Endoglycan, a member of the CD34 family, functions as an L-selectin ligand through modification with tyrosine sulfation and sialyl Lewis x.

During lymphocyte homing to secondary lymphoid organs and instances of inflammatory trafficking, the rolling of leukocytes on vascular endothelium is mediated by transient interactions between L-selectin on leukocytes and several carbohydrate-modified ligands on the endothelium. Most L-selectin ligands such as CD34 and podocalyxin present sulfated carbohydrate structures (6-sulfated sialyl Lewis x or 6-sulfo-sLex) as a recognition determinant within their heavily glycosylated mucin domains. We recently identified endoglycan as a new member of the CD34 family. We report here that endoglycan, like the two other members of this family (CD34 and podocalyxin) can function as a L-selectin ligand. However, endoglycan employs a different binding mechanism, interacting with L-selectin through sulfation on two tyrosine residues and O-linked sLex structures that are presented within its highly acidic amino-terminal region. Our analysis establishes striking parallels with PSGL-1, a leukocyte ligand that interacts with all three selectins, mediating leukocyte-endothelial, leukocyte-leukocyte, and platelet-leukocyte interactions. Since the distribution of endoglycan includes hematopoietic precursors and leukocyte subpopulations, in addition to endothelial cells, our findings suggest several potential settings for endoglycan-mediated adhesion events.

During lymphocyte homing to secondary lymphoid organs and instances of inflammatory trafficking, the rolling of leukocytes on vascular endothelium is mediated by transient interactions between L-selectin on leukocytes and several carbohydrate-modified ligands on the endothelium. Most L-selectin ligands such as CD34 and podocalyxin present sulfated carbohydrate structures (6-sulfated sialyl Lewis x or 6-sulfo-sLex) as a recognition determinant within their heavily glycosylated mucin domains. We recently identified endoglycan as a new member of the CD34 family. We report here that endoglycan, like the two other members of this family (CD34 and podocalyxin) can function as a L-selectin ligand. However, endoglycan employs a different binding mechanism, interacting with L-selectin through sulfation on two tyrosine residues and O-linked sLex structures that are presented within its highly acidic aminoterminal region. Our analysis establishes striking parallels with PSGL-1, a leukocyte ligand that interacts with all three selectins, mediating leukocyte-endothelial, leukocyte-leukocyte, and platelet-leukocyte interactions. Since the distribution of endoglycan includes hematopoietic precursors and leukocyte subpopulations, in addition to endothelial cells, our findings suggest several potential settings for endoglycan-mediated adhesion events.
The selectins are a family of three cell adhesion molecules that participate in the initial interaction of leukocytes with the vascular endothelium (reviewed in Refs. 1 and 2). They function as lectin-like receptors by virtue of C-type lectin domains at their amino termini. While E-and P-selectin are expressed on activated endothelium, and the latter on platelets as well, L-selectin is broadly distributed on leukocytes. Counter-receptors ("ligands") for the selectins contain carbohydrate-based post-translational modifications that are essential for recognition by the lectin domains. The first established function for L-selectin was in the process of lymphocyte homing, where it mediates the tethering and rolling of lymphocytes along high endothelial venules (HEV) 1 of lymph nodes (reviewed in Refs. 3 and 4). Subsequently, this function was extended to Peyer's patches (3). The rolling step is critical as it initiates the recruitment of lymphocytes into secondary lymphoid organs where responses to antigen are triggered. Several HEV-expressed molecules have been identified as potential physiological ligands for L-selectin in mouse lymph lymphoid organs and human tonsils (reviewed in Ref. 5). These include GlyCAM-1 (mouse), CD34 (mouse and human), podocalyxin (human), MAdCAM-1 (mouse and human). Recently, endomucin (mouse and human) has been added to this list of candidates (6). All of these ligands are sialomucin in nature, containing a predominance of O-linked carbohydrate chains. Presented by these O-glycans are three post-translational modifications, i.e. ␣2-3 sialylation, ␣1-3 fucosylation, and carbohydrate 6-O-sulfation, all of which contribute to the interaction with the lectin domain of L-selectin (5). Mounting evidence indicates these features are accommodated in the capping group 6-sulfo sLex in which the sLex tetrasaccharide (Sia␣2,3Gal␤1,4-[Fuc␣1,3] GlcNAc) is modified with sulfation on the C-6 position of GlcNAc (7)(8)(9). This structure can be presented as the terminus of both a core-2 branch and an extended core-1 branch of an O-glycan, and contributes to ligand activity in both positions (9,10). A potential role for Gal-6-sulfate has also been suggested for ligands in the mouse, but the structural context for this modification is presently unknown (7,11).
The critical role of sulfation, in particular the GlcNAc-6sulfate modification, has emerged most clearly through the characterization of the antibody MECA-79 (12). This antibody binds GlcNAc-6-sulfate containing structures within L-selectin ligands in a sulfation-dependent, yet fucosylation-and sialylation-independent manner (9,13). Even though MECA-79 recognizes only a subregion of the L-selectin determinant, it is function-blocking and recognizes all of the L-selectin ligands identified thus far on a molecular level. It was used in the original identification of "peripheral lymph node addressin" (PNAd), a complex of L-selectin ligands expressed in mouse and human secondary lymphoid organs (12,14). Included in the complex are the aforementioned CD34, podocalyxin, Gly-CAM-1, and endomucin. Recent studies have identified an HEV-expressed GlcNAc-6-O-sulfotransferase, known as HEC-GlcNAc6ST (or LSST), that contributes to the formation of the MECA-79 epitope and L-selectin ligand activity within mouse lymph nodes (15)(16)(17). The fucosylation requirements of L-selectin ligands are met primarily by fucosyltransferase VII (FTVII) with some participation by FTIV (18).
While MECA-79 defines a major class of HEV-expressed ligands for L-selectin, it is clear that MECA-79 unreactive ligands exist on vascular endothelium both within lymphoid organs and in inflammatory extralymphoid sites (17, 19 -23). The molecular identify of these ligands is unknown. Furthermore, several studies indicate that L-selectin can mediate leukocyte-leukocyte adhesion (24 -26). Such interactions could amplify primary leukocyte-endothelial interactions at inflammatory sites by facilitating the capture of additional leukocytes from the blood. These secondary L-selectin-dependent interactions are mainly supported by PSGL-1, a mucin-like glycoprotein that is broadly expressed on leukocytes where it serves as the primary ligand for P-selectin (reviewed in Ref. 27). In distinction from the PNAd interactions, which involve sulfated sLex structures within the mucin domains, the interaction of L-selectin with PSGL-1 involves recognition of a sLex determinant and tyrosine sulfates within the amino terminus of the molecule (28 -30). While PSGL-1 is the predominant leukocyte ligand for L-selectin, alternative ligands apparently also exist (31,32).
In the present study, we identify endoglycan as a novel ligand for L-selectin and show that it employs a binding mechanism similar to that of PSGL-1. Endoglycan was originally cloned based on sequence homologies in its cytoplasmic domain to CD34 and podocalyxin (33). It shares a similar overall structure with these molecules, including a mucin-like domain and a membrane proximal globular domain. Based on these structural homologies and similar genomic organizations (34 -36), endoglycan, podocalyxin, and CD34 define a gene family known as the CD34 family. In contrast to CD34 and podocalyxin, endoglycan could not be detected in PNAd isolated from human tonsils (37). Instead endoglycan possesses a 171 amino acid amino-terminal region, containing an abundance of anionic residues and two tyrosine residues that are potential sites for sulfation. In further contrast to CD34 and podocalyxin, endoglycan is a proteoglycan that is modified with chondroitin sulfate (CS) chains. Endoglycan, like CD34 and podocalyxin, is widely expressed. Its distribution includes vascular endothelium, smooth muscle, hematopoietic precursor populations and leukocyte subpopulations (33). Here, we demonstrate that endoglycan with appropriate post-translational modifications interacts with L-selectin through recognition of sLex and tyrosine sulfates in the amino-terminal region of the molecule. Our analysis uncovers striking parallels between endoglycan and PSGL-1.

EXPERIMENTAL PROCEDURES
Endoglycan cDNA Constructs-The generation of the EG/IgG and AD/IgG expression constructs in vector pEF-BOS has been described (33). The AD/IgG mutants YYA and FYT were generated by site-directed mutagenesis following standard protocols. Enzymes and the appropriate buffers were purchased from NEB Inc., the dNTPs from Roche Applied Science. All other AD/IgG mutants were generated by PCR-based, site-directed mutagenesis using the QuikChange mutagenesis kit (Stratagene). Mutagenic primers (Sigma, Genosys) were used to introduce single amino acid exchanges as described in the result section and a selection primer to destroy a single EagI site within an exon of the IgG gene. All nucleotide exchanges were verified by sequencing.
Laminar Flow Assay-Laminar flow assays were performed as described by 38. In brief, the coating densities of the substrate proteins were equalized by ELISA, coating various concentrations of endoglycan F c chimeras or human IgG (Zymed Laboratories Inc.) onto 96-well plates (BD Biosciences) and detecting bound protein with goat antihuman IgG (F c specific) alkaline-phosphatase-conjugated mAb (Caltag). 3.5-cm dishes were coated with the chimeras (input concentration equivalent to 10 -20 g/ml of human IgG) and incorporated in a parallel flow chamber. For inhibition studies, cells were pretreated with 5 g/ml DREG 56 (Caltag Laboratory) or 10 mM EDTA. For enzyme treatments, coated substrates were incubated with 10 milliunits of sialidase (Vibrio cholerae, GLYKO, Inc.) in 50 mM sodium acetate, 4 mM CaCl 2 , and 0.1% bovine serum albumin, pH 5.5 or with 20 milliunits/ml C-ABC (Proteo vulgaris, GLYKO, Inc.) or with buffer alone as a control.
To compare the number of rolling cells on different substrates, Jurkat T cells were perfused through the flow chamber and counts of rolling cells were determined over a range of different shear stresses (1-8 dyn/cm 2 ). Two samples of each substrate preparation were analyzed four times each. Statistical significance was determined by unpaired Student-Newman-Keuls multiple comparison tests. If not otherwise stated, values represent the mean Ϯ S.E. for one representative experiment. For all of the experiments, the same results were obtained in at least three separate determinations employing independently produced protein preparations. Cell counts and velocities were evaluated using the NIH Image program.
Affinity Binding to L-selectin-For precipitation studies, L-selectin/ IgG chimera was coated onto protein A (100 g of protein per 10 l beads) and incubated with 35 SO 4 -conditioned medium collected from transfected COS-7 or CHO cells. To determine the requirement for sialylation, conditioned medium was supplemented with 50 mM NaAc and 4 mM CaCl 2 , adjusted to pH 5.5 with HCl and incubated with or without 100 milliunits/ml sialidase added. The samples were readjusted to pH 8 with Tris/HCl, pH 9 prior to precipitation. To determine the requirement for fucose, conditioned medium produced without FT-VII was used. 35 SO 4 -labeled AD/IgG chimeras were released from the beads with 10 mM EDTA. The concentrated samples were subjected to SDS-PAGE (7.5%). Prior to autoradiography, the gels were pretreated with Fluro-Hance (Research Products Int.).
Biochemical Characterization of AD/IgG and Mutant Proteins-To determine the character of the sulfation, purified 35 SO 4 -labeled AD/IgG was incubated with 20 milliunits/ml C-ABC in 10 mM Tris/HCl, pH 8 for 2 h at 37°C. A fraction of this reaction was adjusted to pH 7.1 and further incubated with 33 milliunits of aryl-sulfatase (Aerobacter aerogenes, E.C. 3.1.6.1, Sigma) in a 20-l volume. The samples were analyzed as above. To determine the 35 SO 4 incorporation, equal amounts of purified AD/IgG as determined by ELISA were digested with C-ABC for 30 min. To investigate the modification of the AD/IgG threonine mutants with sLex, 10 g of HECA 452 (BD Pharmingen) were coated to 10 l of anti-rat kappa -Sepharose beads (Zymed Laboratories Inc.) and used as described above to precipitate 35 SO 4 -labeled AD/IgG from conditioned medium of COS-7 or CHO cells. The cells were transfected with the indicated combination of modifying enzymes FTVII and Core2GlcNAcT-1. AD/IgG was released from the beads by boiling with Laemmli sample buffer.
Immunoprecipitation with Endoglycan-specific Antibody-Stable transfectants of endogycan-expressing cells were generated by transfecting CHO cells with the PCLP-2/pEAK10 plasmid, containing the endoglycan full-length cDNA and a selection marker for puromycin as described above. Cells were maintained for 10 days in selection medium supplemented with puromycin. Transfectants expressing high levels of endoglycan were determined by flow cytometry analysis employing an endoglycan-specific antibody generated against a GST fusion protein of endoglycan (33). For 35 SO 4 labeling, cells were grown to ϳ70% confluence in a 175 cm 2 flask and cultured as described above with 2 mCi of Na 2 35 SO 4 . Labeled cells were lysed in PBS, 2% Triton X-100, 5 mM EDTA, and a protease inhibitor mixture. The lysate was incubated with protein A-agarose, which had been bound with 5 g of the endoglycan-specific antibody or normal rabbit IgG (Zymed Laboratories Inc.). Aliquots of the precipitates were boiled in Laemmli sample buffer with or without 5% ␤-mercaptoethanol and subjected to SDS-PAGE (5%).

Post-translation Requirements for Endoglycan as an L-selectin
Ligand-To determine if endoglycan can serve as an Lselectin ligand, we examined the rolling of L-selectin bearing Jurkat-T cells on recombinant EG/IgG (extracellular domain of endoglycan fused to the Fc region of human IgG) in a parallel plate flow chamber. The optimal binding of L-selectin to PNAd components, such as GlyCAM-1 and CD34, require specific sialylation, fucosylation, and carbohydrate sulfation (5). To determine the contributions of ␣1,3 fucosylation and GlcNAc-6-sulfation, endoglycan was expressed in the presence or absence of cDNAs for FTVII and HEC-GlcNAc6ST, which provide these modifications. Sialylation was provided by endogenous ␣2,3 sialyltransferases present in COS-7 cells. The resulting recombinant proteins were purified from conditioned medium and coated onto the bottom plate of a flow chamber at equal site densities as determined by ELISA.
As shown in Fig. 1A, appropriately modified endoglycan supported tethering and rolling of Jurkat cells over a shear stress range of 1-8 dyn/cm 2 . Since endoglycan was not recognized by MECA-79, we wanted to determine the effect of GlcNAc-6sulfation on its ligand activity. In contrast to GlyCAM-1 and CD34, Jurkat cell rolling on recombinant EG/IgG was not augmented by GlcNAc-6-sulfation imparted by HEC-GlcNAc6ST, as measured either by the number of rolling cells (Fig. 1A) or rolling velocity (Fig. 1B). The interactions were nevertheless L-selectin-dependent, since rolling was completely blocked by L-selectin antibody (DREG-56) or by chelation of calcium with EDTA. There was no rolling in the absence of FTVII, establishing an absolute requirement for fucosylation. Sialylation was also required, since treatment of the EG/IgG substrate with sialidase completely abrogated rolling. The rolling on FTVIImodified endoglycan was comparable to that of HEC-GlcNAc6ST-and FTVII-modified GlyCAM-1/IgG as assayed under identical conditions (38).
Ligand Activity of the Amino-terminal Domain of Endoglycan-These results suggested a different binding mechanism for endoglycan and led us to investigate the functional importance of the distinct amino-terminal region of endoglycan. We therefore employed an Fc fusion protein containing just the amino-terminal domain of the molecule (AD/IgG) (33). Recombinant EG/IgG and AD/IgG chimeras were produced in the presence of exogenously provided FTVII and used as substrates in flow chamber experiments as above. When coated at equal densities, both substrates supported the rolling of Jurkat cells to comparable extents, as measured both by the number of rolling cells ( Fig. 2A) and the velocity of rolling (Fig. 2B).
Earlier experiments with AD/IgG had demonstrated that the amino-terminal domain of endoglycan is modified with CS (33). To investigate a potential contribution of CS chains toward the ligand activity, AD/IgG was treated with chondroitinase ABC (C-ABC) prior to the flow chamber testing. This treatment did not alter the L-selectin mediated rolling of Jurkat cells on AD/IgG, indicating that the CS modification, at least as provided by the COS-7 cell system, did not influence the binding activity of the amino-terminal domain of endoglycan. The interaction of AD/IgG with L-selectin was confirmed to be dependent on the modification with sialic acid and fucose as above (Fig. 3A).
To verify ligand activity of endoglycan by another means, we employed precipitation of AD/IgG with an L-selectin/IgG fusion protein, as has been used in the previous characterizations of ligands in the PNAd category (13, 35, 39, 40). 35 SO 4 -labeled AD/IgG produced in the presence or absence of FTVII cDNA was precipitated by L-selectin/IgG coated beads. As described previously (33), AD/IgG separates into two distinct bands when subjected to SDS-PAGE: a broad relatively high molecular weight component, which represents the CS-modified form and a sharper band of 80 kDa, which lacks the CS modification.

Endoglycan Is a Novel PSGL-1-like Ligand
With FTVII included in the transfection, L-selectin bound AD/ IgG in a calcium-dependent manner as shown by elution of AD/IgG by EDTA (Fig. 3B). In the absence of FTVII, the reactivity of AD/IgG with the L-selectin chimera was dramatically reduced. Sialidase treatment of the FTVII-modified AD/IgG eliminated its interaction with the L-selectin chimera.
Biochemical Analysis of Endoglycan-Since recombinant AD/IgG expressed with exogenous FTVII manifested L-selectin ligand activity by two independent assays, we sought to determine the nature of the critical post-translational modifications of this molecule. As stated above, treatment of 35 SO 4 -labeled AD/IgG with C-ABC eliminated the broad high molecular weight component, leaving a sharper band with a molecular weight of ϳ80 kDa. Subsequent treatment of this component with aryl-sulfatase (Aerobacter aerogens) nearly eliminated the remaining 35 SO 4 signal, indicating that the amino-terminal domain of endoglycan carries tyrosine sulfation in addition to CS modification (Fig. 4A). As described below, this conclusion was substantiated by the analysis of endoglycan mutants lacking the key tyrosines.
To determine whether sialylation and fucosylation occurred together within sLex-related structures, we immunoprecipitated 35 SO 4 -labeled AD/IgG with HECA-452, a sLex reactive mAb (41). HECA-452 precipitated AD/IgG when the protein was co-expressed with FTVII, but no reactivity was detected in the absence of the fucosyltransferase (Fig. 4B). sLex structures are found on both O-and N-linked carbohydrate chains (42). The elaboration of sLex structures on O-glycans, in contrast to that on N-glycans, depends on the presence of a core-2 branching enzyme such as Core2GlcNAcT-I (43). Since COS-7 cells express this activity, whereas CHO cells do not, we transfected CHO cells with cDNAs for AD/IgG, and Core2GlcNAcT-I and FTVII in various combinations. As shown in Fig. 4B, HECA-452 reactivity depended on the presence of both Core2GlcNAcT-I and FTVII, indicating that the modification of AD/IgG with sLex is restricted to O-linked chains. Additionally, we observed that AD/IgG expressed in CHO cells contained a higher proportion of the high molecular species as compared with AD/IgG produced in COS-7 cells. Since the high molecular component remained sensitive to treatment with C-ABC (not shown), this result suggests that that CHO cells can produce AD/IgG with a higher degree of CS modification than COS-7 cells.
Similarities between Endoglycan and PSGL-1-Endoglycan contains three cysteines in its membrane proximal domain. Sequence alignments with family members CD34 and podocalyxin suggest that two of the cysteines are involved in intrachain stabilization of the membrane proximal globular domain. The third cysteine residue is located in close proximity to the transmembrane domain, similar to an unpaired cysteine found in PSGL-1 (44). PSGL-1 forms a disulfide-dependent dimer that is resistant to SDS denaturation (45). To determine whether endoglycan behaved similarly, we transfected CHO cells with a full-length cDNA for endoglycan. The cells were labeled with 35 SO 4 and the lysate was subjected to immunoprecipitation with a polyclonal antibody to endoglycan. SDS-PAGE performed in the presence of ␤-mercaptoethanol revealed a band of 150 kDa molecular mass, while in the absence of ␤-mercaptoethanol, a band of ϳ300 was seen, consistent with a disulfide-linked dimer (Fig. 5).
Further similarities between endoglycan and PSGL-1 were revealed by comparison of amino acid sequences of their aminoterminal domains (Fig 6A). As noted above, this region of PSGL-1 carries post-translational modifications that are criti- cal for its interactions both with P-selectin and L-selectin (29,30). Present in the human sequence of PSGL-1 are 3 tyrosine residues that are modified by sulfation, and a single threonine, which carries a sLex modification (highlighted by asterisks in Fig. 6A). In contrast to PSGL-1 where the 3 tyrosines span 6 amino acids, endoglycan possesses 2 tyrosine residues separated by 16 amino acids. Significant sequence similarity was found around the third tyrosine of PSGL-1. The sequence alignment shows that 5 of 9 amino acids are identical with the endoglycan sequence including the last tyrosine and the downstream threonine.
In order to determine whether the amino acids of endoglycan that align with the critical residues of PSGL-1 were subject to similar post-translational modifications, we generated several point mutants of AD/IgG by site-directed mutagenesis. The resulting proteins carried amino acid exchanges as indicated in Fig. 6B. Either one or both tyrosine residues were replaced by phenylalanine, while Thr-124 was replaced by alanine. The constructs were named according to the amino acid changes in their sequence. Thus, the WT sequence is denoted YYT with the first Y representing Tyr-97, the second Y representing Tyr-118 and the T in the third position representing Thr-124. According to this designation system, the following mutants  were generated: FYT, YFT, FFT, YYA, and FFA.
To examine tyrosine sulfation, we prepared 35 SO 4 -labeled forms of the WT AD/IgG and the tyrosine mutants FYT, YFT, and FFT in COS-7 cells. Equal amounts of the Fc chimeras were subjected to chrondroitinase ABC treatment and analyzed by SDS-PAGE and densitometry. The 35 SO 4 incorporation of mutants FYT and YFT was 56% and 64%, respectively of the WT protein while the FFT was only 18% of the WT signal (Fig. 6C). The low level of residual sulfation might be attribut-able to sulfation of N-linked or O-linked chains within AD/IgG or to sulfation of GAG chain "stubs" after digestion with C-ABC.
To determine whether Thr-124 carried the sLex modification, we subjected equal amounts of WT, YYA, and FFA forms of AD/IgG, expressed in the presence or absence of FTVII, to immunoprecipitation with HECA-452. The reactivity of HECA-452 with sLex structures is fucose-dependent (14). Accordingly, no precipitation was observed for AD/IgG produced without FTVII, while a reduced but still detectable reactivity with mutants YYA and FFA was observed (Fig. 6D). These findings showed that the majority of sLex was associated with Thr-124 but indicated the presence of additional sLex structures within the amino-terminal region of endoglycan.
Analysis of Ligand Activity of AD/IgG Mutants-To investigate the functional importance of single modifications, the endoglycan mutants were analyzed for ligand activity in the flow chamber. When both tyrosine residues were mutated to phenylalanine (FFT), the rolling of Jurkat T-cells on recombinant endoglycan was reduced by more than 50% relative to the WT recombinant protein. A decrease in the number of rolling cells of up to 80% was observed when Thr-124 was mutated to alanine (YYA). Mutation of both tyrosines and the Thr-124 (FFA) did not cause any further decrease in the rolling activity (Fig. 7A). The rolling was nevertheless completely dependent on the coexpression with FTVII indicating that the residual sLex structures within the AD/IgG (shown in Fig. 6D) were contributing to the L-selectin interaction. These effects on rolling were observed over the entire range of shear stresses tested (1-8 dynes/cm 2 ). The mutations also affected rolling velocity. At a shear stress of 1 dyn/cm 2 , the mean rolling velocity was increased by more than 20% for the FFT mutant relative to the WT protein and 36% and 40% for YYA and FFA mutants, respectively (Fig. 7B).  White letters on black ground indicate identical residues, similar residues are highlighted with gray. Arrows denote the amino termini of the processed proteins, and asterisks denote residues within the PSGL-1 sequence that are critical for ligand function. Within the endoglycan sequence, the double underlined S indicates the attachment site for the CS GAG chain and the underlined N indicates a potential N-glycosylation site. B, the AD/IgG mutant constructs were named according to the amino acid residues introduced at positions 97, 118, and 124, replacing tyrosines by phenylalanine and/or threonine by alanine. C, 35 SO 4 -labeled AD/IgG (YYT) and AD/IgG mutants in which one or both tyrosine residues were replaced by phenylalanine (FYT, YFT, and FFT) were purified from COS-7 cell conditioned medium. Equal protein amounts as determined by ELISA were treated with 50 milliunits of C-ABC and subjected to SDS-PAGE (7.5%) and autoradiography. The relative amount of sulfate incorporation is given as a percentage of that for WT AD/IgG. D, equal amounts of 35 SO 4 -labeled AD/IgG expressed with or without FTVII and AD/ IgG mutants lacking Thr-124 (YYA and FFA) expressed with FTVII were precipitated with HECA-452 and analyzed as before.

DISCUSSION
In the present study, we present evidence that endoglycan, the third member of the CD34 family is a novel ligand for L-selectin. Even though endoglycan is closely related to the other members of the family, our experiments reveal that it employs a distinct binding mechanism for interacting with L-selectin. In contrast to CD34 and podocalyxin, endoglycan did not depend on the activity of carbohydrate 6-O-sulfotransferases in order to function as an L-selectin ligand. Thus, the cotransfection of FTVII alone was sufficient to impart ligand activity to endoglycan. Furthermore, the inclusion of a cDNA for HEC-GlcNAc6ST did not detectably enhance the rolling activity for endoglycan, nor did the inclusion of a cDNA for a Gal-6-O-sulfotransferase (KSGal6ST) (data not shown). Additionally, AD/IgG, lacking the mucin domain, exhibited comparable ligand activity to the chimera containing the entire extracellular domain (EG/IgG). However, both fusion proteins needed to be modified with fucose and sialic acid in order to function as L-selectin ligands. The suspicion that sLex-related structures were present within endoglycan was confirmed by showing that HECA-452 was able to precipitate AD/IgG and EG/IgG produced in the presence of FTVII but not in its absence. By using CHO cells in which the presence of the core-2 branching enzyme could be controlled by transfection, we were able to show that the sLex-related structures were restricted to O-linked glycans.
Our earlier characterization of native endoglycan in HUVEC and recombinant endoglycan (both EG/IgG and AD/IgG) in COS-7 cells has revealed a high level of 35 SO 4 incorporation, most of which was attributable to CS chains (33). The fact that AD/IgG contains only one potential attachment side for glycosaminoglycan (GAG) chains at position 79 of the unprocessed protein sequence establishes the utilization of this site.
Whether and to what extent the other potential GAG sides within the endoglycan sequence are modified by CS remains to be determined. The residual sulfation of AD/IgG remaining after C-ABC treatment was established to be due to tyrosine sulfation. First, it was greatly reduced after treatment with aryl-sulfatase. Secondly, we observed a dramatic loss of sulfate incorporation when either one of the two tyrosine residues within the AD region was replaced by phenylalanine, indicating that both tyrosines are subject to this modification.
Studies by Miyasaka and co-workers (46) have shown a subfraction of CS chains, in particular oversulfated chains such as CS-E, exhibit ligand activity for L-selectin. We found that treatment of FTVII-modified AD/IgG with C-ABC did not impair its ability to support lymphocyte rolling. Moreover, the precipitation studies indicated that both forms of AD/IgG could interact with L-selectin. Thus, the 80-kDa component lacking CS chains, bound to L-selectin just as efficiently as the high molecular weight component that contained CS chains. It should be pointed out that only a small fraction of endoglycan expressed in COS-7 cells was modified with CS, as judged by the relative sulfate incorporation of the CS modified and nonmodified forms. The CS-modified portion of endoglycan was higher in AD/IgG produced by CHO cells and endogenous en- FIG. 7. Rolling of Jurkat cells on endoglycan mutants. AD/IgG and mutant AD/IgG proteins were coated at equal site densities and tested in a parallel flow chamber as described in Fig. 1. Values represent the mean Ϯ S.E. for three independent experiments performed with independent preparation of recombinant proteins. A, the number of rolling cells per field at shear stresses 1-8 dyn/cm 2 . The rolling of each of the mutants was significantly reduced compared with the WT protein at shear stresses 1-6 dyn/cm 2 with probability values of p Ͻ 0.01 for 1 and 6 dyn/cm 2 and p Ͻ 0.001 for 2 and 4 dyn/cm 2 as determined by unpaired Student-Newman-Keuls Multiple Comparison. The number of rolling cells at 8 dyn/ cm 2 was too low for a statistical analysis. The YYA and FFA mutants were consistently weaker ligands than the FFT mutant; however, this trend achieved statistical significance only for shear stresses 1 and 2 dyn/cm 2 . B, the velocity of a minimum of three rolling cells was determined for each of the AD/IgG mutants at a shear stress of 1 dyn/cm 2 . The differences are statistically significant between the WT and each of the mutants with p Ͻ 0.0002 for FFT and p Ͻ 0.0001 for YYA and FFA, as well for the comparison of mutant FFT with each of the threonine mutants, YYA (p Ͻ 0.05) or FFA (p Ͻ 0.02). doglycan from HUVEC did not contain detectable amounts of the non-modified form (33). It remains to be determined whether endoglycan in particular vascular beds or on leukocytes carries oversulfated CS chains, which might facilitate interactions with L-selectin or P-selectin. Intriguingly, the same modifications of CS chains that promote selectin interaction also facilitate the binding of certain chemokines (46), raising the possibility that certain forms of endoglycan might be involved in chemokine sequestration. A dual function for a proteoglycan as a selectin ligand and chemokine presentation molecule has recently been proposed for collagen XVIII, which carries heparan sulfate chains (47).
Excluding the CS chain as critical for the ligand activity of endoglycan as produced in COS-7 cells left the tyrosine sulfation for further investigation in this respect. The presence of tyrosine sulfation and O-linked sLex within the amino-terminal region of endoglycan led us to compare endoglycan with PSGL-1. As reviewed above, PSGL-1 is a homodimeric selectin ligand that is expressed on the surface of almost all circulating leukocytes (27). Like endoglycan, it shows an absolute dependence on sialylation and fucosylation for ligand activity (1,27). As another parallel, we established that endoglycan, like PSGL-1, formed a disulfide-dependent homodimer in SDS-PAGE. The amino-terminal anionic region of human PSGL-1 contains 3 tyrosine residues, each of which can be sulfated, and a single sLex modified O-glycan at a proximal threonine residue (Thr-57). Alignment of the two amino-terminal sequences of endoglycan and PSGL-1 revealed a striking correspondence in a critical 9 amino acid peptide region. This peptide carries in identical spacing the last tyrosine (Tyr-51 in PSGL-1 and Tyr-118 in endoglycan) in both sequences as well as the threonine residue (Thr-57 in PSGL-1 and Thr-124 in endoglycan). PSGL-1 can interact with all three selectins (27). While the presence of the sLex structure is necessary for the binding in all cases, the sulfation requirements differ among the selectins. E-selectin binding to PSGL-1 is indifferent to tyrosine sulfation (28,48). In the case of P-selectin, all three tyrosine sulfates can interact but the major contribution is from Tyr-48 (50). 2 For L-selectin, sulfation of Tyr-51 plays the predominant role (30). Notably, this residue corresponds to Tyr-118 in the endoglycan sequence (Fig. 6A).
The parallels between endoglycan and PSGL-1 led us to generate a series of AD/IgG mutants lacking one or several key residues. Our findings established that a sLex modification of an O-glycan at Thr-124 accounts for up to 80% of the ligand activity. This is compatible with the result that the majority of HECA-452 reactivity was found to be associated with this residue. Rolling was strongly supported by sulfation on Tyr-97 and -118. However, about 50% of the activity remained when both tyrosines were mutated, indicating that the modification with sLex is capable of sustaining a degree of ligand activity in the absence of tyrosine sulfation. Parallel findings were reported for PSGL-1 (51). The small amount of ligand activity in the threonine mutant remained fucose-dependent and is probably attributable to O-linked sLex modifications on other threonines or serines in the amino-terminal domain. The existence of these additional sLex modifications for endoglycan is supported by the HECA-452 reactivity of the threonine mutants YYA and FFA. Since both of these mutants exhibited equivalent, albeit weak ligand activity, we conclude that the secondary sLex-bearing sites are not facilitated by the tyrosine sulfates. It is worth pointing out that the binding region of endoglycan (highlighted by the frame in Fig. 6A), containing both tyrosines and Thr-124, is highly conserved within the mouse sequence (GenBank TM accession number XM_194286). It differs only in 2 amino acids out of 33, preserving the potential attachment sites for N-and O-glycans, and tyrosine sulfation.
As reviewed in the introduction, numerous reports indicate the existence of MECA-79-independent L-selectin ligands on vascular endothelium, although their molecular identities have not been revealed so far. Based on the characterization presented herein, endoglycan represents a strong candidate for a number of these sites. Since endoglycan is also present on hematopoietic precursors and leukocyte subpopulations (33) our findings suggest the involvement of endoglycan in leukocyte-leukocyte in addition to leukocyte-endothelial interactions. If it is found that P-selectin and E-selectin can utilize endoglycan as a ligand, as is strongly suspected from its parallels to PSGL-1, the range of potential in vivo functions for this molecule would be further broadened since these selectins are found on platelets and/or activated vascular endothelium. Gene targeting approaches (52,53) and antibody blockade studies (49) have been essential in establishing the importance of PSGL-1 as a selectin ligand. Similar approaches will be necessary to gain a further understanding of endoglycan.