Minimal Sulfated Carbohydrates for Recognition by L-selectin and the MECA-79 Antibody*

Sulfated forms of sialyl-LeXcontaining Gal-6-SO4 or GlcNAc-6-SO4 have been implicated as potential recognition determinants on high endothelial venule ligands for L-selectin. The optimal configuration of sulfate esters on the N-acetyllactosamine (Galβ1→4GlcNAc) core of sulfosialyl-LeX, however, remains unsettled. Using a panel of sulfated lactose (Galβ1→4Glc) neoglycolipids as substrates in direct binding assays, we found that 6′,6-disulfolactose was the preferred structure for L-selectin, although significant binding to 6′- and 6-sulfolactose was observed as well. Binding was EDTA-sensitive and blocked by L-selectin-specific monoclonal antibodies. Surprisingly, 6′,6-disulfolactose was poorly recognized by MECA-79, a carbohydrate- and sulfate-dependent monoclonal antibody that binds competitively to L-selectin ligands. Instead, MECA-79 bound preferentially to 6-sulfolactose. The difference in preferred substrates between L-selectin and MECA-79 may explain the variable activity of MECA-79 as an inhibitor of lymphocyte adhesion to high endothelial venules in lymphoid organs. Our results suggest that both Gal-6-SO4 and GlcNAc-6-SO4 may contribute to L-selectin recognition, either as components of sulfosialyl-LeX capping groups or in internal structures. By contrast, only GlcNAc-6-SO4 appears to contribute to MECA-79 binding.

L-selectin, the "leukocyte selectin," mediates the tethering and rolling of lymphocytes along high endothelial venules (HEVs) 1 in peripheral lymph nodes, a prerequisite for extravasation of the lymphocytes (1)(2)(3). By virtue of a calcium-type lectin domain at its amino terminus, L-selectin functions as a calcium-dependent, carbohydrate-binding receptor that recognizes a set of discrete counter-receptors (generally termed ligands) displayed on the luminal aspect of HEVs (reviewed in Refs. 4 and 5). Several HEV-expressed, L-selectin-reactive ligands (all of which possess mucin-like domains) have been identified as potential physiological ligands for L-selectin (re-viewed in Ref. 6). These include GlyCAM-1 (7), CD34 (8,9), and podocalyxin (10).
A large body of evidence has established that the optimal interaction between L-selectin and these HEV-expressed ligands requires sialylation, fucosylation, and carbohydrate sulfation (11)(12)(13)(14)(15)(16)(17). In an attempt to rationalize these requirements in terms of carbohydrate structures, a detailed analysis of the O-linked oligosaccharide side chains of GlyCAM-1 was conducted (18 -20). Sulfation analysis of acid-hydrolyzed glycans revealed monosulfated monosaccharides and disaccharides (Nacetyllactosamine) with equivalent levels of Gal-6-SO 4 and GlcNAc-6-SO 4 . Analysis of the simplest oligosaccharide side chains identified equivalent levels of two novel sulfated isomers of sialyl-Le X (sLe X ), 6Ј-sulfo-sLe X , containing Gal-6-SO 4 , and 6-sulfo-sLe X , containing GlcNAc-6-SO 4 (see Fig. 1), as capping groups of core-2 branched oligosaccharides. The finding of a sLe X motif was significant because this tetrasaccharide binds to all three selectins (reviewed in Refs. [21][22][23][24][25], although weakly. Following the identification of the 6Ј-and 6-sulfo forms of sLe X within GlyCAM-1, a number of studies have been directed at examining the binding of these and analogous structures to L-selectin (reviewed in Ref. 26). There is general agreement that sulfation at C-6 of GlcNAc, in the context of 6-sulfo-sLe X , enhances L-selectin binding relative to sLe X (27)(28)(29). Similar findings have been obtained using 3Ј-sulfo-Le X and 3Ј-sulfo-Le a as mimetics of sLe X (28, 30 -32). Furthermore, Kannagi and coworkers (33,34) have described 6-sulfo-sLe X -reactive antibodies that stain HEVs and inhibit L-selectin binding to HEVs. With respect to the contribution of Gal-6-SO 4 , several groups have reported that this modification of sLe X or of sLe X mimetics either enhances or does not affect L-selectin binding (29,30,(35)(36)(37). In marked contrast, Feizi and co-workers found that this modification eliminates binding to L-selectin (28,38).
Previously, in an attempt to understand the contribution of the two sulfate esters in question to L-selectin binding, we synthesized a limited series of sulfated lactose derivatives (39). As inferred from the ability of these compounds to compete the binding of an L-selectin/IgG chimera to GlyCAM-1, 6Ј,6-disulfolactose (containing sulfate esters in positions analogous to those in 6Ј,6-disulfo-sLe X ) exhibited significant affinity for Lselectin. In the present study, we have synthesized a more extensive series of mono-and disulfated lactose derivatives with various combinations of sulfate esters on Gal and Glc. These compounds were derivatized with lipid tails, creating neoglycolipids that could be coated onto wells of microtiter plates. Thus, direct binding studies could be done with Lselectin chimeras and L-selectin bearing lymphocytes, allowing specificity controls to verify that the observed binding was functionally relevant. These studies have led to the definition of the minimal sulfated structures that support recognition by L-selectin. In addition, we employed these compounds to investigate the epitope corresponding to MECA-79, a widely used monoclonal antibody (mAb) that recognizes HEV-expressed ligands for L-selectin in human and other species (6,40). The MECA-79 epitope has been shown to be sulfation-dependent, but the structural context for this modification has heretofore not been defined (14 -16). In the present study, we find that L-selectin favors the disulfated lactose derivative with modifications at both C-6 of Gal and C-6 of Glc, whereas MECA-79 favors the monosulfated lactose derivative bearing Glc-6-SO 4 .
Sialyllactose Neoglycolipid-The sialyllactose neoglycolipid was synthesized from 1-␤-O-allyllactose by first forming an N-allyl glycoside (42) and then coupling the glycosylamine to steroyl chloride. A solution of 0.63 g (1.0 mmol) of sialyllactose (NeuAc␣233Gal␤134Glc) (Neose, Horsham, PA) in 10 ml of allylamine was stirred at room temperature for 72 h. The solvent was removed by rotary evaporation, and the resulting residue was suspended in hexanes and evaporated five times. The resulting N-allyl glycoside was suspended in 2.5 ml of dimethylformamide; 0.675 ml of steroyl chloride (0.61 g, 2.0 mmol) in 1 ml of dimethylformamide containing 0.697 ml of diisopropylethylamine (0.52 g, 4.0 mmol) was added dropwise over 10 min; and the solution was stirred vigorously for 60 min. The reaction was quenched by the addition of 1 ml of methanol and extracted in a solution of 5 ml of hexanes and 5 ml of distilled water. The solution was centrifuged at high speed in a bench-top microcentrifuge. The unreacted steroyl chloride and diisopropylethylamine partitioned into the organic layer; the unreacted N-allylsialyllactose and diisopropylethylamine salts partitioned into the aqueous layer; and the neoglycolipid was trapped in a densely packed foam layer at the interface between the two solvents. This extraction was repeated five times, and the recovered foam layer was suspended in 40% aqueous acetonitrile, frozen, and lyophilized to yield 0.242 g (26%) of sialyllactose neoglycolipid. The calculated mass for C 44 H 78 N 2 O 19 was 938.52, and the measured mass was 961.6 [M ϩ Na] ϩ (liquid secondary ionization mass spectrometry, positive mode).
Preparation of GlyCAM-1-GlyCAM-1 was prepared as described previously (43). Briefly, mouse serum from Pel-Freez Biologicals (Rogers, AR) was extracted with 4 volumes of 2:1 chloroform/methanol, and the layers were separated by centrifugation at 2000 ϫ g. The upper aqueous layer was separated from the organic and precipitated protein layers, concentrated to one-half of the original serum volume by boiling, and dialyzed against 20 volumes of Dulbecco's phosphate-buffered saline (PBS) with two changes. This preparation was then diluted with PBS back to the original serum volume and used as an enriched source of GlyCAM-1. GlyCAM-1 was captured onto Immulon II microtiter plate wells using a purified rabbit anti-GlyCAM-1 peptide polyclonal antibody, CAM02 (44).
Selectin/IgM Chimeras-The recombinant murine L-selectin/human IgM chimera cDNA and the tissue culture supernatant containing murine E-selectin/human IgM were gifts from Dr. Lloyd Stoolman and have been described previously (45). The recombinant human L-selectin/human IgM chimera was prepared by subcloning human L-selectin cDNA, a gift of Dr. Thomas Tedder, into the human IgM vector used to make the murine L-selectin/human IgM chimera. Recombinant protein was produced in transiently transfected COS-7 cells grown in Opti-MEM serum-free medium (Life Technologies, Inc.). The level of expression was determined by enzyme-linked immunosorbent assay using immobilized goat anti-human IgM polyclonal antibody to capture the chimera from the supernatant and a goat anti-human IgM/alkaline phosphatase conjugate to detect the immobilized chimera. Human IgM was used as a standard. The chimeras were used as crude tissue culture supernatant.
Selectin Direct Binding Assay-Neoglycolipids were resuspended in 85% aqueous ethanol containing 5 g/ml cholesterol and 2.5 g/ml phosphatidylcholine. Serial dilutions were coated onto polyvinyl chloride microtiter plates at 40 l/well, and the solvent was evaporated at 37°C. It is assumed in enzyme-linked immunosorbent assays of this type that the input concentration of lipid accurately predicts the coating concentration. The finding that the rank orders of reactivity of the neoglycolipids for L-selectin, E-selectin, and MECA-79 are very distinct (see "Results") indicates that the differences are not just due to variations in the amount of lipid coated. The coated wells were washed once with distilled water and then blocked with 3% BSA in PBS for 2 h. Tissue culture supernatant containing the selectin/IgM chimera was diluted 1:1 in 2% BSA in PBS and incubated at 100 l/well for 2 h. After two washes in PBS, goat anti-human IgM/alkaline phosphatase conjugate diluted 1:1000 in 1% BSA in PBS was added at 100 l/well for 45 min. After five washes in PBS, bound selectin/IgM chimera and human IgM were detected by the addition of phosphatase substrate (p-nitrophenyl phosphate (1 mg/ml) in 10% diethanolamine and 0.1 mM MgCl 2 ) at 100 l/well. Absorbances were recorded at 405 nm on a Bio-Rad microplate reader. All incubations and wash steps were performed at ambient temperature.
MECA-79 Direct Binding Assay-Microtiter wells were coated with neoglycolipids and blocked with 3% BSA in PBS as described for the L-selectin direct binding assay. MECA-79 and the OZ-42 control antibody were diluted to 10 g/ml in 1% BSA in PBS and incubated at 100 l/well for 2 h. The wells were washed twice with PBS, and rabbit anti-rat IgM/alkaline phosphatase diluted 1:1000 in 1% BSA in PBS was added at 100 l/well for 45 min. After five washes in PBS, bound antibody was detected by the addition of p-nitrophenyl phosphate at 100 l/well. Absorbances were recorded at 405 nm. All incubations and wash steps were performed at ambient temperature.
Antibody Inhibition Assay-Microtiter plates were coated with 200 M solutions of neoglycolipid in 85% ethanol, 5 g/ml cholesterol, and 2.5 g/ml phosphatidylcholine at 40 l/well (8 nmol/well). The solvent was evaporated at 37°C, and the wells were blocked with 3% BSA in PBS. Anti-L-selectin or control mAb was diluted to 2 g/ml in 2% BSA in PBS and added in equal volumes to tissue culture supernatants containing the L-selectin/IgM chimera. The solutions were incubated for 30 min and then transferred to the neoglycolipid-coated plates at 100 l/well. After a 2-h incubation, bound chimera was probed and detected as described for the direct binding assay. All incubations and wash steps were performed at ambient temperature. EDTA Inhibition Assay-Microtiter plates were coated with neoglycolipid and blocked as described for the antibody inhibition assay. Serial dilutions of EDTA in 2% BSA in PBS were combined in equal volumes with undiluted L-selectin/IgM-containing tissue culture supernatant and added to the neoglycolipid-coated wells at 100 l/well. After a 2-h incubation, bound chimera was probed and detected as described for the direct binding assay. All incubations and wash steps were at ambient temperature. The EqCal software package (Biosoft, Cambridge, United Kingdom) was used to calculate the free calcium and magnesium levels as a function of the amount of EDTA added to the Dulbecco's PBS/Opti-MEM sample buffer. The calculations were based on the following parameters: an initial concentration of 0.90 mM for calcium, an initial concentration of 0.75 mM for magnesium, pH 7.4, 25°C, and an ionic strength of 0.1 M.
Cell Adhesion Assay-L-selectin-expressing Jurkat T-cells were obtained from the laboratory of Dr. Arthur Weiss and maintained in RPMI 1640 medium supplemented with 100 units/ml penicillin, 100 mg/ml streptomycin, 2 mM glutamine, and 5% heat-inactivated fetal calf serum (Hyclone Laboratories, Logan, UT). Neoglycolipids were diluted to 200 M in 85% ethanol, 5 g/ml cholesterol, and 2.5 g/ml phosphatidylcholine and coated onto microtiter plates at 40 l/well (8 nmol/well).
After drying, the wells were blocked with 3% BSA in PBS for 2 h. Jurkat cells were centrifuged and resuspended to 5 ϫ 10 6 /ml in 0.1% BSA in PBS. The fluorescent dye BCECF/AM was added at a dilution of 1:1000 from a 2 mM stock solution in dimethyl sulfoxide and incubated with the cells for 20 min in the dark. The labeled cells were centrifuged, resuspended to 2 ϫ 10 6 /ml in 0.1% BSA in PBS, and transferred to the neoglycolipid-coated plate at 100 l/well. After 30 min, the plate was washed twice with PBS, and 0.1% BSA in PBS was added at 100 l/well. Arbitrary fluorescence intensity units were recorded at 485 nm excitation and 530 nm emission on a CytoFluor II fluorescence multiwell plate reader (PerSeptive Biosystems, Foster City, CA). To demonstrate Lselectin-dependent binding, some cells were incubated with 10 mM EDTA or with DREG-56 or a class-matched control antibody at 10 g/ml for 20 min prior to incubation with the immobilized neoglycolipids. All reactions, washes, and centrifugation steps were performed at ambient temperature.
Statistics-Significant differences among means were determined by one-way analysis of variance. When significant differences were detected, multiple pairwise comparisons were performed using a Tukey test (SigmaStat statistical software, SPSS Inc., Chicago, IL). Unless stated otherwise, the data shown are from representative experiments and are the average of duplicate wells after subtraction of background signal (carrier lipids only), with error bars denoting the range in signal.

Binding of Recombinant L-selectin to Sulfated
Lactose Neoglycolipids-To investigate the contribution of sulfate esters in defined positions to L-selectin binding, serial dilutions of a panel of sulfated lactose neoglycolipids with 25-atom single chain hydrocarbon tails (Fig. 1) were immobilized on microtiter plate wells and assayed for their ability to support the binding of L-selectin/IgM chimeras. For both the human and murine L-selectin chimeras, the best substrate for binding was 6Ј,6disulfolactose ((SO 4 -6)Gal␤134(SO 4 -6)Glc), but significant binding was also observed with the 6Ј-sulfolactose ((SO 4 -6)Gal␤134Glc) and 6-sulfolactose (Gal␤134(SO 4 -6)Glc) derivatives (Fig. 2, A and B). At a coating concentration of 5 nmol/ well, 6Ј,6-disulfolactose generated a signal 2.4-fold greater than 6-sulfolactose (p Ͻ 0.001) and 2.8 -3.6-fold greater than 6Ј-and 3Ј-sulfolactose ((SO 4 -3)Gal␤134Glc) (p Ͻ 0.001). Differences in reactivity between 6Ј,6-disulfolactose and the other sulfated derivatives and between all of the sulfated derivatives and lactose or carrier lipids only (data not shown) were statistically significant (p Ͻ 0.001). These results are consistent with previous data in which similar compounds were used as soluble inhibitors of L-selectin binding to GlyCAM-1 (39). There were no significant differences in the reactivity profiles between the human and murine chimeras; however, the murine chimera consistently generated higher signals than the human chimera at comparable concentrations. There was no detectable binding of either chimera to non-sulfated lactose or sialyllactose (data not shown).
To address the possibility that the greater reactivity of Lselectin for 6Ј,6-disulfolactose relative to the monosulfated derivatives was due to an increase in negative charge rather than to a specific configuration of sulfate esters, L-selectin binding to 3Ј,6Ј-disulfolactose ((SO 4 -3)(SO 4 -6)Gal␤134Glc) and 3Ј,6disulfolactose ((SO 4 -3)Gal␤134(SO 4 -6)Glc) was determined (Fig. 3). The addition of a sulfate ester to C-3Ј of either 6Ј-or 6-sulfolactose did not yield binding greater than that of the parent monosulfated lactose derivative.
To further characterize the binding of L-selectin to the sulfated lactose neoglycolipids, we determined the effects of a FIG. 1. Sulfated lactose derivatives modeled after the sulfated N-acetyllactosamine core of sulfo-sLe X . A, sulfo-sLe X . The N-acetyllactosamine core on which the sulfated lactose neoglycolipids are based in B is shown boxed. Sulfation at C-6 of Gal (R 1 ) defines 6Ј-sulfo-sLe X . Sulfation at C-6 of GlcNAc (R 2 ) defines 6-sulfo-sLe X . Sulfation at C-6 of Gal and GlcNAc defines 6Ј,6-disulfo-sLe X . B, sulfated lactose derivatives and sialyllactose. C, aLe X . D, lipid tail of sulfated lactose neoglycolipids.

FIG. 2. Binding of L-selectin/IgM to sulfated lactose neoglycolipids.
Serial dilutions of sulfated lactose neoglycolipids were dried onto wells of microtiter plates and assayed for their ability to support binding of the human (A) and murine (B) L-selectin/IgM chimeras. The enhanced binding to 6Ј,6-disulfolactose relative to the monosulfated compounds was observed in 12 independent experiments using three different preparations of 6Ј,6-disulfolactose and two different preparations each of 6Ј-and 6-sulfolactose at different times. f, 6Ј,6-disulfolactose; E, 6Ј-sulfolactose; Ⅺ, 6-sulfolactose; ‚, 3Ј-sulfolactose. function-blocking mAb (DREG-56) (52) and divalent cation chelation on the binding of the human L-selectin chimera. As shown in Fig. 4, the binding of L-selectin to all of the sulfated lactose derivatives was strongly inhibited by DREG-56 or by 10 mM EDTA. Similarly, mouse L-selectin binding to these sulfated compounds was inhibited by a function-blocking mAb (MEL-14) or by EDTA (data not shown). The binding of wheat germ agglutinin and ricin toxin agglutinin to lactose or 6Ј,6disulfolactose, respectively, was not inhibited by 20 mM EDTA, indicating that EDTA does not strip the neoglycolipid substrate from the microtiter well (data not shown).
We investigated the divalent cation dependence of L-selectin binding to the most active sulfated lactose neoglycolipids in detail by measuring chimera binding as a function of EDTA concentration. As shown in Fig. 5, L-selectin binding to 6Јsulfolactose, 6-sulfolactose, and 6Ј,6-disulfolactose exhibited the identical divalent cation dependence as observed for its binding to GlyCAM-1. 50% inhibition of binding was achieved at 0.9 mM EDTA, which corresponds to a calcium concentration of 80 M. This value is in accord with previous estimates of the amount of calcium needed for L-selectin function (66).
Binding of Jurkat T-cells-To investigate the binding of Lselectin in a cellular context, we examined the adhesion of an L-selectin-expressing T-cell line (Jurkat) to the same panel of sulfated lactose neoglycolipids. As was the case for the Lselectin chimeras, the Jurkat cells bound best to 6Ј,6-disulfolactose (Fig. 6), and differences in reactivity between 6Ј,6-disulfolactose and the other sulfated derivatives were statistically significant (p Ͻ 0.001). Binding to the other sulfated derivatives was clearly above background levels, (p Ͻ 0.001), but without notable differences in potency among them. Jurkat cell binding to 6Ј,6-disulfolactose was effectively inhibited by EDTA and the DREG-56 mAb (Fig. 6, inset), verifying that the interaction was L-selectin-dependent and exhibited the same characteristics as the chimera binding.
Binding of MECA-79 -To determine whether there was overlap between the MECA-79 epitope and the L-selectin recognition determinant, we assayed MECA-79 binding to the sulfated lactose neoglycolipids (Fig. 7). The strongest binding was observed with 6-sulfolactose, whereas much less binding was observed with 6Ј-and 3Ј-sulfolactose. Surprisingly, the weakest signal among the sulfated compounds was seen with 6Ј,6-disulfolactose. MECA-79 binding, however, was much weaker relative to the L-selectin/IgM chimeras, as evidenced by a significantly longer substrate conversion time. The control rat IgM was not reactive with any of the compounds tested (data not shown).
Binding of E-selectin-E-selectin has previously been shown to bind to sulfated derivatives of Le X (Gal␤134) (Fuc␣133)GlcNAc) or Le a (Gal␤133 (Fuc␣134) GlcNAc) (30,38,53,54). In view of the binding of L-selectin to sulfated lactose derivatives, we wanted to examine the interaction of E-selectin with the same compounds in parallel assays. The preferred structure for the E-selectin/IgM chimera was the non-sulfated sLe X analog, aLe X (Fig. 8A). This compound consists of a Le X -like trisaccharide with Glc substituting for GlcNAc and with a 3Ј-O-acetic acid substituting for the 3Ј-sialic acid of sLe X (Fig. 1C). The L-selectin chimera did not FIG. 3. Binding of L-selectin/IgM to disulfated lactose neoglycolipids. Serial dilutions of neoglycolipids were dried onto microtiter plate wells, and human L-selectin/IgM was assayed for binding. The enhanced binding of 6Ј,6-disulfolactose relative to the other disulfated derivatives was observed in four independent experiments. The data shown for 6Ј,6-disulfolactose are the same as those shown in Fig. 2A; all compounds were tested in the same experiment. At a coating concentration of 5 nmol of neoglycolipid/well, differences in reactivity between 6Ј,6-disulfolactose and the other sulfated derivatives (3Ј,6Ј-and 3Ј,6disulfolactose) were statistically significant (p Ͻ 0.001). f, 6Ј,6-disulfolactose; E, 3Ј,6Ј-disulfolactose; Ⅺ, 3Ј,6-disulfolactose.

FIG. 4. Antibody and EDTA inhibition of L-selectin/IgM binding.
Neoglycolipids were dried onto microtiter wells at a concentration of 8 nmol/well. Prior to incubation with the dried lipids, the L-selectin chimera was incubated with antibody or EDTA for 30 min at room temperature. A, effects of DREG-56 (black bars) or a class-matched control antibody (gray bars); B, effects of 10 mM EDTA. These results were obtained in three independent experiments. 6Ј,6, 3Ј,6Ј, and 3Ј,6, 6Ј,6-, 3Ј,6Ј-, and 3Ј,6-disulfolactose, respectively; 6Ј, 6, and 3Ј, 6Ј-, 6-, and 3Ј-sulfolactose, respectively. show detectable binding to this sLe X analog at a concentration of 8 nmol/well, although (as above) L-selectin bound markedly to various sulfated lactose derivatives at this concentration. A markedly different pattern of E-selectin/IgM binding to the sulfated compounds was observed compared with L-selectin/ IgM (Fig. 8B). The strongest binding was to 6-sulfolactose, followed by 6Ј,6-disulfolactose. In contrast to the results with L-selectin, E-selectin binding was seen only at the highest concentration of neoglycolipid coated (8 nmol/well). 6Ј-Sulfolactose failed to support E-selectin binding at any concentration (Fig. 8A). Despite the ability of E-selectin to bind to 3Ј-sulfo-Le X and 3Ј-sulfo-Le a (31,32,55), E-selectin did not bind to 3Ј-sulfolactose or 3Ј,6-disulfolactose and bound only weakly to 3Ј,6Ј-disulfolactose (Fig. 8A). Like L-selectin binding, E-selectin binding was strongly inhibited by divalent cation chelation using EDTA (Fig. 8A). These comparisons add to the existing evidence for a fundamental difference in the specificities of these two selectins (56 -58). DISCUSSION Interest in the Gal-6-SO 4 and GlcNAc-6-SO 4 modifications within L-selectin ligands was originally prompted by our analysis of the acid hydrolysis products of GlyCAM-1 (18). Recon-stitution experiments performed with recently cloned Gal-and GlcNAc-6-O-sulfotransferases have justified this interest by confirming that both of these modifications (i.e. Gal-6-SO 4 and GlcNAc-6-SO 4 ) can contribute to L-selectin ligand activity (59 -61). Interestingly, transfection of Chinese hamster ovary cells with a combination of GlcNAc-and Gal-6-O-sulfotransferase cDNAs imparted much greater L-selectin ligand activity than either sulfotransferase alone, indicating a synergistic contribution from the two kinds of sulfation. The present study was directed at examining the contribution of Gal-6-SO 4 and Glc-NAc-6-SO 4 to L-selectin ligand activity in isolation from the influence of sialylation and fucosylation. Guided by the analysis of the simplest chains of GlyCAM-1 (18 -20), the minimal structures that we synthesized were based on the N-acetyllactosamine core (Gal␤134GlcNAc) of sulfo-sLe X . Although the C-2 N-acetate of GlcNAc in N-acetyllactosamine is missing in the sulfated lactose (Gal␤134Glc) derivatives studied here, this alteration has been shown to have a minimal effect on L-selectin binding (62).
A significant advantage of the direct binding studies over competition studies is that important specificity controls could be performed. Thus, we were able to establish that the binding of the L-selectin/IgM chimeras to the sulfated neoglycolipids (as well as Jurkat cell binding to 6Ј,6-disulfolactose) could be effectively inhibited by function-blocking antibodies or by divalent cation chelation. EDTA titration demonstrated that the divalent cation requirement for L-selectin binding to the preferred sulfated lactose derivatives (6Ј,6-disulfolactose, 6Ј-sulfolactose, and 6-sulfolactose) was identical to that of native GlyCAM-1. These results strongly argue that these compounds are engaged by a site in the L-selectin calcium-type lectin domain that is critical for recognition of physiological ligands. Further supporting evidence for this conclusion derives from our previous observation that soluble sulfated lactose derivatives (with 6Ј,6-disulfolactose as the best inhibitor) can compete the binding of L-selectin to GlyCAM-1 (39). The fact that specific sulfated lactose derivatives, lacking fucose and sialic acid, can bind to relevant sites in L-selectin again emphasizes the importance of sulfate esters with the appropriate spatial orientation as recognition elements. The observation that L-selectin bound better to the sulfated lactose neoglycolipids than to the non-sulfated sLe X analog is consistent with our previous report that 6Ј,6-disulfolactose is a superior inhibitor of L-selectin binding to GlyCAM-1 than sLe X (39).
Further work will be necessary to understand the relationship between the minimal binding structures defined above and actual recognition determinants of HEV-expressed ligands for L-selectin. As reviewed above, a variety of studies have established that 6-sulfo-sLe X (NeuAc␣233Gal␤134(Fuc-␣133)(SO 4 -6)GlcNAc) is an important recognition determinant (27, 34, 38, 59 -61, 63). Contained within this structure is 6-sulfo-N-acetyllactosamine, which, from the data presented herein, can contribute a significant degree of Lselectin binding by itself. Given that L-selectin functions in the dynamic processes of tethering and rolling of lymphocytes, a complete analysis of 6-sulfo-sLe X must examine the contribution of each modification (including sulfation) to kinetic constants as well as to the overall equilibrium constant. A previous study has shown that sulfated Le X derivatives can exhibit very different inhibitory activities against L-selectin depending upon whether equilibrium or flow chamber assays are used (37).
The structural context for the contribution of Gal-6-SO 4 is less certain at the present time. As reviewed above, there is significant controversy as to whether sulfation at C-6 of Gal augments the affinity of sLe X or 6-sulfo-sLe X for L-selectin. Furthermore, antibody staining studies by Kannagi and coworkers (34) have failed to detect the presence of 6Ј-sulfo-sLe X (NeuAc␣233(SO 4 -6)Gal␤134(Fuc␣133)GlcNAc) or 6Ј,6-disulfo-sLe X (NeuAc␣233(SO 4 -6)Gal␤134(Fuc␣133)(SO 4 -6)Glc-NAc) determinants on HEVs in human lymphoid organs. The present study demonstrates that 6Ј-sulfolactose and especially 6Ј,6-disulfolactose can support L-selectin binding without sialylation or fucosylation. One possibility is that Gal-6-SO 4 may contribute to L-selectin binding in the context of 6Ј-sulfo-or 6Ј,6-disulfo-N-acetyllactosamine, lacking fucose and/or sialic acid. Such structures might exist as glycan capping groups or be present internally in extended oligosaccharide side chains.
In this context, it should be stressed that native GlyCAM-1 possesses extended and multisulfated chains, the structures of which have not been solved (20).
The finding that L-selectin binding to the sulfated lactose derivatives was highly sensitive to divalent cation chelation by EDTA is somewhat surprising. Structural analysis of an engi-neered form of E-selectin complexed with sLe X revealed a role for calcium in coordinating the C-2 and C-3 hydroxyls of fucose, in addition to four amino acids in the lectin domain (64). Furthermore, calcium-independent L-selectin binding has been demonstrated for a number of non-fucosylated anionic molecules, e.g. sulfatide, lipid A, lipopolysaccharide, and cardiolipin (48,65,67). 3 Thus, it was expected that binding to the sulfated lactose neoglycolipids, which lack fucose, would be calciumindependent. It is possible that in addition to coordinating vicinal hydroxyls on fucose, calcium is important for maintaining a properly folded lectin domain for optimal engagement of carbohydrate or specific sulfate esters. Indeed, this is the case with other calcium-type lectins (68,69). Further study of this issue is warranted.
The MECA-79 antibody has been an invaluable probe for L-selectin ligands on HEVs of normal lymphoid organs and on activated endothelium at sites of chronic inflammation (6). Sulfation, but not fucosylation or sialylation, is required for this epitope, but detailed structural information has been lacking (15,17). Recent experiments using a cloned sulfotransferase have revealed an essential contribution of GlcNAc-6-SO 4 for MECA-79 binding, but again did not illuminate a structural context for this modification (60). The experiments conducted here showed that 6-sulfolactose supported MECA-79 binding considerably better than the other sulfated derivatives. This result is consistent with the demonstrated requirement for GlcNAc-6-SO 4 , but the inactivity of the 6Ј,6-disulfated derivative was unexpected. In light of our results, one possibility is that the function-blocking activity of this antibody is achieved by neutralizing recognition determinants such as 6-sulfo-sLe X and 6-sulfo-N-acetyllactosamine. The reduced affinity of MECA-79 for 6Ј-sulfolactose and 6Ј,6-disulfolactose may signify that this antibody cannot directly neutralize the contribution of Gal-6-SO 4 to ligand activity. These considerations may explain the variable ability of MECA-79 to block lymphocyte attachment to HEVs at different anatomical sites (40,70,71).
The studies reported here advance our understanding of the minimal sulfated structures that may be involved in L-selectin binding. A key question for future work will be to define the full recognition determinants of its endothelial ligands. There is considerable indirect evidence that the structures of these determinants differ as a function of the anatomical location of the secondary lymphoid organ (34,72). Varying contributions from Gal-6-SO 4 and GlcNAc-6-SO 4 may underlie this apparent diversity of recognition elements. The need for further information also applies to the inducible L-selectin ligands on nonlymphoid endothelium, which are implicated in the inflammatory trafficking of leukocytes (41,73).