Report

Hexameric Structure and Assembly of the Interleukin-6/IL-6 α-Receptor/gp130 Complex

See allHide authors and affiliations

Science  27 Jun 2003:
Vol. 300, Issue 5628, pp. 2101-2104
DOI: 10.1126/science.1083901

This article has a correction. Please see:

Abstract

Interleukin-6 (IL-6) is an immunoregulatory cytokine that activates a cell-surface signaling assembly composed of IL-6, the IL-6 α-receptor (IL-6Rα), and the shared signaling receptor gp130. The 3.65 angstrom–resolution structure of the extracellular signaling complex reveals a hexameric, interlocking assembly mediated by a total of 10symmetry-related, thermodynamically coupled interfaces. Assembly of the hexameric complex occurs sequentially: IL-6 is first engaged by IL-6Rα and then presented to gp130in the proper geometry to facilitate a cooperative transition into the high-affinity, signaling-competent hexamer. The quaternary structures of other IL-6/IL-12 family signaling complexes are likely constructed by means of a similar topological blueprint.

Gp130 is a shared signal-transducing receptor for a family of more than 10 different four-helix-bundle cytokines, including IL-6, leukemia inhibitory factor (LIF), ciliary neurotrophic factor (CNTF), oncostatin-M (OSM), and others (13). Gp130 engagement leads to activation of Src and JAK/Tyk tyrosine kinases, as well as the STAT family of transcription factors (4, 5). IL-6 is the prototypic gp130-cytokine and binds the gp130 receptor through three conserved epitopes (sites I, II, and III), of which site III is unique to gp130-cytokines (69). IL-6 must first form a complex with a nonsignaling α receptor, IL-6Rα, though site I. Site II is a composite epitope formed by the binary complex of IL-6 and IL-6Rα, which interacts with cytokine binding homology region (CHR; D2D3) of gp130. The subsequent interaction of site III with the gp130 immunoglobulin-like activation domain (D1 or IGD) forms the competent signaling complex (6, 8, 10, 11).

Whereas some functional studies indicate that IL-6–type signaling complexes contain two copies of cytokine, IL-6Rα, and gp130 (2:2:2), others suggest an alternative complex stoichiometry (1215). Recently, a tetrameric complex (2:2) between a viral IL-6 homolog and gp130 revealed a unique interaction that was inferred to be a mimic of the site III interaction of human IL-6 that would be found in a core template for the human IL-6 hexamer (11). However, an alternative model of the human IL-6 receptor complex, based on a recent crystal structure of the unliganded IL-6Rα, has been proposed that is at odds with the viral tetramer structural template (16).

Here we present a structural model, derived from 3.65 Å resolution crystallographic data (Table 1), of the complex between human IL-6, the extracellular binding domains of human IL-6Rα, and the extracellular activation and binding domains of gp130. The ternary complex forms a hexamer containing two IL-6, two IL-6Rα, and two gp130 that assemble sequentially and cooperatively. This hexameric structure reveals a conserved architectural blueprint for assembly of all gp130-cytokine signaling complexes.

Table 1.

Data collection and refinement statistics. ALS BL, Advanced Light Source Beamline.

Crystallographic statistics
Data collection
Space group R32
Unit cell (Å) (a, b, c) 279.8, 279.8, 96.7
Source ALS BL 8.2.1
Resolution (Å) (highest resolution shell) 50-3.65 (3.84-3.65)
Measured reflections 12,0957
Unique reflections 16,144
Completeness (%) 98.3 (98.1)
I/σ(I) 13.0 (1.9)
RmergeView inline(%) 0.065 (0.551)
Refinement statistics
Resolution range (Å) 20-3.65
R cryst View inline 0.33 (49.4)
R free View inline 0.28 (40.6)
No. of atoms (protein) 5,322
rms deviation from ideality
    Bond lengths (Å) 0.005
    Bond angles (°) 01.0
    Dihedral angles 22.4
    Improper angles 1.03
  • View inline* Rmerge = Σhkl|I — 〈I〉|/ΣhklI, where I is the intensity of unique reflection hkl, and 〈I〉 is the average over symmetry-related observation of unique reflection hkl.

  • View inline Rcryst = Σ|FobsFcalc|/ΣFobs, where Fobs and Fcalc are the observed and the calculated structure factors, respectively.

  • View inline Rfree is R with 5% of reflections sequestered before refinement.

  • We expressed a soluble complex of human IL-6, the three N-terminal domains of gp130 comprising the activation (D1) and CHR segments (D2D3), and the CHR (D2D3) of IL-6Rα from insect cells (Fig. 1A). Despite the low resolution of our data (Table 1), the availability of high-resolution crystal structures of the components [gp130 D1D2D3 (2.4 Å) (11), human IL-6 (1.9 Å) (17), and human IL-6Rα (2.4 Å) (16)], combined with the rigidity of the structural modules [<1.5 Å root mean square deviations (rmsds) from their unliganded structures], has resulted in a highly accurate model (Fig. 1B). Well-ordered electron density in the interfaces reveals many side-chain interactions (Fig. 2B). Indeed, previously disordered regions of the unliganded IL-6 structure (17) are now clearly defined in the complex (Fig. 1C).

    Fig. 1.

    Structure of the human Il-6/IL-6Rα/gp130 hexameric complex. (A) Schematized view of the domain structure of IL-6, IL-6Rα, and gp130. (B) Top view of a SigmaA-weighted 2FobsFcalc electron density map contoured at 2σ of the hexamer. The gp130 molecules are colored in cyan and blue, IL-6Rα molecules are colored green and purple, and IL-6 molecules are colored pink and red. The coloring scheme is maintained in all figures. (C) Tilted side view of the hexamer rotated ∼90° toward the viewer and tilted on the diagonal from (B). The five unique interfaces are labeled as sites I, IIa, IIb, IIIa, and IIIb. Figures were prepared with MOLSCRIPT (21) and Raster3D (22).

    Fig. 2.

    Structural anatomy of each interface within the hexameric assembly. (A) “Open book” surface representation of each of the interfaces with related contact surfaces colored purple (site I), red (site IIa), orange (site IIb), yellow (site IIIa), and gold (site IIIb). (B) Close-up view of the secondary structure of each interface including amino acid contact residues. One half of the hexamer, containing five interfaces, is shown within the circle; each interface is boxed in a different color, which is then expanded in a corresponding colored box surrounding the central circle. Residues that contribute the majority of buried surface area in each interface are drawn thicker and colored red. The overall buried surface area is indicated above each box. Electron density from a 2σ omit map of the site I molecular interface highlights well-ordered side chains. Figures were calculated with MOLSCRIPT (21) and Raster3d (22). Surface representations were calculated with VMD (23) and rendered with Raster3d (22).

    The overall shape of the hexamer is crudely reminiscent of a table with the tabletop composed of IL-6, IL-6Rα–D2, and gp130-D1D2 resting on legs composed of IL-6Rα–D3 and gp130-D3 (Fig. 1C). The hexamer is held together by 10 two-fold–related protein-protein interfaces, five of which are unique to each half of the hexamer (i.e., a dimer of five interfaces: sites I, sites IIa and IIb, sites IIIa and IIIb), that results in a total buried surface area of nearly 5500 Å2 (Fig. 2, A and B). The structure of the core of the hexamer (gp130, D1D2D3, and human IL-6) is reminiscent of the viral IL-6 tetramer complex. However, the addition of the IL-6Rα results in a much more complicated network of interfaces than is seen in the viral complex, rather than simply “decorating” the exposed site I of the human cytokine.

    The site I binding epitope of IL-6 is localized to the A and D helices and interacts with IL-6Rα to bury more than 1200 Å2 and form the initial IL-6/IL-6Rα binary complex (Fig. 2, A and B). The D3 domain of IL-6Rα provides the majority of contact surface with IL-6, contributing more that 70% of the total buried surface area. Of the 18 contact residues contributed by the IL-6Rα, Phe229 and Phe279 from the D3 domain dominate the binding interface, contributing 28% (174 Å2) and 20% (129 Å2), respectively, of the total buried surface area and dock into two cavities on the IL-6 surface (Fig. 2B). Mutational data support the structurally critical role of these two phenylalanine residues and have identified Phe229 as a “hotspot” residue at the interface [numbered Phe248 in (7)]. IL-6 also contributes numerous charge interactions within the site I interface surrounding the Phe residues, such as Arg179 and Lys171, which likely serve to endow this interface with hydrogen bonds that restrict its specificity compared to the promiscuous gp130 binding sites.

    The remaining four unique (or eight two-fold–related) protein-protein interfaces in the hexamer can be separated into two energetically coupled composite sites. The composite site II interaction of the IL-6/IL-6Rα complex with gp130 is separated into two spatially distinct interfaces: site IIa is between the IL-6 A and C helical faces and the “elbow” region at the boundary between the D2 and D3 domains of gp130 CHR, and site IIb is between the IL-6Rα D3 domain and the gp130 D3 domain (Fig. 2, A and B). In site IIa, Phe169 contributes the largest fraction (24%, 128 Å2) of the total buried surface area. This residue was originally identified as conserved in gp130, and as crucial for all cytokine interactions from mutational studies (8, 18). The binding interface of IL-6 is also decorated with large polar residues (Arg24, Lys27, Arg30, Asp34) that are flattened in the interface so that the polar head groups are directed outward toward solvent, whereas the nonpolar methylene side chains are used for hydrophobic interactions. In site IIb there is overall shape complementarity with the D3 domain of IL-6Rα, presenting a convex bulge that docks into an extended cavity on the D3 domain of gp130, burying ∼1078 Å2 of surface area.

    The structure of the composite site II interface reveals why IL-6 requires IL-6Rα to bind gp130. The additional surfaces provided by the site IIb interface enhance the overall binding affinity. Neither IL-6 nor IL-6Rα has measurable affinity for gp130 alone. Hence, the D3-D3 interface appears to act as a molecular brace to hold IL-6 against the gp130 CHR binding site.

    Gp130 cytokines possess a third receptor-binding epitope, site III, at one pole of the four-helix bundle, that is not found on other cytokines but is necessary for receptor activation within the gp130, or “IL-6–type,” cytokines. Site IIIa is a broad and discontinuous interface between IL-6 and gp130 where the tip of the IL-6 four-helix bundle (A/B loop and N-terminal region of D helix) abuts into the bottom β sheet of the D1 domain of gp130 and includes residues mapped to the site III epitope of IL-6 (Fig. 2, A and B). The existence of the site IIIb interface was not anticipated; here the tip of the gp130 D1 domain forms a large interaction surface with the side of the D2 domain of IL-6Rα that buries more than 500 Å2. The critical residue in the site IIIa interface is Trp157 of IL-6 (21% of total buried surface area, 141 Å2), which has been defined by mutagenesis as being the critical aromatic site III signature residue (12, 14). The solvent-exposed N-terminal peptide of gp130 (Leu2 through Cys6) intercalates within a groove on the surface of IL-6 formed by the AB and CD interhelical loops and contributes nearly 50% (∼300 Å2) of the buried contact surface from gp130 in site IIIa (Fig. 2, A and B). In total, 20 residues from IL-6 and 21 residues from gp130 make up site IIIa.

    The formation of the site IIIb interaction is enabled by the curvature of the IL-6 B and D helical axes that form site I (Fig. 1B). Because the site I on IL-6 is located toward one end of this curved surface, the IL-6Rα projects back at an angle when bound, diagonal to the gp130 long axis, resulting in a tapered longitudinal end of the hexamer and the close apposition of the side of IL-6Rα to the tip of the gp130 D1 domain (Fig. 1, B and C). Hence, site I and site IIIb appear to be structurally codependent. Site IIIb establishes a structural link between site I and site III, forming a closed loop of five codependent interfaces within each half of the hexamer (→ site I → site IIa/IIb → site IIIa → site IIIb → site I →).

    Assembly of the 10 interlocking interfaces in the hexamer requires a precise geometric alignment of the individual modules. We used isothermal titration calorimetry (ITC) to determine whether the formation of the final complex is cooperative. To deconvolute the site I, II, and III interfaces into discrete bimolecular interactions, we expressed a series of engineered soluble constructs (19). In the first assembly step, the site I interaction, binding of IL-6 to IL-6Rα has a large exothermic value, consistent with the structure of the site I interface, as well as mutational data, which reveal specific polar and charged interactions surrounded by nonpolar, knob-in-holes type of contacts that provide shape complementarity (Fig. 2B).

    For the composite site II interface, we used a single-chain version of the IL-6/IL-6Rα complex to reduce a three-body interaction problem into a two-body system. Titration of the IL-6/IL-6Rα complex with the gp130 D2D3 domain construct of gp130 results in a moderate-affinity [dissociation constant (Kd) ∼40 nM] complex with one copy of the single-chain construct and one copy of gp130 D2D3 (i.e., IL-6/IL-6Rα/gp130, 1:1:1) (Fig. 3). The weak enthalpy (–4.1 kcal/mol) and large favorable entropy [20.0 cal/(molK)] suggest that solvent has been excluded from the interface, which appears in the structure to be stabilized through primarily hydrophobic contacts. For the subsequent site III calorimetric titrations between single-chain IL-6/IL-6Rα and gp130 D1D2D3, we measured the additive effect of the composite site II (D2D3) and site III (D1) interfaces on the higher order assembly. The interaction of single-chain IL-6/IL-6Rα with gp130 D1D2D3 has a 50-fold higher affinity (Kd = 0.8 nM) than with gp130-D2D3, revealing cooperativity mediated through the D1 domain of gp130 (Fig. 3). Additionally, we measured a 90% larger heat capacity that is likely due to the increase in buried surface area in going from the trimolecular site II to the hexameric site II + III complexes (i.e., dimer of trimers). Hence, addition of the D1 domain results in a cooperative and sequential transition from the heterotrimer to the hexamer, which is achieved by the precise spatial presentation of individually moderate- and/or weak-affinity interfaces (Fig. 3).

    Fig. 3.

    Stepwise energetic and structural assembly of the functional human IL-6 hexamer signaling complex. Isothermal titration calorimetry (ITC) was used to measure the thermodynamic parameters for each step in the hexamer assembly pathway. The ITC titrations are designated as follows: site I: IL-6 and IL-6Rα forming a binary complex; site II: single-chain (sc) IL-6/IL-6Rα and gp130 D2D3 form a nonsignaling binary complex in the absence of the gp130 D1 domain; sites II and III: scIL-6/IL-6Rα and gp130 D1D2D3 form the signaling-competent hexamer; sites II and III 6 domain: scIL-6/IL-6Rα and gp130 D1D2D3D4D5D6 resulting in the hexamer with the three membrane-proximal domains of gp130. As discussed in the text, the single-chain version of the IL-6/Rα complex was used for these measurements to deconvolute the trimolecular equilibrium (i.e., IL-6 + IL-6Rα + gp130) into a bimolecular interaction event (i.e., single-chain IL-6/IL-6Rα + gp130). Thermodynamic parameters for each titration are provided in tabular format. Surfaces are calculated from the coordinates of the hexamer components. The D1 domains of IL-6Rα are included as schematic modules.

    Finally, we assessed whether the three membrane-proximal domains of gp130 have an impact on the assembly energetics (Fig. 3). We consistently found that the six-domain gp130 construct has a lower overall free energy in forming the hexamer (Fig. 3) than the three-domain construct in the crystal structure. Hence, the assembly energy of the hexamer is being used to hold the legs against one another, preventing the lower free-energy minimum achieved by the three-domain “legless” complex, which is unimpeded by this steric clash. Given the apparent interaction between the gp130 membrane-proximal domains revealed by ITC, combined with functional data showing cross-linking of the D5 domains during activation of cell-surface gp130 (20), we postulate that a bend exists between the D3 and D4 domains to enable the D4D5D6 domains to join in the final signaling assembly (Fig. 3). Further structural studies of the full-length assembly will be necessary to unambiguously determine the disposition of the gp130 legs in the cell-surface assembly. Although the elucidation of the hexameric structure discussed here enhances our understanding of signaling through shared cytokine receptors, the ultimate question remains how ligand engagement by the ectodomains leads to activation of intracellular signaling cascades.

    Supporting Online Material

    www.sciencemag.org/cgi/content/full/300/5628/2101/DC1

    Materials and Methods

    References

    References and Notes

    View Abstract

    Navigate This Article