Post-translational modifications of recombinant P-selectin glycoprotein ligand-1 required for binding to P- and E-selectin.

P-selectin glycoprotein ligand-1 (PSGL-1) is a mucin-like ligand for P- and E-selectin on human leukocytes. PSGL-1 requires sialylated, fucosylated O-linked glycans and tyrosine sulfate to bind P-selectin. Less is known about the determinants that PSGL-1 requires to bind E-selectin. To further define the modifications required for PSGL-1 to bind P- and E-selectin, we transfected Chinese hamster ovary (CHO) cells with cDNAs for PSGL-1 and specific glycosyltransferases. CHO cells synthesize only core 1 O-linked glycans (Galβ1-3GalNAcα1-Ser/Thr); they lack core 2 O-linked glycans (Galβ1-3(Galβ1-4GlcNAcβ1-6)GalNAcα1-Ser/Thr) because they do not express the core 2 β1-6-N-acetylglucosaminyltransferase (C2GnT). CHO cells also lack α1-3 fucosyltransferase activity. PSGL-1 expressed on transfected CHO cells bound P- and E-selectin only when it was co-expressed with both C2GnT and an α1-3 fucosyltransferase (Fuc-TIII, Fuc-TIV, or Fuc-TVII). Chromatography of β-eliminated O-linked glycans from PSGL-1 co-expressed with C2GnT confirmed synthesis of core 2 structures. Tyrosine residues on PSGL-1 expressed in CHO cells were shown to be sulfated. Phenylalanine replacement of three tyrosines within a consensus sequence for tyrosine sulfation abolished binding to P-selectin but not to E-selectin. These results demonstrate that PSGL-1 requires core 2 O-linked glycans that are sialylated and fucosylated to bind P- and E-selectin. PSGL-1 also requires tyrosine sulfate to bind P-selectin but not E-selectin.

The selectins are a group of three Ca 2ϩ -dependent lectins that mediate rolling adhesion of leukocytes on the vessel wall during inflammation (reviewed in McEver et al. (1)). L-selectin, expressed on leukocytes, binds to constitutively or inducibly expressed carbohydrate ligands on endothelial cells. E-selectin, expressed on cytokine-activated endothelium, and P-selectin, expressed on thrombin-activated endothelial cells and platelets, bind to carbohydrate ligands on myeloid cells and subsets of lymphocytes. Leukocyte rolling requires that selectins rapidly associate and then dissociate from their ligands in a manner largely independent of shear stress (2). The selectins interact weakly with sialylated and fucosylated oligosaccharides such as sialyl Lewis x (sLe x ), 1 but bind with higher affinity/ avidity to only a few glycoproteins (1). Studies with mAbs support a physiologic role for one of these glycoproteins, Pselectin glycoprotein ligand-1 (PSGL-1). PSGL-1 accounts for all the high affinity binding sites for P-selectin on human leukocytes (3). PSGL-1 must interact with P-selectin in order for neutrophils to roll on P-selectin under physiologic shear forces (3,4), and it also contributes to the rolling of neutrophils on E-selectin (5).
PSGL-1 is a membrane glycoprotein with two identical disulfide-linked subunits (6). Each subunit has at most three N-linked glycans, but has many clustered, sialylated O-linked glycans (6,7), some with polylactosamine terminating in sLe x (8). PSGL-1 is a type I membrane protein with an extracellular domain that is rich in serines, threonines, and prolines (9). PSGL-1 binds both E-selectin and P-selectin (8 -11). The structural requirements for binding are not well characterized, although there is evidence that PSGL-1 does not interact identically with P-and E-selectin (5,8). PSGL-1 must be sialylated and fucosylated to bind both molecules (6,9,10). Enzymatic removal of N-linked glycans from human neutrophil PSGL-1 does not affect binding to P-selectin, suggesting that O-linked glycans are required for binding (6). It is not known if O-linked glycans are required for binding to E-selectin. Fab fragments of PL1, an IgG mAb to PSGL-1, block binding of PSGL-1 to P-selectin but only partially inhibit binding to E-selectin (3,5). The PL1 epitope is located near the N terminus (3), 2 near three clustered tyrosines within a consensus motif for tyrosine sulfation (9,12). PSGL-1 is sulfated on tyrosine rather than on carbohydrate, and enzymatic removal of sulfate from tyrosine eliminates binding of PSGL-1 to P-selectin (13). These data demonstrate that tyrosine sulfate functions in conjunction with sialylated and fucosylated glycans to mediate high affinity binding of PSGL-1 to P-selectin. It is not known if tyrosine sulfation is required for PSGL-1 to bind E-selectin.
Antibodies-The anti-human P-selectin mAbs S12 and G1 were prepared as described previously (14,15). G1, but not S12, blocks binding of P-selectin to leukocytes. The anti-human E-selectin mAbs ES1 and ES2 were prepared as described previously (5). ES1, but not ES2, blocks binding of E-selectin to leukocytes. The anti-human PSGL-1 mAbs PL1 and PL2 were prepared as described previously (3). PL1, but not PL2, blocks binding of PSGL-1 to P-selectin. These mAbs are all of the IgG 1 subclass. The hybridoma secreting the IgM anti-sLe x mAb CSLEX-1 (16) was obtained from the ATCC. Rabbit antiserum to a peptide corresponding to residues 42-56 of the cDNA-derived amino acid sequence of human PSGL-1 was prepared as described previously (8). MOPC21 (IgG 1 ) was purchased from Cappel-Organon Technika. Fluorescein isothiocyanate-conjugated goat anti-mouse IgG/IgM was purchased from Caltag Laboratories.
Proteins-Human platelet P-selectin and human neutrophil PSGL-1 were purified as described previously (8,17). Human neutrophil PSGL-1 was radiolabeled with 125 I as described previously (8). A recombinant soluble form of human P-selectin (sPS, formerly called tPS) was purified as described previously (18). A recombinant soluble form of human E-selectin (sES, formerly called tES) was a gift from Dr. David Lyons (8).
cDNAs-The PSGL-1 cDNA was amplified by the polymerase chain reaction from human leukocyte cDNAs as described previously (3). We reported that this cDNA encoded an additional decamer repeat between amino acids 127 and 128 of the original published sequence (9) and changed Ala 128 to Pro. On reexamination, the decamer sequence was found to be inserted between amino acids 128 and 129, which did not change the original sequence. The additional decamer sequence is TEAQTTQPVP. This additional decamer has been observed in most PSGL-1 cDNAs (19). The cDNA insert was excised from the pBK-EF vector with ScaI and KpnI and ligated into the mammalian expression vector pZeoSV (Invitrogen). The pZeoSV plasmid allows selection for resistance to the antibiotic Zeocin in both Escherichia coli and mammalian cells.
A mutant PSGL-1 cDNA was created in which the codons encoding the tyrosines at residues 46, 48, and 51 were replaced with codons encoding phenylalanines. Mutagenesis was performed with a polymerase chain reaction protocol (20). The PSGL-1 cDNA template was inserted into Bluescript (Stratagene) at the XbaI and KpnI sites. The outside primers for the polymerase chain reaction included a restriction site for XbaI on the 5Ј end in the Bluescript vector and for a unique StuI site in the coding region of the PSGL-1 cDNA. The amplified product was sequenced to confirm that it contained the expected substitutions. It was then excised with XbaI and StuI and used to replace the corresponding XbaI-StuI fragment of wild-type PSGL-1 in Bluescript. The entire PSGL-1 mutant cDNA was then excised from Bluescript with XbaI and KpnI and inserted into pZeoSV.
The C2GnT cDNA was amplified by the reverse transcriptase polymerase chain reaction from total RNA from human testis (Clontech). The two oligonucleotides used for polymerase chain reaction (5Ј-GCG GCG GCG TCT AGA CCA CCA TGC TGA GGA CGT TGC TGC GAA G-3Ј and 5Ј-CGC GCG CGT CGA CTC ACA GTC AGT GTT TTA ATG TCT CCA AAG C-3Ј) were designed to match the 5Ј and 3Ј ends of the published sequence of the human C2GnT cDNA (21). The primers also included additional non-complementary sequence that included an XbaI site 5Ј to the initiating methionine and a SalI site 3Ј to the stop codon. The amplified product was ligated into the PCRII vector using the TA cloning kit (Invitrogen). The entire insert was sequenced to confirm its identity with the published sequence. The insert was then excised with EcoRI and ligated into the expression vector pcDNA3 (Invitrogen).
The human Fuc-TIII and Fuc-TIV cDNAs were amplified by the polymerase chain reaction from genomic DNA. 3 The amplified products were sequenced to confirm that they matched the published sequences (22,23) and were then ligated into the expression vector pRC/RSV (Invitrogen). The cDNA encoding human Fuc-TVII in the plasmid pCDM8 (24) was a generous gift from Dr. John Lowe.
Transfections-CHO DHFR(Ϫ) cells and COS-7 cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal calf serum at 37°C in an atmosphere containing 5% CO 2 . CHO DHFR(Ϫ) cells express sLe x when transfected with cDNAs for Fuc-TIII or Fuc-TIV (25). CHO cells were transfected with Fuc-TIII or Fuc-TIV cDNA in pRC/RSV using the Stratagene Transfection MBS kit according to the suggested protocol. CHO cells were transfected with C2GnT cDNA in pcDNA3, or C2GnT cDNA plus Fuc-TIII cDNA, using calcium phosphate precipitation. The transfected cells were selected in medium containing 400 g/ml of G418 (Life Technologies, Inc.). Individual colonies were isolated and recloned by limiting dilution. In this manner, permanently transfected CHO cells were isolated that expressed Fuc-TIII, Fuc-TIV, C2GnT, or Fuc-TIII plus C2GnT. Portions of these cells were then transfected with PSGL-1 cDNA in pZeoSV using Lipofectamine. Some of the Fuc-TIV-expressing cells were transfected with both PSGL-1 cDNA and C2GnT cDNA. All cells transfected a second time were selected in medium containing both G418 (400 g/ml) and Zeocin (250 g/ml). Individual colonies were expanded and, in some cases, recloned by limiting dilution. In this manner, permanently transfected CHO cells were isolated that expressed various combinations of PSGL-1, C2GnT, Fuc-TIII, and/or Fuc-TIV.
Fuc-TVII cDNA was transiently transfected into CHO cells permanently transfected with PSGL-1 cDNA, or with PSGL-1 cDNA plus C2GnT cDNA. In other experiments, wild-type or mutated PSGL-1 cDNA was transiently transfected into CHO cells permanently transfected with Fuc-TIII cDNA or C2GnT cDNA plus Fuc-TIII cDNA. Transiently transfected CHO cells were analyzed 2 days after transfection. COS-7 cells were transfected with Fuc-TIII or Fuc-TIV cDNA using Lipofectamine and selected in medium containing 400 g/ml G418 to obtain cells permanently expressing Fuc-TIII or Fuc-TIV. 3 COS cells expressing Fuc-TIII were transiently transfected with PSGL-1 cDNA, or PSGL-1 plus C2GnT cDNA using Lipofectamine. The cells were analyzed 2 days after transfection.
C2GnT Assay-Cells were detached with 0.02% EDTA in phosphatebuffered saline, washed twice with phosphate-buffered saline, centrifuged, and stored as cell pellets at Ϫ20°C until they were analyzed. Frozen cell pellets were typically resuspended in 100 l of 10 mM sodium cacodylate, pH 7.2, containing 0.1 mg/ml leupeptin (Sigma), 0.2 mg/ml aprotinin (Sigma), and 1 mM Pefabloc SC (AEBSF, 4-(2-aminoethyl)-benzenesulfonyl fluoride HCL) (Boehringer Manneheim), and homogenized with 30 strokes of a hand-held Teflon Potter-Elvehjem homogenizer. Homogenates were brought to 1% Triton X-100 by addition of 0.1 volume of 10% Triton X-100 in buffer. Protein concentrations were determined using the BCA protein assay (reagents from Sigma) with bovine serum albumin as standard. C2GnT activities were assayed with a modification of a previously described procedure (26). Briefly, C2GnT activities were measured by mixing 2 l of cell lysate with 5 l of assay mixture containing 25 mM sodium cacodylate, pH 6.5, 5 mM EDTA, 333 mM GlcNAc, 10% glycerol, 0.8 mM UDP-GlcNAc (100,000 cpm/nmol) (unlabeled UDP-GlcNAc from Sigma, and UDP-6-[ 3 H]-Nacetylglucosamine from American Radiolabeled Chemicals, Inc.), and 8 mM benzyl 2-acetamido-2-deoxy-3-O-␤-D-galactopyranosyl-␣-D-galactopyranoside from Sigma. Mixtures were incubated at 37°C for 4 h. Each reaction mixture was diluted with 0.3 ml of water and loaded onto a C18 SepPak cartridge (Millipore) mounted on a vacuum manifold. The cartridge was washed with 25 ml of water to remove unreacted radiolabeled sugar-nucleotide. The hydrophobic glucosylated product was then eluted with 3 ml of methanol and counted in a scintillation counter. Blank values were obtained by carrying out assays without acceptor and were subtracted from the values obtained with acceptor.
Flow Cytometry-HL-60 cells were maintained in RPMI 1640 containing 20% fetal calf serum, 4 mM L-glutamine, 100 IU/ml penicillin, and 100 g/ml streptomycin. Binding of mAbs to intact HL-60, CHO, or COS cells was measured as described previously (3). Binding of fluidphase P-selectin to intact cells was assessed as described previously (27).
Cell Adhesion Assay-Confluent or near-confluent monolayers of CHO cells were labeled for 16 h in medium containing 10 Ci/ml [ 3 H]thymidine. The cells were washed with Hanks' balanced salt solu-tion (HBSS, Life Technologies, Inc.) and then detached from the substratum by incubation in 0.02% EDTA for 10 min. The cells were washed twice in serum-free Dulbecco's modified Eagle's medium and then suspended at 2 ϫ 10 6 cells/ml in Dulbecco's modified Eagle's medium containing 1% bovine serum albumin and 1% fetal bovine serum. The cells were labeled to a specific activity of 2-7 ϫ 10 Ϫ6 Ci/cell; when the cells were centrifuged, greater than 95% of the radioactivity remained associated with the cell pellet.
Microtiter wells (Immulon I Removawell strips, Dynatech Laboratories) were coated with soluble P-selectin or soluble E-selectin (each at 2.5 g/ml) in 100 l of HBSS overnight at 4°C. The wells were then blocked with 1% human serum albumin in HBSS for 2 h at room temperature. Labeled CHO cells (100 l, 2 ϫ 10 5 cells) were added to each well and incubated on an orbital shaker for 20 min at room temperature. In some experiments, the cells were preincubated with 10 g/ml of the mAbs PL1, PL2, G1, or S12. After aspirating the unbound cells, the wells were gently washed three times with HBSS. Each well was then cut out, and the bound radioactivity was determined by liquid scintillation counting. The number of bound cells was derived from a standard curve generated with known concentrations of cells.
Metabolic Labeling and Immunoprecipitation of PSGL-1-CHO cells were metabolically labeled with [ 3 H]glucosamine or [ 35 S]sulfate as described previously for HL-60 cells (6,13). Cell extracts were immunoprecipitated as described previously (8,13), using the rabbit antiserum to a peptide corresponding to residues 42-56 of PSGL-1 or, as a control, normal rabbit serum. The immunoprecipitates were analyzed by SDS-PAGE under reducing or nonreducing conditions, followed by fluorography. In some gels, 125 I-labeled PSGL-1 from human neutrophils was analyzed in parallel.
Structural Analysis of PSGL-1 Oligosaccharides-Following identification of [ 3 H]glucosamine-labeled PSGL-1 by fluorography, the dried gel was aligned with the exposed x-ray film, and the band from the nonreducing lane was subjected to ␤-elimination (28). The released oligosaccharides were treated with sialidase (2.5 milliunits/ml) in 20 l of 50 mM sodium acetate, pH 5.5, for 18 h at 37°C (28). In some experiments, neutral ␤-eliminated oligosaccharides from PSGL-1 coexpressed with C2GnT and Fuc-TIV were treated with 0.25 milliunit/ml of ␣1-3/4-L-fucosidase in 20 l of 50 mM KH 2 PO 4 , pH 6.0, 0.1 M NaCl, 0.02% sodium azide at 37°C for 18 h. A portion of the oligosaccharides was analyzed by amine-adsorption high performance liquid chromatography on a Varian AX-5 column (4 mm ϫ 30 cm). The column was equilibrated in 65:35 (v/v) acetonitrile:deionized water and was eluted with a linear gradient of increasing water (0.33%/min) (29). Fractions (0.5 ml) were collected, and radioactivity in each fraction was determined by scintillation counting.
Detection of Tyrosine Sulfate-Tyrosine sulfation of PSGL-1 was determined as described previously (13). Briefly, immunoprecipitated 35 S-PSGL-1 was subjected to base hydrolysis, then analyzed by chromatography on a Varian AX-5 column (4 mm ϫ 30 cm) that was eluted with a gradient of NaH 2 PO 4 .
To determine whether both C2GnT and an ␣1-3 fucosyltransferase must modify PSGL-1 for it to bind P-or E-selectin, we first isolated clones of permanently transfected CHO cells that expressed various combinations of PSGL-1, C2GnT, and/or an ␣1-3 fucosyltransferase (Fuc-TIII or Fuc-TIV). C2GnT activity was detected only in lysates of CHO cells that were transfected with the C2GnT cDNA (Table I). Flow cytometric analysis with the anti-sLe x mAb CSLEX-1 indicated that CHO cells expressed no endogeneous sLe x on the cell surface (Fig.  1B). However, expression of Fuc-TIII or Fuc-TIV caused appearance of sLe x on the CHO cell surface (Fig. 1, A, C-E, G, H). Clones of CHO cells transfected with PSGL-1 cDNA expressed comparable levels of the protein, as measured by binding of the anti-PSGL-1 mAbs PL1 and PL2 (Fig. 1, B-D, G, H). Indeed, expression of PSGL-1 on the CHO cells was higher than on HL-60 cells (Fig. 1F).
To determine whether PSGL-1 on CHO cells bound P-selectin, we used flow cytometry to measure binding of fluid-phase P-selectin to the cell surface (Fig. 2). Using this assay, we previously demonstrated that PSGL-1 accounts for all the high affinity binding sites for P-selectin on leukocytes; binding is abolished by the blocking anti-PSGL-1 mAb PL1, but not by the non-blocking mAb PL2 (3) (Fig. 2F). P-selectin did not bind CHO cells lacking PSGL-1, even when they expressed both C2GnT and an ␣1-3 fucosyltransferase (Fig. 2, A and E). Pselectin also did not bind CHO cells expressing PSGL-1 plus C2GnT (Fig. 2B), or CHO cells expressing PSGL-1 plus either Fuc-TIII or Fuc-TIV (Fig. 2, C and G). In contrast, P-selectin bound avidly to CHO cells expressing PSGL-1 plus C2GnT and either Fuc-TIII or Fuc-2TIV (Fig. 2, D and H). PL1, but not PL2, completely inhibited the binding, suggesting that P-selectin interacted with a similar region on PSGL-1 expressed by either CHO cells or leukocytes. These results establish that CHO cells require both C2GnT and an ␣1-3 fucosyltransferase in order for PSGL-1 to bind P-selectin.
We also transiently expressed Fuc-TVII in CHO cell clones that stably expressed PSGL-1, or PSGL-1 and C2GnT. Like Fuc-TIII and Fuc-TIV, Fuc-TVII caused expression of sLe x on the cell surface (Fig. 3, A and B). Fuc-TVII also conferred binding of PSGL-1 to P-selectin, but only when the cells also expressed C2GnT (Fig. 3, C and D). Thus, CHO cells express a functional form of PSGL-1 when it is co-expressed with C2GnT and either of the fucosyltransferases normally expressed in leukocytes, Fuc-TIV or Fuc-TVII.  Fluid-phase P-selectin Does Not Bind PSGL-1 Co-expressed with C2GnT and Fuc-TIII on COS Cells-The oligosaccharide structures are less well characterized in COS cells than in CHO cells. However, as measured by a cell adhesion assay, P-selectin binds PSGL-1 that is co-expressed with Fuc-TIII in transfected COS cells (9). 4 We used flow cytometry to determine whether co-expression of PSGL-1 with Fuc-TIII in transfected COS cells resulted in appearance of high affinity binding sites for fluid-phase P-selectin on the cell surface. PSGL-1 cDNA was transiently transfected into COS cells that had been permanently transfected with Fuc-TIII cDNA. The transfected COS cells expressed both PSGL-1 and sLe x , as measured by binding of the mAbs PL1, PL2, and CSLEX-1 (Fig. 4A). However, fluid-phase P-selectin did not bind to the COS cells (Fig.  4C). PSGL-1 co-expressed with Fuc-TIII in COS cells has many core 1 O-linked glycans that are removed by sialidase and endo-␣-galactosaminidase (8,9). This might be due to the fact that COS cells have slightly lower levels of endogeneous C2GnT activity than HL-60 cells (Table I). To determine whether a deficiency of C2GnT accounted for the relatively poor binding of P-selectin to PSGL-1 on COS/Fuc-TIII cells, we transiently transfected COS/Fuc-TIII cells with cDNAs for both C2GnT and PSGL-1. These cells also expressed PSGL-1 and sLe x (Fig. 4B), and lysates of the cells had significantly increased C2GnT activity (Table I). However, fluid-phase P-selectin still failed to bind to these cells (Fig. 4D). In parallel experiments, P-selectin bound avidly to PSGL-1 on HL-60 cells and on CHO cells expressing PSGL-1, C2GnT, and Fuc-TIII (data not shown). These results demonstrate that P-selectin binds with relatively lower affinity to PSGL-1 co-expressed with Fuc-TIII and C2GnT on COS cells than to PSGL-1 coexpressed with the same glycosyltransferases on CHO cells. The mechanisms accounting for this difference are not known. However, the data reinforced the decision to use CHO cells for these experiments, because the oligosaccharide structures are much better characterized in CHO cells than in COS cells.

PSGL-1 Expressed on CHO Cells Mediates Adhesion to Pand E-selectin Only When It Is Co-expressed with
C2GnT and an ␣1-3 Fucosyltransferase-We next used a cell adhesion assay to measure the modifications required for PSGL-1 expressed on CHO cells to bind P-and E-selectin. CHO cells transfected with an ␣1-3 fucosyltransferase express sLe x on the cell surface, and adhere to immobilized P-or E-selectin under certain experimental conditions (23,25,38). To measure PSGL-1-dependent adhesion, we developed conditions in which CHO cells expressing an ␣1-3 fucosyltransferase did not adhere to P-or E-selectin. Using these experimental conditions, CHO cells adhered to immobilized P-selectin or E-selectin in a PSGL-1-dependent manner, but only when the cells also expressed C2GnT and Fuc-TIII (Fig. 5, A and C). Adhesion of the cells was Ca 2ϩ -dependent and was inhibited by a blocking, but not a nonblocking, mAb to the appropriate selectin (Fig. 5, B  and D). Cell adhesion to P-selectin was blocked by PL1, but not by PL2 (Fig. 5B). In contrast, cell adhesion to E-selectin was not inhibited by PL1 or PL2 (Fig. 5D). CHO cells adhered in a similar PSGL-1-dependent manner to immobilized P-and Eselectin when they expressed C2GnT and Fuc-TIV (data not shown). These data demonstrate that PSGL-1 expressed on CHO cells binds both P-and E-selectin, but only when it is co-expressed with both C2GnT and an ␣1-3 fucosyltransferase.

PSGL-1 Expressed in CHO Cells Is Modified with Core 2 O-linked Glycans When It Is Co-expressed with C2GnT-To
determine if co-expression of C2GnT caused the addition of core 2 O-linked glycans to PSGL-1, we metabolically labeled CHO cell clones with [ 3 H]glucosamine, which is incorporated as amino sugars in oligosaccharides. Radiolabeled PSGL-1 was immunoprecipitated from cell lysates, resolved by SDS-PAGE under reducing and nonreducing conditions, and detected by fluorography (Fig. 6A). We analyzed CHO cells expressing PSGL-1 with either Fuc-TIII or Fuc-TIV, in the presence or absence of C2GnT. PSGL-1 expressed by all the CHO cells migrated as a disulfide-linked homodimer, as did 125 I-PSGL-1 from human neutrophils. The mobilities of PSGL-1 from all  (6), prevented binding to a P-selectin affinity column (data not shown). Sialidase treatment decreased the electrophoretic mobility of PSGL-1 from CHO cells (Fig. 6B), an effect similar to that observed with PSGL-1 from leukocytes (6). Addition of endo-␣-N-acetylgalactosaminidase had no further effect on the mobility of PSGL-1 co-expressed with C2GnT. This lack of effect is similar to that observed with PSGL-1 from leukocytes (8), and is consistent with the resistance of core 2 O-linked glycans to this enzyme. In contrast, addition of endo-␣-Nacetylgalactosaminidase to PSGL-1 expressed without C2GnT resulted in loss of a detectable radiolabeled band on the gel. This indicates that all of the labeled O-linked glycans were core 1 structures that were removed by the enzyme.
To confirm that co-expression of C2GnT caused addition of core 2 O-linked glycans to PSGL-1, immunoprecipitated [ 3 H]glucosamine-labeled PSGL-1 was resolved by SDS-PAGE under nonreducing conditions. The gel slice containing PSGL-1 was excised and subjected to ␤-elimination. The released Fig. 1 were incubated without platelet P-selectin (Control) or with platelet P-selectin, in the presence or absence of the anti-PSGL-1 mAbs PL1 or PL2. Bound P-selectin was detected with a biotinylated nonblocking anti-P-selectin mAb, S12, followed by phycoerythrin-streptavidin. The data are representative of at least three independent experiments.

FIG. 3. Fluid-phase P-selectin binds PSGL-1 on transfected CHO cells when it is co-expressed with C2GnT
and Fuc-TVII. CHO cells permanently transfected with cDNAs for PSGL-1, or for PSGL-1 and C2GnT, were transiently transfected with cDNA for Fuc-TVII. Two days after transfection, the cells were analyzed for binding of mAbs as in Fig. 1 or for binding of fluid-phase P-selectin as in Fig. 2. The data are representative of three independent experiments.

FIG. 4. Fluid-phase P-selectin does not bind PSGL-1 co-expressed with
C2GnT and Fuc-TIII on COS cells. COS cells permanently transfected with cDNA for Fuc-TIII (A and C) or C2GnT and Fuc-TIII (B and D) were transiently transfected with cDNA for PSGL-1. Two days after transfection, the cells were analyzed for binding of mAbs as in Fig. 1 or for binding of fluid-phase P-selectin as in Fig. 2. In parallel assays, fluid-phase Pselectin bound avidly to HL-60 cells and to CHO cells expressing PSGL-1, Fuc-TIII, and C2GnT as in Fig. 1. The data are representative of at least three independent experiments.

FIG. 5. PSGL-1 co-expressed with C2GnT and Fuc-TIII in CHO cells mediates cell adhesion to immobilized P-and E-selectin.
Adhesion of CHO cells permanently transfected with the indicated cDNAs was measured in wells containing immobilized P-selectin (A) or E-selectin (C). Adhesion of CHO cells expressing C2GnT, Fuc-TIII, and PSGL-1 to immobilized P-selectin (B) or E-selectin (D) was measured in the presence of the indicated mAb or in buffer containing EDTA. The data represent the mean Ϯ S.D. of triplicate wells and are representative of three independent experiments.
[ 3 H]glucosamine-labeled O-linked glycans were desialylated and analyzed by amine-adsorption high performance liquid chromatography. The radiolabeled oligosaccharides from PSGL-1, when co-expressed only with FucT-III or Fuc-TIV, eluted as the GalNAcitol and Gal␤1-3GalNAcitol standards, consistent with their predicted core 1 structures (Fig. 7). In contrast, most of the radiolabeled oligosaccharides from PSGL-1, when co-expressed with C2GnT plus either Fuc-TIII or Fuc-TIV, co-eluted with the tetrasaccharide core 2 standard, Gal␤1-3(Gal␤1-4GlcNAc␤1-6)GalNAcitol (Fig. 7). Some of the radioactivity in cells co-expressing C2GnT and Fuc-TIV eluted as a pentasaccharide, indicating the presence of a larger core 2 structure (29). This peak was no longer observed after treatment of the labeled glycans from these cells with an ␣1-3-specific fucosidase (Fig. 7). This suggests that the larger structure is a core 2 tetrasaccharide modified with a fucose in ␣1-3 linkage to the ␤1-6-linked N-acetylglucosamine. These data confirm that co-expression of C2GnT results in addition of core 2 O-linked glycans to PSGL-1 in CHO cells.
PSGL-1 Co-expressed on CHO Cells with C2GnT and a Fucosyltransferase Must Also Be Sulfated on Tyrosine to Bind P-selectin but Not E-selectin-PSGL-1 on leukocytes must be tyrosine sulfated to bind P-selectin (13). To determine whether PSGL-1 expressed by CHO cells was sulfated, we metabolically labeled CHO cells co-expressing PSGL-1, C2GnT, and either Fuc-TIII or Fuc-TIV with [ 35 S]sulfate. PSGL-1 was immunoprecipitated from cell lysates, resolved by SDS-PAGE under non-reducing conditions, and analyzed by fluorography. A radiolabeled protein with the expected mobility for PSGL-1 was immunoprecipitated from cells co-expressing C2GnT and either Fuc-TIII or Fuc-TIV, indicating that PSGL-1 was sulfated in CHO cells (Fig. 8A). To confirm that the sulfate was on tyrosine, the gel slice containing the [ 35 S]sulfate-labeled PSGL-1 from cells co-expressing C2GnT and Fuc-TIV was excised, subjected to strong base hydrolysis, and analyzed by anion exchange chromatography. The radioactivity was recovered in a single peak that co-migrated with the authentic tyrosine sulfate standard (Fig. 8B). These data demonstrate that PSGL-1 expressed on CHO cells, like PSGL-1 on leukocytes, is sulfated on tyrosine rather than on carbohydrate. The oligosaccharides were desialylated and then analyzed by amine-adsorption high performance liquid chromatography. A portion of the glycans from PSGL-1 co-expressed with C2GnT and Fuc-TIV was also treated with ␣1-3 fucosidase prior to analysis. The radioactivity in each fraction was measured by scintillation counting. The retention times of the core 1 and core 2 oligosaccharide standards are shown.
PSGL-1 has three tyrosines at residues 46, 48, and 51 that are located within a consensus sequence for tyrosine sulfation (9,12). To test the importance of these residues for binding P-selectin and E-selectin, we prepared a construct of PSGL-1 in which all three tyrosines were replaced with phenylalanines. Wild-type or mutant PSGL-1 cDNA was transiently transfected into CHO cells that permanently expressed C2GnT and Fuc-TIII. The use of the same permanently transfected cells for transient expression of wild-type or mutant PSGL-1 ensured equivalent levels of C2GnT activity (Table I) and surface expression of sLe x (Fig. 9). Furthermore, surface levels of wildtype and mutant PSGL-1 were comparable (Fig. 9). However, fluid-phase P-selectin bound to CHO cells expressing wild-type PSGL-1, but not to cells expressing mutant PSGL-1 (Fig. 9). Furthermore, CHO cells expressing mutant PSGL-1 did not adhere to immobilized P-selectin, although they did adhere to immobilized E-selectin (Fig. 10). These data demonstrate that PSGL-1 expressed in CHO cells requires at least one of the three N-terminal tyrosines to bind P-selectin, presumably because this residue(s) must be sulfated. In contrast, appropriately glycosylated PSGL-1 does not require the tyrosines, and therefore tyrosine sulfate, to bind E-selectin. DISCUSSION Because their oligosaccharide structures are well characterized, CHO cells are an excellent model for analyzing the effects of glycosylation on protein structure and function. CHO cells synthesize complex-type bi-, tri-, and tetra-antennary N-glycans with type 2 lactosamine (Gal␤1-4GlcNAc) and polylactosamine; the only other modification is terminal sialylation as NeuAc␣2-3Gal␤-4GlcNAc-R. CHO cells synthesize O-glycans with simple core 1 structures that can be mono-or disialylated. Our results establish that PSGL-1 expressed in CHO cells binds P-and E-selectin only when it is co-expressed with C2GnT and an ␣1-3 fucosyltransferase. These enzymes mediate branching and fucosylation of the O-linked glycans, modifications that are critical for PSGL-1 to bind P-and E-selectin. PSGL-1 expressed in CHO cells is also sulfated on tyrosine, a modification required for binding P-selectin but not E-selectin. These studies extend previous findings on PSGL-1 from human leukocytes that 1) sialic acids on PSGL-1 are required for binding P-selectin (6); 2) enzymatic removal of N-linked glycans from PSGL-1 does not affect binding to P-selectin (6); 3) PSGL-1 has many clustered, sialylated O-linked glycans (7), some with polylactosamine terminating in sLe x (8); and 4) enzymatic removal of sulfate from tyrosine on PSGL-1 abrogates binding to P-selectin (13).
We demonstrated previously that Fab fragments of PL1, a mAb to a membrane-distal epitope on PSGL-1, block binding of P-selectin to PSGL-1 on human leukocytes (3). Here we show that PL1 also blocks binding of P-selectin to PSGL-1 that is expressed with C2GnT and an ␣1-3 fucosyltransferase on CHO cells. This suggests that P-selectin binds a similar region on native and recombinant PSGL-1. On leukocytes, the putative N terminus of mature PSGL-1 begins at residue 42, just C-terminal to a consensus propeptide cleavage site (9). The PL1 epitope has been mapped to a region within residues 49 -62. 2 A polyclonal antiserum to residues 42-56 also prevents binding of P-selectin to human leukocyte PSGL-1 and to PSGL-1 co-ex-pressed with C2GnT and an ␣1-3 fucosyltransferase in CHO cells. 2 This region contains three tyrosines within a consensus sequence for tyrosine sulfation (12), and tyrosine sulfation of PSGL-1 on leukocytes is required for high affinity binding to P-selectin (13). Here we show that PSGL-1 expressed in CHO cells is also sulfated on tyrosine and that mutation of the three tyrosines eliminates binding to P-selectin. Taken together, the data suggest that P-selectin preferentially binds an N-terminal region of PSGL-1 that requires at least one tyrosine sulfate and one sialylated, fucosylated, core 2 O-linked glycan. The precise spatial relationship of the tyrosine sulfate(s) to the O-linked glycan(s) remains to be determined.
Competitive binding experiments suggest that E-selectin binds with lower affinity than P-selectin to the region that overlaps the PL1 epitope, but also binds an additional site(s) on PSGL-1 from human leukocytes (5,8). Here we demonstrate that PL1 fails to inhibit PSGL-1-dependent adhesion of transfected CHO cells to E-selectin. Furthermore, substitution of the N-terminal tyrosines, which eliminates adhesion to P-selectin, does not affect adhesion to E-selectin. These data demonstrate that, like leukocyte PSGL-1, PSGL-1 expressed in CHO cells binds differently to E-selectin than to P-selectin, although binding to both selectins requires sialylated, fucosylated, core 2 O-linked glycans.
The O-glycosylation sites and specific structures of the core 2 O-linked glycans required for native or recombinant PSGL-1 to optimally bind P-selectin or E-selectin need further study. For example, native PSGL-1 has polylactosamine (8), which can be extended from core 2 structures (35), but the actual structures of the O-linked glycans on native PSGL-1 have not been determined. It is possible that polylactosamine extension is required for a terminal sLe x structure to interact optimally with Pselectin or E-selectin. Fucosylation of internal N-acetylglucosamine residues might also be required for optimal binding (39). Internal fucosylation renders polylactosamine resistant to cleavage by endo-␤-galactosidase and may explain why treatment of native PSGL-1 with this enzyme does not eliminate binding to P-selectin (8).
Although fucosyltransferases have been intensively studied with regard to their actions on small oligosaccharide acceptors and N-linked glycans (22-25, 40 -44), little is known about how fucosyltransferases modify O-linked glycans. We found that PSGL-1 binds P-and E-selectin when it is co-expressed in CHO cells with C2GnT and either Fuc-TIII, Fuc-TIV, or Fuc-TVII. However, the data do not demonstrate that these fucosyltransferases modify specific O-linked glycans in the same manner. Furthermore, it is not known whether Fuc-TIV and Fuc-TVII modify PSGL-1 similarly in leukocytes, where the two enzymes are normally expressed (23,24,42). Chromatography of the ␤-eliminated glycans of PSGL-1 co-expressed with C2GnT and Fuc-TIV in CHO cells revealed an oligosaccharide eluting as a pentasaccharide that was not observed in ␤-eliminated O-linked glycans from CHO cells co-expressing C2GnT and Fuc-TIII. Based on the change in elution position after digestion with an ␣1-3-specific fucosidase, the pentasaccharide is probably a core 2 tetrasaccharide with a fucose linked ␣1-3 to the N-acetylglucosamine. It should be noted that larger Olinked structures with polylactosamine could be present but not be detected if they do not elute from the amine-adsorption column under the conditions used. The fucosylated polylactosamine-containing N-linked glycans synthesized by CHO cells expressing Fuc-TIV are smaller than those synthesized by CHO cells expressing Fuc-TIII, suggesting that fucosylation by Fuc-TIV competes with polylactosamine chain extension (37).  Fig. 9 were used in the cell adhesion assay. As an additional control, some cells permanently transfected with cDNAs for C2GnT and Fuc-TIII were not transiently transfected with cDNA for PSGL-1. One day after transfection, the cells were labeled with [ 3 H]thymidine. Two days after transfection, adhesion of the CHO cells was measured in wells containing immobilized P-selectin (A) or E-selectin (B). The data represent the mean Ϯ S.D. of triplicate wells and are representative of two independent experiments.
Perhaps Fuc-TIV and Fuc-TIII also differentially modify the core 2 O-linked glycans in CHO cells. If so, Fuc-TIV might produce a core 2 glycan with a single terminal sLe x ; this structure would be detected as a core 2 pentasaccharide after sialidase treatment, as observed in Fig. 7. Fuc-TIII might preferentially fucosylate core 2 O-linked glycans with longer polylactosamine that are not readily identified by amine-adsorption high performance liquid chromatography under the conditions used.
The importance of appropriate core 2 O-linked glycans is underscored by the observation that P-selectin bound better to PSGL-1 co-expressed with C2GnT and Fuc-TIII on CHO cells than to PSGL-1 co-expressed with C2GnT and Fuc-TIII on COS cells. It is possible that PSGL-1 requires modification of only a few specific O-linked glycans to bind optimally to P-selectin. Modifications of other O-linked glycans may ensure optimal binding to E-selectin. PSGL-1 expressed on CHO cells shares several structural and functional features with PSGL-1 on leukocytes. However, detailed characterization of the O-linked structures at specific sites on native and recombinant PSGL-1 is needed to define the requirements for optimal binding to Pand E-selectin.