Structurally Distinct Requirements for Binding of P-selectin Glycoprotein Ligand-1 and Sialyl Lewis x to Anaplasma phagocytophilum and P-selectin*

Colonization of neutrophils by the bacterium Anaplasma phagocytophilum causes the disease human granulocytic ehrlichiosis. The pathogen also infects mice, its natural host. Like binding of P-selectin, binding of A. phagocytophilum to human neutrophils requires expression of P-selectin glycoprotein ligand-1 (PSGL-1) and α1–3-fucosyltransferases that construct the glycan determinant sialyl Lewis x (sLex). Binding of A. phagocytophilum to murine neutrophils, however, requires expression of α1–3-fucosyltransferases but not PSGL-1. To further characterize the molecular features that A. phagocytophilum recognizes, we measured bacterial binding to microspheres bearing specific glycoconjugates or to cells expressing human PSGL-1 and particular glycosyltransferases. Like P-selectin, A. phagocytophilum bound to purified human PSGL-1 and to glycopeptides modeled after the N terminus of human PSGL-1 that presented sLex on an O-glycan. Unlike P-selectin, A. phagocytophilum bound to glycopeptides that contained sLex but lacked tyrosine sulfation or a specific core-2 orientation of sLex on the O-glycan. A. phagocytophilum bound only to glycopeptides that contained a short amino acid sequence found in the N-terminal region of human but not murine PSGL-1. Unlike P-selectin, A. phagocytophilum bound to cells expressing PSGL-1 in cooperation with sLex on both N-and O-glycans. Moreover, bacteria bound to microspheres coupled independently with glycopeptide lacking sLex and with sLex lacking peptide. These results demonstrate that, unlike P-selectin, A. phagocytophilum binds cooperatively to a nonsulfated N-terminal peptide in human PSGL-1 and to sLex expressed on PSGL-1 or other glycoproteins. Distinct bacterial adhesins may mediate these cooperative interactions.

Human granulocytic ehrlichiosis is a tick-transmitted disease caused by a bacterium recently named Anaplasma phagocytophi-lum (1)(2)(3). Clinical manifestations include fever, headache, myalgia, leukopenia, thrombocytopenia, and impaired host defenses that occasionally lead to fatal complications (4 -7). A. phagocytophilum is maintained in a zoonotic cycle between the arthropod vector, Ixodes scapularis, and mice that serve as reservoir hosts (8 -10). Human infection is inadvertent and is not an obligate component of the life cycle of the bacterium.
A hallmark of infection in both mice and humans is the colonization of neutrophils (1,2). This tropism for a particular cell type suggests that the bacteria recognize specific molecular determinants on the neutrophil surface. Current information suggests that these determinants are related to the cell-surface ligands for selectins, a family of cell adhesion molecules that mediate tethering and rolling of leukocytes on vascular surfaces during the earliest steps of inflammation (11,12). P-, E-, and L-selectin bind to ␣2-3-sialylated and ␣1-3-fucosylated glycans such as sialyl Lewis x (sLe x ), 1 which are expressed on most leukocytes. Targeted disruption of the genes encoding FTVII and FTIV, the ␣1-3-fucosyltransferases expressed in murine leukocytes, eliminates selectin-dependent interactions in vivo (13,14). The major ligand for P-and L-selectin on leukocytes is P-selectin glycoprotein ligand-1 (PSGL-1), a homodimeric mucin (12,15,16). Monoclonal antibodies to Nterminal epitopes on human and murine PSGL-1 inhibit leukocyte interactions with P-and L-selectin (17)(18)(19). P-selectin binds in a stereospecific manner to the N terminus of human PSGL-1 through recognition of tyrosine sulfate residues, adjacent peptide determinants, and fucose and sialic acid residues on a properly positioned core-2 O-glycan (20 -23). Murine PSGL-1 appears to use a related group of modifications to bind to P-selectin (24).
Binding of A. phagocytophilum to human myeloid HL-60 cells requires cell surface sialylation and correlates with expression of FTVII (25). Moreover, mAbs to N-terminal epitopes on human PSGL-1 inhibit binding of A. phagocytophilum to human neutrophils or HL-60 cells (26). Expression of both PSGL-1 and FTVII in heterologous cells renders them susceptible to binding and infection by A. phagocytophilum. These data were interpreted to indicate that A. phagocytophilum specifically recognizes fucosylated PSGL-1 (26). However, measurements of bacterial interactions with intact leukocytes did not demonstrate direct binding of A. phagocytophilum to PSGL-1. Studies in mice suggest that the molecular determinants required for A. phagocytophilum binding might be more complex than those originally proposed (27). Bacterial binding to and infection of leukocytes from mice lacking both FTVII and FTIV is markedly diminished in vitro and in vivo, consistent with the observed requirement for ␣1-3-fucosylation of bacterial ligands on human leukocytes. However, A. phagocytophilum binds to and infects leukocytes from PSGL-1-deficient mice much like leukocytes from wild-type mice (27). Thus, A. phagocytophilum infection in both humans and mice requires sialylation and ␣1-3-fucosylation of neutrophils, but PSGL-1 is only required in human neutrophils. These disparities raise the possibility that A. phagocytophilum expresses distinct adhesins, which bind cooperatively to distinct ligands on leukocytes. Some of these ligands might be specific for humans or mice.
In this study, we used purified molecularly defined glycoconjugates coupled to microspheres, in conjunction with experiments with transfected cells, to more precisely determine the structural features required for binding to A. phagocytophilum. Our data reveal clear differences in how P-selectin and A. phagocytophilum bind to PSGL-1. They further suggest that A. phagocytophilum binds cooperatively to distinct ligands on leukocyte surfaces.

MATERIALS AND METHODS
Preparation of Labeled, Fixed A. phagocytophilum-Cell-free A. phagocytophilum were prepared as described (27,28). Bacteria were washed once with Iscove's modified Dulbecco's modified Eagle's media (Invitrogen) and incubated with 10 M CellTracker Green (Molecular Probes) for 45 min at 37°C. Labeled bacteria were centrifuged twice to eliminate unbound probe before fixing with 1% paraformaldehyde (27). In some experiments, bacteria were fixed under identical conditions with Saccomanno fluid (Stablab Medical Products), which does not cross-link proteins. Unless noted otherwise, data from paraformaldehyde-fixed bacteria are presented in this paper.
Binding of A. phagocytophilum to microspheres or cells was meas-FIG. 1. Human sPSGL-1, PSGL-1-derived glyco(sulfo)peptides, or sLe x coupled to microspheres. Biotin was covalently attached to the C-terminal cysteine of sPSGL-1 or each glycopeptide or to a spacer group on sLe x . The biotinylated ligands were bound to streptavidincoated microspheres.
ured as described (27 2. A. phagocytophilum binds to a glycopeptide modeled after the Nterminal region of human PSGL-1 independently of tyrosine sulfation. A, microspheres bearing sPSGL-1 or the indicated glycopeptide were incubated with anti-PSGL-1 mAb or with an isotypematched control mAb. Bound antibody was detected with FITC-conjugated goat anti-mouse IgG. B, microspheres bearing sPSGL-1 or the indicated glycopeptide were incubated with the anti-sLe x mAb HECA-452 or with an isotype-matched control mAb. Bound mAb was detected with FITC-conjugated goat anti-rat IgM. C-F, control microspheres or microspheres bearing sPSGL-1 or the indicated glycopeptide were incubated with fixed, CellTracker Green-labeled A. phagocytophilum in the presence or absence of EDTA or the anti-PSGL-1 mAbs PL1 or PL2. Bound bacteria were measured by flow cytometry. These data are representative of three experiments.

FIG. 3.
A. phagocytophilum binding to a PSGL-1-derived glycopeptide requires ␣1-3-linked fucose on the Oglycan. A, structures of 2-GP-6, which has an ␣1-3-linked fucose, and of 4-GP-5, which lacks an ␣1-3-linked fucose. Each peptide was covalently coupled through its free N-terminal amino group to carboxylated microspheres. B, binding of mAb PL1 to microspheres bearing the indicated glycopeptide was measured as in Fig. 2. C, binding of labeled A. phagocytophilum to control microspheres or to microspheres bearing the indicated glycopeptide was measured as in Fig. 2. These data are representative of three experiments.

A. phagocytophilum Binds to a Glycopeptide Modeled after the N-terminal Region of Human PSGL-1 Independently of
Tyrosine Sulfation-For A. phagocytophilum binding studies, we initially used a group of well defined ligands with varying affinities for P-selectin ( Fig. 1). Recombinant sPSGL-1 was coexpressed in CHO cells with FTVII and core2GlcNAcT-I to confer the glycosylation required for optimal binding to selectins (34). P-selectin binds to a small N-terminal region that includes three sulfated tyrosines and a core-2 O-glycan capped with sLe x at Thr-16. This region contains the epitope for PL1, a mAb that blocks binding of PSGL-1 to P-selectin (37). 2-GSP-6, a glycosulfopeptide modeled after this region, binds with equivalent affinity to P-selectin (20). 2-GP-6, which lacks sulfate on the tyrosines, binds with much lower affinity to P-selectin. 2-GP-1 and 2-GSP-1, which have only a single Nacetylgalactosamine residue attached to Thr-16, do not detectably bind to P-selectin. sLe x , a simple tetrasaccharide that lacks all peptide determinants and the core-2 O-glycan presentation, binds with very low affinity to P-selectin (21).
sPSGL-1 and each glyco(sulfo)peptide contained a C-terminal cysteine, which allowed specific attachment of biotin to the sulfhydryl group. sLe x was attached to biotin through a linker. Each biotinylated ligand was coupled to streptavidin coated on 6-m microspheres. Microspheres were prepared with equivalent densities of sPSGL-1, 2-GSP-6, 2-GSP-1, and 2-GP-6, as measured by flow cytometry with PL1 ( Fig. 2A). The anti-sLe x mAb HECA-452 bound equivalently to microspheres coated with 2-GSP-6 or GP-6 as expected, because each peptide contains a single core-2 O-glycan that is capped with sLe x (Fig.  2B). HECA-452 bound to ϳ5-fold more sites on sPSGL-1 microspheres, presumably because each sPSGL-1 molecule has several fucosylated glycans.
Fluorescence-labeled, paraformaldehyde-fixed A. phagocytophilum were incubated with ligand-coupled microspheres, and binding was measured by flow cytometry. The labeled bacteria bound to sPSGL-1-coupled microspheres but not to control mi- FIG. 4. A. phagocytophilum binding to a PSGL-1-derived glycopeptide does not require a specific orientation of sLe x on the O-glycan. A, structures of 1-GP-6, which presents sLe x on a core-2 O-glycan, and of 1-GP-6Ј, which presents sLe x on an extended core-1 Oglycan. B, binding of mAb PL1 to microspheres bearing the indicated glycopeptide was measured as in Fig. 2. C, control microspheres or microspheres bearing the indicated glycopeptide were pretreated in the presence or absence of sialidase. Binding of labeled A. phagocytophilum was measured as in Fig. 2. These data are representative of three experiments.
FIG. 5. A. phagocytophilum binds weakly to sLe x lacking peptide. A, biotinylated sLe x was bound to streptavidin-coated microspheres at low, medium, and high densities as measured by binding of mAb HECA-452. B, control microspheres or microspheres bearing sLe x at the indicated density were pretreated in the presence or absence of sialidase. Binding of labeled A. phagocytophilum was measured as in Fig. 2. These data are representative of three experiments. crospheres (Fig. 2C). The anti-PSGL-1 mAb PL1 blocked binding, whereas PL2, an mAb that recognizes a more membraneproximal epitope on PSGL-1, had no effect on binding. This result demonstrates that A. phagocytophilum adheres to a PSGL-1-coated surface lacking other glycoproteins. Furthermore, the bacteria specifically bound to microspheres coated with 2-GSP-6, demonstrating that the P-selectin binding region of PSGL-1 was sufficient for productive binding (Fig. 2D). Chelation of divalent cations with EDTA had no effect on binding, as observed previously (25) for bacterial binding to HL-60 cells. This contrasts with the Ca 2ϩ -dependent binding of selectins to PSGL-1 and other sLe x -capped glycoconjugates (11,12). Labeled A. phagocytophilum also bound specifically to 2-GP-6coupled microspheres (Fig. 2E), demonstrating that binding does not require tyrosine sulfation of PSGL-1. This contrasts sharply with the importance of tyrosine sulfation of PSGL-1 for optimal binding to P-selectin (38 -40). The bacteria did not bind to microspheres coated with 2-GSP-1 (Fig. 2F). Thus, the O-glycan component of the PSGL-1-derived glycopeptide is essential for binding. Bacteria fixed with Saccomanno, which does not cross-link proteins, also bound specifically to microspheres coated with sPSGL-1, 2-GSP-6, or 2-GP-6 (data not shown). This indicates that binding was not an artifact because of a specific method of fixation.
A. phagocytophilum Binding to a PSGL-1-derived Glycopeptide Requires ␣1-3-linked Fucose on the O-glycan-Binding of A. phagocytophilum to both human and murine leukocytes correlates with expression of ␣1-3-fucosyltransferases, suggesting that binding requires ␣1-3-fucosylation of cell-surface glycans (25)(26)(27). To directly address the contribution of ␣1-3linked fucose, we compared bacterial binding to 2-GP-6 with that to 4-GP-5, which lacks an ␣1-3-linked fucose on the core-2 O-glycan (Fig. 3A). Because 4-GP-5 has a C-terminal methionine rather than a cysteine, we coupled 4-GP-5 and 2-GP-6 through their N-terminal free amino groups to carboxylated microspheres at matched densities (Fig. 3B). Labeled A. phagocytophilum bound specifically to N-terminal-coupled 2-GP-6 ( Fig. 3C). This closely resembled binding to C-terminal-coupled 2-GP-6 (see Fig. 2E), indicating that the orientation of the 2-GP-6 on the microspheres did not affect bacterial recognition. In contrast, A. phagocytophilum did not bind to 4-GP-5-coupled microspheres (Fig. 3C). These data demonstrate that an ␣1-3linked fucose residue is an essential recognition determinant for A. phagocytophilum. Murine-human GP-6 replaced five unique amino acids in human 2-GP-6 with the corresponding single asparagine found in murine GP-6. Human-murine GP-6 replaced the single asparagine in murine GP-6 with the corresponding five amino acids found in human 2-GP-6. Each glycopeptide had the same core-2 O-glycan capped with sLe x (C2-O-sLe x ) attached to the threonine. Binding of anti-human PSGL-1 PL1 was measured as in Fig. 2. Binding of anti-murine PSGL-1 mAb 4RA10 was measured identically, except that binding was detected with FITCconjugated goat anti-rat IgG. E and F, binding of labeled A. phagocytophilum to control microspheres or to microspheres bearing the indicated glycopeptide was measured as in Fig. 2. These data are representative of three experiments.
A. phagocytophilum Binding to a PSGL-1-derived Glycopeptide Does Not Require a Specific Orientation of sLe x on the O-glycan-P-selectin binds in a stereospecific manner to the N-terminal region of human PSGL-1, as dramatically demonstrated by its inability to bind to an isomeric glycosulfopeptide that presents sLe x on an extended core-1 O-glycan rather than on a branched core-2 O-glycan (21). We measured A. phagocytophilum binding to two isomeric PSGL-1-derived glycopeptides, 1-GP-6, bearing a core-2 O-glycan, and 1-GP-6Ј, bearing an extended core-1 O-glycan (Fig. 4A). The glycopeptides were N-terminally coupled to microspheres at matched densities (Fig. 4B). Labeled A. phagocytophilum bound equivalently to both microspheres (Fig. 4C). The anti-PSGL-1 mAb PL1 blocked binding to both microspheres, confirming the specificity of binding (data not shown). Furthermore, sialidase treatment of microspheres eliminated A. phagocytophilum binding (Fig. 4C). These data confirm that binding requires sialylation, as well as fucosylation of the glycan. They further demonstrate that bacterial binding, unlike P-selectin binding, does not require a specific orientation of the glycan on the PSGL-1-derived peptide.
A. phagocytophilum Binds Weakly to sLe x Lacking Peptide-The above results demonstrate that A. phagocytophilum binds readily to sLe x presented on a PSGL-1-derived peptide independently of tyrosine sulfation or a specific alignment of the O-glycan. To determine whether the peptide itself contributes to bacterial recognition, we measured A. phagocytophilum binding to sLe x lacking peptide, which was coupled to microspheres at three different densities as measured by mAb HECA-452 (Fig. 5A). Sialidase treatment eliminated HECA-452 binding to all three groups of microspheres (data not shown). Labeled A. phagocytophilum bound weakly to sLe x in a sialidase-sensitive manner (Fig. 5B) but only when sLe x was coupled to microspheres at an ϳ100-fold greater density than the densities of glycopeptides (compare Fig. 5A with Fig. 2B). These data demonstrate that the peptide component of human PSGL-1 makes a major contribution to bacterial binding.

A. phagocytophilum Binding to a PSGL-1-derived Glycopeptide Requires an Amino Acid Sequence Present in the N Terminus of Human but Not Murine PSGL-1-A. phagocytophilum
binds to both wild-type and PSGL-1-deficient murine neutrophils (27). The pathogen might have evolved a PSGL-1-independent mechanism to bind murine neutrophils, because it could not bind to murine PSGL-1. Consistent with this hypothesis, A. phagocytophilum bound only weakly to transfected CHO cells expressing murine PSGL-1 and FTVII (27). The N-terminal sequence of murine PSGL-1 sequence differs significantly from that of human PSGL-1 (41). Mutational analysis of the N-terminal region suggests that murine PSGL-1 requires sulfation of one of two tyrosines and O-glycosylation of a closely situated threonine to bind P-selectin (24). We compared A. phagocytophilum binding to biotinylated human and murine PSGL-1-derived glycopeptides that were coupled to streptavidin-coated microspheres. Each glycopeptide contained a core-2 O-glycan capped with sLe x at the respective threonine thought to be important for P-selectin binding. Anti-human PSGL-1 mAb PL1 bound to microspheres coupled with human 2-GP-6 but not with murine GP-6, whereas anti-murine PSGL-1 mAb 4RA10 bound to microspheres coupled with murine GP-6 but not with human 2-GP-6 ( Fig. 6, A and B). We also coupled two chimeric glycopeptides through their C-terminal cysteines to microspheres. Murine-human GP-6 substituted a five-residue peptide segment that separates the tyrosine closest to the O-linked threonine in the human sequence with a single asparagine residue that separates the tyrosine closest to the O-linked threonine in the murine sequence (Fig. 6C). Human-murine GP-6 substituted this single asparagine in the murine sequence with the five-residue human peptide (Fig.  6D). The substitution in murine-human GP-6 eliminated binding of anti-human PSGL-1 mAb but introduced binding of anti-murine PSGL-1 mAb 4RA10. Conversely, the substitution in human-murine GP-6 nearly eliminated binding of 4RA10 but introduced binding of PL1. These data confirm the previously determined epitope for PL1 (37) and further localize the 4RA10 epitope (30). Labeled A. phagocytophilum bound in a PL1inhibitable manner to human 2-GP-6 and, to a slightly lesser extent, to human-murine GP-6 (Fig. 6E). In marked contrast, A. phagocytophilum did not bind detectably to murine GP-6 or murine-human GP-6 (Fig. 6F). These results demonstrate that a peptide sequence found in human but not murine PSGL-1 is required for binding to A. phagocytophilum.

A. phagocytophilum Binds to Cells Expressing Human PSGL-1 and FTVII in Cooperation with N-and O-glycans-
Bacterial binding to glycopeptide-coated microspheres required cooperative interactions with both the human PSGL-1-derived peptide and a sLe x -capped O-glycan attached to Thr-16. Because cooperative binding did not require a specific orientation of the O-glycan at Thr-16, we asked whether A. phagocytophilum might bind to PSGL-1-expressing cells through interactions with sLe x presented on glycans at other locations. For this purpose we used transfected CHO cells expressing wild-type human PSGL-1 or PSGL-1 in which Thr-16 was replaced with an alanine (T16A). The cells coexpressed FTVII and core2GlcNAcT-I, which enables construction of sLe x on both N-glycans and core-2 O-glycans. Other cells coexpressed wildtype human PSGL-1 with FTVII but not with core2GlcNAcT-I. These cells express only short core-1 O-glycans that cannot be ␣1-3-fucosylated (34, 42). PL1 (10 g/ml) bound slightly less to cells expressing T16A than wild-type PSGL-1 (Fig. 7A), but 50 g/ml PL1 bound comparably to both cell lines (data not shown). The reduced affinity of PL1 for T16A is consistent with the mutation introduced in the center of the PL1 epitope. Anti-PSGL-1 mAb PL2 bound equivalently to cells expressing both wild-type and T16A PSGL-1 (Fig. 7B), confirming that the cells expressed comparable levels of PSGL-1. All three cell lines were equivalently fucosylated as measured by binding of anti-sLe x mAb HECA-452 (Fig. 7C). In contrast, CHO-131, a mAb that recognizes sLe x only on core-2 O-glycans (31), bound strongly to CHO cells expressing both FTVII and core2GlcNAcT-I but very weakly to cells expressing only FTVII (Fig. 7D). Labeled A. phagocytophilum bound to all three cell lines (Fig. 7E). PL1 prevented binding, confirming the requirement for PSGL-1 (Fig.  7F). Bacteria bound slightly less to cells expressing T16A than to cells expressing wild-type PSGL-1, suggesting that an O-glycan at Thr-16 partially contributes to binding (Fig. 7E), yet A. phagocytophilum bound equally well to cells expressing wildtype PSGL-1 in the presence or absence of core2GlcNAcT-I. These results demonstrate that A. phagocytophilum binds readily to PSGL-1-expressing cells in the absence of an O-glycan at Thr-16. They further suggest that the bacteria can use fucosylated N-glycans, as well as O-glycans, to cooperatively interact with PSGL-1.
To further explore the relative contributions of N-and Oglycans to binding to A. phagocytophilum, we treated transfected CHO cells with benzyl-GalNAc or kifunensine. Benzyl-GalNAc competitively inhibits extension and modification of O-glycans, resulting in accumulation of short Gal␤1-3GalNAc and GalNAc O-glycans linked to serine or threonine (43). Kifunensine, a mannosidase I inhibitor that prevents trimming of mannose residues from high mannose N-glycans, prevents synthesis of complex N-glycans that are acceptors for ␣1-3-fucosyltransferases (44). Kifunensine treatment was effective, as documented by its ability to markedly reduce binding of PHA-L, a lectin that recognizes tri-and tetraantennary complex N-glycans containing outer galactose residues and an ␣-linked mannose residue substituted at the C-2 and C-6 positions (45) (Fig. 8, A and B). Benzyl-GalNAc treatment was effective, as demonstrated by its ability to markedly reduce binding of mAb CHO-131 to CHO cells expressing both FTVII and core2GlcNAcT-I (Fig. 8C). It had no further effect on the low level binding of CHO-131 to CHO cells expressing only PSGL-1 and FTVII, suggesting that this background binding is not to fucosylated core-2 O-glycans (Fig. 8D).
Benzyl-GalNAc and kifunensine each partially reduced binding of HECA-452 to cells coexpressing PSGL-1, FTVII, and core2GlcNAcT-I, indicating that these cells express sLe x on both N-and O-glycans (Fig. 8E). In contrast, kifunensine virtually eliminated HECA-452 binding to cells coexpressing only PSGL-1 and FTVII, whereas benzyl-GalNAc had no effect (Fig.  8F). This further documents that these cells express sLe x primarily on N-glycans. Benzyl-GalNAc and kifunensine each partially inhibited binding of A. phagocytophilum to cells expressing PSGL-1, FTVII, and core2GlcNAcT-I (Fig. 8G). In cells expressing only PSGL-1 and FTVII, kifunensine eliminated bacterial binding, whereas benzyl-GalNAc had no effect (Fig.  8H). These data demonstrate that A. phagocytophilum can bind to PSGL-1 in cooperation with fucosylated N-and O-glycans. To confirm that this phenomenon was not specific to CHO cells, we studied human myelocytic HL-60 cells. Because A. phagocytophilum infects HL-60 cells as readily as human neutrophils, HL-60 cells are used to propagate the pathogen in vitro (27,28). Kifunensine and benzyl-GalNAc inhibited, respectively, binding of PHA-L and CHO-131 to HL-60 cells (Fig. 9, A  and B), and each partially inhibited binding of HECA-452 (Fig.   9C) and A. phagocytophilum (Fig. 9D). Thus, A. phagocytophilum uses fucosylated N-and O-glycans to interact with human myeloid cells or transfected cells expressing human PSGL-1.
A. phagocytophilum Binds to Microspheres Coupled Independently with PSGL-1-derived Glycopeptide Lacking sLe x and with sLe x Lacking Peptide-Human PSGL-1 has many O-glycans and up to three N-glycans (46 -49). matched as measured by PL1 binding (Fig. 10, A and E) and HECA-452 binding (Fig. 10, B and F), respectively. At the densities studied, A. phagocytophilum did not bind to microspheres bearing either 2-GP-1 or sLe x . In contrast, the bacteria bound in a PL1-inhibitable manner to microspheres bearing 2-GP-6 or 2-GP-1 plus sLe x (Fig. 10, C, D, G, and H). At high densities, the degree of bacterial binding to microspheres coupled with 2-GP-1 plus sLe x approached that to microspheres coupled with 2-GP-6 (Fig. 10G). These results demonstrate that A. phagocytophilum binds cooperatively to PSGL-1 peptide lacking sLe x and to peptide lacking sLe x that are independently coupled to the same microsphere. This suggests that the pathogen might bind to cells through cooperative interactions with the N terminus of PSGL-1 and with sLe x -capped N-or Oglycans on other cell-surface proteins. DISCUSSION The pathogen A. phagocytophilum employs some features that P-selectin uses to bind to neutrophils (25). The dependence of bacterial binding on expression of both human PSGL-1 and an ␣1-3-fucosyltransferase has been interpreted to indicate that A. phagocytophilum binds to fucosylated PSGL-1 (26). The ability of mAbs to the N-terminal region of human PSGL-1 to inhibit bacterial binding is consistent with a model in which A. phagocytophilum binds to the same N-terminal region of PSGL-1 recognized by P-selectin. This model is also supported by the dependence of bacterial binding on expression of ␣1-3fucosyltransferases in murine neutrophils (27). However, A. phagocytophilum binds readily to PSGL-1-deficient murine neutrophils, suggesting more complex models for bacterial binding to neutrophils (27). Here we have further defined the molecular features that A. phagocytophilum recognizes. These features overlap but are clearly distinct from those that Pselectin uses to bind to human and murine neutrophils.
All published studies of A. phagocytophilum binding or infection employ intact bacteria. We measured binding of fixed bacteria to ligand-coupled microspheres or to cells. This allowed determination of requirements for binding independent of signaling or other events that might regulate bacterial in- ternalization or viability within cells. It is important to note that this assay involves multivalent adhesion of one particle (fixed bacteria) to another particle (microsphere or cell). We coupled molecularly defined ligands to the microspheres or used cells with at least partially characterized cell surface structures. However, the adhesin or adhesins on A. phagocytophilum that mediate adhesion are unknown. Thus, interpretations of these binding data and previous binding data must include the possibility that A. phagocytophilum uses more than one adhesin to interact with the target particle and that target cells display more than one ligand. The bacterial binding assay differs from assays that measure binding of purified P-selectin to defined glycoconjugates, as well as to cell surfaces.
Studies with microspheres demonstrated that A. phagocytophilum bound to purified human PSGL-1 or to a small glycosulfopeptide modeled after the extreme N-terminal region of mature human PSGL-1 that binds to P-selectin. Therefore adhesion does not need additional cell surface molecules, and initial examination suggests that the same region of PSGL-1 that binds to P-selectin also binds to A. phagocytophilum. Consistent with this notion, bacterial adhesion required sialylation and ␣-1-3-fucosylation of the O-glycan at Thr-16 of the glycopeptide. Unlike P-selectin binding, however, bacterial adhesion required neither tyrosine sulfation of the glycopeptide nor a specific core-2 alignment of sLe x on the O-glycan at Thr-16. Indeed, significant bacterial binding was observed to microspheres coupled with both sLe x lacking peptide and with a PSGL-1-derived peptide lacking an sLe x -capped O-glycan at Thr-16. Bacterial adhesion required cooperative interactions with sLe x lacking peptide and with peptide lacking sLe x , because neither alone supported significant adhesion even at high densities. Bacterial interaction with the peptide required a five-residue sequence found in human but not murine PSGL-1. The peptide interaction region may be larger, because the amino acid sequence C-terminal to this five-residue segment is nearly identical in humans and mice. This species-dependent difference in bacterial binding to human and murine PSGL-1 may explain why A. phagocytophilum requires a PSGL-1independent mechanism to bind to murine neutrophils (27).
The adhesion of A. phagocytophilum to microspheres that presented sLe x and a human PSGL-1-derived peptide in different orientations suggests that the bacteria might also recognize sLe x presented at different positions on PSGL-1-expressing cells. Our studies with HL-60 cells and transfected CHO cells confirm this interpretation. A. phagocytophilum bound to transfected cells expressing T16A PSGL-1, demonstrating that bacterial adhesion, unlike P-selectin binding, does not require an O-glycan at Thr-16 of PSGL-1. Inhibitors of N-or O-glycosylation revealed that A. phagocytophilum bound to PSGL-1 in cooperation with sLe x presented on both N-and O-glycans. It is formally possible that that these glycans must be on PSGL-1, which is both N-and O-glycosylated (46,47,50). However, it seems likely that bacterial adhesion to neutrophils or transfected cells involves cooperative interactions with the N-terminal peptide sequence of human PSGL-1 and with N-and/or O-glycans on both PSGL-1 and other glycoproteins. The moderately diminished bacterial adhesion to cells expressing T16A PSGL-1 suggests that sLe x presented on an O-glycan at this residue may be a favored but not exclusive interaction site.
It is conceivable that A. phagocytophilum expresses a single adhesin with separate binding sites for the N-terminal peptide of human PSGL-1 and for sLe x or related glycans. This adhesin might simultaneously bind to peptide and sLe x on the same PSGL-1 molecule. However, the ability of bacteria to adhere to microspheres or cells presenting PSGL-1 and sLe x in different orientations suggests that even a single adhesin might bind to peptide on one PSGL-1 molecule and to sLe x on another PSGL-1 molecule or on a different glycoprotein. It is equally conceivable that one adhesin binds to the PSGL-1 peptide and that a molecularly distinct adhesin binds to sLe x . Indeed the available data do not exclude the possibility that different adhesins recognize, respectively, sialic acid and ␣1-3-linked fucose. Other bacteria use multiple adhesins to facilitate infection. A striking example is Helicobacter pylori, which expresses one adhesin that binds to the fucosylated glycan Lewis b and another adhesin that binds to sLe x -related glycans on gastric mucosa (51). The PSGL-1-independent binding of A. phagocytophilum to murine neutrophils strengthens the concept of multiple adhesins. Here binding might require a specific adhesin that recognizes an uncharacterized molecule on murine but not human neutrophils. This putative interaction might cooperate with binding of another adhesin(s) to sLe x or related glycans, allowing the bacterium to infect murine neutrophils as effectively as human neutrophils. Further definition of A. phagocytophilum interactions with neutrophils will require isolation and characterization of the bacterial adhesin or adhesins that recognize protein or carbohydrate structures on the neutrophil surface.