Crystal Structure of the Interleukin-15·Interleukin-15 Receptor α Complex

Interleukin (IL)-15 is a pleiotropic cytokine that plays a pivotal role in both innate and adaptive immunity. IL-15 is unique among cytokines due to its participation in a trans signaling mechanism in which IL-15 receptorα (IL-15Rα) from one subset of cells presents IL-15 to neighboring IL-2Rβ/γc-expressing cells. Here we present the crystal structure of IL-15 in complex with the sushi domain of IL-15Rα. The structure reveals that theα receptor-binding epitope of IL-15 adopts a unique conformation, which, together with amino acid substitutions, permits specific interactions with IL-15Rα that account for the exceptionally high affinity of the IL-15·IL-15Rα complex. Interestingly, analysis of the topology of IL-15 and IL-15Rα at the IL-15·IL-15Rα interface suggests that IL-15 should be capable of participating in a cis signaling mechanism similar to that of the related cytokine IL-2. Indeed, we present biochemical data demonstrating that IL-15 is capable of efficiently signaling in cis through IL-15Rα and IL-2Rβ/γc expressed on the surface of a single cell. Based on our data we propose that cis presentation of IL-15 may be important in certain biological contexts and that flexibility of IL-15Rα permits IL-15 and its three receptor components to be assembled identically at the ligand-receptor interface whether IL-15 is presented in cis or trans. Finally, we have gained insights into IL-15·IL-15Rα·IL-2Rβ·γc quaternary complex assembly through the use of molecular modeling.

Interleukin (IL)-15 is a pleiotropic cytokine that plays a pivotal role in both innate and adaptive immunity. IL-15 is unique among cytokines due to its participation in a trans signaling mechanism in which IL-15 receptor ␣ (IL-15R␣) from one subset of cells presents IL-15 to neighboring IL-2R␤/␥ c -expressing cells. Here we present the crystal structure of IL-15 in complex with the sushi domain of IL-15R␣. The structure reveals that the ␣ receptor-binding epitope of IL-15 adopts a unique conformation, which, together with amino acid substitutions, permits specific interactions with IL-15R␣ that account for the exceptionally high affinity of the IL-15⅐IL-15R␣ complex. Interestingly, analysis of the topology of IL-15 and IL-15R␣ at the IL-15⅐IL-15R␣ interface suggests that IL-15 should be capable of participating in a cis signaling mechanism similar to that of the related cytokine IL-2. Indeed, we present biochemical data demonstrating that IL-15 is capable of efficiently signaling in cis through IL-15R␣ and IL-2R␤/␥ c expressed on the surface of a single cell. Based on our data we propose that cis presentation of IL-15 may be important in certain biological contexts and that flexibility of IL-15R␣ permits IL-15 and its three receptor components to be assembled identically at the ligandreceptor interface whether IL-15 is presented in cis or trans. Finally, we have gained insights into IL-15⅐IL-15R␣⅐IL-2R␤⅐␥ c quaternary complex assembly through the use of molecular modeling.
Interleukin (IL) 2 -15 is a member of the four-␣-helix bundle family of cytokines, which plays a pivotal role in both innate and adaptive immunity (1,2). IL-15 was initially identified due to its ability to mimic the activity of IL-2 in vitro (3,4), and despite a lack of significant sequence homology, IL-15 and IL-2 are able to similarly stimulate the proliferation and differentiation of natural killer (NK), T, and B cells (1,2). Despite these many overlapping functions, IL-15 and IL-2 possess distinct biological activities in vivo, such as the ability of IL-15 to control the development of CD8 ϩ memory T cells and the involvement of IL-2 in the maintenance of regulatory T cells (5).
The cell-surface receptor for IL-15 comprises three subunits: IL-15 receptor (IL-15R) ␣, IL-2R␤ (also known as IL-15R␤, CD122, and p75), and ␥ c (also known as CD132 and p65). The ectodomain of IL-15R␣ consists of a single sushi domain (6) (also known as a short consensus repeat or complement control protein repeat), which is essential for IL-15 binding and IL-15R␣ function (7,8), a membrane-proximal proline-threonine-rich (PT) region, and a linker/hinge region that connects the sushi domain and the PT region (6). IL-2R␣ is structurally related to IL-15R␣ but contains an additional sushi domain that is connected to the first by a linker region. The ectodomain of IL-2R␤ and ␥ c each consist of two fibronectin-type III domains, which participate in IL-15 binding (9), and the cytoplasmic domains of IL-2R␤ and ␥ c sequester signaling molecules such as Jak1 and Jak3 (10,11). In contrast, the short cytoplasmic domains of IL-15R␣ and IL-2R␣ are not thought to play a major role in signaling (12), although the cytoplasmic domain of IL-15R␣ has recently been shown to play a role in receptor recycling (13). Because IL-2R␤ and ␥ c are shared subunits of the receptor for IL-2 (6,12,14), IL-15 and IL-2 are capable of activating similar signaling pathways, including Jak1/Jak3 and Stat3/Stat5, Ras/mitogen-activated protein kinase, and phosphatidylinositol 3-kinase pathways (1), which has been suggested to partially account for the overlapping functions of IL-15 and IL-2.
A possible mechanism by which IL-15 and IL-2 attain their distinct in vivo functions is through differences in the properties of their ␣ chain receptors. IL-15R␣ binds specifically to IL-15 with high affinity (K d ϭ 30 -100 pM) (6,12,15,16), whereas IL-2R␣ specifically binds to IL-2 with a comparatively lower affinity (K d ϭ 10 -30 nM) (17). IL-15, IL-2, and their ␣ chain receptors also exhibit distinct expression profiles, which have been suggested to contribute to their distinct biological activities (18). Furthermore, in contrast to IL-2, which acts as a soluble cytokine, IL-15 is coordinately expressed with IL-15R␣ by activated monocytes and dendritic cells and is extremely difficult to detect in free form in humans and mice due to stable complex formation with IL-15R␣ on the cell surface (13, 19 -23). Recent data have demonstrated that IL-15R␣ from activated monocytes presents IL-15 in trans to IL-2R␤/␥ c and/or IL-15R␣/IL-2R␤/␥ c -expressing cells such as CD8 ϩ memory T cells and NK cells (13,21,24). This finding suggests that the architecture of IL-15R subunits in the IL-15 quaternary complex differs from that of IL-2 receptor subunits in the IL-2 quaternary complex, which are assembled in cis (i.e. on the surface of the same cell) as revealed by the recently reported IL-2⅐IL-2R␣⅐IL-2R␤⅐␥ c crystal structure (25,26). Although trans presentation is currently thought to be the major mechanism by which IL-15 exerts its biological effects in vivo, the possibility for the assembly of a cis IL-15 quaternary complex on cells capable of expressing IL-15R␣, IL-2R␤, and ␥ c , has not been ruled out (18). However, considering the structural similarities between IL-15, IL-2, and their receptors (15,27,28), and data suggesting that IL-15 has only a single binding site for interaction with IL-15R␣ (29), the molecular basis by which trans presentation occurs and can even co-exist with the formation of IL-2-like cis quaternary complex is not clear.
In this report, we present the crystal structure of murine IL-15 in complex with the sushi-domain fragment of IL-15R␣, refined to 2.2-Å resolution. The structure reveals the molecular determinants of the specific, high affinity IL-15⅐IL-15R␣ interaction. Importantly, helix B and the hC-hD loop of IL-15 adopt unique conformations, which permit specific interactions with IL-15R␣ that likely account for its significantly higher affinity relative to the IL-2⅐IL-2R␣ interaction. We also present biochemical data demonstrating that, in addition to trans presentation, IL-15R␣ is capable of presenting IL-15 to IL-2R␤/␥ c in cis, suggesting a potential role for cis IL-15 presentation in certain biological contexts. Finally, analysis of the structure suggests that flexibility of the linker and/or PT region of IL-15R␣ permit IL-15 and its three receptor components to be assembled identically at the ligand-receptor interface whether IL-15 is presented in cis or trans.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The DNA fragments encoding the full-length ectodomain of mature murine IL-15R␣ (residues 1-168; sIL-15R␣ 1-168 ), and the full-length mature murine IL-15 (residues 1-114) were cloned into the TA vector pCR2.1TOPO (Invitrogen) as fusion proteins containing an N-terminal histidine tag and a tobacco etch virus protease cleavage site. Selenomethionine-labeled proteins were synthesized using the large scale dialysis mode of the Escherichia coli cell-free reaction (30). Both IL-15 and sIL-15R␣ 1-168 precipitated during synthesis. The precipitated proteins were denatured in 25 mM Tris-HCl buffer (pH 8.3) containing 6 M guanidine hydrochloride and 1 mM dithiothreitol and were refolded separately or co-refolded by rapid dilution into 50 mM Tris buffer (pH 8.3) containing 1 M arginine hydrochloride, 5 mM reduced glutathione, and 0.5 mM oxidized glutathione. After stirring for 24 h at 4°C, the protein solutions were filtered and dialyzed against 25 mM Tris-HCl buffer (pH 7.5) containing 150 mM NaCl for 12-24 h at 4°C. The refolded proteins were purified to homogeneity by chromatography on affinity (HisTrap) and gel-filtration (Superdex 200 HR 10/30) columns. The histidine tags were removed by overnight digestion with tobacco etch virus protease at 4°C. Due to the location of the tobacco etch virus cleavage site, seven residues from the histidine tag (GSSGSSG) remain at the N termini of the IL-15 and sIL-15R␣ 1-168 proteins after histidine tag cleavage. During the gelfiltration step with 25 mM Tris-HCl buffer (pH 7.5) containing 300 mM NaCl, IL-15 eluted as a mixture of trimeric and monomeric species, and only the monomeric protein was used for subsequent experiments. For complex formation, separately refolded monomeric IL-15 and sIL-15R␣ 1-168 were mixed at a 1.5:1 molar ratio and concentrated using an Amicon Ultra 15 filtration unit. The IL-15⅐sIL-15R␣ 1-168 mixture was loaded onto a Superdex 200 HR 10/30 column equilibrated in 25 mM Tris-HCl buffer (pH 7.5) containing 150 mM NaCl and eluted at a position consistent with 1:1 complex formation, which was confirmed via analytical ultracentrifugation. Co-refolded IL-15⅐sIL-15R␣ 1-168 complex behaved identically to the separately refolded and reconstituted complex during purification. The co-refolding method was utilized for large scale protein preparation due to its consistently higher yields. For large scale protein preparation, an anion exchange step (Mono Q HR 5/5) was added prior to the final gel-filtration step (which was performed using a Superdex 75 HR 16/60 column). All columns were purchased pre-packed from GE Healthcare.
Limited Proteolysis, Crystallization, and Data Collection-Initial crystallization trials of the IL-15⅐sIL-15R␣ 1-168 complex were performed using the standard sitting drop vapor diffusion method. Despite extensive screening efforts, we were unable to obtain crystals of the complex using the full-length ectodomain of IL-15R␣. During protein synthesis and purification, we noted that the C-terminal region of sIL-15R␣ 1-168 was extremely sensitive to proteolytic degradation, suggesting that this region of the receptor is highly flexible.
To obtain boundaries more suitable for crystallization, we performed limited proteolysis experiments on the IL-15⅐sIL-15R␣ 1-168 complex. Using elastase (Nacalai Tesque), a stable fragment of the complex was identified during small scale pilot experiments, and the reaction conditions were scaled-up. The digested protein was purified by gel-filtration chromatography and analyzed using MALDI-TOF and N-terminal sequencing. The results of this analysis showed that the PT region and most of the linker region of IL-15R␣ are rapidly degraded and that the stable fragment of IL-15R␣ corresponds to residues 1-71, which includes the entire sushi domain and the first five residues of the linker region. The results also showed that IL-15 was not proteolytically digested. On the basis of these results a modified IL-15R␣ construct (sIL-15R␣ 1-71 ) was generated, synthesized, and purified in complex with IL-15 as described above for sIL-15R␣  .
Diffraction quality crystals of the selenomethionine-labeled IL-15⅐sIL-15R␣ 1-71 complex (hereafter referred to as IL-15⅐IL-15R␣) were grown by mixing 2 l of IL-15⅐IL-15R␣ (18 mg/ml in 25 mM Tris-HCl buffer (pH 7.5) containing 150 mM NaCl) with 2 l of crystallization buffer (1.7 M sodium/potassium phosphate, pH 6.8, 3% benzamidine-HCl) and incubating at 20°C using hanging drop vapor diffusion. Large, single crystals measuring 0.5 mm in the longest dimension appeared within 5-7 days. Crystals were flash frozen in a final cryoprotectant solution composed of the mother liquor and 25% glycerol. Diffraction data were measured on an ADSC Quantum210 detector at Photon Factory-Advanced Ring, Beamline NW12. The data were indexed, integrated, and scaled using HKL2000 (31). The IL-15⅐IL-15R␣ crystals belong to the tetragonal space group P42 1 2 with unit cell dimensions a ϭ b ϭ 118.5 Å, c ϭ 76.11 Å. There are two IL-15⅐IL-15R␣ complexes in the asymmetric unit, with a solvent content of ϳ58%.
Structure Determination and Refinement-The IL-15⅐IL-15R␣ structure was solved using diffraction data collected at the selenium absorption peak wavelength ( ϭ 0.9793 Å) of one selenomethionine-derivatized crystal. The program SOLVE (32) was used to locate the selenium sites and to calculate the phases using the single wavelength anomalous dispersion method. The program RESOLVE (33) was used for density modification, and ARP/wARP (34) was used to generate an initial model via automatic model building. Positional and B-factor refinements were performed using the program CNS (35), and model building into the 2F o Ϫ F c and F o Ϫ F c electron density maps was performed with the program O (36). The final refined atomic model contains residues 1-114 of both copies of IL-15 in the asymmetric unit, residues 3-71 in one copy of IL-15R␣, and residues 3-70 in the other copy of IL-15R␣ in the asymmetric unit, 90 water molecules, and one benzamidine molecule. The remaining residues of the N-terminal histidine tag are completely disordered in both copies of IL-15R␣, whereas the last three residues of this region are ordered in both copies of IL-15 and are included in the atomic model. Additionally, residues 1-2 are disordered in both copies of IL-15R␣, and residue 71 is disordered in one copy of IL-15R␣ in the asymmetric unit. The atomic coordinates and structure factors have been deposited into the Protein Data Bank under accession code 2PSM. Superimposition of the C ␣ trace of the two copies of IL-15 and the IL-15R␣ shows that they are essentially identical (r.m.s.d. ϭ 0.451 Å and 0.202 Å, respectively). Representations of the structure were created with the programs PyMOL 3 and CCP4MG (38).
Generation of IL-15R␣-and IL-2R␤-expressing FDC3 Cell Line-The cDNA of full-length murine IL-15R␣ and IL-2R␤ were amplified by PCR and introduced into retroviral vectors, MSCV-IRES-GFPRV and pNX-IRES-hCD4RV, respectively. The murine mast cell line, FDC-P1 (39), was purchased from ATCC, re-cloned, and renamed as FDC3. This cell line was cultured in Dulbecco's modified Eagle's medium containing 10% calf serum and conditioned media of P815 cells expressing IL3. PLAT:E packaging cell lines were transfected with IL-15R␣/MSCV-IRES-GFPRV or IL-2R␤/pMX-IRES-hCD4RV by FuGENE according to the manufacture's instructions. After 24 h, the medium was replaced, and the retroviral supernatant was generated by culturing the cells for a further 24 h. The FDC3 cell line was infected by overnight culture in retroviral culture supernatant containing 8 mg/ml Polybrene. Infected FDC3 cell lines were purified by cell sorting for expression of the green fluorescent protein and/or the hCD4 surface markers. Sorted cells were expanded, were Ͼ90% pure, and stably expressed the retroviral markers.
Single Cell Assay-FDC3␣␤ and FDC3␤ cells cultured in IL-3 containing Dulbecco's modified Eagle's medium for 4 days were washed with phosphate-buffered saline three times, and cultured in IL-15-free, IL3-free Dulbecco's modified Eagle's medium with 10% calf serum for 7 h. A single cell was picked up by micropipette and transferred into a well of a Terasaki plate (Nunc, Roskilde, Denmark) containing Dulbecco's modified Eagle's medium with 10% calf serum in the presence or absence of 30 ng/ml IL-15. Cell division was checked under the microscope at 0 and 22 h of incubation.
Thymidine Uptake Assay-The sIL-15R␣ protein used for the thymidine uptake assay was produced using a modified method of a previously published protocol (40). Briefly, the cDNA coding full-length ectodomain of murine IL-15R␣ was amplified by PCR and subcloned into the pQE30 bacterial expression vector (His-sIL-15R␣ 1-168 ). His-sIL-15R␣ 1-168 was expressed in E. coli strain M15 cells and was purified to homogeneity by nickel affinity and gel-filtration chromatography.
Proliferation assays were set up in 96-well microtiter plates. Murine IL-15 (R&D) was diluted across the plates in the presence or absence of 1 g/ml His-sIL-15R␣ 1-168 protein. In the experiment using IL-15⅐IL-15R␣ 1-168 complex or IL-15⅐IL-

RESULTS AND DISCUSSION
To elucidate the structural basis for the high affinity and specificity between IL-15 and IL-15R␣, and to obtain insights into the architecture of the IL-15⅐IL-15R complex, we have determined the crystal structure of IL-15 in complex with IL-15R␣. Initial attempts to crystallize IL-15 in complex with the full-length ectodomain of IL-15R␣ were unsuccessful. Limited proteolysis experiments performed on purified IL-15⅐IL-15R␣ complex led to the identification of a stable IL-15R␣ fragment (encompassing the sushi domain and the first five residues of the linker region) that was amenable to crystallization (see "Experimental Procedures"). The IL-15⅐IL-15R␣ structure was solved using the single wavelength anomalous dispersion method and has been refined to 2.2-Å resolution with working and free R values of 22.4% and 23.6%, respectively ( Table 1). As shown in Fig. 1A, IL-15R␣ interacts with the upper edge of the IL-15 four-helix bundle.
Structure of Receptor-bound IL-15-The crystal structure shows that IL-15 comprises a left-handed antiparallel four-helix bundle (hA, hB, hC, and hD) with the helices arranged in an up-up-down-down topology typical of members of the short-chain subfamily of helical cytokines ( Fig. 1A) (41). As expected based on primary sequence, IL-15 contains two disulfide bridges, which the structure shows help to stabilize the conformation of the hC-hD loop in regions of the loop that engage in contacts with the receptor (Fig. 1A).
Despite sharing only 16% sequence identity overall, the helices that comprise the four-helix bundles of receptor bound IL-15 and IL-2 (PDB ID: 1Z92 and 2B5I) (25,42) superimpose well, with an r.m.s.d. of 1.2 Å over 68 equivalent C␣ positions (Figs. 1B and 2A). The most obvious structural differences between IL-15 and IL-2 result from differences in the length of the helices that form the core of the protein and length and conformational differences in the loop regions between the helices (hA-hB, hB-hC, and hC-hD loops) (Figs. 1B and 2A). Importantly, the hA-hB loop of IL-15 is shorter by four residues compared with that of IL-2, which prevents the formation of a helix corresponding to hAЈ of IL-2 (Figs. 1B and 2A). Another important difference is observed in hB of IL-15, which adopts a more rigid and extended conformation compared with the corresponding region of IL-2 (hB and hBЈ), which is bent due to the presence of a proline residue (Pro-65) at the start of hBЈ (Figs. 1B and 2A). IL-15 lacks a proline residue in its hB region, which permits the C-terminal extension of hB and more rigid conformation of the helix. Because hB and the hA-hB loop region account for a major portion of the ligand-receptor interface in both the IL-15⅐IL-15R␣ and IL-2⅐IL-2R␣ complexes, these conformational differences have an impact on the distinct modes of ␣ receptor binding observed for these ligands (see below).
A final structural difference worth noting is the fact that hA of IL-15 is significantly shorter at its N terminus compared with hA of free or receptor-bound IL-2 from previously determined crystal structures (PDB ID: 1Z92, 2ERJ, and 2B5I) (Figs. 1B and 2A). hA of IL-2 plays a role in the assembly of the IL-2⅐IL-2R␣⅐IL-2R␤⅐␥ c complex (26) (IL-2 quaternary complex), so the significantly shorter N terminus of IL-15 relative to IL-2 has possible implications on the IL-15 mode of binding to IL-2R␤/ ␥ c , which will be discussed in detail below.
Structure of Ligand-bound IL-15R␣-The IL-15R␣ sushi domain has an elliptical structure comprised of six ␤-strands (␤B, ␤C, ␤CЈ, ␤D, ␤E, and ␤EЈ) that envelope a compact hydrophobic core (Fig. 1A). The central ␤-sheet of the IL-15R␣ sushi domain is formed by strands ␤C, ␤D, and ␤E. In typical sushi domain structures two ␤-strands at the N-terminal region of the domain (␤A and ␤B) pack against the central ␤-sheet to form a ␤-sandwich arrangement (43). As noted previously in the NMR structure of unliganded human IL-15R␣ sushi domain (PDB ID: 2ERS), this arrangement is not observed in the IL-15R␣ sushi domain fold due to the presence of three consecutive proline residues (Pro-5, Pro-6, and Pro-7) that prevent the formation of the ␤A strand (Fig. 1A) (27). The IL-15R␣ sushi domain contains two disulfide bridges: the first between Cys-4 at the N terminus of the sushi domain and Cys-46 in ␤D, which serves to pin the highly flexible N-terminal region to the central ␤-sheet of the domain, and the second between Cys-30 at the C-terminal end of the ␤C strand and Cys-64 in the ␤EЈ strand, which seems to stabilize the sushi domain core (Fig. 1A). Comparison of our crystal structure of ligand-bound murine IL-15R␣ to the NMR structure of unliganded human IL-15R␣ (PDB ID: 2ERS) shows that the core of the IL-15R␣ sushi domain does not undergo significant conformational changes upon ligand binding. However, the conformation of loop regions between the ␤-strands (outside of regions involved in ligand binding) are quite different, presumably due to inherent flexibility (data not shown).
Although IL-15R␣ and IL-2R␣ are structurally related in that they are composed of sushi domains, linker, and PT regions, IL-15R␣ contains a single sushi domain that mediates ligand binding, whereas IL-2R␣ contains two sushi domains separated by a linker region (Fig. 2B). The previously determined IL-2⅐IL-2R␣ crystal structure (PDB ID: 1Z92) (42) showed that the two sushi domains of IL-2R␣ engage in strand exchange and that the N-terminal sushi domain (IL-2R␣ D1) participates in most of the interactions with IL-2. Superimposition of IL-2R␣ D1 onto the sushi domain of IL-15R␣ from the IL-15⅐IL-15R␣ structure shows that the conformation of most of the receptor regions involved in ligand binding adopt a similar conformation (␤C-␤EЈ region of the receptors superimpose with an r.m.s.d. of 1.73 Å) (Fig. 2C). In contrast, the strand-exchanged region of IL-2R␣ D1 (residues 102-122), which is involved in some contacts with IL-2, adopts a significantly different conformation from the corresponding region of IL-15R␣ (residues 3-20), which does not make any contacts with IL-15 (Fig. 2C). The effects of these structural differences on the distinct modes of IL-15⅐IL-15R␣ and IL-2⅐IL-2R␣ binding will be discussed below.
IL-15⅐IL-15R␣ Interface-The IL-15R␣ binding surface of IL-15 comprises residues from hB, hCЈ, and the hA-hB and hC-hD loops, whereas the IL-15 binding surface of IL-15R␣ comprises residues from strands ␤C and ␤CЈ and the ␤CЈ-␤D and ␤E-␤EЈ loops (Fig. 1). Calculation of the surface electrostatic potential of IL-15 reveals a groove at the receptor-binding interface that is lined with acidic residues and thus possesses a striking degree of negative potential (Fig. 3A). Calculation of the surface electrostatic potential of IL-15R␣ reveals three prominent "knobs" of positive potential located at its ligandbinding surface that are formed by Arg-25 and Arg-27 in ␤C, and Arg-36 in ␤CЈ (Fig. 3A). In forming the IL-15⅐IL-15R␣ com-  A, ribbon representation of the IL-15⅐IL-15R␣ complex. IL-15 is colored orange, and IL-15R␣ is colored green. Secondary structure assignments were performed using PROCHECK (37). The ␤-strands of IL-15R␣ and helices of IL-15 are labeled according to the conventional nomenclature for these proteins. The side chains of cysteine residues that participate in intramolecular disulfide bridge formation in IL-15 and IL-15R␣ are shown as sticks, and their sulfur atoms are colored yellow. The N and C termini of are labeled NT and CT, respectively. All figures present the copy of the IL-15⅐IL-15R␣ complex from the asymmetric unit that had stronger electron density. All figures were prepared using the program PyMOL unless noted otherwise. B and C, structure-based sequence alignment of selected interleukins (B) and interleukin receptor alpha sushi domains (C). Predicted signal sequences have been omitted from the alignment. Residue numbers are in bold font to the left of the sequence alignment. All proteins are numbered starting from the first residue after predicted signal peptide cleavage. A dash represents a gap introduced to optimize the alignment. The locations of the secondary structural elements are indicated by black boxes. The proline residue in the IL-2 that contributes to the bent conformation of the hB-hBЈ region is circled in red. IL-15 residues that interact with IL-15R␣ are shaded green, IL-15R␣ residues that interact with IL-15 are shaded orange, IL-2 residues that interact with IL-2R␣ are shaded cyan, and IL-2R␣ residues that interact with IL-2 are shaded purple. IL-15 and IL-15R␣ residues that play a particularly important role in determining the unique mode IL-15⅐IL-15R␣ specificity are marked with an asterisk. IL-2 residues that interact with IL-2R␤ or IL-2R␥ c in the IL-2 quaternary complex (25, 26) (PDB IDs: 2B5I and 2ERJ) are indicated with ␤ or ␥, respectively, below the sequence alignment. The sequences of D. rerio IL-15 (zIL-15), G. gallus IL-15 (cIL-15), and G. gallus IL-15R␣ (cIL-15R␣) are shown to highlight the evolutionary conservation of the IL-15⅐IL-15R␣ mode of binding, which is unique among cytokine⅐cytokine receptor complexes. plex, the IL-15R␣ knobs insert into the groove of IL-15 to form an interlocking interface with good charge and shape complementarity (S c ϭ 0.73) and high specificity (14 direct hydrogen bonds). 14 residues from IL-15 and 15 residues from IL-15R␣ bury a total surface area of 1240 Å 2 at the IL-15⅐IL-15R␣ interface.
Analysis of the interactions at the IL-15⅐IL-15R␣ interface reveal two adjacent concentrated clusters of primarily hydrogen bond-mediated interactions (termed sites 1 and 2, Fig. 4, A  and B), which likely account for most of the binding energy of the complex. At the center of site 1, Arg-27 in ␤C of IL-15R␣ forms salt bridges with Glu-53 at the C-terminal tip of hB and Asp-22 in the hA-hB loop of IL-15 (Fig. 4A). As shown in Fig.  4A, these central salt bridges are surrounded by additional contacts between residues in hB of IL-15 and the ␤CЈ-␤D and ␤E-␤EЈ loops of IL-15R␣. The set of interactions observed at site 1 is unique to the IL-15⅐IL-15R␣ complex and constitutes a major determinant of the unique mode of IL-15⅐IL-15R␣ binding (see below).
Arg-36 of IL-15R␣ is at the center of the network of hydrogen bonds that take place at site 2 of the IL-15⅐IL-15R␣ interface.
Here, Arg-36 of IL-15R␣ forms salt bridges with Glu-46 in the middle of hB and Glu-93 in the hC-hD loop of IL-15 (Fig. 4B). Adjacent to this set of interactions, Tyr-26 and Glu-46 of IL-15 engage in hydrogen bonds with the backbone oxygen of Arg-36 and the backbone nitrogen of Gly-39, respectively, in IL-15R␣ (Fig. 4B). Although most of the contacts at the IL-15⅐IL-15R␣ interface are hydrogen bond-mediated, there are a few important hydrophobic interactions that take place at site 2. The first of these hydrophobic contacts is mediated by the methyl group of Ala-38 at the tip of the ␤CЈ-␤D loop of IL-15R␣, which inserts into a prominent hydrophobic pocket formed by Tyr-26, Leu-45, and Val-49 and a partially solvent-exposed disulfide bridge between Cys-42 and Cys-88 in IL-15 (Fig. 4B). This interaction strengthens the site 2 interface directly via binding and contributes to the exclusion of bulk solvent by capping one end of the IL-15⅐IL-15R␣ interface. Adjacent to this interaction, Ile-65 and Pro-68 in the ␤⌭Ј and linker regions, respectively, of IL-15R␣, pack against a hydrophobic patch formed by the aliphatic portion of the side chains of Glu-87, Glu-89, and Glu-90 in hCЈ and the hC-hD loop of IL-15 (Fig. 4B). This set of hydrophobic interactions at site 2 is unique to the IL-15⅐IL-15R␣ interface and, in addition to site 1 interactions, play a role in determining IL-15⅐IL-15R␣ specificity (see below).
The residues involved in the key interactions identified in the murine IL-15⅐IL-15R␣ crystal structure are in good agreement with published mutagenesis studies (15,28). Scatchard analysis using human proteins shows that mutation of IL-15R␣ Arg-36 to threonine results in an 80-fold decrease in affinity for IL-15 (28), and mutation of IL-15 Glu-46 to lysine results in a complete loss in affinity for IL-15R␣ (15), underscoring the importance of the site 2 interactions to general binding affinity. Furthermore, sequence alignment shows that all of the key residues involved in murine IL-15 and IL-15R␣ complex formation, including Arg-27, Arg-36, and Ala-38 of IL-15R␣ and Tyr-26, Glu-46, and Glu-53 of IL-15, are fully conserved in all vertebrate IL-15 and IL-15R␣ orthologs for which sequence information is available (Fig.  1, B and C). This suggests that the IL-15⅐IL-15R␣ mode of binding has been conserved throughout evolution, and thus, our structure provides the molecular basis for the high affinity and specificity of the IL-15⅐IL-15R␣ complex for all vertebrates.
This hypothesis has been reinforced by the recently determined human IL-15⅐IL-15R␣ crystal structure, which was published by Chirifu et al. (44) during the preparation of this report (PDB ID: 2Z3Q and 2Z3R). The structures of the human and mouse orthologs of IL-15 and IL-15R␣ are similar (supplemental Fig. S1, A and B), and as expected the interactions at sites 1 and 2 are conserved (supplemental Fig. S1, C and D). One unexpected structural difference between the human and mouse IL-15 proteins is that, although hB of both orthologs adopts the same extended conformation (which permits Glu-53 at the tip of hB to salt bridge with Arg-27 of IL-15R␣), the orientation of hB (IL-15R␣ binding site) relative to helices A, C, and D (IL2R␤/␥ c binding site), is slightly different between the two structures (supplemental Fig. S1A). This appears to be due to sequence divergence in the hB-hC loop of the ligands, in particular Tyr-54 (which is Ser-54 in human IL-15) (Fig. 1B), which would severely clash with residues in hB, hC, and the hA-hB loop if murine IL-15 adopted the same structure as human IL-15 in this region (supplemental Fig. S2).

IL-15⅐IL-15R␣ Mode of Binding Is Distinct from That of IL-2⅐IL-2R␣
and Other Cytokine⅐Cytokine Receptor Complexes-In marked contrast to the IL-15⅐IL-15R␣ interface, which is dominated by salt bridges/ hydrogen bonds, the IL-2⅐IL-2R␣ interface (PDB ID: 1Z92) is dominated by hydrophobic interactions at its center, which are surrounded by polar interactions (42). As predicted by Lorenzen et al. (27) based on molecular modeling, calculation of surface electrostatic potential of IL-15 and IL-15R␣ from our crystal structure shows that the IL-15⅐IL-15R␣ interface contains more charged residues and exhibits a higher degree of charge and shape complementarity compared with the IL-2⅐IL-2R␣ interface (compare Fig. 3, A and B). Rickert et al. identified two critical interaction sites at the IL-2⅐IL-2R␣ interface: one centered around Phe-42 in the hA-hB loop of IL-2 (patch 1, Fig. 4C) and the second centered around Tyr-45 in the hA-hB loop of IL-2 (patch 2, Fig. 4D) (42). To compare their modes of binding, we superimposed IL-2 from the IL-2⅐IL-2R␣ structure (PDB ID: 1Z92) onto IL-15 from our IL-15⅐IL-15R␣ structure. Interestingly, the superimposition shows that, despite the considerably different nature of their interactions, patch 1 and patch 2 of the IL-2⅐IL-2R␣ interface overlap well with site 1 and site 2 of the IL-15⅐IL-15R␣ interface (Fig. 4, compare A with C, and B with D).
At patch 1 of the IL-2⅐IL-2R␣ interface, hAЈ and hBЈ of IL-2 adopt a significantly different conformation than the corresponding regions at site 1 of IL-15 (Fig. 4, compare C and A). In the IL-2⅐IL-2R␣ structure, hAЈ and hBЈ lie roughly parallel to one another and form an interface with the unique strandswapped regions of sushi domain 1 (particularly ␤G) and sushi domain 2 (particularly the N terminus) of IL-2R␣ (Fig. 4C). Although these interactions account for only a small portion of the IL-2⅐IL-2R␣ interface and are slightly different in each of the three IL-2⅐IL-2R␣-containing crystal structures due to crystal packing (PDB IDs: 2B5I, 1Z92, and 2ERJ), they are significant in terms of binding mechanism, because they result in a slight rotation of IL-2R␣ relative to IL-15R␣, which has only one sushi domain and thus cannot participate in equivalent interactions (Fig. 5). Importantly, Leu-42 and Tyr-43 of IL-2R␣ ␤D are brought into the vicinity of Phe-42 due to the relative rotation of IL-2R␣, where they engage in a hydrophobic interaction that acts as the primary energetic determinant in the IL-2 binding epitope (Figs. 4C and 5) (42). The overall result of this relative rotation is that residues with different physicochemical properties and from different regions of the receptor and ligand form the functional epitopes at the IL-2⅐IL-2R␣ and IL-15⅐IL-15R␣ interfaces (Fig. 4, compare C and A).
Compared with site 1/patch 1, the superimposition shows that the nature of the interactions at site 2/patch 2 of the IL-15⅐IL-15R␣ and IL-2⅐IL-2R␣ interfaces are better conserved (Fig. 4, compare D and B). Patch 2 at the IL-2⅐IL-2R␣ interface is centered around Tyr-45 in the hA-hB loop of IL-2, which packs against the aliphatic part of Arg-35 and Arg-36 from the ␤CЈ strand of IL-2R␣ and engages in a hydrogen bond with the backbone nitrogen of Arg-35 (Fig. 4D). Surrounding this central hydrophobic cluster are several polar interactions, including a salt bridge between Arg-36 of IL-2R␣ and Glu-62 of IL-2 (Fig. 4D). Interestingly, this salt bridge corresponds to the salt bridge between Arg-36 of IL-15R␣ and Glu-46 of IL-15 (Fig. 4,  compare D and B). Tyr-45 of IL-2 is also conserved in IL-15 (Tyr-26, Fig. 1B), however, Tyr-26 of IL-15 engages in a distinct set of interactions with the receptor in the IL-15⅐IL-15R␣ structure (Fig. 4B). Superimposition shows that site 2/patch 2 serves as the pivot point of the relative rotation observed at the IL-2⅐IL-2R␣ and IL-15⅐IL-15R␣ interfaces (Fig. 5), which is not surprising considering the conservation of key interactions at this region mentioned above.   Fig. 2. Regions of the complexes that engage in unique contacts are indicated with arrows. Note that IL-15 and IL-2 superimpose well with one another, but that there is a slight rotation of the receptors relative to one another, which allows the formation of specific contacts in the IL-15⅐IL-15R␣ and IL-2⅐IL-2R␣ complexes. Also note that distinct features of IL-2, including a longer hA-hB loop (and presence of hAЈ) and the kink in hB (resulting in hBЈ) enable unique contacts with the bottom edge of IL-2R␣ D2, the ␤G strand from the strand-swapped region of D1, and the ␤D strand (refer to Fig. 4C). The extended hB of IL-15 and differences in sequence allow for unique contacts with the ␤C strand IL-15R␣ (refer to Fig. 4A). The pivot point of the rotation is marked with a red circle and corresponds to residue Ala-38 of IL-15 and Lys-38 of IL-2, which are located in site 2/patch 2 of the IL-15⅐IL-15R␣ and IL-2⅐IL-2R␣ interfaces (refer to Fig. 4, B and D).
IL-15R␣ and IL-2R␣ have similar domain composition, share limited sequence homology, and together are unique among cytokine receptors in that sushi domains mediate interactions with their specific four-helix bundle cytokines. However, despite these similarities, comparison of the structures shows that the IL-15⅐IL-15R␣ and IL-2⅐IL-2R␣ modes of binding have diverged during evolution. The nature of the interactions at the IL-2⅐IL-2R␣ interface resembles those of cytokine⅐ cytokine receptor complexes such as growth hormone⅐growth hormone receptor (PDB ID: 3HHR) (45) and GCSF⅐GCSFR (PDB ID: 1PGR) (46) complexes, which are dominated by hydrophobic contacts, whereas the nature of interactions at the IL-15⅐IL-15R␣ interface resembles the hydrogen bond-dominated "avocado clusters" observed at the IL4⅐IL4R␣ interface (PDB ID: 1IAR) (47), or the mosaic, interlocking interface of the p35⅐p40 complex of IL12 (PDB ID: 1F45) (48). A functional consequence of this divergence between IL-15⅐IL-15R␣ and IL-2⅐IL-2R␣ is that it results in complexes with different affinities, which may account for the inherently different signaling properties and biological activities of IL-15 and IL-2 by modulating the strength and duration of signaling by the ␤␥ c complex. Indeed, it has recently been demonstrated that IL-15 is unique among cytokines due to its persistent in vivo activity and ability to mediate prolonged signaling in CD8 ϩ T cells (13). Consistent with this observation, IL-15 is predominantly detected in complex with IL-15R␣ at the plasma membrane (effectively increasing its local concentration), and upon endocytosis these IL-15⅐IL-15R␣ complexes are stable and can be effectively recycled back to the cell surface ("intercellular reservoir effect" (13,22)). It has been suggested that both of these unique properties of IL-15 may be mediated by the high affinity and enhanced stability of the IL-15⅐IL-15R␣ complex (13,22,29,49).
The binding kinetics of the IL-15⅐IL-15R␣ and IL-2⅐IL-2R␣ interactions have previously been analyzed using surface plasmon resonance (16,50). These studies have shown that, although the k on of the IL-15⅐IL-15R␣ and IL-2⅐IL-2R␣ interactions are comparable, the k off of the IL-15⅐IL-15R␣ interaction is more that 4 orders of magnitude slower compared with the IL-2⅐IL-2R␣ interaction. Because k off is determined by the strength of short range interactions, it is likely that the additional salt bridges observed in the IL-15⅐IL-15R␣ complex at sites 1 and 2 and the excellent shape complementarity at the IL-15⅐IL-15R␣ interface are the main determinants of its slower k off , and thus, higher affinity. Importantly, the IL-15⅐IL-15R␣ crystal structure will provide an excellent molecular framework for the determination of the precise contributions of the individual interactions to the thermodynamics and kinetics of the IL-15⅐IL-15R␣ interaction in future studies.

IL-15⅐IL-15R␣⅐IL-2R␤⅐␥ c Quaternary Complex Model and Biochemical Evidence for cis Presentation of IL-15-Previously
published mutagenesis studies have suggested that IL-15 and IL-2 share similar interaction surfaces for IL-15R␤ and ␥ c (51). To gain insights into the architecture of IL-15 and IL-15R subunits during receptor activation, we created a model of the IL-15⅐IL-15R␣⅐IL-2R␤⅐␥ c complex by superimposing IL-15 and IL-15R␣ from the IL-15⅐IL-15R␣ crystal structure onto IL-2 and IL-2R␣ from the IL-2 quaternary complex structure (PDB ID: 2B5I) (Fig. 6A). The IL-15⅐IL-15R␣⅐IL-2R␤⅐␥ c complex model shows that the interaction surface of IL-15 and IL-2 for their respective ␣ receptors is also shared, and comparison of the topology of IL-15⅐IL-15R␣ and IL-2⅐IL-2R␣ structures shows that, with respect to the polarity of their N and C termini, IL-15 and IL-15R␣ are oriented similarly to IL-2 and IL-2R␣ at the ligand-receptor interface. Together, these findings are somewhat surprising considering the ability of IL-15R␣ to mediate trans IL-15 presentation and suggest that IL-15R␣ should be capable of presenting IL-15 in cis to IL-2R␤/␥ c in a manner reminiscent of the IL-2⅐IL-2R␣⅐IL-2R␤⅐␥ c complex. With respect to previously published models of the IL-15⅐IL-15R␣ complex, the topology of the IL-15⅐IL-15R␣ complex revealed by our crystal structure most resemble the models proposed by Lorenzen et al. (27) and "docking solution 2" described by Quemener et al. (28).
To test the ability of IL-15R␣ to present IL-15 in cis to IL-2R␤/␥ c on the surface of the same cell, we utilized the mouse mast cell line FDC3, which normally express the ␥ c subunit of IL-15R, but not IL-15R␣ or IL-2R␤. FDC3 cells were infected with retroviruses expressing full-length murine IL-2R␤ only (FDC3␤) or murine IL-15R␣ and IL-2R␤ (FDC3␣␤), and we employed the use of a single cell proliferation assay in which the ability of a single cell to proliferate in response to IL-15 is tested, thereby eliminating the possibility of trans presentation. As shown in Fig. 6B, a single FDC3␣␤ cell was consistently able to proliferate in the presence of 30 ng/ml IL-15, whereas FDC3␤ did not, clearly demonstrating that cis presentation can indeed occur.
We next decided to analyze the efficiency of cis versus trans IL-15 presentation by comparing the effect that soluble IL-15R␣ and soluble IL-15⅐IL-15R␣ complex have on the ability of IL-15 to induce proliferation in FDC3␤ versus FDC3␣␤ cells. As shown in Fig. 6 (C and D), FDC3␤ cells could proliferate only in the presence of high concentration (EC 50 Ϸ 100 ng/ml) of murine IL-15 alone, whereas the proliferative capacity of IL-15 was markedly enhanced in the presence of His-sIL-15R␣  or IL-15⅐sIL-15R␣ 1-168 complex (EC 50 Ϸ 3 ng/ml). Importantly, the IL-15⅐sIL-15R␣ 1-71 complex behaved nearly identically to the IL-15⅐sIL-15R␣ 1-168 complex with respect to the enhancement of IL-15 activity on FDC3␤ cells (Fig. 6D), demonstrating that the IL-15R␣ fragment crystallized in this study (residues 1-71) is responsible for the enhanced biological activity of IL-15. Interestingly, in contrast to the enhancing activity of soluble IL-15R␣ on FDC3␤ cells, the proliferative capacity of IL-15 on FDC3␣␤ cells was decreased in the presence of His-sIL-15R␣ 1-168 (EC 50 Ϸ 3 ng/ml) compared with IL-15 alone (EC 50 Ϸ 0.05 ng/ml).
Cumulatively, the above results indicate that IL-15 is most active when it is presented in cis by membrane-bound IL-15R␣ to IL-2R␤/␥ c on the surface of the same FDC3␣␤ cell. In this context, soluble IL-15R␣ inhibits the activity of IL-15 presumably by competing with membrane-bound IL-15R␣ for IL-15. When trans presentation is forced (on FDC3␤ cells), soluble IL-15R␣ enhances the activity of IL-15 relative to IL-15 alone, but IL-15 signaling in this context is never as effective as cis presentation on FDC3␣␤ cells. The molecular basis by which soluble IL-15R␣ enhances the activity of IL-15 on FDC3␤ cells ates soluble IL-15R sushi domain that forms a complex with IL-15 and is capable of signaling through the IL-2R␤/␥ c (23). Bulanova et al. (23) suggest that, in addition to the membranebound IL-15⅐IL-15R␣ complex, these alternatively spliced sushi domains may also play an important role in IL-15 trans presentation in vivo, and by extension of our model, we propose that these complexes exhibit the same architecture as presented in Fig. 6A.
Insights into the IL-2R␤ and ␥ c Receptor Binding Sites of IL-15-Consistent with the hypothesis that IL-15 and IL-2 share a common interface with IL-2R␤ and ␥ c , analysis of the IL-15 quaternary complex model shows that hA, hC, and hD adopt a similar conformation and that there are no major steric clashes between IL-15 and the IL-2R␤/␥ c receptors (Fig. 6A). Importantly, IL-15 residues Asp-8 in hA and Gln-108 in hD, which are conserved in IL-2 and known to be essential for IL-2R␤ and ␥ c receptor binding, respectively (53), are positioned to make productive interactions (data not shown). The model also provides a likely explanation for the observation that Tyr-103 of ␥ c is more important for IL-15 binding than for IL-2 binding (54). In the IL-2 quaternary complex, Tyr-103 of ␥ c engages in hydrophobic contacts with IL-2 residues Gln-126, Ser-127, and the methyl group of Thr-123. Ser-127 of IL-2 has diverged to a methionine residue in IL-15 (Met-109) (Figs. 1B and 7), which substantially increases the hydrophobicity of this region and likely strengthens the interaction with Tyr-103 of ␥ c . Sequence analysis shows that in addition to Asp-8 and Gln-108, many IL-15 residues at the predicted IL-15⅐IL-2R␤/␥ c interfaces are conserved or conservatively substituted (Fig. 1B), which likely imparts a degree of general binding affinity to IL-2 and IL-15 for the shared ␤ and ␥ receptors. A notable difference observed in the IL-15 quaternary complex model at the IL-15⅐␥ c interface is that an insertion of three residues and altered conformation of the hA-hB loop immediately preceding hB of IL-15 brings this region (IL-15 residues Asp-30, Phe-31, and His-32) in the vicinity of the hC-hCЈ1 loop of ␥ c (Fig. 7). Interestingly, the conformation of this loop region is stabilized by the unique disulfide bridge between Cys-35 in the hA-hB loop and Cys-85 in the hC-hD loop of IL-15. Potential contacts between IL-15 and the hC-hCЈ1 loop of ␥ c would be specific to IL-15, because this loop does not make contacts with IL-2 in the IL-2 quaternary complex. Overall the model is consistent with previous data showing that the ␤ and ␥ c receptors utilize an overlapping but non-identical binding surface for IL-2 and IL-15 and that there are differences in the relative contribution of individual residues to complex formation.
In contrast to formation of the IL-2 quaternary complex, which demonstrates cooperative binding (K d ϭ 10 -30 nM for IL-2⅐IL-2R␣; K d ϭ 10 pM for IL-2 quaternary complex), there is not a significant increase in affinity of the IL-15 quaternary complex compared with the affinity of IL-15 for the ␣ receptor alone (K d ϭ 30 -100 pM) (9,12,55). In the IL-2 quaternary complex, IL-2 residues Leu-12, Glu-15, Leu-18, and Leu-19 from the N-terminal portion of hA participate in interactions at a three-way junction between IL-2⅐IL-2R␤ and ␥ c that is suggested to mediate cooperative binding (26). As mentioned above, one important difference between IL-2 and IL-15 is that hA of IL-15 is shorter by seven residues at its N terminus compared with IL-2 from the IL-2 receptor complex and is shorter by nine residues compared with IL-2 from the IL-2 quaternary complex (Figs. 1B and 7). Due to its shorter N terminus, in IL-15 there is no corresponding residue to Leu-12 of IL-2. In addition, residues Glu-15, Leu-18, and Leu-19 of IL-2 have diverged to Ile-3, Arg-6, and Tyr-7 in IL-15 (Figs. 1B and 7). It is tempting to suggest that, because of these differences in sequence and structure, the corresponding interactions at the IL-15⅐IL-2R␤/␥ c receptor junction cannot take place or are weakened in the IL-15 quaternary complex, which may partially account for the reduced cooperativity and stability of the IL-15 quaternary complex (6,56). The precise details of the interactions at the IL-15⅐IL-2R␤ and IL-15⅐␥ c interfaces, and further insights into the molecular basis for the promiscuity of the IL-2R␤ and ␥ c await the determination of the IL-15 quaternary complex structure.  Fig. 6A. Note that the distinct conformation of the hA-hB loop of IL-15 places this region of the ligand into the vicinity of the hC-hCЈ1 loop of ␥ c where unique interactions can potentially take place. Also note the position and length of hA of IL-15 and IL-2 quaternary complex interface. IL-2 residues Leu-12, Glu-15, Leu-18, and Leu-19 from the N-terminal portion of hA participate in interactions at a three-way junction between IL-2⅐IL-2R␤ and ␥ c that is suggested to mediate cooperative binding. Due to the shorter hA of IL-15, it has no homologous residue to Leu-12 of IL-2. Furthermore, IL-2 residues Glu-15, Leu-18, and Leu-19 have diverged to Ile-3, Arg-6, and Tyr-7 in IL-15 making it unlikely that a homologous set of interactions takes place at the IL-15⅐IL-2R␤/␥ c junction.
Finally, there are data suggesting that human and mouse IL-15 exhibit some degree of species specificity in terms of signaling through IL-2R␤/␥ c (55). As mentioned above, comparison of our murine IL-15⅐IL-15R␣ structure to the recently reported human IL-15⅐IL-15R␣ structure (44) shows that the predicted IL-2R␤ and ␥ c binding sites of human and mouse IL-15 adopt a similar overall conformation. In contrast to mouse and human IL-15 residues involved in IL-15R␣ binding, which are highly conserved, sequence analysis shows that there are several prominent amino acid substitutions at the predicted IL-2R␤ and ␥ c interfaces of mouse and human IL-15 (Fig. 1B). At the predicted ␥ c binding site, mouse IL-15 residues Arg-7, Glu-11, and Arg-105 have diverged to Ile-7, Arg-11, and His-105 in human IL-15, and at the predicted IL-2R␤ binding site, mouse Arg-64 has diverged to Glu-64 in human IL-15 (Fig. 1B). As mentioned above, the orientation of the alpha receptor binding site (hB) relative to that of the predicted IL-2R␤/␥ c binding site (hA, hC, and hD) is also slightly different in human and murine IL-15 (supplemental Figs. S1 and S2). Taken together, these sequence and structural differences may play some role in the observed species specificity for the ability IL-15 to signal through IL-15R␤/␥ c (55).