Cell-Cell Interaction-dependent Regulation of N-Acetylglucosaminyltransferase III and the Bisected N-Glycans in GE11 Epithelial Cells

Changes in oligosaccharide structures are associated with numerous physiological and pathological events. In this study, the effects of cell-cell interactions on N-linked oligosaccharides (N-glycans) were investigated in GE11 epithelial cells. N-glycans were purified from whole cell lysates by hydrazinolysis and then detected by high performance liquid chromatography and mass spectrometry. Interestingly, the population of the bisecting GlcNAc-containing N-glycans, the formation of which is catalyzed by N-acetylglucosaminyltransferase III (GnT-III), was substantially increased in cells cultured under dense conditions compared with those cultured under sparse conditions. The expression levels and activities of GnT-III but not other glycosyltransferases, such as GnT-V and α1,6-fucosyltransferase, were also consistently increased in these cells. However, this was not observed in mouse embryonic fibroblasts or MDA-MB231 cells, in which E-cadherin is deficient. In contrast, perturbation of E-cadherin-mediated adhesion by treatment with EDTA or a neutralizing anti-E-cadherin antibody abolished the up-regulation of expression of GnT-III. Furthermore, we observed the significant increase in GnT-III activity under dense growth conditions after restoration of the expression of E-cadherin in MDA-MB231 cells. Our data together indicate that a E-cadherin-dependent pathway plays a critical role in regulation of GnT-III expression. Given the importance of GnT-III and the dynamic regulation of cell-cell interaction during tissue development and homeostasis, the changes in GnT-III expression presumably contribute to intracellular signaling transduction during such processes.

It has been well known that sugar chains have various effects on the functional aspects of glycoproteins and play important roles in cell dif-ferentiation, adhesion, and proliferation (1,2). The sugar chains of glycoproteins are produced via catalysis by various glycosyltransferases. As a result, a specific structure is determined by the expression pattern of these enzymes. The expression of glycosyltransferases is regulated not only by physiological conditions, such as developmental stages, but by pathological conditions, such as rheumatoid arthritis and tumorigenesis (3,4). The increasing body of evidence suggests that the structures of the glycan components, such as sialyl Lewis antigens (5,6), mucin-type O-glycans (7), or N-glycans (3,8), which are controlled by glycosyltransferases, greatly contribute to cell adhesion, cell invasion, and cancer metastasis. A malignant phenotype has been reported to be highly associated with N-linked oligosaccharides (N-glycan) containing ␤1,6branching, which is catalyzed by N-acetylglucosaminyltransferase V (GnT-V) 4 by transfer of GlcNAc from UDP-GlcNAc to a ␤1,6 mannose in N-glycans (8). It has also been reported that GnT-V activities and ␤1,6-branched N-glycan levels are increased in highly metastatic tumor cell lines (9,10). Conversely, cancer metastasis is greatly suppressed in GnT-V knock-out mice (11). These results suggest that GnT-V is strongly associated with cancer metastasis.
In contrast to GnT-V, N-acetylglucosaminyltransferase III (GnT-III) has been found to play an important role in the suppression of metastasis. GnT-III catalyzes the attachment of a GlcNAc to a core mannose of N-glycan via a ␤1,4-linkage to form the bisecting GlcNAc structure (Fig. 4A), which does not serve as substrate for GnT-V to form a ␤1,6branch (12,13). Therefore, GnT-III has been proposed to be antagonistic to GnT-V. In fact, lung metastasis of highly metastatic mouse melanoma B16 cells, in which GnT-V activities are much higher than those in other cell types, are significantly suppressed by overexpression of the GnT-III gene (14). Interestingly, an enhancement in cell-cell adhesion through the prolonged turnover of E-cadherin on the cell surface was reported in these GnT-III transfectants (15). On the other hand, the overexpression of GnT-V results in decreased N-cadherin clustering on the cell surface and therefore the down-regulation of cadherin-associated cell-cell adhesion (16). These results strongly suggest that the remodeling of N-glycan structures modified by glycosyltransferases function to modulate cell adhesion and cancer metastasis.
The cadherins are a superfamily of adhesion molecules that function in cell recognition, tissue morphogenesis, and tumor suppression (17). E-cadherin is the prototypical member of these calcium-dependent cell adhesion molecules and mediates homophilic cell-cell adhesion. The loss of E-cadherin expression or function in epithelial carcinoma has long been thought to be a primary reason for the disruption of tight epithelial cell-cell contacts and the release of invasive tumor cells from the primary tumor (18). E-cadherin serves as a widely acting suppressor of invasion and growth of epithelial cancers, and its functional elimination represents a key step in the acquisition of the invasive phenotype for many tumors. Indeed, E-cadherin is found in epithelia, where it promotes the formation of tight cell-cell associations known as adherens junctions. In contrast, N-cadherin is found primarily in neural tissues and fibroblasts, where it is thought to mediate a less stable and more dynamic form of cell-cell adhesion (19). Therefore, cell-cell adhesion is believed to be both temporally and spatially regulated during development. The issue of whether or not cell-cell interaction affects N-glycan structures in a reverse manner is a subject of debate. In this study, we investigated this issue using a dense or sparse culture model, and found that bisected N-glycans were dramatically induced in cells under conditions of a dense culture through an E-cadherin involved pathway.

EXPERIMENTAL PROCEDURES
Cell Cultures-Epithelial GE11 cells, derived from ␤1 integrin knockout embryonic stem cells (20), were maintained at 37°C in Dulbecco's modified Eagle's medium, supplemented with 10% fetal bovine serum, 100 units/ml penicillin G, and 0.1 mg/ml streptomycin under a humidified atmosphere containing 5% CO 2 . Mouse embryonic fibroblasts of Fut8 ϩ/ϩ and Fut8 Ϫ/Ϫ were previously established in our laboratory (21). Madin-Darby canine kidney cells and MDA-MB231 cells were maintained under the same conditions. Cells at 5 ϫ 10 6 and 5 ϫ 10 5 were plated on 15-cm dishes for a dense culture and a sparse culture, respectively, followed by incubation for 4 days. Cell densities were confirmed by phase-contrast observations, as shown in Fig. 1 Expression Plasmid and DNA Transfection-Human E-cadherin cDNA was first cloned, and then p-E-cadherin-ECFP-N for E-cadherin fused to ECFP was prepared by inserting the cDNA fragment encoding human E-cadherin into the 5Ј end of the cDNA fragment encoding ECPF in pECFP-N1 (Invitrogen). All of the constructs used here were confirmed by sequencing. p-E-cadherin-ECFP-N1 or pECFP-N1 was transfected into MDA-MB231 cells by using Lipofectamine 2000 (Invitrogen). The stable transfectants were selected by G418 (Nacalai Tesque, Inc.).
Preparation of Pyridylaminated (PA) N-Linked Oligosaccharide from Cells-Cells were cultured and harvested at the indicated times. After washing twice with phosphate-buffered saline, the cells were lysed by sonication and then lyophilized. The N-glycans were liberated from glycoproteins by hydrazinolysis at 100°C for 10 h. After reacetylation and lyophilization, the reducing ends of the N-glycans were labeled with 2-aminopyridine (Nacalai Tesque). Excess 2-aminopyridine was removed with a cellulose cartridge glycan preparation kit (Takara Bio Inc., Japan).
Analysis of N-Glycans by the Reversed and Normal Phase HPLC-The PA N-glycans prepared from cells were analyzed on a reversed phase HPLC system (Shimazu Co., Japan) using an ODS80-TM column (4.6 ϫ 150 mm; Tosoh). Elution was performed at a flow rate of 1.0 ml/min at 55°C using 20 mM ammonium acetate buffer (pH 4.0) as solvent A and the same buffer containing 1% 1-butanol as solvent B. The column was pre-equilibrated with 4% solvent B, and after injection of a sample, the PA N-glycans were separated by means of a linear gradient of 4 -25% of solvent B for 60 min. The eluents were then subjected to a normal phase HPLC for further analysis. An Amide 80 TM column (4.6 ϫ 250 mm; Tosoh) was used, and elution was performed at a flow rate of 1.0 ml/min at 40°C using 0.1% trifluoroacetic acid in 80% acetonitrile as solvent A and 0.1% trifluoroacetic acid in MilliQ water as solvent B. The column was pre-equilibrated with 0% solvent B, and after injection of a sample, the PA N-glycans were separated by a linear gradient of 0 -100% solvent B for 50 min. PA N-glycans were detected by a fluorescence detector (Shimazu) at excitation and emission wavelengths of 320 and 400 nm, respectively.
Glycosidase Digestion for N-Glycans-The PA N-glycans were sequentially digested at 37°C with sialidase (Nacalai Tesque), ␤-galactosidase, and fucosidase (Seikagaku Co., Japan) in 0.1 M citrate-phosphate buffer (pH 5.0). After the digestion, the mixture was boiled for 3 min, and the supernatant, after centrifugation at 15,000 rpm for 10 min, was analyzed by reversed phase HPLC, as described above.
Analysis of N-Glycan Structures by Mass Spectrometry-N-Glycans treated with dialidase were purified by reverse and normal phase HPLC and then subjected to mass spectrometric (MS) analysis. Mass measurements were carried out using a matrix-assisted laser desorption/ionization (MALDI) time-of-flight (TOF) mass spectrometer equipped with a pulsed ion extraction system (Reflex IV; Bruker-Daltonik GmbH, Germany). All spectra were obtained using the reflectron mode with a delayed extraction of 200 ns and were the result of the signal averaging of 200 laser shots. For sample preparation, a 0.5-l volume of the matrix solution (2,5-dihydroxybenzoic acid; 10 g/liter in 30% ethanol) was deposited on the stainless steel target plate and allowed to dry. Then 0.5 l of appropriately diluted analyte solution was used to cover the matrix on the target plate and allowed to dry. Finally, 0.5 l of matrix solution was added to the deposited sample/matrix mixture on the target plate and allowed to dry. Structural analyses of oligosaccharides were achieved by a multistage tandem mass spectral matching strategy using a MALDI-quadrupole ion trap-TOF mass spectrometer (AXIMA-QIT; Shimadzu, Japan) (22).
Glycosyltransferase Activity Assays-Cells were grown to 30, 80, and 100% confluence, respectively, and then harvested. After washing with phosphate-buffered saline, the cells were lysed by sonication. The protein concentrations of cell lysates were determined by means of a BCA protein assay kit, and the same amounts of proteins were used as the source for the GnT-III and GnT-V activity assays as described previously (23). For the Fut8 activity assay, a cell lysate was incubated with a biantennary N-glycan labeled with pyridylaminoethylsuccinamyl as an acceptor substrate and reaction mixture as described previously (24). The substrate we used was GlcNAc␤1-2Man␣1-6(GlcNAc␤1-2Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc-PA, which was produced according to Hase et al. (25). For each assay, 100 pmol of the substrate (in 100 l of total reaction solution) were utilized. The specific activity was expressed as pmol of N-acetylglucosamine transferred/h/mg of proteins (23).
PCR for mRNA Expression Analysis-Total RNA was prepared with TRIzol (Invitrogen) and real time PCR was performed in a Smart Cycler II (Takara Bio). The sequences of the sense and antisense primers of mouse GnT-III were 5Ј-CGAGGACACCACCGAGTATT-3Ј and 5Ј-ACTCGTGGTTGACGTTGATG-3Ј, respectively. The ␤-actin mRNA was used as a control in PCR runs. The product sizes obtained by the PCR were 350 bp for mouse GnT-III and 700 bp for ␤-actin. Quantitative PCR was performed by monitoring the real time increase in fluorescence of the SYBR Green dye on an ABI PRISM 7000 sequence detector system (Applied Biosystems) according to the instructions from the manufacturer. Gene-specific primers were designed by Takara Bio, and the sequences of sense and antisense primers of mouse GnT-III were TCAACGCCATCAACATCAAC and CCTTCGAGTACATC-CGCCAC, respectively. The mean number of cycles to the threshold (C T ) of fluorescence detection was calculated for each sample, and the results were normalized to the mean C T of glyceraldehyde-3-phosphate dehydrogenase for each sample tested. Relative expression levels of GnT-III are expressed as the -fold increase in the dense culture. For detection of mRNA in MDA-MB231 cells, the primers of human GnT-III were used: 5Ј-AAGACCCTGTCCTAT-3Ј for sense and 5Ј-GTTG-GCCCCCTCAGG-3Ј for antisense (26). The product size obtained by the PCR was 345 bp for human GnT-III. This detection was also confirmed by PCR using plasmid DNA containing human and rat GnT-III as controls.
Immunoprecipitation, Immunoblotting, and Lectin Blot Analysis-Cells cultured under different conditions as indicated were washed with phosphate-buffered saline and then lysed with lysis buffer (10 mM Tris-HCl, 1% Triton-X, 150 mM NaCl, aprotinin, leupeptin, and 1 mM phenylmethylsulfonyl fluoride). Insoluble materials were removed by centrifugation at 15,000 rpm for 10 min at 4°C. The supernatant (2.5 mg of protein) was incubated with anti E-cadherin monoclonal antibody (3 g/ml) (BD Biosciences) for 1 h at 4°C, and protein G beads (20 l in 50% slurry) were then added, followed by incubation for another 2 h at 4°C with a rotator. After washing three times with lysis buffer, the immunoprecipitates were subjected to 7.5% SDS-PAGE, and the separated proteins were transferred to a nitrocellulose membrane. The membrane was incubated with a lectin for a lectin blot analysis or an antibody for immunoblot analysis.
Perturbation of E-cadherin-mediated Interactions by EDTA or a Neutralized Anti-E-cadherin Antibody-GE11 cells (1 ϫ 10 7 ) were plated on 15-cm dishes and cultured in the presence or absence of EDTA at a final concentration of 1 or 0.5 mM. After incubation for 2 days, the cells were harvested for analysis of GnT-III activity, mRNA expression, and N-glycan structures. On the other hand, GE11 cells (4 ϫ 10 5 ) were plated on 3.5-cm dishes and incubated for 2 days with or without a neutralized anti-E-cadherin antibody (ECCD-1 clone; Takara Bio) at a final concentration of 50 g/ml, and the activities of GnT-III were then examined.

RESULTS
In order to investigate the effects of cell-cell interactions on N-glycan biosynthesis, we compared N-glycans obtained from cells cultured under dense or sparse culture conditions. GE11 cells were selected as a model cell to reduce the possible influence of ␤1 integrin-mediated cell adhesion, since GE11 cells are derived from ␤1 integrin knock-out embryonic stem cells (20). As shown in Fig. 1, cells cultured under sparse conditions have adequate space available for cell spreading. In contrast, cells cultured under dense conditions have no space for full spreading, resulting in a reduction in cell size, as confirmed by phasecontrast microscopy.
Different Patterns of N-Glycans Released from Cells Cultured under Dense and Sparse Cultures-N-Glycans were liberated from glycoproteins by hydrazinolysis, and the reducing ends of N-glycans were then labeled with 2-aminopyridine. The N-glycans purified with cellulose-Sepharose were digested with sialidase and analyzed by reversed phase HPLC as described under "Experimental Procedures." Interestingly, a peak eluting at 67 min was dramatically increased in the dense culture, compared with that in the sparse culture (Fig. 2, A and B). We digested the mixture with some glycosidases to confirm whether it corresponded to an N-glycan. The peak eluted at 67 min was shifted to 60 and 39 min after additional digestion with ␤-galactosidase (Fig. 2C) or ␤-galactosidase plus fucosidase (Fig. 2D), suggesting that the peak corresponds to a complex type of sugar chain. To elucidate the structure of the peaks, the elution pattern of those peaks with several standard N-glycans was compared by reversed phase HPLC. As shown in Fig. 3B, the peaks that eluted at 67, 60, and 39 min correspond to bisected biantennary N-glycans, G(Gn)GF-bi-PA, Gn(Gn)GnF-bi-PA, and Gn(Gn)Gn-bi-PA, respectively, suggesting that bisecting GlcNAc-containing oligosaccharide levels are increased in cells cultured under dense conditions. The peak that eluted at 64 min might not be an N-glycan, because digestion with glycosidases could not make it shift.
Analysis of N-Glycan Profiles by Mass Spectrometry-We recently developed a rapid and accurate approach to the high throughput analysis of oligosaccharides utilizing an observational multistage tandem mass spectral (MS n ) library of a large variety of the structurally defined oligosaccharides (22). To obtain oligosaccharide profiles in this study, the N-glycans were purified by reverse and normal phase HPLC. As shown in Fig. 3, mass spectra of desialylated N-glycans were obtained from GE11 cells cultured under the two different conditions of cell density. The N-glycan profiles of GE11 cells cultured under high and low cell density conditions are clearly different. Carbohydrate compositions of the nine major peaks in the MS spectra ( Fig. 3A) are summarized in Table 1. The peak intensities of (M ϩ Na) ϩ for neutral oligosaccharides reflect the relative concentrations of the constituents of a mixture over the mass range from m/z 1000 to 2400 for MALDI-TOF MS (27). In high density culture conditions, the intensities of peaks 2, 3, and 6, which contain two HexNAc residues on the N-glycan core structure, are decreased, whereas peaks 4, 5, and 7 with three HexNAc residues are increased, compared with those for the low cell density culture condition, suggesting that the oligosaccharides were extended by one HexNAc in GE11 cells cultured under dense conditions. Structural analyses of these peaks with multistage tandem mass spectral matching were performed (22), and the peaks (1, 2, 5, 7, 8, and 9) were successfully identified as shown in Fig. 3B. The data reveal that peaks 5 and 7, which were increased under high cell density conditions, corresponded to peaks 1 and 2 with an added bisecting GlcNAc, respectively. This accounts for the most significant difference in the N-glycans between the low and high cell densities of GE11 cells, GGF-bi-PA of peak 2, and G(Gn)GF-bi-PA of peak 7. The three peaks (3, 4, and 6) could not be identified by the spectral matching strategy, which might be due to their being a mixture of isomeric structures. Some peaks (m/z 2304.97, 2347.00, and 2369.97) observed from m/z 2300 to m/z 2400 in the spectrum of high density GE11 cells might have arisen from nonoligosaccharide matter, since these could not be matched by a GlycoMod search (28). These MS results correspond to reversed phase HPLC results  or dense (B) cultures, as described above, were digested with sialidase and then subjected to reversed phase HPLC. The oligosaccharide mixtures from cells under a dense culture were sequentially digested with sialidase ␤-galactosidase (C) and sialidase ␤-galactosidase plus fucosidase (D) in 0.1 M citrate-phosphate buffer (pH 5.0). After the digestion, the mixtures were analyzed by the reversed phase HPLC. The arrow indicates a major increased peak in densely cultured cells, which shifts after sequential digestion. The star indicates that the peak is a contaminant, which could not be digested by endo-and exoglycosidases. described above, strongly suggesting that bisecting GlcNAc is increased as cell-cell interactions increase in GE11 cells.
In order to determine whether the change in N-glycans with an increase in cell-cell interactions is specific to epidermal cells like GE11 cells or not, we composed the same cell culture models using embryonic fibroblasts of Fut8 ϩ/ϩ and Fut8 Ϫ/Ϫ , derived from wild type and Fut8 knock-out mice (21). No significant differences in N-glycan profiles were found in Fut8 ϩ/ϩ (Fig. 3C) as well as in Fut8 Ϫ/Ϫ cells (Fig. 3D) between low and high density culture conditions. This result indicates that an increase in bisected N-glycans with increasing cell-cell adhesion may be specific for epidermal cells and not for fibroblasts. It is also noteworthy that the N-glycans released from Fut8 Ϫ/Ϫ cells do not contain any core fucose, compared with those from wild type cells, demonstrating that Fut8 is a unique glycosyltransferase that transfers a core fucose to N-glycans.
Enhancements in Cell Densities are Concomitant with Increased Activities of GnT-III but Not Fut8 and GnT-V-To investigate the underlying mechanism for the enhanced expression of bisected N-glycans, GnT-III activities were examined. Interestingly, GnT-III activities were increased in a cell density-dependent manner (Fig. 4B). The GnT-III activities in cells cultured under dense conditions were ϳ8-fold higher than those in cells cultured under sparse condition. In previous studies, we found that forskolin, an adenylyl cyclase activator, enhanced GnT-III at the transcriptional level (29), whereas the elevation of GnT-III expression was increased in the M phase of the cell cycle at the posttranslational level but not at the transcriptional level (30). Here, the mRNA expression levels of GnT-III were up-regulated in cells cultured under dense conditions. Reverse transcription-PCR and real time PCR clearly showed that the expression level of GnT-III mRNA in densely cultured cells is higher than that in sparsely cultured cells (Fig. 4, D and  E). However, the activities of GnT-V (data not shown) and Fut8 (Fig.  4C) were not affected by these culture conditions, suggesting that cellcell interactions specifically regulate GnT-III activation.
It is well known that the expression levels of E-cadherin can be upregulated by cell-cell adhesion in most epithelial cells (31). The loss or decrease of E-cadherin expression is associated with the invasion and metastasis of cancers. The E-cadherin-catenin complex is devoted to maintaining cell formation and cell-cell adhesion. The E-cadherin intracellular domain binds ␤-catenin, which in turn binds ␣-catenin. To confirm that this also occurs in GE11 cells, we examined E-cadherin expression and E-cadherin-associated complex. As expected, the expressions of E-cadherin and its complex formation, as confirmed by the use of anti ␣and anti ␤-catenin antibodies, were consistently enhanced in the cells cultured under dense conditions (Fig. 5, A and B). In cell lysates, the expression of E-cadherin, but not ␤-catenin and ERK1/2 as control proteins, was dramatically increased by a dense cul-ture, suggesting that the GE11 cells, as epithelia cells, respond normally to cell-cell interaction.
Overexpression of GnT-III in B16 mouse melanoma cells resulted in an enhancement in cell-cell adhesion and a prolonged turnover of E-cadherin on the cell surface as described previously (14,15). To examine whether E-cadherin is targeted for modification by endogenous GnT-III in the GE11 cells cultured under dense conditions, E-cadherin was immunoprecipitated and analyzed by E 4 -PHA lectin, which specifically recognizes bisected N-glycans. As shown in Fig. 5C, the intensity of E 4 -PHA staining was significantly increased in densely cultured cells, demonstrating that E-cadherin is one of target proteins for endogenous GnT-III.
Enhanced GnT-III expression levels are regulated by cell-cell interactions in an E-cadherin-mediated manner. E-cadherin is a member of the classic cadherin family of single pass transmembrane glycoproteins that mediate Ca 2ϩ -dependent cell-cell interactions. We hypothesized that cell density-dependent increases in GnT-III activities were due to a signal mediated by E-cadherin-dependent cell-cell interaction. To explore the mechanism of the regulation of GnT-III expression, we blocked the function of E-cadherin by adding EDTA, a metal chelator, to the culture medium to disrupt cell-cell interaction through chelation of the extracellular calcium required for E-cadherin homophilic binding between cells. Following EDTA treatment, the cells had no cell-cell interaction, but the cells remained attached to the culture dishes. Importantly, treatment with EDTA in a final concentration at 1 mM resulted in a reduction in GnT-III expression, as confirmed by PCR and Western blot analysis (Fig. 6, A and B) as well as bisecting GlcNAc containing biantennary N-glycans indicated by an arrow (Fig. 6C). EDTA treatment resulted in the down-regulation of GnT-III and E-cadherin expression but not ␤-catenin as a control. Furthermore, treatment with the neutralized anti-mouse E-cadherin antibody (ECCD-1), which blocks E-cadherin-dependent cell-cell contact (32,33), reduced GnT-III activities to 35% of the control (Fig. 6D). Taken together, these results suggest that an E-cadherin-dependent pathway was involved in the regulation of GnT-III expression in cells under dense culture conditions. To confirm this hypothesis, we also analyzed the elution pattern of N-glycans released from MDA-MB231 breast cancer cells, which are devoid of E-cadherin, and found that there were no apparent differences between high and low density cell culture conditions (Fig. 7A). The expression and activity of endogenous GnT-III in MDA-MB231 cells are shown in Fig. 7, B and C. mRNA of endogenous GnT-III was confirmed by PCR using rat and human GnT-III plasmid as negative and positive control, respectively (Fig. 7B), and the activity of endogenous GnT-III was measured by HPLC (Fig. 7C). To more clearly show the importance of E-cadherin in regulation of

Up-regulation of GnT-III by Cell Adhesion
GnT-III expression, we further rescued the expression of E-cadherin in MDA-MB231 cells by transfection of pE-cadherin-ECFP-N1, which has been reported to induce a calcium-dependent cell-cell adhesion (34). The positive transfectants were confirmed by SDS-PAGE, followed by blotting with anti-E-cadherin antibody (data not shown). Then GnT-III activities in parent cells, mock, and E-cadherin transfectants were investigated under dense growth condition. As shown in Fig. 7C, parent and mock cells showed a slight increase in GnT-III activity accompanied by dense growth. However, in the E-cadherin transfectants, significant increase can be observed. These data strongly supported the conclusion above that E-cadherin plays a crucial role in regulation GnT-III under dense growing conditions. As mentioned in reference to Fig. 3, the change in bisected N-glycans was not observed in fibroblasts, which are devoid of E-cadherin expression. Therefore, we asked whether or not the change is a phenomenon common to epidermal cell lines during cell-cell interactions. We analyzed the elution pattern of N-glycans released from Madin-Darby canine kidney cells (data not shown), a typical epidermal cell line, and found that does not always appear to be the case. However, GnT-III activities regardless of cell density were not detected in Madin-Darby canine kidney cells, indicating that the increase in bisected N-glycans in a dense culture requires basal levels of not only E-cadherin but GnT-III as well.

DISCUSSION
Evidence is presented that GnT-III and its product, bisected N-glycans, are up-regulated by E-cadherin-mediated cell-cell interaction. The disruption of E-cadherin-mediated cell-cell interactions by treatment with EDTA or a neutralizing antibody blocks this regulation. Consistent with that, the change of N-glycans is not observed in E-cadherin-

. Enhanced expression of GnT-III in cells cultured under a dense culture.
A, reaction pathway for biosynthesis of the bisecting GlcNAc by GnT-III; GE11 cells were grown to different cell confluences as indicated, harvested, and lysed by sonication. The same amounts of proteins in cell lysates were used as the enzymatic source for GnT-III (B) and Fut8 (C) as described under "Experimental Procedures." S, substrate; P, product. D, reverse transcription-PCR analysis of GnT-III expression. Total RNA was prepared with TRIzol. The ␤-actin mRNA was used as a control. E, quantitative PCR was performed by monitoring in real time the increase in fluorescence of the SYBR Green dye on an ABI PRISM 7000 sequence detector system. The mean number of cycles to the threshold (C T ) of fluorescence detection was calculated for each sample, and the results were normalized to the mean C T of glyceraldehyde-3-phosphate dehydrogenase for each sample tested. Relative expression levels of the target genes are expressed as the -fold increase, compared with the expression level in a sparse culture, which was taken as 1. deficient cells. Moreover, reintroduction of E-cadherin into E-cadherindeficient cells rescued the significant increase in GnT-III activity under dense growth condition. These results clearly demonstrate that E-cadherin is the major mediator involved in the regulation of GnT-III by cell-cell interactions. Given the important biological functions of GnT-III, as reported previously (35,36), the present study may provide a new insight into molecular mechanism of the relationship between cell-cell interactions, normal development, and cancer metastasis.
Regulation of cadherin-mediated adhesion and associated adherens junctions is thought to underlie the dynamics of the adhesive interac-tion between cells, which are regulated during tissue development and homeostasis as well as during the progression of tumor cells. During normal development, E-cadherin-mediated cell adhesion is vital to the gastrulation movement that reorganizes embryonic germ layers as well as to the development of other migratory cell types such as the neural crest (37). In fact, expression of E-cadherin could be greatly regulated by epithelia cell-cell interaction (31) (this study). On the other hand, the disruption of E-cadherin-mediated cell adhesion appears to be a central event in the transition of noninvasive to invasive carcinomas. Some studies have focused on identifying and characterizing transcriptional  repressors of E-cadherin expression in epithelial tumor cells. The most prominent factors to arise from these studies, including the related factors Slug, Snail, SIP1, and Twist, are best known for their roles in early embryogenesis and tumor progression (38). However, in an earlier study, we found that E-cadherin-mediated cell-cell adhesion is regulated by post-transcriptional modification with N-glycans. The overexpression of GnT-III in B16 melanoma cells resulted in the suppression of lung metastasis by the up-regulation of cell-cell adhesion through two possible pathways: (i) the delayed turnover of E-cadherin and therefore increased E-cadherin expression levels on the cell surface (15) and (ii) the decreased tyrosine phosphorylation of ␤-catenin (39), which is believed to negatively regulate cell-cell adhesion by the perturbation of E-cadherin-catenin complex formation. Conversely, the overexpression of GnT-V blocks N-cadherin-mediated cell-cell adhesion (16). These regulations of cadherin expression or cell-cell interaction on the cell surface by GnT-III or GnT-V were not at transcriptional levels. Thus, our studies taken together demonstrate that the expression of E-cadherin is closely regulated by not only transcriptional factors but also by post-transcriptional modifications.
Interestingly, E-cadherin-mediated interactions were found to conversely up-regulate GnT-III expression in this study. Such significant and obvious regulation was only observed in epithelial cells that express basal levels of E-cadherin and GnT-III, but not in MDA-MB231 cells, an E-cadherin-deficient cell line, not in Madin-Darby canine kidney, of which GnT-III express is undetectable, and not in fibroblasts, which lack E-cadherin. In addition, E-cadherin is one of target proteins for endogenous GnT-III. Collectively, our results suggest that E-cadherin and GnT-III could be regulated by each other in a positive feedback. It has been reported that the association of E-cadherin with ␤1 integrin has an important role in cell-cell adhesion and intracellular signaling (40,41). In order to exclude the possible influence of ␤1 integrin on the regulation of GnT-III, GE11 cells were utilized as our experimental model. We also investigated the activity of GnT-III in MDA-MB231 cells, which have endogenous ␤1 integrin (42) but lack E-cadherin. We found only a slight increase in GnT-III activity in this cell line under dense conditions. However, the introduction of E-cadherin into MDA-MB231 cells resulted in a significant increase in GnT-III (Fig. 7C). Therefore, the absence of ␤1 integrin did not appear to be related to the difference in GnT-III products under different culture conditions. Although E-cadherin played a major role in the regulation in GE11 cells, E-cadherin expression alone seems to be not sufficient for modulating density-dependent increases in GnT-III expression. This was demonstrated not only by the perturbation experiment with the ECCD-1 antibody, but also by the results shown in Fig. 7C. As shown in Fig. 7C, the slight increase in GnT-III activity was also observed in MDA-MB231 cells under dense growth conditions, although such an increase was minute compared with that of E-cadherin transfectants. Therefore, we assumed that besides the major regulator, E-cadherin, some other cellcell contact-mediated molecules might be also involved in the regulation of GnT-III to a degree.
An increasing body of evidence suggests that the remodeling of N-glycosylations by transfection or knock-out as well as the RNA interference of certain glycosyltransferases could specifically modify the functions of some glycoproteins and therefore affect the biological functions of cells (43). So far, GnT-III is usually believed to be a "negative regulator" for receptor tyrosine kinases and adhesive receptors, such as integrins, which synergize with growth factor receptors for transducing cellular signals (44). For example, epidermal growth factor-induced neurite outgrowth through the Ras/mitogen-activated protein kinase activation pathway was completely blocked in GnT-III-transfected PC12 cells (45). Similarly, phosphorylation of the epidermal growth factor receptor was not increased in U373 MG glioma or HuH7 cells that overexpress GnT-III (46). Nerve growth factor-stimulated TrkA receptor phosphorylation and signaling was disrupted due to failure of the TrkA receptor to dimerize in GnT-III-overexpressing PC12 cells (47). On the other hand, introduction of GnT-III also suppresses ␣5␤1 integrin-mediated biological functions: cell spreading, migration, and the phosphorylation of the focal adhesion kinase (48). These observations indicate that the modulation of N-glycan structures by GnT-III could control not only growth factor-mediated signals but cell-extracellular matrix adhesion as well. Considering the up-regulation of GnT-III in the dense culture model in the current study, the GnT-III presumably participates in controlling intracellular signaling during E-cadherinmediated cell-cell adhesions.
For cadherin, recent studies focusing on cellular signaling have shown that several pathways are activated by cadherin-mediated cellcell interactions. Cadherin-mediated adhesions activate the Rho family GTPases, regulate the availability of ␤-catanin for participating in Wnt signaling, and function in receptor tyrosine kinase signaling (49). E-cadherin also forms complexes with integrins, tetraspanin CD151, and protein kinase C, which are believed to be involved in the regulation of a variety of biological events, including actin reorganization (50). It remains to be studied which signal pathway(s) participates in the upregulation of GnT-III by cell-cell interactions. N-Glycosylation of glycoproteins is known to be essential in the progression of the cell cycle and cell proliferation, as indicated by temperature-sensitive N-glycosylation-defective mutant cells (51) and the effects of tunicamycin, an N-glycosylation inhibitor (52). This evidence suggests that N-glycans play a role in the cell cycle in which an involvement by some oligosaccharide structures is possible. The cell cycle is mainly regulated by a variety of cyclins and their inhibitors, which are downstream targets of signals of a number of growth factors, which could be modulated by N-glycosylation as described above. To a certain extent, cells under sparse and dense culture conditions can be interpreted as cells at the proliferation and differentiation maintenance states, respectively. In this study, GnT-III expression was significantly up-regulated by cell-cell interactions, which might be reasonable for maintenance of cell differentiation rather than cell proliferation, since growth factor-mediated activation can be suppressed by the up-regulation of GnT-III. In fact, the results of several studies suggest that E-cadherin can induce a ligand-independent activation of the epidermal growth factor receptor and subsequent activation of Rac1 as well as mitogen-activated protein kinase, which apparently promotes cell migration and proliferation (53). Thus, it is possible to speculate that the up-regulation of GnT-III by cell-cell interaction might neutralize such signals for the maintenance of cell differentiation phenotype.
In summary, this is the first clearly delineated description of the effects of cell-cell interaction on N-glycosylation, which indicate that bisected N-glycans are dramatically induced in cells under dense culture conditions through an E-cadherin-mediated pathway. The regulation upon cell-cell interaction is relatively specific to GnT-III, since other glycosyltransferases, GnT-V and Fut8, do not participate, suggesting that the bisecting GlcNAc, a unique structure, is of critical importance in the regulation of the biosynthesis of N-glycans and plays an important role in cell biological functions. Further studies, to identify the possibility of other target proteins involved in regulation of GnT-III and downstream signaling, could shed some light on the regulation of cell-cell adhesion and the function of bisecting GlcNAc. We believe that such studies might contribute to the development of a remedy for cancer metastasis by using sugar chains.