Molecular basis of leukocyte rolling on PSGL-1. Predominant role of core-2 O-glycans and of tyrosine sulfate residue 51.

Interactions between the leukocyte adhesion receptor L-selectin and P-selectin glycoprotein ligand-1 play an important role in regulating the inflammatory response by mediating leukocyte tethering and rolling on adherent leukocytes. In this study, we have examined the effect of post-translational modifications of PSGL-1 including Tyr sulfation and presentation of sialylated and fucosylated O-glycans for L-selectin binding. The functional importance of these modifications was determined by analyzing soluble L-selectin binding and leukocyte rolling on CHO cells expressing various glycoforms of PSGL-1 or mutant PSGL-1 targeted at N-terminal Thr or Tyr residues. Simultaneous expression of core-2 beta1,6-N-acetylglucosaminyltransferase and fucosyltransferase VII was required for optimal L-selectin binding to PSGL-1. Substitution of Thr-57 by Ala but not of Thr-44, strongly decreased L-selectin binding and leukocyte rolling on PSGL-1. Substitution of Tyr by Phe revealed that PSGL-1 Tyr-51 plays a predominant role in mediating L-selectin binding and leukocyte rolling whereas Tyr-48 has a minor role, an observation that contrasts with the pattern seen for the interactions between PSGL-1 and P-selectin where Tyr-48 plays a key role. Molecular modeling analysis of L-selectin and P-selectin interactions with PSGL-1 further supported these observations. Additional experiments showed that core-2 O-glycans attached to Thr-57 were also of critical importance in regulating the velocity and stability of leukocyte rolling. These observations pinpoint the structural characteristics of PSGL-1 that are required for optimal interactions with L-selectin and may be responsible for the specific kinetic and mechanical bond properties of the L-selectin-PSGL-1 adhesion receptor-counterreceptor pair.

Previous observations indicated the involvement of O-glycans attached to Thr-57 and tyrosine sulfate residues in supporting L-selectin-and P-selectin-mediated rolling (35). In the present study, we characterized the PSGL-1 determinants that interact with L-selectin. Adhesion studies indicated that Oglycosylation of Thr-57 and sulfation of Tyr-46 and -51 play a critical role in supporting recombinant L-selectin binding to PSGL-1 and leukocyte rolling on PSGL-1. In addition, these determinants were shown to play a major role in stabilizing rolling velocity, a key feature for the regulation of leukocyte exposure to chemotactic stimuli that lead to cell arrest and firm adhesion. By contrast, Tyr-48 played only a minor role in supporting L-selectin-mediated adhesion whereas it was shown to be critical for P-selectin-mediated interactions with PSGL-1 (23,34).
cDNAs-PSGL-1 cDNA was a gift from Genetics Institute (Boston, MA) (39), fucosyltransferase VII (FucT-VII) cDNA from J-B Lowe (Howard Hughes Institutes, Ann Harbor, MI) and core-2 ␤1,6-N-acetylglucosaminyltransferase transferase (C2GnT) from M. Fukuda (the Burnham Institute, La Jolla Cancer Research Center, San Diego, CA). The cDNA sequence encoding the internal ribosome entry site (IRES) of the encephalomyocarditis virus sequence was a gift from P. Aebischer (EPFL, Lausanne, Switzerland). The pZeoSV and the pcDNA3.1 vectors were from Invitrogen (Groningen, The Netherlands). The IRES sequence was inserted in the multiple cloning sites of the pZeoSV vector (pIRES Zeo SV vector). C2GnT and FucT-VII cDNAs were then subcloned in the pIRES Zeo SV vector to permit the translation of C2GnT and FucT-VII sequences from one mRNA. The expression cassette was constructed by inserting the C2GnT sequence followed by the IRES and the FucT-VII cDNA sequences in the multiple cloning site of the pZeo SV vector. This type of vector allows the expression of the target gene (C2GnT) and the selection marker (sLe x expression associated to FucT-VII activity) from the same promoter so that virtually all transfected cells expressing the selection marker also express the gene of interest (40 -42).
Cells and Transfections-Heparinized blood samples were obtained from healthy donors. Lymphocytes were isolated by blood centrifugation on Ficoll-Hypaque and monocyte depletion by adherence on plastic (43). Neutrophils were isolated from Ficoll-Hypaque pellets by dextran sedimentation and erythrocyte hypotonic lysis. CHO/dhfr Ϫ cells (ATCC number: CRL 9096) were stably transfected with cDNAs encoding wildtype or mutant PSGL-1, subcloned in pCDNA3.1 vector. When indicated, CHO cells were co-transfected with FucT-VII cDNA subcloned in pZeoSV vector (Invitrogen) or in the pIRES Zeo SV expression vector, which allows the simultaneous translation of C2GnT and FucT-VII sequences. Transfections were performed using LipofectAMINE™ Plus (Invitrogen). CHO dhfr Ϫ were cultured in MEM␣ medium containing ribonucleotides, deoxyribonucleotides, and 10% fetal calf serum; COS-7 cells were cultured in Dulbecco's modified Eagle's medium/10% fetal calf serum. Transfectants were selected in medium containing 400 g/ml G418 (Invitrogen) and, when required, 200 g/ml Zeocin (Invitrogen). Individual clones expressing high levels of the various forms of PSGL-1, C2GnT, and FucT-VII-dependent expression of sLe x and the cutaneous lymphocyte antigen (CLA) were isolated by limiting dilutions and identified by immunophenotypic analysis. The expression of PSGL-1, sLe x , and CLA was assessed using, respectively, PL2, CSLEX-1, and HECA-452 mAbs. CHO cells, which were selected for adhesion studies, expressed similar levels of PSGL-1 and of sLe x /CLA. The expression of C2GnT and FucT-VII mRNAs in CHO cells expressing wild-type or mutant PSGL-1 was detected by RT-PCR using the one tube Titan System™ (Roche Diagnostic, Rotkreuz, Switzerland). Sequences of forward and reverse primers used for C2GnT PCR amplification were GGCAGTGCCTACTTCGTGGTCA and ATGCTCATCCAA-ACACTGGATGGCAAA; for FucT-VII, CCCACCGTGGCCCAGTACCG-CTTCT and CTGACCTCTGTGCCCAGCCTCCCGT; for PSGL-1, ATG-CCTCTGCAACTCCTCCT and CTGCTGAATCCGTGGACAGGTT. The expression levels of the various forms of PSGL-1 by CHO cells were determined by the measurement of antigen site density using the DAKO QIFIKIT ® (Dako, Glostrup, Denmark). Antibody binding site density was calculated (44) using PL2 mAb to assess PSGL-1 expression. PL2 binding sites were found equal to 115 Ϯ 37 (mean Ϯ S.D.) binding sites/m 2 in CHO cells expressing constructs used to perform experiments illustrated in Figs. 4 -8. CHO cells used in experiments illustrated in Figs. 2 and 3 expressed higher levels of PSGL-1 (247 Ϯ 19 PL2 binding sites/m 2 ). The determination of sLe x and CLA expression indicated that CHO cells used for L-selectin/ binding studies and adhesion assays expressed comparable levels of FucT-VII activity.
Immunophenotypic Analysis-One or two color flow cytometry analysis was carried out by incubating cells with appropriate unlabeled mAbs, FITC-, PE-conjugated mAbs (10 g/ml) or soluble adhesion receptors (L-selectin/ or CD4/ chimera at 50 g/ml) (37,38). When required goat anti-mouse IgG-FITC (Tago BIOSOURCE Europe S.A., Nivelles, Belgium), goat anti-mouse IgM-FITC (Tago BIOSOURCE), or rabbit anti-rat IgM-FITC were used as secondary antibodies (Dako). L-selectin/ and CD4/ chimeric proteins were suspended in phosphate-buffered saline containing 1% bovine serum albumin and 1 mM CaCl 2 . Cell surface binding of chimeric proteins was detected using a polyclonal FITC-conjugated rabbit anti-human IgM heavy chain antibody (Dako). The specificity of L-selectin/ binding to PSGL-1 was indicated by the abrogation of L-selectin/ binding in presence of 5 mM EDTA or 100 g/ml anti-L-selectin mAb LAM1-3 or 10 g/ml anti-PSGL-1 mAb KPL1. In experiments performed to evaluate the role of sialic acid residues in supporting neutrophil rolling on PSGL-1, CHO-PSGL-1 cells, co-expressing C2GnT and FucT-VII, were cultured for 30 min at 37°C in PBS containing 0.1 units/ml Vibrio cholerae neuraminidase (Roche Diagnostics). The reactivity of the mAb CSLEX-1 with sLe x was abrogated by this treatment. Flow cytometry was performed with a Epics XL-MCL cytofluorimeter (Coulter Electronics, Hialeah, FL). A total of 5000 cells were analyzed in each experiment.
Cell Adhesion Assays-A laminar flow was generated in a parallel plate flow chamber (GlycoTech Corp Rockville, MD) mounted on a glass coverslip (Polylabo SA, Plan-les-Ouates, Switzerland) covered with a confluent monolayer of transfected CHO cells. Neutrophils suspended in RPMI 1640/1% fetal calf serum at 0.5 ϫ 10 6 /ml, lymphocytes (10 6 /ml), 300.19 pre-B cells stably expressing L-selectin (300.19-L-selectin cells, 0.5 ϫ 10 6 /ml) or K-562 cells cells stably expressing P-selectin (K562 P-selectin cells, 0.5 ϫ 10 6 /ml), were perfused through the chamber using a syringe pump (Harvard Apparatus, Indulab AG, Gams, Switzerland) for 5 min, at room temperature, under a constant shear stress. Leukocyte interactions with CHO cells were visualized using a phase contrast microscope (Leica Leitz DM IL, Renens, Switzerland) and a high resolution Sony CCD-IRIS videocamera (Japan). Images were recorded on an S-VHS-recorder (Panasonic MD 830, Telecom Lausanne, Switzerland). Sequential images of leukocyte interactions with transfected CHO cells were analyzed using a digital image analysis system cells every 0.032 s are illustrated in Figs. 7a and 8a and were used to assess the mean velocity Ϯ S.D. of each tracked cell over 2-5 s observation periods. The mean velocity of frame-by-frame tracked cells was included between percentile 25-75 of the velocity of each cell population illustrated in Fig. 6. The S.D. value of the mean velocity of each tracked cell was then used to calculate the mean Ϯ S.D. of cell-rolling velocities of each cell population. The mean Ϯ S.D. was used as an indicator of the variation of cell-rolling velocity. 373-1477 independent determinations of frame-by-frame velocity were measured for each tested conditions. The anti-L-selectin mAb LAM 1-3 and the anti-PSGL-1 mAbs PL1 or KPL1 were used as inhibitors of L-selectin interaction with PSGL-1. The isotype-matched mAbs LAM1-14 and PL2 were used as controls. In experiments evaluating the inhibitory effect of HECA-452, a rat IgM mAb was used as control (Dako). Leukocytes expressed similar levels of L-selectin before and at the end of each experiment.
Model of PSGL-1 Interactions with L-selectin and P-selectin-P-selectin and L-selectin interactions with PSGL-1 were compared using a L-selectin homology model based on the crystal structure of P-selectin co-complexed with PSGL-1 (PDB Data Bank ID: 1G1S) (23). 66% sequence identity and 85% sequence similarity were disclosed between P-selectin and L-selectin using a dynamic programming method implemented in the MODELLER program (45). With this sequence homology, the probability that both L-selectin and P-selectin share the same fold is very high (46 -48). A root mean-squared deviation (RMSD) of 0.67 Å was calculated for all C␣ atoms between the L-selectin model and the P-selectin template with the Swiss PDB viewer program (49). Assuming that PSGL-1 interactions with L-selectin and P-selectin are highly similar, we used the coordinates of PSGL-1 (PDB Data Bank ID: 1G1S) and superimposed L-selectin on P-selectin to analyze L-selectin interactions with PSGL-1. Hydrogen-bonding pattern was analyzed using the HBPLUS program and standard geometric definitions considering the distance and the angle between the hydrogen atom and the acceptor/donor atoms (50). To compute hydrogen bonds, we used, as criteria, 3,9 Å as maximal distance between heavy atoms and 90°as minimal angle between the donor (D) atom, the hydrogen (H) atom, and the acceptor (A) atom (DHA angle), the probability of finding energetically favorable hydrogen bonds being smaller at longer distances or smaller angles (51).
The mobility of P-selectin loops was studied in detail with the Swiss PDB viewer (49) and the MOLMOL programs (52) using the crystal structure of P-selectin co-complexed with its ligands PSGL-1 (1G1S), sLe x (1G1Q), and the uncomplexed form (1G1R) (23). PSGL-1-sulfated Tyr-48 was previously reported to interact with a first loop constituted of P-selectin amino acids 42-48 and with a second loop constituted of amino acids 108 -114. PSGL-1-sulfated Tyr-51 interacts with a third loop constituted of amino acid 64 -89 (23). The mobility of each loop was assessed by calculating with the MOLMOL program the local RMSDs of the lectin domain amino acid backbone atoms between the different P-selectin structures (52)(53)(54). Regions constituted of residues with local RMSDs higher than the mean local RMSD (1.3 Å) were defined as mobile.
Statistical Analysis-Analysis of variance and the Bonferroni multiple comparison test or the Kruskal-Wallis non parametric ANOVA test were used to assess statistical significance of differences between groups. The Mann-Whitney test was used to compare the medians of two unpaired groups. p values of Ͻ0.05 were considered as significant. L-selectin/ chimera weakly interacted with CHO cells expressing FucT-VII cDNA alone or co-expressing C2GnT and FucT-VII cDNAs (Fig. 2b, upper panels). L-selectin/ bound to a much higher percentage of CHO cells when PSGL-1 was co-expressed with both FucT-VII and C2GnT (Fig. 2b, lower right panel).
Regulation of Neutrophil-rolling Velocity by Core-2 O-Glycans-Neutrophil-rolling velocity on CHO-PSGL-1/FucT-VII or CHO-PSGL-1/C2GnT/FucT-VII cells was analyzed under a constant shear stress of 1.25 dyn/cm 2 (Fig. 3b) , sLe x , and CLA. L-selectin/ strongly reacted with wildtype PSGL-1 or PSGL-1 T44A (Fig. 4a, left and center panels) whereas it only weakly bound to PSGL-1 T57A (Fig. 4a, right  panel). L-selectin/ binding was not completely inhibited by the replacement of Thr-57 by Ala indicating that additional structures support L-selectin/ binding to PSGL-1. The N-terminal tyrosine sulfation consensus is the most likely alternate potential binding site. Sialyl Le x /CLA determinants expressed at the surface of CHO cells may also play a role (Fig. 2b, upper  panels).
Additional analysis was performed to examine the frame-byframe velocity of 300.19-L-selectin cells rolling on PSGL-1 mutants under a constant shear stress of 1.25 dyn/cm 2 . The velocity of tracked cells was determined by measuring cell displacements within successive video frames (0.032 s) in the flow direction. Each increase in velocity is represented by a peak and each decrease by a valley (Fig. 7a). The replacement of tyrosine sulfate residues by Phe strongly increased the variations of cell-rolling velocity indicated by higher irregularity in "peaks" and "valleys", as illustrated in Fig. 7a. The observed instability of cell rolling was quantified by calculating the S.D. of the mean velocity of each tracked cell. The pooled data obtained from the whole cell population were used to determine the mean S.D. of rolling velocities on wild-type and each PSGL-1 mutant. More irregular rolling velocities were observed on PSGL-1 mutants devoid of tyrosine sulfate residues or expressing Tyr-48 as single tyrosine sulfate residue (mean Ϯ S.D. of 300.19-L-selectin cell-rolling velocities on PSGL-1 Y46F/Y48F/Y51F: 170 m/s versus 163 m/s on PSGL-1 Y46F/Y51F, n ϭ 5, p Ͻ 0.001) than on wild-type PSGL-1 (mean Ϯ S.D.: 60 m/s, n ϭ 5, p Ͻ 0.001, Fig.  7a). The stability of rolling velocity was less affected on mutant PSGL-1 expressing Tyr-46 (mean Ϯ S.D. on PSGL-1 Y48F/Y51F: 111 m/s, n ϭ 5) or Tyr-51(mean Ϯ S.D. on PSGL-1 Y46F/Y48F: 117 m/s, n ϭ 6) than Tyr-48 (Fig. 7a).
Lack of Inhibition of L-selectin-dependent Rolling on PSGL-1 by CSLEX-1 and HECA-452 mAbs.-The anti-CLA mAb HECA-452 was reported to inhibit by Ͼ90% L-selectin-dependent lymphocyte rolling on the human vascular endothelial cell line EA hy926 transfected with FucT-VII cDNA suggesting that CLA, a sLe x determinant, is a major determinant of endothelial L-selectin ligand(s) (55). In contrast to these observations, neutrophil rolling was not significantly reduced by  Table I and Fig. 9. Two potential hydrogen bonds were disclosed between Ser-47 of L-selectin and the sulfate group of Tyr-48, whereas P-selectin binding to PSGL-1sulfated Tyr-48 is supported (1) by 4 hydrogen bonds located between Ser-46, Ser-47, His-114, and the sulfate group of Tyr-48 and (2) by an additional hydrogen bond located between the peptidic backbones of Lys-112 and Tyr-48 ( Fig. 9 and Table  I). Importantly, the basic residue present at position 114 of P-selectin (His-114) is absent on L-selectin. DISCUSSION The determinants of PSGL-1 that mediate L-selectin and P-selectin binding include tyrosine sulfate residues and Oglycans attached to Thr-57 (24, 33, 34, 56 -62). The crystal structure analysis of P-selectin co-complexed with the N-terminal peptide of PSGL-1 showed that O-glycans terminated by sLe x /CLA as well as Tyr-48 and -51 play an essential role in supporting P-selectin binding. In addition, although this interaction had not been highlighted in the crystallographic study mentioned above, rolling adhesion assays and glycosulfopeptide binding studies have suggested a role for PSGL-1 Tyr-46 in adhesion to P-selectin (23,24,34,35). The results presented here show that P-selectin and L-selectin use similar mechanisms to bind to PSGL-1. However, the three tyrosine sulfate residues of PSGL-1 do not contribute in an equal fashion to Land P-selectin binding. Specifically, our results indicate that Tyr-48 is of key importance in supporting P-selectin-mediated rolling on PSGL-1 whereas it only plays a minor role in mediating L-selectin binding. In addition, the present study shows that: 1) sialylated and fucosylated core-2 O-glycans attached to Thr-57 are essential to allow optimal L-selectin binding to PSGL-1 and 2) to support and stabilize leukocyte rolling on CHO-PSGL-1 cells; 3) at least 2 or 3 N-terminal Tyr sulfate residues are required to optimally support leukocyte recruitment and rolling; and 4) Tyr-51 plays a predominant role in recruiting and stabilizing leukocyte rolling on PSGL-1.
We have defined here the minimal molecular requirements supporting L-selectin interactions with PSGL-1. L-selectin/ weakly bound to sLe x /CLA-expressing CHO cells in the absence of PSGL-1 (Fig. 2b, upper panel). However, this interaction was not sufficient to support neutrophil rolling (Fig. 3a). This observation is consistent with the notion that low affinity interactions between L-selectin and sLe x cannot efficiently support L-selectin-mediated rolling (13). Interestingly, a low number of neutrophils rolled on CHO-PSGL-1/FucT-VII cells even in the absence of C2GnT suggesting that the presentation of sLe x / CLA by core-2 O-linked glycans is not essential to support L-selectin-dependent rolling (Fig. 3a). The strong reduction in neutrophil rolling and L-selectin/ binding observed after the replacement of Thr-57 (but not of Thr-44) by Ala confirmed the essential role played by Thr-57 in presenting O-glycans involved in L-selectin binding (Fig. 4). Interestingly, leukocyte rolling was less affected by the replacement of the whole tyrosine sulfation consensus by Phe residues, suggesting a predominant role for core-2 O-glycans (Fig. 5, b and c). A critical role for sialic acid residues was indicated by the abrogation of neutrophil rolling on PSGL-1 after sialidase treatment of CHO cells. Similar observations were previously reported for P-selectin (22,34,37,63). Sialic acid residues presented by sLe x / CLA determinants most likely play a key role in this interaction.
C2GnT co-expression with FucT-VII and PSGL-1 strongly increased leukocyte recruitment and decreased rolling velocity (Fig. 3). Similar observations were made by others who studied PSGL-1 interactions with P-selectin in vitro using leukocyte isolated from C2GnT-deficient mice or in vivo in C2GnT-deficient mice (25,28). Higher rolling velocities on P-selectin were observed in the absence of C2GnT suggesting that core-2 Olinked glycans have a major role in regulating leukocyte-rolling The hydrogen bonding pattern was obtained using the HBPLUS program (69). The donor (D) and the acceptor (A) atoms are defined by the three letter code amino acid, residue numbers and atom types (in the X-ray structure (23), the Tyr-SO 3 -48 corresponds to Tys607 and Tyr-SO 3 -51 to Tys 610). D-A distance: distance between the donor and the acceptor atoms. DHA: angle centered on the hydrogen (H) and linking the donor and acceptor atoms.  velocity. The present study extends these observations to Lselectin. By stabilizing and reducing leukocyte-rolling velocity on L-selectin or P-selectin, core-2 O-glycans may improve the exposure of rolling cells to cytokines or chemokines at site of inflammation and promote leukocyte arrest and firm adhesion following integrin activation.
Whereas the results of this study confirm that L-selectin interaction with PSGL-1 is dependent on the expression of O-glycans attached to Thr-57 and of N-terminal tyrosine sulfate residues (Figs. 4 and 5) (35), they also revealed that PSGL-1 mutants exhibiting a single tyrosine sulfate residue were not as efficient as wild-type PSGL-1 in supporting leukocyte rolling. Moreover, these results showed that tyrosine sulfate residues did not contribute equally to L-selectin interactions with PSGL-1. A predominant role for Tyr-51 was indicated by: 1) the lower rolling velocities (Fig. 6), 2) the more efficient recruitment of leukocytes on PSGL-1 Y46F/Y48F expressing Tyr-51 as single tyrosine sulfated residue (Fig. 5), 3) the more efficient binding of L-selectin/ to PSGL-1 Y46F/Y48F (Fig. 5a), and 4) the increased stability of leukocyte-rolling velocity on this mutant PSGL-1 (Fig. 7).
Experiments performed with glycosulfopeptides previously indicated that Tyr-48 plays a major role in mediating the interactions of the N-terminal peptide of PSGL-1 with P-selectin (34). Adhesion studies performed with K-562-P-selectin cells demonstrated with whole cells that Tyr-48 plays a predominant role in supporting P-selectin-mediated rolling interactions with PSGL-1 whereas it had only a minor role in mediating L-selectin-dependent rolling (Fig. 6). These observations are in keeping with crystal structure analysis that revealed a major role for His-114 in mediating P-selectin binding to PSGL-1 Tyr-48 (23). The analysis of hydrogen-bonding pattern disclosed additional potential bonds, which may strengthen P-selectin interactions with Tyr-48 ( Fig. 9 and Table I). The absence of a basic amino acid residue at position 114 of L-selectin and the lower number on hydrogen bonds may explain why Tyr-48 of PSGL-1 plays only a minor role in supporting L-selectin binding to PSGL-1 ( Fig. 9 and Table I). Of note, molecular modeling analysis of L-selectin interactions with Tyr-48 is consistent with results of adhesion studies performed under flow at various shear stress and validate this model.
L-selectin and P-selectin are likely to bind to PSGL-1 in a similar fashion because of the conservation of residues within the sLe x binding site and the presence of a basic residue (Lys) at position 85 (23). Electrostatic interactions most likely play a major role in supporting the negatively charged Tyr-51 binding to L-selectin Lys-85 and to P-selectin Arg-85. An important role for Lys-85 in supporting L-selectin binding to PSGL-1 is suggested by the elevated partial charge of the ammonium group of Lys-85 (ϩ0.69) (64). In comparison, a lower charge is associated to the iminium group of Arg-85 (ϩ0.12) of P-selectin. The absence of a basic amino acid residue at position 114 of Lselectin and the presence of Lys-85 may explain why sulfated Tyr-51 plays a predominant role in supporting L-selectin interactions with PSGL-1 whereas sulfated Tyr-48 is less important. L-selectin/ binding studies and adhesion assays indicated that Tyr-46 plays an important role in supporting L-selectin interactions with PSGL-1. Leukocyte rolling was more stable and slower on PSGL-1 Y48F/Y51F than on PSGL-1 Y46F/Y51F (Figs. 6 and 7). Interestingly, the involvement of Tyr-46 in P-selectin binding was not revealed by crystal structure analysis (23) whereas binding studies of PSGL-1 glycosulfopeptides to P-selectin (34) and adhesion studies performed with K-562-P-selectin cells (Figs. 5e and 6d) indicate that Tyr-46 supports this reaction. The role of Tyr-46 in mediating L-selectin and P-selectin binding was recently further supported by the ability of the anti-PSGL-1 mAb PS-4, which reacts with Tyr-46 but not with Tyr-48 or -51, to inhibit L-selectin-dependent rolling on PSGL-1. 2 Rolling is an important step during which leukocytes are exposed to chemoattractants at sites of inflammation, a reaction that leads to integrin activation and leukocyte firm adhesion. Frame by frame analysis of cell displacements indicates that cell rolling occurs through a series of steps or jerks that appear to represent receptor-ligand dissociation events (65)(66)(67). Rolling through selectins is unaffected by alterations in selectin density and hydrodynamic forces acting on the cell (65). This surprising stability of rolling has been explained by the ability of leukocyte to reach a dynamic balance between formation and breakage of bonds between selectins and their ligands over a wide range of wall shear stress and ligand densities (66,67). The stereospecific interactions of tyrosine sulfate and O-glycans with P-selectin and L-selectin create a high affinity binding site (23,27,34), which contributes to rolling stabilization. The highly irregular cell rolling observed on PSGL-1 mutants devoid of tyrosine sulfate residues or expressing Tyr-48 as single tyrosine sulfate residue (Fig. 7a) indicates that Tyr-46 and -51 play a major role in the ability of PSGL-1 to stabilize L-selectin-mediated rolling. Similar observations were made for core-2 O-glycans attached to Thr-57 which present sLe x /CLA to L-selectin (Fig. 8a). Post-translational modifications of PSGL-1 may facilitate endothelium surveillance for signs of inflammation and thereby represent important additional levels of regulation of leukocyte traffic.