Evolutionary conservation of P-selectin glycoprotein ligand-1 primary structure and function

Background P-selectin glycoprotein ligand-1 (PSGL-1) plays a critical role in recruiting leukocytes in inflammatory lesions by mediating leukocyte rolling on selectins. Core-2 O-glycosylation of a N-terminal threonine and sulfation of at least one tyrosine residue of PSGL-1 are required for L- and P-selectin binding. Little information is available on the intra- and inter-species evolution of PSGL-1 primary structure. In addition, the evolutionary conservation of selectin binding site on PSGL-1 has not been previously examined in detail. Therefore, we performed multiple sequence alignment of PSGL-1 amino acid sequences of 14 mammals (human, chimpanzee, rhesus monkey, bovine, pig, rat, tree-shrew, bushbaby, mouse, bat, horse, cat, sheep and dog) and examined mammalian PSGL-1 interactions with human selectins. Results A signal peptide was predicted in each sequence and a propeptide cleavage site was found in 9/14 species. PSGL-1 N-terminus is poorly conserved. However, each species exhibits at least one tyrosine sulfation site and, except in horse and dog, a T [D/E]PP [D/E] motif associated to the core-2 O-glycosylation of a N-terminal threonine. A mucin-like domain of 250–280 amino acids long was disclosed in all studied species. It lies between the conserved N-terminal O-glycosylated threonine (Thr-57 in human) and the transmembrane domain, and contains a central region exhibiting a variable number of decameric repeats (DR). Interspecies and intraspecies polymorphisms were observed. Transmembrane and cytoplasmic domain sequences are well conserved. The moesin binding residues that serve as adaptor between PSGL-1 and Syk, and are involved in regulating PSGL-1-dependent rolling on P-selectin are perfectly conserved in all analyzed mammalian sequences. Despite a poor conservation of PSGL-1 N-terminal sequence, CHO cells co-expressing human glycosyltransferases and human, bovine, pig or rat PSGL-1 efficiently rolled on human L- or P-selectin. By contrast, pig or rat neutrophils were much less efficiently recruited than human or bovine neutrophils on human selectins. Horse PSGL-1, glycosylated by human or equine glycosyltransferases, did not interact with P-selectin. In all five species, tyrosine sulfation of PSGL-1 was required for selectin binding. Conclusion These observations show that PSGL-1 amino acid sequence of the transmembrane and cytoplasmic domains are well conserved and that, despite a poor conservation of PSGL-1 N-terminus, L- and P-selectin binding sites are evolutionary conserved. Functional assays reveal a critical role for post-translational modifications in regulating mammalian PSGL-1 interactions with selectins.

Comparison of these mammal sequences shows that the transmembrane and cytoplasmic domains are highly conserved [9,27,28]. Little information is however available on the intra-and inter-species evolution of decameric motives and on the conservation of PSGL-1 N-terminus. Multiple sequence alignment of a large number of mammalian PSGL-1 sequences is necessary to examine these points and define motives associated with the core-2 Oglycosylation of the N-terminal threonine, homologous to Thr-57 on human PSGL-1.

Conservation of PSGL-1 sequence
Multiple alignment of mammalian PSGL-1 sequences is presented in figure 1. A signal peptide (SP) cleavage site is predicted between residues 17 and 18 in most sequences. Equine PSGL-1 is an exception with a predicted cleavage site between residues 18 and 19. Nine sequences including human have a propeptide predicted to be cleaved by paired basic amino acid-converting enzymes (PACE/furin; [9,43] at residue 41 (38 for northern tree shrew). By contrast, the PACE consensus sequence, RX [R/K]R is not observed in bovine, sheep, cat, bat and equine PSGL-1 (Fig. 1).
A mucin-like domain is present in all studied species. It lies between the conserved N-terminal O-glycosylated threonine (Thr-57 in human) and the transmembrane domain, and contains a central region exhibiting decameric repeats (DR). This region was analyzed using the MEME program, whose parameters were applied to each sequence individually and/or simultaneously to all sequences. DR-containing central regions were aligned considering the intra-and inter-species evolution of decameric motives. The degree of inter-species conservation in the N-and C-terminal ends of the mucin-like domain (which sometimes contains traces of mutated decamers) is low. The mucin-like domain is composed of 247 to 322 residues and the number of DR varies from 7 in pig to 18 in chimpanzee and rhesus monkey ( Table 2). The number of DR varies in human from 14 to 16 repeats [47][48][49][50]. We also observed a polymorphism in rat. One of the three Amino acid sequences of mammalian PSGL-1 Figure 1 Amino acid sequences of mammalian PSGL-1. (A) Multiple alignment of 12 complete PSGL-1 amino acid sequences and (B) of N-terminal sequences from putative PSGL-1 proteins of cat and sheep, which were inferred respectively from partial genomic sequence and EST sequences identified through homology searches (EMBL/Genbank/DDBJ respective accession numbers: AANG01098304 and DY506895). The putative signal peptide (SP) and propeptide (Pro) cleavage sites are indicated by arrows. The consensus sequence for propeptide cleavage (RX [R/K]R), recognized by PACE, is boxed. Threonine homologous to human Thr-57 is indicated by a black arrowhead. Arbitrary gaps have been added in each sequence in order to isolate and align separately the mucin-like region containing the decameric repeats, which is surrounded by a frame. The transmembrane domain (TM) is marked by a bar. Asterisks indicate the amino acids involved in moesin binding.
available sequences contains 12 DR [51], whereas only 11 repeats have been observed in the sequences cloned by us ( Fig. 1) and others [25]. This polymorphism suggests a dynamic intraspecies evolution of this region.
The analysis of the sequences of PSGL-1 mucin-like regions showed that several constitutive repeats of 10 amino acids can be identified in the center of these regions, while both ends are made up with unconserved amino acids. The best permutation motif, which is the most consistent with the different sequences and which optimizes the number of repeated units per sequence, is AATEAQTTQP.
Interestingly, in canine PSGL-1, 3 DR strongly differ in their sequences from the others (Fig. 1A). These units are identical to each other and are located every 30 positions. Combining decamera to form repeats of 30 amino acids displays a greater consensus between repeats suggesting that duplication of 30 amino acid units (itself created by two duplications of 10 amino acid units followed by mutations in the third one) arose at least twice in the evolution of dog PSGL-1. The same kind of phenomenon is observed in bat, where the best repeated unit has a length of 15 amino acids. Similarly, equine repeated units exhibit a greater consensus when they are formed of 20 residues units instead of 10 [52].
A transmembrane domain of 23 residues is predicted in all sequences immediately after the conserved cysteine involved in PSGL-1 dimerization (Fig. 1A). A short extracellular juxta-membrane region is involved in binding versican G3 domain, whose interaction with PSGL-1 promotes leukocyte aggregation [53]. Interestingly, three positions in this region are perfectly conserved in all studied species (Asp-313, Val-317, Lys-318). The transmem-brane domain is followed by a cytoplasmic tail, which is made up of two highly conserved regions. Over the 31 first positions of the cytoplasmic domain, 20 are completely conserved and 5 contain conservative substitutions (Fig.  1A). Among the conserved positions, Ser-346, Arg-347, Lys-348 and Ser-358 (Fig. 1A) are involved in moesin binding to the cytoplasmic domain of human PSGL-1 [54]. In all sequences, the C-terminal region is ended by 11 almost perfectly conserved residues.
Human L-and P-selectin interact with human, rat, bovine, pig or equine CHO-PSGL-1 cells CHO cells co-expressing human FucT-VII and C2GnT-I and human, bovine, pig, rat or equine PSGL-1 were prepared. The five transfectants expressed similar levels of sLe x and CLA. PSGL-1 expression was detected using a mAb reacting with PSGL-1 C-terminal 6 × His tag (Invitrogen). The anti-human PSGL-1 mAbs PL1, KPL1 and PL2 [28,55] did not react with bovine, pig, rat or equine PSGL-1 (data not shown). Flow cytometric analysis of human Por L-selectin/μ binding to the various CHO-PSGL-1 transfectants showed that P-and L-selectin/μ bind similarly to human, bovine, pig, rat or equine PSGL-1 expressed by transfected CHO cells. As the reactivity of mouse PSGL-1 with human selectins was previously described [23], we did not repeat these analyses (Fig 2).
Human L-, P-and E-selectin bind heterogeneously to human, bovine, pig or rat neutrophils PSGL-1 expressed by CHO transfectants differ in their glycosylation pattern from mammalian neutrophil PSGL-1. In CHO transfectants, the various mammalian PSGL-1 are glycosylated by FucT-VII and C2GnT-I of human origin, while in mammalian neutrophils PSGL-1 is glycosylated by their own glycosyltransferases. As the glycosylation pattern may affect PSGL-1 interactions with L-or P-selectin, we examined the reactivity of human selectins with mammalian neutrophils (Fig. 3). L-and P-selectin/μ chimera strongly reacted with human and bovine PSGL-1, while a weaker reaction was observed with pig and rat. The L-and P-selectin carbohydrate ligands sLe x and CLA, identified by CSLEX-1 and HECA-452 mAbs respectively, were strongly expressed by human neutrophils and also, surprisingly, by equine neutrophils (mean fluorescence intensity ± SD: human: 74 ± 1, n = 2 and 79 ± 12, n = 2; equine: 173 ± 9, n = 2 and 108 ± 34, n = 2). By contrast, despite significant selectin binding, sLe x and CLA were undetectable on bovine, pig and rat neutrophils (not illustrated). As selectin binding is dependent on cell surface expression of fucosylated ligands, we examined FucT-VII mRNA expression by RT-PCR amplification of total RNA from bovine, pig, rat and equine neutrophils. FucT-VII mRNA transcripts were detected in all investigated species (data not shown). Thus, as previously established for mouse leukocytes [56], the lack of reactivity of mAbs CSLEX-1 and HECA-452 with most mammalian PSGL-1 is likely due to the strong specificity of these mAbs for human oligosaccharides. Moreover, the observation that mAbs CSLEX-1 and HECA-452 strongly react with equine neutrophils suggests that human and equine neutrophils exhibit common carbohydrate structures, which are not detectable in mouse, rat, pig or bovine.

CHO cells expressing mammalian PSGL-1 efficiently roll on human L-or P-selectin
The role of PSGL-1 in regulating CHO-PSGL-1 cell rolling on human L-or P-selectin was assessed under hydrodynamic flow conditions. Human PSGL-1 expressing cells were less recruited on human P-selectin than CHO cells expressing bovine PSGL-1. Moreover, on human L-selectin, cell recruitment of CHO cells expressing human PSGL-1 was less efficient than that of cells expressing bovine, pig or rat PSGL-1 (Fig. 4A). Surprisingly, CHO cells expressing equine PSGL-1 did not roll on P-selectin and were weakly recruited on L-selectin.
Previous studies showed that N-terminal tyrosine sulfate residues are involved in supporting human PSGL-1dependent rolling on L-and P-selectin [16,20,45]. Human, bovine, rat and pig PSGL-1 exhibit two or three potential N-terminal tyrosine sulfation sites, whereas equine PSGL-1 contains only one single site (Fig. 1A). The contribution of PSGL-1 sulfation to cell rolling was assessed by comparing recruitment of CHO cells expressing control or desulfated human, bovine, rat, pig and equine PSGL-1 on L-or P-selectin (Fig. 4B). Inhibition of PSGL-1 sulfation strongly reduced L-and P-selectindependent rolling. The recruitment of CHO cells expressing human PSGL-1, on P-selectin, was inhibited by 88 ± 5%, whereas the recruitment of cells expressing bovine, rat and porcine PSGL-1 was almost abrogated (Fig. 4B). Rolling inhibition induced by desulfation was also seen on L-selectin (although to a lesser degree than on P-selectin). Thus, as previously described for human PSGL-1, sulfation of bovine, pig, rat or equine PSGL-1 N-terminal tyrosine residues is required to support PSGL-1-dependent rolling on L-or P-selectin.
Interestingly, multiple sequence alignment of mammalian L-or P-selectin shows partial or complete conservation of amino acid residues that regulate human selectin binding to PSGL-1 tyrosine sulfate residues [16,18]. Ser-47, Lys-112 and His-114 on human P-selectin bind to human PSGL-1 Tyr-48, while human L-selectin Lys-85 and P-selectin Arg-85 interact with Tyr-51 ( Fig. 4C) [16,18]. In mammalian P-selectins, Ser-47 is conserved, except for bat and rhesus monkey, and Lys-112 and His-114 is either conserved or replaced by arginine, which may interact with sulfated Tyr-48. Except for pig and Human L-and P-selectin/μ chimera cross-react with CHO cells expressing mammalian homologues of PSGL-1 horse, Arg-85, which binds to human PSGL-1 Tyr-51, is conserved or replaced by lysine (Fig. 4C). L-selectin Ser-47, which binds to human PSGL-1 Tyr-48, is conserved or replaced by a threonine, except for mouse, tree shrew and cat, while L-selectin Lys-85, which interacts with human PSGL-1 Tyr-51, is perfectly conserved. Results of Fig. 4B and alignment of Fig. 4C suggest that, like in human PSGL-1, tyrosine sulfation of mammalian homologues is critical for L-and P-selectin interactions. Sulfation of a unique tyrosine sulfate residue was sufficient to support equine PSGL-1-dependent rolling on human L-selectin (Fig. 4B). However, the recruitment of CHO cells expressing equine PSGL-1 on L-selectin was much less efficient than that of all other CHO cell transfectants (Fig. 4A).

Mammalian neutrophil recruitment on human L-or Pselectin is heterogeneous
The impact of PSGL-1 glycosylation by mammalian FucT-VII and C2GnT-I on PSGL-1-dependent rolling on human L-or P-selectin was assessed under various shear stresses (0.5 to 2.0 dynes/cm 2 ; Fig. 5). The recruitment of bovine, porcine, rat and equine neutrophils on human L-or Pselectin strongly differed from that of the corresponding CHO-PSGL-1 cells (Kruskal-Wallis test, P < 0.0001; Fig.  4A). At 1.5 and 2.0 dynes/cm 2 , bovine and human neutrophils rolled similarly on P-selectin. However, at lower shear stresses, bovine neutrophils were significantly less recruited than human neutrophils. Human and bovine neutrophil recruitment on L-selectin was similar at 1.0 and 2.0 dynes/cm 2 (Fig. 5). Above 0.5 dynes/cm 2 , porcine, rat and equine neutrophils were less recruited on L-or Pselectin than human or bovine neutrophils (Fig. 5). At 1.5 dynes/cm 2 , recruitment of porcine, rat and equine neutrophils was respectively 4-, 290-and 3-fold lower on Lselectin and 53-and 36-fold lower on P-selectin than that of human neutrophils. As observed with CHO cells expressing equine PSGL-1, equine neutrophils did not roll on P-selectin. At all shear stresses, rat neutrophils poorly rolled on human L-or P-selectin. These observations are in agreement with results of human selectin chimera binding to neutrophils (Fig. 3); both assays showed that bovine neutrophil PSGL-1 strongly interacts with human L-or P-selectin whereas interactions are weaker between human selectins and porcine neutrophil PSGL-1 and almost absent with rat and equine PSGL-1 (Fig. 5). These results are in contrast with those obtained with CHO cells expressing pig or rat PSGL-1, which are much more efficiently recruited on human selectins (Fig. 4A). Interspecies differences in PSGL-1 core-2 O-glycosylation may explain these observations.
0.001). Compared to CHO cells expressing human PSGL-1, increased velocities of CHO cells expressing bovine PSGL-1 on P-selectin may have resulted in increased cell recruitment on human selectins (Fig. 4).
The stability of rolling velocities was assessed by measuring CHO-PSGL-1 cell and neutrophil displacements on human L-selectin within successive video frames (0.1 ms). Peaks represent increases in velocity and valleys decreases (Fig. 7). The stability of CHO-PSGL-1 cell rolling velocities on human L-selectin was heterogeneous among the studied species. Although CHO cells expressing human and bovine PSGL-1 had similar mrv, rolling velocities of CHO cells expressing bovine PSGL-1 (mean SD ± SD: 21 ± 3 μm/s) were less stable than those of cells expressing human PSGL-1 (11 ± 2 μm/s). CHO cells expressing rat PSGL-1 were the least stable (48 ± 9 μm/s, Fig. 7A). The stability of neutrophil rolling velocities was also highly heterogeneous among the studied species (Fig. 7B). Human and equine neutrophils exhibited the most stable rolling velocities (mean SD ± SD: 33 ± 4 μm/s vs. 32 ± 6 μm/s, n = 10, NS, Fig. 7B), whereas rat neutrophils were the least stable (67 ± 40 μm/s, n = 4). Interestingly, pig neutrophils exhibited periods of very slow rolling (mean velocity < 10 μm/s) alternating with sudden acceleration, rapidly followed by deceleration (Fig. 7B). Bovine and equine neutrophils had similar behaviors. Despite the presence of oligosaccharides recognized by HECA-452 and CSLEX-1 mAbs on both CHO cells and neutrophils expressing equine PSGL-1, transfected CHO cells rolled significantly faster and less stably than equine neutrophils, suggesting that other structures regulate equine neutrophil rolling.

Discussion
Selectins and PSGL-1 play a critical role in regulating leukocyte migration in inflammatory lesions [4]. Whether human selectins can bind to mammalian PSGL-1 had not been previously studied. As PSGL-1 is an attractive target for anti-inflammatory therapy [29][30][31][32][33]35,[37][38][39], the identification of conserved PSGL-1 functional regions may be helpful to design selectin inhibitors mimicking PSGL-1. We therefore analyzed PSGL-1 amino acid sequences of several mammals (5 previously described sequences; [9,[24][25][26] as well as 9 new sequences described herein, 7 complete and 2 N-terminal sequences; Fig. 1A-B) identified by us or others [25,26]. Multiple sequence alignments (Fig. 1) show that conservation of sequence is not homogeneous along the protein, and that the primary sequence of the site of interaction of L-or P-selectin [16,18] is not perfectly conserved. All sequences contain a threonine homologous to the core- dicting sulfation; [57] its length in the mature protein, as well as the number (1 to 3) and positions of the potentially sulfated tyrosine residues are variable. Desulfation and sulfation inhibition studies suggest that tyrosine sulfation plays a key role in supporting mammalian PSGL-1 interaction with human L-and P-selectin (Fig. 4B). Data presented here indicate that L-and P-selectin binding sites on PSGL-1 are evolutionary conserved and emphasize the role of threonine-linked core-2 O-glycans and tyrosine sulfate residues in supporting mammalian PSGL-1 interactions with human selectins. PACE cleaves PSGL-1 propeptide on human neutrophils. In contrast, the propeptide remains on CHO-PSGL-1 cells, which do not express the PACE protease. The lack of PSGL-1 cleavage by PACE in bovine and equine neutrophils did not prevent PSGL-1 interactions with selectins. The importance of propeptide cleavage is unclear: whether it may facilitate tyrosine sulfation or Nterminal O-glycosylation remains to be determined [21].
The T [D/E]PP [D/E] sequence, which is associated on human and mouse PSGL-1 with threonine O-glycosylation [9,18,58], is observed in most mammals except dog and horse, in which it is respectively replaced by TDAPE and TDLLK sequences. Despite these changes, equine neutrophils rolled on human L-selectin (Fig. 5). By contrast, neither equine neutrophils nor CHO cells expressing equine PSGL-1 significantly interacted with human Pselectin (Fig. 4

, 5). This suggests that the T [D/E]PP [D/E] motif may be important for mammalian leukocyte rolling on human P-selectin.
We observed that the sequence AATEAQTTQP is the best permutation motif to optimize the number of decameric units per sequence and that the most similar units lie at the center of the mucin-like region, while unconserved amino acids are more frequently present at both ends. This suggests that decamera located at the center of the mucin-like domain might be the most recent and that the evolution of this region might have proceeded by duplications of decameric units, followed by mutations and deletions. This process allowed the conservation of the length of the mucin-like domain with a 250-280 amino acid length (except in bushbaby), despite a variable number of repeated units among species (from 7 DR in pig to 18 DR in monkeys, Table 2). The preservation of PSGL-1 length may play a role in supporting the rolling on human selectins of leukocytes or CHO cells expressing human, bovine, pig or rat PSGL-1 (Fig. 4).
Transmembrane and cytoplasmic domain sequences are well conserved (Fig. 1). The juxta-membranous cysteine residue, involved in human PSGL-1 dimerization and in stabilizing leukocyte rolling on P-selectin [27,28,59] is perfectly conserved. A role for PSGL-1 as signaling molecule was indicated by its involvement in activating GTPase Ras and mitogen-activated protein kinases, as well as in inducing the secretion of inflammatory molecules [60][61][62] or in activating αMβ 2 or αLβ 2 integrins [63][64][65]. The high degree of conservation of the cytoplasmic domain suggests that PSGL-1-mediated intracellular signaling is evolutionary conserved. Human PSGL-1 engagement induces Syk phosphorylation and recruitment in lipid rafts as well as the expression of the early-immediate gene c-fos [54,66]. Syk activity, which is critically involved in regulating PSGL-1-dependent rolling on P-selectin [66], is dependent on the binding of PSGL-1 cytoplasmic domain to moesin, which serves as adaptor between PSGL-1 and Syk [54]. Importantly, the moesin binding residues, corresponding to Ser-346, Arg-347, Lys-348, and Ser-358 in human PSGL-1 [54] are perfectly conserved in all analyzed mammalian sequences. Of note, these amino acids are located within a group of 31 amino acids, among which 20 are identical and 5 similar.
On L-selectin, rolling velocities of CHO cells expressing human, bovine, and pig PSGL-1 were similar, whereas the median rolling velocity of CHO cells expressing rat or equine PSGL-1 was 4-and 5-fold higher respectively than that of CHO cells expressing human PSGL-1 (Fig. 6). The increased rolling velocities of CHO cells expressing bovine, pig or rat PSGL-1 on P-selectin may partially explain the preserved cell recruitment on P-selectin (Fig.  4A). As all CHO-PSGL-1 transfectants are glycosylated by human C2GnT-I and FucT-VII, differences in CHO-PSGL-1 cell recruitment and rolling velocities may mainly result from differences in N-terminal amino-acid residues interacting with the lectin domain of human L-or P-selectin. Among these residues, tyrosine sulfate residues may critically regulate PSGL-1 interactions with L-or P-selectin, like in human PSGL-1 [16,18,22,45]. The strong inhibition of CHO-PSGL-1 cell interactions with P-or L-selectin after desulfation and inhibition of sulfation supports this possibility (Fig. 4B). In addition, in most studied mammals, the amino acids regulating selectin interactions with potentially sulfated tyrosine residues are conserved (Fig.  4C). In mouse, Tyr-54 and Thr-58 regulate PSGL-1 interactions with P-selectin. [23] Because only one tyrosine is used, it was suggested that mouse PSGL-1 binding may rely more on O-glycans attached to Thr-58 than does human PSGL-1. [23] This may also occur in other mammals, which exhibit a single tyrosine residue (tree shrew, bat and horse, Fig. 1).
Differences in tyrosine sulfation and O-glycosylation may affect the stability of rolling velocities on L-selectin. Thus, the patterns of bovine, pig and equine neutrophil displacements differed from those of CHO cells expressing mammalian PSGL-1. In particular, pig neutrophils, and also bovine and equine neutrophils, exhibited periods of very slow rolling velocity, alternating with rapid accelerations and decelerations (Fig. 7). These observations emphasize the role of post-translational modifications in regulating PSGL-1 binding to human selectins.

Conclusion
Data presented here indicate that mammalian PSGL-1 share a common primary structure and has evolutionary conserved interactions with L-and P-selectin. As in human, PSGL-1-dependent rolling is regulated by core-2 O-glycosylation of a conserved threonine residue and by tyrosine sulfation. The high degree of conservation of PSGL-1 cytoplasmic domain suggests, as for human PSGL-1, a potential involvement in signal transduction and in regulating cell rolling. These results provide additional insights into the structure and function of PSGL-1 and may be helpful to design PSGL-1 peptidomimetics.

Bovine, porcine, murine and equine PSGL-1 and FucT-VII cDNAs
RNA was extracted from mammalian lymphocytes using TRIzol ® (Invitrogen, Basel, Switzerland). Bovine, pig and rat homologues of human PSGL-1 cDNAs were generated from lymphocyte total RNA using GeneRacer™ Kit (Invitrogen), according to the manufacturer protocols. Primer design was based on sequence homologies between human and mouse PSGL-1 [9,24]. Primers are listed in Table 1.

Immunophenotypic analysis
Cell staining with mAbs or L-, P-, or E-selectin/IgM heavy chain (μ) chimera was performed and analyzed with a Cytomics™ FC 500 cytofluorimeter (Beckman Instruments, Nyon, Switzerland), as described [19].  [16,19,69]. CHO-PSGL-1 cell and neutrophil interactions were recorded for 5 min by videomicroscopy [16,19,69]. Rolling velocities illustrated in Fig. 6 were measured by tracking individual cells every 0.1 s, for 1-20 s, using a digital image analysis system (Mikado software, GPIL SA, Martigny, Switzerland; [16,19,69]. 183-755 independent determinations of cell rolling velocities were measured to analyze the velocities of transfectants and 25-381 determinations for the analysis of neutrophil velocities. Frameby-frame velocities (Fig. 7) were measured by tracking cells every 0.1 s for 6 s, within 0.28 mm 2 microscopic fields. The mean velocity of frame-by-frame tracked cells was included between percentiles 40-60 of the velocity of each cell population illustrated in Fig. 6. L-and P-selectindependent rolling was inhibited (>95%) by 10 mM EDTA or LAM1-3 or WAPS12.2 mAbs (data not shown). Mocktransfected CHO cells did not roll on L-or P-selectin. CHO transfectants used in adhesion assays expressed similar levels of cell surface PSGL-1 and sLe x [19].

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 < 0.05 were considered as significant.

Authors' contributions
OS, BB and FG designed research, analyzed data and wrote the paper. BB and SG performed experiments. FG performed multiple alignment. MS critically reviewed the manuscript and contributed to writing. All authors read and approved the final manuscript