Binding of Glycosulfopeptides to P-selectin Requires Stereospecific Contributions of Individual Tyrosine Sulfate and Sugar Residues*

P-selectin glycoprotein ligand-1 (PSGL-1) is a mucin on leukocytes that binds to selectins. P-selectin binds to an N-terminal region of PSGL-1 that requires sulfation of at least one of three clustered tyrosines (TyrSO3) and an adjacent core-2-based O-glycan expressing sialyl Lewis x (C2-O-sLex). We synthesized glycosulfopeptides (GSPs) modeled after this region of PSGL-1 to explore the roles of individual TyrSO3 residues, the placement of C2-O-sLexrelative to TyrSO3, the relative contributions of fucose and sialic acid on C2-O-sLex, and the function of the peptide sequence for binding to P-selectin. Binding of GSPs to P-selectin was measured by affinity chromatography and equilibrium gel filtration. 2-GSP-6, which has C2-O-sLex at Thr-57 and TyrSO3 at residues 46, 48, and 51, bound to P-selectin with high affinity (K d ∼ 650 nm), whereas an isomeric trisulfated GSP containing C2-O-sLex at Thr-44 bound much less well. Non-sulfated glycopeptide (2-GP-6) containing C2-O-sLex at Thr-57 bound to P-selectin with ∼40-fold lower affinity (K d ∼25 μm). Proteolysis of 2-GP-6 abolished detectable binding of the residual C2-O-sLex-Thr to P-selectin, demonstrating that the peptide backbone contributes to binding. Monosulfated and disulfated GSPs bound significantly better than non-sulfated 2-GP-6, but sulfation of Tyr-48 enhanced affinity (K d ∼ 6 μm) more than sulfation of Tyr-46 or Tyr-51. 2-GSP-6 lacking sialic acid bound to P-selectin at ∼10% that of the level of the parent 2-GSP-6, whereas 2-GSP-6 lacking fucose did not detectably bind; thus, fucose contributes more than sialic acid to binding. Reducing NaCl from 150 to 50 mm markedly enhanced binding of 2-GSP-6 to P-selectin (K d ∼ 75 nm), demonstrating the charge dependence of the interaction. These results reveal a stereospecific interaction of P-selectin with PSGL-1 that includes distinct contributions of each of the three TyrSO3residues, adjacent peptide determinants, and fucose/sialic acid on an optimally positioned core-2 O-glycan.

The N terminus of mature human PSGL-1 begins at residue 42, after removal of the signal peptide from residues 1-18 and a propeptide from residues 19 -41 (11). Sulfation of tyrosine residues and O-glycosylation in the mature N-terminal region of PSGL-1 appear necessary for high affinity binding of PSGL-1 to P-selectin (4), but the importance of modifications at specific amino acid residues is not well understood. Studies employing site-directed mutagenesis of recombinant PSGL-1, which was co-expressed in Chinese hamster ovary cells or COS cells with specific glycosyltransferases, suggested that a core-2 based Oglycan with the sialyl Lewis x (C2-O-sLe x ) antigen at Thr-57 and at least one of the Tyr residues at Tyr-46, -48 or -51 are required for measurable binding of PSGL-1 to P-selectin (12)(13)(14). Whether the individual Tyr residues make distinct contributions to P-selectin recognition has not been examined. Mutation of Thr-44 to alanine did not inhibit binding of recombinant PSGL-1, suggesting that an O-glycan at this position does not contribute to binding. However, this residue might be O-glycosylated in hematopoietic cells but not in transfected Chinese hamster ovary cells.
Very little is known about the stereospecific contributions of tyrosine sulfation and O-glycosylation to binding of PSGL-1 to P-selectin. Substitution of any two of the three tyrosines with phenylalanines in recombinant PSGL-1 impairs the mechanical properties of its bonds with P-selectin in shear flow, but these substitutions could affect peptide structure as well as prevent sulfation (18). Synthetic conjugates that express sLe x on different types of O-glycans bind differentially to P-selectin (16,19). Notably, a GSP modeled after the N terminus of PSGL-1 that contains sLe x expressed on a core-2 based Oglycan binds much better to P-selectin than an isomeric GSP that contains sLe x expressed on an extended core-1 based Oglycan (16). It is not clear whether the position of the O-glycan relative to TyrSO 3 residues, i.e. at Thr-44 versus Thr-57, contributes to binding.
The interpretation of the effects of amino acid substitutions on post-translational modifications of PSGL-1 has significant limitations because it is difficult to structurally define specific post-translational modifications on recombinant glycoproteins. There is usually microheterogeneity of glycosylation that complicates the interpretation of binding affinities. The use of synthetic, homogeneously glycosylated glycosulfopeptides offers marked advantages, but the tyrosyl-protein sulfotransferase adds sulfate from PAPS donor to all available Tyr residues within the consensus sequence. Thus, enzymatic sulfation does not readily allow sulfation of individual Tyr residues within a peptide that contains multiple Tyr residues (16).
To gain more insight into the contributions of individual TyrSO 3 residues, the peptide backbone, specific sugar residues, and their site-specific contributions, we generated synthetic GSPs modeled after the mature N terminus of human PSGL-1 that contain TyrSO 3 residues and O-glycans at specific sites. Our results demonstrate that P-selectin differentially recognizes GSPs containing one, two, or three TyrSO 3 residues and that the relative contributions of each TyrSO 3 residue differ. P-selectin binds weakly to non-sulfated anionic peptides containing C2-O-sLe x , but the O-glycan must be placed at Thr-57 rather than at the alternative site at Thr-44. These studies further extend our knowledge about the roles of peptide, sulfate, and carbohydrate determinants in binding of PSGL-1 to P-selectin.
Enzymatic Sialyl Lewis x on a core-2-based O-glycan was synthesized at Thr-57 in each peptide using recombinant or highly purified glycosyltransferases essentially as described (16). The reactions were accomplished by adding one glycosyltransferase and one donor at a time; after the reaction was Ͼ95% complete a new glycosyltransferase and a donor were added. At each step of the synthesis, a small aliquot of the reaction mixture was analyzed by HPLC. The completeness of each reaction was easily monitored by HPLC, because the addition of each monosaccharide reduced the retention time for the peptides (data not shown). At the end of the reaction sequence the reaction mixtures were deproteinated by chloroform-methanol (2:1) extraction, and samples were further purified by HPLC. The masses of the final products (2-GSP-6 series) and of peptide products after sialyltransferase reaction (2-GSP-5 series) were verified by electrospray mass spectrometry (see Table I). The sialyltransferase products served as acceptors for the synthesis of radiolabeled peptides. Enzymatic Synthesis of Isomeric Glycosulfopeptides 3-GSP-6 and 3-GSP-6Ј-Isomeric glycopeptides 3-GP-6 and 3-GP-6Ј were synthesized as described (16). Matrix-assisted laser desorption ionization/time of flight mass spectrometric analysis confirmed the masses of the HPLC-purified products (observed m/z 3637.3 for 3-GP-6 and observed m/z 3636.6 for 3-GP-6Ј; calculated m/z 3637.7 for both peptides). Glycopeptides 3-GP-6 and 3-GP-6Ј were sulfated enzymatically using human tyrosyl-protein sulfotransferase-1 and either [ 35 S]PAPS (544 000 cpm/ nmol) (NEN Life Science Products, Inc.) or nonlabeled PAPS (Sigma) as a sulfate donor as described (16). The fully sulfated products 3-GSP-6 and 3-GSP-6Ј were purified using HPLC, and the masses of the final products were verified using electrospray mass spectrometry (Table I).
Equilibrium Gel Filtration Chromatography-Hummel-Dreyer equilibrium gel filtration experiments were conducted in 2 ml of Sephadex G-100 columns (0.5 ϫ 10 cm) as described (16). 3 H-Labeled peptides (specific activity 1000 cpm/pmol) were used in the buffer at a concentration of 8000 cpm/ml. Two different concentrations of NaCl were used: 150 mM or 50 mM NaCl in 20 mM MOPS, pH 7.5, containing 2 mM CaCl 2 , 2 mM MgCl 2 , and 0.02% NaN 3 . Inhibition experiments with EDTA were conducted using 1 mM EDTA with 150 mM or 50 mM NaCl in 20 mM MOPS, pH 7.5, containing 0.02% NaN 3 . The amount of soluble Pselectin (sPS) (21) used was 500 pmol in high salt buffer or 50 pmol in low salt buffer. Inhibition experiments with anti-P-selectin monoclonal antibody G1 (22) were performed using the same molar amount of G1 and sPS (500 pmol or 50 pmol) in buffers containing Ca 2ϩ .
P-selectin Affinity Chromatography-sPS was coupled to Ultralink Biosupport Medium (Pierce) at a density of 6.5 mg/ml. An sPS column (0.9 ml, 0.5 ϫ 4.5 cm) was equilibrated with 20 ml of 20 mM MOPS, pH 7.5, containing 150 mM NaCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 0.02% NaN 3 . Radiolabeled peptides dissolved in 200 l of equilibration buffer (1000 -2000 cpm, 1-10 pmol) were chromatographed in the P-selectin column by collecting 0.5-ml fractions at a flow rate of 250 l/min. Bound peptides were eluted using 10 mM EDTA instead of divalent cations in buffer. In some experiments, 50 mM NaCl was used instead of 150 mM NaCl in buffers containing Ca 2ϩ or EDTA (1 mM). To derive the ⌬ value for each bound peptide, the elution volume of unbound peptide (2 ml) was subtracted from the elution volume of bound peptide.
Mass Spectrometric Analysis of Glyco(sulfo)peptides-Electrospray mass spectra were collected in the negative ion mode using an API365 triple quadrupole mass spectrometer (PerkinElmer Sciex Instruments, Thornhill, Ontario, Canada). Samples were dissolved in 1% triethylamine in 50% aqueous methanol to a concentration of 5 pmol/l and injected into the mass spectrometer with a nanoelectrospray ion source (MDS Protana A/S, Odense M, Denmark).

Synthesis of Glyco(sulfo)peptides-
We explored the possibility that only one or two TyrSO 3 residues together with a nearby O-glycan can support high affinity binding to P-selectin. To this end we synthesized a series of glyco(sulfo)peptides corresponding to the extreme N terminus of human PSGL-1 (residues 45-61) with one, two, or three TyrSO 3 residues at defined positions or without TyrSO 3 residues (Fig. 1). Each glyco(sulfo)peptide contained a C-terminal Cys residue for future coupling of the peptide to artificial supports. Each glyco(sulfo)peptide was subsequently modified by purified or recombinant glycosyltransferases to generate a core-2-based O-glycan expressing the sLe x antigen at Thr-57 (Fig. 2). To compare the roles of sialic acid and fucose for binding to P-selectin, peptides with three TyrSO 3 residues but having incomplete glycosylation were also synthesized ( Fig. 1).
Peptides with one, two, or three TyrSO 3 residues (Tyr-46, -48, and/or -51) and an ␣-linked GalNAc residue at Thr-57 were synthesized chemically using Fmoc derivatives of TyrSO 3 and tri-O-acetyl-GalNAc␣-Thr during the solid phase peptide synthesis. The sulfated peptides were cleaved from the solid support under mild conditions because of the acid lability of TyrSO 3 residues (20). After peptide cleavage, the C-terminal Cys of each peptide was protected by converting the free sulfhydryl into S-S-CH 3 to prevent oxidation and dimerization of the peptide. After deacetylation of tri-O-acetyl-GalNAc, the peptides were purified by reversed phase HPLC, characterized by mass spectrometry, and used as acceptors for enzymatic glycosylation. sLe x on a core-2-based O-glycan was synthesized on each peptide at Thr-57 using recombinant or highly purified glycosyltransferases (Fig. 2). The completeness of each glycosyltransferase reaction was easily monitored by HPLC, because the addition of each monosaccharide reduced the retention time for the peptides by ϳ0.5-2 min (data not shown). Electrospray mass spectrometry was used to analyze the masses of the final products. The mass of each glyco(sulfo)peptide matched the calculated mass (Table I). Radiolabeled peptides were synthesized using either GDP-[ 3 H]Fuc or GDP-[ 14 C]Fuc in the reactions with ␣1,3-fucosyltransferase, which is the final step in the synthesis. In addition, radiolabeled glycosulfopeptides with three TyrSO 3 residues but with incomplete O-glycan structure were synthesized. These glycosulfopeptides are 2-GSP-2 (core-1 O-glycan), desialylated 2-GSP-6, and 2-GSP-5, which lacks a fucose residue (Fig. 1).
We previously showed that a core-2-based O-glycan containing sLe x at Thr-57 together with three TyrSO 3 residues supported high affinity binding of a glycosulfopeptide (GSP-6) to P-selectin (16). However, the extreme N terminus of PSGL-1 contains another potential glycosylation site at Thr-44. To explore the possibility that an isomeric glycosulfopeptide with sLe x O-glycan at Thr-44 together with three TyrSO 3 residues may bind to P-selectin, we synthesized two isomeric glycosulfopeptides corresponding to residues 43-61 of the extreme N terminus of human PSGL-1; these peptides contained three TyrSO 3 residues and a sLe x O-glycan at either Thr-57 (3-GSP-6) or Thr-44 (3-GSP-6Ј) (Fig. 1). O-Glycans were synthesized enzymatically as shown in Fig. 2, starting from the nonsulfated peptides. As a final step, all three Tyr residues of 3-GP-6 and 3-GP-6Ј were enzymatically sulfated using recombinant human tyrosyl-protein sulfotransferase-1. Electrospray mass spectrometer verified the masses of the fully sulfated glycosulfopeptides, 3-GSP-6 and 3-GSP-6Ј (Table I).
A Glycosulfopeptide with C2-O-sLe x at Thr-57 but Not at Thr-44 Binds to P-selectin-Two isomeric 35 SO 3 -labeled glycosulfopeptides containing three TyrSO 3 residues and a C2-O-sLe x at either Thr-57 (3-GSP-6) or Thr-44 (3-GSP-6Ј) were chromatographed on an affinity column containing sPS at a coupling density of 6.5 mg/ml (Fig. 3). 3-GSP-6 bound to immobilized sPS and was eluted with 10 mM EDTA, whereas 3-GSP-6Ј did not detectably bind to P-selectin. This result demonstrates that a core-2-based sLe x O-glycan at Thr-57, but not at Thr-44, confers high affinity binding of the GSP to P-selectin. Subsequent experiments were carried out with shorter peptides where the O-glycan was present at Thr-57 and the two N-terminal amino acids Ala-43 and Thr-44 were deleted.
Each Additional TyrSO 3 Residue Enhances the Binding Af-  finity of GSPs for Immobilized P-selectin-As an initial test to determine whether incompletely sulfated GSPs containing an intact sLe x O-glycan interact with P-selectin, [ 3 H]Fuc-labeled GSPs were chromatographed on a sPS affinity column (Fig. 4). At physiological salt concentration (150 mM NaCl) nonsulfated peptide 2-GP-6 had only a slight affinity for immobilized sPS (Fig. 4C, ⌬ ϭ 0.2; the ⌬ reflects the difference in elution volume between the test sample and that of a control peptide, with no affinity for P-selectin or the test peptide analyzed in the presence of 1 mM EDTA). The monosulfated peptide 2-GSP(48)-6 had higher affinity for immobilized P-selectin (⌬ ϭ 1.2) than the other isomeric peptides 2-GSP(46)-6 and 2-GSP(51)-6 (⌬ ϭ 0.7) (Fig. 4A). Disulfated peptides were more retarded in their elution from the sPS column than monosulfated peptides (Fig.  4B). Two of the disulfated peptides (2-GSP(46,48)-6, ⌬ ϭ 4.3; 2-GSP(48,51)-6, ⌬ ϭ 3.7) had higher affinity for P-selectin than 2-GSP(46,51)-6 (⌬ ϭ 2.2). This indicates that Tyr(48)-SO 3 is more important for binding of disulfated as well as monosulfated peptides to P-selectin. Trisulfated 2-GSP-6 had the highest affinity for immobilized P-selectin and was eluted with EDTA ( Fig. 4C). Because of the reversible nature of the binding of the ligands to immobilized P-selectin, bound ligands eventually eluted from the column without the necessity of chelating Ca 2ϩ . When chromatography was continued without the addition of EDTA, 2-GSP-6 eluted as a very symmetrical broad peak beginning at fraction 20 (⌬ ϭ 14, the inset in Fig. 4C). In contrast, replacement of divalent cations in the buffer with 1 mM EDTA completely inhibited the binding of 2-GSP-6 to Pselectin (⌬ ϭ 0, dashed line in Fig. 4), demonstrating that binding of 2-GSP-6 to P-selectin is dependent on Ca 2ϩ . These results demonstrate that each additional TyrSO 3 residue enhances the binding affinity of glycosulfopeptides to P-selectin and that Tyr-48 is the most optimal tyrosine sulfate position for binding. ␣1,3-Fucose Contributes More than ␣2,3-Sialic Acid to Bind-ing of 2-GSP-6 to P-selectin-The binding of trisulfated but incompletely glycosylated GSPs to immobilized P-selectin was also studied (Fig. 4, D-F). These peptides were desialylated 2-GSP-6, 2-GSP-5 (no fucose), and 2-GSP-2 (core-1 O-glycan only). Desialylated 2-GSP-6 showed a weak affinity for sPS (⌬ ϭ 0.3, Fig. 4D), whereas the other two GSPs did not detectably bind. These results indicate that ␣1,3-fucose in the absence of sialic acid contributes to binding, whereas ␣2,3-sialic acid in the absence of fucosylation does not detectably contribute to binding. Three TyrSO 3 Residues Increase the Binding Affinity of 2-GSP-6 for P-selectin by 40-Fold-Hummel-Dreyer equilibrium gel filtration (23,24) was used to directly measure the dissociation constants (K d ) for binding of GSPs to soluble Pselectin. We examined those mono-, di-, and trisulfated glycosulfopeptides that clearly interacted with immobilized P-selectin (i.e. ⌬ Ͼ 1). Experiments were conducted using a constant amount of [ 3 H]Fuc-labeled sample in Ca 2ϩ -containing buffer and different concentrations of sPS. The K d was derived from the binding data as shown for 2-GSP-6 ( Fig. 5A) and 2-GSP(46,48)-6 ( Fig. 5C). In typical equilibrium gel filtration experiments (23,24), a plateau is usually observed between the peak elution position of a large-sized complex of protein-bound ligand and the trough elution of the small-sized free ligand. However, we did not observe a significant plateau between the peak and trough in equilibrium binding studies with sPS and the GSPs (Fig. 5, A and C). This was due to the close elution positions of the sPS versus the GSPs (Fig. 5A, inset), thus resulting in an absence of a significant plateau. This absence of a plateau does not materially affect the results, since the elution of the complex of sPS with GSP is clearly distinguishable from the elution of the free GSPs. 2-GSP-6 bound to sPS with a K d of ϳ650 nM (Fig. 5B). The binding was Ca 2ϩ -dependent, since it was blocked by inclusion of EDTA, and specific to the lectin domain of P-selectin, since it was inhibited by the monoclonal antibody G1, which recognizes the lectin domain of Pselectin (not shown). Scatchard analysis of the binding data with the assumption that each molecule of sPS binds a single molecule of 2-GSP-6, indicated that at saturation, 2-GSP-6 was quantitatively complexed with sPS and the number of binding sites is n ϭ 1.2 (data not shown). This assumption is consistent with previous studies demonstrating that sPS is a soluble, monomeric protein (21). K d values of 1.5 and 1.9 M were determined for the disulfated peptides 2-GSP(46,48)-6 and 2-GSP(48,51)-6, respectively, which showed the highest affinity for immobilized sPS relative to 2-GSP-6 ( Fig. 5, C-E). The disulfated 2-GSP(46,51)-6 bound to P-selectin with a K d of ϳ5.9 M (Fig. 5F). The binding affinity for 2-GSP(48)-6, which exhibited the highest affinity for immobilized sPS among the monosulfated peptides, was approximately the same as for 2-GSP(46,51)-6 (K d ϳ 6.1 M, Fig. 5G). The dissociation constant for the nonsulfated peptide 2-GP-6 was estimated by Hummel-Dreyer equilibrium gel filtration using a single concentration of sPS (3.2 nmol). The concentration of bound 2-GP-6 (2 nM) and free sPS (7.6 M) were compared with the equilibrium gel filtration data for 2-GSP-6, 2-GSP(46,48)-6, and 2-GSP(48,51)-6 ( Fig. 5, A-E), yielding an estimated K d of ϳ20 -30 M for 2-GP-6. These results show that the three TyrSO 3 residues increase the binding affinity of 2-GSP-6 for P-selectin by 40-fold over the non-sulfated glycopeptide.
A direct linear relationship was found between elution positions (⌬) of GSPs on immobilized P-selectin and the association constants (K a or 1/K d ) derived from equilibrium gel filtration data (Fig. 5H). This direct relationship demonstrates that the elution position on affinity chromatography is a direct reflection of the equilibrium binding in solution and shows that the immobilized P-selectin column can be used to estimate dissociation constants by calibrating the column with GSPs having known dissociation constants.
Sub-physiological Concentration of NaCl Enhances the Binding Affinity of 2-GSP-6 for P-selectin-The binding of selectins to sLe x is inhibited by increasing the concentration of NaCl (25). To study the effect of sub-physiological salt concentration on the binding affinity of 2-GSP-6 for sPS, we determined the K d for binding of 2-GSP-6 to sPS at 50 mM NaCl concentration (Fig. 6A). The K d was significantly lowered to ϳ76 nM, which is nearly a 9-fold higher affinity than that observed at 150 mM NaCl (Fig. 5B). The binding of 2-GSP-6 to sPS at 50 mM NaCl was completely inhibited by EDTA and by monoclonal antibody G1, demonstrating that binding was Ca 2ϩ -dependent and involved the lectin domain of P-selectin (Fig. 6B).
Monosulfated GSPs as Well as Nonsulfated 2-GP-6 Bind to Immobilized P-selectin in a Ca 2ϩ -dependent Manner at Subphysiological Salt Concentration-We further explored the binding of GSPs to immobilized P-selectin using a sub-physiological salt concentration to aid in comparing the affinities of relatively weak ligands. When affinity chromatography on immobilized P-selectin was performed in Ca 2ϩ -containing buffer with 50 mM NaCl, all isomeric monosulfated glycosulfopeptides were more retarded in their elution on the column compared with conditions employing 150 mM NaCl (Fig. 7A). 2-GSP(48)-6 clearly bound with higher affinity to immobilized P-selectin than the two other isomeric GSPs. These results are consistent with the results obtained using 150 mM NaCl, as described above in Fig. 4A. Binding of monosulfated peptides to P-selectin was Ca 2ϩ -dependent, since replacement of divalent cations with 1 mM EDTA in the elution buffer completely inhibited the binding (Fig. 7B). Monosulfated GSPs had higher affinity for immobilized P-selectin in comparison to non-sulfated 2-GP-6 under low salt conditions. However, 2-GP-6 was detectably retarded in its elution under low salt conditions, and the binding was inhibited with 1 mM EDTA (Fig. 7C). These data confirm the results obtained at 150 mM NaCl, which indicated that monosulfated GSPs bind to immobilized P-selectin with higher affinity than nonsulfated 2-GP-6 and that Tyr-48 is the most optimal sulfate position for binding.
The Nonsulfated Anionic Peptide Backbone Together with C2-O-sLe x Contributes to Binding to P-selectin-Since 2-GP-6, which lacks TyrSO 3 residues, bound weakly to immobilized P-selectin, we tested whether the peptide itself contributed to this interaction. Proteolytic cleavage of 2-GP-6 with the broadly specific protease Pronase abolished binding of the residual [ 3 H]Fuc-labeled C2-O-sLe x -Thr (Fig. 7D). This result demonstrates that the nonsulfated anionic peptide backbone together with a C2-O-sLe x at Thr-57 is important for binding to P-selectin and that the C2-O-sLe x is not sufficient for detectable binding under these conditions. DISCUSSION A set of glycosulfopeptides based on the primary sequence of the N-terminal region of human PSGL-1 were synthesized to study the effect of site-specific tyrosine sulfation as well as the structures of O-glycans and their peptide locations for binding to P-selectin (Fig. 1). The results indicate that each TyrSO 3 residue contributes to the binding affinity of glycosulfopeptides for P-selectin, although their contributions are not equivalent (Fig. 8). For example, the disulfated 2-GSP(46,51)-6 binds only slightly better than the monosulfated 2-GSP-(48)-6. The three TyrSO 3 residues together increase the binding affinity of 2-GSP-6 for P-selectin ϳ40-fold over the non-sulfated glycopeptide. The ability of the non-sulfated 2-GP-6, but not its proteolytic fragments, to bind P-selectin, indicates that the highly anionic peptide backbone of the peptide contributes to the interaction. Quantitative measurements of binding affinities indicate that the ␣1,3-fucosyl residue of 2-GSP-6 has a contribution to P-selectin binding that is equivalent to the contribu-tions of the combined three TyrSO 3 residues. By contrast, the ␣2,3-sialic acid residue has an important, but subordinate role, in comparison to either ␣1,3-fucose or tyrosine sulfation. Finally, an isomeric GSP containing a C2-O-sLe x at Thr-57, but not at Thr-44, binds to P-selectin. Thus, the stereospecific placement of both the O-glycan and the TyrSO 3 residues on the peptide is critical for optimal binding to P-selectin.
Although previous studies indicated that TyrSO 3 residues and a specific C2-O-sLe x are required for PSGL-1 to bind to P-selectin, the contributions of individual TyrSO 3 residues and the peptide backbone had not been quantitatively evaluated. Replacement of the three Tyr residues at positions 46, 48, and 51 with Phe in recombinant PSGL-1 abrogated its binding to P-selectin (12)(13)(14)(15). Site-directed replacement of any two of the three Tyr residues with Phe did not eliminate binding, indicating that a single TyrSO 3 residue may be sufficient for binding (14). But recombinant PSGL-1 expressing single TyrSO 3 residues is not functionally equivalent to recombinant wild-type PSGL-1, because cells expressing P-selectin roll more irregularly and with less mechanical strength on single-Tyr forms of PSGL-1 (18).
Our results quantify the contributions of sulfation of each of the three Tyr residues at positions 46, 48, and 51. The trisulfated 2-GSP-6 binds with high affinity to P-selectin (K d ϳ 650 nM), whereas mono-or disulfated GSPs bind with significantly lower affinity (K d ϳ 1.5-10 M) (Fig. 8). The non-sulfated 2-GP-6 also binds to P-selectin, but with even lower affinity (K d ϳ 25 M); thus, sulfation of all three Tyr residues on 2-GSP-6 increases binding affinity by ϳ40-fold. Only 25% of 35 S-labeled tryptic fragments of PSGL-1 from HL-60 cells bind to a Pselectin affinity column (17). The bound fraction may represent the fully sulfated and glycosylated peptides, and the nonbound material may represent inefficiently sulfated and/or glycosylated peptides. Partial characterization of P-selectin-bound 35 S-labeled tryptic fragments of recombinant PSGL-1 co-expressed with specific glycosyltransferases in Chinese hamster ovary cells indicated that the fragments have only minor structural differences from that of GSP-6 (17). The differences are mainly due to charge heterogeneity among the tryptic fragments. Taken together, these studies suggest that native or recombinant PSGL-1 might be partly sulfated or represent a population of mono-, di-, and trisulfated species. Differential sulfation of PSGL-1 might affect the kinetics of leukocyte rolling during inflammation (18).
The method used here to generate GSPs is particularly advantageous to explore contributions of individual TyrSO 3 residues and O-glycan positioning in interactions with P-selectin. Previously we used recombinant human tyrosyl-protein sulfotransferase-1 to enzymatically sulfate the three Tyr residues in a glycopeptide. However, tyrosyl-protein sulfotransferases do not readily sulfate specific Tyr residues, and the separation and characterization of partially sulfated isomeric peptides is extremely difficult. Here we incorporated TyrSO 3 residues into the glycopeptide chain using Fmoc-Tyr(SO 3 Na)-OH and tri-Oacetyl-GalNAc␣-Fmoc-Thr as building blocks during the solid phase peptide synthesis. Fmoc-Tyr(SO 3 Na)-OH was utilized earlier to generate mono-and disulfated peptides, but these peptides lacked O-glycosylated residues (20). The disadvantage of this method is that the acidic deprotection conditions must be mild because of the acid lability of the TyrSO 3 residues, which can lead to inefficient peptide cleavage from the solid support. The recoveries of the sulfated peptides can be improved using the acid-sensitive 2-chlorotrityl resin for the synthesis of TyrSO 3 -containing peptides (26).
The dissociation constant for binding of the original GSP-6 to  NaN 3 ). B, chromatography in buffer containing EDTA (20 mM MOPS, pH 7.5, 50 mM NaCl, 1 mM EDTA, 0.02% NaN 3 .) C, radiolabeled 2-GP-6 was loaded into the Pselectin column in buffers containing either Ca 2ϩ or EDTA (see panels A and B). D, radiolabeled 2-GP-6 digested with Pronase was loaded into the sPS column in buffer containing Ca 2ϩ (see panel A). When duplicate analyses were performed, elution profiles were identical. P-selectin is ϳ350 nM, as determined by equilibrium gel filtration (16). Our present results show that the K d for binding of 2-GSP-6 to P-selectin is ϳ650 nM. The differences in the amino acid sequences between GSP-6 (23-mer) and 2-GSP-6 (18-mer) are three extra amino acids (QAT) at the N terminus and two extra amino acids (ML) at the C terminus of GSP-6 (Met is replaced with Cys in 2-GSP-6). The difference in K d between GSP-6 and 2-GSP-6 may result from small conformational and/or charge differences that make the binding of GSP-6 to P-selectin more favorable than that of 2-GSP-6. Salt concentrations above the physiological level inhibit binding of selectins to isolated sLe x tetrasaccharides, whereas low salt concentrations enhance binding (25,27). Our present data show that lowering the NaCl concentration by 3-fold (from 150 mM to 50 mM NaCl) enhances the binding affinity of 2-GSP-6 for P-selectin by 9-fold. EDTA blocks the binding of 2-GSP-6 to P-selectin at low salt as well as at physiological salt concentrations, demonstrating the importance of Ca 2ϩ for binding under both conditions. EDTA also blocks the binding of nonsulfated 2-GP-6 to P-selectin, which indicates the involvement of Ca 2ϩ for binding of C2-O-sLe x to P-selectin. Because we were unable to detect binding of the trisulfated peptide with only a core-1 O-glycan (2-GSP-2) to P-selectin (K d Ͼ 30 M), we currently do not know whether the binding of TyrSO 3 residues to P-selectin is dependent on Ca 2ϩ . These interactions may not require Ca 2ϩ , since glycosaminoglycan ligands have been shown to bind P-selectin in a cation-independent manner (28).
Indirect experiments have suggested that sialic acid and fucose residues are critical monosaccharides for PSGL-1 binding to P-selectin, but the distinct contributions of sialic acid and fucose residues have not been quantitatively evaluated. Neuraminidase treatment of human leukocytes dramatically reduced, but did not eliminate, their binding to P-selectin, but it was not determined whether the residual binding was due to weak recognition of the resulting Le x determinant (8,29,30). Neutrophil binding to P-selectin on stimulated platelets was partially inhibited by a Le x -containing oligosaccharide, but inhibition was not compared with sLe x -containing oligosaccharides (31). Interestingly, treatment of HL-60 cells with high concentrations of ␣-fucosidase was more potent in blocking adhesion to P-selectin than treatment of the cells with neuraminidase (32). These earlier results suggested that P-selectin might bind weakly to Le x -containing ligands. Our current study quantitatively validates this suggestion. The desialylated 2-GSP-6, which expresses the Le x determinant, binds to P-selectin with measurable affinity (K d ϳ 10 -20 M) (Fig. 8). In contrast, a sialic acid-containing glycosulfopeptide without fucose (2-GSP-5) does not detectably bind to P-selectin (K d Ͼ 30 M). Thus, an ␣1,3-fucose contributes more than an ␣2,3-sialic acid to glycosulfopeptide binding to P-selectin. Collectively, the binding studies with GSPs suggests that the lectin domain of P-selectin has a binding site for ␣1,3-fucose, a secondary site for ␣2,3-sialic acid, and a binding site for peptide determinants and TyrSO 3 residues.
A model depicting these stereospecific interactions is illustrated in Fig. 9. The model suggests that the position of a C2-O-sLe x relative to tyrosine sulfation and other peptide determinants has a major impact on binding of PSGL-1 to Pselectin. The stereospecific recognition of these sugar residues is consistent with our previous observation that P-selectin binds to a glycosulfopeptide containing the sLe x epitope on a core-2 O-glycan, but not on an extended core-1 O-glycan, at Thr-57 (16). The location, orientation, and distance of the TyrSO 3 groups relative to the C2-O-sLe x in PSGL-1 are critical for optimal binding affinity. In this context it should be noted that the mature N terminus of murine PSGL-1 has only two FIG. 8. Estimated dissociation constants and relative binding affinities of glyco(sulfo)peptides for P-selectin. Dissociation constants and relative binding for each glyco(sulfo)peptides were derived from equilibrium gel filtration data. a , in these samples, the dissociation constants and relative binding were derived from P-selectin affinity chromatography data. The hyphens in each of the right two columns indicate that binding was not detectable (i.e. K d Ͼ 30 M).

FIG. 9.
A model for the stereospecific binding of 2-GSP-6 to the lectin domain of P-selectin. Three TyrSO 3 residues along with a cluster of negatively charged amino acid residues bind to a specific region within the lectin domain of P-selectin. The sLe x -capped core-2based O-glycan binds to a different location in the lectin domain of P-selectin. The results suggest that the ␣1,3-fucose residue contributes more than ␣2,3-sialic acid to binding to P-selectin. EGF, epidermal growth factor; CR, complement regulatory.
Tyr residues that are potential sites for sulfation (33). Furthermore, the positions of these Tyr residues relative to Thr residues that are potential sites of O-glycosylation are also very different from those of human PSGL-1. Thus, the model in Fig.  9 suggests that murine PSGL-1 may bind to human P-selectin with lower affinity than does human PSGL-1, because the stereochemistry of the interaction would be different. In our assays, binding of P-selectin to the peptide and TyrSO 3 determinants is not detectable in the absence of a C2-O-sLe x . However, very low affinity binding of P-selectin to these determinants may be analogous to low affinity binding of P-selectin to the platelet glycoprotein Ib-IX-V complex, which requires multiple TyrSO 3 residues, but not glycosylation, on GPIb␣ (34). In shear flow, the very high density of GPIb␣ on platelets may allow it to support platelet translocation on P-selectin despite its low binding affinity.
The specific residues within the P-selectin lectin domain that recognize the TyrSO 3 residues, peptide determinants, and sugar residues on PSGL-1 are not known. Mutagenesis and chemical modification experiments have suggested that P-selectin requires conserved lysine residues at positions 111 and 113 to bind to sLe x -containing ligands (27,(35)(36)(37)(38). The crystal structure of the lectin and epidermal growth factor domains of E-selectin has been determined (39), but no structure of Eselectin bound to a selectin ligand has been reported. However, the structure of an sLe x oligosaccharide bound to a chimeric form of mannose-binding protein has been determined (40). This structure revealed that Ca 2ϩ is coordinated to the 2-and 3-OH of fucose, which forms a network of cooperative hydrogen bonds with amino acid side chains that also coordinate Ca 2ϩ . Interestingly, the sialic acid of the sLe x oligosaccharide does not directly contact the protein. Fucose might dock in an analogous manner to specific Ca 2ϩ -coordinating residues in the lectin domain of E-or P-selectin. Basic residues in a different region of the lectin domain of P-selectin might form salt bridges with acidic peptide and TyrSO 3 determinants in PSGL-1, but other interactions such as hydrogen bonding could be equally or more important. For example, hydrogen-bonding networks, rather than salt bridges, mediate interactions of the cysteinerich region of the mannose receptor with sulfated GalNAc residues (41). Structural elucidation of the molecular contacts between P-selectin and PSGL-1 will require analysis of complexes between P-selectin and GSPs by crystallography and nuclear magnetic resonance.