|Home | About | Journals | Submit | Contact Us | Français|
Interleukin-4 and Interleukin-13 are cytokines critical to the development of T cell-mediated humoral immune responses, which are associated with allergy and asthma, and exert their actions through three different combinations of shared receptors. Here we present the crystal structures of the complete set of Type I (IL-4Rα/γc/IL-4) and Type II (IL-4Rα/IL-13Rα1/IL-4, IL-4Rα/IL-13Rα1/IL-13) ternary signaling complexes. The Type I complex reveals a structural basis for γc’s ability to recognize six different γc-cytokines. The two Type II complexes utilize an unusual top-mounted Ig-like domain on IL-13Rα1 for a novel mode of cytokine engagement that contributes to a reversal in the IL-4 versus IL-13 ternary complex assembly sequences, which are mediated through substantially different recognition chemistries. We also show that the Type II receptor heterodimer signals with different potencies in response to IL-4 versus IL-13, and suggest that the extracellular cytokine-receptor interactions are modulating intracellular membrane-proximal signaling events.
T cell-mediated immunity occurs in successive antigen-driven and cytokine-driven steps. In the first step, T cell receptors are activated by specific antigens presented by peptide-MHC complexes on the surface of antigen presenting cells. In a second step, these activated T cells are induced to clonally expand by the cytokine Interleukin-2 (IL-2), and subsequently differentiate into CD4+ Th1 or Th2 cells as a result of a specific subset of cytokines engaging and activating their respective cytokine receptors (Leonard, 1999). IL-4 is the primary cytokine implicated in the development of Th2-mediated responses (Seder and Paul, 1994), which are also associated with allergy and asthma (Barnes, 2002). The cytokines IL-4 and IL-13 are produced by Th2 cells, and recruit and activate IgE-producing B cells, and enhance IgE-mediated responses. The strong association of Th2-cytokines with allergic disease has generated interest in targeting their receptors for therapeutic intervention (Foster et al., 2002; Mueller et al., 2002).
IL-4 alpha receptor (IL-4Rα) is a unique member of the common-gamma chain (γc) family of receptors (Nelms et al., 1999), with the ability to signal within three different receptor complexes, known as Type I and Type II receptors (Andrews et al., 2006; Kelly-Welch et al., 2003; Mueller et al., 2002). On cells of hematopoietic stem cell in origin, the Type I receptor comprises IL-4Rα and the common gamma-chain (γc), which is also shared by the cytokines IL-2, -7, -9, -15 and -21 (Nelms et al., 1999). On cells of non-hematopoietic stem cell in origin, IL-4 can use the Type II complex, comprising IL-4Rα and IL-13Rα1 (Hilton et al., 1996; Obiri et al., 1995). This receptor complex is also a functional receptor for IL-13 (Aman et al., 1996), which shares only ~25% homology with IL-4, and this largely explains the overlap of the biological effects of IL-4 and IL-13 (Wills-Karp, 2004). Type I receptor complexes can only be formed by IL-4 and are more active in regulating Th2 development. In contrast, the Type II receptor complex formed by either IL-4 or IL-13 is not found on T cells and is more active in regulating cells that mediate airway hypersensitivity and mucus secretion (Andrews et al., 2006). In fact, IL-13 appears to have its own distinct role in allergic inflammation, acting as a key regulator of allergen-induced airway inflammation, and goblet cell metaplasia (Wills-Karp, 2004). The molecular basis for the cytokines differing functional responses is unclear. Both Type I and Type II receptor complexes signal through the Jak/STAT cascade, with IL-4Rα associating with Jak1, γc with Jak3, and IL-13Rα1 with Tyk2. A second mechanism of signal transduction activated by IL-4 and IL-13 leads to the insulin receptor substrate (IRS) family (Kelly-Welch et al., 2003).
IL-4 and IL-13 are prototypical four-helix bundle short chain cytokines (Moy et al., 2001; Mueller et al., 2002; Walter et al., 1992; Wlodawer et al., 1992). The IL-4/IL-13 receptors γc, IL-4Rα and IL-13Rα1 contain the tandem Fibronectin-III domains and the WSXWS box that form the classical elbow-shaped “Cytokine binding homology region (CHR)” which binds to the helical faces of the cytokine (Bazan, 1990), as originally defined in the human Growth Hormone (hGH) system (de Vos et al., 1992). A previous structure of a non-signaling binary complex between IL-4 and IL-4Rα indicated a similarity to the hGH-R site I paradigm (Hage et al., 1999). While the structure of a complete short-chain cytokine receptor heterodimeric signaling complex was reported for IL-2 (Wang et al., 2005), the Type I and Type II IL-4/13 receptors have several novelties that pose interesting structural questions (Mueller et al., 2002). First, in the Type II complex, the IL-13Rα1 contains an extra N-terminal Ig-like domain (D1) not found in other receptors of the γc subfamily, that is required for IL-13 signaling but not by IL-4 (Arima et al., 2005). It is unknown how this top-mounted Ig-like domain participates in cytokine recognition.
A second novelty pertains to cytokine receptor ‘sharing’ and the evolutionary relationships between IL-13Rα1, γc and IL-4Rα (Boulay et al., 2003; Leonard et al., 1994; Ozaki and Leonard, 2002). While γc is the primary short chain cytokine shared receptor, IL-4Rα is also shared in three signaling complexes. IL-13Rα1 bears an evolutionary relationship to γc (Boulay et al., 2003), but its ligand specificity is narrower and it diverged from a common ancestral family as γc with the acquisition of the N-terminal Ig-like domain. It will be interesting to determine if these receptors possess structural properties conducive to degenerate recognition and signaling in comparison to their highly specific alpha receptor counterparts. In addition, as only one γc-family complex has been solved (IL-2), the Type I complex could reveal a molecular basis for γc cross-reactivity.
Finally, IL-4 and IL-13 can induce distinct functional responses through the identical Type II receptor heterodimers (Kelly-Welch et al., 2003), raising the question of how ligand-specificity is conferred by shared receptors and can influence signaling. Thus, the IL-4/13 system presents an opportunity to investigate whether downstream signaling differences through common receptors have origins in the structural/biophysical properties of the extracellular receptor-cytokine complexes. In order to probe these questions, we present structural, biochemical, and signaling studies on the complete set of Type I and Type II ternary receptor complexes, IL-4Rα/γc/IL-4, IL-4Rα/IL-13Rα1/IL-4, IL-4Rα/IL-13Rα1/IL-13.
To obtain diffraction-quality crystals, we engineered glycan-minimized recombinant molecules (see Methods) that behaved similarly to wild-type glycosylated proteins, and determined the complex structures to resolutions between 2.9–3.0Å (Table S1, Figs. S1, S2). All three ternary complexes (Fig. 1) globally resemble the canonical cytokine-receptor complex architecture seen in several other cytokine receptor complexes (Stroud and Wells, 2004), albeit with substantial deviations apparent in the two Type II complexes (Figs. 1b, c, S4) (discussed below). In the three complex structures, we refer to the cytokine/IL-4Rα interfaces as site I, and the γc and IL-13Rα1 interfaces will be referred to as sites IIa, IIb and III (Fig. 1). The cytokines bridge the receptor heterodimers, resting in Y-shaped forks (sites I and IIa) that are buttressed underneath by extensive receptor-receptor contact forming the “stems” of the Y’s (site IIb) between the D2-D2 (Type I complex) and D2-D3 domains (Type II complexes). In each complex extensive amounts of surface are buried within the cytokine-receptor and receptor-receptor interfaces (Table S2, Fig. S4) (3980Å2 for IL-4 Type I, 4400Å2 for IL-4 Type II, and 4030Å2 for IL-13 Type II).
In both Type II complexes there is a novel structural feature that we term site III (Figs. 1b, c). The extra N-terminal IL-13Rα1 D1 ‘Ig-like’ domain contacts the dorsal surfaces of both IL-4 and IL-13. The structure of the IL-13Rα1 D1 domains is an “s-type” Ig-like fold (Fig. S3) (Bork et al., 1994), where the conventional d strand in a top sheet of an Ig-fold now becomes a c’ strand of the bottom sheet due to a switch in pairing (Fig. S3). This has important implications for the Type II complexes in that, by virtue of the strand swap that occurs in s-type Ig folds, the c’ strand of IL-13Rα1 is positioned to contact the C–D loop of the cytokine through a β-sheet like interaction.
IL-13Rα1 is derived from a common ancestral subgroup as γc, the divergence of IL-13Rα1 is linked to the acquisition of the Ig-like domain (Boulay et al., 2003). The γc -cytokines IL-2 and IL-15 require unusual ‘sushi’ domain alpha receptors for formation of the “high affinity” IL-2/IL-15-Rαβγc complexes (Fig. 2a) (Waldmann, 2006). The IL-2Rα and IL-15Rα bind to the dorsal surfaces of the cytokines (Fig. 2a, c) (Chirifu et al., 2007; Rickert et al., 2005), at locations overlapping with the IL-13Rα1 D1 binding sites on IL-4 and IL-13 (Fig. 2b, c). From the ancestral relationship of IL-13Rα1 and γc, it is tempting to speculate that the top-mounted D1 domain of IL-13Rα1 has structural and functional analogies to IL-2/IL-15Rα. Possibly the divergence of three-domain IL-13Rα1 from the two-domain γc subgroup is the result of a gene fusion event between a separate Ig-like domain receptor (analogous to IL-2/IL-15Rα) and the IL-13Rα1 D2D3 module (Fig. 2b), consolidating both receptors on a single polypeptide chain.
The three complexes form by the sequential, and cooperative assembly, of composite cytokine-receptor binding surfaces that exhibit striking differences in the extent and distribution of polar and apolar surfaces (Fig. S4, and Tables S2, contact Tables S4–S7). In order to determine whether this structural diversity is reflected in diverse binding chemistries, we investigated the complex assembly thermodynamics using Isothermal Titration Calorimetry (ITC) with soluble receptor extracellular domains (ECDs) (Fig. 3, Table S3). Consistent with previous studies (Andrews et al., 2002; Obiri et al., 1995; Russell et al., 1993), we find in the Type I complex, IL-4 first binds to IL-4Rα, followed by recruitment of γc to form a ternary complex (Fig. 3a). For the Type II IL-4 complex, IL-4 again first binds to IL-4Rα followed by recruitment of IL-13Rα1 (Fig. 3b). In contrast, for the Type II IL-13 complex, IL-13 first binds IL-13Rα1 with an affinity of Kd = 30nM (similar to previous BIAcore studies (Kd = 34nM, (Andrews et al., 2002)) and this binary complex then recruits IL-4Rα (Fig. 3c). Strikingly, in the Type I and II IL-4 complexes, the same IL-4/IL-4Rα binary complex engages γc and IL-13Rα1 through opposite thermodynamic profiles (Figs. 3a,b – far right, Table S3), in which the Type I complex shows a strongly favored enthalpy ΔH= −11.7 kcal/M, and a disfavored entropy ΔS = −10.5cal/molK−1, compared to the Type II complex showing a weakly favored enthalpy ΔH= −4.8 kcal/M and strongly favorable entropy ΔS = 13.0cal/molK−1. Further, even in a case where several key interfacial residues persist, such as the site I interface of IL-4Rα with IL-4 (ΔH= −11.2kcal/mol, ΔS= +3.5cal/molK−1) versus IL-13 (ΔH= −5.8kcal/mol, ΔS = +16.0cal/molK−1), distinct thermodynamic binding signatures are observed (Figs. 3a, b, c, Table S3). Although in our structures it is difficult to visualize trapped water molecules potentially impacting the entropy of the interactions, our results show that these shared receptors utilize diverse binding thermodynamics tailored to each ligand, in contrast to the basis of cross-reactivity in the gp130 system, which has converged on a desolvation-driven mechanism for all of its ligands (Boulanger et al., 2003).
With respect to formation of the ternary signaling complexes, the IL-13 Type II appears to be the most stable by virtue of a higher affinity of the IL-13/IL-13Rα1 binary complex for IL-4Rα (Kd = 20nM) (Fig. 3c – far right, Table S3). The IL-4 Type I and Type II complexes both have very low affinity interactions with γc (Kd = 559nM), similar to BIAcore measurements (Andrews et al., 2006; Zhang, 2002) and IL-13Rα1 (Kd = 487nM), respectively, resulting in lower overall complex stabilities (Fig. 3a,b – far right). Thus, the recruitment of the second ‘trigger’ receptors appears energetically limiting for the IL-4 complexes and could play a role in the differential signaling properties of the IL-4 versus IL-13 complexes (discussed later).
The interactions seen in the Type I and Type II IL-4/IL-4Rα interfaces are similar to the previously reported IL-4/IL-4Rα binary complex (Hage et al., 1999) (Fig. 4a), except that in the ternary complexes the IL-4Rα FN-III domains are rotated several degrees ‘counterclockwise’ around the IL-4 helical bundle, likely due to the interactions formed between the bases of the receptors (i.e. the stem of the Y). The principal feature of the large IL-4/IL-4Rα interface (Figs. 4a, S4) remains a charge complementarity between patches centered on “hotspot” residues Glu9 and Arg88 on the IL-4 A and C helices (Fig. 4a) (Zhang et al., 2002), respectively, which form polar contacts with Tyr13, Tyr183, Ser70, and Asp72 on IL-4Rα.
The site I interface of IL-13 with IL-4Rα (Fig. 4b) is presented on a more compact IL-13 helical scaffold (see Figs. 1b,c top view) (two helical turns shorter in both A and C helices than IL-4). Although the binding surface and amino acids used on IL-4Rα to bind IL-13 are nearly identical to that used to contact IL-4 (Fig. 4a), their four-helix bundles superimpose poorly when the complexes are aligned on IL-4Rα (Fig. 4c). Yet, IL-13 exactly recapitulates two critical charged residues by presenting Glu12 and Arg65 to IL-4Rα in the same positions to form similar receptor contacts as seen for IL-4 residues Glu9 and Arg88 (Figs. 4b–d), which have been shown by mutagenesis to be energetically important (Kraich et al., 2006). However, the composition of remaining IL-13 contact residues with IL-4Rα is different (Fig. 4c, d), in accord with the distinct binding thermodynamics the two cytokines use to bind IL-4Rα(Figs. 3b, c, Table S3). In the IL-13/IL-4Rα site I, the interface lacks interactions ringing the charged hotspots observed in the IL-4/IL-4Rα interface, such as the loss of IL-4 residue Trp91 on the C helix (Fig. 4c). The number of polar bonds per 100Å2 of surface buried in the IL-13/IL-4Rα interface is 1.0, denoting a substantial loss in specificity relative to the extraordinarily specific IL-4/IL-4Rα complex (2 polar bonds per 100Å2). Given that IL-4 binds to IL-4Rα with sub-nanomolar affinity, and IL-13 has no measurable affinity for IL-4Rα alone, the energetic value of the two ‘mimicked’ charged contacts on IL-13 (Kraich et al., 2006) is apparently manifested within the context of the ‘composite’ interface formed together with the receptor-receptor contacts (site IIb). In this fashion, the IL-4Rα/IL-13 site I works in concert with site IIb to engage IL-4Rα through tandem, multi-point attachment of two weak interfaces.
The formation of the ternary signaling complex is completed by the recruitment of a second receptor, either γc or IL-13Rα1 through structurally separated binding contacts designated sites IIa (fork of the Y) (Table S5), IIb (stem of the Y) (Table S6) and, in the cases of the Type II complexes, site III (Table S5).
The three site IIa interfaces are uniformly characterized by the convex elbow of the receptors occupying a canyon on the A and D helices of the cytokines (Figs. 5a–c), exhibiting excellent knob-in-holes shape complementarity (Table S2)(Sc = 0.82, 0.73 and 0.65 for the IL-4 Type I, and IL-4 and IL-13 Type II complexes, respectively). Several hydrogen bonds flank the sides of the apolar canyons in each complex, and likely stabilize the respective docking orientations. Consistent with their ancestral relationship, IL-13Rα1 and γc present binding sites with convergent structural features (Figs. 5a, b). The site IIa contacts in each of the receptor complexes are primarily formed by the FG2 loop of the membrane proximal FN-III domains of γc and IL-13Rα1 (D2 and D3, respectively), but γc also uses the EF1 loop in its D1. The receptor binding surfaces present a somewhat chemically inert main chain surface that inserts into a cleft in the bases of cytokines. Both γc and IL-13Rα1 insert an unusual surface-exposed disulfide bond linking the FG2 and BC2 loops, together with a neighboring side chain (Leu208 for γc, Leu319 for IL-13Rα1) into the cavity on the cytokine surface (Figs. 5a–c). One substantial deviation between γc and IL-13Rα1 is that γc also forms extensive van der Waals contacts with IL-4 through Tyr103 on the EF1 loop of the D1 domain, resulting in a two-point site IIa attachment to IL-4 (Fig. 5a). This EF-loop interaction is missing in both Type II complexes (Figs. 5b, c), consistent with mutational data on the IL-13Rα1 EF loop showing minor effects on binding (Arima et al., 2005).
A stripe of amino acids on the IL-4 and IL-13 A and D helices demarcates the hydrophobic canyon lined by the alkyl moieties of both polar and non-polar amino acids (Figs. 5d, e). These side chains part to both sides to form clefts into which the narrow receptors insert to contact the main chains of the cytokine helices, facilitating degenerate recognition. (Figs. 5b,c). The canyon on IL-4 engages non-identical amino acids on γc versus IL-13Rα1 in structurally similar fashions (Figs. 3a, b). For example, IL-4 residues Gln8, Ile11, Asn15 interact with γc Pro207 and Leu208, which appear to serve as structural equivalents Lys318 and Leu319 of IL-13Rα1 (Figs. 5a,b). Although the binding canyons on IL-4 and IL-13 utilize different amino acids to engage their receptors, they retain similar surface and shape properties, suggesting a manner of convergent structural evolution to enable cross-reactivity.
The most unique structural feature in the Type II complexes is site III (Figs. 5f, g). The principal structural characteristic of site III is an anti-parallel beta sheet between the c’ strand of IL-13Rα1 and the C–D strand of IL-4 and IL-13. Similar sheet-like contacts form between IL-4 strand residues Asn105 to Thr108, or IL-13 residues Thr88 to Glu91, to the c’ strand residues of IL-13Rα1 Lys76 to Ile78. However, there is an additional source of contact in the IL-13 site III that may explain the requirement of the IL-13Rα1 D1 for signaling in the IL-13 Type II complex, in contrast to IL-4 (Arima et al., 2005). We also found in mutational studies that after deletion of IL-13Rα1 D1, the D2D3 CHR module alone does not detectably bind to IL-13, whereas IL-13Rα1 D2D3 does form a ternary complex with IL-4/IL-4Rα(Fig. S5). In the IL-13 complex site III (Fig. 5g), a prominent saucer-shaped, hydrophobic patch on the IL-13 surface is formed by the alkyl sides chain of Met 33, Asp87, Lys89, and Trp35, creating a surface apposing an hydrophobic patch on the tip of IL-13Rα1 D1 composed of Trp65 and Ile78 on the c’ strand (Table S2, Fig. S4). By contrast, the analogous region of IL-4 lacking the complementary hydrophobic patch (Figs. 5f, S4), is apparently relatively devoid of energetic value. However, as for the site IIa energetics, the energetic basis of the respective IL-13Rα1 D1-dependencies for IL-4 and IL-13, is not explained by the structure of site III alone, but likely highly energetically coupled to sites IIa (for IL-13) and sites IIa/IIb (for IL-4).
In all three complexes, extensive contacts (~1200–1300Å2 BSA) are formed between the membrane-proximal domains of the receptors (Fig. S4 & Table S2). Although the sequences of γc and IL-13Rα1 are highly divergent (26% identity), the site IIb interactions appear to utilize a limited degree of molecular mimicry to engage the shared IL-4Rα in that several pair-wise amino acid contacts are replicated. For example, both γc and IL-13Rα1 use structurally equivalent Phe186 and Phe297, and Pro189 and Pro300, respectively, to interact with a hydrophobic stretch of IL-4Rα d strand residues Tyr150 through Tyr159 (Table S6). The receptor-receptor interfaces have very poor shape complementarity values of 0.49, 0.52 and 0.57 for the three complexes. By comparison, the excellent shape complementarity of the site IIa interfaces would appear to suggest that the cytokine-receptor contact is the primary driving force for site II formation. Collectively, the three weak interactions alone (sites IIa, IIb and III), cooperatively form a tripartite binding site where productive affinity is achieved through multi-point attachment.
γc is a signaling receptor for at least six different short-chain cytokines, including IL-2, -4, -7, -9, -15 and -21 (Ozaki and Leonard, 2002). Previously we determined the structure of the quaternary complex of IL-2 with IL-2Rα, IL-2Rβ, and γc (Wang et al., 2005). With a second γc ternary complex, we analyzed how γc can engage six cytokines with less than 15% sequence identity (Fig. 6). The IL-2 and IL-4 interfaces with γc are strikingly similar, both burying ~1000 Å2, having similar ratios of polar/apolar surface, and low specificity indices (0.7–1.0 polar bonds/100Å2). The γc binding sites on IL-2 and IL-4 exhibit apolar ‘canyons’ that receive the protruding elbow of γc through near-ideal shape complementarity (Fig. 6b) (Sc for IL-4 site IIa is 0.82, for IL-2 is 0.82). γc residues Tyr103 and the disulfide Cys160-Cys209 engage the IL-4 canyon in a coupled manner as seen in the IL-2-γc interface (Fig. 6b). Cytokine contact residues on γc loops BC2 and FG2 maintain similar conformations in both IL-2 and IL-4 complexes, suggesting that cross-reactivity is not occurring through structural plasticity of the receptor, but rather a rigid γc surface encounters extraordinarily complementary shape features on γc-cytokines, which are further enhanced by specific hydrogen bonds peripheral to the canyon.
When IL-2 and IL-4 carbon-α (abbrev. Cα) positions are compared in the complexes (Fig. 6c), five residues each on helices A and D overlap with a very close r.m.s. deviation. of 0.2 Å and 0.3 Å, respectively, as compared with the overall IL-2/IL-4 r.m.s.d. of 1.7 Å. The structurally equivalent side chains of IL-2 residues Glu15, Leu19, Gln126, Ser130 and IL-4 residues Gln8, Lys12, Arg121, Ser125 collectively line the complementary canyon, making analogous interactions with γc (Fig. 6c). Several key hydrophobic residues within the helical core, which dictate the resultant positions of “external” interacting side chains on the helices, are conserved (Fig. 6d). We propose these cytokine residue positions constitute a γc “recognition motif” that likely demarcates residues serving similar structural roles in other γc-cytokine complexes.
Given the shared use of receptors in the IL-4/13 system, we asked what the relative potencies of receptor activation were for IL-4 versus IL-13 through the Type I and II receptors on cell lines (Fig. 7). Since engagement of the Type I and Type II receptor complexes leads to the activation of STAT6 (Kelly-Welch et al., 2003), we compared the IL-4-and IL-13-induced phosphorylation of STAT6 as a measure of receptor complex activation (Fig. 7). In addition, we analyzed the phosphorylation of STAT3 (Fig. S6). We analyzed signaling by the Type I IL-4 receptor in the human B-cell line Ramos, which express γc and IL-4Rα. We analyzed signaling by the Type II receptors in the human epithelial carcinoma cell line A549, which express IL-4Rα and IL-13Rα1, but do not express the γc or the decoy receptor IL-13Rα2 (Fig. 7a). Treatment of A549 cells with either IL-4 or IL-13 for 30 minutes stimulated the potent tyrosine phosphorylation of STAT6 and the relatively weak phosphorylation of STAT3 (Figs. 7b, S6). IL-4 was able to stimulate phosphorylation at significantly lower doses (five to ten-fold) than IL-13. Treatment of Ramos cells with IL-4 stimulated the tyrosine phosphorylation of STAT6 with a dose-response similar to that observed on A549. We also examined the kinetics of STAT6 and STAT3 phosphorylation at several cytokine concentrations (Figs. 7c, S6, S7). At all concentrations tested (1, 5, 50 ng/ml), IL-4 rapidly induced the tyrosine phosphorylation of STAT6 in A549 reaching plateau levels between 10–15 minutes of stimulation. However, the response to IL-13 was substantially slower. Treatment of A549 cells with 1 ng/ml, 5 ng/ml or 50 ng/ml stimulated the tyrosine phosphorylation of STAT6 to plateau levels at >60, 30, and 15 minutes respectively. The relative delay in STAT6 phosphorylation induced by IL-13 via the Type II receptor complex was most apparent at lower concentrations of cytokine (1 ng/ml). (Figs. S6, S7). The stimulation of STAT3 phosphorylation by either IL-4 or IL-13 was quite weak in these cells, however, the pattern of responsiveness was similar to what we observed for STAT6 (Fig. S6). Treatment of Ramos cells with IL-4 stimulated the tyrosine phosphorylation of STAT6 with kinetics similar to that observed on A549; IL-4 achieved maximal phosphorylation between 10–15 minutes of stimulation (Figs. 7c, S7). We did not observe the induction of STAT3 phosphorylation by IL-4 in these cells (Fig. S6). The implications of these signaling data in the context of the structural and thermodynamic data are discussed further below.
The results presented here on the IL-4/13 system bear on several coupled issues in ligand-receptor interactions, namely: 1) the structural basis of degeneracy versus specificity of shared receptors, 2) the assembly energetics and sequences of hetero-dimeric receptor complexes, and 3) whether differences in these extracellular parameters could qualitatively influence the coupling of ligand engagement to intracellular receptor activation. Surprisingly, we found that that the identical Type II receptor heterodimer, coupled to the same intracellular signaling molecules, responds to different ligands, IL-4 versus IL-13, with different signaling potencies and kinetics (Figs. 7b, c, S6, S7). One can consider the initial cytokine-receptor interaction as the ‘driver,’ and the second receptor recruited for ternary complex signaling as the ‘trigger’, akin to terminology previously proposed for the assembly of γc heterodimeric complexes (Lai et al., 1996). The IL-4/13 system has the unique property that IL-4 and IL-13 have swapped their trigger and driver. The evolutionary expansion of receptor usage by switching 'driver' and 'trigger' modules has also occurred in other receptors in the family, such as IL-7Rα/γc (Noguchi et al., 1993) and IL-7Rα/TSLP-R (Pandey et al., 2000). In the IL-4/13 system this switch is mainly due to the acquisition of the energetically critical site III in IL-13, and the ‘hobbling’ of its site I. From previous studies we know that the equilibrium binding affinities of IL-4 and IL-13 for their heterodimeric receptor complexes on cell surfaces are similar (Kd ranging from 30–400 pM) (Aman et al., 1996; Andrews et al., 2002; Hilton et al., 1996; Park et al., 1987), so the explanation for the potency differences is not simply receptor occupancy. The fact that IL-4 binds to the IL-4Rα with very high affinity (Kd~ sub-nanonomolar) (Andrews et al., 2006; Park et al., 1987), and the presence of γc or IL-13Rα1 provides little additional affinity (Murata et al., 1998; Russell et al., 1993; Zhang, 2002) suggests that driver association is dictating IL-4 Type II signaling potency. Yet, we find that the high affinity driver complex (IL-4/IL-4Rα) recruits the IL-13Rα1 ‘trigger’ receptor with very low affinity (Kd ~ 487nM), which would be energetically rate-limiting for signaling. In contrast, IL-13 initially binds to its driver receptor (IL-13Rα1) on the surface of cells with a relatively slow on-rate, but still with a moderately strong binding constant (Kd ~30nM) (Aman et al., 1996; Hilton et al., 1996). However, the presence of its trigger IL-4Rα increases the cell surface affinity for IL-13 from Kd ~ 10 nM to 30 pM (Miloux et al., 1997). Thus, the recruitment of the trigger receptor (IL-4Rα) by the IL-13/IL-13Rα1 driver complex is more energetically favorable (Kd ~20nM) than is the converse trigger receptor (IL-13Rα1) recruitment by the IL-4/IL-4Rα driver complex (Kd ~487nM). Yet, paradoxically, we find that IL-4 is more potent at stimulating the tyrosine phosphorylation of STAT6 in several cell lines.
From the aforementioned energetic results we would predict that if the trigger receptor becomes limiting, the sensitivity and kinetics of STAT6 phosphorylation may change. Based on estimates in cell lines, the numbers of IL-4Rα chains (50–5000 molecules/cell) appears to be limiting as compared to the numbers of IL-13Rα1 chains (5000–150,000 molecules/cell) (Park et al., 1987, Obiri et al., 1995). In the two cell lines we analyzed herein, expression of the IL-4 trigger receptor (γc or IL-13Rα1) was greater than IL-4Rα (Fig. 7a). However, in cells where IL-13Rα1 is limiting, IL-13 can become more potent than IL-4 (William Paul, Ilkka S. Juntilla, Kiyoshi Mizukami, Harold Dickenstein, Martin Meier-Schellersheim, Raymond P. Donnely (NIH), personal communication). Thus, it appears that when the IL-4 trigger receptor (IL-13Rα1) is abundant, the high affinity IL-4/IL-4Rα driver complex determines the ultimate signaling potency. By comparison, the affinity of IL-13 for its driver IL-13Rα1 (Kd ~30nM), results in efficient formation of the IL-13/IL-13Rα1 driver complex even when IL-13Rα1 is limiting, and the substantially higher affinity of the IL-13/IL-13Rα1 driver complex for its trigger receptor (IL-4Rα) (Kd ~ 20nM) could result in more potent IL-13 signaling versus IL-4. Thus, the potency of signal transduction can be influenced by a variety of physical-chemical properties that are communicated through the cytokine receptor ECD interactions, including the concentration of cytokine and expression levels of receptor, the order of assembly of the signaling complexes, and the affinities of ‘trigger’ and ‘driver’ receptor recruitment. In vivo, the potential for higher order assemblies of receptors on the cell surface, soluble shed receptors in the extracellular millieu, and the expression of the non-signaling IL-13Rα2 serving as a decoy, could play unknown roles in IL-4/13 signaling regulation.
Although highly speculative, another possible contributing factor to the enhanced potency of IL-4 may lie in the relative orientations of the receptor heterodimers in the two complexes. It has been shown for gp130 and the Erythropoietin receptor (EPO-R) that perturbations in the rotational “pitch” of the membrane-proximal intracellular regions, which are proximal to the Jak binding sites, can modulate signaling potency (Constantinescu et al., 2001; Greiser et al., 2002). Additionally, one study examined structures of complexes of EPO-R ectodomains with synthetic peptide agonist and antagonist mimics of EPO (Livnah et al., 1998), and found the antagonist complex was rotated approximately 14 degrees away from the two-fold axis compared to the agonist complex (Fig. S8b). We find that, when the IL-4 and IL-13 Type II complexes are superimposed on either IL-4Rα or IL-13Rα1, the relative orientation of the membrane-distal domains of the ‘partner’ receptor differs by ~7–8 degrees (Fig. S8a). The origin of this difference appears to lie mainly in the IL-4 versus IL-13 site I interfaces, which have diverged more in their relative cytokine/receptor positioning than the membrane-proximal regions. This perturbation in orientation could (Fig. S8), in principle, exert a modulatory role on signaling in a fashion analogous to the EPO peptide mimetics, although this remains speculative.
Collectively, our structural, biochemical and signaling results invite revision of the concept that extracellular ligand engagement of cytokine receptors does not relay ‘instructive’ information intracellularly. Experiments in which the ECD of EPO-R was replaced with the Prolactin ECD, resulting in red blood cell formation, concluded that structurally indiscriminate dimerization was adequate to initiate cytokine receptor signaling (i.e ‘permissive signaling’) (Socolovsky et al., 1998). While we concur that downstream signaling specificity is manifested intracellularly, our results appear to suggest that the potency of receptor signaling can be influenced by the biophysical and structural properties of the extracellular receptor-ligand interactions, which may, in this case, be utilized as a means for different ligands to induce divergent signaling responses through shared receptors in different cellular contexts.
Human IL-4, IL-13, and the ectodomains of γc, IL-4Rα, and IL-13Rα1 cDNA were sub-cloned into pAcGP67-A with C-terminal hexahistidine tags (BD Biosciences, San Diego, CA), and secreted from High-Five™ insect cells in Insect-Xpress™ medium (Lonza, Walkersville, MD). In order to reduce protein heterogeneity due to Asn-linked glycosylation, we expressed the molecules in various glycan-minimized forms. IL-4 was expressed in the presence of tunicamycin (2.5 mg/L) (Calbiochem, San Diego, CA) to suppress Asn-linked glycosylation. For IL-4Rα we mutated Asn➔Gln in four, out of a possible six, N-linked glycosylation sites (N28Q, N73Q, N109Q, N151Q). IL-13 and IL-13Rα1 were co-expressed in the presence of tunicamycin (0.65 mg/L). For the Type I complex, each component was separately purified by gel filtration FPLC Superdex 200 (GE Heathcare, Piscataway, NJ) after which a 1:1:1 molar ratio of IL-4, IL4Rα and γc was mixed, digested with carboxypeptidase A (EMD Biosciences, San Diego, CA) to remove C-terminal His-tags, and then re-purified by gel filtration. The IL-4 Type II complex was produced in a similar fashion for crystallization trials, using IL-13Rα1 expressed in the presence of tunicamycin. For the IL-13 Type II complex, IL-13 and IL-13Rα1 co-expressed in the presence of tunicamycin was digested with carboxypeptidase A and purified as a binary complex by gel filtration. The IL-4Rα tetra-glycan mutant was separately expressed and purified, and added to the binary complex in a 1:1:1 ratio, and purified by MonoQ FPLC. Ternary complex molecular masses were confirmed by MALS analysis.
Calorimetric titrations were carried out on a VP-ITC calorimeter (MicroCal, Northhampton, MA) at 27 °C. Gel filtration purified samples were dialyzed into sterile filtered 10 mM Hepes, pH 7.2, 150mM NaCl. Data were processed with OriginLab software. Protein concentrations were determined using the BCA method. The n, or stoichiometry values determined by ITC vary by ~25% (e.g. n values between 0.7 and 1.3 for an n value of 1.0) and are consistent with gel filtration molecular weight estimation. In the titrations of binary complexes, the titrand was used in the cell at a concentration of 5 µM (IL-13Rα) or 8 µM (IL-4Rα), and the titrant in the syringe at a concentration 81 µM (IL-13) or 65 µM (IL-4). In the titration of the ternary complexes two approaches were used in order to compare similar complexes. Preformed and purified 1:1 IL-4/IL-4Rα binary complex (55–57 µM) was titrated into either γc (8 µM) or IL-13Rα1 (3 µM). Conversely, preformed and purified 1:1 IL-13/IL-13Rα1 binary complex (3 µM) was the titrand in the cell while IL-4Rα (22 µM) was the titrant in the syringe.
The IL-4Rα/γc/IL-4 ternary (8–12 mg/ml) was screened for crystallization using sparse matrix crystallization reagent Wizard III (deCODE Biostructures, Bainbridge Island, WA). Crystals grew in 8–12% PEG 8000, 0.1 M HEPES pH 7.5, and 8% ethylene glycol. Data collection was performed at the beamline 11-1, Stanford Synchrotron Radiation Laboratory (SSRL). The crystals diffracted to 2.95 Å resolution in space group P212121, with unit cell dimensions of a = 52.58 Å, b = 86.65 Å, c = 175.67 Å. The data set was indexed, integrated, and scaled with HKL2000 (Otwinowski and Minor, 1997).
The IL-4Rα/IL-13Rα1/IL-4 ternary complex grew in conditions containing 5–10% PEG 8000, 0.1 M cacodylate pH 6.5, 0.16 M calcium acetate, and 20% glycerol. Data collection was performed at beamline 8.2.2, Advanced Light Source (ALS), University of California, Berkeley. The crystals diffracted to 3.0 Å resolution in the space group P21, with unit cell dimensions of a = 61.64 Å, b = 62.84 Å, c = 115.13 Å, and β = 96.31°. The data set was indexed, integrated, and scaled with HKL2000 (Otwinowski and Minor, 1997).
For the IL-4Rα/IL-13Rα1/IL-13 ternary complex two crystal forms were obtained, using either the fully glycosylated IL-4Rα(form I), or the quadruple glycosylation knock-out version as described above (form II). The form I crystals grew in 10–15 % PEG 8000, 8% ethylene glycol, 0.1 M ammonium citrate pH 6.0, and were cryopreserved in 23% ethylene glycol. Data collection was performed at ALS beamline 8.2.1. The crystals diffracted to 3.5 Å resolution with anisotropic diffraction extending to 3.2 Å, and belong to space group P212121, with unit cell dimensions of a = 91.98 Å, b = 125.10 Å, c = 148.27 Å, with two copies of the complex per asymmetric unit. Data was indexed, integrated, and scaled with HKL2000 (Otwinowski and Minor, 1997).
The higher resolution, glycan-minimized form II crystals were obtained in 12% PEG 6000, 0.2 M ammonium citrate, and 20% glycerol, and cryo-protected with 15% ethylene glycol. These crystals diffracted at SSRL BL 11-1 to 3.0 Å with anisotropic diffraction extending to 2.8 Å. They belong to the space group C2, with unit cell dimensions of a = 211.51 Å, b = 58.17 Å, c = 64.24 Å, with one copy of the complex per asymmetric unit. The data set was indexed, integrated and scaled with Mosflm and SCALA as implemented in CCP4 (CCP4, 1994).
The IL-4Rα/γc/IL-4 ternary complex structure was determined by molecular replacement (MR) with the program Phaser (McCoy, 2007) from the CCP4 program suite (CCP4, 1994), using the binary complex of IL-4/IL-4Rα (pdb access code 1IAR) and the γc chain from the IL-2 quaternary complex (pdb access code 2B5I) as search models. Rigid-body refinement on the molecular replacement solution, followed by simulated annealing and individual B-factor refinement were performed using CNS. After model rebuilding with COOT based on composite omit maps, and further refinement with CNS and Refmac5 (Brunger et al., 1998; Emsley and Cowtan, 2004; Murshudov et al., 1997), the final Rcryst and Rfree were 22.6% and 29.7%, respectively. For this as well as the two structures described below, Procheck (Laskowksi, 1993) indicates that Ramachandran statistics and the geometry of the protein models are compatible with structures at similar resolutions (Table 1).
For the IL-4Rα/IL-13Rα1/IL-13 ternary complex, MR solutions were obtained using the diffraction data of form I crystals (with fully glycosylated IL-4R), with IL-4R as the search model (pdb access code 1IAR). These solutions enabled the establishment of a non-crystallographic symmetry (NCS) operator relating molecules. Phaser also identified one MR solution for IL-13, by searching with the averaged IL-13 NMR model (pdb access code 1IJZ), and the second copy of IL-13 was properly placed into the crystal lattice using the NCS operator. Rigid-body refinement was performed on the MR solutions of the IL-4Rα/IL-13 binary complex. Subsequently, the D2D3 portion of IL-13Rα1 was obtained by Molrep (Vagin and Teplyakov, 2000) using a “culled” receptor model based on the D2D3 domains of gp130 (pdb access code 1BQU). The second copy of the IL-13Rα1 D2D3 domains was then placed using the NCS operator, and the partial model was rigid-body refined using CNS. The diffraction data of higher resolution from the form II crystals of the IL-4Rα/IL-13Rα1/IL-13 ternary complex was used for model rebuilding and refinement. After molecular replacement with Phaser to place the partially refined ternary complex model into the current crystal lattice, several rounds of multi-crystal averaging improved the density maps, allowing the rebuilding of the D2D3 domains of IL-13Rα1. The remaining D1 domain of 13Rα1 was placed into the complex with Phaser, using a fibronectin type III domain structure (pdb access code 1FNA) as search model. Following rigid body refinement, reiterations of model rebuilding with COOT and structure refinement using CNS and Refmac5, the final Rcryst and Rfree for this ternary complex were 25.6% and 31.2%, respectively (Table 1).
For the remaining IL-4Rα/IL-13Rα1/IL-4 ternary complex, the MR search was performed with Phaser, using models from the other two complexes. Following rigid body refinement, model rebuilding using COOT and structure refinement with CNS and Refmac5, the Rcryst and Rfree were 23.8% and 30.4%, respectively (Table 1).
Expression of receptor subunits was analyzed by FACS. PE-conjugated antibodies to human IL-4Rα, γC, and IL-13Rα1 were from BD Pharmingen (San Diego, CA); PE-conjugated anti- IL-13Rα2 was from Cell Sciences (Canton, MA). The signaling experiments were performed as described previously (Zamorano et al., 1996). Briefly, A549 or Ramos cells were washed 3 times and treated with various concentrations of IL-4 or IL-13 (R & D Systems Minneapolis, MN) for various times. Cells were harvested and lysed. Equal amounts of protein (~1 mg) were immunoprecipitated with rabbit anti-STAT6 (S-20, Santa Cruz Biotechnology, CA). The samples were separated on 7.5% SDS-PAGE, and transfered to polyvinylidene difluoride (PVDF) membrane. The membranes were probed with an anti-phosphotyrosine antibody, RC20-H (Transduction Labs, Lexington, KY). The bound antibody was detected using enhanced chemiluminescence. Where indicated the blots were stripped and probed with anti-STAT6. In some cases, total cell lysate (30 µg/lane) was analyzed using anti-phospho-specific STAT6 (Y641) or STAT3 (Y705) antibodies (Cell Signaling Technology, Danvers, MA). The films were scanned and analyzed using NIH Image. The density of signal from the phospho-STAT6 was normalized to the density of total STAT6, and expressed as phosphorylated STAT6/total STAT6 ratios. The calculated percent maximum from 3 independent experiments was averaged and graphed +/− SEM. The student's T-test was used to determine the significance of differences between groups.
The authors gratefully acknowledge Jean-Louis Boulay, William Paul, Ilkka S. Juntilla, Kiyoshi Mizukami, Harold Dickenstein, Martin Meier-Schellersheim, Raymond P. Donnely (NIH) for discussion and sharing of unpublished data. This work was supported by the Sandler Program for Asthma Research (KCG, SL), the American Cancer Society (SL), National Science Foundation (LC), Howard Hughes Medical Institute (KCG) and NIH (AI51321) (KCG), AI38985 (ADK) and T32 HL007698 (NMH).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.