Activation of β1,3-N-Acetylglucosaminyltransferase-2 (β3Gn-T2) by β3Gn-T8

Enzymatic activities of some glycosyltransferases are markedly increased via complex formation with other transferases or cofactor proteins. We previously showed that β1,3-N-acetylglucosaminyltransferase-2 (β3Gn-T2) and β3Gn-T8 can form a heterodimer in vitro and that the complex exhibits much higher enzymatic activity than either enzyme alone (Seko, A., and Yamashita, K. (2005) Glycobiology 15, 943–951). Here we examined this activation and the biological significance of complex formation in differentiated HL-60 cells. β3Gn-T2 and -T8 were co-immunoprecipitated from the lysates of both-transfected COS-7 cells, indicating their association in vivo. We prepared inactive mutants of both enzymes by destroying the DXD motifs. The mixture of mutated β3Gn-T2 and intact β3Gn-T8 did not exhibit any activation, whereas the mixture of intact β3Gn-T2 and mutated β3Gn-T8 had increased activity, indicating the activation of β3Gn-T2 via complex formation. Next, we compared expression levels of β3Gn-T1-T8 in HL-60 cells and DMSO-treated differentiated HL-60 cells, which produce larger poly-N-acetyllactosamine chains. The expression level of β3Gn-T8 in the differentiated cells was 2.6-fold higher than in the untreated cells. Overexpression of β3Gn-T8, but not β3Gn-T2, induced an increase in poly-N-acetyllactosamine chains in HL-60 cells. These results raise a possibility that up-regulation of β3Gn-T8 in differentiated HL-60 cells increases poly-N-acetyllactosamine chains by activating intrinsic β3Gn-T2.

Glycosyltransferases are present in the endoplasmic reticulum/Golgi membranes, cytoplasm, cell surface, and body fluids. In the presence of appropriate sugar donors, they work for the biosynthesis of various glycoconjugates. Recently, it has been shown that some glycosyltransferases form protein complexes with other glycosyltransferases and/or non-glycosyltransferase proteins (reviewed in Ref. 1). Complex formation contributes to enzymatic activation, stable expression in the Golgi apparatus, correct localization in intracellular vesicles, efficient biosynthesis of glycan chains, and modification of substrate specificities. Enzymatic activation has been proven for protein O-mannosyltransferases (2)(3)(4), N-acetylglucosaminyltransferases and glucuronyltransferases involved in heparan sulfate biosynthesis (5)(6)(7)(8)(9), ST8Sia-I (GD 3 synthase) and ␤4GalNAc-T1 (GM 2 /GD 2 synthase) (10), and chondroitin synthase (11)(12)(13). In these cases, glycosyltransferases exhibit little enzymatic activity when expressed alone, but their catalytic activities emerge if their respective cofactor proteins are simultaneously expressed. Because the activation does not occur by in vitro mixing of the glycosyltransferases and cofactor proteins, the process of complex formation appears to involve intermolecular disulfide bond formation or complicated interactions during early stages of polypeptide synthesis. In contrast, we previously found that ␤1,3-N-acetylglucosaminyltransferase-2 (␤3Gn-T2) 2 and ␤3Gn-T8 can form a heterodimer in vitro and that the enzymatic activity of the dimer is much higher than the sum of the individual activities (14). Because in vitro mixing of individually expressed fractions of ␤3Gn-T2 and -T8 is sufficient for enzymatic activation, dimer formation and enzymatic activation should occur with the completely folded proteins. However, it has been unclear whether the two enzymes associate with each other in vivo and which of the one or more enzymes are catalytically activated in the complex.
Poly-N-acetyllactosamine (polyLacNAc) is a linear glycan chain consisting of repeating N-acetyllactosamine units (Gal␤1-4GlcNAc␤1-3) n . The glycan chains occur in glycosphingolipids and N-linked/O-linked glycan chains of specific glycoproteins. In some cases, the 3-OH and/or 6-OH of galactose (Gal) and N-acetylglucosamine (GlcNAc) residues are modified by sialic acids, fucose (Fuc), and/or sulfate residues, which serve as determinants for various carcinoembryonic antigens and ligands for various cell recognition-associated lectins (15). In HL-60 cells, polyLacNAc chains exist primarily on lysosomal membrane glycoproteins (lamps) (16). HL-60 cells can differentiate to granulocytic cells in the presence of DMSO. * This work was supported in part by Grant-in-Aid for Scientific Research on Lee et al. (17) showed that polyLacNAc chains increase in the DMSO-treated differentiated HL-60 cells. They also showed that ␤3Gn-T activities in the differentiated HL-60 cells are 1.5fold higher than those in the undifferentiated cells, suggesting that ␤3Gn-T is a rate-limiting enzyme for the biosynthesis of longer polyLacNAc chains. This linear glycan is biosynthesized by the repeating action of ␤1,4-galactosyltransferase and ␤3Gn-T. ␤3Gn-T1, -T2, -T3, -T4, -T7, and -T8 have been shown to possess the ability to synthesize polyLacNAc chains (14, 18 -21). However, it is unclear which of the one or more enzymes are responsible for the increase in polyLacNAc chains in the differentiated HL-60 cells.
In this study, we have addressed the following issues: whether ␤3Gn-T2 and -T8 associate with each other in vivo; whether ␤3Gn-T2 and/or -T8 are catalytically activated in a complexed state; and which one or more ␤3Gn-Ts are associated with the increase in polyLacNAc chains in differentiated HL-60 cells.  (22).
Co-immunoprecipitation-The plasmids (5 g) pFLAG-T2 and pFLAG-T8 were transfected into semi-confluent COS-7 cells on 10-cm dishes using 20 g of Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Stable transfectants were isolated by selection using 400 g/ml G418 sulfate (Calbiochem, Darmstadt, Germany). A second transfection for transient expression was performed with 10 g of the appropriate plasmids and Lipofectamine 2000. The cells were harvested after 24 h and washed twice with phosphatebuffered saline. After adding 1 ml of lysis buffer (50 mM Tris-HCl (pH 7.5), 0.15 M NaCl, 1% (v/v) Nonidet P-40, 1 mM phenylmethanesulfonyl fluoride, 1 g/ml pepstatin, and 1 g/ml leupeptin), the cell pellets were suspended and lysed on ice for 20 min. After centrifugation, the supernatants were collected and incubated with 1 g of anti-myc antibody (Invitrogen) and 15 l of Protein G-Sepharose TM 4 Fast Flow (GE Healthcare, Buckinghamshire, England) at 4°C for 1 h. The resins were washed with the lysis buffer four times. Equal aliquots of the resins were used for SDS-PAGE, and the proteins were trans-ferred onto a nitrocellulose membrane (Bio-Rad Trans-Blot Transfer Medium, Hercules, CA). Tag-conjugated proteins on the membranes were treated with horseradish peroxidase)conjugated anti-FLAG-M2 antibody (0.3 g/ml) (Sigma-Aldrich) or horseradish peroxidase-conjugated anti-myc antibody (1 g/ml, Invitrogen) and detected with ECL Western blotting Detection Reagents (GE Healthcare).
Tomato-lectin Blotting-The plasmids (10 g) were transfected into HL-60 cells (1 ϫ 10 6 ) using 20 g of Lipofectamine 2000 according to the manufacturer's instructions. After 48 h, the cells were collected and washed with phosphate-buffered saline three times. Aliquots of cell homogenates (3 g of protein) were used for SDS-PAGE, and the proteins were transferred onto a nitrocellulose membrane. Parts of membranes were digested with Escherichia freundii endo-␤-galactosidase (Seikagaku Co., Tokyo, Japan) (50 milliunits of enzyme in 0.1% bovine serum albumin-50 mM sodium acetate buffer (pH 5.5) at 25°C for 16 h). After blocking with 1% bovine serum albumin/ phosphate-buffered saline, the membrane was treated with 10 g/ml biotin-conjugated lectin from Lycopersicon esculentum (tomato, Sigma-Aldrich) in phosphate-buffered saline/0.1% Tween 20 at 4°C for 2 h. After washing, the membrane was treated with horseradish peroxidase-conjugated streptavidin (GE Healthcare) at 4°C for 1 h. The detection was performed using ECL Western blotting Detection Reagents. Chemiluminescence was quantified using a LAS-1000 multi-imager (Fuji Photo Film, Japan).

RESULTS
␤3Gn-T2 Associates with ␤3Gn-T8 in Vivo-We previously found that the mixture of ␤3Gn-T2 and ␤3Gn-T8 exhibited much higher enzymatic activity than either enzyme alone and that the two enzymes could form a heterodimer in vitro (14). However, it remained unclear whether the two enzymes interacted with each other in vivo. To assess this, co-immunoprecipitation was performed. ␤3Gn-T2 FLAG tagged at the C terminus (T2-FL) and ␤3Gn-T8 myc tagged at the C terminus (T8-myc) were simultaneously expressed in COS-7 cells, and the cell lysates were immunoprecipitated by anti-myc antibody. The precipitates were analyzed by Western blotting (Fig. 1A). T2-FL co-immunoprecipitated with T8-myc and was detected by anti-FLAG antibody, indicating that ␤3Gn-T2 associates with ␤3Gn-T8 in vivo.
␤3Gn-T2 Is Activated by ␤3Gn-T8-Increased enzymatic activity in the mixture of ␤3Gn-T2 and -T8 indicated that the enzyme(s) were activated in the complex. To assess which ␤3Gn-T was activated, we prepared mutated proteins that were enzymatically inactive, but could form the complex. Glycosyltransferases generally have a DXD motif (23), which is involved in binding to divalent cations and sugar nucleotides and is thus essential for their catalytic activities. ␤3Gn-T2 and -T8 contain ␤3Gn-T8 Activates ␤3Gn-T2 DDD 247 and 246 QDD 248 , respectively, as DXD motifs. We constructed expression vectors with mutated ␤3Gn-T2 (T2-DA) and -T8 (T8-QA), which had 245 ADD 247 and 246 ADD 248 , respectively. The ability of T2-DA and T8-QA to form a complex in vivo was examined by co-immunoprecipitation as above. Fig. 1 (B and C) shows that the substitutions in their DXD motifs did not affect their ability to interact with their wild-type counterpart. Next, we examined the enzymatic activities of T2-DA, T8-QA, and 1:1 mixtures with wild-type ␤3Gn-T2 and -T8. Soluble forms of (His) 6 -tagged T2-DA and T8-QA were produced using with the P. pastoris protein expression system, as previously described (14), and purified by Ni-NTA agarose chromatography (Fig. 2). Both mutant proteins electrophoresed as smeared bands, similar to the wildtype proteins. Peptide:N-glycosidase F treatment showed that the polypeptide moieties of T2-DA and T8-QA had the same molecular masses (ϳ45 kDa) as the wild-type enzymes (Fig. 2). The smeared profiles were attributed to heterogeneity in yeast large N-linked glycans. The enzymatic activities are shown in Fig. 3. Neither T2-DA nor T8-QA had any enzymatic activity, indicating that the DXD motifs are required for catalytic activity. The mixture of wild-type ␤3Gn-T2 and T8-QA had 6.3-fold higher enzymatic activity (1.9 Ϯ 0.05 pmol/min) than ␤3Gn-T2 alone (0.30 Ϯ 0.04 pmol/min), whereas the mixture of wild-type ␤3Gn-T8 and T2-DA had the same level of activity as ␤3Gn-T8 alone (0.014 Ϯ 0.002 pmol/min). The same results were obtained from another independent experiment using individually prepared enzymes. These results indicate that ␤3Gn-T2 is activated by ␤3Gn-T8, and that the mutant T8-QA, which has no enzymatic activity, is also able to activate ␤3Gn-T2.
Next, we examined whether murine ␤3Gn-T8 could activate murine and human ␤3Gn-T2. Amino acid similarities for catalytic domains of ␤3Gn-T2 and ␤3Gn-T8 between human and murine were 87 and 75%, respectively. As shown in Table 1, the V max /K m value of the mixture, murine ␤3Gn-T2 and murine ␤3Gn-T8, was approximately twice higher than that of murine ␤3Gn-T2 alone. Interestingly, murine ␤3Gn-T8 could activate human ␤3Gn-T2, although the degree of the activation by murine ␤3Gn-T8 was lower than that by human ␤3Gn-T8. In converse, human ␤3Gn-T8 could also activate murine ␤3Gn-T2. These results indicate that the ability of ␤3Gn-T8 to activate ␤3Gn-T2 is conserved between human and murine. Furthermore, to assess whether ␤3Gn-T8 could stabilize ␤3Gn-T2 or not, the stability of the enzymatic activities of human ␤3Gn-T2 in the presence or absence of human ␤3Gn-T8 was examined. As shown in Fig. 4, the enzymatic activities of ␤3Gn-T2 alone and the mixture of ␤3Gn-T2 and ␤3Gn-T8 decreased in the same manner. This suggests that ␤3Gn-T8 does not have the ability to stabilize the enzymatic activity of ␤3Gn-T2 at least in vitro.

DISCUSSION
We have clearly demonstrated in this study that (i) ␤3Gn-T2 and -T8 associate in vivo, (ii) ␤3Gn-T2 is activated by ␤3Gn-T8, and (iii) the increase in polyLacNAc chains in differentiated HL-60 cells is due primarily to up-regulation of ␤3Gn-T8. We previously showed that soluble forms of ␤3Gn-T2 and -T8, lacking transmembrane and cytoplasmic regions, could form an activated heterodimer in vitro (14). This result suggested that catalytic domains and/or stem regions are necessary for the complex formation and enzymatic activation. However, it remained unclear which regions of the two enzymes interacted with each other. Because activation occurs between ␤3Gn-T2 and DXD-mutated T8-QA, the binding of ␤3Gn-T8 to UDP-GlcNAc is not necessary for the complex formation, and therefore the two enzymes may interact in a polypeptide region outside of the catalytic region. Their interaction seems to change the conformation of the catalytic region of ␤3Gn-T2 and elevate its catalytic activity. Several studies have examined the  ␤3Gn-T8 Activates ␤3Gn-T2 molecular mechanism of complex formation by glycosyltransferases: stem regions are important for the interaction of ␤1,2-N-acetylglucosaminyltransferase-I and ␣-mannosidase II (25) and for the oligomerization of ␤1,3-glucuronosyltransferase (26), ␤1,6-N-acetylglucosaminyltransferase-V (27), heparan sulfate 6-O-sulfotransferases (28), and GlcNAc 6-O-sulfotransferase-1 (29). Transmembrane regions are also important for oligomerization of ␤1,4-galactosyltransferase-I (30), ␣1,3-fucosyltransferase VI (31), and ␣2,6-sialyltransferase-I (32). Catalytic domains are involved in oligomerization of ␣2,6-sialyltransferase-I (33) and dimerization of GM 2 synthase (34). Some glycosyltransferases form oligomer/multimer complexes with rather high molecular weights, whereas other complexes are formed via intermolecular disulfide bond(s), most likely in the process of polypeptide biosynthesis. The heterodimer between ␤3Gn-T2 and -T8 may be suitable for tertiary structural studies of glycosyltransferase complexes, because this complex can be formed in vitro. Such studies will lead to a better understanding of the molecular mechanism of complex formation by glycosyltransferases.
The V max /K m value of the human T2/T8 complex is 9.3-fold higher than that of ␤3Gn-T2 alone (14). By complex formation with ␤3Gn-T8, the K m value of human ␤3Gn-T2 decreases 2.4fold, and the V max value increases 3.9-fold (14). In contrast, the substrate specificity of the complex is almost the same as that of ␤3Gn-T2 (14), and ␤3Gn-T8 cannot stabilize the enzymatic activity of ␤3Gn-T2 (in this study), indicating that ␤3Gn-T8 can augment turnover velocity of ␤3Gn-T2. Interestingly, murine ␤3Gn-T8 can also activate murine ␤3Gn-T2 (Table 1). In this case, the K m value of ␤3Gn-T2 decreases 4.6-fold by complex formation, and the V max value also decreases 2.3-fold. In murine T2/T8, the K m value is altered by the complex formation more than the V max value. In contrast, the V max value is altered more than the K m value in human T2/T8. The biological significance of this difference is unclear, but it may be possible that the availability of acceptor substrates is different between in human and murine cells. It should be noted that murine and human ␤3Gn-T8 can activate human and murine ␤3Gn-T2, respectively. Human ␤3Gn-T8 is more effective for the activation of murine ␤3Gn-T2 than murine ␤3Gn-T8. This result suggests that putative binding sites between ␤3Gn-T2 and -T8 could be conserved in human and murine.
What is the biological significance of complex formation between ␤3Gn-T2 and -T8 and the resulting enzymatic activation? It should be noted that ␤3Gn-T2 alone has substantial activity. In fact, when tetra-antennary N-linked glycan is used as a substrate, ␤3Gn-T2 has the highest specific activity of ␤3Gn-T1, -T2, -T3, -T4, -T7, and -T8, all of which are able to synthesize polyLacNAc chains (14, 18 -21). Moreover, although tetra-antennary glycan is the best substrate for ␤3Gn-T2, this enzyme can also efficiently act on tri-, bi-, and monoantennary N-linked glycans (14). Both ␤3Gn-T2 and -T8 are expressed in various human tissues, but their relative expression levels differ between tissues. In particular, ␤3Gn-T8 is poorly expressed in colon, prostate, and brain, whereas ␤3Gn-T2 is substantially expressed in those tissues (19,20). Considering these facts, we speculate that ␤3Gn-T2, even in the absence of ␤3Gn-T8, usually synthesizes rather shorter polyLacNAc chains or elongates lower branching N-linked glycans. In contrast, expression of ␤3Gn-T8 may be required for elongated-polyLacNAc-chain synthesis in some specific tissues, at specific developmental stages, and in carcinogenesis. Ishida et al. (20) have reported that expression of ␤3Gn-T8 is quite low in normal colon, but increases markedly in colon cancer tissues. Although it has been unclear whether expression of polyLacNAc chains [33Gal␤134GlcNAc␤13] n increases in colon cancer, Terada et al. (35) have recently reported that novel fucosylated polyLacNAc chains, [33Gal␤133(Fuc␣13 4)GlcNAc␤13] n , occur in colon cancer SW1116 cells and serve as ligands for mannan-binding protein. These results suggest that ␤3Gn-T8 is responsible for the biosynthesis of this type of polyLacNAc chain in malignant tumor cells.
Recently Togayachi et al. (36) reported on ␤3Gn-T2 knockout mice, in which the expression of polyLacNAc chains detected by tomato lectin is markedly reduced, at least in thymus, spleen, lymphocytes, and macrophages, suggesting that ␤3Gn-T2 is predominantly involved in the synthesis of polyLacNAc chains in these tissues. Their results are in accordance with our results showing that the increased enzymatic activity is attributable to the ␤3Gn-T2 portion of the ␤3Gn-T2/T8 complex.
Lee et al. (17) performed quantitative analysis of N-linked glycan chains of lamps in HL-60 cells and HL-60 differentiated cells and clearly showed the increase of polyLacNAc chains in the latter. Because the level of N-linked glycans binding to tomato lectin increases only 1.5-to 1.6-fold, this may not seem so remarkable. However, our results are in agreement, because the polyLacNAc chains detected by tomato-lectin staining in ␤3Gn-T8-transfected HL-60 cells increased ϳ2-fold (Fig. 6). Because the molecular weights of lamps derived from ␤3Gn-T8-or T8-QA-transfected cells are comparable with those from untreated HL-60 cells, the increase in LacNAc units in the transfected cells does not seem particularly large. Actually, Lee et al. (17) showed that the molecular masses of lamp proteins increase from 110 -150 kDa to 130 -170 kDa in the differentiation of HL-60 cells. We observed the same results as them (data not shown). This increase seems to be due to the increase of polyLacNAc chains, because Lee et al. (17) showed by structural studies that levels of other sugar modifications such as sialylation and fucosylation in lamp proteins don't change between the differentiated and undifferentiated HL-60 cells. This fact suggests that the expression level of ␤3Gn-T8 in the differentiated HL-60 cells may be higher than those in the transfectants for wild and mutated ␤3Gn-T8 in Fig. 6. On the FIGURE 7. A hypothetical schematic model for the increase in polyLacNAc chains during differentiation of HL-60 cells. In HL-60 cells, ␤3Gn-T2 is more highly expressed than ␤3Gn-T8, and some ␤3Gn-T2 (circles) is free from the complex with ␤3Gn-T8. After DMSO treatment, ␤3Gn-T8 is up-regulated. Newly synthesized ␤3Gn-T8 forms a complex with free ␤3Gn-T2 and activates ␤3Gn-T2 (hexagons). The arrow sizes indicate the intensity of the relative enzymatic activities.
PolyLacNAc is known to be expressed in specific cells/tissues associated with development and carcinogenesis and to serve as a cell-recognition molecule by binding to several lectin proteins. The next step is to determine whether these biological phenomena are related to ␤3Gn-T2/T8 complex formation.