Recombinant Human Interleukins IL-1 a , IL-1 b , IL-4, IL-6, and IL-7 Show Different and Specific Calcium-independent Carbohydrate-binding Properties*

A method was developed for the determination of pu-tative lectin activities of cytokines. It involved the immunoblotting measurement of the quantity of these cytokines unbound to a series of different immobilized glycoconjugates and displacement of the bound cytokines with oligosaccharides of known structures. This method allows demonstrating that the following interleukins specifically recognize different oligosaccharide structures in a calcium-independent mechanism: inter-leukin-1 a binds to the biantennary disialylated N -gly-can completed with two Neu5Ac a 2–3 residues; interleu-kin-1 b to a GM 4 sialylated glycolipid Neu5Ac a 2–3Gal b 1-Cer having very long and unusual long-chain bases; interleukin-4 to the 1,7 intramolecular lactone of N -acetyl-neuraminic acid; interleukin-6 to compounds having N -linked and O -linked HNK-1-like epitopes; and interleukin-7 to the sialyl-Tn antigen. Because the glycan ligands are rare structures in human circulating cells, it is suggested that such activities could be essential cerebellum (phospholipids were eliminated using mild alkaline methanolysis). Sciatic nerve extracts were obtained by homogenization of freshly dis-sected sciatic nerves from adult Wistar albino rats in 1% SDS at the concentration of 10 mg of protein/ml. Myelin-associated glycoprotein (MAG) was purified from adult rat brain myelin (22) according to Quarles et al. (23). The P 0 glycoprotein was prepared from rat sciatic nerve myelin according to Kitamura et al. ( 24). Oligosaccharides were isolated from different sources, and their structures were determined by NMR as pre- present in the mixture (for example separated RNase B and bLTF for IL-1 a or individual mucins for IL-4 and IL-7 or purified glycoproteins for IL-6 or purified gangliosides for IL-1 b ). When a specific binding to one of these compounds was observed, attempts were made to inhibit this binding first with mixtures of oligosaccharides isolated from the glycoconjugate ligand and then with separated individual oligosaccharides. In crude inhibition assays, oligosaccharides were added to the wells at a concentration of 10 m g/ml in PBS for mixtures, or 1 m g/ml in PBS for purified compounds, prior to addition of the cytokine. For testing the efficiency of the different inhibitory glycans, serial 10-fold dilutions of these compounds in PBS were performed.

Cytokines are modulators of the activity of the immune system, their mechanism of action remaining, for the most part, not decrypted. As a general rule, the action of a cytokine results from its binding to membrane receptors, a series of molecules coupled to signaling systems involving kinases and/or phosphatases (1)(2)(3)(4)(5). The binding of a cytokine to its receptor(s) generally results in the phosphorylation/dephosphorylation of the intracytoplasmic domain of the receptor(s), the first step of the signaling. In general, the phosphorylation/dephosphorylation mechanism is cell type-specific, the kinases/phosphatases involved in these processes being quite specifically associated with surface molecular complexes different from the cytokine receptor complex (3,6). Even when two different cytokines use the same receptor, the signal transduction pathways may be specific of the cytokines (7).
We made the hypothesis that the specific association of interleukin receptors with other surface complexes could be due to carbohydrate-binding properties of these cytokines, a property already suggested in the literature (8 -14). The lectin activity of interleukin-2 (IL-2) 1 for specific oligomannosides (15) appeared to be essential, because IL-2 behaves as a bifunctional molecule able to extracellularly associate its ␤ receptor (IL-2R␤) to other surface receptor complexes bearing N-glycans recognized by IL-2. This is the case for the CD3⅐TCR complex in which a N-glycosylated form of CD3 is an IL-2 ligand. This specific extracellular association is responsible for the specific phosphorylation of the IL-2R␤ by the CD3⅐TCR-associated kinase p56 lck (15), considered as a first step in the antigenspecific activation process of CD4 ϩ T cells. As a consequence of this carbohydrate-binding property, it was suggested that oligomannosides accumulated in specific diseases or bound to specific microorganisms could alter this essential role of IL-2 (16,17). Because several groups reported lectin-like activities for cytokines (8, 10 -13), we undertook a study of such properties for recombinant human interleukins. This work demonstrates that IL-1␣, IL-1␤, IL-4, IL-6, and IL-7 recognize very specific oligosaccharide ligands in a calcium-independent mechanism. The implications of these findings are discussed.

EXPERIMENTAL PROCEDURES
Biochemicals-Recombinant human cytokines (produced in bacteria) and their respective polyclonal rabbit antibodies were from Chemicon International Inc. (Temecula, CA). Alkaline-phosphatase-labeled antirabbit IgG and normal goat serum were from Sigma. Ovalbumin, ribonuclease B (from bovine pancreas), fetuin, and glycosaminoglycans (chondroitin sulfate A, B, and C, heparan sulfate, heparin, dermatan sulfate, and hyaluronic acid) were from Sigma. Bovine lactotransferrin was from Euromedex (Souffelweyersheim, France). Equine and ovine submaxillary mucins were prepared according to Tettamanti and Pigman (18), and Bufo bufo egg mucins were prepared according to Morelle and Strecker (19). Gangliosides were extracted from young rat cerebella as previously described (20). Neutral glycolipids were extracted from human meconium by Folch extraction and Folch partition followed by separation by silica gel chromatography (21). Asialo-GM 1 was prepared by limited acid hydrolysis (aqueous formic acid, pH 2.0, during 30 min at 50°C) followed by a Folch partition. Cerebrosides and sulfatides were isolated from the rat cerebellum (phospholipids were eliminated using mild alkaline methanolysis). Sciatic nerve extracts were obtained by homogenization of freshly dissected sciatic nerves from adult Wistar albino rats in 1% SDS at the concentration of 10 mg of protein/ml. Myelin-associated glycoprotein (MAG) was purified from adult rat brain myelin (22) according to Quarles et al. (23). The P 0 glycoprotein was prepared from rat sciatic nerve myelin according to Kitamura et al. (24). Oligosaccharides were isolated from different sources, and their structures were determined by NMR as pre-* This work was supported by Grant 9925 of the French Association pour la Recherche sur le Cancer. 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.
Glycoproteins and glycosaminoglycans were dissolved in PBS (25 mM sodium phosphate buffer, pH 7.2, containing 0.15 M NaCl), each at the concentration of 1 mg/ml. Sciatic nerve extract was dissolved (1 mg of protein/ml) in PBS containing 0.1% SDS. Gangliosides and glycolipids were dissolved in methanol (10 g of hexose/ml of each compound). 50 l of the previous mixtures were added to the wells and put in an oven at 50°C until dried. After immobilization, the microwells were washed thrice with 300 l of PBS and then saturated thrice during 2 h with periodate-treated BSA (pBSA). As reported by Glass et al. (43), the periodate treatment followed by elimination of excess of periodate with glycerol of BSA, reduced by a factor 100 the quantity of lectin necessary to reveal specific ligands because of the absence of binding to oligosaccharide impurities of BSA (44). Wells were washed with PBS and then water and dried at room temperature. The quantity of bound glycoconjugates was calculated from the amount of hexose equivalents determined directly in the microwells using the resorcinol sulfuric acid micromethod (45) relative to dilutions of a galactose solution treated in parallel. Measurements of the absorbance were performed using a Bio-Rad model 550 microplate reader at 430 nm and applying the coefficient of relative absorbance as reported by the previous authors (45). A mean value of 0.5 g galactose equivalent was found for fetuin, ovalbumin, OSM, gangliosides, neutral glycolipids, and RNase B, 1.2 g for equine submaxillary mucin and the mucin of B. bufo, and 1.5 g for the mixture of glycosaminoglycans. These quantities of bound glycans remained the same after incubations with the cytokines as determined in the wells after recovery of the supernatant.
Detection of the Binding of Cytokines to Immobilized Glycoconjugates-The wells containing the immobilized glycoconjugates were saturated once as above, washed thrice in PBS, and then supplemented with 50 l of PBS containing 0.3% pBSA and 5 mM EDTA. Cytokines (0.1-0.5 g in 5 l of PBS) were added, and the plates were incubated for 2 h at room temperature with a rotatory agitation. The supernatants were carefully recovered and diluted with 25 l of the Laemmli dissociating buffer (46). Samples were placed for 15 min in boiling water. Identical volume aliquots (10 -25 l) corresponding to the same cytokine were submitted to polyacrylamide gel electrophoresis (13% acrylamide) in the presence of SDS (46) followed by blotting onto nitrocellulose (47). After saturation with 3% BSA in PBS containing a 1/20 dilution of normal goat serum, blots were revealed using the proper polyclonal anti-cytokine antibody followed by the alkaline phosphataselabeled second antibody and then staining with the nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate reagent (48). The binding of a cytokine to one or several glycoconjugate mixtures was evidenced by its decreased amount on blots. For quantitation, blots were scanned using a Scanjet 4c (Hewlett Packard), and integrations were performed using the Quantiscan 32 software. Because for most cytokines the polyclonal antibodies were allowed to detect down to 10 ng of cytokines, optimal results (difference in intensity between samples for which a binding was observed and the others) were obtained using 50 to 100 ng of cytokines in each incubation. All determinations were performed in triplicate. Incubations in the presence of 0.3% pBSA also reduced nonspecific binding. However, the very high amount of pBSA produced specific electrophoretic artifacts; these were frequently curving of the electrophoretic front and presence of a double band for a single compound. These artifacts could be substantially reduced but not eliminated by decreasing the electric power of the electrophoresis during the concentration step of the sample.
Determination of the Specificity of the Cytokine-Glycoconjugate Interactions-Once a binding of a cytokine to a specific mixture was observed, a refinement of the nature of the interaction was performed. The first step consisted in studying the binding of the cytokine to individual glycoconjugates present in the mixture (for example separated RNase B and bLTF for IL-1␣ or individual mucins for IL-4 and IL-7 or purified glycoproteins for IL-6 or purified gangliosides for IL-1␤). When a specific binding to one of these compounds was observed, attempts were made to inhibit this binding first with mixtures of oligosaccharides isolated from the glycoconjugate ligand and then with separated individual oligosaccharides. In crude inhibition assays, oligosaccharides were added to the wells at a concentration of 10 g/ml in PBS for mixtures, or 1 g/ml in PBS for purified compounds, prior to addition of the cytokine. For testing the efficiency of the different inhibitory glycans, serial 10-fold dilutions of these compounds in PBS were performed.

RESULTS
The rationale of the method used for studying the lectin activities of interleukins was to detect the cytokines unbound to different immobilized glycoconjugates with known glycan structures using an immunoblotting technique as schematized in Fig. 1. The reasons for choosing such a method were as follows: 1) because cytokines could loose their carbohydratebinding properties upon chemical or radiochemical labeling (49), we used unlabeled recombinant human cytokines having a preserved biological activity; 2) because no precise carbohydrate-binding properties of most cytokines were found in the literature, we studied the binding of cytokines to a large variety of glycoconjugates or mixtures of glycoconjugates, the structure of their glycans being in large part determined; and 3) because the presence of calcium (above 1 mM in Tris-buffered saline) induced a nonspecific fixation of all interleukins to all immobilized glycoconjugates, the experiments were performed in the presence of 5 mM EDTA. Consequently, the data reported in this manuscript were concerned only with Ca 2ϩ -independent carbohydrate-binding properties.
Interleukin 1␣ Is a Lectin Specific for Disialylated Biantennary N-Glycans with Two ␣2,3-Linked Sialic Acid Residues-As shown in Fig. 2a, IL-1␣ binds significantly to fetuin and to the mixture of RNase B and bLTF, whereas no binding was observed to the other immobilized glycoconjugates. These data were surprising, because the major glycans of RNase B and bLTF were of the oligomannosidic type, in contrast with fetuin, which contains essentially sialylated O-and N-glycans. When incubations were performed separately on RNase B and bLTF, the binding of IL-1␣ was observed only to the latter. From the quantity of IL-1␣ bound to bLTF, it could be calculated that IL-1␣ binding sites were present on about 7% of the bLTF molecules, a result that indicated that the ligands of IL-1␣ corresponded to minor glycans of this molecule, as they were minor glycans of fetuin. In fact, the O-glycan fraction obtained from fetuin after reductive ␤-elimination and Biogel P2 chromatography, as well as the two major O-glycans of fetuin, were devoid of inhibitory activity, in contrast with the N-glycan-enriched fraction obtained by extensive Pronase digestion of fetuin. However, the purified major triantennary N-glycan of fetuin and the products obtained after its partial or complete desialylation did not show any inhibitory activity. Because bLTF possessed minor biantennary N-glycans with the Neu5Ac␣2,6GalNAc␤1,4GlcNAc sequence (35), we tested the inhibitory activity of this compound. As for the previous ones, no significant inhibitory activity was obtained at the 10 Ϫ4 M concentration range. We therefore tested a variety of biantennary oligosaccharides isolated from the urine of patients with sialidoses (25,26). The one containing two ␣2,6-linked Neu5Ac residues was ineffective at 10 Ϫ4 M, whereas the binding of IL-1␣ was completely inhibited using 10 Ϫ6 M of the biantennary N-glycan containing two ␣2,3-linked Neu5Ac residues (Fig. 2, b  and c). A quasi-equimolar mixture of the two isomers of biantennary N-glycans containing ␣2,3and ␣2,6-linked Neu5Ac residues showed a weak inhibition only at the 10 Ϫ4 M concentration range. Therefore, it was concluded that IL-1␣ is a calcium-independent lectin endowed with a higher affinity for the biantennary N-glycan having the Neu5Ac␣2,3Gal␤1, 4GlcNAc␤1,2 sequence on its two branches (Fig. 2d).
Previous authors (8) reported that IL-1␣ was able to bind to uromodulin, a urine glycoprotein rich in sialylated N-glycans, the binding being inhibited by the bulk of glycopeptides isolated by Pronase digestion of fetuin. These authors suggested that IL-1␣ was a lectin specific for the major glycan of fetuin, the trisialylated triantennary N-glycan. Our experiments demonstrated that the inhibitor of the interaction was not the previous glycan but the biantennary N-glycan with two ␣2,3linked Neu5Ac residues. 50% inhibition of the binding to fetuin was obtained using 0.5-1 M of this compound, i.e. 100 -200fold higher efficiency than the total N-glycans of fetuin. The reason why the biantennary N-glycan was a high affinity li-gand, whereas ␣2,3-sialyl-lactose or the mixture of the linear monoantennary ␣2,3-sialylated glycan isolated from patients with sialidosis (25,26) was ineffective, remained speculative. This might be related to the tendency of IL-1␣ to associate into dimers, with each subunit being able to bind one of the branches of the biantennary N-glycan. Such a specifically increased affinity upon formation of oligomers of lectins was previously demonstrated for the asialo-fetuin receptor (50).
IL-1␤ Is Likely a Lectin Specific for the GM4 Glycolipid-IL-1␤ binds only significantly to the fraction containing the mixture of gangliosides isolated from the rat cerebellum (Fig.  3a). Because these gangliosides were previously fractionated according to their charges by ion-exchange chromatography (20), we examined the binding of IL-1␤ to the different classes of mono-, di-, tri-, and tetrasialo-gangliosides, respectively. Only the monosialo-ganglioside fraction showed a binding of IL-1␤ (Fig. 3b). The different constituents of the fraction of monosialo-gangliosides isolated by preparative TLC (20) were immobilized individually. As shown in Fig. 3c, IL-1␤ binds to a single fraction, previously identified as the sialylated galactosylceramide GM 4 (20).
This binding was surprising, because GM 4 is a very simple sialylated glycolipid, Neu5Ac␣2,3Gal␤1-Cer. It was therefore expected that the binding could be inhibited by small oligosaccharides with similar structures. In fact, sialyl-lactoses (with ␣2,3or ␣2,6-linked Neu5Ac) were without inhibitory effects at the concentration of 10 Ϫ3 M. Because it was described that IL-1␤ binds to glycolipids of GPI anchors (13), we tested the possibility that the GM 4 sample used in this study was contaminated by GPIs. This possibility was ruled out using GC/MS analysis of this GM 4 sample (30,31). Using a new procedure allowing the complete liberation of all GPI constituents, 3 the analysis showed the total absence of D-mannose, inositol, and GlcN, which are essential constituents of GPIs. To further document this point, the binding of IL-1␤ was tested to an immobilized glycosylphosphotydylinositol-ceramide isolated from Tulamen (kindly provided by Drs. J. and L. Previato). No binding was observed. Furthermore, the binding of IL-1␤ to GM 4 was not inhibited by Man-6-P at the concentration of 10 Ϫ3 M, indicating that the binding to GM 4 observed here was unrelated to the binding to GPI oligosaccharides (13).
The question remained to know how a compound with a simple carbohydrate composition could be a specific ligand of a cytokine. Based on the observations of Karlsson (51) on the influence of the lipid moiety on the binding of microbe lectins, we made the hypothesis that part of the specificity could be due to the lipid portion of the rat cerebellum GM 4 . This question was answered using GC/MS analysis (30). Although the fatty acid composition was classical (C16:0 and C18:0 in equivalent amounts representing more than 98% of total fatty acids), the long-chain base composition was unusual. Indeed, the linear C22:0 phytosphingosine (60%) and the linear C22:1 sphingenine (30%) were the major constituents. A minor constituent (10%) was identified as 18-O-ethyl-C18:1 sphingenine. Although it was not demonstrated that the long-chain base composition of the GM 4 was important for the binding of IL-1␤, this remained a stimulating possibility. Because of the lack of an inhibitor, the lectin activity of IL-1␤ could not be ascertained as such. Nevertheless, the completely different behavior as lectins of IL-1␣ and IL-1␤ could explain why these interleukins, endowed with a common receptor, have different biological functions (see "Discussion").  (Neu5Ac-1,7L)-IL-4 binds strongly to the mucin mixture (Fig. 4a). When the binding was studied for the three different isolated mucins, IL-4 binds strongly to the mucin of the eggs of B. bufo and at a lesser extend to OSM (Fig. 4b). The binding was not inhibited using the mixtures of the acidic or neutral oligosaccharide alditols isolated from the mucin by reductive ␤-elimination (see Ref. 27 and Fig. 4c). This suggested two possibilities; either the interaction was carbohydrate-independent, or an essential determinant was lost during the alkaline treatment of the ␤-elimination procedure. In fact, the binding of IL-4 to the B. bufo mucin was eliminated when the mucin was first exposed to NH 3 gas for 72 h at room temperature, indicating that an important determinant for the binding was an O-acyl group, possibly located on a sialic acid residue of this mucin. This was actually the case, because the binding of IL-4 to the mucin of the eggs of B. bufo was inhibited using the sialic acids liberated from the mucin by mild acid hydrolysis (Fig. 4c, arrow). Previous studies (27) showed that the sialic acids detected in this mucin after the procedure of   3. Evidence for the specific binding of IL-1␤ to GM 4 . a, immunoblotting screening of the binding of 0.1 g of IL-1␤ to the different classes of immobilized glycoconjugates. Note the binding of IL-1␤ to the ganglioside mixture. b, study of the binding of IL-1␤ to the families of mono-, di-, tri-, and tetrasialo-gangliosides. Note the specific binding to the monosialo-ganglioside fraction (arrow head). c, comparative binding of IL-1␤ to the different isolated monosialylated glycosphingolipids. Note the decreased amount of the cytokine in the supernatant issued from the microwell in which GM 4 was immobilized (arrow head). The amount of immobilized gangliosides were proportional to that present in the initial mixture of monosialo-gangliosides in which GM 4 represented only 5% of the total compounds. d, structure of the three long-chain bases identified in the immobilized GM 4 . In a, b, and c, the direction of the electrophoreses was to the left. ␤-elimination were Neu5Ac and Neu5Gc in equivalent amounts. Because the commercial compounds were not inhibitory, it was concluded that the ligand of IL-4 was an Oacylated derivative of one of these two compounds.

N-Acetyl Neuraminic Acid
When GC/MS analyses of the sialic acids present in this mucin were performed, 2 four major peaks were detected (Fig.  5). Two of them corresponded to Neu5Ac and Neu5Gc, respectively, but Neu5Ac was a very minor constituent of the mucin. In contrast, the sialic acids derived from the oligosaccharidealditols obtained by reductive ␤-elimination showed an equivalent ratio of Neu5Ac and Neu5Gc (27). Classical O-acetylated and O-lactylated sialic acids were absent from this mucin. In contrast, the major peak corresponded to a compound with a mass of 879 (as detected in the chemical ionization mode, i.e. a deficit of mass of 228 relative to Neu5Ac) suggesting that the major compound corresponded to a lactone derived from Neu5Ac. Based on the fine fragmentation pattern of the compound in the electron impact mode of ionization, it was concluded that this compound was the 1,7 lactone of Neu5Ac (Fig.  4d). The fourth peak corresponded to the 1,7 lactone of Neu5Gc. The existence of such intramolecular lactones of sialic acids was previously suggested in the literature but never obtained in sufficient amounts to be identified with security.
The problem remained to know whether the ligand of IL-4 in the B. bufo mucins was the 1,7 lactone of Neu5Ac or that of Neu5Gc or both. Because OSM, showing only a weak binding for IL-4, contains a significant amount of the 1,7 lactone of Neu5Ac but no trace of the corresponding lactone of Neu5Gc, it was concluded that the ligand of IL-4 in the B. bufo mucins could not be the 1,7 lactone of Neu5Gc but was the 1,7 lactone of Neu5Ac. Fig. 6a, IL-6 binds only to the sciatic nerve extract, suggesting the presence of a unique ligand in this heterogeneous fraction. We made the hypothesis that a possible ligand was the HNK-1 epitope (SO 3 H (3-)GlcA) present both on glycolipids and glycoproteins found in this tissue (52). IL-6 also binds to the mucins isolated from the eggs of Rana temporaria, also rich in glycans bearing this epitope (see Ref. 28 and Fig. 6b). Because the two major glycoproteins of rat peripheral nervous system myelin (MAG and overall P 0 ) possessed this epitope, we immobilized these two isolated glycoproteins on plastic and measured the binding of IL-6 relative to the other immobilized glycoconjugates. IL-6 actually binds to these two isolated glycoproteins. The binding of IL-6 was not inhibited by glycopeptides isolated from RNase B or from fetuin. In contrast, the binding was inhibited using 10 Ϫ4 M of a glycopeptide fraction obtained from the rat brain synaptosomal fraction (53) in which the HNK-1 epitope was present as minor constituents based on matrixassisted laser desorption ionization/time of flight analysis. 4 Because a series of reduced oligosaccharides possessing the SO 3 H (3-)GlcA motive were isolated from the mucins of the eggs of R. temporaria (28), we tested these compounds for their inhibitory activity of the binding of IL-6 to the rat P 0 glycoprotein. Interestingly, the best inhibitor was a compound comprising a Fuc residue (see Ref. 28 and Fig. 6, c-e), whereas compounds lacking the Fuc residue were not inhibitory at 10 Ϫ4 M. The presence of an additional Gal residue also abolished the inhibitory potency of the oligosaccharides (Fig. 6, c-e). Because none of the other oligosaccharides lacking the SO 3 H (3-)GlcA group were inhibitory (data not shown), it was concluded that, although the SO 3 H (3-)GlcA motive was necessary for the binding to IL-6, it was not sufficient. Considering the structure already determined of the N-glycans possessing the HNK-1 epitope (41) and the structure of the O-glycans inhibiting the binding of IL-6 to these N-glycans, it was suggested that the recognition domain of IL-6 consisted in the sequence SO 3 H (3-)GlcA␤1,3Gal␤1,4 followed by a monosaccharide sequence possessing a methyl group in the vicinity of this structure (issued either from the acetamido group in GlcNAc in N-glycans or from the methyl group in the Fuc residue in R. temporaria oligosaccharide alditols). Fig. 7a, IL-7 binds to fetuin and the mixture of mucins and not to the glycosaminoglycan mixture, as contrasted from the data of previous authors (12). When the analysis was subsequently performed on the individual mucins, IL-7 binds only to the ovine submaxillary mucin (Fig. 7b). The binding was not changed when the OSM was submitted to partial or total deacylation with NH 3 gas, indicating that O-acylation of sialic acid residues was not important for the binding. In contrast this binding was inhibited when the mucin was desialylated by mild acid hydrolysis, indicating that a sialic acid residue was important for the binding. However, this binding could not be inhibited using 10 Ϫ4 M of purified Neu5Ac (the major sialic acid present in OSM) or purified Neu5Gc or by the mixture of sialic acids isolated from OSM by mild acid hydrolysis. Furthermore, O-glycans isolated from OSM by a reductive ␤-elimination procedure did not show any inhibitory activity. This suggested that the presence of a GalNAc-OH instead of GalNAc residue abolished the binding or that the binding of IL-7 involved a portion of the polypeptide chain. These assumptions were sustained by the observation that complex glycopeptide mixtures obtained by prolonged Pronase digestion of OSM were inhibitory of the interaction of IL-7 with OSM. As shown in Fig. 7c, an active inhibitory compound was found as a low M r fraction isolated by Biogel P2 chromatography. This fraction contained a single resorcinol-positive spot migrating faster than the major monosialo-glycoserine of fetuin Neu5Ac␣2,3Gal␤1,3GalNAc␣1-Ser and containing one Ser, one GalNAc, and one Neu5Ac residue (Fig. 7d). Consequently, these data were compatible with the hypothesis that the IL-7 ligand was the sialyl-Tn antigen, an onco-fetal antigen relatively abundant on the ovine submaxillary mucin and comprising both an oligosaccharide and peptide determinant (54). 1␣, IL-1␤, IL-4, IL-6, and IL-7. Similar studies have been undertaken for other cytokines or chemokines. However, lectin activity has not been demonstrated for some of them, possibly because we lack the specific ligand. The case of IL-4 is very relevant to this view, because its lectin activity would have been missed in the absence of the B. bufo mucin as a ligand. Several authors have demonstrated that cytokines could loose their lectin activity upon chemical or radiochemical labeling. Indeed, as demonstrated for IL-2, the labeling of the cytokine by iodine or biotin inhibited its biological function, although the binding to the receptor was not affected (49). Thus, the principal interest of this technique is to allow studies with proteins having a fully preserved biological activity. The discovery of the high affinity ligand will permit the labeling of the cytokine in the presence of the carbohydrate epitope to protect the carbohydrate recognition domain and leads to further investigations, like quantitative studies and identification of the endogenous ligands of cytokines on cells or on blots.

IL-7 Binds Specifically to a Glycopeptide of the Ovine Submaxillary Mucin-As shown in
We demonstrate that except for IL-1␤ (for which a soluble high affinity oligosaccharide ligand was not yet isolated), the other interleukins have high affinity oligosaccharide or glycopeptide ligands inhibiting the binding of the cytokine to the immobilized ligands at a concentration range lower than 10 Ϫ6 M. Although the K d of the interaction could not be determined using the present methodology, the identified ligands were of high affinity. For example, the ligand of IL-6 from R. temporaria was able to inhibit the binding of IL-6 to MAG at concentration as low as 10 Ϫ10 M. This suggested that if endogenous ligands are present on cells of the immune system, the lectin activity of the human interleukins may provide another targeting system different from the classical interleukin receptors. Although high affinity ligands of these interleukins were identified, the endogenous ligands of the different cytokines are still not identified. They may be structurally different from (or at least not identical to) the endogenous ligands but should share at least a common determinant. For example, IL-6 binds to N-linked HNK-1 epitope found on MAG and P 0 glycoproteins, the structure being HSO 3 (3-)GlcA␤1,3Gal␤1,4GlcNAc␤1,2 (44). The ligands found in R. temporaria had a significantly different structure, HSO 3 (3-)GlcA␤1,3Gal␤1,4[Fuc␣1,2]Ga-l␤1,3, whereas the linear oligosaccharide lacking the Fuc residue was inactive at 10 Ϫ4 M. This suggested that, besides a site for the SO 3 H (3-)GlcA determinant, the carbohydrate recognition domain of IL-6 contains a domain interacting with a methyl group provided either by the 2-acetamido group of Glc-NAc or the methyl group of Fuc, the HSO 3 (3-)GlcA␤1,3Gal␤1,4 being not sufficient for a high affinity interaction. For IL-4, the ligand, the 1,7 lactone of Neu5Ac, is a compound actually present in glycoproteins of the human lymphocyte membrane. 5 The question of the function of such lectin activities of cytokines remains largely unanswered. As suggested by the function of the IL-2 lectin activity (15), an essential role in cytokine signaling is expected, consisting in association of the cytokine receptor complex with specific glycoprotein or glycolipid ligands of another surface complex. As a consequence of this hypothesis, only cells having both the receptor and the ligand of the cytokine could respond to the cytokine. This could explain why different cytokines having the same receptor can stimulate specifically certain cell types and not the others. As an example, IL-1␣ and IL-1␤ have the same receptors but have different signaling profiles in different cells. IL-1␤, but not IL-1␣, is able to stimulate human astrocytes, a mechanism responsible for the nervous regulation of fever (55). It is noteworthy considering that astrocytes are the only cells of the central nervous system producing IL-1␤ and possessing both IL-1 receptors (56) and the GM 4 glycolipid (57).
The experimental approaches to answer the question of the 5 C. Mariller, A. Pous, and J.-P. Zanetta, manuscript in preparation.
FIG. 6. Evidence for the high affinity binding of IL-6 to specific compounds possessing the HSO3(3-)GlcA motive. a, immunoblotting screening of the binding of 0.1 g of IL-6 to the different mixtures of immobilized glycoconjugates. Note that a significant decreased in the amount of IL-6 in the supernatant was only observed for the sciatic nerve extract (arrow). b, binding of IL-6 to R. temporaria mucins (RtM). Wells were coated by 50 g of mucins, except in control wells (ctrl). Note (arrow) the strong fixation of IL-6 to the mucins. c, inhibition experiments of the binding of IL-6 to the immobilized rat P 0 glycoprotein using different oligosaccharide-alditols obtained from the jelly coat mucins of the eggs of R. temporaria. Note (arrows) that only compounds I and IV were active. d, quantitation of the experiments shown in c. e, structures of the different compounds used in these inhibition experiments. The presence of an additional Gal (compound II) abolishes the inhibitory potency of compound IV observed in c.
biological function of the lectin activity of interleukins appear relatively simple, because the binding of a cytokine to its receptor induces specific changes in phosphorylation/dephosphorylation, as observed for IL-2 (15). Indeed, the high-affinity oligosaccharide ligand should inhibit the changes of phosphorylation/dephosphorylation induced by the cytokine. This is actually the case for the IL-6 ligand from R. temporaria, because this compound at a 10 Ϫ10 concentration range inhibits the dephosphorylation of a few tyrosine-phosphorylated proteins induced by IL-6 in a whole population of resting human lymphocytes. 6 Therefore, this study provides new concepts and tools for understanding the mechanisms of the regulation of the human immune system and suggests that fine modulations of the immune system could be obtained using specific oligosaccharides or glycans isolated in consistent amounts from natural sources. These substances may be an alternative to therapies using antibodies against interleukins or interleukin receptors in different pathologies. FIG. 7. Carbohydrate-binding properties of IL-7. a, immunoblotting screening of the binding of 0.1 g of IL-7 to the different immobilized glycoconjugates. Note the low but significant decrease in the amount of IL-7 in the supernatant of incubations with fetuin and the mixture of mucins. b, demonstration that IL-7 binds to OSM. c, inhibition of the IL-7 binding to OSM using glycopeptide fractions obtained by extensive Pronase digestion of OSM and fractionation by Biogel P2 gel filtration. Note that the very low M r fraction OSM5 completely inhibited the binding of IL-7 to immobilized OSM. Control corresponds to incubation of IL-7 to a well containing only periodate-treated BSA. d, high performance thin-layer chromatography migration in the solvent system propanol/methanol/0.25% aqueous KCl (60/10/35 v/v/v) 7 of the sialylated glycopeptide showing an inhibition of the binding of IL-7 to immobilized OSM (arrow head). Note that this compound migrates faster than the Neu5Ac␣2,3Gal␤1,3GalNAc␣1-Ser (1) and Neu5Ac-␣2,3Gal␤1,3[Neu5Ac␣2,6]GalNAc␣1-Ser (2), suggesting a Neu5AcGalNAc-Ser sequence.