beta 1,3-Galactosyltransferase beta 3Gal-T5 acts on the GlcNAcbeta 1-->3Galbeta 1-->4GlcNAcbeta 1-->R sugar chains of carcinoembryonic antigen and other N-linked glycoproteins and is down-regulated in colon adenocarcinomas.

We attempted to determine whether beta1,3-galactosyltransferase beta3Gal-T5 is involved in the biosynthesis of a specific subset of type 1 chain carbohydrates and expressed in a cancer-associated manner. We transfected Chinese hamster ovary (CHO) cells expressing Fuc-TIII with beta3Gal-T cDNAs and studied the relevant glycoconjugates formed. beta3Gal-T5 directs synthesis of Lewis type 1 antigens in CHO cells more efficiently than beta3Gal-T1, whereas beta3Gal-T2, -T3, and -T4 are almost unable to direct synthesis. In the clone expressing Fuc-TIII and beta3Gal-T5 (CHO-FT-T5), sialyl-Lewis a synthesis is strongly inhibited by swainsonine but not by benzyl-alpha-GalNAc, and sialyl-Lewis x is absent, although it is detected in the clones expressing Fuc-TIII and beta3Gal-T1 (CHO-FT-T1) or Fuc-TIII and beta3Gal-T2 (CHO-FT-T2). Endo-beta-galactosidase treatment of N- glycans prepared from clone CHO-FT-T5 releases (+/-NeuAcalpha2-->3)Galbeta1-->3[Fucalpha1-->4]GlcNAcbeta1-->3Gal but not GlcNAcbeta1-->3Gal or type 2 chain oligosaccharides, which are found in CHO-FT-T1 cells. This result indicates that beta3Gal-T5 expression prevents poly-N-acetyllactosamine and sialyl-Lewis x synthesis on N-glycans. Kinetic studies confirm that beta3Gal-T5 prefers acceptors having the GlcNAcbeta1-->3Gal end, including lactotriosylceramide. Competitive reverse transcriptase mediated-polymerase chain reaction shows that the beta3Gal-T5 transcript is expressed in normal colon mucosa but not or poorly in adenocarcinomas. Moreover, recombinant carcinoembryonic antigen purified from a CHO clone expressing Fuc-TIII and beta3Gal-T5 reacts with anti-sialyl-Lewis a and carries type 1 chains on oligosaccharides released by endo-beta-galactosidase. We conclude that beta3Gal-T5 down-regulation plays a relevant role in determining the cancer-associated glycosylation pattern of N-glycans.

Type 1 chain oligosaccharides found in N-and O-glycans, as well as in glycolipids, contain the distinctive Gal␤133GlcNAc disaccharide as their core structure. It is synthesized by ␤1,3galactosyltransferases (␤1,3Gal-Ts), 1 a family of enzymes whose genes have been cloned very recently (1). The functional role of type 1 chains is not known, but several studies have indicated that some of them are differentially expressed in cancer. In particular, CEA expressed in colon cancer is characterized by the absence of such chains (2), which are abundantly present on the N-glycans of its normal counterparts synthesized by healthy colon mucosa, also referred to as the nonspecific cross-reacting antigen-2 (3) and the normal fecal antigen-2 (4). At this regard, a ␤1,3Gal-T activity measured using GlcNAc␤133Gal␤134Glc as acceptor was found lower in adenocarcinomas than in normal mucosa (5). However, type 1 chain Lewis antigens sialyl-Lewis a (sLe a ) and Le b are considered tumor markers (6), and serum levels of sLe a (CA19-9 antigen) are used for clinical diagnosis and follow-up of epithelial cancers of the gastrointestinal tract (7). sLe a from cancer patient serum was found on mucins (8), and it is reported on both mucins (9) and glycolipids (10) in adenocarcinoma cell lines. Five ␤3Gal-T cDNAs are presently available. ␤3Gal-T1, cloned first from melanoma cells (11) and then from colon carcinoma cells (12), as well as ␤3Gal-T2 (13), were found to synthesize sLe a in CHO cells but to be very poorly expressed in cancer cell lines expressing sLe a . ␤3Gal-T3 and -T4 (14) were found expressed in various tissues, including colon and pancreas and in some cancer cell lines, but the expression levels do not correlate with those of type 1 chain Lewis antigens. Moreover, there is no evidence yet for their ability to synthesize the type 1 chain. In particular, ␤3Gal-T3 was very recently reported as a GalNAc transferase involved in globoside biosynthesis and renamed ␤3GalNAc-T1 (15). ␤3Gal-T5, cloned from colon carcinoma cells, was found to direct synthesis of sLe a in different cell lines, and its expression levels correlate with those of Lewis type 1 antigens in cancer cell lines from colon and pancreas (16). This enzyme was suggested to be the one previously characterized as the ␤1,3Gal-T active on glycolipids (17) and GlcNAc (18,19), and involved in the synthesis of core 3 O-glycans (20). The above data led to the working hypothesis that the biosynthesis of different type 1 chain oligosaccharides may depend on the expression of different and differentially regulated ␤3Gal-Ts. In particular, an enzyme able to act on N-glycans would be expected to be less active in cancer than in normal mucosa, whereas another active on O-glycans should be more expressed in cancer.
To evaluate whether ␤3Gal-T5 is actually involved in the biosynthesis of a specific subset of type 1 chain oligosaccharides and expressed in a cancer-associated manner, we have transiently transfected CHO cells expressing Fuc-TIII with ␤3Gal-T cDNAs and compared the relative ability of enzymes to direct synthesis of Lewis type 1 antigens. We then constructed CHO clones permanently expressing Fuc-TIII and ␤3Gal-T5, Fuc-TIII and ␤3Gal-T2, or Fuc-TIII and ␤3Gal-T1, determined the effect of drugs affecting glycosylation on antigen expression, and characterized the oligosaccharides that became radioactive in some clones upon metabolic labeling with tritiated Gal. Moreover, we studied the substrate specificity of ␤3Gal-T5 by calculating the kinetic constants toward different acceptors, and measured the expression levels of the transcript in colon adenocarcinomas and surrounding normal mucosa by competitive RT-PCR. We further modified the CHO clone expressing ␤3Gal-T5 to make it able to stably express human CEA and investigated the presence of type 1 chains in CEA purified from such clone.
Cell Cultures, Transfections, and Treatments-CHO cells expressing Polyoma virus T antigen and Fuc-TIII (CHO-T-FT), COLO-205, and WM266-4 cells were cultured as described previously (12,19). Human gastric adenocarcinoma MKN-45 cells (a gift of C. Ponzetto, University of Turin), were cultured in RPMI 1640 medium containing 10% fetal bovine serum, 100 units/ml penicillin, 1.0 mg/ml streptomycin, and 2 mM L-Glu. For transient transfection, 2.0 ϫ 10 5 CHO-T-FT cells were plated in 12-well plates 20 h before transfection, washed with serum free ␣-minimal essential medium, and incubated with 0.5 ml of transfection solution for 3 h under usual growing conditions. Transfection solutions, containing 1 g/ml ␤3Gal-T cDNA in pcDNAI or pCDM8 vectors plus 0.065 g/ml pcDNAI-Luc and 18 l/ml DOTAP (Roche Molecular Biochemicals) were prepared and used as reported previously (12). Seventy-two hours after transfection, cells were harvested, washed, and resuspended with PBS. One-tenth of resuspended cells was processed for luciferase activity determination, using a commercial kit (Luciferase assay system, Promega) according to the manufacturer's recommendations. The remaining material was stained and analyzed by flow cytometry as previously reported (12,19). For treating cell lines and clones with drugs affecting glycosylation, 1 ϫ 10 5 cells were plated in 12-well plates and incubated overnight, and the medium was replaced with medium containing 0.1 g/ml swainsonine (Sigma), 2 mM benzyl-␣-GalNAc (Sigma), or 2 mM sodium butyrate. After 48 h, medium was replaced with fresh medium containing drugs. After additional 48 h, cells were collected, stained, and analyzed by flow cytometry (19).
Construction of CHO Clones-CHO clones expressing either Fuc-TIII and ␤3Gal-T5 (CHO-FT-T5), Fuc-TIII and ␤3Gal-T2 (CHO-FT-T2), or Fuc-TIII and ␤3Gal-T1 (CHO-FT-T1) were obtained by the calcium phosphate transfection method (23), using a modification of the procedure reported (24). Briefly, the DNA mixture (20 g) contained 1 g of EcoRI-linearized pSV2Neo, 10 g of ScaI-linearized pcDNAI/ Fuc-TIII, and 10 g of ScaI-linearized pCDM8/␤3Gal-T5, or 10 g of KpnI-linearized pCDM8/␤3Gal-T2, or 10 g of ScaI-linearized pcDNAI/␤3Gal-T1, respectively. Upon selection with 0.4 mg/ml active G418, 30 (CHO-FT-T5 and CHO-FT-T1) or 60 colonies (CHO-FT-T2) were collected using cloning cylinders, grown in tissue culture slides, stained with anti-sLe a or anti-Le a antibody followed by secondary FITC-conjugated anti-mouse IgG, and analyzed by fluorescence microscopy. Two or three positive colonies were subcloned by limiting dilution in 96-well plates, and several subclones were analyzed as above. sLe a and Le a expression on positive colonies was quantitated by flow cytometry. Single colonies expressing a constant level of sLe a and Le a , named CHO-FT-T5, CHO-FT-T2, and CHO-FT-T1, respectively, were selected and used for further characterization and experiments. CHO cells expressing CEA (CHO-FT-T5-CEA) were obtained from clone CHO-FT-T5 by the transfection procedure described above but using 20 g of KpnIlinearized pCDM8-CEA and 1 g of pHA58 plasmid in the DNA mixture, and 0.5 mg/ml hygromycin B (Roche Molecular Biochemicals) for selection. pHA58 plasmid, a generous gift of P. Morandini (University of Milan), was derived from pSV72 vector and had the hygromycin B gene under the control of mouse phosphoglycerate kinase-1 promoter. Monoclonal anti-CEA antibody was used, at 1:100 dilution, for cell staining.
DNA Preparation-␤3Gal-T1 cloning in pcDNAI vector has been previously reported (12). For ␤3Gal-T2, -T3, -T4, -T5, and CEA cloning, the corresponding coding sequences were amplified using singlestranded cDNA as template, obtained by reverse transcription of total RNA extracted (12)   ␤3Gal-T5 Acts on CEA and Is Down-regulated in Cancer units/ml avian myeloblastosis virus reverse transcriptase (Amersham Pharmacia Biotech), and freshly denatured total RNA (250 g/ml) as template. Reactions were kept for 90 min at 42°C. Amplifications were performed in the presence of a commercially available "high fidelity" Taq polymerase (LA Taq, Takara), according to the manufacturer's protocol, using 2.5 g/ml of each oligonucleotide primer, deduced from published sequences, and 4 l of reverse transcription reaction for each 100 l of amplification mixture. PCR reactions were incubated as follows: a single treatment at 94°C for 3.5 min followed by a cycle consisting of 1.5 min at 94°C (melting), 2.0 min at 62°C (annealing), and 2.5 min at 72°C (extension), repeated 30 (␤3Gal-Ts) or 15 times (CEA); a final extension step was performed at 72°C for 8 min. The amplified DNA was digested with HindIII and XbaI (␤3Gal-T3) or ligated to BstXI adaptors (all others), and cloned in pcDNAI (␤3Gal-T3) or pCDM8 vectors (all others) using a procedure reported previously (12). Direct DNA sequencing of the obtained constructs, performed by the dideoxynucleotide chain-termination method using an automated procedure, indicated that the coding sequences are identical to those published. pcDNAI-Luc, expressing the luciferase gene, was constructed by removing the luciferase coding sequence from plasmid pGL3 (Promega) using HindIII and XbaI, and subcloning into the corresponding sites of pcDNAI vector. Enzyme Assays and Reaction Product Characterization-␤3Gal-T5 and Fuc-TIII were assayed in cell clones as previously reported (19). For kinetic analysis, ␤3Gal-T5 was assayed upon transfection of pCDM8-␤3Gal-T5 in COS-7 cells. Cells were transfected, harvested, washed, resuspended, and used as the enzyme source, as described previously (21). Enzyme activity was determined in a reaction mixture containing, in a final volume of 10 l, 0.1 M Tris/HCl buffer, pH 7.0, 10 mM MnCl 2 , 0.5 mg/ml Triton X-100, 1 mg/ml ␣-lactalbumin, 5 mM CDP-choline, 0.8 mM donor UDP-[ 3 H]Gal, specific radioactivity (10 mCi/mmol), and 0.2-1.0 mg/ml cell protein, in the presence of different acceptors at various concentrations. In the case of glycolipid acceptor, it was dissolved in chloroform/methanol, 4:1 (v/v), mixed in the reaction tube with 15 g of Triton CF-54 dissolved in the same solvent, and dried before adding the reaction mixture. Incubations were done at 37°C for 60 min. At the end of incubation, reaction products were assayed by Dowex chromatography (oligosaccharide acceptors) or descending paper chromatography (glycolipid acceptor) according to previously reported protocols (19,25). The oligosaccharide reaction products were identified by pooling several Dowex eluates, which were lyophilized and purified by Bio-Gel P-2 chromatography (21). The obtained saccharides were treated with D. pneumoniae or Xanthomonas manihotis ␤-galactosidases, specific for ␤1,4and ␤1,3-linkages, respectively, and analyzed using a Bio-Gel P-2 column (19). Reaction product with glycolipid acceptor, pooled from several reactions, was purified by Sep-Pack C-18 cartridge, desalted by partitioning in chloroform/methanol/water, 2:1:1 (v/v), and analyzed by HPTLC, using chloroform/methanol/water, 60:35:8 (v/v), as eluting solvent system, and visualized by fluorography (25).
Metabolic Labeling and Carbohydrate Analysis-CHO clones (2.0 ϫ 10 6 cells) were plated in 60-mm dishes containing 0.1 mCi of [ 3 H]Gal (Amersham Pharmacia Biotech) in 2.5 ml of culture medium and incubated 40 h under regular conditions. Labeled cells were harvested as for immunostaining, resuspended in PBS at a density of 4 ϫ 10 6 cells/0.1 ml, and kept frozen until used. N-Glycans were released from cell suspension by N-glycanase (Glyko) digestion of the material denatured by heating in the presence of SDS, under the conditions recommended by the manufacturer. Released N-glycans were isolated by a Sep-Pak C-18 cartridge and further purified by Sephadex G-50 chromatography, as reported (24). Oligosaccharides released by endo-␤-galactosidase were characterized following the procedure reported by Seuyoshi et al. (26) with some modifications. Purified N-glycans were passed through a Bio-Gel P-4 column (0.7 ϫ 50 cm), and only the radioactivity eluted with water as a peak close to the exclusion volume was collected, lyophilized, and used for characterization. Endo-␤-galactosidase digestion was performed on radioactive N-glycans, 10,000 -20,000 cpm/l, using the enzyme from Bacteriodes fragilis (Glyko), 0.5 milliunit/l, in the buffer supplied by the manufacturer, for 20 h at 37°C. The reaction mixture was then applied to the same Bio-Gel P-4 column as above, and the radioactivity was eluted with the exclusion volume collected and referred to as the endo-␤-galactosidase resistant N-glycans, whereas the radioactivity eluted as oligosaccharides was applied to a QAE-Sephadex column to separate neutral and charged sugars, according to a previous procedure (26). Material collected in the flow-through was referred to as the neutral oligosaccharides, whereas that eluted with NaCl was collected, desalted on a Bio-Gel P-2 column, and treated with Newcastle disease virus sialidase. Neutral and de-sialylated oligosaccharides were analyzed by a Bio-Gel P-4 column (0.7 ϫ 100 cm), eluted with water at a flow rate of 0.1 ml/min, 4.5 min/fraction. The obtained peaks were collected, lyophilized, and treated with glycohydrolases for characterization. X. manihotis ␤-galactosidase (Glyko), almond meal ␣-fucosidase (Oxford Glycosystem), D. pneumoniae ␤-N-acetylglucosaminidase (Glyko), D. pneumoniae ␤-galactosidase (Sigma), and Newcastle disease virus sialidase (Glyko) digestions were performed on radioactive oligosaccharides, 1000 -2000 cpm/l, according to the procedures reported (19,26). The amount of Gal released by glycohydrolase treatment of endo-␤-galactosidase-resistant N-glycans was determined as reported (19). Glycohydrolase digestions of the tetrasaccharide obtained from clone CHO-FT-T5 were analyzed using a Bio-Gel P-2 column (0.7 ϫ 100 cm) eluted with water at a flow rate of 0.1 ml/min, 4.5 min/fraction, whereas those of the peaks were collected from CHO-FT-T1 using a Bio-Gel P-4 column (0.7 ϫ 100 cm) under the same conditions.
CEA Purification and Characterization-CEA was extracted and purified following a published procedure (27). Briefly, about 2.5 ϫ 10 7 metabolically labeled cells, prepared as above described, were diluted to 1.5 ϫ 10 7 cells/ml in PBS, homogenized by vortexing 1 min, and treated with 0.1 unit/ml phosphatidylinositol phospholipase C (Glyko) at 37°C for 2 h. Total reaction was spun at 14,000 rpm for 10 min, and the supernatant was collected, made 0.6 M with perchloric acid, and centrifuged as above. The clear supernatant was neutralized with NaOH, freeze-dried, resuspended, and loaded onto a Sepharose CL6B column equilibrated and eluted with 50 mM PBS, pH 5.2. Elution was monitored by liquid scintillation counting, and relevant fractions were analyzed by SDS-PAGE and visualized by fluorography. Western blotting ␤3Gal-T5 Acts on CEA and Is Down-regulated in Cancer was performed as previously reported (24) using monoclonal anti-CEA (1:200), anti-sLe a (1:500), and anti-sLe x (1:100) antibodies, followed by peroxidase-conjugated goat anti-mouse immunoglobulin (Amersham Pharmacia Biotech) (1:2000). ECL was developed with Renaissance chemiluminescence reagent (PerkinElmer Life Sciences). Homogeneous fractions were collected, lyophilized, and desalted on a Bio-Gel P-2 column. Purified [ 3 H]CEA was treated with C. perfringens sialidase and almond meal ␣-fucosidase, and then with endo-␤-galactosidase, under the conditions reported for N-glycan digestion. The reaction mixture was then applied to a Bio-Gel P-2 column (0.7 ϫ 50 cm), eluted with water at a flow rate of 0.13 ml/min, 2.5 min/fraction. The obtained peaks, as well as the total purified CEA, were treated with ␤-galactosidases and analyzed by Bio-Gel P-2 chromatography as above described.
Competitive RT-PCR Analysis-For the analysis of human samples, bioptic specimens were collected at surgery, immediately frozen in dry ice, and placed in liquid nitrogen until used. For RNA extraction, 1-2 mm 3 of material was homogenized with a rotary homogenizer in 0.5 ml of the lysis buffer from a Qiagen RNeasy minikit and processed in the presence of DNase according to the manufacturer's recommendations. First strand cDNA was synthesized by Moloney murine leukemia virus reverse transcriptase (Amersham Pharmacia Biotech) as previously reported (12). Control reactions were prepared by omitting the reverse transcriptase in the mixture. First strand cDNA was amplified in a 50-l reaction and in the presence of 20 fg of competitor for 35 cycles (glycosyltransferases) or 5 pg of competitor for 25 cycles (␤-actin), under the conditions reported (12). No amplification was detected when the control reactions were used as template. Human ␤-actin, Fuc-TIII, FUT2, and ␤3Gal-T1 competitors and oligonucleotide primers used were the same as those described in a previous study (12). ␤3Gal-T2, -T3, -T4, -T5, and mouse ␤-actin oligonucleotide primers, as well as the restriction enzymes used for competitor preparation, are listed in Table  I. Quantification of human ␤3Gal-T5, ␤3Gal-T2, ␤3Gal-T1, and Fuc-TIII transcripts in CHO clones was performed by the same procedure but using 60 fg of each glycosyltransferase competitor, or 20 pg of mouse ␤-actin competitor, in 50-l reactions that were amplified for 30 (glycosyltransferases) or 25 cycles (␤-actin).

Biosynthesis of sLe a and Le a in CHO Cells
Transiently Transfected with ␤3Gal-T cDNAs-The ability of different cloned ␤3Gal-Ts to synthesize the type 1 chain was assessed to determine the amount of sLe a and Le a antigens expressed upon cDNA transfection on the surface of CHO-T-FT cells, a clone permanently expressing Fuc-TIII and Polyoma virus T antigen. For this purpose, cells were transiently cotransfected with ␤3Gal-T and reporter luciferase cDNAs both placed in expression vectors having the Polyoma virus origin of replication. Luciferase activity from different transfections ranged between 0.6 and 1.4 units. Despite such differences in transfection efficiency, both sLe a and Le a antigens were easily detected in cells transfected with ␤3Gal-T5 (very bright) and ␤3Gal-T1, poorly detected with ␤3Gal-T2, and not at all detected with ␤3Gal-T3 and ␤3Gal-T4. In a typical experiment (Fig. 1), a peak representing up to 35% of cells transfected with ␤3Gal-T5 appears intensely bright mostly after anti-sLe a antibody staining. Fluorescence intensity on positive cells was much lower upon ␤3Gal-T1 transfection, and faintly or not detectable in the other cases, although transfection efficiency was similar or sometimes higher, as assessed by luciferase activity assay. These results suggest that different substrate specificity exists among the different ␤3Gal-Ts, and only ␤3Gal-T5 finds large amount of suitable acceptors in CHO cells.
Construction and Characterization of CHO Clones Expressing ␤3Gal-T5 and Fuc-TIII, ␤3Gal-T2 and Fuc-TIII, or ␤3Gal-T1 and Fuc-TIII-To evaluate the hypothesis that the amount of sLe a and Le a antigens expressed in CHO cells transfected with ␤3Gal-Ts reflects the availability of the proper precursors, we constructed three CHO clones permanently expressing the antigens upon stable transfection with either ␤3Gal-T5 and Fuc-TIII, ␤3Gal-T2 and Fuc-TIII, or ␤3Gal-T1 and Fuc-TIII. All clones, named CHO-FT-T5, CHO-FT-T2, and CHO-FT-T1, re-spectively, express human Fuc-TIII transcript in a similar amount, as determined by competitive RT-PCR, whereas each clone expresses only its corresponding human ␤3Gal-T transcript ( Fig. 2A). Flow cytometry analysis (Fig. 2B)  Fuc-TIII and ␤3Gal-T1 (CHO-FT-T1) were constructed as described under "Experimental Procedures." A, total RNA was extracted from the clones and reverse-transcribed, and the obtained first strand cDNA was amplified by PCR using primers specific for human glycosyltransferases or mouse ␤-actin, in the presence of the proper competitor DNA. One-fifth aliquot of each PCR reaction was analyzed by 1% agarose gel electrophoresis, visualized by staining with ethidium bromide, and photographed, and the spots were quantitated by densitometric scanning of the negative film. The amounts of amplified target cDNAs were calculated from their respective standard curves and normalized by those for ␤-actin. B, cell clones were stained with monoclonal anti-sLe a (solid line), anti-Le a (dotted line) (both IgG), or anti-sLe x (stippled line) (IgM) monoclonal antibody, followed by FITC-conjugate anti-mouse IgG or IgM, respectively, and analyzed by flow cytometry. Negative controls (shaded areas) represent the cell clones stained with FITC-conjugate anti-mouse IgG alone.
␤3Gal-T5 Acts on CEA and Is Down-regulated in Cancer those obtained by the transient transfection experiments (Fig.  1). Moreover, CHO-FT-T2 and CHO-FT-T1 cells are homogeneously bright with anti-sLe x staining, as expected in CHO cells expressing Fuc-TIII, whereas CHO-FT-T5 cells are totally negative (Fig. 2B). This last result is surprising and suggests a competition between Gal-Ts in the synthesis of type 1 versus type 2 chain Lewis antigens.
Effect of Drugs Affecting Glycosylation on the Expression of sLe a and sLe a in CHO Clones-CHO cells are known to express mostly complex type N-glycans and simple core 1 type O-glycans (28). Because the latter are not precursors of type 1 chains, our findings suggest that ␤3Gal-T5 has a distinctive ability to act on complex type N-glycans. To address this issue and to study the contribute of N-and O-linked glycosylation on the expression of sLe a and sLe x , cells were treated with swainsonine and benzyl-␣-GalNAc, selectively affecting complex type N-glycan and O-glycan biosynthesis, respectively, before stain-ing with anti-sLe a or anti-sLe x antibodies. sLe a staining is 80% inhibited by swainsonine treatment in CHO-FT-T5, whereas benzyl-␣-GalNAc treatment is ineffective and sLe x staining is absent (Fig. 3). Conversely, in clone CHO-FT-T2 sLe a staining is almost absent, whereas sLe x staining is reduced by swainsonine treatment (about 65%) and slightly stimulated by sodium butyrate. In clone CHO-FT-T1 sLe a staining is almost 70% inhibited by benzyl-␣-GalNAc, whereas swainsonine treatment is ineffective; moreover, sodium butyrate treatment stimulates sLe a synthesis 5-fold. sLe x staining in this clone is reduced by swainsonine treatment and slightly stimulated by sodium butyrate, as in CHO-FT-T2. In control COLO-205 cells, sLe a and sLe x reactivity is totally abolished by benzyl-␣-Gal-NAc, and only about 30% inhibited by swainsonine treatment (Fig. 3). These results indicate that sLe a expressed in CHO-FT-T5 cells is carried by complex type N-glycans, presumably  The N-glycans were prepared from the total incorporated radioactivity and treated with endo-␤-galactosidase, and the released oligosaccharides were separated in the neutral and charged fractions. As shown in Table II, the radioactivity recovered after each step is rather similar in the three clones. On the other hand, the composition of the neutral saccharides released by endo-␤-galactosidase treatment of CHO-FT-T5 N-glycans is dramatically different from that of CHO-FT-T1 N-glycans. In fact, almost all radioactivity derived from clone CHO-FT-T5 is eluted as a single peak in the area of tetrasaccharides by Bio-Gel P-4 fractionation, whereas that from clone CHO-FT-T1 is distributed in several peaks, including some of higher molecular weight as well as one in the disaccharide area (Fig. 4, upper panel). The radioactive peak obtained from CHO-FT-T5 N-glycans is converted to a trisaccharide by almond meal ␣-fucosidase, to an equal mixture of disaccharide and monosaccharide by the action of both almond meal ␣-fucosidase and X. manihotis ␤-galactosidase, and to a monosaccharide by the combination of the above enzymes and D. pneumoniae ␤-N-acetylglucosaminidase (Fig. 4, lower panel). Digestion of the tetrasaccharide peak with almond meal ␣-fucosidase and D. pneumoniae ␤-galactosidase determines the formation of a single peak in the area of trisaccharides but not detectable monosaccharides (not shown). The elution profile obtained after de-sialylation of the charged oligosaccharides is identical, indicating that Gal␤133[Fuc␣134]GlcNAc␤133Gal and NeuAc␣233Gal␤133[Fuc␣134]GlcNAc␤133Gal account for the vast majority of oligosaccharides released by endo-␤-galactosidase digestion of the N-glycans formed in clone CHO-FT-T5. However, the elution profile of the oligosaccharides released by endo-␤-galactosidase treatment of CHO-FT-T1 and CHO-T-FT cells are very similar. Among such peaks, the one eluted as a disaccharide was identified as GlcNAc␤133Gal, because it is converted to a monosaccharide by D. pneumoniae ␤-N-acetylglucosaminidase. Such a disaccharide is formed by the action of endo-␤-galactosidase on multiple lactosamine repeats (29). The peak eluted at the size of tetrasaccharides was identified as Gal␤134[Fuc␣133]GlcNAc␤133Gal, because it is converted to a monosaccharide only by the sequential action of almond meal ␣-fucosidase, D. pneumoniae ␤-galactosidase, and D. pneumoniae ␤-N-acetylglucosaminidase. The larger peaks (fractions 51-55 and 46 -50 in Fig. 4, upper panel) contain one and two other major oligosaccharides, respectively. They were identified by sequential exoglycosidase digestions and Bio-Gel P-4 filtration as the pentasaccharide, hexasaccha-ride, and heptasaccharide already reported (26). Such oligosaccharides contain double lactosamine repeats fucosylated on the inner GlcNAc, a substitution that makes the Gal␤134GlcNAc linkage resistant to endo-␤-galactosidase under usual reaction conditions (30). These results strongly suggest that the (ϮNeuAc␣233)Gal␤133[Fuc␣134]GlcNAc␤133Gal oligosaccharide synthesized in CHO-FT-T5 replaces the poly-Nacetyllactosamine chains and sLe x structure present in CHO-FT-T1 as well as in CHO cells expressing Fuc-TIII only. To assess the presence of type 1 chain carbohydrates in endo-␤galactosidase-resistant N-glycans, we have determined the amount of radioactive Gal released by X. manihotis ␤-galactosidase. As shown in Table III, X. manihotis ␤-galactosidase is active only on the glycans derived from CHO-FT-T5, but not on those from CHO-FT-T1. The amount of Gal released by the enzyme is strongly dependent on the concurrent action of sialidase, but not on that of ␣-fucosidase, and is much lower than that released by D. pneumoniae ␤-galactosidase under the same reaction conditions. These results indicate that type 1 chains synthesized on endo-␤-galactosidase-resistant N-glycans by ␤3Gal-T5 are present in limited amount and are not significantly substituted by fucose residues.
In Vitro Properties of ␤3Gal-T5-Using a cell homogenate prepared from COS cells transiently transfected with ␤3Gal-T5 cDNA as the enzyme source, we determined the optimal reaction conditions using GlcNAc as acceptor and found that the activity requires Mn 2ϩ , is maximal at neutral pH, is saturated by donor UDP-Gal concentrations above 0.5 mM, and is not affected by ␣-lactalbumin. Under the same reaction conditions, various oligosaccharides are also used. The obtained reaction products are over 95% affected by X. manihotis ␤-galactosidase and almost unaffected by D. pneumoniae ␤-galactosidase. In the presence of the proper detergent concentration, the same homogenate very efficiently transfers Gal to the glycolipid lactotriosylceramide, and the reaction product was identified as lactotetraosylceramide (Gal␤133GlcNAc␤133Gal␤134 Glc␤131ЈCer) by HPTLC mobility and differential sensitivity to ␤-galactosidases. The calculated K m and V max values (Table  IV) indicate that the enzyme prefers the GlcNAc␤133Gal end but also distinguish among the different carrier molecules,  H]Gal released by D. pneumoniae or X. manihotis ␤-galactosidase treatment of endo-␤-galactosidase resistant N-glycans prepared from CHO clones The whole reaction mixtures (about 5000 cpm) were applied to Bio-Gel P-2 column, and radioactivity in the eluted fractions was determined by liquid scintillation counting. Results are the means for two independent experiments and are presented as percentage of recovered radioactivity in the galactose peak. ␤3Gal-T5 Acts on CEA and Is Down-regulated in Cancer with a preference for the glycolipid acceptor. GlcNAc linked to ␣-Man through ␤1,2or ␤1,6-linkages is also a suitable acceptor, but with much lower affinity. These in vitro results are in good agreement with those obtained by the structural analysis of CHO-FT-T5 clone, where we found that endo-␤-galactosidase-sensitive saccharides, having the GlcNAc␤133Gal acceptor sequence, are preferentially operated upon by ␤3Gal-T5, whereas endo-␤-galactosidase-resistant N-glycans, mostly having GlcNAc␤132/6Man as acceptor sequences, are poorly utilized.
Expression of ␤3Gal-T5 Transcript in Normal Colon Mucosa and in Adenocarcinomas-To evaluate whether the ␤3Gal-T5 transcript is differentially expressed during carcinogenesis, we analyzed its amount by competitive RT-PCR performed on total RNA extracted from human colon specimens collected at surgery, representing both adenocarcinomas and surrounding normal mucosa. Clinical features and tumor staging are outlined in Table V. The ␤3Gal-T5 transcript is detected in all normal mucosa samples, although it is faintly detectable or undetectable in adenocarcinomas whose cDNA is well amplified using control primers such as those for ␤-actin or other glycosyltransferases (Fig. 5). In quantitative terms (Fig. 6), the ␤3Gal-T5 transcript is on the average 30-fold less expressed in adenocarcinomas than in normal mucosa. In individual cases, the levels in adenocarcinomas range from 4-fold to over 100-fold less than in normal mucosa. For comparison we looked at the expression levels of the other ␤3Gal-T transcripts, as well as of Fuc-TIII and FUT2, two fucosyltransferases involved in type 1 chain fucosylation. ␤3Gal-T4 transcript is expressed at high and heterogeneous levels in normal mucosa and remains detectable in all adenocarcinomas. ␤3Gal-T1 transcript is detectable in eight normal mucosa and four adenocarcinoma cases, whereas ␤3Gal-T2 transcript is detectable in four normal mucosa and seven adenocarcinoma cases. ␤3Gal-T3, Fuc-TIII, and FUT2 transcripts are heterogeneously expressed in both normal mucosa and adenocarcinomas. Altogether, these data suggest that the ␤3Gal-T5 transcript, almost undetectable in colon adenocarcinomas, is strongly down-regulated during carcinogenesis.
Characterization of CEA Purified from a CHO Clone Expressing ␤3Gal-T5-To study the ability of ␤3Gal-T5 to act on CEA, a CHO clone permanently expressing this antigen, as well as ␤3Gal-T5 and Fuc-TIII, was obtained by transfecting CHO-FT-T5 cells with CEA (CHO-FT-T5-CEA). As assessed by flow cytometry, the expression levels of sLe a and Le a antigens in this clone are identical to those of the starting one, whereas those of CEA were comparable to that of human cancer cell lines. In particular, we measured a mean fluorescence intensity lower than in MKN-45 cells, but higher than in COLO-205 cells (not shown). The tritiated material purified from clone CHO-FT-T5-CEA metabolically radiolabeled with [ 3 H]Gal runs on SDS-PAGE as a single radioactive band of an apparent molecular mass of 200 kDa (Fig. 7). Immunoblot analysis shows that this band strongly reacts with anti-CEA antibody. Sequential stripping and reprobing of the filter with anti-Lewis antigen antibodies reveals that the purified CEA reacts with anti-sLe a but not with anti-sLe x antibodies, indicating that type 1 chains are specifically present on CEA expressed in the clone. To directly assess the presence of type 1 chain carbohydrates on CEA and to determine whether they are preferentially bound to GlcNAc␤133Gal␤134GlcNAc␤13 R outer sequences, the purified [ 3 H]CEA, once de-fucosylated and de-sialylated, was treated with different ␤-galactosidases before and after endo-␤-galactosidase treatment. To this end, the endo-␤-galactosidase reaction mixture was submitted to Bio-Gel P-2 filtration and the endo-␤-galactosidase-resistant and -sensitive radioactivity was recovered with the exclusion volume and at the size  5. Competitive RT-PCR analysis of ␤3Gal-T5 transcript in normal mucosa and in adenocarcinomas. RNA extracted from adenocarcinomas and surrounding normal mucosa, collected at surgery, was reverse-transcribed, and the obtained first strand cDNA was mixed with competitor (truncated) cDNAs (5 pg and 20 fg for ␤-actin and glycosyltransferases, respectively), and subjected to PCR (25 and 35 cycles for ␤-actin and glycosyltransferases, respectively). Primers and PCR product length are indicated in Table I. One-fifth aliquot of each amplification reaction was electrophoresed in 1% agarose gel and visualized by staining with ethidium bromide. N, normal mucosa; A, adenocarcinoma. Numbers denote patients according to Table V. Samples 1-6 and 7-10 were run on different gels. ␤3Gal-T5 Acts on CEA and Is Down-regulated in Cancer of trisaccharides, respectively (Fig. 8, upper panel). A small but detectable amount of Gal was removed by X. manihotis ␤-galactosidase from total de-sialylated and de-fucosylated CEA, and the amount released from the endo-␤-galactosidase-resistant material (fractions 18 -22 in Fig. 8) was minimal, whereas that released by D. pneumoniae ␤-galactosidase was more consistent in both cases. Conversely, almost 40% of the radioactivity from the endo-␤-galactosidase-sensitive trisaccharide (fractions 32-39 in Fig. 8) was released by X. manihotis ␤-galactosidase, but poorly affected by the diplococcal enzyme (Fig.  8, lower panel). These data indicate that ␤3Gal-T5 preferentially acts on the GlcNAc␤133Gal␤134GlcNAc␤13 R side chains present in CEA. DISCUSSION This report shows that ␤3Gal-T5, in transfected CHO cells, directs synthesis of type 1 chain oligosaccharides on CEA and other N-glycans having the GlcNAc␤133Gal␤13 4GlcNAc␤13 R outer sequence, prevents poly-N-acetyllactosamine and sLe x synthesis on N-glycans, and is downregulated in colon adenocarcinomas.
We have found that ␤3Gal-T5 synthesizes sLe a and Le a antigens very efficiently in CHO cells, much more than ␤3Gal-T1, which is still more efficient than ␤3Gal-T2, whereas ␤3Gal-T3 and -T4 are unable to synthesize these antigens at all. In the stable clone CHO-FT-T5, sLe a expression is selectively affected by swainsonine, an inhibitor of complex type N-glycan processing (31), but not by benzyl-␣-GalNAc, an inhibitor of O-glycan biosynthesis (32), although the opposite occurs in clone CHO-FT-T1. These data support the hypothesis that ␤3Gal-T5 acts on complex type N-glycans in CHO cells and suggest that ␤3Gal-T1 presumably affects unknown O-glycans available in these cells in a low amount. In this regard, it is interesting that sLe a expression in CHO-FT-T1 is strongly stimulated by sodium butyrate, a commonly used activator reported to enhance core 2 GnT activity in CHO cells (33). However, results obtained with CHO-FT-T2 suggest that ␤3Gal-T2 requires dedicated substrates not available in these cells. Moreover, CHO-FT-T1 cells, as well as CHO-FT-T2, were found to express a large amount of sLe x that is undetectable in CHO-FT-T5. We found that the oligosaccharides released from N-glycans by endo-␤-galactosidase are completely different in the clones. In fact, (ϮNeuAc␣233)Gal␤133[Fuc␣134]GlcNac␤133Gal is the only oligosaccharide identified in CHO-FT-T5, whereas in clone CHO-FT-T1 several oligosaccharides are released, which correspond to those reported in CHO expressing Fuc-TIII only (26). They include the sLe x tetrasaccharide as well as poly-Nacetyllactosamine side chains, but not type 1 chain oligosaccharides. Among endo-␤-galactosidase-resistant N-glycans, type 1 chain oligosaccharides are present in small amounts and not fucosylated in clone CHO-FT-T5 but not detectable in CHO-FT-T1 cells. On this basis we conclude that the expression of Radioactive material recovered upon phosphatidylinositol-specific phospholipase C extraction, perchloric acid precipitation, and Sepharose CL6B chromatography was desalted on a Bio-Gel P-2 column, and 10,000-cpm aliquots were loaded on two lanes of a 6% SDS-PAGE. After the run, one lane was processed for fluorography using liquid En 3 Hance (PerkinElmer Life Sciences) (10day exposure). The other one was blotted onto a nitrocellulose filter that was probed with monoclonal anti-CEA antibody using peroxidase-conjugated goat anti-mouse Ig as secondary antibody, and visualized by ECL. The same filter was stripped and reprobed sequentially with monoclonal anti-sLe a and anti-sLe x antibodies.
␤3Gal-T5 Acts on CEA and Is Down-regulated in Cancer ␤3Gal-T5 in CHO cells inhibits the synthesis of poly-Nacetyllactosamines and sLe x in N-glycans, replacing them with a short type 1 chain, whereas that of ␤3Gal-T1 does not affect N-glycans. Expression of N-linked poly-N-acetyllactosamines, which are often modified to express functional oligosaccharides such as sLe x , is considered to be associated with tumor progression and malignancy (34). Recent work on poly-N-acetyllactosamine extension in N-glycans (35) demonstrated that it is achieved mainly by ␤4Gal-T1 and ␤1,3-N-acetylglucosaminyltransferase iGnT (36) and is favored by ␤1,6GlcNAc branching, because the branched structure serves as a much better acceptor for such enzymes (37). Expression of GlcNAc-TV, responsible for branching, also correlates with metastatic potential of cell lines (38) as well as with metastasis and poor prognosis in colon cancer (39). Because ␤3Gal-T5 has a very high affinity for acceptors having the GlcNAc␤133Gal outer sequence, but much lower for those with the GlcNAc␤132Man or GlcNAc␤136Man sequence, the following scenario can be envisaged in CHO cells. In the presence of ␤3Gal-T5, ␤4Gal-T1 still acts on GlcNAc linked to Man, and complex type N-glycans are elongated by a single lactosamine unit. This is acted upon by iGnT, presumably on a ␤1,6-branched chain, which forms the GlcNAc␤133Gal␤134GlcNAc␤13 R outer group. At this stage, ␤3Gal-T5 competes with ␤4Gal-T1 adding a ␤1,3-galactosyl residue that prevents further chain elongation by iGnT. Consequently, poly-N-acetyllactosamine chains could not be efficiently extended nor sLe x synthesized.
It is worth noting that the K m values of ␤3Gal-T5 for GlcNAc␤133Gal␤134Glc and GlcNAc␤132Man are virtually identical to those reported for a ␤1,3Gal-T expressed in normal colon mucosa whose activity was found low in adenocarcinomas (5). These data prompted us to evaluate the expression level of ␤3Gal-T5 transcript in normal mucosa versus colon adenocarcinomas. Competitive RT-PCR analysis performed on 10 colon adenocarcinoma cases indicate that ␤3Gal-T5 transcript is constantly down-regulated in cancer, confirming that it is the enzyme previously detected by Seko et al. (5) and corroborating the hypothesis that in vivo it would prevent the expression of poly-N-acetyllactosamines and sLe x on N-glycans. This behavior is unique among ␤3Gal-T and other glycosyltransferase transcripts studied to date (40 -43).
Because CEA synthesized in normal mucosa is reported to express N-glycans having type 1 chains as Gal␤133GlcNAc␤13 3Gal␤134GlcNAc␤13 R outer sequences (3,4), whereas CEA produced by colon adenocarcinomas lacks type 1 chains but have poly-N-acetyllactosamine and ␤1,6 branched type 2 chains on N-glycans (2,44), our results suggest that ␤3Gal-T5 may act on CEA in vivo. To assess this hypothesis, we permanently expressed CEA in clone CHO-FT-T5 (CHO-FT-T5-CEA). We found that CEA purified from the clone reacts with anti-sLe a antibody by Western blot. Moreover, the type 1 chain carbohydrates present in the molecule are more abundant in endo-␤-galactosidase-sensitive oligosaccharides than in those resistant. These results confirm that ␤3Gal-T5 acts on CEA and suggest that it is responsible for the differential glycosylation pattern of this protein in colon cancer.
The kinetic data calculated in this paper for lactotriosylceramide, as well as those reported (20) for GlcNAc␤133GalNAc as acceptors, suggest that ␤3Gal-T5 is able to act on glycolipids as well as on core 3 O-linked glycoproteins. The possibility of a broad range of specificity of the enzyme is confirmed by the recent report that it also acts on the GalNAc end of globotetraosylceramide (45). On this basis, the down-regulation of ␤3Gal-T5 in cancer is expected to affect other glycoconjugates synthesized by the enzyme in normal mucosa. However, ␤3Gal-T5 expression is maintained in some colon adenocarcinoma cell lines (16), and we suggest that the results obtained with these cells must be cautiously interpreted, or reinterpreted (12,19). In particular, further experimental data are needed to establish whether or not the CA19.9 antigen present in patient serum is synthesized by ␤3Gal-T5 and secreted from the tumor mass.