Molecular cloning and characterization of UDP-GlcNAc:lactosylceramide beta 1,3-N-acetylglucosaminyltransferase (beta 3Gn-T5), an essential enzyme for the expression of HNK-1 and Lewis X epitopes on glycolipids.

A new member of the UDP-N-acetylglucosamine:beta-galactose beta1,3-N-acetylglucosaminyltransferase (beta3Gn-T) family having the beta3Gn-T motifs was cloned from rat and human cDNA libraries and named beta3Gn-T5 based on its position in a phylogenetic tree. We concluded that beta3Gn-T5 is the most feasible candidate for lactotriaosylceramide (Lc(3)Cer) synthase, an important enzyme which plays a key role in the synthesis of lacto- or neolacto-series carbohydrate chains on glycolipids. beta3Gn-T5 exhibited strong activity to transfer GlcNAc to glycolipid substrates, such as lactosylceramide (LacCer) and neolactotetraosylceramide (nLc(4)Cer; paragloboside), resulting in the synthesis of Lc(3)Cer and neolactopentaosylceramide (nLc(5)Cer), respectively. A marked decrease in LacCer and increase in nLc(4)Cer was detected in Namalwa cells stably expressing beta3Gn-T5. This indicated that beta3Gn-T5 exerted activity to synthesize Lc(3)Cer and decrease LacCer, followed by conversion to nLc(4)Cer via endogenous galactosylation. The following four findings further supported that beta3Gn-T5 is Lc(3)Cer synthase. 1) The beta3Gn-T5 transcript levels in various cells were consistent with the activity levels of Lc(3)Cer synthase in those cells. 2) The beta3Gn-T5 transcript was presented in various tissues and cultured cells. 3) The beta3Gn-T5 expression was up-regulated by stimulation with retinoic acid and down-regulated with 12-O-tetradecanoylphorbol-13-acetate in HL-60 cells. 4) The changes in beta3Gn-T5 transcript levels during the rat brain development were determined. Points 2, 3, and 4 were consistent with the Lc(3)Cer synthase activity reported previously.

The expression of the Lewis x (Le x ; CD15) carbohydrate structure, ␣1,3-fucosyl-N-acetyllactosamine, Gal␤1-4(Fuc␣1-3)GlcNAc␤1-R, is also regulated developmentally and region specifically in the brain (23)(24)(25)(26)(27)(28)(29)(30)(31). The Le x also functions as a cell-cell recognition molecule in the highly organized structures of the central nervous system (30,32,33). The Le x expression on GSLs in brain is also regulated by the Lc 3 Cer synthase which synthesizes a root structure of the GSLs carrying the Le x epitope (34). Thus, SGGLs and GSLs carrying the Le x epitope are of interest in terms of their biosynthetic regulation. Lc 3 Cer synthase is an important enzyme with respect to hematopoietic cell differentiation (35). HL-60 cells, a human promyelocytic leukemic cell line, are capable of bidirectional differentiation into a monocytoid or granulocytoid lineage (35). The Lc 3 Cer synthase and GM3 synthase (lactosylceramide: ␣2,3-sialyltransferase), which share the acceptor substrate lactosylceramide (LacCer); Gal␤1-4Glc␤1-1Cer, are key enzymes in determining the biosynthetic flow of GSLs in the two different directions. During the monocytic differentiation of HL-60 cells induced by 12-O-tetradecanoylphorbol-13-acetate (TPA) treatment, ganglioside GM3 was markedly increased by upregulation of GM3 synthase and down-regulation of Lc 3 Cer synthase. In contrast, during the granulocytic differentiation of HL-60 cells induced by all-trans-retinoic acid (RA), neolactoseries GSLs were markedly increased by up-regulation of Lc 3 Cer synthase. Stults and Macher (36) concluded that mature myeloid cells exhibited ␤3Gn-T activity toward both acceptor substrates, leading to the expression of several types of neolacto-neutral GSLs in the cells, while lymphoid cells exhibited activity for nLc 4 Cer, but not LacCer. The lack of Lc 3 Cer synthesizing activity in lymphoid cells explains the absence of neolacto-neutral GSLs.
Lc 3 Cer synthase is the enzyme controlling the expression of neolacto-series GSLs and so plays important roles in many cells. In the present study, we cloned and characterized a fifth member of the ␤3Gn-T family (␤3Gn-T5), and identified ␤3Gn-T5 as the most likely candidate for Lc 3 Cer synthase.

EXPERIMENTAL PROCEDURES
Tumor Cell Lines and Monoclonal Antibodies (MAbs)-The tumor cell lines were cultured in RPMI 1640 medium (Life Technologies, Inc., Rockville, MD) supplemented with 10% fetal bovine serum.
Isolation of Rat and Human ␤3Gn-T5 cDNAs-We constructed a rat shank cDNA library for a random sequencing project. During the sequencing, we found a rat cDNA encoding a partial sequence of ␤3Gn-T5. This novel sequence (738 bp) did not encode the full ORF. But it had the ␤3Gal-T (␤3Gn-T) motifs which are shared by the known ␤3Gal-Ts and ␤3Gn-Ts. The cDNA library of Colo205 cells constructed in a previous study (6) was screened with the rat partial ␤3Gn-T5 cDNA as a probe to isolate the full-length human ␤3Gn-T5 cDNA which possessed a 3.7kilobase pair insert DNA. We did not clone a rat full-length ␤3Gn-T5 in this study.
Quantitative Analysis of the Four ␤3Gn-T Transcripts in Human Tissues and Tumor Cell Lines by Competitive RT-PCR-The principle behind the competitive RT-PCR method for quantification of transcripts was described in detail in our previous studies (6,38). Total cellular RNAs of various human tissues were purchased from CLONTECH. Those of various tumor cell lines were extracted and purified in our laboratory. As to the measurement of the transcripts for three cloned ␤3Gn-T genes, ␤3Gn-T2, -T3, and -T4, the primers and the PCR conditions were also reported previously (1).
Regarding the ␤3Gn-T5 gene, a standard DNA plasmid was constructed by subcloning a full-length ORF cDNA into a pBluescript SK(Ϫ) vector. A competitor DNA plasmid carrying a small deletion (245 base pairs) within the ORF of the ␤3Gn-T5 cDNA was constructed as follows. The standard DNA plasmid was double digested with appropriate restriction endonucleases, MscI and BglII, and was blunt-ended with T4 DNA polymerase. This was followed by self-ligation of the deleted DNA plasmid. As to the measurement of the ␤3Gn-T5 transcripts by competitive RT-PCR, we used the following primer set: forward primer, 5Ј-TCTTATGACTGCTGATGATGACAT-3Ј, and reverse primer, 5Ј-CTTTAGGATCTGTAGCATTCTTCC-3Ј.
Transfection Experiments to Express Each of the Four Human ␤3Gn-T Genes in Namalwa Cells-Expression of glycosyltransferase genes subcloned in pAMo was described in detail in a series of previous papers (6, 7, 39 -42). The construction of pAMo vector inserted with each of three cloned ␤3Gn-T genes, ␤3Gn-T2, -T3, and -T4, was also reported previously (1). Regarding the ␤3Gn-T5 gene, the ␤3Gn-T5 ORF fragment was excised from the standard ␤3Gn-T5 DNA in pBluescript SK(Ϫ), and then blunt-ended with T4 DNA polymerase. The blunt-ended fragment was flanked with SfiI linker and inserted into the SfiI site of pAMo vector. Each of the four ␤3Gn-T genes subcloned into a pAMo vector was transfected by electroporation into Namalwa cells. These cells were selected in the presence of geneticin (G418) (Life Technologies, Inc.) at a concentration of 0.8 mg/ml. Stable transfectant cells were obtained after 25 days exposure to geneticin. The levels of the transcripts expressed in the Namalwa transfectant cells were measured by means of competitive RT-PCR.

Separation of Glycolipids by Thin Layer Chromatography (TLC) and
Immuno-TLC with Anti-nLc 4 Cer (Anti-paragloboside) MAb-For glycolipid analysis, 1.0 ϫ 10 8 of the Namalwa transfectant cells with each ␤3Gn-T gene were collected, washed twice with phosphate-buffered saline, and then lyophilized. Total glycolipids were extracted from the lyophilized cells. Crude glycolipids were extracted twice from the transfectant cells, first with chloroform/methanol (2:1, v/v), and then with chloroform/methanol/water (30:60:8, v/v/v). Samples dried with the N 2 evaporator were dissolved in methanol, then subjected to mild alkaline treatment in 0.1 N KOH/methanol at 40°C for 2 h, and neutralized with 1 N acetic acid. After the free fatty acids had been removed with n-hexane, the remaining fractions were dried with the N 2 evaporator and subjected to Folch's partition. The lower neutral glycolipid fractions were dried with the N 2 evaporator and subjected to immuno-TLC analysis. Neutral glycolipid equivalent to 1.0 ϫ 10 7 cells was applied to each lane of TLC. Neutral glycolipids were separated by TLC (HPTLC Kieselgel 60, 5641; MERK, Germany) with mixtures of chloroform/methanol/water (60:35:8, v/v/v) and immuno-TLC analysis was performed with the antibody 1B2 as described (36).
Construction and Purification of ␤3Gn-T Proteins Fused with FLAG Peptide-The putative catalytic domain of each of ␤3Gn-T2, -T3, -T4 had been expressed as a secreted protein fused with FLAG peptide in insect cells as described in a previous study (1). In the present study, a 1.1-kilobase pair DNA fragment encoding a COOH-terminal portion of ␤3Gn-T5 (amino acids 39 to 378) was amplified by PCR. The PCR was performed with Platinum Pfx DNA Polymerase (Life Technologies, Inc.), according to the supplier's manual. The 5Ј and 3Ј primer sequences were flanked with BamHI and XbaI sequences, respectively, to create the restriction sites. Those sequences were as follows: forward primer, 5Ј-CGGGATCCATTGTGAGCCATATGAAGTCATAT-3Ј, and reverse primer, 5Ј-GCTCTAGATGACAGTGAAAACATACAACATTC-3Ј. The amplified fragment was first inserted between the BamHI and XbaI sites of the pBluescript SK(Ϫ) vector. Subsequently, the insert DNA was excised by BamHI and NotI, and was inserted between the BamHI and NotI sites of pVL1393-F2 to yield pVL1393-F2G5. pVL1393-F2 is an expression vector derived from pVL1393 (Pharmingen) and contains a fragment encoding the signal peptide of human immunoglobulin (MHFQVQIFSFLLISASVIMSRG) and FLAG peptide (DYKDDDDK).
We prepared recombinant viruses as described previously (1). Sf21 insect cells (Pharmingen) were infected with each individual recombinant virus at a multiplicity of infection of 10 and incubated at 27°C for 72 h to yield conditioned media containing recombinant ␤3Gn-T proteins fused with FLAG peptide. Bacu3GnT proteins were readily purified by anti-FLAG M1 antibody resin (Sigma) and eluted with 50 mM TBS (50 mM Tris-HCl, pH 7.4, 150 mM NaCl), 4 mM CaCl 2 buffer (pH 7.4). The recombinant proteins obtained in this system were named Bacu3GnT2, Bacu3GnT3, Bacu3GnT4, and Bacu3GnT5.
Bacu3GnT proteins separated by 10% SDS-polyacrylamide gel electrophoresis were transferred to an Immobilon PVDF membrane (Millipore, Bedford, MA). The membrane was probed with anti-FLAG monoclonal antibody, and stained with the ECL Western blotting detection reagents (Amersham Pharmacia Biotech). The intensity of positive bands on Western blotting was measured by densitometer (43) to determine the relative amounts of each Bacu3GnT protein.
Assaying of ␤3Gn-T Activity-Two types of each recombinant ␤3Gn-T, the soluble enzyme produced in the baculoexpression system and the membrane-bound form expressed in the cell homogenates of Namalwa transfectant cells and other cultured cell lines, were used to determine the relative activities of each ␤3Gn-T toward various substrates. Each soluble Bacu3Gn-T purified was adjusted to the same amount as described above and used for the assay. The various cultured cells were solubilized in a 20 mM HEPES buffer (pH 7.2) containing 2% Triton X-100, and 20 g of total protein in the cell homogenates was used for the enzyme reaction. Various oligosaccharides were fluorescently labeled with 2-aminobenzamide (2AB) or pyridylaminated (PA) (1,44) and used for acceptors.
The ␤3Gn-T activity was assayed in a 20-l reaction mixture containing 150 mM sodium cacodylate buffer (pH 7.2), 50 mM UDP-GlcNAc, 10 mM MnCl 2 , 0.4% Triton CF-54, 1 M 2AB-acceptor or PA-acceptor substrate, and the enzyme source. After incubation at 37°C for 16 h, the reaction was terminated by boiling for 3 min, and the mixture was diluted with 80 l of water. After centrifugation at 15,000 rpm for 5 min, the supernatant was filtered using an Ultrafree-MC column (Millipore). A 10-l aliquot of each supernatant was subjected to high performance liquid chromatography on a TSK-gel ODS-80T S QA column (4.6 ϫ 300 mm; Tosoh, Tokyo, Japan). The reaction products were eluted with 20 mM ammonium acetate buffer (pH 4.0) containing 7% methanol at a flow rate of 1.0 ml/min at 50°C and monitored with a fluorescence spectrophotometer, JASCO FP-920 (JASCO, Tokyo, Japan) (1,44).
We conducted an assay of the incorporation of radioactive sugar into glycolipid acceptor substrates. The ␤3Gn-T activity for glycolipids was assayed in a 25-l reaction mixture containing 150 mM sodium cacodylate buffer (pH 7.2), 480 M UDP-GlcNAc, 175 nCi of UDP-[ 14 C]GlcNAc (Amersham Pharmacia Biotech), 10 mM MnCl 2 , 0.4% Triton CF-54, 10 nmol of glycolipid acceptor substrate, and the enzyme source. After incubation at 37°C for 16 h, the reaction was terminated by boiling for 3 min, and 200 l of water containing 0.1 M KCl was added.
The reaction mixture was centrifuged at 15,000 rpm for 5 min. Radioactive glycolipid products were separated from the free radioactive UDP-[ 14 C]GlcNAc using a Sep-Pak Plus C 18 Cartridge (Waters). For conditioning, a Sep-Pak Plus C 18 Cartridge was washed with 10 ml of 100% methanol, and then washed twice with 10 ml of water. The supernatant of the reaction mixture was loaded on the equilibrated cartridge. Elution of the radioactive products was achieved using 5 ml of 100% methanol. Eluted products were dried with evaporator, and then the residues were dissolved in an adequate volume of 100% methanol. The dissolved residues were separated on a HPTLC plate with a solvent system of chloroform/methanol/0.2% CaCl 2 (65:35:8, v/v/v). The radioactive intensities of the bands were measured with a BAS 2000 Imaging Analyzer (Fuji Film, Tokyo, Japan).
Differentiation of HL-60 Cells in in Vitro Culture-Human myelogenous leukemia HL-60 cells were seeded at an initial density of 2 ϫ 10 5 cells/ml and incubated with either 1 M RA or 8 nM TPA (35). At various points over the course of the culture, the cells were harvested, and the total RNA in the cells was recovered for competitive RT-PCR.
Measurement of the Rat ␤3Gn-T5 Transcript in Developing Brain-Brains of SD rats (Charles River Japan, Tokyo, Japan) were separated into cerebral cortex and cerebellum, and the total RNA recovered from each tissue was subjected to competitive RT-PCR assay. Total RNA was extracted from a mixture of 4 -6 rat brains at each stage of development, and was used for cDNA synthesis.
The cDNA (738 bp; GenBank TM accession number AB045279) encoding the partial ORF of rat ␤3Gn-T5 was used as a standard, from which a competitor DNA was generated by deletion of a BsmBI fragment (185 bp). A primer set, a forward primer, 5Ј-CAAGATTTCACTGATTCTT-TCCAC-3Ј and a reverse primer, 5Ј-GTCCTGTAGGTCTTGTGAGT-GTCC-3Ј, was used for the competitive RT-PCR assay. The competitive RT-PCR experiment was performed twice on each sample.

RESULTS
A Novel cDNA Homologous to the Cloned ␤3Gal-Ts or ␤3Gn-Ts-A novel cDNA sequence encoding ␤3Gn-T motifs was found during the random sequencing of a rat shank cDNA library. We cloned a full-length human cDNA homologous (AB045278) to the rat cDNA (AB045279), and named it ␤3Gn-T5. The amino acid sequence homology in the corresponding region between human and rat was 82.9%. This indicated that the two cDNAs are orthologous.
The human ␤3Gn-T5 cDNA contains a full-length ORF encoding a protein of 378 amino acids, as shown in Fig. 1. ␤3Gn-T5 is predicted to be a typical type II membrane protein consisting of a NH 2 -terminal cytoplasmic domain of 12 residues, a transmembrane segment of 20 residues, and a stem region and catalytic domain of 346 residues. ␤3Gn-T5 had the three motifs typical of members of the ␤3Gal-T and ␤3Gn-T families. On ClustalW analysis (Fig. 1), four cysteine residues were found to be conserved in the four ␤3Gn-Ts which indicates that some of these cysteines are essential for maintenance of the tertiary structure of ␤3Gn-Ts. In the second ␤3Gn-T motif, a triplet of aspartic acid residues, DDD, which may be a dication binding site as proposed in a crystallization study of other glycosyltransferases (45), was conserved. Four possible N-glycosylation sites were found in the primary sequence of ␤3Gn-T5. One of them was conserved in all ␤3Gn-Ts. The ␤3Gn-T5 gene was found to be localized to a draft genome sequence (GenBank TM accession number AC025833) and the 3.7-kilobase pair cDNA containing the ORF is composed of a single exon.
On a phylogenetic tree (Fig. 2), the four members of the ␤1,3-N-Acetylglucosaminyltransferase Synthesizing Lc 3 Cer ␤3Gn-T family apparently formed a cluster which is separated from ␤3Gal-T members. ␤3Gn-T5 is positioned in the ␤3Gn-T family branch, however, it is in an outer branch away from the cluster of the other members. Three enzymes, ␤3Gn-T2, -T3, and -T4, form a subfamily in the phylogenetic tree and ␤3Gal-T4 and iGn-T also form a subfamily. The divergence of ␤3Gn-T5 occurred earlier than that of any other ␤3Gn-T member.
Quantitative Analysis of ␤3Gn-T5 Transcripts in Human Tissues and Various Cell Lines-As summarized in Fig. 3 and Table I, this gene was expressed in almost all tissues and cell lines with very few exceptions, although the expression level was different depending on the tissue and cell line. The tissues expressing ␤3Gn-T5 at a considerably high level were lung, colon, placenta, testis, pituitary gland, and cerebellum. Brain, liver (very low), spleen, lymph node, and thymus expressed ␤3Gn-T5 at a low level. Colonic adenocarcinoma, Colo205, SW620, lung cancer cell lines, EBC-1, HAL8, LX-1, PC-7, and RERF-LC-MS, and stomach cancer cell lines, KATOIII, MKN7, and HSC43 expressed the ␤3Gn-T5 transcripts at a high level. All neuroblastoma cells examined, except for NAGAI and GOTO cells, expressed the ␤3Gn-T5 transcript at a relatively high level. On the other hand, all leukemic cells derived from lymphocytes, except for NALL-1 (lymphoblastic leukemia), expressed ␤3Gn-T5 almost at an undetectable level. U937 (monocyte-like) and HL-60 (promyelocytic leukemia) cells expressed it at a considerable level. HepG2 cells did not express it at all. The expression levels in the cultured cells reflected those in the original tissues derived therefrom.
Relative Activities of Four Recombinant ␤3Gn-Ts, Which Were Produced as Truncated Forms of Fusion Protein with the FLAG Peptide in a Baculoexpression System, Toward Oligosaccharides-The recombinant enzymes of four ␤3Gn-Ts, ␤3Gn-T2, -T3, -T4, and -T5, produced as fusion proteins with a FLAG peptide in the baculoexpression system were named Bacu3GnT2, -3GnT3, -3GnT4, and -3GnT5, respectively. The amounts of each enzyme were made equal for assaying the ␤3Gn-T activity.
We observed no galactosyltransferase activity of the four ␤3Gn-Ts toward LNnT-PA and agalacto-LNnT-PA (data not shown). The ␤3Gn-T activity of Bacu3GnT2 toward LNnT-PA was strongest among all combinations of the enzyme and the substrate. Thus, the activity of Bacu3GnT2 toward LNnT-PA is presented as 100%, and all other activities are given as relative values.
Bacu3GnT2 exhibited relative activity toward the following ␤1,3-N-Acetylglucosaminyltransferase Synthesizing Lc 3 Cer tivity for LNT-PA and LNFP-V-PA. Bacu3GnT2 and -3GnT5 also exhibited weak activity for LNFP-III-PA. LNFP-II-PA and LNDFH-II-PA could not be utilized as acceptor substrate for any Bacu3GnT examined. Bacu3GnT4 activity was almost undetectable for all substrates except a very faint activity for LNFP-V-PA. In a previous study, we confirmed that Bacu3GnT3 and Bacu3GnT4 apparently exhibited detectable levels of ␤3Gn-T activity toward LNnT-PA and LNT-PA because we used an excess of recombinant enzyme for the reaction (1).
The oligosaccharide substrates having the polylactosamine structures, repeats of units of lactosamine (Gal␤1-4GlcNAc; LN), were labeled with 2AB and used as acceptor substrates (44). The LNnT-2AB oligosaccharide was used as a control .  2LN, 3LN, 4LN, and 5LN in Table II indicate that each oligosaccharide has 2-, 3-, 4-, or 5-repeating lactosamine (LN) units, respectively. The ␤3Gn-T activity toward LNnT-2AB of Bacu3GnT2 was again the strongest among all combinations of enzyme and substrate, therefore its activity is expressed as 100%, and the activities of the other combinations are expressed relative to this value in Table II. Bacu3GnT2 transferred a GlcNAc with almost the same level of activity to all polylactosamine substrates regardless of the number of LN units. Bacu3GnT3 exhibited low, but apparently positive activity for all lengths of polylactosamine substrate. The activity of Bacu3GnT4 was again hardly detected with the amount of recombinant protein used in the present study. Interestingly, Bacu3GnT5 preferred the shorter substrates, i.e. 2LN-2AB and 3LN-2AB. The activities of Bacu3GnT5 for the longer polylactosamine chains, 4LN-2AB and 5LN-2AB, were almost onetenth of those for the shorter chains.
Relative Activities of Four ␤3Gn-Ts, Which Were Produced in  As seen in Fig. 4, the homogenates of Namalwa-3GnT5 cells exhibited strong activity for the synthesis of both Lc 3 Cer and nLc 5 Cer. The homogenates of HL-60 cells also showed positive activities toward both acceptors, LacCer and nLc 4 Cer. Relative activities were obtained by measurement of the positive bands in Fig. 4. The activity of Namalwa-3GnT5 toward LacCer is presented as 100%, and all other activities are given as relative values. Namalwa-3GnT5 exhibited strong activity toward Lac-Cer resulting in the synthesis of Lc 3 Cer (100%), however, the other three ␤3Gn-Ts showed no activity toward LacCer. The wild-type Namalwa cells and the mock-transfected Namalwa cells showed faint activity, 2.6 and 2.5%, respectively, for the synthesis of nLc 5 Cer. The other transfectant cells did not show an increase in either activity, although the transcript for the transfected gene was apparently overexpressed in the cells. Their activities were 3.2, 2.8, and 2.4%, respectively. In contrast, Namalwa-3GnT5 exhibited strong activity toward nLc 4 Cer. The relative activity of Namalwa-3GnT5 for nLc 5 Cer synthesis was almost the same, 105.2%, as that for Lc 3 Cer synthesis.
Second, we measured the relative activities of the four Bacu3GnTs, three of which, Bacu3Gn-T2, -T3, and -T4, were the same recombinant enzymes used in the previous experiments (1) toward glycolipid acceptors. Bacu3GnT5 was prepared in this study. As summarized in Table III, the activity of Bacu3GnT5 toward LacCer is presented as 100%. Bacu3GnT5 exhibited very strong activities toward both LacCer and nLc 4 Cer in comparison with the other enzymes. Interestingly, Bacu3GnT5 preferred nLc 4 Cer over LacCer as an acceptor, and transferred GlcNAc to nLc 4 Cer 4.67-fold more efficiently than to LacCer. Bacu3GnT5 did not utilize galactosylceramide (Gal-Cer) as an acceptor. Bacu3GnT2 exhibited considerable activity toward nLc 4 Cer, 21.8%, but not toward LacCer. Bacu3GnT3 showed very faint activity toward nLc 4 Cer, 0.8%. Again, Bacu3GnT4 did not show any activity toward the glycolipid acceptors.

Immuno-TLC Analysis of Glycoshingolipids (GSLs) Extracted from Namalwa Cells Transfected with Each ␤3Gn-T
Gene-As seen from the orcinol staining results of neutral GSLs (Fig. 5A), among multiple bands of GSLs, the band intensity of LacCer of Namalwa-3GnT5 apparently decreased as compared with that of the other transfectants. By immunostaining using the 1B2 mAb (Fig. 5B), positive bands were detected in all transfectants including the wild-type and mock transfectant cells. However, the nLc 4 Cer band of Namalwa-3GnT5 cells showed a very strong intensity as compared with that of the other transfectants. The positive band faintly detected below that of nLc 4 Cer of Namalwa-3GnT5 cells corresponded to nLc 6 Cer. It is known that mAb 1B2 reacts with both nLc 4 cer and nLc 6 Cer (37). The above results are interpreted as follows. The overexpressed ␤3Gn-T5 in the cells consumed the substrate, LacCer, and thereafter, the product Lc 3 Cer was galactosylated by (an) endogenous ␤1,4-galactosyltransferase(s) to produce nLc 4 Cer, because Namalwa cells are known to endogenously possess an excess amount of ␤4Gal-T1. Some of the nLc 4 Cer produced in the cells was further converted to nLc 5 Cer by the overexpressed ␤3Gn-T5 and again galactosylated to produce nLc 6 Cer by endogenous ␤4Gal-T(s).
Correlation of Lc 3 Cer Synthesizing Activity with the Amount of ␤3Gn-T5 Transcript in Various Cultured Tumor Cells-The Lc 3 Cer synthesizing activity was highest in KATOIII cells among various cultured cancer cells examined in this study (Table IV). Thus, the activity of KATOIII was set as 100%, and the values for Lc 3 Cer synthesizing activity of the other cells were calculated relative to this (see Table IV). The level of Lc 3 Cer synthesizing activity was almost correlated with the amount of ␤3Gn-T5 transcript (the mathematical correlation coefficient is 0.83), but not with the levels of the other ␤3Gn-T transcripts (the mathematical correlation coefficients of other ␤3Gn-Ts are all under 0.45). KATOIII cells possessed the most ␤3Gn-T5, followed by Colo205, EBC-1, LS180, and AOI cells in decreasing order both for activity and the amount of transcript.

␤1,3-N-Acetylglucosaminyltransferase Synthesizing Lc 3 Cer
However, SK-N-SH cells are an exception which exhibited low activity, despite the relatively high expression of ␤3Gn-T5 transcripts. We do not know why the Lc 3 Cer synthesizing activity of SK-N-SH cells conflicted with the amount of ␤3Gn-T5 transcript. Some cancer cells, HepG2, Jurkat, Daudi, Ramos, and GOTO cells, did not express the ␤3Gn-T5 gene at all, and exhibited no activity of Lc 3 Cer synthase.
On the other hand, Colo205 cells, exhibiting weaker activity than KATOIII cells, expressed larger amounts of ␤3Gn-T2 transcript than KATOIII cells. HepG2 cells expressed considerable amounts of ␤3Gn-T2 and -T3 transcripts, but they exhibited no activity of Lc 3 Cer synthase. ␤3Gn-T4 was expressed at a very low or undetectable level in all cell lines examined. The above results indicated that the Lc 3 Cer synthesizing activity is mainly directed by ␤3Gn-T5 in these cell lines.
Expression of the Human ␤3Gn-T5 Transcripts during Differentiation in the Human Promyelocytic Cell Line HL-60 -As seen in Fig. 6, the expression level of the ␤3Gn-T5 transcript during the granulocytic differentiation by RA treatment apparently increased in comparison to that of the wild-type HL-60 cell. On the other hand, it markedly declined during the monocytic differentiation induced by TPA treatment.
Change of the Transcript Level of ␤3Gn-T5 in the Developing Rat Brain-As shown in Fig. 7, in the rat cerebral cortex, the expression level of the ␤3Gn-T5 transcript peaked at around ED19, and had almost completely disappeared by PD14. The adult cerebral cortex expressed no ␤3Gn-T5 transcript. In contrast, cerebellum showed biphasic changes in the transcript level. The level of ␤3Gn-T5 transcript decreased from ED19 to PD3, and then increased until PD8. Subsequently, the level again decreased till PD14. In the adult cerebellum, the expression level was comparatively high. DISCUSSION Lc 3 Cer synthase plays a key role in the control of carbohydrate synthesis in GSLs during cell differentiation and development. As demonstrated in the present study, ␤3Gn-T2, -T3, and -T4 are not the Lc 3 Cer synthase. We did not examine activities of the recombinant enzymes of iGn-T. However, the transcript levels of iGn-T in various tumor cell lines were not correlated with the Lc 3 Cer and nLc 5 Cer synthesizing activities in the respective cell lysates (data not shown). So, we can rule out iGn-T as a candidate for the Lc 3 Cer synthase. We concluded that ␤3Gn-T5 is the most feasible candidate for the following reasons. 1) Bacu3GnT5 and the homogenates of Namalwa-3GnT5 cells exhibited strong activity to synthesize Lc 3 Cer in vitro. 2) LacCer was consumed to be converted to neolactoseries GSLs in the Namalwa-3GnT5 cells.
3) The transcript levels of ␤3Gn-T5 in various tissues and cell lines were consistent with the Lc 3 Cer synthesizing activity as reported previously. 4) The expression level of the ␤3Gn-T5 transcript in various cultured cancer cells was well correlated with the Lc 3 Cer synthesizing activity. 5) The changes in the ␤3Gn-T5 transcript level during HL-60 differentiation and during rat brain development were consistent with those of the Lc 3 Cer synthesizing activity as reported by others (15,20,35,46).
The truncated enzyme expressed in the insect cells preferred nLc 4 Cer which has a longer carbohydrate chain than LacCer, while the membrane-bound form exhibited almost the same activity toward both substrates. The difference in activity between the two forms may be related to structural difference or the presence of detergent in the reaction mixture. Glycosyltransferases are the Golgi enzymes bound to the Golgi membrane by the transmembrane domain. We assume that the truncated form, which is soluble due to the absence of transmembrane domain, may easily access GSLs with a long carbohydrate chain, which are more hydrophilic than GSLs with shorter carbohydrate chains. The membrane-bound form expressed in Namalwa cells probably has more physiological activity than the truncated form. In the reaction mixture for ␤3Gn-T assay, the membrane-bound form probably exists as a micellar penetrating the Golgi membrane. It is likely that the membrane-bound enzyme rather than the truncated soluble form interacts with glycolipid substrates.
The homogenates of Namalwa-3GnT5 cells exhibited strong activity for both GSL substrates, LacCer and nLc 4 Cer, as well as toward LNnT-2AB and the two shorter polylactosamine chains, 2LN-2AB and 3LN-2AB. This indicated that ␤3Gn-T5 effectively recognizes a polylactosamine structure within two units of lactosamine. The marked reduction in the ␤3Gn-T5 activity for the longer polylactosamine chain would suggest that GSLs with a long polylactosamine chain, such as nLc 6 Cer, are not good substrates for ␤3Gn-T5. In previous studies (15,20,46), the activity of ␤3Gn-T toward LacCer and nLc 4 Cer was measured using tissue homogenates of rat brain during development. Both activities showed not only very similar profiles of change during the development, but almost the same level. This may be consistent with the present results showing that the two activities are directed by a single enzyme, ␤3Gn-T5. The wild-type and mock-transfected Namalwa cells showed weak ␤3Gn-T activity for nLc 4 Cer, but no activity for LacCer. We could not identify which enzyme directs this activity in the wild-type Namalwa cells. The activity may be directed by endogenous ␤3Gn-T2 or there may be some unknown ␤3Gn-T(s) in the Namalwa cells.
The Lc 3 Cer synthase is a key enzyme in the expression of a series of neolactoglycolipids, i.e. nLc 4 Cer and its derivatives. In particular, the expression level of two SGGLs, SGGL-1 and SGGL-2, carrying the HNK-1 epitope is determined by the Lc 3 Cer synthase (15, 19 -22, 46). The change in the level of ␤3Gn-T5 transcript in developing rat brain almost paralleled that in Lc 3 Cer synthesizing activity reported previously (15,20,46). This strongly indicated that ␤3Gn-T5 is the Lc 3 Cer synthase. To confirm this, we will examine whether or not ␤3Gn-T5 is co-localized with HNK-1 on SGGLs and CD15 on neolacto-series GSLs in a future study. Lc 3 Cer synthase is also an important enzyme in hematopoietic cell differentiation. The changes in the level of ␤3Gn-T5 transcript during HL-60 differentiation shown in Fig. 6 are consistent with the results of Nakamura et al. (35). Stults et al. (36) reported that lymphoid cell lines lack Lc 3 Cer synthesizing activity, but possess nLc 5 Cer synthesizing activity, whereas myeloid cell lines express both activities. Almost all lymphoid cell lines we examined in this study showed very low or undetectable levels of ␤3Gn-T5 transcript, while HL-60 cells (promyelocytic leukemia) and U937 cells (monocyte-like) expressed substantial amounts of ␤3Gn-T5. This again supported that ␤3Gn-T5 is responsible for the synthesis of Lc 3 Cer in hematopoietic cells.
Holmes (50) reported the ␤3Gn-T activities for the synthesis of GSL in the cells homogenates of a human colon cancer cell line, SW403. The SW403 cell homogenates showed almost the same level of ␤3Gn-T activity toward two GSL acceptors, Lac-  Cer and nLc 4 Cer. This could be interpreted to mean that both activities were directed by a single enzyme, ␤3Gn-T5, as in the case of rat brain. In the present study, we demonstrated that recombinant ␤3Gn-T5 effectively catalyzes the biosynthesis of both Lc 3 Cer and nLc 5 Cer at almost the same efficiency. There is some controversy (52) over whether a single enzyme, ␤3Gn-T5, synthesizes both Lc 3 Cer and nLc 5 Cer in colon tissue. Based on the Basu study (52), there appear to be unknown ␤3Gn-Ts in colon tissue, which differentially catalyze the synthesis of Lc 3 Cer or nLc 5 Cer. Another controversy with the present study relates to the report by Holmes et al. (51) in which significantly high activity of Lc 3 Cer synthase was detected in colonic adenocarcinoma tissues of patients and cell lines derived therefrom, but in contrast, the activity was undetectable in normal colonic epithelial cells. In the present study, all colonic adenocarcinoma cell lines expressed substantial amounts of ␤3Gn-T5 transcript. This is consistent with the results of Holmes et al. (51). However, normal colon tissue also expressed substantial amounts of the transcript. We will examine whether or not the ␤3Gn-T transcripts are markedly up-regulated in the colorectal cancer tissues of patients.
Regarding the other ␤3Gn-Ts examined in this study, ␤3Gn-T2 was most active toward polylactosamine acceptors, and it effectively extended the polylactosamine chain even on the longer chain acceptors. Thus, ␤3Gn-T2 is the most probable candidate for the enzyme which functions to extend the polylactosamine chain. ␤3Gn-T5 is involved only in the synthesis of short polylactosamine chains or initiation of polylactosamine synthesis. ␤3Gn-T3 and ␤3Gn-T4 exhibited very little activity, almost undetectable, toward all substrates examined. They apparently showed positive ␤3Gn-T activity in a previous study (1), because excess amounts of enzyme were used for the assay. The native acceptor substrates for ␤3Gn-T3 and -T4 may be different from the Gal residue acceptor, and some unknown native acceptors may exist for ␤3Gn-T3 and -T4.
The results of the present study strongly indicated that ␤3Gn-T5 is the most feasible candidate for Lc 3 Cer synthase. In the future, we will assess the biological functions of this synthase, the key enzyme determining the expression of biologically functional epitopes on GSLs.