Aberrant glycosylation of E-cadherin enhances cell-cell binding to suppress metastasis.

Introduction of the β1-4 N-acetylglucosaminyltransferase (GnT-III) gene was reported to suppress metastasis in highly metastatic B16-hm murine melanoma cells (Yoshimura, M., Nishikawa, A., Ihara, Y., Taniguchi, S., and Taniguchi, N. (1995) Proc. Natl. Acad. Sci. U. S. A. 92, 8754-8758). In this study, the effect of GnT-III gene transfer on E-cadherin was studied, since E-cadherin acts as a suppressor of metastasis. E-cadherin expression at cell-cell contacts of B16-hm cells expressing high GnT-III activity was greater than controls without affecting transcription. Lectin blotting showed that E-cadherin from GnT-III transfectants was glycosylated by ectopically expressed GnT-III. The glycosylated E-cadherin exhibited the delayed turnover and the decreased release from cell surface, as compared with the native E-cadherin, resulting in the elevated expression at the cell-cell border of GnT-III transfectants. Furthermore, cell-cell aggregation was enhanced in GnT-III transfectants, indicating that the glycosylated E-cadherin is biologically functional. These results suggest that the glycosylated E-cadherin contributes to the suppression of metastasis by the introduction of GnT-III gene into melanoma cells.

Introduction of the ␤1-4 N-acetylglucosaminyltransferase (GnT-III) gene was reported to suppress metastasis in highly metastatic B16-hm murine melanoma cells (Yoshimura, M., Nishikawa, A., Ihara, Y., Taniguchi, S., and Taniguchi, N. (1995) Proc. Natl. Acad. Sci. U. S. A. 92, 8754-8758). In this study, the effect of GnT-III gene transfer on E-cadherin was studied, since E-cadherin acts as a suppressor of metastasis. E-cadherin expression at cell-cell contacts of B16-hm cells expressing high GnT-III activity was greater than controls without affecting transcription. Lectin blotting showed that E-cadherin from GnT-III transfectants was glycosylated by ectopically expressed GnT-III. The glycosylated E-cadherin exhibited the delayed turnover and the decreased release from cell surface, as compared with the native E-cadherin, resulting in the elevated expression at the cellcell border of GnT-III transfectants. Furthermore, cellcell aggregation was enhanced in GnT-III transfectants, indicating that the glycosylated E-cadherin is biologically functional. These results suggest that the glycosylated E-cadherin contributes to the suppression of metastasis by the introduction of GnT-III gene into melanoma cells.
The malignant phenotype, including metastatic potential, has been reported to be associated with the ␤1-6 branches of N-oligosaccharides, the product of ␤1-6 N-acetylglucosaminyltransferase (GnT-V, EC 2.4.1.155) 1 in ras-transformed rat fibroblasts, rat mammary carcinoma cells, rat lymphoma cell line (1,2), and human colon cancer cells (3). The composition of ␤1-6 branches is also reported to correlate with the progression and staging of human breast and colon neoplasia (4). Ectopical expression of GnT-V gene in mouse lung epithelial cell line resulted in loss of contact inhibition, progressive growth of tumors, and increase of lung metastasis after subcutaneous inoculation into nude mice (5). This report also supports that the ␤1-6 structure plays a causative role in tumorigenicity and metastasis.
As shown in Fig. 1, both ␤1-4 N-acetylglucosaminyltransferase (GnT-III, EC 2.4.1.144) and GnT-V use the biantennary structure of N-oligosaccharides as a substrate, and substrate specificity studies showed that GnT-V is not able to form any further triantennary structure in the presence of a bisecting GlcNAc residue (6,7). On the basis of nuclear magnetic resonance data (8), the biantennary structure of a core mannose was twisted in the presence of bisecting GlcNAc. This conformational change rendered the substrate inaccessible to GnT-V to form the ␤1-6 structure, which raised the possibility that ␤1-6 branch formation could be suppressed by the introduction of the GnT-III gene, thereby resulting in changes in the metastatic potential and of malignant phenotypes. Actually, in our investigation, the decrease of ␤1-6 structure by the competition between the endogeneous GnT-V and ectopically expressed GnT-III led to the suppression of lung metastasis by mouse melanoma cells (9). The metastatic process is a complex of many biological events, and many molecules are responsible for metastasis (10). As almost all of these molecules are glycoproteins, the alteration of enzyme activity, which catalyzes the formation of the sugar chain such as the introduction of the GnT-V gene (5) or the GnT-III gene (9), may change the structure and composition of N-glycans and may affect the physiological function of the molecules responsible for metastasis.
The present investigation was undertaken to examine whether the reduced metastatic potential of B16 murine melanoma cells expressing ectopic GnT-III was due to the altered biological function of E-cadherin that mediates homotypic cellcell adhesion, since the E-cadherin expression correlates inversely to metastatic phenotype in many cancer cells (11)(12)(13). We found that the E-cadherin, which was glycosylated by GnT-III, was increased due to the prolonged turnover rate and the decreased release from cell surface. Furthermore, cell aggregation was increased in GnT-III transfectants, indicating that the glycosylated E-cadherin participates in the suppression of metastasis by GnT-III gene transfer.
Cell Lines-B16-hm displaying highly metastatic potential, GnT-IIItransfected B16-hm cells displaying low metastatic potential, and mock-transfected B16-hm cells, which had been already established (9), were used in this investigation. They were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum and antibiotics. Cells in exponential growth and between 4 and 10 passages were used throughout this investigation, and the results described here were independent of passage number.
Immunofluorescence Microscopy-B16-hm cells and GnT-III transfectants were seeded on 3.5-cm dishes with a coverglass on the bottom of the dish. Cells were washed once with phosphate-buffered saline * This work was supported in part by grants-in-aid for cancer research and scientific research on priority areas from the Ministry of Education, Science, and Culture of Japan and the Uehara Memorial Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
(PBS), fixed in 4% paraformaldehyde in PBS for 20 min at room temperature, and permeabilized with Tris-HCl-buffered saline, pH 7.4 (TBS) containing 0.5% Triton X-100 for 15 min at room temperature. After the nonspecific binding sites were blocked by incubation for 1 h with 3% bovine serum albumin in TBS, cells were incubated with ECCD-2 (10 g/ml) in TBS containing 5 mM CaCl 2 for 4 h at 4°C and then stained with FITC-conjugated anti-rat IgG antibody (10 g/ml) for 1 h at 4°C. The fluorescence of the cells was visualized with a photomicroscope with epifluorescence (IIRS, Olympus, Japan) and photographed with a 20-s exposure. Cells were also observed and photographed with a microscope camera (ELWD 0.3, Nikon, Japan).
Lectin Blot and Immunoprecipitation-Cells were pelleted and lysed in the lysis buffer (150 mM NaCl, 20 mM Tris, pH 7.4) containing 0.5% SDS, 0.5% Triton X-100, 2 mM CaCl 2 , 10% (w/v) glycerol, 100 g/ml phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, and 1 g/ml aprotinin for 20 min on ice. Following clarification at 15,000 rpm for 20 min at 4°C, the protein concentration was measured using a BCA kit (Pierce). For immunoprecipitation, cell lysates (200 g of protein) were precleared with normal rabbit serum (Vector) and protein G-Sepharose beads (Pharmacia, Uppsala, Sweden) and then incubated with ECCD-2 (10 g/ml). Immune complexes were collected with protein G-Sepharose beads and were released by boiling in Laemmli's sampling buffer, separated by 8% SDS-polyacrylamide gel electrophoresis (PAGE), and electrotransferred onto nitrocellulose membranes (Schleicher and Schuell, Germany). The blots were blocked in TBS with 0.05% Tween 20 containing 3% bovine serum albumin and incubated for 4 h with biotinylated E-PHA (10 g/ml) or probed with ECCD-2 (10 g/ml) followed by biotinylated anti-IgG (10 g/ml). After washing three times with TBS with 0.05% Tween 20, the blots were incubated for 1 h with horseradish peroxidase-avidin complex (Vector) and developed using an ECL detection system (Amersham Corp.) according to the manufacturer's protocols. In using ECCD-2, the probing of the Western blot was performed in the presence of 5 mM CaCl 2 , because Ca 2ϩ was necessary for this antibody to recognize the epitope (12). Signals by E-PHA binding and ECCD-2 binding were judged by densitometric scanning (Shimazu, CS-9000, Tokyo, Japan) and standardized as to the amount of E-cadherin.
Northern Blot-Total cellular RNA was isolated from cell samples by acid guanidine thiocyanate and phenol/chloroform extraction (14). Extracted RNAs (20 g) were electrophoresed on 1.0% formaldehydeagarose gel, which was stained with ethidium bromide to confirm the amount of RNA loaded, and RNAs were transferred onto a Zeta-probe membrane (Bio-Rad). After hybridization of the membrane with a 32 Plabeled mouse E-cadherin full-length cDNA (12) (donated by Dr. Takeichi, Kyoto University, Kyoto, Japan) overnight, the membrane was washed and autoradiographed, as described (12).
Metabolic Labeling and Chase Study-Subconfluent monolayers of B16-hm cells and GnT-III transfectants in 6-well tissue cultures were preincubated for 2 h at 37°C with methionine-free DMEM (Life Technologies, Inc.) with dialyzed 10% fetal calf serum. For pulse-chase studies, L-[ 35 S]methionine and L-[ 35 S]cysteine (Pro-mix TM , Amersham Corp.) were added at the concentration of 50 Ci/ml to the culture media and incubated for 15 min at 37°C for protein labeling. After rinsing five times with cold PBS to remove excess [ 35 S]methionine/ cysteine, cells were incubated at 0, 2, 4, 8, 12, and 24 h in DMEM with 10% non-dialyzed fetal calf serum. At the various incubation times, supernatant was collected from each well before the well was washed five times with PBS. Then, E-cadherin molecules, which were immunoprecipitated from cell lysates and supernatants, were subjected to 10% SDS-PAGE in the same procedure as described above. After electrophoresis, the gel was completely dried using a Gel Dryer (Bio-Rad, model 583) and autoradiographed for 2 days using x-ray film.
Cell Aggregation Assay-The effect of ECCD-2 on the cell aggregation assay was evaluated, as described (15). In brief, B16-hm cells or GnT-III transfectants were washed once with PBS and incubated in the assay buffer (150 mM NaCl, 0.6 mM Na 2 HPO 4 , 10 mM glucose, and 10 mM Hepes, pH 7.4) containing 0.01% trypsin and 2 mM CaCl 2 for 30 min at 37°C. Then, cells (3 ϫ 10 4 cells) were washed twice with assay buffer, resuspended in 300 l of assay buffer containing 2 mM CaCl 2 , and transferred into each well of 24-well plates (Costar). An antibody to rat IgG as a control or the ECCD-2 was added to each well at 1, 2, 5, 10, 15, and 30 g/ml in triplicate. After a 60-min incubation with occasional shaking at 37°C, an equal volume of 8% paraformaldehyde in PBS was added to each well, and the plates were centrifuged for fixation. Photos were taken at random under a phase contrast microscope (Nikon) to count single cells or cell aggregates (10 or more cells). After counting cells (Ͼ200), cell aggregation was determined by R cont Ϫ R e , where R cont and R e are the ratio of cells in aggregate per total cell count in the presence of anti-rat IgG and ECCD-2, respectively. Cell aggregation assay was repeated three times. The R cont values were always above 0.90 and were not significantly different at any concentration of the antibody.

Accumulation of E-cadherin at the Cell-Cell Border in GnT-
III Transfectants-Two clones with elevated GnT-III activity designated as B16-hm-III-1 and -2, and one clone with no detectable GnT-III activity designated as B16-hm-neo-1 were used in this investigation as positive transfectants and a control transfectant, respectively. Morphologically, the B16-hm cells and the control transfectant appeared fibroblastoid with loose cell-cell contacts, whereas positive transfectants were epithelioid and proliferated in compact organization ( Fig. 2A).
E-cadherin expression was examined by indirect immunofluorescence (Fig. 2B). In B16-hm cells and a control transfectant, E-cadherin was weakly expressed at the cell-cell contacts. Positive transfectants, however, showed intense fluorescence with condensation at the cell-cell contacts, indicating elevated expression of E-cadherin at cell-cell contacts of positive transfectants.
Increased Glycosylated E-cadherin in GnT-III Transfectants-E-PHA lectin was used to analyze the alterations of carbohydrate structures of E-cadherin, since E-PHA has a high affinity for bisecting GlcNAc structure (16). Western blotting studies showed that the expression of E-cadherin was more increased in positive transfectants than in B16-hm cells and control transfectants (Fig. 3A, upper panel), which was consistent with the results of immunofluorescence microscopy. The signal for E-PHA binding to the positive cells was detected as the bands at 125 kDa, corresponding to immunoprecipitated E-cadherin, while these bands were almost undetectable in the B16-hm cells and a control transfectant (Fig. 3A, lower panel).
The relative E-PHA binding ratio (%) normalized by E-cadherin signal density was as follows: B16-hm, 0.9; B16-hmneo-1, 1.5; B16-hm-III-1, 13.4; and B16-hm-III-2, 16.9; this indicated that the E-cadherin in the positive transfectants was glycosylated by GnT-III and few, if any, bisected oligosaccharides were attached to E-cadherin in B16-hm cells and a control transfectant. As compared with the B16-hm cells and the control transfectant, E-cadherin transcripts were not increased in positive transfectants, suggesting that the raised E-cadherin expression in GnT-III transfectants was not due to an increase of the transcriptional level (Fig. 3B).
Prolonged Turnover and Decreased Release of E-cadherin in GnT-III Transfectants-The turnover rate of E-cadherin was examined by chase studies. In B16-hm cells, immunoprecipi- tated surface E-cadherin showed the most intense signal at 4 h and then declined to an undetectable level at 12 h (Fig. 4A,  lower panel). Immunoprecipitated surface E-cadherin from B16-hm-III-1 cells (upper panel) showed the most intensified signal at 8 h, and the signals of the immunoprecipitated Ecadherin were maintained at the detectable level during all the chase time periods. Released E-cadherin was detected as a band of 104 kDa in the supernatants from both B16-hm cells and positive transfectants, which was less in size than the surface E-cadherin (Fig. 4B). In the supernatant of B16-hm cells, released E-cadherin was first detected at 2 h, gradually increased to show the most intensified signal at 8 h, and maintained to be at the detectable level during the incubation (upper panel). In B16-hm-III-1 cells, released E-cadherin was undetectable until 4 h during the incubation. The signals of released E-cadherin at 4 and 8 h were faint and much weaker than that detected in the supernatants of B16-hm cells and declined to the undetectable level at 12 and 24 h (lower panel). Turnover and release of E-cadherin of a control transfectant were similar to that of B16-hm cells, showing that transfection procedures did not affect the turnover of E-cadherin (data not shown). Collectively with these results, expression of GnT-III prolonged the turnover of E-cadherin and inhibited the release of Ecadherin from cell surface, which resulted in an increased level and accumulation of E-cadherin molecules at cell-cell contacts.
Increased Cell Aggregation in GnT-III Transfectants-To determine whether the increased E-cadherin expression was involved in the homotypic cell adhesion of B16-hm cells, cell aggregation was assayed using an antibody to E-cadherin that blocks E-cadherin-mediated adhesion (Fig. 5). In B16-hm cells and a control transfectant, more than 90% of cell aggregation was inhibited by 5 g/ml ECCD-2. The inhibitory effect by ECCD-2 was, however, considerably lower in the positive transfectants, showing that the increased E-cadherin expression was associated with increased cell aggregation and that raised E-cadherin was biologically functional. DISCUSSION In many glycoproteins, the composition of oligosaccharides, especially N-linked oligosaccharides, exists in the structural components, which contribute to the folding, stability, and biological function of the molecules (17-19). Therefore, the activity and stability of the glycoprotein could be affected by subtle  , ϫ 200). B, distribution of E-cadherin on B16-hm cells and GnT-III transfectants, detected by indirect immunofluorescence microscopy. Cells seeded on the coverglass were fixed, blocked in 3% bovine serum albumin, and stained with ECCD-2. After incubation with secondary FITC-conjugated anti-mouse IgG, the stained cells were observed under an epifluorescence microscope. All the photographs were taken by 20-s exposure and photographed (original magnification, ϫ 200). changes in oligosaccharide components. For example, an Noligosaccharide on CD2 stabilizes the protein structure for the interaction with CD58 (19). Replication of the hepatitis B virus is suppressed by GnT-III gene transfer (20). Non-glycosylated growth hormone, which is secreted both from the apical and basolateral sides of Madin-Darby canine kidney cells, is secreted from the apical side when glycosylated (21). In this study, we examined the altered biological function of N-glycosylated E-cadherin by means of the introduction of the GnT-III gene (22) into mouse melanoma B16-hm cells. As the exogeneous expression of GnT-III affects the processing of glycoproteins with complex-type oligosaccharides, the GnT-III gene transfection gives a global effect on many glycoproteins, including E-cadherin.
Deduced from the amino acid sequence, mouse E-cadherin has four putative N-glycosylation sites (12). Lectin blot analysis revealed that immunoprecipitated E-cadherin from both B16-hm cells and a control transfectant had undetectable bisecting GlcNAc residues. In contrast, E-cadherin from the positive transfectant did bind E-PHA, indicating bisecting GlcNAc residues were added to E-cadherin molecules through the glycosylation by GnT-III in the positive transfectants. Immunoblot analysis revealed that the amount of E-cadherin on the cell surface was elevated in the positive transfectants, whereas Northern analysis showed that the expression level of E-cadherin transcripts in positive transfectants was not significantly increased in comparison with that of B16-hm cells and a control transfectant, suggesting that ectopically expressed GnT-III may have a significant influence to E-cadherin not at the transcription level but at the post-translational level.
Next, we quantitated the turnover of E-cadherin by means of pulse-chase studies. Positive transfectants showed the delayed turnover of E-cadherin and the elevated E-cadherin amount on the cell surface, as determined by a pulse-chase study. Moreover, release of E-cadherin from the cell surface was decreased in positive transfectants, suggesting that the increased bisecting GlcNAc structure is likely to be suppressive to the proteolytic cleavage of E-cadherin on the cell surface. The prolonged turnover and inhibited release led to the elevation and accumulation of E-cadherin at cell-cell contact regions. As the expression level of E-cadherin on the cell surface has been reported to inversely correlate with cell invasion and metastatic potential (11), homotypic adhesion is likely to be enhanced in the positive transfectants, which lead to the suppression of metastasis. As the introduction of N-glycosylation sites is shown to modify protein sorting (19), the turnover of E-cadherin also may be affected by the aberrant glycosylation by GnT-III.
To examine further whether the increased E-cadherin at the cell-cell border of positive transfectants was biologically functional as a homotypic adhesion molecule, we performed cell aggregation assay and evaluated adhesive function of glycosylated E-cadherin. Compared with B16-hm cells and control transfectants, cell-cell binding was more enhanced in positive transfectants, as judged by cell aggregation assay. Thus, this indicates that cell-cell aggregation was quantitatively enhanced by the increased expression of E-cadherin, although there also remains a possibility that the glycosylation may give the qualitative effect on the adhesive function of E-cadherin. We conclude that the suppression of metastasis in B16-hm cells expressing GnT-III (9) is at least partly due to the increased level of glycosylated E-cadherin.
Taken together, the results in this study indicate that the introduction of GnT-III gene raised the surface expression of glycosylated E-cadherin due to the delay of its turnover rate and the inhibition of its release from cell surface to result in the enhanced cell aggregation in GnT-III transfectants. These findings are supposed to lead to the suppression of metastasis in mice, as previously reported (9). We believe that our study is the first demonstration that aberrant glycosylation of an adhesion molecule through transferred gene for a glycosyltransferase could modify cell adhesion to suppress metastasis. This indicates that glycosylation is one of the important events in the process of metastasis. FIG. 5. Cell aggregation assay involving an inhibitory antibody to E-cadherin. Cell-cell aggregation was assayed at the various concentrations of ECCD-2, and cell-cell aggregation was calculated as described under "Experimental Procedures." The assay was performed in triplicate, and the results are represented as the mean of three independent experiments.