Research Article

Structure of the Quaternary Complex of Interleukin-2 with Its α, ß, and γc Receptors

See allHide authors and affiliations

Science  18 Nov 2005:
Vol. 310, Issue 5751, pp. 1159-1163
DOI: 10.1126/science.1117893

Abstract

Interleukin-2 (IL-2) is an immunoregulatory cytokine that acts through a quaternary receptor signaling complex containing alpha (IL-2Rα), beta (IL-2Rβ), and common gamma chain (gc) receptors. In the structure of the quaternary ectodomain complex as visualized at a resolution of 2.3 angstroms, the binding of IL-2Rα to IL-2 stabilizes a secondary binding site for presentation to IL-2Rβ. γc is then recruited to the composite surface formed by the IL-2/IL-2Rβ complex. Consistent with its role as a shared receptor for IL-4, IL-7, IL-9, IL-15, and IL-21, γc forms degenerate contacts with IL-2. The structure of γc provides a rationale for loss-of-function mutations found in patients with X-linked severe combined immunodeficiency diseases (X-SCID). This complex structure provides a framework for other γc-dependent cytokine-receptor interactions and for the engineering of improved IL-2 therapeutics.

The cytokine IL-2 is mainly produced by antigen-activated T cells and promotes the proliferation, differentiation, and survival of mature T and B cells as well as the cytolytic activity of natural killer (NK) cells in the innate immune defense (1, 2). IL-2 is used therapeutically as an immune adjuvant in certain types of lymphoproliferative diseases and cancers, and IL-2 antagonists can prevent organ transplant rejection (3, 4). However, severe dose-limiting toxicity has limited its effectiveness in the clinic. These deleterious side effects are mediated through different combinations of IL-2 receptors, which suggests that structure-based engineering of receptor-selective variants could have clinical benefit (5).

IL-2 exerts its pleiotropic activities through binding to different receptor complexes, depending on which of the components are expressed on the cell surface: the alpha chain (IL-2Rα), beta chain (IL-2Rβ), and common cytokine receptor gamma chain (γc) (610). Isolated IL-2Rα has been termed the “low-affinity” IL-2 receptor (binding affinity Kd ≈ 10 nM) and is not involved in signal transduction (11). A complex of IL-2Rβ and γc binds with intermediate affinity (Kd ≈ 1 nM) and is the receptor form on NK cells, macrophages, and resting T cells (2), although IL-2Rβ alone has very low affinity (Kd ≈ 100 nM) and γc alone has no detectable binding affinity for IL-2 (12). The association of IL-2Rβ and γc in the presence of IL-2 is necessary and sufficient for effective signal transduction through the heterodimerization of their cytoplasmic domains and subsequent kinase activation of multiple signaling pathways (13, 14). A complex with three subunits—IL-2Rα, IL-2β, and γc—binds with high affinity (Kd ≈ 10 pM) and is the receptor form on activated T cells (10). The high-affinity receptor complex mediates most biological effects of IL-2 in vivo (2).

Whereas IL-2Rα is a specific receptor for IL-2, IL-2Rβ is also a component of the IL-15 receptor and γc is shared by cytokines IL-4, IL-7, IL-9, IL-15, and IL-21 (15). Mutations in γc can abolish the activity of all γc-dependent cytokines and result in X-linked severe combined immunodeficiency diseases (X-SCID), in which the T and NK cells are absent or profoundly reduced in number (16). Because the six γc-dependent cytokines have low sequence homology, structural information will be helpful to delineate shared versus ligand-specific binding determinants that could be exploited therapeutically. Previously, we reported the structure of the binary complex of IL-2 with IL-2Rα (17). We now present the crystal structure, at 2.3 Å resolution, of the quaternary complex of IL-2 with the extracellular domains of receptors IL-2Rα, IL-2Rβ, and γc.

Overall structure. Because of the heterogeneity of the fully glycosylated proteins expressed from insect cells, we crystallized a glycan-minimized quaternary complex, which had five potential Asn-linked glycosylation sites mutated (18). This material behaved identically to the fully glycosylated proteins and yielded crystals that diffracted to 2.3 Å resolution (18).

The quaternary complex is composed of one copy each of IL-2, IL-2Rα, IL-2Rβ, and γc (Fig. 1A). The orientation of IL-2Rα explains the necessity for the long connecting peptide, disordered in the structure, between the IL-2Rα globular head and the transmembrane segment. This allows the IL-2Rα binding domain to extend away from the cell surface and reach the dorsally located binding site on IL-2 (Fig. 1B). The bases of the receptors IL-2Rβ and γc, both class I–type cytokine receptors, converge to form a Y shape and IL-2 binds in the fork (Fig. 1, A and B). Formation of the quaternary complex is mediated by four binding sites—IL-2/IL-2Rα, IL-2/IL-2Rβ, IL-2/γc, and IL-2Rβ/γc—burying a total of 5700 Å2 of surface area (fig. S1). The IL-2/IL-2Rα and IL-2/IL-2Rβ contacts are independent, whereas IL-2 and IL-2Rβ form a composite interface with γc, reflecting the cooperative nature of complex assembly.

Fig. 1.

Structure of the human IL-2/Rαβγ quaternary signaling complex. (A and B) Ribbon diagram of the complex structure shown in two views related by a ∼90° rotation about the vertical axis. IL-2 is shown in violet, and the receptors are shown in cyan (IL-2Rα), blue (IL-2Rβ), and gold (γc). The observed N-linked carbohydrates at Asn123 of IL-2Rβ and at Asn49, Asn62, and Asn137 of γc are shown in gray (20). Disulfide bonds are shown in red. The disordered peptides connecting the C terminus of the receptors to the cell membrane are shown as dotted lines in their respective colors. The program PyMol (43) was used to make all figures.

IL-2Rα has been shown to deviate from typical cytokine receptor structure and mode of interaction with IL-2 (17). It is composed of two domain-swapped “sushi” modules, essentially miniature β-sheet sandwich domains. IL-2Rβ and γc are prototypical members of the class I cytokine receptor superfamily (19). Both are composed of N- and C-terminal fibronectin-III domains (D1 and D2, respectively), which are characterized by a β-sandwich sheet consisting of seven antiparallel strands arranged in a three-on-four topology. In IL-2Rβ, the D1 and D2 domains are connected by a helical linker and are bent at ∼90°, whereas in γc the D1 and D2 domains are bent at ∼120° (Fig. 1A). Both IL-2Rβ and γc contain the two disulfide bonds in the N-terminal domain (D1) and a “WSXWS” motif (20) in the C-terminal domain (D2) that are characteristic of class I cytokine receptors (19). However, a third disulfide bond in the γc D2 domain is unusual because of its central position in the interface with IL-2 and its role in enabling degenerate cytokine recognition (Cys160 to Cys209) (discussed below).

IL-2/IL-2Ra. In the “low-affinity” complex, the atomic interactions between IL-2 and IL-2Rα, now visualized at 2.3 Å, are unchanged from the binary complex at 2.8 Å (17). The binding interface between IL-2 and IL-2Rα in the quaternary complex is composed of helices A′ and B′ and part of the AB loop in IL-2 and strands G, C, and D in the D1 domain and strand A in the D2 domain in IL-2Rα (table S2A). The two prominent hydrophobic ridges around residues Phe42 and Tyr45 of IL-2 insert into grooves between the IL-2Rα beta strands. Superposition of the two IL-2Rα structures in the binary and quaternary complexes shows a significant shift in the D2 domain of IL-2Rα (∼2 Å), which is most likely a result of crystal packing and reflects some flexibility in the D1-D2 junction.

IL-15 is the only other cytokine that uses an atypical sushi-domain alpha receptor (IL-15Rα), which is expressed primarily on NK cells (21). However, IL-15Rα is only a single sushi domain analogous to the IL-2Rα D1 domain (22, 23). By analogy with IL-2, the IL-15/IL-15Rα complex likely forms first, followed by binding to IL-2Rβ and γc to form the quaternary signaling complex.

IL-2Rα does not appear to make any contact with either IL-2Rβ or γc. This is rather surprising, given that the IL-2/IL-2Rα complex binds with much higher affinity to IL-2Rβ (Kd ≈ 30 pM) than does IL-2 binding to IL-2Rβ alone (Kd ≈ 100 nM) (12, 24) and the on-rate of IL-2 for IL-2Rβ is faster in the presence of IL-2Rα by a factor of 3 to 20 (11, 25). We see no evidence of a composite receptor binding surface for IL-2, so what is the basis of the cooperativity? One possibility would be simple entropy reduction, wherein IL-2Rα captures and concentrates free IL-2 at the cell surface for presentation to IL-2Rβ and γc. Another possibility would be an IL-2Rα–induced conformational change in IL-2 that stabilizes the formation of the ternary complex.

To address the latter mechanism, we compared IL-2 structures in the quaternary complex, binary complex, and unbound states. The root mean square deviations for Cα atoms in the helical core between the different IL-2 molecules indicate nearly identical structures, ranging from 0.29 Å to 0.57 Å. One notable exception is at the beginning of helix C of IL-2, where, in the binary and quaternary complexes, several turns of helix C are slightly unwound and translated forward by ∼1.0 Å toward IL-2Rβ (Fig. 2A). This local conformational adjustment moves IL-2 residue Asn88 into hydrogen-bonding distance to IL-2Rβ residue Arg42 (Fig. 2B). This movement possibly “primes” the next step in complex assembly by forming a more complementary IL-2Rβ binding site. Consistent with this, mutation of Asn88 in IL-2 ablates binding to IL-2Rβ (5). IL-2Rα may stabilize a favorable IL-2Rβ–binding conformation of IL-2 helix C, reducing a conformational entropy penalty that would be incurred during binding to IL-2Rβ. This priming of a “quiescent” IL-2Rβ binding site in IL-2 by IL-2Rα could effectively increase the on-rate for the IL-2 interaction with IL-2Rβ, as has been observed.

Fig. 2.

IL-2Rα binding results in local conformational changes within IL-2 helix C. (A) Backbone superposition of IL-2 structures in quaternary complex (violet), binary complex (orange) (PDB 1Z92) (17), and three unbounded states: PDB 1M4C (green), 1M47 (dark green), and 3INK (gray) (44, 45). (B) IL-2 residue Asn88 in helix C forms a hydrogen bond with Arg42 from IL-2Rβ in quaternary and binary complexes as a result of closer proximity induced by IL-2Rα binding.

IL-2/IL-2Rb. The interface between IL-2 and IL-2Rβ buries ∼1350 Å2 formed by residues from helices A and C in IL-2 and residues from loops CC′1, EF1, BC2, and FG2 in IL-2Rβ (table S2B). The interface is highly polar, with eight hydrogen bonds directly between IL-2 and IL-2Rβ residues. Strikingly, there are seven water molecules buried in the interface that bridge interactions between IL-2 and IL-2Rβ by forming bonds with protein atoms (Fig. 3A) (table S2B). Solvent exchange with the layer of water molecules between IL-2Rβ and IL-2 could explain the fast on- and off-rates and the weak affinity of the IL-2/IL-2Rβ binary complex. Two residues of IL-2 that have been shown by mutagenesis to be critical for IL-2Rβ binding, Asp20 and Asn88, are involved in hydrogen bonding networks to both water molecules and side chains on IL-2Rβ (Fig. 3A). The side chains of IL-2Rβ residues His133 and Tyr134 insert into a complementary cavity in IL-2 to form hydrogen and ionic bonds with Asp20 of IL-2 (Fig. 3B).

Fig. 3.

A polar interface and hydration layer between IL-2 and IL-2Rβ. (A) All interactions between IL-2 and IL-2Rβ. The buried water molecules in the interface are shown as green spheres. The hydrogen bonds between IL-2 and IL-2Rβ are in black; those between water molecules and protein atoms are in green. (B) Close-up view of the shape complementarity in the interface, as viewed from above.

IL-2Rβ is also used by IL-15 to form a quaternary complex along with IL-15Rα and γc (15). IL-15 has limited sequence identity (19%) with IL-2, so its contact with IL-2Rβ is probably through a unique set of interactions. The bridging water molecules may contribute to the ability of IL-2Rβ to cross-react by accommodating the different IL-15 residues through remodeling of the intervening solvent layer.

IL-2/gc. Neither IL-2 nor IL-2Rβ alone have measurable affinities for γc (12). Therefore, two very weak interactions, IL-2/γc and IL-2Rβ/γc, combine to produce an intermediate-affinity IL-2/IL-2Rβ/γc complex. In the quaternary complex structure, the interaction surface of the IL-2/Rαβ complex with γc is composed of two interfaces: a small one between IL-2 and γc, and a larger one between IL-2Rβ and γc.

The IL-2/γc interface buries ∼970 Å2 of surface area and is the smallest of the four protein-protein interfaces in the complex. The γc binding surface is striking in its absence of extended side chain–specific interactions with IL-2 and in the preponderance of main-chain contacts. Although there are several apparent “hotspots,” the γc binding surface is remarkable in its flatness and almost tangential contact with IL-2 (Fig. 4A). The γc structure contains an unusual disulfide bond in the heart of the interface with IL-2 that connects loops FG2 with BC2 and supports the conformations of Ser207 to Pro211 that form direct contacts with IL-2 (Fig. 4A). The disulfide also contributes to the apparent rigidity of the cytokine-binding surface, which is surprising given that one prevailing assumption for receptor cross-reactivity is structural plasticity (26). The γc binding surface does not appear to contain mobile structural elements, although we do not have a structure of the unliganded receptor for comparison.

Fig. 4.

Interactions between γc and IL-2. (A) Contacting residues in the IL-2/γc interface. (B) Surface representation of the relative contact patch around Tyr103 from γc, as viewed from above.

The overall interface involves residues from helices A and D in IL-2 and residues from loops CC′1, EF1, BC2, FG2, and the linker between strands G1 and A2 in γc (table S2C). In contrast to the broad array of specific polar interactions between IL-2 and IL-2Rβ, small contact patches dominate the IL-2/γc interface. The first one is composed of residue Tyr103 from γc and residues Ser127 and Ser130 from IL-2. The Tyr103 aromatic ring packs flat against the side chains of Ser127 and Ser130 in IL-2 (Fig. 4, A and B). The second is around residue Gln126 in IL-2, which has been shown by mutagenesis to be a critical energetic hotspot. Similar to Tyr103, the side chain of Gln126 is almost parallel to the surface formed by main-chain atoms of residues Pro207 to Ser211 in γc, and this orientation is further fixed by two hydrogen bonds with the receptor, to Pro207 O and Ser211 OG (Fig. 4A). A single bridged water molecule in the IL-2/γc interface forms hydrogen bonds with Gln126 of IL-2 and with Gln127 and Asn128 of γc, respectively (table S2C).

Previous mutagenesis studies have found that two of the flat γc patches we see in the structure that are involved in binding IL-2—residue Tyr103 and residues from Leu208 to Ser211 in γc—are also important for binding IL-4, IL-7, IL-15, and IL-21 (2729). There is also evidence that γc binding sites for different cytokines overlap but are not identical (30). We propose that these two patches form the central degenerate recognition surfaces that participate in binding all cytokines in the γc-dependent family by using their flat surfaces, and that the peripheral polar interactions modulate specificity for individual cytokines.

IL-2Rb/gc. The second part of the composite interface between IL-2/Rβ and γc is formed by extensive interactions between the D2 domains of IL-2Rβ and γc, burying more than 1750 Å2 of surface area (Fig. 5A). The D2 domains from IL-2Rβ and γc are related by almost exact two-fold symmetry, and the interface is formed by 21 residues from IL-2Rβ and 19 residues from γc from strands C2, C′2, and E2 and loop C′E2 (table S2D). The interface is highly polar, with a peripheral ring of 17 hydrogen bonds surrounding a hydrophobic stripe in the center dominated by Trp166 from IL-2Rβ and Tyr167 from γc (Fig. 5B) (table S2D).

Fig. 5.

Extensive receptor-receptor contact between IL-2Rβ and γc. (A) Surface representation of the quaternary complex shows the shape complementarity in the IL-2Rβ/γc interface. (B) Hydrogen-bonding interactions between IL-2Rβ and γc.

The D2-D2 interaction between IL-2Rβ and γc is the largest buried surface seen so far in cytokine-receptor complexes, and it underscores the role of receptor-receptor contact in the cooperative assembly of the quaternary complex. Although it is surprising that IL-2Rβ and γc have no measurable affinity toward one another given this extensive contact surface, a lack of interaction would prevent the receptors from heterodimerizing and signaling in the absence of cytokine. Given the structural observations of a small IL-2/γc interface and a large and tightly packed IL-2Rβ/γc interface, we suggest that the receptor-receptor (i.e., D2-D2) contact may serve as an important energetic determinant. If so, the role of the cytokine would be to stabilize complex formation by guiding a perfect geometrical alignment of the numerous interatomic contacts (hydrogen bonds, van der Waals contacts, etc.) in the D2-D2 interface. In this respect, considering the relatively flat and chemically inert IL-2/γc interface, it may be that specificity is largely provided by receptor-receptor contact with IL-2Rβ rather than cytokine.

This model can in part be rationalized by the “shared” function of γc. γc is expressed on most immune cell types, but the tissue and cytokine specificity are regulated by the coordinate expression of different α receptors (or, in the case of IL-2, the β receptor). Given the lack of sequence homology between γc-dependent cytokines, the capacity of γc to discriminate among (and to cross-react with) these ligands would be more easily achieved by spreading the energetics of the interaction over the combined ∼2600Å2 of surface area presented by the IL-2/IL-2Rβ composite surface, rather than focusing it all on the small portion of this surface contributed by cytokine alone (∼970 Å2). By comparison, in the structures of more ligand-specific cytokine receptors such as human growth hormone receptor (hGHbp) and erythropoietin receptor (epoR) complexed with their ligands, there is much less receptor-receptor contact (∼900 Å2 for hGHbp, no contact for epoR) (31, 32).

X-SCID mutations. X-linked severe combined immunodeficiency (SCID) is a syndrome of profoundly impaired cellular and humoral immunity caused by mutations in the gene encoding the common gamma chain (33). The mutated gene results in faulty signaling through several cytokine receptors; thus, T, B, and NK cells can be affected by a single mutation. We mapped extracellular γc mutations that have been found in X-SCID patients in which γc is expressed but is not competent for activation by any of the γc cytokines. Many mutations appear to concentrate near the γc cytokine-binding site, and several of these—Y103N, Y103C, L208P, C209R, C209Y, G210R, G210V, and C160R (20, 33)—map to the γc binding interface with IL-2 (Fig. 6A). Mutation of Cys209 or Cys160, which participate in the disulfide bond in the γc cytokine-binding surface, would be particularly destabilizing. The interface mutations would effectively ablate cytokine recognition by γc, but it seems likely that the receptor would still appear to be competent to signal if heterodimerized. Although the database X-SCID mutations map to all other parts of the γc structure, none of the X-SCID mutations map to the IL-2Rβ/γc interface, possibly implying a structural necessity for this area to be preserved in the expressed receptor (Fig. 6B).

Fig. 6.

Mapping known X-SCID mutations in the structure of γc. (A) Five missense mutations that are located in the γc cytokine-binding epitope, and make contact with IL-2 in the structure, are shown as red sticks. (B) Distribution of all missense mutations in the X-SCID mutation database (http://genome.nhgri.nih.gov/scid) in γc. The γc area participating in the D2-D2 interaction with IL-2Rβ is free of mutations and is indicated within the dashed line.

Degenerate cytokine recognition by gc. The γc-dependent cytokines have, on average, 19% sequence identity to one another, with most of the homology concentrated inside the helical cores. Although we currently know the structures of only IL-2 and IL-4 in the γc-dependent cytokine family, we sought to identify conserved residues that might serve as a recognition code for γc binding throughout the family. Sequence alignment between IL-2 and other γc-dependent cytokines (fig. S3) indicates that residue Gln126, which plays a key structural role in IL-2 interactions with γc, is conserved in IL-2, IL-9, IL-15, and IL-21, whereas IL-4 and IL-7 have Arg121 and Lys139 in this position, respectively. Superposition of the IL-4/IL-4Rα complex (34) with the IL-2 quaternary complex indicates that Arg121 may play a structural role similar to that of Gln126 in IL-2 in contacting γc. Although position 126 in helix D may serve as a common contact point with γc, there are not obvious constellations of conserved residues that allow us to dock the different cytokines with γc. It appears that each cytokine uses distinct structural solutions for γc recognition.

Cytokine recognition by shared receptors. The flat and apparently rigid surface in the common binding epitope of γc suggests that it uses somewhat chemically inert complementary surfaces to interact with divergent cytokine residues. Although this contrasts with notions of receptor promiscuity through binding site flexibility (26), it parallels structural results for gp130, the shared cytokine receptor for long-chain cytokines (35), in complex with three different cytokines: LIF, viral IL-6, and human IL-6 (3638). In the gp130 system, thermodynamic compensation between rigid surfaces, rather than conformational change, enables cross-reactivity with a broad range of chemically diverse cytokine surfaces (35). We predict, on the basis of direct thermodynamic measurements of the quaternary complex assembly (12), that γc also uses such a mechanism for cross-reactivity. Such a large range of energetic compensation appears to be a property of binding sites found in shared receptors, which are tuned for degenerate recognition through a mechanism that bypasses the entropic penalty for conformational change (39).

Therapeutic implications. A recombinant human IL-2 (rIL-2) analog (Aldesleukin, Proleukin, Chiron Inc., Emeryville, CA) is currently licensed in the United States for the treatment of metastatic melanoma and renal cell carcinoma and is undergoing clinical trials for patients with HIV/AIDS (40, 41). Treatment of cancer patients with rIL-2 results in robust responses but is associated with life-threatening toxicity, which limits its use (40). The antitumor efficacy of rIL-2 therapy has been shown to be mediated by the high-affinity quaternary complex containing IL-2Rαβγ expressed on T cells, whereas the toxic side effects are mediated through the IL-2Rβγ form of the receptor on NK cells (42). This hypothesis suggests that it might be possible to dissociate efficacy and toxicity by generating an IL-2 variant with selectivity for the IL-2Rαβγ receptor complex, versus the IL-2Rβγ complex of NK cells (5). Proof-of-concept was demonstrated with an IL-2 variant bearing an Asn88 → Arg mutation that conferred a factor of 3000 selectivity increase for the IL-2Rαβγ complex by crippling the interaction between IL-2 and IL-2Rβ (5).

In the structure we see that Asn88 is the side chain brought into hydrogen-bonding distance to IL-2Rβ by the structural perturbation of helix C in response to IL-2Rα binding, and is involved in an extensive hydrogen-bonding network (Fig. 3A). Such an energetically critical residue may not be the most desirable choice for generating a receptor-selective IL-2, because it may not be necessary to completely ablate binding to the IL-2Rβγ receptors. Rather, weakening the IL-2Rβ interaction, or even contact with γc, while maintaining near wild-type affinity for the IL-2Rαβγ complex appears tenable through structure-guided engineering. It is our hope that this quaternary complex structure can be used to design IL-2 variants that will allow its powerful clinical potential to be more fully realized.

Supporting Online Material

www.sciencemag.org/cgi/content/full/310/5751/1159/DC1

Materials and Methods

Figs. S1 to S3

Tables S1 and S2

References

References and Notes

View Abstract

Navigate This Article