Oversulfated Chondroitin/Dermatan Sulfates Containing GlcAβ1/IdoAα1–3GalNAc(4,6-O-disulfate) Interact with L- and P-selectin and Chemokines*

We previously reported that versican, a large chondroitin/dermatan sulfate (CS/DS) proteoglycan, interacts through its CS/DS chains with adhesion molecules L- and P-selectin and CD44, as well as chemokines. Here, we have characterized these interactions further. Using a metabolic inhibitor of sulfation, sodium chlorate, we show that the interactions of the CS/DS chains of versican with L- and P-selectin and chemokines are sulfation-dependent but the interaction with CD44 is sulfation-independent. Consistently, versican's binding to L- and P-selectin and chemokines is specifically inhibited by oversulfated CS/DS chains containing GlcAβ1–3GalNAc(4,6-O-disulfate) or IdoAα1–3GalNAc(4,6-O-disulfate), but its binding to CD44 is inhibited by all the CS/DS chains, including low-sulfated and unsulfated ones. Affinity and kinetic analyses using surface plasmon resonance revealed that the oversulfated CS/DS chains containing GlcAβ1/IdoAα1–3GalNAc(4,6-O-disulfate) bind directly to selectins and chemokines with high affinity (K d 21.1 to 293 nm). In addition, a tetrasaccharide fragment of repeating GlcAβ1–3GalNAc(4,6-O-disulfate) units directly interacts with L- and P-selectin and chemokines and oversulfated CS/DS chains containing GlcAβ1/IdoAα1–3GalNAc(4,6-O-disulfate) inhibit chemokine-induced Ca2+ mobilization. Taken together, our results show that oversulfated CS/DS chains containing GlcAβ1/IdoAα1–3GalNAc(4,6-O-disulfate) are recognized by L- and P-selectin and chemokines, and imply that these chains are important in selectin- and/or chemokine-mediated cellular responses.

Proteoglycans are ubiquitous components of cell surface membranes, basement membranes, and extracellular matrices in various tissues. They belong to a family of macromolecules that consist of core proteins to which glycosaminoglycans (GAGs), 1 sulfated polysaccharides, are attached. GAGs are lin-ear polysaccharides made up of disaccharide units composed of hexosamine and hexuronic acid (or hexose). They are classified into chondroitin sulfate (CS), dermatan sulfate (DS), heparin, heparan sulfate (HS), keratan sulfate (KS), and hyaluronic acid (HA). Because of the high sulfate and carboxyl group content of their GAG moieties, proteoglycans have strong negative charges. This property allows them to interact with a wide range of proteins, including growth factors, enzymes, cytokines, chemokines, lipoproteins, and adhesion molecules (1,2).
We previously showed that a large CS/DS proteoglycan, versican (also called PG-M), that was derived from a renal adenocarcinoma cell line, ACHN, interacts through its CS/DS chains with adhesion molecules such as L-and P-selectin and CD44 (3,4), and various chemokines (5), all of which have been implicated in leukocyte trafficking. Versican possesses a hyaluronic acid-binding domain at its N terminus, a GAG attachment domain in the middle, and a set of epidermal growth factor-like, C-type lectin-like, and complement regulatory protein-like domains at its C terminus (6). Alternative splicing of the versican gene generates four versican isoforms: V0, V1, and V2, which bear GAG attachment domains of different lengths, and V3, which is without a GAG attachment domain (7). Versican is widely expressed in many tissues, including the kidney, skin, brain, and aorta (8). Our previous studies indicated that versican from certain cell types, such as ACHN and Vero cells, but not that from 293T cells or human skin fibroblasts, interacts with L-selectin, suggesting that at least one glycoform of versican species is reactive with L-selectin (3). We have also shown that a subset of GAGs, including CS B from pig skin, CS E from squid cartilage, and HS from bovine kidney, interacts with Land P-selectin and chemokines and inhibits the interaction with versican (4,5). Thus, it seems likely that a nonspecific electrostatic interaction is not the sole factor determining the interaction, and that a specific carbohydrate structure is rec-* This work was supported in part by grants-in-aid for the Center of Excellence Research and for Scientific Research on Priority Areas, Sugar Remodeling and Cellular Communications from the Ministry of Education, Science and Culture, Japan, the Science Research Promotion Fund of the Japan Private School Promotion Foundation, and Taisyo Pharmaceutical Co. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
In the present study, we have extended our previous work to examine the structural requirement for the interaction of CS/DS chains with adhesion molecules and chemokines. We report here affinity and kinetic parameters for the interaction of various GAGs with L-and P-selectin and chemokines. We also provide evidence that a tetrasaccharide fragment composed of two GlcA␤1-3GalNAc(4,6-O-disulfate) units directly interacts with L-and P-selectin and chemokines.
Production of Monomeric CD44 -CD44 cDNA was produced by PCR using the sense primer 5Ј-TTTAAGCTTATGGACAAGTTTTGGTG-GCAC-3Ј, which contained a HindIII restriction site (in bold) and the codons of the first seven amino acids of human CD44 (underlined), and the antisense primer 5Ј-TTTTCTAGAAACACGTCATCATCAG-TAGGGTT-3Ј, which contained an XbaI restriction site (in bold) and the codons for amino acids 172-178 (underlined), inserted into expression vector pcDNA3.1/Myc-His(ϩ)B (Invitrogen), and sequenced. Monomeric CD44 was produced by transfecting the expression vector into 293T cells with LipofectAMINE reagent (Invitrogen) and culturing the cells for 4 days, followed by immunoaffinity chromatography of the conditioned medium using an anti-CD44 mAb BRIC 235-conjugated CNBractivated-Sepharose 4B. The purity of the soluble CD44 was more than 95% as assessed by SDS-PAGE and silver staining. Immunoprecipitation-Immunoprecipitation was performed as described previously (3), except that the beads were washed with buffer A (0.05% Tween 20, 20 mM HEPES-NaOH, 0.15 M NaCl, 1 mM CaCl 2 , 1 mM MgCl 2 , pH 6.8) in this study.
Enzyme Treatment and Derivatization of the Disaccharide Products with 2-AB-Versican-derived GAGs were incubated with chondroitinase ABC (0.38 units/ml) in combination with or without chondro-6sulfatase (0.31 units/ml) in 3% acetic acid adjusted to pH 7.0 with triethylamine (buffer B) at 37°C for 1 h. After evaporation, the derivatization of the disaccharide products with 2-AB (2-aminobenzamide) was performed according to the method of Kinoshita and Sugahara (14). They were then purified by partition between chloroform and distilled water (1:1). The aqueous phase, containing the derivatized disaccharides, was analyzed by high performance liquid chromatography (HPLC) as described previously (15). Eluates were monitored using a fluorescence detector with excitation and emission wavelengths of 330 and 420 nm, respectively.
Preparation of the Conditioned Medium of ACHN Cells-ACHN cells were cultured for 2 days in the presence or absence of 100 mM sodium chlorate in RPMI 1640 containing 10% fetal calf serum. After the condi- In lanes 19 and 20, the conditioned medium was treated with hyaluronidase SD (50 munits/ml) before the incubation. The precipitates were subjected to SDS-agarose-PAGE and analyzed as described under "Experimental Procedures." IP, immunoprecipitation. tioned medium was removed, the cells were further cultured in a serumfree medium, EX-CELL 610 HSF (JRH Bioscience, Lenexa, KS), in the presence or absence of 100 mM sodium chlorate for 4 days. The conditioned medium was collected and spun at 10,000 ϫ g for 15 min at 4°C.
Sandwich ELISA-Wells of 96-well flat-bottomed microtiter plates (Costar EIA/RIA plate number 3690, Corning Inc., Corning, NY) coated with BSA, anti-versican mAb 2B1, Ig chimeras or chemokines, and blocked with 3% BSA in phosphate-buffered saline were incubated for 1 h with conditioned medium from ACHN cells treated with sodium chlorate or untreated. After washing the wells with buffer A, 1 g/ml biotinylated anti-D antibody was added and incubated for 1 h. After washing, the binding was detected with alkaline phosphatase-conjugated streptavidin and Blue Phos TM substrate as described previously (4).
Kinetic Analysis Using BIAcore-This experiment was performed on a BIAcore TM biosensor (BIAcore AB, Uppsala, Sweden). All experiments were performed at 25°C. The running buffer, which was used for the washing and dissociation phase, was buffer C (20 mM HEPES-NaOH, 0.15 M NaCl, 1 mM CaCl 2 , 1 mM MgCl 2 , pH 6.8). In the first step, about 1.8 -2.0 kilo resonance units (1 kilo resonance units ϭ 1 ng/mm 2 ) streptavidin was covalently immobilized on the B1 sensor chip via its primary amine groups, using an amine coupling kit (Amersham Bioscience) according to the instructions provided by the manufacturer. The remaining activated groups were blocked with 150 l of 1 M ethanolamine-HCl, pH 8.5. Each GAG that had been biotinylated at the reducing end with EZ-link TM biotin-LC-hydrazide (Pierce) according to the method of Sadir et al. (16) was then injected over the sensorchip surface to obtain an immobilization level of about 150 resonance units. For the binding assays, different concentrations of monomeric selectins, monomeric CD44, or chemokines were injected at 30 l/min. The sensorchip surface was regenerated with 300 l of 1 M NaCl for chemokines and CD44, or with 300 l of 1 M NaCl followed by 100 l of 50 mM EDTA, pH 8.0, for selectins. No significant change in the baseline was observed after surface regeneration. Affinity kinetic parameters were determined with the BIAevaluation 3.0 software (Pharmacia Biosensor) using a single-site binding model.

Preparation of Streptavidin-conjugated Alkaline Phosphatase Coupled to Biotinylated Oligosaccharide Fragments from CS A, CS C, and
CS E-CS E from squid cartilage (1 mg) was suspended in 50 mM sodium acetate, 133 mM NaCl, and 0.04% gelatin, pH 5.0, containing 0.6 mg (1,800 units) of sheep testicular hyaluronidase (Sigma), and incubated at 37°C for a total of 68.5 h. An additional 2 mg (6,000 units) of the enzyme was added at time 24 and 45.5 h. CS A from whale cartilage and CS C from shark cartilage were suspended in 50 mM sodium acetate, pH 5.0, containing 0.6 mg of sheep testicular hyaluronidase, and incubated at 37°C for 24 h. The digests were fractionated by HPLC on an amine-bound Silica PA-03 column with a linear gradient from 16 mM to 1 M NaH 2 PO 4 . Each fraction was applied to a Sephadex G-25 column (1 ϫ 30 cm, Amersham Bioscience) equilibrated with distilled water, and the fractions containing oligosaccharides were collected and evaporated to dryness. To each fraction containing 20 -80 g of oligosaccharides, 125 mM EZ-link TM biotin-LC-hydrazide and 1 M NaCNBH 3 in dimethyl sulfoxide/acetic acid (7:3) was added, and the reaction mixture was incubated for 3 h at 65°C and then for 12.5 to 18.5 h at 37°C. To remove unreacted biotin-LC-hydrazide, each reaction mixture was applied to a Sephadex G-25 column as described above and the fractions containing biotinylated oligosaccharides were collected and evaporated to dryness. Under these conditions, about 100% of the oligosaccharides appeared to be biotinylated, because nonreducing terminus-derived disaccharides but not reducing terminus-derived unsaturated disaccharides were detected after the chondroitinase ACII digestion of a portion of the biotinylated oligosaccharide fractions, followed by 2-AB derivatization and HPLC analysis (data not shown). A portion of the fraction containing 16 pmol of each biotinylated oligosaccharide was dissolved in 0.5 ml of buffer A and incubated overnight with 1 l (2 pmol) of streptavidin-alkaline phosphatase (Promega) at 4°C. The reaction mixture was diluted 3 times with buffer A and applied to wells of 96-well flat-bottomed microtiter plates (Costar EIA/RIA plate number 3690) coated with selectin-or CD44-Igs or chemokines. Binding was detected with the Blue Phos TM substrate as described previously (4). After incubation, the plate was washed and the binding was determined by enzyme-linked immunosorbent assay as described previously (4).

RESULTS
Sulfation-dependent Interaction of Versican with L-and Pselectin-We previously reported that a CS/DS proteoglycan, versican, which was derived from ACHN cells, binds L-and P-selectin and CD44 in a manner fully dependent on versican's CS/DS chains (4). To determine the sulfation requirement for the interaction of the CS/DS chains of versican with L-and P-selectin and CD44, we first treated ACHN cells with sodium chlorate, a metabolic inhibitor of sulfation. This treatment resulted in the inhibition of more than 90% of sulfation ( Fig. 1, lanes 1-4) but allowed the core protein of versican to be synthesized (lanes [5][6][7][8]. This treatment inhibited the interaction of versican with L-and P-selectin-Igs (lanes 11, 12, 15, and 16), suggesting that sulfation of the CS/DS chains of versican is required for the interaction with L-and P-selectin. The unsulfated versican did not interact with E-selectin-Ig as described previously (lanes 13 and 14, Ref. 4).
In contrast, the interaction between versican and the CD44-Ig was not inhibited by the sodium chlorate treatment (lanes 17 and 18). It could be argued, however, that this was due to the formation of a trimolecular complex that consisted of versican, HA, and CD44-Ig; i.e. versican bound HA to which CD44 also bound, and neither the versican-HA nor the CD44-HA interaction required sulfation, while the versican-CD44 interaction did require it. We thus treated the [ 35 S]methionine-labeled culture supernatant of ACHN cells with hyaluronidase for 3 h as described previously (4), and then incubated it with CD44-Ig. Although the HA was almost completely digested by this treatment (data not shown), the interaction of versican with CD44-Ig was not affected at all (lanes 19 and 20), confirming the idea that the sulfation of versican is not required for the interaction with CD44; i.e. the CS/DS chains of versican interact with CD44 in a sulfation-independent manner.
Effects of Various CS/DS Chains on the Binding of Versican to L-and P-selectin and CD44 -We next performed an inhibition assay using various CS/DS chains to characterize the structure that interacts with selectins and CD44 (Fig. 2). Table  I shows the disaccharide composition of the CS/DS chains used in this study. CPS was produced by the selective 6-O-sulfation (12) of CS A from whale cartilage, and the chondroitin (CH) was produced by chemical desulfation (13) of CS A. Gel filtration analyses of CPS (12) and CH (13) showed that these reactions took place with a minimal degree of cleavage of the glycosidic linkages and that these GAGs did not aggregate in the solution, excluding the possibility that these GAGs function in a multivalent manner. Thus, the chain length and valency of CPS, CS A, and CH appeared to be the same, whereas the extent of sulfation of these GAGs was different (Table I). Similarly, the DPS, DS, and dermatan differed only in their sulfation. As shown in Fig. 2, the binding of biotinylated versican to the Land P-selectin-Igs was inhibited dose dependently by CPS, DPS, and CS E, all of which contain GlcA␤1/IdoA␣1-3Gal-NAc(4,6-O-disulfate) as a major disaccharide component, but not by low or unsulfated CS/DS chains such as CH, CS A, dermatan and DS, or KS. In contrast, the binding of biotinylated versican to CD44-Ig was inhibited by all the CS/DS chains examined, including low-sulfated and unsulfated ones, but not by KS. These results are consistent with the above results that sulfation plays a critical role in the interaction of the CS/DS chains of versican with L-and P-selectin but not with CD44.
Structural Analysis of the GAG Moieties of Versican-Our previous results showed that versican from ACHN cells contains at least CS B and CS C (4). To characterize the GAG structure of versican in more detail, we next performed a highly sensitive disaccharide composition analysis of versican-derived GAGs using a fluorescence substance, 2-AB (14). As shown in Fig. 3A, five peaks corresponding to the elution positions of standard disaccharides were detected after treating versican's GAGs with chondroitinase ABC. The addition of chondro-6sulfatase with chondroitinase ABC yielded three peaks corresponding to ⌬Di-0S, ⌬Di-UA-2S, and ⌬Di-4S (Fig. 3B), confirming that the five peaks detected in Fig. 3A are ⌬Di-0S, ⌬Di-6S, ⌬Di-4S, ⌬Di-di(2,6)S, and ⌬Di-di(4,6)S. Peak areas of ⌬Di-0S, ⌬Di-6S, ⌬Di-4S, ⌬Di-di(2,6)S, and ⌬Di-di(4,6)S in Fig. 3A were 0.8, 15.7, 77.6, 1.4, and 4.5%, respectively. Similar results were obtained using chondroitinase ACII instead of chondroitinase ABC, although minor additional peaks (6.3%) that did not correspond to the disaccharide standards were detected (data not shown). These results collectively indicate that versican's GAG contains GlcA␤1-3GalNAc(4,6-O-disulfate) and is a heteropolymer composed of a mixture of the major CS and the minor DS chains, the latter of which are resistant to chondroitinase ACII.
Interaction of Versican with Chemokines Is Also Sulfationdependent and Inhibited by Oversulfated CS/DS Chains-We previously showed that versican can interact with not only adhesion molecules but also certain chemokines (5). To explore the sulfation requirement for the interaction of versican with chemokines, we first generated undersulfated versican by culturing ACHN cells with 100 mM sodium chlorate. Under these conditions, only ⌬Di-0S was detected as a major peak (82.8%) in versican GAGs, while ⌬Di-6S (7.2%) and ⌬Di-4S (10.0%) were detected as minor peaks, and no ⌬Di-di(2,6)S or ⌬Didi(4,6)S was detected (Fig. 4A). Thus, the sodium chlorate treatment yielded undersulfated versican bearing mainly unsulfated CS/DS chains.
We then examined the binding of intact and undersulfated versican to chemokines. Control experiments showed that both intact and undersulfated versican bound anti-versican mAb, 2B1, and CD44-Ig, whereas only the intact versican bound Land P-selectin-Igs (Fig. 4B), consistent with the results shown in Fig. 1. Under these conditions, intact but not undersulfated versican significantly bound chemokines such as SLC, ␥-interferon inducible protein-10 (IP-10), platelet factor 4 (PF4), and SDF-1␤, suggesting that the sulfation of the CS/DS chains of versican is required for the interaction with chemokines. Neither form of versican bound much, if at all, recombinant trun- cated SLC (SLC-T) that lacked the C-terminal 32 amino acids containing basic amino acid clusters. In addition, neither form bound SDF-1␣ that naturally lacks the C-terminal 4 amino acids of SDF-1␤. These results suggest that the CS/DS chains of versican interact with the C-terminal region of SLC and SDF-1␤. We next examined the effects of oversulfated CS/DS chains on the binding of versican to chemokines. As shown in Fig. 5, the binding of biotinylated versican to SLC, IP-10, PF-4, and SDF-1␤, and to L-and P-selectin was inhibited by oversulfated CS/DS chains such as CPS, DPS, and CS E to a similar extent. As shown above, the binding of biotinylated versican to CD44-Ig was inhibited by CH, CS A, DPS, and CS E, although the effect of CPS was not significant in the dose range used in this series of experiments. The binding of versican to anti-versican mAb 2B1 was not affected by the addition of any of the GAGs examined. Collec-tively, these results suggest that sulfation of the CS/DS chains is required for the interaction with chemokines and that oversulfated CS/DS chains containing GlcA␤1/IdoA␣1-3GalNAc(4,6-Odisulfate) interact with chemokines.

Affinity Kinetics of the Interactions of Oversulfated CS/DS Chains with L-and P-selectin and
Chemokines-Using surface plasmon resonance, we next examined the affinity kinetics of the interactions of various CS/DS chains with L-and P-selectin, CD44, and chemokines. As shown in Fig. 6, SLC, IP-10, SDF-1␤, and monomeric L-and P-selectin bound dose-dependently to CS E that was immobilized on the sensorchip surface. Similar affinity kinetics was observed when CPS or DPS was used instead of CS E. In contrast, selectins and chemokines failed to interact with CS A or CH. These results were consistent with the inhibition assay described above (Figs. 2 and 5). Evaluation of the affinity kinetic parameters (Table II) indicated that L-and P-selectin as well as chemokines interacted with oversulfated CS/DS chains containing GlcA␤1/IdoA␣1-3GalNAc(4,6-O-disulafide) with high affinity (K d 21.1 to 293 nM). In contrast, the interaction of CD44 with these GAGs was quite different from the interactions of selectins and chemokines. Monomeric CD44 bound to CS A and CH with low affinity (K d 85.2 to 129 M). Monomeric CD44 also bound weakly to CS E (K d 211 M), and it failed to bind to CPS or DPS. Monomeric E-selectin, SLC-T, and SDF-1␣ did not interact much if at all with any of the CS/DS chains examined (data not shown).
Having obtained these results, we next examined whether these oligosaccharides could interact with selectins, CD44, and chemokines. To this end, we first biotinylated each oligosaccharide at the reducing terminus with biotin-LC-hydrazide, then used the biotinylated oligosaccharide to form a complex with streptavidin-conjugated alkaline phosphatase. We subsequently examined their binding to L-and P-selectin-Igs, CD44-Ig, or chemokines. Interestingly, only the e-2 fraction but no others bound to L-and P-selectin-Igs, SLC, IP-10, and SDF-1␤ (Fig. 7D). The e-1 fraction bound to PF4 moderately but not to other chemokines or L-and P-selectins. These results indicate that the repeating GlcA␤1-3GalNAc(4,6-O-disulfate) units are specifically recognized by L-and P-selectins as well as the majority of the chemokines examined and that a single GlcA␤1-3GalNAc(4,6-O-disulfate) is probably sufficient for the interaction with PF4.

CS E Inhibits Ca 2ϩ Responses Induced by SLC but Not by SLC-T-We previously showed that versican inhibits integrin activation and intracellular Ca 2ϩ mobilization induced by a chemokine, SLC (5).
To investigate whether oversulfated CS/DS chains also inhibit chemokine activity, we next examined CS E's ability to affect the Ca 2ϩ mobilization induced by SLC. As shown in Fig. 8, whereas SLC or SLC that had been preincubated with CS A induced significant Ca 2ϩ mobilization in L1.2 cells transfected with a receptor for SLC, CCR7, SLC that had been preincubated with CS E failed to induce Ca 2ϩ mobilization. Similarly, SLC that had been preincubated with CPS or DPS did not induce Ca 2ϩ mobilization (data not shown). On the other hand, the Ca 2ϩ mobilization induced by SLC-T was not affected by any of these oversulfated CS/DS chains. These results suggest that the oversulfated CS/DS chains inhibit the biological activity of SLC by interacting with the C-terminal region of SLC. DISCUSSION In this study, having demonstrated that sulfation is essential for the interaction of CS/DS chains with L-and P-selectin and chemokines, we analyzed the carbohydrate structures that bind to these molecules and have shown that oversulfated FIG. 5. Binding of biotinylated versican to chemokines is inhibited by oversulfated CS/DS chains. Biotinylated versican (0.25 g/ml) was added to wells coated with BSA (5 g/ml), anti-versican mAb 2B1 (5 g/ml), CD44-Ig (1 g/ml), L-selectin-Ig (L-Ig, 3 g/ml), P-selectin-Ig (P-Ig, 3 g/ml), or chemokines (1 g/ml) in the presence or absence (gray) of 100 g/ml GAGs (blue, CH; green, CS A; yellow, CPS; orange, DPS; pink, CS E). The binding was determined by enzymelinked immunosorbent assay as described previously (4). Each bar represents the mean Ϯ S.D. of triplicate determinations.

FIG. 4. Sulfation of versican is required for its interaction with chemokines.
A, disaccharide composition analysis. The sodium chlorate-treated or untreated conditioned medium was incubated with 20 turbidity reducing units/ml hyaluronidase (Streptomyces hyalurolyticus) at 37°C for 4.5 h. Thereafter, versican was precipitated from each conditioned medium with protein A-Sepharose beads (10 l beads) coupled to 10 g of anti-D antibody. After washing, the beads were incubated in the presence of 1 unit/ml chondroitinase ABC and 1 unit/ml chondroitinase ACII at 37°C for 2 h. The disaccharide fragments were derivatized with 2-AB and analyzed by HPLC. The elution positions of 2-AB-derivatized standard disaccharides shown are as described in the legend for Fig. 3. B, sandwich enzyme-linked immunosorbent assay. Binding of versican from untreated (solid bars) or sodium chlorate-treated (striped bars) conditioned medium to wells coated with BSA (6 g/ml), anti-versican monoclonal antibody 2B1 (3 g/ml), L-selectin-Ig (L-Ig, 3 g/ml), E-selectin-Ig (E-Ig, 3 g/ml), P-selectin-Ig (P-Ig, 3 g/ml), CD44-Ig (3 g/ml), SLC (3 g/ml), SLC-T (3 g/ml), IP-10 (3 g/ml), PF4 (6 g/ml), SDF-1␤ (6 g/ml), or SDF-1␣ (6 g/ml) was determined as described under "Experimental Procedures." Each bar represents the mean Ϯ S.D. of quadruplicate determinations. FIG. 6. Sensorgrams recording the interactions of chemokines, L-and P-selectin, and CD44 with immobilized GAGs in the BIAcore. Various concentrations of SLC, IP-10, SDF-1␤, monomeric L-and P-selectin, and monomeric CD44 were continuously injected over sensorchip surfaces coupled to CS E or CS A at a flow rate of 30 l/min from time 0 to 90 s; after this, running buffer was injected and the response in resonance units was recorded as a function of time. Sulfation plays an important role in the interactions of L-and P-selectin with the majority of ligands hitherto reported. For example, the tyrosine sulfation of P-selectin glycoprotein ligand-1 is required for its interaction with L-and P-selectin (17)(18)(19). The ligands for L-selectin on the high endothelial venules bind Lselectin in a sulfation-dependent manner (20 -22). HNK-1reactive sulfoglucuronyl glycolipids (23), heparin oligosaccharides (24), and HS GAGs (25) bind L-and P-selectin. Our results showing that sulfation is required for versican's binding to L-and P-selectin ( Figs. 1 and 4) are thus consistent with these previous findings. Extending these observations, we also showed that the sulfation of versican is required for the interaction with chemokines but not CD44 (Figs. 1 and 4). The absence of a sulfation requirement for versican's binding to CD44 is reminiscent of the binding of unsulfated GAG and HA to CD44 (26).
Surface plasmon resonance analysis showed that the binding affinities of soluble monomeric L-and P-selectin to oversulfated CS/DS chains are higher than those for known ligands ( Fig. 6 and Table II). It has been reported that monomeric L-selectin binds to immobilized glycosylation-dependent cell adhesion molecule-1 (GlyCAM-1) with low affinity (K d ϭ 108 M) and with very fast association (Ն10 5 M Ϫ1 s Ϫ1 ) and dissociation (Ն10 s Ϫ1 ) rates (27). P-selectin has been reported to bind to P-selectin glycoprotein ligand-1 with relatively high affinity (K d ϳ300 nM) and with fast association (4.4 ϫ 10 6 M Ϫ1 s Ϫ1 ) and dissociation (1.4 s Ϫ1 ) rates (28). These properties have been proposed to be critical for the dynamic selectin-mediated rolling adhesion that is mediated by rapid adhesion and deadhesion (27,28). We speculate that the high-affinity binding of L-and P-selectin to oversulfated CS/DS chains with a slow dissociation rate (Table II)  The oversulfated CS/DS chains interact with certain chemokines with high affinity as observed with the selectins (Table  II). The high affinity binding of chemokines SLC, IP-10, and SDF-1␤ suggest that these chemokines might be readily trapped by oversulfated CS/DS chains in vivo. This hypothesis is supported by the surface plasmon resonance kinetic analysis, which demonstrated that the oversulfated CS/DS-chemokine complex formation is characterized by fast association rates (0.864 to 4.15 ϫ 10 4 M Ϫ1 s Ϫ1 ). It is conceivable that the oversulfated CS/DS-bound form of chemokines may not function as FIG. 8. CS E inhibits Ca 2؉ responses induced by SLC but not by SLC-T. L1.2/CCR7 cells (1 ϫ 10 6 cells/ml) were loaded with Fura-2 and stimulated with SLC or SLC-T (100 ng/ml) in the presence or absence of GAGs (100 g/ml). Arrowheads indicate the time of application of the stimulators. Intracellular calcium concentration was monitored by measuring the fluorescence ratio as described previously (5). The scale is shown as a bar on the right. streptavidin-conjugated alkaline phosphatase coupled to or without (pink) biotinylated fractions a (red), c (yellow), e-1 (green), or e-2 (blue) to wells coated with BSA (10 g/ml), L-selectin-Ig (L-Ig, 5 g/ml), E-selectin-Ig (E-Ig, 5 g/ml), P-selectin-Ig (P-Ig, 5 g/ml), CD44-Ig (5 g/ml), SLC (5 g/ml), SLC-T (5 g/ml), IP-10 (10 g/ml), PF4 (2.5 g/ml), SDF-1␤ (5 g/ml), or SDF-1␣ (5 g/ml). Binding was detected with the Blue Phos TM substrate as described previously (4). Each bar represents the mean Ϯ S.D. of quadruplicate determinations.
agonists for chemokine receptors, because those GAGs inhibit chemokine activity (Fig. 8). Rather, the oversulfated CS/DSchemokine complex may function as a reservoir for chemokines in vivo. The slow dissociation rate (2.78 ϫ 10 Ϫ4 to 5.30 ϫ 10 Ϫ3 s Ϫ1 ) that is observed in the interaction of chemokines with oversulfated CS/DS chains supports this idea.
Hints about the mechanism of CS/DS binding to chemokines are provided by the observation that versican interacts with SDF-1␤ but not SDF-1␣ (Figs. 4 and 5). Because SDF-1␣ and SDF-1␤ are produced by an alternative splicing of a single gene with the latter lacking only 4 amino acid residues in the C terminus (29), versican, or its CS/DS chains, appears to interact with the chemokine's C terminus. Consistent with this idea, versican failed to bind to a mutant SLC that lacked the Cterminal basic amino acid clusters (Fig. 4), and CS B and CS E also failed to bind the C terminus truncated SLC. 2 Therefore, these data strongly suggest that CS/DS interacts with the C terminus of these chemokines.
Oversulfated CS/DS chains containing GlcA␤1/IdoA␣1-3GalNAc(4,6-O-disulfate) have been found in squid cartilage (30), mast cells (31,32), neutrophils (33,34), monocytes (35,36), glomeruli (37), and mesangial cells (38), and have been reported to interact with various biologically active molecules and regulate their functions in vitro. For example, CS E inhibits the adhesion of cortical neuronal cells to the neurotrophic factor midkine through direct interaction with midkine (39), promotes neurite outgrowth in embryonic day 18 rat hippocampal neurons (40), serves as a receptor for PF4 on the surface of neutrophils (34), and inhibits procoagulant activity (36). Our results showing that oversulfated CS/DS chains containing GlcA␤1/IdoA␣1-3GalNAc(4,6-O-disulfate) bind to proinflammatory molecules, such as selectins and chemokines, and regulate chemokine activity, further indicate the versatility of these oversulfated chains in regulating biological responses. It is now extremely important to understand the exact tissue distribution of the oversulfated CS/DS chains bearing the abil-ity to interact with these proinflammatory molecules and the regulatory mechanisms whereby these GAGs are generated.