Function and Molecular Modeling of the Interaction between Human Interleukin 6 and Its HNK-1 Oligosaccharide Ligands*

Interleukin 6 (IL-6) is endowed with a lectin activity for oligosaccharide ligands possessing the HNK-1 epitope (3-sulfated glucuronic acid) found on some mammalian glycoproteinN-glycans (Cebo, C., Dambrouck, T., Maes, E., Laden, C., Strecker, G., Michalski, J. C., and Zanetta, J. P. (2001)J. Biol. Chem. 276, 5685–5691). Using high affinity oligosaccharide ligands, it is demonstrated that this lectin activity is responsible for the early dephosphorylation of tyrosine residues found on specific proteins induced by interleukin 6 in human resting lymphocytes. The gp130 glycoprotein, the signal-transducing molecule of the IL-6 pathway, is itself a molecule possessing the HNK-1 epitope. This indicates that IL-6 is a bi-functional molecule able to extracellularly associate its α-receptor with the gp130 surface complex. Computational modeling indicates that the lower energy conformers of the high affinity ligands of IL-6 have a common structure. Docking experiments of these conformers suggest that the carbohydrate recognition domain of IL-6 is localized in the domain previously identified as site 3 of IL-6 (Somers, W., Stahl, M., and Seehra, J. S. (1997) EMBO J. 16, 989–997), already known to be involved in interactions with gp130.

These data did not take into account the possibility that IL-6 could be, as other interleukins (17), a bi-functional molecule having, beside a receptor-binding domain, a carbohydrate recognition domain (CRD). As demonstrated for interleukin 2 (IL-2; Ref. 18) in resting human lymphocytes (which do not express the alpha IL-2 receptor (IL-2R␣), when IL-2 is bound to its ␤-receptor (IL-2R␤) through its receptor-binding domain, it associates the later to the T cell receptor complex through its CRD. Indeed, in resting human T cells, IL-2 recognizes with a high affinity oligomannosidic N-glycans with 5 and 6 mannose residues found on one N-glycosylated form of CD3. The IL-2 signaling (tyrosine phosphorylation of IL-2R␤by p56 lck ) is entirely dependent upon this specific lectin/carbohydrate interaction, and interference with oligomannosides of pathogens could be responsible for severe immunodeficiencies (19,20). The same lectin activity was shown recently to be involved in the carbohydrate-dependent association between IL-2R␤ and IL-2R␣ in a mouse cell line constitutively expressing IL-2R␣ (21), a process resembling the second step in the activation process of human lymphocyte occurring after the internalization of the TCR complex (22) subsequent to the initial action of IL-2 on resting cells mentioned above.
By using this IL-6 high affinity ligand in cultures of resting human lymphocytes, it is demonstrated that the lectin activity of IL-6 is responsible for the IL-6-induced dephosphorylations of tyrosine-phosphorylated proteins. One major glycoprotein ligand of IL-6 having the HNK-1 epitope is gp130 itself. Docking experiments of the lower affinity conformers (determined by computational calculations) of the identified oligosaccharide ligands of IL-6 suggest that the CRD of IL-6 overlaps with site 3 of IL-6, previously identified as a site implicated in the association with gp130. These data propose new concepts for the interactions between the IL-6 receptor and its signal-transducing molecule gp130.

MATERIALS AND METHODS
Chemicals-The recombinant human IL-6 (produced in bacteria) and its polyclonal rabbit antibody were from Chemicon International Inc. (Temecula, CA). The monoclonal mouse anti-phosphotyrosine antibody was from Upstate Biotechnology (Lake Placid, NY). Anti-phosphoserine, anti-phosphothreonine, alkaline phosphatase-labeled anti-rabbit and anti-mouse IgG, normal goat serum, bovine serum albumin (BSA), RPMI 1640 culture medium, Brij97 detergent, phenylmethylsulfonyl fluoride, p-tosyl-arginine methyl ester, nitro blue tetrazolium, 5-bromo-4-chloro-3-indolyl phosphate, ␣ 2 -microglobulin, and leupeptin were from Sigma. The serum from the patient with IgM paraproteinemia was kindly provided by Dr. N. Baumann (Laboratoire de Neurobiologie, Hôpital de la Salpétrière, Paris). MAG was purified from adult rat brain myelin (46) according to Quarles and Pasnak (47). The P0 glycoprotein was prepared from rat sciatic nerve myelin according to Kitamura et al. (48). The HNK-1 oligosaccharide alditol ligands of IL-6 were obtained from the mucins of the eggs of R. temporaria (45) as their homologues (49,50). Nitrocellulose (0.45 m pore size) was from Schleicher & Schuell (Dassel, Germany). Periodate-treated BSA was obtained according to Glass et al. (Ref. 51; this treatment allows the elimination of interferences due to the presence of glycoconjugates in commercially available BSA preparations).
Isolation and Cultures of Human Lymphocytes-Peripheral blood lymphocytes of normal individuals were isolated by Ficoll-Paque centrifugation. After elimination of adherent cells on plastic for 30 min in RPMI 1640 containing 10% fetal calf serum, cells (2 ϫ 10 7 cells) were maintained at 37°C in a water-saturated atmosphere containing 5% CO 2 before treatment. These cells were treated as follows.
(i) For testing the effects of IL-6 and of IL-6 and its R. temporaria ligand on phosphorylations, cells were washed three times in serumfree RPMI 1640 and supplemented or not with IL-6 (20 ng/ml) or IL-6 plus its ligand (1 nM). After 15 min at 37°C, cells were harvested in ice-cooled phosphate-buffered saline, washed twice in ice-cold phosphate-buffered saline, boiled for 5 min in the Laemmli dissociating buffer (52), and analyzed by SDS-PAGE. After transfer onto nitrocellulose (53), the blots were revealed using anti-phosphotyrosine, antiphosphoserine, or anti-phosphothreonine monoclonal antibodies, followed by AKP-labeled rabbit anti-mouse IgG and nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate staining (54).
(ii) For testing the presence of HNK-1 epitope on gp130, 10 7 cells were washed in phosphate-buffered saline and lysed at 4°C in 750 l of 10 mM Tris-HCl buffer at pH 7.4 containing 150 mM NaCl, 5 mM EDTA, 50 g/ml periodate-treated BSA, and 1% (v/v) Brij97 (55) supplemented with protease inhibitors (10 mM phenylmethylsulfonyl fluoride, 10 mM p-tosyl-arginine methyl ester, 20 g/ml microglobulin, and 10 M leupeptin). After elimination of the nuclei by centrifugation (12,500 ϫ g for 15 min at 4°C), the supernatants were diluted once with an equal volume of 10 mM Tris-HCl buffer containing 150 mM NaCl, 1% Triton X-100, and 1% deoxycholate and immunoprecipitated by addition of a 1/1000 dilution of the serum of a patient with paraproteinemia (overnight at 4°C). After centrifugation, the pellet was centrifuged and washed ten times by centrifugation (10 min at 12,000 rpm) with the immunoprecipitation buffer and finally with phosphate-buffered saline. The final pellet was boiled in the Laemmli dissociating buffer and submitted to 10% SDS-PAGE, followed by blotting on nitrocellulose. Blots were revealed using the mouse monoclonal anti-gp130 antibody, followed by an AKP-labeled anti-mouse IgG antibody. Controls were performed using normal human serum instead of the pathological serum.
In another series of experiments, resting human lymphocytes and HepG2 cells were lysed as above and immunoprecipitated using the serum of the patient with IgM paraproteinemia. After extensive washing in the lysis buffer, the immunoprecipitates were submitted to SDS-PAGE and silver stained.
Computational Procedures-The conformations of two different oligosaccharide ligands of IL-6 were studied (see "Results"): the N-glycan glycoprotein ligand found on P0 and MAG and the higher affinity ligand of the eggs of R. temporaria oligosaccharide alditol. For these two molecules, the random search technique was used as implemented in the Sybyl version 6.6 software to sample the conformational space (Tripos, Inc., St. Louis, MO (www.tripos.com)). The OPLS All-Atom Force Field (56) was used except for the sulfated monosaccharide (3-sulfated GlcA) for which a charge determination was performed using quantum chemical procedures available in the Jaguar software (Jaguar Version 4.0; Schrödinger, Inc., Portland, OR 97204 (www.schrodinger.com)). The Density Functional Theory was used with the B3LYP hybrid method and the 6-31G basis set. The molecular electrostatic potential was computed on a spherical grid and fit to a set of point charges located at the atomic centers. The fit was constrained to also reproduce the dipole moment. Fourteen torsional angles of rotatable bonds were changed for the glycoprotein ligand (compound IV in Fig. 1B), 14 and 17 for compounds I and II, respectively, which consist of the glycosidic linkages and the CH 2 OH, NHCOCH 3 , COOH, and SO 3 groups. In both cases, 20,000,000 configurations were examined using the bump checking (with a Van der Waals scaling factor of 0.7) and energy as criteria. For compound IV, 40311 conformers were found, among which the 10 of lowest energy (in a 6 kcal/mol range) were retained for docking with human IL-6). For compound II only 149 conformers were found due to its low flexibility as compared with compound IV. Again the 10 lowest in energy were kept for docking (in a 10 kcal/mol range). The docking with IL-6 was performed using the SO 4 2Ϫ sulfate group found in the crystal as a template (9). A random search was performed after replacing the corresponding sulfate in the IL-6 crystal by the oligosaccharidebound sulfate group. No new conformers were found.

The IL-6 Lectin Activity Is Responsible for Early Tyrosine
Dephosphorylations of Quiescent Human Lymphocytes-To examine the function of the lectin activity of IL-6 for glycans having the HNK-1 epitope, a population of quiescent human lymphocyte was incubated with IL-6 for 15 min at 37°C, and the phosphorylation pattern was analyzed using anti-phosphotyrosine, anti-phosphoserine, and anti-phosphothreonine monoclonal antibodies. As shown in Fig. 1A, lane 2, the IL-6treated material showed a significant decrease of phosphotyrosine residues of tyrosine-phosphorylated proteins of M r 77,000, 73,000, 69,000, 60,000, and 56,000, respectively, whereas not significant variations were observed using the two other antibodies (not shown). When the samples containing IL-6 were co-incubated with IL-6 ligands isolated from the mucins of the eggs of R. temporaria (Fig. 1B), the decrease in tyrosine phosphorylation was suppressed, and the profile of the tyrosine-phosphorylated proteins was identical to the profiles of untreated cells (Fig. 1A, lanes 3 and 4). This indicated that, at low concentrations (1 ng/ml for compound II, i.e. ϳ1 nM and 0.8 g/ml, i.e. ϳ1 M for compound I), the IL-6 oligosaccharide ligands were able to inhibit the earliest dephosphorylations already observed (57) in the IL-6-dependent signaling. In contrast, compound III was ineffective at the concentration of 10 M.
gp130 (IL-6R␤) Has a HNK-1-containing Glycan-The previous data suggested that the lectin activity of IL-6 was of fundamental importance for the IL-6-dependent association between IL-6R␣ and its signal-transducing molecule gp130. Therefore it was necessary to determine whether the HNK-1 epitope was present in molecules of the gp130 surface complex, especially on gp130. This point was analyzed in the following way. Resting human lymphocytes (not supplemented with IL-6) were lysed in a mild detergent (55), supplemented with a stronger detergent mixture (known to dissociate surface molecular complexes (55)), and then immunoprecipitated using a serum from a human patient with IgM paraproteinemia (43). The immunoprecipitated material was submitted to SDS-PAGE, blotted on nitrocellulose, and the blots were revealed using a mouse monoclonal anti-gp130 antibody followed by an AKP-labeled anti-mouse IgG antibody. As shown in Fig. 2B, the anti-HNK-1 immunoprecipitate actually contained a high M r band of 131,000 specifically revealed using the anti-gp130 antibody. This demonstrated that at least one of the molecules bearing the HNK-1 epitope in the gp130 molecular complex was gp130 itself.
In fact, silver staining of the SDS gels performed in reducing conditions of the immunoprecipitated material indicated that the number of subunits (besides heavy and light chains of IgM) was extremely reduced. Indeed, four protein bands were detected with M r of 57,000 63,000, 85,000 and 130,000, respectively (Fig. 2D). The latter, migrating as gp130, was the more abundant of the subunits immunoprecipitated using the patient serum. These data were totally different from those obtained with the HepG2 cell line (Fig. 2C), in which more than 20 protein subunits were immunoprecipitated, all being expressed at a higher level when expressed per mg of total cell protein.
Computational Calculations of the IL-6 Ligand Conformations-The previous experiments demonstrated that: (i) the lectin activity of IL-6 was essential for its biological function; (ii) this function was the extracellular association between the IL-6R␣ complex and the gp130-containing complex, HNK-1containing glycan being present on gp130 itself. Therefore, it was of interest to make a theoretical model of the interaction between IL-6 and its ligands. The compound II isolated from R. temporaria egg mucins was the higher affinity ligand so far identified. Consequently, it was considered as the reference structure for the determination of the conformation of IL-6 oligosaccharide ligands. Because IL-6 bound to the P0 and MAG glycoproteins possessing the HNK-1 epitope on specific N-glycans of known structures (24), the extended HNK-1 part of this compound (compound IV in Fig. 1B) was taken as the putative endogenous glycoprotein ligand of IL-6. Based on studies (17) of different O-glycans isolated from the eggs of Rana arvalis (49) and of R. temporaria (45), it was clear that the 3-O-SO 3 H group on GlcA was a fundamental determinant of the interaction between IL-6 and its ligand. The presence of an additional residue (Gal and/or more complex sequences) on the Gal␤1,4 of compound II (compound III in Fig. 1B) strongly suppressed the affinity for IL-6. The same was observed when the Gal␤1,4 of compound II was replaced by a Gal␤1,3. The absence of the Fuc␣1,2 residue linked to the Gal␤1,3 of compound II (compound I) decreased the affinity by a factor of about 10 3 . Therefore, it was suggested that these two compounds possessed in common a special conformation allowing the specific interaction with a domain of IL-6. One possibility was that, beside the common nonreducing carbohydrate sequence SO 3 H-3-GlcA␤1,3Gal␤1-, these compounds had a hydrophobic residue (methyl group of Fuc and acetamido group of GlcNAc, respectively) increasing the affinity for IL-6. Consequently, two key structures were chosen for computational conformational analysis: compound II and the compound IV reduced to the Man␣1,6 branch (Fig. 1B).
The representations of the lower energy conformations of the R. temporaria ligand, compound II (Fig. 3), showed that they corresponded to very rigid structures. The conformation shown in Fig. 3A was significantly different from all other conformations, since its energy was 7 kCal/mol lower than the other  lower energy conformation shown in Fig. 3B. It presented a very rigid and compact tripod-like structure, contrasting with the other low energy conformation showing a more extended structure. The glucuronic acid 3-sulfate motif was rotated by about 90°relative to the previous compound, whereas the acetamido group of GalNAc-ol was rotated by 180°relative to the previous compound. The two lowest energy conformations of the glycoprotein HNK-1 epitope (portion of compound IV) were found to be very similar (Fig. 3, C and D) and presented very close energetic levels. These two structures presented a great similarity with the lowest energy conformer of R. temporaria for the motif SO 3 H-3-GlcA␤1,3Gal␤1-. This was evidenced in Fig. 4, where these groups were placed in similar positions. Interestingly, in all these compounds, the methyl groups were in opposition to the sulfate group and at different distances from the sulfate group in the R. temporaria and in the glycoprotein ligand of IL-6. This suggested that the hydrophobic residues were not involved directly in the binding to IL-6. Rather, these experiments suggested that the important domain for the interaction with IL-6 was the conformation of the SO 3 H-3-GlcA␤1,3Gal␤1 motif. This was perfectly compatible with the fact that the addition of a residue of Gal␤ on the hydroxyl in position 4 of Gal inhibited the interaction with IL-6. It might be also understood why that the absence of the Fuc residue in compound II (compound I) dramatically decreased the binding of IL-6, modifying the conformation of the SO 3 H-3-GlcA␤1,3Gal␤1 motif of interaction with IL-6. Indeed, the lower energy conformers of compound I showed different conformations as compared with those of compounds II and IV. Especially, the Gal residue vicinal to the GlcA residue of compound I was turned by 180°C compared with that of the previous compounds, explaining its reduced affinity for IL-6 compared with compound II.
Docking of Oligosaccharide Ligands into the IL-6 Carbohydrate Recognition Domain-The biochemical, crystallographic, and computational data obtained on the ligands of IL-6 (Refs. 7, 9, and 15 and this work) allowed us to determine important features for the putative carbohydrate recognition domain of IL-6. Indeed, the three-dimensional structure of IL-6 obtained by x-ray crystallography showed 4 sulfate groups constituting putative binding sites for the sulfate group of the HNK-1 common to all IL-6 ligands. One of these sulfate groups (SO 4 2Ϫ ) was strongly hydrogen-bonded to Gln 156 , suggesting that this position could be a privileged site for the binding of the sulfate group of the HNK-1 epitope. Furthermore, this sulfate-binding site was close to a water-exposed tryptophane residue (Trp 157 ), an amino acid frequently encountered in the CRD of lectins (58), and especially in the CRD of calcium-independent lectins (59). Therefore, the glycoprotein HNK-1 (portion of compound IV) conformer was computationally docked into the IL-6 molecule, placing the sulfate group of HNK-1 at the position of SO 4 2Ϫ . Computational analysis of the lower energy structure of the complex allowed placing the HNK-1 glycan as shown in Fig. 5. In this position, the sulfated glycan appeared to be specifically hydrogen-bonded to the protein. Indeed, besides three hydrogen bonds involving SO 4 2Ϫ with Gln 156 , the calculations of the interatomic distances indicated the possibility of two strong hydrogen bonds between the nitrogen atom of the side chain of Asn 155 and the hydroxyl groups of the C-4 and C-6 carbon atoms of the galactose residue part of the SO 3 H-3-GlcA␤1,3Gal␤ motif (3.08 and 2.86 Å, respectively). A third weak hydrogen bond (3.43 Å) could be formed between the oxygen atom of the amido group of Asn 155 and the oxygen atom of the C-2 carbon atom of GlcA. Furthermore, the two other pyranic rings of the HNK-1 ligand (GlcNAc␤1,4 and Man␣1,6) were found in the vicinity (mean distance 5 and 4.5 Å for GlcNAc and Man, respectively) from the indolic cycle of Trp 157 , a situation allowing strong interactions between the pyranic rings and Trp. Additional hydrogen bonds could be formed between the oxygen atoms of the C-4 of Gal and C-2 of GlcA and a water molecule (H 2 O 57 (9)), itself hydrogen-bonded to the nitrogen atom of the acetamido group of Asn 155 .

Importance of the Lectin Activity of IL-6 for Its Biological
Function-The signaling pathway of IL-6 involves the association of the molecular complex containing its receptor (IL-6R␣) to the surface molecular complex including the signal-transducing gp130 glycoprotein (IL6-R␤). Several studies (60 -63) demonstrated that this IL-6-induced association between the two types of molecular complexes initiated a reduction of specific phosphotyrosine residues of several proteins. This was due to a specific phosphatase, SHP-2, endowed with two binding sites for phosphotyrosine residues (62, 64 -67) associated with the gp130 complex. In a previous paper, we demonstrated (17) that IL-6 has, as several other cytokines, a calcium-independent lectin activity. Indeed, IL-6 recognizes glycans having the HNK-1 epitope as a major determinant. Since this property was never detected before, the question was asked to find out FIG. 3. Results of molecular modeling of the lower energy conformations of the high affinity ligand of IL-6 (compound II in Fig. 1B) isolated from the R. temporaria mucins (A and B) and of the portion (compound IV in Fig. 1B)  the biological role of the lectin activity of IL-6. The present study demonstrates that the lectin activity of IL-6 is essential for the first steps of its signaling pathways. Indeed, the addition of minute amounts (ϳ1 nM) of the high affinity oligosaccharide ligand (compound II) completely inhibited the early dephosphorylations of phosphotyrosine residues induced by IL-6 on resting human lymphocytes. The inhibition of tyrosine dephosphorylations was dose-dependent, and the effect of the different oligosaccharides from R. temporaria on these tyrosine dephosphorylations was identical to their ability to inhibit the binding of IL-6 to MAG and P0 myelin glycoproteins. Indeed, compound I showed in both conditions only 0.1% of the efficiency of compound II, whereas compound III was totally inactive at the concentration of 10 M. The parallelism of action of the different oligosaccharides in vitro and in vivo indicated that the inhibitory effect on tyrosine dephosphorylations was due to the lectin activity of IL-6 for specific oligosaccharides having the HNK-1 epitope.
It was previously shown (7-9) that the fixation of IL-6 to IL-6R␣ provoked the association with the gp130 (IL-6R␤), the signal-transducing molecule of the IL-6 system. One way to explain the effect of the HNK-1 oligosaccharides was that gp130, or one molecule of the gp130 complex, possessed one HNK-1 containing glycan. Although it cannot be excluded that other molecules of the gp130 complex have the HNK-1 epitope, the immunoprecipitation experiments indicate that gp130 it-self possesses this epitope. Furthermore, silver staining of the immunoprecipitated material indicated that gp130 is the major glycoprotein of resting human lymphocytes possessing the HNK-1 epitope. Therefore, the inhibition of the IL-6 signaling by HNK-1 oligosaccharides can be understood as an inhibition of the carbohydrate-dependent association between IL-6R␣ and one glycan of the signaling molecule gp130.
This suggests that IL-6 behaves as a bi-functional molecule endowed with a receptor-binding domain and a carbohydrate recognition domain and able to make a bridge between its receptors and a glycan of its signal-transducing molecule gp130 (this mechanism is schematized in Fig. 6). Although the consequences on the early tyrosine phosphorylations mechanisms on resting human lymphocytes of the association are different from that observed for IL-2 (dephosphorylation for IL-6 instead of tyrosine phosphorylations; Ref. 18), the lectin activity of these two cytokines is necessary for the specific associations between the interleukin receptor and the signal-transducing complex. Based on the discovery of different lectin activities of several cytokines, these observations could lead to the definition of a general mechanism of action of cytokines on resting and/or normal human lymphocytes in the specific association of their receptors to signal-transducing molecules.
The data observed for the immunoprecipitates of the cancer cell line HepG2, in which more than 20 different immunoprecipitated subunits were detected at similar levels as gp130, might explain the extreme responsiveness of these cells to IL-6. However (68,69), these cells did use other signaling systems not involving gp130. The overexpression of the HNK-1 epitope in HepG2 cells might result in a polysemous signaling.
Putative Localization of the CRD of IL-6 -Because of the importance of the biological function of the lectin activity of IL-6, we decided to go further into the definition of the conformation of its oligosaccharide ligands and of its carbohydrate recognition domain. IL-6 belongs to a family of cytokines, including interleukin-11, the ciliary neurotrophic factor, the oncostatin M, and the cardiotrophin-1, sharing similar features in their structural organization (9). Interleukin-6 is a protein of 184 amino acid residues long containing four ␣-helixes organized in a classical four-helix bundle and a supplementary mini-helix E located in the CD loop (66). Studies based on site-directed mutagenesis (7,15) and/or crystal analysis (9) were performed on IL-6 and led to the design of three sites of crucial importance for the binding of the cytokine to IL-6R␣, on the one hand, and to gp130, on the other hand.
Site 1 is located on the C-terminal part of the D-helix of IL-6 and is implicated in the binding of IL-6 to its ␣-receptor. molecules. Site 2 consists of amino acid residues in the A and C helices, whereas site 3 is located in the terminal part of the CD loop and the N-terminal part of the D-helix. Two types of mutants bearing point mutations at sites 2 and 3 were generated (15); the first group possessed fourth amino acid substitutions at site 2 (Y31D/G35F/S118R/V121D). The following point mutations, W157R/D160R and T162D, designed the site 3 variants. Interestingly, combined site 2 and 3 mutants normally bound to the IL-6R␣, but have no biological activity, failing to bind gp130. Site 2 or 3 mutants, although they were able to bind one single molecule of gp130 in the presence of the IL6R␣, also lack to transduce the intracellular signal generated by the cytokine, similarly unable to form a gp130 dimer. These observations led the authors to identify two independent binding sites for gp130 on the IL-6 molecule (9,16,70).
Our data propose site 3 of IL-6 as its carbohydrate recognition domain. Indeed, it possesses a sulfate-binding site, the sulfate group being strongly attached to Gln 156 through three hydrogen bonds (the third hydrogen bond involving a protonated form of sulfate is likely at neutral pH because of the partial ionization of this relatively weak acidic group). It possesses a water-exposed Trp 157 residue included in a cavity, the function remained unknown, although its replacement by other amino acids eliminated the IL-6-dependent signaling. Such a role of Trp residues in the CRD of lectins is a common feature (58,59), the indolic ring allowing strong interactions with the pyranic rings of monosaccharides. Furthermore, two strong hydrogen bonds can be formed between the hydroxyl groups of C-6 and C-4 carbon atoms of Gal substituted in position 3 by SO 3 H-3-GlcA. These interactions could be stabilized by a third hydrogen bond involving the hydroxyl group of the C-2 carbon atom of GlcA. Based on the use of different glycans isolated from R. temporaria and R. arvalis, the binding to IL-6 requires a strict conformation of the SO 3 H-3-GlcA␤1,3Gal␤1 motif, corresponding to the lowest energy conformers calculated both for the R. temporaria and the glycoprotein HNK-1 epitope. This conformation is not dependent upon hydrophobic interactions involving methyl groups of the ligands, these methyl groups being likely important for directing the conformation of the ligand. Such a specificity of hydrogen bonds in determining the binding or not of a mono-or oligosaccharide and in changing the carbohydrate binding properties of lectins has been demonstrated previously by molecular engineering (58,71). In this way, the present data reinforce the concepts based on the previous knowledge on the mechanisms and on the determination of the carbohydrate specificities of protein-sugar interactions. Based on the results of docking experiments (Fig. 5), the lower energy conformation of the glycoprotein HNK-1 group showed that the semi-acetalic group of the C-1 carbon atom of Man is turned outside the IL-6 structure, indicating that the interaction observed with a portion of the HNK-1 containing N-glycan will still occur with a complete glycan because of the absence of steric hindrance of the of the more internal parts of the glycan and of the protein to which it is attached (gp130). Furthermore, the data of the docking experiments explained why compound III is not a ligand, and compound I is a poor ligand, of IL-6. Therefore, it may be hypothesized that site 3 of IL-6, defined as a site of interaction with gp130, corresponds to the carbohydrate recognition domain of IL-6.
The question remains as to how these data can fit with previous studies (biochemical and structural) and, especially, with those obtained with recombinant soluble forms of IL-6 receptors (72). Indeed, in the presence of a soluble form of gp130, the complex IL-6⅐IL-6R␣, which is present on the form of a heterodimer, is transformed into a hexamer. This is likely due to the fact that the soluble gp130 associates into dimers spontaneously and can bind two IL-6⅐IL-6R␣ complexes. The question remained to explain how the lectin activity could participate to the formation of these hexamers. The recombinant soluble form of gp130 used in the previous studies for detecting the formation of hexamers (73)(74)(75)(76) was produced in Chinese hamster ovary cells. This point seems of importance, since these cells are (as HepG2 cancer cell types) competent for the synthesis of the HNK-1 epitope. Therefore, it may be suggested that the formation of the hexamer is directed by the presence of one N-glycan possessing the HNK-1 epitope on each gp130 molecule, i.e. two glycans endowed with the HNK-1 epitope per dimer of soluble gp130. This allows the fixation of two IL-6 molecules bound to IL-6R␣ and, consequently, the formation of the hexamer comprising two gp130, two IL-6 and two IL-6R␣ molecules. The localization of the CRD of IL-6 in site 3 of the molecule fits with the theoretical model of the formation of the hexamer proposed by Somers et al. (9). The recombinant soluble form of gp130 used in the previous studies possessed three potential N-glycosylation sites (Asn 13 , Asn 39 , and Asn 109 ) susceptible to having the HNK-1 epitope. Based on the stoichiometry of the molecules in the hexamers (74), it is suggested that only one of these N-glycosylation sites has the HNK-1 epitope (otherwise the stoichiometry would have been changed to higher order of associations). However, it is not certain that one of these three N-glycosylation sites found on the soluble recombinant form of gp130 is actually the same as that bearing the HNK-1 epitope on the complete gp130 produced by quiescent human lymphocytes. Indeed, there is a general consensus that membrane anchoring plays important roles in the pattern of N-glycosylation of glycoproteins. Therefore, it may be that, in human lymphocytes, the HNK-1 epitope could be built on N-glycans present on more internal parts of the complete gp130. Although the precise signals in the polypeptide chain necessary for the synthesis of the HNK-1 epitope are not known, it may be that, if such signal sequence exist, it would be the same in the soluble gp130 and in the total membrane-bound gp130. Therefore, it is possible that the HNK-1 bearing N-glycan would be the same in the two types of molecules, a point that could be easily solved by site-directed mutagenesis of the individual three potential N-glycosylation sites of the soluble gp130. A question remains to know how other cytokines, which use the same gp130 signal-transducing molecule, can associate their receptors to gp130. Based on the three-dimensional structures, it is unlikely that they recognize also the HNK-1 epitope, although they could be able to associate their receptor with gp130 through other specific lectin activities. The high level of N-glycosylation of gp130 may be a key for understanding different carbohydrate-dependent association with the cytokine receptors.
Therefore, this study proposes new concepts in the mechanism of the IL-6 function and structure. The involvement of the HNK-1 epitope in the function of IL-6 may be of importance for pathology, since HNK-1 is especially expressed in the nervous tissue and more specifically expressed in myelinating cells in adult brain (28,29,39). This overexpression of the HNK-1 epitope could be a clue for understanding demyelinating diseases. For example, the knockout of the IL-6 gene in mice suppresses the experimental allergic encephalomyelitis and its associated demyelination observed in specific mouse strains sensitive to experimental allergic encephalomyelitis (77)(78)(79).