N-Acetylglucosaminyltransferase III Antagonizes the Effect of N-Acetylglucosaminyltransferase V on α3β1 Integrin-mediated Cell Migration*

N-Acetylglucosaminyltransferase V (GnT-V) catalyzes the addition of β1,6-GlcNAc branching of N-glycans, which contributes to metastasis. N-Acetylglucosaminyltransferase III (GnT-III) catalyzes the formation of a bisecting GlcNAc structure in N-glycans, resulting in the suppression of metastasis. It has long been hypothesized that the suppression of GnT-V product formation by the action of GnT-III would also exist in vivo, which will consequently lead to the inhibition of biological functions of GnT-V. To test this, we draw a comparison among MKN45 cells, which were transfected with GnT-III, GnT-V, or both, respectively. We found that α3β1 integrin-mediated cell migration on laminin 5 was greatly enhanced in the case of GnT-V transfectant. This enhanced cell migration was significantly blocked after the introduction of GnT-III. Consistently, an increase in bisected GlcNAc but a decrease in β1,6-GlcNAc-branched N-glycans on integrin α3 subunit was observed in the double transfectants of GnT-III and GnT-V. Conversely, GnT-III knockdown resulted in increased migration on laminin 5, concomitant with an increase in β1,6-GlcNAc-branched N-glycans on the α3 subunit in CHP134 cells, a human neuroblastoma cell line. Therefore, in this study, the priority of GnT-III for the modification of the α3 subunit may be an explanation for why GnT-III inhibits GnT-V-induced cell migration. Taken together, our results demonstrate for the first time that GnT-III and GnT-V can competitively modify the same target glycoprotein and furthermore positively or negatively regulate its biological functions.

Malignant transformation is accompanied by increased ␤1,6-GlcNAc branching of N-glycans attached to Asn-X-Ser/ Thr sequences in mature glycoproteins (1)(2)(3). N-Acetylglucosaminyltransferase V (GnT-V) 3 catalyzes the addition of ␤1,6-linked GlcNAc (see Fig. 8) and defines this subset of N-glycans (4,5). A relation between GnT-V and cancer metastasis has been reported by Dennis et al. (6) and Yamashita et al. (1). Studies on transplantable tumors in mice indicate that the product of GnT-V directly contributes to the growth of cancer and subsequent metastasis (7,8). On the other hand, somatic tumor cell mutants that are deficient in GnT-V activity produce fewer spontaneous metastases and grow more slowly than wildtype cells (6,9). The suppression of tumor growth and metastasis has been reported in GnT-V-deficient mice (3). Moreover, Partridge et al. (10) reported that GnT-V-modified N-glycans with poly-N-acetyllactosamine, the preferred ligand for galectin-3, on surface receptors oppose their constitutive endocytosis and result in promoting intracellular signaling and consequently cell migration and tumor metastasis. These results indicate that inhibition of GnT-V might be useful in the treatment of malignancies by targeting their roles in metastasis.
N-Acetylglucosaminyltransferase III (GnT-III) participates in the branching of N-glycans (see Fig. 8), catalyzing the formation of a unique sugar chain structure-bisecting GlcNAc (11). GnT-III is generally regarded to be a key glycosyltransferase in the N-glycan biosynthetic pathway, since in vitro the introduction of the bisecting GlcNAc results in the suppression of further processing and the elongation of N-glycans as the result of catalysis by other glycosyltransferases, which are unable to use the bisected oligosaccharide as a substrate (12,13). It is interesting to note that the metastatic capabilities of B16 mouse melanoma cells are down-regulated by introduction of the GnT-III gene (14). E-cadherin, a homophilic type of adhesion molecule (15), is highly associated with the prevention of metastasis (16), and E-cadherin on GnT-III-transfected cell surfaces was found to be resistant to proteolysis, resulting in an extended half-life of turnover (17). Thus, GnT-III, contrary to GnT-V, has long been thought to inhibit cancer metastasis.
Cell-extracellular matrix (ECM) interactions play essential roles during the acquisition of migration and invasive behavior of cells. Cell surface transmembrane glycoprotein-integrin is a major receptor for ECM and connects many biological functions, such as development, control of cell proliferation, protection against apoptosis, and malignant transformation (18). Integrin ␣3␤1, the major laminin 5 (LN5) receptor, is widely distributed in almost all tissues, and it has been proposed to be involved in tumor invasion (19 -21). In some malignant tumors, ␣3␤1 integrin was found to be the most predominant integrin expressed (22) and made an important contribution to pulmonary metastasis (23). On the other hand, the glycosylation of integrins contributes to the tumor metastasis. Guo et al. reported that an increase in ␤1,6-GlcNAc sugar chains of the integrin ␤1 subunit resulted in the stimulation of cell migration (24). Interestingly, it has also been reported that the ␣3␤1 integrin expressed by the metastasis human melanoma cell lines, contained a higher level of ␤1,6-branched structures than that expressed in a nonmetastasis parent cell line (25).
Although it had been assumed that the reaction of GnT-V can be inhibited by the action of GnT-III, as evidenced by substrate specificity studies in vitro, the hypothesis of competition between GnT-III and GnT-V in cell migration and tumor metastasis has not been directly verified so far. In the present study, we examined the functions of ␣3␤1 integrin, which is believed to be highly associated with tumor metastasis, and found that ␣3␤1 integrin can be modified by either GnT-III or GnT-V. Our finding clearly shows that GnT-III inhibits the effects of GnT-V on ␣3␤1 integrin-mediated cell migration by competing with GnT-V for the modification of ␣3 subunit.
Plasmids and Transient Virus Transfection-cDNAs encoding full-length human GnT-III or GnT-III inactive mutant (D317A) were ligated into adenoviral vector, constructed using an adenoviral expression vector kit (Takara Bio). The 3 ϫ 10 5 MKN45 GnT-V transfectants were then infected with 150 l of virus solution (2 ϫ 10 9 plaque-forming units/ml). After a 24-h incubation, the cultured medium was replaced with a fresh medium. 48 h later after infection, cells were subjected to various experiments.
Construction of Small Interfering RNA Vector and Retroviral Infection-Small interfering oligonucleotides specific for GnT-III were designed on the Takara Bio site on the World Wide Web, and the oligonucleotide sequences used in the construction of the small interfering RNA vector were as follows: 5Ј-GATCCGTCAACCACGAGTTCGACCTTCA-AGAGAGGTCGAACTCGTGGTTGACTTTTTTAT-3Ј and 5Ј-CGATAAAAAAGTCAACCACGAGTTCGACCTCTCT-TGAAGGTCGAACTCGTGGTTGACG-3Ј. The oligonucleotides were annealed and then ligated into BamHI/ClaI sites of the pSINsi-hU6 vector (Takara Bio). A retroviral supernatant was obtained by transfection of human embryonic kidney 293 cells using the retrovirus packaging kit Ampho (Takara Bio) according to the manufacturer's protocol. CHP134 cells were infected with the viral supernatant, and the cells were then selected with 500 g/ml G418 for 2-3 weeks. Stable GnT-III knockdown clones were selected and confirmed by GnT-III activity and gene expression. Quantitative real time PCR analyses of GnT-III mRNA expression in these clones were performed with a Smart Cycler II System and the SYBR premix Taq (Takara Bio). Reverse transcription was carried out at 42°C for 10 min, followed by 95°C for 2 min using random primers, followed by PCR for 45 cycles at 95°C for 5 s and 60°C for 20 s with the following primers: 5Ј-GCGTCATCAACGCCAT-CAA-3Ј 5Ј-TGGACTCGCACACCACAAAG-3Ј. Normalization of the data were performed using the glyceraldehyde-3phosphate dehydrogenase mRNA levels.
GnT-III and GnT-V Activity Assay-The activities of GnT-III and GnT-V were assayed as described previously (29,30). Briefly, cell lysates were homogenized in phosphate-buffered saline (PBS) containing protease inhibitors. The supernatant, after removal of the nucleus fraction by centrifugation for 15 min at 900 ϫ g, was used in the assays, which involved high performance liquid chromatography methods using a pyridylaminated biantennary sugar chain as an acceptor substrate. Protein concentrations were determined using a bicinchoninic acid kit (BCA kit) (Pierce) with bovine serum albumin as a standard.
Western Blot and Lectin Blot Analysis-Cell cultures were harvested in lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton, 10 g/ml leupeptin, 10 g/ml aprotinin, 1 mM phenymethylsulfonyl fluoride). Cell lysates were centrifuged at 15,000 ϫ g for 10 min at 4°C, the supernatants were collected, and the protein concentrations were determined using a BCA protein assay kit. Proteins were then immunoprecipitated from the lysates using a combination of 2 g of anti-integrin ␣3 subunit antibody and 15 l of protein G-Sepharose 4 Fast Flow (Amersham Biosciences) for 1 h at 4°C. Immunoprecipitates were suspended in reducing sample buffer, heated to 100°C for 3 min, resolved on 7.5% SDS-PAGE, and electrophoretically transferred to nitrocellulose membranes (Schleicher & Schuell). The blots were then probed with anti-␣3 antibody or biotinylated E 4 -or L 4 -PHA. Immunoreactive bands were visualized using the Vectastain ABC kit (Vector Laboratories, CA) and an ECL kit (Amersham Biosciences). For GnT-III, GnT-V, cell lysate, and actin blotting, an equal amount of cell lysates was subjected to SDS-PAGE and then transferred to nitrocellulose membranes. The membranes were incubated with the corresponding primary antibodies and secondary antibodies for 1 h each, and detection was performed by an ECL kit.
Migration Assay-Transwells (BD Biosciences) were coated with 5 nM recombinant LN5 as described previously (32), 10 g/ml human plasma FN, and collagen I (COL) (Sigma) in PBS by an incubation overnight at 4°C. Serum-starved cells (2 ϫ 10 5 cells/well in 500 l of 5% fetal calf serum medium) were seeded in the upper chamber of the plates. After incubation overnight at 37°C, cells in the upper chamber of the filter were removed with a wet cotton swab. Cells on the lower side of the filter were fixed and stained with 0.5% crystal violet. Each experiment was performed in triplicate, and counting was done in three randomly selected microscopic fields within each well.
Functional Blocking Assay-To identify which integrin is involved in cell migration on LN5, functional blocking antibodies against different types of integrins were individually preincubated with cells for 10 min at 37°C. The preincubated cells were transferred into transwells coated with LN5 and then incubated overnight at 37°C. The migrated cells were then quantified as described above.
Statistical Analysis-Statistical evaluations were performed using Student's t test; differences among experimental groups were considered significant for p Ͻ 0.05. Data were expressed as mean values Ϯ S.D.
Analysis of N-Glycan Structures by Mass Spectrometry (LC/MS n )-Purified ␣3␤1 integrin was applied to SDS-PAGE, and the ␣3 subunit was excised from the gel and then cut into pieces. The gel pieces were destained and dehydrated with 50% acetonitrile. The protein in the gel was reduced and carboxymethylated by the incubation with dithiothreitol and sodium monoiodoacetate (34). N-Glycans were extracted from the gel pieces as reported by Kustar et al. (35) and reduced with NaBH 4 . Half of the extracted oligosaccharides were incubated with ␣-neuraminidase from Arthrobacter ureafaciens in 50 mM phosphate buffer, pH 5.0, at 37°C for 18 h and desalted with Envi-carb (Supelco, Bellefonte, PA). LC/MS and LC/multistage MS (MS n ) was carried out on a quadrupole liner ion trap-Fourier transform ion cyclotron resonance mass spectrometer (FT-ICR MS; Finnigan LTQ FTTM, Thermo Electron Corp., San Jose, CA) connected to a nano-LC system (Paradigm, Michrom BioResource, Inc., Auburn, CA). The eluents were 5 mM ammonium acetate, pH 9.6, 2% CH 3 CN (pump A) and 5 mM ammonium acetate, pH 9.6, 80% CH 3 CN (pump B). The borohydridereduced N-linked oligosaccharides were separated on a Hypercarb (0.1 ϫ 150 mm, Thermo Electron Corp.) with a linear gradient of 5-20% B in 45 min and 20 -50% B in 45 min. A full MS 1 scan (m/z 450 -2000) by FT-ICR MS followed by datadependent MS 2,3 for the most abundant ions was performed in both negative and positive ion modes as previously reported (36).

Overexpression of GnT-V Stimulated ␣3␤1
Integrin-mediated Cell Motility-It has been reported that overexpression of GnT-V in epithelial cells results in a loss of contact inhibition, increased cell motility in athymic nude mice (7), and an enhanced metastasis (8). In this study, experiments were first designed to determine whether GnT-V overexpression could affect cell migration on different ECMs. The extent of haptotaxis toward LN5, FN, and COL, specific ligands for ␣3␤1, ␣5␤1, and ␣1␤1 and ␣2␤1 integrin, respectively, was observed in MKN45 cells transfected with mock, GnT-III, or GnT-V. In the case of the GnT-V transfectants on LN5, the number of transwell cells migrating to the lower surface of the membrane was considerably increased (p ϭ 0.001), the overexpression of GnT-III resulted in a decrease in cell migration on LN5 compared with mock (p ϭ 0.0013) (Fig. 1A). However, the migration of these three types of cells on FN was barely detectable. Although GnT-III transfection resulted in a decreased cell migration on COL compared with mock (p ϭ 0.007), GnT-V transfection failed to induce a significant increase in cell migration on COL (Fig. 1B), suggesting that MKN45 cells may favor LN5 as an ECM for cell migration induced by GnT-V. These results further supported the view that ␣3␤1 integrin, one of the most abundant integrins in epithelial cells, is distinct from other integrins, such as ␣5␤1 integrin, and preferentially pro-motes cell migration (37). Moreover, the cell migration of GnT-V transfectant on LN5 was strongly inhibited by the presence of function-blocking antibodies against integrin ␣3 or/and ␤1 subunit, suggesting that the GnT-V-induced cell migration on LN5 was mainly mediated by ␣3␤1 integrin (Fig. 2). These results indicated that overexpression of GnT-V resulted in an increase in ␣3␤1 integrin-mediated cell motility.
Overexpression of GnT-III Inhibited ␣3␤1 Integrin-mediated Cell Migration Induced by GnT-V-The Overexpression of GnT-III has been reported to inhibit cell migration by enhancement of E-cadherin-mediated homotypic adhesion (17) and by inhibiting ␣5␤1 integrin-mediated cell migration (38). In addition, in vitro GnT-V cannot use the product of GnT-III, a bisected oligosaccharide, as a substrate (12), so experiments were then designed to determine whether the introduction of GnT-III prevents ␣3␤1 integrin-mediated cell migration enhanced by GnT-V. The efficiency of transfection was confirmed by immunostaining with anti-GnT-III antibody and determined to be more than 80% (data not shown). As shown in Fig. 3, the transfection of GnT-III into the GnT-V transfectant resulted in a significant decrease in cell migration compared with the GnT-V transfectant (p ϭ 0.002). However, the inhibition was not observed after transfection of the GnT-III-inactive mutant, suggesting that the activity of GnT-III was essential for the negative regulation of GnT-V-induced cell migration. Therefore, we proposed that GnT-III directly counteracted the effect of GnT-V on ␣3␤1 integrin-mediated cell migration.
Transfection of GnT-III Had No Effect on the Expression of GnT-V and Integrin ␣3 Subunit-To explore the possible mechanisms involved in the inhibition of GnT-III-to GnT-Vinduced cell migration, we first attempted to determine whether the overexpression of GnT-III affected the expression of GnT-V and ␣3 subunit expressed on the cell surface by means of blotting a total cell lysate with the GnT-III antibody and the biotinylation of cell surface proteins followed by immunoprecipitation of ␣3 using the corresponding antibody, since N-glycosylation plays an important role in the quality control of  the expression of glycoproteins. As shown in Fig. 4A, the levels of expression of GnT-V were not influenced by the introduction of GnT-III, and equivalent amounts of loaded proteins were verified by blotting an actin antibody. On the other hand, the expression of integrin ␣3 subunit on the cell surface also remained unchanged among the transfectants of GnT-III plus GnT-V, GnT-III mutant plus GnT-V, and GnT-V (Fig. 4B). These results suggested that the inhibition of GnT-III-to GnT-V-induced cell migration could not be ascribed to a change in the expression levels of GnT-V and/or ␣3 subunit on the cell surface.

Transfection of GnT-III Had No Effect on the Activity of GnT-V-
Since the introduction of GnT-III had no effect on the expressions of GnT-V and ␣3 subunit, we further determined if the overexpression of GnT-III suppressed the activity of GnT-V. Since this was a transient transfection, the activity of GnT-III was checked at six time points from 24 to 144 h after the transfection. We found that GnT-III activity reached the highest level 48 h after transfection (Fig. 5A), and there was no corresponding activity in GnT-III mutant (data not shown). The expression level of GnT-III mutant was similar to that of wild-type GnT-III confirmed by blotting with GnT-III antibody, and equivalent amounts of loaded proteins were verified by blotting with anti-actin antibody (Fig. 5B). As shown in Fig. 5C, GnT-V activity was found to be stable, even in the period (48 h after transfection) where the activity of GnT-III reached the highest level in these double-transfected cells. This result indicated that GnT-III inhibited GnT-V-induced cell migration not due to the suppression of GnT-V activity.
Increased GnT-III Product but Decreased GnT-V Product on Integrin ␣3 Subunit-The modification of N-glycosylation contributes to the functions of integrins (39). Here, we checked whether changes of ␣3␤1 integrin modification had occurred in these transfectants. The integrins were immunoprecipitated from these transfectants and then probed with E 4 -PHA lectin, which preferentially binds to bisecting GlcNAc residues in N-glycans, or L 4 -PHA lectin, which binds to ␤1,6-branched GlcNAc. Fig. 6A (top) shows that the transfection of GnT-III to the GnT-V transfectant resulted in an increase in the GnT-III product on the integrin ␣3 subunit. More interestingly, the level of GnT-V product on ␣3 was decreased in the double transfectants (Fig. 6A, middle). Consistent with this observation, transfection of the GnT-III mutant failed to induce such changes. Equivalent amounts of the ␣3 subunit were verified by blotting ␣3-immunoprecipitated lysates (Fig. 6A, bottom). Moreover, cell lysates were subjected to SDS-PAGE, followed by a lectin blot. A comparison of bands especially around 117-200 and 60 -89 kDa among these transfectants consistently indicated that increased GnT-III products but decreased GnT-V products presented on the glycoproteins after the introduction of GnT-III to the GnT-V transfectant (Fig. 6B). Furthermore, to further confirm such competition on the ␣3 sub-

. No effects of GnT-III transfection on expression levels of both
GnT-V and integrin ␣3 subunit expressed on cell surface. A, double-transfected cells were lysed, and whole lysates were subjected to 7.5% SDS-PAGE and then transferred to a nitrocellulose membrane and blotted with GnT-V antibody (top) or actin antibody (bottom). B, transfected cells were biotinylated, whole lysates were immunoprecipitated (IP) with anti-␣3 antibody, and the samples were subjected to 7.5% SDS-PAGE and transferred to a nitrocellulose membrane. The biotinylated proteins were then detected as described under "Experimental Procedures." WT, wild type; WB, Western blot. unit, we purified this integrin from GnT-III, GnT-V, and GnT-III plus GnT-V transfectants using a GD6 peptide affinity column combined with a wheat germ agglutinin affinity column. The purity was evaluated by SDS-PAGE followed by silver staining (data not shown). The purified ␣3 subunit was cut from gels and then subjected to LC/MS n as described under "Experimental Procedures." As shown in Fig. 6, C and D, mass spectra of desialylated N-glycans were obtained from the ␣3 expressed in GnT-III, GnT-V, and GnT-III plus GnT-V transfectants, respectively, by a full MS1 scan (m/z 450 -2000). Carbohydrate structures of the major peaks were deduced from the m/z values of protonated ions in the full MS 1 spectra obtained by FT-ICR MS and product ions in MS 2,3 spectra (Fig. 6D). Based on the presence of [HexNAc-Hex-HexNAc-HexNAc-OH ϩ H] ϩ (m/z 792) and [HexNAc-Hex-HexNAc-(dHex)HexNAc-OH ϩ H] ϩ (m/z 938) in MS2,3 spectra, peaks 4, 5, 7, 8, 10, and 11 were determined as bisected glycans. Peak 4 was deduced to be a biantennary oligosaccharide, the major peak in the GnT-III transfectant. After the transfection of GnT-III into the GnT-V transfectant, peak 4 was increased compared with that of the GnT-V transfectant, whereas peak 6, which is the major peak in the GnT-V transfectant, was decreased. For the present technique, the branched form is determined by analyzing the sialylated oligosaccharides by LC/MS n in the negative ion mode. Referring to the result of the L 4 -PHA lectin blot and the fact that peak 6 is the major one in the GnT-V transfectant, peak 6 could be deduced the ␤1,6-branched GlcNAc form, although only bisialylated forms were detected by MS. The MS data also revealed that peaks 7 or 8 and 9, 10, or 11 were triantennary and tetraantennary oligosaccharides, respectively, from the pres-ence of their corresponding trisialylated and tetrasialylated forms. Peaks 1 and 2 were high mannose oligosaccharides. To further quantify the competition, we used the MS results to show that for GnT-III products (represented by the sum of the peaks 4, 5, 7, 8, 10, and 11), the proportion was, respectively, 79.5, 29.5, and 48.5% among the transfectants of GnT-III, GnT-V, and GnT-III plus GnT-V; for GnT-V products (represented by the sum of peaks 6 and 9), the proportion was 1.2, 34.9, and 18.1%, respectively, among the transfectants of GnT-III, GnT-V, and GnT-III plus GnT-V. Consistent with the results shown in Fig.  6A, these data strongly suggested that GnT-III transfection resulted in increased bisecting GlcNAc but decreased ␤1,6-branched GlcNAc on the ␣3 subunit. However, the N-glycan proportions partially, but not totally, are correlated with the extent of the modification in cell migration observed (Fig. 3), since only N-glycans located on some motifs of integrins have been proposed to influence their conformations and therefore to regulate their functions (40). 4 Taken together, these results suggested the following; ␣3 was a common target of GnT-III and GnT-V, and the priority taken by GnT-III in the competition resulted in the inhibition of GnT-V modification.
Increased ␤1,6-Branched GlcNAc as Well as Cell Migration in GnT-III Knockdown Cells-To further identify the competition of GnT-III and GnT-V definitely, we developed an RNA interference strategy to efficiently silence GnT-III expression in CHP134 cells, which express endogenous GnT-III and GnT-V. After retroviral infection, CHP134 cells were selected based on their resistance to G418 as described under "Experimental Procedures." GnT-III activity was effectively down-regulated by 70%, compared with those in parent and mock cells (Fig. 7A), whereas GnT-V activity, as a control, showed no significant changes (data not shown). A quantitative real time PCR analysis also indicated the down-regulation of RNA interference-directed GnT-III mRNA expression in these cells (Fig. 7B). We then tested cell migration on LN5 and found that GnT-III knockdown resulted in an increased cell migration compared with mock cells (Fig. 7, C and D). We further investigated the N-glycans on the ␣3 subunit. As shown in Fig. 7E, increased ␤1,6-branched GlcNAc but decreased bisecting GlcNAc on ␣3 was found in the GnT-III knockdown cells, compared with those in the mock cells. Together with the data in Fig. 6, these data provided the evidence to show that GnT-III inhibited

DISCUSSION
It has long been thought that the product of GnT-V, ␤1,6-GlcNAc branching of N-glycans, contributes directly to cancer progression and metastasis (6). Animal studies have shown that GnT-V-deficient transgenic mice experience attenuated tumor growth and metastasis (3). In human, the activity and/or expression of GnT-V is elevated in multiple types of tumors (41,42), and high levels of these enzymes or their cognate sugars are correlated with metastasis and a poor patient prognosis (41,43). In addition, GnT-V-modified cell surface receptors prolonged the turnover by inhibiting endocytosis (10) or resistance to degradation by protease (26). These results suggest that GnT-V may contribute to cancer metastasis through stabilizing target proteins. On the other hand, the introduction of GnT-III leads to a reduced metastatic potential. Moreover, those trans-fectants displayed decreased cell motility and attachment to laminin and collagen (14). Thus, it appears that GnT-V and GnT-III regulate cell migration and invasion as well as metastasis in opposite manners. In fact, GnT-III could be considered to be an antagonist of GnT-V, because bisecting GlcNAc renders the biantennary substrate inaccessible to GnT-V in vitro (13).
The ␣3␤1 integrin is one of the most important proteins that mediate cell motility and invasion and appears to be one of the plausible target proteins of GnT-V in promoting cancer metastasis. In fact, Pochec et al. (25) reported that ␤1,6-branched structures were highly expressed in high metastatic melanoma, compared with low metastatic melanoma. In the present study, we found for the first time that GnT-III and GnT-V competitively modify the same target, integrin ␣3 subunit, thereby regulating its functions. We demonstrated that GnT-III transfection to the GnT-V transfectant resulted in the inhibition of ␣3␤1 integrin-mediated cell migration, due to an increase of bisecting GlcNAc, but a decreased ␤1,6-GlcNAc, on the ␣3 subunit. However, the transfection of the GnT-III inactive mutant failed to induce such changes. Conversely, the competition was further confirmed by an RNA interference strategy to silence GnT-III in CHP134 cells, which express endogenous GnT-V and GnT-III. We found that GnT-III knockdown resulted in increased GnT-V product on the ␣3 subunit. Taken together, to the best of our knowledge, we presented a previously uncharacterized demonstration of the existence of competition for the same substrate between GnT-III and GnT-V in living cells (Fig. 8).
Two mechanisms have been proposed for the inhibition of cell motility by the overexpression of GnT-III: an enhancement in cell-cell adhesion and the down-regulation of cell-ECM adhesion (39). Our present study suggested one more; GnT-III competed with GnT-V for the modification of ␣3 subunit, causing a decrease in the product of GnT-V on ␣3 subunit. Luo et al. (40) had suggested that the changes in the glycan structure of integrin can affect its conformation and activity. They reported that in CHO-K1 cells, the addition of a glycan at the ␤1 I-like domain caused an increase in the distance between the ␤1 head and stalk domains, therefore inducing the integrin dimer to be a more extended (activated) integrin conformation (40). We suggested that the competition of GnT-III and GnT-V for the modification of ␣3 may cause changes in the glycan within key regions of this integrin, therefore causing the decreased cell migration. Details of the effect of glycans on this integrin are a subject of further investigation in our future study.
It is noteworthy that GnT-III and GnT-V do not always oppositely regulate all glycoproteins. In this study, we found that GnT-III transfection causes a similar decrease, but to a lesser extent, in cell migration on COL compared with the result on LN5. However, on COL, GnT-V transfection did not result in an increase in cell migration, compared with mock. This suggested that ␤1,6-GlcNAc modification has little effect or only mild effects on ␣1␤1 and ␣2␤1 integrin, which are receptors for COL. In fact, we reported that the introduction of the bisecting GlcNAc to the ␣5 subunit resulted in a reduced affinity in the binding of ␣5␤1 integrin to FN, resulting in a decreased cell migration (38). We thus assumed that the GnT-III affects the ␣1, ␣2, and ␣3 subunits similarly, which caused the decreased cell migration on LN5 and COL. However, the modification of ␤1,6-GlcNAc to ␣1 and ␣2 subunits may not affect their binding to COL. Considering that ␣3␤1 integrin is a strong adhesive receptor that promotes cell migration (37), the selective competition between GnT-III and GnT-V for ␣3 might play an important role in cancer metastasis.
Concerning metastasis, other important glycosyltransferases cannot be overlooked (e.g. sialyltransferases). The modification of the ␤1 subunit by sialyltransferase makes this integrin capped with the negatively charged sugar, sialic acid. The abundance of sialic acids, especially elevated ␣2,6-sialylation (44), contributes  On the other hand, the product of GnT-III suppresses cell migration. More importantly, this product cannot be utilized as a substrate by GnT-V, which is represented with a cross. Therefore, ␣3␤1-mediated cell migration induced by GnT-V can be blocked due to competition with GnT-III. E, mannose; f, N-acetylglucosamine.
to cell motility and invasion (25,(45)(46)(47). Thus, it is possible that GnT-V mediates at least some of its effects on cell behavior via increased sialylation (41). The effect of GnT-III on sialylation is a topic that also merits further exploration.
In conclusion, this study reports for the first time that GnT-III competes with GnT-V for the modification of integrin ␣3 subunit in living cells (Fig. 8). This competition results in the inhibition of ␣3␤1 integrin-mediated cell migration induced by GnT-V. The finding suggests that the competition between both enzymes occurs not only in vitro but also in vivo and might provide a new insight into unraveling the molecular mechanism of tumor metastasis.