Single chain human interleukin 5 and its asymmetric mutagenesis for mapping receptor binding sites.

Wild type human (h) interleukin 5 (wt IL5) is composed of two identical peptide chains linked by disulfide bonds. A gene encoding a single chain form of hIL5 dimer was constructed by linking the two hIL5 chain coding regions with Gly-Gly linker. Expression of this gene in COS cells yielded a single chain IL5 protein (sc IL5) having biological activity similar to that of wt IL5, as judged by stimulation of human cell proliferation. Single chain and wt IL5 also had similar binding affinity for soluble IL5 receptor alpha chain, the specificity subunit of the IL5 receptor, as measured kinetically with an optical biosensor. The design of functionally active sc IL5 molecule. Such mutagenesis was exemplified by changes at residues Glu-13, Arg-91, Glu-110, and Trp-111. The receptor binding and bioactivity data obtained are consistent with a model in which residues from both IL5 monomers interact with the receptor alpha chain, while the interaction likely is asymmetric due to the intrinsic asymmetry of folded receptor. The results demonstrate a general route to the further mapping of receptor and other binding sites on the surface of human IL5.


Wild type human (h) interleukin 5 (wt IL5) is composed of two identical peptide chains linked by disulfide bonds. A gene encoding a single chain form of hIL5 dimer was constructed by linking the two hIL5 chain coding regions with a Gly-Gly linker. Expression of this gene in COS cells yielded a single chain IL5 protein (sc IL5) having biological activity similar to that of wt IL5, as judged by stimulation of human cell proliferation.
Single chain and wt IL5 also had similar binding affinity for soluble IL5 receptor ␣ chain, the specificity subunit of the IL5 receptor, as measured kinetically with an optical biosensor. The design of functionally active sc IL5 allowed asymmetric mutagenesis of the symmetrical IL5 molecule. Such mutagenesis was exemplified by changes at residues Glu-13, Arg-91, Glu-110, and Trp-111. The receptor binding and bioactivity data obtained are consistent with a model in which residues from both IL5 monomers interact with the receptor ␣ chain, while the interaction likely is asymmetric due to the intrinsic asymmetry of folded receptor. The results demonstrate a general route to the further mapping of receptor and other binding sites on the surface of human IL5.
Human interleukin 5 (hIL5) 1 is a T cell-derived cytokine which plays an important role in the differentiation, proliferation, and activation of eosinophils (Sanderson et al., 1992;Bentley et al., 1992). Natural hIL5 is a disulfide-linked, homodimeric glycoprotein with 115 residues per chain. The high resolution crystal structures of both Escherichia coli-expressed (Milburn et al., 1993) and Drosophila-expressed hIL5 (Johanson et al., 1995) have revealed a core of two four-helix bundles. Each four-helix bundle resembles the four-helix bundle seen in IL2 (Bazan and McKay, 1992), IL4 (Smith et al., 1992), and GM-CSF (Diederichs et al., 1991). However, the bundle organization in IL5 is unique in that helix D of one monomer combines with helices A, B, and C of the second monomer, and vice versa. In the two-bundle structure, the A and D helices form one face, the B and C helices a second. The organization of structural features on each face is palindromic, as observed for restriction sites in DNA.
Human IL5 receptor is composed of two different chains, denoted ␣ and ␤ (Tavernier et al., 1991). The ␣ chain is specific for IL5 (Murata et al., 1992) and has a K d in the 0.1-1 nM range depending on the assays and receptor forms examined (Li et al., 1996). In contrast, the ␤ chain of hIL5R is not cytokine-specific but is shared with the receptors for IL3 and GM-CSF (Tavernier et al., 1991) and appears needed for signal transduction. A soluble form of hIL5 receptor ␣ chain (shIL5R␣), which has only the extracellular domain of IL5R␣, also has been described. This binds to hIL5 with nanomolar affinity (Tavernier et al., 1991;Johanson et al., 1995). Despite the dimeric nature of IL5, a 1:1 binding stoichiometry between shIL5R␣ and hIL5 has been reported (Devos et al., 1993;Johanson et al., 1995).
New insights into the binding mechanism of IL5 to its receptor are emerging with the availability of high resolution structure and mutagenesis techniques. Data from hybrid constructs of mouse/human IL5 suggest that the carboxyl-terminal 36 residues of IL5 interact directly with the IL5R␣ and confer species specificity (McKenzie et al., 1991). By Ala-scanning mutagenesis of the carboxyl-terminal region of hIL5, we found previously that Glu-110 and Trp-111 contribute significantly to receptor binding . In addition, mutation of residues (Glu-89 and Arg-91) in the loop between helices C and D have been found to affect binding to IL5R␣ (Tavernier et al., 1995;Graber et al., 1995). All of these residues cluster around the interface between the two 4-helix bundles of hIL5 and appear to constitute a central patch for binding to a single molecule of hIL5R␣ . In contrast, residue Glu-13, which is at the distal ends of the IL5 dimer away from the helix bundle interface, was suggested to interact with the ␤ chain of IL5R, since mutation at this position resulted in loss of biological activity but did not affect the binding affinity to the ␣ chain (Tavernier et al., 1995;Graber et al., 1995).
While emerging data suggest models for the topography of receptor binding sites in hIL5, a more defined understanding is impeded by the homodimeric nature of the protein. Because of this, mutagenesis of wild type hIL5 inevitably has resulted in symmetrical changes in side chains on both sides of the 4-helix bundle interface. It is thus unclear whether one or both of the residues found to be important participate equally in receptor binding and, furthermore, whether the topology of binding is different for hIL5R␣ versus hIL5R␤. One way to overcome this limitation is asymmetric mutagenesis. Here, we report construction of an active single chain form of hIL5 dimer, denoted sc IL5, and its use to construct asymmetric mutations at residues Glu-110, Trp-111, Arg-91, and Glu-13.

Construction of the Single Chain hIL5 COS Expression Plasmid
Construction of the COS expression vector pCDNIL5 containing the hIL5 coding sequence was described before . For construction of the sc IL5 gene, a DraIII/BglII linker containing the amino-terminal 5 amino acids of hIL5 and 2 glycine residues was mixed with a BglII/SacI fragment (14 kilobases) from pCDNIL5, and the mixture was ligated into pCDN-IL5 which had been digested with * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
DraIII and SacI. The product of the 3-piece ligation, called pCDN-IL5(sc), encodes sc IL5, a monomeric protein containing two tandem hIL5 sequences joined by a 2-amino acid linker, Gly-Gly. The sequence of this construction was verified directly by the dideoxy method using the DyeDeoxy terminator kit from Applied Biosystems.

Mutagenesis
Site-directed mutagenesis was first carried out on pCDN-IL5 by "cassette mutagenesis" (Wells et al., 1985). To make a single site mutagenesis in sc IL5, the mutant pCDN-IL5 was digested with BglII and ligated with the 350-base pair BglII fragment of pCDN-IL5(sc). To make double mutations in single chain hIL5, PCR mutagenesis (Landt et al., 1990) was used to mutate the residue on the amino-terminal half site of sc IL5. The template for the PCR reaction is pCDN-IL5. The resulting PCR fragment was then ligated into the BglII site of mutant pCDNIL5(sc) in which the corresponding residue on the carboxyl-terminal half site had been previously mutated. Presence of the desired mutations and absence of PCR mistakes were verified by DNA sequencing.
Receptor Binding Analysis of Mutants in Crude Expression Supernatants-Kinetic and equilibrium constants for the interaction between hIL5R␣ and different forms of hIL5 were measured using an IAsys optical biosensor (Fisons) in an assay similar to that described previously . Briefly, the non-neutralizing monoclonal antibody 24G9 (Ames et al., 1995) was first immobilized onto the sensor surface. The expressed hIL5 from COS supernatants was anchored noncovalently but tightly to the non-neutralizing antibody. The binding of various concentrations of shIL5R␣ to the attached hIL5 was then measured. All mutants studied here were found to bind to the anchoring antibody as tightly as wt hIL5 and dissociated only very slowly from the antibody within the time of each run of the assay. The very slow dissociation had no significant affect on the binding analysis .
Biological Activity Assays: TF 1 Cell Proliferation-Biological activity was measured using a subclone of the human erythroleukemia cell line TF-1 (subclone TF-1.28), which is highly responsive to recombinant hIL5. For assay, cells were cultured in RPM1 1640 medium supplemented with L-glutamine, penicillin-streptomycin, and 10% heattreated fetal calf serum (Life Technologies, Inc.). Plates with 96 roundbottomed wells were seeded with 5000 cells/well and incubated for 48 h in triplicate in the presence of serially diluted wt IL5 or sc IL5. Cultures were pulsed with 0.5 Ci of [ 3 H]thymidine (Amersham) for the final 4 h and processed for scintillation counting. Data were fitted to a 4-parameter logistic curve (Grafit 3.0), and EC 50 values were calculated. Results are given as the mean of at least 3 determinations. Coefficients of variation were between 5 and 25%. Concentrations of wt IL5, sc IL5, and mutants were determined by Western blot analysis (Johanson et al., 1995) and by quantitative IL5 enzyme-linked immunosorbent assay using the monoclonal antibodies 24G9 and TRFK-5 (R & D Systems).

Design and Expression of Single Chain IL5-
The crystal structure of hIL5 has shown that the carboxyl terminus of one hIL5 monomer is very close to the amino terminus of the second monomer (Milburn et al., 1993). Also, both the amino and carboxyl termini of hIL5 are exposed to the solvent and are very flexible (Milburn et al., 1993;Johanson et al., 1995). Therefore, we joined the two hIL5 monomers together covalently with a peptide linker. Since Gly residues are flexible and often found in protein ␤-turns (Chou and Fasman, 1978), we engineered a single chain hIL5 (sc IL5) by connecting two hIL5 monomers with a Gly-Gly linker (Fig. 1A).
sc IL5 and wt IL5 were expressed in COS cells, and the supernatants were analyzed by SDS-PAGE followed by immunoblotting with a polyclonal anti-hIL5 antiserum . As shown in Fig. 1B, sc IL5 had the same molecular mass (ϳ34 kDa) as that of the wt IL5 under nonreducing conditions. The levels of expression of wt IL5 and sc IL5 were similar. Under reducing conditions, sc IL5 still ran as a dimer (34 kDa) while wt IL5 was reduced to two monomers (ϳ17 kDa) (Fig. 1B). The fact that hIL5 and sc IL5 had the same molecular masses under nondenaturing conditions indicated that both forms were glycosylated similarly and that no intermolecular disulfide bonds had formed in sc IL5.
Receptor Binding Activities of Single Chain IL5-Binding of shIL5R␣ to antibody-anchored wt IL5 or sc IL5 was measured using a sandwich biosensor assay (Fig. 2). The linear portions of the association and dissociation phases of sensorgrams for a series of shIL5R␣ concentrations were analyzed to give k on and k off , as described previously . As summarized in Table I, the k on and k off rates of sc IL5 were very close to those of wt IL5. The K d values of sc IL5 and wt IL5 were also similar (4.2 nM and 4.4 nM, respectively). This indicates that the single chain form of hIL5 retains full binding activity to the hIL5 receptor ␣ chain.
The binding of COS-expressed sc IL5 and wt IL5 to the full length IL5R␣ was also compared by competition for binding of 125 I-IL5 to Drosophila cell membranes containing expressed IL5R␣ (Johanson et al., 1995). Single chain IL5 was equally as effective as wt IL5 in inhibiting the binding of 125 I-IL5 to the cell membranes (data not shown), consistent with the biosensor data, suggesting that sc IL5 and wt IL5 can bind to the same site(s) of full-length IL5R␣ with similar affinity.
The ability of sc IL5 to induce signal transduction was measured by cell proliferation. Single chain IL5 showed activity comparable to that of wt IL5 in the TF-1 cell assay (Table I).
Overall, there were no major differences in either receptor binding or bioactivity of single chain and wild type IL5.
Asymmetric Mutagenesis of Residues Affecting IL5R␣ Binding and IL5R Activation-The design of active sc IL5 made feasible asymmetric IL5 mutagenesis. We chose several residue positions for such mutations in this study to exemplify the approach. The mutations chosen were based on the crystal structure of IL5 and previous results showing a role in receptor binding or signal transduction (Devos et al., 1993;Johanson et al., 1995;Graber et al., 1995;Tavernier et al., 1995;. These residues included Glu-110, Trp-111, and Arg-91 for IL5R␣ binding and Glu-13 for signal transduction but not receptor ␣ chain binding. When Ala residues were substituted in these positions symmetrically in the wt IL5 system, receptor binding and signal transduction activities were observed as shown in Table I bottom, sc IL5. The gene of the single chain IL5 was constructed by linking two hIL5 genes in tandem separated by a spacer that encoded the dipeptide -Gly-Gly-. The amino-terminal half of the molecule is defined as a, the carboxyl-terminal half as b. The wtIL5 contains the full sequence of hIL5 except that the amino-terminal sequence was NH 2 -GARSEIPTSALVKET (Johanson et al., 1995). The amino terminus of IL5 (b) of the single chain IL5 is RSEIPTSALVKET. B, expression of single chain hIL5 (Western blot). The supernatant (10 l) of COS cells transfected with indicated expression vectors (see "Materials and Methods") was quenched with SDS-PAGE loading buffer (with or without 100 mM dithiothreitol), run on a 15% SDS-PAGE gel, and then transferred to Immobilon. The blot was probed with rabbit anti-hIL5 antibodies.

Asymmetric Mutagenesis of sc IL5 for Mapping Receptor Binding Sites
metrical mutants were consistent with previously reported data . Similar overall properties were observed (Table I) for symmetrical mutants made with sc IL5 for G110A and W111A, sc IL5 (G110A (a, b)) and sc IL5 (W111A (a, b)), respectively, although the extents of decrease in receptor binding and bioactivities were less for the sc IL5 mutants.
The effects of asymmetric Ala mutagenesis in sc IL5 for Glu-110, Trp-111, Arg-91, and Glu-13 are shown in Table I. In all cases, with the exception of Glu-13, single site mutations in the b domain of sc IL5 led to small but finite decreases (ϳ50%) in shIL5R␣ binding activity compared with wt IL5. However, for each mutation, the reduction in binding affinity for the asymmetric construct was 4 -19-fold less than that obtained with the corresponding double mutant (for example, sc IL5 (Glu-110 (b)) versus sc IL5 (Glu-110 (a, b))). Consistent with the binding affinity data, asymmetric mutagenesis of or Arg-91 all resulted in an increased EC 50 value compared with that of sc IL5, but these values were 4 -30-fold less than those obtained with the corresponding double mutants.
Mutagenesis of Glu-13 in sc IL5 did not affect IL5R␣ binding activity, as expected from earlier-reported results (Tavernier et al., 1995;Graber et al., 1995), but did cause a marked decrease in biological activity. The effect on activity of the single site E13A mutant in sc IL5 was nearly 10-fold less than that with the corresponding double mutant in wt IL5. Finally, we also formed two asymmetric double mutants in the sc IL5 system, namely (E13A (a), G110A (b)) and (E13A (b), G110A (b)). In the former, the mutations were in one 4-helix bundle, while in the latter, the two mutations were in different bundles. In both cases, the effects on IL5R␣ binding were similar to that seen in the asymmetric E110A-alone mutants, while the decrease in bioactivity was far greater when each bundle contained one mutation than when one bundle had both while the other had none.

DISCUSSION
In this study, we constructed a tethered dimer of hIL5 with a two-Gly linker. This single chain hIL5 has properties very similar to native hIL5 dimer in both receptor ␣ chain binding and biological activity. Single chain hIL5 offers an opportunity The straight lines show a line fit to the linear part of the data. The slopes of these lines give values for k s at each concentration (see Morton et al. (1995)). C, plot of k s versus concentration. The slope of the line to these points gives the association rate constant. D, determination of dissociation rate constant. The dissociation phase of the sensorgram at 60 nM shIL5R␣ in A was replotted as ln(response at time zero of dissociation/response at time n) versus time. The straight line shows the line fit to the early part of the data. The slope of the line gives the dissociation constant.
to study the effect of changes in one of the two monomers, and hence one of the two 4-helix bundles, on properties of hIL5.
Since hIL5 has a 2-fold palindromic symmetry and binds to shIL5R␣ with a 1:1 stoichiometry, we were interested in investigating whether the IL5R␣ binding site is formed by one or both monomers and how the topography of binding sites for receptor ␣ and ␤ chains could lead to signal transduction. Previous mutagenesis studies have mapped the IL5R␣ binding site to residues near the central symmetry axis of the dimer, with residues Glu-110, Trp-111, and Arg-91 being the most important (Tavernier et al., 1995;Graber et al., 1995;. Accordingly, we made asymmetric mutations at these positions on sc IL5. Our data do not fit well to a simple half-site reactivity model, since mutants with residues changed only on one monomer did not have complete wt activity (all showed ϳ50% lower affinity). Also, the increased K d values caused by these mutations were mainly due to an increase in the dissociation rate k off (Table I). This is not expected for a half-site reactivity model, since the interaction face between ligand and receptor should remain the same for an ideal halfsite model. However, each of the single-site mutants showed a much weaker decrease in IL5R␣ binding than the double-site mutants. This result argues against the possibility that residues on both monomers are equally involved in IL5R␣ binding. Our data are consistent with a model in which residues from both monomers form a central shared patch to interact with IL5R␣ while the asymmetric receptor ␣ chain uses structural elements from this patch asymmetrically to stabilize the IL5-IL5R␣ complex. However, we do not believe it possible to conclusively define the detailed way in which ␣ chain binding occurs from this first asymmetric mutagenesis study. Even less can be ventured for ␤ chain binding from the E13A mutants, since we do not yet know the stoichiometry of the IL5-IL5R␤ interaction.
A significant caveat in any interpretation of our asymmetric mutagenesis data is that we cannot be certain to what extent mutations may affect the folded structure of sc IL5. At least for the E110A and W111A mutants, several observations argue that gross conformational changes are unlikely. 1) Symmetrical mutants for these residues in wt IL5 have stabilities similar to that of native sequence wt IL5 . 2) Single chain IL5 had receptor binding and signal transduction activities equivalent to those of native sequence wt IL5. 3) The asymmetric mutants in sc IL5 for E110A and W111A were likely no more perturbed structurally than the double mutants. Furthermore, all of the mutants reported here were found to bind to the monoclonal antibody 24G9; this antibody appears to be conformationally dependent since it does not bind to reduced and alkylated hIL5 (data not shown). Hence, it is likely that the sc mutants we have made so far are not grossly unfolded. Nonetheless, we cannot rule out more subtle conformational changes as a factor in altered functional properties. This must await more deliberate structural analysis of purified mutants, a planned objective of future work.
In conclusion, we have designed an IL5 system for asymmetric mutagenesis and have exemplified the approach of asymmetric mutagenesis with sc IL5. The results obtained suggest a model for IL5-IL5R␣ receptor recognition in which residues from both 4-helix bundle domains contribute to the binding of a single molecule of IL5R␣, possibly by formation of a central patch as suggested previously . However, the receptor may use this symmetrical patch asymmetrically to stabilize the IL5-IL5R␣ complex due to the asymmetry of the receptor molecule. It remains for further mutagenesis and, importantly, structural analysis of key mutants, to deduce a more refined and certain understanding of binding site topography for IL5R␣ and ultimately the way this topography leads to signal transduction through IL5R␤.