Identification of a Receptor Binding Site in the Carboxyl Terminus of Human Interleukin-6”

To identify a receptor binding site of human inter- leukin-6 (IL-6), we created a library of IL-6 variants with single amino acid substitutions in the last 15 residues (171-186) in the COOH terminus of IL-6. Twenty-seven IL-6 variants were tested for biological activity on a human hepatoma and a mouse hybridoma cell line. Most variants were additionally tested in a receptor binding assay using a human myeloma cell line. Several single amino acid substitutions in the COOH terminus of IL-6 were found to decrease biological activity significantly. This is especially seen in variants with amino acid substitutions that alter the postulated amphipathical a-helix structure between residues 178 and 183. The two highly conserved Arg residues at positions 180 and 183 seem to play a very important role in biological activity. The loss of biological activity in all inactive variants is completely par- alleled by a decrease of IL-6 receptor binding, as determined by competition binding experiments. One mutant (Leu171) displayed a higher activity on human cells and a higher binding affinity to the receptor and can be considered an IL-6 agonist. It is concluded that the amphipathical a-helix structure in the COOH terminus of IL-6 is critical for ligand receptor interaction. Fur-thermore, the region between residues Serl” and ArglS3 (Ser-Leu-Arg-Ala-X-Arg) is identified as a receptor binding site in the COOH terminus

To identify a receptor binding site of human interleukin-6 (IL-6), we created a library of IL-6 variants with single amino acid substitutions in the last 15 residues (171-186) in the COOH terminus of IL-6. Twenty-seven IL-6 variants were tested for biological activity on a human hepatoma and a mouse hybridoma cell line. Most variants were additionally tested in a receptor binding assay using a human myeloma cell line. Several single amino acid substitutions in the COOH terminus of IL-6 were found to decrease biological activity significantly. This is especially seen in variants with amino acid substitutions that alter the postulated amphipathical a-helix structure between residues 178 and 183. The two highly conserved Arg residues at positions 180 and 183 seem to play a very important role in biological activity. The loss of biological activity in all inactive variants is completely paralleled by a decrease of IL-6 receptor binding, as determined by competition binding experiments. One mutant (Leu171) displayed a higher activity on human cells and a higher binding affinity to the receptor and can be considered an IL-6 agonist. It is concluded that the amphipathical a-helix structure in the COOH terminus of IL-6 is critical for ligand receptor interaction. Furthermore, the region between residues Serl" and ArglS3 (Ser-Leu-Arg-Ala-X-Arg) is identified as a receptor binding site in the COOH terminus of human IL-6.
Interleukin-6 (IL-6)' is a multifunctional cytokine that is produced by a variety of cells such as B-cells, T-cells, monocytes, fibroblasts, and endothelial cells (f6r review see Refs. 1, 2, and 3). IL-6 exerts several activities including terminal differentiation of B-cells (4), proliferation and differentiation of T-cells (5), regulation of the acute phase response (6), and growth regulation of epithelial cells (7). It also stimulates the * Supported by National Heart, Lung and Blood Institute Grant HL31012. 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.
$ Present address Department of Internal Medicine, Hirosaki University School of Medicine, Hirosaki 036, Japan. differentiation of hematopoietic progenitor cells alone or in combination with other cytokines such as IL-3 and IL-4 (8,9). IL-6 also plays a role in the differentiation of megakaryocytes and seems to be the most important factor for thrombopoiesis (10). Another important feature of IL-6 is the ability to act as a growth factor of precancerous cells, such as in myeloma and plasmacytoma (11,12). In accordance with these properties, the target cells for IL-6 include B-cells, T-cells, myeloma cells, megakaryocytes, monocytes, early stem cells, and hepatocytes. The various biological effects of IL-6 on the different cell types are initiated by the interaction of IL-6 with two distinct receptor subunits. First, IL-6 will form a complex with an 80-kDa receptor followed by the binding of this complex to a non-ligand binding subunit, the membrane glycoprotein gp130 (13,14). The binding of the IL-6-80-kDa receptor complex to gp130 results in signal transduction (14). The signal transduction pathways upon IL-6 receptor binding are still unclear; however, it has been recently demonstrated that the biological effect of IL-6 on B-cell hybridoma cells is mediated by a novel protein kinase signaling pathway resulting in the activation of JunB and TIS11 gene transcription (15).
In the absence of a tertiary structure of IL-6, the ligandreceptor model for IL-6 is based on the homology with other proteins. Bazan (16) has extensively analyzed the homology of various evolutionarily related proteins by means of computer modeling and indicated that a amphipathical a-helical structure is found in the COOH terminus of several cytokines including IL-6, erythropoietin, myelomonocytic growth factor, granulocyte-colony stimulating factor (G-CSF), as well as growth hormone and prolactin. Structure-function studies on growth hormone, G-CSF, and IL-3 suggest that this amphipathic &-helix structure at the COOH termini of these proteins is involved in receptor binding (17)(18)(19).
Several studies on IL-6 have indicated that the COOH terminus of IL-6 is important for biological activity. Brakenhoff et al. (20) showed by epitope mapping that neutralizing monoclonal antibodies against IL-6 were directed against a region in the COOH terminus of IL-6, suggesting a role of the COOH terminus in biological activity. More direct evidence for the importance of the COOH terminus was obtained by structure-function analysis of IL-6. Two studies showed that IL-6 variants with deletions of the COOH-terminal 52 or 20 amino acids do not have any detectable IL-6 activity on human and mouse cell lines (20,21). Kriittgen et al. (22) found that even a deletion of only the last three amino acids of IL-6 results in a significant loss of biological activity. None of these IL-6 variants was, however, tested for binding to the IL-6 receptor. Therefore, it remains unclear whether the loss of activity is due to a loss of receptor binding or to a defective signal transduction upon IL-6 receptor interaction.
This led us to investigate the exact role of the COOH terminus of IL-6 for biological activity as well as receptor binding. Specifically we have attempted to identify the individual residues involved in receptor binding. We created a library of IL-6 variants with single amino acid substitutions in the COOH terminus using saturation mutagenesis. We found that single amino acid changes in the COOH terminus of IL-6 can completely abolish biological activity and receptor binding. It is concluded that the receptor binding site of the COOH terminus of IL-6 is located between the residues Ser17' and ArglR3.

Construction of Recombinant ZL-6
The construction of the wild-type IL-6 gene has been described previously (21). Briefly an IL-6 gene containing 4 serines instead of the naturally occurring 4 cysteines was constructed from 22 synthetic oligonucleotides and cloned into a modified pBS M13+ cloning vector (Stratagene, La Jolla, CA). By replacing an EcoRV-StuI fragment with a fragment assembled from 6 oligonucleotides containing four codons corresponding to the four cysteines in natural IL-6, the wildtype IL-6 gene was constructed. This gene was cloned into the expression vector p340. This vector is used for high level expression of an IL-6-@-galactosidase fusion protein with a small collagen linker between the two proteins. The construction of the expression vector has been previously reported (23). A synthetic minicistron was inserted downstream of the promoter to increase levels of protein expression (23). The expression vector containing wild-type IL-6 is called p478.

Construction of ZL-6 Mutants
A library of point mutations in the COOH terminus of IL-6 was made by saturation mutagenesis, as described earlier (24). In this method, two complementary oligonucleotides are synthesized that cover the region to be investigated. By contaminating the phosphoramidites with a small amount of each of the other three phosphoramidites during synthesis of the oligonucleotide, an oligonucleotide is obtained that has several mutations in its sequence compared to the wild-type gene to be studied.
To reduce the background of nonmutated IL-6 in subsequent cloning procedures a BglII-PmlI fragment of the p478 vector was replaced by a 29-bp segment constructed from two oligonucleotides. The oligonucleotides (a = GAT CTT AAG TAG TAG GTA CCA

A G C T T C A C a n d b = A A T T C A T C A T C C A T G G T T C G A A G
TG) were made on a DNA synthesizer (Applied Biosystems 380A, Foster City, CA) and purified by reverse-phase high performance liquid chromatography. Reagents were obtained from American Bionetics (Hayward, CA). By inserting this fragment the @-galactosidase gene is out of frame with the initiating methionine. This vector (p560) was used for further cloning experiments.
Four oligonucleotides c, d, e, and f that replaced the BglII-PmlI segment of the original expression vector p478 were synthesized. This segment contains the last 46 bases of the IL-6 gene and part of the collagen linker of the fusion protein. The comulete seauence of these oligonucleotides is: c = GAT CTT TGG CCC AAG AGG TCT CGC TGG CCC ACA AGG TCC AC; e = GTG GAC CTT GTG GGC CAG CGA GAC CTC TTG GGC CAA

A A G C A C C A G C A G G A C C A A C A G G A C C : f = A A C A G G A C C A G G A T C T G A C A T T T G C C G A A G A G C T C T C A G G C T A G A
CTG CAG GAA CTC CTT AAA A. The nonunderlined part of the oligonucleotides c and f as well as the complete oligonucleotides d and e are identical to the original segments in the expression vector p478. The underlined part of the oligonucleotides code for the COOHterminal 46 bases of the IL-6 gene, which was mutagenized by using a mixture of phophoramidites in the synthesis mixes during synthesis of the oligonucleotides. The synthesis mixtures contained 96% of pure nucleotide and a 4% equimolar mixture of all phosphoramidites. It was calculated that about 20% of the IL-6 constructs should be wild-type IL-6 and 80% should have single or multiple mutations (24). A 129 bp fragment was obtained by ligating the pairs of oligonucleotides d-f and c-e. The fragment was cloned into p560, which resulted in an expression vector with the complete IL-6-P-galactosidase fusion protein that has several point mutations in the COOHterminal part of IL-6. After transformation of XL-1 Blue cells (Stratagene, La Jolla, CA), the transformation mixture was plated out on Luria broth culture plates, coated with ampicillin and tetracyclin as selection markers and isopropyl-@-D-thiogalactopyranoside (IPTG) for induction of fusion protein synthesis and 5-bromo-4-chloro-3indolyl (3-D-galactopyranoside (BCIG) as a substrate for @-galactosidase to be able to select for dark blue colonies. This first transformation resulted in mixed colonies because the complementary oligonucleotides c and f are likely to have different sequences due to the mutagenesis procedure. The colonies were washed off the plates and DNA was purified and used for a consecutive transformation to obtain pure colonies. A total of 250 blue colonies were checked for having the 129-bp insert by the polymerase chain reaction (PCR) (25). The plasmids that had the correct size insert were used for further analysis. Unfortunately this mutagenesis procedure created IL-6 variants with single amino acid substitutions in 14 of the last 15 residues of IL-6. Therefore an additional variant was made using site-directed mutagenesis. This variant (Arg'" to Lys) was made using the T7-GENTM Mutagenesis kit (U. S. Biochemicals, Cleveland, OH) according to instructions provided by the kit (26). Because we wanted to have a non-conservative replacement of residue Glu'", a mutant was made by oligonucleotide-directed mutagenesis of one of the obtained double-point mutants ( G~u '~~ to Val and ArglS3 to Lys) using PCR. One of the oligonucleotides was annealed over the last 24 bases of the coding sequence of IL-6, thereby changing these last bases to the original IL-6 coding sequence and the LyslR3 to the original residue (Arg). After ligation of the PCR product in the vector, an IL-6 variant with only G~u '~~ replaced by Val was created.

Expression, Purification, and Quantitation of Recombinant ZL-6
Variants Detailed description of our bacterial expression system has previously been published (21, 23,26). Briefly, the expression vectors containing the gene for the fusion protein with IL-6 mutants were transformed in Escherichia coli JM101. Single ampicillin-resistant colonies were picked to inoculate 10-ml broth cultures. At log phase growth of the bacteria, expression of the fusion protein was induced by addition of 100 ml of IPTG (25 mg/ml). @-Galactosidase activity was measured and at maximum levels of activity, bacteria were pelleted by centrifugation and stored at -20 "C until further use. Bacteria were resuspended in 1 ml of 0.5 X phosphate-buffered saline and lysed by repeated freezing and thawing after lysozyme treatment. T o reduce viscosity, the lysate was sonicated and the fusion protein was pelleted by centrifugation. The pellet was washed to remove soluble contaminants and the fusion protein was solubilized in 2% sodium lauroyl sarcosine. Insoluble contaminants were removed by centrifugation and the fusion protein was precipitated by two rounds of selective ammonium sulfate precipition. The IL-6 variants were cleaved from the p-galactosidase by collagenase treatment prior to quantitation or use in bioassays. Because of the high amounts of IL-6 expressed in the induced bacteria, we were able to quantitate the protein preparations by denaturing sodium dodecyl sulfate-polyacrylamide gel electrophoresis under reducing conditions followed by Coomassie staining and scanning laser densitometry.

Bioassays
Hybridoma Growth Assay-The hybridoma growth factor activity was measured using an IL-6-dependent, mouse-mouse hybrid cell line (7TD1) as described previously (27). Briefly, 7TD1 cells were suspended a t a density of 2 X lo4 cells/ml in Iscove's-DMEM medium containing 10% fetal bovine serum, L-arginine (0.55 mM), L-asparagine (0.24 mM), L-glutamine (1.5 mM), P-mercaptoethanol (50 mM), hypoxanthine (0.1 mM), thymidine (16 mM), and penicillin/streptomycin. Cells were plated in 96-well tissue culture plates (100 ml/ well). IL-6 samples to be tested were diluted into the same medium, and each dilution was tested by adding 100 ml/well to the plated cells. After 4 or 5 days, the number of cells was evaluated by colorimetric determination of the hexosaminidase levels (28). Plates were read a t a wavelength of 405 nm on a Vmax ELISA plate reader (Molecular Devices Corp., Palo Alto, CA).
Hepatocyte Stimulation Assay-Hepatocyte stimulation activity was measured using a human hepatoma cell line HEP3B2. The cells were maintained in DMEM-H medium supplemented with 10% fetal bovine serum and penicillin/streptomycin. On the day before treatment cells were trypsinized and plated on 96-well tissue culture plates at a density sufficient to give confluent monolayers after overnight growth. The next day, the IL-6 samples to be tested were diluted in maintenance medium supplemented with M dexamethasone, and each dilution was tested in duplicate by adding 100 ml/well to the plated cells. Conditioned medium was collected after 48 h and assayed for fibrinogen production by a sandwich ELISA using a commercially available capture antibody (IgG fraction, goat anti-human fibrinogen; Cappel, Durham, NC) and a peroxidase-conjugated detection antibody (peroxidase conjugated IgG fraction, goat anti-human fibrinogen; Cappel). Color was developed using a peroxidase substrate (Kirkegaard & Perry, Gaithersburg, MD) and plates were read at a wavelength of 450 nm on the Vmnx plate reader.

Calculations
All assays were carried out in duplicate serial dilutions of three or more independent preparations of each IL-6 variant. The biological activities were determined from dose-response curves prepared for each protein preparation. To minimize variation in these biological assays a duplicate protein preparation of wild-type IL-6 was used on each 96-well tissue culture plate to calculate the activity of two (HEP3B2) or three (7TD1) IL-6 variants on the same plate as a percentage of wild-type IL-6. For the 7TD1 assay ODlos was assumed to be directly proportional to the cell number and was plotted directly against the log of the IL-6 concentration. For the HEP3B2 assay the OD450 was converted to fibrinogen concentration using the V, , , plate reader software before being plotted. For each assay the roughly linear portion of each dose-response curve was fitted with a second order polynomial curve using a computer program. The IL-6 concentration that gave half-maximal proliferation in the 7TD1 assay was calculated for each mutant, and this value was used to calculate hybridoma growth activity relative to the wild-type recombinant IL-6. For the HEP3R2 assay, the IL-6 concentration causing a doubling of fibrinogen secretion was calculated, since this value allowed a more accurate comparison of biological activities, especially for the relatively inactive mutants. The activity of each IL-6 mutant was calculated by taking the mean (and standard error of the mean) of the activities determined for the independent protein preparations.

i n~i~ Assays
Binding competition studies were performed using human recombinant IL-6 derived from E. coli (generously provided by Dr. Steve Clark, Genetics Instititute, Cambridge, MA). IL-6 was radioiodinated using the Bolton-Hunter reagent (Du Pont-New England Nuclear) as was described in detail earlier (29). The specific activity of the radiolabeled IL-6 was 22 mCi/gg. For the binding experiments we used a human myeloma cell line U-266. These cells were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine, and penicillin/streptomycin. This medium was also used for making the dilutions of the IL-6 variants and as binding medium. The fetal bovine serum batches were routinely tested for IL-6 activity and only batches that showed negligible IL-6 activity were used for competition binding experiments. Cells were harvested by low speed centrifugation, washed once in Ca*+/Mg'-free phosphatebuffered saline and resuspended at a concentration of IO' celIs/mi medium of 4 "C. The IL-6 variants were treated with collagenase and diluted immediately afterward to the desired concentrations in binding medium. '251-IL-6 was diluted to a concentration of 16 ng/ml. Thereafter 100 p1 of cells, 50 gl of IL-6 variants, and 50 p1 of Iz5I-IL-6 were transferred in rapid succession using a 12-well multichannel pipet into round-bottomed, polyvinyl 96-well plates (Dynatech, Chan- Data for Scatchard analysis and the competition binding experiments were obtained by measuring the binding in the presence of a constant concentration of '""I-IL-6 and varying concentrations of unlabeled IL-6 variants. The results of the Scatchard and competition binding studies were analyzed using the nonlinear curve fitting pro-gram LIGAND (30). After the Kd of wild-type IL-6 and the IL-6 variants for the receptor were calculated, we expressed the receptor binding capacity of the IL-6 variants as a percentage of receptor binding of wild-type human IL-6 by calculating the quotient of Kd wild-type IL-6/& variant IL-6. This allowed us to compare the binding capacity of the IL-6 variants with the biological activity of the variants on human cell lines which was also expressed as a percentage of wild-type IL-6. For several mutants that competed poorly with wild-type IL-6 for the receptor binding sites, even using very high concentrations of the mutant proteins, we could not calculate a Kd. Using the EBDA/LIGAND program we were able to calculate an apparent IC, that was used to estimate the residual receptor binding capacity of these variants (in most of these variants <2% of wild-type IL-6).

C o~t r~c~i o~
of the IL-6 Variants-For the studies described here we have used a previously constructed vector (~478) for high level expression of an IL-6-/3-galactosidase fusion protein. This vector allows us to produce large quantities of IL-6 variants in IO-ml cultures (21, 26). To obtain a large number of IL-6 variants with different mutations in the COOH terminus, we used the saturation mutagenesis technique (24). First we made some alterations in our expression vector p478 to reduce the background after cloning. Afterward a library was constructed by cloning a 129-bp fragment composed of two pairs of annealed oligonucleotides into this altered vector. Two of these oligonucleotides coding for the 15 COOH-terminal amino acids of IL-6 contained several mutations compared to the original IL-6 gene. These mutations were introduced during oligonucleotide synthesis by contaminating the synthesis mixes with the other three phosphoramidites. After ligation of the fragment into the expression vector and two consecutive transformations, numerous blue colonies were obtained. A total of 250 colonies were screened using PCR for having the correct insert. Afterward 201 plasmids were further analyzed by dideoxy sequencing. Wild-type IL-6 was found in 50 (25%) of the clones, 68 (34%) had one amino acid substitution and 83 (42%) showed two or more amino acid changes. Repeated mutations were found in 24% of the mutated clones and 30% of the IL-6 variants were not useable for further analysis because of deletions or insertions, the introduction of stop codons or the loss of collagenase cleavage sites. We obtained 63 unique variants of which 32 had a single amino acid substitution and 31 had two or more substitutions in the COOH terminus of IL-6. We selected IL-6 variants with the most interesting single amino acid mutations for further analysis of the biological activity compared to re~ombinant IL-6. In addition we made two other IL-6 variants by site-directed mutagenesis to cover all 15 COOHterminal amino acids of IL-6. All IL-6 variants are shown in Fig. 1. The variants with two or more amino acid substitutions were not used for further analysis.
Biological Activity of the IL-6 Variants-After expressing the proteins in our high level bacterial expression system, we performed two assays with human and mouse cell lines to determine the biological activity of our IL-6 mutants: a fibrinogen induction assay using the human hepatoma cell line HEP3B2 and a hybridoma proliferation assay using an IL-6 dependent murine hybridoma cell line 7TD1. The results of both assays are summarized in Table I. The biological activities of all variants are shown as a percentage of the activity of wild-type recombinant IL-6, which has been defined at 100% for both assays.
As is shown in Table I only nonconservative mutations such as changing these charged residues with noncharged residues result in a reduced activity, but also replacement of Arg"' by Lys abolishes the activity completely. Substitution of the noncharged Ser'78 and Led7' with nonconservative Arg residues decreases IL-6 activity significantly. It seems that the nonconservative replacements of amino acids is more deleterious for biological activity in the last 8 amino acids than for the more NH2-terminal amino acids that we studied. IL-6 variants with nonconservative mutations, L Y S '~~ t o Gln and G l P to Lys maintained more than 50% activity. We did not find a variant with an amino acid substitution at position 171-177 that had an activity less than half of that of recombinant IL-6. This might be due to the kind of substitutions made, but more likely shows that this region is functionally less important than the last 8 amino acids. The Ser'77 to Pro mutation is important in this respect, because Pro is known to be able to disrupt helices. This IL-6 variant still had a biological activity comparable with wild-type.
Besides the importance of the charge of the side groups of the various residues, the size of the side group seems to be important for proper functioning of IL-6. This is particularly shown at position 181. The replacement of the Ala"' with a hydrophilic Ser that has about the same size side group as Ala, does not affect activity. In contrast, replacement by a hydrophilic Thr that has only one more methyl group than Ser reduces the activity by 93%.
We found only two variants that displayed a higher IL-6 activity on the human hepatoma cell line as compared to wildtype IL-6. The substitutions were relatively conservative, Phe"' to Leu and Leu'75 to Met. These IL-6 variants did not have a higher activity than human wild-type IL-6 on the mouse cell line 7TD1. A species specificity of the IL-6 variants was also noticed in our previous studies with IL-6 mutants (21,26). Most of those variants, however, had a higher activity in the mouse cell line assays than in the human cell line assays. In the present study we also found that most IL-6 variants displayed a higher activity on mouse than on human cell lines. The residues G~u '~~, Phe'74, Led7', Arg'"', Argl", and GlnIa4 are conserved in the COOH terminus in mouse, rat, and human IL-6. Despite their conservation and therefore apparent importance, we only found a significant decrease of activity in the variants with Arg"' and ArglR3 replaced. Interestingly, these two residues are also conserved in the COOH termini of human G-CSF andchicken myelomonocytic growth factor that both have a significant homology with the COOH terminus of IL-6 (16,31) Receptor Binding of the IL-6 Variants-Binding assays were performed using the human myeloma cell line U-266. These cells have a high number of IL-6 receptors and are therefore very suitable for IL-6 receptor binding analysis (13). U-266 cells were incubated for 3 h at 4 "C with a constant concentration of radiolabeled wild-type IL-6 and differing concentrations of unlabeled IL-6 or IL-6 variants for Scatchard analysis and the competition binding studies. Scatchard analysis of the specific binding of '2s1-IL-6 to U-266 cells revealed 19,850 binding sites/cell with a Kd value of 1.6 X 10"' M. The binding capacity of our IL-6 variants was expressed as a percentage of our wild-type IL-6 by calculating the quotient of the dissociation constants of wild-type IL-6/ variant IL-6 ( Fig. 2). Except for four IL-6 variants with amino acid substitutions that displayed normal biological activities, a t least one IL-6 variant for every residue in the COOH terminus that had been substituted was tested for receptor binding capacity.
The results are summarized in Fig. 2. In general, the binding studies were in concordance with the findings of the biological activity assays. However, one variant (Led7' to Met) displayed a 3-fold higher biological activity than receptor binding capacity. One variant (Phe17' to Leu) with a higher biological activity on the human hepatoma cell line compared to wildtype IL-6 also had an increased affinity to the IL-6 receptor. With the exception of Leu"', the variants with substitutions in the region between residues Ser'I8 and ArglS3 competed poorly against wild-type IL-6 for the IL-6 receptor. As indicated in Fig. 2 most of these variants had less than 2% receptor binding capacity compared to wild-type IL-6. This indicates that the loss of biological activity in those IL-6 variants can be explained by an inability to bind to the IL-6 receptor and is not due to a defective signal transduction upon IL-6 receptor binding. It can be deduced from Fig. 2 that the receptor binding region in the COOH terminus of IL-6 is defined to the region between residues Ser17' and ArgIs3 (Ser-Leu-Arg-Ala-X-Arg).

DISCUSSION
In order to study the importance of individual residues in the COOH terminus of IL-6 for biological activity and receptor binding we constructed a library of mutants with single amino acid substitutions in the last 15 amino acids of IL-6 using saturation mutagenesis. This technique is a very efficient tool to make several mutations in a region of the protein of interest (24,32). In theory it is possible to obtain every mutation in the region of interest by contaminating the phosphoramidites with a low amount of each of the other three during chemical synthesis of oligonucleotides covering the target region to be studied (22). We were interested mostly in creating a library of single amino acid mutations to obtain information on the importance of each individual residue.
After analyzing 201 clones by dideoxy sequencing, all 15 amino acids had been replaced and a total of 63 unique mutants, of which 32 had single amino acid substitutions, were obtained. The random mutagenesis procedure allowed us to study the role of each residue in the COOH terminus of IL-6 in detail.
We were interested in studying the COOH terminus of IL-6 because several reports suggested its importance for biological activity. A number of deletion mutants with truncations at the COOH terminus have a decreased biological activity (20)(21)(22), whereas truncations up to 28 residues at the NH2 terminus did not alter biological activity (21,33). Additionally, the COOH terminus has recently been identified to be important in receptor binding for various other cytokines homologous to IL-6 (18,19). Based on computer modeling studies, Bazan proposed earlier that several cytokines and colonystimulating factors have a common helical structure at the COOH terminus that might be involved in receptor binding (16). This amphipathical a-helical structure is defined as an a-helix with opposing polar and nonpolar faces oriented along the long axis of the helix (34). A cartoon of the amphipathical helix structure of the COOH terminus of IL-6 is shown in Fig. 3. As is indicated, two distinct faces can be seen at the surface of the helical wheel: a hydrophobic face mostly consisting of Leu residues, and a highly hydrophilic face with mostly charged residues at the surface. Amphipathical structures are found in a variety of proteins. Their functional properties are multiple and may include enhancement of receptor binding probably by providing the proper geometry for the ligand to interact with the receptor (35).
A detailed study on the importance of an a-helical amphipathic structure for ligand-receptor interaction was reported by Cunningham and Wells (17). They studied the growth hormone-growth hormone binding protein interaction by alanine scanning mutagenesis. This is of particular interest for our study, because it is suggested, based on the secondary structure prediction of IL-6, that IL-6 has a similar tertiary structure as growth hormone, including the receptor binding regions. This tertiary structure includes an antiparallel bundle of four a-helices (A-D) with a distinctive loop topology (36). In growth hormone, the central portion of the D-helix to the COOH terminus was found to be important for receptor binding. Disruption of the COOH-terminal a-helical structure in growth hormone by replacing individual residues with alanines resulted in a significant loss of growth hormone receptor binding and it was concluded that T Y~"~ plays a central role. Based on the homology of IL-6 with growth hormone, Bazan proposed t h a t t h e L y~'~~-G l n '~~-P h e '~~ region in the COOH terminus of IL-6 and specifically the conserved aromatic residue in Phe'74 are important for IL-6 receptor binding (16,31). In our study we made several mutations in that region and found that the biological activity and receptor binding of those IL-6 variants were not reduced. However, we identified that the primary sequence Ser-Leu-Arg-Ala-X-Arg (positions 178-183) is essential for IL-6 to bind to its receptor. Mutations affecting the amphipathical nature of the helix in this region decreased biological activity and receptor binding significantly. One mutation that actually keeps the amphipathic helix intact, Arg"' to Lys, nevertheless resulted in a loss of receptor binding, indicating a crucial role for this specific residue in receptor binding. The importance of the other Arg at position 183 was already suggested by the biological inactivity of an IL-6 deletion mutant that lacked the last three residues (22). We show that this residue is indeed critical for biological activity, because several IL-6 variants with a substitution of this residue were inactive due to a loss of receptor binding. We suggest that Led7' is important for receptor binding, whereas Leu''' does not seem to be important. However, our random mutagenesis procedure resulted only in nonconservative substitutions at position 179 and only hydrophobic residues at position 182 as a substitution for the leucines. Therefore we cannot exclude the possibility that a hydrophobic residue at position 179 may be sufficient to keep the amphipathical nature of the helix intact and consequently result in normal IL-6 receptor binding. As can be seen in Fig. 3, the binding region (residues 178-183) is confined to less than two turns in the helix and the side chains project to two different faces of the helix. This is in contrast t o growth hormone, where all important residues project to the same face of the helix. This might indicate that in IL-6 both faces of the a-helix are involved in receptor binding.
Recently, the tertiary structure of granulocyte macrophagecolony-stimulating factor (GM-CSF) was elucidated by x-ray crystallography (37). It has a similar bundle structure of four helices as is determined for growth hormone and IL-4 (36,38), and predicted for IL-3, IL-6, erytropoietin, and G-CSF (16,19). The receptors of all these proteins belong to the same class I cytokine receptor family (16,39). Based on the x-ray structure, two putative receptor binding sites are located in the center of the GM-CSF molecule. No involvement of the COOH terminus in receptor binding was indicated, which is in accordance with earlier structure-function studies on GM-CSF (40). So although proteins such as IL-6, growth hormone and GM-CSF share striking similar structural patterns, the location of their receptor binding domains seems to be different. All identified receptor binding regions, however, share an amphipathical helical structure necessary for a strong and specific interaction between ligand and receptor (17,37,41).
The interaction of IL-6 with its receptor starts by forming a complex with an 80-kDa receptor subunit followed by the binding of this complex to a second subunit, gp130, resulting in signal transduction (14). IL-6 by itself cannot bind to gp130, nor can the 80-kDa receptor (14). It is not known whether complex formation of IL-6 with the 80-kDa receptor results in a conformational change in IL-6 or in the 80-kDa receptor that makes a domain available on one of them, or that they mutually form a domain that recognizes gp130. This consecutive binding of a ligand to one subunit, followed by the binding of this complex to an accessory receptor subunit is common in various cytokines, including IL-2 and IL-3 (42,43). Structure-function analysis of mouse IL-2 indicated that two distinct regions in IL-2 bind to different IL-2 receptor subunits (42). Recently it was shown that two distinct regions in growth hormone bind in a similar way to two growth hormone receptor molecules, resulting in dimerization of the receptor by one growth hormone molecule (44). So for both these ligands several domains that are important for receptor binding have been identified. Because several other regions in IL-6 besides the COOH terminus are important for biological activity, other receptor binding sites have also been predicted for IL-6, including the region between residues 29 through 34 and residues 153 trough 162 (33,45). These different regions might be involved in creating the epitope on IL-6 that binds to gp130 after interaction of IL-6 with the 80-kDa receptor.
For our receptor binding studies we used the human myeloma cell line U-266, which has a high number of IL-6 receptors (14). Although the results of the activity assay are in agreement with the results of the binding assays for most variants, one IL-6 variant had a 3-fold higher biological activity than receptor binding capacity. This may be explained by the fact that we used a human hepatoma cell line for the determination of the biological activity and a human myeloma cell line for receptor binding. The difference in the relative amount of the two IL-6 receptor subunits on these two cell lines may influence the binding capacity of the IL-6 variants and result in a tissue specific variation. Because the results of the activity assay were in agreement with the binding assay for all the other variants, we feel that our results indicate that a change in biological activity of our IL-6 variants compared to wild-type IL-6 can be explained by a change in receptor binding. Clearly, the biologically inactive variants that we constructed by substituting amino acids in the COOH terminus of IL-6 are not binding to the IL-6 receptor. Future structure-function studies on the different regions of IL-6 that are suggested to be involved in receptor binding and the use of soluble receptor subunits for binding assays might provide more insight in the interaction of IL-6 with both the receptor subunits. This is especially important for the development of IL-6 antagonists and agonists.
In recent years, naturally occurring or engineered variants of several cytokines such as IL-1 and IL-2 have been identified that have a reduced or defective biological activity, but still maintain receptor binding (46,47) These variants have been used as antagonists for cytokine activity and have shown very useful in animal disease models (46). Antagonists of IL-6 might also be of therapeutic use in several disease states in which IL-6 is involved in the pathogenesis, such as multiple myeloma, proliferative glomerulonephritis, and sepsis (48)(49)(50). Because IL-6 can act as a potent growth factor for myeloma cells it is considered of main pathogenetic importance in multiple myeloma (11,51). Studies using neutralizing murine monoclonal antibodies against IL-6 have shown promising results in patients with end-stage multiple myeloma (51). The immunization encountered during treatment with monoclonal antibodies, as has been reported in several patients, stresses the need of a nonantigenic IL-6 antagonist (52). So far we have not identified antagonists for IL-6, nor have others performing structure function experiments. In all the IL-6 variants that have been constructed and appropri-ately tested, the decrease of the biological activity was paralleled by a decreased receptor binding (19,40). In this study we have identified a potential agonist for IL-6 activity; the IL-6 variant with Phe17' replaced by Leu had a higher biological activity on a human hepatoma cell line compared to wildtype IL-6 that was associated with an increased binding capacity of this variant to the IL-6 receptor on a human myeloma cell line. An agonist of IL-6 might be of therapeutic value as an inducer of platelet production in patients with thrombopenia, in combination with other cytokines useful for patients undergoing bone marrow transplantation or used as an inhibitor of cell growth in patients with myeloid leukemia (10,531.