Like all mucins, PSGL-1 has O-glycans attached to many serine and threonine residues in the extracellular domain. The structures of the O-glycans have been determined for PSGL-1 derived from the human myeloid HL-60 cell line . Most have relatively short, sialylated core 1 or branched core 2 structures. None are detectably sulfated. Remarkably, even though all PSGL-1 molecules from HL-60 cells bind to P-selectin, only a minority of the O-glycans are fucosylated . Fucosylation of O-glycans on murine leukocytes appears to be even more rare . This suggests that a1-3-fucosyltransferases have site-specific preferences for modifying O-glycans on PSGL-1. It further implies that most of the O-glycans function only to extend the polypeptide backbone.
The first suggestion of a site-specific location of a critical fucosylated O-glycan arose from the discovery that mAbs with epitopes at the extreme N-terminal region of human and murine PSGL-1 block binding to both P- and L-selectin [13, 14, 2125]. The mAbs to the N-terminal region of human and murine PSGL-1 that block binding to P- and L-selectin in biochemical assays also block rolling of leukocytes on P- or L-selectin under flow conditions. Such studies first demonstrated that PSGL-1 is the dominant ligand for leukocyte rolling on these two selectins. In vitro, leukocytes use interactions of PSGL-1 with L-selectin to roll on adherent leukocytes or to initiate leukocyte aggregation [21, 24]. These leukocyte-leukocyte interactions lead to secondary tethering of leukocytes to a P- or E-selectin surface , a potential mechanism to amplify leukocyte recruitment to the vessel wall.
In human PSGL-1, the binding site for P- and L-selectin comprises a peptide sequence containing three tyrosine sulfate residues near a threonine to which a specific O-glycan is attached (Fig. 1B). Biochemical assays and expression of recombinant forms of PSGL-1 have demonstrated that optimal binding to P-selectin and L-selectin requires sulfation of the tyrosines and addition of a branched, core 2 O-gly-can capped with sLex to the threonine [27-32]. Targeted deletion of the murine gene encoding core2GlcNAcT-I, the major core 2 |31-6-N-acetylglucosaminyltransferase in leukocytes, eliminates binding of leukocytes to P-selectin . This suggests that core2GlcNAcT-I has the essential role in constructing core 2 O-glycans on PSGL-1. There are two tyrosyl protein sulfotransferases that catalyze addition of sulfate esters to tyrosine residues. Targeted deletion of the gene for either enzyme does not impair leukocyte binding to selectins [34, 35], suggesting that both enzymes contribute to tyrosine sulfation of PSGL-1.
The structural requirements for binding P- and L-selectin have been more definitively mapped by synthesis of glycosulfopeptides modeled after the N-termi-nal region of PSGL-1 [36-38] (Fig. 1B). Each sulfate contributes to binding affinity, and the peptide itself confers weak binding. The fucose moiety is essential for binding, whereas the sialic acid plays a lesser role. The position of the O-glycan in relation to the tyrosine sulfates is critical. Even the core backbone of the O-glycan is important: a short core 2 O-glycan capped with sLex supports binding, whereas an isomeric core 1 O-glycan does not . Furthermore, a glycosulfopeptide with an extended core 2 branch containing fucosylated polylactosamine capped with sLex binds poorly . These studies reveal specific stereochemical requirements for optimal binding of PSGL-1 to P- and L-selectin. Relative to P-selectin, L-selectin binds to PSGL-1 with lower affinity and more rapid dissociation kinetics [36-39]. Therefore, there are subtle but important differences in specificity that may be dictated by specific residues in the lectin domain of each selectin.
X-ray crystallography was used to solve the structure of the lectin and EGF domains of P-selectin bound to a recombinant N-terminal fragment of PSGL-1 . The PSGL-1 fragment is sulfated on all three tyrosines, and the sequence of the core 2 O-glycan is nearly identical to the one in the synthetic glycosulfopeptide that confers optimal binding to P-selectin (Fig. 1B). The PSGL-1 fragment binds to a large but shallow surface on the lectin domain, opposite to where the EGF domain is attached. The fucose has multiple binding interactions, some of which participate in coordinating the single Ca2+ ion in the lectin domain. The galactose and sialic acid residues make fewer contacts. Other regions of the lectin domain contact certain amino acids plus the sulfates of the middle and C-terminal tyrosines of PSGL-1. Monomeric P-selectin binds with equivalent affinity to native or recombinant PSGL-1 and to the synthetic glycosulfopeptides, with dissociation constants estimated at ~80-800 nM [3, 36, 37, 40, 41]. In contrast, P-selectin binds to carbohydrates containing only sLex with dissociation constants of 1-10 mM [42-44]. The multiple interactions of carbohydrate, amino acids, and sulfate residues with the lectin domain explain why PSGL-1 binds to P-selectin with much higher affinity than do peptide-free oligosaccharides that contain sLex. The shallow binding site on P-selectin may contribute to the rapid binding kinetics. The N-terminal tyrosine sulfate is not visualized in the co-crystal structure, which is surprising since a recombinant form of PSGL-1 or a glycosulfopeptide with sulfation restricted to this tyrosine still binds to P-selectin with appreciable affinity [31, 32, 37]. Interestingly, the N-terminal region of murine PSGL-1, which can bind to both human and murine P-selectin, has only two tyrosines that might be sulfated . Compared to the human sequence, these tyrosines are positioned much closer to the two threonines that are the best candidates for attachment of a fucosylated O-glycan. Mutagenesis studies suggest that only one tyrosine and one threonine contribute to binding . Thus, there may be additional ways in which P-selectin can bind to both sLex and a sulfated peptide segment.
PSGL-1 binds differently to E-selectin than to P- and L-selectin. Sialylation and a1-3-fucosylation of glycans are required for binding, and at least some of these modifications appear to be on core 2 O-glycans . However, tyrosine sulfation of PSGL-1 is not required for binding to E-selectin [28, 29]. Because mAbs to the N terminus of PSGL-1 do not block interactions with E-selectin , there may be other binding sites on PSGL-1 [47, 48], although the limited number of fucosylated O-glycans should restrict the number of these sites .
All myeloid cells bind to all three selectins, suggesting that they express the full complement of glycosyltransferases required to construct selectin ligands on PSGL-1 and other glycoproteins . In contrast, most circulating T and B cells do not bind to selectins [14, 49]. During the transition to the effector cell phenotype, T cells acquire selectin-ligand binding function, which appears to be primarily due to up-regulated expression of core2GlcNAcT-I and Fuc-TVII . In vitro, cytokines such as TNF-a induce endothelial cells to express glycosyltransferases that modify PSGL-1 so that it can bind selectins . In vivo, however, available evidence does not support a general increase in expression of functional PSGL-1 on endothelial cells at all sites of inflammation.
Unlike most leukocytes, some subsets of natural killer cells and dendritic cells express glycoforms of PSGL-1 that have glycans with sulfate esters attached to the C-6 position of N-acetylglucosamine residues [51, 52]. The functional significance of these modifications is not clear, although PSGL-1 on natural killer cells has been reported to bind L-selectin . This might be due to attachment of a sulfate ester to the C-6 position of the N-acetylglucosamine residue of sLex to produce 6-sulfo-sLex, a well-characterized binding determinant for L-selectin on the glycans of mucins from lymph node endothelial cells .
Was this article helpful?