Structures of the O-glycans on P-selectin glycoprotein ligand-1 from HL-60 cells.

P-selectin glycoprotein ligand-1 (PSGL-1) is a disulfide-bonded homodimeric mucin-like glycoprotein on leukocytes that interacts with both P- and E-selectin. In this report we describe the structures of the Ser/Thr-linked O-glycans of PSGL-1 synthesized by HL-60 cells metabolically radiolabeled with 3H-sugar precursors. In control studies, the O-glycans on CD43 (leukosialin), a mucin-like glycoprotein also expressed by HL-60 cells, were analyzed and compared to those of PSGL-1. O-Glycans were released from Ser/Thr residues by mild base/borohydride treatment of purified glycoproteins, and glycan structures were determined by a combination of techniques. In contrast to expectations, PSGL-1 is not heavily fucosylated; a majority of the O-glycans are disialylated or neutral forms of the core-2 tetrasaccharide Galβ1→4GlcNAcβ1→6(Galβ1→3)GalNAcOH. A minority of the O-glycans are α-1,3-fucosylated that occur as two major species containing the sialyl Lewis x antigen; one species is a disialylated, monofucosylated glycan, and the other is a monosialylated, trifucosylated glycan having a polylactosamine backbone. CD43 lacks the fucosylated glycans found on PSGL-1 and is enriched for the nonfucosylated, disialylated core-2 hexasaccharide. These results demonstrate that PSGL-1 contains unique fucosylated O-glycans that are predicted to be critical for high affinity interactions between PSGL-1 and selectins.

P-selectin is a Ca 2ϩ -dependent cell adhesion molecule expressed by activated platelets and endothelium and is a member of the selectin family of receptors that also includes Land E-selectin. During early steps in the inflammatory response, P-selectin, which is rapidly redistributed to the surface of activated endothelial cells, initiates attachment and rolling events between these cells and circulating leukocytes (1,2). E-selectin, which is inducibly synthesized and expressed on activated endothelial cells in a delayed fashion, also binds leukocytes and mediates rolling (1,2). P-selectin is similar to other selectins in its ability to interact weakly with a variety of sialylated, fucosylated, and sulfated glycans, including those containing the sialyl Lewis x antigen (sLe x ) 1 NeuAc␣233Gal␤134(Fuc␣133)GlcNAc␤1-R (3). However, Pselectin binds with high affinity to a specific cell surface glycoprotein on human leukocytes, termed P-selectin glycoprotein ligand-1 (PSGL-1) (4). Studies with mAbs indicate that PSGL-1 is the primary determinant mediating rolling of leukocytes on P-selectin under physiological shear forces (5)(6)(7). PSGL-1 is also recognized by E-selectin, although the binding sites for E-selectin are not identical to those for P-selectin (7)(8)(9)(10)(11).
PSGL-1 is a mucin-like, homodimeric, disulfide-bonded glycoprotein with ϳ120-kDa subunits (4). The cDNA for PSGL-1 isolated from a cDNA library of HL-60 cells predicts a trans-membrane glycoprotein of 402 amino acids (9). The mature polypeptide, after signal sequence removal and amino-terminal processing, has a predicted extracellular domain of 267 amino acids with three consensus sites for N-glycosylation (9). In this extracellular domain are also 56 Thr and 14 Ser residues, many of which may be O-glycosylated (4,8,12). A number of experimental results have indicated that the O-glycans on PSGL-1 are responsible for high affinity interactions between PSGL-1 and P-selectin. Enzymatic removal of N-glycans from PSGL-1 does not affect binding to P-selectin (4). Recombinant forms of PSGL-1 lacking N-glycans still bind P-selectin, whereas mutations in Thr residues in the amino-terminal domain of PSGL-1 decrease binding to P-selectin (13,14).
The structures of the glycans on PSGL-1 responsible for binding to P-and E-selectin are not known. Both sialylation and fucosylation of PSGL-1 are important for interactions with P-and E-selectin (4,9,10,18). Recent studies on the posttranslational glycosylation of recombinant PSGL-1 indicate that O-glycans containing the core-2 motif GlcNAc␤136-(Gal␤133)GalNAc␣13 Ser/Thr are necessary for high affinity binding of PSGL-1 to P-and E-selectin (19). However, proper glycosylation of PSGL-1 is necessary but not sufficient for high affinity binding to P-selectin. PSGL-1 also contains tyrosine sulfate residues near the amino terminus that are essential for high affinity interactions with P-selectin, but not with Eselectin (13,14,19,20).
Although carbohydrates on PSGL-1 are critical for binding to selectins, no detailed chemical structures of the glycans are available. Much of the information about the glycosylation of the molecule has been obtained by enzymatic treatments of the native ligand and by studies on recombinant forms of PSGL-1 expressed in various cell types. While these indirect methods can provide valuable information about critical determinants on the ligand, detailed structural information on O-glycans from native PSGL-1 is essential to identify glycans that are important for ligand function and to provide a clearer understanding of why PSGL-1 is a ligand for P-and E-selectin, whereas other mucins such as CD43 are not. Here we describe the structures of the O-glycans on PSGL-1 synthesized by HL-60 cells that were metabolically radiolabeled with 3 H-sugar precursors. We have compared the glycosylation of PSGL-1 with that of CD43 to determine whether two sialomucins expressed by the same cells are O-glycosylated differently. The HL-60 cell line was used in these studies because the posttranslational modifications known to be important for binding to P-and E-selectin on both HL-60 and neutrophil PSGL-1 are comparable (4,5,9,20).
Our studies demonstrate that the majority of O-glycans of PSGL-1 are disialylated or neutral forms of the core-2 tetrasaccharide. Less than 15% of the O-glycans are fucosylated and these contain the structural determinant for the sLe x antigen. These results demonstrate that PSGL-1 is glycosylated differently from CD43 and that PSGL-1 contains unique O-glycans that are likely to be critical for high affinity interactions with P-and E-selectin.
Immunoprecipitation of CD43-CD43 was immunoprecipitated from 3 H-sugar-labeled HL-60 cells using a CD43-specific mAb, H5H5 (IgG1). The H5H5 hybridoma cell line was produced by Dr. T. August and obtained from the Developmental Hybridoma Bank (The Johns Hopkins University School of Medicine). CD43 was immunoprecipitated using described procedures (8). The immunoprecipitates were analyzed by SDS-PAGE under nonreducing conditions, followed by fluorography.
To define the Man:Fuc ratio and provide standard high mannosetype N-glycans, radiolabeled glycoproteins were prepared from [ 3 H]Man-labeled HL-60 cell extracts. For the preparation of glycopeptides, the cell extracts were precipitated with trichloroacetic acid and treated with Pronase as described previously (24). From these glycopeptides, high mannose-type N-glycan standards (Man 9 GlcNAc 1 , Man 8 GlcNAc 1 , Man 7 GlcNAc 1 , Man 6 GlcNAc 1 , and Man 5 GlcNAc 1 ) were prepared following treatment with endo-␤-N-acetylglucosaminidase H (31).

Relative Molar Ratios of Radiolabeled Monosaccharides in [ 3 H]GlcNlabeled
Glycans-In the [ 3 H]GlcN-labeled, disialylated hexasaccharide NeuAc␣233Gal␤134GlcNAc␤136(NeuAc␣233Gal␤133)GalNAcOH, prepared from HL-60 cells, the NeuAc, GlcNAc, and GalNAcOH residues are radiolabeled (32). During equilibrium radiolabeling over a 48-h period, the relative molar ratios for these residues should be close to unity. Longer labeling times do not affect the distribution of radioactivity between GlcN, GalN, and NeuAc, confirming that equilibrium has been attained. To determine the relative molar ratio for GlcNAc and GalNAc residues, this [ 3 H]GlcN-hexasaccharide was desialylated to generate a [ 3 H]GlcN-tetrasaccharide; this tetrasaccharide was hydrolyzed in strong acid and the released radioactivity was identified by high pH anion exchange chromatography (HPAEC) as described below. All radioactivity was recovered in [ 3 H]GlcN and [ 3 H]GalNOH in the ratio of [ 3 H]GlcN:[ 3 H]GalNOH of 1.0:0.8. This ratio was used as a correction factor for calculating relative molar ratios of isolated glycans, i.e. the radioactivity recovered in [ 3 H]GalN(OH) in a sample after hydrolysis was divided by 0.8. The relative molar ratio we observed is consistent with other studies on HL-60 cells metabolically radiolabeled with the [ 3 H]GlcN precursor (17). To determine the relative molar ratio for NeuAc and GlcNAc residues, the [ 3 H]GlcN-hexasaccharide was desialylated with neuraminidase, and the released radioactivity in NeuAc was determined. The radioactivity recovered in GlcN and NeuAc was in the ratio of 1.0:1.4, respectively. Since NeuAc radioactivity was derived from two residues of the disialylated hexasaccharide, this gave a final value for the relative molar ratio for GlcN:NeuAc of 1.0:0.7.
Determination of Monosaccharide Composition-The ratios of GlcN: GalN and Man:Fuc in PSGL-1 and CD43 were determined following strong acid hydrolysis of excised gel slices containing purified PSGL-1 and CD43. The gel slices were treated with 250 l of 2 N trifluoroacetic acid at 121°C for 2 h. The released, radiolabeled monosaccharides in the hydrolysate were identified by HPAEC on a CarboPac PA-1 column (4 ϫ 250 mm) in a Dionex system and elution with 16 mM NaOH for 30 min. The relative molar ratio for GlcN:GalN in glycans was calculated by dividing the radioactivity in the GlcN peak by the radioactivity in the GalN peak and correcting for differences in specific activity of [ 3 H]GlcN versus [ 3 H]GalN. The Man:Fuc ratio in [ 3 H]Man-glycans was also determined following acid hydrolysis as described above, using a Man:Fuc ratio of 1.0:1.0, which is typically observed after equilibrium radiolabeling of cells with [2-3 H]Man (24,32,33).
Miscellaneous Procedures-Enzymatic treatments of glycans with ␤-N-acetylhexosaminidase, ␤-galactosidase, Arthrobacter neuraminidase, Streptomyces ␣-1,3/4-fucosidase, and E. freundii endo-␤-galactosidase were performed as described previously (34,35). Digestion with Newcastle disease virus (NDV) neuraminidase was performed in 20 l of 10 mM phosphate, pH 7.0, with 20 milliunits of enzyme for 24 h at 37°C, followed by addition of another 20 milliunits of enzyme and further incubation for 24 h. O-Glycans were analyzed and purified by descending paper chromatography on Whatman filter paper for the times noted using the pyridine/ethyl acetate/water/acetic acid (5:5:1:3) solvent system, as described previously (31). Glycans were chemically defucosylated by treatment with 0.1 N trifluoroacetic acid at 100°C for 1 h as described elsewhere (36). (Fig. 1A). The [ 3 H]GlcN-PSGL-1 was purified by affinity chromatography on a column of immobilized P-selectin. The bound, EDTA-eluted fraction isolated after the first chromatography (1ϫ) was repurified by a second chromatography (2ϫ) on the P-selectin column. Virtually all (90 -99%) of the radiolabel in the 2ϫ-purified material bound P-selectin. This two-step procedure was necessary to remove a contaminating glycoprotein of ϳ120 kDa under nonreducing conditions that remained after the first step. The purified PSGL-1 behaved as a dimer of ϳ250 kDa in nonreducing conditions and ϳ120 kDa in reducing conditions. CD43 was immunoprecipitated using a CD43 mAb, and electrophoretically separated under nonreducing conditions. A single band of 120 kDa for CD43 was observed in both nonreducing (Fig. 1B) and in reducing conditions (data not shown).

Purification of PSGL-1 and CD43-PSGL-1 and CD43 were purified from HL-60 cells metabolically labeled with [ 3 H]GlcN
Composition of Radiolabeled Sugars in PSGL-1 and CD43-In the initial assessment of the glycosylation of PSGL-1 and CD43, we determined the ratio of GlcN:GalN in the purified glycoproteins. [ 3 H]GlcN is metabolized by animal cells into radiolabeled GlcNAc, GalNAc, and sialic acids (32). Gel slices containing the [ 3 H]GlcN-glycoproteins were treated with strong acid (which results in destruction of sialic acids), and the radiolabeled GlcN and GalN were identified by Dionex HPAEC. The GlcN:GalN ratio was determined to be Ϸ2:1 for PSGL-1 and Ϸ1:1 for CD43. The GlcN:GalN ratio of Ϸ1:1 for CD43 is consistent with published evidence that a majority of the glycans in CD43 have the simple core-2 motif Gal␤13 4GlcNAc␤136(Gal␤133)GalNAcOH (15,17). These results demonstrate that the glycans of PSGL-1 contain higher amounts of GlcNAc relative to GalNAc than the glycans of CD43.
The presence of the sLe x determinant on PSGL-1 has led to expectations that PSGL-1 might be heavily fucosylated (8). We assessed the amount of Fuc present on PSGL-1 and CD43 isolated from HL-60 cells metabolically radiolabeled with [2-3 H]Man. This labeled precursor is metabolized by cells to [2-3 H]Fuc, and the relative specific activity of Man and Fuc after equilibrium labeling is equivalent (32). [ 3 H]Man-PSGL-1 and -CD43 were isolated by SDS-PAGE and fluorography. The corresponding bands were subjected to strong acid hydrolysis, and the released monosaccharides were separated by HPAEC on a Dionex system. The Man:Fuc ratio was determined to be 3:5 for PSGL-1 and 3:2 for CD43. As a control, the Man:Fuc ratio was also determined for the total unpurified glycoproteins from HL-60 cells and found to be 3:1. Thus, PSGL-1 contains more Fuc residues than CD43 and more Fuc residues than average glycoproteins in HL-60 cells.
Using this information it is possible to estimate the number of Fuc residues on PSGL-1. The cDNA sequence of PSGL-1 predicts that PSGL-1 has three potential N-glycosylation sites (9). PSGL-1 contains only complex-type N-glycans, each of which should have 3 Man residues (4,8,12). Thus, 3 complextype N-glycans on PSGL-1 represent 9 Man residues per mol and, correspondingly, there are Ϸ15 Fuc residues per mol of PSGL-1. In contrast, CD43 contains only a single N-linked glycan (16) and has much less Fuc in comparison to PSGL-1. Taken together, the compositional analyses from [ 3 H]GlcNand [ 3 H]Man-glycoproteins demonstrate that PSGL-1 is glycosylated differently than CD43.
The P-4-I samples from both PSGL-1 and CD43 were further purified by chromatography on a column of Bio-Gel P-10. Radioactivity in the P-4-I fraction from PSGL-1 separated into two major peaks on Bio-Gel P-10 designated P-10-1 and P-10-2 (Fig. 2B). This population of glycans is larger in size than the disialylated core-2 hexasaccharide standard NeuAc␣23 3Gal␤134GlcNAc␤136(NeuAc␣233Gal␤133)GalNAcOH, whose elution position is indicated by the dashed line. The major material in the P-4-II fraction from PSGL-1 eluted on Bio-Gel P-10 in a position slightly smaller (designated P-10-3) than the disialylated core-2 hexasaccharide standard (Fig. 2B). In contrast, most of the material in both the P-4-I and P-4-II fractions of CD43 eluted identically on Bio-Gel P-10 as the disialylated core-2 hexasaccharide standard (Fig. 2B).
The glycans recovered in P-4-I from CD43 were analyzed using exoglycosidase treatments and anion exchange chromatography, as described previously (15). The glycans were shown to be the expected disialylated core-2 hexasaccharide (data not shown). A small fraction of the P-4-I sample from CD43 was recovered in larger sized glycans on Bio-Gel P-10 ( Fig. 2B), consistent with previous studies showing that a small fraction of O-glycans from CD43 have an extended polylactosamine structure on the core-2 motif (17). Since the structures of glycans in CD43 have been described (15,17), they were not further analyzed.
Compositional Analyses of P-10-1, P-10-2, and P-10-3 Glycans from PSGL-1-O-Glycans released from Ser/Thr residues by ␤-elimination should contain [ 3 H]GalNAcOH at the reducing terminus, which is recoverable as [ 3 H]GalNOH following strong acid hydrolysis. To identify which glycans in the mixture from PSGL-1 represent the O-glycans, the GalNOH, GalN, and GlcN content of each glycan from Bio-Gel P-10 was determined by HPAEC on a Dionex system following strong acid hydrolysis. More than 95% of the total GalN(OH) in PSGL-1 P-10-2 and P-10-3 fractions was recovered as GalNOH, demonstrating that ␤-elimination was efficient and that the O-glycans of PSGL-1 are represented in these fractions. The ratio of GlcN to GalNOH in P-10-2 and P-10-3 was 1.2:1 and 0.8:1, respectively. The glycans in P-10-1 lack GalNOH and do not represent O-glycans released by ␤-elimination. Instead, the P-10-1 fraction contains N-glycans still attached to peptide. This was confirmed by the presence of [ 3 H]Man recovered in these glycans from [ 3 H]Man-PSGL-1 (data not shown). P-10[hyphen1 contained a small amount of unreduced GalN, which might arise from GalNAc residues present in N-glycans (37), or from a small amount of residual O-glycans still linked to peptide and not released during ␤-elimination procedures.
To determine whether the anionic character of the glycans was due to sialic acid and to define the linkage of sialic acid, portions of the [ 3 H]GlcN-labeled O-glycans were treated with neuraminidase from NDV. This enzyme displays high specificity for ␣-2,3-linked sialic acid residues and will not efficiently cleave sialic acid in other linkages (38). After NDV neuraminidase treatment, the glycans in P-10-1 were less charged, consistent with a loss of sialic acid (Fig. 3B). However, the presence of residually charged glycans is indicative of the profile expected for N-glycans in glycopeptides. In contrast, the glycans in P-10-2 and P-10-3 became neutral following NDV neuraminidase treatment, demonstrating that all sialic acids in these glycans are ␣-2,3-linked (Fig. 3, D and F). The peak of material eluting with 20 mM NaCl following NDV neuraminidase treatment was quantitatively recovered as free [ 3 H]NeuAc, as shown by its co-elution with standard NeuAc on descending paper chromatography (Fig. 3, D and F,  insets). In a previous report we established that the sialic acid on PSGL-1 from [ 3 H]GlcN-labeled HL-60 cells is Neu5Ac (12). These results demonstrate that the O-glycans in P-10-2 are mono-and disialylated species and that the O-glycans in P-10-3 are a combination of neutral and sialylated species, with sialic acid in ␣2,3-linkage to glycans. Furthermore, these results demonstrate that the O-glycans of PSGL-1 are not sulfated, since neutral species result following treatment with NDV neuraminidase.
To further confirm the presence of fucose in these glycans and to facilitate the identification of Fuc in other glycans, HL-60 cells were metabolically radiolabeled with [ 3 H]Fuc. The [ 3 H]Fuc-labeled O-glycans were recovered by ␤-elimination as described for the [ 3 H]GlcN-glycans, following the same procedures shown in Figs. 1 and 2. The [ 3 H]Fuc recovered in the desialylated P-10-2b fraction co-migrated with the desialylated [ 3 H]GlcN-P-10-2b 1 glycans (Fig. 5E). Taken together, these results demonstrate that the P-10-2b 1 glycans from PSGL-1 are fucosylated and contain the sLe x structure.
Endo-and Exoglycosidase Treatments of P-10-2a Glycans-Because of their relatively large size, we considered the possibility that the P-10-2a glycans contain polylactosamine (3Gal␤134GlcNAc␤1-) n . To assess this possibility, the P-10-2a glycans were desialylated and the sialic acid was removed by QAE-Sephadex chromatography. The resulting neutral glycans were treated with endo-␤-galactosidase, an enzyme that cleaves internal ␤134 galactosyl residues within a type 2 polylactosamine (40). The glycans were resistant to this treatment (Fig. 6A). The desialylated P-10-2a glycans were also resistant to combined treatment with ␤-galactosidase and ␤-Nacetylhexosaminidase (Fig. 6A). We then considered the possibility that the P-10-2a glycans might contain a polylactosamine backbone in which all internal GlcNAc residues are fucosylated. Such polyfucosylated, polylactosamine structures are resistant to endo-␤-galactosidase (41). We characterized polyfucosylated polylactosamines in the parasitic helminth Schistosoma mansoni and found that complete defucosylation was necessary before endo-␤-galactosidase could digest the chains (35).
The desialylated P-10-2a glycans were chemically defucosylated and then treated with endo-␤-galactosidase. After defucosylation, the enzyme quantitatively digested the P-10-2a glycans to release three major compounds in an approximate equimolar ratio identified as the trisaccharide Gal␤13 4GlcNAc␤133Gal, the residual core-2 trisaccharide GlcNAc␤136(Gal␤133)GalNAcOH, and the disaccharide GlcNAc␤133Gal (Fig. 6B). The generation of such products from the specific action of endo-␤-galactosidase is predicted for a glycan with the backbone structure (Structure 1), where the arrows indicate specific cleavage sites for endo-␤galactosidase of a defucosylated polylactosamine-containing glycan (40). The length of the polylactosamine chain can be deduced from the radioactivity recovered in the disaccharide GlcNAc␤133Gal relative to the other fragments. The recovery of the trisaccharide GlcNAc␤136(Gal␤133)GalNAcOH following endo-␤-galactosidase treatment demonstrates that the polylactosamine is extended from the GlcNAc in ␤1-6 linkage to the GalNAcOH residue. The recovery of the trisaccharide Gal␤134GlcNAc␤133Gal demonstrates that Gal residues are present in the nonreducing terminus of the polylactosamine. As expected, treatment of the desialylated and defucosylated P-10-2a glycans with a combination of ␤-galactosidase and ␤-N-acetylhexosaminidase resulted in complete digestion to free [ 3 H]GlcNAc and the 3 H-labeled disaccharide Gal␤13 3GalNAcOH (Fig. 6B).
The P-10-2a glycans are monosialylated (Fig. 4B) and could have one of two possible structures (Structure 2, a or b), To distinguish between these possibilities, the [ 3 H]GlcN-P-10-2a sialylated glycans were treated with a mixture of ␤-galactosidase, ␤-N-acetylhexosaminidase, and Streptomyces ␣-1,3/4fucosidase, with or without neuraminidase. It would be predicted that structure (a) would require neuraminidase in addition to the other enzymes for complete digestion, whereas structure (b) would be degraded by the exoglycosidases in the absence of neuraminidase. Treatment of the [ 3 H]GlcN-P-10-2a with neuraminidase alone released radiolabeled NeuAc, as expected (Fig. 8). When the glycans were treated with a mixture of ␤-galactosidase, ␤-N-acetylhexosaminidase, and Streptomyces ␣-1,3/4-fucosidase, in the absence of neuraminidase, no radioactivity was released. Inclusion of neuraminidase with other exoglycosidases resulted in complete degradation of the glycans to free [ 3 H]NeuAc, free [ 3 H]GlcNAc, and the [ 3 H]disaccharide Gal␤133GalNAcOH (Fig. 8). Taken together, these results demonstrate that the single sialic acid residue on the P-10-2a glycans is present in a terminal position on the polylactosamine chain, shown in structure (a), and that these glycans have the sLe x structure (Table I).
Endo-and Exoglycosidase Treatments of P-10-3 Glycans-The P-10-3 glycans were primarily neutral or monosialylated species (Fig. 3E). The predominant species (Ϸ80% of the radioactivity) was nonsialylated. The P-10-3 fraction was treated with neuraminidase, and released sialic acid was removed by chromatography on QAE-Sephadex. Approximately 90% of the desialylated species co-migrated with the standard core-2 tetrasaccharide Gal␤134GlcNAc␤136(Gal␤133)GalNAcOH and the remainder co-migrated with the core-1 disaccharide Gal␤133GalNAcOH (Fig. 9). Treatment of the desialylated P-10-3 glycans with ␤-galactosidase caused a shift in the migration of the major peak to that of the expected trisaccharide standard (Fig. 9). Treatment with a combination of ␤-galactosidase and ␤-N-acetylhexosaminidase generated free [ 3 H]Glc-NAc and [ 3 H]Gal␤133GalNAcOH (Fig. 9). Approximately 20% of the radiolabel in the P-10-3 glycans was found in predominately monosialylated species that were converted to neutral species by neuraminidase treatment (Fig. 3F). These results demonstrate that the P-10-3 glycans are primarily neutral core-2 tetrasaccharides and some monosialylated core-2 pentasaccharides (Table I). In addition, some of the neutral glycans have the core-1 disaccharide structure (Table I).
Relative  (Table I). This type of analysis was successfully used in previous studies on CD43 (15,17). The majority of O-glycans in PSGL-1 contain the core-2 structure. A minority of the glycans, recovered in P-10-2b 1 and P-10-2a, contain the sLe x determinant. DISCUSSION This study demonstrates that PSGL-1 from human HL-60 cells contains O-glycans with a core-2 motif. A majority of the glycans are not fucosylated and are mixtures of neutral and sialylated species. A minority of the O-glycans (Ϸ14%) are fucosylated and contain the terminal sLe x structure. Two types of fucosylated O-glycans are present; one type is a disialylated heptasaccharide lacking polylactosamine and the other is a unique monosialylated, trifucosylated glycan that contains polylactosamine. The presence of core-2 O-glycans on PSGL-1 from HL-60 cells is consistent with results of studies on the glycosylation of PSGL-1 from human neutrophils. Desialylated PSGL-1 from both human neutrophils and HL-60 cells is resistant to treatment with endo-␣-N-acetylgalactosaminidase (O-glycanase), which cleaves only desialylated core-1 glycans (4,8). The direct demonstration of sLe x determinants and polylactosamine on O-glycans of HL-60-derived PSGL-1 reinforces indirect evidence that these structures are present on O-glycans from neutrophil-derived PSGL-1 (8,12 has many neutral core-2 tetrasaccharides, whereas CD43 has mostly disialylated, core-2 hexasaccharides. The core-2 structure is a precursor for polylactosamine synthesis in O-glycans (43), but only PSGL-1 has significant amounts of polylactosamine. This indicates that the core-2 structure is necessary but not sufficient for polylactosamine addition. Furthermore, CD43 lacks the two species of fucosylated O-glycans found in PSGL-1. Although a monofucosylated O-glycan was identified in CD43, this species represents only 0.5% of the O-glycans in CD43 (17). The trifucosylated monosialylated O-glycan we have identified on PSGL-1 is not found on CD43.
The pathways for biosynthesis of the two types of fucosylated O-glycans on PSGL-1 are not known. Myeloid cells express both FTIV and FTVII (28 -30, 49, 50), and both enzymes can synthesize the sLe x determinant when expressed in appropriate cells (49 -51). In studies on recombinant PSGL-1 expressed in CHO cells, co-expression of C2GnT with either FTIII, FTIV, or FTVII generates a functional ligand that promotes static adhesion of transfected cells to immobilized P-selectin and binding to fluid-phase P-selectin (19). However, the specificities of these fucosyltransferases for O-versus N-glycans and for terminal lactosaminyl units versus polylactosamine have not been well studied. FTIV does not efficiently fucosylate internal Glc-NAc residues in long polylactosamines on N-glycans when expressed in CHO cells, whereas FTIII, the Lewis enzyme, does efficiently fucosylate such long polylactosamines (52). We analyzed the glycans on [ 3 H]GlcN-PSGL-1 synthesized by transfected CHO cells co-expressing C2GnT and FTIV; among the neutral glycans generated by neuraminidase treatment was a fucosylated core-2 pentasaccharide lacking polylactosamine (similar to the P-10-2b 1 glycan) (19). Perhaps the short sLe xcontaining O-glycan (P-10-2b 1 ) of PSGL-1 is generated by FTIV, and the longer sLe x -containing glycan (P-10-2a) is generated by FTVII. Studies on the structures of the O-glycans on recombinant PSGL-1 expressed in CHO cells with either FTIV or FTVII will determine whether these two enzymes differ in the fucosylation of polylactosamine sequences in O-glycans.
P-selectin can bind weakly to a variety of sulfated glycans, FIG. 6. PSGL-1 fraction P-10-2a consists of a sialylated, polylactosamine-containing O-glycan that is trifucosylated. Fraction P-10-2a derived from [ 3 H]GlcN-PSGL-1 was treated with exoglycosidases and analyzed by descending paper chromatography for 24 h. The ␤-eliminated glycans were first desialylated with neuraminidase, and the released sialic acid was separated from neutral glycans by ion-exhange chromatography. A, chromatography of desialylated glycans before and after treatments with either endo-␤-galactosidase, or a combination of ␤-galactosidase and ␤-N-acetylhexosaminidase. B, desialylated glycans were chemically defucosylated and analyzed before and after treatments with either endo-␤-galactosidase or a combination of ␤-galactosidase plus ␤-N-acetylhexosaminidase. Migration of authentic standards is indicated: 3, Gal␤134GlcNAc␤133Gal␤1; 3*, GlcNAc␤136(Gal␤133)GalNAcOH; 2, Gal␤134GlcNAc; 1, GlcNAc.  and these glycans inhibit binding of P-selectin to human myeloid cells (53)(54)(55)(56). However, the O-glycans of PSGL-1 are not sulfated. Instead, PSGL-1 contains tyrosine sulfate that is required for interactions with P-selectin but not with E-selectin (13,14,19,20). Three consensus sites for tyrosine sulfation occur at the amino terminus of PSGL-1 at residues 46, 48, and 51 (9). PL1, a mAb to PSGL-1, blocks binding to P-selectin and recognizes an epitope spanning residues 49 -62 that overlaps the tyrosine sulfation sites (57). Near the tyrosine sulfation sites are two Thr residues that represent potential O-glycosylation sites at residues 44 and 57. Mutations in these residues reduce binding of PSGL-1 to P-selectin when PSGL-1 is coexpressed in COS cells with either FTIII or FTVII (13,14). These results suggest that only one or two O-glycans in conjunction with tyrosine sulfate residues may be sufficient to promote high affinity binding of PSGL-1 to P-selectin. However, O-glycans in other regions of the molecule may also contribute to interactions with P-and E-selectin.
It was originally suggested that mucin-like glycoproteins act as convenient scaffolds upon which many O-glycans can be clustered for recognition by selectins (1). However, our data, in conjunction with other studies, indicate that mucin-like glycoproteins are differentially glycosylated. Further studies are required to identify the factors regulating differential glycosylation of sialomucins and to address the possibility that there are site-specific differences in the structures of O-glycans in these mucins. With regard to PSGL-1, further studies are needed to identify whether the fucosylated O-glycans occur at specific sites and to determine which structural features of the fucosylated O-glycans are required for recognition by P-and E-selectin.