Spatial orientation of the alpha and betac receptor chain binding sites on monomeric human interleukin-5 constructs.

Interleukin-5 (IL-5), a disulfide-linked homodimer, can be induced to fold as a biological active monomer by extending the loop between its third and fourth helices (Dickason, R. R., and Huston, D. P. (1996) Nature 379, 652-655). We have designed eight monomeric IL-5 proteins to optimize biological activity and stability of the monomer. This was achieved by (i) inserting the joining loop at three different positions, (ii) by introducing an additional intramolecular disulfide bridge onto these backbones, and (iii) by creating circular permutations to fix the position of the carboxyl-terminal helix relative to the three other helices. The proteins dimerize with Kd values ranging from 20 to 200 microM and are therefore monomeric at the picomolar concentrations where they are biologically active. Introduction of a second disulfide confers increased stability, but this increased rigidity results in lower activity of the protein. Contrary to wild type IL-5, mutation of the betac contact residue on the first helix, Glu12, to Lys, into the circularly permutated constructs, did not abolish TF-1 proliferative and eosinophil activation activities. These results indicate that activation of the IL-5 receptor complex is not mediated solely by Glu12 on the first helix, and alternative mechanisms are discussed.

Interleukin-5 (IL-5) 1 is the key cytokine involved in the differentiation and maturation of eosinophil precursors and the activation and survival of mature eosinophils (1)(2)(3)(4)(5). The association of eosinophilia with chronic inflammatory conditions such as asthma, rhinitis, and atopic dermatitis (6 -8) indicates that blocking the action of IL-5 may provide therapeutic benefit in these allergic disorders. Indeed, neutralizing antibodies to IL-5 have been shown to reduce pulmonary eosinophilia, tissue damage, and bronchial hyperactivity in animal models of asthma (9 -11). Experiments with mice in which the IL-5 gene has been deleted have further validated the central role of IL-5 in eosinophilia (12).
IL-5 was originally identified from a murine T cell culture supernatant (13) and was shown to be a disulfide-linked ho-modimer consisting of two glycosylated subunits (14). Glycosylation is not required for activity, as the human recombinant protein produced in Escherichia coli is fully active (15). The three-dimensional crystal structure of the E. coli-derived protein showed that the disulfide-linked dimer forms two domains, each containing four helices which pack with the cytokine fold (16). The cytokine fold is common to many other cytokines including granulocyte-macrophage colony stimulating factor (GM-CSF) and growth hormone (17). However, all the other cytokines are monomeric. The dimeric topology of IL-5 is unique in that each four helix bundle consists of three helices from one subunit, whereas the fourth is provided by the other subunit.
IL-5 binds to a heterodimeric receptor complex composed of an ␣-chain that binds IL-5 uniquely and a common signaling chain ␤ c , which is also a component of the GM-CSF and IL-3 receptors (18,19). Eosinophils bear 300 -1000 such binding sites, to which IL-5 binds with K d of 200 -400 pM (20,21). Surprisingly, given the homodimeric structure of IL-5, it can bind to the receptor ␣-chain to form a 1:1 complex of IL-5 dimer to receptor monomer (22,23). Residues of both IL-5 and the IL-5 receptor ␣-chain required for receptor-ligand interaction have been mapped by extensive mutagenesis studies (24 -27). The residues of IL-5 that bind the ␣-chain are found in the carboxyl-terminal region as follows: Glu 109 and Trp 110 on the fourth helix and Glu 88 and Arg 90 on the ␤-sheet preceding it, while activation of the ␤ c -chain has been shown to be transmitted by a single glutamic residue on the first helix. The two other cytokines that share the ␤ c -chain similarly have a Glu on the first helix as a ␤ c -chain contact, Glu 21 for GM-CSF (28) and Glu 22 for IL-3 (29). Mutating this acidic residue in the IL-5 protein to glutamine which is polar (25) or the positively charged lysine (30) and to arginine in GM-CSF (31) results in protein antagonists that are able to bind the ␣-chain but are unable to activate the signaling ␤ c -chain.
Asymmetric mutagenesis of single chain IL-5 molecules suggests that a single IL-5 domain suffices for biological activity but that optimal binding of the ␣-chain may require residues from both of the four helix bundles (32,33). Extension of the loop linking the third and fourth helices of IL-5 in a manner analogous to GM-CSF allows the protein to fold as a monomer (34). However, this monomeric IL-5 was 15-fold less active than the wild type molecule. This may reflect either a lack of receptor interaction points normally contributed by the second four helix bundle or reduced stability of the monomeric protein.
We have shown that in single chain dimers of IL-5 only a single copy of the receptor ␣-chain contact residues, Arg 90 and Glu 109 , is necessary to attain full biological activity. This observation led us to design a series of monomeric IL-5 constructs that we have used to separate effects of protein stability from those that may be due to interaction of the second domain of IL-5 with its receptor. We also created monomers stabilized by an additional disulfide bond and two circularly permutated monomeric proteins designed to introduce tighter packing of helix D with the other three helices. When the ␤ c -chain blocking mutation, E12K, was introduced into the circularly permutated constructs, monomeric IL-5 proteins were found to retain agonist activity in in vitro bioassays. This is in marked contrast to the results observed for the wild type protein containing this mutation and suggests that residues other than E12K play a role in ␤ c -chain activation and that the E12K mutant may act by introducing negative interactions with the ␤ c -chain rather than removing positive interactions.

EXPERIMENTAL PROCEDURES
Reagents-Unless otherwise stated, all chemicals were purchased from Sigma. Enzymes were from New England Biolabs, and chromatographic material was from Pharmacia Biotech Inc.
Construction, Expression, and Purification of IL-5 Mutants-IL-5 constructs were made using a synthetic gene coding for the mature sequence of human IL-5 (35) which begins with sequence encoding MTEIP . . . , where T is Thr 22 of the human IL-5 precursor. Residue numbering begins with the amino-terminal methionine. The single chain dimer was constructed by the insertion of a Gly residue between two copies of the synthetic human IL-5 gene using overlap PCR. PCR products corresponding to the single chain length were subcloned into pLT4ex4 and expressed as described previously (15). Point mutations were inserted into one IL-5 gene of the single chain by using previously mutated IL-5 genes as template in the second of the two monomer PCRs.
Monomers were constructed by inserting sequences encoding the sequence (S/C)PPTEPTS, which corresponds to residues Ser 105 to Ser 112 of GM-CSF, just after Gln 81 (IL-5.1 and 5.2), just following Lys 84 (IL-5.3 and 5.4), or in place of Lys 82 -Lys 84 (IL-5.5 and 5.6). In the even numbered constructs (IL-5.2, IL-5.4, and IL-5.6), two cysteine residues have been inserted to allow a potentially stabilizing disulfide bond to form between the first residues of the insert sequence and a cysteine introduced by an I112C mutation at the carboxyl-terminal end of the protein.
Insertion mutants and circular permutations were made by megaprimer and overlap extension PCR, respectively. In all cases, the genes were inserted into NcoI/HindIII-digested pET23d (Novagen) and expressed in E. coli BL21(DE3).
After the first TF-1 proliferation bioassay, the IL-5.3 backbone was identified as producing the most active protein and the mutants IL-5.3, IL-5.4, IL-5.cT29, and IL-5.cT63, which all had the loop inserted after Lys 84 , were subsequently fermented in 5-liter fermentors for further characterization. The proteins were purified from inclusion bodies and renatured as described for the recombinant wild type protein (15,36). Renaturation was also carried out by rapid dilution of the purified protein in 6 M guanidine/HCl into 0.1 M Tris/HCl, pH 8.5.
Disulfide Bond Determination-Disulfide bond formation was determined by the analysis of the amino acid composition of the proteins after alkylation of reduced and oxidized samples and derivatization of free Cys residues with dithiodipropionic acid (37). The proteins were hydrolyzed at 112°C for 24 h, and the analysis was carried out using the Waters AccQ.Tag Chemistry Package.
Analysis of Apparent Molecular Weight and Aggregation State-The proteins were analyzed by gel permeation chromatography using a SMART system equipped with a Superdex 75 column equilibrated with 0.1 M Tris/HCl, pH 8.5, containing 0.15 M NaCl. The column was calibrated with standards of known molecular weights, as well as recombinant IL-5 and GM-CSF. 50 l were applied at concentrations between 35 and 150 g/ml.
Sedimentation equilibrium analytical ultracentrifugation of native IL-5, the various IL-5 mutants, and GM-CSF was performed using a Beckman XL-A (Palo Alto, CA) centrifuge with six-channel 12-mm charcoal-filled epon centerpieces. Runs were performed at 25,000, 30,000, 32,500 and 35,000 rpm at 4°C with scans taken at 220 or 280 nm at 1-h intervals. Equilibrium was judged to be achieved by the absence of change between plots of several successive scans after approximately 20 h. 100 l of each sample in 100 mM Tris/HCl, pH 8.0, was centrifuged against 120 l of the equivalent buffer blank. Solvent density was determined empirically at 4°C using a Mettler DA-110 density/specific gravity meter calibrated against water. The partial specific volume of each protein was calculated using the method of Cohn and Edsall (38). Temperature differentials were incorporated using the appropriate equation modified from values of each amino acid at 25°C (39). Data sets were obtained as radial distance versus absorbance and later converted to concentration units using an empirically derived extinction coefficient. Raw data were analyzed by the Beckman/Microcal Origin non-linear regression software package using multiple iterations of the Marquardt-Levenberg algorithm (40) for parameter estimation or by global fitting routines kindly provided by the National Analytical Ultracentrifuge Facility at Storrs, CT.
Analysis of Secondary Structure-Circular dichroism (CD) spectral analysis was performed using an Aviv model 62DS CD spectropolarimeter. Proteins were scanned repetitively in 0.1-cm quartz cuvettes from 199 to 340 nm in 1-nm wavelength increments. Ellipticity was converted to molar ellipticity for comparisons.
Thermal Stability Determination Using Circular Dichroism-Thermal transitions were analyzed with the CD instrument described above by monitoring the proteins at 222 nm over a temperature range of 5-95°C. Data were collected in 1°C increments with a slope of 10°C/ min. The half-point of the thermal transition, T1 ⁄2 , was determined by iterative fitting using the Boltzmann equation. Data were fitted to the following thermodynamic model (see Equations 1 and 2).
and where T is ellipticity at T (kelvin), N is the native protein ellipticity, D is the unfolded protein ellipticity, R is the gas constant, T D is the temperature at which the protein unfolding transition is halfcomplete, ⌬H TD is the enthalpy change at T D , and ⌬c p is the heat capacity change.
In Vitro IL-5 Bioassays-The mutants were assayed for activity in two bioassays, induction of TF-1 proliferation and eosinophil activation as measured by adhesion to recombinant VCAM-1 (41). Their affinity for receptor binding was measured in equilibrium competition binding assays with recombinant ␣-chain and TF-1 cells expressing the heterodimeric receptor complex (42). All assays were carried out as described (24).

RESULTS
Protein Design-Increasing the length of the loop between helices C and D of IL-5 to resemble the loop of GM-CSF allows IL-5 helix D to fold back onto its parent monomer (Fig. 1). The amino acid sequences used for creating the monomeric IL-5s are depicted schematically in Fig. 2. In each case 8 amino acids, corresponding to the loop separating helices C and D in GM-CSF, were introduced into the analogous position in IL-5. The sequence (S/C)PPTEPTS was inserted just after Gln 81 (IL-5.1 and IL-5.2), following Lys 84 (IL-5.3 and IL-5.4) or in place of Lys 82 -Lys 84 (IL-5.5 and IL-5.6). In the even numbered constructs (IL-5.2, IL-5.4, and IL-5.6) two cysteine residues have been inserted to allow a potentially stabilizing disulfide bond to form between the first residue of the insert sequence and a cysteine introduced by an I112C mutation at the carboxylterminal end of the protein. Circular permutations were made as an alternative to the cystine bridge in an attempt to increase protein stability by restricting movement of helix D. The new amino termini were introduced at Thr 29 and Thr 63 , residues located on exposed loops at regions known not to be involved in receptor binding (24,25). Finally, we attempted to completely eliminate any possibility of dimerization by mutating the polar residues Thr 42 and Asn 43 that are involved in hydrogen bonding at the dimeric interface to the acidic Glu to create electrostatic repulsion.
Protein Characterization-Wild type IL-5 is expressed to 15% total protein in E. coli. The monomeric constructs expressed to approximately 40% total protein as shown in Fig. 3. They were easily purified and renatured from inclusion bodies giving yields of 8 mg/g E. coli cells compared with 1 mg/g for wild type IL-5. In fact, they could be renatured by a simple dilution from the guanidine denaturant, a process which was not feasible for the wild type dimer. 2 However, the yield using this procedure was lower than the longer protocol for the dimeric protein, which was therefore used for scale up purifications. The introduction of the charge reversal mutation at the ␤ c -chain binding residue, Glu 12 , to Lys, in the wild type dimeric IL-5 protein, caused an approximately 20-fold decrease in the expression level (lane 3, Fig. 3). However, when this mutation was introduced into the IL-5.cT29 and IL-5.cT63 circularly permutated constructs, the expression level was significantly higher than that observed for the E12K mutation in the wild type protein.
Correct folding of the purified proteins was established by CD spectroscopy. CD spectra (200 -300 nm) indicated that the helical content of all the monomeric constructs was very similar to the dimeric protein (Fig. 4). Introduction of the charge reversal mutation, corresponding to E12K in the IL-5 sequence, into the circularly permutated constructs does not appear to perturb the overall secondary structure elements, as the spectrum for IL-5.cT63(E12K) overlays well with the other spectra as shown in Fig. 4.
Disulfide bond formation was measured by analysis of amino 2 A. Proudfoot, unpublished results. acid composition. All of the constructs had one disulfide bond. The three constructs, IL-5.2, IL-5.4 and IL-5.6, which had an additional pair of Cys residues introduced with the aim of forming a disulfide bridge analogous to the second disulfide bond in GM-CSF, were found to effectively contain this second disulfide. Nonreducing SDS-polyacrylamide gel electrophoresis demonstrated that there was no formation of inter-molecular disulfide bonds (results not shown).
The quaternary state of the IL-5 constructs was analyzed by size-exclusion chromatography and analytical ultracentrifugation. When subjected to gel filtration at concentrations around 1 mg/ml (approximately 70 M), the proteins eluted at the volume observed for IL-5 indicating that they were associating as dimers (data not shown). However, at concentrations of 150 g/ml or less (Ͻ10 M), the proteins co-eluted with GM-CSF, indicating that a weak monomer-dimer self-association was occurring at concentrations well above the concentrations at which IL-5 exhibits biological activity. To further refine these observations, analytical centrifugation was employed to determine dissociation constants (K d ) for these interactions ( Table  I). The insertion of the loop after Gln 81 in IL-5.1 and IL-5.2 produced proteins that had the least tendency to dimerize, since they had K d values of 400 M, whereas the other monomers analyzed had dissociation constants between 20 and 60 M. The circular permutations shared intermediate dissociation constants, IL-5.cT29 having a K d of 200 M and IL-5.cT63 a K d of 67 M. In each case the K d for dimerization is far greater than the concentrations at which biological activity was measured. Replacement of the polar Thr 42 and Asn 43 residues located at the dimer interface of wild type IL-5 with Glu did not prevent dimerization since both the single T42E and N43E mutations as well as the double T42E/N43E mutants all showed a K d of 40 M. This suggests that the dimerization process in these mutants may involve a difference in the quaternary packing compared with the wild type protein.
The stability of four monomeric constructs determined by thermal denaturation followed by CD showed that the selffolding monomers were significantly less stable that the parent IL-5 dimer and GM-CSF monomer. The T1 ⁄2 values for IL-5.3, IL-5.cT29, and IL-5.cT63 were 58.8, 51.6, and 55.9°C respectively, compared with 71.1°C for IL-5 and 71.4°C for GM-CSF. However, the introduction of the second disulfide inferred increased stability to the protein, as demonstrated by the T1 ⁄2 of 64.9°C obtained for IL-5.4.
Biological Activity-The single chain protein was almost equipotent to wild type IL-5 in the TF-1 proliferation assay, where it had an EC 50 of 2.8 pM compared with 1.6 pM for the wild type. Introduction of a single R90A or E109A mutation at   (Fig. 5).
The insertion of the 8 amino acid loop that enabled IL-5 to fold as a monomer resulted in proteins that elicit full biological activity in TF-1 proliferation, with a single exception, IL-5.cT29, where the protein was a partial agonist in the range of concentrations tested. The results are summarized in Table I. The most favorable position for insertion of the loop was after Lys 84 . IL-5.3 showed the highest activity in the TF-1 proliferation assay, with an 11-fold decrease compared with the wild type. This backbone was therefore chosen for the design of the circular permutations. The effects of inserting the loop after Gln 81 or by replacing Lys 82 -Lys 84 are small with respect to biological activity, as IL-5.1 and IL-5.5 had 17-and 44-fold increases in EC 50 values in the TF-1 proliferation assay, respectively. Although the creation of the second disulfide bond was favorable in terms of stability, these more rigid conformations were not advantageous with respect to bio-activity; IL-5.2, IL-5.4, and IL-5.6 had EC 50 values 3-8-fold higher than their parent constructs. The circular permutations similarly were active in this assay, where IL-5.cT63 had an EC 50 33-fold higher than the wild type, but IL-5.cT29 was consistently only a partial agonist. Mutation of the polar residues Gln 42 and Thr 43 to Glu into the IL-5.5 backbone had little effect in this assay.
Eosinophil activation was used as a second in vitro bioassay. The monomers were more active in their capacity to induce eosinophil adhesion when compared with wild type IL-5 than in the TF-1 proliferation assay. IL-5.3 had an EC 50 of 5.3 nM and is only 4-fold less active than wild type IL-5, which has an EC 50 of 1.3 pM, and again the second disulfide introduced in IL-5.4 results in an approximately 4-fold drop in activity, with an EC 50 of 20 pM. Both circular permutation constructs exhibited full agonist activity in the induction of eosinophil adhesion, and in fact IL-5.cT63, with an EC 50 of 3.4 pM, was almost as potent as the wild type protein.
Receptor binding assays showed that the reduction in affinity of the monomeric proteins for either the recombinant ␣-chain or the ␣⅐␤ complex was significantly larger than the reduction in potency in biological assays. The IC 50 values for competition of 125 I-IL-5 were between 100-and 1000-fold higher than the wild type IL-5. IL-5.3 had an IC 50 140-fold larger than the wild type in the SPA assay for ␣-chain binding, which competes for 125 I-IL-5 with an IC 50 of 1.2 nM and a 210-fold increase for competition for the ␣⅐␤ complex, where wild type IL-5 competes for 125 I-IL-5 with an IC 50 of 0.16 nM. The introduction of the second disulfide into this backbone in IL-5.4 caused an additional 10-fold reduction in both binding assays. Of the two circular mutations, IL-5.cT63 showed a 10-fold greater affinity to both the ␣-chain and to the ␣⅐␤ complex than IL-5.cT29 compared with wild type IL-5. IL-5.cT63 had a 70-fold lower affinity for the ␣-chain and 135-fold lower for the ␣⅐␤ complex, whereas IL-5.cT29 showed decreases in affinity of 800 and 1300, respectively.
The introduction of the ␤ c contact residue Glu 12 to Lys mutation in the circular permutations did not abrogate agonist activity in either TF-1 proliferation or the induction of eosinophil adhesion, contrary to the results obtained by this mutation in the wild type protein (Figs. 6 and 7). IL-5.cT63(E12K) was a partial agonist in both assays, whereas IL-5.cT29(E12K) showed partial agonist activity in the TF-1 proliferation assay but was a full agonist with an EC 50 of 910 nM in the induction of eosinophil adhesion. The introduction of this mutation into IL-5.cT29 had very little effect on its binding to the receptor, whereas IL-5.cT63(E12K) showed a 15-fold decrease in affinity for the ␣-chain and a 10-fold increase in IC 50 for the competition of 125 I-IL-5 from the ␣⅐␤ complex on TF-1 cells compared with their parent constructs.  Table I and are the results of three separate experiments.

DISCUSSION
The residues of IL-5 contributing to binding the specific ␣-chain of the IL-5 heterodimeric complex have been identified by extensive alanine scanning mutagenesis studies (24,25) and used to define the spatial location of essential groups, or pharmacophore, of human IL-5. These residues are located in the carboxyl-terminal region, a region previously shown to be responsible for the specificity between human and murine species (43). Glu 88 and Arg 90 are located on the ␤-sheet linking the third and fourth helices, and Glu 109 and Trp 110 are located toward the distal end of the fourth helix. In the three-dimensional structure solved for the E. coli protein (16), Trp 110 is buried. Mutation of Trp 110 to Ala is therefore thought to affect the orientation of Glu 109 . However, the approximate distances between the other functional side chains can be determined. Glu 88 and Arg 90 are separated by 8 Å, and both are 26 Å from Glu 109 on the same subunit but also from the Glu 109 residue located on the other subunit. Because the original mutagenesis studies were carried out on the wild type protein, it is impossible to differentiate which Glu 109 is involved in the pharmacophore.
Several lines of evidence have indicated that both four helix bundle domains may be not essential for bio-activity. First, the ␣-chain binds to the IL-5 dimer with a 1:1 stoichiometry as demonstrated using the recombinant receptor in an in vitro binding assay (22). Second, the construction of fully active single chain IL-5 proteins has allowed single copy mutations of the residues involved in ␣-chain binding (32, 33, and this work). We have shown here that at least for TF-1 proliferation, despite the close proximity in the three-dimensional structure of the Glu residues located at the distal ends of the fourth helices, receptor activation requires only one of the Glu 109 residues. Similarly, only one of the Arg 90 residues is required. Since we were unable to produce IL-5 in E. coli bearing the E88A mutation, we did not investigate the effect of producing it in single copy. Several combinations of alanine mutations of both the ␣-and ␤ c -chain binding sites have been made (32,33), and these mutants were analyzed for their effects on binding to the ␣-chain and their capacity to induce TF-1 proliferation. Although mutation of the ␣-chain binding site residues in single copy lowers affinity for binding to the receptor, a factor attributed to increases in the dissociation constant, k off , the effects on TF-1 proliferation are small suggesting that a single domain is sufficient for biological function.
IL-5 has been induced to fold with a monomeric topology by extension of the loop linking the third and fourth helices in a manner analogous to GM-CSF (34) as depicted in the model shown in Fig. 1. However, this protein was shown to have a 15-fold lower activity as measured by TF-1 proliferation compared with wild type dimeric IL-5, and little is known about the stability and oligomerization of the monomeric protein. We were interested to see if a fully active monomer could be formed by improving the packing of the fourth helix in the IL-5 monomer. The fourth helix carries one of the essential ␣-chain binding sites, Glu 109 , but has also been shown important in maintaining the integrity of the four helix bundle structure. Successive removal of the first two helical turns in the dimeric protein causes significant losses in activity, which are correlated to extensive changes in the structure, rather than the loss of the residue Glu 109 (44). Monomeric IL-5 was made by inserting sequence encoding the loop between GM-CSF helices C and D, Ser 105 to Ser 112 , into the analogous location of IL-5 in a region corresponding to the junction of exons 3 and 4 in IL-5. To obtain maximal packing of helix D, we engineered a disulfide bond designed to maintain a close packing of the helix D in the four helix bundle, as is found in GM-CSF. Finally, we made two circular permutations of the monomeric IL-5. This results in covalent attachment of helix D to the beginning of helix A.
It is obvious from the previously published report and this work that the principal factor required to achieve activation of the IL-5 receptor complex is the packing of the four helices into the "cytokine fold" so that the ␣and ␤ c -chain binding sites are correctly oriented to interact with their respective receptor subunits. Subtle effects can be seen on the positioning of the loop into the primary IL-5 sequence. Insertion of the loop after Lys 84 resulted in the most active protein, since it was 11-fold less active than the wild type in TF-1 proliferation and 5-fold less in eosinophil adhesion. Using this backbone to create the circular permutations, one, IL-5.cT63, was only 3-fold less active than wild type IL-5 in the eosinophil adhesion assay. In general, higher activities for all the constructs were observed in the eosinophil adhesion assay than in the TF-1 proliferation. It could be reasoned that in view of their lower stability compared with the native dimer, the short incubation time of 30 min in the adhesion assay is more favorable than the 3-day assay of TF-1 proliferation, during which time protein destabilization and degradation could easily occur.
The receptor binding assays may also reflect the lower stability of the monomeric proteins, presumably due to a suboptimal packing. Although their overall conformation resembles that of IL-5 as demonstrated by circular dichroism, the monomers are over 100 times less efficient at competing for 125 I-IL-5 (with a single exception, IL-5.cT63 which has a 70-fold increase in IC 50 ) on both the ␣-chain and the ␣⅐␤ complex. Equilibrium competition of the iodinated ligand from the receptor may be considered to be more demanding on structure,  Table I and are the results of three separate experiments. especially in view of the fact that activity is triggered by picomolar concentrations, whereas competition occurs at nanomolar concentrations.
By introducing the charge reversal E12K into the IL-5 sequence, we have produced a potent antagonist of both TF-1 proliferation and eosinophil adhesion (30), but we were hampered in our attempts to continue our studies in animal models of allergic disorders by the difficulty of producing the protein in E. coli. Attempts to express this mutant at a high level in a baculovirus expression system were similarly unsuccessful. We were therefore interested in using the monomeric scaffolds, in particular the circular permutations where the Glu residue in question was no longer proximal to the amino terminus of the sequence, as a means of obtaining large amounts of the antagonist. Although the permutated proteins possess the characteristics necessary to confer IL-5 activity, introduction of this mutation surprisingly did not abolish activity. We had previously made the observation that the E12K mutation in the wild type protein, while creating a potent antagonist against IL-5 induced TF-1 proliferation and eosinophil adhesion, retained the ability to induce eosinophil survival, albeit with a 50,000fold reduction in potency (30). This suggested that in the eosinophil there may be separate signaling pathways involved in adhesion and survival. Moreover, there may be other residues that are involved in activating and triggering the ␤ c -chain or, alternatively, that the ␣-chain itself may signal in the induction of eosinophil survival. However, in the two constructs described in this work bearing this charge reversal mutation activity is achieved in the two functional assays, TF-1 proliferation and induction of eosinophil adhesion, for which the wild type E12K mutant was devoid of activity.
We believe that these results support the hypothesis that the Glu residue on the first helix is not the only point of contact required for ␤ c activation. This argument is strengthened by the fact that mutations of the Glu residues identified as being ␤ c contact points for GM-CSF and IL-3, also on the first helix, have been found to retain their capacity as agonists. In the case of IL-3, the charge reversal mutation E21R retains agonist activity to induce TF-1 proliferation, with a 20,000 reduction in potency (45). The mutation E22A in murine GM-CSF likewise does not abrogate agonist activity, whereas the E22R mutation in human GM-CSF is devoid of agonist activity and produces a specific GM-CSF antagonist. Taken together, these results imply that Glu 12 residue does not directly transmit a signal to ␤ c per se but that this residue is more probably involved with contacts allowing the conformational change to take place in ␤ c that is required for this subunit to interact with the next component in the transduction pathway. Recent evidence has been proposed for such a conformational change by mutagenesis studies of ␤ c , revealing that disruption of the interaction of certain hydrophobic residues by their mutation to polar residues results in constitutive activation of ␤ c (46). These authors propose that such a mutation forces ␤ c into the conformation that would be normally induced by its interaction with the ␣-chain-ligand complex. It is therefore possible that the circularly permutated IL-5 monomers bearing the charge reversal mutation have a structure that is sufficiently similar to the IL-5 four helix bundle to recognize the ␣-chain and yet is sufficiently modified so that the charge reversal does not impair the subsequent conformational change required by ␤ c .
While not having elucidated why IL-5 is the only four helix bundle cytokine that is dimeric, we believe that the characterization of the active monomeric constructs described here support the hypothesis that a change in splicing pattern between the 3rd and 4th exons probably gave rise to the dimeric topology of IL-5 during the course of evolution but that the gene continued to evolve so that the dimeric configuration is thermodynamically favored over a monomeric structure. In addition, the monomeric forms in which the amino termini have been permutated are not rendered inactive by a charge reversal of the Glu on helix A, the only ␤ c contact site identified to date. We believe that these monomeric mutants will prove useful in studies of the signaling pathways by which IL-5 activates ␤ c and elicits its effects on its target cell, the eosinophil, often considered to be one of the major factors causing tissue damage in the late phase of asthma.