Synergistic interactions of the two classes of ligand, sialyl-Lewis^a/x fuco-oligosaccharides and short sulpho-motifs, with the P- and L-selectins: implications for therapeutic inhibitor designs

Abstract

The E-, L- and P-selectins are carbohydrate-recognizing cell-adhesion molecules mediating selective leucocyte recruitment in inflammation. The 3′-sialyl- and 3′-sulpho-oligosaccharides of Lewis^x (Le^x) and Lewis^a (Le^a) series are bound by them, but for high-avidity binding of P- and L-selectins to the glycoprotein counter-receptor known as P-selectin glycoprotein ligand, PSGL-1, there is a requirement for sulpho-tyrosines neighbouring a sialyl-Le^x glycan. The two selectins can also bind 3-O- or 6-O-sulphated galacto-lipids (sulphatides). Here we compare some features of the interactions of P- and L-selectins with a novel lipid-linked sulpho-tyrosine probe, and with the sulphatides and neoglycolipids of sialyl- and sulpho-Le^x/Le^a fuco-oligosaccharides. The sulpho-tyrosine probe is bound by both selectins. There are close similarities in the interactions of the two selectins with sulpho-tyrosine and the sulphatides; the binding is relatively resistant to chelation of calcium ions, in contrast to the absolute requirement of calcium ions with the long fuco-oligosaccharides, including 6-sulpho-sialyl-Le^x. With both selectins, there is striking synergy in binding signals elicited by the two ligand types when presented as equimolar mixtures on a matrix. Thus, there are two operationally distinct binding sites on both L- and P-selectin; and the binding sites for sulphate groups in the two ligand types are probably distinct. When sulpho-tyrosine and sialyl-Le^x are presented on liposomes, a potent inhibitory activity is generated toward the binding of P-selectin to HL60 cells, with 50% inhibitory concentration (IC₅₀) values in the nanomolar range. These features of the lipid-linked ligand analogues, and the simple approach for their display on liposomes, may have applications in designs and screening of selectin inhibitors as anti-inflammatory compounds.

Introduction

Three cell adhesion molecules – E- and P-selectins on microvascular endothelia, and L-selectin on leucocytes – are key players at the initial stages of inflammatory responses to infection and injury, and lymphocyte emigration into lymphoid organs.^1–4 Their role is to initiate the processes of extravasation of selected leucocytes by means of the cell tethering and rolling that they mediate under conditions of shear force at the microvascular endothelial surface. Cascades of other, much tighter, cell adhesion and activation events can then occur, which lead to migration of the leucocytes through endothelium and underlying basement membrane. Regulated expression of the selectins, and of their ligands, thus serves to initiate and also to terminate the inflammatory response. Inappropriate expression, however, of these adhesion molecules contributes to the pathology of inflammatory disorders.^5,6 Knowledge of the molecular basis of the selectin interactions is considered important as a possible lead to novel therapeutic interventions.

Since the discovery of lectin domains at the tips of the three selectins, much effort has been directed at identifying their carbohydrate ligands. Initial developments occurred rapidly. Knowledge that human E- and P-selectins bind granulocytes and monocytes, served to focus research on 3′-fucosyl-N-acetyllactosamine (Le^x), and sialyl-Le^x sequences, which are distinctive markers of myeloid cells.^7,8 The 3′-sialyl-Le^x and the isomeric sequence 3′-sialyl-Le^a were readily shown to be recognized by all three selectins.^2,9,10 A picture has emerged, however, of differences in the binding specificities of the selectins, such that variant carbohydrate sequences related to sialyl-Le^a/x, and additional elements on the counter-receptors, elicit preferential binding by one or other of these receptors.^11–13 Carbohydrate ligands identified for E-selectin, apart from the above-mentioned sialyl sequences, include the 3′-sulpho-Le^a and 3′-sulpho-Le^x sequences as found on epithelial mucins.¹⁴ The sulpho-Le^a and -Le^x are also bound by the L- and P-selectins (reviewed in ref. 15). However, as discussed below, other sulphate-containing motifs have been described on glycoprotein counter-receptors for L- and P-selectins. These have been shown to be critical elements for high-avidity binding of L-selectin to the glycoprotein GlyCAM-1 (abbreviation for glycoprotein cell adhesion molecule-1) and of P-selectin to PSGL-1 (abbreviation for P-selectin glycoprotein ligand-1), the sulphate being carried on carbohydrate and on protein, respectively.^11,12

Direct binding experiments with structurally defined oligosaccharide sequences have revealed¹⁶ that the sialyl-Le^x sequence that is modified by 6-O-sulphation at N-acetylglucosamine, as found on the L-selectin counter-receptor, GlyCAM-1,¹⁷ rather than 6-O-sulphation at the outer galactose, is the most potent ligand thus far for L-selectin. The potency is enhanced if the sialic acid is non-acetylated.^16,18,19 On PSGL-1, the major counter-receptor for P-selectin, the presence of sulphated tyrosine residues, in addition to a sialyl-Le^x glycan, assembled at the N-terminal tip of the counter-receptor, is required for high-avidity binding.^6,20–24 The interaction of the lectin domain of P-selectin with the sialyl-Le^x-bearing sulpho-glycopeptide portion of PSGL-1 has now been demonstrated by X-ray crystallography.²⁵ PSGL-1 is also bound by L- and E-selectins; for L-selectin binding, but not for E-selectin, again there appears to be a requirement for both the sialyl-Le^x and the sulpho-tyrosine.^6,12 Thus, the P- and L-selectins differ from E-selectin in their recognition of an amino-acid-associated determinant, as well as carbohydrate.

Interactions of L-selectin with several other acidic compounds have been described when these are lipid linked and presented in the clustered state, i.e. in oligomeric form through clustering of the lipid-linked motifs, which generate the avidities required for readily detectable binding.²⁶ These include sulphated galactosyl-ceramides (sulphatides), sulphated ganglio-series glycolipids, and neoglycolipids derived from glycosaminoglycan disaccharides (reviewed in refs ¹⁵ and ²⁷). P-selectin also binds to these compounds (ref. ²⁸ and R. A. Childs, unpublished). Here, the position of the sulphate on the sulphatides²⁹ and the glycosaminoglycan disaccharides³⁰ is not critical. This is in contrast to the requirement for 3-O-sulphation at the terminal galactose³¹ of Le^a and Le^x and the 6-O-sulphation of the N-acetylglucosamine of the sialyl-Le^x sequences.

The binding of all three selectins to the sialyl-Le^a and -Le^x sequences is calcium dependent, whereas the binding of L- and P-selectin to the sulphatides and glycosaminoglycans is less sensitive to calcium chelation.^28,30 P-selectin binding to the N-terminal sulpho-glycopeptide domain of PSGL-1, bearing sulphate at tyrosines 46, 48 and 51, is also dependent on the presence of calcium ions, as shown by recent studies with synthetic tyrosine-sulphated analogues of this domain.²⁴ Binding could not be detected, however, to the sulpho-peptide in the absence of the fuco-oligosaccharide; thus, it was not possible to determine whether the binding to the sulpho-tyrosine is dependent on the presence of calcium ions.

Here we examine some features of the interactions of the P- and L-selectins with sialyl- or sulpho-Le^x or Le^a, and 6-sulpho-sialyl-Le^x sequences, sulpho-tyrosine and the 3- and 6-sulphated sulphatides. The interactions with the ligands, in the form of lipid-linked probes, are examined individually or as mixtures in binding and inhibition of binding experiments. Two classes of ligands can thus be discerned:

• the fuco-oligosaccharides (sialylated or sulphated) that are bound with an absolute requirement for calcium ions; and

• the short sulpho-ligands whose binding is relatively resistant to chelation of calcium ions.

We observe that the two types of ligand act synergistically with both the L- and P-selectins, and demonstrate the application of this synergy in the design of potent selectin inhibitors.

Materials and Methods

Chemicals

L-1,2-dihexadecyl-sn-glycero-3-phosphoethanolamine (DHPE), N,N-diisopropylethylamine and N,N-dimethylformamide were obtained from Aldrich (Gillingham, Dorset, UK). 2-(1H-benzotriazole-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate and N-hydroxybenzotriazole were from Novabiochem (Laufelfingen, Switzerland). N-Fmoc-O-sulphate-tyrosine-Na was from Bachem (UK) Ltd. (St Helens, Merseyside, UK). Silicagel 60 aluminium-backed, high performance thin-layer chromatography (TLC) plates. (Art. 5547) were from Merck (Lutterworth, Leicester, UK). Cholesterol sulphate (Su chol) was from Sigma (Poole, Dorset, UK).

Recombinant soluble selectins

Soluble forms of murine E-, L- and P-selectins fused to the CH2, CH3 and CH4 domains of human immunoglobulin M (IgM) and expressed in transfected COS-7 cells³² were used as culture supernatants (kindly provided by Dr John Lowe, University of Michigan Medical School, Ann Arbor, MI). The culture medium was Dulbecco's modified Eagle's minimal essential medium (DMEM) containing 10% fetal calf serum (FCS), 100 IU/ml of pencillin and 100 µg/ml of streptomycin. Human L-selectin fused to the Fc portion of human immunoglobulin G (IgG) and expressed in transfected Chinese hamster ovary cells (kindly provided by Dr Gray Shaw at the Genetics institute, Cambridge, MA) was isolated from tissue culture supernatant, as described previously.¹⁶

Compounds investigated

The compounds investigated are shown in Table 1.

Table 1. Compounds investigated.

Lipid-linked saccharides

The glycolipids containing 3′-sialyl-Le^x trisaccharide (SLe^x tri) sequence (GSC161), the 3′-sulpho-Le^x trisaccharide [(SuLe^x tri), GSC150], the 6-sulpho-sialyl-Le^x pentasaccharide [(6SuSLe^x), GSC269] and the sulphatides, 3-sulpho-galactose (3Su gal) and 6-sulpho-galactose (6Su gal), which were previously designated GSF-1 and GSF-19, respectively, linked to a 2-(tetradecyl)hexadecyl group (B30), were chemically synthesized.^33–36 The 3′-sialyl-Le^a and 3′-sulpho-Le^a pentasaccharides, abbreviated here to SLe^a penta and SuLe^a penta, respectively, were used as neoglycolipids, linked to DHPE.^31,37 Galactosylceramide (Gal cer) was from Sigma.

Lipid-linked sulpho-tyrosine

Sulpho-tyrosine linked to DHPE (Su tyr) was prepared according to the following principle: the carboxyl group of N-Fmoc-O-sulpho-l-tyrosine was reacted with the amino group of DHPE under peptide coupling conditions to give N-Fmoc-O-sulpho-l-tyrosinyl-DHPE. This intermediate was isolated by preparative TLC. Treatment with piperidine removed the Fmoc protective group and the resulting O-sulpho-l-tyrosinyl-DHPE was isolated by preparative TLC. The details are as follows.

A mixture of N-Fmoc-O-sulpho-l-tyrosine, 2-(1H- benzotriazole-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate and N-hydroxybenzotriazole (40 µmol each), and N,N-diisopropylethylamine (80 µmol), all in dimethylformamide (200 µl), was added to DHPE (10 µmol) in dichloromethane (2 ml). Dichloromethane (1 ml) and dimethylformamide (400 µl) were then added to clarify the solution. After incubation for 2 hr at room temperature, with occasional sonication, the reaction mixture was washed with water and the dichloromethane solution containing lipid was evaporated to dryness. The product was dissolved in piperidine (200 µl) and, after 5 min, chloroform (2 ml) and glacial acetic acid (400 µl) were added, and the mixture washed several times with water. After evaporation of solvent, the product was purified by TLC using chloroform–methanol–water (130:50:9, by vol), and quantified by TLC scanning densitometry using DHPE as the standard, as described previously for neoglycolipids.²⁶ The mass of the purified product, determined by liquid secondary ion mass spectrometry, was 906·6 Da, in accordance with the molecular formula shown in Fig. 1.

Figure 1.

Structure and molecular mass of O-sulpho-l-tyrosinyl-DHPE. (DHPE = l-1,2-dihexadecyl-sn-glycero-3-phosphoethanolamine)

Microwell binding assays

The binding of the E-, L- and P-selectins to lipid-linked compounds immobilized in microwells was assayed, as described previously,³¹ using galactosyl ceramide and cholesterol sulphate as negative controls. In brief, dilutions of the lipid-linked compounds dissolved in methanol, singly or as equimolar mixtures, were immobilized by drying in plastic microwells (Falcon 3912; Becton-Dickinson UK Ltd., Oxford, UK). After blocking, the selectin–IgM chimeras (50 µl per well of culture supernatants containing ≈1 µg/ml of selectin and 2 mm calcium) were added in the presence or absence of EDTA. The concentrations of the selectins in the culture supernatants were determined by comparison of the intensities of Coomassie Blue staining of bands with those of protein standards (from Sigma). Binding was detected using biotinylated F(ab′)₂ fragments of goat anti-human IgM, followed by a biotin–streptavidin–horseradish peroxidase system. L-selectin IgG chimera was complexed with biotinylated goat anti-IgG at a ratio of 1:3 (wt/wt), and binding was detected as described previously.³¹

Cell binding assays

Binding of P-selectin to HL60 cells was detected by flow cytometry. The HL60 cells (kindly provided by Dr Geoffrey Brown, University of Birmingham, Edgbaston, Birmingham, UK) were grown in RPMI medium containing 10% FCS, 100 IU/ml of penicillin and 100 µg/ml of streptomycin. For binding experiments, cells were suspended at 1·5 × 10⁶ cells/ml in flow cytometry buffer consisting of 10 mm phosphate-buffered saline (PBS), pH 7·4, containing 1% (vol/vol) FCS, 0·05% (wt/vol) sodium azide and 2 mm CaCl₂. Fifty microlitres of cell suspension and 25 µl of PBS were added to Falcon 2052 tubes and incubated at 4° for 1 hr with 100 µl of P-selectin culture supernatant. Cells were washed three times with flow cytometry buffer and incubated for 30 min with 100 µl of a 2-µg/ml solution of biotinylated F(ab′)₂ fragments of goat anti-human IgM diluted in flow cytometry buffer. Cells were washed and incubated for 30 min with 20 µl of streptavidin-conjugated phycoerythrin (Becton-Dickinson, UK Ltd.). The percentage of cells staining with P-selectin was determined with the Becton-Dickinson fluorescence-activated cell sorter (FACscan) using Cellquest software. For inhibition experiments, individual or equimolar mixtures of lipid-linked ligands were incorporated into liposomes (100–500 nm diameter) containing cholesterol: egg lecithin: lipid-linked ligands (or control compounds) at a molar ratio of 1:0·7:1.^31,38 Under the bath sonication conditions³⁹ the liposomes are predominantly unilamellar. Serial dilutions of the liposomes in PBS (25 µl) were added to the cells followed by P-selectin and 100 µl of culture supernatant, and the selectin binding was detected as described above. The percentage of inhibition of cell binding in the presence of inhibitors was determined as follows:

$$\begin{array}{c}\%\text{Inhibition of cell binding}=\left[\right(MF of cells without inhibitors -\; MF of cells with inhibitors)\\ ÷\text{MF of cells stained without inhibitors}]\times \mathrm{100,}\end{array}$$

where MF = median flourescence.

Results

Effect of EDTA on P- and L-selectin binding to the two classes of immobilized ligands

The binding of the P- and L-selectins to PSGL-1 is calcium dependent, the presence of both sialyl-Le^x and sulpho-tyrosine being required on the counter-receptor to elicit a significant binding signal.^{20,21,24,40,41} To examine features of the binding to sulpho-tyrosine in the absence of neighbouring sialyl-Le^x, we performed selectin-binding experiments using lipid-linked sulpho-tyrosine immobilized in microwells. P-selectin gave strong binding signals with the sulpho-tyrosine thus immobilized. Under the conditions of these experiments, binding was unaffected in the presence of EDTA as high as 20 mm (Fig. 2a, Fig. 3a). With L-selectin, however, binding to sulpho-tyrosine was impaired in the presence of EDTA when tested at the lower levels of the immobilized ligand. At higher ligand levels (200–300 pmol per well), there was less of an effect; in different experiments, the binding signals were 70–100% of those observed in the presence of calcium (Fig. 2b, Fig. 3b). Only when suboptimal concentrations of P-selectin were used was there a perceptible impairment of binding to sulpho-tyr in the presence of EDTA, analogous to that observed under the optimal assay condition with L-selectin (results not shown). This relatively calcium-independent binding contrasted with the well-established EDTA sensitivity of the binding of selectins to sialyl-Le^x and -Le^a sequences (Fig. 2g, Fig. 3).

Figure 2.

Binding of P- and L-selectins to immobilized lipid-linked ligands in the presence of calcium or EDTA. Different amounts of lipid-linked compounds were dried down in microwells and probed with immunoglobulin M (IgM) chimeras of P-selectin (panels a, c, e, g, i, k) and L-selectin (panels b, d, f, h, j, l) in culture medium containing 2 mm calcium, in the presence or absence of 20 mm EDTA. Binding of selectins was detected as described in the Materials and methods. Closed and open symbols represent experiments carried out with 2 mm calcium and 20 mm EDTA, respectively. Results are expressed as mean values in duplicate wells with the range indicated by error bars. SLe^a penta, sialyl-Le^a pentasaccharide; SLe^x tri, sialyl-Le^x trisaccharide; 3Su gal, 3-sulpho-galactose; SuLe^a penta, sulpho-Le^a pentasaccharide; SuLe^x tri, sulpho-Le^x trisaccharide; Su tyr, sulpho-tyrosine.

Figure 3.

Binding of P- and L-selectins to immobilized lipid-linked ligands in the presence of different concentrations of EDTA. Three hundred picomoles of the lipid-linked ligands were added to each well, and P- or L-selectin, in culture medium containing 2 mm calcium, was applied in the presence or absence of different concentrations of EDTA. The insets show, in each case, a separate data set using additional concentrations of added EDTA, closely spaced between 0 and 4 mm. Binding of selectins was detected as described in the Materials and methods. Results are expressed as the percentage of binding by the selectins in the absence of EDTA, with the range in duplicate wells indicated by error bars. SLe^a penta, sialyl-Le^a pentasaccharide; 3Su gal, 3-sulpho-galactose; Su tyr, sulpho-tyrosine.

We also examined the influence of EDTA on the binding of the two selectins to sulphated saccharides. The binding to sulpho-Le^a pentasaccharide was abolished in the presence of EDTA (Fig. 2k, 2l), in agreement with previous observations with this and the sulpho-Le^x analogue.³¹ In contrast, the binding curves of the trisaccharide analogue of sulpho-Le^x (Fig. 2e, 2f) shared features with that of the sulphatides 3-sulpho-galactose and 6-sulpho-galactose (results with the former are shown here in Figs 2c, 2d and Fig. 3). With L-selectin, there was a marked impairment of binding in the presence of EDTA at all concentrations of immobilized ligand tested, whereas with P-selectin, binding was impaired only at the lower concentrations of immobilized ligand, at ≤30 pmol added per well. The binding of the P- and L-selectins to the short sulphated compounds in the presence of EDTA was not a non-specific type of interaction as, under the same assay conditions, no E-selectin binding signals were elicited with these compounds (results not shown).

The effects of different concentrations of EDTA on P- and L-selectin binding were studied with the two groups of ligand using the sialyl-Le^a pentasaccharide as representative of the calcium-dependent ligands, and 3-sulpho-galactose and sulpho-tyrosine as representative of the relatively calcium-independent ligands. Maximal inhibitory effects of EDTA were manifest at a concentration of 4 mm; the binding signals decreased steeply between EDTA concentrations of 2 and 4 mm (Fig. 3). This corresponds to a calculated concentration of free calcium of ≤5 nm.⁴² Thus, with 4 mm EDTA present in the culture medium, and at a ligand loading level of 300 pmol per well, the binding to sialyl-Le^a was abolished with both selectins, and the binding to 3-sulpho-galactose by L-selectin was diminished to ≈30% of the maximum binding. Under these conditions, the binding of P-selectin to both 3-sulpho-galactose and to sulpho-tyrosine, and of L-selectin to sulpho-tyrosine, was unaffected.

In separate experiments (results not shown) we have ruled out the possibility that the diminished binding signals are caused by leaching of a proportion of the immobilized ligands in the presence of EDTA. The initial washing of the ligand-coated wells with 20 mm EDTA, prior to the binding assays, had no perceptible effect on the binding signals with the P- and L-selectins. Thus, the binding of sulphatide by the two selectins, and of sulpho-tyrosine by L-selectin, is impaired by the chelation of calcium by EDTA. Overall, the effects of the chelation by EDTA are more pronounced with L-selectin than P-selectin.

Co-operativity in the binding signals elicited by the two classes of ligand

We reasoned that if the binding sites (on the selectins) for the two classes of ligand are structurally and operationally distinct, there would be an increased avidity and amplification of the binding signals elicited when the two ligand types are presented as clustered mixtures. If, on the other hand, the two classes of ligand occupy a single site, a mixture of the two ligand types would elicit binding that is no more than a sum of the binding to the compounds immobilized. When an equimolar mixture of the long fuco-oligosaccharide ligands sialyl-Le^a and sulpho-Le^a pentasaccharides were immobilized, the binding signals were those expected for the sum of the amounts of the individual components (Fig. 4a, 4d). Similarly, there was no amplification of binding signals with P-selectin when the short sulphated ligands 3-sulpho-galactose and sulpho-tyrosine were immobilized as equimolar mixtures (results not shown). In contrast, when the immobilized ligands were equimolar mixtures of sialyl-Le^x (or sialyl-Le^a) and a short sulpho-ligand (sulpho-tyrosine or 3- or 6-sulpho-galactose), there was an amplification of the binding signals with both P- and L-selectins. Results of representative experiments are presented in Fig. 4, where the x-axis represents the sum of concentrations of the lipid-linked compounds.

Figure 4.

Binding of P- and L-selectins to the two classes of ligands linked to lipid, and immobilized individually or as mixtures. In (a), (b), (d) and (e), different amounts of lipid-linked compounds were added to microwells, individually and as mixtures, and probed with immunoglobulin M (IgM) chimeras of P-selectin (panels a and b) and L-selectin (panels d and e) in culture medium containing 2 mm calcium, as described in the Materials and methods. In panels (c) and (f) the binding of IgM chimeras of P- and L-selectin, respectively, are shown in the presence of 2 mm calcium or 20 mm EDTA: the lipid-linked ligands were immobilized individually or as 1 : 1 mixtures; the total amount of ligand in each well was 30 pmol. Results are expressed as means of values in duplicate wells with the range indicated by error bars. Gal cer, galactose; SLe^a penta, sialyl-Le^a pentasaccharide; SLe^x tri, sialyl-Le^x trisaccharide; Su chol, cholesterol sulphate; 3Su gal, 3-sulpho-galactose; 6Su gal, 6-sulpho-galactose; 6Su-SLe^x, 6-sulpho-sialyl-Le^x pentasaccharide; SuLe^a penta, sulpho-Le^a pentasaccharide; Su tyr, sulpho-tyrosine.

We also examined, with both P-and L-selectins, the effect of mixing 6-sulpho-sialyl-Le^x with 3-sulpho-galactose. Here also, there was amplification of binding signals. Results with L-selectin are shown in Fig. 4(e) (inset). The amplification of binding signals was not confined to the two selectin–IgM chimeras, but also occurred with the L-selectin–IgG chimera when tested with equimolar mixtures of sialyl-Le^a and 3-sulpho-galactose (results not shown).

We evaluated the requirement of calcium in the synergy between the two types of ligand. With all of the compounds tested, the amplification of the binding signals with both P- and L-selectins was calcium dependent, as the signals in the presence of EDTA corresponded to those observed with the singly immobilized ligands. Representative results are shown in Fig. 4(c), 4(f). With P-selectin, the binding signal elicited by the sulpho-tyrosine and sialyl-Le^x mixture, in the presence of EDTA, corresponded to that predicted for the 15 pmol of sulpho-tyrosine included in the mixture. With L-selectin, the mixture elicited no binding signals in the presence of EDTA, in agreement with the binding data of the individual ligands at the levels tested.

By displaying the two types of lipid-linked ligands on liposomes, we examined the potential for exploiting their co-operativity in the design of selectin inhibitors. In assays of the inhibition of P-selectin binding to PSGL-1-expressing HL60 cells, the co-operativity in the interactions of the sulpho-tyrosine and sialyl-Le^x was clearly manifest (Fig. 5). In repeated experiments, sialyl-Le^x liposomes gave little or no inhibition of the binding of P-selectin to the cells. Sulpho-tyrosine liposomes had, at best, a 50% inhibitory concentration (IC₅₀) value of 3 × 10⁻⁶ m. However, liposomes containing a 1:1 mixture of sulpho-tyrosine and sialyl-Le^x trisaccharide had IC₅₀ values ranging from 3 × 10⁸ to 6 × 10⁻⁸ m, in different experiments, and were thus at least 100 times more effective as inhibitors of the P-selectin binding than the liposomes containing the components separately.

Figure 5.

Inhibition of the binding of P-selectin to HL60 cells by liposomes, displaying lipid-linked ligands individually or as equimolar mixtures. The binding of P-selectin to HL60 cells was assayed by flow cytometry in the presence of phosphate-buffered saline (PBS) (positive control) or liposomes in PBS incorporating various concentrations of compounds, singly or as equimolar mixtures as described in the Materials and methods. The flow cytometric data are presented as percentage inhibition of binding (median fluorescence values) relative to the positive control. Results are expressed as means of values obtained from duplicate tubes. Gal cer, galactose; SLe^x tri, sialyl-Le^x trisaccharide; Su tyr, sulpho-tyrosine.

Discussion

The availability of a lipid-linked sulpho-tyrosine probe has allowed us to compare the interactions of sulpho-tyrosine, the sulphatides and fuco-oligosaccharide ligands with the P- and L-selectins. The salient conclusions are as follows:

1. The lipid-linked sulpho-tyrosine, in the absence of fuco-oligosaccharides, is bound by P- and L-selectins and there are similarities in the interactions of sulpho-tyrosine and sulphatides with the two selectins in that the binding is relatively resistant to chelation of calcium ions by EDTA. This contrasts with the interactions of long fuco-oligosaccharides, which are calcium dependent.

2. When the sulpho-tyrosine (or a sulphatide) is presented as a mixture with sialyl-Le^x, sialyl-Le^a or 6-sulpho-sialyl-Le^x sequences, there is a striking amplification of the binding signals with P- as well as L-selectin.

3. The amplification is calcium dependent and is the result of separate contributions of the two ligand types.

While our investigations were in progress, the crystal structure was published of the recombinant lectin and adjoining EGF domains of the E- and P-selectins complexed with sialyl-Le^x tetrasaccharide and, in addition, of P-selectin with the N-terminal sulpho-glycopeptide domain of PSGL-1.²⁵ In the structure, the binding of P-selectin (and also of E-selectin) to the sialyl-Le^x oligosaccharide occurs through 3-OH and 4-OH groups of the fucose via co-ordination bonds with a calcium ion at an area of the protein characterized by neutral and negative electrostatic potential, whereas the binding of P-selectin to the sulpho-tyrosine-containing polypeptide moiety is at an area of positive electrostatic potential, and involves hydrophobic contacts.²⁵ Our experimental data are in accord with the two binding regions in the P-selectin structure: one for oligosaccharide and the other for sulpho-tyrosine. Structural data are not yet available on L-selectin. The striking similarities we have observed in the binding patterns of L-selectin and P-selectin with the various ligand analogues, including 6-sulpho-sialyl-Le^x and sulpho-tyrosine, indicate that there are two ligand-binding regions on L-selectin also. The possibility that there may exist two binding sites on L-selectin, one for sialyl-Le^x and the other for sulphatides, was raised previously.²⁹ The co-operativity that we observed between the 6-sulpho-sialyl-Le^x and 3-sulpho-gal indicates that the sulphate groups in the two ligand analogues are bound at different sites on the selectins.

The molecular basis of the impairment of P-selectin (and also L-selectin) binding to short sulpho-ligands in the presence of EDTA is not yet known. It is possible that this is a consequence of general conformational changes at the lectin surface of the type documented in the crystal structures of the prototype C-type lectins, mannose-binding proteins, in the absence of calcium.^43,44 The binding curves with the short sulpho-ligands in the presence and absence of EDTA suggest that P-selectin has a higher affinity than L-selectin for these short sulpho-ligands, and that the affinities of the two selectins for sulpho-tyrosine are greater than for the sulphatides. The impaired L-selectin binding to short sulpho-ligands in the presence of EDTA has been observed also with a series of lipid-linked sulphated derivatives of lactose.⁴²

One of the short sulpho-motifs investigated in the present study is the 6-sulpho-galactose. This motif can act synergistically with sialyl-Le^x in eliciting binding signals with the L-selectin (also with P-selectin). The synergism is calcium dependent (Fig. 4f). This finding offers a possible explanation for the results obtained from dual transfection of a galactose:6′-sulphotransferase and an N-acetylglucosamine:6-sulphotransferase into Chinese hamster ovary cells.⁴⁵ These authors observed greater L-selectin binding to the cells transfected with both enzymes than to those transfected with the individual enzymes, and the binding was calcium dependent. The N-acetylglucosamine:6-sulphotransferases are known to be involved in the generation of 6-sulpho-sialyl-Le^x,⁴⁶ the potent L-selectin ligand.¹⁶ We are unaware as to whether the products of galactose:6′-sulphotransferase are involved in the generation of any 6′-sulpho-sialyl-Le^x. The cellular products of this enzyme are not yet known. Based on observations with the chemically synthesized 6′-sulpho-sialyl-Le^x pentasaccharide,¹⁶ we predict that any 6′-sulpho-sialyl-Le^x that might be generated would not elicit binding signals with L-selectin. Rather, our present findings are in accord with the possibility, raised previously,¹³ that the galactose:6′-sulphotransferase might generate short sulpho-ligands on the glycoprotein counter-receptors or on other membrane components, and that these could act synergistically with the fuco-oligosaccharide ligands. This view is supported by results of recent in vitro experiments,⁴⁷ in which chemically synthesized 3′-sialyl-neolactotetraose and three of its sulphated analogues were tested as substrates for the fucosyltransferase, Fuc-TVII, which is a key enzyme in the formation of the acidic Le^x-based fuco-oligosaccharide ligands for the selectins.⁴⁸ The sulphated analogues examined were the 6′-sulphated, the 6-sulphated, or the 6,6′-disulphated forms of 3′-sialyl-neolactotetraose. The results indicate that only the 6-sulphated analogue can serve as an acceptor for fucosyltransferase.

The approach in the present study, with the two ligating elements (sialyl-Le^x and sulpho-tyrosine) linked to lipid and displayed on liposomes, demonstrates that it is possible on the fluid membrane to eliminate restrictions of the type imposed by the polypeptide moiety of PSGL-1, and generate substantial binding signals. There is a precedent for application of the liposome approach for generating bifunctional high-avidity ligand analogues for the P- and L-selectins.^49,50 The simple liposome technology we used lends itself to the rapid screening of ligand analogues as selectin inhibitors in the generation of ‘designer’ substances for medical use in anti-inflammatory therapeutics. The lipid-linked ligand approach may also help to evaluate possible dual ligands, saccharide- and peptide-borne, for other lectin-like receptors. Such dual ligands have been considered a possibility on lectin-like receptors of lymphocytes, but unequivocal binding data are awaited.^51–53 The neoglycolipid approach would be applicable, particularly where the potential oligosaccharide ligands, for example those associated with specific major histocompatibility complex (MHC) molecules, may be available only in limited amounts.

Glossary

Abbreviations

DHPE

l-1,2-dihexadecyl-sn-glycero-3-phosphoethanolamine

Le^a

Lewis^a

Le^x

Lewis^x

MF

median fluorescence