INTRODUCTION

Selectins are part of a family of cell adhesion molecules involved in immune cell trafficking (1-7). Cytokine activated endothelial cells express on their lumenal surface E- and P-selectins, which bind to carbohydrate ligands related to the Lewis blood group antigens expressed on monocytes, neutrophils, and certain subsets of T-cells. When engaged, the receptor-ligand complexes facilitate leukocyte rolling on the endothelium, which later gives way to strong adhesion and extravasation of cells into the underlying tissue. L-selectin on the surface of the leukocytes plays a similar role, possibly by facilitating aggregation of neutrophils (8-10). Although these adhesion events help defend against infection and facilitate tissue repair, they sometimes go awry and cause inadvertent tissue destruction (e.g. during ischemia-reperfusion or trauma-induced tissue damage). Therefore, much interest exists in developing agents to inhibit the adherence of leukocytes to endothelium as a way of modulating deleterious inflammatory reactions.

Strategies for inhibiting selectin-carbohydrate interactions include competition by soluble recombinant forms of selectins (11), peptides based on the primary sequence of the carbohydrate binding site (12, 13), anti-selectin antibodies (14-16), oligosaccharides related to Lewis^A and Lewis^X (7, 17-24), inositol polyanions (25), sulfated lactose derivatives (26), heparin (27, 28), and molecular mimics of sialyl Lewis^X (sLe^x)¹ (e.g. glycyrrhizin derivatives; Ref. 29), including oligonucleotides (30, 31). Inhibiting the glycosyltransferases that participate in forming the carbohydrate ligands would provide another way to block leukocyte attachment to the endothelium. Several glycosyltransferase inhibitors have been described, which resemble natural carbohydrate substrates but lack critical hydroxyl groups or contain methyl groups at key positions (32-34). These compounds bind to the glycosyltransferases and competitively inhibit the enzymes, sometimes with micromolar K_i values. Unfortunately, none of the available compounds inhibits glycosylation in intact cells, presumably because the hydrophilicity of sugars prevents them from permeating cell membranes.

Glycoside-based primers represent another class of potential inhibitors. These compounds resemble biosynthetic intermediates and act as substrates for oligosaccharide assembly, thereby diverting the synthesis of chains from endogenous proteins and lipids. Examples include N-acetylgalactosaminides, which prime oligosaccharide chains similar to those found O-glycosidically linked to serine and threonine residues of glycoproteins (35, 36). Incubation of HL-60 promyelocytic leukemia cells with GalNAc-O-benzyl² inhibited the synthesis of sLe^x on O-linked glycoproteins, which in turn blocked selectin-mediated adhesion to activated endothelial cells (37). Peracetylated fluorinated analogs of N-acetylglucosamine had similar effects on selectin-mediated adhesion of tumor cells to endothelial cells (38, 39). Recently, we showed that a peracetylated disaccharide ((Ac)₄Gal1-4(Ac)₂ GlcNAc-O-naphthalenemethanol) inhibited sLe^x expression on HL-60 cells and worked at a low concentration (50 µM) (40). In this study, we have extended this finding to another disaccharide, peracetylated GlcNAc1-3Gal-O-naphthalenemethanol, and to human U937 cells. We show that the acetylated disaccharides inhibit the E-selectin-dependent adhesion of U937 cells to cytokine-activated endothelial cells by blocking the formation of sLe^x.

EXPERIMENTAL PROCEDURES

Acetylated Gal1-4GlcNAc-O-naphthalenemethanol (Gal1-4GlcNAc-NM), GlcNAc1-3Gal-O-naphthalenemethanol (GlcNAc1-3Gal-NM), and Gal1-4(3-OMe)GlcNAc-O-naphthalenemethanol were chemically synthesized from monosaccharide units using standard orthogonal blocking chemistry, coupling, and deblocking strategies. The details of their synthesis, purification, and chemical characterization will be presented elsewhere.³ They were >98% pure by ¹H NMR, ¹³C NMR, and thin layer chromatography. [6-³H]Gal1-4GlcNAc-O-naphthalenemethanol was synthesized by oxidizing Gal1-4GlcNAc-NM with galactose oxidase and then reducing the product with NaB³H₄ (Amersham, 50-75 Ci/mmol; 1 Ci = 37 GBq) (41). It was then purified through a reverse phase Sep-Pak C18 column (Waters). The trisaccharide Gal1-4([¹⁴C]Fuc1-3)GlcNAc-O-naphthalenemethanol was synthesized enzymatically using Gal1-4GlcNAc-NM, GDP-[¹⁴C]fucose (200 mCi/mmol, NEN Life Science Products) and growth medium from COS-1 cells transfected with fucosyltransferase IV (a gift from J. Lowe, University of Michigan Medical Center). The trisaccharide standard, [6-³H]Gal1-4GlcNAc1-3Gal-O-naphthalenemethanol was synthesized by galactosylation of GlcNAc1-3Gal-NM with bovine 1-4 galactosyltransferase and UDP-[6-³H]galactose (32 Ci/mmol, NEN Life Science Products) according to the manufacturer's directions (Boehringer Mannheim). [6-³H]Fuc (83 Ci/mmol), [6-³H]Gal (29.5 Ci/mmol), [6-³H]GlcN-HCl (33.3 Ci/mmol), and [³H-methyl]thymidine (83 Ci/mmol) were from NEN Life Science Products. [6-³H]Mannosamine (20 Ci/mmol) was from American Radiochemicals, and H₂[³⁵S]O₄ (25-40 Ci/mg) was obtained from Amersham. All other chemicals were purchased from Aldrich or Sigma unless stated otherwise.

U937 human histiocytic lymphoma cells were from the American Type Culture Collection (CRL 1593.2). They were grown in RPMI 1640 medium supplemented with 10% (v/v) fetal bovine serum (Hyclone Laboratories, Logan, UT), glutamine (0.3 g/liter), streptomycin sulfate (100 µg/ml), and penicillin (100 units/ml) (42, 43). The cells were passaged by dilution (1:20) when they reached ~10⁶/ml. Human umbilical vein endothelial cells (HUVEC) were from Clonetics Corp. (CC-2519). They were grown in Medium 199 containing 20% (v/v) fetal bovine serum, heparin (100 µg/ml, Sigma) and endothelial cell growth supplement (100 µg/ml, Collaborative Biomedical). The cells were passaged every 5 days with a solution of 0.025% trypsin and 0.01% EDTA in phosphate-buffered saline (PBS) (44). All cell lines were maintained at 37 °C in a humidified incubator containing 5% CO₂ and 95% air.

The disaccharides were dissolved in Me₂SO and added to growth medium to achieve the concentrations indicated in the figures and tables. The concentration of Me₂SO was adjusted to 0.5% in all samples, and then cells were added to a starting density of ~1 × 10⁵ cells/ml. For labeling studies, the cells were incubated with the disaccharides and 10 µCi/ml [6-³H]fucose, 10 µCi/ml [6-³H]galactose, 10 µCi/ml [6-³H]GlcN, 15 µCi/ml [6-³H]mannosamine, or 1 µCi/ml [³H-methyl]thymidine. The cells were sedimented by centrifugation, and the pellets were treated with 10% trichloroacetic acid to precipitate glycolipids and glycoproteins. The pellets were dissolved in 0.1 M NaOH and counted by liquid scintillation spectrometry using UltimaGold (Packard). In priming studies, the conditioned medium was adjusted to 0.5 M NaCl with a stock solution of 1.5 M NaCl. Supernatants were then applied to 0.2-cc Sep-Pak Vac RC C18 cartridges (Waters), which had been prewashed with 100% methanol, water, and 0.5 M NaCl. After applying the samples to the columns, they were washed sequentially with 0.5 M NaCl (2.5 ml) and water (25 ml). Radioactive oligosaccharides were eluted with 40% methanol in water (2.5 ml). This fraction was concentrated by vacuum centrifugation (Savant SpeedVac concentrator), dissolved in water, and counted by liquid scintillation. Total cell protein was estimated from the cell pellet using the Bio-Rad protein assay kit and bovine serum albumin (BSA) as standard.

Primed oligosaccharides that eluted in the 40% methanol fraction were analyzed by QAE-Sephadex (Sigma) anion-exchange chromatography to separate charged and neutral species (45). Samples were dissolved in 2 mM Tris base and applied to a column of QAE-Sephadex (0.5 ml) pre-equilibrated with the same buffer. The column was first washed with 2 mM Tris base (10 ml) and then sequentially with solutions of 2 mM Tris base containing 5 mM NaCl, 10 mM NaCl, 20 mM NaCl, 50 mM NaCl, 100 mM NaCl, or 200 mM NaCl (2.5 ml each). Radioactive oligosaccharides were recovered in the fraction containing 2 mM Tris base (neutral) and the 20 mM NaCl and 2 mM Tris base wash (1 charged). These samples were concentrated, dissolved in 2.5 ml of 0.5 M NaCl solution, and applied to a Sep-Pak C18 cartridge, which was pre-equilibrated with 0.5 M NaCl. The column was washed with water (5 ml), and primed material was eluted with 40% methanol in water (1 ml). This material was then used for subsequent analytical work.

All enzymatic treatments were done at 37 °C in a final volume of 25 µl. Samples were treated with 2 milliunits of Fusarium oxyporum fucosidase (Seikagaku) for 8 h in 50 mM citrate buffer (pH 5) (46), 25 microunits of Streptomyces sp. fucosidase (Sigma) for 24 h in 50 mM potassium phosphate buffer (pH 6) (47), 5 milliunits of Diplococcus pneumonia 1-4 galactosidase (Boehringer Mannheim) for 8 h in 50 mM potassium phosphate-sodium citrate buffer (pH 6.1) (48, 49), or NDV sialidase (2.5 milliunits, Oxford Glycosystems) for 4 h in 50 mM sodium acetate buffer (pH 6) (50). The enzymes were inactivated by heating the samples for 2 min in a boiling water bath.

Samples were dissolved in water (200 µl) and injected into a reverse phase C18 column (TosoHaas RP18, 4.6 mm × 25 cm) connected to a Rainin HPXL solvent delivery system. The column was first washed with water (flow rate 0.5 ml/min) and then with increasing concentrations of acetonitrile in water, as shown by the dashed line in Fig. 4. Radioactivity in the eluant was monitored by Radiomatic Flo-one Beta detector connected in line to the column.

To measure sLe^x on the cell surface, U937 cells were grown in the presence and absence of primers for 42-48 h, washed with PBS, and fixed with 1% p-formaldehyde in PBS (5 min, room temperature). The fixed cells were rinsed once with PBS and suspended (1 × 10⁵) in 100 µl of PBS containing 1% (w/v) BSA. CSLEX1 mAb (Becton-Dickinson) was added (5 µg/ml). A control incubation contained nonspecific mouse IgM (5 µg/ml) instead of CSLEX1. Human IgG (500 µg/ml) was added to prevent reaction of CSLEX1 with Fc receptors on the cells. Anti-mouse IgM coupled to horseradish peroxidase was added (1:100). After 30 min at room temperature, the cells were washed with BSA/PBS buffer three times by centrifugation and resuspended in 250 µl of reaction buffer containing 25 mM citric acid, 50 mM Na₂HPO₄, 3.7 mM o-phenylenediamine, and 30 µl of 30% H₂O₂. The reaction was stopped after 5 min at room temperature by adding 250 µl of 3 M sulfuric acid, and the absorbance at 490 nm was recorded.

The expression of sialic acids on the cell surface was estimated by lectin binding and flow cytometry. Cells were washed three times with cold PBS containing 2 mg/ml BSA and 0.05% sodium azide, and then treated for 15 min at room temperature with fluorescein isothiocyanate-conjugated Maackia amurensis lectin (10 µg/ml, E-Y Laboratories). After rinsing the cells at 4 °C, they were fixed with 1% (w/v) p-formaldehyde and analyzed by flow cytometry (FACScan, Becton Dickinson).

HUVEC (1 × 10⁵ cells/well) at the second to fourth passage were grown for 48 h in a 24-well tissue culture plate (Corning). The monolayers were washed three times in Medium 199 (1 ml) without growth supplement and heparin. Medium 199 (0.5 ml) containing 20% fetal bovine serum with and without 20 ng/ml TNF- (R&D Systems) was added next (51). After 5 h, the monolayers were washed three times with minimal essential medium (MEM).

U937 cells (2 × 10⁵/well) in RPMI 1640 medium were incubated with and without acetylated disaccharide primers and [³H-methyl]thymidine (1 µCi/ml) for 2 days in a six-well plate. The cells were washed three times with MEM containing 2% heat-inactivated fetal bovine serum (adjusted to pH 7.4 with Tricine) and resuspended at 1 × 10⁵ cells/ml. The cells were then added to TNF--activated HUVEC and incubated for 35 min at 4 °C. The plates were washed gently three times with 1 ml of cold MEM, and the cells were solubilized with 0.1 M NaOH (250 µl). Radioactivity was measured after neutralizing the solution with 1 M acetic acid (25 µl). The values were normalized to the radiospecific activity of the cells determined on a separate aliquot. Cell adhesion was >80% dependent on E-selectin under these conditions, based on the inhibitory activity of anti-ELAM-1 mAb (R&D Systems).

RESULTS

Previous studies showed that F9 cells take up and rapidly deacetylate peracetylated Gal1-4GlcNAc-NM (Fig. 1), resulting in the stimulation of oligosaccharide synthesis on the exogenous disaccharide (40). Priming of oligosaccharides in this way inhibited the expression of sLe^x on the surface of HL-60 cells, presumably due to diversion of chain assembly on endogenous glycoconjugates. To determine whether other disaccharides would behave similarly, we prepared acetylated GlcNAc1-3Gal-NM and compared it to Gal1-4GlcNAc-NM. Both compounds were quite hydrophobic and limited in solubility in aqueous growth medium to 0.2 mM. U937, a human histiocytic cell line that expresses sLe^x, grew well in the presence of the disaccharides, up to 50 µM (Fig. 2). Above this concentration, some diminution in growth rate was observed, possibly due to the detergent properties of the compounds at high concentration.

Growth of U937 cells in presence of acetylated Gal1-4GlcNAc-NM and GlcNAc1-3Gal-NM.

To test whether the disaccharides primed oligosaccharide chains, U937 cells were incubated with acetylated Gal1-4GlcNAc-NM or GlcNAc1-3Gal-NM and [6-³H]Fuc, [6-³H]Gal, or [6-³H]GlcN. Radioactive oligosaccharides assembled on the primer and secreted into the growth medium were collected on Sep-Pak C18 cartridges, which bound the compounds due to the hydrophobic aglycone (naphthalenemethanol). In addition, some endogenous material produced by the cells bound to the resin (Table I). However, these counts were not recoverable after analytical C18 reverse phase HPLC and probably represented unincorporated sugars or cellular glycoconjugates. Over 95% of the radioactive products generated on the primer were secreted into the growth medium. Therefore, in all subsequent experiments, the conditioned medium was the source of material for further analysis.

Table I. Radiolabeling of oligosaccharides assembled on acetlated primers in U937 cells U937 cells (~2 × 105) were incubated with each glycoside (50 µM) for 48 h in presence of 10 µCi/ml [6-3H]Fuc, 10 µCi/ml [6-3H]Gal, or 25 µCi/ml [6-3H]GlcN. Oligosaccharide chains on the primers were isolated by chromatography on a Sep-Pak C18 column, and the amount of radioactive material recovered in the methanol fraction was quantitated (see "Experimental Procedures"). The values were expressed relative to the amount of cell protein in the dish. The data varied by ~20% between experiments. Glycoside [6-3H]Fuc [6-3H]Gal [6-3H]GlcN 3H cpm/µg cell protein None 11 37 84 Acetylated Gal1-4GlcNAc-NM 52 43 77 Acetylated GlcNAc1-3Gal-NM 86 340 72 Acetylated Gal1-4(3-OMe)GalNAc-NM 9 32 78

As shown in Table I, both acetylated Gal1-4GlcNAc-NM and acetylated GlcNAc1-3Gal-NM stimulated the incorporation of [6-³H]fucose into products that bound to Sep-Pak C18 cartridges compared with the control which did not contain added disaccharides. The incorporation of [³H]Fuc into oligosaccharide products rose with increasing concentration of added primer (Fig. 3). Priming on GlcNAc1-3Gal-NM showed a trend toward saturability above 50 µM, whereas priming on Gal1-4GlcNAc-NM did not. Interestingly, the incorporation of [³H]Fuc into oligosaccharides assembled on GlcNAc1-3Gal-NM was greater at all concentrations than those assembled on Gal1-4GlcNAc-NM. This difference was also observed in F9 teratocarcinoma cells (data not shown).

To characterize the primed products, the [³H]Fuc-labeled oligosaccharides eluted from Sep-Pak C18 cartridges were separated into charged and uncharged species by passage through a QAE-Sephadex column (45). Essentially all of the products generated on Gal1-4GlcNAc-NM were neutral (Table II), whereas about one-half of the material generated on GlcNAc1-3Gal-NM was neutral and the remainder eluted at a salt concentration indicative of a 1 charge (45).

Table II. Neutral and charged oligosaccharides were formed on acetylated disaccharides U937 cells (~2 × 105) were incubated with the disaccharides (50 µM) in presence of [6-3H]Fuc or [6-3H]Gal, and the radiolabeled products were isolated by Sep-Pak C18 chromatography (see "Experimental Procedures"). The charged and neutral oligosaccharides were separated by chromatography on QAE-Sephadex (see "Experimental Procedures"). The values varied by less than 10% in individual experiments. , none detected. Glycoside [6-3H]Fuc oligosaccharides [6-3H]Gal oligosaccharides Neutral Charged Neutral Charged % of total Acetylated Gal1-4GlcNAc-NM >98 <2     Acetylated GlcNAc1-3Gal-NM 52 48 44 56

Analysis by reversed-phase HPLC of the neutral fucosylated products formed on Gal1-4GlcNAc-NM revealed a single peak of material eluting exactly at the same position as standard Le^X glycoside (Gal1-4([¹⁴C]Fuc1-3)GlcNAc-O-NM; Fig. 4B). The linkage stereochemistry of the fucose residue was established by two criteria. First, treatment of the radioactive material with -fucosidase from Streptomyces sp. (which cleaves 1,3/1,4-linked fucose) (47) liberated all of the radioactivity, as measured by step elution from a C18 cartridge. In contrast, treating the material with F. oxyporum -fucosidase (specific for 1,2/1,4-linked fucose) (46) had no effect, indicating that the linkage was 1-3. Second, feeding cells a modified disaccharide, in which the 3-OH of the GlcNAc residue was blocked by methylation (Gal1-4(3-OMe)GlcNAc-NM), did not result in [³H]Fuc addition to oligosaccharides (Table I). These findings demonstrated that the primary oligosaccharide assembled on Gal1-4GlcNAc-NM was Le^x, i.e. Gal1-4(Fuc1-3)-GlcNAc-NM. The lack of any [³H]Gal or [³H]GlcN incorporation into oligosaccharides above the control confirmed the absence of any chain extension products (Table I). Furthermore, treatment of the sample with mild base did not alter the elution of the product by reverse phase HPLC (data not shown), indicating that all of the acetyl groups in the original primer had been removed.

The fucosylated products generated on GlcNAc1-3Gal-NM were more complex inasmuch as both neutral and charged species were detected (Table II). The neutral species eluted from a C18 reverse-phase HPLC column as a single peak ~2 min earlier than the Le^x glycoside standard, suggesting that it was more polar possibly due to the presence of another sugar residue (Fig. 4C). Labeling cells with [³H]Gal yielded a peak with similar retention time (large peak in Fig. 5A), indicating the presence of a Gal residue and suggesting the structure, Gal1-4(Fuc1-3)GlcNAc1-3Gal-NM. This structure was confirmed by sequential enzyme digestion. (i) [³H]Fuc was released quantitatively when a sample was treated with Streptomyces sp. fucosidase, but not with F. oxyporum fucosidase, confirming that it was linked 1-3 to the GlcNAc residue. (ii) Treatment of the [³H]Gal-labeled material with -fucosidase caused the material to shift to the position of the trisaccharide standard, [³H]Gal1-4GlcNAc1-3Gal-NM (Fig. 5B and arrow in Fig. 5A). (iii) Initial attempts to release the [³H]Gal residue with D. pneumonia -galactosidase (cleaves terminal 1-4-linked galactose; Refs. 48 and 49) were not successful (Fig. 5C), but after -fucosidase treatment the galactose residue was released quantitatively (data not shown). Together, these results suggest the structure Gal1-4(Fuc1-3)GlcNAc1-3Gal-NM for the major neutral product generated on GlcNAc1-3Gal-NM. The more retarded peak of [³H]Gal-labeled material shown in Fig. 5A co-eluted with the standard [³H]Gal1-4GlcNAc1-3Gal-NM, and it was sensitive to -galactosidase (Fig. 5C), suggesting that it was the trisaccharide, Gal1-4GlcNAc1-3Gal-NM.

Reverse phase chromatography of [6-³H]Gal-labeled oligosaccharides built on GlcNAc1-3Gal-NM.

The [³H]Fuc-labeled species with a 1 charge eluted from the C18 reverse phase column as a single peak at an earlier time than the neutral products described above (Fig. 4D). After digestion with NDV sialidase, the charged material comigrated with the neutral [³H]Fuc-labeled species (Fig. 4, compare E and C), and the neutralized material was sensitive to 1,3/1,4-fucosidase (data not shown). These findings suggested the structure NeuAc2-3Gal1-4(Fuc1-3)GlcNAc1-3Gal-NM. To confirm its identity, we isolated the charged material labeled biosynthetically with [³H]Gal. This material eluted from the C18 column at the same position as the charged species labeled with [³H]Fuc (compare Figs. 5D and 4D). However, the [³H]Gal-labeled material eluted in a somewhat broader peak than the [³H]Fuc-labeled material, suggesting the presence of more than one charged species in the [³H]Gal-labeled products. Treating the [³H]Gal-labeled samples with NDV sialidase followed by reverse phase HPLC yielded two peaks, which comigrated with the two neutral [³H]Gal-labeled oligosaccharides (Fig. 5E). Fucosidase and galactosidase treatment had the same effects on these oligosaccharides as they had on the neutral species that were generated on this primer. Together, these findings suggested that the charged oligosaccharides were NeuAc2-3Gal1-4GlcNAc1-3Gal-NM and NeuAc2-3Gal1-4(Fuc1-3)GlcNAc1-3Gal-NM. These data and the recovery of material in the individual peaks are summarized in Fig. 6.

Biosynthesis of fucosylated and sialylated oligosaccharides on acetylated Gal1-4GlcNAc-NM and GlcNAc1-3Gal-NM.

To test whether the priming activity of the disaccharides inhibited glycosylation of endogenous glycoconjugates, cells were labeled in the presence and absence of the primers with [³H]Fuc and [³H]ManNH₂, a precursor of sialic acid residues. Precipitation of labeled cellular glycoconjugates with trichloroacetic acid revealed that both disaccharides reduced the incorporation of [³H]Fuc by 35-45% (Table III). In contrast, neither compound affected the incorporation of [³H]ManNH₂ into sialic acid residues. Because the extent of mannosamine labeling was considerably lower than with the other radioactive sugars, we also examined the reactivity of the cells toward M. amurensis lectin, which binds terminal 2-3-linked sialic acid residues (52). Cell sorting with fluorescein isothiocyanate-conjugated lectin showed that disaccharide-treated and untreated cells had identical profiles (Fig. 7). As expected, the reactivity was diminished by previous treatment with NDV sialidase. Thus, the primers selectively reduced fucose incorporation without altering the addition of sialic acid.

Table III. Glycosides inhibit fucosylation of endogenous glycoconjugates U937 cells (2 × 105) in RPMI 1640 growth medium were incubated with or without glycosides (35 µM). After 32 h, the cells were labeled with [3H]thymidine (1 µCi/ml), [6-3H]fucose (15 µCi/ml), or [6-3H]mannosamine (15 µCi/ml) for 20 h. The cells were then centrifuged and washed three times with cold PBS containing 2 mg/ml BSA, and cellular glycoconjugates were precipitated with 10% trichloroacetic acid. The precipitates were dissolved in 0.1 M sodium hydroxide and counted. Each labeling scheme was done in duplicate, and the average values are given. The individual measurements varied by 10% from the average. Glycoside [3H]Thy [3H]Fuc [3H]ManNH2 cpm × 103 None 220 ± 2 54 ± 5 10 ± 1  Acetylated Gal1-4GlcNAc-NM 210 35 10 Acetylated GlcNAc1-3Gal-NM 220 30 9

These findings suggested that the primers might reduce expression of fucose containing determinants on endogenous glycoconjugates, such as sLe^x. To test this hypothesis, we measured sLe^x expression on U937 cells using an ELISA (see "Experimental Procedures"). As shown in Fig. 8, the primers inhibited sLe^x expression in a dose-dependent manner. Acetylated Gal1-4GlcNAc-NM at its highest concentration (50 µM) inhibited expression of sLe^x by ~30%, whereas the more potent primer, acetylated GlcNAc1-3Gal-NM, inhibited sLe^x expression by ~4-fold at 25 µM. In general, the relative inhibitory activity of the disaccharides paralleled their priming activity (cf. Fig. 3).

sLe^x on O-linked glycoproteins is thought to be responsible for adhesion of U937 cells to E-selectin expressed on TNF--activated endothelial cells (37). Thus, the inhibition of sLe^x expression by the disaccharides should reduce their adhesion to activated endothelial cells. To test this possibility, we grew U937 cells in the presence of acetylated Gal1-4GlcNAc-NM and GlcNAc1-3Gal-NM at various concentrations and challenged them to adhere to TNF--activated HUVEC. Adhesion was dependent on previous activation of the endothelial cells with TNF-, and attachment under these conditions was inhibited by 80% with anti-ELAM-1 (E-selectin) mAb (Fig. 9). Pretreatment of U937 cells with NDV sialidase showed that adhesion depended on sialic residues. Cells treated with acetylated disaccharides adhered to a lesser extent than untreated cells and the extent of inhibition depended on concentration. Acetylated Gal1-4GlcNAc-NM inhibited cell adhesion by ~50% at 50 µM, whereas acetylated GlcNAc1-3Gal-NM inhibited adhesion more dramatically. In general, the more potent primer (acetylated GlcNAc1-3Gal-NM) inhibited adhesion to a greater extent at all concentrations tested.

To confirm that the inhibition of adhesion was due to priming per se as opposed to some other effect, cells were incubated with the methylated disaccharide, Gal1-4(3-OMe)GlcNAc-NM, in which the site for fucosylation (3-OH of GlcNAc) was blocked by a methyl group. The methylated derivative did not prime a fucosylated product (Table I). Furthermore, it did not inhibit expression of cell surface sLe^x (Fig. 10A) or affect the adhesion of U937 cells to activated HUVEC (Fig. 10B). Methylated derivatives of GlcNAc1-3Gal-NM (i.e. (3-OMe)GlcNAc1-3Gal-naphthalenemethanol and (4-OMe)GlcNAc1-3Gal-naphthalenemethanol) were also tested. Both disaccharides failed to prime any oligosaccharides and also did not inhibit sLe^x expression or cell adhesion (data not shown).

DISCUSSION

We reported previously that HL-60 and F9 cells readily take up acetylated Gal1-4GlcNAc-NM, remove the acetyl groups, and use the deblocked disaccharide for oligosaccharide synthesis. In this report we have extended this finding to U937 cells and to a second disaccharide, GlcNAc1-3Gal-NM, which resembles another intermediate in the assembly of lactosaminoglycans (Fig. 1). Priming by these disaccharides results in the formation and secretion of oligosaccharides, sufficient in mass to divert the assembly of carbohydrate chains from endogenous glycoproteins and glycolipids. Priming resulted in diminution of sLe^x expression and a parallel decrease in adhesion dependent on sLe^x. Thus, the disaccharides act as metabolic inhibitors and have the interesting property of affecting a key step involved in an inflammatory reaction.

Several important features of these compounds deserve additional attention. First, in order for the compounds to be effective, the cells had to remove the O-linked acetyl groups, which were required for passage of the compounds across the plasma membrane (40). All of the oligosaccharide products generated by the cells lacked the O-acetyl groups, suggesting that deacetylation occurred very efficiently. In other studies, we have found that deacetylation occurs rapidly in the cytoplasm and in microsomal membranes,⁴ presumably mediated by one or more carboxylesterases (53-57). Because the glycosyltransferases responsible for oligosaccharide biosynthesis reside in the Golgi, only the deacetylated compounds arising in that compartment may be accessible to the enzymes. If correct, then the effective concentration of active, deacetylated primers may be much less than the disaccharide concentration added to the cells. However, a recent study of -xyloside priming in isolated Golgi vesicles indicates that the primary neutral disaccharide product (Gal1-4Xyl-O-methylumbelliferol) could exit the Golgi and appear in the incubation medium (58). Furthermore, incubation of Golgi vesicles with non-acetylated Gal1-4Glc-O-methylumbelliferol gave rise to sialylated products trapped inside the vesicles (59). These studies suggest that the Golgi may be much more permeable to polar compounds than the plasma membrane. The permeability limit of the Golgi with respect to oligosaccharide size and structure deserves further study since it may be possible to design larger, more specific primers.

As shown in Figs. 4 and 5, Gal1-4GlcNAc-NM and GlcNAc1-3Gal-NM primed the formation of specific oligosaccharides in U937 cells. Gal1-4GlcNAc-NM underwent selective fucosylation to form the Le^x analog, Gal1-4(Fuc1-3)GlcNAc-NM, as determined by enzymatic defucosylation. That Gal1-4(Fuc1-3)GlcNAc-NM was the sole product formed indicates that the inhibition of sLe^x expression and diminished cell adhesion by Gal1-4GlcNAc-NM was due to inhibition of fucosylation of endogenous glycoconjugates rather than inhibition at another biosynthetic step (e.g. sialylation). Analysis of endogenous glycoconjugates confirmed this idea (Table III and Fig. 7). A more complex array of oligosaccharides arose on GlcNAc1-3Gal-NM, including fucosylated and sialylated oligosaccharides. Since sialylation of endogenous glycoconjugates also was not altered by GlcNAc1-3Gal-NM, both compounds appear to inhibit sLe^x expression by altering fucosylation.

The specificity of oligosaccharide synthesis on the primers provides insights into the substrate preference and capacity of the biosynthetic enzymes involved in sLe^x formation in vivo. Myeloid cells express nearly equal amounts of Fuc TIV and Fuc TVII mRNA, and both enzymes play a role in forming fucosylated selectin ligands (60, 61), although Fuc TVII may predominate (62). Interestingly, Gal1-4GlcNAc-O-R (where R = a hydrophobic aglycone) serves as an acceptor in vitro for Fuc TIV, but not Fuc TVII because the latter requires prior sialylation for activity (e.g. NeuAc2-3Gal1-4GlcNAc-O-R is a substrate; Refs. 60, 62, and 63). These observations suggest that Gal1-4(Fuc1-3)GlcNAc-NM was produced on Gal1-4GlcNAc-NM by the action of Fuc TIV rather than Fuc TVII. Thus, Fuc TIV would appear to play a significant role in forming E-selectin ligands in U937 cells.

The absence of any sialylated products from Gal1-4GlcNAc-NM suggests that the K_m for the 2-3 sialyltransferase may be higher that the apparent K_m for Fuc TIV. Since extension products also were not detected, the K_m for the 1-3 GlcNAc transferase responsible for the formation of polylactosaminoglycan chains also may be high (64). These data do not take into account competition by endogenous fucosylated and sialylated glycoprotein and glycolipid intermediates which may differ in concentration and affinity. Analysis of the endogenous glycoconjugates that accumulate in the presence of the primers may provide more insight into this issue. Moreover, such studies would reveal whether the disaccharides inhibit specific glycoproteins and whether the pathways of N-linked and O-linked oligosaccharide formation are affected similarly.

In contrast to the single product generated on Gal1-4GlcNAc-NM, GlcNAc1-3Gal-NM first underwent extension by the addition of a Gal residue to form Gal1-4GlcNAc1-3Gal-NM and then the trisaccharide underwent fucosylation and sialylation. This finding supports the idea that Fuc TIV and Fuc TVII will not add to a terminal GlcNAc residue (65-69). Sialylation also took place, giving rise to NeuAc2-3Gal1-4GlcNAc1-3Gal-NM and NeuAc2-3Gal1-4(Fuc1-3)GlcNAc1-3Gal-NM, which contains the sLe^x determinant. The failure to sialylate Gal1-4GlcNAc-NM, therefore, may reflect a preference of the 2-3 sialyltransferase for substrates with three (or more sugars).

Inasmuch as these disaccharides resemble biosynthetic intermediates involved in the formation of sLe^x, they compete with endogenous substrates. Thus, they act like N-acetylgalactosaminides, which mimic -GalNAc containing glycoprotein intermediates (35). However, the disaccharide primers reported here are nearly 50-fold more potent than the monosaccharide primer. This difference is not related to the aglycone, because GalNAc-O-naphthalenemethanol has the same dose profile as the benzyl derivative.⁴ Interestingly, inhibition of sLe^x expression occurred at lower doses of GlcNAc1-3Gal-NM compared with Gal1-4GlcNAc-NM, which correlated well with the fact that the former primed fucosylated oligosaccharide chains at lower concentration than the latter (Fig. 3). A priori, one could not have predicted these differences in priming efficacy by merely comparing kinetic constants. Thus, other disaccharide-based primers may prove even more potent than the ones reported here (e.g. GlcNAc1-6Gal-O-NM, GlcNAc1-6GalNAc-O-NM, or Gal1-3GalNAc-O-NM). Use of higher order oligosaccharide primers also may prove useful, and varying the aglycone may be advantageous (70-72).

Finally, the disaccharides described here represent starting points for designing specific inhibitors of glycosyltransferases and drugs for treating diseases or disorders that depend on protein-carbohydrate interactions (73). The selective addition of fucose to Gal1-4GlcNAc-NM suggests that modifying the disaccharide by deoxygenation, fluorination, or by the addition of reactive groups may convert a primer into an active-site directed inhibitor (32, 33, 74). It is unclear how these compounds will behave in animals and whether they have useful therapeutic value. Preliminary studies indicate that intraperitoneal injection of Gal1-4GlcNAc-NM into mice had no immediate deleterious effects, but more studies are needed to determine its bioavailability, clearance, and ability to block formation of sLe^x in vivo.