The Structure of Interleukin-2 Complexed with Its Alpha Receptor

See allHide authors and affiliations

Science  03 Jun 2005:
Vol. 308, Issue 5727, pp. 1477-1480
DOI: 10.1126/science.1109745


Interleukin-2 (IL-2) is an immunoregulatory cytokine that binds sequentially to the alpha (IL-2Rα), beta (IL-2Rβ), and common gamma chain (γc) receptor subunits. Here we present the 2.8 angstrom crystal structure of a complex between human IL-2 and IL-2Rα, which interact in a docking mode distinct from that of other cytokine receptor complexes. IL-2Rα is composed of strand-swapped “sushi-like” domains, unlike the classical cytokine receptor fold. As a result of this domain swap, IL-2Rα uses a composite surface to dock into a groove on IL-2 that also serves as a binding site for antagonist drugs. With this complex, we now have representative structures for each class of hematopoietic cytokine receptor–docking modules.

Interleukin-2 (IL-2), which is one of the first cytokines identified and a member of the four-helix bundle cytokine superfamily, acts at the heart of the immune response (1). IL-2 and its alpha receptor, IL-2Rα, are expressed by T cells after the activation of T cell receptors by peptide–major histocompatibility complexes. The subsequent autocrine interaction of IL-2 with its receptors leads to the stimulation of signal transduction pathways resulting in T cell, B cell, and natural killer (NK) cell proliferation and clonal expansion (2).

The pleiotropic biological activities of IL-2 are mediated by three cell surface receptors: the IL-2Rα chain; the IL-2Rβ chain; and the common gamma chain (γc), which is also a receptor for IL-4, IL-7, IL-9, IL-15, and IL-21 (3). These cell surface receptors form a complex that signals through the intracellular activation of the Janus tyrosine kinase 3 (Jak3) and the signal transducer and activator of transcription 5 (STAT5) (4). The IL-2Rα chain, originally identified as the Tac antigen (CD25) (57), is enigmatic in that it lacks signature features of the cytokine receptor superfamily (8). IL-2Rβ (p75) and the γc are both members of the hematopoietic growth factor receptor family, containing the signature cytokine-binding homology region (CHR), which is composed of two fibronectin type-III (FN-III) repeats (2, 8).

Biochemical studies show that the assembly of the IL-2 receptor complex is initiated by the interaction of IL-2 with IL-2Rα, followed by sequential recruitment of IL-2Rβ and γc (9, 10). IL-2Rα alone is the “low-affinity” receptor (dissociation constant Kd ∼ 10 nM). When expressed together, IL-2Rα and IL-2Rβ form the pseudo–high-affinity receptor (Kd ∼ 30 pM). Finally, the IL-2Rαβγc complex forms the high-affinity receptor (Kd ∼ 10 pM) that is the signaling complex found on activated T cells (2). The IL-2Rβ and γc binding sites on IL-2 have been mapped to locations analogous to the site I and site II cytokine-binding sites originally established in the human growth hormone (hGH) system (11). However, based on sequence analysis and mutagenesis studies, IL-2Rα is predicted to differ from other cytokine receptors in both its structure and its mode of interaction with IL-2 (12).

IL-2Rα is a target for therapeutic modulation because it is not expressed on resting T and B cells but is continuously expressed by the abnormal T cells of patients with forms of leukemia, autoimmunity, and organ transplant rejection (13, 14). An antagonistic monoclonal antibodytoIL-2Rα (anti-Tac) (Daclizumab) is effective in preventing the rejection of organ transplants (15). Small-molecule inhibitors of IL-2Rα have also been developed (16, 17). IL-2 (Proleukin) itself is used to augment immune system function and has efficacy in treating metastatic renal carcinoma and melanoma (18). However, there is severe dose-limiting toxicity that is largely attributed to activation of the βγc form of the receptor on NK cells (18). Currently, no structural information exists for any of the IL-2 receptors, and this information could assist in the design of improved therapeutics. Here we present a crystal structure at 2.8 Å resolution of human IL-2 in complex with the extracellular domain of IL-2Rα (19).

In the complex structure, the IL-2Rα ecto-domain resembles an arm bent ∼90° at the elbow between the N- and C-terminal domains (D1 and D2, respectively), engaging IL-2 along the length of the underside of D1 (Fig. 1, A and B). The IL-2 binding surface comprises helix A′, helix B′, and part of the AB loop. The long axes of the IL-2Rα β sheets are aligned parallel with the helical axes of the cytokine. This differs from the typical cytokine receptor interaction. For example, in the complex of hGH with its receptor (hGH-R), the CHR module of hGH-R, composed of two β-sandwich FN-III domains, forms a considerably larger, but similar, bent arm–like structure (Fig. 1B) (20). However, the protruding elbow region of hGH-R exposes loops that bind to the sides of the hGH four-helix bundle (20) (Fig. 1B). Although substantially different, the closest similarity can be found in the recently elucidated site III contact between gp130-class cytokines and the gp130 D1 domain (21). However, in that interaction, the cytokine forms contacts exclusively through loops at the tip of the cytokine, rather than with residues on the helical surfaces.

Fig. 1.

Structure of the human IL-2/IL-2Rα complex. (A) Side view of the complex showing IL-2Rα covering the groove between the AB loop and helix B′. The IL-2Rα D1 domain is green, the D2 domain is cyan, and IL-2 is violet. Disulfide bonds of IL-2Rα are pink. Hydrophobic core residues between D1 and D2 are drawn as yellow and blue sticks, respectively. Disordered connecting regions, which are not part of the experimental model, are shown as black dotted lines. This coloring scheme is maintained in Figs. 1, 2, 3, 4. (B) The binding mode between IL-2 and IL-2Rα versus the classical site I/II paradigm of hGH and its receptor, in which the elbow of the receptor binds to the sides of the cytokine four-helix bundle (20). A semitransparent surface representation with receptor chains in blue and ligands in violet is shown. Black dotted vees indicate the elbow regions of IL-2Rα and hGH-R.

In the IL-2Rα structure, the core D1 and D2 domains are separated by a 42-residue inter-domain linker peptide (Thr65 to His103), and the second domain has an additional C-terminal 54-residue connecting peptide leading to the cell membrane (Gly165 to Glu217). Neither of these linkers is visible in the electron density map, and they are therefore not included in the IL-2Rα model (Fig. 1A). These linker segments do not contribute to ligand binding. Thus, 120 residues (D1 and D2 domains) of the 217–amino acid extracellular domain are structured. It is possible that this extensive amount of flexible polypeptide contributed to the difficulties in obtaining crystals suitable for a structure determination, as crystallization of the complex was originally reported in 1989 (22).

IL-2Rα deviates from the classical cytokine receptor fold, both in the structures of the individual domains and in the interdomain folding topology (Fig. 1A). Each of the two IL-2Rα domains exhibits ∼30% amino acid homology to a group of β-sandwich protein domains variously called sushi domains, short consensus repeats (SCRs), or complement control protein (CCP) repeats (23) (Fig. 2B). Canonical sushi domains (∼65 amino acids in a 2-on-3 sandwich) can be considered “mini” FN-III–type domains (∼110 residues in a 3-on-4 sandwich). Like sushi domains, IL-2Rα D1 (residues 1 to 64) and D2 (residues 104 to 165) show an elliptical β-sandwich structure, containing several residues that are highly conserved in sushi domains of several complementary related proteins (Cys3/30/46/61, Pro7, Tyr20, Gly33, Phe34, and Trp55) (24). However, both IL-2Rα domains deviate substantially from canonical sushi topology; they are 1-on-4, instead of 2-on-3, β sandwiches. The folding topology of IL-2Rα also exhibits swapping of β strands across the N- and C-terminal sushi-like domains (Fig. 2A). Strands A and B exchange with strands F and G, respectively, to give the folding topology F-on-G-C-D-E for D1 and A-on-B-H-I-J for D2. Interdomain disulfide bonds between Cys3 and Cys46 of D1 and Cys147 and Cys104 of D2 enforce the strand exchange by pinning strand A of D1 to strand I of D2 and strand G of D2 to strand D in D1 (Figs. 1A and 2A). This strand exchange has not been seen among β–sandwich–type structures (Fig. 2B), or in classical FN-III–type cytokine receptors. The IL-15 α receptor is a single sushi domain, so there is not the possibility for a strand exchange in that case.

Fig. 2.

Domain swapping in IL-2Rα. (A) The IL-2Rα model is separated into D1 and D2 domains revealing the swapped A and B, and F and G strand locations, respectively. Semitransparent ovals in two diagrams of the separated sushi domains highlight the corresponding strands, which are involved in the strand swap. A surface representation of IL-2Rα in the center shows the complete receptor molecule. (B) A representative structure of a sushi domain, or CCP module [second domain of β2 glycoprotein-1 (Protein Data Bank entry 1QUB)] (23, 32). The analogous strands to those involved in the strand swap between IL-2Rα D1 and D2 domain are blue and highlighted by a semitransparent oval.

Despite the unstructured 42–amino acid linker peptide in IL-2Rα, the D1 and D2 domains are in intimate contact with one another through an extensive interdomain interface. A large hydrophobic core is formed between the swapped top strands in each domain, which likely imparts rigidity to the overall structure (Fig. 1A). These exposed aromatics on the surface of the top strands of the β sandwich would serve as the inner hydrophobic core between the top and bottom sheets of a canonical sushi domain, if the strands in IL-2Rα were not domain-swapped.

The IL-2 molecule has the common cytokine fold (8), with the typical up-up-down-down four-helix topology. The four-helix bundle structure of the IL-2 molecule in the IL-2/IL-2Rα complex closely superimposes with the free IL-2 molecule [0.74 Å root mean square deviation (RMSD) on 76 Cα positions], but with minor structural adaptations in the receptor binding site (25, 26). The IL-2 CD loop, which is usually disordered in uncomplexed IL-2 structures, is completely ordered in the receptor-bound IL-2 structure because residues Glu106 to Asp109 are involved in binding to IL-2Rα. The first and last turn of helix A′ as well as the last turn of helix B′ are partially unwound as compared with free IL-2 (fig. S1).

The interaction between IL-2 and IL-2Rα occurs between the four-stranded β sheet, strands G-C-D-E, and the IL-2 A′ and B′ helices (Fig. 3A). As a result of complex formation, 20 IL-2 residues and 21 IL-2Rα residues bury a total of 1868 Å (fig. S2 and table S2). Two prominent hydrophobic patches on IL-2 are consistent with thermodynamic measurements indicating that the desolvation of apolar surface is the primary energetic driving force of this interaction (red amino acids in Fig. 3B and red patches in Fig. 4A). The first patch is composed of Tyr45 on the AB loop of IL-2, which packs into a pocket formed by the methylene groups of Arg35 and Arg36 of the IL-2Rα C′ strand. This hydrophobic cluster is surrounded by polar interactions. Mutations of IL-2Rα Arg35 and Arg36 disrupt interaction with IL-2 (27).

Fig. 3.

Molecular anatomy of the IL-2/IL-2Rα interface. (A) Amino acid contact residues within the complex interface. IL-2 (violet) is at the top and the IL-2Rα D1 domain (green) and D2 domain (cyan) are at the bottom. Hydrogen bonds appear as red dotted lines; disulfide bonds are pink; corresponding cysteine residues are numbered; and β sheets and connecting loops of IL-2Rα are labeled as indicated. (B) “Footprint” surface representation of the IL-2Rα interface viewed through the IL-2 helices onto the IL-2Rα. Contact residues of IL-2 (violet sticks) are projected onto the buried surface (orange) of IL-2Rα. The hydrophobic anchor residues Phe42 and Tyr45 of IL-2 are colored red to orient the reader throughout the interface.

Fig. 4.

Comparison of receptor versus drug binding to IL-2. (A) “Footprint” representation of IL-2 interface as viewed through the IL-2Rα β strands onto the IL-2 surface. Contact residues of IL-2Rα (green and cyan sticks) are projected onto the buried surface (orange) of IL-2. The hydrophobic anchor residues Phe42 and Tyr45 of IL-2 are red. (B) Analogous footprint view of the drug compound 1 bound to IL-2. Compound 1 is depicted with blue sticks [Protein Data Bank entry name 1M48 (17)] projected onto the buried surface (orange) of IL-2. Compound 1 uses IL-2 Phe42 (red patch) as an anchor residue, thereby preventing IL-2Rα from binding to IL-2.

The most conspicuous and centrally located hydrophobic cluster is composed of Phe42 and Leu72 on the IL-2 AB loop and helix B′, respectively, inserting into a recessed pocket on IL-2Rα composed of Leu42 to Tyr43, and Met25 (fig. S4 and Figs. 3B and 4A). Several salt bridges and hydrogen bonds surround this hydrophobic patch (Fig. 3A and table S2). Mutational studies identify this second hydrophobic region around IL-2 Phe42 as the primary energetic determinant in the IL-2–binding epitope (27, 28). T cells can express IL-2 splice variants, which are natural inhibitors of IL-2 signaling through the high-affinity αβγc receptor (29). The splice variants lack either exon 2 (IL-2δ2), which encodes Asn30 to Lys49, or exon 3 (IL-2δ3), which encodes Ala50 to Lys97. Because these regions are involved in the interaction between IL-2Rα and full-length IL-2, the splice variants would almost certainly not bind IL-2Rα. Thus, the mechanism of inhibition appears to be the occupation of the IL-2Rβ and γc receptors, preventing recruitment of IL-2Rα by wild-type IL-2 and inhibiting signaling on activated T cells through the αβγc complex.

The disruption of protein-protein interactions with small molecules has proven to be an exceedingly difficult goal for the pharmaceutical industry in comparison to the inhibition of enzyme active sites. However, potent small-molecule inhibitors of the IL-2/IL-2Rα interaction have been developed (17). One of these small molecules, compound 1 (16), fits into a groove between the IL-2 AB loop and helix B′, wrapping around IL-2 Phe42 as a hydrophobic anchor residue (Fig. 4B) (17). In binding to IL-2, the small molecule induces the rearrangement of residues Lys35, Arg38, Met39, and Phe42, compared with several unliganded IL-2 structures (17). In particular, a large change in conformation of the IL-2 Phe42 aromatic ring creates the recessed channel used for binding the drug (Fig. 4B). This conformational adjustment does not occur upon receptor binding; in contrast, Phe42 is a prominent knob fitting into a pocket on IL-2Rα. Superposition of the drug/IL-2 complex on the IL-2/IL-2Rα complex reveals that although a substantial portion of the drug fits within open space in the interface, there are steric clashes with the IL-2 Phe42 binding pocket. The drug apparently uses highly favorable enthalpic interactions to compensate for the entropic penalty of binding (17) and to overcome its small buried surface area. In contrast, binding of IL-2Rα to IL-2 is entropically favorable. Hence, even though the drug and receptor target a similar hydrophobic hot spot, they use opposite thermodynamic solutions for binding.

Previously determined binding modes between hematopoietic cytokines and their receptors involve combinations of the classical site I/II and the recently determined site III mode for gp130 cytokines (21). We suggest that the distinctive docking geometry of the IL-2/IL-2Rα interaction is a fourth binding mode and now gives us representative examples of all binary recognition modules used by hematopoietic receptors to recognize four-helix cytokines. These four modules can be used as building blocks, in different combinations, to construct topological models for complexes between all known cytokines and their receptors. As discussed, IL-2 also binds to two additional receptor subunits, the IL-2Rβ and γc chains, to form a quaternary signaling complex. Functional studies and molecular modeling placed the binding epitopes of IL-2Rβ and γc on the faces of adjacent helices A (IL-2 Asp20) and D (IL-2 Gln126), respectively (11, 14, 30) (figs. S1 and S5). It has been postulated that IL-2Rβ will bind in site I–type and γc will bind in site II–type geometries (Fig. 1B, hGH complex). The binding site of IL-2Rα is ideal for an initial engagement of cytokine, leaving sides of the helical faces open for engagement of the IL-2Rβ and γc components. By initially capturing IL-2 on the cell surface, the IL-2Rα would reduce the entropic cost for the subsequent recruitment of additional receptors (31).

Supporting Online Material

Materials and Methods

Figs. S1 to S5

Tables S1 and S2


References and Notes

View Abstract

Navigate This Article