Epitope Randomization Redefines the Functional Role of Glutamic Acid 110 in Interleukin-5 Receptor Activation*

Sequence randomization through functional phage display of single chain human interleukin (IL)-5 was used to investigate the limits of replaceability of the Glu110 residues that form a part of the receptor-binding epitope. Mutational analysis revealed unexpected affinity for IL-5 receptor α chain with variants containing E110W or E110Y. Escherichia coli-expressed Glu110variants containing E110W in the otherwise sequence-intact N-terminal half, including a variant with an E110A replacement in the sequence-disabled C-terminal half, were shown by their CD spectra to be folded into secondary structures similar to that of single chain human IL-5 (scIL-5). Biosensor kinetics analysis revealed that (E110W/A5)scIL-5 and (E110W/A6)scIL-5 had receptor α chain binding affinities similar to that of (wt/A5)scIL-5. However, (E110W/A6)scIL-5 had a significantly reduced bioactivity in TF-1 cell proliferation compared with both (wt/A5)scIL-5 and (E110W/A5)scIL-5, and this activity reduction was disproportionately greater than the much smaller effect of Glu110 mutation on receptor binding affinity. The marked and disproportionate decrease in TF-1proliferation observed with (E110W/A6)scIL-5 suggests a role for Glu110 in the biological activity mediated by the signal transducing receptor βc subunit of the IL-5 receptor. This is also consistent with the lack of stimulation of JAK2phosphorylation by the (E110W/A6)scIL-5 mutant in recombinant 293T cells, as compared with the concentration-dependent stimulation seen for scIL-5. The results reveal the dispensability of charge in the Glu110 locus of IL-5 for receptor α chain binding and, in contrast, its heretofore underappreciated importance for receptor activation.

The cytokine interleukin-5 (IL-5) 1 plays a key role in controlling the maturation, proliferation, and activation of eosino-phils, which have been implicated in the pathogenesis of asthma and other allergic inflammatory diseases associated with hypersensitivity reactions in the lung (1)(2)(3). IL-5 exerts its biological functions through binding to a heteromeric cell surface receptor composed of two types of subunit, ␣ and ␤ c , each containing a glycosylated extracellular domain with repeating fibronectin-like sequences, a single transmembrane domain, and a cytoplasmic domain. The ␣ chain is IL-5-specific, although structurally related to the ␣ chains for GM-CSF and IL-3 receptors. The ␤ c chain is identical to the ␤ chains of GM-CSF and IL-3 receptors (4,5), hence the denotation ␤ c for common ␤. The extracellular domain of ␣ subunit can bind to IL-5 in the absence of ␤ c chain (4,6,7). Human ␣ chain alone binds IL-5 with a K d of 0.3-0.6 nM when expressed in COS7 cells (6). That ␤ c subunit also contributes to ligand binding (7,8) is indicated by several findings, including the increased (2-4 fold) hIL-5 binding affinity when ␣ and ␤ c chains are co-expressed (9). However, direct binding of IL-5 to ␤ c alone has never been quantitated. The ␤ c subunit is essential for signaling.
IL-5 is a member of a broad family of structurally related cytokines that contain a helical bundle core (10). The crystal structure of the IL-5 dimer shows it is composed of a pair of four-helix bundles (11,12). Each bundle is similar to the growth hormone fold (13) found in many cytokines including IL-2 (14), IL-4 (15,16), GM-CSF (17), macrophage CSF (18), and granulocyte CSF. In IL-5, helix D of one polypeptide chain combines with helices A, B, and C of a second, identical chain by domain swapping (19) to form the stable dimer.
Understanding of the IL-5 receptor activation process is still incomplete but is advancing. A growing body of evidence suggests that helices A (residues  and D (residues 93-118) of IL-5 both contribute important components to receptor binding. Evidence obtained in our own work by site-directed mutagenesis (20 -22) as well as in other reports (23,24) confirms the involvement of residues in the C-terminal D helix region of human IL-5 (in particular Glu 110 ) and nearby residues in the CD loop (in particular Glu 89 and Arg 91 ) in receptor ␣ chain interaction. Evidence for involvement of the IL-5 A helix in recruiting ␤ c has been obtained with a chimeric protein containing an N-terminal segment (residues 5-29) of mouse IL-5 and the remaining sequence from human GM-CSF (25). A role for Glu 13 of the A helix in biological activity but not ␣ chain interaction has been shown by mutagenesis of wild type IL-5 (23,24) and of scIL-5 (21). The Glu 13 of IL-5 is analogous to the functionally important Glu 22 in IL-3 and Glu 21 in GM-CSF. This is consistent with a role of Glu 13 in ␤ c chain interaction. The CD loop and Glu 110 compose an epitope around the fourhelix bundle interface of the IL-5 cylinder, whereas the Glu 13 residues are at the distal ends of the cylinder (Fig. 1). A view has arisen in which receptor ␣ chain recruitment occurs around the four-helix bundle interface, whereas ␤ c recruitment leading to signaling occurs at the distal ends (26).
Although site-directed mutagenesis studies have identified important residues for receptor recruitment and activation, the one-residue-at-a-time substitution allowed by this technique limits a full mechanistic understanding of what specific structural and electrostatic features of the IL-5 surface are required for receptor binding. Such understanding can be obtained more completely by examining the replaceability of individual side chains, or sets of side chains, by all other side chain types, in other words by examining libraries of sequences containing all possible side chains in local regions of the surface and determining which combinations allow productive binding. Such random epitope mutagenesis can be achieved with sequence libraries formed by phage display. This technique has been applied successfully to in vitro antibody maturation and protein engineering (27). Typically a foreign gene is fused in frame with phage coat protein pIII encoded by a phagemid vector. The surface displayed recombinant foreign protein can be selected by affinity selection procedures (biopanning). Human growth hormone has been displayed on phage and higher receptor affinity variants selected through random mutagenesis and phage panning (28).
We have displayed scIL-5 on phage using a two-stage construction that now allows asymmetric randomization mutagenesis of the IL-5 molecule and therefrom a deeper understanding of the IL-5-receptor recognition process (29). An asymmetrically disabled but functional scIL-5 mutant was displayed on phage. This mutant was constructed of an N-terminal half containing the original five charged residues ( 88 EERRR 92 ) in the CD loop combined with a C-terminal half containing a disabling CD loop sequence ( 88 AAAAA 92 ) in which the charged side chains of both Glu 89 and Arg 91 (identified by site-directed mutagenesis as important for ␣ chain recruitment) are replaced. This asymmetrically disabled variant was used as a starting point to generate an asymmetric scIL-5 library in which the 88 -92-residue N-terminal CD loop was randomized to identify functional IL-5 variants. From this epitope library, a receptor-binding variant of IL-5 was selected in which the only charged residue in the CD loop is Arg 90 . The results obtained argue that the key receptor recognition element in the CD loop is an Arg residue, that the position of this charged residue may be mutable within the loop from its position in wild type IL-5, that the CD loop can be simplified in considering design of structure-based mimetics, and that phage display has utility to examine the IL-5 receptor recruitment mechanism.
Based on these epitope randomization results, we have now used the sequence randomization approach to examine the third residue in the putative IL-5R␣ recruitment epitope of IL-5, namely Glu 110 . Mutational analysis of Glu 110 in phage displayed, asymmetrically disabled scIL-5 showed unexpectedly that residue replacements that enabled retention of IL-5R ␣ binding included Trp and Tyr. The substantial binding efficacy upon E110W mutation was verified with purified, Escherichia coli-expressed mutated proteins. In contrast, depletion of both negative charges at residue 110 led to a major reduction in cell proliferation and JAK2 phosphorylation activity. The results of this work show that the Glu 110 negatively charged locus is important for receptor activation, although it is dispensible for ␣ chain binding, that the negative charge of Glu 110 can aid activation either within its four-helix bundle domain or through cross-domain cooperativity, and that charge-depleting mutations at Glu 110 may help lead to IL-5 antagonists that bind ␣ chain but do not induce receptor activation. We consider two mechanistic possibilities by which Glu 110 may function in receptor activation, including its direct involvement in ␤ c contact or alternatively its more indirect impact on ␤ c recruitment via the mode of ␣ chain binding that it helps to effect.

Construction of Glu 110 Mutants by Site-directed Mutagenesis-
The phagemid vector, pMK-wt/A5-G3 (29), was used as a starting point to prepare different Glu 110 mutants. Two working vectors, pCR-IL-5A and pCR-IL-5B encoding one copy of IL-5 sequence, have been used for asymmetric mutation as reported previously (29). To generate mutations in Glu 110 position, the codon of Glu 110 in pCR-IL-5A was mutated to 19 other amino acids using a QuickChange site-directed mutagenesis kit (Stratagene). The combination of mutated Glu 110 in the N-terminal half and a wild type Glu 110 in the C-terminal half of scIL-5 yielded an asymmetric mutant phagemid, designated pMK-E110X/A5-G3. X represents all residues with the exception of glutamic acid. The resulting constructs were verified by DNA sequencing.
PRO-BIND ELISA plates (Falcon) were coated with 100 l (10 mg/ ml) of shIL-5R␣-Fc (31). Protein-charged ELISA plates were incubated with 100 l of phage samples in PBS buffer for 1 h at 37°C, and the wells then were washed with PBS plus 0.5% Tween 20. Phage binding was detected with a 1:2500 dilution of horseradish peroxidase-conjugated sheep anti-M13 IgG (Amersham Pharmacia Biotech) and color development with 3,3Ј,5,5Ј-tetra-methyl-benzidine dihydrochloride. This reagent was dissolved in 0.05 M phosphate-citrate buffer, pH 5.0, containing 0.03% sodium perborate (Sigma). The reaction was measured by the absorbance at 450 nm. Null phage, which did not carry any insert, was used as background control.
Competitive phage ELISA on microtiter plates coated with shIL-5R␣-Fc (20 g/ml) were performed to confirm the relative receptor binding affinities of different phage selectants. The experiments were carried out according to Jones et al. (32). A concentration of phage sample, predetermined to elicit 60% signal in titration assays, was incubated with different concentrations of soluble IL-5R␣, ranging from 0 to 100 nM. These phage mixtures were then added to the shIL-5R␣-Fc-coated microwells. Following incubation, microtiter plates were washed thoroughly, and bound phage were detected with horseradish peroxidase-conjugated anti-phage antibody (1:900 dilution). EC 50 values were determined by fit to a four-parameter equation to determine the concentration of competing shIL-5R␣ that resulted in half-maximal phage binding.
Expression and Purification of scIL-5 and Glu 110 Variants from E. coli-The phagemid vector pHage 3, which is derived from M13, was used for protein expression (Maxim Biotech, Inc., San Francisco, CA). Transcription was under the control of the lac promoter. This vector contains an Amber stop codon before Gene III. The scIL-5 or Glu 110 mutant genes were amplified using polymerase chain reaction, with primers that also introduced mutations to the NcoI and BamHI restriction sites, converting them to SalI and ApaI, respectively. The resulting polymerase chain reaction fragment was digested with SalI/ ApaI and ligated into the SalI/ApaI sites of pHage 3 phage vector, yielding pH3-(E110X/A5)scIL-5.
The mutant plasmid was electroporated into TOPP3 cells (Stratagene). The recombinant cells were grown in super broth medium and induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside as described previously (29). The lysate supernatant from the induced culture was dialyzed against PBS buffer (pH 7.4) and then loaded onto two monoclonal anti-IL-5 affinity columns (in series) equilibrated in PBS buffer. The two affinity columns were packed with mAb 2E3-Sepharose 4B matrix and mAb 4A6-Sepharose 4B matrix, respectively. The 2E3 mAb is a neutralizing antibody that blocks IL-5 interaction with the ␣ chain of the IL-5 receptor complex (29,33). Conversely, 4A6 mAb is an activity-neutralizing antibody whose exact binding epitope is presently unknown but that does not block IL-5 interaction with receptor ␣ chain (29). Protein was eluted with 0.1 M glycine, pH 2.5. The final yield of scIL-5 and Glu 110 mutants was analyzed by SDS-10% polyacrylamide gel electrophoresis (PAGE) according to Laemmli (34).
Kinetic Analysis of Receptor Binding by E. coli-expressed scIL-5-Kinetic and equilibrium constants for the interaction between hIL-5R␣ and E. coli-expressed scIL-5 or Glu 110 mutants were measured using a BIACORE X optical biosensor (Biacore Inc., Piscataway, NJ). The monoclonal antibody 4A6 was first immobilized onto the biosensor chip (31). The expressed scIL-5 or Glu 110 variants from bacterial supernatants were anchored noncovalently to the antibody. The binding of various concentrations of soluble hIL-5R␣-Fc (12) to the antibody-anchored scIL-5 or mutant protein was then measured. Alternatively, soluble hIL5R␣-Fc was directly immobilized on the sensor chips, and scIL-5 variants were used as analytes. Conditions for immobilization and the sensor assay were the same as those described by Morton et al. (20). For the calculation of rate constants, the association and dissociation phases of sensorgrams obtained for a series of soluble hIL-5R␣-Fc or scIL-5 variant at different concentrations were fitted to a 1:1 Langmuir model, AϩB 7 AB, to yield, respectively, an association rate constant k on and a dissociation rate constant k off . Data analysis was conducted using the BIAcore evaluation software version 3.0. The equilibrium dissociation constant, K d , was determined from ratios of k off /k on (35).
Concentrations of E. coli-expressed scIL-5 or mutant proteins were determined by Western blot analysis and by quantitative IL-5 ELISA using the monoclonal antibodies 4A6 and TRFK-5 (R & D Systems).
Circular Dichroism-Measurements were made on an Aviv 62A DS Circular Dichroism Spectrometer in a 0.1-cm water-jacketed cuvette. Solution conditions were 10 M protein (pH 7.4), 20 mM sodium phosphate, and 150 mM NaCl at 25°C. Spectra shown are the averages of three scan runs at 24 nm/min with a 5-s response time. Thermal stabilities of scIL-5 constructs were evaluated by monitoring the ellipticity at 222 nm. Temperature was increased in each case at a rate of 30°C/h. JAK2 Phosphorylation Assays-293T cells were a gift from Martin Carroll (Department of Hematology/Oncology, University of Pennsylvania). Cells were maintained in Dulbecco's modified Eagle's medium/10% fetal bovine serum. Cloning of the cDNA for IL-5R␣ was performed as described (37). The ␤ c chain clone was provided by Dr. Richard Cook (SmithKline Beecham). The full-length IL-5 receptor ␣ was amplified by polymerase chain reaction and cloned into the mammalian expression vector pEF-BOS-CS (a gift from Dr. Jerry Langer, Rutgers University). The common ␤ chain was cloned into the mammalian expression vector pCDEF (provided by Dr. Jerry Langer). Human JAK2 cDNA was provided by Dr. Chaim Roifman (Hospital for Sick Children, Toronto, Canada) and was cloned into the expression vector pCDNA3. 293T cells (2 ϫ 10 8 ) were transfected for 6 h with 160 g each of IL5R␣, ␤ c , and hJAK2 DNA samples by using CaPO 4 method. After 48 h, cells were harvested and analyzed by immunoprecipitation and flow cytometry for receptor expression.

RESULTS
Phage Display of Glu 110 scIL-5 Variants-The two glutamic acid 110 residues in the IL-5 dimer form a negatively charged patch on the surface of the cytokine (Fig. 1) and have been proposed from previous Ala replacement mutagenesis studies to be a key part of the pharmacophore for receptor ␣ chain binding that also has been proposed to include Glu 89 and Arg 91 . Recent sequence randomization study of the CD loop region of IL-5 (29) has led to a significant reevaluation of how Glu 89 and Arg 91 participate in ␣ chain binding. The current study was undertaken to more deeply understand the role of Glu 110 through a similar randomization approach.
To perform a rapid and efficient randomization of Glu 110 in the N-terminal domain of (wt/A5)scIL-5, an asymmetric construct reported previously (29), we combined site-directed mutagenesis and phage display technology to screen the IL-5 receptor ␣ chain binding ability. Phage particles displaying Glu 110 scIL-5 variants were detected by mAb 4A6 immobilized on microtiter plates (29). These assays were used to verify that the Glu 110 variants were displayed as proteins on the phage particles and accessible for receptor binding. Fig. 2A shows the phage ELISA results of this initial phage screening for receptor ␣ chain binding. To measure relative binding affinities of Glu 110 phage variants, phage titrations were used, as suggested before (38), to identify phage concentrations that produced a similar signal for all 20 Glu 110 variants. Representative curves are shown in Fig. 2B. Clearly E110D, E110W, and E110Y stand out as having relatively higher binding affinity. Hydrophobic residue replacements show a poor or nonexistent The core structure is that of scIL-5 as modeled from that of wild type IL-5 (11,12) using the program Insight II (Molecular Simulations Inc., San Diego, CA). binding affinity versus the other mutated forms. Competitive phage ELISA was carried out to further evaluate the relative receptor binding affinities for selected Glu 110 variants. The shIL-5R␣ binding affinities of phage displayed (E110W/ A5)scIL-5 and (E110Y/A5)scIL-5 were determined by this assay, giving EC 50 values of 21.2 Ϯ 0.4 and 19.6 Ϯ 0.3 nM, respectively (Fig. 3).
Expression of Glu 110 Mutants in E. coli-To confirm the binding affinities inferred from phage ELISA, we expressed and purified Glu 110 mutants for bioactivity and kinetic analysis. The phage vector pH3-(E110X/A5)scIL-5, in which scIL-5 and Glu 110 mutants were fused to a Gene III leader sequence under the control of a lac promoter, was engineered for protein expression. Soluble scIL-5 protein was detected in the supernatant after isopropyl-1-thio-␤-D-galactopyranoside induction, as expected for recombinant proteins that are expressed and secreted into the periplasm. The construction of (E110W/ A5)scIL-5 is shown in Fig. 4A. The purified, E. coli-expressed scIL-5, (wt/A5)scIL-5 and Glu 110 mutant proteins were analyzed by SDS-PAGE, revealing a single band of approximately 30 kDa (Fig. 4B). Unfortunately, (E110Y/A5)scIL-5 mutant was not sufficiently stable during the expression and purification process to allow for further characterization. Because of the unexpected receptor ␣ chain binding activity detected with (E110W/A5)scIL-5, we also constructed and produced (E110W/ A6)scIL-5, as shown in Fig. 4A. In this construct, the Glu 110 in the domain with CD loop mutated is replaced by Ala, resulting in a protein with no negative charges in the Glu 110 locus (Figs. 1 and 4A). This additional mutant was expressed and purified similarly as the A5 mutant (Fig. 4B).
Protein Stabilities of Glu 110 Mutants-We directly compared the secondary structures and stabilities of E. coli expressed scIL-5 mutants by their CD spectra. As shown in Fig. 5A, the similarity of CD for scIL-5, (wt/A5)scIL-5, (E110W/A5)scIL-5 and (E110W/A6)scIL-5 indicates that the Glu 110 variants likely are folded into secondary structures similar to that of scIL-5. The shapes of the CD spectra are consistent with large ␣-helical contents, consistent with the known structure of IL-5 as a four-helix bundle protein (11). Thermal stabilities of Glu 110 scIL-5 mutants were evaluated by monitoring the effects of increasing temperature on molar ellipticity at 222 nm (Fig. 5B). The melting profiles show that the E110W mutants have a melting temperature similar to that of (wt/A5)scIL-5, although the progressive decrease in negative ellipticity for the E110W mutants, especially (E110W/A6)scIL-5, suggest some instability in these proteins at lower temperatures (31). Importantly, though, all of the mutants were predominantly folded into helix-rich conformations at the temperatures used for subsequent receptor binding and bioactivity assays, namely 25 and 37°C, respectively.
Receptor Binding Activities of Glu 110 Mutants-Kinetics of receptor binding were determined using a sandwich biosensor assay that measured binding of shIL5R␣-Fc to antibody-anchored scIL-5 or Glu 110 mutants. Representative sensorgrams for (E110W/A5)scIL-5 and (E110W/A6)scIL-5 are shown in Fig.  6 (A and B). For all cases, sensorgrams were obtained at a series of shIL5R␣-Fc concentrations, ranging from 25 to 100 nM. The calculated kinetic values are summarized in Table I. There was no significant difference in receptor binding affinity among (wt/A5)scIL-5, (E110W/A5)scIL-5, and (E110W/A6)-scIL-5; the K d values determined were 8.4 Ϯ 0.8, 4.1 Ϯ 0.4, and 5.0 Ϯ 0.2 nM, respectively. The IL-5R␣-Fc affinities for these asymmetric mutants, with a C-terminal half containing a disabled CD loop sequence ( 88 AAAAA 92 ) missing the key recognition residues (Fig. 4A), were all lower by close to an order of magnitude than that for scIL-5 protein, with the major difference being the dissociation rate (Table I). This reduced binding is reminiscent of that found before for monomeric versus wild type IL-5 and may reflect the fact that all of the asymmetric mutants have only a single functional four-helix bundle rather than the two of wild type or scIL-5.
In an alternative methodology, the kinetics of binding of scIL-5 variants to shIL5R␣-Fc was examined with the receptor immobilized directly on the sensor chips of an optical biosensor. This was done to evaluate whether the equivalent binding affinities of the Glu 110 mutants might be skewed by 4A6 mAb anchoring. Overlays of sensorgrams for (E110W/A5)scIL-5 and (E110W/A6)scIL-5 are given in Fig. 6 (C and D), and calculated biosensor-derived binding parameters are summarized in Ta-ble I. The affinities obtained with this alternative configuration do suggest a modestly gradual weakening of affinity upon losing the charges of the Glu 110 residues. However, the differences in affinity are small among the Glu 110 mutants and (wt/ A5)scIL-5. Overall, the biosensor results confirm that the receptor ␣ chain binding affinities of (wt/A5), (E110W/A5), and (E110/A6) forms of scIL-5 are quite similar. The higher affinities seen with the assay of IL-5R␣-Fc binding to 4A6-anchored IL-5 than with IL-5 binding to directly immobilized IL-5R␣-Fc are likely due to the bivalency of the analyte (IL-5R␣-Fc) in the former configuration versus the monovalency of the analyte (IL-5) in the latter.
To further evaluate the bioactivity of Glu 110 mutants, JAK2 phosphorylation analysis was conducted using 293T cells transfected with IL5-R␣-Fc and JAK2. The JAK2 phosphorylation assay for scIL-5 shows a concentration-dependent pattern (Fig. 8). In contrast, this dose-dependent pattern was not found for (E110W/A6)scIL-5 mutant. These results correlate with those of TF-1 proliferation assays and suggest that Glu 110 residue plays a significant role in receptor activation. DISCUSSION In this study, we have used epitope randomization to identify the limits of sequence replaceability of Glu 110 for IL-5 function. Randomization on phage displayed scIL-5 showed that replacements by E110Y and E110W, residues without negatively charged side chains, were only fractionally less effective than E110D in inducing IL-5R␣ binding, despite the previously held view (20,21) that Glu 110 charge formed a key component for receptor ␣ chain recognition. When E110W mutations were expressed in soluble proteins, more quantitative measurement of binding properties than achievable with phage displayed proteins confirmed the effectiveness of E110W mutants in IL-5R␣ binding. However, Trp replacement abruptly reduced bioactivity, except when Glu 110 was present in the neighboring four-helix bundle domain. These results argue that the negative charge at Glu 110 or the hydrogen bonding network provided by the carboxylate of Glu 110 is dispensible for ␣ chain binding but is important for receptor activation. The data further argue that Glu 110 can function in receptor activation from either four-helix domain, that is either within-domain or crossdomain. To what extent flexibility in the residue 110 side chain is important in enabling the interactions between IL-5 and receptor subunits that lead to receptor activation cannot be determined by the current data.
The acceptability of E110W for IL-5R␣ binding, for example in (E110W/A6)scIL-5, contrasts with the significant loss of such binding activity with dual E110A replacement seen previously in wild type IL-5 (20,21). This could suggest that the Trp side chain contributes stabilizing contacts which are lost upon Ala replacement. Two possibilities are -stacking and hydrogen bonding. The former as a sole explanation seems unlikely as judged by the lower replaceability of Glu 110 by Phe (Fig. 2). However, the latter also seems unlikely as a sole determinant given the lower apparent efficacy of Ser and Thr in replacing Glu (Fig. 2). The unique acceptability of W and Y suggests that perhaps it is a combination of hydrogen bonding and -stacking that may enable effective ␣ chain binding by the mutants. Obviously, -stacking is not available with the native sequence Glu 110 , whereas hydrogen bonding is. In this work, bioactivity was measured in two ways, mainly by proliferation of TF-1 cells and also by phosphorylation of JAK2 in recombinant 293T cells for one of the key mutants. For the mutant case compared, (E110W/A6)scIL-5, suppression of bioactivity was noted by concentrations of (E110W/A5)scIL-5 (analyte) to IL-5R␣-Fc that was covalently immobilized on the sensor surface. The increase in response represents the binding of scIL-5; the decay represents the dissociation of bound scIL-5 upon washing with running buffer alone. Association rates and dissociation rates were evaluated separately using a 1:1 Langmuir binding model in BIAcore evaluation software version 3.0. The best fit curves are shown as lines. For dissociation, the initial 25 s of data were fitted using the 1:1 model. D, sensorgrams and data fits for binding by various concentrations of (E110W/A6)scIL-5 to sensor-iommobilized IL-5R␣-Fc. Experimental protocol is as described for C. The lines through the experimental data show global fits of association and dissociation phases using a 1:1 Langmuir binding model in BIAcore evaluation software version 3.0. B, sensorgrams and data fits for binding of IL-5R␣-Fc to (E110W/A6)scIL-5. Experimental protocol is as described for A. C, overlays of sensorgrams showing binding by various both assays. Intriguingly, though, the activity observed with the nonmutated scIL-5 was distinctly different in dose response dependence. For cell proliferation, the EC 50 was in the pM range, in keeping with the previously found discrepancy from the nM affinity of IL-5 for human receptor ␣ and even ␣ϩ ␤ c (9). This discrepancy has been interpreted to suggest that cell proliferation response can occur at fractional receptor occupancy, likely because of amplification of binding through the signaling cascade. In contrast, that the JAK2 phosphorylation response was in the nM range suggests that this early step in the signaling cascade is triggered with a more 1:1 linkage to receptor occupancy. This suggests that the amplification of signal leading to pM proliferation EC 50 values occurs beyond the JAK2 phosphorylation step.
The overall results of this work argue that the negative charge at position 110 plays a role in receptor activation. Two possible mechanisms may be envisioned: (i) Glu 110 makes a direct contact with ␤ c in recruiting the latter into an activation complex with IL-5 and IL-5R␣ and (ii) Glu 110 participates in the orientation of R␣ into an alignment that is productive for recruitment and activation of ␤ c through interactions other than at Glu 110 . It is not possible to distinguish between these two possibilities by the current data. However, the latter seems more likely, because the former would require close overlap of R␣ and ␤ c binding residues at the four-helix bundle interface. One might ask why, if mechanistic choice II were operative, mutations with no Glu at 110 were not observed to have lower R␣ binding affinity (Table I). One reason could be the possibility raised above that Trp might compensate for loss of negative charge interaction through a combination of -stacking and hydrogen bonding. That the negative charge at Glu 110 can play a role in R␣ binding affinity in the absence of such compensating interactions is consistent with the lower binding efficacy effected by simple Ala replacements, in E110A wtIL-5 (20). Intriguingly, the possibility that Glu 110 plays a role in receptor activation was hinted by the nonlinearly reduced bioactivity versus receptor ␣ chain affinity of this E110A wtIL-5 mutation (20).
The results of this work have significant implications for designing IL-5 antagonists of potential use in asthma and other eosinophilia-associated disease (1)(2)(3). The heretofore held view that the receptor ␣ chain-binding epitope was distributed over residues Glu 89 , Arg 91 , and Glu 110 (26) has suggested that the receptor ␣ chain-binding epitope is large and hence that antag-

and E110 variants
The rate constants k on and k off are from BIAcore optical biosensor assays (Fig. 6); the equilibrium dissociation constant K d is calculated from the ratio of k off /k on . TF-1 cell proliferation activities are given as the concentrations required for half-maximal stimulation. The binding parameters and EC 50 values given are the means Ϯ S.D. of three determinations. onizing this interaction with a small molecule would have a low likelihood of success. However, the current data suggest that the Glu 89 -Arg 91 -Glu 110 epitope is divisible into two parts and that each may provide an opportunity for local contact disruption. In this view, interfering with the local CD loop contacts with receptor ␣ chain might be sufficient for effective antagonism of receptor ␣ chain binding. Alternatively, interfering with the Glu 110 local epitope might interfere sufficiently with ␤ c recruitment so that even though ␣ chain binding might persist bioactivity would not result. The combined mutations of Glu 110 and Glu 13 , the latter of which has previously been shown to lead to IL-5 antagonism, may produce a more effective IL-5 antagonist than achievable by mutation of either Glu position alone.