A Novel Glycosulfopeptide Binds to P-selectin and Inhibits Leukocyte Adhesion to P-selectin*

P-selectin glycoprotein ligand-1 (PSGL-1) is a dimeric membrane mucin on leukocytes that binds selectins. The molecular features of PSGL-1 that determine this high affinity binding are unclear. Here we demonstrate the in vitro synthesis of a novel glycosulfopeptide (GSP-6) modeled after the extreme N terminus of PSGL-1, which has been predicted to be important for P-selectin binding. GSP-6 contains three tyrosine sulfate (TyrSO3) residues and a monosialylated, core 2-based O-glycan with a sialyl Lewis x (C2-O-sLex) motif at a specific Thr residue. GSP-6 binds tightly to immobilized P-selectin, whereas glycopeptides lacking either TyrSO3 or C2-O-sLex do not detectably bind. Remarkably, an isomeric glycosulfopeptide to GSP-6, termed GSP-6′, which contains sLex on an extended core 1-based O-glycan, does not bind immobilized P-selectin. Equilibrium gel filtration analysis revealed that GSP-6 binds to soluble P-selectin with aK d of ∼350 nm. GSP-6 (<5 μm) substantially inhibits neutrophil adhesion to P-selectin in vitro, whereas free sLex (5 mm) only slightly inhibits adhesion. In contrast to the inherent heterogeneity of post-translational modifications of recombinant proteins, glycosulfopeptides permit the placement of sulfate groups and glycans of precise structure at defined positions on a polypeptide. This approach should expedite the probing of structure-function relationships in sulfated and glycosylated proteins, and may facilitate development of novel drugs to treat inflammatory diseases involving P-selectin-mediated leukocyte adhesion.

Each 120-kDa subunit of human PSGL-1 contains numerous sialic acids and approximately 70 extracellular Ser and Thr residues, which are potential sites for O-glycosylation, plus three potential sites for N-glycosylation (6, 7) (Fig. 1). These features suggested that the large amount of carbohydrate on the mucin might promote high avidity binding to P-selectin. However, indirect evidence suggests that the extreme N-terminal extracellular region of mature PSGL-1, which begins at residue 42, is important for high affinity binding to P-selectin (reviewed in Ref. 3). Specifically, tyrosine sulfate residues and O-glycans within that region have been considered essential for binding ( Fig. 1). A monoclonal antibody directed to a peptide epitope spanning residues 49 -62 of PSGL-1 blocks cell adhesion to P-selectin, whereas antibodies to other domains of PSGL-1 do not block adhesion (8,9). Site-directed mutagenesis of recombinant PSGL-1 supports an important role for at least one Tyr residue at positions 46, 48, and 51 plus a Thr at position 57 in generating adhesion of cells to P-selectin (10 -13). Treatment of human neutrophil PSGL-1 with aryl sulfatase removes sulfate from tyrosine residues and decreases binding of PSGL-1 to immobilized P-selectin (14). These combined results were taken to predict a potential role of tyrosine sulfate plus an O-glycan at Thr 57 of PSGL-1 in binding to P-selectin.
Little is known about the specific glycan structures on human leukocyte PSGL-1 that contribute to its binding to Pselectin. Structural analyses of the total O-linked glycans of PSGL-1 from HL-60 cells revealed that many of these O-glycans have a core 2-based structure, although only a minor fraction carries the sLe x determinant (15). Generation of recombinant PSGL-1 capable of binding P-selectin in CHO cells requires co-expression of a core 2 ␤1,6 N-acetylglucosaminyltransferase (␤1,6-GlcNAcT) and an ␣1,3-fucosyltransferase (12,16). These observations led to the prediction that sLe x on core 2-based O-glycans on PSGL-1 is required for its binding to P-selectin. In support of this prediction, leukocytes from mice lacking expression of either the core 2 ␤1,6-GlcNAcT or the ␣1,3-fucosyltransferase FucT-VII bind weakly to P-selectin (17,18). Finally, P-selectin binds to chimeric glycoproteins containing only a small N-terminal domain of PSGL-1 that lacks N-glycans (10,11,19).
These studies point to an important role of the N-terminal domain of PSGL-1 in binding to P-selectin. However, the specific post-translational modifications of this domain and the relative binding affinity of this domain for P-selectin independent of the remainder of the dimeric mucin structure are not yet known. There is no direct evidence that O-glycans containing sLe x on a core 2 motif in the extreme N terminus of PSGL-1 contribute to P-selectin recognition. Furthermore, the putative relationship between tyrosine sulfation and O-glycosylation in the extreme N terminus of PSGL-1 on binding to P-selectin has not been directly examined. The overall role of sLe x itself is unclear. Some cells reportedly deficient in sLe x expression bind well to P-selectin, suggesting that glycans other than sLe x may serve as ligands for both E-and P-selectin (20). Additionally, polyanionic compounds lacking peptide, sialic acid, or fucose can inhibit P-selectin-mediated cell adhesion (21)(22)(23). Finally, recent studies have suggested that covalent dimerization of PSGL-1 is essential for its binding to P-selectin (24,25). Thus, there are many questions regarding the direct role of the Nterminal domain of mature PSGL-1 in binding to P-selectin and the specific contribution of tyrosine sulfate and O-glycosylation in this region for binding.
The interaction of P-selectin with simple sLe x -containing glycans is relatively weak (26), and concentrations of sLe x in the millimolar range are required to inhibit adhesion of cells to P-selectin (27). In contrast, surface plasmon resonance measurements indicate that monomeric P-selectin binds to neutrophil-derived PSGL-1 with relatively high affinity (K d ϳ300 nM) and rapid on/off kinetics (28). This affinity is particularly high relative to typical carbohydrate-binding proteins, where the K d for monovalent carbohydrate ligands is usually in the range of 0.1-1 mM (29). Indeed, L-selectin binds to GlyCAM-1, a heavily sialylated and sulfated mucin, with a K d of ϳ100 M (30).
We sought to directly address the functional significance of the extreme N-terminal domain of PSGL-1 and its putative post-translational modifications in binding to P-selectin. To this end, we used purified and recombinant enzymes to modify synthetic peptides based on the N-terminal sequence of human PSGL-1. The resultant glycosulfopeptides contain tyrosine sulfate residues and O-glycans with defined carbohydrate residues that allow a direct exploration of their importance in binding to P-selectin. This new strategy of glycosulfopeptide synthesis was chosen because it has several advantages over the expression of recombinant glycoproteins. Recombinant glycoproteins expressing the post-translational modifications described here are difficult, if not impossible, to produce, because of heterogeneity in glycosylation and sulfation and because of the uncertain requirements of glycosyltransferases for sitespecific initiation and modification of O-glycan structure. Here we use synthetic glycosulfopeptides to show that both tyrosine sulfate residues and a specific core 2-based O-glycan with sLe x (C2-O-sLe x ) on a nearby Thr residue are required for high affinity binding to P-selectin. The binding affinity of one of these glycosulfopeptides is similar to that of native neutrophilderived PSGL-1. Moreover, micromolar concentrations of this glycosulfopeptide are sufficient to block neutrophil attachment to immobilized P-selectin. These results help define the molecular nature of the interaction between P-selectin and PSGL-1 and pave the way for novel glycosulfopeptides that may be useful in treating inflammatory diseases.

EXPERIMENTAL PROCEDURES
Deacetylation of GP-1-Crude glycopeptide 1 (GP-1) was synthesized at the Protein Resource Facility of Oklahoma State University. Tri-Oacetylated GalNAc was incorporated into the peptide during the solid phase synthesis using tri-O-acetyl-GalNAc␣-Fmoc Thr derivative (Oxford GlycoSciences, Oxford, United Kingdom). The crude GP-1 (2 mg) was de-O-acetylated with 6 mM methanolic sodium methylate as described (31). The deacetylated peptide was purified by reversed phase HPLC. The retention time of deacetylated GP-1 (34.6 min) was clearly different from the tri-O-acetylated GP-1 (45.3 min) (Fig. 3). The yield of the pure GP-1 was 1.1-1.5 mg. In matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectra, the observed m/z for the [M Ϫ H] Ϫ molecular ion was 2952.9 (calculated m/z 2953.2) (Fig.  4). In addition the [M Ϫ H] Ϫ molecular ion of oxidized GP-1 (m/z 2969.5) was present where the methionine residue had been oxidized.
Enzymatic Synthesis of GSP-6Ј-␤1,3-GlcNAcT-Extension of GP-2 was carried out using purified recombinant ␤1,3-GlcNAcT from Neisseria meningitidis IgtA (34). Acceptor GP-2 (0.1 mM) was incubated for 20 h at 25°C in the presence of 1 mM UDP-GlcNAc and 8 nmol/h of ␤1,3-GlcNAcT (activity assayed using lactose as an acceptor) in a total volume of 100 l in 100 mM sodium cacodylate, pH 7.5, containing 5 mM ATP, 15 mM MnCl 2 , and 0.5% bovine serum albumin. After deproteination and desalting the reaction mixture was analyzed by HPLC. The retention time of the product GP-3Ј was 31.9 min (data not shown). The conversion of GP-2 to GP-3Ј was 62%. In MALDI-TOF analysis, the observed m/z for the [M Ϫ H] Ϫ molecular ion of GP-3Ј was 3317.8 (calculated m/z 3318.5) (data not shown). ␤1,4-GalT, ␣2,3-sialylT, and ␣1,3-FucT-VI reactions were carried out essentially as described above. Each glycosyltransferase reaction was followed by HPLC. GP-6Ј had a retention time of 30.3 min in HPLC (data not shown). In MALDI-TOF analysis, the observed m/z for the [M Ϫ H] Ϫ molecular ion of GP-6Ј was 3917.6 (calculated m/z 3918.1) (data not shown). The TPST-1 reaction using GP-6Ј as an acceptor and [ 35 S]PAPS (400,000 cpm/nmol) as a donor was performed as described for GP-6. The retention time of 35 SO 3 -GSP-6Ј was 18.7 min in HPLC. Partially sulfated slower moving products were not present.
Reversed Phase High Performance Liquid Chromatography-Glycopeptide samples were filtered on a Spin-X membrane (Corning Costar, Cambridge, MA) and were subsequently analyzed in a reversed phase C-18 HPLC column (Vydac, Hesperia, CA) on a Beckman System Gold HPLC. The following solvent system was used at a flow rate of 1 ml/min: 1-10 min, 20% acetonitrile, 80% water, 0.1% trifluoroacetic acid; 10 -70 min, linear acetonitrile gradient from 20% to 45% in water, 0.1% trifluoroacetic acid. The UV absorbance at 215 nm or 275 nm was monitored and/or the radioactivity of the collected fractions was measured. The pooled fractions were dried under vacuum.
a flow rate of 70 l/min. Radioactivity in the fractions was determined by liquid scintillation counting. Control experiments to test inhibitors of binding between sPS and GSP-6 were performed using 400 pmol of sPS and 5,000 cpm/ml of 35 SO 3 -GSP-6. The EDTA concentration was 1 mM in 20 mM MOPS, pH 7.5, 150 mM NaCl, 0.02% NaN 3 . Anti-P-selectin monoclonal antibodies (G1 and S12) (800 pmol each) were preincubated for 30 min with sPS in buffer A before 35 SO 3 -GSP-6 was added. Elution was performed by using buffer A with 35 SO 3 -GSP-6.
Mass Spectrometric Analysis-MALDI-TOF mass spectrometry was performed in the linear negative ion delayed extraction mode with a Biflex TM time-of-flight instrument (Bruker-Franzen Analytik, Germany) equipped with a nitrogen laser operating at 337 nm. HPLCpurified glycopeptide samples, except GSP-2 and GSP-6, were dissolved in 30% aqueous acetonitrile, and a 0.5-l aliquot (about 2.5 pmol) was mixed with 0.5 l of 2,4,6-trihydroxyacetophenone matrix (3 mg/ml in acetonitrile, 20 mM aqueous diammonium citrate, 1:1, by volume) and immediately dried under reduced pressure. The spectra were externally calibrated with insulin [M Ϫ H] Ϫ and [M Ϫ 2H] 2Ϫ signals. Electrospray ionization mass spectra were collected in the negative ion mode using a Q-TOF hybrid quadrupole/orthogonal acceleration time-of-flight mass spectrometer (Micromass Ltd., Manchester, UK). GSP-2 and GSP-6 were dissolved in 50% aqueous acetonitrile and injected into the mass spectrometer with a nanoelectrospray ion source. Instrument calibration was performed with sodium trifluoroacetate ion clusters (38).
The reaction mixture was desalted and deproteinated before analysis by chromatography on sPS columns.
Construction, Expression, and Purification of Recombinant, Soluble Core 2 ␤1,6-GlcNAcT-A fusion protein was constructed that contained the catalytic and stem region of human core 2 ␤1,6-GlcNAcT with the 12-amino acid HPC4 epitope at both the N and C termini. The epitope is bound in a Ca 2ϩ -dependent manner by the monoclonal antibody HPC4 (40). The catalytic and stem region of the core 2 ␤1,6-GlcNAcT was amplified by polymerase chain reaction using a pcDNA3 plasmid containing the full-length cDNA of the human core 2 ␤1,6-GlcNAcT (type L) as a template (12). The following primers were used for amplification: 5Ј primer containing BamHI cleavage site, 5Ј-GCCTGAATTT-GTAAGGGATCCACACTTAGAGCTTGCTGGGGAGAATCC-3Ј and 3Јprimer containing EcoRI site and HPC4 epitope, 5Ј-GTAGAATTCTTA-TCACTTGCCGTCGATCAGCCTGGGGTCCACCTGGTCCTCGTGTTT-TAATGTCTCCAAAGC-3Ј. The polymerase chain reaction product (1.2 kilobase pairs) was cloned into pCR-TOPO 2.1 vector (Invitrogen, Carlsbad, CA) and used to transform E. coli strain JM109 for plasmids preparation. The construct was released from pCR-TOPO 2.1 vector by digestion with BamHI and EcoRV and purified by agarose gel electrophoresis. The construct (1.2 kilobase pairs) was ligated into a BamHI/EcoRV site of modified pcDNA 3.1(ϩ) vector (pcDNA 3.1 (ϩ)T-H), which contains an NH 2 -terminal transferrin signal sequence and HPC4 epitope (33) and used to transform Escherichia coli strain DH5␣. The resulting plasmid, pcDNA 3.1(ϩ)TH-sC2 (6.7 kilobase pairs), was isolated and sequenced and used to transfect CHO/dhfr Ϫ cells using LipofectAMINE (Life Technologies, Inc.). Clonal selection was carried out under neomycin resistance, and the cells were maintained in Dulbecco's modified Eagle's medium (Cellgro, Herndon, Virginia) containing 10% fetal calf serum and G418 (600 g/ml). Stable clones of cells expressing core 2 ␤1,6-GlcNAcT activity in the media (50 nmol/h/ml) were selected and grown to 100% confluence. The medium was changed to Dulbecco's modified Eagle's medium containing 2% fetal calf serum and incubated for 2-3 days. The medium was collected and adjusted to 1 mM CaCl 2 and 5 mM benzamidine. Soluble core 2 ␤1,6-GlcNAcT containing an HPC4 epitope tag was purified from the conditioned medium (60 ml) using a HPC4-mAb affinity column (3.5-ml column of 5 mg/ml HPC4-mAb coupled to Ultralink TM biosupport medium) at 4°C as described (41). The purified enzyme was stabilized by adding 0.1% bovine serum albumin, and the enzyme was concentrated using Centricon-30 ultrafiltration tubes (Amicon, Beverly, MA). The purified enzyme was used directly or aliquoted and stored at Ϫ20°C. The activity (8.2 mol/h/ml) was stable at Ϫ20°C for at least 2 months. Core 2 ␤1,6-G-lcNAcT assays were performed using 1 mM Gal␤133GalNAc␣-p-nitrophenyl (Toronto Research Chemicals Inc., Canada) and 1 mM UDP-[ 3 H]GlcNAc (specific activity 1000 cpm/nmol). The assays were carried out at 37°C with 2.5-10 l of the purified enzyme for 30 min or 25 l of cell culture medium for 2-3 h in a total volume of 50 l of 50 mM sodium cacodylate, pH 7.0. The radiolabeled reaction product was separated from the radiolabeled donor using Sep Pak cartridges (Waters, Milford, MA).
Neutrophil Isolation and Labeling-Human neutrophils were isolated from healthy volunteers as described (42) and labeled with Calcein-AM (Molecular Probes, Inc., Eugene, OR) according to the manufacturer's instructions.
Neutrophil Adhesion Assay-The adhesion assay was performed essentially as described (37), with the following modifications. Calceinlabeled neutrophils were used. sPS was coated directly on wells of Immulon 1 microtiter plates by incubating the wells with 2 g/ml sPS in 0.1 M sodium carbonate buffer at 4°C overnight (100 l/well). For GSP-6 inhibition, the wells were preincubated with 50 l of different dilutions of GSP-6 in Hank's balanced salt solution containing 0.1% human serum albumin at room temperature for 15 min. In control experiments, wells were preincubated with mAbs against P-selectin. In other controls, mAbs against PSGL-1 or fluid-phase sPS were preincubated with 25 l of the cell suspension at room temperature for 15 min. The neutrophils (25 l) were then added to the sPS-coated wells. The number of adherent cells was quantified using an Fmax fluorescence plate reader (Molecular Devices).

Synthesis of Glyco(sulfo)peptides-
The possibility that the N-terminal region of PSGL-1 is independently capable of high affinity interactions with P-selectin was explored by synthesizing the target glycosulfopeptide designated GSP-6. The synthetic route to GSP-6 is shown schematically in Fig. 2. This glycosulfopeptide was targeted to test the hypothesis that a single O-glycan in conjunction with TyrSO 3 is required for high affinity binding to P-selectin. The synthesis of this complex glycosulfopeptide has not been described previously, partly because current chemical methods for synthesis are exceedingly complicated. In addition, the sulfotransferase and glycosyltransferases capable of modifying a peptide to generate GSP-6 have only recently become available.
A key enzyme catalyzing the first step in the biosynthesis of O-glycans is a polypeptide:N-acetylgalactosaminyltransferase (␣-GalNAcT) that adds GalNAc from UDP-GalNAc to Ser and Thr residues. A large family of these enzymes has recently been discovered (43)(44)(45). Some demonstrate recognition of Ser/Thr residues within specific polypeptide domains, and only one (␣-GalNAcT-4) appears to recognize the Thr residue within the N-terminal peptide domain of PSGL-1 (44). Rather than face uncertainty as to which ␣-GalNAcT can add GalNAc to a specific Ser or Thr on a peptide, we incorporated an acetylated Fmoc derivative of GalNAc-Thr at a specific site during peptide synthesis. We recently observed that conversion of Thr 57 to Ala, but not of Thr 44 to Ala, blocks binding of full-length recombinant PSGL-1 to P-selectin (13). Thus, we initiated O-glycan synthesis on the Thr residue corresponding to Thr 57 rather than Thr 44 (Fig. 2). This deacetylated glycopeptide, designated GP-1, served as the starting material for the synthesis.
The stable synthesis of peptides containing multiple tyrosine sulfate residues (TyrSO 3 ) is chemically complex. Preliminary studies demonstrated inefficient synthesis of such peptides when Fmoc derivatives of TyrSO 3 were utilized in the solidphase synthesis of peptides. Therefore, a soluble recombinant form of tyrosyl-protein sulfotransferase-1 (TPST-1) (32, 33) was used to add sulfate to Tyr residues to form the glycosulfopeptide GSP-6.
At each step of the synthesis, the glycopeptide products were analyzed by reversed phase HPLC (Fig. 3). The addition of each monosaccharide caused a significant reduction in retention time for the glycopeptides, allowing the completeness of each reaction step to be easily monitored. At each step of the reactions shown in Fig. 2, the glycopeptide products were purified by reversed phase HPLC, and mass spectral analyses were performed to verify the sizes of the synthesized products. As shown in Fig. 4, the predicted and observed masses for each glycopeptide were identical within experimental error. Thus, a purified form of each of the glycopeptides (GP-1 through GP-6) and the glycosulfopeptide (GSP-6) was obtained. In control studies some of the glycopeptide intermediates were enzymatically sulfated by TPST-1. For example, GP-2 was enzymatically sulfated to generate GSP-2. Only microgram quantities of each glyco(sulfo)peptide were required for the current studies. Because of the yield and efficiency of synthesis of these compounds, it is possible to synthesize larger quantities.
GSP-6 Binds to Immobilized P-selectin-To test whether these synthetic glyco(sulfo)peptides interact with P-selectin, a series of affinity columns containing recombinant sPS at different coupling densities were prepared. The different column densities allow estimation of the relative affinities of glyco(sulfo)peptides for P-selectin.
Radiolabeled glyco(sulfo)peptides were applied to immobilized sPS in Ca 2ϩ containing buffer (Fig. 5). In columns containing sPS at densities ranging from 1.0 to 2.0 mg/ml, the only glyco(sulfo)peptide retarded or bound was GSP-6. The elution of GSP-6 was retarded on columns containing 1.3 and 1.6 mg/ml sPS. GSP-6 bound to the column containing 2.0 mg/ml sPS and could be eluted with EDTA. GP-6 lacking the sulfates on tyrosines and GSP-2 lacking the sLe x determinant had no detectable affinity for sPS. The results demonstrate the dual importance of sulfated tyrosines and sLe x for binding. Interestingly, neither GSP-5, which lacks the fucosyl residue, nor the desialylated form of GSP-6 bound detectably to immobilized sPS. These results demonstrate that both sialic acid and fucose in sLe x are necessary for high affinity binding of GSP-6 to immobilized sPS.
An Isomer of GSP-6 Does Not Bind to Immobilized P-selectin-To test whether a core 2-based O-glycan is essential for binding of GSP-6 to immobilized sPS, we synthesized a novel glycosulfopeptide that is isomeric in structure to GSP-6. This glycosulfopeptide, designated GSP-6Ј, has sLe x on an extended core 1-based O-glycan (C1-O-sLe x ) rather than on a core 2-based O-glycan (Fig. 6A). It was synthesized by a series of steps as outlined under "Experimental Procedures." A key step in the synthesis of GSP-6Ј is the addition of GlcNAc in ␤1-3 linkage to the Gal residue in the core 1 O-glycan by a recombinant ␤1,3-GlcNAcT from N. meningitidis IgtA (34). This glycopeptide, designated GP-3Ј, was subsequently modified by the action of ␤1,4-GalT, ␣2,3-sialylT, and ␣1,3-FucT to generate a glycopeptide designated GP-6Ј, which has sLe x on the extended core 1 O-glycan. GP-6Ј was converted to GSP-6Ј by action of TPST-1. Mass spectral analysis confirmed the predicted size of the final product.
Unexpectedly, GSP-6Ј did not bind to immobilized sPS (Fig.  6B). To confirm the presence of sLe x on the extended core 1 O-glycan, enzyme-linked immunosorbent assays were performed using 2H5, a monoclonal antibody that recognizes the sLe x determinant (46). 2H5 bound to immobilized GP-6 and GP-6Ј, but not to the control glycopeptide GP-2, which lacks sLe x (data not shown). This verifies the expression of sLe x determinants on both GP-6 and GP-6Ј. Taken together, these results demonstrate that sLe x must be expressed on a core 2-based O-glycan for GSP-6 to bind immobilized sPS.
GSP-6 Binds with Relatively High Affinity to Soluble Pselectin-The dissociation constant (K d ) for binding of GSP-6 to soluble sPS was determined using an equilibrium gel filtration technique (35,36). Different amounts of fluid-phase sPS (25-1000 pmol) were loaded into a small gel filtration column equilibrated with 35 SO 3 -GSP-6 in Ca 2ϩ -containing buffer (Fig.  7A). The binding data were plotted to derive the equilibrium binding constant, yielding an estimated K d of ϳ350 nM (Fig.  7B). Binding of GSP-6 to sPS was inhibited with EDTA and with the inhibitory anti-P-selectin mAb G1, which binds to the lectin domain of P-selectin (Fig. 7B, inset). Binding was not inhibited with anti-P-selectin mAb S12, which binds to one of the consensus repeats of P-selectin (47,48). These results demonstrate that GSP-6 binds with relatively high affinity to sPS in a Ca 2ϩ -dependent manner.
GSP-6 Is a Potent Inhibitor of Neutrophil Adhesion to Immobilized Soluble P-selectin-The ability of GSP-6 to inhibit neutrophil adhesion to P-selectin was tested in microtiter wells coated with sPS (Fig. 8). We first validated the specificity of adhesion. Adhesion was inhibited by EDTA and the anti-P-selectin mAb G1, but not by the anti-P-selectin monoclonal antibody S12. Adhesion was also inhibited by PL1, a mAb directed to an N-terminal epitope of PSGL-1 that blocks binding of PSGL-1 to P-selectin. In contrast, PL2, which recognizes an epitope within the mucin decapeptide repeats of PSGL-1, did not inhibit adhesion. These results demonstrate that adhesion in this assay requires binding of PSGL-1 to sPS. Low concentrations of fluid-phase sPS (5.67 M) inhibited neutrophil adhesion. A similar concentration of GSP-6 (4.7 M) also significantly inhibited neutrophil adhesion to immobilized sPS. In marked contrast, a pure sLe x -containing tetrasaccharide (NeuAc␣233Gal␤134[Fuc␣133]GlcNAc) only minimally inhibited neutrophil adhesion even at very high concentrations (5.3 mM). Taken together, these results demonstrate that GSP-6 binds specifically to P-selectin and strongly inhibits PSGL-1-dependent neutrophil adhesion to P-selectin. DISCUSSION This study demonstrates that the small glycosulfopeptide GSP-6 binds with high affinity to P-selectin. GSP-6 represents a portion of the primary sequence of the extreme N terminus of PSGL-1 and was modified to contain a specific C2-O-sLe x on Thr 57 and TyrSO 3 residues at Tyr 46 , Tyr 48 , and Tyr 51 . Comparison of binding of various glyco(sulfo)peptides demonstrates that both C2-O-sLe x and TyrSO 3 are required for high affinity binding of GSP-6 to P-selectin. GSP-6 binds to soluble P-selectin with a K d of ϳ350 nM and inhibits PSGL-1-dependent adhesion of neutrophils to immobilized P-selectin.
Requirement of Tyrosine Sulfation for Glycosulfopeptide Binding to P-selectin-The inability of glycopeptides lacking TyrSO 3 to bind immobilized P-selectin demonstrates that TyrSO 3 plays a critical role in glyco(sulfo)peptide recognition by P-selectin. Previously, indirect evidence suggested an im- portant role of TyrSO 3 residues in promoting high affinity binding of PSGL-1 to P-selectin. Treatment of native PSGL-1 with bacterial arylsulfatase blocks its binding to immobilized sPS (14). Blockade of overall sulfation of PSGL-1 by treating cells with chlorate to prevent formation of the phosphoadenosine phosphosulfate, the substrate for sulfation reactions, reduces adhesion of cells to P-selectin (10,11). Replacement of the three tyrosine residues at positions 46, 48, and 51, which fall within the tyrosine sulfation motif, with phenylalanines inhibits binding of recombinant PSGL-1 to P-selectin (10 -13). However, replacement of two of the three Tyr residues with Phe within the extreme N-terminal domain of PSGL-1 permits binding of recombinant PSGL-1 to P-selectin (13). The present study demonstrates directly that sulfation of tyrosine residues 46, 48, and 51 in GSP-6 promotes high affinity binding of GSP-6 to P-selectin. The GSP technology will allow testing of the importance of the number and relative locations of TyrSO 3 to binding of glycosulfopeptides to P-selectin.
Requirement of C2-O-sLe x for Glycosulfopeptide Binding to P-selectin-Our results directly demonstrate that an sLe x -bearing O-glycan on Thr 57 in GSP-6 is required for high affinity binding to P-selectin. Previous indirect evidence suggested a requirement for an O-glycan in this region. Chimeric glycoproteins containing only the extreme N-terminal region of PSGL-1 and lacking N-glycosylation sites bind to P-selectin (10,11). Substitution of Thr 57 , but not Thr 44 , blocks binding of fulllength recombinant PSGL-1 to P-selectin (12,13). Other observations suggested that ␣1,3-fucosylation of PSGL-1 is also required for binding to P-selectin. Expression of recombinant human PSGL-1 in COS cells capable of binding P-selectin required co-expression of an ␣1,3-fucosyltransferase (7). Leukocytes from mice lacking FucT-VII also fail to bind P-selectin (18).
Unexpectedly, we found that sLe x on a core 2-based Oglycan, but not an extended core 1-based O-glycan, promotes high affinity binding of GSP-6 to P-selectin. Previous indirect evidence suggested that expression of recombinant human PSGL-1 in CHO cells capable of binding P-selectin required co-expression of the core 2 ␤1,6-GlcNAcT (12,16). HL-60 cell-derived PSGL-1 contains many O-glycans with the core 2 motif, but only a minor subset of these O-glycans contain the sLe x antigenic determinant (15). Core 2-based O-glycans may also be important for neutrophil adhesion to P-selectin in vivo, since leukocytes from mice lacking expression of the Different amounts of sPS were loaded into a Sephadex G-75 gel filtration column equilibrated with 35 SO 3 -GSP-6 (10,000 cpm/ml) in buffer. Elution was carried out using the same specific activity of 35 SO 3 -GSP-6 in buffer. Fractions of 140 l were collected. B, the bound and free sPS concentrations were calculated from the equilibrium gel filtration data shown in Fig. 7A. Nonlinear curve fitting was plotted using a rectangular hyperbola equation. The calculated dissociation constant (K d ) is ϳ350 nM. The inset shows the specificity of binding of sPS to 35 SO 3 -GSP-6. Gel filtration was carried out using 400 pmol of sPS and a constant amount of 35 SO 3 -GSP-6 (5000 cpm/ml) in the presence of the indicated inhibitor. The EDTA concentration in the buffer was 1 mM, and 800 pmol of the anti-P-selectin mAbs G1 or S12 were mixed with sPS. core 2 ␤1,6-GlcNAcT fail to bind to fluid-phase P-, L-or E-selectin (17). Our results directly demonstrate that the expression of a C2-O-sLe x is required for high affinity binding of GSP-6 to P-selectin.
It is remarkable that GSP-6Ј, which has C1-O-sLe x , does not bind P-selectin. This result suggests that the ␤1,6 GlcNAc residue of the core 2 branch may enhance accessibility of the sLe x moiety to P-selectin, perhaps through a more flexible linkage (49). It is also possible that the ␤1,6 GlcNAc residue subtly alters the secondary structure of the glycosulfopeptide, thereby enhancing the exposure of the TyrSO 3 along with the sLe x moiety. Alternatively, it is possible that the underlying core 2 O-glycan may bind directly to P-selectin. The availability of GSPs with various core O-glycan modifications will now allow a direct testing of these possibilities.
Taken together, these results suggest a model in which ␣1,3fucosyl and ␣2,3-sialyl residues within C2-O-sLe x plus one or more TyrSO 3 residues on GSP-6 interact with the lectin domain of P-selectin (Fig. 9). Such interactions may not occur efficiently if sLe x is present in an extended core 1 O-glycan, as for GSP-6Ј, where the ␣1,3-fucosyl and ␣2,3-sialyl residues may be displaced in their orientations relative to the TyrSO 3 residues (Fig. 9). This model suggests that the lectin domain of P-selectin has independent binding sites for these determinants on GSP-6 and that a specific stereochemical interaction occurs. This possibility is indirectly supported by the observation that the affinity of GSP-6 for P-selectin far exceeds that of the simple tetrasaccharide sLe x . Mapping of the specific binding sites in the lectin domain of P-selectin for GSP-6 will require structural characterization by crystallography and other approaches.
Previous structural analyses demonstrated that some core 2-based O-glycans from PSGL-1 from HL-60 cells contain a polyfucosylated, polylactosamine sequence (33Gal␤134 (Fuc␣133)GlcNAc␤13) n capped with sLe x (15). The present results show that a glycosulfopeptide lacking this polyfucosylated, polylactosamine sequence binds P-selectin with high affinity. Core 2 O-glycans with polyfucosylated, polylactosamine sequences in PSGL-1 may bind to P-selectin with altered kinetics or affinity compared with core 2-based O-glycans lacking this feature. Alternatively, the polyfucosylated polylac-tosamine may promote binding of PSGL-1 to other selectins. The synthesis of GSPs with core 2-based O-glycans containing polyfucosylated, polylactosamine sequences may help to address these possibilities. In vitro, FucT-VII can generate a terminal sLe x moiety on a polylactosamine sequence, but cannot add internal fucosyl residues to generate the polyfucosylated, polylactosamine sequence found in the O-glycans of PSGL-1 (50,51). Human and murine leukocytes contain a second ␣1,3-fucosyltransferase, termed FucT-IV, that adds fucosyl residues to internal sequences of polylactosamine (50). The combined actions of both FucT-VII and FucT-IV may be necessary to efficiently synthesize such polyfucosylated polylactosamines.
Comparison of Glycosulfopeptide and PSGL-1 Binding to P-selectin-Our results indicate that a synthetic glycosulfopeptide binds to P-selectin with an affinity equivalent to that of neutrophil-derived PSGL-1, which occurs as a disulfide-bonded dimer in the membrane. These results contrast with recent proposals that covalent dimerization of PSGL-1 is required for binding to fluid-phase P-selectin (24,25). Surface plasmon resonance measurements demonstrate that monomeric, soluble recombinant P-selectin binds immobilized neutrophil-derived PSGL-1 with a K d of ϳ300 nM (28). This compares favorably with that of GSP-6, which binds to P-selectin with a K d of ϳ350 nM, as determined by Hummel-Dreyer equilibrium gel filtration. The discrepancies between our results and those suggesting a requirement for covalent dimerization of PSGL-1 for binding to P-selectin are unclear. The use of recombinant molecules synthesized in a cellular environment presents special difficulties, since the precise post-translational modifications required for binding of recombinant PSGL-1 to P-selectin are not readily controlled nor easily quantified. In agreement with the properties of GSP-6, soluble tryptic fragments derived from the N terminus of either HL-60-cell-derived PSGL-1 or recombinant PSGL-1 co-expressed in CHO cells with core 2 ␤1,6-GlcNAcT and FucT-VII bind to immobilized sPS. 2 Micromolar concentrations of GSP-6 inhibit neutrophil attachment to P-selectin. By contrast, even millimolar concentrations of free sLe x tetrasaccharide inhibit adhesion poorly, in agreement with previous studies (27). P-selectin binds with low affinity to cells expressing sLe x in the absence of PSGL-1, whereas it binds with much higher affinity to HL-60 cells or neutrophils (26,52). PSGL-1 is perhaps a unique macromolecular ligand for P-selectin because it presents TyrSO 3 and C2-O-sLe x in the appropriate configuration and orientation to promote high affinity binding. PSGL-1 also binds to L-selectin, and the anti-PSGL-1 antibody PL1 inhibits binding by L-selectin (53)(54)(55)(56). This suggests that L-selectin also binds to the extreme N-terminal domain of PSGL-1. Whether GSP-6 or its analogs can bind to L-selectin is not yet known.
A variety of in vivo studies have confirmed that PSGL-1 is an important selectin ligand. Antibodies to PSGL-1 inhibit interactions of leukocytes with P-selectin and recruitment of leukocytes to areas of inflammation in animal models (57)(58)(59). Recombinant soluble forms of PSGL-1 inhibit selectin-mediated inflammatory responses in models of inflammation and thrombosis in vivo (60 -63). The ability of GSP-6 to block PSGL-1-dependent neutrophil adhesion to P-selectin at micromolar concentrations in vitro suggest that GSP-6 might block selectinmediated inflammatory responses in vivo.
Implications of Glycosulfopeptide Generation-The ability to synthesize glycosulfopeptides with specific TyrSO 3 residues and O-glycans may help define the roles of such modifications in other glycoproteins. For example, the chemokine receptors CXCR4 and CCR5 are also co-receptors with CD4 for the human immunodeficiency virus-1 (HIV-1). Sulfation of N-terminal tyrosine residues contributes to the efficiency of HIV-1 entry through both chemokine co-receptors (64). CCR5 also contains N-terminal O-glycans, but their potential functions are unclear (64). Synthetic glycosulfopeptides may allow direct exploration of the roles of O-glycans and TyrSO 3 residues for HIV-1 binding to the chemokine receptors. The Alzheimer ␤/A4 amyloid precursor protein (APP) contains N-and O-glycans and TyrSO 3 residues (65, 66). Although little is known about the possible role of TyrSO 3 in APP, the O-glycans on APP may modulate its proteolytic processing in the late Golgi (67). Synthetic glycosulfopeptides modeled after APP domains may help clarify the proteolytic processing events giving rise to ␤-amyloid deposits.
The glycosulfopeptides we have generated could not have been derived easily by expression of recombinant glycoproteins. Precise definition of the glycosylation and tyrosine sulfation in recombinant glycoproteins is exceedingly difficult, because of the inherent heterogeneity in glycan structures and the variable efficiency of tyrosine sulfation. Furthermore, cellular initiation of O-glycosylation by ␣-GalNAcTs is complex. Each of the many different enzymes in this family may require a slightly different peptide motif for initiation of O-glycosylation (43)(44)(45). The technology we have used for synthesis of glycosulfopeptides allows the complete control of the O-glycan sites and structures without regard to O-glycosylation motifs. The in vitro synthesis of these glycosulfopeptides also allows the introduction of modified monosaccharides, e.g. sialic acid derivatives, precise modifications of glycan structures, as in GSP-6Ј, modifications in the peptide length, and exploration of the possible contribution of peptide primary and secondary structure to the binding of a glycosulfopeptide to its receptor. Thus, this methodology offers important new avenues to explore the structure/function relationships of O-glycosylation and tyrosine sulfation.