Deoxygenated Disaccharide Analogs as Specific Inhibitors of β1–4-Galactosyltransferase 1 and Selectin-mediated Tumor Metastasis*

The disaccharide peracetylated GlcNAcβ1–3Galβ-O-naphthalenemethanol (disaccharide 1) diminishes the formation of the glycan sialyl Lewis X (Neu5Acα2–3Galβ1–4(Fucα1–3) GlcNAc; sLeX) in tumor cells. Previous studies showed that the mechanism of action of disaccharide 1 involves three steps: (i) deacetylation by carboxyesterases, (ii) action as a biosynthetic intermediate for downstream enzymes involved in sLeX assembly, and (iii) generation of several glycans related to sLeX. In this report, we show that GlcNAcβ1–3Galβ-O-naphthalenemethanol binds to the acceptor site of human β1–4-galactosyltransferase much like the acceptor trisaccharide, GlcNAcβ1–2Manβ1–6Man, which is present on N-linked glycans. The 4′-deoxy analog, in which the acceptor hydroxyl group was replaced by -H, did not act as a substrate but instead acted as a competitive inhibitor of the enzyme. The acetylated form of this compound inhibited sLeX formation in U937 monocytic leukemia cells, suggesting that it had inhibitory activity in vivo as well. A series of synthetic acetylated analogs of 1 containing -H, -F, -N3, -NH2, or -OCH3 instead of the hydroxyl groups at C-3′- and C-4′-positions of the terminal N-acetylglucosamine residue also blocked sLeX formation in cells. The reduction of sLeX by the 4′-deoxy analog also diminished experimental tumor metastasis by Lewis lung carcinoma in vivo. These data suggest that nonsubstrate disaccharides have therapeutic potential through their ability to bind to glycosyltransferases in vivo and to alter glycan-dependent pathologic processes.

The sialylated, fucosylated tetrasaccharide, sLe X , 3 is a common carbohydrate determinant present in many O-GalNAclinked mucins and N-linked glycans that act as selectin ligands (see Ref. 1 and references therein). Expression of sLe X endows tumor cells with the capacity to bind to platelets and endothelial cells in the vasculature via P-and E-selectins, thus facilitating hematogenous metastasis possibly through protection against innate immune cells and by adhesion to the blood vessel wall. Strategies for blocking selectin-carbohydrate interactions include (i) competition by soluble recombinant forms of selectins, glycoprotein ligands, and glycolipids, (ii) peptides based on the primary sequence of the carbohydrate binding site, (iii) anti-selectin antibodies, (iv) oligosaccharides related to Lewis X , (v) inositol polyanions and sulfated sugars, (vi) heparin, and (vii) molecular mimics of sLe X , including oligonucleotides (reviewed in Refs. 2 and 3). Analogs of acceptor substrates of the various glycosyltransferases involved in glycan biosynthesis provide another class of potential inhibitors (reviewed in Refs. 4 and 5). Although many of these analogs are effective in vitro, they generally do not exhibit inhibitory activity in cells due to poor membrane permeability. The large number of polar hydroxyl groups and the lack of membrane transporters for oligosaccharides in most cells presumably prevent their uptake (6).
In contrast to many of the inhibitors described above, peracetylated disaccharides (e.g. acetylated Gal␤1-4GlcNAc␤-Onaphthalenemethanol (NM), acetylated Gal␤1-3GalNAc␣-O-NM, and acetylated GlcNAc␤1-3Gal␤-O-NM) inhibit sLe X biosynthesis in cells (6 -9). These compounds are taken up by cells by passive diffusion and acted on by cytoplasmic or membrane-associated carboxyesterases, which remove the acetyl groups. The compounds gain access to the biosynthetic enzymes located in the Golgi complex, where they serve as substrates, priming oligosaccharide synthesis and generating products related to O-GalNAc-linked mucin oligosaccharides. Priming in this manner diverts the assembly of the O-linked chains from endogenous glycoproteins, resulting in inhibition of expression of terminal Lewis antigens that are recognized by selectins. Inhibition occurs at a much lower dose than for monosaccharide-based agents, such as GalNAc␤-O-benzyl (ϳ25 M versus 1-2 mM, respectively) (10,11). Furthermore, the disaccharides appear to selectively affect sLe X formation, since sLe a expression was unaffected (12). By blocking selectin ligand expression, these compounds block both experimental and spontaneous metastasis (12,13).
In this study, we have examined acetylated disaccharide analogs that have been modified so that after deacetylation their activity as substrates would be altered. Characterization of the 4Ј-deoxy derivative using ␤1-4-galactosyltransferase 1 as a model showed that it acts by competitively inhibiting the enzyme. Interestingly, the peracetylated form of this analog maintains the capacity to inhibit sLe X expression in U937 lymphoma cells and Lewis lung carcinoma (LLC) cells and block tumor formation in vivo. Thus, the deoxy analog presumably inhibits one or more galactosyltransferases in vivo, thereby blocking sLe X formation and experimental tumor cell metastasis without generation of oligosaccharide products.

EXPERIMENTAL PROCEDURES
General Chemical Methods-Optical rotations were measured at room temperature using an Autopol III polarimeter or a PerkinElmer Life Sciences 241-Mc polarimeter. Melting points were measured with Unimet Thomas Hoover. NMR experiments were carried out on a Bruker AMX-300, AMX-400, AMX-500, or Varian Mercury 400-MHz spectrometer. Chloroform was used as an internal standard for chemical shift calibration unless otherwise noted; the 1 H chemical shift of CHCl 3 was set to 7.262 ppm, and the 13 C chemical shift of CDCl3 was set to 77.0 ppm. Mass spectrometric data were recorded using Electrospray API-100, Electrospray LCQDECA, or FAB with 3-nitrobenzyl alcohol as the matrix and polyethylene glycol as the reference. All reaction products were analyzed by thin layer chromatography on Silica Gel 60-F 254 (Merck) plates and detection by UV light or by charring with 5% (v/v) H 2 SO 4 in ethanol. Column chromatography was performed on Silica Gel 60 (40 -63 mm; Mallinckrodt V 150). Molecular sieves (4A) were heated at 180°C for 24 h prior to use. All solvents were distilled prior to use and stored over molecular sieves. Chemical reactions were performed in oven-dried glassware under an inert argon atmosphere. All evaporations were carried out at 40 -50°C under reduced pressure, unless otherwise noted.
Chemical Synthesis of C-3Ј-and C-4Ј Hydroxyl-modified Glycosides-Disaccharide 1 was prepared as described (8). (8), and per-O-acetylated 3-methoxy-GlcNAc␤1-3Gal␤-O-NM (9) were chemically synthesized and characterized by NMR spectroscopy and mass spectrometry as outlined below. The data were in good accordance with the proposed structures.
Cell Lines, Radiolabeling, and HPLC Analysis of Oligosaccharides-U937 cells (CRL 1593.2) and murine LLC cells (LLC1, CRL-1642) were purchased from the American Type Culture Collection (Manassas, VA) and cultured in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal bovine serum (Hyclone Laboratories, Logan, UT), glutamine (0.3 g/liter), streptomycin sulfate (100 g/ml), and penicillin (100 units/ml). Cells were maintained at 37°C in a humidified incubator containing 5% CO 2 and 95% air and passaged every 4 days using ATV TM trypsin solution (Invitrogen). The various disaccharides were dissolved in Me 2 SO and added to growth medium to achieve the indicated concentrations. The final concentration of Me 2 SO was adjusted to 0.5% (v/v) in all samples. Cells were added to the supplemented medium at a density of 2 ϫ 10 5 cells/ml. Cells were radiolabeled in growth medium containing 1 mM glucose. Samples of radiolabeled oligosaccharide products were analyzed by reverse phase HPLC as described (9). Nonradioactive samples were analyzed by mass spectrometry on a Thermo Finnigan LCQ classic mass spectrometer (San Jose, CA). Samples 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.
Enzyme-linked Immunosorbent Assay and Galactosyltransferase Assays-To measure the presence of sLe X and P-selectin ligand, cells were grown to confluence for 3-4 days in 6-well plates with and without 1-9. Binding of monoclonal antibody CSLEX-1 and P-selectin Ig chimera to cells fixed with 1% paraformaldehyde was measured by an enzyme-linked immunosorbent assay as described (9,13). Background binding was measured after treatment with sialidase, which showed that nonspecific binding was less than 5%.
The standard galactosyltransferase reaction (25 l) contained 10 milliunits of purified bovine milk ␤1-4-galactosyltransferase (Sigma), 50 mM MES, pH 6.0, 0.3 Ci of UDP-[1-3 H]galactose, 250 M UDP-galactose, 15 mM MnCl 2 , 50 mM KCl, and varying amounts of deacetylated 1 as acceptor. After incubation at room temperature for 15 min, the reaction products were diluted with 1 ml of 0.5 M NaCl and applied to a Sep-Pak C18 (100 mg; Waters). After washing with 5 ml of water, the products were eluted with 50% methanol, dried, and counted by liquid scintillation. Ovalbumin was tested as substrate at 10 mg/ml as described (19). The inhibitory activities of deacetylated 1 and 2 were tested at the indicated concentrations.
Experimental Metastasis-LLC cells were grown for 4 -5 days in the presence or absence of 50 M 1 or 2. The cells were harvested with EDTA, resuspended in sterile 0.9% saline, and injected (2 ϫ 10 5 cells/150 l) into the lateral tail vein of anesthetized (inhaled isoflurane, Janssen Pharmaceuticals, Titusville, NJ) 8 -12-week-old Es1(e) mice (bred in a C57Bl/6 background, a kind gift from P. Potter, University of North Carolina via Charles River Laboratories) (20). These mice carry a mutant allele of the esterase locus Es-1 designated Es1(e) (21) that results in reduced plasma esterase activity (22). After 10 days, lung metastases were detected by visible nodules on the surface of the lungs and quantitated under a dissecting microscope. Representative photographs were taken.
␤4Gal-T1 forms Gal␤1-4GlcNAc units in various glycoconjugates and has been shown to participate in the formation of sLe X and selectin ligands (24). Human ␤4Gal-T1 has been studied by x-ray crystallography, which has led to detailed insights into its kinetic mechanism (25)(26)(27). To determine how deacetylated 1 interacts with ␤4Gal-T1, we co-crystallized it with the open form of the human enzyme, hM340H-Gal-T1, in the presence of Mn 2ϩ and UDP-hexanolamine (Fig. 2a) (15,28,29). The apo-␤4Gal-T1 enzyme exists in an open conformation, where the acceptor binding site is absent. Upon binding of Mn 2ϩ and UDP-Gal, but not UDP and Mn 2ϩ , the wild type ␤4Gal-T1 undergoes conformational changes that create the acceptor binding site (29). It has been shown that mutation of the Met-344 residue in bovine ␤4Gal-T1 (corresponding residue Met-340 in human ␤4Gal-T1) to a His residue can facilitate these conformational changes readily in the presence of Mn 2ϩ and UDP-hexanolamine, creating the acceptor binding site (28). Therefore, in the past (15,28) and in the present crystallographic studies, instead of wild-type human ␤4Gal-T1, the

Disaccharide Analog Inhibitors
mutant hM340H-Gal-T1 in the presence of Mn 2ϩ , UDP-hexanolamine was used to determine the acceptor substrate binding to ␤4Gal-T1. In the presence of deacetylated 1, ␤4Gal-T1 was found in closed conformation with the Trp-310 side chain inside the catalytic pocket interacting with the bound UDPhexanolamine (Fig. 2a). The long flexible loop in this structure is also found in the closed conformation, with His-343 coordinated with Mn 2ϩ covering the UDP-hexanolamine and expos-ing the acceptor-substrate binding site to facilitate the binding of deacetylated 1. The N-acetylglucosamine residue of deacetylated 1 is tightly bound in the acceptor sugarbinding site, and the ␤-linked galactose residue hydrophobically stacks on the aromatic side chain of Tyr-282 (Fig. 2b). The aglycone (naphthalenemethanol) extends out from the binding site. A structural water molecule forms a hydrogen bond with the side-chain amino group of Arg-355 as well as between the N-acetylglucosamine and galactose residues stabilizing the complex (Fig. 2b). A molecular surface depiction shows the similarity of binding of deacetylated 1 and the more natural substrate analog, GlcNAc␤1-2Man␣1-6Man␤- (Fig. 2, c-e).
Deacetylated 1 (GlcNAc␤1-3Gal␤-O-NM) acted as a substrate exhibiting an apparent K m of ϳ10 M (Fig. 3a), a value lower than that reported for GlcNAc␤1-2Man␣1-6Man (K m ϳ 60 M) (15). Although the GlcNAc moiety exhibits only  hydrophobic interactions with the ␤4Gal-T1 molecule (Fig. 2, c  and d), the differences in the interaction of the galactose residue in deacetylated 1 versus the corresponding mannose residue in the trisaccharide substrate and additional contributions from the interaction of the naphthalenemethanol moiety with the enzyme may explain why deacetylated 1 has a lower K m than that of GlcNAc␤1-2Man␣1-6Man (Fig. 2e). Since ␤4GalT-1 transfer galactose to the C-4Ј hydroxyl group of GlcNAc-containing acceptors, one would predict that the deacetylated 4Ј-deoxy derivative of 1 would not act as a substrate. As expected, this compound lacked acceptor activity (deacetylated 2) when incubated with bovine ␤4GalT-I.
Previous studies showed that alterations of the C-4Ј hydroxyl groups in acceptors for ␤4Gal-T1 (i.e. by deoxygenation, halo-genation, alkylation, or amination of the N-acetylglucosamine residue) affected binding and reactivity of GlcNAc-containing substrates (14, 30 -36). However, the enhanced interaction of deacetylated 1 with the enzyme compared with other substrates suggested that deacetylated 2 might act as a competitive inhibitor. As shown in Fig. 3b, deacetylated 2 acted as an inhibitor using deacetylated 1 as substrate. Replotting the data in reciprocal fashion (1/v versus 1/[S]) showed that it acted in a competitive manner with an apparent K i of 192 Ϯ 83 M. Interestingly, both deacetylated 1 and 2 also inhibited the transfer of galactose to a glycoprotein substrate, ovalbumin, with comparable efficacy (Fig. 3c). These data differ from previous findings showing that 4-deoxy-GlcNAc-O-methyl or 4-deoxy-Glc-NAc␤1-3Gal-O-methyl were not inhibitory (31,33). One explanation for this discrepancy is that the large aromatic aglycone in deacetylated 2 compared with the O-Me group in previous derivatives may partially compensate for the loss of binding energy due to removal of the C-4Ј hydroxyl group, perhaps by interaction with Tyr-282 in the acceptor binding site (Fig.  2b). The importance of the naphthalenemethanol moiety is also consistent with other studies showing that substrates containing aromatic aglycones can inhibit the enzyme from acting on free GlcNAc or GlcNAc-O-methyl with K i values lower than K m for GlcNAc (K i ϭ 10 -22 M versus K m ϭ 1-10 mM) (33,34).
Acetylated Disaccharide Analogs Block sLe X Formation in Tumor Cells-To test if the 4Ј-deoxy analog 2 had inhibitory activity in vivo, the compound was added to the culture medium of U937 human monocytic lymphoma-derived cells, and the expression of sLe X on the cell surface was monitored with the monoclonal antibody CSLEX-1. Analog 2 diminished cell surface sLe X expression in a dose-dependent manner and to a similar extent as the parent compound 1 (Fig. 4). The 4Ј-deoxy analog 2 also inhibited binding of a P-selectin-IgG chimera to cells in a dose-dependent manner similar to 1 with an ED 50 value of ϳ30 M (Fig. 5).
Since compound 2 lacks the acceptor hydroxyl group for galactosylation, in theory, any further modifications by fucosy- FIGURE 4. Inhibition of sLe X and ␤4GalT-1 by disaccharide analogs. U937 cells were grown in the absence or presence of various concentrations of each peracetylated disaccharide analog for 4 days, as indicated. sLe X was measured by an enzyme-linked immunosorbent assay using monoclonal antibody, CSLEX-1. The values represent the average of triplicate measurements, which varied by Ͻ10%. Treatment of control cells with Arthrobacter ureafaciens sialidase (AUS) abrogated binding of CSLEX-1. The y axis scale differs for various compounds because the individual compounds were tested in sets on different days. Black bar graphs, disaccharide 1; dark gray bar graphs, C-4Ј hydroxyl-modified analogs 2-6; light gray bar graphs, C-3Ј hydroxyl-modified analogs 7-9. U937 cells were grown in the absence or presence of various concentrations of 1 or 2 for 4 days, as indicated. The binding of P-selectin-Ig chimera was measured by enzyme-linked immunosorbent assay, as described (13). The values represent the average of triplicate measurements, which varied by Ͻ10%. Treatment of control cells with 5 mM EDTA abrogated binding of P-selectin chimera due to its dependence on Ca 2ϩ . lation and sialylation should not occur, since the relevant transferases require Gal␤1-4GlcNAc-R core as a substrate (9). As predicted, treating U937 cells with 2 in the presence of [ 3 H]galactose and [ 3 H]fucose did not yield any radiolabeled oligosaccharides generated on the glycoside (Fig. 6, b and d), whereas multiple neutral and charged (sialylated and sulfated) oligosaccharides were assembled on 1 (Fig. 6, a and c) (7,9). The lack of any glycosylated products was confirmed in nonradioactive experiments using mass spectrometry (data not shown). Since the parent compound 1 has been previously shown to inhibit sLe X expression in U937 cells by preventing fucosylation of endogenous glycoprotein substrates (9), these new findings suggest that analog 2 inhibits the formation of sLe X by a different mechanism.
Inspection of the crystal structure of ␤4Gal-T1 suggested that other C-4Ј derivatives might have potent inhibitory activity by affecting its interaction with Asp-314, which hydrogen-bonds with the C-4Ј hydroxyl group of the GlcNAc acceptor moiety. To this end, we synthesized a series of peracetylated analogs in which the C-3Ј and C-4Ј hydroxyl groups of the GlcNAc moiety were replaced by -H, -F, -N 3 , -NH 2 , or -OCH 3 . Incubation of U937 cells with each peracetylated disaccharide revealed that all of the analogs diminished cell surface sLe X expression in a dose-dependent manner and to a similar extent as 1 (Fig. 4).
Analog 2 Inhibits Experimental Tumor Metastasis-Since sLe X expression on tumor cells correlates with poor prognosis and aggressive metastasis (1), inhibitors of sLe X expression could have therapeutic value. Disaccharide 1 was previously shown to prevent experimental and spontaneous metastasis of murine LLC cells in mice due to inhibition of sLe X expression and diminished binding of tumor cells with platelets (12,13). Neither 1 nor 2 was toxic to LLC cells up to 50 M based on trypan blue exclusion and growth rates (Fig. 7a). Both compounds 1 and 2 decreased binding of monoclonal antibody CSLEX-1 to LLC cells by 49 and 30%, respectively (Fig. 7b). Treatment of LLC cells with analog 2 also dramatically reduced lung-metastatic tumor formation (77 Ϯ 12 tumors (n ϭ 5) to 13 Ϯ 6 (n ϭ 5) in an experimental metastasis model using compoundtreated versus vehicle (DMSO)-treated cells; p Ͻ 0.001) (Fig. 7,  c and d). Some animals receiving compound-treated cells had only one or two tumor nodules in the lung (Fig. 7d). Treatment of cells with compound 1 yielded similar results, indicating that the antimetastatic activity of these compounds is related to their ability to bind galactosyltransferase(s) and either act as a primer (compound 1) or as a competitive inhibitor (compound 2).
In summary, analogs of GlcNAc␤1-3Gal␤-O-NM block sLe X expression presumably by competitively inhibiting the action of ␤4Gal-T1 or related galactosyltransferases. In contrast, the parent compound 1 inhibits cell surface sLe X expression by a mechanism that involves priming of oligosaccharides and diversion of downstream enzymes involved in sLe X biosynthesis on tumor cells, such as ␣2-3sialyltransferases and ␣1-4-  fucosyltransferase (9). The deacetylated form of the 4Ј-deoxy analog can bind to ␤4Gal-T1 in vitro, and assuming that this is the primary target in vivo, binding leads to effective inhibition of the enzyme for natural glycoprotein substrates. We presume that the deacetylated form of compound 2 is responsible for inhibition, but formally we cannot exclude the possibility that a partially deacetylated derivative also might have activity.
Further studies are needed to determine if ␤4Gal-T1 is the primary target or if other galactosyltransferase might be inhibited (23). Regardless, 2 and other analogs listed in Fig. 1 provide more selective agents for blocking cell surface sLe X expression than inhibitors that act earlier in the pathway, such as N-acetylgalactosaminides, which block the formation of all O-GalNAclinked glycans. Our finding that the IC 50 value of 2 in vivo is ϳ5-fold less than the K i value in vitro of the deacetylated compound for ␤4Gal-T1 indicates that if this enzyme is indeed the target, partial inhibition may suffice to block multivalent interactions between selectins and tumor cell glycoconjugates involved in cell adhesion. These data suggest that nonsubstrate disaccharides have therapeutic potential through their ability to alter glycan-dependent pathologic processes. Further studies are warranted to explore whether other types of disaccharides and their derivatives can act in a similar manner in other glycosylation pathways.