Receptor and antibody interactions of human interleukin-3 characterized by mutational analysis.

Human interleukin-3 (hIL-3) is a regulator of proliferation and differentiation of multipotent hemopoietic progenitor cells. Mutants of hIL-3 have been constructed by oligonucleotide-directed mutagenesis and expressed in Escherichia coli and Bacillus licheniformis. Purified muteins were assayed for induction of DNA synthesis in IL-3-dependent human cells and for binding to the IL-3 receptor. Residues at the NH2 and COOH termini together comprising one-quarter of the molecule could be removed without loss of biological function. Deletions of 6-15 residues within the central part of the molecule caused a large reduction (up to 5 logs) but no complete loss of activity. Substitution of evolutionary conserved residues resulted in a strong decrease of biological activity and demonstrated that the S-S bridge is an essential structural element in hIL-3. Interestingly, four muteins displayed a significantly higher potency of binding to the IL-3 receptor than in stimulating DNA synthesis. These results demonstrate that receptor binding may be (partly) disconnected from activation of DNA synthesis. Analysis of hIL-3 muteins demonstrated that the majority of monoclonal antibodies are directed against a small portion of the IL-3 molecule. The neutralizing potential of individual monoclonal antibodies could be increased by a combination of antibodies directed against nonoverlapping epitopes.


Human interleukin-3 (hIL-3) is a regulator of proliferation and differentiation of multipotent hemopoietic progenitor cells. Mutants of hIL-3 have been constructed by oligonucleotide-directed mutagenesis and expressed in Escherichia coli and Bacillus licheniformis.
Purified muteins were assayed for induction of DNA synthesis in IL-3-dependent human cells and for binding to the IL-3 receptor. Residues at the NH2 and COOH termini together comprising one-quarter of the molecule could be removed without loss of biological function. Deletions of 6-15 residues within the central part of the molecule caused a large reduction (up to 5 logs) but no complete loss of activity. Substitution of evolutionary conserved residues resulted in a strong decrease of biological activity and demonstrated that the S-S bridge is an essential structural element in hIL-3. Interestingly, four muteins displayed a significantly higher potency of binding to the IL-3 receptor than in stimulating DNA synthesis. These results demonstrate that receptor binding may be (partly) disconnected from activation of DNA synthesis. Analysis of hIL-3 muteins demonstrated that the majority of monoclonal antibodies are directed against a small portion of the IL-3 molecule. The neutralizing potential of individual monoclonal antibodies could be increased by a combination of antibodies directed against nonoverlapping epitopes.
Interleukin-3 (IL-3)' is a hemopoietic growth factor that regulates proliferation and differentiation of immature blood cell progenitors (Metcalf, 1986;Clark and Kamen, 1987;Wagemaker et al., 1990a). Murine IL-3 was shown to stimulate proliferation of multipotent hemopoietic progenitors (Ihle and Weinstein, 1986). The human counterpart was identified through molecular cloning of the corresponding gene (Yang et al., 1986;Dorssers et al., 1987;Otsuka et ab, 1988). Recombinant human interleukin-3 (hIL-3) stimulates proliferation and differentiation of a broad range of hemopoietic progenitors (Dorssers et al., 1987;Bot et al., 1988) and also supports * 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.
$ Supported by the Dutch Cancer Society.
DNA mutagenesis (Botstein and Shortle, 1985) has been applied extensively as a tool to study the mechanism of action of proteases, enzymes, growth factors, receptors, transcription factors, oncogenes, and other cell components (Knowles, 1987;Stone et al., 1988;Russell and Fersht, 1987;Bass et al., 1988;Cohen et al., 1986a;Kuga et al., 1989;Yanofski and Zurawski, 1990;Ibiiiez et al., 1990). Analysis of fusion proteins, deletion, insertion, and substitution mutants has enabled the precise localization of functional domains in some cases. Amino acid residues involved in receptor binding of IL-2 (Cohen et al., 1986a;Zurawski, 1988, 1989) and growth hormone  have been identified by using these procedures. Such information may result in the development of mutant proteins with either enhanced function (Russell and Fersht, 1987), reduced antigenicity, or even antagonistic activity (Marcucci and De Maeyer, 1986;Baird et al., 1988;Hannum et al., 1990;Eisenberg et al., 1990;Carter et al., 1990).
The stimulatory effect of IL-3 as well as other growth factors is mediated by binding to specific cell surface receptors (Metcalf, 1986;Nicola and Metcalf, 1988). To assess which regions of the protein are involved in receptor binding we have constructed a series of deletion and substitution mutants. These mutants were expressed in Escherichia coli and Bacillus licheniformis, purified and tested for biological function and receptor binding. In addition, these muteins were used to define the binding sites of neutralizing monoclonal antibodies (mAbs) directed against hIL-3.

MATERIALS AND METHODS
Bacterial Expression Vectors-All modifications on human IL-3 were performed on the IL-3 cDNA insert, contained in the eukaryote expression plasmid pLB4 (Dorssers et aL, 1987) after removal of the repeat sequences within the 3"noncoding region by blunt end ligation of filled AvaI (nucleotide residue 545) and XhoI (nucleotide residue 857) sites. The shortened IL-3 cDNA was inserted into the polylinker of pUC8 in phase with the NH2-terminal amino acids of the locZ polypeptide. Since the 5'-noncoding sequences, as well as the sequences encoding the IL-3 leader polypeptide, were included in this construct (pUC/hmulti), a 175-amino acid fusion polypeptide of 19,686 daltons was produced. This fusion protein was produced efficiently in E. coli and used for raising antibodies.
To produce a polypeptide closely resembling the mature human IL-3, a construct was made lacking the 5"nontranslated and leader IL-3 sequences. The IL-3 cDNA insert of pUC/hmulti was digested with Hind11 and HindIII, ligated to a synthetic oligonucleotide com-  (Sanger et al., 1977). the complete IL-3 insert on the SalI-Hind111 fragment was introduced into pUC8 (digested with Sal1 and HindIII) for protein production. T o allow for direct sequencing and mutagenesis, the fl-ori (PuuI fragment) of plasmid pTZ18R replaced the corresponding PuuI fragment of pUC8. The resulting bacterial expression plasmid (pPH1) produced a fusion protein of 145 amino acids with an apparent M, of 16,384. The DNA and protein sequence of the heterologous leader are shown.

M T M I T N S R G S V D ATGACCATGATTACGAATTCCCGGGGATCCGTCGAC
An alternative expression plasmid was constructed for transfer of interesting mutants into large scale production vectors for different hosts. For this purpose, the p P H l plasmid was digested with HpaI and HindIII to remove the 3"terminal part of the IL-3 cDNA, made blunt, and ligated with a blunt fragment carrying the corresponding IL-3 sequences (nucleotide region 137-497) and the B. licheniformis a-amylase terminator contained in plasmid pGB/IL-322 (van Leen e t al., 1991). The resulting plasmid (pPH3) was digested with RamHI and Hind11 to substitute part of the polylinker and 5"terminal IL-3 sequences for kinased synthethic oligonucleotides. Thus the IL-3 gene was reconstructed with a heterologous a-amylase leader peptide (148 amino acid residues, M, = 16,541). The DNA sequence of this plasmid (pPH4) was verified. The DNA and protein sequence of the heterologous leader are shown.

ATGACCATGATTACGAATTCCCGGGGATCCTCTGCAGCAGCGGCG
In Vitro Mutagenesi-Zn vitro mutagenesis was performed using synthetic oligonucleotides (mutations were flanked by 12-15 homologous nucleotides) according to the procedure developed by Kunkel et al. (1987). Single-stranded template DNA was prepared by transformation of p P H l or pPH4 DNA into E. coli strain CJ236 (Rio-Rad) and superinfection with M13K07 (Pharmacia) helper phage. Phage DNA was prepared and fractionated on 0.6% low melting agarose to remove the M13K07 phage single-stranded DNA. Purified pPHl/pPH4 single-stranded DNA (f100 ng) was annealed with 20 ng of phosphorylated primer at 65 'C and cooled slowly to room temperature. Synthesis of the complementary strand was performed using 1-2 pg of gene 32 protein, 4 units of T4 DNA polymerase, and 2.5 units of T4 DNA ligase. Reaction was carried out a t 0 "C for 5 min, a t 20 "C for 10 min, and a t 37 "C for 90 min. After completion, one-third of the reaction mixture was mixed with thawed competent JM109 cells and plated. Individually picked colonies were grown and superinfected with the M13K07 helper phage to produce singlestranded DNA. The sequence of these clones was verified using the M13 reverse sequencing primer and two IL-%specific sequence primers (nucleotides 118-137 and 261-280).
The double cysteine mutant (9045) was constructed by ligation of the corresponding RstRI-RarnHI fragments of the plasmids carrying the mutant IL-3 genes 904 and 905. Mutant 932/9 was derived from 939 following HpaI and RarnHI digestion and ligation with oligonu-cleotides containing the n-amylase leader sequence. Mutants X and j! were prepared by ligation of fragments carrying the 5'-and 3'terminal deletions.
Of all selected transformants. plasmid DNA was isolated for a second round of transformation into JM109 cells to exclude potential contamination with the parental IL-3 construct. After sequence verification these clones were used for protein production. For expression in Rucillu~, the fragment of the pPH4-derived mutant containing the a-amylase signal peptide, the mutant 11,-3 cDNA, and the a-amylase terminator was isolated and transferred to the Bacillus expression vector pGB/IL-322 (van k e n et al., 1991).
Molecular biology procedures not detailed were performed according to the standard procedures (Maniatis et al., 1982;Herger and Kimmel, 1987;Sambrook et al., 1989). Enzymes and other molecular biology reagents were obtained from GIHCO-HRL and Pharmacia.
Purification of Recombinant IL-3 Protein--I;. coli JM109 cultures (100 ml) were inoculated with 0.5 ml of a fresh overnight culture of the (mutant) IL-3 clone and grown a t 37 "C until an OL) of 0.4-0.6 at 550 nm. Plasmid-directed protein synthesis was induced by addition of 1 mM isopropyl 1-thio-fi-D-galactopyranoside (Pharmacia). After culture for 3-16 h the cells were collected by centrifugation (10 min. 3,000 X g a t 4 "C) and stored frozen. Lysozyme was added (500 pg/ ml) to the E. coli suspension in I0 mM of T E (10 mM Tris/HCI. pH 8, 1 mM EDTA). After incuhation for 30 min a t room temperature. M&IZ and DNase were added to final concentrations of 10 mM and 20 pglml, respectively. After incubation a t 37 "C for 15 min, Tween 20 (0.2%), DTT (2 mM), and phenylmethylsulfonyl fluoride (0.1 mM) were added. The suspension was cooled on ice and sonified vigorously (two X 35 s). The homogenate was clarified by centrifugation (BO min a t 15,000 X g) a t 4 "C, and the supernatant was discarded. The pellet was resuspended in 4 ml of 55% sucrose in buffer T P D (.50 mM Tris/ HCI, pH 8, 0.1 mM phenylmethylsulfonyl fluoride, and 2 mM DTT) by sonication and layered onto a discontinuous sucrose gradient (2ml portions of 75 and 60% of sucrose in buffer TPD). After centrifugation a t 200,000 X g a t 25 "C for 2 h the inclusion bodies containing the IL-3 proteins were recovered from the 7srh sucrose interphaw (essentially as described by Clark et al. ).' After a t least 4-fold dilution the inclusion bodies were pelleted a t 25,000 X g (30 min) and sonificated in 5 ml of 8 M urea containing 50 mM Tris/HCI. pH 8.9, and 2 mM DTT and left overnight at 4 "C. The clarified solution was next applied to a 3-ml DEAE-Sepharose Fast Flow column (I'harmaria). equilihrated with 8 M urea, 50 mM Tris/HCl, pH 8.9. and 1 mM D T T buffer. The IL-3 protein was bound to the column and step eluted with 75 mM NaCl in the same buffer. The eluted protein was dialyzed against several portions of 10 mM Tris/HCI. pH 8.0, and 1 mM IYIT buffer and made isotonic by adding IO-fold concentrated RPMI cell culture medium (Sigma) containing 1% HSA. The filter-sterilized solution was used for biochemical and biological characterization. fluoride, 1 M ammonium sulfate and passed over a 15-20-ml Fractogel TSK Butyl 650(C) column (Merck) equilibrated with 1 M ammonium sulfate, 10 mM Tris/HCl, pH 7.0, buffer. The bound IL-3 protein was eluted with 10 mM Tris/HCl buffer and subsequently passed over a 1.5-ml DEAE-Sepharose Fast Flow column equilibrated with 10 mM Tris/HCl, pH 8.0, buffer. The flow-through was collected and adjusted to 70% ammonium sulfate to concentrate the IL-3 protein. The precipitate was collected by centrifugation at 15,000 X g, dissolved, and dialyzed against 10 mM Tris/HCI, pH 8.0, 1 mM DTT buffer.
Protein samples were analyzed on 13.5% SDS-polyacrylamide gels (acrylamide/bisacrylamide = 291). Proteins bands were visualized by either Coomassie Brilliant Blue G-250 staining or immunological methods. Proteins were quantified (+15%) using densitometric analysis (model 620, Bio-Rad) with the purified recombinant hIL-3 preparation (0.3-1 pg) as a reference. For preparation of large amounts of 21-kDa fusion protein (pUC/hmulti) for antibody production, the bacterial pellet was homogenized in TE buffer containing 0.2% Nonidet P-40 using sonication. After centrifugation the pellet was solubilized in 6 M urea, 1% SDS, 1% 8-mercaptoethanol and fractionated on preparative 12.5% SDS-polyacrylamide gels. After completion of the run, I-cm side strips were cut from the gel and stained briefly. The 21-kDa fusion protein band was located in the gel using the stained strips, excised, and used for immunization.
Immunological Procedures-Monoclonal antibodies were raised by injecting BALB/c mice with the gel-purified IL-3 fusion protein. The excised gel band was minced in saline with a mortar and emulsified in a 1:l ratio in complete Freund's adjuvant containing 1 mg of Mycobacterium tuberculosis H37RA/ ml.
Booster injections were given at week 2 in incomplete Freund's adjuvant. Splenic lymphocytes were fused 3 days later with SP2/0 myeloma cells according to standard procedures (Galfre and Milstein, 1981). Hybridoma supernatants were screened with an enzyme-linked immunosorbent assay with a lysate of E. coli cells containing the fusion protein as a positive control and a lysate of E. coli/pUC8 as negative control. Hybridoma cultures specific for IL-3 were selected and stabilized. Monoclonal antibodies were mostly of IgG1, IgGPA, and IgG2B type. A limited number of hybridoma cell lines was injected into mice for large scale production of antibodies. The ascites fluids were purified using protein A affinity chromatography.
For immunological detection of IL-3 proteins the gel-fractionated proteins were transferred onto nitrocellulose (0.2 pm) using a semidry blotting system (Pharmacia) with a continuous buffer system (39 mM glycine, 48 mM Tris, 0.0375% SDS, and 20% methanol) at 1.2 mA/ cm2 for 90 min. The nitrocellulose filter was subsequently air dried, preincubated with BSA, and incubated for 0.5-16 h with the monoclonal antibodies directed against hIL-3. Immunological complexes were visualized using biotinylated anti-mouse IgG, streptavidin-biotinylated horseradish peroxidase (Amersham Corp.), or alkaline phosphatase-linked anti-mouse IgG (Promega, Madison, WI) according to the protocols provided by the manufacturer.
[3H]Thymidine (0.1 pCi, 2 pCi/mmol) was added in 20 pl of SFM, and cells were harvested 16 h later. A unit of IL-3 activity is defined as the amount required to give 50% of maximal DNA synthesis in this assay. A protein preparation from &galactosidase containing inclusion bodies, produced in E. coli JM109 using the expression vector pUR288 (Riither and Mijller-Hill, 1983), was purified in the same manner as the IL-3 mutants and used as a negative control. No stimulation of DNA synthesis was observed using this backrial extract. To eliminate dayto-day variations in the bioassay a recombinant wild-type hIL-3 preparation produced in Bacillus (van Leen et al., 1991) was included as a standard.
IL-3 receptor binding assays were performed using purified recombinant IL-3 radiolabeled with the Bolton-Hunter reagent (Budel et al., 1989). The specific activity was estimated at 80,000 cpm/ng of IL-3. As target, either donor leukocytes or patient AML cells were used. Cells were washed in Hanks' balanced salt solution and suspended in a-minimum essential medium plus 1% BSA. Usually, 3-10 X IO6 cells were incubated with 200-300 p~ of radiolabeled IL-3 and various concentrations of unlabeled (mutant) protein in a total volume of 200 pl for 1 h at 37 "C. Cell-bound labeled IL-3 was centrifuged through a 0.5-ml cushion of bovine calf serum in an Eppendorf tube for 10 min at 1,000 X g at 4 'C (Budel et aL, 1989). The tubes were frozen in liquid nitrogen and the tips cut off for counting in a Packard y-counter. Competition efficiencies of mutant IL-3 preparations are expressed relative to the reference IL-3 preparation.

RESULTS
Protein Production and Purification-The expression vectors pPHl and pPH4 carrying the IL-3 cDNA fused to the NHn-terminal amino acids of @-galactosidase were modified by the oligonucleotide-directed mutagenesis procedure described by Kunkel et al. (1987). A total of 36 mutant IL-3 proteins, consisting of 18 deletion mutants and 18 substitution mutants, has been generated and analyzed (see Table 11). The expression of pPH4-derived IL-3 proteins in E. coli was generally slightly higher than the same mutant protein from the pPHl expression vector. The employed purification procedures were not designed for complete purification of IL-3 proteins but to remove toxic bacterial components and allow for quantification. In general, minor high molecular weight contaminants were observed on SDS gels. Using densitometric scanning of stained SDS gels (Fig. l), the amount of IL-3 protein was determined in preparations that were also used for bioassays and to which BSA was added. The total yield was generally 0.1-1 mg of semipurified IL-3 protein from 100ml cultures. Exceptions were mutants 811 and 933 carrying deletions between residues 29 and 49, which gave no detectable fusion protein production in either pPH1-or pPH4derived expression vectors. Northern blot analysis indicated a strong reduction of mutant-specific RNA in the bacteria without a significant difference in plasmid DNA content. These results suggest that transcription or mRNA stability is reduced in E. coli through the deletions introduced in the eukaryote IL-3 cDNA sequence for reasons as yet unknown.   Introduction of these mutant DNA sequences into a different host ( R . licheniformis), using an expression vector carrying the a-amylase promoter, also resulted in poor production levels.

Antibody Interactions
Characterization of Monoclonal Antibodies-Monoclonal antibodies of 17 hybridomas directed against hIL-3 were characterized by Western blot analysis of all deletion mutants. The majority of the monoclonal antibodies raised against SDS gel-purified fusion protein is directed against epitopes located between residues 30 and 55 of the mature IL-3 polypeptide (Table I). Since single deleted domains gave loss of binding of most mAbs these epitopes are likely to reflect linear amino acid chains. Further characterization by competition enzymelinked immunosorbent assay of monoclonal antibodies from ascites fluids showed clear cross-competition between A1 and A18, between A5 and A8, and between A8 and A18, indicating that the binding epitopes of A1 and A18 overlap but are distinct from the overlapping epitopes of A5 and A8. Crosscompetition was not observed with mAb A4. These data are in agreement with the immunoblotting experiments (Tahle I). Furthermore, all monoclonal antibodies were found to react with hIL-3 preparations from mammalian, hacterial, and yeast cultures, irrespective of glycosylation state (data not shown).
Analysis of key residues in the binding epitopes was performed using various substitution mutants and recomhinant rhesus monkey IL-3 protein, which shows 19.55 amino acid sequence divergence with hIL-3 (Burger et al. 19908, 1990b: see also Fig. 5). The Western blot analysis (Fig. 2, right panel) revealed that ascites mAbs A1 and A18 recognize a similar domain. Binding is restricted to residues 31 to 37 and is slightly reduced upon substitution of 36D + R (1480, lane 5 ) and strongly affected by substitution of 34LL + SP in RhIL-3 ( l a n e 9). Binding of mAb A8 is strongly influenced hy substitution of residues 43ED + KR (1481, fanP 4 ) and moderately by 46D -P R (1482, lane 3 ) , indicating that its epitope is located between residues 37 and 46. Suhstitutions in RhIL-3 at positions 42G + E and 46D -P T apparentlv do not interfere with A8 binding (lane 9). Analysis of deletion mutants showed that mAb A5 binding was located between amino acid residues 37 and 55. Substitutions of residues 46D + R (1482, lane 3), 50E + K (1483, lane 2), and 54RR ED (1484, lane 1 ) showed clear reduction of binding whereas other NH2-terminal substitutions (such as 43ED + KR, lane 4 ) did not. This suggests that binding is confined to residues 44-55 and is sensitive to substitutions at positions 46 (D + T), 49 (M 4 V), and 51 (N + K) in RhIL-3 (lane 9). mAb A4 seems to bind to all available IL-3 mutants and variants (Fig. 2, right panel).
Experiments to test the efficacy of the purified ascites mAbs to block biological function of hIL-3 in vitro were performed on AML193 cells (Fig. 3). Monoclonal antibodies Al, A8, and A5 gave 50% inhibition of IL-3-stimulated (1.3 ng/ml, i e . & 20 units/ml) DNA synthesis at final dilutions of 1.0-4 X lod3 whereas A4 showed little inhibition (&3 X lo-'). Mixtures of A1 + A5 and A1 + A8 were effective in 2.5 X dilution whereas combination of all three mAbs showed 50% inhibition at 7 X dilution (Fig. 3). Inhibition of AML193 DNA synthesis by the highest antibody concentration used (1%) could be completely reversed by the addition of excess IL-3, indicating the absence of nonspecific inhibitory substances. These results show that combinations of two or three monoclonal antibodies directed against different epitopes provide more potent inhibition (in comparison with the best single antibody: 4-14-fold, respectively) of in vitro biological function of human IL-3.
Biological Activity of IL-3 Mutants-IL-3 fusion proteins derived from both pPHl and pPH4 expression plasmids containing the complete sequence for the mature human IL-3 protein were virtually identical in biological activity and not significantly different from the B. licheniformis "wild-type" reference hIL-3 (van Leen et al., 1991) preparation (Table 11). Apparently, the heterologous NHZ-terminal peptides of 12-15 amino acid residues exert no negative effect on biological activity. However, molecules with longer NHz-terminal peptides showed reduced biological activity. In-frame fusion of the complete IL-3 cDNA to lac2 of pUC8 (pUC/hmulti) resulted in a 175-amino acid protein (including the wild-type leader peptide) displaying 100-fold reduced biological activity. Fusion of the mature IL-3 protein sequence to the COOH  1-14/120-133 1-14/116-133 29-37 3.7 f 1.5 X 10" 7.9 f 6.9 X 10" 7.1 f 4.5 X 10" 2.3 f 0.9 X 10" 1.1 f 0.6 X 5.7 f 3.2 X 10-3h f 3.9 X 1.6 (0.5-3.2) "Deletion of amino acid numbers of mature human IL-3 is indicated. Substitutions of amino acids (numbers refer to the first residue) are indicated in single letter code. * Mean specific biological activity (fS.D.) is expressed relative to the reference IL-3 preparation (average +S.D., 1.35 & 0.71 X lo7 units/mg of IL-3 protein; n = 12). Mutant proteins were tested in two to five independent experiments. e Relative receptor binding activity divided by relative biological activity (mean value) is given. The mean and range (between parentheses) of values from two or three independent binding experiments are presented.
'Mutants were prepared using the pPHl expression vector. e NC, no competition was observed for the particular mutein preparation.
f Inclusion bodies were pelleted from sucrose solutions and solubilized in urea. Muteins were diluted directly in culture medium and tested for biological activity. Receptor binding experiments were performed once.
Mutant proteins were expressed in Bacillus and lack a (heterologous) leader polypeptide sequence.
Mutein protein concentration was too low for densitometric scanning and was estimated from gels and/or Western blots.
terminus of a full-length 8-galactosidase protein resulted in an inactive polypeptide (not shown).
IL-3 deletion mutants were mostly significantly reduced in biological activity (Table I1 and Fig. 4A). Deletions of 6-15 amino acids between residues 47 and 106 resulted in more than 10,000-fold reduction in specific activity. However, in all cases residual biological activity was detected, which was  Fung et al., 1984), and rat (Ra, Cohen et al., 198613) were aligned using the Clustal algorithm (Higgins and Sharp, 1988). Alignment was based on conservation of splice sites in these IL-3 species. Gaps in the aligned protein sequence are represented by blanks, and identical residues are indicated by dashes.

The numbering refers to hIL-3 protein sequence. Cylinders at the top represent proposed helices for human IL-3 as predicted by the algorithms of Garnier et al. (1978), Chou and Fassman (1978), and Rose (1978) available on the DNASIS (Pharmacia) software. Helix ends were chosen according to the amino acid preferences given by Richardson and Richardson (1988).
absent in control preparations from E. coli. These results indicate the sensitivity of the bioassay and the lack of large amounts of toxic components in the protein preparations used at high concentration. Biological activity was retained for deletion mutants 932 (1-14) and 939 (120-130) and the corresponding double mutant 932/9 (Fig. 4A). To demonstrate that the heterologous NHz-terminal peptide did not substitute for the natural amino acids, minimal mutants X (1-14/120-133) and 2 (1-14/116-133) were expressed in B. licheniforrnis.
These muteins exhibited close to full biological activity (Table   11). In contrast, single substitutions of both cysteine residues (mutants 904,16C + A, and 905,84C + A) resulted in 3,000fold reduction of activity (Fig. 4A). A similar effect was seen with a double cysteine mutant (9045,16/84C + A), indicating that the S-S bridge plays an important stabilizing role in the IL-3 molecule. Substitution of (individual) amino acids by residues with opposing charge or function also resulted in a significant decrease of biological activities. Substitutions at positions 59E + K (1485) and 75E + G or R (1487A/B) gave moderate reduction of function whereas alterations of residues 43ED * KR (1481), 54RR + ED (1484), 63RA + PG (1486), 108RRK + EDE (1492), and 113FY + AT (1493) had extensive effects (Table 11).
Receptor Binding Capacity-Receptor binding of these mutant IL-3 proteins was determined to assess the relationship between binding affinity and the ability to stimulate DNA synthesis. The binding experiments do not allow for detailed comparison of receptor binding activities because of the low numbers (25-200) of high affinity receptors (Kd c 50-200 PM) on normal and leukemic target cells (Budel et al., 1989). Variations in competition efficiencies were observed in different experiments and are indicated. A typical example of a competition binding experiment is presented in Fig. 4B. In general, there is a good correlation between biological activity and receptor binding (Table 11). Mutants with full or partial biological activity competed for binding of labeled wild-type IL-3 with comparable efficiency. Most mutants that showed a large reduction (>10,000-fold) in biological activity were unable to compete for binding of wild-type hIL-3 because of the reduced sensitivity of the receptor binding assay. The cysteine substitution mutants (904 and 9045) and mutants 1483 (50E + K) and 1488 (79K + E) appeared to be more potent (37-100-fold) in receptor binding than in stimulating DNA synthesis (Table I1 and Fig. 4B).

DISCUSSION
Production of hIL-3 fusion proteins and mutants thereof in E. coli proved to be very efficient. The heterologous NHzterminal amino acids did not affect the biological activity significantly. Analysis of the deletion mutants demonstrated that the NHz-and COOH-terminal amino acids are nonessential to biological function. The boundaries of the essential domain of hIL-3 are presented by the cysteine residue on position 16 and the hydrophobic domain at position 113-115.
The cysteine residue contributes to the single S-S bridge, which is an essential structural element in hIL-3 architecture as shown by alanine substitution (Table 11). The cysteine residues in hIL-3 are strictly conserved in all studied species (Fig. 5) and were also shown to be required for biological function in murine 1L-3 (Clark-Lewis et al., 1988). The hydrophobic residues at the COOH-terminal boundary mentioned above are also completely conserved in primate and rodent IL-3 species (Fig. 5) and could thus be essential for function. Binding of monoclonal antibodies to the essential domain of hIL-3 neutralized the biological function by interference with receptor binding and signal transduction. Deletion of amino acid residues may result in removal of residues involved in receptor binding or affect protein folding and similarly reduce the affinity of the mutein for the IL-3 receptor. Amino acid deletions within the essential domain of hIL-3 either had a moderate effect on activity (between residues 20 and 47) or resulted in at least 10,000-fold reduction of biological activity (Table 11). The finding that all deletion mutants had retained some specific biological activity was unexpected. Contamination of mutant proteins with wildtype IL-3 derived from the parental pPH vector is unlikely. Plasmid DNA of characterized mutants has been recloned in E. coli prior to protein production. This implies that the residual biological activity of the mutant proteins must be caused by specific interaction with the receptor and subsequent activation of intracellular signalling pathways.
Analysis of the substitution mutants shows that alteration of highly conserved residues (Fig. 5) in both primate (human, gibbon, and rhesus monkey) and rodent (mouse and rat) IL-3 species gives more than 3 orders of magnitude reduction of biological activity (1481, 43ED + K R 1484, 54RR + ED; 1491, 106E + K; 1492, 108RRK + EDE; 1493, 113FY + AT). Substitution of residues that have not been conserved during evolution shows moderate reduction (2-3 logs) of activity. Since a functional species barrier exists between primate and rodent IL-3 species (Cohen et al., 1986b;Dorssers et al., 1987;Burger et al., 1990a) it is likely that conserved residues are important in protein architecture rather than in specific interaction with the receptor. Circular dichroism measurements of purified recombinant hIL-3 have indicated a high proportion (260%) of a-helix ~o n t e n t .~ Combining computer-assistedpredictions (Chou and Fasman, 1978; Rose, V. Smit, ITRI-TNO and A. J. W. G. Visser, LU Wageningen, personal communication. Garnier et al., 1978) and observed amino acid preferences at the ends of helices (Richardson and Richardson, 1988), we propose a-helices between residues 18 and 29 (I), 42 and 54 (11), 57 and 82 (111), 85 and 91 (IV), and 105 and 121 (V) of hIL-3 (Fig. 5). Introduction of helix-breaking residues (Pro-Gly) at residues 63/64 (1486) strongly affects biological activity. This may be explained by disruption of the long helix 111, which is predicted for all species by the mentioned algorithms. Helix stability may be increased by properly charged residues at its ends (Presta and Rose, 1988;Richardson and Richardson, 1988). Substitution of the conserved residues 43/44 (1481), 54/55 (1484), and 106 (1491) at the termini of the postulated helices I1 and V by residues with opposing charge resulted in loss of activity and may be explained by this model. Charge reversals of conserved charged residues at positions 108-110 (1492) also strongly affected activity. These residues are not located at the end of a helix and are not supposed to contribute notably to helix stability but could be involved in long range interactions. Substitution of the completely conserved hydrophobic residues 113/114 (1493) severely reduced biological function and may be explained by disruption of hydrophobic interactions within the molecule and/or with the receptor. Two-dimensional nuclear Overhauser effect spectroscopy-'H NMR measurements on recombinant hIL-3 and RhIL-3 showed virtually identical nuclear Overhauser effects from the aromatic side chains of Phe113,4 indicating that the aromatic protons have highly conserved long range interactions (10.5 nm) in both IL-3 species.
The majority of the monoclonal antibodies reacted with continuous epitopes within two small regions of the hIL-3 protein. The largest group of mAbs is capable of identifying the 933 (29-37) mutant but failed to react with the 811 (31-49) mutant (Table I). The epitopes recognized by these mAbs must reside between residues 37 and 50, which is a highly hydrophilic region and could be predicted to be an antigenic determinant (Hopp and Woods, 1981). Actually, mAb A5 appears to interact with the hydrophilic face (residues 46, 50, and 54) of helix I1 since alterations in these residues strongly affected binding. The other group of antibodies is characterized by the absence of reaction with mutant 933, which lacks 9 amino acids. This hydrophobic proline-rich region between helices I and I1 would not have been predicted to be a major antigenic site. Alternatively, the deletion of these residues may cause structural changes that preclude binding of the antibodies to amino acid residues outside this deletion. This is unlikely since amino acid changes further to the COOH terminus did not affect antibody binding. Apparently, these mAbs interact with a rather hydrophobic domain of IL-3, which is much smaller than the average length of 15-22 amino acid residues constituting an epitope (Laver et al., 1990). However, kinetic parameters have not been measured, and thus only the most prominent residues participating in the binding energy have been identified. The weak capacity of the individual mAbs to inhibit hIL-3-induced DNA synthesis in AML cells may be attributed to a relative low affinity of the antibodies in comparison with the high affinity binding of the growth factor to its receptor. The combination of mAbs directed against different epitopes did compensate for this low affinity and increased the neutralizing potential.
The residual biological activity of all deletion mutants implies that single (small) deletions are insufficient to abrogate receptor binding. It may thus be speculated that more than one domain of the hIL-3 polypeptide is involved in interaction with the receptor on AML193 target cells and that D. Schipper, Gist-brocades, unpublished results. deletion mutants do not cover both domains simultaneously. This hypothesis is supported further by the apparent discrepancy in receptor binding and induction of DNA synthesis of some mutants, indicating that the different receptor binding domains on the growth factor do not induce signal transduction with identical efficiencies. Discontinuous polypeptide domains essential for biological function have also been suggested for other growth factors such as growth hormone , GM-CSF (Shanafelt and Kastelein, 1989;Kaushansky et al., 1989) and tumor growth factor-a (Defeo-Jones et al., 1988). The principle of multiple domain interaction between growth factor and receptor has been demonstrated in detail for IL-2 (Collins et al., 1988;Robb et al., 1988;Zurawski, 1988,1989). The high affinity receptor of IL-2 consists of two polypeptides that interact with different domains of the growth factor (Collins et al., 1988;Zurawski and Zurawski, 1989). The IL-3 receptor also seems to be a multicomponent complex. A low affinity murine IL-3-binding protein with structural homology to other growth factor receptors (such as IL-2RP and erythropoietin receptor) has been molecularly cloned but lacks biological function (Itoh et al., 1990;Bazan, 1990). Apparently this molecule associates with a second polypeptide to constitute a high affinity receptor capable of transducing the signal (Hibi et al., 1990). A further complication is presented by receptor molecules on specific lineages of human hemopoietic cells (not murine) capable of binding both IL-3 and GM-CSF Lopez et al., 1989;Budel et al., 1990). Thus it may be speculated that the low affinity IL-3-binding protein can form a complex with at least two different polypeptides resulting in distinct high affinity receptors. Since the AML cells used in the binding experiments expressed both types of IL-3 receptors, differential binding of mutant proteins to either IL-3 receptors or common receptors may have been masked. However, preliminary experiments revealed no discrepancy in binding by various mutants (e.g. 810, 812, 822, and 904) to AML cells with specific IL-3 receptors and IL-3/GM-CSF common re~eptors.~ The mutational analysis presented here has shown that approximately 24% of the amino acid residues at the extremes of the mature hIL-3 are not essential to biological function. Similar results have been obtained for murine IL-3 (Clark-Lewis et al., 1986) and other growth factors Zurawski, 1988 Shanafelt andKastelein, 1989;Yanofsky and Zurawski, 1990), indicating that the terminal residues are often less important in protein folding and function. Comparison of primate IL-3 sequences revealed that the COOHterminal 9 amino acids are absent in rhesus monkey IL-3, without affecting its stimulatory effect on human cells . In contrast, hIL-3 displays 100-fold reduced activity on rhesus monkey hemopoietic cells in vitro and in uiuo Wagemaker et al., 1990b). These results indicate that the protein sequence divergence in RhIL-3 does not significantly affect interaction with the various types of IL-3 receptors on human cells but also suggest that the RhIL-3 receptor(s) may have diverged notably during evolution. Finally, the observed higher binding activity of some mutants in comparison with their capacity to induce DNA synthesis suggests that receptor binding and induction of cell proliferation may be disconnected and that antagonists may be devised.