The structure of the extracellular domains of human interleukin 11α receptor reveals mechanisms of cytokine engagement

Interleukin (IL) 11 activates multiple intracellular signaling pathways by forming a complex with its cell surface α-receptor, IL-11Rα, and the β-subunit receptor, gp130. Dysregulated IL-11 signaling has been implicated in several diseases, including some cancers and fibrosis. Mutations in IL-11Rα that reduce signaling are also associated with hereditary cranial malformations. Here we present the first crystal structure of the extracellular domains of human IL-11Rα and a structure of human IL-11 that reveals previously unresolved detail. Disease-associated mutations in IL-11Rα are generally distal to putative ligand-binding sites. Molecular dynamics simulations showed that specific mutations destabilize IL-11Rα and may have indirect effects on the cytokine-binding region. We show that IL-11 and IL-11Rα form a 1:1 complex with nanomolar affinity and present a model of the complex. Our results suggest that the thermodynamic and structural mechanisms of complex formation between IL-11 and IL-11Rα differ substantially from those previously reported for similar cytokines. This work reveals key determinants of the engagement of IL-11 by IL-11Rα that may be exploited in the development of strategies to modulate formation of the IL-11–IL-11Rα complex.

Interleukin (IL) 11 activates multiple intracellular signaling pathways by forming a complex with its cell surface ␣-receptor, IL-11R␣, and the ␤-subunit receptor, gp130. Dysregulated IL-11 signaling has been implicated in several diseases, including some cancers and fibrosis. Mutations in IL-11R␣ that reduce signaling are also associated with hereditary cranial malformations. Here we present the first crystal structure of the extracellular domains of human IL-11R␣ and a structure of human IL-11 that reveals previously unresolved detail. Disease-associated mutations in IL-11R␣ are generally distal to putative ligand-binding sites. Molecular dynamics simulations showed that specific mutations destabilize IL-11R␣ and may have indirect effects on the cytokine-binding region. We show that IL-11 and IL-11R␣ form a 1:1 complex with nanomolar affinity and present a model of the complex. Our results suggest that the thermodynamic and structural mechanisms of complex formation between IL-11 and IL-11R␣ differ substantially from those previously reported for similar cytokines. This work reveals key determinants of the engagement of IL-11 by IL-11R␣ that may be exploited in the development of strategies to modulate formation of the IL-11-IL-11R␣ complex.
Interleukin 11 (IL-11) is a member of the IL-6 family of cytokines, which includes IL-6, leukemia inhibitory factor, oncostatin M, ciliary neutrophilic factor, IL-27, IL-31, cardiotrophin-1, cardiotrophin-like cytokine, and neuropoietin (1). Activation of downstream signaling pathways by these cytokines is generally initiated via the formation of oligomeric receptor complexes that include the ␤-subunit signaling receptor, gp130, and one or more cytokine-specific co-receptors (2,3). The majority of our structural and mechanistic understanding of this cytokine family is based on structural information available for IL-6, leukemia inhibitory factor, and their receptors (4 -6).
Characterization of the in vivo source of IL-11 has only recently begun, as a result of emerging links to multiple pathologies. IL-11 has classically been associated with hematopoiesis (7); however, it has more recently been identified as the major cytokine involved in gastrointestinal tumorigenesis and is a promising therapeutic target (8). IL-11 also has emerging roles in cardiovascular and liver fibrosis (9,10). Mutations in the IL-11-specific ␣-receptor, IL-11R␣, have gained increased interest as a result of their causative role in hereditary diseases that are typified by craniosynostosis and delayed tooth eruption (11)(12)(13)(14). Several of these mutations have been shown to impair IL-11 signaling in vitro (11).
Following secretion, IL-11 is believed to interact with IL-11R␣, which is expressed in tissue-specific cell populations (15). This binary complex is thought to subsequently engage gp130 (6,16). Previous mutagenesis and structural studies indicate that IL-11 interacts with its receptors through three independent sites on its surface (17). Site I is responsible for IL-11R␣ binding; site II binds a gp130 molecule and contributes to the formation of a trimeric complex; and site III engages with a second gp130 molecule, resulting in the cooperative formation of a hexameric signaling complex containing two copies of each component (16).
Upon formation of the signaling complex, Janus kinases (JAKs) associated with the cytoplasmic regions of gp130 are activated, although the exact mechanisms of activation remain unclear (18). Because IL-11R␣ does not bind JAKs at its cytoplasmic domain, signaling is thought to result from transactivation of JAK molecules bound to the cytoplasmic domains of the two gp130 molecules of the hexameric signaling complex. JAK activation then leads primarily to phosphorylation and activation of signal transducer and activator of transcription (STAT) 3. Activation of other signaling pathways, including the extracellular signal-regulated kinase/mitogen-activated protein kinase pathway and the phosphatidylinositol 3-kinase pathway, is less well-understood.
The structural basis of IL-6 signaling has been well-studied, and the structure of the hexameric IL-6 signaling complex has This article contains supporting information. ‡ These authors contributed equally to this work. * For correspondence: Michael D. W. Griffin, mgriffin@unimelb.edu.au. cro ARTICLE been solved (6). Low-resolution EM studies of the IL-11 signaling complex suggest that the overall arrangement is likely similar to that of IL-6 (19). We previously reported the first crystal structure of human IL-11 (17) and showed that although the topology is similar to IL-6, IL-11 is significantly elongated, suggesting different geometry of the signaling complex. Despite the growing biological importance of IL-11 signaling, molecular understanding of the structure and assembly of the IL-11 signaling complex remains in its infancy.
Here, we present the first crystal structure of human IL-11R␣ and a new, more complete structure of IL-11 that reveals structural details of functionally important regions. Disease-associated mutations in IL-11R␣ are generally located distal to putative binding surfaces of the receptor. Molecular dynamics simulations reveal the mechanisms by which several of these mutations disrupt the structure of IL-11R␣ and thereby prevent signaling. We present a model of the IL-11-IL-11R␣ complex and in combination with biophysical and mutagenic characterization of the cytokine-receptor interaction show that IL-11R␣ and IL-6R␣ engage their cognate cytokines with similar affinities but use surprisingly different thermodynamic and structural mechanisms. Our work provides structural and mechanis-tic detail of the first step of formation of the IL-11 signaling complex that may be exploited in the development of molecules that can modulate complex formation.

The structure of the extracellular domains of the interleukin 11 ␣-receptor
The complete extracellular region of IL-11R␣ (IL-11R␣ EC ; residues 1-341 of the mature protein after signal peptide cleavage) was expressed in the insect cell line Sf21 and purified from the cell culture supernatant. To reduce formation of disulfide-linked dimers, the C226S mutation (20) was present in all IL-11R␣ constructs described in this work. Crystals of IL-11R␣ EC were in space group P6 5 22. Initial phase estimates were obtained by molecular replacement using domains from unpublished Fab-bound structures of IL-11R␣, and the structure was refined at a resolution of 3.43 Å (PDB code 6O4P). Data and refinement statistics are presented in Table 1, and representative electron density is shown in Fig. S1 (A and B).
The structure of IL-11R␣ EC consists of an N-terminal Ig-like domain (D1) and two fibronectin type III (FnIII) domains (D2, The values for the highest resolution shell are given in parentheses.

Data collection
Pearson correlation coefficient between independently merged half data sets.
D3) that form the cytokine-binding homology region (CHR) (Fig. 1A). By homology to other cytokine receptors, IL-11 likely binds to the loops present at the D2/D3 junction (Fig. 1A). The receptor is L-shaped, with D2 and D3 forming the CHR. The arrangement of the three domains is similar to other IL-6 family cytokine receptors (Fig. S1C). The ␣-carbon RMSD between IL-11R␣ and IL-6R␣ (PDB code 1N26) is 5.5 Å, and that between IL-11R␣ and D1-D3 of gp130 (PDB code 1I1R) is 5.8 Å, indicating moderate structural similarity. The primary deviations between the three structures are in the position of D1. The putative cytokine-binding region in IL-11R␣ shows less surface charge than that of IL-6R␣ (Fig. S1D). Two protein molecules are present in the asymmetric unit (␣-carbon RMSD of 2.0 Å), forming a crystallographic dimer in a "head-to-head" configuration through an interaction between D2 of each receptor molecule (Fig. 1B). The C terminus of the A, two views of the structure of IL-11R␣ EC . Each of the domains and the section of the C-terminus that is defined in the electron density are indicated. The transmembrane domain is at the C-terminal region of the receptor. B, the asymmetric unit of the IL-11R␣ crystal structure, formed by two IL-11R␣ molecules, with an extensive contact between D2 of the two molecules. C, the structure (left panel) and topology (top right panel) of D1 from chain B of IL-11R␣ EC with disulfide bonds indicated. Loops are colored pink, the two strands contributing to the smaller, anti-parallel ␤-sheet are blue, and the five strands contributing to the larger, mixed parallel/anti-parallel ␤-sheet are orange. A topology diagram of the typical s-type Ig domain is also shown (bottom right panel). D, continuous sedimentation coefficient (c(s)) distributions for IL-11R␣ EC at three concentrations, showing that IL-11R␣ EC is primarily monomeric in solution under the conditions tested. Slight concentration dependence in the sedimentation coefficient suggests the formation of a transient oligomer. E, c(s) distributions for IL-11R␣ D1-D3 at several concentrations. F, small-angle X-ray scattering data for IL-11R␣ D1-D3 , overlaid with the theoretical scattering profile calculated from molecule A of the crystal structure of IL-11R␣ EC ( 2 ϭ 1.05).

Structure of human interleukin 11␣ receptor
receptor is more complete in chain A, forming a crystal contact with a protein molecule in a neighboring asymmetric unit. The absence of density for the complete C terminus may be a result of disorder or caused by the presence of endoproteinase Glu-C during the crystallization experiment. N-Linked glycans are observed at Asn 105 and Asn 172 . D1 of IL-11R␣ forms an Ig-like domain with an unusual s-type topology (22) (Fig. 1C). Strand A in the ␤ sandwich forms a non-canonical mixed parallel/anti-parallel ␤ sheet with strands G, F, C, and D. This is a similar overall topology to D1 of IL-6R␣; in both cases the Ig-fold is distorted (21). Two disulfide bonds are present in D1: one between Cys 26 in the strand B/strand C linker and Cys 72 in strand F. A disulfide bond in a similar position is present in the D1 of IL-6R␣ (21). A second disulfide bond is present between Cys 4 and Cys 25 in strand B, which was not predicted from sequence analysis or homology to other receptors. The disulfide bond is well-supported in the electron density and confirmed in a simulated-annealing omit map (Fig. S1A). The unusual fold of D1 may be a consequence of these disulfides, with the Cys 4 -Cys 25 disulfide serving to sterically constrain strand A, preventing the formation of a typical anti-parallel ␤-sheet with strands B and E.
D2 of IL-11R␣ contains the two disulfide bonds expected for this domain (between Cys 98 and Cys 108 and between Cys 148 and Cys 158 ). D3 of IL-11R␣ contains the conserved tryptophanarginine ladder (comprising tryptophan residues 246, 282, and 285 and arginine residues 235, 239, 270, and 274), which includes the strongly conserved WSXWS sequence motif. Like other cytokine receptors, the sequence containing the WSXWS motif forms a short polyproline type II helix that is stabilized by side chain-main chain interactions and the tryptophanarginine ladder.
The interface formed between the two IL-11R␣ monomers in the asymmetric unit of the crystal structure has a buried surface area of 1088 Å 2 ( Fig. 1B) (23). To establish whether IL-11R␣ EC self-associates in solution, we used sedimentation velocityanalytical ultracentrifugation (SV-AUC) at protein concentrations of 6.5-19.5 M (0.25-0.75 mg/ml) ( Fig. 1D and Fig. S1E, panel i). These experiments show that IL-11R␣ EC is predominantly monomeric in solution with a standardized sedimentation coefficient (s 20,w ) of 2.70 at 13.0 M. This represents a molecular mass of 41.3 kDa, with a frictional ratio (f/f 0 ) of 1.57 calculated from the fit to the SV data (Fig. S1E, panel i), in good agreement with the expected molecular mass from the sequence (38.2 kDa). The theoretical sedimentation coefficient, calculated from the crystal structure coordinates of chain A using HYDROPRO (24) was 2.92, consistent with the experimental value. A small, concentration-dependent increase in weight-average sedimentation coefficient was observed, from 2.67 S at 6.5 M to 2.72 S at 19.5 M, likely indicating the formation of a weak-affinity dimer. It is possible that any weakaffinity dimerization is increased at the cell membrane, where the receptor may be concentrated in lipid rafts, analogous to other cytokine receptors (25, 26), and can diffuse in only two dimensions, increasing its effective concentration.
To study the solution properties of IL-11R␣ without the C-terminal extension, we generated a construct comprising only domains D1-D3 (IL-11R␣ D1-D3 ; residues 1-297 of the mature protein). The standardized sedimentation coefficient of IL-11R␣ D1-D3 measured at a protein concentration of 15.5 ⌴ (0.5 mg/ml) was 2.62 S (Fig. 1E and Fig. S1E, panel ii), corresponding to a molecular mass of 34.9 kDa, with a f/f 0 value of 1.47, in good agreement with the sequence molecular mass (32.1 kDa). Similar to IL-11R␣ EC , a small concentration-dependent increase in weight average sedimentation coefficient was observed from 2.62 S at 7.8 M to 2.69 S at 23.3 M, suggesting that possible weak dimerization is mediated by the structured, extracellular domains of IL-11R␣ and is not a consequence of the disordered C terminus. The small-angle X-ray scattering (SAXS) profile of IL-11R␣ D1-D3 agrees well with the monomer of the crystal structure coordinates ( 2 ϭ 1.05) (Fig. 1F, Table  S1, and Fig. S1F), confirming that the crystal structure accurately represents the solution conformation of the structured, extracellular domains of IL-11R␣.

Pathogenetic mutations disrupt the structure of IL-11R␣
A number of pathogenic mutations have been identified in the gene for IL-11R␣, IL11RA, resulting in point-substitution mutations in IL-11R␣ that cause a genetic disease featuring craniosynostosis and delayed tooth eruption (11)(12)(13)(14). Mapping the disease-associated mutations onto our structure of IL-11R␣ indicates that very few of the mutations are in the putative IL-11 or gp130 binding sites ( Fig. 2A). The P178T, P199R, and R274W mutations have previously been studied in vitro (11); however, the lack of structural information on IL-11R␣ has hindered understanding of the molecular impact of the mutations.
To investigate the effects of the mutations on the structure of IL-11R␣, we ran a series of short (50 ns) all-atom molecular dynamics (MD) simulations on IL-11R␣ (Fig. 2B) and several of the disease mutants. In IL-11R␣, the C␣ RMSD and backbone amide bond order parameters (S 2 ) calculated from the MD trajectory indicate a low level of local disorder and overall local rigidity within each of the three domains ( Fig. 2B and Fig. S2A). However, the three domains were dynamic with respect to each other throughout the simulation ( Fig. 2B and Fig. S2B). The loops comprising the putative IL-11 binding site were relatively rigid and do not undergo large motions on the time scale of the simulation.
MD simulations of IL-11R␣ with the disease-associated mutations suggest that several of them destabilize key structural elements in the receptor or destabilize interdomain interfaces. One mutation, C72F, removes a disulfide bond in D1, which likely has a role in stabilizing the unusual Ig fold of D1. Introducing this mutation to D1 resulted in the loop joining strands F and G adopting a markedly different conformation, which may alter the stability of the domain (Fig. 2C, Fig. S2C, and Movie S1).
A second mutation, P178T, is located in a loop in D2 that faces D1. This mutation resulted in a shift in the relative pose of D1 and D2, likely because of removal of the interaction of Pro 178 with a pocket on D1 that stabilizes the D1/D2 interface. However, in each replicate simulation, the final relative orientation of D1 and D2 differed. (Fig. 2D, Fig. S2, D and E, and Movie S2).
The R274W mutation is situated within the tryptophanarginine ladder in D3 of the receptor. This mutation destabi-lized the tryptophan-arginine ladder and resulted in the destabilization of the membrane-distal region of D3. Arg 274 also contributes to a hydrogen-bonding network at the D2/D3 interdomain interface in the WT receptor (Fig. 2E, Fig. S2F, and Movie S3). The mutation thus destabilizes the D2/D3 linker and results in an increase in flexibility at the D2/D3 interface, potentially disrupting the IL-11-binding interface and reducing cytokine affinity.

Structure of human interleukin 11␣ receptor
The P199R mutation is located in the D2-D3 interdomain linker. The mutation causes a slight increase in the D2-D3 interdomain distance but does not otherwise greatly alter the interdomain pose or dynamics of IL-11R␣ (Fig. 2F, Fig. S2G, and Movie S4).
Several other pathogenic mutations have little appreciable impact on the structural dynamics within the time scale of the simulation. For example, P43T does not greatly alter the flexibility of the affected loop in D1, C108S does not appear to significantly alter D2 through the simulation, nor does R239C destabilize D3 or the tryptophan-arginine ladder in which it is situated (Fig. S2, H-J). One mutation (H276R) is close to the putative gp130-binding region of D3 and thus may act by directly altering signaling complex formation at this interface.
Together our simulations show that the effect of a subset of the craniosynostosis mutations in IL-11R␣ is to destabilize the structure of IL-11R␣. The P178T, R274W, and P199R mutations have previously been shown to result in incomplete glycosylation, leading to retention in the endoplasmic reticulum and poor cell surface expression, contributing to reduced IL-11-mediated STAT3 activation (11). Our results suggest that destabilization of the structure caused by the P178T and R274W mutations is sufficient to stall correct trafficking of the receptor. D1 of IL-6R␣ has previously been shown to be involved in intracellular trafficking of the receptor (27). Thus, destabilization of D1 or the D1/D2 interface in IL-11R␣ by the P178T mutation may result in a lack of correct processing of the receptor. In the case of the R274W and P199R mutations, destabilization of the cytokine-binding surface at the junction between D2 and D3 may also reduce the IL-11-binding capacity of mutant IL-11R␣ that is correctly expressed at the cell surface, further reducing the potential for formation of the active signaling complex. The apparently minor structural effects of some mutations, such as P199R, that are positioned distal to the putative cytokine-and gp130-binding regions of the receptor, suggest alternative mechanisms that impair IL-11 signaling, for example, disruption of protein expression or global receptor folding.

The high-resolution structure of IL-11
In our previous structure of human IL-11, parts of the long loops between helices A and B and between helices C and D were poorly defined (17). Mutagenesis suggests that the AB loop is involved in binding IL-11R␣ (28), and in the structure of the IL-6 signaling complex, the AB loop forms contacts to both IL-11R␣ and gp130 (6). To gain insight into these loops, we solved a higher-resolution structure of IL-11.
To facilitate growth of crystals that diffracted to high resolution, we truncated IL-11 by 10 residues at the N terminus. We named this new construct IL-11 ⌬10 (residues 11-178 of the mature protein) and the full-length protein IL-11 FL . Both IL-11 FL and IL-11 ⌬10 have similar high thermal stability, as measured by differential scanning fluorimetry (Fig. S3A) (29). Stimulation of human colon cancer cell line, DLD1 with either IL-11 ⌬10 or IL-11 FL results in similar levels of activation of STAT1 and STAT3 (Fig. 3A), indicating that they have similar biological activity. We note that N-terminally truncated IL-11 constructs have been used previously with no reported alteration in biological activity (30,31).
Crystals of IL-11 ⌬10 were rod-like plates in space group P2 1 2 1 2. Initial phase estimates were obtained by molecular replacement using our previous structure of IL-11 (PDB code 4MHL), and the new structure was refined at a resolution of 1.62 Å (PDB code 6O4O; see summary statistics in Table 1 and representative electron density in Fig. S3B). Overall, the structure of IL-11 ⌬10 is similar to our previously solved structure of IL-11 (RMSD 1.5 Å, Fig. S3C), forming a typical cytokine four-␣-helical bundle (Fig. 3B). The three receptor-binding sites of the cytokine are not significantly altered in the structure (Fig.  3C) (17). A cis proline (Pro 103 ) is observed at the C-terminal end of the 3 10 helical section of helix C. The equivalent proline in our previous structure of IL-11 is in the trans configuration (Fig. S3D). Both proline isomers are strongly supported by electron density in their respective structures, suggesting that Pro 103 can adopt either the cis or trans isomer and that the 3 10 helix is dynamic in solution.
Our high-resolution structure of IL-11 ⌬10 allows the extended loops joining helices A and B and helices C and D to be included in the model. The AB loop is formed by 26 residues between Phe 43 and Leu 69 (Fig. 3D). The position of the loop is stabilized by a hydrogen bond between Ser 53 and His 86 in helix B, and this region of the loop is thus well-defined in the election density. Mutagenesis has previously implicated the C-terminal end of the loop in receptor binding (28). This portion of the loop is adjacent to site I and poorly defined in the electron density.
The CD loop of IL-11 forms an unusually long polyproline type II (PP2) helix (Fig. 3E), comprising 14 residues. The CD loop is stabilized by several contacts between the loop and the core of the cytokine (Fig. 3E). To our knowledge, an equivalently long polyproline helix has not been observed in the structure of any other cytokine. The role of the PP2 helix is likely structural, to efficiently join the C-terminal end of helix C and the N-terminal end of helix D, which are 44 Å apart, with a relatively short sequence of 21 residues.
To further study the dynamic nature of the loops of IL-11, we ran a series of short (100 ns) molecular dynamics simulations on IL-11 (Fig. 3F). In the time scale of the simulation, the four-␣-helical bundle was stable and did not undergo large movements ( Fig. 3F and Fig. S3E). In agreement with NMR studies of other IL-6 family cytokines, the ␣-helices showed "helical fraying" and were more dynamic at the ends of the helices, compared with the core (Fig. 3F) (32,33). The PP2 helix structure of the CD loop was preserved throughout the simulation, although the loop underwent lateral movements. The AB loop was generally highly dynamic on the time scale of the simulation, although the central portion of the loop was stabilized by interactions with the ␣-helical core. The C-terminal end of the loop, which is implicated in IL-11R␣ binding, was highly dynamic on the time scale of the simulation.
We also used SV-AUC to show that IL-11 ⌬10 is monomeric, with no concentration-dependent increase in sedimentation coefficient. The sedimentation coefficient was measured as 1.7 S (Fig. 3G and Fig. S3F), representing a molecular mass of 17.2 kDa (f/f 0 1.28), in agreement with the sequence molecular mass Structure of human interleukin 11␣ receptor (18.2 kDa). The theoretical sedimentation coefficient calculated from the crystal structure was 1.8 S, in good agreement with the experimental value. SAXS data for IL-11 ⌬10 also agrees well with the theoretical scattering profile calculated for the crystal structure coordinates ( 2 ϭ 1.43) (Fig. 3H, Table S1, and Fig. S3G). These experiments confirm that IL-11 ⌬10 is monomeric in solution. Similar experiments show that IL-11 FL is monomeric in solution (Fig. S4, A-D).

Structure of human interleukin 11␣ receptor IL-11 and IL-11R␣ interact with nanomolar affinity
We used SV-AUC to investigate the interaction between IL-11 and IL-11R␣. For these experiments, the complex was formed by mixing 5 M IL-11 ⌬10 and 5 M IL-11R␣ EC immediately prior to the experiment, with no further purification. The appearance of a peak in the c(s 20,w ) distribution with a sedimentation coefficient of 3.2 S, larger than either IL-11 ⌬10 and IL-11R␣ EC alone, indicated formation of a complex between IL-11R␣ EC and IL-11 ⌬10 (Fig. 4A, panel i). The estimated molecular mass of this species was 60.8 kDa, with f/f 0 of 1.71, consistent with a complex forming with 1:1 stoichiometry (Fig.  S5A). A similar complex was formed between IL-11 ⌬10 and IL-11R␣ D1-D3 (Fig. 4A, panel   We used multi-angle light scattering coupled with size-exclusion chromatography (SEC-MALS) to provide additional evidence for the formation of a 1:1 complex between IL-11 and IL-11R␣. We measured the absolute molecular mass of IL-11 ⌬10 as 21.0 kDa, that of IL-11R␣ D1-D3 as 36.3 kDa (Fig.  S6B), and that of the IL-11 ⌬10 -IL-11R␣ D1-D3 complex as 51.1 kDa (Fig. 4B), consistent with a 1:1 complex.
To determine the dissociation constant for the IL-11-IL-11R␣ interaction, we used fluorescence-detected SV-AUC (FD-AUC), which can accurately measure proteins present at nanomolar and picomolar concentrations (34). We expressed IL-11 FL N-terminally fused to a monomeric, ultrastable GFP (muGFP) (35). SV-AUC showed that muGFP-IL-11 is monomeric across a wide concentration range (Figs. S6C and S8E) and forms a complex with IL-11R␣ EC in a 1:1 stoichiometry at concentrations of 5 M of each component ( Fig. 4C and Fig.  S5B). Complex formation was apparent at concentrations of IL-11R␣ EC in the nanomolar range, with two peaks observed in c(s 20,w ) distributions corresponding to free muGFP-IL-11 and muGFP-IL-11 in complex with IL-11R␣ EC (Fig. 4D and Fig.  S5C). We generated a sedimentation coefficient isotherm for the titration of IL-11R␣ EC against muGFP-IL-11, which, when fit to a 1:1 binding model, gave a K D of 22 nM (68% confidence interval, 14 -35 nM) ( Fig. 4E and Fig. S5C). This is consistent with the dissociation constant for similar site I cytokine-␣ receptor interactions. For example, IL-6 and IL-6R␣ interact with a K D of 9 nM (6), IL-2 and IL-2R␤ interact with a K D of 144 nM (36), and IL-7 interacts with IL-7R␣ with a K D of ϳ50 nM (37). In each of these cases, the complete signaling complex is formed by further high-affinity interactions between the cytokine-␣ receptor complex and other receptors. These experiments show that the IL-11-IL-11R␣ EC interaction also fits into this paradigm; an initial low-nanomolar affinity step to form the complex between IL-11 and IL-11R␣ occurs first, allowing subsequent engagement by gp130.
The tendency of GFP to form weakly associating dimers with a K D of ϳ100 M has previously limited the use of GFP in quantitative biophysical experiments (38). The monomeric, ultrastable GFP used here does not detectably dimerize (35), allowing it to be used as a genetically encoded fluorescent tag for biophysical experiments. Previous efforts to use FD-AUC to measure high-affinity protein-protein interactions have generally relied on covalent modification of one of the interacting part-ners with a fluorescent dye, with previous studies noting that the use of covalent dyes as fluorescent labels alters the binding properties of the proteins under investigation (39). The use of a genetically encoded, monomeric fluorescent fusion tag overcomes this limitation, allowing the accurate measurement of nanomolar-affinity dissociation constants in the analytical ultracentrifuge, without requiring the covalent modification of one of the proteins involved in the interaction.

The IL-11-IL-11R␣ interaction is entropically driven
We used isothermal titration calorimetry (ITC) to complement our FD-AUC binding experiments above and to examine the thermodynamic basis of cytokine-receptor engagement ( Table 2). ITC showed that IL-11 ⌬10 interacts with IL-11R␣ EC and IL-11R␣ D1-D3 with similar affinities, with K D values of 40 Ϯ 20 and 23 Ϯ 3 nM, respectively (n ϭ 3, standard error; Fig. 5A, panels i and ii). These values are consistent with our SV-AUC experiments and show that the C-terminal extension of IL-11R␣ does not affect IL-11 binding. We also measured the affinity for the interaction between IL-11 FL and IL-11R␣ EC , K D of 55 Ϯ 14 nM (n ϭ 3, standard error; Fig. S6D), showing that deletion of the N terminus of IL-11 does not significantly alter affinity for IL-11R␣ (p ϭ 0.58). The thermodynamics of the IL-11 ⌬10 -IL-11R␣ D1-D3 interaction are strongly driven by entropy (⌬H ϭ Ϫ25 Ϯ 2 kJ/mol, ⌬S ϭ 66 Ϯ 7 J/(mol⅐K)). We also measured the IL-11 ⌬10 -IL-11R␣ D1-D3 interaction using ITC at two additional temperatures (283 and 298 K) to determine the heat capacity of the reaction, ⌬C p (Fig. S6E, panels i-iii, and Table 2). The heat capacity was measured as Ϫ3.3 Ϯ 0.07 kJ/(mol⅐K) (means Ϯ S.E.). An empirical relationship exists between heat capacity and total buried surface area, a large negative ⌬C p being consistent with a large buried surface area (40 -42). This suggests that the IL-11-IL-11R␣ interaction is hydrophobic in nature, resulting in the burying of a large hydrophobic surface.
The cytokine-binding site of IL-11R␣ EC lacks large charged or hydrophilic regions, consistent with a hydrophobic interaction that is primarily driven by a positive change in entropy. This contrasts strongly with the IL-6 -IL-6R␣ interaction, which is strongly exothermic, with a corresponding unfavorable entropy change (⌬H Ϫ100 kJ/mol, ⌬S Ϫ192 J/(mol⅐K) at 10°C), a consequence of the structural differences between the two cytokines and receptors (6). Thus, despite apparent structural similarity, IL-6R␣ and IL-11R␣ employ different thermodynamic mechanisms to engage their cognate cytokines.

Structure of human interleukin 11␣ receptor A model of the IL-11-IL-11R␣ binary complex provides detail of the structural mechanism of engagement
Cytokines generally bind to the CHR surface at the junction between FnIII domains D2 and D3, with D3 also involved in interacting with other receptors comprising the complete signaling complex (1). This region of IL-11R␣ is made up of four loops, formed by residues 98 -106 (between strands A and B), residues 129 -145 (between strands C and D) and 160 -169 (between strands D and E) in D2, and residues 220 -232 (between strands B and C) in D3. Part of the loop between strands C and D of D2 (residues 132-139 of chain A and 132-141 of chain B) was not defined in the electron density. To our knowledge, a similar large and disordered loop in the CHR has not yet been described for any other cytokine receptor. The membrane-proximal region of D3 serves to engage gp130, to complement the site II interaction on the cytokine. This region is similar in topology and surface charge in both IL-6R␣ and IL-11R␣, suggesting that the mechanism of ␣-receptor engagement with gp130 is similar between the two receptors.
The configuration of the CHR differs between IL-11R␣ and IL-6R␣ (Fig. S1C, panel i). In IL-11R␣, the relative positioning of D2 and D3, which is more similar to that of gp130 (Fig. S1C, panel ii), creates a smaller cytokine-binding surface than IL-6R␣. The electrostatic surface potential in the cytokinebinding sites also differ between the two proteins (Fig. S1D). The IL-6 -binding site in IL-6R␣ is noticeably more charged than that of IL-11R␣, with a negatively charged patch formed by several acidic residues in the loop formed between strands F and G in D3, which mediate a number of electrostatic contacts to IL-6 in the IL-6 signaling complex (Fig. S1D) (6, 21). These structural differences suggest that IL-11R␣ employs different structural mechanisms from IL-6R␣ to engage its cognate cytokine at Site-I.
To investigate the structural mechanism of IL-11 binding by IL-11R␣, we constructed a model of the IL-11-IL-11R␣ complex. Using the structure of the IL-6 signaling complex (PDB code 1P9M (6)), we aligned IL-11 and IL-11R␣ to their homologous chains in the IL-6 complex and refined this model using RosettaDock of the Rosie server (43,44). Models were scored using RosettaDock, and the top-scoring model was taken as the representative model (Fig. 5B). An overlay of the initial model and the final model is shown in Fig. S9A, the top 10 scoring models are shown in Fig. S9B. Relative to the initial model, the docked model shows a significant rotation of the pose of cytokine with respect to the binding site on the receptor. The model shows that the missing CD loop in D2 of IL-11R␣, which we did not include in the model, is in close proximity to the binding site.
Our model has a buried surface area of 567 Å 2 at the interface between IL-11 and IL-11R␣, similar to that of the IL-6 -IL-6R␣ interface in the IL-6 signaling complex (706 Å 2 ). This is consistent with the initial cytokine-receptor interaction forming a transiently stable complex. The pose of D2 with respect to D3 of IL-11R␣ is different from that of IL-6R␣, resulting in a differently shaped cytokine-binding surface (Fig. S9C), which may account for the small difference in buried surface area. The binding mode of the cytokine is overall similar, consistent with previous mutagenesis on IL-11 and our structure of IL-11 (17,(45)(46)(47).
SAXS analysis of the IL-11-IL-11R␣ complex supports the docked model. The complex was formed by mixing IL-11R␣ D1-D3 and IL-11 ⌬10 at an equimolar ratio, prior to SAXS measurement. The molecular mass was measured as 50.1 kDa, consistent with a 1:1 complex, and in agreement with the mass and stoichiometry determined by SV-AUC and SEC-MALS (Table S1). Theoretical scattering for the docked model fits the experimental SAXS data well ( 2 ϭ 1.03) (Fig. 5C, Table  S1, and Fig. S7A), showing that the model accurately represents the overall shape of the binary IL-11R␣-IL-11 complex. Similarly, the model agrees well with an ab initio model of the complex, generated using DAMMIN (Fig. S9, D and E). Likewise, the theoretical sedimentation coefficient (3.3 S) matches the experimentally determined sedimentation coefficient of the IL-11R␣ D1-D3 -IL-11 ⌬10 complex (3.3 S) (Fig. 4A, panel ii), further supporting the 1:1 stoichiometry of the complex.
We used the PISA server to analyze the interactions formed between the two proteins in the docked model. The major interacting residues of IL-11 are Arg 33 , Met 59 , Ala 61 , Gly 62 , and several residues in the C terminus of the cytokine, particularly Arg 169 (Fig. 5B). Arg 33  Because Arg 169 of IL-11 makes a key intermolecular contact in our model, we constructed and purified the IL-11 ⌬10/R169A mutant. SV-AUC analysis of 5 M IL-11R␣ D1-D3 in the presence of 5 M IL-11 ⌬10/R169A results in a peak in the sedimentation coefficient distribution of ϳ2.7 S (Fig. 5D and Fig. S8A), less than that for the IL-11 ⌬10 -IL-11R␣ D1-D3 complex (3.3 S), showing that the R169A mutation substantially decreases affinity for IL-11R␣. Stimulation of DLD1 cells with IL-11 ⌬10/R169A showed greatly reduced potency in activation of STAT3 than the WT cytokine (Fig. S7B), consistent with the reduction in IL-11R␣ binding leading to impaired formation of the active signaling complex. Residues important for biological activity of IL-11 have previously been identified by site-directed mutagenesis of human and mouse IL-11 (45)(46)(47), and these mutagenesis experiments further support our model (Fig. S9F). Substitution of Arg 33 , Asp 165 , Trp 166 , Arg 169 , Leu 172 , and Leu 173 all reduce the biological activity of IL-11 (45)(46)(47). The N-terminal region of the AB loop of IL-11, containing the interacting residues Met 59 , Ala 61 , and Gly 62 has previously been targeted by phage display and mutagenesis to alter the binding of IL-11 to IL-11R␣; thus, this region has also been shown to be key for the interaction.
Our model predicts that IL-11 forms interfaces of 225 and 369 Å 2 with D2 and D3, respectively. Previously, the isolated D3

Structure of human interleukin 11␣ receptor
of IL-11R␣ was reported to bind IL-11 with an affinity of 48 nM (20). We expressed, purified, and refolded D3 of IL-11R␣ (IL-11R␣ D3 ; residues 192-315 of the mature protein) from Escherichia coli inclusion bodies. SV-AUC analysis of IL-11R␣ D3 showed a single, narrow peak in the c(s 20,w ) distribution with a sedimentation coefficient of 1.5 S (calculated from the fit to the data at 28 M) and no concentration-dependent change (Figs. S7C and S8F), indicating a homogenous product that did not self-associate in the concentration range measured. CD spectra of the refolded protein showed a characteristic all-␤ spectrum, with a positive peak at 230 nm, likely because of -stacking interactions in the WSXWS motif (48) (Fig. S7D). 15 N-1 H heteronuclear single quantum coherence spectra from the purified, refolded IL-11R␣ D3 were well-dispersed and showed seven resolved tryptophan indole NH resonances of different intensities and line widths, indicating that the protein is folded (Fig. S7E).
SV-AUC analysis of IL-11R␣ D3 (5 M) with increasing concentrations of IL-11 ⌬10 showed no concentration-dependent increase in sedimentation coefficient, with weight average s 20,w values of 1.67 S at 10 M IL-11 ⌬10 , 1.66 S at 20 M, and 1.62 S at 40 M (Fig. 5E and Fig. S8B). Because the theoretical sedimentation coefficient of the IL-11-IL-11R␣ D3 complex is 2.62 S, these data suggest that IL-11R␣ D3 does not bind IL-11 with high affinity.
An apparently unique feature of the IL-11-binding site in IL-11R␣ is a dynamic loop between strands C and D in D2 of the receptor. Our model of the binary complex suggests that this loop may contact bound cytokine and therefore could have a role in binding IL-11, through the formation of polar contacts between the loop and cytokine or by providing additional buried surface area. To investigate this, we generated a construct, IL-11R␣ D1-D3/⌬loop , in which residues 132-140 in the loop were removed and replaced by two glycine residues. SV-AUC showed that IL-11R␣ D1-D3/⌬loop is monomeric in solution and formed a complex with IL-11 ⌬10 with the expected 1:1 stoichiometry ( Fig. 5F and Fig. S8C). ITC showed that the K D of the interaction between IL-11 ⌬10 and IL-11R␣ D1-D3/⌬loop is 8 Ϯ 4 nM (n ϭ 3; Fig. 5A, panel iii), not significantly different from that of IL-11 ⌬10 and IL-11R␣ D1-D3 (p ϭ 0.21). Thus, removal of the loop does not significantly alter the affinity for the interaction between IL-11R␣ and IL-11 ⌬10 , suggesting that the loop does not participate directly in cytokine binding.
It is possible that the dynamic loop functions to partially shield the hydrophobic regions of the cytokine-binding surface in the absence of cytokine, thereby reducing the potential of this region to participate in deleterious, nonspecific interactions. This function would be consistent with our observation that other cytokine receptors that have more hydrophilic character at their cytokine-binding regions, such as IL-6R␣, do not possess this dynamic loop structure.

Conclusion
The increasing identification of roles for IL-11 in a broad range of diseases underscores the need to thoroughly understand the structure of IL-11, its receptors, and the overall molecular mechanism of IL-11 signaling complex formation. Here, we have solved the crystal structure of human IL-11R␣ and a new structure of human IL-11 that reveals detail of functionally important loop regions. We show that several mutations in IL-11R␣ that are associated with disease act to disrupt key structural elements in IL-11R␣, for example through disrupting interdomain interfaces or conserved structural motifs within the receptor. We present a model of the complex and support this model through biophysical and mutagenic analysis. We propose that a dynamic loop proximal to the cytokinebinding region of IL-11R␣ functions to protect this region from nonspecific interactions. Our data elucidate the structural and thermodynamic mechanisms of IL-11 binding by IL-11R␣ and show that this engagement is mediated by both D2 and D3 of the receptor. Together, this work reveals key structural determinants of cytokine engagement by IL-11R␣ on the pathway to formation of the active signaling complex. This molecular detail can be exploited in future development of agents that can modulate this process.

Protein expression and purification
Human IL-11R␣ EC with N-terminal honeybee-melittin signal peptide and C-terminal His 8 tag, was expressed in Sf21 insect cells. Recombinant protein was purified using nickelaffinity chromatography and gel-filtration chromatography. IL-11R␣ D1-D3 and IL-11R␣ D1-D3/⌬loop with N-terminal honeybee-melittin signal peptide, His 8 tag, and TEV cleavage site were expressed in Sf21 cells. Recombinant protein was purified from conditioned media using nickel-affinity chromatography and gel-filtration chromatography. Cleavable tags were removed using TEV protease. IL-11R␣ D3 was refolded and purified from bacterial inclusion bodies as previously described (20). All IL-11R␣ constructs contained the C226S mutation to reduce formation of disulfide cross-linked dimers (20).

Crystallization and X-ray diffraction data collection
IL-11R␣ EC was crystallized using the sitting-drop vapor-diffusion method. Initial crystals were obtained at 293 K in the precipitant 28% PEG 400, 0.2 M calcium chloride, and 0.1 M sodium HEPES, pH 7.5. Crystallization drops were produced by mixing 1.1 mg/ml IL-11R␣ EC in a ratio of 1:0.9:0.1 with the precipitant and the endoproteinase Glu-C. Spherulites appeared after 24 h and were used to prepare a microseed stock (49). Subsequent seeding gave needle clusters in the condition 20% PEG 3350, 0.2 M lithium citrate. Seeding using these needle crystals produced single crystals in the condition 0. IL-11 ⌬10 was crystallized using the sitting-drop vapor-diffusion method. Crystals were obtained at 293 K in the precipitant 30% PEG 3350, 0.2 M ammonium sulfate, 0.1 M Tris, pH 8.5. Crystals appeared after 24 h as thick bundles of two-dimensional plates. These crystals were used for microseeding, providing single crystals in the precipitant 18% PEG 3350, 0.1 M Bis-Tris propane, pH 9, 0.2 M ammonium sulfate, 5 mM praseodymium chloride. Crystallization drops were produced by mixing 1.5 l of IL-11 ⌬10 (5 mg/ml), 1.5 l of precipitant, and 0.5 l of seed. Plates appeared overnight and grew to the approximate dimensions 500 ϫ 20 ϫ 5 m after equilibration against precipitant for 1 week. Crystals were flash-cooled in liquid nitrogen directly from crystallization drops, and X-ray diffraction data were collected at 100 K at the Australian Synchrotron MX2 Beamline (50).

X-ray diffraction data processing and structure refinement
Diffraction data were indexed, integrated and scaled using XDS (51), analyzed using POINTLESS (52) and merged using AIMLESS (53) from the CCP4 suite. Initial phase estimates for IL-11R␣ were obtained by molecular replacement with Phaser (54), using individual domains of IL-11R␣ from unpublished Fab-bound structures as the search models. Refinement was performed using phenix.refine with noncrystallographic symmetry torsion restraints (55), followed iteratively by manual building using Coot (56). Several cycles of simulated annealing were performed early in the refinement to reduce potential model bias. Translation/libration/screw (TLS) refinement was performed in the final rounds, with each domain defined as a separate TLS group. Simulated annealing composite omit maps were calculated using Phenix.
Initial phase estimates for IL-11 ⌬10 were obtained using molecular replacement with Phaser (54), using our previous structure of IL-11 (PDB code 4MHL) (17) as the search model. Auto-building with simulated annealing was performed in phenix.autobuild to reduce phase bias from the search model. Refinement was performed in phenix.refine (55) with iterative manual building using Coot (56). TLS refinement was performed using a single TLS group containing all protein atoms. Explicit riding hydrogens were used throughout refinement and included in the final model; the atomic position and B factors for hydrogens were not refined. Structures were visualized in PyMOL and aligned with the CE (57) algorithm in PyMOL. Buried surface area was determined using the PISA server (23) Residues of both structures are numbered according to the mature protein sequence after cleavage of signal peptides.

Absorbance-detected sedimentation velocity analytical ultracentrifugation
Absorbance-detected SV-AUC experiments were conducted using a Beckman Coulter XL-I analytical ultracentrifuge, equipped with UV-visible scanning optics. Reference and sample solutions were loaded into double-sector 12-mm cells with quartz windows and centrifuged using an An-60 Ti or An-50 Ti rotor at 50,000 rpm (201,600 ϫ g) and 20°C. Radial absorbance data were collected at 280 nm, in continuous mode. All exper-iments were conducted in TBS (20 mM Tris, 150 mM sodium chloride), pH 8 or 8.5. IL-11 ⌬10 and muGFP-IL-11 was centrifuged at concentrations of 0.8, 0.4, and 0.2 mg/ml. IL-11R␣ was centrifuged at concentrations of 0.75, 0.5, and 0.25 mg/ml. Complexes of IL-11 and IL-11R␣ were prepared by mixing 5 M each of IL-11 and IL-11R␣ and centrifuged without further purification. Sedimentation data were fitted to a continuous sedimentation coefficient, c(s), model, and the frictional ratio (f/f 0 ) was fit using SEDFIT (58). Buffer density, viscosity, and the partial specific volume of the protein samples were calculated using SEDNTERP (59). For the complexes between IL-11 and IL-11R␣ and between muGFP-IL-11 and IL-11R␣, the partial specific volume used was 0.73 ml/g. The theoretical sedimentation coefficients of IL-11 ⌬10 and IL-11R␣ were calculated using HYDROPRO, using standard conditions (water, 20°C) (24).

Fluorescence-detected sedimentation velocity analytical ultracentrifugation
Fluorescence-detected SV experiments were conducted using a Beckman XL-A analytical ultracentrifuge, equipped with an Aviv Biomedical fluorescence detection system. Sample solutions were loaded into double-sector 12-mm cells with quartz windows and centrifuged using an An-50 Ti rotor. Experiments were conducted at 50,000 rpm (201,600 ϫ g) and 20°C. muGFP-IL-11 was centrifuged at a concentration of 150 nM (0.007 mg/ml).
To generate the sedimentation coefficient isotherm, the concentration of muGFP-IL-11 was 150 nM, and a 1.5-fold serial dilution series of IL-11R␣ was prepared starting from a concentration of 1 M in TBS, pH 8.0. To prevent nonspecific absorption of muGFP-IL-11 to cell components, 0.2 mg/ml -casein (Sigma-Aldrich) was added to the samples (34). Sedimentation velocity data were processed in SEDFIT as above. c(s) distributions were integrated between 1.0 and 6.0 S. The isotherm was fitted to a 1:1 heteroassociation model in SEDPHAT, with K A and the sedimentation coefficients of muGFP-IL-11 and the complex floated in the analysis (60). The 68% confidence interval was estimated using SEDPHAT.

Structure of human interleukin 11␣ receptor
reduction was performed using the Scatterbrain software, SEC-SAXS analysis using CHROMIXS (64), and the ATSAS suite (64, 65). Theoretical scattering profiles from the crystal structure coordinates were calculated and fit to the experimental scattering data using CRYSOL (66). Ab initio models were calculated using DAMMIF (67) and DAMMIN (68,69). Ten models were calculated using DAMMIF, the models were averaged using DAMAVER, and the averaged model was used as a starting model for DAMMIN. A summary of the SAXS data acquisition and processing is given in Table S1.

Molecular dynamics simulations
All MD simulations were performed using NAMD 2.1.3b1 (70) and the CHARMM22 force field (70,71) at 310 K in a water box with periodic boundary conditions. Simulations were analyzed in VMD 1.9.3 (72). A model of IL-11R␣ was created based on chain B of the crystal structure. The missing loop (residues 132-141) was rebuilt using the PHYRE2 server (73). The missing loop was excluded from all representations of the trajectories and the analysis. The disordered C terminus was not simulated in the model. The structures were solvated (box size, 88.8 ϫ 126.6 ϫ 53.9 Å), and ions were added to an approximate final concentration of 0.15 M NaCl. Simulations of IL-11R␣ was carried out with 10-ps minimization, followed by 50-ns MD. Mutations were introduced to this equilibrated model, and a further simulation was carried out with 10-ps minimization and then 50-ns MD. An additional 50-ns MD was also performed for the unmutated IL-11R␣. The interdomain distance distributions were calculated using a script in VMD, which defined a centroid for each of the three domains and measured the change in distance through the MD simulation. A model of the complete IL-11 ⌬10 structure was created based on the crystal structure. For residues with multiple orientations, only one orientation was selected. The structure was solvated (box size, 53.6 ϫ 53.1 ϫ 74.9 Å), and ions were added to an approximate final concentration of 0.15 M NaCl. A MD simulation was performed using a 10-ps minimization time, followed by 100-ns MD.

Differential scanning fluorimetry
Protein samples were analyzed by differential scanning fluorimetry at a concentration of 0.1 mg/ml in TBS, pH 8.5, with 2.5ϫ SYPRO Orange dye (Sigma-Aldrich). 20 l of the sample was loaded into a 96-well quantitative PCR plate (Applied Biosystems), and four technical replicates of each sample were analyzed. The plates were sealed, and the samples were heated in an Applied Biosystems StepOne Plus quantitative PCR instrument, from 4 to 95°C, with a 1% gradient. The unfolding data were analyzed using a custom script in MATLAB r2016a. The temperature of hydrophobic exposure (T h ) was defined as the minimum point of the first derivative curve and used to compare the thermal stability of different proteins (29).

Isothermal titration calorimetry
Protein samples were buffer exchanged into TBS, pH 8.5, using gel filtration before analysis by ITC. ITC data were collected at 303 K using a MicroCal iTC200 (GE Healthcare). Titrations were performed using 15 injections of 2.5 l of IL-11, after an initial injection of 0.8 l. IL-11R␣ was present at a concentration of 10 M, and the concentration of IL-11 was 10-fold greater than the concentration of IL-11R␣. Titration data were integrated using NITPIC (74,75) and analyzed in SEDPHAT using a 1:1 interaction model (60). Each titration was conducted in triplicate. The values stated are the means Ϯ S.E.

In vitro cell culture
DLD1 cells were grown in RPMI ϩ 10% fetal calf serum, in a 5% CO 2 atmosphere. The cells were grown to confluency in 6-well plates, the medium was removed, and the cells were treated with IL-11 ⌬10 or IL-11 FL at a concentration of 50 ng/ml in RPMI or RPMI as a vehicle control and incubated for 1 h. The cells were then washed with cold PBS and lysed in radioimmune precipitation assay buffer. Protein concentration was determined by the bicinchoninic acid assay. Lysates were diluted with SDS-PAGE loading buffer, resolved on a 10% polyacrylamide gel, and wet-transferred to a nitrocellulose membrane. The membranes were blocked, incubated with the indicated primary antibodies (for phospho-STAT3 CST catalog no. 9145; for phospho-STAT1 CST catalog no. 9167, for total STAT3 CST catalog no. 4904 for total STAT1 CST catalog no. 9172, and for actin Sigma catalog no. A1978), then detected using conjugated fluorescent secondary antibodies (Odyssey catalog no. 926-32211/926-68072), and visualized using the Odyssey IR imaging system (LI-COR Biosciences). Original membranes are shown in Fig. S10.

Docking
In silico docking was performed using the Docking2.0 algorithm, part of the ROSIE server (43,44,76). An initial approximation of the complex orientation was generated by overlaying the IL-11R␣ and IL-11 structures with IL-6R␣ and IL-6 in the IL-6 signaling complex structure (6). This model was used as input to the docking_local_refine protocol of RosettaDock, which limits rotations/transitions of the complex components. The models were scored by RosettaDock. The top-scoring model was taken as the representative model. Buried surface area and interacting residues were determined using the PISA server (23).

CD spectroscopy
CD experiments were conducted using an Aviv CD spectrometer (410-SF). Spectra were collected for 12 M IL-11R␣ D3 at 20°C, in 50 mM sodium phosphate, pH 8.0, over a wavelength range of 260 -190 nm in 1-nm steps with an averaging time of 4 s, using a 1-mm-path length quartz cuvette. Each measurement (sample and blank) was collected in triplicate. The buffer signal was subtracted, and the data were converted to mean residue ellipticity.

Multi-angle light scattering
SEC-MALS data were collected using a Shimadzu LC-20AD HPLC, coupled to a Shimadzu SPD-20A UV detector, Wyatt Dawn MALS detector and Wyatt Optilab refractive index detector. The data were collected following in-line fractionation with a Zenix-C SEC-300 4.6 ϫ 300-mm SEC column Structure of human interleukin 11␣ receptor (Sepax Technologies), pre-equilibrated in 20 mM Tris, 150 mM sodium chloride, pH 8.5, running at a flow rate of 0.35 ml/min. 10 l of sample was applied to the column at a concentration of ϳ2 mg/ml. The IL-11R␣ D1-D3 -IL-11 ⌬10 complex was prepared by mixing equimolar concentrations of the components prior to the experiment. MALS data were analyzed using ASTRA, version 7.3.2.19 (Wyatt). The detector response was normalized using monomeric BSA (Pierce, catalog no. 23209). Protein concentration was determined using differential refractive index, using a dn/dc of 0.184. 15 N-IL-11R␣ D3 was expressed using the method of Marley et al. (77), purified, and refolded as previously described (20), and successful incorporation of 15 N was confirmed using electrospray ionization-TOF MS. 15 N-1 H heteronuclear single quantum coherence spectra were collected on an 18.8 T Bruker Avance II spectrometer ( 1 H resonance frequency 800 MHz), at 283 K, on 130 M 15 N-IL-11R␣ D3 , 20 mM Bis-Tris, 50 mM arginine, 10% 2 H 2 O, pH 7. The spectra were processed using NMRPipe (78) and visualized using NMRFAM-SPARKY (79).

Statistical analysis
Statistical significance was determined using a two-tailed, paired t test in Microsoft Excel, version 16.27 for Mac OSX.

Data availability
Coordinates and structure factors for IL-11R␣ EC and IL-11 ⌬10 have been deposited in the Protein Data Bank with accession codes 6O4P and 6O4O, respectively. SAXS data and models for IL-11 ⌬10 , IL-11 FL , IL-11R␣ D1-D3 , and the IL-11 ⌬10 -IL-11R␣ D1-D3 complex have been deposited in the Small Angle Scattering Biological Data Bank with accession codes SASDGH2, SASDGJ2, SASDGG2, and SASDGK2, respectively. All other data are contained within the manuscript and the supporting information.