Noncovalent association of P-selectin glycoprotein ligand-1 and minimal determinants for binding to P-selectin.

P-selectin glycoprotein ligand-1 (PSGL-1) is a disulfide-bonded, homodimeric mucin ( approximately 250 kDa) on leukocytes that binds to P-selectin on platelets and endothelial cells during the initial steps in inflammation. Because it has been proposed that only covalently dimerized PSGL-1 can bind P-selectin, we investigated the factors controlling dimerization of PSGL-1 and re-examined whether covalent dimers are required for binding its P-selectin. Recombinant forms of PSGL-1 were created in which the single extracellular Cys (Cys(320)) was replaced with either Ser (C320S-PSGL-1) or Ala (C320A-PSGL-1). Both recombinants migrated as monomeric species of approximately 120 kDa under both nonreducing and reducing conditions on SDS-polyacrylamide gel electrophoresis. P-selectin bound similarly to cells expressing either wild type or mutated forms of PSGL-1 in both flow cytometric and rolling adhesion assays. Unexpectedly, chemical cross-linking studies revealed that both C320S- and C320A-PSGL-1 noncovalently associate in the plasma membrane and cross-linking generates dimeric species. Chimeric recombinants of PSGL-1 in which the transmembrane domain in PSGL-1 was replaced with the transmembrane domain of CD43 (CD43TMD-PSGL-1) could not be chemically cross-linked, suggesting that residues within the transmembrane domain of PSGL-1 are required for noncovalent association. Cells expressing CD43TMD-PSGL-1 bound P-selectin. To further address the ability of P-selectin to bind monomeric derivatives of PSGL-1, intact HL-60 cells were trypsin-treated, which generated a soluble approximately 25-kDa NH(2)-terminal fragment of PSGL-1 that bound to immobilized P-selectin. Because N-glycosylation of PSGL-1 hinders trypsin cleavage, a recombinant form of PSGL-1 was generated in which all three potential N-glycosylation sites were mutated (DeltaN-PSGL-1). Cells expressing DeltaN-PSGL-1 bound P-selectin, and trypsin treatment of the cells generated NH(2)-terminal monomeric fragments (<10 kDa) of PSGL-1 that bound to P-selectin. These results demonstrate that Cys(320)-dependent dimerization of PSGL-1 is not required for binding to P-selectin and that a small monomeric fragment of PSGL-1 is sufficient for P-selectin recognition.

Earlier studies on recombinant chimeric forms of PSGL-1 suggested that P-selectin recognizes the NH 2 -terminal region of PSGL-1 in both static cell adhesion assays (12,13) and affinity capture precipitation assays (12). Soluble Ig chimeras containing various lengths of the NH 2 terminus of PSGL-1 bound fluid-phase P-selectin in affinity capture assays when immobilized on Protein A beads. Cells expressing P-selectin also adhered to plates coated with the Ig chimeras under static conditions. In both assays, however, binding of the smallest Ig chimera containing the first 19 residues of fully processed PSGL-1 (residues 42-60) to P-selectin appeared to be reduced compared with that of longer constructs (12). Appendage of amino acid residues 38 -57 to CD43, a membrane-bound sialomucin, promoted cell adhesion to immobilized P-selectin (13). In contrast with these results, it was recently reported that mutation of the single extracellular Cys residue in the context of full-length PSGL-1 prevents dimerization and eliminates binding of PSGL-1 to P-selectin in both flow cytometric and cell adhesion assays (18,19). Based on these studies it was proposed that only covalently dimerized forms of PSGL-1 can bind to P-selectin. However, we recently generated a series of semisynthetic, monomeric glycosulfopeptides based on the NH 2terminal region of human PSGL-1 (24) and tested them for their binding to P-selectin. The results demonstrated that a glycosulfopeptide containing 3 Tyr-SO 3 Ϫ residues and a core-2 based O-glycan expressing the sLe x antigen binds P-selectin with an affinity equivalent to that of the native, dimeric PSGL-1 from human neutrophils (20,24).
Here we report our studies on the factors controlling dimerization of membrane PSGL-1 and our re-examination of the issue of whether monomeric forms of PSGL-1 can bind to Pselectin. Our results confirm that the Cys residue at position 320 in full-length PSGL-1 is required for covalent dimerization. However, recombinant forms of the full-length PSGL-1 in which Cys 320 has been mutated to either Ser or Ala retain the ability to bind P-selectin in either fluid-phase binding assays or in cell adhesion assays under shear conditions. Unexpectedly, these mutated forms of PSGL-1 efficiently associate noncovalently in the membrane and chemical cross-linking generates dimeric species. Replacement of the transmembrane domain in PSGL-1 with that of CD43 generated a recombinant form of PSGL-1 that no longer self-associates in the membrane, but is able to support P-selectin binding. Proteolytic cleavage of membrane native and recombinant PSGL-1 generated small glycosulfopeptides derived from the NH 2 terminus of PSGL-1 that retain high affinity binding to P-selectin. Taken together, the results demonstrate that dimerization of membrane PSGL-1 is not required for P-selectin binding.
PCR1-A small extracellular region of PSGL-1 immediately adjacent to the TMD was amplified. The sense primer (primer 1) is homologous to a region just upstream of the PSGL-1 TMD and contains the BsmI restriction site, thus defining the extracellular region of PSGL-1 to be manipulated. The chimeric antisense primer (primer 2) contains sequence encoding amino-terminal residues of the CD43 TMD at its 5Ј end, thus incorporating the proper splice junction between the extracellular residues of PSGL-1 adjacent to the TMD and the new CD43based TMD into the PCR product. Primers 1 and 2 were used with PSGL-1/pZeoSV as the template. An annealing temperature of 55°C was used. The 101-bp product was gel purified.
PCR2-In a similar manner, the cytoplasmic region of PSGL-1 adjacent to the TMD was amplified using a chimeric sense primer (primer 3) containing sequence encoding carboxyl-terminal residues of the CD43 TMD at its 5Ј end. The antisense primer (primer 4) is homologous to a region just downstream of the PSGL-1 TMD and contains the EcoNI restriction site, thus defining the extent of the cytoplasmic region of PSGL-1 to be manipulated. Primers 3 and 4 were used with PSGL-1/pZeoSV as the template and an annealing temperature of 72°C. The 159-bp product was gel-purified.
PCR3-The TMD of CD43 was amplified using chimeric sense and antisense primers that contained 5Ј sequence encoding the residues which encompass the TMD of PSGL-1. Primers 5 and 6 were used with CD43/pBKEF as the template and an annealing temperature of 68°C. The program included a 1 o /4-s slope in the transition from annealing to extension temperature (72°C). The 97-bp product was gel-purified.
PCR4 -The 3 purified PCR products were mixed together in 1ϫ PCR buffer with 1.25 mM dNTPs and Pfu polymerase. They were allowed to denature for 3 min, anneal at 62°C for 1 min, and extend at 72°C for 10-min in order to generate the chimeric template. Primers 1 and 4 were added after the extension period and amplification was performed with an annealing temperature of 62°C. The 301-bp product was gel purified and digested with BsmI and EcoNI. The resultant 269-bp BsmI-EcoNI product was gel-purified and ligated into PSGL-1/pZeoSV following the strategy outlined for C320S-PSGL-1/pZeoSV. After restriction enzyme screening, several potential clones were sequenced. All of the clones had correct splice sites at the PSGL-1/CD43 TMD junctions, but the internal portion of the CD43 TMD was not correct due to mispriming of primer 5. Fortunately, the incorrect sequence was flanked by AlwNI and NarI sites, so the following strategy was used to fix the sequence: CD43/pBKEF was digested with AlwNI and NarI, and the 55-bp AlwNI-NarI fragment encoding the internal portion of the CD43 TMD was gel purified and saved for later use. The partial CD43TMD-PSGL-1/pZeoSV clone (pCD43TMD-PSGL-1/pZeoSV) was digested with AlwNI to generate two species. The 3674-bp AlwNI-AlwNI product containing the bulk of the pZeoSV vector and the sequence encoding the extracellular domain of PSGL-1 up to the AlwNI site in the CD43 TMD was stored for later use. The 1015-bp AlwNI-AlwNI fragment containing sequence encoding the rest of the CD43 TMD (including the incorrect portion), the cytoplasmic domain of PSGL-1, and the remainder of plasmid was further digested with NarI to generate a 51-bp AlwNI-NarI fragment containing the incorrect CD43TM sequence and a 954-bp NarI-AlwNI fragment containing the CD43TM-PSGL-1 junction, cytoplasmic domain of PSGL-1, and pZeoSV sequence. The 50-bp fragment was discarded and the 954-bp NarI-AlwNI fragment was ligated to the 55-bp AlwNI-NarI from CD43/ pBKEF. Due to the nature of the AlwNI cut site, the 2 fragments were only complementary at the NarI site, so mis-ligation at the AlwNI sites did not occur. After a 12-h ligation period, the 3674-bp AlwNI-AlwNI fragment was added to the ligation mixture and ligation was allowed to continue for several hours. Sequence analysis of the resultant construct confirmed that it contained the CD43 TMD properly spliced into the corresponding region of PSGL-1. For all constructs, the regions that were manipulated by PCR or restriction digestion were confirmed to be properly mutated and ligated by sequence analysis of both strands.
Proteins-sPS, a recombinant, soluble form of P-selectin truncated after the ninth consensus repeat, was purified by immunoaffinity chromatography of conditioned medium from stably transfected human 293 kidney cells as described previously (31). Membrane P-selectin (mPS) was immunoaffinity purified from human platelets as described previously (32).
For the rolling adhesion assays, stable cell lines were made by transfecting CHOdhfr Ϫ cells with either wt-PSGL-1/pZeoSV or C320S-PSGL-1/pZeoSV. Selection medium (low salt DME containing 10% FBS, 1 ϫ HT, 1 ϫ NEAA, 2 mM Gln, and 250 g/ml Zeocin) was added 48 h after transfection. Clones were chosen for the rolling adhesion assay based on similar levels of expression of wt-and C320S-PSGL-1 as assessed by flow cytometric staining with PL2. The stable cell lines were transiently transfected with C2GnT/pcDNA3 and FucT-III/pRc/ RSV 48 h before the rolling adhesion assay was performed.
Preparation of Cell Extracts and SDS-PAGE-Cells were washed three times in HBSS, 0.1% NaN 3 and pelleted. Extracts were prepared by resuspending the pellet in an equal volume of 2ϫ solubilization buffer (40 mM MOPS, pH 7.5, 200 mM NaCl, 4 mM CaCl 2 , 4 mM MgCl 2 , 1% protein grade Triton X-100, 20 g/ml leupeptin, 20 g/ml aprotinin, 8 g/ml pepstatin A, 10 mM benzamidine, 0.04% NaN 3 ). Phenylmethylsulfonyl fluoride was added to 1 mM. The mixture was incubated on ice for 45 min and then sonicated in an ice bath sonicator for 20 min. A final sonication step was done for 1-2 s using a Branson Sonifier/Cell Disruptor 185. The extracts were then spun at 16,000 ϫ g for 10 min, and the supernatant was transferred to a fresh tube. The protein concentration was determined using a micro BCA assay (Bio-Rad). 250 g of each cell extract was boiled in sample buffer with or without 5% ␤-mercaptoethanol (33) and loaded onto SDS-polyacrylamide gels. Electrophoresis was carried out at 10 mA until samples reached the separating gel, at which time the current was increased to 30 mA.
Western Blotting-After SDS-PAGE, proteins were transferred overnight at 20 volts to nitrocellulose membranes (0.2 m) using the Bio-Rad minigel blotting apparatus. Membranes were stained for 2 min with Ponceau S to determine the quality of the transfer, and the stacking gel interface was marked. Blots were destained in TBS (20 mM Tris, 300 mM NaCl, pH 7.5) and blocked with 5% non-fat dry milk in TTBS (TBS ϩ 0.05% Tween 20) for 1 h. Blots were probed with 1 g/ml PL1 in 0.5% non-fat dry milk/TTBS. The secondary antibody was peroxidase-conjugated sheep anti-mouse IgG. Antibody incubations and washes were performed according to ECL manufacturer's instructions. After addition of the ECL substrate, the blots were exposed to Bio-Max MR film. For the cross-linking assays, the blots were stripped according to instructions supplied with the ECL substrate and reprobed with the anti-CD43 mAb H5H5 (neat).
Flow Cytometry-Flow cytometric analyses were performed as described previously (21). In this assay, sPS is biotinylated on attached oligosaccharides using biotin hydrazide. Biotinylation of sPS with amine reactive reagents interferes with its PSGL-1 binding activity, whereas biotinylation of the carbohydrate on sPS does not affect its ability to bind PSGL-1.
Rolling Adhesion Assays-The ability of CHOdhfr Ϫ cells expressing recombinant PSGL-1, FucT-III, and C2GnT to interact with P-selectin under shear stress was analyzed. A dual chamber, parallel-plate flow apparatus was used to simulate shear stresses of 0.5 and 1 dyn/cm 2 (15,34). The flow chamber consisted of a 35-mm tissue culture dish coated with mPS. Dishes were coated with equivalent concentrations of mPS as described previously (35), resulting in site densities of approximately 200 sites/m 2 . The CHO cells were resuspended at 5 ϫ 10 5 cells/ml in HBSS, 0.5% human serum albumin and incubated on ice with buffer alone or with buffer containing 10 g/ml of the mAb PL1 10 min prior to perfusion. The cells were perfused through the flow chamber at a shear stress of 0.5 or 1 dyn/cm 2 . Rolling was allowed to equilibrate for 4 min prior to data acquisition. The number of rolling CHO cells was quantified by digitizing image frames and determining the number of rolling CHO cells in six selected fields using a ϫ10 objective. The experiments were performed at 4°C to minimize clumping of the resuspended CHO cells.
Cross-linking Assays-Recombinant CHO cells were detached using 0.02% EDTA. HL60 and recombinant CHO cells were washed three times in ice-cold HBSS. Cells were resuspended at a concentration of 2.3 ϫ 10 6 cells/ml in HBSS, and 1 ml was aliquoted into two Eppendorf tubes for each cell line. A stock solution of 10 mM BS 3 in HBSS was made just prior to use and then added to a final concentration of 2 mM to each tube. Control tubes received HBSS without BS 3 . The tubes were kept on ice for up to 90 min. The reaction was quenched by the addition of 1 M Tris/HBSS, pH 7.5. After an additional 10-min incubation on ice, cells were pelleted and the supernatant was aspirated. Cell extracts were prepared and 250 g of each extract was loaded onto 5% SDS-PAGE gels. After electrophoresis, the proteins were transferred to nitrocellulose and the membranes were probed with mAb PL1. After addition of the ECL substrate, the blots were exposed to Bio-Max MR film. The blots were then stripped according to instructions supplied with the ECL substrate and reprobed with the anti-CD43 mAb H5H5 (neat).
Efficiency of cross-linking was determined from densitometric scans of the ECL films. The percent of monomer versus dimer was determined for each sample by the following formula: percent of monomer (or dimer) ϭ peak area for monomer (or dimer)/(peak area for monomer ϩ peak area for dimer). It is common to see a residual 240-kDa PSGL-1 band under reducing conditions in SDS-PAGE (11,14,15). To correct for dimeric PSGL-1 that did not represent chemical cross-linking, the percent of dimeric PSGL-1 present in control (mock-treated) samples was subtracted from the percent of dimeric PSGL-1 present in BS 3treated samples under reducing SDS-PAGE conditions. The corrected values represent the efficiency of cross-linking.
Purification of NH 2 -terminal Fragments of PSGL-1-The metabolically labeled cells were collected and washed five times with HBSS. 7CN-CHO cells were detached from the tissue culture plates with EDTA prior to washing them. The cells were treated with 75 M trypsin/ HBSS for 1 h at 37°C in a 5% CO 2 environment. After trypsinization, the cells were pelleted at 500 ϫ g and the supernatant was removed to a fresh tube. The cell pellet was washed once with HBSS and the wash supernatant was combined with the first tryptic supernatant. The supernatant was filtered using a Maxi-Spin 0.45-m cellulose-acetate filtration device to remove residual cells. One EDTA-free, Mini-cØmplete tablet (protease inhibitor mixture) and soybean trypsin inhibitor in 3-fold excess over the mass of trypsin used to treat the cells was added to the filtrate. After incubation overnight at 4°C, the tryptic supernatant was subjected to immunoaffinity chromatography. The supernatant was applied to a blocked Emphaze pre-column (2-ml bed volume) hooked in-line to a 5 mg/ml PL1-Emphaze column (1-ml bed volume). After the sample was completely loaded, the column system was washed with several column volumes of equilibration buffer (10 mM phosphate, pH 6.8, 300 mM NaCl, 0.02% NaN 3 ). The pre-column was then disengaged, and the PL1 column was washed extensively with equilibration buffer. Bound tryptic fragments were eluted with 100 mM triethylamine, pH 11.5. Each 1-ml fraction was immediately neutralized with 200 l of 1 M NaH 2 PO 4 already present in the collection tube prior to elution. The eluted material was pooled and subjected to concentration and buffer exchange using Centriplus 3 concentrators. An aliquot of the material was analyzed by electrophoresis on a 5-22% SDS-polyacrylamide gel, followed by fluorography.
Partial Characterization of ⌬N-PSGL-1 Tryptic Fragments-Sizing of the PL1-bound material was performed by gel filtration on a Pharmacia FPLC system using a Superose TM 6 HR 10/30 column equilibrated in 10 mM Tris-HCl, 150 mM NaCl, 6 M guanidine HCl, 0.02% NaN 3 , pH 7.5. The flow rate was 0.25 ml/min and 1-min fractions were collected. Elution was monitored by scintillation counting of 30-l aliquots from each fraction. A comparative analysis of the hydrophobic character of the fragments was done using reverse phase chromatography on a Beckman HPLC system using a Vydac C 18 column (4.6 ϫ 250 mm) with a 5-35% acetonitrile, 0.1% trifluoroacetic acid gradient. The flow rate was 1 ml/min and 1-min fractions were collected. Elution was monitored by scintillation counting of 150-l aliquots from each fraction. A semi-synthetic glycosulfopeptide termed GSP-6 was used as a standard. GSP-6 is a small glycosulfopeptide peptide (4,159.2 daltons) containing the primary sequence of the extreme NH 2 terminus of human PSGL-1 and also containing three Tyr-SO 3 Ϫ residues and a single Thr residue with a monosialylated, monofucosylated, core-2 O-glycan that expresses the sialyl Lewis x antigen (24). GSP-6 has been shown to bind to P-selectin with a relatively high affinity equivalent to that of native, dimeric PSGL-1 purified from human neutrophils (20). A comparative analysis to determine the anionic character of the tryptic fragments was also performed using GSP-6 as a standard. Anion exchange chromatography was done on a Pharmacia SMART TM System using a Mini Q TM PC 3.2/3 column in 20 mM Bis-Tris propane, pH 7, with a 0 -2 M NaCl gradient. The flow rate was 0.5 ml/min and 0.25-ml fractions were collected. Elution was monitored by scintillation counting of each fraction.

C320S-PSGL-1 Migrates as a Monomeric Species in SDS-PAGE-
The only cysteine in the extracellular domain of the 16-repeat form of PSGL-1 resides near the predicted transmembrane domain at position 320 (Fig. 1A). The model put forward suggests that PSGL-1 covalently dimerizes through disulfide bonding via Cys 320 . This residue is numbered as Cys 310 in the 15-repeat form of PSGL-1 originally reported (9). Two groups reported that substitution of Cys 320 with Ala prevented covalent dimerization of PSGL-1 (18,19). In assessing the role of this Cys residue for covalent dimerization of PSGL-1, we mutated the Cys to Ser, which is a more conserv-ative substitution (C320S-PSGL-1) (Fig. 1B). Cell extracts from HL60 cells, which constitutively express PSGL-1, and from CHOdhfr Ϫ cells expressing recombinant FucT-III, C2GnT, and either wt-PSGL-1 or C320S-PSGL-1, were electrophoresed on 7.5% SDS-polyacrylamide gels. Following electrophoresis, the proteins were transferred to nitrocellulose membranes and probed with the anti-PSGL-1 mAb PL1, which recognizes a peptide epitope in the extreme NH 2 -terminal domain of the molecule (Fig. 1A). As expected, under nonreducing conditions, HL60-and wt-recombinant PSGL-1 both migrated with a molecular mass of ϳ240 kDa (Fig. 2). C320S-PSGL-1, on the other hand, migrated as a monomeric species with a molecular mass of ϳ120 kDa. Under reducing conditions, HL60-, wt-, and C320S-PSGL-1 all migrated with a molecular mass of ϳ120 kDa. Nontransfected CHOdhfr Ϫ cells did not express PSGL-1 and were not stained by PL1. These results confirm that Cys 320 is responsible for covalent dimerization of PSGL-1.
C320S-and C320A-PSGL-1 Are Expressed at the Cell Surface and Bind to Fluid-phase P-selectin-To assess the effect of the Cys to Ser mutation on cell surface expression of PSGL-1 and its binding to P-selectin, CHOdhfr Ϫ cells stably expressing FucT-VII and C2GnT were either mock transfected or transfected with wt-or C320S-PSGL-1 and stable clones were isolated. The anti-PSGL-1 mAb PL2 was used to determine the level of expression of C320S-PSGL-1 compared with wt-PSGL-1. Although the level of expression of C320S-PSGL-1 is lower than that of wt-PSGL-1 in the clones chosen for these studies (Fig. 3, top panel), transiently transfected cells repeatedly had equivalent levels of expression as assessed by PL2 staining, indicating that the Cys 3 Ser mutation did not alter surface expression of C320S-PSGL-1 (data not shown). As expected, mock transfected cells were not stained by PL2. Since the cells were stably transfected with FucT-VII, the cells were also monitored for the presence of the sLe x epitope by staining with CSLEX-1. All cells expressing FucT-VII were positive for the sLe x epitope (Fig. 3, middle panel).
To determine if C320S-PSGL-1 retained the ability to bind to P-selectin, a fluid-phase binding experiment was conducted using flow cytometric analysis with sPS. Both wt-and C320Stransfected cells bound sPS, and the binding was inhibited by the anti-human PSGL-1 mAb PL1 (Fig. 3, bottom panel). Mock transfected cells did not bind sPS. Consistent with the levels of PSGL-1 expression, the wt-PSGL-1 clone was stained more brightly by P-selectin than the C320S-PSGL-1 clone. In transiently transfected populations, the level of PSGL-1 expression was equivalent between the two constructs and the level of P-selectin binding was also equivalent (data not shown). Similar results were obtained with transfection of CHOdhfr Ϫ cells expressing FucT-III and C2GnT with wt-and C320S-PSGL-1 (data not shown). These data demonstrate that C320S-PSGL-1 is expressed at the cell surface and binds to fluid-phase P-selectin.
These results differ from those recently reported by two separate groups, who found that mutation of the extracellular Cys in PSGL-1 to Ala (C320A or the equivalent C310A mutation) resulted in loss of P-selectin binding activity in assays employing flow cytometry (18, 19), low-shear cell adhesion, or rolling cell adhesion (19). Therefore, we also generated the C320A construct to address the possibility that the Cys 3 Ala mutation was more severe in its effect on PSGL-1 binding activity than the Cys 3 Ser alteration (Fig. 1B). However, cells expressing C320A-PSGL-1, FucT-VII, and C2GnT also bound sPS in flow cytometric assays, and sPS binding was blocked by PL1 (Fig. 3). These results show that replacement of Cys 320 with either Ser or Ala does not prevent binding of PSGL-1 to fluid-phase P-selectin.
It has been reported that the homologous mutant C310A-PSGL-1, when co-expressed in human 293 cells with FucT-III, did not bind to platelet-derived P-selectin (18). Since the 293 cell system chosen to study the C310A mutation was different from our CHO cell system, we also expressed C320S-PSGL-1 in human 293 cells. The 293 cells presumably express endogenous C2GnT, since only co-expression of an ␣1,3-fucosyltransferase with PSGL-1 is needed for recognition by P-selectin (18). Human 293 cells were transiently transfected with either wt-or C320S-PSGL-1 and FucT-III. Mock transfectants received plasmids lacking the insert. Cells expressing either wt-or C320S-PSGL-1, but not mock transfectants, bound fluid-phase sPS and the binding was inhibited by the anti-P-selectin mAb G1 (data not shown). These results further strengthen the conclusion that covalent dimerization of PSGL-1 on cell surfaces is not required for binding fluid-phase P-selectin.
C320S-PSGL-1 Binds to P-selectin Under Shear Stress-To further assess the ability of C320S-PSGL-1 to bind P-selectin, we measured cellular interactions under shear conditions. CHOdhfr Ϫ cells stably expressing similar levels of either wt-or C320S-PSGL-1 were transiently transfected with FucT-III and C2GnT. In mock transfectants, the clone stably expressing wt-PSGL-1 was transiently transfected with pRC-RSV and pcDNA3. Forty-eight hours after transfection, the cells were perfused over mPS-coated plates. Cells expressing either wt-or C320S-PSGL-1 attached to and rolled on mPS at shear stresses of 0.5 dyn/cm 2 (Fig. 4A) or 1.0 dyn/cm 2 (data not shown). Attachment of the cells was inhibited by PL1 (Fig. 4A), but not by PL2 (data not shown). Mock-transfected cells did not attach. CHOdhfr Ϫ cells expressing FucT-III, C2GnT, and either wt-or C320S-PSGL-1 detached from the P-selectin surface in a similar fashion in response to increasing shear stress (Fig. 4B). Since the cells were transiently transfected with FucT-III and C2GnT, the cells were monitored for expression of PSGL-1, sLe x expression, and ability to bind to fluid-phase P-selectin by flow cytometry. Cells expressing either wt-or C320S-PSGL-1 stained similarly for all three parameters, although because of the transient transfection experiment, expression of sLe x was lower than for stably transfected cells. Staining of the cells by sPS was inhibited by the anti-P-selectin mAb G1 (Fig. 4C). These data demonstrate that cell surface PSGL-1 lacking intermolecular disulfide bonds interacts with immobilized P-selectin under shear stress.
C320S-PSGL-1 or C320A-PSGL-1 Self-associates in the Plasma Membrane-As shown above, C320S-PSGL-1 migrates as a monomeric species in SDS-PAGE. To test whether it might noncovalently dimerize, HL60 cells and cells stably expressing either wt-or C320S-PSGL-1 in the C2GnT, FucT-III CHOdhfr Ϫ environment were treated for 90 min with the homobifunctional cross-linker BS 3 , then lysed and subjected to immunoblotting with the anti-PSGL-1 mAb PL1. BS 3 reacts with primary amines; the sulfate moiety on BS 3 prevents it from crossing the lipid bilayer, so only extracellular lysine residues and the NH 2 -terminal residue are available for cross-linking. The locations of Lys available for cross-linking in PSGL-1 and in CD43, a membrane sialomucin with a similar size to PSGL-1, are indicated in Fig. 5A. PSGL-1 has three to four Lys available in its extracellular domain. In leukocytes, the re- moval of the propeptide, which contains one Lys, from PSGL-1, leaves three Lys available for cross-linking. The propeptide may not be efficiently cleaved by the endogenous PACE homolog in CHO cells (27,36). The Lys in PSGL-1 (excluding the propeptide Lys) are localized near the transmembrane domain. CD43 contains five Lys residues that are distributed throughout the extracellular domain.
Both HL60-and wt-CHOdhfr Ϫ cell-derived PSGL-1 migrated with a molecular mass of ϳ240 kDa under nonreducing conditions after cells were treated with BS 3 or mock treated (Fig.  5B). Mock-treated C320S-PSGL-1 migrated as a monomeric species with a molecular mass of ϳ120 kDa. However, after treatment with BS 3 , a ϳ240 kDa species was generated (Fig.  5B). No higher oligomeric species were observed, although a small amount of very high molecular weight oligomers might escape detection in these assays. Under reducing conditions, there was an increase in the amount of PSGL-1 migrating as the dimeric ϳ240 kDa species in HL60-and wt-PSGL-1 in samples treated with BS 3 . This ϳ240-kDa species was retained in the BS 3 -treated C320S-PSGL-1 under reducing conditions, since the cross-linker is not reducible (Fig. 5B). The gel films from the chemiluminescence-based Western blots were scanned, and the efficiency of cross-linking was determined. Wild-type PSGL-1, which has a disulfide bond, served as an internal control for determining the efficiency in which C320S-PSGL-1 is cross-linked. In this experiment, ϳ29% of the wt-PSGL-1 was cross-linked by BS 3 . In comparison, ϳ22% of the C320S-PSGL-1 molecules were cross-linked, indicating that noncovalent association is efficient in this mutated form of PSGL-1. As a control, the blots were stripped of PL1 and probed with an antibody directed against CD43. The migration pattern of CD43 was not altered upon treatment with BS 3 , indicating that CD43 is not cross-linked under the conditions that crosslink PSGL-1 (Fig. 5C).
To determine if the less conservative Cys 320 3 Ala mutation affected noncovalent association of PSGL-1, CHOdhfr Ϫ cells expressing either wt-PSGL-1 or C320A-PSGL-1, FucT-VII, and C2GnT were also treated with BS 3 (Fig. 5D). As seen with C320S-PSGL-1, a 240-kDa C320A-PSGL-1 species was produced after treatment of the cells with BS 3 . It should be noted that in the course of these studies we used both ␣1,3/4-fucosyl-transferase FucT-III and the ␣1,3-fucosyltransferase FucT-VII. All our studies show that these enzymes provide equivalent results. In the cross-linking experiments, the clone expressing wt-PSGL-1 in the C320S-PSGL-1 study shown in Fig. 5, B and C, expresses FucT-III, whereas the clone expressing wt-PSGL-1 in the C320A-PSGL-1 study shown in Fig. 5D expresses FucT-VII. The wt-PSGL-1, C2GnT, and FucT-VII clones contain three distinct species of PL1-reactive material that migrate slightly slower than the Cys mutants (Fig. 5D). These species are presumed to be degraded and/or differentially glycosylated subunits of PSGL-1. This artifact is clonedependent and possibly arises due to inherent differences between clones, or it could be due to overexpression of any of the components added to the recombinant system. Furthermore, the smear seen in the BS 3 -treated wt-PSGL-1 sample (Fig. 5D) should not be taken for higher oligomers resulting from crosslinking, since cross-links formed by BS 3 are not reducible and higher oligomers are not apparent under reducing conditions. In time course measurements, we found that cross-linking of wt and mutated forms of PSGL-1 is extremely rapid, and is apparent within the first minute after addition of BS 3 (data not shown). Furthermore, these experiments were done with a relatively high concentration of BS 3 (2 mM), which is the concentration recommended for cell surface cross-linking. Thus, neither the C320A nor the C320S mutation inhibits self-association of PSGL-1. These results demonstrate that PSGL-1 non-covalently associates in the absence of a disulfide bond.
CD43TMD-PSGL-1 Does Not Self-associate in the Plasma Membrane but Retains the Ability to Bind P-selectin-Since C320S-and C320A-PSGL-1 continued to form noncovalent associated or dimeric forms at the cell surface, we pursued studies to eliminate the region responsible for self-association in order to more fully address the role of dimerization in P-selectin binding. In early studies, we truncated C320S-PSGL-1 after the first 5 residues of the cytoplasmic domain to generate the tail-less construct C320STL-PSGL-1. This construct was crosslinkable by BS 3 and bound P-selectin (data not shown), so we proceeded to the transmembrane domain. Since CD43 is a monomeric sialomucin that is not cross-linkable by BS 3 (Fig.  5C), we decided to replace the transmembrane domain in PSGL-1 with that of CD43 to generate the chimeric CD43TMD- PSGL-1 (Fig. 5A). Since Cys 320 is the second residue in the putative transmembrane, replacement of the TMD automatically eliminated the disulfide bond in PSGL-1. CHOdhfr Ϫ cells stably expressing C2GnT and FucT-VII were transfected with CD43TMD-PSGL-1 and subjected to cross-linking analysis. Cells expressing C320A-PSGL-1, C2GnT, and FucT-VII were used as a positive control. Upon treatment with BS 3 , C320A-PSGL-1 was cross-linked into a ϳ240-kDa species while CD43TMD-PSGL-1 remained monomeric at ϳ120 kDa (Fig.  6A). This result indicated that the transmembrane domain is required for self-association of PSGL-1 subunits. The Western blot shown in Fig. 6A indicates that the level of expression of CD43TMD-PSGL-1 is lower than that of C320A-PSGL-1. However, upon overexposure of the Western blot film in Fig. 6A we still did not observe a ϳ240-kDa species after attempted crosslinking of CD43TMD-PSGL-1 (data not shown). Furthermore, cross-linking analysis of cells expressing low levels of C320Aand C320S-PSGL-1 revealed that they continue to self-associate at a variety of cell surface expression levels (data not shown). The cells were analyzed by flow cytometry to ascertain the levels of expression for PSGL-1 and FucT-VII epitopes by PL2 and CSLEX-1 staining, respectively (Fig. 6B, top and middle panel). The cells expressing CD43TMD-PSGL-1 were not yet clonal, resulting in the lower expression level seen in the Western blot. The major population has an expression level similar to that of the C320A-PSGL-1 clone (Fig. 6B, top panel). FucT-VII activity was apparent by CSLEX-1 staining (Fig. 6B,  middle panel). The ability of cells expressing the monomeric CD43TMD-PSGL-1 to bind P-selectin was analyzed by flow cytometry. Fluid-phase, soluble P-selectin bound to cells expressing CD43TMD-PSGL-1. The P-selectin staining of cells expressing CD43TMD-PSGL-1 was broader than that seen for cells expressing C320A-PSGL-1, presumably due to the mixed nature of the CD43TMD-PSGL-1 cell population. Binding of P-selectin to cells expressing CD43TMD-PSGL-1 was inhibited by the anti-PSGL-1 mAb PL1 (although in the experiments shown inhibition was incomplete due to sub-saturating levels of PL1) (Fig. 6B, bottom panel). These results further demonstrate that a covalent dimer of PSGL-1 is not necessary for binding to P-selectin. Furthermore, CD43TMD-PSGL-1, which FIG. 5. C320S-or C320A-PSGL-1 self-associates as a noncovalent dimer in the absence of the disulfide bond. A, primary structural model of PSGL-1 and CD43 and chimeric forms of PSGL-1 demonstrating the location of lysine residues. Lysine residues are highlighted to show that the domain swap between PSGL-1 and CD43 does not eliminate lysine residues reactive to the cross-linking reagent. The sequence of the transmembrane domain plus the three flanking residues is shown for CD43, PSGL-1, and CD43TMD-PSGL-1. ProP, propeptide; CD, cytoplasmic domain; K, lysine residue. Since the cross-linking reagent BS 3 does not cross cell membranes, only those lysine residues in the extracellular domain are available for cross-linking. B, HL60 cells and CHOdhfr Ϫ cells expressing either wt-or C320S-PSGL-1, C2GnT, and FucT-III were incubated in the presence (ϩ) or absence (Ϫ) of BS 3 , as described under "Experimental Procedures." Following treatment, the samples were quenched with 1 M Tris/HBSS, pH 7.5, and cell extracts were prepared. 250 g of each cell extract was loaded onto 5% SDS-polyacrylamide gels under nonreducing (NR) and reducing (R) conditions. Western blotting analysis was performed using PL1. The gel is shown from the stacking gel interface. C, the Western blots in B were stripped as described under "Experimental Procedures" and re-probed with the anti-CD43 mAb H5H5. D, CHOdhfr Ϫ cells expressing either wt-or C320A-PSGL-1, C2GnT, and FucT-VII were treated as in B. SGI, stacking gel interface. is unable to be chemically cross-linked and is presumed to be monomeric, is able to bind to P-selectin. Finally, the transmembrane domain of PSGL-1 is critical for the noncovalent dimerization of PSGL-1 subunits observed in our studies.

Isolation of [ 35 S]Sulfate-labeled NH 2 -terminal Tryptic
Fragments of PSGL-1 from HL60 Cells-To further address the issue of whether dimerization of PSGL-1 is required to bind P-selectin, we investigated whether small monovalent fragments of PSGL-1 could be recognized with high affinity by P-selectin. HL60 cells were metabolically radiolabeled with [ 35 S]sulfate for 48 h, which efficiently labels Tyr-SO 3 Ϫ residues at positions 46, 48, and 51 in the extreme NH 2 terminus of PSGL-1 (11). The cells were washed and then treated with trypsin for 1 h at 37°C. The cells were then pelleted and the tryptic supernatant was collected. After addition of trypsin inhibitors, the tryptic supernatant was applied to a PL1 affinity column. PL1 recognizes a peptide spanning residues 49 -62 in the extreme NH 2 terminus of human PSGL-1 (Fig. 1A), whereas mAb PL2 recognizes a region within the decameric repeats between residues 188 -235 (27) (Fig. 1A). After exten-sive washing of the PL1 affinity column, bound material was eluted with 100 mM triethanolamine, pH 11.5 (Fig. 7A). SDS-PAGE analysis revealed two 35 S-labeled NH 2 -terminal fragments of PSGL-1 migrating as ϳ25 and ϳ100 kDa species (Fig.  7A, inset). Both PL1 and PL2 immunoprecipitated the 100-kDa fragment, whereas only PL1 precipitated the 25-kDa fragment (data not shown). Upon N-glycanase treatment of the 100-kDa fragment and re-treatment with trypsin, more 25-kDa fragment was generated (data not shown). These data demonstrate that both fragments contain the extreme NH 2 -terminal domain of PSGL-1 recognizable by PL1 and that the two fragments are not associated and can be independently immunoprecipitated. NH 2 -terminal tryptic fragments of PSGL-1 were then subjected to P-selectin affinity chromatography. About 20% of the radiolabeled material bound to P-selectin (Fig. 7B). Upon rechromatography of the P-selectin unbound material on the P-selectin affinity column, all of the material remained unbound, indicating that the first chromatography was quantitative (data not shown). The P-selectin-bound material from HL60 cells contained both the ϳ25and 100-kDa labeled spe- FIG. 6. CD43TMD-PSGL-1 does not self-associate as a noncovalent dimer but retains the ability to bind P-selectin. A, CHOdhfr Ϫ cells stably expressing recombinant C2GnT and FucT-VII were stably transfected with cDNAs encoding either C320A-or CD43TMD-PSGL-1 as described under "Experimental Procedures." The cells were incubated in the presence (ϩ) or absence (Ϫ) of BS 3 , as described under "Experimental Procedures." Following treatment, the samples were quenched with 1 M Tris/HBSS, pH 7.5, and cell extracts were prepared. 250 g of each cell extract was loaded onto a 5% SDS-polyacrylamide gel under reducing (R) conditions. Western blotting analysis was performed using PL1. B, the cells were stained with the anti-human PSGL-1 mAb PL2 to assess the level of PSGL-1 expression (top panel); anti-sLe x mAb CSLEX-1 to demonstrate functionality of FucT-VII (middle panel); and sPS to determine which constructs bound fluidphase P-selectin (bottom panel). The specificity of sPS binding was demonstrated by inhibition with the anti-human PSGL-1 mAb PL1. cies (Fig. 7B, inset). These results demonstrate that soluble tryptic fragments of PSGL-1 containing the NH 2 -terminal domain bind with high affinity to immobilized P-selectin.

Isolation and characterization of [ 35 S]Sulfate-labeled NH 2terminal Tryptic Fragments of Recombinant PSGL-1 from CHO
Cells-Human PSGL-1 contains three potential N-glycosylation sites at Asn 65 , Asn 111 , and Asn 302 (Fig. 1, A and B). The site at Asn 65 flanks the trypsin cleavage site at Arg 64 (Fig. 1A), possibly causing wt PSGL-1 to be insensitive to trypsin at that site due to steric hindrance. We reasoned that a recombinant form of PSGL-1 lacking the N-glycosylation sites might make PSGL-1 more sensitive to trypsin at the Arg 64 site, allowing us to purify smaller fragments. The three Asn-X-Ser/Thr sequons for N-glycosylation of PSGL-1 were deleted by mutating the third residue of the sequon (Ser or Thr) to Ala (Fig. 1B). This mutant (T67A, S113A, S304A; referred to as ⌬N-PSGL-1) was used to create a stable CHOdhfr Ϫ cell line expressing ⌬N-PSGL-1, C2GnT, and FucT-VII (referred to as 7CN-CHO cells hereafter). The 7CN-CHO cells were metabolically labeled with [ 35 S]sulfate for 48 h and treated with trypsin following the protocol established for HL60 cells. The tryptic supernatant was applied to a PL1 affinity column. After extensive washing, the column was eluted with triethanolamine (Fig. 8A). SDS-PAGE analysis of the immunoaffinity purified material revealed 3 classes of 35 S-labeled NH 2 -terminal fragments: a minor ϳ90 kDa band, a smear of molecular mass around 30-kDa and smaller species which migrated with molecular masses Ͻ10-kDa (Fig. 8A, inset). These species were subjected to gel filtration on a Superose-6 column and the fractions representing separate peaks were pooled (Fig. 8B). Three pools were made: fractions 38 -43 (pool 1), 44 -50 (pool 2), and 51-68 (pool 3). The major peak is represented by pool 3. SDS-PAGE analysis revealed that pool 1 contained the ϳ90-kDa band and the 30-kDa smear. Pool 3 contained ϳ3 fragments Ͻ10 kDa (Fig.  8B, inset). Pool 3 was then subjected to P-selectin affinity chromatography. All of the PL1-bound, ⌬N tryptic fragments in pool 3 also bound to P-selectin (Fig. 8C). The ⌬N tryptic fragments bound by P-selectin were next subjected to denaturing FPLC gel filtration chromatography. The material eluted as a single symmetrical species of ϳ9 kDa (Fig. 8D).
While these studies were in progress, we succeeded in synthesizing a glycosulfopeptide termed GSP-6, which contains the first 22 residues of mature, human PSGL-1 (residues 42-63 of unprocessed PSGL-1) (24). To satisfy the presumed basic components needed for P-selectin binding, a core-2 based Oglycan expressing sLe x was enzymatically generated at Thr 57 . The glycopeptide was then enzymatically sulfated at Tyr 46 , Tyr 48 , and Tyr 51 to generate the fully modified GSP-6, whose structure is shown in Fig. 9A. In our recent study (24), GSP-6 was shown to bind P-selectin with relatively high affinity (K d was ϳ350 nM), which is equivalent to that of native, dimeric PSGL-1 purified from human neutrophils (20). Since the components of GSP-6 are precisely defined, we used it as a standard for comparison in reverse-phase and anion-exchange chromatography to more fully characterize the recombinant ⌬N-PSGL-1 tryptic fragments. Reverse phase chromatography of the ⌬N fragments revealed that three peaks were readily separable in a 5-35% acetonitrile gradient (Fig. 9A). The most hydrophilic peak, ⌬N-I, did not bind to the C18 column, while ⌬N-II and ⌬N-III eluted at 50.5 and 54.5 min, respectively (Fig.  9A). In this gradient system, GSP-6 eluted as 2 species at 52.5 and 56.5 min. The 52.5-min peak represents an oxidized form of GSP-6, as previously documented (24). Since oxidation of GSP-6 results in a 4-min decrease in retention time, it is possible that ⌬N-II is the oxidation product of ⌬N-III. Since addition of a single monosaccharide was sufficient to alter the elution of the intermediates in the synthesis of GSP-6 by as much as 1.8 min in reverse phase chromatography (24), the 2-min difference between GSP-6 and ⌬N-III suggests that the structural differences are small. ⌬N-I may represent structural heterogeneity in oligosaccharide structure or polypeptide length. Anion-exchange chromatography of the ⌬N fragments resulted in partitioning of the sample into 3 peaks. ⌬N-1 coeluted with GSP-6 at a salt concentration of 0.67M. ⌬N-2 and ⌬N-3 eluted at concentrations of 0.93 and 1.1 M NaCl, respectively. GSP-6 has potential, net overall charge of Ϫ10. The estimated charges for ⌬N-1, ⌬N-2, and ⌬N-3 are Ϫ10, Ϫ14, and Ϫ16.5, respectively. These results indicate that at least one of the tryptic fragments is structurally similar to GSP-6. Furthermore, the amino-terminal fragments appear to behave as monomeric species since the peak co-migrating with GSP-6 repre- ]sulfate-labeled HL60 cells were treated with trypsin as described under "Experimental Procedures." The supernatant was collected, and labeled NH 2 -terminal fragments were purified on a PL1 immunoaffinity column. Bound material was eluted with 100 mM triethylamine, pH 11.5. Inset, the PL1-bound material was analyzed on a 5-22% gradient SDS-PAGE under nonreducing conditions, followed by autoradiography. B, the PL1-bound material was loaded onto a 5 mg/ml soluble P-selectin affinity column. Bound material was eluted with 5 mM EDTA. sents the smallest possible charged species (Ϫ10), and a species carrying a Ϫ20 charge is not apparent. DISCUSSION This study demonstrates that PSGL-1 does not require covalent dimerization to bind to P-selectin. Although the derivatives in which the disulfide bond was eliminated by site-directed mutagenesis migrated as monomers in SDS-PAGE, cross-linking studies established that C320S-or C320A-PSGL-1 self-associate as dimers at the cell surface. Replacement of the PSGL-1 TMD with that of CD43 eliminated crosslinking, suggesting that the TMD is critical for this noncovalent dimerization. Furthermore, the chimeric CD43TMD-PSGL-1 retained the ability to bind P-selectin, indicating that neither covalent nor noncovalent dimerization of PSGL-1 is required for P-selectin binding. Trypsin treatment of cells expressing PSGL-1 generated small NH 2 -terminal fragments of PSGL-1 that retained the ability to bind P-selectin. These results demonstrate that a small glycosulfopeptide derived from the extreme NH 2 terminus of full-length PSGL-1 is sufficient for binding to P-selectin.
Our results are consistent with several lines of evidence that the NH 2 -terminal region of PSGL-1 is required for binding to P-selectin (11-13, 15, 17, 21, 27, 37). These prior results demonstrated that the extreme NH 2 terminus of PSGL-1 is necessary for binding to P-selectin, but they did not address whether ]sulfate-labeled 7CN-CHO cells were treated with trypsin as described under "Experimental Procedures." The supernatant was collected and labeled NH 2 -terminal fragments were purified on a PL1 immunoaffinity column. Bound material was eluted with 100 mM triethylamine, pH 11.5. Inset, the PL1-bound material was analyzed on a 5-22% gradient SDS-polyacrylamide gel under nonreducing conditions. B, the PL1 bound material was fractionated by gel filtration on a Superose TM 6 HR 10/30 column equilibrated in 10 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 0.02% NaN 3 with a flow rate of 0.25 ml/min. Three separate pools were collected: pools 1 and 2 representing fractions 38 -43 and 44 -50, respectively, and pool 3, representing the major peak and containing fractions 51-68. Inset, pools 1 and 2 and an aliquot of pool 3 were analyzed on a 5-22% gradient SDS-polyacrylamide gel under nonreducing conditions, followed by autoradiography. C, the PL1-bound material from pool 3 was loaded onto a 5 mg/ml soluble P-selectin affinity column. Bound material was eluted with 5 mM EDTA. D, the [ 35 S]sulfate-labeled, sPS-bound NH 2 -terminal fragment of ⌬N-PSGL-1 was analyzed by denaturing gel filtration on a Superose TM 6 HR 10/30 column equilibrated in 6 M guanidine-HCl, as described under "Experimental Procedures." The radioactivity in each fraction collected was determined. The elution position of sizing markers is indicated. A graphical estimation of the molecular mass of the ϳ9.0 kDa fragment is shown in the inset and indicated by the arrow.
the NH 2 -terminal domain must be dimeric to bind. Some prior studies relied on bivalent Ig-chimeras (12) or membrane-bound chimeras of unknown valency (13). The issue of valency is especially important, since two separate groups have reported that prevention of disulfide-bond dimerization of PSGL-1 generates membrane-anchored PSGL-1 that does not bind P-selectin. It was reported that C310A-PSGL-1, when co-expressed with FucT-III in human 293 cells, did not bind to fluid-phase P-selectin in flow cytometric assays (18). Likewise, another group reported that C320A-PSGL-1, when co-expressed in K562 cells with FucT-VII, did not bind P-selectin in fluid-phase flow cytometric assays, nor did the mutated PSGL-1 support rolling adhesion of cells on immobilized P-selectin under shear stress (19).
We obtained markedly different results from those of Fujimoto et al. (18) and Snapp et al. (19). We eliminated the potential disulfide bond by the conservative mutation of Cys 320 to Ser, in addition to the change to Ala. Both mutations prevent covalent dimerization of recombinant PSGL-1, but they have no effect on its ability to bind fluid-phase P-selectin in flow cytometric assays or to support rolling cell adhesion on Pselectin-coated plates under shear stress. To address the possibility that the discrepancy between our results and that of Fujimoto et al. (18) was due to the different cell types used for expression of the constructs (CHOdhfr Ϫ versus 293 cells), we also co-transfected 293 cells with C320S-PSGL-1 and FucT-III. Soluble P-selectin bound similarly to cells expressing C320S-PSGL-1 or wt-PSGL-1.
We can only speculate as to why other groups failed to observe that PSGL-1 lacking the disulfide bond can still bind P-selectin (18,19). Binding of recombinant PSGL-1 to P-selectin requires correct co-expression of C2GnT and an ␣1,3-fucosyltransferase responsible for synthesis of core-2 O-glycans and sLe x antigen, respectively (17). Some clones that express recombinant C320A-PSGL-1 may express low levels of either or both of these glycosyltransferases, thus lowering the efficiency of the required post-translational modifications. There may also be unidentified clonal differences in expression of other post-translational modifications that are required for PSGL-1 to bind P-selectin.
We found that cell surface C320S-PSGL-1 is efficiently crosslinked by BS 3 , indicating that the individual polypeptides of PSGL-1 self-associate, perhaps as dimers, in the absence of a disulfide bond, but higher order oligomers were not detected. We considered the possibility that the less conservative C320A mutation used by others (18,19) might prevent noncovalent dimerization of the two polypeptides, whereas the more conservative C320S mutation would not disturb noncovalent dimerization. We found, however, that both C320A-PSGL-1 and C320S-PSGL-1 form noncovalent dimers in the membrane that are efficiently cross-linked by BS 3 . From these studies, it is clear that covalent dimerization through disulfide bond linkage is not necessary for PSGL-1 to bind P-selectin.
Our observation that cells expressing CD43TMD-PSGL-1 are capable of binding fluid-phase P-selectin, strengthens the overall interpretation that oligomerization of PSGL-1 is not required for it to bind P-selectin. We are continuing to characterize CD43TMD-PSGL-1 and other chimeric forms of PSGL-1 we have generated for their binding to P-, L-, and E-selectin under conditions of shear stress to further address the possible role of oligomerization of PSGL-1. These studies suggest that residues within the TMD of PSGL-1 are important for noncovalent association, but the specific nature of this self-association is under investigation.
To further investigate the role of self-association of PSGL-1 in P-selectin recognition, we explored whether small, proteo-lytically derived fragments from PSGL-1 retained the ability to bind P-selectin. Trypsin treatment of HL60 cells generated two major NH 2 -terminal fragments with molecular masses of ϳ100and ϳ25-kDa that bound to P-selectin affinity columns. These studies revealed that native PSGL-1 is relatively resistant to complete degradation by trypsin, probably because it is highly glycosylated. One possible reason for the resistance of PSGL-1 to trypsin is the degree of N-glycosylation of the molecule. Specifically, we noted that the NH 2 -terminal tryptic site at Arg 64 is immediately adjacent to a potential N-glycosylation sequon beginning at Asn 65 . We therefore created a recombinant form of PSGL-1 in which all N-glycosylation sequons are altered (⌬N-PSGL-1). Trypsin treatment of cells co-expressing this construct with C2GnT and FucT-VII generated small NH 2terminal glycosulfopeptide fragments that bound P-selectin. This demonstrates that dimerization of full-length PSGL-1 is not required to bind P-selectin, as proposed by other investigators (18,19).
We are currently investigating the detailed structure of the NH 2 -terminal-derived tryptic fragments generated from PSGL-1, since its glycosylation and tyrosine sulfation status may shed more light on the specific features required for binding to P-selectin. Because of the exceedingly small quantities of this small NH 2 -terminal tryptic fragment, we generated a semi-synthetic glycosulfopeptide modeled after what we assume to be the tryptic fragment. This small glycosulfopeptide GSP-6, as shown in Fig. 9A, contains three Tyr-SO 3 Ϫ residues and a monosialylated and monofucosylated core-2 O-glycan expressing the sialyl Lewis x antigen. The trypsin-released NH 2 -terminal glycosulfopeptide from ⌬N-PSGL-1 behaves similarly to GSP-6 on both reverse-phase and ion exchange chromatography, indicating that the trypsin-derived and semi-synthetic glycosulfopeptide are highly similar in structure. This further indicates that the trypsin-derived glycosulfopeptide, like GSP-6, is monomeric in structure. As we have recently shown, GSP-6 binds with an equivalent affinity to P-selectin as wild-type, dimeric PSGL-1 purified from human neutrophils (20,24).
Although our results demonstrate that a small NH 2 -terminal fragment derived from intact PSGL-1 is sufficient to bind Pselectin, dimerization of cell surface PSGL-1 may have important functions. In addition to mediating cell adhesion to P-and L-selectin, PSGL-1 also functions as a signaling receptor (38 -41). PSGL-1 is concentrated in the tips of leukocyte microvilli (15); upon leukocyte activation PSGL-1 is redistributed to the uropods of the polarized cells (42,43). Dimerization of PSGL-1 may enhance cell adhesion, cell signaling, and/or interactions with cytoskeletal elements. It is interesting that the position of the extracellular Cys residue is conserved in human and mouse PSGL-1 (44). Covalent dimerization through disulfide bonds may stabilize noncovalent dimers of PSGL-1 in the membrane and enhance their biological activities. The cells expressing monomeric or oligomeric membrane-anchored forms of PSGL-1 are currently under investigation to address the biological significance of PSGL-1 oligomerization.