Expression Patterns of α2,3-Sialyltransferases and α1,3-Fucosyltransferases Determine the Mode of Sialyl Lewis X Inhibition by Disaccharide Decoys*

A variety of human adenocarcinomas express sialylated, fucosylated Lewis blood group antigens on cell surface and secreted mucins. Binding of these antigens to P-selectin on platelets is thought to facilitate formation of platelet-tumor emboli in the circulation, which in turn allows sequestration of the tumor cells in the microvasculature. Here we report a pharmacologic approach for blocking these interactions through metabolic inhibition of sialylation. Peracetylated forms of Galβ1,4GlcNAcβ-O-naphthalenemethanol and GlcNAcβ1,3Galβ-O-naphthalenemethanol were taken up by LS180 human colon carcinoma cells, O-deacetylated, and utilized as biosynthetic intermediates, resulting in heterogeneous oligosaccharides. The primed oligosaccharides included sialylated, sulfated, and fucosylated products based on mass spectrometry. Assembly of free oligosaccharides on the glycosides decoyed glycosylation of cellular glycoproteins, as assessed by altered binding of lectins and carbohydrate-specific antibodies. Expression of α2,3-sialylated oligosaccharides on the cell surface was diminished specifically, whereas α2,6-sialylation and fucosylation were not. In U937 lymphoma cells, the glycosides decreased fucosylation without affecting sialylation. The differential inhibitory activities correlated inversely with fucosyltransferase and sialyltransferase activity based on enzyme assays and microarray analysis. Regardless of the mechanism, the disaccharides blocked the cells from forming selectin ligands and inhibited adhesion to immobilized selectins, suggesting that the glycosides might prove useful for interfering with tumor cell adhesion and metastasis.

Oligosaccharides containing sialyl Lewis X (sLe x , 1 NeuAc␣2, 3Gal␤1,4[Fuc␣1,3]GlcNAc-) and sialyl Lewis A (sLe a , NeuAc␣2, 3Gal␤1,3[Fuc␣1,4]GlcNAc-) are commonly associated with glycoproteins on the surface of tumor cells (for review, see Ref. 1). Clustered arrangements of these antigens can mediate binding of tumor cells to selectin adhesion receptors on activated endothelia, platelets, or leukocytes, which is thought to increase the probability that tumor cells lodge as "emboli" in the microvasculature of distant organs during metastasis (2)(3)(4)(5)(6)(7). Postsurgical survival studies of patients bearing colon and lung adenocarcinomas indicate higher mortality for individuals whose tumors express sLe x or sLe a (e.g. Refs. 8 -10), suggesting that inhibition of the formation of selectin carbohydrate ligands on tumor cells might reduce tumor metastasis. Reduced metastasis occurs after the removal of selectin ligands from tumor cells genetically (11)(12)(13) or enzymatically (2,3), by blockade of selectin interactions with heparin (14), or by knock-out of Pselectin in the host (3). Thus, a pharmacologic strategy for metabolically altering selectin ligand expression would be desirable, since it might lead to novel therapeutic agents for treating metastatic disease.
Analysis of Primed Oligosaccharides-AcGGn-NM and AcGnG-NM were dissolved in Me 2 SO and added to growth medium to achieve the concentrations indicated in the The conditioned medium was adjusted to 0.5 M NaCl and centrifuged to remove any particulate material, and the supernatants were applied to C18 Sep-Pak cartridges (Waters, Milford, MA) that had been prewashed sequentially with methanol, water, and 0.5 M NaCl. After application of the samples, the columns were washed with 0.5 M NaCl (2.5 ml) and water (25 ml). Radioactive oligosaccharides were eluted with 40% methanol. The products were analyzed by anion-exchange chromatography, (QAE-Sephadex; Sigma) to separate the charged and neutral radiolabeled products (27). A sample of charged radiolabeled oligosaccharides was treated with Arthrobacter ureafaciens sialidase (20 milliunits/10 6 cells (Calbiochem)) and analyzed by reverse phase HPLC (TosoHaas RP18, 4.6 mm ϫ 25 cm, Tosoh Biosep LLC, Montgomeryville, PA) connected to a HPXL solvent delivery system (Rainin Instrument Company, Oakland, CA). The column was first washed with water (flow rate 0.5 ml/min) and then with increasing amounts of acetonitrile in water. Radioactivity in the eluant was monitored by the Radiomatic Flo-one Beta detector (PerkinElmer Life Sciences) connected in-line to the column.
Mass spectrometric data were recorded on a Thermo Finnigan LCQ Lung classic mass spectrometer (San Jose, CA). Samples containing primed products in water (10 l) were introduced into the electrospray ionization ion trap using an autosampler connected to a TSP P400 quaternary pump using 50% acetonitrile containing 0.1% formic acid. The data were fit to compositions of known oligosaccharides present in animal cells.
Binding of CSLEX-1 and Selectin Chimeras to Tumor Cells-To measure the presence of selectin ligands, cells were grown to confluence for 4 -7 days in six-well plates. Binding of CSLEX-1 and selectin/IgG chimeras (Ps-Ig and Es-Ig) (28) was performed as described (24), except the enzyme-linked immunosorbent assay incubation buffer consisted of Dulbecco's phosphate-buffered saline with calcium (Invitrogen), 1% bovine serum albumin, 0.01% NaN 3 , and 1% goat serum. U937 cells were incubated with CSLEX-1 for 30 min, PC-3 cells for 3-4 h, and all other cell lines for 2 h in order to maximize binding. Incubations with Es-Ig (0.125-0.625 g/ml) and Ps-Ig (0.1-0.5 g/ml) were done for 1 h. In some experiments, cells were first treated for 1 h at 37°C with A. ureafaciens sialidase in 100 l of 0.05 M HEPES (pH 6.9) or 2 mM EDTA in phosphate-buffered saline.
Cell Adhesion to Immobilized Selectins-96-Well plates were coated overnight at 4°C with recombinant E-or P-selectin (2-16 g/ml) and blocked with 1% bovine serum albumin in phosphate-buffered saline. The plates were coated with E-selectin at 2 g/ml for U937, LS180, BT-20, WiDr, and A549; 4 g/ml for Hal-8 and A427; and 16 g/ml for H460. For P-selectin, the optimal concentrations were 1 g/ml for A427; 2 g/ml for LS180, A549, and Hal-8; and 6 g/ml for U937. Tumor cells were harvested with EDTA (5 mM, 20 min), resuspended in Dulbecco's modified Eagle's medium containing 1% fetal bovine calf serum, labeled with Calcein AM (5 M; Molecular Probes, Inc., Eugene, OR), and added to the selectin-coated wells (5 ϫ 10 4 cells/well). The cells were allowed to settle for 25 min at room temperature, and the plates were then stirred at 75 rpm on an orbital shaker for 15 min. Nonadherent cells were removed at unit gravity by inverting the plates for 15 min in a vessel filled with phosphate-buffered saline (29). For Hal-8, A427, A549, H460, and MV522, the plates were inverted for only 2-5 min, since the cells were much less adherent. After inversion, the wells were aspirated, and the amount of fluorescence was measured using a computerized fluorimeter (CytoFluor II; Promega, Madison, WI), averaging the values of four wells for each condition. Controls included treating cells for 1 h at 37°C with A. ureafaciens sialidase (20 milliunits/1 ϫ 10 6 cells) in 0.05 M HEPES buffer (pH 6.9), pretreating wells with anti-E-selectin or anti-P-selectin monoclonal antibody (1 g/well), or growth of cells in a 50 M concentration of the inactive disaccharide, AcGal␤1,3Gal␤-O-NM. Cell viability was judged to be Ͼ90% by trypan blue exclusion at the completion of each experiment.
Sialyltransferase and Fucosyltransferase Assays-Total sialyltransferase and fucosyltransferase (FucT) activities were assayed in cell lysates prepared from U937 and LS180 cells. After washing the cartridge with 25 ml of water, the products were eluted with 50% methanol, dried, and counted by liquid scintillation. Reactions containing Sia␣2,3Gal␤1,4GlcNAc were diluted with 0.5 ml of 5 mM PO 4 (pH 6.8) and applied to a small column of Dowex 1 (X8 -400, phosphate form). Reaction products eluted in the flow-through and the wash (2 ml of water) (30).
Sample Preparation and Processing for Microarray Analysis-U937 and LS180 cells were grown in triplicate in 150-mm diameter culture dishes until confluent. RNA from each dish was isolated using TRIzol reagent (Invitrogen) and further purified using Rneasy (Qiagen, Valencia, CA) and kept frozen at Ϫ80°C. For each sample, 5 g of total RNA was prepared using standard Affymetrix (Santa Clara, CA) Gene-Chip protocols for all labeling, staining, and scanning procedures (available on the World Wide Web at affymetrix.com) (31). Each labeled sample was hybridized to individual glyco-v1 chips, an oligonucleotide array custom designed for the Consortium for Functional Glycomics (available on the World Wide Web at web.mit.edu/glycomics/consortium). Briefly, each transcript was detected by one or more probe sets generally consisting of 11 probe pairs. Each probe pair is made up of one 25-base pair Perfect Match oligonucleotide that matches the sequence of the targeted transcript and an oligonucleotide designed with a mismatch at the center position. A complete list of probe sets and annotation for the glyco-v1 oligonucleotide array is available on the World Wide Web at web.mit.edu/glycomics/consortium/resources/resource coree.shtml.
All GeneChip data were analyzed using Microarray Suite 5.0 (MAS 5.0) (Affymetrix) and DNA Chip Analyzer (dChip) (32,33) software. The hybridization signal intensities were generated in dChip using the "Perfect Match-only" model and normalized to the median array of the analyzed set. "Present," "Marginal," and "Absent" calls were based on "detection" p values determined using Affymetrix MAS 5.0 software with all user-definable parameters set at the Affymetrix default values. The dChip software used all of the Affymetrix .CEL files generated in this study as a training set. Hierarchical clustering using centered correlation and average linkage was performed with the software BRB ArrayTools (available on the World Wide Web at linus.nci.nih.gov/BRB-ArrayTools.html) developed by Dr. Richard Simon and Amy Peng at the Biometric Research Branch of the NCI, National Institutes of Health.
Changes in gene expression between the study groups were determined using the following methodologies. To identify statistically significant changes in gene expression, significance analysis of microarrays was used (34). Delta values were chosen that minimized the median false discovery rate. Two additional methods were used to filter the gene list. First, we applied the limit -fold change model, which systematically bins genes by signal intensity; those genes within the top 5-10% of the highest -fold changes for each bin are selected (35). Second, Affymetrix MAS 5.0 present/absent calls were used to filter the list. When comparing an expression measurement for a gene between two groups, we required the majority of calls in the up-regulated group to be "present." Probe sets not meeting this criterion were eliminated from further analysis.

RESULTS AND DISCUSSION
The carbohydrate antigens sLe x and sLe a are expressed on leukocytes, epithelial cells, and many carcinomas in configurations that can be recognized by one or more selectin receptors. A survey of established tumor lines using a monoclonal antibody to sLe x (CSLEX-1) showed that many express sLe x -containing antigens, with the exception of the breast carcinomas, MCF-7 and SK-BR-3 (Table I). The pattern of reactivity correlated in most cell lines with binding of recombinant P-selectin-Ig and E-selectin-Ig chimeras, but not in all cases (e.g. A427 and A549 lung carcinomas, SK-BR-3 breast carcinoma, and PC-3 prostate carcinoma), consistent with the heterogeneity of carbohydrate ligands for these selectins. The LS180 human colon carcinoma was chosen for further study, since it expresses ligands that bound to CSLEX-1 and the selectin chimeras (Table I).
Peracetylated disaccharides, such as AcGGn-NM and AcGnG-NM, metabolically inhibit the expression of sLe x in U937 lymphoma cells. Cells take up the glycosides by passive diffusion, deacetylate them with endogenous carboxyesterases, and use them as substrates for glycosyltransferases, resulting in the generation of oligosaccharides containing Lewis type antigens or precursor structures (23)(24)(25). Both compounds are nontoxic up to 100 M based on trypan blue exclusion and normal growth rates. To determine whether the disaccharides had similar properties in LS180 colon carcinoma cells, they were added to the culture medium along with radioactive fucose, galactose, glucosamine, or sulfate to label newly made oligosaccharides. Both compounds stimulated the incorporation of radiolabeled sugars and sulfate into glycans assembled on the exogenous primers, confirming that they were taken up, deacetylated, and made available to the glycosyltransferases in the cell (data not shown). Control assays with no added acceptor were also performed, and the signal was subtracted from samples that contained added substrate. The assays were done in duplicate and varied by less than 10%.
The acetylated disaccharides resemble in behavior N-acetylgalactosaminides (e.g. GalNAc␣-O-Bn), which without prior acetylation are taken up by most types of cells. However, Gal-NAc␣-O-Bn impacts the first committed step in O-linked mucin biosynthesis, whereas the acetylated disaccharides tap into the biosynthetic pathways at later stages of oligosaccharide assembly. Furthermore, the disaccharide-based compounds are active at much lower concentrations (Ն10 M versus Ն0.5 mM, respectively). As shown below, the acetylated disaccharides show more selective effects than GalNAc␣-O-Bn.
To examine the full array of oligosaccharides generated on the disaccharide primers, samples were analyzed by electrospray mass spectrometry (Table II). Several patterns were present, such as one or more Gal␤1,4GlcNAc␤1,3 units on GGn-NM and the addition of a Gal residue to GnG-NM, consistent with the presence of ␤1,4Gal and ␤1,3GlcNAc transferases. Fucose was added to several oligosaccharides primed on GGn-NM and GnG-NM, presumably in the latter case after the addition of a Gal residue (24). The oligosaccharides generated on GGn-NM and GnG-NM consisted of both neutral and charged species, and the charged products contained sialic acid and/or sulfate residues. Most of the charged species consisted of sialylated oligosaccharides based on their conversion to neutral oligosaccharides by treatment with a nonspecific sialidase. Fig.  1 shows that the charged fucosylated oligosaccharides were mostly sialylated (the arrows indicate the position of neutral oligosaccharides derived from the charged compounds after treatment with sialidase). The resistant material contained sulfate as assessed by labeling studies with 35 SO 4 (data not shown). Both primers generated oligosaccharides that resembled Lewis and sialylated Lewis type antigens (Table II).
Disaccharide Treatment Alters Glycosylation-To assess whether priming of oligosaccharides on the glycosides altered glycosylation of endogenous glycoconjugates, treated cells were analyzed by flow cytometry after reaction with fluorescent plant lectins and monoclonal antibodies that bind to specific carbohydrates. The signal intensity of untreated cells was set to 100% for each lectin or antibody, and the relative value for treated cells was determined (Fig. 2). In LS180 cells, staining by M. amurensis hemagglutinin, which recognizes terminal ␣2,3-linked sialic acid residues and 3-O-sulfated galactose (36), was reduced by both disaccharides. Binding was completely abolished by sialidase treatment (data not shown), indicating that the major determinant expressed by LS180 cells was a sialylated component. Binding of the monoclonal antibody CD15, which recognizes Le x (Gal␤1,4(Fuc␣1,3)GlcNAc-), increased or did not change, suggesting that fucosylation was not inhibited. Peanut agglutinin, which recognizes Gal␤1, 3GalNAc␣-, strongly stained disaccharide-treated cells, suggesting a parallel decrease in sialylation of core 1 O-glycans. In contrast, staining by S. nigra agglutinin increased. This lectin recognizes ␣2,6-linked sialic acid-containing structures, suggesting that the disaccharides selectively decreased ␣2,3-sialylation.
Malignant transformation is often accompanied by increased ␤1,6 branching of Asn-linked glycans in glycoproteins, which has been correlated with tumor growth (37). No difference between control and disaccharide-treated cells was observed by P. vulgaris leukoagglutinin binding, suggesting that the formation of the branched structures was unaffected. Staining with L. esculentum lectin, S. tuberosum lectin, and E. cristagalli lectin, which recognize poly-N-acetyllactosamine, and R. communis agglutinin I (terminal ␤-linked galactose) revealed modest reduction in reactivity (Fig. 2). In general, treatment of LS180 cells with AcGGn-NM had similar but less profound effects on lectin staining than AcGnG-NM. On the other hand,  GalNAc␣-O-Bn had very different effects, profoundly reducing both ␣2,3-sialylation (M. amurensis hemagglutinin) and fucosylation (CD-15) and increasing the reactivity of cells to P. vulgaris leukoagglutinin, L. esculentum lectin, S. tuberosum lectin, E. cristagalli lectin, and in particular S. nigra agglutinin and peanut agglutinin. U937 cells primed many of the same oligosaccharides as in LS180 cells (Table II and Ref. 24), but they differed significantly in lectin reactivity after disaccharide treatment (Fig. 2). In U937 cells, both disaccharides significantly increased MAH staining and only modestly affected S. nigra agglutinin binding. Fucosylation decreased, based on reaction with A. aurantia lectin, which recognizes ␣1,3-linked fucose residues, in agreement with previous studies (24). GalNAc␣-O-Bn decreased both sialylation and fucosylation in these cells.
In order to examine whether specific transferases were affected, mRNA samples were isolated from U937 and LS180 cells and analyzed using the Glyco-v1 GeneChip microarray. Expression signals were calculated using the Perfect Matchonly model of dChip (33). Microarray expression signals were analyzed by cluster analysis in order to reveal similarities among microarray data sets, and changes in gene expression were tested statistically across the two cell types. Affymetrix MAS 5.0 was used to make present, absent, and marginal calls for each gene expression measurement.
Hierarchical cluster analysis of expression data showed that a number of genes can be identified with high confidence as differentially expressed between the two cell types (data not shown). Interestingly, only ST3GalIV was detected in LS180 cells (Table III) linked oligosaccharides (Gal␤1,3GalNAc␣-) and oligosaccharides that terminate in Gal residues in a tissue-and glycoprotein-specific manner (38,39). Absence of the enzyme results in enhanced binding of peanut agglutinin, like that seen in disaccharide-treated cells (Fig. 2). ST3GalIV also appeared to be expressed at comparable levels in U937 cells, and ST3GalV and ST3GalVI were also present (40,41). These isozymes are thought to be involved in ganglioside biosynthesis. The lack of inhibition of sialylation in these cells suggests that the individual isozyme is less important than the overall level of activity (Fig. 3).
The expression data for the ␣1,3-fucosyltransferases also showed striking differences (Table IV). U937 cells expressed FucTIV and FucTVII, in agreement with previous studies of myeloid cells (42)(43)(44). The disaccharides affected fucosylation in this cell line (Fig. 2), suggesting that one or both isozymes utilize the disaccharides in vivo. LS180 cells expressed FucTIII and FucTVI (Table IV). These enzymes also apparently utilize the disaccharides, since fucosylated products were generated (Table II), but priming was not sufficient to block fucosylation of glycoconjugates on the cell surface (Fig. 2). The high overall level of fucosyltransferase activity in LS180 cells compared with U937 cells may explain the apparent lack of inhibition by the disaccharides.
Inhibition of Selectin Ligand Presentation-To test whether the decrease in ␣2,3-sialylation in LS180 cells led to depression of sLe x expression, cells were reacted with CSLEX-1, a monoclonal antibody that reacts with sLe x -containing oligosaccharides, and chimeras of E-selectin-Ig and P-selectin-Ig. The amount of bound reagent was assessed by enzyme-linked immunosorbent assay (Fig. 4). Both AcGGn-NM and AcGnG-NM decreased binding in a dose-dependent fashion. In general, AcGnG-NM was more effective than AcGGn-NM, with inhibition occurring at doses as low as ϳ15 M. Control experiments showed that treatment of the cells with sialidase or 2 mM EDTA abrogated binding of CSLEX-1 and E-selectin and P-selectin chimeras, as expected for selectin-dependent processes. Treatment of cells with 50 M AcGG-NM, which is not related in structure to oligosaccharides bearing sLe x , did not affect binding of CSLEX-1 (data not shown).
Many of the tumor lines expressing P-selectin and E-selectin ligands bound to plates coated with recombinant selectins (Table I). Cell adhesion to immobilized selectins is a multivalent process and therefore should be very sensitive to alterations in ligand presentation due to decreased sialylation. When treated with AcGGn-NM or AcGnG-NM, adhesion of LS180 cells was diminished in a dose-dependent manner, reaching ϳ80% inhibition at the highest dose tested (Fig. 5). Inhibition was similar to that caused by pretreatment of the cells with sialidase or blocking antibodies. Cells treated with acetylated Gal␤1, 3Gal␤-O-NM, which did not diminish sLe x expression, did not affect cell adhesion, indicating the specificity of the effects.
Summary and Perspective-Acetylated disaccharides that resemble the core structure of Lewis antigens prime oligosaccharides in LS180 colon carcinoma cells. Priming in this way inhibits ␣2,3-sialylation and inhibits the formation of sialylated Lewis antigens recognized by selectins. In U937 cells, the disaccharides also caused an inhibition of selectin ligands, but in this case, the effect is due to altered fucosylation. The inhibition of sialylation and fucosylation by disaccharides validates the relevant ␣2,3-sialyltransferase (ST3GalIV) and ␣1,3-fucosyltransferases (FucTIV and FucTVII) expressed in sensitive tumor lines as a target for anti-selectin-based therapies.
Previous studies have shown that GalNAc␣-O-Bn can inhibit sialylation or fucosylation dependent on cell type (45)(46)(47)(48). Gal-NAc␣-O-Bn in LS180 and U937 cells suppressed both ␣2,3sialylation and ␣1,3-fucosylation and dramatically increased ␣2,6-sialylation, indicating a lack of specificity. Thus, acetylated disaccharides have several distinct advantages over monosaccharide primers such as GalNAc␣-O-Bn, including efficacy at much lower concentrations and greater specificity. Furthermore the presence of two monosaccharide residues in the disaccharide primers provides a larger framework for making functional group analogs, which may prove successful as active site-directed inhibitors of the relevant ␣2,3-sialyltransferase and ␣1,3-fucosyltransferases.
The differential sensitivity of sialylation and fucosylation in LS180 and U937 cells appears to reflect differences in the level of expression of the relevant fucosyltransferases and sialyltransferases, with an inverse correlation to their overall activity. The microarray data provided additional insight by providing evidence for the selective effects on subsets of enzymes expressed in each cell type. Together, the findings suggest that one might be able to predict how pharmacological agents will affect glycosylation by simply measuring overall transferase activities and by profiling the enzyme mRNA levels. Since expression of the transferases may be regulated both transcriptionally and translationally, deviations from the expected patterns could occur. Other factors may play a role as well, including availability of nucleotide sugars and flux of endogenous substrates coming through the system. Nevertheless, the combination of priming data, lectin binding, enzyme assays, and microarray data provides a way to evaluate the potential efficacy of at least primer-based decoys and perhaps other inhibitory agents that act on glycosylation. Application of this information to particular tumor cell types may optimize drug discovery and development for treating cancer and other disorders, such as inflammation, in which selectin-carbohydrate interactions play a pathophysiological role.