6'-Sulfo sialyl Lex but not 6-sulfo sialyl Lex expressed on the cell surface supports L-selectin-mediated adhesion.

In order to determine if a sulfated oligosaccharide on the cell surface can function as an L-selectin ligand, a novel approach for in vitro transfer of oligosaccharides was utilized (Srivastava, G., Kaun, K. J., Hindsgaul, O., and Palcic, M. M. (1992) J. Biol. Chem. 267, 22356-22361). CHO cells were incubated with synthetic 6′-sulfo sialyl Lex, NeuNAcα2→3(sulfate-6)Galβ1→4(Fucα1→3)GlcNAc or 6-sulfo sialyl Lex, NeuNAcα2→3Galβ1→4[(Fucα1→3)sulfate→6GlcNAc] oligosaccharide linked to C-6 of a fucose residue in GDP-fucose and a milk fucosyltransferase. The resultant CHO cells expressing 6′-sulfo sialyl Lex or 6-sulfo sialyl Lex on their cell surface were tested for adhesion to E-selectin and L-selectin chimeric proteins coated on plates. The results indicate that 6′-sulfo sialyl Lex supports L-selectin-mediated adhesion much better than sialyl Lex similarly tagged on the cell surface. In contrast, 6-sulfo sialyl Lex containing a sulfate group on the N-acetylglucosamine residue did not support adhesion with either selectin. These combined results suggest that 6′-sulfo sialyl Lex is a much better ligand than sialyl Lex oligosaccharide for L-selectin.

It has been suggested that among the members of the selectin family, E-and P-selectin bind to sialyl Le x1 NeuNAc␣233Gal␤134(Fuc␣133)GlcNAc3 present on neutrophils and other leukocytes (1)(2)(3)(4), although the identity of the real physiological epitope remains to be clarified (5). In contrast, the natural ligand for L-selectin has not been determined yet (6,7). Several lines of studies indicate, however, that a sulfated sialylated and fucosylated oligosaccharide(s) is the natural ligand for L-selectin. For example, the binding of Lselectin was abolished by pretreatment of high endothelial venules with neuraminidase (8). Fucoidin, a highly sulfated fucose polymer, efficiently inhibits L-selectin binding (9). Moreover, the inhibition of sulfation in those cells expressing Lselectin ligands resulted in the loss of L-selectin binding to ligands (10).
The binding of L-selectin and P-selectin with high affinity, however, can be achieved only when the ligands are presented on specific glycoproteins. For L-selectin, it has been shown that GlyCAM-1, CD34, and MadCAM-1 are the major glycoproteins that express L-selectin ligands with high affinity (11)(12)(13). These three glycoproteins contain amino acid sequences that carry mucin-type O-glycans, allowing multiple presentation of oligosaccharide ligands. It has been also demonstrated that sialyl Le x in O-glycans can be formed only in N-acetyllactosamine extensions formed by core 2 branches (14,15). Later it was shown that core 2 branches could be newly expressed in CHO cells by introducing core 2 ␤-1,6-N-acetylglucosaminyltransferase, C2GnT (16).
Recent studies have shown that GlyCAM-1 contains the sulfated forms of sialyl Le x , and the major sulfated oligosaccharide appears to be NeuNAc␣233(sulfate36)Gal␤134(Fuc␣133) GlcNAc3, 6Ј-sulfo sialyl Le x (17,18). In addition, sulfate36 GlcNAc was detected in the acid hydrolysate of the oligosaccharides derived from GlyCAM-1 (19), suggesting that Neu-NAc␣233Gal␤134(sulfate36)GlcNAc, 6-sulfo sialyl Le x could be present in GlyCAM-1 oligosaccharides. Although these studies demonstrate that 6Ј-sulfo sialyl Le x and possibly 6-sulfo sialyl Le x are present in GlyCAM-1, no studies have been published so far demonstrating that either of these structures actually supports the L-selectin-mediated adhesion.
In order to form a sulfated, sialylated Le x from the precursor N-acetyllactosamine, a sialyltransferase(s), a fucosyltransferase(s), and a sulfotransferase(s) are required. Although a sialyltransferase (20,21) and fucosyltransferase (Fuc-TIII and Fuc-TVII) (22)(23)(24) have been cloned, so far no cDNA encoding the sulfotransferase to form this structure has been cloned. In order to test if a sulfated sialylated oligosaccharide serves as a ligand for L-selectin, we thus elected in the present study to employ a novel method for transferring oligosaccharides to the cell surface of CHO cells by a combination of synthetic chemistry and enzymatic transfer (25). In this method, oligosaccharides attached to C-6 of GDP-fucose can be transferred to C-3 or C-4 of N-acetyllactosamine or Gal␤133GlcNAc as an intact oligosaccharide unit linked to C-6 of fucose (25) (see Fig. 1). Because a milk fucosyltransferase can add fucose to 3Ј-sialylated oligosaccharides as well (26,27), many terminal oligosaccharides can be acceptors for this reaction. We then tested the resulting CHO cells expressing 6Ј-sulfo sialyl Le x or 6-sulfo sialyl Le x for adhesion to E-or L-selectin chimeric protein. The results clearly demonstrate that 6Ј-sulfo Le x is a better ligand for L-selectin than sialyl Le x , and 6-sulfo sialyl Le x does not support the adhesion to either L-selectin or E-selectin.

EXPERIMENTAL PROCEDURES
Establishment of CHO Cells Stably Expressing CD34 and C2GnT-CHO DG44 cells expressing C2GnT and leukosialin, CHO-leu⅐C2GnT, were established as described (16). CHO-leu⅐C2GnT cells were transfected by the LipofectAMINE method (28) with pCDM8-CD34 (29) and pHyg in a 10:1 molar ratio, as described (30). Clonal cell lines were * This work was supported by Grants PO1 CA71932, RO1 CA48737, and in part by R37 CA33000. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This article is dedicated to the late Dr. Thomas Kuhn, who is responsible for revolutionary changes in the philosophy of science.
‡ Initially supported by a fellowship from the Toyobo Biotechnology Foundation.
Transfer of Oligosaccharides in GDP-Fuc-Oligosaccharides Using a Milk Fucosyltransferase-␣133/4-Fucosyltransferase was partially purified from human milk as described (25). CHO-leu⅐C2GnT⅐CD34 (4 ϫ 10 6 ) cells, attached to each of 6-well chambers, were incubated with 2.4 milliunits of the purified fucosyltransferase and 174 nmol of the GDP-Fuc-oligosaccharide derivatives in 1 ml of Opti-minimum essential medium containing 7 mM MnCl 2 , which was titrated to pH 7.0 by 1 N HCl. After incubation in a CO 2 incubator at 37°C for 16 h, the cells were washed with phosphate-buffered saline and resuspended in ␣-minimum essential medium. Resultant CHO cells were labeled with Tran 35 S as described (30). These cells tagged with sialyl Le x , 6Ј-sulfo sialyl Le x , and 6-sulfo sialyl Le x oligosaccharides are named CHO-sialyl Le x , CHO-6Јsulfo sialyl Le x , and CHO-6-sulfo sialyl Le x , respectively.
Adhesion Assays-E-selectin-Fc chimera and L-selectin-Fc chimera were purified as described (30,33). The above-mentioned monodispersed CHO cells (5 ϫ 10 4 cells) were suspended in 100 l of Dulbecco's modified Eagle's medium containing 5% fetal calf serum and added to each of 96 wells coated with E-selectin-Fc chimera protein or L-selectin-Fc chimera protein, coated as described (33). In the present study, the solution containing 500 g of L-selectin/ml was used for its coating since no efficient adhesion was achieved at its lower concentrations. After incubation for 20 min at 4°C (30,33), the unbound cells were washed with 100 l of minimum essential medium four times. The remaining adherent cells were released by the cell dissociation solution (Specialty Media, Lavallette, NJ), and the radioactivity was determined by scintillation counting (30,33).
Estimation of the Number of Sialyl Le x and Sulfo Sialyl Le x Oligosaccharides on the Cell Surface-No antibodies specific to 6Ј-sulfo sialyl Le x or 6-sulfo sialyl Le x are available. Moreover, anti-sialyl Le x antibody (CSLEX-1, Becton Dickinson) is IgM, which is a pentameter, and the number of oligosaccharides cannot be accurately determined by using anti-sialyl Le x antibody. This antibody did not react with 6Ј-sulfo sialyl Le x or 6-sulfo sialyl Le x , either.
In order to circumvent the problem, we determined the number of binding sites of Maachia amurensis agglutinin (MAA) (34) to ␣-2,3linked sialic acid in the oligosaccharides. For this, sialyl Le x -BSA conjugates (25) were first coated on plates, and the number of binding sites of mouse monoclonal anti-BSA antibody (Sigma) to the conjugates was measured according to the procedure of Ho and Springer (35), as described (36). Rabbit serum albumin (Organon Teknika) was used to block nonspecific binding because anti-BSA antibody did not cross-react with rabbit serum albumin. From this experiment, the amount of BSA in each well was estimated as 0.034 pmol of BSA. Since, on average, 15 oligosaccharides were attached to each molecule of BSA, 2 this predicts that 0.509 pmol of oligosaccharides are present in each coated well. In parallel, the number of binding sites of MAA to sialyl Le x -BSA conjugates was determined by addition of biotinylated MAA (Vector) followed by 125 I-streptavidin (Amersham). At the saturation levels, the amount of bound MAA should be equivalent to 0.509 pmol/well. This measurement thus allowed us to build a calibration curve.
To estimate the amount of the sialyl Le x or sulfo sialyl Le x oligosaccharides on the cell surface, various amounts of biotinylated MAA were added at 4°C to various CHO cells tagged with oligosaccharides. The amount of bound MAA was determined by binding of 125 I-streptavidin at 4°C.

Transfer of Oligosaccharides Conjugated to GDP-Fuc-CHO
cells were incubated with the fucosyltransferase and GDP-Fucsialyl Le x , GDP-Fuc-6Ј-sulfo sialyl Le x , or GDP-Fuc-6-sulfo sialyl Le x . As shown in Fig. 2A, the resultant CHO-sialyl Le x cells express a substantial amount of sialyl Le x , detected by immunofluorescent staining using anti-sialyl Le x antibody. Fig. 2B shows the corresponding immunostaining of CHO-FTIII cells, which were produced by gene transfer of Fuc-TIII (22,31). The staining pattern of CHO-sialyl Le x is distinct from that of CHO-FTIII cells in that CHO-sialyl Le x cells show more clustered staining ( Fig. 2A).
In order to estimate how much sialyl Le x , 6Ј-sulfo sialyl Le x , or 6-sulfo sialyl Le x oligosaccharides were transferred, the binding assays using MAA were performed on these cells. Control CHO cells were found to have 1.84 ϫ 10 5 binding sites of MAA (Fig. 3A), since they express ␣-2,3-linked sialic acid (37,38). Assuming that ␣-2,3-linked sialic acid in sialyl Le x , 6Ј-sulfo sialyl Le x , and 6-sulfo sialyl Le x equally bind to MAA, the binding sites of MAA in these three CHO cells were determined. The results demonstrated almost identical numbers of binding sites of MAA (4.39 ϫ 10 5 , 4.35 ϫ 10 5 , and 4.64 ϫ 10 5 ) in CHO-sialyl Le x , CHO-6Ј-sulfo sialyl Le x , and CHO-6-sulfo sialyl Le x (Fig. 3, B, C, and D). By subtracting the amount of ␣-2,3-linked sialic acid present in control CHO cells, it can be concluded that 2.51 to 2.80 ϫ 10 5 /cell of sialyl Le x , 6Ј-sulfo sialyl Le x , and 6-sulfo sialyl Le x were transferred to the cell surface of CHO cells. These results also corroborated our hypothesis that ␣-2,3-linked sialic acid in different oligosaccharides bind almost equally to MAA, providing that the efficiency 2 Y. Isogai and O. Hindsgaul, manuscript in preparation. of oligosaccharide transfer is not influenced by their structural differences. Fig. 4 demonstrated clearly that CHOsialyl Le x adhered well to E-selectin chimera (column 2), while the adhesion was minimum to control human IgG proteins (column 3). The adhesion was completely inhibited by preincubation of the cells with anti-sialyl Le x antibody (column 4), anti-E-selectin antibody, or by the addition of 5 mM EDTA (data not shown), but not by control mouse IgM (column 5). Surprisingly, CHO-6Ј-sulfo sialyl Le x cells adhered well to E-selectin (column 7). However, almost no adhesion was detected for CHO-6-sulfo sialyl Le x cells (column 10).

Adhesion of CHO-Sialyl Le x and CHO-6-Sulfo Sialyl Le x Cells to E-selectin-
Adhesion of CHO-6Ј-Sulfo Sialyl Le x and CHO-Sialyl Le x to L-selectin-As shown in Fig. 5, CHO cells expressing 6Ј-sulfo sialyl Le x adhered well to L-selectin chimera (column 5). This adhesion could be inhibited by preincubation with anti-L-selectin antibody (column 7), but not by control mouse IgG (column 8). In contrast, sialyl Le x oligosaccharides on CHO cells modestly supported the adhesion to L-selectin (column 2). Apparently, the expression of CD34 is critical since CHO cells which were not transfected with CD34 cDNA did not support the adhesion to L-selectin (data not shown). More strikingly, CHO cells expressing 6-sulfo sialyl Le x did not adhere to L-selectin at all (column 10). These combined results indicate that 6Ј-sulfo sialyl Le x is a much better ligand for L-selectin than sialyl Le x while it is a slightly less efficient ligand for E-selectin than sialyl Le x . 6-Sulfo sialyl Le x on the other hand is hardly bound to either E-or L-selectin.
Previous studies have demonstrated that GlyCAM-1, which carries L-selectin ligands, can have a sulfate group in C-6 of galactose, 6Ј-sulfo sialyl Le x and C-6 of N-acetylglucosamine, 6-sulfo sialyl Le x (17,19). The present study clearly indicates that 6Ј-sulfo sialyl Le x is a better ligand for L-selectin than sialyl Le x . These results are consistent with the previous report that sulfation was required for efficient binding of L-selectin to GlyCAM-1 (10). However, we also found in the present study that sialyl Le x is an inefficient yet moderate ligand for Lselectin. These results are consistent with the recent report on the inhibition of L-selectin-mediated adhesion using synthetic oligosaccharides; their results indicate that sialyl Le x is as effective as sulfate33(sulfate36)Gal␤134Glc for inhibiting the binding of L-selectin to GlyCAM-1 (39). The present study also demonstrated that 6-sulfo sialyl Le x is not an efficient ligand for either E-or L-selectin. It is possible that the 6-sulfo sialyl Le x structure present in GlyCAM-1 does not serve well as a ligand for L-selectin. It was also reported that 6-sulfo sialyl Le x inhibits L-selectin binding to adressin (42). In those studies, however, only inhibition assay was employed, and the comparison with 6Ј-sulfo sialyl Le x was not tested.
Recent crystallographic studies showed the structure of the carbohydrate-binding domain of E-selectin (40). In that study, sialyl Le x tetrasaccharide was tentatively modeled based on the structure of the complex of mannose oligosaccharide-mannose-binding protein (41) and NMR data on sialyl Le x oligosaccharide (40). These results suggest that one of the critical amino acids involved in its binding may be glutamic acid 92 (40,41). It is thus tempting to speculate that a sulfate group at C-6 of N-acetylglucosamine in 6-sulfo sialyl Le x causes steric hindrance as well as charge repulsion to glutamic acid 92. It is also noteworthy that E-and L-selectin have a strong homology in the carbohydrate-binding domain, including those amino acids critical for their binding to carbohydrate ligands shown for mannose-binding protein (40,41,43). It is thus not surprising that only quantitative differences in efficiency can be found between sialyl Le x and 6Ј-sulfo sialyl Le x for L-selectin-and E-selectin-mediated adhesion, as shown in the present study.
As described in the introduction, the chemical and enzymatic tagging is powerful when testing the activity of oligosaccharides that cannot be synthesized on cells by gene transfer. We expect that the method employed in the present study will be an important tool in determining the roles of oligosaccharides on the cell surface in various studies.