Disulfide Bond Structure and N-Glycosylation Sites of the Extracellular Domain of the Human Interleukin-6 Receptor*

The high affinity interleukin-6 (IL-6) receptor is a hexameric complex consisting of two molecules each of IL-6, IL-6 receptor (IL-6R), and the high affinity converter and signaling molecule, gp130. The extracellular “soluble” part of the IL-6R (sIL-6R) consists of three domains: an amino-terminal Ig-like domain and two fibronectin-type III (FN III) domains. The two FN III domains comprise the cytokine-binding domain defined by a set of 4 conserved cysteine residues and a WSXWS sequence motif. Here, we have determined the disulfide structure of the human sIL-6R by peptide mapping in the absence and presence of reducing agent. Mass spectrometric analysis of these peptides revealed four disulfide bonds and two free cysteines. The disulfides Cys102-Cys113 and Cys146-Cys157 are consistent with known cytokine-binding domain motifs, and Cys28-Cys77with known Ig superfamily domains. An unusual cysteine connectivity between Cys6-Cys174, which links the Ig-like and NH2-terminal FN III domains causing them to fold back onto each other, has not previously been observed among cytokine receptors. The two free cysteines (Cys192 and Cys258) were detected as cysteinyl-cysteines, although a small proportion of Cys258 was reactive with the alkylating agent 4-vinylpyridine. Of the four potentialN-glycosylation sites, carbohydrate moieties were identified on Asn36, Asn74, and Asn202, but not on Asn226.

Interleukin-6 (IL-6) 1 is a multifunctional cytokine that plays a central role in host defense due to its wide range of immune and hematopoietic activities, as well as its potent ability to induce the acute phase response (1)(2)(3). Since overexpression of IL-6 has been implicated in the pathology of a number of diseases (for reviews, see Refs. 4 and 5), it is anticipated that selective antagonists of IL-6 action may offer therapeutic benefits in the treatment of IL-6-related diseases.
The biological activities of IL-6 are mediated by the IL-6 receptor system which comprises two receptor proteins: the specific ligand-binding ␣-subunit receptor (IL-6R) and the signal transducing ␤-subunit, gp130. Gp130 also forms part of the receptor complexes of leukemia inhibitory factor, ciliary neurotrophic factor, oncostatin M, cardiotrophin-1, and IL-11 (6) which, in part, provides a molecular basis for the functional redundancy of these cytokines. IL-6 first binds the IL-6R with an affinity of ϳ1 nM and the IL-6⅐IL-6R complex then binds gp130 with a resulting affinity of ϳ10 pM (7). The ternary complex of the IL-6 receptor system is a hexamer, comprising two molecules each of IL-6, IL-6R, and gp130 (8,9).
The cDNA of the human IL-6R encodes a protein of 468 amino acids (7), including a signal peptide of 19 amino acids, an extracellular region of 339 amino acids, a transmembrane domain of 28 amino acids, and a short cytoplasmic domain of 82 amino acids. This sequence shows 54 and 57% overall amino acid identity with the cDNA sequences for mouse (10) and rat (11) IL-6R, respectively. The mature 80-kDa IL-6R is a glycosylated form of the predicted 50-kDa precursor (12) and contains six potential N-linked glycosylation sites. The extracellular region has a modular structure, consisting of three domains of approximately 100 amino acids. The amino acid sequence of the NH 2 -terminal domain is characteristic of the immunoglobulin superfamily (Ig-like) (13,14). Members of this family share a common ␤-sheet folding topology called a Greek Key (15), whereby neighboring ␤-strands form hydrogen bonds in an anti-parallel fashion to form a ␤-pleated sheet. Two ␤-sheets are then packed against each other to produce a hydrophobic core. Similarly, the two COOH-terminal domains of the IL-6R are classified as fibronectin type III-like (FN III) modules, a subclass of the ␤-sandwich fold (14). The topology of these domains is similar to those of Ig-like modules, with the notable exception of the "sheet switching" of ␤-strand D from the first ␤-sheet of an Ig-like domain to form ␤-strand CЈ on the second ␤-sheet of FN III domains. Together, the two FN III domains form a cytokine-binding domain (CBD) which is characteristic of class I cytokine receptors (16) (e.g. receptors for interleukins-3, -5, -6, -11, gp130, erythropoietin (EPO), ciliary neurotrophic factor, granulocyte-colony stimulating factor, growth hormone (GH), and prolactin (PRL)). Generally, these receptors are characterized by two conserved disulfide bonds located in the NH 2 -terminal FN III domain and a conserved WSXWS motif located in the COOH-terminal FN III domain.
The cytoplasmic and transmembrane domains of the IL-6R are not required for IL-6 signaling (17) and biologically active soluble forms of IL-6R (sIL-6R) are naturally found in low concentrations in human urine (18) and serum (19,20) of healthy individuals. In contrast to many other soluble cytokine receptors that act as inhibitors by competing for ligand binding with cellular receptors (e.g. tumor necrosis factor, IL-1, -2, -4, interferon-␥, nerve growth factor, leukemia inhibitory factor, granulocyte-stimulating factor and granulocyte macrophagecolony stimulating factor) (for reviews, see Refs. 21 and 22), the sIL-6R acts as an agonist of IL-6 activity (17). It is not clear whether sIL-6R is generated by proteolytic shedding of membrane-bound IL-6R (23), or from an alternatively spliced mRNA species (24,25), or both. In certain disease states, for example, patients with human immunodeficiency virus infection or multiple myeloma, increased levels of sIL-6R have been reported (26,27). Therefore, inhibition of the IL-6⅐sIL-6R complex has been labeled as a key target to antagonize the in vivo action of IL-6 (28).
To elucidate the tertiary structure of the IL-6R extracellular region, we have purified human sIL-6R using a Chinese hamster ovary (CHO) cell expression system (29,30). This form of the sIL-6R contains four potential N-linked glycosylation sites and 10 cysteine residues. Three cysteines are located in the Ig-like domain, six in the NH 2 -terminal FN III domain, and one in the COOH-terminal FN III domain. Previously, we have shown that the ligand affinity purified sIL-6R bound IL-6 and gp130 with a 2:2:2 stoichiometry of IL-6, sIL-6R, and sgp130 (8) and was bioactive as determined by the ability of the IL-6⅐sIL-6R complex to prevent the differentiation of embryonic stem cells (31). Here, reversed-phase HPLC peptide mapping under nonreducing and reducing conditions, in combination with mass spectrometric and NH 2 -terminal sequence analysis, was used to determine the disulfide structure and carbohydrate attachment sites of sIL-6R. On the basis of these results, we have created a model that depicts the topology of the extracellular region of the IL-6R and predicts its interactions with IL-6 and gp130.
Purification of the Extracellular or "Soluble" Domain of the Human IL-6 Receptor (sIL-6R)-sIL-6R was purified from the conditioned medium of CHO cells transfected with an expression vector (pECEdhfr344) which encodes the extracellular binding domain of the IL-6R (truncated at residue 345) (29). The sIL-6R was concentrated from CHO cell conditioned media using a Sartocon Miniapparatus (Sartorius, Goettingen, Germany) equipped with a 30,000 molecular weight cut-off membrane and purified by ligand affinity chromatography using an IL-6-Sepharose column (30).
Labeling of Free Cysteine Residues-Prior to disulfide determination, free cysteine residues in the sIL-6R (200 g) were treated with a 5-fold molar excess of 4-vinylpyridine in 50 mM Tris-HCl buffer, pH 8.4, for 1 h at 25°C in the dark. The modified protein was purified by gel permeation chromatography on a 100 ϫ 10-mm inner diameter fast desalting column (G-25 Sephadex, Pharmacia).
Ion-trap Mass Spectrometry-On-line MS analysis of peptide fractions was performed on a Finnigan-MAT LCQ quadrupole ion trap mass spectrometer equipped with an ESI source (San Jose, CA). "Triple play" experiments, consisting of MS/zoom scan/and MS/MS, were performed as described elsewhere (37). Source CID/single ion monitoring (sCID/ SIM) was employed to identify S-pyridylethyl cysteine-containing peptides (38). For sCID/SIM, the relative collision energy in the source region was set at 70% (arbitrary value) and the mass range was scanned from m/z 104.5-107.5 to detect S-pyridylethyl fragment ions (m/z 106). Peptides were identified using the Finnigan PEPMAP program, and from their CID product ion spectra using the MS-Tag and MS-Product algorithms (Prospector pacific rim mirror site, http://jpsl. ludwig.edu.au).
Homology Modeling-The Ig-like domain of the IL-6R was modeled using the structure of the mouse monoclonal antibody FAB D44.1 V L domain (39) which showed 23% sequence identity, the highest for all known Ig three-dimensional structures. The coordinates for the template were taken from the Protein Data Bank (40), entry 1MLB (chain A, residues 1-109). Template structures for the CBD of IL-6R were the CBD of gp130 (41), Protein Data Bank entry 1BQU; GHR, chain B (first binding receptor) of Protein Data Bank entry 3HHR (42); EPOR, Protein Data Bank entry 1EBP (43); and PRLR, Protein Data Bank entry 1BP3 (44). The sequence alignment was prepared in two parts. The Ig-like domain of IL-6R was manually aligned with the template FAB D44.1 V L structure. For the CBD, a structure-based multiple sequence alignment was performed manually. The ␤-sheets of the known CBD structures were superimposed to provide the basis of the alignment. The remaining sequences of unknown structure were manually aligned with these structures, conserving the disulfide patterns, WSXWS motif, and hydrophobic patterns of the ␤-sheets.
The MODELLER program (45) was used to generate separate models of the Ig-like domain and CBD. The quality of the models was assessed as described previously (46); in particular using the ProsaII program (47). MODELLER was also used to determine the relative orientations of the Ig-like domain and CBD. A disulfide restraint between Cys 6 and Cys 174 of the Ig-like domain and FN III domain, respectively, was introduced in accordance with our experimental results. Fifty models were generated with a range of orientations between the Ig-like domain and the CBD. The final two models were chosen on the basis of the quality checks described above and agreement with experimental data. The model of the sIL-6R, complexed with the crystal structures of IL-6 (48) (Protein Data Bank entry 1ALU) and the gp130 CBD (41), was constructed by superimposing these moieties over the human GH receptor complex (42).

Initial Characterization of Recombinant sIL-6
Receptor-NH 2 -terminal sequence analysis of the first 20 residues of the purified sIL-6R was in agreement with the published sequence (7) (data not shown). The purified sIL-6R yielded a single broad band on SDS-PAGE with an apparent molecular mass of ϳ52,000 (Fig. 1A, lane 1), a value significantly higher than 36,368 Da calculated from the amino acid composition (7). Upon treatment with either neuraminidase or a combination of neuraminidase and an endoglycosidase mixture, the M r of the sIL-6R was reduced to ϳ50,000 and ϳ40,000, respectively (Fig.  1A, lanes 2 and 3). These data suggest that the increased M r of sIL-6R (ϳ12,000) is due to glycosylation of the CHO cell-derived protein. Pronounced charge heterogeneity of the mature sIL-6R was observed upon IEF (Fig. 1B, lane 1). Treatment with neuraminidase (Fig. 1B, lane 2) and neuraminidase plus an endoglycosidase mixture (Fig. 1B, lane 3) reduced this complexity to one or two major bands, respectively, indicating that the heterogeneity of the sIL-6R preparation is primarily due to differential N-linked glycosylation.
Labeling of Free Cysteine Residues-Free cysteine residues in sIL-6R (ϳ200 g) were modified with 4-vinylpridine at pH 8.5. Following enzymatic deglycosylation, the treated sIL-6R was digested with trypsin at pH 6.0, and subjected to RP-HPLC/ESI-IT-MS analysis as described under "Experimental Procedures." The total ion current profiles of nonreduced and reduced tryptic digest of sIL-6R are shown in Fig. 2 (panels A  and B, respectively). sCID/SIM of the nonreduced digest for S-pyridylethyl ions (m/z 106) revealed a single peak at retention time 31.42 min (Fig. 2C). The mass spectrum of the peak at retention time 31.42 min (Fig. 2D) Table I), consistent with cysteinylation of Cys 258 .
Sequence data of peptide T1b was not available due to the low abundance of this ion (ϳ20% compared with peptide T1a). However, the 14.8-Da difference observed between peptides T1a and T1b is consistent with the difference in mass between cysteinylated (ϩ119 Da) and S-pyridylethylated (ϩ105 Da) Cys 258 of peptide Asp 253 -Arg 268 . These data suggest that the peak observed in the sCID/SIM profile (Fig. 2C) emanates from an S-pyridylethylated form (Cys 258 ) of peptide T1b. Peptide T1b was also observed (in low abundance) at approximately the same retention time in the TCEP-reduced total ion current profile (Fig. 2B). Taken together, these data indicate that the majority of sIL-6R purified from CHO cell conditioned medium contain a modified Cys 258 (cysteinylated), and only a small portion (ϳ20%) remains as the unmodified Cys 258 (free sulfhydryl).
Determination of Disulfide Linkages of sIL-6R by Tryptic Peptide Mapping-A portion (25%) of the tryptic digest of deglycosylated/S-pyridylethylated sIL-6R was reduced with 10 mM TCEP at pH 6.0 and subjected to on-line RP-HPLC/ESI-IT-MS analysis (Fig. 2B) using the same chromatographic conditions described for the nonreduced digest. Inspection of the nonreduced and reduced tryptic peptide maps of sIL-6R (Fig. 2, panels A and B) revealed that upon reduction of the digest, the retention times of five peptide fractions (T1-T5) in the nonreduced tryptic map ( Fig. 2A) changed with the concomitant appearance of nine peptide fractions (T6-T14) in the reduced tryptic map (Fig. 2B). A summary of peptide masses found in fractions T1-T14 is shown in Table I.
As mentioned above, MS analysis of peptide fraction T1 revealed tryptic peptide Asp 253 -Arg 268 (Fig. 2, A, D, and E) containing a cysteinyl-cysteine at position 258 in the sequence (T1a), as well as the small proportion of the peptide that had  Table I) was also confirmed by MS/MS analysis (data not shown). Modification of Cys 192 with 4-vinylpyridine was not observed.
Inspection of the nonreduced and reduced total ion current profiles for the tryptic digest of sIL-6R (Fig. 2) indicated that peptide fraction T3 (6544.0 Da, Fig. 2A, Table I) was comprised of three peptides (T6, Arg 5 -Arg 13 ; T8, Lys 133 -Lys 154 ; and T14, Phe 155 -Lys 182 ) linked by disulfide bridges involving Cys 6 , Cys 146 , Cys 157 , and Cys 174 (Fig. 2B and Table I). MS/MS analysis of peptides T6, T8, and T14 confirmed their identity (data not shown). Analysis of the CID spectrum of the ϩ5 ion (m/z 1309.6) of T3 (Fig. 3A) Fig. 2A and Table I) was tentatively identified by a comparison of peptide masses (see Table I The mass of peptide fraction T5 (4259.9 Da, Table I) (Fig. 3C). Automated Edman degradation of tryptic peptide fractions T1-T5 ( Fig. 2A) confirmed the NH 2 -terminal sequence of the peptide components in these fractions (Table I).
N-Linked Glycosylation Sites in sIL-6R-The N-linked glycosylation sites in sIL-6R were determined by the non-appearance of asparagine residues during automated Edman degradation of tryptic and Asp-N endopeptidase peptides containing Asn-Xaa-(Ser/Thr) motifs (data not shown). Of the four potential N-glycosylation sites in the extracellular domain of the IL-6R, NAT (residues 36 -38), NIT (residues 202-204), and NYS (residues 74 -76) were found to be glycosylated, whereas one potential N-glycosylation site, NSS (residues 226 -228), was not modified. Further confirmation of N-linked glycosylated asparagine residues was provided by Asn/Asp conversion following endoglycosidase F treatment (49); see Asn/Asp conversion at position 74 in peptide T7 (Fig. 4).
Homology Modeling-Two separate models of the sIL-6R were created that depict different orientations between the a Tryptic peptide fractions are labeled according to their order of retention as shown in Fig. 2A. b Peptide sequences identified from MS/MS spectral data using the programs MS-Tag, MS-Product, and manual assignment. Underlined sequences were confirmed by NH 2 -terminal sequencing. One-letter abbreviations were used for amino acids. Tryptic peptides shown in parentheses refer to peptides produced upon reduction of the tryptic digest prior to RP-HPLC (Fig. 2B).
c Numbers denote amino acid positions in the sequence of the mature protein (Fig. 6). d Numbers refer to observed mass while those in parentheses are calculated from the amino acid sequence minus 2 daltons for each disulfide bond.
e Numbers refer to observed mass while those in parentheses are calculated from the amino acid sequence. f Difference in mass between the sum of observed reduced masses and the observed nonreduced mass. g The difference in mass between the nonreduced and reduced peptide is consistent with S-cysteinylated cysteine. h The difference in mass between the nonreduced and reduced peptide is consistent with S-pyridylethylated cysteine.
Ig-like domain and the CBD (Fig. 5). In model A (Fig. 5A), the linker region between these domains follows the gp130 template and orients the Ig-like domain in the plane of the CBD. In model B (Fig. 5B), the linker region follows the PRLR template and the Ig-like domain is orientated across the plane of the CBD. The GHR and EPOR structures were not chosen as templates for the linker region since the GHR has no linker region and the orientation of the EPOR linker region would not allow the formation of the Cys 6 -Cys 174 connectivity. The template used for the Ig-like domain was the V L domain of FAB D44.1; a V-type Ig-like fold comprised of nine ␤-strands (39). The amino acid sequence of FAB D44.1 aligned well with the IL-6R Ig-like domain in ␤-strands B, C, D, and F (Fig. 6A). These strands form the hydrophobic core of the domain and contain a number of conserved hydrophobic and aromatic residues, including the disulfide bond linking ␤-strands B and F. In contrast, the peripheral strands A, E, and G give less favorable alignments. The absence of ␤-strands CЈ and CЈЈ of the FAB D44.1 V L domain in the IL-6R suggests that the Igmodeled domain forms a C1-type rather than a V-type fold (for review, see Ref. 14). Regardless of the assignment of the particular Ig-like fold, the overall ␤-sandwich structure is retained. In particular, the relative position of Cys 6 will not vary. The second and third domains of the IL-6R are classified as FN III domains. Sequence analysis of the sIL-6R revealed 25% sequence identity to the CBD of gp130, which is higher than any other template in the data base (Fig. 6B).
The MODELLER program works in such a way that the calculated model follows most closely the best template (50). Therefore, our models for the tandem FN III modules of sIL-6R are most closely aligned to the recently published crystal structure for the CBD of gp130 (41). Models A and B of the sIL-6R were shown to be of good quality as assessed by the structural quality checks described under "Experimental Procedures." The ProsaII Z-scores of the CBD of models A and B were Ϫ4.8 and Ϫ5.3, respectively, comparing favorably with a score of Ϫ4.8 for the CBD of gp130. The ProsaII plot of the IL-6R Ig-like domain is less favorable than the profile of the template, FAB D44.1, a general feature observed in homology model building with low sequence identity.

DISCUSSION
Knowledge of the cysteine connectivity pattern and carbohydrate attachment sites of sIL-6R provides a foundation for the determination of the three-dimensional structure of this molecule and, ultimately, of the IL-6⅐IL-6R⅐gp130 complex. Mass spectrometric and NH 2 -terminal sequence analysis of a tryptic digest of the sIL-6R identified four disulfide linkages (Cys 28 -Cys 77 , Cys 102 -Cys 113 , Cys 146 -Cys 157 , and Cys 6 -Cys 174 ), two "free" cysteines (Cys 192 and Cys 258 ), and three N-linked glycosylation sites (Asn 36 , Asn 74 , and Asn 202 ).
The Cys 6 -Cys 174 disulfide is novel among the cytokine receptors. Sequence alignment of the IL-6R with GHR, PRLR, EPOR, and gp130, whose CBD structures are known (Fig. 6), predicts that Cys 174 is located on ␤-strand F of the NH 2 -terminal FN III domain, while Cys 6 is situated at the NH 2 -terminal end of ␤-strand A of the Ig-like domain. Therefore, the disulfide bond formed between these two residues provides an additional link between the Ig-like and FN III domains. Cys 6 is conserved in the IL-11R (54,55), however, the IL-11R has no corresponding cysteine partner in the NH 2-  Cys 192 and Cys 258 do not form intramolecular disulfide bonds and occur as either free or in a cysteinylated form. These residues are predicted to be located at the COOH-terminal end of ␤-strand G of the NH 2 -terminal FN III domain and ␤-strand E of the COOH-terminal FN III domain, respectively. The three-dimensional models of the sIL-6R shown in Fig. 5 predict that Cys 192 and Cys 258 are spatially distant from each other and are unable to form a disulfide bond. Cys 258 is conserved among human, mouse, and rat IL-6 receptor species, while Cys 192 is replaced by a leucine in the mouse and rat sequences (Fig. 6). Cys 258 is also conserved in the EPOR (58) and cytokine-like factor-1, a recently cloned soluble member of the class I cytokine receptor family whose function is presently unknown (59). The presence of free cysteine residues in the membraneproximal region of the IL-6R extracellular domain suggests that the IL-6R may homodimerize via intermolecular disulfide bond formation, in a manner similar to that of gp130 (60), or EPOR (61). Moreover, disulfide linkages between specific ligand-binding (␣-subunit) and signaling (␤-subunit) receptors have been reported for the IL-3/IL-5/granulocyte macrophagecolony stimulating factor subfamily of cytokine receptors (62)(63)(64). These observations indicate that the potential for disulfide-linked oligomerization of the IL-6R warrants further investigation.
Three N-linked glycosylation sites in the sIL-6R were identified at Asn 36 , Asn 74 , and Asn 202 . Asn 36 and Asn 74 are pre-dicted to be located on the loop region connecting ␤-strands B-C and on ␤-strand F of the Ig-like domain, respectively, while Asn 202 is expected to be on ␤-strand A of the COOH-terminal FN III domain. These carbohydrate moieties do not directly affect ligand binding as shown by the expression of functional sIL-6R in Escherichia coli (65). However, the presence of carbohydrate chains in these regions almost certainly excludes them from protein-protein interaction sites within the IL-6R complex. The absence of glycosylation at Asn 226 is consistent with the known involvement of neighboring residues Ser 228 -Leu 232 in ligand binding (17,66).
Implications of the cysteine connectivities and carbohydrate attachments on the overall topology of the IL-6R extracellular region are modeled in Fig. 5. The inter-domain disulfide bond which links the Ig-like and NH 2 -terminal FN III domains is distant from the hinge region connecting these two modules. This causes the Ig-like domain to fold back onto the NH 2terminal FN III domain either in the plane (Fig. 5A) or across the plane of the CBD (Fig. 5B). A different type of inter-domain disulfide bond has been reported for the extracellular region of the interferon-␥ receptor, a class II cytokine receptor (67,68). Here, the inter-domain disulfide bond is located in the hinge region of the CBD, close to the linker sequence connecting the two FN III domains. This maintains a hinge angle of approximately 120°, whereas the inter-domain disulfide bond found in the IL-6R imposes a hinge angle approaching 0°between the Ig-like and NH 2 -terminal FN III domains.
A model of the sIL-6R (Fig. 5A) complexed with the crystal structure of IL-6 (48) and CBD of gp130 (41), superimposed over the crystal structure of the growth hormone receptor complex (42), is shown in Fig. 7. This model is consistent with our previously proposed hexameric IL-6 receptor complex model which is based upon the association of two GH/GHR-like trimers (2). While other IL-6/IL-6R/sgp130 trimer models have been reported (69 -71), these were generated prior to the publication of the IL-6 and gp130 CBD three-dimensional structures and most likely contain significant errors in the sequence alignments for gp130 and IL-6R. For instance, Kalai et al. (71) align Cys 174 of the IL-6R with Ile 109 of the GHR, thereby directing the side chain of Cys 174 toward the core of the NH 2 -terminal FN III domain which would prevent the formation of an inter-domain disulfide bond with Cys 6 . In our model, Cys 174 is aligned with Thr 112 of the GHR and Glu 177 of gp130 (Fig. 6). This orientates the side chain of Cys 174 outwards from the core of the protein, which is consistent with a Cys 6 -Cys 174 connectivity. The formation of a disulfide bond between Cys 6 and Cys 174 restrains the Ig-like domain away from the IL-6-binding site (Fig. 7). This suggests that the Ig-like domain is unlikely to be involved in ligand binding.
Our model predicts that cysteinylated Cys 192 is located within the IL-6-binding site of the IL-6R. However, this modification seemingly does not interfere with ligand binding since the modified receptor was purified by affinity chromatography using IL-6-Sepharose. This finding is in accord with a previous FIG. 7. Ribbon model of a trimeric IL-6⅐sIL-6R⅐sgp130 (CBD) receptor complex. This model is based on the growth hormone receptor complex (42). A, top view, and B, side view of the receptor complex. IL-6 is colored green, sIL-6R (model A) orange, with the Ig-like domain of the IL-6R yellow, and the CBD of gp130 magenta. Residues of IL-6 known to be involved in binding to IL-6R and gp130 are colored green, with binding site III, involved in binding to a second gp130 receptor colored cyan. IL-6R residues involved in binding to IL-6 are colored blue. The disulfide between the Ig-like and FN III domains of the sIL-6R (Cys 6 -Cys 174 ), Cys 192 and Cys 258 are colored yellow. Potential N-linked glycosylation sites are colored pink. (Coordinates of this model are available at http://www.liba.ludwig.edu.au.) study, which demonstrated that mutation of Cys 192 to alanine did not inhibit IL-6 binding (17). It is unlikely that the IL-6R forms an inter-molecular disulfide bond with IL-6 upon binding, since biophysical experiments on both human (72) and mouse (73) IL-6 have shown that it does not contain free cysteines. These observations suggest that despite its location, Cys 192 is not involved in ligand binding. However, the potential for disulfide exchange has not yet been addressed. Asn 226 is also located inside the predicted ligand-binding site of the IL-6R which is consistent with this residue not being glycosylated, since carbohydrate chains in this region would be expected to sterically inhibit ligand binding. N-Linked carbohydrate chains located on Asn 36 , Asn 74 , and Asn 202 do not interfere with any of the protein-protein interaction sites of the receptor complex predicted in our model. The role of the Ig-like domain and the free cysteine residues in the extracellular domain of the IL-6R must await further studies.