Biosynthesis of GlyCAM-1, a mucin-like ligand for L-selectin.

L-selectin, a member of the selectin family of leukocyte-endothelial adhesion proteins, mediates the initial attachment of lymphocytes to lymph node high endothelial venules during lymphocyte recirculation. One of the endothelial-associated ligands for L-selectin is GlyCAM-1, a mucin-like glycoprotein, which presents novel sulfated, sialylated and fucosylated O-glycans. In order to understand the generation of these glycans, we have examined the biosynthesis of GlyCAM-1 in lymph node organ culture. Using peptide-specific antibodies, lectins, and recombinant L-selectin, we detected the following species of GlyCAM-1: unglycosylated (<28 kDa); modified with GalNAc only (28-33 kDa); modified with sialic acid, fucose, and sulfate but lacking L-selectin reactivity (40-50 kDa); and mature (L-selectin-reactive) ligand (50-60 kDa). Pulse-chase labeling at 15°C suggested that GalNAc is added in a pre-Golgi compartment. Treatment with brefeldin A almost completely blocked sulfation, indicating that this modification occurs in the trans-Golgi network. Two distinct sialylation events occurred in the presence of brefeldin A, while fucosylation was partially blocked. We conclude that sialylation precedes both fucosylation and sulfation during biosynthesis. This ordering will help to identify the critical acceptor structures recognized by lymph node glycosyltransferases and sulfotransferases.

L-selectin, a member of the selectin family of leukocyte-endothelial adhesion proteins, mediates the initial attachment of lymphocytes to lymph node high endothelial venules during lymphocyte recirculation. One of the endothelial-associated ligands for L-selectin is Gly-CAM-1, a mucin-like glycoprotein, which presents novel sulfated, sialylated and fucosylated O-glycans. In order to understand the generation of these glycans, we have examined the biosynthesis of GlyCAM-1 in lymph node organ culture. Using peptide-specific antibodies, lectins, and recombinant L-selectin, we detected the following species of GlyCAM-1: unglycosylated (<28 kDa); modified with GalNAc only (28 -33 kDa); modified with sialic acid, fucose, and sulfate but lacking L-selectin reactivity (40 -50 kDa); and mature (L-selectin-reactive) ligand (50 -60 kDa). Pulse-chase labeling at 15°C suggested that GalNAc is added in a pre-Golgi compartment. Treatment with brefeldin A almost completely blocked sulfation, indicating that this modification occurs in the trans-Golgi network. Two distinct sialylation events occurred in the presence of brefeldin A, while fucosylation was partially blocked. We conclude that sialylation precedes both fucosylation and sulfation during biosynthesis. This ordering will help to identify the critical acceptor structures recognized by lymph node glycosyltransferases and sulfotransferases.
L-selectin is a lectin-like receptor that is widely expressed on the surface of circulating leukocytes (1,2). It plays a central role in lymphocyte-endothelial cell interactions in the normal recirculation of lymphocytes between the blood and secondary lymphoid organs (1,3). In addition, L-selectin, as well as the other members of the selectin family, E-and P-selectin, participate in the recruitment of various leukocytes to sites of inflammation (reviewed in 4 -7). In lymphocyte recirculation, bloodborne lymphocytes interact with the specialized endothelial cells of post-capillary high endothelial venules (HEV) 1 and ultimately extravasate across the endothelium into secondary lymphoid organs. L-selectin is essential for the initial adhesive interaction of lymphocytes with HEV of lymph nodes and also appears to be involved in the recruitment of lymphocytes to Peyer's patches (1,3,8,9). Leukocyte integrins and their endothelial counter-receptors participate in later steps of recruitment (10 -14).
A fundamental property of L-selectin is that it recognizes carbohydrate-based ligands via its lectin-like domain (reviewed in 7,15). An L-selectin/Ig chimera has been used to identify two HEV-associated ligands in mouse as GlyCAM-1 and CD34 (formerly designated Sgp50 and Sgp90) (16 -18). More recently, GlyCAM-1 has been observed in HEV-like vessels that are induced at sites of chronic inflammation (19). 2 MAdCAM-1, a member of the Ig superfamily, exists in mouse mesenteric lymph node HEV in a glycoform that is recognized by L-selectin (20). GlyCAM-1, CD34, and MAdCAM-1 are found at other sites as glycoforms (19,21,22) that do not exhibit high affinity binding with L-selectin. A fourth distinct ligand, Sgp200 is also present in mouse lymph nodes (23) but has not been identified at the molecular level. All of these HEV-associated glycoproteins possessing ligand activity for L-selectin are recognized by the function-blocking monoclonal antibody known as MECA 79 (16,23,24). Finally, a heparin-like ligand for L-selectin has been identified intracellularly in cultured endothelial cells (25).
Both GlyCAM-1 and CD34 are sialylated, fucosylated, and sulfated glycoproteins, and their primary sequence indicates that they are serine/threonine-rich mucin-like glycoproteins with many potential sites for O-linked glycosylation (16 -18). MAd-CAM-1 also possesses a short mucin domain (26) which is proposed to bear the carbohydrate recognition determinants for Lselectin. In the case of GlyCAM-1, all of the oligosaccharides are O-linked, adding approximately 35 kDa to a predicted core protein of 14 kDa (17,27). GlyCAM-1 is present at high levels in the conditioned medium of lymph node organ cultures and in serum (17,28). By EM immunocytochemistry, GlyCAM-1 is undetectable on the apical plasma membrane of the endothelial cells of HEV but is found in large cytoplasmic vesicles (29). Taken together, these observations suggest that GlyCAM-1 is a secreted product. In contrast, CD34 is an integral membrane protein (30).
Sialic acid and sulfate are critical components of the oligosaccharide ligands for L-selectin (16,31), and an essential role for fucosylation is strongly suspected (32). L-, E, and P-selectin recognize the sialyl Lewis x tetrasaccharide (sLe x , Neu5-Ac␣233Gal␤1-4(Fuc␣1-3)GlcNAc), and related structures (reviewed in Refs. 15,33), although each selectin has preferred biological ligands (34 -36). There has been recent interest in the possibility that sulfation may define a unique modification of L-selectin ligands, which greatly enhances their interaction with L-selectin. Direct structural analysis of GlyCAM-1 (37) has identified Gal-6-sulfate and GlcNAc-6-sulfate as the major sulfated monosaccharides in the context of N-acetyllactosamine, i.e. Gal␤134GlcNAc. Further structural studies revealed that 6Ј-sulfo sialyl Lewis x, i.e. Sia␣233(SO 4 -6)Gal-␤134[Fuc␣133]GlcNAc and 6-sulfo sialyl Lewis x, i.e. Sia␣23 3Gal␤134[Fuc␣133]GlcNAc-6SO 4 are major capping groups of this ligand (27,38). Structures of two of the simplest O-glycans of GlyCAM-1 are predicted (38) as Structures 1 and 2: These oligosaccharides contain the T-antigen, i.e. Gal␤13 3GalNAc, which is incorporated into the core-2 structure (39), Although there is increasing information about the carbohydrate structure of the biological ligands of L-selectin, as well as those for the two endothelial selectins (40,41), there have been no reports on the biosynthesis of these structures. The present study investigates the biosynthesis of O-linked oligosaccharides of GlyCAM-1 as it relates to the elaboration of functional ligand activity. We employ lectin analysis, pulse-chase labeling, and the inhibition of membrane transport (via reduced temperature and brefeldin A) for this analysis. We report the identification of GlyCAM-1 biosynthetic intermediates that represent distinct stages of the O-linked biosynthetic pathway. These studies provide a view of how L-selectin binding activity may be regulated at the level of O-glycan biosynthesis. Arachis hypogaea (peanut) agglutinin-agarose (5 mg PNA/ml gel), Vicia villosa agglutinin-agarose (3 mg VVA/ml gel), Sambucus nigra (elderberry bark) agglutinin-agarose (3 mg SNA/ml gel), and Lycopersicon esculentum agglutinin agarose (2 mg LEA/ml gel) were purchased from Vector (Burlingame, CA). Limax flavus agglutinin was obtained from Calbiochem. Aleuria aurantia agglutinin (AAA) was from Boehringer Mannheim, and Maackia amurensis agglutinin (MAA) was from Sigma. The latter three lectins were immobilized on CNBr-activated Sepharose 4B (Sigma) at a concentration of 2 mg lectin/ml gel (for Limax agglutinin and AAA) or 10 mg/ml gel (for MAA) following the protocol recom-mended by the manufacturer. The rabbit polyclonal anti-GlyCAM-1 preimmune sera, and anti-peptide 2 and anti-peptide 3 antibodies were produced as described previously (17). A recombinant L-selectin-human IgG1 chimera protein (LEC-IgG) was prepared as described previously (42) and generously provided by Larry Lasky and Susan Watson of Genentech Inc. LEC-IgG-Protein A-Sepharose (substituted at 10 mg LEC-IgG/ml of packed beads) was prepared with dimethyl pimelimindate (Pierce) following the procedure for cross-linking antibodies to Protein A, described by Harlow and Lane (43).

Metabolic Labeling and Immunoprecipitation
Steady-state Labeling-Mesenteric and peripheral (axillary, brachial, and cervical) lymph nodes were dissected from 18 to 24 mice (ICR) and diced in 1-mm 2 slices with a razor blade. The lymph nodes were divided into portions (lymph nodes from three mice/portion, referred to as three lymph node equivalents) and were radiolabeled in wells of a 24-well tissue culture plate. When indicated, the lymph nodes were preincubated with or without BFA and/or chlorate prior to labeling (see below). Before use, the 3 2 . The conditioned medium was then removed and the tissue was washed twice with 0.5 ml of Dulbecco's phosphate-buffered saline (PBS). The first wash was combined with the conditioned medium, diluted to 1.25 ml with PBS, and centrifuged at 10,000 ϫ g for 5 min at 4°C. The tissue from each well was homogenized in 1.25 ml of 2% lysis buffer (PBS with 2% Triton X-100 and 0.02% NaN 3 ) with protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1% (v/v) aprotinin, 10 g/ml pepstatin, and 2 g/ml leupeptin) with a Potter-Elvehjem homogenizer on ice. Lysis was continued for 1 h on a rocker at 4°C. The lysate was centrifuged at 10,000 ϫ g for 1 h at 4°C and the supernatant saved. Detergent lysates and conditioned medium generated for different conditions within the same experiment were normalized for total protein (see below) and subjected to preclearing and immunoprecipitation as described below.
Pulse-Chase Labeling-Lymph nodes (mesenteric, axillary, brachial, and cervical) were dissected from 25 mice and starved in 50 ml of Ser/Thr-free medium for 45 min at 37°C in 5% CO 2 . To 5 wells of a 24-well plate, 0.5 ml of Ser/Thr-free medium with 300 Ci each of [ 3 H]Ser and [ 3 H]Thr was added and prewarmed to 37°C. The lymph nodes were diced into 1-mm 2 pieces with a razor blade, divided into five portions, and placed on stainless steel wire grids in the wells of a 24-well plate containing the prewarmed labeling medium. After a 5 or 10 min incubation, the 24-well plate was placed on ice, and the wire grids holding the lymph nodes were quickly removed with forceps and vigorously dipped up and down in a centrifuge tube containing 50 ml of ice-cold chase medium (complete medium supplemented with a 200 molar excess of unlabeled Ser and Thr) to release the lymph nodes into the medium. The lymph nodes were pelleted by centrifugation at 150 ϫ g for 30 s at 0°C, and the supernatant decanted. The tissue was placed on a glass plate on ice and quickly divided into two times the number of portions needed for the chase points. Two independent portions were combined for each sample, which decreased the error in apportioning tissue. Each sample was divided in half and added to two wells of a 24-well plate containing 0.75 ml of chase medium/well and incubated for various lengths of time at 15, 20, or 37°C. Final chase times were calculated by adding the pulse-labeling time (i.e. 5 or 10 min) to the incubation time in chase medium. At the end of each time point, the conditioned medium was collected and placed on ice. The lymph nodes were quickly added to 5 ml of chase medium and centrifuged at 150 ϫ g for 30 s at 0°C. After aspirating the medium, the tissue was snap frozen in liquid N 2 . The conditioned medium was centrifuged at 10,000 ϫ g for 5 min at 4°C and also snap frozen in liquid N 2 . After all time points had been collected, the lymph nodes were thawed, and the tissue for each time point was homogenized in 1.25 ml of 2% lysis buffer with protease inhibitors (see above) with a Potter-Elvehjem homogenizer on STRUCTURE 1.
ice. The lysate was prepared as described in steady-state labeling and subjected to preclearing and immunoprecipitation in parallel with the conditioned medium, as described below.
Protein Normalization-When comparing multiple samples within the same experiment, the protein concentration of each detergent lysate was determined by a standard colorimetric protein assay (Pierce). The conditioned medium and detergent lysates were normalized for total protein based on the detergent lysate protein concentrations.
Immunoprecipitation and Peptide Elution-Lysate and conditioned medium samples were precleared by incubation with Protein A-Sepharose beads (50 l of packed beads/1.25 ml of lysate or conditioned medium) on a rocker overnight at 4°C. GlyCAM-1 was immunoprecipitated from lysate and conditioned medium with an anti-GlyCAM-1 rabbit polyclonal Ab (anti-peptide 2 Ab, also known as CAM02) directed to peptide 2 of GlyCAM-1 (CKEPSIFREELISKD, 17). Protein A-Sepharose was coated with preimmune or anti-peptide 2 Ab (25 l of serum/25 l of beads diluted to 1.0 ml with PBS) on a rocker overnight at 4°C (referred to as preimmune beads and anti-peptide 2 beads, respectively). The next day, lysate and conditioned medium were incubated with preimmune beads for a second preclearing step (25 l of packed beads/ 1.25 ml of lysate or conditioned medium) on a rocker for 3 h at 4°C. Precleared lysate and conditioned medium were incubated with antipeptide 2 beads (25 l of packed beads/1.25 ml of lysate or conditioned medium) on a rocker for 3 h at 4°C. The beads were washed three times in 2% lysis buffer and three times in 0.5% lysis buffer (PBS with 0.5% Triton X-100). The materials bound to the anti-peptide 2 beads were directly solubilized in Laemmli sample buffer or specifically eluted with 200 l of elution buffer (1 mg/ml peptide 2, 0.25% Triton X-100 in PBS with protease inhibitors) on a rocker for 3-4 h at 4°C. Samples solubilized in Laemmli sample buffer were directly analyzed by SDS-PAGE on 10% SDS-polyacrylamide gels under non-reducing conditions and subjected to fluorography with EN 3 HANCE (DuPont NEN). The beads were removed from peptide eluates by centrifugation, and the eluates were diluted to 400 l with 0.25% lysis buffer (PBS with 0.25% Triton X-100). GlyCAM-1 eluates were dialyzed against PBS (for direct reprecipitation, see below) or sialidase buffer (for sialidase treatment, see below) for 8 -16 h at 4°C and further analyzed by reprecipitation with a second rabbit polyclonal anti-GlyCAM-1 Ab, preimmune serum, lectins, or LEC-IgG (see reprecipitation procedures described below).

Treatment with Inhibitors and Enzymes
BFA Treatment-The diced lymph nodes were preincubated in 1.0 ml of complete medium with 0 -2.5 g/ml BFA (added from a 0.5 mg/ml stock in 100% MeOH) in wells of a 24-well plate (three lymph node equivalents/well) for 1 h at 37°C in 5% CO 2 prior to labeling. Six lymph node equivalents (2 wells) were used for each BFA concentration. After the preincubation, the lymph nodes were washed two times in Ser/Thrfree medium with BFA, low sulfate medium with BFA, or complete medium with BFA for [ 3 H]Ser/Thr, [ 35 S]sulfate, or [ 3 H]Gal labeling, respectively. The lymph nodes were then radiolabeled (as described for steady-state labeling) in the appropriate medium in the presence of BFA.
Chlorate Treatment-The diced lymph nodes were preincubated in 1.0 ml of low sulfate medium (1/10 normal sulfate) with 10 mM sodium chlorate for 1 h prior to labeling, as described for BFA treatment. For [ 3 H]Ser/Thr labeling, the lymph nodes were washed two times in low sulfate Ser/Thr-free medium with 10 mM chlorate. For [ 3 H]Gal labeling, the lymph nodes were washed two times in low sulfate medium with 10 mM chlorate. The lymph nodes were then radiolabeled as described for steady-state labeling in the appropriate low sulfate medium in the presence of 10 mM chlorate. In the experiment in which the lymph nodes were radiolabeled with [ 3 H]Ser/Thr in the presence of both chlorate and BFA, the washes and radiolabeling were conducted in low sulfate, Ser/Thr-free medium.
Sialidase Treatment-The lysate and conditioned medium eluates (400 l) from the anti-peptide 2 beads (see immunoprecipitation above) were dialyzed overnight against sialidase buffer (100 mM NaCl, 50 mM sodium acetate, 4 mM CaCl 2 , pH 6.5). The eluate pH was reduced to 5.5 with 10% (v/v) glacial acetic acid, and equal aliquots (200 l of each) were incubated with or without 0.05 unit of A. ureafaciens sialidase (10 units sialidase/ml) and 0.02 unit of V. cholerae sialidase (Oxford Glycosystems, Rosedale, NY; 2 units sialidase/ml) for 4 h at 37°C. In one experiment, the samples were further treated with exo-␤(134)galactosidase with or without fucosidase (see below). In all other cases, the samples were diluted to 1.0 ml with 0.25% Triton X-100 in PBS containing protease inhibitors, adjusted to pH 7.4 with 1 N NaOH and further analyzed by anti-peptide 3, preimmune serum, LEC-IgG, or lectins as described below.

Reprecipitation of GlyCAM-1
In most experiments, aliquots of GlyCAM-1, isolated by anti-peptide 2, were subjected to another round of immunoprecipitation with a second anti-GlyCAM-1 rabbit polyclonal Ab (anti-peptide 3 Ab, also known as CAM05) directed to peptide 3 (CIISGASRITKS) of GlyCAM-1 (17). Anti-peptide 3 Protein A-Sepharose (anti-peptide 3 beads) was prepared as described for anti-peptide 2 beads. Aliquots of GlyCAM-1 (typically, the eluate from 10 l of anti-peptide 2 beads, containing 1.2 mouse lymph node equivalents of GlyCAM-1), with or without prior desialylation and/or defucosylation, were diluted to 0.25 ml in O.25% lysis buffer (0.25% Triton X-100 in PBS) and incubated with 25 l of anti-peptide 3 beads or preimmune beads on a rocker for 3-16 h at 4°C. In many experiments, equal aliquots of GlyCAM-1 were diluted to 0.25 ml in 0.25% lysis buffer and incubated in parallel with lectins (VVA, PNA, MAA, Limax agglutinin, AAA, or SNA) or LEC-IgG immobilized on Sepharose or agarose beads (25 l of packed beads/sample; see above). After the incubation, the beads were centrifuged and washed three times with 0.25% lysis buffer. The material bound by anti-peptide 3 or preimmune serum beads was directly solubilized in Laemmli sample buffer for analysis by SDS-PAGE or 5% SDS for direct liquid scintillation counting. The material bound by lectins was eluted with 100 mM of the appropriate mono-or disaccharide competitor (GalNAc, VVA; lactose, PNA, MAA, and SNA; N-acetylneuraminic acid, Limax agglutinin; L-fucose, AAA) in 0.3 ml of 0.25% lysis buffer containing protease inhibitors on a rocker for 4 h at 4°C. The material bound by LEC-IgG was eluted in parallel with 10 mM EDTA in 0.3 ml of 0.25% lysis buffer containing protease inhibitors. Eluates were directly counted by liquid scintillation or precipitated with 4 volumes of acetone on ice for 30 min with 5 g of bovine serum albumin carrier protein, centrifuged at 10,000 ϫ g for 15 min at 4°C, and solubilized in Laemmli sample buffer for analysis by SDS-PAGE and fluorography. In some cases, aliquots of the anti-peptide 3, preimmune serum, lectin, and LEC-IgG precipitates prepared for SDS-PAGE were directly counted.

Identification of GlyCAM-1 Biosynthetic
Intermediates-In order to characterize intermediates along the GlyCAM-1 biosynthetic pathway, the protein core of GlyCAM-1 was radiolabeled with [ 3 H]Ser/Thr. Mouse lymph nodes were radiolabeled in organ culture, and GlyCAM-1 was isolated from the lymph node detergent lysate and conditioned medium with a rabbit polyclonal antibody (Ab) raised against peptide 2 from the deduced GlyCAM-1 protein core (17). Bound material was specifically eluted with peptide 2 and reprecipitated with a second anti-peptide Ab directed against the C terminus of GlyCAM-1 (peptide 3; 17). As shown in Fig. 1A (anti-peptide 3 lanes), multiple [ 3 H]Ser/Thr-labeled GlyCAM-1 proteins were isolated from the detergent lysate. The pattern of precipitated proteins was highly reproducible, consisting of a broad band at 40 -60 kDa and strongly labeled low molecular mass proteins migrating between 28 -33 kDa (28, 29, 31, 32, and 33 kDa proteins, collectively denoted by *). Faintly labeled proteins also were visible at 22-27 and 15 kDa, the latter of which closely approximates the predicted mass of the core protein. These proteins were all precipitable by an independent anti-peptide Ab against the N terminus of GlyCAM-1 (peptide 1, (17,31)). We also determined that individual electroeluted bands could be reprecipitated with anti-peptide 2 Ab (data not shown). The recognition of these discrete proteins by anti-peptide Abs raised against three independent deduced GlyCAM-1 peptides indicated that they all contained the GlyCAM-1 core protein.
When conditioned medium was analyzed as above (Fig. 1B, anti-peptide 3), a major broad band was seen at 40 -60 kDa while the low molecular mass proteins were detected only at trace levels.
Evaluation of O-Linked Oligosaccharides of GlyCAM-1-As observed previously (16,31), functional GlyCAM-1, defined as the subset that binds to the L-selectin-human IgG 1 chimera (LEC-IgG), migrated as a broad band at 50 -60 kDa (designated by #, Fig. 1, A and B). In order to characterize the oligosaccharide structures on the multiple forms of GlyCAM-1 precipitated by the anti-peptide Abs, we examined the binding of these proteins to a panel of lectins with defined carbohydrate specificities (Table I). VVA, which preferentially binds GalNAc␣13 Ser/Thr, was used to identify the first glycosylation step in the biosynthesis of O-linked oligosaccharides for mucin-type glycoproteins (reviewed in Ref. 45). PNA was used to detect the presence of the core Gal␤133GalNAc structure (referred to as T-antigen or core-1), commonly found on mucinlike proteins, and Limax agglutinin was used to identify sialylated species. AAA, MAA, and SNA recognize discrete features of the Sia␣233(SO 4 -6)Gal␤134[Fuc␣133]GlcNAc capping structure of GlyCAM-1 (27). These lectins require fucosylation, sialylation, and sulfation, respectively, for their binding (Table I).
For the lectin and LEC-IgG binding studies, [ 3 H]Ser/Thrlabeled GlyCAM-1, isolated from detergent lysates or conditioned medium with anti-peptide 2 Ab, was treated with or without sialidase. Equal aliquots were subsequently incubated with the lectin panel, anti-peptide 3 Ab, or LEC-IgG, all immobilized on agarose or Sepharose. The material bound was analyzed by SDS-PAGE (Fig. 1) and was quantified by scintillation counting (Table II). Fig. 1A presents the analysis of precipitates of GlyCAM-1 from the detergent lysates. We established that precipitation was efficient, since a second round of precipitation yielded only 10 -20% of the initial values (Table II). VVA precipitated proteins up to 45 kDa, including the discrete low molecular mass species migrating between 28 -33 kDa (denot-kDa proteins reprecipitated by VVA are shifted slightly upward in the gel by 1 kDa. The VVA and AAA precipitates from desialylated Gly-CAM-1 (data not shown) were identical to the those shown for untreated GlyCAM-1. Each lane contains intracellular (A) or secreted (B) radiolabeled GlyCAM-1 from the lymph nodes (axillary, brachial, cervical, and mesenteric) of one mouse. In both A and B, approximately 50% of the Limax agglutinin precipitate was lost prior to SDS-PAGE. In A, the exposure time of the anti-peptide 3 precipitates (lanes 2 and 3) was one-third less than the rest of the precipitates to adequately visualize the 28 -33 kDa proteins. Molecular mass markers were phosphorylase b (94,000), bovine serum albumin (67,000), ovalbumin (43,000), carbonic anhydrase (30,000), soybean trypsin inhibitor (20,000); df, dye front.  Table II) and subjected to analysis by SDS-PAGE on 10% gels under non-reducing conditions (band profiles did not differ with reduction), and fluorography using EN 3 HANCE. Discrete low molecular mass proteins migrating between 28 -33 kDa are denoted by *. Although the spacing is the same, 28 -33  (Table II), and the precipitated proteins ranged from 34 to 45 kDa (Fig. 1A). As shown in Table II, 24% of the total available GlyCAM-1 was precipitated by PNA after desialylation, confirming that the sialylated T-antigen (Sia␣233Gal␤133GalNAc) is a significant structure on Gly-CAM-1 (46). PNA did not bind the 28 -33 kDa proteins that were precipitated by VVA, suggesting that this cluster of proteins contained only GalNAc and may represent the earliest glycosylated precursors of GlyCAM-1. There are 38 potential sites for the initiation of O-linked glycosylation of GlyCAM-1 (17), and it is likely that the 28 -33 kDa species contained different numbers of GalNAc␣1-Ser/Thr modifications. The precipitation of higher molecular mass species (Ͼ33 kDa) by VVA suggested that non-extended GalNAc residues existed on these proteins. From both lysate and conditioned medium, Limax agglutinin, MAA, and AAA precipitated only those proteins migrating at 40 -60 kDa, thus identifying the sialylated and fucosylated forms of GlyCAM-1 (designated å in Fig. 1A). The 40 -60 kDa proteins also contained the (SO 4 -6)Gal␤134GlcNAc determinant as indicated by SNA binding. It has been previously shown that SNA recognition of GlyCAM-1 is inhibited by ␣233 sialylation of Gal in the SNA epitope (27). Without prior desialylation of GlyCAM-1, SNA weakly precipitated the 40 -60 kDa band ( Fig. 1A and B). With sialidase treatment, SNA binding increased by approximately 4-fold (Table II) and precipitated a broad band at 38 kDa (Fig. 1, A and B), representing the desialylated form of the 40 -60 kDa proteins. Thus, the Ͼ40 kDa proteins contained the sialic acid, fucose, and sulfate modifications found in the 6Ј sulfated, sLe x capping structure.
LEC-IgG precipitated a diffuse 50 -60 kDa band (designated #, Fig. 1, A and B), indicating that only the highest molecular mass subset of the 40 -60 kDa proteins possessed the features necessary for L-selectin binding. These results may indicate that the lower molecular weight forms of GlyCAM-1 contained incompletely processed oligosaccharides or an insufficient density of the mature oligosaccharide to support L-selectin binding. Secreted GlyCAM-1 was enriched in the more fully processed oligosaccharide structures (Fig. 1B, Table II). This conclusion is supported by the higher proportion of AAA, MAA, Limax agglutinin, SNA and LEC-IgG binding, and the lower proportion of VVA binding. The representation of the T-antigen remained constant, which is consistent with the formation of this core oligosaccharide structure at an early stage of the O-glycan biosynthesis.
Precursor-Product Relationship of Glycosylation Intermediates-Pulse-chase studies were undertaken to establish directly a precursor-product relationship among the putative GlyCAM-1 glycosylation intermediates seen in the above analysis. Lymph nodes were pulse-labeled with 3 [H]Ser/Thr for 5 min and then chased for various times in the presence of excess unlabeled serine and threonine. GlyCAM-1 was isolated from detergent lysates and conditioned medium with anti-peptide Abs as above and analyzed by SDS-PAGE (Fig. 2) and direct counting (Fig. 3). Fig. 2 shows the distribution of GlyCAM-1 proteins immunoprecipitated from the detergent lysate and conditioned medium at each time point. The 28 -33 kDa cluster of low molecular mass proteins were synthesized within 5 min. These proteins were converted to a molecular mass of 40 -50 kDa with a half-time of approximately 30 min. The 40 -50 kDa species were detected in the conditioned medium at half-maximal levels by approximately 75 min (quantified in Fig. 3). The mature 50 -60 kDa subset of GlyCAM-1, containing ligand activity for L-selectin, was first detected in the conditioned medium at 90 min in parallel with the broader 40 -50 kDa species. Only a trace level of the 50 -60 kDa protein was present in the lysate, suggesting that it was secreted soon after synthesis.
Effect of Reduced Temperature on Processing of GlyCAM-1 Precursors-Reduced temperature has been a useful tool for studying the intracellular organelles involved in the transport H]Ser/Thr-labeled GlyCAM-1 with or without prior desialylation were incubated with lectins, LEC-IgG or anti-peptide 3 antibody as described in Fig. 1 and under "Experimental Procedures." The percent of GlyCAM-1 bound was calculated as described in Fig. 5. The data are the mean of duplicates from two independent experiments (deviations were Ͻ5% of the mean). The percent of [ 3 H]Ser/Thr-labeled GlyCAM-1 recovered from the detergent lysates in a second round of precipitation were as follows: VVA (Ϫ), 1.2; PNA (ϩ), 1.6; MAA (Ϫ), 0.9; Limax agglutinin (Ϫ), 3.2; AAA (Ϫ), 1.5; SNA (ϩ), 1.8; LEC-IgG (Ϫ), 0.8. These values indicate that the recovery of GlyCAM-1 in the first round of precipitation was almost quantitative (80 -90%). Ϫ, indicates without sialidase treatment; ϩ, indicates with sialidase treatment.  The samples were normalized for total protein, and GlyCAM-1 was isolated by immunoprecipitation with anti-peptide 2 Ab. The material bound was eluted with peptide 2, dialyzed against PBS, and reprecipitated with preimmune serum and anti-peptide 3 Ab. Aliquots of each sample were counted in duplicate by liquid scintillation (see Fig. 3) and subjected to analysis by SDS-PAGE. The anti-peptide 3 Ab immunoprecipitates are shown; the preimmune serum immunoprecipitates were completely negative (data not shown). Each lane contains intracellular (lysates) or secreted (conditioned medium) pulse-labeled Gly-CAM-1 from the lymph nodes of 1.5 mice. of membrane and secretory proteins. Transport of viral glycoproteins from the TGN to the plasma membrane is blocked at 20°C (47,48), and transport from the rough endoplasmic reticulum to the Golgi apparatus is blocked at 15°C (48). Treatment at 15°C causes the accumulation of proteins in transitional elements between the endoplasmic reticulum and cis-Golgi (48,49).
We undertook temperature block studies in order to localize the GlyCAM-1 glycosylation intermediates to particular intracellular compartments. Lymph nodes were pulsed with [ 3 H]Ser/Thr for 5 min, as described for Fig. 2; however in this case, the chase was performed at 37, 20, or 15°C for both a short (20 min) and long interval (90 min). As shown in Fig. 4, at 37°C a small fraction of the 28 -33 kDa cluster was chased to 40 -60 kDa by 20 min. By 90 min, the 40 -60 kDa proteins increased in the lysate and accumulated in the conditioned medium, recapitulating what was demonstrated in Fig. 2. Reducing the temperature to 20°C significantly slowed the processing of the 28 -33 kDa proteins. At 20°C, the 40 -60 kDa proteins were completely absent at 20 min and substantially reduced at 90 min. Secretion of GlyCAM-1 was completely blocked. Furthermore, the level of the 50 -60 kDa protein within the lysate was increased at 20°C relative to 37°C at 90 min, suggesting that mature GlyCAM-1 was accumulating, presumably in the TGN. With further reduction of the temperature to 15°C, processing of the 28 -33 kDa proteins to 40 -60 kDa, as well as secretion, were completely blocked at both time points. The 28 -33 kDa proteins were generated within the 5-min pulse-labeling period at 37°C (see Fig. 2). The demonstration that these proteins were not further processed at 15°C strongly suggested that they acquired their additional molecular mass before being transported to the Golgi cisterna. The 28 -33 kDa proteins that accumulated at 15°C comigrated precisely with the GlyCAM-1 proteins precipitated by VVA in Fig. 1A. These results suggest that the initiation of O-linked glycosylation on the GlyCAM-1 core protein occurs in a pre-Golgi compartment.
Time Course of the Biosynthesis of Oligosaccharides of Gly-CAM-1-The pulse-chase studies were extended by evaluating the time course of the biosynthesis of the oligosaccharides of GlyCAM-1. Lymph nodes were pulse-labeled with [ 3 H]Ser/Thr and chased for various lengths of time (as in Fig. 2), and GlyCAM-1 was immunoprecipitated with anti-peptide 2 Ab. In this case, the material recovered from the lysate and conditioned medium were combined, and aliquots were reprecipi-tated with LEC-IgG, AAA, Limax agglutinin, SNA, or PNA. The data were plotted as the percent of total available Gly-CAM-1, as defined by antibody precipitation. As shown in the inset of Fig. 5, maximal binding of GlyCAM-1 to AAA and SNA was achieved at 180 min, with half-maximal binding at 55-60 min. Half-maximal binding was estimated at 55-60 min for both PNA and Limax and at 65-70 min for LEC-IgG (Fig. 5). These studies demonstrate a close temporal relationship between the acquisition of LEC-IgG binding activity and the fucosylation, sialylation, and sulfation of GlyCAM-1 oligosaccharides, but do not satisfactorily resolve the order of these modifications.
Brefeldin A Inhibits Sulfation of GlyCAM-1-We used BFA to gain further information on the order of oligosaccharide processing steps. BFA causes the cis-, medial-, and trans-Golgi cisternae to disassemble and fuse with the endoplasmic reticulum (reviewed in Ref. 50). By inducing the redistribution of resident processing enzymes from the Golgi stacks back to the endoplasmic reticulum and blocking membrane transport into the TGN (50), BFA provides the opportunity to distinguish oligosaccharide processing steps occurring in the combined ER/ Golgi compartment ("BFA compartment") from those taking place in the TGN. The effect of BFA on the sulfation of Gly-CAM-1 was examined by pretreating lymph nodes with BFA (0 -2.5 g/ml) for 1 h, followed by metabolic labeling with [ 3 H]Ser/Thr or [ 35 S]SO 4 in the presence of BFA. GlyCAM-1 was then isolated from the lymph node lysates and conditioned medium with anti-peptide 2 Ab and counted by liquid scintillation. The counts recovered at each BFA concentration were plotted as a percentage of the total counts recovered without BFA treatment. As shown in Fig. 6A, the level of intracellular [ 3 H]Ser/Thr-labeled GlyCAM-1 increased to 135% at the highest BFA concentration (2.5 g/ml), indicating that BFA did not prevent the synthesis of the GlyCAM-1 core protein. Secretion Ab. The material bound by these matrices was counted by liquid scintillation and computed as the fraction of total GlyCAM-1 (as determined by anti-peptide 3 Ab) at each BFA concentration to account for differences in available intracellular GlyCAM-1. The relative lectin reactivity (BFA versus non-BFA) was computed as the fraction of GlyCAM-1 precipitated by a given lectin in the BFA condition divided by the fraction of GlyCAM-1 that was precipitated by the lectin without BFA (Fig. 6B).The values are converted to percentages. Thus, a value of 50% for a particular BFA treatment indicates that BFA caused a net reduction of 50% in lectin reactivity. Precipitations with PNA and SNA were performed after sialidase treatment, whereas the deter-minations for AAA, MAA, Limax agglutinin, and LEC-IgG binding were made without prior sialidase treatment. As shown in Fig. 6B, the reactivity of GlyCAM-1 with SNA decreased by 99% at 2.5 g/ml BFA (Fig. 6B), paralleling the inhibition of [ 35 S]SO 4 incorporation demonstrated in Fig. 6A. The reactivity of GlyCAM-1 with AAA decreased by 50%, indicating that BFA had a partial inhibitory effect on overall fucosylation. In contrast, reactivity for PNA, MAA, and Limax agglutinin increased to 495, 335, and 195%, respectively. PNA binding was negligible without sialidase treatment (data not shown) demonstrating that the T-antigen was efficiently sialylated in the presence of BFA as it was in its absence (Fig. 6B FIG. 5. The maturation of O-linked oligosaccharides of Gly-CAM-1. Lymph nodes were pulse-labeled for 10 min with [ 3 H]Ser/Thr and chased for various times. Detergent lysates and conditioned medium were generated for each time point, normalized for total protein, and subjected to immunoprecipitation with anti-peptide 2 Ab and peptide elution. The GlyCAM-1 recovered from the detergent lysate and conditioned medium were combined for each time point, dialyzed against sialidase buffer, and treated with or without A. ureafaciens and V. cholerae sialidase. Equal aliquots of sialidase treated and untreated GlyCAM-1 were then reprecipitated with anti-peptide 3 Ab, PNA, Limax agglutinin, AAA, SNA, or LEC-IgG. The material bound was specifically eluted as described under "Experimental Procedures" and counted by liquid scintillation. The data are plotted as the percent of total GlyCAM-1 bound by the lectins and LEC-IgG over time. The percent of GlyCAM-1 bound was calculated by dividing the counts bound with the lectin or LEC-IgG by the counts bound with anti-peptide 3 Ab. For Limax agglutinin, AAA, and LEC-IgG, data are shown for precipitates of non-sialidase-treated GlyCAM-1, whereas for PNA and SNA, data are shown for precipitates of sialidase-treated GlyCAM-1. Data are derived from the mean of duplicate reprecipitations and have a standard deviation of less than 5% of the mean value. Inset shows the percent of GlyCAM-1 bound by AAA and SNA in an independent experiment with chase times up to 360 min. These data are the mean of reprecipitations performed in triplicate with a standard deviation of less than 1% of the mean value. and Table II). The increase in binding to Limax agglutinin with BFA was also consistent with this conclusion. The BFA-induced increases in PNA, MAA, and Limax agglutinin binding may reflect the increased residency of GlyCAM-1 intermediates in the fused ER/Golgi compartment, which would allow longer exposure to the relevant glycosyltransferases.
The enhanced MAA reactivity with BFA treatment may represent the combination of two factors. First, ␣233 sialylation of Gal␤134GlcNAc may be increased. Second, decreased sulfation of Sia␣233Gal␤134GlcNAc at C-6 of Gal enhances reactivity with MAA (27). To determine the relative contribution of these effects, a base line for maximal GlyCAM-1 binding to MAA was established using sodium chlorate, which inhibits GlyCAM-1 sulfation by 90% without interfering with sialylation (31). As shown in Table III, MAA precipitated 4.6% of the [ 3 H]Ser/Thr-labeled GlyCAM-1 from a detergent lysate of untreated lymph nodes. With chlorate treatment, MAA binding increased to 15.2%, while SNA binding decreased from 14.6 to 1.5%, reflecting the decrease in sulfation. When BFA was added in combination with chlorate, MAA still bound 10.4% of the available GlyCAM-1. Thus, with the sulfation effect controlled for, MAA recognition of GlyCAM-1 was not substantially diminished (15.2 versus 10.4%). As predicted from its binding specificity (Table I), the enhanced MAA binding seen with BFA, chlorate, or the two combined drugs was prevented by sialidase treatment of GlyCAM-1 (Table III).
LEC-IgG binding was completely inhibited by BFA treatment (Fig. 6B). The dramatic loss of LEC-IgG binding under this condition provides further evidence that sulfation is essential for recognition by L-selectin.
The Generation of BFA-induced Glycosylation Intermediates-The nearly complete inhibition of sulfation by BFA implied that BFA would cause the accumulation of distinct glycosylation intermediates. To identify such species, GlyCAM-1 was radiolabeled with [ 3 H]Ser/Thr in the presence or absence of BFA. Equal aliquots of GlyCAM-1 were then treated with or without sialidase, reprecipitated with anti-peptide 3 Ab or lectins as in Fig. 6B, and then analyzed by SDS-PAGE. As shown in Fig. 7 (anti-peptide 3 lane), BFA treatment dramatically altered the size distribution of the immunoprecipitated Gly-CAM-1 precursors. The relative amount of the 28 -33 kDa protein cluster was considerably reduced, and a new dominant 40 kDa protein (range is 35-46 kDa) and a faint 50 kDa species were generated (lane 3). These latter two species were sialylated as indicated by their binding to Limax agglutinin (lane 15); additionally, sialidase treatment reduced the new 40 and 50 kDa proteins by 5 kDa, producing a broad 35 kDa band and a much fainter 45 kDa band, respectively (lane 4). In addition, these two new species were fucosylated as shown by precipitation with AAA (data not shown). In accord with the results above, neither of the proteins were precipitated by LEC-IgG or SNA. Therefore, these proteins represented sialylated and fucosylated but non-sulfated biosynthetic intermediates of Gly-CAM-1. We are not certain what accounts for the difference in size between the 40 and 50 kDa BFA intermediates. Both of these proteins were precipitated by MAA (lane 11), and by PNA after sialidase treatment (lane 8), indicating that they contained sialylated N-acetyllactosamine (Sia␣233Gal␤134-GlcNAc), as well as sialylated T-antigen (Sia␣233Gal␤133-Gal). It is possible that differences exist in oligosaccharide branching or the number of elongated chains on these two species.
The Effect of BFA on the Generation of Gal␤134GlcNAc-Gal␤134GlcNAc (N-acetyllactosamine) is modified by sulfate, sialic acid, and fucose on the major oligosaccharide capping groups of GlyCAM-1, i.e. 6Ј-sulfo sLe x and 6-sulfo sLe x . This disaccharide is a component of the MAA epitope (Sia␣233Gal␤134GlcNAc). The ability of MAA to bind the BFA intermediates at a level similar to undersulfated Gly-CAM-1 (chlorate-generated) indicates that this structure was synthesized normally in the presence of BFA. As an independent approach for detecting this disaccharide, we analyzed the ability of Diplococcus exo-␤(134)galactosidase to release Gal from the BFA intermediates. This enzyme specifically hydrolyzes Gal in a ␤134 linkage to GlcNAc (51), an activity that is blocked by sulfation at the 6-position of Gal (38). Based on the knowledge that chlorate blocks the sulfation of GlyCAM-1 but does not diminish its galactosylation (38), we used chlorategenerated GlyCAM-1 as a positive control to establish the maximum possible release of Gal by the ␤(134)galactosidase. GlyCAM-1 was radiolabeled with [ 3 H]Gal in the presence of chlorate, BFA, or no inhibitor and isolated from lysates with anti-peptide 2 Ab. The ␣233 sialylation of Gal and the ␣133 fucosylation of GlcNAc within the capping structures of Gly-CAM-1 also inhibit Gal release by Diplococcus exo-␤(1-4)galactosidase (27). Accordingly, aliquots of the [ 3 H]Gal-labeled GlyCAM-1 were treated with or without sialidase and digested with the exo-␤(134)galactosidase with or without ␣(133/4)fucosidase treatment. As shown in Table IV, without chlorate or BFA treatment, exo-␤(134)galactosidase released 4% of [ 3 H]Gal from desialylated/defucosylated GlyCAM-1. With chlorate treatment, the amount of [ 3 H]Gal released from desialylated/defucosylated GlyCAM-1 increased to 15%, and this hydrolysis was completely dependent on prior sialidase treatment. With BFA treatment, exo-␤134 galactosidase released 12% of [ 3 H]Gal, indicating that ␤134-linked Gal was only slightly reduced (20%) in the presence of BFA. Moreover, this hydrolysis was also completely dependent on prior desialylation, confirming that Gal␤134GlcNAc was efficiently sialylated in the presence of BFA, in agreement with the MAA binding data shown in Fig. 6B and Table III. The effect of ␣(133/4)fucosidase on exo-␤(134)galactosidase release of Gal allowed an independent assessment of the degree of fucosylation of GlyCAM-1 in the presence of BFA. As shown in Table  IV, there was a substantial increase (7-12%) in the amount of Gal released by this enzyme when the GlyCAM-1 intermediates were treated with ␣(133/4)fucosidase plus sialidase as compared to sialidase alone. This result concurs with the AAA analysis (Fig. 6B) in establishing that a substantial degree of fucosylation occurred in the presence of BFA. DISCUSSION GlyCAM-1 is an HEV-derived, secreted ligand for L-selectin. Its functional role in lymphocyte-HEV binding has not been determined as yet. Nonetheless, detailed biochemical analysis of GlyCAM-1 is warranted, since it shares a carbohydratebased recognition determinant with the other known HEV ligands for L-selectin (23) and is associated with sites of chronic  (19). We have defined the following discrete stages in biosynthesis of GlyCAM-1: 1) unglycosylated species of Ͻ28 kDa; 2) discrete 28 -33 kDa proteins containing GalNAc-terminating chains; 3) a broadly migrating 40 -50 kDa species containing the T-antigen, sialic acid, fucose, and sulfate, but not reactive with LEC-IgG; and 4) a 50 -60 kDa sialylated, fucosylated, and sulfated protein, reactive with LEC-IgG.
By pulse-chase analysis, we established that the first three groups of proteins were biosynthetic intermediates of mature GlyCAM-1. The low molecular mass proteins (28 -33 kDa) were synthesized within 5 min and processed to 40 -50 kDa with a half-time of approximately 30 min. The half-time for the acquisition of L-selectin binding was Ϸ65 min. An unexpected finding was that up to 75% of the 40 -50 kDa species was secreted into the medium without attaining the capacity to bind L-selectin.
Using temperature blocks, we have gained information about the initiation of O-glycosylation in GlyCAM-1. A reduction of the temperature to 15°C during the chase period completely blocked the processing of the rapidly synthesized 28 -33 kDa cluster. In multiple cell types, membrane transport into the Golgi stacks is blocked at 15°C, and glycoproteins accumulate in pre-Golgi transitional elements of the endoplasmic reticulum (48,49,(52)(53)(54). Since the 28 -33 kDa proteins contain GalNAc-terminating chains, our findings indicate that the initiation of glycosylation on GlyCAM-1 occurs in a pre-Golgi compartment. In some systems, the addition of GalNAc to nascent proteins occurs in the endoplasmic reticulum or transitional elements of the endoplasmic reticulum (55)(56)(57), whereas in others, initiation appears to take place in the Golgi apparatus (58 -60). Thus, the site for the initiation of O-linked glycosylation appears to differ for different cell types and perhaps for different core proteins.
The metabolic inhibitor BFA is a valuable experimental tool that permits discrimination of processing events in the ER/ Golgi compartment from those in the TGN. We employed this drug to dissect the terminal processing events for GlyCAM-1, which could not be adequately resolved by pulse-chase analysis. As expected, BFA completely blocked the secretion of Gly-CAM-1 into conditioned medium. BFA caused the accumulation of biosynthetic intermediates of GlyCAM-1 with oligosaccharides that were efficiently sialylated, partially fucosylated, and almost completely lacking in sulfate.
The epitope for PNA (the T-antigen) increased approximately 5-fold with BFA treatment, and it was fully sialylated since prior desialylation was required for binding. The increased level of sialylated T-antigen with BFA may have been due to the increased contact of GlyCAM-1 with the appropriate glycosyltransferases in the BFA-induced compartment. BFA treatment also allowed the formation of N-acetyllactosamine (Gal␤134GlcNAc) within GlyCAM-1 and its efficient sialylation. Thus with BFA, the total amount of [ 3 H]Gal released by Diplococcus exo-(␤134)galactosidase decreased only marginally relative to the control. Additionally, the release of [ 3 H]Gal completely depended upon the prior desialylation of GlyCAM-1, consistent with a fully sialylated state of the terminal Gal␤134 residues. Finally, MAA reacted with BFA-generated GlyCAM-1 comparably to chlorate-generated GlyCAM-1, indicating the formation of Sia␣233Gal␤134GlcNAc.
Taken together, these results argue that the sialyltransferases that form the Sia␣233Gal␤133GalNAc and Sia␣233Gal␤134GlcNAc structures are localized in a pre-TGN compartment. This conclusion is consistent with previous studies on several glycoproteins in which BFA treatment does not impede sialylation of O-linked oligosaccharides (61-63). The ␤134 galactosyltransferase involved in the formation of  N-acetyllactosamine is localized to the trans-Golgi cisternae in a number of cells (64,65). Thus, the Gal␤134GlcNAc ␣233 sialyltransferase pertinent to GlyCAM-1 is likely to reside in the trans-Golgi cisternae, in distinction to the apparent TGN localization of sialyltransferases that act on N-linked oligosaccharides (63,66,67). Our data cannot distinguish the subcellular localization of the Gal␤133GalNAc ␣233 sialyltransferase relative to the Gal␤134GlcNAc ␣233 sialyltransferase. However, the T-antigen-specific ␣233 sialyltransferase involved in the synthesis of another sialomucin has been mapped to a compartment proximal to the trans-Golgi cisternae (65). Fucose is added in an ␣133 linkage to GlcNAc in the N-acetyllactosamine of GlyCAM-1. BFA inhibited fucosylation of GlyCAM-1 by Ϸ50% as determined by direct precipitation with AAA and by the defucosylation requirement for exo-(␤134)galactosidase action. Thus, the accessibility of nascent oligosaccharides to the fucosyltransferase was clearly affected by BFA. In contrast to the sialyltransferases, the fucosyltransferase appears to reside in a compartment that was partially redistributed by BFA. Given the apparent greater efficiency in the redistribution of the sialyltransferases, the fucosyltransferase is likely to reside in a more distal region of the biosynthetic pathway. This conclusion is consistent with the general finding that ␣233 sialylation precedes ␣133 fucosylation during the synthesis of sLe x (68,69).
Gal-6-sulfate and GlcNAc-6-sulfate occur equally in Gly-CAM-1 (37). In the presence of BFA, the sulfation of GlyCAM-1 was almost completely suppressed, as demonstrated by the 85% reduction in 35 [S]sulfate labeling of GlyCAM-1 and the 99% reduction in SNA binding. Since BFA allowed the synthesis of the Gal␤134GlcNAc structure, the inhibition of sulfation is likely attributable to the inaccessibility of GlyCAM-1 biosynthetic intermediates to the TGN where the relevant sulfotransferases reside. In a number of other systems, BFA has been employed to reach the same conclusion about the subcellular localization of sulfotransferases that modify O-glycans (70 -73).
Taken together, our biosynthetic analysis of GlyCAM-1 argues that the sialylation events precede both fucosylation and sulfation. As noted above, the ordering of sialylation versus fucosylation is consistent with previous studies. The relationship of fucosylation to sulfation is more problematic. BFA produced partial inhibition of fucosylation (ϳ50%) and almost complete inhibition of sulfation, which would argue for fucosylation occurring before the two sulfation modifications. In support of this possibility, Jain et al. (74) have reported that several of the known ␣133/4 fucosyltransferases are unable to fucosylate Sia233(SO 4 -6)Gal␤134GlcNAc to form the 6Ј-sulfo sLe x capping structure, whereas these enzymes are active on the non-sulfated structures. However, Scudder et al. (75) have reported that a lymph node N-acetylglucosamine-6-O-sulfotransferase is unable to add sulfate to GlcNAc-containing oligosaccharides if the C-3 position of GlcNAc is substituted with fucose, arguing that sulfation cannot precede fucosylation on this sugar. Clearly, further studies are necessary to define the temporal relationship of the two sulfation events to fucosylation during the biosynthesis of GlyCAM-1.
The present study has identified glycosylation intermediates of GlyCAM-1 as it is synthesized in mouse peripheral lymph nodes. Our analysis helps to elucidate acceptor structures for the endothelial enzymes that form the ligand. An important future challenge is to determine the molecular identity of these enzymes and to understand how their activities are regulated in lymphoid organs and at sites of inflammation.