Cloning, Expression, and Characterization of a Novel UDP-galactose:β-N-Acetylglucosamine β1,3-Galactosyltransferase (β3Gal-T5) Responsible for Synthesis of Type 1 Chain in Colorectal and Pancreatic Epithelia and Tumor Cells Derived Therefrom*

The sialyl Lewis a antigen is a well known tumor marker, CA19-9, which is frequently elevated in the serum in gastrointestinal and pancreatic cancers. UDP-galactose:N-acetylglucosamine β1,3-galactosyltransferase(s) (β3Gal-Ts) are required for the synthesis of the sialyl Lewis a epitope. In the present study, a novel β3Gal-T, named β3Gal-T5, was isolated from a Colo205 cDNA library using a degenerate primer strategy based on the amino acid sequences of the four human β3Gal-T genes cloned to date. Transfection experiments demonstrated that HCT-15 cells transfected with the β3Gal-T5 gene expressed all the type 1 Lewis antigens. In gastrointestinal and pancreatic cancer cell lines, the amounts of β3Gal-T5 transcripts were quite well correlated with the amounts of the sialyl Lewis a antigens. The β1,3Gal-T activity toward agalacto-lacto-N-neotetraose was also well correlated with the amounts of β3Gal-T5 transcripts in a series of cultured cancer cells, and in Namalwa and HCT-15 cells transfected with theβ3Gal-T5 gene. Thus, the β3Gal-T5 gene is the most probable candidate responsible for the synthesis of the type 1 Lewis antigens in gastrointestinal and pancreatic epithelia and tumor cells derived therefrom. In addition, β3Gal-T5 is a key enzyme that determines the amounts of the type 1 Lewis antigens including the sialyl Lewis a antigen.

CA19-9 in serum is a well known tumor marker, which is frequently used for the clinical diagnosis of cancer, in particular, colorectal, pancreatic, and gastric cancers (1,2). The 1116NS19-9 (19-9) 1 monoclonal antibody detects a CA19-9 an-tigen, of which the antigenic epitope has been defined as the carbohydrate structure of sialyl Lewis a (sLe a ) (2)(3)(4). Besides its usefulness as a tumor marker, sLe a antigen is known to be a ligand for selectins (5,6). Clinical statistical analysis demonstrated that cancer patients who express abundant sLe a antigens have a worse prognosis as to liver metastasis than patients who do not express sLe a antigens (7,8). Thus, it is of interest that sLe a antigens may confer some metastatic capacity on cancer cells.
Regarding ␤3Gal-Ts, we have reported for the first time the cloning of a ␤3Gal-T gene from human WM266 -4 melanoma cells using an expression cloning method (10). The recent rapid growth of data bases of expressed sequence tags (ESTs) and the Human Genome Project enabled us to find novel genes homologous to the original one. Thus, three human ␤3Gal-T genes homologous to the original one were cloned very recently (11,12). The four ␤1,3GalTs, including the original one, are named ␤3Gal-T1 to -T4 (12). Expression studies on the four human ␤3Gal-Ts demonstrated that two of them, ␤3Gal-T1 and T2, apparently transfer Gal to GlcNAc with a ␤1,3-linkage resulting in type 1 chain synthesis, but ␤3Gal-T4 transfers Gal to an N-acetylgalactosamine (GalNAc) residue, resulting in the synthesis of the type 3 chain, Gal␤1,3GalNAc (12). The human ␤3Gal-T4 did not transfer Gal to a GlcNAc residue for the type 1 chain synthesis (12). The human ␤3Gal-T4 is likely to be the human homologue of the rat G M1 /G D1 synthase (13), since the amino acid sequence of human ␤3Gal-T4 shows very high homology, 79.4%, to that of the rat G M1 /G D1 synthase, and the human ␤3Gal-T4 apparently transfers Gal to the GalNAc residue of asialo-GM2 and GM2, resulting in the asialo-G M1 and G M1 synthesis, respectively (12). The activity of human ␤3Gal-T3 has not been detected toward any of the acceptor substrates used in their study (12). Three mouse ␤3Gal-T genes have been cloned and named m␤3GalT-I, m␤3GalT-II, and m␤3GalT-III, corresponding to human ␤3Gal-T1, ␤3Gal-T2, and ␤3Gal-T3, respectively (14). m␤3Gal-TII and m␤3GalT-III were found to exhibit the ␤3Gal-T activity toward both GlcNAc and GalNAc residues; however, they showed quite low activities for the type 1 chain synthesis, i.e. about 3% of the activity of m␤3GalT-I (14).
It has not been elucidated which ␤3Gal-T determines the expression of the sLe a epitopes in gastrointestinal and pancreatic cancers. The tissue distributions of the four ␤3Gal-Ts were determined by Northern analysis (11,12), it being found that neither ␤3Gal-T1 nor -T2 is expressed in the pancreas, which indicated that there may be unknown ␤3Gal-T(s) synthesizing the type 1 chain in the pancreas. They did not examine the expression of those ␤3Gal-Ts in the gastrointestinal tissues, such as colon and stomach, which frequently produce the sLe a antigens when they become cancerous.
In this study, we first noticed that none of the four human ␤3Gal-Ts cloned to date, ␤3Gal-T1 to -T4, is responsible for the sLe a expression in gastrointestinal and pancreatic cancers, and successfully cloned a novel ␤3Gal-T gene, named ␤3Gal-T5, from Colo205 cells. ␤3Gal-T5 is the most probable candidate participating in the synthesis of the sLe a epitopes, i.e. CA19-9 antigens, in gastrointestinal and pancreatic cancer cells.
Cloning of the Four Cloned ␤3Gal-T Genes from Various Human cDNA Libraries and Construction of Expression Plasmids-The cDNA encoding ␤3Gal-T1 was cloned by the expression cloning method used in our previous study (10). We found three sequences homologous to that of ␤3Gal-T1 in the EST data bases. The full-length cDNAs encoding the other three homologous sequences were cloned from various human cDNA libraries using probes encoding the fragment sequences in the EST data base. They were identical to ␤3Gal-T2, -T3, and -T4, which were reported by Amado et al. (12), and Kolbinger et al. (11).
Construction of a cDNA Library from Colo205 Cells-Total cellular RNA was isolated from Colo205 cells using the acid guanidium thiocyanate-phenol-chloroform method (19). Poly(A) ϩ -rich-RNA was isolated with Oligotex TM -dT30 (Super) (Roche, Tokyo, Japan). Complementary DNAs were synthesized with oligo(dT) primers from poly(A) ϩ -rich-RNA using a Superscript Choice System for cDNA Synthesis (Life Technologies, Inc.). A cDNA library was constructed by inserting size-fractionated cDNAs (more than 1.5 kilobase pairs) into an expression vector, pAMo, using SfiI adaptors (17,18). We obtained about 1 ϫ 10 6 inde-pendent clones as a cDNA library and extracted plasmid DNAs from the library.
PCR for Cloning of a Fragment Encoding a Novel ␤3Gal-T-On alignment of the amino acid sequences of the four cloned ␤3Gal-Ts, we found conserved amino acid sequences at three positions, and named them Motifs 1, 2, and 3 (Table I) The cDNAs of the Colo205 cDNA library described above were used as templates for PCR amplification to obtain a DNA fragment. Two PCRs were performed with two sets of degenerate primers, respectively, i.e. the first PCR was performed with primers at-1 and at-2, and the second PCR with primers at-3 and at-4. The amplified PCR products were inserted into a pBluescript SK (Ϫ) (pBS) vector (Stratagene, La Jolla, CA), and the DNA fragments obtained were sequenced by the dideoxynucleotide chain termination method using an ALF DNA sequencer (Amersham Pharmacia Biotech, Uppsala, Sweden). Two fragment DNAs contained novel nucleotide sequences, which, however, were homologous to the corresponding regions of the cloned ␤1,3GalT genes. On an additional PCR involving Colo205 cDNAs as templates using primers encompassing the two fragment sequences, both fragments were found to be encoded by one species of cDNA.
Cloning of Full-length cDNAs Encoding a Novel ␤3Gal-T-The two DNA fragments obtained through the two PCRs, i.e. those with the Motif 1 and 2 primers, and the Motif 2 and 3 primers, respectively, were mixed and used as the probe for hybridization to isolate full-length cDNA clones. We screened the Colo205 cDNA library and isolated several distinct clones having inserts of different sizes. All inserts encoded the same sequence of one species of cDNA, this sequence being found to be homologous to those of the known four ␤3Gal-Ts. Thus, we named this novel gene the human ␤3Gal-T5 gene. After the cDNA sequences had been completed, we searched the data base of the Human Genome Project to determine whether the same sequence or homologous ones were registered or not. We found a genome sequence completely identical to the cDNA sequences in the data base, which was very recently registered (June 2nd, 1998). Its registration number is AF064860. By comparison between the cDNA sequences and the genome one, we determined the genomic organization of the ␤3Gal-T5 gene.
Quantitative Analysis of the five ␤1,3GalT Transcripts in Human Tumor Cell Lines and Human Tissues by Competitive RT-PCR-The principle of the competitive RT-PCR method was described in detail in our previous papers (9,18). Competitor DNA plasmids each carrying a small deletion within the respective full-length ORF cDNA were constructed by appropriate restriction endonuclease digestion as shown in Table II. For instance, a competitor DNA plasmid of the ␤3Gal-T1 gene was prepared by deleting the 212-bp BanII-EcoRV fragment from the standard plasmid DNA containing the full-length cDNA of ␤3Gal-T1.
Total cellular RNA was isolated from various tumor cell lines and human tissues. Complementary DNAs were synthesized with an oligo(dT) primer from 6 g of DNase I-treated total RNA in a 20-l (total The competitive RT-PCR was performed with AmpliTaq Gold TM (Perkin Elmer) in a 50-l (total volume) reaction mixture comprising 10 l of standard plasmid DNA or sample cDNA, 10 l of competitor DNA at the optimal concentration, which differs with the transcript, and 0.2 M amounts of each primer of the gene-specific primer sets listed in Table  II. The PCR buffer for the competitive RT-PCR comprised 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 1.5 mM MgCl 2 , 0.2 M of each dNTP, and 0.001% (w/v) gelatin. PCR was performed with a pre-PCR heat step at 95°C for 11 min, followed by the optimal number of PCR cycles, each of which comprised 1 min at 95°C, 1 min at the optimal annealing temperature (Table II), and 2 min at 72°C. After the competitive RT-PCR, a 10-l aliquot was electrophoresed in a 1% agarose gel and the bands were visualized by ethidium bromide staining. The intensities of the amplified fragments were quantified by scanning positive pictures using the public domain NIH Image program. 2 Measurement of the ␤-actin transcript in each sample was performed using the same competitive RT-PCR method as for the ␤3Gal-T transcripts. Each value for the ␤3Gal-T and ␤-actin transcripts was plotted on the respective standard curve to obtain the actual amount of each transcript. The actual amount of each ␤3Gal-T transcript was divided by that of ␤-actin for normalization.
Transfection Experiments to Express the Five Human ␤3Gal-T Genes in Namalwa (Burkitt Lymphoma) and HCT-15 Cells-Each of the five ␤3Gal-T genes subcloned into the pAMo vector was stably transfected by the electroporation method into Namalwa or HCT-15 cells. These cells were selected in the presence of Geneticin (G418) (Life Technologies, Inc.) at a concentration of 0.8 mg/ml in RPMI 1640 medium supplemented with 10% heat-inactivated fetal bovine serum. Stable transformant cells were obtained after 25 days of exposure to Geneticin. Cell homogenates of stable transformants expressing each of the ␤3Gal-T genes were subjected to assaying of ␤3Gal-T activity. The levels of the transcripts expressed in the transformant cells were measured by means of competitive RT-PCR to normalize the ␤3Gal-T activity. The stable transformants of HCT-15 cells were subjected to limiting dilution to obtain single transformant clones.
Flow Cytometry Analysis-The expression of type 1 Lewis antigen epitopes, i.e. the Le a , Le b , sLe c , and sLe a epitopes, on the surface of the cultured tumor cells and the cells transfected with each of the ␤3Gal-T genes was examined by flow cytometry analysis using an Epics Elite (Coulter, Tokyo, Japan). The transfected cells (1 ϫ 10 6 ) were incubated with a first antibody (10 g/ml) for 1 h on ice, and then washed twice with PBS (pH 7.4) containing 1% BSA and 0.1% sodium azide, followed by incubation with fluorescein isothiocyanate-conjugated anti-mouse IgM or IgG (Bio-Rad). Then, the cells were washed again with PBS-BSA and finally subjected to flow cytometry analysis.
Western Blotting Analysis-Cell pellets were solubilized in 20 mM HEPES buffer (pH 7.2) containing 2% Triton X-100 by brief sonication. Proteins separated on 6% SDS-polyacrylamide gel electrophoresis were transferred to an Immobilon PVDF membrane (Millipore, Bedford, MA) in a Transblot SD cell (Bio-Rad). The membrane was blocked with PBS containing 5% skim milk at 4°C overnight and then incubated with 10 g/ml 19-9. The membrane was stained according to the manual with the ECL Western blotting detection reagents (Amersham Pharmacia Biotech).

Discrepancy between the Expression Levels of three ␤3Gal-T Transcripts, i.e. the ␤3Gal-T1, -T2, and -T3 Transcripts, and the Amounts of Type 1 Lewis Antigens Expressed in Various
Tumor Cells-Various tumor cell lines derived from different human tissues were examined as to the transcript levels of the four ␤3Gal-T genes that were cloned previously, and their expression levels were compared with the amounts of type 1 Lewis antigens, i.e. the sLe a , Le a (7LE), and Le b (TT42) antigens, expressed in these cancer cells (Fig. 1). Flow cytometry analysis revealed that Colo205, Colo201, and SW1116 (colon cancer) cells, and Capan-2 (pancreatic cancer) cells expressed large amounts of type 1 Lewis antigens, these results being consistent with those of a previous study (20). All the above cell lines except for Capan-2 were strongly stained with the three antibodies, i.e. 19-9 (anti-sLe a ), 7LE (anti-Le a ), and TT42 (anti-Le b ) antibodies. Capan-2 cells were also strongly stained with 19-9, but not with 7LE or TT42. However, these four types of cells did not express ␤3Gal-T1 or ␤3Gal-T2 transcripts at all. The expression of ␤3Gal-T1 was abundantly detected in PC-1 (lung cancer) cells, and faintly detected in Jurkat (T cell leukemia) and PC-3 (prostatic cancer) cells. ␤3Gal-T2 was expressed in Namalwa and SK-N-MC (neuro- We observed a significant discrepancy between the expression of the three ␤3Gal-Ts, i.e. ␤3Gal-T1, -T2, and -T3, and the expression of the type 1 Lewis antigens in these cell lines. In contrast, the expression of ␤3Gal-T4 appeared to be correlated with the type 1 Lewis antigen expression, i.e. the cells expressing type 1 Lewis antigens, i.e. Colo201, Colo205, SW1116, HT-29, and Capan-2 cells, also expressed substantial amounts of the ␤3Gal-T4 transcript. ␤3Gal-T1, -T2, and -T3 could synthesize the type 1 chain (11,12,14); however, they were not correlated with the expression of type 1 Lewis antigens in the present study. The expression of ␤3Gal-T4 seemed to be correlated with the expression of type 1 Lewis antigens in some tumor cells; however, it could not synthesize the type 1 chain (12). From these results, we concluded that none of the four ␤3Gal-Ts is responsible for type 1 chain synthesis, resulting in sLe a (CA19-9) antigen expression, in gastrointestinal and pancreatic cancer cells.
Cloning and Sequence of a Novel cDNA Homologous to the Cloned ␤3Gal-Ts-As described under "Experimental Procedures," we obtained two DNA fragments encoding novel sequences, which, however, are homologous to the corresponding regions of the four cloned ␤3Gal-T genes. The sequences of the two DNA fragments were found to be encoded by a single cDNA species. We named this gene ␤3Gal-T5. By use of the DNA fragments as probes, full-length cDNA clones were obtained from the Colo205 cDNA library. Complementary DNA sequenc-  ing analysis revealed that the ␤3Gal-T5 cDNA contains an ORF encoding a protein of 310 amino acids (Fig. 2). The position of the AUG start codon was assigned according to the Kozak consensus sequence (21). A hydropathy profile based on the Kyte and Doolittle method (22) indicated that the ORF encodes a type II membrane protein, which is a typical feature of glycosyltransferases (data not shown). The three motifs of amino acid sequences, Motifs 1, 2, and 3, which we employed for the design of degenerate primers in this study, were conserved in the sequence of ␤3Gal-T5. Four cysteine residues were conserved in the five ␤3Gal-Ts, which indicates that some of these cysteines are essential for maintenance of the tertiary structures of ␤3Gal-Ts. Fifty-four of the 310 amino acid residues of ␤3Gal-T5 were conserved in comparison with the sequences of the other four ␤3Gal-Ts. Three possible N-glycosylation sites were found in the primary sequence of ␤3Gal-T5.
Genomic Structure of the ␤3Gal-T5 Gene and Alternatively Spliced Isoforms of Transcripts-By comparison of the fulllength cDNA sequence with the genome sequence, which has been registered in the Genome Project Database (registration no. AF064860), the chromosomal localization and the genomic structure of the ␤3Gal-T5 gene were determined (Fig. 3A). According to the description in GenBank, this gene is localized to human chromosome 21q22.3. The ORF of the ␤3Gal-T5 gene was found to be encoded by a single exon, as in the cases of the four cloned ␤3Gal-T genes. The A nucleotide of the translation initiation codon, ATG, was found to be the first nucleotide of exon 4 encoding the ORF.
The sequence of the 5Ј-flanking region was extended by the 5Ј-RACE method using Colo205 transcripts. All subclones obtained with the 5Ј-RACE method had a common nucleotide sequence at the 5Ј end, i.e. all of them started at nucleotide position 85153 in the genome sequence AF064860 (Table III). Thus, five isoforms of ␤3Gal-T5 transcripts, isoforms 1, 2, 3, 4, and 5, were identified with the 5Ј-RACE method (Fig. 3B). All intron sequences at the exon-intron junctions complied with the acceptor and donor site sequences of splicing rule, i.e. the GT-AG rule (Table III). Each transcript of the five isoforms in Colo205 cells was quantified by determining the intensity of its band on a RT-PCR gel (Fig. 3C). As shown in Fig. 3A, exon 2 or 3 contains an XbaI or BsmI restriction site, respectively. The amplified bands on RT-PCR were digested with XbaI or BsmI to confirm exon 2 or 3, respectively. As shown in Fig. 3C, two bands were obtained on RT-PCR using the primer set, si-2 and si-1. The upper and lower bands correspond to isoform 1, consisting of exons 1, 3, and 4, and isoform 2, consisting of exons 1Ј, 3, and 4, respectively, and both materials were digested by BsmI, but not by XbaI. The ratio of the band intensities of isoforms 1 and 2 was about 1 to 1. The other three isoforms, isoforms 3, 4, and 5, were not detected on this RT-PCR, which ctctctagAGAACCCT GTTTGGAGgtagggct indicated that isoforms 3, 4, and 5 are minor transcripts in Colo205 cells. RT-PCR with primers si-4 and si-1 gave a single band for isoform 3, which was digested by both BsmI and XbaI. A band for isoform 4 was not detected on this RT-PCR, because the amount of isoform 4 transcripts may be very small. RT-PCR with primers si-2 and si-3 gave a single band for isoform 5. Isoform 1 and 2 were abundant among the five isoforms, and both isoforms amounted to approximately 50% of the total transcripts of the ␤3Gal-T5 gene, respectively. The other three isoforms, i.e. isoforms 3, 4, and 5, were only present in trace amounts.

Correlation of the Expression Levels of the ␤3Gal-T5 Transcripts with the Amounts of Type 1 Lewis Antigens in Various
Tumor Cells-As can be seen in Fig. 4, the expression levels of ␤3Gal-T5 transcripts and the amounts of CA19-9 antigens in various cancer cells were determined by the competitive RT-PCR method and Western blot analysis, respectively. The results of flow cytometry analysis in Fig. 1 are well consistent with those of Western blot analysis in this section, i.e. the four types of cells, i.e. Colo205, Colo201, SW1116, and Capan-2 cells, that were stained strongly with 19-9 on flow cytometry also gave strong positive bands, which were smear ones with high molecular weights indicating they are mucins, with 19-9 on Western blotting (Fig. 4). These four cell lines also expressed abundant ␤3Gal-T5 transcripts (Fig. 4). The other cell lines, HT-29, WiDr, and Capan-1 cells, intermediately expressed the ␤3Gal-T5 gene.
The Ability of Type 1 Chain Synthesis of ␤3Gal-T5 in Transfected Cells-Namalwa cells do not possess the type 1 chain and lack ␣1,3/4-fucosyltransferase (Fuc-TIII), which is the only enzyme capable of the synthesis of type 1 Lewis antigens, such as the Le a , Le b , and sLe a antigens. Therefore, Namalwa cells transfected stably with the ␤3Gal-T5 gene were stained with DU-PAN-2 (anti-sLe c ), which recognizes the precursor structure, SA␣2,3Gal␤1,3GlcNAc, of the sLe a epitope (3,4). As can be seen in Fig. 5, Namalwa cells transfected stably with the ␤3Gal-T5 gene gave positive peaks with DU-PAN-2. HCT-15 cells were chosen as the host cells for the transfection experi-ment with the ␤3Gal-T5 gene for the following reasons. First, they are cancer cells derived from colon tissue. Second, they are known to express substantial amounts of Fuc-TIII and ST3GalIV (data not shown), but not to express ␤3Gal-T5 at all (Fig. 4). A single transformant clone of HCT-15 cells, which had been transfected stably with the ␤3Gal-T5 gene, was obtained by the limiting dilution method and named HCT-3GT5H. Flow cytometry analysis of HCT-3GT5H cells apparently showed positive peaks with the antibodies against all type 1 Lewis antigens (Fig. 5). These results confirmed that ␤3Gal-T5 can synthesize the type 1 chain in transformant cells, i.e. not only in Namalwa cells but also in colon cancer (HCT-15) cells.
␤3Gal-T Activity toward Agalacto-LNnT in Various Tumor Cells and HCT-3GT5 Cells-Agalacto-LNnT-PA was used as an acceptor substrate to measure ␤3Gal-T activity, resulting in the synthesis of lacto-N-tetraose-PA (LNT-PA). The ␤3Gal-T activity, i.e. the LNT-PA synthesizing activity, in Namalwa cells transfected with the ␤3Gal-T5 gene, Namalwa-3GT5, was strongest among all samples of transfected cells and cultured cancer cells examined. Thus, the activity of Namalwa-3GT5 is expressed as 100% activity, and the ␤3Gal-T activities of the other samples relative to that of Namalwa-3GT5 cells are presented in Table IV. The amounts of transcripts for individual ␤3Gal-T genes were measured by competitive RT-PCR, and the relative amounts of individual transcripts normalized as to the amounts of ␤-actin transcripts are shown in Table IV. The level of ␤3Gal-T activity synthesizing LNT-PA was quite parallel to the amounts of ␤3Gal-T5 transcripts, i.e. the cell homogenate of Colo205 showed the strongest activity among the cultured cancer cells, followed by those of SW1116 and Capan-2. This indicated that the LNT-PA synthesizing activity is mainly directed by ␤3Gal-T5. Mock-transfected HCT-15 (HCT-mock) cells did not exhibit any activity, but HCT-3GT5H cells exhibited strong activity almost equal to that of Colo205 cells. HCT-3GT5L cells, which expressed almost one-third of the amount of ␤3Gal-T5 transcripts in the HCT-3GT5H cells, showed one-third of the ␤3Gal-T activity of HCT-3GT5H cells. The cells expressing substantial amounts of transcripts for the other four ␤3Gal-T genes, i.e. the ␤3Gal-T1, -T2, -T3, and -T4 genes, did not exhibit LNT-PA synthesizing activity at all. Namalwa cells stably expressing ␤3Gal-T1, -T2, -T3 and -T4, which were named Namalwa-3GT1, -3GT2, -3GT3, and -3GT4 cells, respectively, also did not exhibit any LNT-PA synthesis activity.
The above results confirmed that the ␤3Gal-T activity synthesizing LNT-PA is mainly directed by ␤3Gal-T5 in these cells, not by the other four ␤3Gal-Ts.
Tissue Distribution and Quantitative Measurement of the ␤3Gal-T5 Transcripts-As can be seen in Fig. 6, ␤3Gal-T5 transcripts were substantially detected in the stomach, jejunum, colon, and pancreas, which are known to express sLe a antigens frequently when they become cancerous. On the other hand, they were hardly detected in the lungs, liver, spleen, adrenal glands, and peripheral blood leukocytes, which rarely produce sLe a antigens when they become malignant. This strongly suggested that ␤3Gal-T5 is responsible for the type 1 chain synthesis, resulting in the sLe a antigen synthesis in gastrointestinal and other tissues.
Multiple Sequence Alignment (ClustalW) of the Five Members of the Human ␤3Gal-T Family-Multiple amino acid sequence alignment of the five ␤3Gal-Ts was constructed by ClustalW method (Fig. 7). Three conserved amino acid motifs, which were employed for design of the degenerate primers, and four conserved cysteine residues were indicated in Fig. 7. One possible N-glycosylation site was also conserved in all five h␤3Gal-Ts.
A phylogenetic tree of the five ␤3Gal-Ts was constructed by means of the neighbor-joining method based on the amino acid sequences (data not shown) (23). The position of ␤3Gal-T5 in the tree is closer to those of ␤3Gal-T1 and -T2, which can synthesize the type 1 chain as reported by us and others (10 -12), than to those of ␤3Gal-T3 and -T4. These three enzymes, ␤3Gal-T1, -T2, and -T5, form a subfamily on the phylogenetic tree.

DISCUSSION
Expression of the sLe a and sLe c epitopes, which are mainly carried on the carbohydrate chains of mucins, is frequently elevated in gastrointestinal and pancreatic cancers. The sLe a antigens are known to be some of the factors determining the prognoses of colorectal and gastric cancer patients (7,8,24). In this sense, it is very important to identify the ␤3GalT(s) responsible for type 1 chain synthesis, which results in the synthesis of the sLe c and sLe a epitopes in gastrointestinal and pancreatic cancers. The novel ␤3Gal-T, i.e. ␤3Gal-T5, isolated in the present study was demonstrated to be responsible for the type 1 chain (including the sLe a and sLe c antigen) synthesis in gastrointestinal and pancreatic cancers by the following evidence. 1) ␤3Gal-T5 could synthesize type 1 chains, leading to expression of Le a , Le b , sLe c and sLe a in transfected cells, i.e. in Namalwa-3GT5 and HCT-3GT5H cells. 2) ␤3Gal-T5 was expressed in cultured colon and pancreatic cancer cells expressing substantial amounts of type 1 chains, whereas the other three ␤3Gal-Ts, i.e. ␤3Gal-T1, -T2, and -T3, were not expressed in these cells.   activities of three mouse ␤3Gal-Ts, i.e. m␤3GalT-I, -II and -III (14). In contrast, we could not detect LNT-PA synthesis activity in the Namalwa cells transfected stably with each of the ␤3Gal-T1, -T2, or -T3 genes. This was probably due to the amounts or structures of the enzymes used or the sensitivity of the assay system. In a previous study, we detected the activity of ␤3Gal-T1 using purified ␤3Gal-T1 that was expressed as a secreted form fused with the IgG binding domain of Staphylococcus aureus protein A (10). The other three groups, i.e. Hennet et al. (14), Kolbinger et al. (11), and Amado et al. (12), assayed ␤3Gal-T activity by measuring radioisotope incorporation using recombinant enzymes produced in soluble forms with a baculo-expression system or the S. aureus protein A fusion system, whereas we used cell homogenates as enzyme sources and a pyridylaminated acceptor substrate in the present study.
We apparently detected LNT-PA synthesizing activity in Namalwa-3GT5 cells and cultured cancer cells expressing ␤3Gal-T5, whereas Namalwa cells transfected with the other four ␤3Gal-T genes and cultured cells endogenously expressing ␤3Gal-T1 (PC-1), ␤3Gal-T2 (Namalwa), or ␤3Gal-T3 (MKN45) did not exhibit LNT-PA synthesizing activity at all. This indicated that ␤3Gal-T5 possesses the strongest activity as to type 1 chain synthesis among the four ␤3Gal-Ts, i.e. ␤3Gal-T1, -T2, -T3, and -T5. The ␤3Gal-T activity synthesizing LNT-PA in Colo205, SW1116, and Capan-2 cells decreased in accordance with the amount of the ␤3Gal-T5 transcript expressed in these cells (Table IV), and the activity was not detected in cells that did not express ␤3Gal-T5. These findings strongly indicated that the endogenous LNT-PA synthesizing activity in cell lines such as Colo205, SW1116, etc., is mainly directed by ␤3Gal-T5. Holmes reported that partially purified ␤3Gal-T(s) from Colo205 cells exhibit preferential activity toward lactotriaosylceramide (Lc 3 ), GlcNAc␤1-3Gal␤1-4Glc␤1-1Cer (25). Valli et al. found that the ␤3Gal-T activity in homogenates of human colorectal cancer cell lines is correlated with the expression levels of type 1 Lewis antigens (26). Their results are consistent with the ␤3Gal-T5 activity demonstrated in this study in these cultured cell lines. The ␤3Gal-T activity detected in their studies may be attributed to that of ␤3Gal-T5.
Western blot analysis revealed that the amounts of ␤3Gal-T5 transcripts were also well correlated with those of sLe a antigens on proteins, probably on mucins in colorectal and pancre-atic cancer cells. This means that ␤3Gal-T5 utilizes carbohydrate chains on proteins as acceptor substrates. In the future, the substrate specificities of ␤3Gal-T5, as well as the other ␤3Gal-Ts, should be examined by employing substrates as analogous as possible to physiological carbohydrate structures.
␤3Gal-T5 was demonstrated to be physiologically expressed in a set of gastrointestinal and other tissues. The substantial amounts of ␤3Gal-T5 transcripts are expressed in stomach, jejunum, colon, and pancreas, strongly suggesting that ␤3Gal-T5 is responsible for expressing sLe a antigens when those tissues become cancerous.
We previously determined the expression levels of 12 glycosyltransferase genes, i.e. those of five ␣1,3-fucosyltransferases (Fuc-TIII, -TIV, -TV, -TVI, and -TVII), four ST3Gals (ST3Gal I, -II,-III, and -IV), one ST6Gal (ST6Gal I), ␤4Gal-T1, and core2-GlcNAc transferase, in colorectal cancer tissues in order to correlate them with the amounts of the sLe x and sLe a antigens (9). Although Fuc-TIII and ST3Gal IV are essentially required for the sLe a synthesis in colorectal cancers, no single enzyme among the 12 was correlated with the amounts of sLe a antigens. Therefore, we conjectured in the previous study that the combinatorial up-regulated expression of multiple enzymes determines the amounts of the antigens. However, it is a noteworthy finding in the present study that the expression levels of ␤3Gal-T5 were well correlated with the amounts of sLe a antigens and the other type 1 Lewis antigens in the cultured cancer cells. This means that ␤3Gal-T5 is a key enzyme determining the expression levels of type 1 Lewis antigens including sLe a antigens in these cells. The above, together with the results of the present study, indicated that ␤3Gal-T5 and ST3Gal IV are responsible for sLe c synthesis, and Fuc-TIII is further required for sLe a synthesis in colorectal cancers in addition to ␤3Gal-T5 and ST3Gal IV.
On the other hand, the type 2 chain, Gal␤1,4GlcNAc, was found to be expressed in all cell lines examined in the present study, since ␤4Gal-T1 is a ubiquitous enzyme, and was substantially expressed in all cell lines (data not shown).
It is of interest to determine whether or not the up-regulation of the ␤3Gal-T5 gene expression determines the levels of sLe a antigens in native cancer tissues. If this is the case, transcriptional regulation of the ␤3Gal-T5 gene will be an attractive subject in the future. In this study, we determined the transcription initiation site of the ␤3Gal-T5 gene in Colo205 cells. The nucleotide sequence in the upstream region was examined for the binding sites of transcription factors using the TFSEARCH (transcription factor search) program, 3 based on the data bases deposited by Heinemeyer et al. (27). We searched the 1-kilobase pair upstream region from the transcription initiation site of the ␤3Gal-T5 gene, but found no TATA box. Within 150 bp upstream of the transcription initiation site, two CdxA sites, an AP-1 site, and a myeloid zinc finger 1 protein, MZF1, site were found. AP-1 and MZF1 have been reported to be potential targets of neoplastic transformation (28,29). CdxA, known as a chicken homeobox-containing gene related to caudal in Drosophila, was previously shown to be expressed in the endoderm-derived gut epithelium during early embryogenesis (30). In the future, we will examine whether or not these transcription factors function in regulation of the ␤3Gal-T5 gene.
Finally, the results of the present study strongly indicate that ␤3Gal-T5 is the most probable candidate responsible for the sLe a antigen synthesis in gastrointestinal and pancreatic cancer cells. In the future, it will be interesting to determine whether or not expression of the ␤3Gal-T5 gene changes some characteristics of cancer cells, especially those related to malignancy.