for 20 min at 4°C in a buffer containing 0.5% Triton X-100, 20 mM Tris-Cl (pH 7.6), 5 mM EDTA, 1 mM ␤-mercaptoethanol, 0.5 g/liter deoxycholic acid, 0.1% SDS, 40 mM sodium chloride, 0.15 mM PMSF, 2 mM leupeptin, and 10 g/ml pepstatin, pH 7.4. Cell extracts were centrifuged for 20 min at 16,000 ϫ g, the pellets were discarded, and the supernatants were saved for immunoprecipitations, SDS-PAGE, and Western blots.
For purified proteins, cross-linking was performed using 16 g of GA733-FL or GA733-EC in 100 l of PBS containing different detergents (i.e. 0.1% or 0.5% Triton X-100, 0.5 mM C 12 E 8 , 10 mM C 8 E 5 , 0.1% Tween 20, 0.5% CHAPS, 1% n-octyl glucoside) with 0.1 mM BS 3 or DSG for 5 min at room temperature. After quenching with 50 mM Tris-Cl, pH 7.6, for 15 min, samples were frozen on dry ice, lyophilized, and separated by SDS-PAGE, and the protein bands were visualized using Coomassie Blue R-250.
Immunoprecipitations, SDS-PAGE, and Western Blots-GA733 was immunoprecipitated from supernatants of cell lysates using GA733 mAb-Sepharose. The supernatants were first precleared for 1 h with a bovine serum albumin-Sepharose resin, and proteins were adsorbed to 25 l of GA733 mAb-Sepharose for 1 h at 4°C with gentle rotation. Precipitates were washed five times with 400 l of cold lysis buffer, and bound proteins were eluted with 200 l of 0.5% SDS using 0.22-m filter microcentrifuge tubes (Millipore, Bedford, MA). Bound proteins (20 l) were boiled for 2 min in 1% Laemmli sample buffer and separated by 6% Tris-Tricine SDS-PAGE under nonreducing and reducing conditions according to Schä gger and Von Jagow (39).
Western blotting was performed after electroblotting proteins to polyvinylidene difluoride Immobilon-P membranes (Millipore, Bedford, MA), using a Trans-Blot cell (Bio-Rad, Hercules, CA) at 250 mA for 2 h, as described by Mozdzanowski et al. (40). Membranes were blocked overnight with 5% nonfat dry milk and probed either with GA733 mAb or GA733 pAb, followed by an appropriate second step antibody and detection was performed using either the 5-bromo-4-chloro-3-indolyl phosphate-nitro blue tetrazolium (Promega, Madison, WI) or ECL chemiluminescence reagents (Amersham Pharmacia Biotech, Buckinghamshire, United Kingdom).
Flow Cytometry Analysis-FACS analysis was performed as described previously (41). Briefly, 10 6 cells in 1 ml per condition in PBS containing 1% BSA (PBS-B) were incubated for 20 min on ice with 20 g of GA733 mAb. After two washes with 5 ml of PBS-B, a rabbit antimouse secondary antibody coupled to FITC (Calbiochem, San Diego, CA) was added to the cells for another 20 min on ice (dilution 1:40). Excess antibody was removed by washing three times with PBS-B. Cells were analyzed on an EPICS XL flow cytometer (Coulter Corp., Hialeah, FL) using forward scatter and side scatter to exclude dead cells. For the negative control condition, the cells were labeled with the secondary antibody-FITC only. Data analysis was performed using the WinMDI software available on the web from the Scripps Institute.
Cell Aggregation Assay-Cells were detached either with EDTA containing 0.01% trypsin (for Caco-2) or with EDTA alone (for Colo-205) as described above for subcultivation. The aggregation assay was carried out in 24-well plates according to Litvinov et al. (3) with minor modifications. To prevent cell adhesion to the bottom of the wells, 1 ml of 1% agarose was poured in each well and allowed to solidify. Typically, 5 ϫ 10 5 cells in 2 ml of HCMF buffer (Hanks' buffer containing 100 mM Hepes, 1% bovine serum albumin, 100 g/ml DNase I, and 0.35 g/liter NaHCO 3 ) without Ca 2ϩ and Mg 2ϩ were placed in each well and incubated in a rotating platform (100 rpm) at 37°C and 5% CO 2 , for different time periods. For inhibition studies, recombinant GA733 proteins were added at time zero to the single cell suspension at different concentrations. The cells were allowed to aggregate for 2 h and 500-l samples were analyzed in a Coulter Counter "CC" (Coulter Corp.) to determine the number of particles. The extent of aggregation was represented by the degree of aggregation (D) calculated according to Shimoyama et where N t is the number of remaining particles at the time point t ϭ 2 h, and N 0 is the initial number of particles corresponding to the total number of cells at time zero.

Purification and Characterization of Recombinant
GA733-EC and GA733-FL-Recombinant proteins were expressed and purified as described under "Materials and Methods". Both recombinant proteins exhibited moderate covalent cross-linking to dimers and higher oligomers during expression in insect cells and affinity purification, presumably resulting from transglutaminases in the insect cells. Fig. 1A shows a Coomassie Blue stained gel of GA733-EC after immunoaffinity chromatography and after HPLC gel filtration. The retention time of GA733-EC (MW ϳ 28 kDa) in the HPLC column falls between those of protein standards with molecular weights of 44 and 17 kDa ( Fig. 2A), suggesting that GA733-EC is possibly a monomer. After HPLC gel filtration, the minor amount of cross-linked dimers was efficiently separated from the uncrosslinked GA733-EC as shown in Fig. 1A.
Affinity purification of GA733-FL was more problematic and several procedures gave poor yields and resulted in more extensive cross-linking of the protein (up to 50% as determined by SDS-PAGE) with the presence of several low and high molecular weight contaminants. Extensive washes of protein bound to the mAb column using buffer with 0.05% Triton X-100 substantially reduced the amount of most contaminants. However, a 220-kDa band, identified as myosin, was difficult to eliminate totally (Fig. 1B). The amount of cross-linked protein could be reduced to about 5% (Fig. 1B) by using 100 mM Tris buffer with 10 mM EDTA, 10 mM EGTA throughout the affinity purification as described under "Materials and Methods". HPLC gel filtration of GA733-FL in the presence of 0.5 mM C 12 E 8 lead to a high protein recovery (Fig. 1B) compared with the use of other detergents (10 mM C 8 E 5 , 0.1% Triton X-100, 0.1% Tween 20, 1% n-octyl glucoside) (data not shown). As shown in Fig. 1B, myosin and minor low molecular weight contaminants were efficiently separated from GA733-FL on the HPLC gel filtration column, whereas cross-linked and uncross-linked GA733-FL coeluted.
MALDI mass analysis of GA733-EC after HPLC gel filtration chromatography is shown in Fig. 3A. Two broad peaks were observed, with average molecular masses of about 28,331 and 29,348 Da. Since the sequence molecular weight of the extracellular domain is 27,372 Da after removal of the signal peptide, these results suggest that the molecule is heterogeneously glycosylated, resulting in two major populations with about 960 and 1,977 Da carbohydrate, with substantial mass heterogeneity within these peaks as reflected by their shape. Attempts to obtain GA733-FL mass spectra with affinity-purified, detergent-free protein or protein after HPLC gel filtration in 0.5 mM C 12 E 8 were not successful. The data shown in Fig. 3B was obtained using affinity-purified protein concentrated in the presence of 10 mM C 8 E 5 and shows two peaks with average masses of 34,291 and 35,962 Da. Comparison of these masses with the expected sequence mass for GA733-FL (32,675 Da) suggests heterogeneous glycosylation, which contributes ϳ1,616 and 3,287 Da to the mass, although some of this mass difference might be contributed by tightly bound detergent or lipid molecules, or other uncharacterized post-translational modifications.
Sedimentation Equilibrium Analyses of GA733 Recombinant Proteins-To systematically evaluate the oligomeric state of GA733-EC, sedimentation equilibrium experiments were performed at 4°C and 30°C, and the concentration versus radius data was fitted with various models using nonlinear regression (38). At both temperatures, the data were described well by a single ideal species model; representative 30°C data is shown in Fig. 4A. The estimated molecular mass obtained by the fitting program was 28,705 Da, in excellent agreement with the average molecular mass of the two-glycosylated species observed using mass spectrometry (28,839 Da). The data clearly demonstrates that the extracellular domain is monomeric up to a concentration of at least 6 mg/ml. Sedimentation equilibrium analyses of GA733-FL were initially performed on samples in the presence of 10 mM C 8 E 5 . This detergent is a useful alternative to C 12 E 8 for sedimentation equilibrium as it has a partial specific volume of 0.993 cm 3 /g (43), eliminating the need for density correction of the solvent with D 2 O (44). Affinity purified samples could not be further purified by HPLC gel filtration due to poor sample recovery in this detergent, and therefore the affinity purified protein was used directly for sedimentation equilibrium analyses. Table I summarizes the analyses of GA733-FL using different conditions. Various models were used to fit the sedimentation equilibrium data. For all six experiments performed in C 8 E 5 , the data was fitted well with models where the predominant species was a dimer. However, in four experiments there was also a small proportion of a large molecular weight species (n ϭ 8 to 40) that was not in chemical equilibrium with the dimer and apparently represented irreversible aggregates. The presence of this large species interfered with unambiguous identification of other species that may have been present. However, in one experiment with no detectable aggregate, the data was fitted well by a reversible monomer-dimer-tetramer model, with estimated K d 's of 98 nM for the monomer-dimer association and 49 M for the dimer-tetramer association.
In the presence of 0.5 mM C 12 E 8 , the small proportion of aggregates and minor contaminants in the affinity-purified GA733-FL could be removed by HPLC gel filtration. In contrast to the experiments on the affinity purified protein, three sedimentation equilibrium experiments with GA733-FL further purified by HPLC gel filtration, consistently fitted well to models containing only dimer and tetramer species, as shown in Fig. 4B and 4C. Models that included monomer or oligomers other than tetramer (e.g. trimer, hexamer, octamer) gave significantly poorer fits as judged by the size of the variance and randomness of the residuals. When the six data sets shown in Fig. 4B and 4C (representing three different loading concentrations of protein and two rotor speeds) were fitted simultaneously, the data was fitted well by a model describing a reversible dimer-tetramer association, with an estimated K d of 10 M. When the data sets were fitted individually, allowing a different equilibrium constant for each set, the randomness of the residuals improved slightly, and K d 's ranged between 5 M and 14 M, but the improvement in the variance was not significant at a 67% confidence level. Thus the data is consistent with a predominantly reversible association of GA733-FL dimers to tetramers with a K d ϳ10 M. Two other experiments on different preparations of GA733-FL were also fitted well by the same model and returned a K d for the dimer-tetramer association of ϳ7 M. The slight improvement in the fit with individual equilibrium constants suggests that a very small amount of one or more species is unable to associate or dissociate. This is most likely due to the small proportion of covalently cross-linked dimer present in most samples of GA733-FL. The association of GA733-FL monomers to dimers appears to be very strong, with no detectable dissociation observed in all samples in C 12 E 8 , and very little dissociation observed in C 8 E 5 with K d ϳ 98 nM. The absence of detectable monomer in the C 12 E 8 experiments is consistent with a K d Ͻ 10 nM for the monomer-dimer association. The slightly weaker association in C 8 E 5 (K d ϳ 98 nM) could be a detergent effect, but is more likely due to the presence of contaminating low molecular weight proteins in these less pure samples.
Chemical Cross-linking of GA733 Recombinant Proteins in Solution-The oligomeric states of GA733 recombinant proteins, GA733-EC and GA733-FL, in the presence of various detergents were evaluated using chemical cross-linking. Fig. 5 shows a representative cross-linking experiment of GA733 recombinant proteins in the presence of 0.5% Triton X-100. GA733-FL was affinity-purified and eluted in the presence of 0.5% Triton X-100, as described under "Materials and Methods", to minimize artifactual cross-linking during purification. Triton X-100 was added to the GA733-EC preparations after purification to have consistent conditions for both proteins. Both membrane permeable (DSG) and membrane impermeable (BS 3 ) cross-linkers were evaluated (see intact cell experiment below) and an optimal concentration of 0.1 mM was determined in preliminary experiments. As seen in Fig. 5, cross-linking of GA733-FL produces predominantly dimers with additional bands at the trimers and tetramers positions. On the other hand, no cross-linking of GA733-EC was observed under similar conditions, confirming that it is a monomer in both the presence and absence of detergents. Similar results to those shown for Triton X-100 were obtained using BS 3  GA733 Oligomeric State in Intact Colon Carcinoma Cells-To characterize the molecular organization of GA733 in intact cell membranes, chemical cross-linkers (BS 3 and DSG) were added to colon carcinoma cells either on monolayers or in single cell suspension. Two human colon carcinoma cell lines expressing high levels of GA733 were evaluated: 1) a homogeneous clone of Caco-2 epithelial cells (C2BBe1; passages 45-76), which forms tight and polarized monolayers, with an apical brush border morphologically comparable to that of the human colon (45,46) and; 2) the colorectal adenocarcinoma cell line Colo-205 (47), which is characterized by reduced cell-cell and cell-substrate adhesion. When cultured in tissue culture flasks, Colo-205 cells exhibit a minimally adherent morphology, form small aggregates, and a substantial proportion of the cells grows in suspension.
Treatment of intact Caco-2 cells with the membrane-impermeable BS 3 cross-linker at different concentrations, followed by protein extraction, immunoprecipitation, and Western blots (Fig. 6, A and B) resulted in appearance of multiple crosslinked species with a major band at ϳ80 kDa in GA733 mAb immunoprecipitates. When Caco-2 cells in monolayer are treated with high cross-linker concentrations (Fig. 6A), a 160-kDa band corresponding to GA733 tetramers is observed. When single cell suspensions are cross-linked using the same conditions (Fig. 6B), a strong dimer band is still observed but the tetramer band is substantially reduced. Quantification of tetramer/dimer ratios using densitometry showed a ratio of 0.4 versus 0.1 for cells in monolayer and single cell suspensions, respectively, when 0.5 mM BS 3 was used. Similarly, tetramer/ dimer ratios were 0.8 versus 0.2 at 2 mM BS 3 for cells in monolayer and single cell suspension, respectively. The noncross-linked controls (Cont) only showed the monomeric GA733 band (Fig. 6). Similar cross-linking results were obtained for cells in monolayers and cell suspensions using the membranepermeable DSG cross-linker (data not shown), suggesting that GA733 was not cross-linked to cytoplasmic proteins under these conditions.
Similarly, Colo-205 colon carcinoma cells were cross-linked either in situ, in a tissue culture flask (Fig. 6C) or in single cell suspensions (Fig. 6D). Addition of BS 3 to nondetached cells resulted in a decrease in the intensity of the GA733 monomer band in a concentration-dependent manner and the appearance of one major band with the molecular weight of a GA733 dimer (Fig. 6C). GA733 pAb also recognized minor bands with approximate molecular sizes of 120 and 160 kDa, presumably corresponding to GA733 trimers and tetramers. Compared with adherent Caco-2 cells, the ratio of tetramers to dimers for Colo-205 was very low, i.e. 0.07 at 0.5 mM BS 3 and 0.12 at 2 mM BS 3 . When single cell suspensions were cross-linked using the same conditions, dimers but not tetramers were observed (Fig.  6D). Cross-linking was also performed using DSG at the same concentrations and produced similar results both in detached and nondetached cells (data not shown).
GA733 in polarized monolayers of Caco-2 was less extensively cross-linked compared with Colo-205 (compare Fig. 6A with 6C), even when high concentrations of cross-linkers were used. The decreased cross-linking efficiency on Caco-2 cells is likely due to reduced accessibility of the reagent to GA733 protein complexes in these cells. GA733 is known to be predominantly present at the cell-cell boundaries of epithelial cells and carcinoma cells (lateral membranes) and absent from the exposed apical surface (3,48,49), which is consistent with the observed reduced accessibility of the cross-linkers to GA733 molecules in Caco-2 cells compared with Colo-205 cells in these experiments.
Trypsin and EDTA Treatments Have No Effect on GA733 Expression at the Surface of Colon Carcinoma Cells-Before testing the ability of GA733 recombinant proteins to inhibit cell-cell aggregation, we evaluated whether EDTA or trypsin/ EDTA treatments had any effect on GA733 expression at the cell surface. Nose et al. (50) showed that trypsin/EDTA treatment, but not trypsin/CaCl 2 treatment affected the membrane expression of cadherins, which were temporarily removed from the cell surface most likely by internalization. Furthermore, Litvinov et al. (3) reported that GA733 was internalized in the absence of calcium on some human mammary carcinoma cells. The three data sets were fitted simultaneously using the nonlinear regression program NONLIN as described under "Materials and Methods." The data were fitted with a model describing monomers. The raw data (circles) and the global fit of an ideal single species model (lines) are shown. In the upper panels, the residuals of the fitted curve to the data points for the three protein concentrations, from highest to lowest (top to bottom) are shown. Sedimentation equilibrium data of GA733-FL at 14,000 rpm (B) and 19,000 rpm (C), at 20°C. The six data sets were globally fitted with a model describing high affinity dimers with no detectable dissociation to monomers and much weaker self-association to tetramers (K d ϭ 10 M), using the nonlinear regression program NONLIN. The lower panels show the concentration versus radius data for three loading concentrations of GA733-FL (circles). The solid lines represent the calculated fit. The upper panels show the residuals of the fitted curves to the data points for the three protein concentrations, from highest to lowest (top to bottom).

TABLE I Sedimentation equilibrium experiments performed with GA733-FL in the presence of C 8 E 5 and C 12 E 8
Different preparations of GA733-FL were used for sedimentation equilibrium experiments. Affinity-purified protein was used for experiments conducted using C 8 E 5 , whereas experiments using C 12 E 8 utilized samples after further purification by HPLC gel filtration in the presence of C 12 E 8 . The experiments were performed at different rotor speeds ranging from 15,000 to 23,000 rpm at 20°C. Various models were used to fit the sedimentation equilibrium data. The model that best fit the data is indicated. The dissociation constants (K d ) for different interactions, when detected, are represented. For all six experiments performed in C 8 E 5 , the data were fitted well with models where the predominant species was a dimer. However, in four experiments there was also a small proportion of a large molecular weight species, presumably irreversible protein aggregates (agg., where n is the average number of oligomers), which was not in chemical equilibrium with the dimer. Experiments performed in the presence of C 12 E 8 consistently fit well to models containing only dimer and tetramer species.

Role of GA733-2 Oligomerization in Cell-Cell Adhesion
To test whether total cellular GA733 levels were affected by trypsin and/or EDTA treatments, single cell suspensions of Caco-2 cells (dissociated with trypsin/EDTA) and Colo-205 cells (dissociated with EDTA alone) were lysed as described under "Materials and Methods" and GA733 cellular protein levels were compared with the corresponding cell line lysed in monolayers (without dissociation) by Western blot using GA733 mAb. As shown in Fig. 7A, quantitative Western blot reveals a single major GA733 band at the expected molecular mass (ϳ40 kDa), the intensity of which is not appreciably affected by trypsin/EDTA (Caco-2) or EDTA (Colo-205) treatments.
To further confirm that the GA733 molecules detected by Western blots are on the surface of detached colon carcinoma cell lines, flow cytometry analysis using GA733 mAb was performed. Both dissociated Caco-2 and Colo-205 cells were highly and homogeneously positive for GA733 expression (Fig. 7B). These results are consistent with data obtained with L cells transfected with GA733 (3), where the authors showed that both trypsin/EDTA and trypsin/CaCl 2 treatments are not able to remove GA733 molecules expressed at the surface of these transfectants.
GA733-FL Inhibits Cell Aggregation, Whereas GA733-EC Has No Effect-Recent work from our group showed that GA733-FL was active and GA733-EC was inactive in solidphase binding assays, where the recombinant protein was immobilized on nitrocellulose-coated plates (12). Although its validity was previously demonstrated for GA733 (12) and other adhesion molecules (51), the solid-phase binding assay is not fully representative of adhesion of two living cells. Furthermore, a truncation mutant of GA733 lacking the 26-amino acid cytoplasmic tail was not able to mediate aggregation when transfected into L cells, although adhesion of this transfectant to solid-phase-adsorbed GA733 remained unaffected (48).
Therefore, cell-cell aggregation assays were performed to further confirm the biological activity of the GA733 recombinant proteins analyzed above. Specifically, the inhibitory effect of GA733 recombinant proteins on cell-cell aggregation was correlated with their oligomeric states using Caco-2 and Colo-205 cells. Cell-cell aggregation is observed as early as 30 min, and the size of the cell aggregates increases progressively with time. An optimal aggregation time of 2 h was chosen for counting particles and was used for most experiments.
The photographs in Fig. 8A, show representative fields of Caco-2 cell aggregation under different conditions. In the absence of added protein, the cells were extensively aggregated after 2 h (Fig. 8A, b). Aggregation was greatly reduced in the presence of GA733-FL (c) but was unaffected by addition of GA733-EC (d). Similar results were observed when Colo-205 cells were tested (data not shown). As shown in Fig. 8B, addition of affinity purified GA733-FL to the cells inhibits aggregation, whereas GA733-EC has no effect, indicating that dimers, but not monomers, are capable of blocking cell-cell adhesion. Fig. 8C shows the concentration-dependent effects of GA733-FL and GA733-EC on the inhibition of Caco-2 cell aggregation. Even at concentrations of ϳ1 M, GA733-EC does not significantly inhibit cell aggregation, whereas 50% inhibition by GA733-FL occurred at less than 0.5 M. Cell-cell adhesion is a complex and poorly understood process involving multiple CAM systems. In addition, other proteins that may interact with GA733, including cytoskeletal proteins, may affect GA733-mediated cell-cell adhesion in vivo. Nonetheless, the inhibition of cell-cell adhesion by GA733-FL in the low M range as shown in Fig. 8C is consistent with the K d ϳ10 M measured for the dimer-tetramer association using the analytical ultracentrifuge. DISCUSSION Intercellular interactions mediated by cell surface CAMs are known to be involved in a wide variety of dynamic processes, including cell movement, proliferation, and differentiation. The control of these different processes plays critical roles in embryogenesis, wound healing, maintenance of normal tissue morphogenesis, and tumor progression. GA733 is a transmembrane glycoprotein that mediates homotypic Ca 2ϩ -independent cell-cell adhesion. GA733 targeting has shown promising results in the immunotherapy of colon cancer (21,22), and recent data demonstrated an inhibitory effect of GA733 on tumor invasion in vitro (12). Yet, the precise function of GA733-mediated cell-cell adhesion, its role in normal epithelial cells, and its involvement in colon cancer are poorly understood.
In the present study, we explored the oligomeric state of GA733 and the mechanism of GA733-mediated cell-cell adhesion. Our data show: (i) recombinant GA733-EC is a monomer in solution and is inactive in cell-cell aggregation assays; (ii) recombinant GA733-FL forms high affinity dimers (K d Ͻ 10 nM) and moderate affinity tetramers (K d ϳ 10 M) in solution and exhibits inhibitory activity in cell aggregation assays in the low micromolar range; and (iii) GA733 in human colon carcinoma cells exists primarily as noncovalent cis-dimers in single cell suspensions. These results demonstrate that GA733 dimerization is essential for GA733-mediated cell-cell adhesion. The monomeric nature of GA733-EC indicates that the cytoplasmic and/or the transmembrane domain of GA733 are needed for this cis-dimer formation. Further studies with truncated recombinant GA733 proteins will be needed to map the specific regions of these domains responsible for controlling dimer formation.
The extent of cross-linking to dimers, in both Caco-2 and Colo-205 colon carcinoma cells, was similar for adherent monolayers versus single cell suspensions. This indicates that GA733 dimerization occurs by cis-interactions within the membranes of individual cells, rather than trans-interactions between GA733 monomers present in two different cells. Other CAMs exhibit similar interactions, e.g. the carcinoembryonic antigen was found to form noncovalent dimers (52), and there are a growing number of other CAMs that require dimer formation for ligand binding. Dimerization of PECAM-1, E-cadherin, VCAM-1, and ICAM-1 seems to be important for their adhesion function and may represent an important regulatory mechanism of signal transduction pathways (26 -28, 53, 54). The small amount of GA733 tetramers detected when Caco-2 cells were cross-linked in monolayers are most likely due to the head-to-head association between GA733 dimers on opposing cells, because GA733 is a homotypic CAM (see model in Fig. 9). The dimer-tetramer association is a moderate affinity interaction (K d ϳ 10 M), which is consistent with reversible intercellular adhesion.
Minor bands at the trimer and tetramer position are observed when Caco-2 cells are cross-linked in suspension, which could result from the fast cell-cell aggregation kinetics of Caco-2 cells during cross-linking. Alternatively, these bands could represent minor cross-linking of GA733 to other proteins. The failure to detect large amounts of GA733 tetramers in Colo-205 colon carcinoma cells cross-linked while attached to the culture flask is not surprising. Despite their normal expression of E-cadherin and catenins as well as GA733, Colo-205 cells do not form monolayers in culture, but grow as small aggregates or dispersed cells with minimal cell-cell contacts, and hence the majority of the surface of individual cells is not engaged in adhesion with other cells. Furthermore, these cells exhibit poor adhesion to extracellular matrix proteins, do not form tight interactions with each other, and dissociate easily in PBS during the cross-linking experiments (data not shown).
Other minor bands are also detected at different positions after cross-linking of Caco-2 and Colo-205 cells, especially a band between the monomer and the dimer at ϳ60 kDa. This band may be the result of one of the following: (i) gel shift of GA733 due to internal cross-linking and side chain modifications; (ii) coincidental random cross-linking with other proteins due to the high density of GA733 at the surface of colon carcinoma cells (about 10 6 copies of GA733 per cell (55)); or (iii) specific interactions with other membrane proteins.
The oligomerization of cell surface receptors plays a central role in the regulation of cellular functions by modulating signal transduction pathways. GA733-1 (Trop-2) that shares 50% amino acid identity with GA733 was shown to be involved in cell signaling. A recent study (14) showed that monoclonal antibodies against GA733-1 induce calcium fluxes in MCF-7 and OvCa-432 cells. Another study (56) demonstrated that GA733-1 is phosphorylated on serine 303, and that protein kinase C is involved in phosphorylating this protein in vitro. The exact mechanism by which homotypic GA733 engagement might activate signal transduction pathways in epithelial cells and the nature of these pathways remain intriguing questions for future studies.
The inhibitory effects of GA733 expression on growth and invasion of transfected mouse tumor cells are surprising in view of the beneficial effects of mAbs against GA733 for treating colorectal carcinoma patients. It is tempting to speculate on how these apparently conflicting observations might be reconciled. In addition to tumor cell elimination by the immune system, mainly via the antibody-dependent cellular cytotoxicity pathway, mAbs to GA733 in clinical therapy may have two other beneficial effects: (i) inhibit cancer cell dissociation from the primary tumor by increasing cell-cell association; (ii) decrease cell growth by mimicking GA733 self-ligation and activating signal transduction pathways leading to inhibition of cell proliferation and/or induction of apoptosis. We are currently generating a soluble fusion protein, a covalent GA733 dimer containing the GA733-EC fused to an Fc fragment of an immunoglobulin (EC-Fc), to further study the role of dimerization in homotypic cell-cell adhesion and signal transduction pathways mediated by GA733.
In addition to colon cancer, an increase of GA733 expression was observed in cervical cancer lesions (11) and a very recent paper reported increased GA733 levels in lung cancer (57). Understanding the precise function of GA733 in normal epithelial cells and its role in the biology of tumors should contribute to the development of new strategies for cancer treatment or a minima substantially improve the existing GA733based therapy of colorectal tumors.