Two Modes of β-Receptor Recognition Are Mediated by Distinct Epitopes on Mouse and Human Interleukin-3*

The cytokine interleukin-3 (IL-3) is a critical regulator of inflammation and immune responses in mammals. IL-3 exerts its effects on target cells via receptors comprising an IL-3-specific α-subunit and common β-subunit (βc; shared with IL-5 and granulocyte-macrophage colony-stimulating factor) or a β-subunit that specifically binds IL-3 (βIL-3; present in mice but not humans). We recently identified two splice variants of the α-subunit of the IL-3 receptor (IL-3Rα) that are relevant to hematopoietic progenitor cell differentiation or proliferation: the full length (“SP1” isoform) and a novel isoform (denoted “SP2”) lacking the N-terminal Ig-like domain. Although our studies demonstrated that each mouse IL-3 (mIL-3) Rα isoform can direct mIL-3 binding to two distinct sites on the βIL-3 subunit, it has remained unclear which residues in mIL-3 itself are critical to the two modes of βIL-3 recognition and whether the human IL-3Rα SP1 and SP2 orthologs similarly instruct human IL-3 binding to two distinct sites on the human βc subunit. Herein, we describe the identification of residues clustering around the highly conserved A-helix residue, Glu23, in the mIL-3 A- and C-helices as critical for receptor binding and growth stimulation via the βIL-3 and mIL-3Rα SP2 subunits, whereas an overlapping cluster was required for binding and activation of βIL-3 in the presence of mIL-3Rα SP1. Similarly, our studies of human IL-3 indicate that two different modes of βc binding are utilized in the presence of the hIL-3Rα SP1 or SP2 isoforms, suggesting a possible conserved mechanism by which the relative orientations of receptor subunits are modulated to achieve distinct signaling outcomes.

Interleukin-3 (IL-3) 3 is a cytokine produced principally by activated T-cells during immune responses that is known to be a key regulator of inducible hematopoiesis and inflammation. Importantly, IL-3 serves a vital role in stimulating basophil and mast cell responses to parasite infections (1). Because basophils are now known to play a critical role in Th2 immune responses (2)(3)(4), IL-3 is a critical regulator of allergic inflammation. In addition, IL-3 is expressed in the major embryonic vessels and serves an essential function in regulating the survival and proliferation of hematopoietic stem cells in the early stages of embryonic development (5). Despite these critical biological functions, our understanding of the mechanism by which IL-3 can stimulate its cell surface receptor remains limited.
IL-3 acts upon target cells via cell surface receptors composed of an IL-3-specific ␣-subunit and a common ␤-subunit (h␤c) that is shared with the related cytokines granulocytemacrophage colony-stimulating factor (GM-CSF) and interleukin-5 (IL-5). Although ␣-subunits bind their cognate ligands with low affinity (K d , 1-100 nM) (6 -8), formation of a high affinity signaling complex with the ␤-subunit (K d , 50 -500 pM) (9,10) is essential for the activation of Janus kinase 2 (Jak2), which is constitutively associated with the intracellular portions of the receptor subunits, and induction of several signaling pathways (11,12). In mice, but not humans, an additional ␤-subunit, known as ␤ IL-3 , is also utilized that specifically binds mouse IL-3 (mIL-3), but not GM-CSF or IL-5 (13). This receptor provides a useful model for studying the molecular details of cytokine recognition of the ␤-subunits, because mIL-3 binds ␤ IL-3 directly with an affinity of ϳ10 -20 nM in the absence of the mIL-3R␣ subunit, even though the latter is absolutely required for the formation of a signaling complex (14,15).
Recently, we reported that both the human IL-3 (hIL-3) and mIL-3R␣ subunits exist as two naturally occurring alternative splice variants: the archetypal full-length receptor with an ectodomain comprising an N-terminal Ig-like domain and a cytokine receptor homology module, designated the "SP1" variant, and a novel form lacking the N-terminal Ig-like domain designated "SP2" (16). These isoforms add additional complexity to the receptor system, because the two IL-3R␣ isoforms have the capacity to elicit different signaling outcomes (16). In the case of mIL-3, we established that mIL-3R␣ SP2 directs mIL-3 binding to the ␤ IL-3 ligand-binding interface formed at the "elbow" between domains 1 and 4 (16), analogous to the ligand-binding interface of human ␤c (h␤c) (15,(17)(18)(19), whereas the mIL-3R␣ SP1 isoform directs mIL-3 binding to a distinct epitope on ␤ IL-3 (15), which remains to be identified.
We sought to characterize the molecular mechanisms of receptor recognition and activation by studying the ligand, mIL-3. In the absence of structures of mIL-3 and its receptors, we prepared a structural model of the hIL-3⅐h␤c interaction based on the recently solved hGM-CSF receptor complex crystal structure (20) and identified putative h␤c-interacting residues within hIL-3. From alignment of the hIL-3 and mIL-3 sequences, we identified the homologous, candidate receptorbinding residues in mIL-3 and prepared a panel of mIL-3 mutants in which the highly conserved residue, Glu 23 , and neighboring residues in the A-and C-helices were individually mutated to alanine. The A-helix glutamic acid residue is very highly conserved throughout IL-3 orthologs (21), and the homologous A-helix glutamic acid residues in the related cytokines, hGM-CSF, hIL-5, and hIL-3, are known to be to central to h␤c recognition (22)(23)(24)(25). Importantly, we found that the mIL-3 Glu 23 and the surrounding cluster comprising Lys 27 (A-helix), Glu 65 (adjacent to C-helix), Val 69 , Asn 73 , and Lys 76 (all C-helix) are critical for ␤ IL-3 binding and receptor activation in the presence of the mIL-3R␣ SP2 isoform, whereas an overlapping epitope involving Lys 27 and Glu 65 is required for binding and activation of the mIL-3R␣ SP1/␤ IL-3 receptor. Our studies implicate these distinct epitopes on mIL-3 as important contributors to the two different modes of ␤ IL-3 binding utilized in the presence of the SP1 or SP2 mIL-3R␣ isoforms. Our subsequent studies implicated the homologous A-helix glutamate, Glu 22 , within the human ortholog, hIL-3, as playing a critical role in h␤c binding in the presence of the hIL-3R␣ SP2 isoform, but with a less critical role in the presence of the SP1 isoform. Consistent with a role for the hIL-3R␣ SP2 isoform directing hIL-3 interaction with the established domain 1-domain 4 interface in h␤c, mutation of domain 4 BЈ-CЈ loop residues similarly disrupted high affinity ligand binding in the presence of hIL-3R␣ SP2, but not SP1. Overall, these studies indicate that two distinct IL-3-binding sites exist on both ␤ IL-3 and h␤c, and their utilization within the respective signaling complexes is dictated by the two IL-3R␣ isoforms, SP1 and SP2, and involves distinct ␤-receptor-binding epitopes within IL-3. Our findings suggest a molecular mechanism by which isoform-specific ectodomain engagements could influence the relative receptor orientations, leading to initiation of distinct intracellular signaling programs.

EXPERIMENTAL PROCEDURES
Cell Lines-The generation of CTLL2 cell lines stably expressing wild-type or mutant mIL-3R␣ SP1, mIL-3R␣ SP2, hIL-3R␣ SP1, or hIL-3R␣ SP2 and wild-type or mutant h␤c or ␤ IL-3 subunits was described previously (16). These cell lines were maintained in RPMI 1640 containing 10% (v/v) fetal bovine serum (FBS), 50 M 2-mercaptoethanol, and murine IL-2 at 10 units/ml in the presence of 2 g/ml puromycin and 0.4 mg/ml G418 to select for cDNAs introduced via the pEFIRES-P and pEFIRES-N vectors, respectively, as before (16). COS7 cells were maintained in Dulbecco's modified Eagle's medium containing 10% FBS. Receptor expression was achieved by electroporation of cells with 10 g of wild-type hIL-3R␣ SP1 or SP2 and 25 g of wild-type or mutant h␤c DNA constructs at 200 V and 960 microfarads, after the cells were harvested by trypsinization, as described previously (15,18). COS7 binding studies were performed on cells 64 -68 h after transfection. The growth medium for the CTLL2 and COS7 cell lines was supplemented with 60 g/ml benzyl penicillin, 100 g/ml streptomycin, and 10 g/ml gentamicin.
Expression and Purification of mIL-3 and hIL-3-Wild-type and mutant mIL-3 and hIL-3 were expressed using the baculovirus/insect cell system according to established protocols (27). Briefly, cDNAs encoding C-terminally His 6 -tagged mIL-3 or hIL-3 within the pBacPak8 vector (Novagen) were co-transfected with FlashBAC DNA (Oxford Biosystems) into Sf9 cells. Subsequently, viruses were amplified and used to infect High5 insect cells to produce cytokines. Wild-type and mutant mIL-3 were secreted into the culture medium and used without further purification as growth stimuli for BaF/3 and SCF ER-Hoxb8 cell lines as indicated. For ligand binding and proliferation studies, the secreted mIL-3 and hIL-3 proteins were dialyzed extensively before purification by nickel chromatography and gel filtration.
Hot and Cold Saturation Assays-Murine IL-3 hot and cold saturation binding assays were performed on CTLL2 cells stably expressing the relevant receptors as described before (15). Human IL-3 hot saturation assays were performed on CTLL2 cells stably expressing hIL-3R␣SP1 or SP2 via the pEFIRES-N vector and h␤c via the pEF-IRES-P vector as described before (15,16). COS7 cells transiently transfected with hIL-3R␣SP1 or SP2 in pEFIRES-N vector and h␤c in pCEX-V3-Xba vector were also used in hIL-3 hot saturation binding assays. 10 6 cells were resuspended in binding medium (RPMI1640 supplemented with 10 mM HEPES, pH 7.5, and 0.5% (w/v) bovine serum albumin) and incubated with increasing concentrations of radioligand in a total volume of 200 l for 2-3 h at 4°C. Nonspecific binding was determined by the addition of a 100fold excess of unlabeled ligand to samples in which high radioligand concentrations were added. The nonspecific binding was interpolated for lower radioligand concentrations using linear regression. The assays were terminated by centrifugation of each binding solution through 0.2 ml of 2:1 (v/v) dibutylphthalate (Aldrich):dinonylphthalate (Fluka) at 12,000 ϫ g for 4 min. The tips of the tubes, containing the visible pellet and associ-ated radio-iodinated ligand, were counted using a Packard 5780 Auto-gamma counter. The dissociation constant (K d ) and Scatchard transformations were calculated from specific binding data using the EBDA (28) and LIGAND (29) programs that are contained in the KELL software (Biosoft, Cambridge, UK). Scatchard plots were used to provide a graphical representation of the binding data, and iterative curve fitting by LIGAND was used to estimate K d values. Multiple data files were coanalyzed to obtain more accurate estimates of K d values. One-and twosite models were evaluated in LIGAND, and only when statistically significant (p Ͻ 0.05), a two-site model was used to determine K d values.
Cold Competition Binding Assays-To assess the ability of wild-type and mutant mIL-3 or hIL-3 variants to replace 125 Ilabeled wild-type mIL-3 or hIL-3, respectively, to saturate relevant receptors stably expressed in CTLL2 cells, cold competition binding assays were performed on CTLL2 cells stably transfected with relevant receptors. Stable transfectants expressing ␤ IL-3 and either mIL-3␣ SP1 or SP2 used in this assay showed a single binding site in hot saturation assays as evidenced by a linear Scatchard plot (not shown), as observed before (16). Likewise, CTLL2 cells stably expressing h␤c and either hIL-3␣ SP1 or SP2 also exhibited a one-site binding model in hot saturation hIL-3 binding assays (see Fig. 6, C and D). The cells were prepared as described above. 100 pM, 1 nM, and 1 nM 125 I-labeled wild-type mIL-3 and a serial dilution of unlabeled wild-type or mutant mIL-3 (ranging from 20 pM to 200 nM) were added to 10 6 cells expressing ␤ IL-3 and mIL-3R␣ SP1, ␤ IL-3 and mIL-3R␣ SP2, and ␤ IL-3 only, respectively, resuspended in binding medium to give a total volume of 200 l. 100 pM 125 I-labeled wild-type hIL-3 and a serial dilution of unlabeled wild-type or hIL-3 E22A were added to 10 6 cells expressing hIL-3R␣ SP1 or SP2 and h␤c resuspended in binding medium to give a total volume of 200 l. After incubation for 4 h at 4°C with intermittent agitation, the cells were washed and processed as described above. All of the assays were run in triplicate, in three independent experiments.
The data were analyzed using the nonlinear regression analysis program GraphPad Prism (version 5.0; GraphPad Software Inc., San Diego, CA). Inhibition constants (mean K i values with 95% confidence interval) of competitors for radiolabeled ligand-binding sites were calculated according to the formula where L is the concentration of radiolabeled ligand, K d is the dissociation constant of the radioligand, and IC 50 denotes the concentration of unlabeled ligand producing 50% inhibition of specific binding by the radiolabeled ligand. The K d of radiolabeled ligand for relevant receptors was determined in parallel hot or cold saturation studies carried out as described elsewhere (15,18,30).
Proliferation Assays-Growth responses of CTLL-2 cells expressing IL-3R␣ and relevant ␤-subunit receptors to wildtype or mutant hIL-3 or mIL-3 were measured in triplicate by [ 3 H]thymidine incorporation as described previously (18). For CTLL-2 cells, proliferation was measured after 2 days of incubation. The proliferation data were analyzed by GraphPad Prism (version 5.0; GraphPad Software Inc.). The EC 50 values were determined from sigmoidal dose response curve fitting and expressed relative to the wild-type response as ED 50 values.

Identification and Mutation of Putative Receptor-binding
Residues within mIL-3-In the absence of a structure of mIL-3 or a model of mIL-3 interaction with the ␤ IL-3 receptor, we generated a structural model of the hIL-3⅐h␤c complex (Fig. 1, A and B) by superimposing the solution structure of a hIL-3 analog onto that of hGM-CSF within the experimentally determined hGM-CSF receptor complex crystal structure (20). The orthologous residues in mIL-3 were then identified by alignment of the human and mouse IL-3 sequences (Fig. 1C). The suitability of using the hIL-3 analog structure in such a model in the absence of an mIL-3 structure is validated by our recent NMR spectroscopy studies of an mIL-3 analog (21), where chemical shift indices confirm that the positions of the key ␣-helices, A-D, align with those in hIL-3 (32). The identified candidate ␤ IL-3 -interacting residues comprise the mIL-3 residues, Ile 20 , Lys 27 (A-helix), Glu 65 (preceding the C-helix), Tyr 68 , Val 69 , Ser 72 , Asn 73 , and Lys 76 (C-helix), all of which cluster around the key A-helix residue, Glu 23 (Fig. 1, A and B), which is conserved in all known IL-3 orthologs except the tree shrew, Tupaia belangeri (Fig. 1C), and in the related cytokines, IL-5 and GM-CSF (Fig. 1D). Although the functional role of mIL-3 Glu 23 has not been examined experimentally to date, the homologous residue in hIL-3, Glu 22 , was shown to be important for receptor binding and activation (33,34). We mutated each of these putative receptor-binding residues individually to alanine within full-length mIL-3 and expressed each mutein and wild-type mIL-3 from High5 insect cells using the baculovirus system. Unfortunately, mIL-3 I20A and Y68A mutants did not express (data not shown), and thus the functional consequences of these mutations could not be examined. The side chains of the Ile 20 and Tyr 68 human IL-3 homologs are buried in the hydrophobic core of the hIL-3 four-helix bundle, which indicates that I20A and Y68A mIL-3 were likely to be not secreted from insect cells because of the loss of structural integrity and consequent degradation. The remaining mutants, E23A, K27A, E65A, V69A, S72A, N73A, and K76A, were successfully expressed from High5 cells and purified for functional studies ( Fig. 2A).

Mutation of the Cluster Surrounding Glu 23 Affects mIL-3
Growth Signaling via the SP2 but Not SP1 mIL-3R␣ Isoform-Initially, we examined whether each of the mIL-3 mutants could stimulate the proliferation of the factor-dependent cell line, CTLL2, stably expressing ␤ IL-3 and either the full-length (SP1 isoform) mIL-3R␣ (Fig. 2B) or a naturally occurring splice variant of mIL-3R␣ lacking the N-terminal Ig-like domain (D1; the SP2 isoform; Fig. 2C). Remarkably, the CTLL2 cells coexpressing the mIL-3R␣ SP1 and ␤ IL-3 subunits were relatively insensitive to mutation of Glu 23 and residues in the surrounding cluster (Fig. 2B and Table 1). Only mIL-3 E65A exhibited FIGURE 1. Structural model of the hIL-3⅐h␤c interface and IL-3 alignments. A, a schematic depicting a model of hIL-3 in complex with the domain 1-domain 4 interface of the h␤c ectodomain. hIL-3 is drawn in green, domain 1 from chain A of the h␤c homodimer is drawn in red, and domain 4 from chain B of the h␤c homodimer is drawn in blue. The model was prepared by superimposition of the solution structure of a hIL-3 analog (Protein Data Bank accession code 1jli) on to hGM-CSF within the hGM-CSF receptor complex crystal structure (Protein Data Bank accession code 3cxe) using Coot (31). This figure was drawn using PyMOL. B, close-up view of the boxed region in A. The side chains of candidate h␤c-interacting residues within hIL-3 are drawn as orange sticks and labeled according to the hIL-3 sequence. C, multi-species alignments of IL-3 amino acid sequences. hIL-3 residues identified as interactors of h␤c by inspection of the model in A and B and their homologs in mIL-3 selected for alanine substitution mutagenesis are highlighted. The residue numbers correspond to the mIL-3 sequence. A more comprehensive alignment of IL-3 ortholog sequences is published in Ref. 21. D, structural alignment of mIL-3, mGM-CSF, and mIL-5 amino acid sequences, with the highly conserved A-helix glutamic acid residue highlighted. The positions of helices, marked with red bars above the sequences, were inferred from chemical shift indices determined by NMR spectroscopy for an mIL-3 analog (32) and mGM-CSF alignment with the hGM-CSF structure (Protein Data Bank accession code 2gmf). The mIL-5 secondary structure was deduced from the crystal structure (Protein Data Bank accession code 3b5k).

Two Modes of mIL-3 and hIL-3 ␤-Receptor Binding
any marked deficit in receptor activation relative to wildtype mIL-3, with 29-fold higher concentrations of mIL-3 E65A than wild-type mIL-3 required to give comparable growth stimulation.
In contrast, each of these mIL-3 mutants, except S72A, less effectively stimulated proliferation of CTLL2 cells co-expressing ␤ IL-3 and the mIL-3R␣ isoform lacking D1 (SP2) relative to wild-type mIL-3 ( Fig. 2C and Table 1). The most severely affected mIL-3 mutants were E23A and K27A: two mutations that led to complete abolition of growth signaling, even at mIL-3 concentrations Ͼ500-fold higher than those required for half-maximal stimulation with wild-type mIL-3. For both of these mutants, no proliferative response was observed for ␤ IL-3 /mIL-3R␣SP2-expressing CTLL2 cells in growth titrations extending to mIL-3 concentrations in excess of 100 times the wild-type mIL-3 effective concentration required for half-maximal growth stimulation (EC 50 ). Mutation of neighboring residues located in the C-helix of mIL-3 led to profound, but less severe, defects in the sensitivity of mIL-3R␣SP2/␤ IL-3 -expressing CTLL2 cells to mIL-3 stimulation. Of these mutants, mIL-3 E65A, V69A, and N73A were the least effective stimuli for CTLL2 mIL-3R␣SP2/␤ IL-3 growth, with respective 9.6-, 19-, and 41-fold higher mIL-3 concentrations required to achieve growth stimulation comparable with wild-type mIL-3 ( Fig. 2C and Table 1). mIL-3 S72A, in comparison, retained stimulatory activity toward CTLL2 mIL-3R␣SP2/␤ IL-3 cells comparable with wild-type mIL-3, albeit with reductions in the maximal levels of cell proliferation, whereas mIL-3 K76A exhibited a modest reduction in proliferative activity (2-fold relative to wild-type mIL-3) (Fig. 2C).
Receptor Binding Activities of mIL-3 Mutants-To further probe the molecular basis for mIL-3 receptor recognition, we performed binding assays to measure the capacity of our panel of mIL-3 mutants to bind receptors composed of ␤ IL-3 and either the mIL-3R␣ SP1 or SP2 isoform. These studies utilized a cold competition assay in which a titration of unlabeled mIL-3 (either wild type or mutant) ranging from 20 pM to 200 nM competed with a fixed amount of 125 I-radiolabeled wild-type mIL-3 for binding to the appropriate mIL-3 receptors stably expressed on CTLL2 cells (100 pM, 1 nM, and 1 nM radiolabeled mIL-3 for mIL-3R␣SP1/␤ IL-3 , mIL-3R␣SP2/␤ IL-3 , and ␤ IL-3,  respectively). From these data, we calculated the concentration of wild-type or mutant mIL-3 required for 50% inhibition of 125 I-labeled wild-type mIL-3 binding to the relevant receptor (IC 50 values), which were converted to inhibition constants (K i values). Initially, we performed a detailed examination of the mIL-3 E23A receptor binding properties (Fig. 3, A-C, and Table  1). These studies revealed that the mIL-3 E23A shows only a modest decrease in affinity for CTLL2 mIL-3R␣SP1/␤ IL-3 cells (K i ϭ 0.30 nM versus 0.17 nM for wild-type mIL-3), whereas mIL-3 E23A did not detectably bind CTLL2 mIL-3R␣SP2/␤ IL-3 cells (Fig. 3, A and B, and Table 1). We used the cold competition assay to measure the affinity of mIL-3 E23A for CTLL2 cells expressing solely the ␤ IL-3 receptor. In this assay, mIL-3 E23A did not detectably bind directly to the ␤ IL-3 , whereas the K i for wild-type mIL-3 binding was measured as 11.8 nM (Fig.  3C). These data suggest that Glu 23 is an essential determinant of mIL-3 direct binding to ␤ IL-3 and, similarly, mIL-3 binding to ␤ IL-3 in the presence of the mIL-3R␣ SP2 isoform. These findings mirror our prior work showing that mutagenesis of ligandbinding residues in ␤ IL-3 domain 1-domain 4 functional epitope abolishes direct binding to mIL-3 as well as mIL-3 binding in the presence of mIL-3R␣ SP2 (16). As a result, because ␤ IL-3 utilizes the same epitope to bind mIL-3 both directly and in the presence of mIL-3R␣ SP2, we measured binding for the remaining mIL-3 mutants to CTLL2 cells co-expressing ␤ IL-3 and either the mIL-3R␣ SP1 or SP2 isoform, but not ␤ IL-3 alone. As shown in Fig. 3 (D and E), mIL-3 K27A exhibited dramatically reduced binding to CTLL2 mIL-3R␣SP1/␤ IL-3 cells (K i , 1.56 nM relative to 0.17 nM for wild-type mIL-3; Table 1) and no detectable binding to CTLL2 cells co-expressing ␤ IL-3 and mIL-3R␣SP2 subunits. Thus, Lys 27 is a critical residue for high affin-ity binding to the mIL-3R␣SP2/␤ IL-3 receptor, but in contrast to Glu 23 , it also plays a key role in the formation of the mIL-3⅐mIL-3R␣ SP1⅐␤ IL-3 complex.
We then examined the capacity of mIL-3 C-helix mutants to bind CTLL2 cells co-expressing ␤ IL-3 and either mIL-3R␣ SP1 or SP2. Of these mutants, only mIL-3 E65A showed any substantial deficit in binding to CTLL-2 cells co-expressing the ␤ IL-3 and mIL-3R␣ SP1 subunits, with a 10-fold increase in K i relative to wild-type mIL-3 ( Fig. 4A and Table 1). In contrast, binding affinities of the mIL-3 mutants, V69A, S72A, N73A, and K76A, for CTLL2 mIL-3R␣ SP1/␤ IL-3 cells were comparable with wild-type mIL-3 (Fig. 4, B-E, and Table 1). We also evaluated the capacity of the mIL-3 C-helix mutants to bind CTLL2 cells co-expressing mIL-3R␣ SP2 and ␤ IL-3 subunits. These experiments revealed an essential role for Asn 73 of mIL-3 in receptor binding, because mIL-3 N73A does not detectably bind the mIL-3R␣ SP2/␤ IL-3 receptor system (Fig. 4I), and mutation of the neighboring residues, Glu 65 and Val 69 , led to Ͼ3and Ͼ9-fold elevations in K i values relative to wild-type mIL-3. However, by comparison, mutation of either Ser 72 or Lys 76 does not substantially compromise the affinity of mIL-3 for the mIL-3R␣ SP2/␤ IL-3 receptor (Fig. 4, H and J, and Table 1).
It should be noted that internalization of receptor⅐ligand complexes will have minimal bearing on the correlation between ligand binding and receptor activation in the present study. First, all binding experiments were conducted at 4°C to minimize the possibility of internalization over the 2-3-h time course of binding assays. Second, studies of the related IL-5 receptor have shown that internalization of human ␤c occurs only after IL-5 binding and the activation of downstream sig-

Two Modes of mIL-3 and hIL-3 ␤-Receptor Binding
naling pathways (35), and as a result, we can assume that receptor⅐ligand complex internalization during the course of our proliferation assays is a key step in "turning off" a signal and will correlate with the capacity of wild-type or mutant IL-3 to activate the IL-3 receptor and initiate intracellular signaling.
Activation of Jak2 and Erk1/2 Signaling by Wild-type and E23A mIL-3-To examine the roles of the two distinct mIL-3 binding modes in governing the initiation of intracellular signaling, we investigated the capacity of wild-type and E23A mIL-3 to activate Jak2 and Erk1/2 in two mIL-3-dependent selfrenewal cell line models: Ba/F3, a pro-B cell line; and SCF ER-Hoxb8, a neutrophil progenitor cell line prevented from undergoing differentiation when cultured in the presence of ␤-estradiol, which is required for the activity of a retrovirally introduced Hoxb8 gene (26). We compared the signals arising from wild-type mIL-3 with mIL-3 E23A, because unlike most other mIL-3 mutants described herein, mIL-3 E23A can elicit a wild-type growth response in CTLL2 cells co-expressing ␤ IL-3 and mIL-3␣ SP1 but no detectable growth response in an analogous cell line expressing mIL-3␣ SP2. Ba/F3 cells exhibit comparable proliferation responses when stimulated with wild-type and E23A mIL-3, 4 and because of its insensitivity to the E23A mutation, we classified Ba/F3 as a cell line that signals primarily via mIL-3R␣ SP1. In contrast, SCF ER-Hoxb8 is a neutrophil progenitor cell line that showed a Ͼ100-fold reduction in proliferative response to the E23A mIL-3 mutant relative to wildtype mIL-3, 4 thereby indicating that this cell line signals primarily via the mIL-3R␣ SP2 isoform. We conducted stimulations of these cell lines with 1% (v/v) conditioned insect cell media containing of wild-type or E23A mIL-3, lysed the cells, and examined the activation status of the key signaling effector kinases, Jak2 and Erk1/2, by Western blot, using antibodies that recognize phosphorylation of Tyr 1007 /Tyr 1008 of Jak2 and Thr 202 /Tyr 204 of Erk1/2 (Fig. 5). Consistent with our findings in the proliferation assay (Fig. 2), Ba/F3 cells showed no measurable differences in Jak2 and Erk1/2 signaling when stimulated with wild-type or E23A mIL-3. In contrast, SCF ER-Hoxb8 cells showed a dramatic reduction in Jak2 and Erk activation when stimulated with E23A mIL-3 (Fig. 5).
hIL-3⅐h␤c Binding Interfaces in Complexes with hIL-3R␣ SP1 and SP2-Although the involvement of hIL-3 Glu 22 in receptor activation and binding to h␤c within the hIL-3R␣ SP1⅐h␤c high affinity complex has been demonstrated (33,34), its role in receptor binding and activation with h␤c in the presence of the novel hIL-3R␣ SP2 isoform has not been examined. In the present work, we compared the capacity of hIL-3 E22A to bind and activate h␤c in the presence of either the hIL-3R␣ SP1 or SP2 isoform (Fig. 6A). Growth signaling was measured using CTLL2 cells stably expressing hIL-3R␣ SP1/h␤c or hIL-3R␣ SP2/h␤c for both wild-type and hIL-3 E22A. In the case of wild-type hIL-3, 16-fold higher concentrations were required for full growth stimulation via the hIL-3R␣ SP2/h␤c receptor relative to the hIL-3R␣ SP1/h␤c receptor, in agreement with our previous findings (16). In contrast, hIL-3 E22A gave a much reduced growth stimulation of cells expressing hIL-3R␣ SP1/h␤c (115fold increased ED 50 for hIL-3 E22A relative to wild-type hIL-3) and no growth stimulation of cells expressing hIL-3R␣ SP2/h␤c (Fig. 6A). These data illustrate that Glu 22 is essential for growth signaling via the hIL-3R␣ SP2/h␤c receptor and plays an important, albeit nonessential, role in signaling via the hIL-3R␣ SP1/ h␤c receptor. Consistent with the growth signaling data described above, binding of hIL-3 E22A to the hIL-3R␣ SP1/ h␤c receptor was measured using cold competition assays and was found to be dramatically reduced relative to wild-type hIL-3 (Fig. 6B) (K i ϭ 123 nM versus wild type 0.125 nM). However, because of the low levels of 125 I-radiolabeled hIL-3 binding to CTLL2 cells co-expressing the hIL-3R␣ SP2 and h␤c subunits, it was not possible to determine the K i for hIL-3 E22A binding using cold competition assays.
We measured the K d for high affinity binding to the hIL-3R␣ SP2/h␤c receptor stably expressed in CTLL2 cell lines and transiently expressed in COS7 cells using hot saturation binding assays with 125 I-radiolabeled hIL-3. In CTLL2 cells, the K d values for hIL-3R␣ SP1/h␤c and hIL-3R␣ SP2/h␤c binding to wild-type hIL-3 were quite similar (157 and 150 pM, respectively), with similar values obtained from binding studies in COS7 cells (172 and 126 pM, respectively) ( Table 2). This is the first time that the K d has been determined for high affinity binding with the naturally occurring hIL-3R␣ SP2 isoform. Intriguingly, these data reveal for the first time that hIL-3 binds to the hIL-3R␣ SP2/h␤c receptor with an affinity comparable with hIL-3R␣ SP1/h␤c binding when expressed in either CTLL2 or COS7 cells (Fig. 6C and Table 2). Notably, far fewer high affinity binding sites/cell were detected in CTLL2 cells (78 sites with SP2 and 1864 sites with SP1), albeit with a more comparable number detected in COS7 cells (88 sites with SP2 and 275 sites with SP1). This is an intriguing observation, because comparable SP1 and SP2 isoform expression can be detected by flow cytometry on the surface of these cells (16), 4 and comparable numbers of low affinity hIL-3-binding sites were measured in COS7 cells (10,928 sites for SP2 and 14,965 sites for SP1). Overall, we conclude that despite being properly transported to the cell surface, the hIL-3R␣ SP2 subunit exhibits a lower propensity to engage in high affinity hIL-3 binding in CTLL2 cells. This situation clearly differs from the Ig-like domain deletion mutants of the mouse and human IL-6␣ receptor, which fail to localize to the cell surface (37), and the human GM-CSF recep-FIGURE 5. Activation of Jak2 and Erk1/2 signaling in Ba/F3 and SCF ER-Hoxb8 cells by wild-type and E23A mIL-3. BaF/3 cells cultured in mIL-3 (left three lanes) and SCF ER-Hoxb8 cells cultured in SCF (right three lanes) were starved for 4 h before stimulation with wild-type (wt), E23A (mt), or no (-) mIL-3 for 10 min. Western blots were performed to detect phospho-Jak2 (Tyr 1007/1008 ) and total Jak2 (upper two panels) and phospho-Erk1/2 (Thr 202 / Tyr 204 ) and total Erk1/2 (lower two panels). The data shown are representative of the outcomes of a minimum of two independent experiments.

Two Modes of mIL-3 and hIL-3 ␤-Receptor Binding
tor ␣-subunit, which requires its Ig-like domain for high affinity GM-CSF binding and normal receptor activation (38).
In earlier work, the key ligand-binding interface on h␤c was shown to comprise residues at the elbow formed between domains 1 and 4, contributed by the two chains of the h␤c homodimer (15,17,18). Previously, we showed that the domain 1 h␤c residues Tyr 15 (A-B loop) and Phe 79 (E-F loop) play important roles in hIL-3 binding and growth signaling with hIL-3R␣ SP1/h␤c (15,18), and other studies have demonstrated that the domain 4 h␤c residue, Tyr 403 (FЈ-GЈ loop), is also critical for hIL-3 binding (36). Interestingly, the domain 4 residues Tyr 347 , His 349 , and Ile 350 in the BЈ-CЈ loop, which are critical for IL-5 and GM-CSF high affinity binding, do not appear to play a role in IL-3 binding with hIL-3R␣ SP1/ h␤c (39). We investigated the role of the h␤c residues Tyr 15 , Phe 79 , Tyr 347 , and His 349 in high affinity binding with the hIL-3R␣ SP2 isoform relative to the hIL-3R␣ SP1 isoform. These experiments were carried out in COS7 cells using transient expression. Neither h␤c Y15A nor h␤c F79A gave detectable hIL-3 binding when co-expressed with either hIL-3R␣ SP1 or hIL-3R␣ SP2, indicating that the domain 1 residues Tyr 15 and Phe 79 are both critical for high affinity binding with each of the IL-3R␣ isoforms. The lack of binding is not due to poor expression of h␤c Y15A or h␤c F79A, because the expression of these two mutants was comparable with wild-type h␤c expression in COS7 cells, as shown previously (Ref. 18 and data not shown). Interestingly, with hIL-3R␣ SP1, the h␤c Y347A and H349A mutants resulted in only modest reductions in wild-type high affinity binding, as described in earlier studies (39). The h␤c Y347A and H349A mutants were previously shown to be expressed at levels comparable with wild-type h␤c (39), a finding supported by the high affinity hIL-3 binding observed in the present work. In contrast, when co-expressed with hIL-3R␣ SP2, the h␤c Y347A and H349A mutants gave no detectable high affinity hIL-3 binding. These data clearly illustrate differences in the relative contributions of functional epitope residues located at the h␤c domain 1-domain 4 elbow region to hIL-3 high affinity in the presence of the two hIL-3R␣ isoforms. With both hIL-3R␣ SP1 and SP2, mutation of the key domain 1 residues, Tyr 15 and Phe 79 , completely abrogates high affinity hIL-3 binding. However, upon mutation of Tyr 347 or His 349 in the domain 4 BЈ-CЈ loop, high affinity hIL-3 binding is ablated in the presence of the hIL-3R␣ SP2 isoform but only modestly reduced with hIL-3R␣ SP1. These data mirror the differential contribution of hIL-3 Glu 22 to recognition and activation of h␤c co-expressed with the hIL-3R␣ SP1 and SP2 isoforms, as described above, where hIL-3R␣ SP1/h␤c receptors could be activated by hIL-3 E22A, and hIL-3R␣ SP2/h␤c receptors could not. Collectively, FIGURE 6. Mutational analysis of the hIL-3⅐h␤c interface. A, growth stimulation of CTLL2 cells expressing the hIL-3R␣ SP1/h␤c or hIL-3R␣ SP2/h␤c receptors by hIL-3 E22A and wild-type hIL-3. B, competition binding assays measuring the ability of hIL-3 E22A and wild-type hIL-3 to compete with 125 I-radiolabeled wild-type hIL-3 for binding to the hIL-3R␣ SP1/h␤c receptor. C-H, hot saturation binding assays using 125 I-radiolabeled wild-type hIL-3 on CTLL2 cells co-expressing hIL-3R␣ SP1/h␤c (C) and hIL-3R␣ SP2/h␤c (D) and COS7 cells co-expressing hIL-3R␣ SP1/h␤c (E), hIL-3R␣ SP2/h␤c (F), hIL-3R␣ SP1/h␤c Y347A (G), and hIL-3R␣ SP1/h␤c H349A (H).

Two Modes of mIL-3 and hIL-3 ␤-Receptor Binding
these data support the role of the h␤c domain 1-domain 4 elbow region in hIL-3 binding but suggest that there are differences in the epitopes bound by hIL-3 depending on whether the complex involves hIL-3R␣ SP1 or SP2.

DISCUSSION
The molecular mechanisms underlying the capacity of the IL-3 receptor to achieve differential signaling outcomes in response to IL-3 stimulation, such as the proliferation or differentiation of hematopoietic progenitors, have remained an enigma ever since the discovery of IL-3. By identifying a naturally occurring splice variant of the IL-3 ␣-subunit termed SP2 that lacks the N-terminal Ig-like domain that is otherwise present in the full-length receptor (termed SP1) in addition to the membrane-proximal cytokine receptor homology module (16), we have uncovered additional complexity within the IL-3 receptor system that would enable IL-3 to activate distinct signaling outcomes. Interestingly, whereas the IL-3R␣ subunits in both mouse and humans exist as either SP1 or SP2 splice variants, no such naturally occurring isoforms have been identified in the related GM-CSF or IL-5 ␣ receptors (38), consistent with the essential roles of the N-terminal Ig-like domains within these ␣-subunits for receptor activation (38,40).
In the present work, we sought to establish which mIL-3 residues are crucial to recognition and activation of the ␤ IL-3 receptor in the presence of either the mIL-3R␣ SP1 or SP2 isoforms. In the absence of the detailed analyses of mIL-3, it has remained unclear: (a) which residues in the ligand itself are critical for ␤ IL-3 binding and (b) whether the same key mIL-3 residues contribute to the two distinct modes of ␤ IL-3 recognition, as dictated by the mIL-3R␣ SP1 or SP2 splice variants. Initially, we identified candidate ␤ IL-3 -interacting residues based on a model of the hIL-3 interaction with the related h␤c receptor (Fig. 1A), which was derived from the recent hGM-CSF⅐receptor complex crystal structure (20). Residues located in the A-and C-helices of hIL-3, which cluster around Glu 22 , a key residue for receptor binding and activation (33,34), were selected from the model the hIL-3/h␤c interaction, and the homologous residues in mIL-3 were identified by sequence alignment (Fig. 1C). We prepared a panel of mIL-3 mutants, comprising E23A and the neighboring residues K27A (A-helix), E65A (preceding C-helix), V69A, S72A, N73A, and K76A (C-helix), and examined their capacities to activate and bind ␤ IL-3 in the presence of either mIL-3R␣ SP1 or SP2 relative to wild-type mIL-3 ( Figs. 3 and 4). Of these mutants, only mIL-3 S72A behaved comparably with wild-type mIL-3 in all of our experiments, which is surprising because Ser 72 is one of the few residues conserved in this region between mIL-3 and hIL-3 with the exception of hydrophobic core contributors. Among the other mIL-3 mutants, we observed an excellent agreement between their receptor binding properties and their capacity to activate the receptor. The only possible exception is mIL-3 K27A, which exhibits a modest defect in its capacity to stimulate proliferation of CTLL2 mIL-3R␣ SP1/␤ IL-3 cells relative to wild-type mIL-3 (a 2-fold increase in EC 50 ) but with a profound reduction in binding affinity for these cells relative to wild-type mIL-3 (a 9-fold increase in K i ).
It is notable that only the mIL-3 K27A and E65A mutants showed reduced affinities for CTLL2 cells co-expressing mIL-3R␣ SP1 and ␤ IL-3 , and they were the only two mutants among our panel that showed reduced stimulation of this cell line. These findings suggest that Glu 65 is a critical residue for mIL-3R␣ SP1/␤ IL-3 receptor binding and activation, and Lys 27 plays a secondary role. Moreover, these data suggest the existence of an epitope on mIL-3 that mediates ␤ IL-3 recognition in the presence of the mIL-3R␣ SP1 subunit that is likely to be distinct from, but still overlap, the canonical receptor-binding epitope clustering around Glu 23 . A complete understanding of the ␤ IL-3 epitope that mediates mIL-3 binding with mIL-3R␣ SP1 awaits determination of a complex crystal structure.
The receptor composed of the mIL-3R␣ SP2 and ␤ IL-3 subunits was remarkably sensitive to mutation of Glu 23 and the surrounding residues in mIL-3, consistent with an important role of the Glu 23 cluster in mediating recognition of the domain 1-domain 4 ligand-binding interface that we have previously described for direct mIL-3 binding to ␤ IL-3 (15), the same interface that is utilized for mIL-3 binding by ␤ IL-3 in the presence of the mIL-3R␣ SP2 isoform (Fig. 3C and Ref. 16). Proliferation of CTLL2 cells expressing the mIL-3R␣ SP2/␤ IL-3 receptor system Binding was determined using the hot saturation binding assay, and K d Ϯ S.E. was determined by co-analysis of data from multiple experiments using LIGAND (29). b Binding was not detectable above nonspecific binding. c The one-site binding model was statistically significant. d To improve the accuracy of determination of the high affinity K d , the low affinity site K d (for hIL-3R␣SP1 binding) was fixed at 100 nM when a two-site binding model was found to be statistically significant, according to established protocols (36).
Our data are reminiscent of prior studies of the hIL-3, where substitution of residues located adjacent to the conserved A-helix glutamate, in particular His 26 , the ortholog of mIL-3 Lys 27 , was poorly tolerated and diminished hIL-3 receptor activation and bioactivity (25,34). In the present study, we unveiled a key role for mIL-3 Lys 27 in the recognition and activation of ␤ IL-3 in the presence of either mIL-3R␣ SP1 or SP2. The importance of this residue is especially interesting considering that Lys 27 is poorly conserved between IL-3 orthologs (Fig. 1C) and, unlike the highly conserved Glu 23 , is not conserved in the related cytokines, GM-CSF and IL-5 (Fig. 1D). This raises the possibility that Lys 27 in mIL-3 is at least partly responsible for conveying the capacity of mIL-3 to directly bind the ␤ IL-3 receptor and may govern the specificity of mIL-3, in preference to GM-CSF and IL-5, for binding to this mIL-3-specific receptor. Our data also indicate that the C-helix residue, Asn 73 , is an important contributor to mIL-3 recognition of the ␤ IL-3 domain 1-domain 4 ligand-binding epitope (Fig. 4). The role of the mIL-3 C-helix in receptor binding and activation was largely unexpected, because prior mutagenesis studies of the orthologous hIL-3 clearly showed that the C-helix was very tolerant of amino acid substitution (34). One notable exception, however, was alanine substitution of hIL-3 Ile 77 , the homolog of mIL-3 Asn 73 , which resulted in a 15-fold reduction in potency (25), consistent with the mIL-3 homolog Asn 73 serving a similar, conserved function in receptor binding and activation.
Our recent discovery of the hIL-3R␣ SP2 splice variant, which complements the existing full-length or SP1 isoform, led us to investigate whether a similar mechanism to that deduced for ␤ IL-3 binding and activation by mIL-3 operates within the human IL-3 receptor system. The hIL-3-binding epitope on h␤c has remained a curiosity, because although hIL-3 binds h␤c via the domain 1-domain 4 elbow interface, mutations within the h␤c domain 4 BЈ-CЈ loop do not abrogate hIL-3 binding in the presence of the hIL-3R␣ SP1 isoform (36) (Fig. 6, G and H).
In the present work, we confirmed that the h␤c domain 4 BЈ-CЈ loop mutants, Y347A and H349A, do not abrogate high affinity hIL-3 binding when co-expressed with hIL-3R␣ SP1 but, conversely, when co-expressed with hIL-3R␣ SP2, no high affinity hIL-3 binding could be detected. In comparison, mutation of other established ligand-binding residues in h␤c domain 1, Tyr 15 and Phe 79 (15), abrogated hIL-3 high affinity binding when co-expressed with either of the hIL-3R␣ isoforms. These data indicate that hIL-3 recognition of h␤c within the hIL-3R␣ SP1⅐h␤c and hIL-3R␣ SP2⅐h␤c signaling complexes involves overlapping, but subtly different, interaction interfaces on the h␤c receptor. The same appears to be true of hIL-3 itself, because alanine substitution of the highly conserved Glu 22 within hIL-3 completely abolishes activation of the hIL-3R␣ SP2/h␤c receptor, whereas activation of the hIL-3R␣ SP1/h␤c receptor is severely reduced but still detectable.
The studies described in the present work have greatly enhanced our knowledge of the molecular details underlying mIL-3 recognition of ␤ IL-3 within the signaling complexes formed with the mIL-3R␣ SP1 or SP2 isoforms. These studies provide further support for the notion that the mIL-3R␣ SP1 or SP2 subunits can direct mIL-3 binding to distinct epitopes on the ␤ IL-3 receptor and suggest an elegant mechanism by which the relative orientations of receptor subunits could potentially be tuned to initiate different intracellular signaling programs to elicit distinct signaling outcomes. In addition, the subtle differences in human IL-3 interaction epitopes on the h␤c receptor, as governed by the hIL-3R␣ SP1 and SP2 isoforms, suggest that a mechanism similar to that deduced for the mIL-3 receptor operates in the human IL-3 receptor system. This concept of tunable receptor activation and signaling outputs has been elegantly explored for the erythropoietin and growth hormone receptors, where constraining the relative orientations of the transmembrane/intracellular domains led to preferential activation of distinct signaling effectors, such as Jak2 or MAPK (41,42). Although the studies described herein have greatly advanced our knowledge of the mechanisms underlying the mouse and human IL-3 receptor activation, a comprehensive understanding of ligand binding and signal transduction by this fascinating receptor system awaits the crystal structures of IL-3 in complex with the ectodomains of the IL-3R␣ SP1 or SP2 and ␤-subunits.