Characterization of an Antagonist Interleukin-6 Dimer by Stable Isotope Labeling, Cross-linking, and Mass Spectrometry*

The homodimeric form of a recombinant cytokine in-terleukin-6 (IL-6 D ) is known to antagonize IL-6 signal- ing. In this study, spatially proximal residues between IL-6 chains in IL-6 D were identified using a method for specific recognition of intermolecular cross-linked peptides. Our strategy involved mixing 1:1 15 N-labeled and unlabeled ( 14 N) protein to form a mixture of isotopically labeled and unlabeled homodimers, which was chemically cross-linked. This cross-linked IL-6 D was subjected to proteolysis by trypsin and the generated peptides were analyzed by electrospray ionization time-of-flight mass spectrometry (MS). Molecular ions from cross-linked peptides of intermolecular origin are labeled with [ 15 N/ 15 N] (cid:1) [ 15 N/ 14 N] (cid:1) [ 14 N/ 15 N] (cid:1) [ 14 N/ 14 N] yielding readily identified triplet/quadruplet MS peaks. All other peptide species are labeled with [ 15 N] (cid:1) [ 14 N] yielding doublet peaks. Intermolecular cross-linked peptides were identified by MS, and cross-linked residues were identified. This intermolecular cross-link detection method, which we have designated

Previously, we have shown that a dimeric form of recombinant IL-6 (IL-6 D ) is a potent antagonist for IL-6 signaling (16). Recombinant IL-6 D binds tightly to soluble IL-6R (sIL-6R) to form a 1:2 IL-6 D (sIL-6R) 2 complex (16). In contrast to the binary IL-6⅐sIL-6R complex, IL-6 D (sIL-6R) 2 binds gp130 weakly and does not show significant biological activity in the signal transducer and activator of transcription 3 (STAT3) phosphorylation assay (16). Natural (glycosylated) human IL-6 is also known to form a dimer that makes up a substantial part of IL-6 in blood or fibroblast secretions (17)(18)(19) and has also been shown to interact with membrane-bound IL-6R (15,20,59). Recently, glycosylated natural human IL-6 D , identified by immunoblotting and size exclusion chromatography, was shown to be a survival factor secreted by epithelial cells that inhibited the apoptosis of B-chronic lymphocytic leukemia cells (21). Significantly, recombinant human IL-6 D from Escherichia coli acted as a survival factor in a similar way (21). Taken together, these results suggest that natural and recombinant IL-6 D may have similar biological activity.
Elucidation of the IL-6 D structure will be critical to understanding the basis of its antagonistic properties. Whereas the structure of IL-6 is known to be a 4-␣-helical bundle (22), the structure of IL-6 D is unknown. Previous biophysical studies of the sedimentation properties and the unfolding-dissociation relationship of IL-6 D (23) have shown it is likely to form a metastable domain-swapped dimer (24,25) in which adjacent subunits have the IL-6 structure, but contain interchanged ␣-helical bundle domain elements.
Here, we investigate the arrangement of domain-swapped IL-6 chains within IL-6 D using a technique based on crosslinking and mass spectrometry. Although the established techniques of x-ray crystallography and NMR spectroscopy yield high resolution data, often this takes months or years to obtain (26,27). Techniques in mass spectrometry (MS) combined with cross-linking (28 -30) or other chemical labeling techniques such as hydrogen/deuterium exchange (31,32) have been evaluated for rapid low-resolution three-dimensional study of proteins (30) or protein complexes (33,34). Cross-linking/MS methods involve chemically or photochemically cross-linking a protein complex (35), followed by digestion of the cross-linked complex and MS analysis of the resulting peptide mixture (36). Cross-linked peptides can be identified by parent ion mass and/or the fragmentation pattern produced by tandem mass spectrometry (MS/MS), thereby locating adjacent protein regions and enabling assembly of low-resolution models of proteins or protein complexes. Cross-linking/MS experiments are * 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.
Despite the advantages of cross-linking methods, namely high molecular weight capability, speed, and the small quantities of protein required, the large number of peptide species that are seen from the digestion of cross-linked proteins makes it difficult to identify relevant intermolecular cross-linked peptides from MS data. This problem has been partially addressed by "tagging" methodologies that allow rapid visual MS location of cross-linked species within complex peptide mixtures (34,(37)(38)(39). For example, the use of a 1:1 mixture of undeuterated and deuterated (d 0 /d 4 -labeled) cross-linking reagent readily allows mass spectrometric detection of all cross-linked species by the presence of d 0 /d 4 -isotope tags (39). However, for studying interactions between proteins, even these tagging methodologies fall short in that they fail to distinguish inter-and intra-cross-linked peptides. This results in cross-linked peptide species being tagged that do not yield useful information on intermolecular interactions, such as intramolecular crosslinked peptides or peptides modified by partially hydrolyzed cross-linking reagent (30,40). Moreover, these tagging methodologies make cross-linking studies of oligomeric proteins, such as IL-6 D , ambiguous because identical cross-linked peptides can in principle arise from inter-or intramolecular origins (38).
Here we present a method for visualizing intermolecular cross-linked peptides in the IL-6 homodimer that we have designated mixed isotope cross-linking (MIX). Applied to IL-6 D , the MIX method requires preparation of uniformly 15 N-labeled and unlabeled ( 14 N) IL-6, combined as a 1:1 mixture and reassociated to form a population of 14 N-, mixed 14 N/ 15 N-, and 15 N-labeled IL-6 D . Cross-linking and mass spectrometric peptide mapping on this mixture allows intermolecular crosslinked peptides to be identified easily as they form distinctive triplet or quadruplet mass spectrum peaks because of the distribution of 14 N-and 15 N-labeled peptides within these crosslinked peptides. In contrast, all intramolecular cross-linked and noncross-linked peptides are seen as doublet mass spectrum peaks. This ability to discriminate between inter-and intra-cross-linked species makes the MIX technique a uniquely useful new tool for studying intermolecular interactions. We describe the application of this technique to determine proximal intermolecular residues within the homodimeric form of IL-6 and to deduce the mode of three-dimensional domain swapping, based on the known structure of monomeric human interleukin-6 (22).

EXPERIMENTAL PROCEDURES
Materials-Trifluoroacetic acid (HPLC/Spectro Grade) and bis(sulfosuccinimidyl)suberate (BS 3 ) were from Pierce. Sequencing grade trypsin (EC 32.4.21.4) was from Roche Molecular Biochemicals. HPLCgrade solvents were from Mallinckrodt, and all other buffers and reagents (Analar grade) were from BDH. All buffers and solutions were prepared with deionized water purified by a tandem Milli Q and Milli RO system (Millipore).
Expression and Purification of IL-6 M , IL-6 D and Mixed 15 N/ 14 N-Labeled IL-6 D -Recombinant human IL-6 was expressed in E. coli as a recombinant fusion protein consisting of the first six N-terminal amino acids of bacterial ␤-galactosidase (Thr-Met-Ile-Thr-Asn-Ser) and residues 48 to 212 of immature human IL-6 (Swiss-Prot P05231) as described elsewhere (42). Purified recombinant [ 15 N]IL-6 was prepared as described by Morton et al. (41). 400 g of monomeric, dissociated  14 N/ 14 N-labeled IL-6 D was isolated from this mixture by preparative size exclusion chromatography (SEC) at 25°C as previously detailed for [ 14 N]IL-6 (14). The column was developed with 20 mM phosphate buffer, pH 7.4, containing 150 mM NaCl, and proteins were detected by absorbance at 215 nm. Both IL-6 M and IL-6 D were Ͼ90% homogenous based on the 215-nm peak area ratio between monomer and dimer as judged by analytical SEC, and were stable for over 3 months when stored at 4°C.
Cross-Linking of IL-6 M and IL-6 D -The bifunctional amine-specific cross-linking reagent BS 3 was added as a 10 mM solution in 20 mM phosphate buffer, pH 7.4, containing 150 mM NaCl, to 400 g of purified IL-6 M or IL-6 D at a concentration of 0.5 mg/ml in the same buffer. Protein to cross-linker molar ratio was 1:10. The cross-linking reaction was allowed to proceed for 2 h at 4°C and was quenched by adding Tris to 50 mM and incubating for 15 min at 25°C. The extent of cross-linking was monitored by SDS-PAGE and analytical RP-HPLC (42).
Peptide Mapping of Cross-linked Protein-Tryptic digestion of crosslinked IL-6 D was performed at an enzyme-to-substrate ratio of 1:40 (by weight) at 37°C for 16 h, in 20 mM phosphate buffer, pH 7.4, containing 150 mM NaCl. Digests containing equivalent masses of noncross-linked IL-6 M /IL-6 D and cross-linked IL-6 M were used as controls for the digest of cross-linked IL-6 D . For peptide mapping, tryptic digests were loaded onto a Brownlee RP-300 7 m, 300-Å octylsilica column (100 ϫ 4.6 mm inner diameter) (Applied Biosystems) and subjected to RP-HPLC on a HP-1090 liquid chromatograph (Agilent Technologies) equipped with a manual injector fitted into the column compartment and a diode-array UV detector for wavelength monitoring of the column eluent. Peptides were eluted at a flow rate of 1 ml/min using a linear gradient of 0 -100% solvent B developed over 85 min, where solvent A consisted of 0.1% aqueous trifluoroacetic acid and solvent B was 0.093% aqueous trifluoroacetic acid in 85% acetonitrile. For on-line LC/MS, a T-junction was inserted downstream of the column operating at a 1:100 split ratio and introduced into a Mariner Biospectrometry workstation (PerSeptive Biosystems Inc.) electrospray ionization/time of flight mass spectrometer at 10 l/min. Eluting peptide fractions were collected manually (ϳ500 l) for MS/MS analysis. The electrospray source was set at a capillary temperature of 160°C and a nozzle potential of 120 V using a spray voltage of 2.8 kV, and curtain and nebulizer gas flow rates were set at 1.6 and 0.4 liter/min, respectively. Mass spectra were acquired for 2 s/spectrum over the m/z range 300 -2000 in profile mode.
Tandem Mass Spectra-MS/MS spectra were acquired on a Q-Tof 2 mass spectrometer (Micromass) by infusion of samples collected off-line from RP-HPLC after diluting 1:1 with 50% aqueous methanol, 0.1% formic acid to enhance sensitivity. Conditions used were: cone voltage, 45 V; collision gas, argon; collision energy, 15 to 35 V. The ESI voltage used was 3.5 kV and the ion source was maintained at 80°C.
Assignment of Mass Spectra-Where a mass peak was shown by isotopic labeling (as described below) to be an intermolecular crosslinked peptide, potential cross-linked peptide species corresponding to the parent mass were identified using a Java program "X-Link" (available on request, from T.T.). X-Link generates a searchable list of all possible cross-linked peptide molecules from a given set of protein sequences, including modifications by partially hydrolyzed cross-linker. All possible cross-linked species having a mass within 0.5 Da of the observed molecular mass were considered, and the correct species was identified by comparing the expected fragmentation patterns to observed MS/MS data.
Molecular Modeling-Molecular models of IL-6 D were generated using the average NMR structure of monomeric IL-6 (Protein data bank code 1IL6) (22). NMR structural data was used in preference to the crystal structure (44) because it contains no chain breaks corresponding to missing loop regions. The dimer orientation was set by superimposing the IL-6 D "subunits" (the IL-6 M -like domains) onto those of IL-10 (45). The cross-linked dimer was generated using the MODELLER program (46) with two IL-6 monomers used as the templates. The five N-terminal residues of the model, which had no template structure available, were positioned in a manner consistent with the observed N terminus cross-linking. In addition, to enable the crossover of helices, residues 127-135 in the C-E loop were released from the IL-6 M structure constraints and were allowed to move freely. Apart from the C-E loop, the two IL-6 M -like domains of the IL-6 D model are structurally identical to IL-6 M .

RESULTS
Experimental Strategy-The experimental strategy used in this study is shown in Fig. 1 14 N]]IL-6 dimers were isolated from this mixture by SEC and were then cross-linked with BS 3 , cleaved with trypsin, and analyzed by loading onto an HPLC column eluting directly into a mass spectrometer. The cross-linking reagent BS 3 reacts with amine groups on lysine and with the N terminus, modifying single residues (ϩ158 Da) or producing interresidue cross-links (ϩ138 Da). Tryptic cleavages are seen at Lys and Arg, but not at BS 3 -modified Lys residues. Within the mixture of cross-linked peptides formed by tryptic digestion, only intermolecular cross-linked peptides gave triplet or quadruplet MS peaks because of the distribution of 14 N and 15 N in otherwise chemically identical peptides within cross-links. The MS peak shape of intermolecular cross-linked peptides is determined by the difference between the number of nitrogen atoms in the cross-linked peptides. If this difference is small, the two possible mixed 14 N/ 15 N-labeled cross-linked peptide species will have similar molecular masses, resulting in a triplet MS peak comprised of 14  All other species, namely intramolecular crosslinked and noncross-linked peptides, were seen as 14 N-and 15 N-labeled forms, resulting in doublet MS peaks. This allowed quick visual identification of molecular ions from intermolecular cross-linked peptides. The identities of the cross-linked tryptic peptides in these intermolecular species were then assigned based on parent mass using the in-house program X-Link.
Cross-linking-The purified [ 15 N/ 15 N, 15 N/ 14 N, 14 N/ 14 N]IL-6 D was judged homogeneous based on co-elution with unlabeled purified IL-6 D on analytical SEC (Fig. 2). Furthermore, the [ 15 N/ 15 N, 15 N/ 14 N, 14 N/ 14 N]IL-6 D could be cross-linked in an identical way to homogeneous [ 14 N]IL-6 D as seen on SDS-PAGE (Fig. 2, inset). The refolding conditions used for [ 15 N/ 15 N, 15 N/ 14 N, 14 N/ 14 N]IL-6 D were identical to those previously used to isolate, assay, and characterize IL-6 D through biological and biophysical assays (16). Previous cross-linking studies on the related cytokine erythopoietin (47) showed that the level of BS 3 cross-linking used here is unlikely to affect protein structure. The cross-linking reaction was confirmed by SDS-PAGE to be specific for dimeric IL-6, as a predominant band of relative molecular mass (M r ) 40,000 was seen on cross-linking IL-6 D , whereas for cross-linked IL-6 M only a M r 20,000 band was observed (Fig. 2, inset).
Peptide were found (see below) to cross-link interchain residues, N terminus Lys 128 , Lys 171 and Lys 27 , and Lys 171 and Lys 66 , respectively (Table I). Importantly, the intermolecular cross-linked peptide species X1 was abundant and was easily observed on a chromatogram, whereas a control digest of cross-linked monomeric IL-6 showed no corresponding intramolecular species resulting from a cross-link between the N terminus and Lys 128 . The species X2 and X3 were both observed at much lower concentrations. Additionally, 13 tryptic peptides and three cross-linked intramolecular peptides were observed as doublet peaks, which did not yield additional structural information on intermolecular interactions within IL-6 D (data not shown). The ratios of integrated peak areas within multiplets, and the separation between multiplet peaks, were consistent with a 14 N: 15 N ratio of 1:0.7. Under ESI-time of flight MS/MS conditions (Q-Tof2), informative fragmentation reactions resulting in sequence coverage of both cross-linked peptides were observed on the 2ϩ and 3ϩ ions of the cross-linked peptides. Furthermore, MS/MS spectra of peaks within multiplets displayed isotopic shifts between MS/MS fragment ion peaks consistent with assignments. For example, the triply charged ion of the cross-link "X3" (see Table I) at m/z 765.78 presented as a quadruplet peak under MS (Fig. 3A). Tandem mass spectrometry (MS/MS) on the lowest m/z peak of this quadruplet showed a series of b and y fragment ions from both tryptic peptides joined in the crosslinked species X3 (Fig. 3B). This MS/MS spectrum unambiguously identified the sequences of the cross-linked peptides and the location of the cross-link between them. For the crosslinked species X3, its constituent peptides were Leu-Pro-Lys 66 -Met-Ala-Glu-Lys, denoted T1, and Ser-Phe-Lys 171 -Glu-Phe-Leu-Gln-Ser-Ser-Leu-Arg, denoted T2, cross-linked via Lys 66 -Lys 171 (Fig. 3C). In addition, MS/MS on the second lightest quadruplet peak of X3 (Fig. 3D) showed the expected sequence ions from the cross-linked peptide [ 14 N]T1-[ 15 N]T2. These results show that the MIX-tagged mass peaks correspond to cross-linked peptides of intermolecular origin.

DISCUSSION
Structure of IL-6 D -Previously, it has been shown that IL-6 D has a frictional coefficient consistent with two IL-6 M subunits in an end-to-end arrangement (23). Furthermore, unfolding/ dissociation relationships (23) indicate that IL-6 D is a metastable dimer and probably exhibits three-dimensional domain swapping, in which the two adjacent domains are IL-6 M -like in structure but have domain elements (␣-helices and/or loops) interchanged symmetrically via the top or bottom "face" on the IL-6 M structure (23,25). In particular, the structure of the ␣-helical bundles making up the IL-6 M -like domains is expected to be essentially conserved between IL-6 D and IL-6 M , the only structural difference of note being in the loops that cross between the two domains (23,25).
Given a three-dimensional domain-swapped structure for IL-6 D , the mode of domain element interchange can be identified from the intermolecular cross-links identified through the MIX experiment. The cross-linked intermolecular peptide X1, which links the N terminus to Lys 128 , was not detected from the control digest of cross-linked (monomeric) IL-6 M . Rather, the species X1 was observed only from the digest of the crosslinked dimer, IL-6 D . Hence in the IL-6 D , the N terminus of one IL-6 chain must be in close proximity to Lys 128 of the other chain. This indicates that X1 must originate from a cross-link at the interface of the IL-6 dimer between the two IL-6 M -like domains. Otherwise, if the cross-linked peptide X1 were intradomain, then it would also be expected from the control digest of IL-6 M because (i) the IL-6 M structure is contained within the domain-swapped IL-6 D structure (23)(24)(25) and (ii) the cross-linked residues in X1 will not be directly involved in the crossover loop region, because Lys 128 is part of the C-helix. Consequently, the two IL-6 M -like domains in IL-6 D must be oriented "head to head," having the N terminus/C terminus containing faces of the IL-6 M subunits adjacent as shown in cartoon models II, III, and IV in Fig. 4A.  Table I Table I). The peak "T1-XL" has mass consistent with an unknown fragmentation within the crosslinker moiety derived from BS 3 . Panel C, possible MS/MS fragmentation schemes from X3, with nomenclature indicated for the peptide fragmentation products. A standard nomenclature is used (54) for fragment ions from cross-linked peptides (55,56). a Cross-links are via -amino groups for lysines or the N terminus. b Cross-linked peptides are denoted by the start and end residue positions in parentheses. Residue numbering used is for mature human IL-6 (23). an IL-6 dimer in which IL-6 M subunits are oriented in this head to head fashion (models II, III, and IV). Although the intermolecular cross-linked peptide X1 links together residues at the dimerization interface in IL-6 D , the other intermolecular cross-linked peptides identified by MS, namely X2 and X3 (Table I), link together lysine residues located in the middle of the IL-6 M -like subunits, as seen in cartoon models II-IV (Fig. 4A). The cross-linked peptides X2 and X3 cross-link between the D-helix and the A-helix/A-B loop, and hence are consistent with an IL-6 D structure in which the D-helix is swapped relative to the A-helix and A-B loop in each IL-6 M -like domain. Only model IV (Fig. 4A) contains this structural feature and is consistent with the intermolecular nature of cross-linked peptides X1, X2, and X3, as shown by the MIX experiment. Furthermore, the domain-swapped models II and III are inconsistent with the conserved intramolecular disulfide bond Cys 42 -Cys 54 , which tethers the A-B loop to the B-helix. Model IV (Fig. 4A) is characterized by symmetrical swapping of the two helices nearest to the C terminus, the Dand E-helices, between IL-6 M -like subunits in IL-6 D . This mode of domain element interchange corresponds closely to that seen in the known structures of the naturally dimeric short chain cytokines IL-10, interferon-␥, and IL-5, which have the "interferon-␥" fold (25) and are distant homologs of IL-6. Moreover, a recent survey of 40 domain-swapped proteins shows that almost all are swapped via regions adjacent to the N or C terminus, which is a property of this model (24).
A molecular model of IL-6 D was constructed (Fig. 4B) based on the cross-linking data and the NMR structure of IL-6 M (22) (see "Experimental Procedures"), in which IL-6 M -like domains were connected to one another via the C-E loops. One IL-6 Mlike domain was composed of the A-, B-, and C-helices from one chain and the D-and E-helices of the other chain. The crosslinking data does not provide information on the relative orientation of IL-6 M -like domains within IL-6 D . However, the IL-6 D model was constructed with an orientation between IL-6 M -like domains similar to that between the "monomer-like" domains in IL-10 (25). IL-10 was selected from the three possible dimeric cytokines mentioned above, on the basis of its elongated shape (45) (the IL-10 inter-domain angle is ϳ135°), for consistency with the known high axial ratio of IL-6 D (23). The five N-terminal residues, which do not appear in the NMR structure of monomeric IL-6, are displayed in a conformation consistent with the cross-linking data (Fig. 4B). This model satisfies the distance constraints, namely Ͻ24 Å between C␣-C␣ of lysines and the N terminus, imposed on IL-6 D by the cross-linking of ⑀-amino groups by BS 3 (30,48) as seen in cross-linked peptides X1, X2, and X3.
Implications for the Inhibitory Role of IL-6 D -IL-6 M in vivo has diverse roles in mediating proliferative signals to B-and T-cells in the immune system. Signaling occurs via sequential binding to a two-receptor system made up of the membranebound proteins IL-6R and gp130. IL-6 first binds to IL-6R for the IL-6⅐IL-6R complex to recruit gp130 in the second step, which then forms a hexameric complex active in signaling and consisting of two molecules each of IL-6, IL-6R, and gp130 (14). The topology of this hexameric IL-6 receptor complex is not yet known. Mutagenesis and structural studies on IL-6 have shown that it contains one site for IL-6R binding, site I, and two sites for gp130 binding, sites II and III (22,44). Site I consists of the C-terminal end of the D-helix, the C-terminal part of the long AB-loop, and the N-terminal part of the B-helix (49). Site II is a region halfway along the A-and C-helices, whereas site III is located on the DE-loop. This topic has also been reviewed by Simpson et al. (12).
Previously, recombinant IL-6 D has been shown to inhibit IL-6 signaling (16,23), which potentially has therapeutic applications. This inhibition is likely to be biologically important, as IL-6 in vivo is partly homodimeric (17)(18)(19) and recent studies suggest that naturally occurring and recombinant IL-6 D may have similar biological activity (21). Thus, IL-6 D in vivo seems likely be a natural inhibitor for IL-6 signaling. Recombinant IL-6 D has been shown to bind strongly to the soluble extracellular domain of IL-6R (sIL-6R) to form a stable IL- FIG. 4. A, models I-IV are cartoon representations of possible modes of domain swapping between subunits in IL-6 D based on a head to head alignment of IL-6 D subunits consistent with the intermolecular crosslink X1 between the N terminus and Lys 128 (see Table I). The two IL-6 chains are black and white. The cross-linked peptides X1, X2, and X3 (Table I) are positioned on the cartoon of IL-6 D and shown as intermolecular (red-blue) or intramolecular (blue-blue or red-red). B, ribbon representation of the IL-6 D structural model. The D-and E-helices are swapped between IL-6 M -like domains as shown in model IV (Fig. 4A). The IL-6 chains are colored dark blue and light blue. C␣ atoms of lysine residues involved in cross-linking are shown as red and orange spheres for the dark and light blue chains, respectively. Regions for which the chain orientation/conformation is unknown are colored magenta. These regions form part of the crossover C-E loop (residues 127-135, which are involved in crossover) and the five N-terminal residues. The N termini of the two chains are also represented by magenta spheres. The ␣-helices of the two chains are labeled A-E. Intermolecular cross-links corresponding to cross-linked peptides X1 (N terminus and Lys 128 ), X2 and X3 (Lys 171 with Lys 27 and Lys 66 , respectively) are represented as dashed lines between C␣ atoms of the cross-linked residues. The figures were prepared using Molscript (57) and Raster3D (58). Coordinates of the model are available from the authors on request. C, ribbon representation of the IL-6 D structural model showing IL-6R and gp130binding sites. Chains are colored as in B. Binding sites are shown using space filling representation of residues implicated in IL-6 receptor binding by previous mutagenesis studies. Site I (red) is involved in IL-6R binding. Sites II and III (orange and green, respectively) are involved in gp130 binding. The exact orientation of the IL-6 M -like domains is unknown, but the extended conformation of IL-6 D (23) is inconsistent with direct blocking of IL-6 receptor binding sites II and III by IL-6 dimerization (see "Discussion"). 6 D (sIL-6R) 2 complex (16). This complex binds gp130 only weakly, inhibiting IL-6 signaling. Although the structures of the IL-6⅐IL-6R complex and of the signaling IL-6/IL-6R/gp130 hexamer are not known, it is possible to rationalize the sIL-6R binding properties of IL-6 D through binding site data. The IL-6 D model with IL-6 receptor-binding sites highlighted (Fig.  4C) shows that site I of IL-6 lies on the same end of the IL-6 molecule as the putative dimer interface. Inspection of this region in the IL-6 D model reveals that two closely situated, antiparallel site I regions are exposed that can bind two sIL-6R molecules (Fig. 4C). Hence, the model of IL-6 D proposed here is compatible with the observed strong binding of 2 equivalents of sIL-6R to IL-6 D (16).
The proposed structure of IL-6 D can also account for its inhibitory nature with respect to IL-6 signaling (16). In structural terms, the lack of strong binding of soluble gp130 to the stable IL-6 D (sIL-6R) 2 complex is likely to be because of restriction of the relative orientations of gp130-binding sites II and III on IL-6 D (50) within IL-6 D (sIL-6R) 2 that precludes cooperative soluble gp130 binding. This is consistent with the model of IL-6 D obtained though the cross-linking data (Fig. 4C), because although no gp130-binding sites on IL-6 (51, 52) appear to be directly blocked by IL-6 dimerization, their relative orientations are constrained. Furthermore, the orientation of IL-6 Mlike domains within the IL-6 D molecule is very different from that seen for the viral IL-6 homolog, vIL-6, within the crystal structure of the 2:2 vIL-6⅐gp130 complex, which is capable of signaling (53). In this complex, the vIL-6 chains are widely separated, the minimum distance between them being ϳ35 Å (for comparison, IL-6 M is ϳ30 Å ϫ 30 Å ϫ 50 Å). Thus, constrained orientation of gp130-binding sites within IL-6 D , rather than dimerization induced blocking of IL-6 receptor-binding sites as previously proposed (23), seems likely to account for the IL-6 D -mediated inhibition of IL-6 signaling. A more detailed structural understanding of the inhibitory role of IL-6 D must await the three-dimensional structures of the signaling (IL-6 M ) and inhibitory (IL-6 D ) complexes with their associated receptors. Although the full description of the IL-6 D structure awaits analysis by x-ray crystallography or NMR spectroscopy, the present structural model may form a basis for rational design of therapeutic antagonists.