Single Glycosyltransferase, Core 2 β1→6-N-acetylglucosaminyltransferase, Regulates Cell Surface Sialyl-Lex Expression Level in Human Pre-B Lymphocytic Leukemia Cell Line KM3 Treated with Phorbolester*

Sialyl-Lex (sLex) antigen expression recognized by KM93 monoclonal antibody was significantly down-regulated during differentiation induced by 12-O-tetradecanoylphorbol-13-acetate (TPA) in human pre-B lymphocytic leukemia cell line KM3. The sLex determinants were almost exclusively expressed on O-linked oligosaccharide chains of an O-glycosylated 150-kDa glycoprotein (gp150). A low shear force cell adhesion assay showed that TPA treatment significantly inhibited E-selectin-mediated cell adhesion. Transcript and/or enzyme activity levels of α1→3-fucosyltransferase, α2→3-sialyltransferase, β1→4-galactosyltransferase, and elongation β1→3-N-acetylglucosaminyltransferase did not correlate with sLex expression levels. However, transcript and enzyme activity levels of core 2 GlcNAc-transferase (C2GnT) were significantly down-regulated during TPA treatment. Following transfection and constitutive expression of full-length exogenous C2GnT transcript, C2GnT enzyme activities were maintained at high levels even after TPA treatment and down-regulation of cell surface sLex antigen expression by TPA was completely abolished. Furthermore, in the transfected cells, the KM93 reactivity of gp150 was not reduced by TPA treatment, and the inhibition of cell adhesion by TPA was also blocked. These results suggest that sLexexpression is critically regulated by a single glycosyltransferase, C2GnT, during differentiation of KM3 cells.

established ligands of CD62E, CD62P, and CD62L, also known as E-, P-, and L-selectins, respectively, and are expressed on glycoproteins and glycosphingolipids (for review, see Ref. 1). In human leukocytes and leukemic cells, only sLe x structures are expressed, and there is no sLe a expression. sLe x is expressed on granulocytes, monocytes, and lymphocytes, and importantly its expression on lymphocytes is regulated in a differentiation stage-specific or an activation stage-specific manner (2). The terminal tetrasaccharide structures are synthesized on N-acetyllactosamine unit repeats by sequential glycosyltransferase actions (reviewed in Ref. 3). Glycosyltransferases involved in sLe x structure synthesis are ␣133-fucosyltransferases (␣133FucT), ␣233-sialyltransferases (␣233ST), ␤13 4-galactosyltransferase (␤134GalT) (4 -5) and elongation ␤133-N-acetylglucosaminyltransferase (elongation ␤133GnT) (6). Among the five human ␣133FucTs, FucT-VII (7) is expressed in human leukocytes and leukemic cells and has been reported to regulate sLe x synthesis in human T lymphocytes during T cell activation (8,9). FucT-IV, termed the "myeloid" ␣133FucT gene is expressed in human granulocytes and myeloid cells and suggested to synthesize Le x and VIM-2 structures and sLe x structures (10 -11). Four human ␣233STs have been cloned, and among them ST3GalIII and ST3GalIV were reported to synthesize the NeuAc␣233Gal␤134GlcNAc structure that is the substrate for ␣133FucTs (12,13). Attempts to correlate the cell surface expression of sLe x with the expression of genes encoding ␣133FucTs and ␣233STs and to detect a rate-limiting step for sLe x structure expression during cell differentiation, transformation, and activation have not been definitive (8,14).
Mechanisms that regulate functional glycosphingolipid expression have been more extensively investigated. For example, we have demonstrated that functional sugar structure synthesis is not regulated through terminal and intermediate glycosyltransferases during differentiation and transformation of human and murine myelogenous leukemia HL-60, K562, and NFS60 cells. Instead, the most upstream glycosyltransferases critically determine terminal carbohydrate structure expression by modulating total flow of glycosphingolipid biosynthesis at upstream branching steps (15)(16)(17)(18)(19)(20). In the case of N-glycans, branching UDP-GlcNAc:␣-D-mannoside ␤136-N-acetylglucosaminyltransferase was reported to correlate closely with metastasis (21). In addition, the synthesis of some O-glycans is also regulated at a branching step by UDP-GlcNAc:Gal␤133-GalNAc (GlcNAc to GalNAc) ␤136-N-acetylglucosaminyltransferase (core 2 GlcNAc-transferase; C2GnT) in T cell activation and Wiscott-Aldrich syndrome (22,23). In this context, it would be of great interest to know whether cell surface sLe x structure expression could be regulated at some branching steps but not at the terminal and elongation steps during cell differentiation.
Human pre-B lymphocytic leukemia cell lines differentiate to a more mature stage upon treatment with 12-O-tetradecanoylphorbol-13-acetate (TPA) (24). KM3 is one of these pre-B lymphoid cell lines with such differentiation capability. In the present study, we have investigated down-regulation mechanism of cell surface sLe x expression using the differentiation system of the KM3 cells. We present here the direct correlation of a branching upstream GlcNAc-transferase, C2GnT, with expression of cell surface sLe x antigen determinants recognized by KM93 monoclonal antibody (mAb) during differentiation. Furthermore, we report that overexpression of C2GnT cDNA overcomes the effect of phorbolester on the synthesis of sLe x without changing other phenotypes of the transfected cells.
Cells and Cell Cultures-Human pre-B leukemia cell line KM3 was kindly supplied by Dr. Jun Minowada (Fujisaki Cell Center, Hayashibara Biochemical Research Institute, Fujisaki, Japan). KM3 cells, the human Burkitt lymphoma cell line Raji, and human myelogenous leukemia HL-60 cells were cultured in RPMI 1640 medium supplemented with 10% fetal calf serum. Differentiation of KM3 cells was carried out by incubating the cells with 8 nM TPA at an initial concentration of 2-5 ϫ 10 5 cells/ml (24). Biosynthesis of glycolipid sugar chains was inhibited by culturing the cells in the presence of 20 M PDMP for 3 days. For inhibition of O-linked sugar chain biosynthesis on glycoprotein, the cells were cultured with 4 mM benzyl-␣-GalNAc (Bz-␣-GalNAc; Sigma) for 3 days. Inhibition of N-linked sugar chain processing was conducted by culturing the cells with 10 g/ml swainsonine (Genzyme, Cambridge, MA) for 3 days. COS-1 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum.
Immunoblot Analyses-Immunoblot analyses were performed using anti-sLe x KM93 mAb or horseradish peroxidase-conjugated anti-human ␤134GalT mAb 8628. Protein was solubilized from cells by a 30-s sonication in 20 mM HEPES buffer (pH 7.2) containing 1% Triton X-100, heated at 100°C for 5 min in the presence of 2-mercaptoethanol, and subjected to 5% or 10% polyacrylamide gel electrophoresis (PAGE) containing 0.1% SDS. The protein was transferred to Immobilon-P SQ polyvinylidene fluoride membrane (Millipore Corp., Bedford, MA). The antigens were stained and detected by a chemiluminescence reagent kit, Renaissance (DuPont) according to the manufacturer's instructions. As a second antibody, horseradish peroxidase-conjugated goat antimouse IgM was used for the detection of KM93-reactive glycoprotein. For calculation of molecular size, prestained and calibrated Kaleidoscope molecular mass standards (Bio-Rad) were used.
Glycosyltransferase Assays-The total membranous fractions were prepared, aliquoted, and stored at Ϫ80°C until use (18). Protein was determined using a Bio-Rad protein assay kit with bovine serum albumin as a standard.
␣133FucT activities were assayed essentially by the method described for the nLcOse 4 ␤133GlcNAc-transferase assay (17) with modification. GDP-[fucose-U-14 C]Fuc (5.4 L) was added in a microtube and then dried. In the microtube, 1.68 nmol of cold GDP-Fuc, 20 nmol of NeuAc␣233Gal␤134GlcNAc␤133Gal␤134Glc-PA, sodium cacodylate buffer (pH 6.8, 3.75 mol), MnCl 2 (0.63 mol), Triton CF-54 (75 g), and the enzyme preparation (100 -300 g of protein) were mixed and incubated at 37°C for 1-3 h in a total volume of 25 l. The reaction was stopped by heating at 100°C for 2 min. The samples were then passed through a 0.22-m Millipore filter, and an aliquot of each filtrate was applied to a TSK gel ODS-80TM column (4.6 ϫ 150 mm, TOSOH, Tokyo, Japan) and separated by high performance liquid chromatography (HPLC) with a fraction collector. Elution was performed at 50°C with a 0.1 M acetate buffer (pH 4.0) containing 0.15% n-butanol at a flow rate of 1.0 ml/min. Transfer of the radioactive Fuc to the acceptor was determined.
␤134GalT activities were assayed as follows. Cold UDP-Gal (7.5 nmol), 20 nmol of GlcNAc-Gal␤134Glc-PA, sodium cacodylate buffer (pH 6.8, 3.75 mol), sodium deoxycholate (2.5 g), MnCl 2 (0.25 mol), and the enzyme preparation (100 -200 g of protein) were mixed and incubated at 37°C for 1-3 h in a total volume of 25 l. The reaction products were processed to HPLC analyses, and the eluates were subjected to quantitation by fluorescence intensity using Gal-PA as a standard.
Semiquantitative Reverse Transcribed PCR Analyses-Total RNA was extracted by the guanidinium CsCl method. Semiquantitative reverse transcribed PCR (RT-PCR) analysis was performed as follows. One g of RNA was reverse-transcribed by SuperScript II reverse transcriptase (Life Technologies, Inc.) using oligo(dT) primers (Amersham Pharmacia Biotech, Uppsala, Sweden). One-twentieth volume of the reaction mixture was subjected to PCR (total of 50 l). The reactions were performed using 60 mM Tris-HCl buffer (pH 9.0 for FucT-VII and ␤134GalT, pH 10.0, for ST3GalIII, ST3GalIV, and C2GnT and pH 8.5 for glutalaldehyde-phosphate dehydrogenase (GAPDH)), MgCl 2 (1. 5% PAGE was conducted in Tris borate-EDTA buffer at 50 V using one-fifth volume (10 l) of the PCR reaction mixtures, and the signals were visualized by autoradiography. The results were quantitated with a BAStation Bio-Image Analyzer (Fuji Film, Tokyo, Japan). By extensive control experiments, the PCR cycle numbers were determined for the respective primer pairs so that the amplification efficiency remained constant and the amplified PCR product was directly proportional to the quantity of the used RNA. The selected cycle numbers were 27 for FucT-VII and ST3GalIV, 26 for ST3GalIII and C2GnT, 25 for ␤134GalT, and 20 for GAPDH. The reasons why we took this procedure instead of conventional RNA blot methods or competitive RT-PCR analyses were as follows. (i) The transcript level of glycosyltransferase was very low, and we could hardly observe in detail their changes during TPA treatment by Northern analyses. (ii) It is easy to detect the timedependent precise changes of the relative amount of each transcript by the semiquantitative RT-PCR method.
Preparation of Full-length C2GnT cDNA and Subcloning to Mammalian Expression Vectors-Full-length human C2GnT cDNA was prepared by PCR cloning. Total RNA from HL-60 cells was reverse-transcribed as above. The cDNA was amplified by 32 cycles of PCR in 60 mM Tris-HCl buffer (pH 8.5), 1.5 mM of MgCl 2 , 15 mM (NH 4 ) 2 SO 4 , 250 M each of dATP, dGTP, dTTP, and dCTP. Primers used were 5Ј-TTA TTG TTT GAA ATG CTG AGG ACG-3Ј (forward) and 5Ј-TAA TGG TCA GTG TTT TAA TGT CT-3Ј (reverse). The amplified C2GnT cDNA was directly subcloned to pCR3 mammalian expression vector and a senseoriented clone was chosen and designated as pCR3-C2GnT. Sequences of the insert were analyzed by the dideoxy chain termination method (29), and the deduced amino acid sequence was confirmed to be identical to that of human C2GnT (22).
Overexpression of C2GnT cDNA in KM3 Cells-pCR3-C2GnT plasmid was transfected into KM3 cells by electroporation method as described (19). Over 10 monoclonal transfectants were selected and established by G418 resistance and by limiting dilution. Among them, a clone that expressed a maximum level of exogenous C2GnT transcript was chosen and designated as KM3c1e2. The transfected cell line with pCR3 vector alone was also established and designated as KM3m1e3. Expression of exogenous C2GnT was confirmed by PCR analyses using a reverse primer specific to pCR3 vector and a forward primer common to both endogenous and exogenous C2GnT (see Fig. 5A; pCR3 vectorspecific primer (V) and common forward primer (C)). Sequences of the common forward primer for C2GnT were the same as that described above. Differential detection of the endogenous C2GnT expression from the exogenous one was conducted by PCR analysis using the common sense primer C and a specific reverse primer for the 3Ј-untranslated region of C2GnT that were not included by exogenous plasmid sequence (see Fig. 5A; reverse primer specific to endogenous transcript (EN)). The neomycin resistance selection marker (neo R ) expressed in the transfectants was detected by PCR analyses. Primer sequences were as follows: exogenous C2GnT (pCR3), 5Ј-GTG ATG GAT ATC TGC AGA ATT C-3Ј (reverse); endogenous C2GnT (22), 5Ј-CTT CTT GTT CAT AAA ATT GCC CG-3Ј (reverse); neo R (30), 5Ј-TTG TCA CTG AAG CGG GAA GGG A-3Ј (forward) and 5Ј-AAT CGG GAG CGG CGA TAC CGT A-3Ј (reverse).
Preparation of E-selectin-transfected COS-1 Cells-Total RNA was extracted from human umbilical vein endothelial cells treated with recombinant IL-1␤ at 10 units/ml concentration for 4 h, and 1 g of RNA was reverse-transcribed as described above. The full-length human E-selectin cDNA was prepared by 32 cycles of PCR amplification in 60 mM Tris-HCl buffer (pH 9.0), 1.5 mM of MgCl 2 , 15 mM (NH 4 ) 2 SO 4 , 250 M each of dATP, dGTP, dTTP, and dCTP. Used primers were 5Ј-AAG TCA TGA TTG CTT CAC AGT TT-3Ј (forward) and 5Ј-AAC TTA AAG GAT GTA AGA AGG C-3Ј (reverse) (31). The amplified cDNA was directly subcloned to pCR3 expression vector, and a sense-oriented clone was chosen and designated as pCR3-E-selectin. The plasmid was transfected into COS-1 cells by the electroporation method as described (19), and the monoclonal transfectant was selected as above and designated as COS1E5 cells. E-selectin expression was confirmed by flow cytometry analysis and RNA blot methods (data not shown). The COS1 derivative cell line transfected with the vector alone, pCR3, was also prepared and designated as COS1m cells.
Low Shear Force COS Cell Adhesion Assay-After treatment with 4 mM Bz-␣-GalNAc for 3 days or 8 nM TPA for 4 days, KM3 cells were harvested, washed with PBS three times, labeled with BCECF-AM at 37°C for 15 min (32), washed with PBS twice, pretreated with or without anti-sLe x mAb (KM93) or the control mAb (mouse anti-IgG; Sigma), washed with PBS twice, and then resuspended in unsupplemented RPMI 1640 medium. Separately, 2.2 ϫ 10 6 COS1E5 cells were seeded and cultured on a cover glass in 35-mm dishes (assay plate) for overnight. After washing the COS1E5 cells twice with unsupplemented RPMI 1640 and pretreating the cells with or without anti-E-selectin mAb (1.2B6; T Cell Diagnostics, Cambridge, MA) or the control mAb (mouse anti-IgG; Sigma), a low shear force COS cell adhesion assay was performed as reported (33) with sight modification. On the layer of COS1E5 cells, 1 ϫ 10 6 of the above labeled KM3 cells (in 0.6 ml of medium) were placed. The cells were incubated on a constantly rocking platform for 15 min at 4°C, and the plates were washed five times with unsupplemented RPMI 1640 followed by fixation with cold 0.4% formaldehyde/RPMI 1640. The cover glass was removed from the assay plate bottom and placed on the slide glass. The fluorescence-labeled cells were counted on a fluorescence microscope system (BX-60/BX-FLA; Olympus, Tokyo, Japan), and the mean number of cells bound to COS1E5 cells/cm 2 was determined.
RESULTS sLe x Antigen Expression in KM3 Cells before and after TPA Treatment-Upon treatment of KM3 cells with 8 nM TPA, cell growth was significantly suppressed within a day (24). Expression of CD10, CD22, and CD20 was characterized before and after treatment (Fig. 1A, left). All three antigens are used as pre-B cell markers. However, CD22 and CD20 are known as more mature markers than CD10. While CD10 expression was very high in KM3 cells before treatment, the expression significantly decreased after treatment for 3 days. On the other hand, CD22 was obviously up-regulated after treatment. In addition, while CD20 was negative in KM3 cells before treat- FIG. 1. A, expression analyses of CD10, CD22, CD20, and CD15s in KM3 cells by FACScan. Left column (from top to bottom), CD10, CD22, and CD20 reactive cells, respectively. Solid, broken, and dotted lines in the left column represent the histograms of positive cells without TPA ment, the expression slightly increased after treatment. Consequently, it was thought that KM3 cells were differentiated into a more mature stage by TPA treatment as described (24). Subsequently, sLe x expression was investigated in the cells. As shown in Fig. 1A (middle), the cells were highly reactive with KM93 mAb before treatment. However, the reactivity significantly decreased in a time-dependent manner: no significant change at day 1, slightly decreased at day 2, and markedly down-regulated at day 3. After that, however, no further decrease was observed at days 4 and 5. The extent of the decreased KM93 reactivity at days 3-5 was about 1 ⁄6 to 1 ⁄12 of the control (calculated by the mean fluorescence intensity), although the results fluctuated from experiment to experiment. By contrast to the mAb KM93, KM3 cells were not stained at all by other mAbs against sLe x structures, CSLEX-1, and 2H5.
Expression of sLe x Structures on O-Glycosylated Protein-The highly expressed sLe x antigen significantly decreased by Bz-␣-GalNAc treatment for 3 days as well, while 3 days' treatment of the cells by swainsonine or PDMP did not show any change in KM93 reactivity at all (Fig. 1A, right). This clearly indicates that sLe x structures on KM3 cells are exclusively located on the O-linked oligosaccharide chains. Subsequently, we tried to detect glycoproteins expressing sLe x antigen determinant by immunoblot analyses (Fig. 1B). While no signal was observed in sLe x -negative Raji cells, a glycoprotein with an apparent molecular size of ϳ150 kDa (designated as gp150) was clearly detected in the positive KM3 cells. Although two minor proteins with molecular sizes of ϳ250 and ϳ200 kDa were also visualized, no other band was noticed on the 5% gel (third lane; No Tx) and even on the 10% gel (data not shown). Upon treatment with Bz-␣-GalNAc, the major and minor glycoproteins almost completely disappeared as shown in the second lane. This also indicates that sLe x structures are located on O-linked oligosaccharide chains of glycoproteins and correlates with the results presented by flow cytometry analyses (Fig. 1A,  right). In addition, we also analyzed the effects of TPA treat-ment for 1, 3, and 5 days (Fig. 1B, right three lanes). The intensity of gp150 and the other minor proteins significantly reduced in a time-dependent manner and completely disappeared at day 5. These results also correlated with the flow cytometry results (Fig. 1A, middle).
Low Shear Force COS Cell Adhesion Analyses of KM3 Cells-Roles of cell surface sLe x on KM3 cells were investigated as summarized in Fig. 2 using COS1E5 cells that permanently expressed E-selectin. KM3 cells adhered to COS1E5 cells ( Fig.  2A, uppermost bar), while sLe x -negative Raji cells did not exhibit cell adhesion (bottom bar). For another control, sLe xpositive KM3 cells did not adhere to COS1m cells transfected with the vector pCR3 alone (Fig. 2C). As shown in Fig. 2A, this adhesion was blocked by anti-E-selectin mAb or by anti-sLe x mAb KM93. In addition, Bz-␣-GalNAc treatment significantly abolished KM3 cell adhesion to COS1E5 cells, while swainsonine and PDMP had no effect on the adhesion. The number of adherent KM3 cells decreased with TPA treatment (Fig. 2B,  uppermost bar). Although the percentage of adherent KM3 cells treated with TPA was about 40% and higher than those of Bz-␣-GalNAc-treated cells (25%), the decrease in the adherent  Forty g of protein was prepared and subjected to 5% PAGE followed by transfer to Immobilon P SQ membrane and by staining with KM93 mAb. The signal was detected by the chemiluminescence method. The arrows indicate the positions of molecular mass standards.

FIG. 2. Low shear force COS cell adhesion analyses of KM3 cells.
Cells were treated with TPA or inhibitors, labeled with BCECF-AM (32), pretreated with or without anti-sLe x mAb (KM93) or control mAb (mouse anti-IgG; Sigma), and resuspended in unsupplemented RPMI 1640. Separately, COS1E5 cells were pretreated with or without anti-Eselectin mAb or the control mAb. Then low shear force COS cell adhesion assay was performed (33)   cell numbers was significant compared with those of no treatment control. To investigate why 40% binding remains after TPA treatment despite the dramatic down-regulation of sLe x expression (Fig. 1, A and B), we conducted another cell adhesion blocking experiment. Neither anti-E-selectin nor anti-sLe x mAb further blocks the binding of TPA-treated KM3 cells (Fig.  2B). Since this suggested that the remaining 40% binding is due to other factor(s) than E-selectin, we examined possible cell adhesion molecule expression in TPA-treated KM3 cells using FACScan. As summarized in Table I, we found no significant up-regulation of VLA4, LFA1␣, integrin ␤1, L-selectin, and CD44 by TPA treatment. Therefore, in addition to E-selectin, we think that alternative adhesion molecule(s) may mediate binding in our COS cell assay after TPA treatment.

Semiquantitative RT-PCR Analyses of Glycosyltransferase
Expression-The above results suggested that sLe x determinants are mainly located on O-glycosylated gp150 and that the sLe x structures on the protein play an important role in Eselectin-mediated cell adhesion. Therefore, expression of glycosyltransferase gene transcripts involved in the biosynthesis of sLe x was analyzed by semiquantitative RT-PCR analyses. Fig.  3A is a typical example of a control experiment demonstrating that the amplified PCR product (C2GnT) was directly proportional to the quantity of used cDNA content. For other PCR products, we have conducted similar experiments and determined that the products were proportional to cDNA amounts (data not shown). Subsequently, expression of FucT-VII, ST3GalIII, ST3GalIV, ␤134GalT, and C2GnT was examined using total RNA of TPA-treated KM3 cells, as shown in Fig. 3, B and C. Standard GAPDH was constitutively amplified, and F and R, forward and reverse primers for full-length cDNA, respectively. C, common forward primer for endogenous and exogenous transcript detection; EN, reverse primer specific to the endogenous transcript; V, reverse primer specific to the vector. The numbers after each character represent the start and end nucleotide positions of the primer, respectively. The position of the 5Ј-untranslated side is expressed as a negative number, and positions of the 3Ј-untranslated side in the vector sequence are expressed as primed numbers. The full-length coding region was prepared from cDNA from the endogenous transcript using the primers F and R and subcloned to pCR3 vector (indicated by the vertical arrows from the upper to lower illustrations). B, RT-PCR analyses of overexpressed C2GnT in the transfected cells. RNA was reverse-transcribed, and cDNA was subjected to PCR reaction. PAGE was conducted, and the bands were visualized by autoradiography. C2GnT (Endo) and C2GnT (Exo), endogenous (0.7 kb) and exogenous C2GnT (0.7 kb), respectively. neo R was 0.5 kb, and GAPDH wsa 0.6 kb. C, C2GnT activities in the transfected KM3 cells. The transfected cells were treated with or without 8 nM TPA for 4 days. Activities of C2GnT were measured using PNP-oligosaccharide as an acceptor (22). C2GnT enzyme activity is expressed as mean Ϯ S.D. of three separate experiments. its radioactivity levels were used for normalization of glycosyltransferase expression. FucT-VII PCR product was clearly observed at day 0, and the expression did not show significant change during differentiation, although its level was slightly enhanced at days 2 and 3. PCR product of ST3GalIII was also observed and constitutively expressed during differentiation by TPA. On the other hand, the expression of ST3GalIV and ␤134GalT transcripts was down-regulated to 2 ⁄3 and 1 ⁄2 levels during differentiation, respectively, especially at days 2 and 3. At day 5, however, expression of ␤134GalT recovered to its original level. Surprisingly, C2GnT expression showed more drastic change than ST3GalIV and ␤134GalT. Expression level of C2GnT transcript was down-regulated in a time-dependent manner to 1 ⁄5 at day 1 and to 1 ⁄10 at days 2 and 3, although the level recovered slightly at days 4 and 5.
Immunoblot Analyses of ␤134GalT Protein-To examine expression of ␤134GalT in protein level during TPA treatment, immunoblot analyses were performed using anti-human 8628 mAb (Fig. 3D). A substantial amount of ␤134GalT protein was exhibited before TPA treatment. While expression of the transcript was down-regulated to 1 ⁄2 the level of the control, ␤134GalT enzyme protein level did not show any decrease during differentiation. Instead, time-dependent increase was observed during TPA treatment.

Activities of Glycosyltransferases That Are Involved in sLe x
Expression-Subsequently, activities of ␣133FucT, ␣233ST, ␤134GalT, and elongation ␤133GnT were examined (Table  II). ␣133FucT activities after TPA treatment were almost the same level as the control. Despite down-regulation of ST3GalIV transcript and cell surface sLe x levels at days 3 and 4, ␣233ST activities did not markedly decrease after TPA treatment. By contrast, ␤134GalT activities significantly increased after TPA induction, although ␤134GalT transcript and cell surface sLe x levels were down-regulated after treatment. Taken together with the results of semiquantitative RT-PCR and immunoblot analyses for ␤134GalT, it was suggested that ␣133FucT, ␣233ST, and ␤134GalT do not hold any key role on the control of sLe x expression during KM3 cell differentiation. Activities of another intermediate glycosyltransferase, elongation ␤133GnT, were slightly decreased after differentiation. However, the decrease was not significant.
Activities of Glycosyltransferases Involved in Core Structure Synthesis of O-Glycans-Since activity and/or transcript levels of terminal and intermediate glycosyltransferases involved in sLe x structure synthesis did not completely correlate with sLe x expression and sLe x determinants were mainly on O-glycosylated gp150, it was of great interest to analyze the mechanism of O-glycan core structure synthesis. Therefore, we determined activities of C1GalT, C2GnT, C3GnT, and C4GnT before and after TPA-induced differentiation (Table II). Activities of C3GnT and C4GnT were not detected at all in control and TPA-treated KM3 cells, while homogenate from porcine colon mucosa had very high activity (27). On the other hand, substantial C1GalT and C2GnT activities were exhibited in control KM3 cells. Moreover, a significant decrease in C2GnT activities after TPA treatment was observed, while no such decrease was detected in C1GalT activities as shown in Table II. To confirm further the involvement of C2GnT in sLe x expression, we conducted time course experiments on the enzyme activity levels. The activities were significantly down-regulated in a time-dependent manner as shown in Fig. 4. Compared with the control, the activities decreased to about 40% at day 1, and 10 -18% at days 2-4. However, the activities at day 5 were slightly recovered from that at day 4. These results strongly suggest a direct correlation of C2GnT with sLe x expression, taken together with the results of flow cytometry, immunoblot (see Fig. 1), and RT-PCR analyses (see Fig. 3).
Overexpression of C2GnT in KM3 Cells-Since the suggested direct correlation of C2GnT with expression of sLe x -bearing O-glycan on gp150 seemed to be the most critical, we conducted overexpression experiments of C2GnT cDNA in KM3 cells. C2GnT-and mock-transfected KM3c1e2 and KM3m1e3 cells were established, respectively. First, we tried to perform differential detection of endogenous and exogenous C2GnT transcripts using RNA blot analyses. However, there were two endogenous transcripts in KM3 cells (4.8 and 2.1 kb); the overexpressed exogenous transcript was almost the same size as the smaller endogenous one, and we could not exactly differentiate the exogenous transcript from the endogenous one. Therefore, differential detection was conducted by RT-PCR analyses as illustrated in Fig. 5A (see the figure legend and "Experimental Procedures"). As summarized in Fig. 5B, the exogenous C2GnT expression was not confirmed in the parental KM3 and mock-transfected KM3m1e3 cells but in the KM3c1e2 cells, while the endogenous C2GnT transcript was detected in all three cell lines. On the other hand, the neo R gene transcript was detected in KM3c1e2 and KM3m1e3 cells, while the transcript was not exhibited in the parental KM3 cells. In addition, about 3-fold higher C2GnT activities were detected in the transfected KM3c1e2 cells than in the parental KM3 cells, as RNA was reverse-transcribed, and cDNA was subjected to PCR. Differential detection of the exogenous (Exo; 0.7 kb) from endogenous (Endo; 0.7 kb) C2GnT was carried out as illustrated in Fig. 5A. PAGE was conducted, and the bands were visualized by autoradiography (A). The quantitated value of each PCR product was normalized by GAPDH and is expressed as mean Ϯ S.D. of three separate experiments and as relative intensity (percentage) compared with the data at day 0 (B).
shown in Fig. 5C. On the other hand, the mock-transfected KM3m1e3 cells expressed the same amount of activities as the parental cells. These results indicate that the exogenous C2GnT is expressed not only in the transcript form but also enzymatically or functionally in the transfected KM3c1e2 cells.
Effect of Phorbolester on Glycosyltransferase and sLe x Expression in KM3c1e2-Using KM3c1e2 cells, expression of gly-cosyltransferase transcripts involved in sLe x structure synthesis were characterized during differentiation. Upon TPA treatment, the endogenous C2GnT transcript level was downregulated in the transfectant, KM3c1e2 cells, in a time-dependent manner just as in the parental KM3 cells (Fig. 6, A and B; compare Fig. 3, B and C). However, exogenous C2GnT was expressed constitutively, and the expression level did not change at all during TPA treatment (Fig. 6, A and B). Moreover, enzyme activities of C2GnT in KM3c1e2 cells were downregulated by TPA treatment, but the activity levels in the TPA-treated or -nontreated KM3c1e2 cells were significantly higher than those in the nontreated parental KM3 and mocktransfected KM3m1e3 cells (Fig. 5C). On the other hand, expression patterns of ␣133FucT, ST3GalIII, ST3GalIV, and ␤134GalT transcripts in TPA-treated KM3c1e2 cells were almost the same as those in the parental KM3 cells, especially in that the levels were not augmented at days 2-4 (data not shown).
Subsequently, effect of exogenous C2GnT transfection on sLe x expression was characterized using the transfected cell liens treated with TPA. Results of the flow cytometry analyses are summarized in Fig. 7A, left column. sLe x expression was down-regulated in the mock-transfected KM3m1e3 cells as in the parental KM3 cells upon TPA treatment (top and middle). In C2GnT-transfected KM3c1e2 cells, however, sLe x expression was not down-regulated by TPA treatment, and the expression level was almost the same as that of no-treatment control (bottom). There was a possibility that this block of sLe x expression down-regulation in the transfected cells was due to the cell surface phenotypic change before the TPA treatment; i.e. according to the transfection, KM3c1e2 cells could be thought to lose capability of differentiation and become resistant to the TPA treatment. To exclude the possibility, CD10 and CD22 expression was analyzed in the transfected KM3c1e2 cells after TPA treatment (Fig. 7A, middle and right columns). CD10 expression was down-regulated after TPA treatment in KM3c1e2 cells as well as KM3m1e3 and parental KM3 cells (middle). On the other hand, CD22 expression was up-regulated after TPA treatment in KM3c1e2 cells as well as the control cell lines (right). These results suggest that the block of sLe x expression down-regulation in KM3c1e2 cells is not due to acquiring the resistancy to TPA treatment but to the direct effect of C2GnT transfection on sLe x synthetic machinery.
Effect of Phorbolester on gp150 Expression and Cell Adhesion of KM3c1e2-We conducted immunoblot analyses in the transfected cells after TPA treatment, as presented in Fig. 7B. While gp150 was significantly decreased in its intensity after TPA treatment in the parental KM3 and mock-transfected KM3m1e3 cells, the protein of the C2GnT-transfected KM3c1e2 cells did not disappear. Instead, the intensity of gp150 in TPA-treated KM3c1e2 cells was as strong as that of the no-treatment control. This was also in a good agreement with the flow cytometry data (Fig. 7A, left column) and indicates that C2GnT transfection results in remodeling of the sLe x synthetic machinery in KM3c1e2 cells.
Finally, we tested the possible effect of C2GnT transfection on cell adhesion capability using low shear force COS cell adhesion assay (Fig. 8). Like KM3 parent cells, adhesion of KM3m1e3 cells was significantly inhibited by TPA treatment. By contrast, TPA treatment had no effect on cell adhesion of KM3c1e2 cells to E-selectin-transfected COS1E5 cells. This indicated that C2GnT transfection completely blocked the TPA-induced inhibition on cell adhesion of KM3c1e2 cells to COS1E5 cells. Since KM3c1e2 cells had no phenotypic change from the parental cells except for sLe x synthetic machinery (Fig. 7), the block of TPA-induced inhibition on cell adhesion must be due to a direct effect of C2GnT transfection on the sLe x synthetic machinery. DISCUSSION T and B lymphocytes are reported to undergo an antigenic shift from immature sLe x positive status to negative status along with differentiation (2). In this study, sLe x antigen expression on pre-B KM3 cells was down-regulated during B cell maturation with TPA (Fig. 1). This was in good agreement with previous results. There are several mAbs against sLe x structures including FH-6, CSLEX-1, and 2H5 (2). Among them, mAb KM93 has been introduced by Hanai et al. (34) and used for the expression cloning of FucT-VII (7). Although sLe x is highly expressed and easily detected by the different antibodies in human granulocytes, monocytes, and myelogenous and monocytic leukemia cells, expression of CD15s recognized by each mAb is somewhat different from each other in human lymphocytes (2). According to the present results, KM3 cells are strongly positive for KM93 but negative for CSLEX-1 and 2H5. There must be some structural differences between the epitopes recognized by KM93 and by CSLEX-1 or 2H5. As far as relationship between KM93-reactive sLe x and glycosyltransferase expression is concerned, however, C2GnT is surely thought as a "key" glycosyltransferase.
O-Glycans have been reported to play important roles on cell adhesion, and multivalency presented by O-glycans is thought as one of the essential characteristics to potentiate cell adhesion strength (35). Therefore, sLe x structures on the O-glycosylated gp150 may play a central role for E-selectin-mediated cell adhesion in our system. Structures of O-linked oligosaccharides have been classified on the basis of their fundamental core sequences; core 1, core 2, core 3, and core 4 oligosaccharides. Since C1GalT activities were significantly high in KM3 cells and were not down-regulated after TPA treatment (Table  II), the synthesis of core 2 sequences would take place in the presence of abundant C2GnT activities and its acceptor substrate, and it would determine the synthetic velocity of the remaining structures in the downstream biosynthetic pathways. Therefore, sLe x may be created on the termini of the N-acetyllactosamine units extended from the GlcNAc␤136 branch of core 2 structures in human leukemic cell lines. On the other hand, we do not exclude the possibility that type 2 chains are extended from the Gal␤133 residue of core 1 and/or the GlcNAc␤136 branch of I antigen structures as well as from the GlcNAc␤136 branch of core 2 structures. sLe x epitopes may possibly be present on their termini and play an important role in collaboration with the same type of epitopes on the type 2 units from the core 2 backbones. However, proof of this requires further experiments.
For ␤134GalT expression, there was a discrepancy between the results using semiquantitative RT-PCR and immunoblot analyses. This would be due to a change of translational efficiency (36) or due to other possible ␤134GalT. However, further elucidation is required. The expression of C2GnT transcript and enzyme activity levels in TPA-treated KM3 cells were about 1 ⁄10 of those in control cells (Fig. 3, B and C, and Fig.  4). Despite the 1 ⁄10 down-regulation, cell surface sLe x expression was not completely suppressed in flow cytometry analyses (Fig. 1A). This would be because the minimal level of C2GnT enzyme activities after TPA treatment was functionally enough to synthesize a certain (minimal) amount of sLe x structures in Since sLe x determinant-bearing gp150 was not detected after TPA treatment for 5 days, the minimal detection of sLe x antigen determinant might be due to greater sensitivity of flow cytometry analyses than immunoblot.
Following transfection and permanent expression of the C2GnT gene, we could remodel the sLe x structure synthetic machinery from phorbolester-responsive to phorbolester-resistant. In the case of CD43, leukosialin, however, C2GnT gene transfection could only result in remodeling CD43 O-glycans from tetrasaccharides to hexasaccharides, from the 110-to 135-kDa glyco-form (23). In addition, only minor portions of O-linked oligosaccharides have been demonstrated to have poly-N-acetyllactosaminyl extensions and sLe x structures in CD43 glycoprotein (37). Moreover, molecular size of our gp150 was distinct from that of CD43, and expression of sLe x antigen determinants was significantly influenced by C2GnT gene transfection in KM3 cells during TPA treatment. Despite the overexpression of C2GnT transcript and enzyme activities, however, further augmentation of cell surface sLe x expression could not be detected in nontreated KM3c1e2 cells. This may be because C2GnT level is saturated for influencing the sLe x synthetic machinery in nontreated KM3c1e2 cells or may be due to some other inhibiting mechanisms on the machinery.
Together with determinative action of FucT-VII, C2GnT has been reported to play an important role for presenting sLe x on P-selectin glycoprotein ligand-1 (PSGL-1) (38). However, we could not detect PSGL-1 transcript in pre-B KM3 cells (data not shown). The size of PSGL-1 is 250 kDa in nonreduced form and 120 kDa in its reduced condition (39), and it is different from that of our gp150 in KM3 cells. In addition, the recent report said that PSGL-1 is essential for adhesion to P-selectin but not E-selectin (33). Therefore, E-selectin-mediated cell adhesion of pre-B KM3 cells presented in this report may be distinct from the one through PSGL-1. Moreover, a direct and dynamic involvement of a single glycosyltransferase, C2GnT, in the control of CD15s expression during differentiation with constitutive and static FucT-VII expression has not been elucidated in any cell systems so far.
The molecular size of gp150 was also different from those of L-selectin (ϳ76 kDa) and lysosomal membrane glycoproteins (lamp-1 and lamp-2) (ϳ115 kDa) (40,41). Lymphocyte L-selectin was not suggested to react with E-selectin-transfected cells (42), and lamp-1 and lamp-2 were reported to carry sLe x structures on N-glycans but not on O-glycans (43). Therefore, it is less possible that our gp150 is identical to one of those previously known core proteins. On the other hand, gp150 could be a human counterpart of murine E-selectin ligand-1, judging from the molecular size (44). However, E-selectin ligand-1 is suggested to be N-glycosylated and at most only weakly Oglycosylated (44). By contrast, our gp150 is suggested to be essentially O-glycosylated. Therefore, further identification and characterization of gp150 including direct E-selectin accessibility to the glycoprotein is highly necessary.