Synthesis of disialyl Lewis a (Le(a)) structure in colon cancer cell lines by a sialyltransferase, ST6GalNAc VI, responsible for the synthesis of alpha-series gangliosides.

Biosynthesis of disialyl Lewis a (Lea) was analyzed using previously cloned ST6GalNAc V and ST6GalNAc VI, which were responsible for the synthesis of alpha-series gangliosides. Among lactotetraosylceramide (Lc4), neolactotetraosylceramide, and their sialyl forms, only sialyl Lc4 was sialylated with ST6GalNAc V and ST6GalNAc VI. The products were confirmed to be disialyl Lea in TLC-immunostaining. Compared with the original substrate GM1b, the synthetic rates of disialyl Lea were 22 and 38% with ST6GalNAc V and ST6GalNAc VI, respectively. Since sialyl Lea could not be converted to disialyl Lea, disialyl Lea was produced only from disialyl Lc4. Therefore, it appears that ST6GalNAc V/VI and fucosyltransferase III (FUT-3) compete for sialyl Lc4, their common substrate. The results of either one transfection or co-transfection of two genes into COS1 cells revealed that both ST6GalNAc VI and FUT-3 contributed in the synthesis of disialyl Lea but partly compete with each other. Many colon cancer cell lines expressed the ST6GalNAc VI gene more or less, and some of them actually expressed disialyl Lea. None of them expressed ST6GalNAc V. These results suggested the novel substrate specificity of ST6GalNAc VI, which is responsible for the synthesis of disialyl Lea but not for alpha-series gangliosides in human colon tissues.

Carbohydrate structures conjugated to proteins and ceramides on the cell surface are involved in the modification of cell-cell or cell-extracellular matrix interaction (1). Consequently, they play important roles in the regulation of cell proliferation, cell adhesion, cancer metastasis, tissue differentiation, and apoptosis (2,3). In particular, sialylation of sugar chains has been suggested to be a very important process during development, cancer evolution, and progression (4), and sialic acid is often responsible for tumor-associated antigenicity.
Monosialyl Lewis a (Le a ) 1 has been suggested to be a repre-sentative tumor-associated antigen in cancers of the pancreas and colon. This antigen is recognized by monoclonal antibody (mAb) 19-9 (5) and has been utilized as an indicator of the tumor size in those tumor patients. In fact, the majority of colon cancer tissues express monosialyl Le a , whereas normal mucosa rarely expresses the antigen (6). On the other hand, both normal and malignant pancreatic tissues express sialyl Le a , although it can be used as a tumor-associated antigen when measured for its serum levels.
In addition to monosialyl Le a antigen, disialyl Le a was identified in the disialoganglioside fraction of human adenocarcinoma of the colon (7). Serum levels of disialyl Le a antigen were measured using mAb FH7, and they turned out to be increased in patients with colonic and pancreatic cancers (8). These reports suggested that this antigen is also applicable as a tumorassociated antigen. Moreover, immunohistochemical analysis of Le a , monosialyl Le a , and disialyl Le a antigens in human colorectal and pancreatic tissues revealed that disialyl Le a might affect the expression mode of sialyl Le a antigen (i.e. an oncofetal antigen) by competing with each other in the biosynthesis of individual structures (9). The synthesis and expression of disialyl Le a might result in masking the expression of sialyl Le a . However, little has been known about the biosynthetic pathway of disialyl Le a or about the sialyltransferase responsible for the transfer of a sialic acid onto GlcNAc with ␣2,6 linkage.
To identify the ␣2,6-sialyltransferase that catalyzes the synthesis of disialyl Le a from monosialyl Le a or the synthesis of disialyl lactotetraosylceramide (disialyl Lc4) from monosialyl Lc4 as a precursor for the synthesis of disialyl Le a , a sialyltransferase assay was performed using previously cloned ␣2,6sialyltransferases (10,11). We demonstrated here that ST6GalNAc V/VI, which were cloned as the sialyltransferases responsible for the synthesis of ␣-series gangliosides, could significantly synthesize disialyl Lc4, indicating the main synthetic pathway of disialyl Le a . Based on the analyses of expression levels of the transferase genes and resulting antigens in human colon cancer cell lines, ST6GalNAc VI appears most likely to be a key enzyme in the synthesis of disialyl Le a structure.
Construction of the Expression Vector-The expression vector pcDNA3.1-ST6GalNAc VI was prepared by inserting an XbaI and XhoI fragment from pMIKneo-ST6GalNAc VI (11) into XbaI and XhoI sites of pcDNA3.1(ϩ) vector.
Preparation of Membrane Fraction-L cells (3 ϫ 10 6 ) were plated in 10-cm dishes at least 48 h prior to transfection. Cells were transiently transfected with an expression plasmid (4 g) by the DEAE-dextran method (18). After 48 h of culture in Dulbecco's modified Eagle's medium containing 7.5% fetal calf serum, the cells were harvested by trypsinization. Cells were pelleted, washed with phosphate-buffered saline (PBS), and lysed in ice-cold PBS containing 1 mM phenylmethylsulfonyl fluoride using a nitrogen cavitation apparatus (Parr Instrument Co., Moline, IL) at 400 p.s.i. for 30 min. Nuclei were removed by low speed centrifugation, and supernatant was centrifuged at 100,000 ϫ g for 1 h at 4°C. The pellet was resuspended in ice-cold 100 mM sodium cacodylate buffer, pH 7.0, and used as an enzyme source for the fucosyltransferase assay as described below.
Fucosyltransferase Assay-The fucosyltransferase assay was performed in a mixture containing 100 mM sodium cacodylate buffer, pH 7.0, 10 mM MgCl 2 , 0.25% Triton X-100, 5 mM CDP-choline (Sigma), 0.1 mM GDP-fucose (Sigma), 25,000 dpm/nmol GDP-[ 14 C]fucose (Amersham Biosciences), 50 g of a membrane fraction, and 10 g of acceptors in a total volume of 50 l. The reaction mixture was incubated at 37°C for 1 h, and the enzyme products were isolated and analyzed by TLC as described above.
Preparation of Soluble Forms of ST6GalNAc V and VI-As described previously (10,11), soluble forms of ST6GalNAc V and VI as fusion proteins with protein A were prepared for the sialyltransferase assay.
Sialyltransferase Assay-The sialyltransferase assay was performed in a mixture containing 100 mM sodium cacodylate buffer, pH 6.0, 10 mM MgCl 2 , 0.3% Triton CF-54, 0.66 mM CMP-NeuAc (Sigma), 6,000 dpm/l CMP-[ 14 C]NeuAc (Amersham Biosciences), the enzyme solution, and 10 g of acceptors in a total volume of 50 l. The reaction mixture was incubated at 37°C for 2 h. The products were isolated using a C 18 Sep-Pak cartridge (Waters, Milford, MA) and analyzed by TLC with a solvent system of chloroform/methanol/0.2% CaCl 2 (55:45: 10). The radioactivity on each plate was visualized with a BAS 2000 image analyzer (Fuji Film, Tokyo, Japan). For kinetic analysis, incubation was performed using various concentrations of acceptor substrates, 0 -0.2 mM GM1b 2 or sialyl Lc4.
TLC-Immunostaining-TLC-immunostaining was performed as de-scribed previously (19). Disialyl Lc4 was detected using anti-disialyl Lc4 mAb FH9 (kindly provided by S. Hakomori (Pacific Northwest Research Institute, University of Washington, Seattle)) (8) at a 1:3 dilution as a primary antibody. Briefly, after chromatography of the glycolipids, the TLC plate was heat-blotted onto a polyvinylidene difluoride membrane. The membrane was incubated with mAb for 1 h, washed, and incubated with biotinylated horse anti-mouse IgG at a 1:200 dilution for 1 h. The antibody binding was visualized with an ABC-PO kit (Vector, Burlingame, CA) and HRP-1000 (Konica, Tokyo, Japan). Transfection for Flow Cytometric Analysis and Immunofluorescence Assay-COS1 cells and colon cancer cell lines in a 6-cm dish (Falcon) were transiently transfected with pcDNA3.1, pcDNA3.1ϪST6GalNAc VI, and/or pcXN2-FUT-3 (provided by S. Nishihara and H. Narimatsu (Soka University, Tokyo)) (1 g/l) by the DEAE-dextran method and cultured for 48 h in Dulbecco's modified Eagle's medium containing 7.5% fetal calf serum before observation.
Flow Cytometric Analysis-The cell surface expression of disialyl Le a and sialyl Le a was analyzed using the transfectant cells after transient transfection of expression vectors. Two days after transfection, cells were trypsinized and washed twice with PBS and then used for flow cytometric analysis using anti-disialyl Le a mAb FH7, anti-sialyl Le a mAb 1H4 (Seikagaku Corp.), anti-sialyl Le a /sialyl Lc4 mAb 2D3 (Seikagaku Corp.), anti-GD1␣ mAb KA-17 (presented by Y. Hirabayashi (RIKEN Brain Science Institute, Wako, Japan)) (20), and anti-GQ1b␣ mAb GGR41 (presented by T. Tai (Tokyo Metropolitan Institute of Medical Science)) (21) on FACScalibur with Cell Quest TM version 3.1f software (Becton Dickinson). Fluorescein isothiocyanate-conjugated anti-mouse IgG (whole) antibody (ICN/Cappel) or anti-mouse IgM antibody (ZYMED) was used as second antibodies.
Immunofluorescence Assay-COS1 cells were cultured on cover glasses in 24-well plates and incubated at 37°C for 24 h and transiently transfected with expression vectors as described above. The cells were fixed with cold acetone for 10 min or with 3.6% paraformaldehyde in PBS for 5 min. In the latter case, they were permeabilized with 0.1% Triton X-100 in PBS for 10 min. Then they were processed for indirect immunofluorescence analysis as described previously (11). The staining was observed using the Radiance TM confocal imaging system (Bio-Rad) and also with ORCA-ER-1394 imaging systems (Hamamatsu Photonics, Hamamatsu, Japan).
Analysis of ST6GalNAc V/VI Gene Expression-The expression levels of the ST6GalNAc V/VI gene in human colon cancer cell lines were determined by RT-PCR and Northern blotting. RT-PCRs were performed with the ST6GalNAc V and VI gene-specific primers, 5Ј-GGT-CTGGCAGTGTGTTTAGC-3Ј (nucleotides 22-41 in the coding sequence) and 5Ј-AACTGGGCACGGACATTCAA-3Ј (nucleotides 916 -935) for ST6GalNAc V, 5Ј-CAGACGCCGGAGAGAAATGA-3Ј (nucleotides 60 -79 in the coding sequence) and 5Ј-GCCCCCAGAAGATG-AACACG3Ј (nucleotides 526 -507) for ST6GalNAc VI. For Northern blot analysis, total RNA was prepared using TRIZOL Reagent TM (Invitrogen) according to the manufacturer's instructions. Fifteen micrograms each of total RNA was electrophoresed and blotted onto a nylon membrane (GeneScreen Plus) (PerkinElmer Life Sciences). They were hybridized with [ 32 P]dCTP-labeled ST6GalNAc V or ST6GalNAc VI cDNA probes as previously described (11).
FIG. 1. Acceptor specificities of ST6GalNAc V/VI. A, sialyltransferase assay was performed using ProtA-ST6GalNAc V and VI as described under "Materials and Methods," and the products were analyzed by TLC. Acceptors used are indicated: GM1b, sialyl nLc4, sialyl Lc4, nLc4, and Lc4. B, to detect disialyl Lc4, TLC-immunostaining was performed with mAb FH9 (hybridoma supernatant) at a 1:3 dilution as described under "Materials and Methods." The antibody binding was detected using the ABC-PO kit (Vector, Burlingame, CA) and HRP-1000 (Konica, Tokyo, Japan).

ST6GalNAc V and VI Synthesized Disialyl
Lc4 -To analyze the novel substrate specificity of ST6GalNAc V and VI, soluble fusion enzymes fused with protein A were prepared, and their sialyltransferase activities toward various lacto-and neolactoseries glycolipids were examined. In previous studies, we reported that these enzymes synthesized ␣-series gangliosides such as GD1␣, GT1a␣, and GQ1b␣. Among Lc4, nLc4 and their sialylated forms, only sialyl Lc4 was utilized with both enzymes (Fig. 1A). The relative incorporation rates were summarized in Table I. Incorporation of [ 14 C]NeuAc toward sialyl Lc4 was lower than toward GM1b in both ST6GalNAc V and VI.
To further confirm that the enzyme products are disialyl Lc4, TLC-immunostaining of the products using an anti-disialyl Lc4 mAb FH9 was performed. As shown in Fig. 1B, the products with ST6GalNAc V and VI clearly showed bands stained with mAb FH9. Anti-disialyl Lc4 mAb FH9 stained the chemically synthesized disialyl Lc4 and a band with the same migration in the products of ST6GalNAc VI (Fig. 2). Thus, the products were confirmed to be disialyl Lc4. Then, we determined their kinetic parameters using various concentrations of two acceptor substrates (i.e. GM1b and sialyl Lc4). The apparent K m showed no marked differences, but V max values were fairly low for sialyl Lc4, resulting in lower V max /K m , compared with GM1b in both enzymes (Table II).
Synthetic Pathway of Disialyl Le a -FUT-3 exhibits both ␣1,3and ␣1,4-fucosyltransferase activity and is the only enzyme that is responsible for the synthesis of type I antigens, such as Le a , Le b , and sialyl Le a (15). Sialyl Lc4, which could be fucosylated by FUT-3, is a precursor of sialyl Le a known as a cancer-associated antigen (22). Disialyl Le a was also suggested to be one of the cancer-associated antigens (7), but its synthetic pathway and function are not clear. To clarify the order of ␣2,6-sialylation and ␣1,4-fucosylation in the synthesis of disialyl Le a , substrate specificity of FUT-3 toward Lc4, sialyl Lc4, and disialyl Lc4 was examined using GDP-[ 14 C]fucose as a donor. FUT-3 mainly acted on Lc4 and sialyl Lc4 and produced Le a and sialyl Le a , respectively (Fig. 3A). It also fucosylated disialyl Lc4, resulting in the new bands, probably disialyl Le a . The enzyme product synthesized from disialyl Lc4 was stained with anti-disialyl Le a mAb FH7 (Fig. 3B), showing its identity to be disialyl Le a . Furthermore, ST6GalNAc V and VI were able to sialylate sialyl Lc4 but not sialyl Le a (Fig. 3C). These results suggest that FUT-3 acted to complete the synthesis of disialyl Le a at the final step and competed with ST6GalNAc V or VI for the common substrate, sialyl Lc4 (Fig. 4).
Synthesis of Disialyl Le a in Cultured Cells-To explore the function of the ST6GalNAc VI gene in vivo, we transiently transfected the expression vectors for ST6GalNAc VI and/or FUT-3 into COS1 cells, which have sialyl Lc4 as a precursor of sialyl Le a and disialyl Lc4, while having no inherent FUT-3 (15). Expression vectors pcDNA3.1-ST6GalNAc VI and pcXN2-FUT-3 were used. The flow cytometric analysis of the transfectant cells with ST6GalNAc VI alone exhibited no expression of sialyl Le a or disialyl Le a antigen as the mock transfectants (Fig. 5, A and B (a, b, e, and f)). When FUT-3 was transfected, not only sialyl Le a but disialyl Le a was expressed, suggesting the inherent ST6GalNAc VI gene expression (Fig. 5, A and B (c  and g)). When ST6GalNAc VI was co-transfected with FUT-3, the expression level of disialyl Le a increased, whereas that of sialyl Le a decreased (Fig. 5, A and B (d and h)). The results of immunocytostaining revealed that disialyl Le a antigen mainly localized in the cytoplasm (Fig. 5B).
Expression of ST6GalNAc VI in Human Colon Cancer Cell Lines-To determine the expression pattern of ST6GalNAc V/VI and FUT-3 mRNA, RT-PCR analysis and Northern blot analysis were performed with seven human colon cancer cell lines (CACO-2, Colo320, DLD-1, HT-29, Lovo, SW1080, and SW1116). RT-PCR was also performed for ␤3Gal-T5 and ST3GalIV genes. To analyze the correlation between the expression levels of disialyl Le a and those of glycosyltransferase genes involved in the synthesis of disialyl Le a , RT-PCR analysis was conducted using the primers based on the human cDNA of ST6GalNAc V/VI, FUT-3, ␤3Gal-T5, and ST3Gal IV. As shown in Fig. 6A, a PCR product of ST6GalNAc VI with 446 base pairs was detected in all human colon cancer cell lines, although the intensity of the bands varied. Lovo cells expressed the ST6GalNAc VI gene at the highest level, whereas SW1080 cells showed a very weak band. In contrast, no bands were detected for the ST6GalNAc V gene in any of them, although the band as positive control was clearly detected with 912 base pairs. Northern blot analysis (Fig. 6B) also indicated the expression of ST6GalNAc VI mRNA with proportional levels to those in RT-PCR but not of V mRNA as shown in Fig. 6A. The  2 is chemically bonded to the oligosaccharides instead of ceramide.

FIG. 2. TLC-immunostaining of enzyme products.
To confirm the structure of the enzyme products, immunostaining was performed using mAb FH9 to detect disialyl Lc4. Lane 1, chemically synthesized disialyl Lc4; lane 2, products of ST6GalNAc VI. The detection was carried out as described under "Materials and Methods." FUT-3 gene was expressed strongly in DLD-1 and SW1116, moderately in HT29, and weakly in Colo320 and Lovo. ST3Gal IV was broadly expressed, and ␤3Gal-T5 was expressed highly in DLD-1 and Lovo and weakly in CACO-2 and HT-29 (Fig. 6C).
To investigate the expression of final enzyme products (i.e. cancer-associated antigens and ␣-series ganglioside antigens in these cell lines), flow cytometric analysis and immunofluorescence assay were carried out. Table III summarizes the expression patterns of the ST6GalNAc V/VI mRNA, cancer-associated antigens, and ␣-series ganglioside antigens in these cell lines.
It was found that ␣-series ganglioside antigens (GD1␣ and GQ1b␣) were minimal or zero in these cell lines, indicating that the main products of ST6GalNAc VI enzyme might be those of type I lacto-series glycolipids in human colon tissues. The expression levels of these antigens on the cell surface appeared to not necessarily correlate with the expression levels of inherent ST6GalNAc VI gene. Two lines with high levels of ST6GalNAc VI gene showed almost null expression of disialyl Le a . But the expression of sialyl Le a was also low in these two lines. Sialyl Le a is also at low levels, suggesting that these two lines lack precursors. On the other hand, cell lines expressing very low levels of ST6GalNAc VI did not express high levels of disialyl Le a except for SW1116. Consequently, there were no critical controversial points in these results to take ST6GalNAc VI as a responsible enzyme for the synthesis of disialyl Le a if we consider other factors such as low precursor levels or low activity of FUT-3.
To inquire about the correlation between the levels of glycosyltransferase genes and the levels of disialyl Le a expression, we further evaluated the expression of responsible genes (␤3Gal-T5, ST3GalIV, and FUT-3) and that of relevant antigens in these cell lines. Fig. 7A showed the expression pattern of these genes, and Fig. 7B is a summary of the antigen expression. The expression levels of FUT-3 correlated rather well with those of disialyl Le a . Otherwise, there was no clear correlation between the expression levels of glycosyltransferase genes and relevant sialyl compounds.
Substrate Competition between ST6GalNAc VI and FUT-3-As we did for COS1 (Fig. 5), we transiently transfected the ST6GalNAc VI and/or FUT-3 expression vectors into six colon cancer cell lines, which originally expressed both genes more or less. Flow cytometric analysis with mAbs FH7 and 1H4 revealed that the overexpression of ST6GalNAc VI cDNA resulted in the elevation of disialyl Le a expression and suppres-  3. Disialyl Le a is synthesized from sialyl Lc4 via disialyl Lc4. A, fucosyltrasferase assay was performed using membrane fraction of cells transfected with FUT-3 as described under "Materials and Methods," and the products were analyzed by TLC. Acceptors were Lc4, sialyl Lc4, disialyl Lc4, nLc4, and Lc4. B, to detect disialyl Le a , TLCimmunostaining was done with mAb FH7 (hybridoma supernatant) at a 1:3 dilution using the TLC plate in A. C, sialyltransferase activities of ST6GalNAc V/VI for fucosylated or nonfucosylated acceptors were analyzed, and the TLC pattern of the products is shown as an autofluorogram.

FIG. 4. Proposed pathway for the biosynthesis of disialyl Lc4
and disialyl Le a . Note that disialyl Le a is synthesized from sialyl Lc4 via disialyl Lc4 but not from sialyl Le a . sion of sialyl Le a expression in many cell lines (DLD-1, Lovo, and SW1116). Two lines (CACO-2 and Colo320) showed no change with transfection of any genes. On the other hand, the transfection of FUT-3 cDNA enhanced the expression levels not only of sialyl Le a but also of disialyl Le a (DLD-1, Lovo, and SW1080). When ST6GalNA VI and FUT-3 cDNA were co-transfected, disialyl Le a levels generally increased compared with the single gene transfection, and sialyl Le a levels were higher than those with the transfection of ST6GalNA VI and lower than those with FUT-3 alone (Fig 8, A and B). Thus, it was confirmed that both ST6GalNAc VI and FUT-3 contribute in the synthesis and expression of disialyl Le a , and they partly compete to share sialyl Lc4 as a common acceptor substrate. DISCUSSION In the present study, we elucidated that ST6GalNAc V/VI, which were isolated as synthases of ␣-series gangliosides (10,11), could catalyze the synthesis of disialyl Lc4, leading to the synthesis of disialyl Le a . ST6GalNAc V was defined as GM1bspecific ␣2,6-sialyltransferase to generate GD1␣ and was expressed specifically in the brain (10). ST6GalNAc VI was also a member of the ␣2,6-sialyltransferase family with the substrate specificity toward not only GM1b but also GD1a and GT1b, leading to the synthesis of GD1␣, GT1a␣, and GQ1b␣, respectively (11). This gene is expressed in many tissues. Sialyl Lc4 appears quite similar to GM1b, and disialyl Le a also resembles GD1␣ in the three-dimensional structure, although sialylated GalNAc at C6 is not fucosylated in GD1␣. Correspondingly, these enzymes discriminated the core structure of sialyl Lc4 and sialyl nLc4 (i.e. type I structure (Gal␤1,3GlcNAc) and type II structure (Gal␤1,4GlcNAc)). Namely, NeuAc␣2,3Gal␤1, 3HexNAc is important for the acceptor recognition. Compared with the synthesis of GD1␣ from GM1b, the efficiency of the synthesis of disialyl Le a is sufficiently high to be expected to actually play roles in cells. Consequently, these enzymes contain a multifunctional character as GM2/GD2/GA2 synthase (23) or GM1/GD1b/GA1 synthase (19) showed. However, this case is very rare since the directly substituted sugars by enzymes are variable (i.e. GalNAc and GlcNAc), although the whole steric connects are very similar between GM1b and sialyl Lc4.
In this study, we clearly demonstrated, for the first time, the biosynthetic pathway of disialyl Le a and the possibility of competition of ST6GalNAc VI with FUT-3 (i.e. synthesis of monosialyl Le a and that of disialyl Lc4 (and disialyl Le a )). The synthesis of disialyl Le a was achieved in COS1 cells by transfection of both ST6GalNAc VI and FUT-3 or FUT-3 alone. The expression rate of disialyl Le a was not so high as expected, partly because the efficiency of co-transfection might be not so high. Sialyl Le a was not converted to disialyl Le a with ST6GalNAc VI. Taken together, sialyl Lc4 is only one substrate examined so far leading to the synthesis of disialyl Le a . Therefore, synthesis of sialyl Le a and disialyl Lc4 (and disialyl Le a ) should compete with each other sharing a common substrate, monosialyl Lc4. Transfection of the expression vector of ST6GalNAc VI into a sialyl Le a -expressing cell line could induce a mild reduction in the expression level of sialyl Le a in addition to new expression of disialyl Le a . These results really suggested the possibility that overexpression of disialyl Le a , based on the action of ST6GalNAc VI, might result in the suppression of the expression level of sialyl Le a . If this is the case, the expression level of ST6GalNAc VI gene might be low in fetal and colon cancers but high in normal colonic mucosa, corresponding with an onco-fetal nature of sialyl Le a . This issue is now under investigation in our laboratory.
Among ST6GalNAc families (I-VI), ST6GalNAc I and II preferentially act on nonsialylated substrates. ST6GalNAc III was poorly active for a type I structure (14), and a human homolog has not yet been reported. ST6GalNAc IV was a protein-dom-inant (O-glycan) enzyme (14). Consequently, we concentrated our efforts on the function of ST6GalNAc V and VI. The results obtained in this study elucidated that these sialyltransferases could act on multiple substrate structures including ganglioseries and lacto-series. Thus, we have defined a novel substrate specificity of previously cloned sialyltransferases.
Whether these sialyltransferases actually exert a catalytic activity in the tissues is a critical point to be clarified. Analyses with seven human colon cancer cell lines showed that ST6GalNAc VI but not V was expressed at various levels as analyzed with RT-PCR. Northern blotting with these cell lines also demonstrated similar results. The intensity in flow cytometry or immunocytostaining of disialyl Le a did not correlate well with the expression levels of ST6GalNAc VI gene. However, two cell lines with poor expression of the ST6GalNAc VI gene also showed only low level expression of disialyl Le a . Two cell lines with high levels of the ST6GalNAc VI gene (CACO-2 and Colo320) scarcely showed disialyl Le a expression. However, they expressed the lowest levels of sialyl Le a among the cell lines examined, suggesting that FUT-3 level and/or sialyl Lc4 level are very low, resulting in inefficient synthesis of disialyl Le a despite high levels of ST6GalNAc VI expression. ␣-Series gangliosides such as GD1␣ and GQ1b␣ were very poor in all of these cell lines. Taken together, it appears quite likely that ST6GalNAc VI is really active and contributes to the synthesis of disialyl Lc4 and disialyl Le a in human colon tissues.
As for the correlation between sialyl Le a /disialyl Le a and FUT3/ST6GalNA VI, it seemed difficult to find a simple competitive relation between these two enzymes by comparing the expression patterns of these genes and their products among cell lines as shown in Fig. 7. This is because the intra-Golgi localization of the ST6GalNA VI and FUT-3 might bias one product over another, resulting in the deviated efficiency of the utilization of the common substrate. Furthermore, the expression levels of other glycosyltransferase genes involved in the synthesis of sialyl Lc4 were various, depending on the cell lines, probably forming different situations in the individual lines. On the other hand, when either one or both of these transferase genes were transfected, we could find the reasonable effects of the expression of individual genes on the expression of sialyl Le a and disialyl Le a as shown in COS1 (Fig. 5) and colon cancer cell lines (Fig. 8). Namely, ST6GalNAc VI is essential for the expression of disialyl Le a , and it partly competes with FUT-3 for the acceptor substrate, sialyl Lc4. However, it simultaneously needs the help of FUT-3 to generate disialyl Le a . The importance of FUT-3 in the synthesis of disialyl Le a was indicated by the fact that FUT-3 expression levels fairly well correlated with disialyl Le a levels (Fig. 7A). Our results of tran-  Fig. 8 demonstrate well the dual aspects of FUT-3 in the synthesis of disialyl Le a .
Beyond the result obtained in this study, the regulatory mechanisms for the expression of disialyl Le a in normal/cancer tissues and the biological roles of the structure in the normal tissues and in colonic malignant cells remain to be investi- gated. New findings obtained in this study will contribute to promote those studies.