Mutations in the immunoglobulin-like domain of gp190, the leukemia inhibitory factor (LIF) receptor, increase or decrease its affinity for LIF.

The leukemia inhibitory factor (LIF) receptor comprises the low affinity binding chain gp190 and the high affinity converter gp130. The ectodomain of gp190 is among the most complex in the hematopoietin receptor family, because it contains two typical cytokine receptor homology domains separated by an immunoglobulin-like (Ig-like) domain. Human and murine gp190 proteins share 76% homology, but murine gp190 binds human LIF with a much higher affinity, a property attributed to the Ig-like domain. Using alanine-scanning mutagenesis of the Ig-like domain, we mapped a LIF binding site at its carboxyl terminus, mainly involving residue Phe-328. Mutation of selected residues into their orthologs in the murine receptor (Q251E and N321D) significantly increased the affinity for human LIF. Interestingly, these residues, although localized at both the amino and carboxyl terminus, make a spatially unique LIF binding site in a structural model of the Ig-like module. These results demonstrate definitively the role of the Ig-like domain in LIF binding and the potential to modulate receptor affinity in this family with very limited amino acid changes.

The leukemia inhibitory factor (LIF) 1 low affinity receptor gp190 belongs to the large family of the hematopoietin receptors, which are characterized by a consensus cytokine receptor homology (CRH) domain. The extracellular region of gp190 is unusual in that it has two CRH domains, herein called D1 for the amino-terminal membrane distal domain and D2 for the membrane proximal domain. D1 and D2 are separated by an immunoglobulin-like (Ig-like) module of around 100 amino acids, and D2 is followed by three type III fibronectin modules (1). The gp190 receptor participates in the high affinity receptor complex for 5 human cytokines (reviewed in Ref. 2), namely LIF, oncostatin M (OSM), ciliary neurotrophic factor (CNTF), cardiotrophin-1 (CT-1), and neurotrophin-1/B cell-stimulating factor 3 (NNT-1/BSF-3). The gp190 also exists in a soluble form capable of binding to LIF, behaving as a competitor of membrane gp190 and as an inhibitor of LIF's biological effects.
Little information relative to the function of each module of gp190 is available to date. The CRH domain is usually implicated in binding to the ligand, and for this reason has been called the cytokine binding domain. The function of the Ig-like module is less clear. In some cases, it has also directly participated in ligand binding. Indeed the Kaposi's sarcoma-associated herpes virus-derived interleukin (IL)-6 (vIL-6) homolog directly binds to the Ig-like domain of its unique receptor chain gp130 (3), this latter being also the high affinity converting chain for the LIF/gp190 complex. Similarly, human OSM interacts with both the CRH and the Ig-like domains of its low affinity receptor chain gp130 (4). Alternatively, the Ig-like module of gp130 does not directly interact with the human cytokine IL-6, but is required when two gp80/IL-6/gp130 complexes dimerize to form the signal transducing hexamer (5). As a final example, the Ig-like domain of the granulocyte-colony stimulating factor (G-CSF) receptor has also been reported to participate in the dimerization of two receptor chains triggered by the ligand (6) and also directly to ligand binding (7).
With respect to human gp190 (hgp190), we ascribed to the Ig-like module the role of a hinge allowing the two CRH domains to interact with each other (8,9). On the other hand, murine gp190 (mLIFR), which shares 76% sequence homology with hgp190, displays a much higher affinity for human LIF (hLIF) than hgp190 does (10). This could be reproduced by the replacement of the Ig-like module of hgp190 with its murine homolog (11). In addition, when analyzing a series of deletion mutants lacking one or several modules of the hgp190 ectodomain, we identified a mutation in the carboxyl terminus of the Ig-like module (mutation 002, i.e. F328T/A329S), which led to a dramatic decrease of hLIF binding (8). Given these preliminary results, we decided to investigate more precisely the role of the Ig-like module of hgp190 in binding to hLIF. Site-directed mutagenesis was performed, and the single point mutants generated were expressed as soluble and membrane-bound forms. They were analyzed for their ability to bind hLIF and trigger LIF-dependent proliferation. In a first series of mutants the * This work was supported by the Ligue Nationale contre le Cancer (Comités de la Dordogne et des Pyrénées-Atlantiques). 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.

EXPERIMENTAL PROCEDURES
Cell Culture-All eukaryotic cell lines were grown in 5% CO 2 at 37°C in a water-saturated atmosphere either in Dulbecco's modified Eagle's medium (Invitrogen) for COS and CHO DUCKX cells or RPMI 1640 medium for Ba/F3 cells, in the presence of 8% fetal calf serum (Invitrogen) and 2 mM L-glutamine. For CHO DUCKX cells 10 g/ml of adenosine, deoxyadenosine, and thymidine were added to the medium. Ba/F3 transfectants were grown in the presence of hLIF at 50 ng/ml. For functional studies, the cells were first starved of hLIF for 3 days by culture in medium supplemented with murine IL-3 (mIL-3) in the form of a COS cell supernatant diluted 1:200 and then washed three times to remove mIL-3.
Construction of the Human gp190 Mutants-Site-directed mutagenesis of the extracellular region of hgp190 (soluble hgp190 or shgp190) was performed using the pAlter-1 phagemid system (Promega, Charbonnières, France) following the manufacturer's recommendations as previously described (8). Mutational oligonucleotides were synthesized (Cybergene, Evry, France) that allowed for the creation of the desired mutation together with a new restriction site, or the inactivation of an existing one wherever possible, in order to allow for easy screening of the mutated plasmids. All punctual shgp190 mutants were subcloned in the pEDr vector, either as a carboxyl-terminal c-Myc-tagged soluble receptor (shgp190myc) or fused to the transmembrane and intracellular region of hgp130 (hgp190/130 chimeras) as previously described (8). The mutants created were all verified by sequencing (ESGS, Evry, France).
Transfection of Cells-COS cells were transiently transfected with 5 g of plasmid DNA encoding the shgp190myc constructs using the DEAE-dextran method as previously described (8). Culture supernatants were collected at day 4, and the recombinant receptor was quantitated by ELISA. Stable transfectants were obtained by electroporating CHO DUCKX DHFR(Ϫ) cells with the cDNAs encoding the soluble receptors and selecting the transfectants in nucleoside-free medium based on the DHFR gene carried by the pEDr plasmid. For stable expression of the transmembrane forms of hgp190 mutants in Ba/F3 cells, the chimeric receptors were cotransfected by electroporation into Ba/F3 cells in the presence of wild-type hgp130 cDNA. Transfected cells were selected in the presence of G418 (Invitrogen) and by progressive replacement of mIL-3 by hLIF as previously described (8).
Flow Cytometry Staining of Cells-The membrane expression of the hgp130 and of the hgp190 mutants on the Ba/F3 cell lines was detected by flow cytometry using the anti-D1 mAb 6G8 and the anti-D2 mAbs 8C2 and 12D3 for hgp190 (12), and using the anti-hgp130 mAbs AM64 and H1 for hgp130. An IgG1 irrelevant isotype-matched mAb was used as a negative control. The staining was performed as previously reported (8) and was analyzed with a three color FACScalibur flow cytometer (BD PharMingen) equipped with the CellQuest software. To analyze the turnover of the transmembrane form of the LIF receptor or its mutants from the cell surface, the indicated Ba/F3 cell lines were incubated with hLIF at 100 ng/ml or without any added cytokine for the indicated times at 37°C, before being stained with the anti-hgp190 8C2 mAb or an isotype-matched negative control mAb.
ELISAs for the Quantitation of shgp190 and of the LIF/shgp190 Complexes-Secreted shgp190myc and its mutants were quantitated using a sandwich ELISA specific for hgp190, according to the previously reported procedure (12,13). The capture mAb was the anti-D1 mAb 6G8, and the biotinylated developing mAb was the anti-D2 mAb 8C2. In vitro hLIF binding to shgp190myc or its mutants was assessed as follows. All steps were performed at room temperature, and plate washes were performed in PBS containing 0.05% Tween 20. ELISA plates were coated with the anti-D1 mAb 6G8, and saturated with PBS/BSA for 1 h. Plates were washed once and then incubated with supernatants containing shgp190myc or its point mutants adjusted at a constant and saturating concentration of 150 ng/ml for 2 h and washed three times. Some of the mutants with lower secretion levels were concentrated to reach 150 ng/ml using polyethylene glycol 35000 (Fluka, Sigma). CHO-derived hLIF ranging from 0.256 ng/ml to 4000 ng/ml was added for 2 h. After three washes, plates were incubated with the biotinylated non-blocking anti-hLIF mAb 1F10 (14) at a final concentration of 2 g/ml for 1.5 h and washed again three times. Subse-quent steps, i.e. incubation of peroxidase-labeled streptavidin and addition of substrate were as described for the ELISA for quantification of shgp190.
Proliferation Assay of Ba/F3 Cells-The proliferative response to hLIF, mLIF, hOSM, and mIL-3 of the transfected Ba/F3 cell line was assessed by a colorimetric proliferation assay using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, Sigma), after 3 days of culture, as described previously (8). To analyze the inhibitory activity of the soluble receptor toward hLIF action, we used Ba/F3 cells expressing wild-type full-length hgp190 and hgp130. A mixture of hLIF (1 ng/ml final concentration) and shgp190myc or soluble mutant Q251E/N321D at the indicated final concentrations, or murine serum as a source of mouse LIF receptor (mLIFR) at the indicated final dilutions, was added to the cells. Cell proliferation was analyzed 3 days later. Whenever specified, cell culture medium was also supplemented with extra amounts of soluble receptors on day 1 and day 2.
Affinity Chromatography for the Purification of shgp190myc Mutants-The purified anti-D1 mAb 10B2 (5 mg) was coupled to a 1-ml Hi-Trap-NHS-activated column (Amersham Biosciences) according to the manufacturer's instructions. Culture supernatant (500 ml) of CHO cells stably producing the recombinant soluble receptor was passed through the column. After washing with 5 ml of PBS at pH 7.5 containing 1 M KCl and 0.02% Tween 20, bound receptor was eluted with 0.1 M glycine at pH 2.5 as 1-ml fractions, and acidity was immediately neutralized with a one-third volume of Tris-HCl at pH 8.0. After extensive dialysis against PBS, the receptor was quantitated by ELISA, and the fractions with the highest concentrations were pooled.
Radioiodination Experiments-Stably transfected Ba/F3 cells were cultured for 3 days in the presence of mIL-3 instead of hLIF. After one PBS wash, cells were resuspended in PBS/BSA with or without the blocking anti-hgp130 B-R3 mAb (20 g/ml) and incubated at 4°C for 30 min. Binding experiments were carried out in PBS/BSA for experiments using labeled Escherichia coli-derived hLIF (PeproTech Inc., Rocky Hill, NJ), or in PBS/BSA containing 5 mM mannose-6-phosphate for experiments using CHO-derived hLIF, to avoid binding to the mannose-6-phosphate/insulin-like growth factor II receptor (Man 6-P/IGFII-R) (15,16). hLIF was iodinated using the chloramine T method as described (17), at a specific radioactivity of around 35,000 cpm/fmol. Regression analysis of the binding data was accomplished using a oneor two-site equilibrium binding equation (Grafit, Erathicus Software, Staines, UK).
Surface Plasmon Resonance Analysis of LIF Binding-These experiments were performed with the BIACore 2000 optical biosensor (BIA-Core, Uppsala, Sweden). hLIF purified from transfected CHO cells was covalently coupled to a carboxymethyl dextran flow cell (CM5, BIACore) as recommended by the manufacturer. The level of immobilization obtained was 1,500 resonance units. Binding of purified CHO-derived soluble receptors was assayed at concentrations ranging from 0.67 to 33 nM in Hepes-buffered saline and at a flow rate of 40 l/min. Association was monitored for 5 min before initiating the dissociation phase for another 11 min with Hepes-buffered saline. Regeneration of the flow cells was achieved with 5 mM glycine-HCl at pH 2.0. The resulting sensorgrams were analyzed using the BIAEvaluation software (BIACore).
Analysis of STAT3 Phosphorylation-Ba/F3 cells expressing wildtype hgp190 and hgp130 were used. CHO-derived hLIF at a final concentration of 10 ng/ml giving maximal STAT3 phosphorylation was incubated with or without the purified soluble receptors, or with normal mouse serum as a source of mLIFR, at twice the indicated concentrations at 37°C for 20 min, before the cells (2 ϫ 10 5 per condition tested) were added in an identical volume. After 5 min at 37°C, the cells were lysed in 50 l of Triton X-100 lysis buffer as described previously (9). After centrifugation, the supernatant was harvested, and the total protein concentration was determined using the bicinchoninic acid method (Sigma). The cell lysate (10 g of protein/lane) was boiled 5 min and separated by SDS-PAGE on 8% gels and then transferred to a nitrocellulose membrane (Amersham Biosciences). The immunoblotting was performed as previously reported (9) with the rabbit anti-phospho-STAT3 antibody (Cell Signaling Technology, Ozyme, Saint-Quentin-en-Yvelines, France) or with the rabbit anti-actin antibody (Sigma) as a control. The proteins were visualized using the chemiluminescence ECL system (Amersham Biosciences).
Design of a Three-dimensional Model of the Ig-like Domain of hgp190 -A structural model of the Ig-like domain of hgp190 was constructed using the program MODELLER (18) using a sequence alignment of the Ig-like domain of hgp130 with the corresponding region of hgp190 and the crystallographic coordinates of the Ig-like domain of hgp130 (PDB accession number: 1I1R, Ref. 3). The quality of the model was assessed using ProsaII.

Production and Expression of shgp190myc and Its Single
Point Mutants-In order to elucidate the role of the Ig domain in LIF binding, the region of hgp190 around mutation 002 (8), which abrogates its binding, was subjected to alanine-scanning site-directed mutagenesis (Fig. 1A). A total of 12 mutations were carried out extending from residues 321 to 334, except for the existing alanine at position 329, which was mutated to glycine (Fig. 1B). Since the junction between the Ig-like module and D2 lies between residues 330 and 331, as inferred from the analysis of intron-exon organization of the hgp190 gene (1), 8 mutations were therefore localized in the Ig-like module and 4 were in the D2 domain (Fig. 1B). Additionally, we also replaced 6 amino acids in the human sequence of the Ig-like module by their murine orthologs as deduced from the alignment of the sequences from both species (Matcher Program, Ref. 19). Among these, 4 affected residues close to mutation 002 (i.e. N321D, I322V, F323Y, and I327V), and 2 affected residues localized in the amino terminus of the Ig-like module (i.e. Q251E and D266N) were chosen by virtue of electrostatic charge changes between both species (Fig. 1B). A double mutant Q251E/N321D was also constructed. The cDNAs for shgp190 and its mutants were fused carboxyl-terminal to the sequence encoding the c-Myc epitope recognized by mAb 9E10. To examine whether the resulting proteins were correctly produced and released in the cell culture supernatant, each cDNA construct was transiently expressed in COS cells, and the secreted receptor was quantitated with a sandwich ELISA specific for hgp190 using mAbs 6G8 and 8C2, which are directed against D1 and D2, respectively. We used mAbs specific for epitopes outside the Ig-like domain to ensure that their reactivity could not be affected by the mutations. In addition, these mAbs recognize conformation-dependent epitopes (12), giving information on the ternary conformation of the secreted receptor. All the hgp190 mutants were expressed at levels grossly comparable to that of shgp190myc, which was produced at 161 Ϯ 42 ng/ml, albeit mutants 002, F328A, A329G, G330A, Y331A, P332A, P333A, and D334A were at approximately 2-4fold lower levels. 2 Binding of hLIF to shgp190myc Mutants as Determined by ELISA-To determine the relative hLIF binding abilities of shgp190myc and its mutants, we developed an ELISA to detect the ligand/low affinity receptor complex in solution. Each mutant at a saturating concentration was trapped on a microtitration plate coated with the mAb 6G8 directed at the D1 domain. Serial dilutions of hLIF were then added, and hLIF bound to the receptor was detected with biotinylated anti-hLIF mAb 1F10, a mAb that does not affect hLIF binding (20). Fig.  2A (left panel) displays the dose-response curves obtained with representative mutants V326A, F328A, G330A, and P332A. For shgp190myc, 50% of the maximal binding (C50) was reached with hLIF at 14.0 Ϯ 3.2 ng/ml (mean Ϯ S.D. of three experiments), corresponding to 0.35 Ϯ 0.08 nM given that the molecular mass of CHO-derived hLIF is 40 kDa. The C50 obtained for all the mutants are listed in Fig. 2A (right panel) and compared (in percent) to the C50 value obtained with shgp190myc. Alanine mutants could be sorted into two groups. A first group contained mutants N321A, A329G, Y331A, P332A, P333A, and D334A, each displaying an ability to bind hLIF close to that of shgp190myc (C50 ratio between 50 and 120%). A second group consisted of F323A, T325A, V326A, I327A, F328A, 002, and G330A, each showing a severely impaired ability to bind hLIF (C50 ratio of 10% and lower). Among these, no hLIF binding could be detected at all for mutants 002 and F328A (indicated by two asterisks in Fig. 2A and shown with mutant F328A in the left panel) even at the highest hLIF concentration tested (4000 ng/ml). In contrast, a faint binding could be detected at high concentrations of hLIF for mutants F323A, T325A, V326A, I327A, and G330A (indicated by one asterisk in Fig. 2A and shown for mutant G330A in the left panel). Mutations that impaired hLIF binding were all localized in the Ig-like domain, because none of the mutations generated downstream of the Ig-like/D2 junction altered hLIF binding.
The human-to-mouse mutants could be sorted into three groups (Fig. 2B). Mutant D266N displayed the same hLIF binding ability as shgp190myc, whereas I322V, F323Y, and I327V showed a decreased binding. A third group consisted of Q251E, N321D, and the double mutant Q251E/N321D, which all possessed a significantly increased hLIF binding ability, with respectively 7, 14, and 35 times that of wild-type shgp190myc.
There was no correlation between the production levels of the mutants in the cell culture supernatants and their ability to bind hLIF, strongly suggesting that the mutations performed did not alter the conformation of the receptors or not to an extent that could impair their function.
hLIF, hOSM, and mLIF Binding to the Membrane-bound Form of the Alanine-scanning hgp190 Mutants-To analyze the ability of the hgp190 single point mutants to trigger a LIF-dependent signal, the c-Myc tag of the mutants was replaced with the transmembrane and intracellular regions of hgp130, and mIL-3-dependent Ba/F3 cells were transfected with wild-type hgp130 and either each of the mutants or a chimeric nonmutated hgp190/130 as a positive control. Stable cell lines were obtained with all the mutants. In contrast, transfection of hgp130 or hgp190/130 alone never allowed the selection of cells in the presence of hLIF.
Membrane expression of the receptor chains was analyzed by flow cytometry. ative Ba/F3 cell lines bearing hgp130 in combination with either F328A, G330A, or P332A chimeric mutants of hgp190. The expression level of the receptor chains may vary from one transfection experiment to another, but the variations observed were not associated with particular mutations.
All the transfectants obtained were then tested for their ability to proliferate in the presence of hLIF and hOSM. Doseresponse proliferation curves are shown in Fig. 3B for the representative mutants F328A, G330A, P332A, and the hgp190/130 control cell line. Cytokine concentrations giving half-maximum proliferation (EC 50 ) are given in Table I. All the cell lines proliferated in a dose-dependent manner in the presence of hLIF (Table I and Fig. 3B, left panel). The non-mutated hgp190/130 receptor supported Ba/F3 cell proliferation with an EC 50 of 0.18 Ϯ 0.05 ng/ml, i.e. 4.5 Ϯ 1.25 pM. As expected, the mutations leading to conserved hLIF binding ability in the ELISA also behaved like the non-mutated receptor in the Ba/F3 proliferation assay (see mutant P332A in Fig. 3B). Mutation F328A, which led to complete abrogation of hLIF binding ability in the ELISA resulted in a sharp (80 -100-fold) decrease in Ba/F3 cells proliferation in the presence of hLIF (EC 50 of 18.67 Ϯ 6 ng/ml), and fully accounted for the decrease observed with mutation 002 (EC 50 of 16.38 Ϯ 4.7 ng/ml). In contrast, mutations that only partially impaired hLIF binding in the ELISA had no influence on cellular proliferation, as shown for mutant G330A in Fig. 3B. The proliferation curves presented depict a representative experiment, and the apparent differences in response to hLIF, which can be seen between the mutants, are not biologically significant with the exception of mutant F328A (see also Table I). The cell transfectants did not proliferate at all in the absence of added cytokine and proliferated comparably in the presence of mIL-3, demonstrating that their intrinsic ability to proliferate was fully retained after transfection. 2 All the Ba/F3 cell lines also proliferated in a dose-dependent manner in the presence of hOSM (Fig. 3B, middle panel for representative mutants and Table I), and no difference was noted between all the mutants and the hgp190/ 130 cell line (EC 50 of 2.02 Ϯ 0.37 ng/ml, i.e. 100 Ϯ 18 pM). Of note, mutation F328A had no effect on the response to hOSM. The response to mLIF was also analyzed for the F328A cell line and the hgp190/130 cell line (Fig. 3B, right panel and Table I). As with hLIF, mutation F328A also strongly impaired the response to mLIF, and the barely detectable residual activity did not allow the calculation of the EC 50 for mLIF. Therefore, residue Phe-328, which is conserved between the human and mouse species, is crucial for binding both human and murine LIF.
Using iodinated hLIF, we measured the high and low affinity dissociation constants of the receptors expressed by Ba/F3 cell lines harboring hgp130 in association with hgp190/130 or the F328A and G330A mutants (Table II). The low affinity binding component (representing hLIF binding to isolated hgp190) was measured in the presence of the anti-hgp130 blocking mAb B-R3, which is known to inhibit the heterodimerization of hgp130 with hgp190 (21,22). On the hgp190/130 cell line, the high and low affinity binding components were characterized by equilibrium dissociation constants (K d ) of 0.21 Ϯ 0.03 nM and 2.4 Ϯ 0.8 nM respectively, values which are consistent with previous reports (17). For mutant G330A the K d for the high affinity binding was unchanged (K d , 0.23 Ϯ 0.07 nM), whereas the K d for the low affinity binding was slightly higher (K d , 6 Ϯ 1.4 nM), in agreement with its lower hLIF binding ability in the ELISA. For mutants F328A and 002, a very low hLIF binding was observed, which did not permit the determination of the high and low affinity binding components.
hLIF, hOSM, and mLIF Binding to the Membrane-bound Form of the Human-to-Mouse hgp190 Mutants-Membranebound forms of the human-to-mouse mutants were also constructed and transfected with hgp130 in Ba/F3 cells. We used as a control the native full-length mLIFR. Cell lines were obtained with all these constructs. Membrane expression of the mutated receptors and hgp130 was confirmed by flow cytometry, and Fig. 4A depicts the staining obtained for three representative cell lines Q251E, N321D, and Q251E/N321D with the anti-hgp190 mAb 6G8 and the anti-hgp130 mAb AM64.
All the cell lines raised were tested for their ability to proliferate in the presence of hLIF, hOSM, and mLIF. Dose-response proliferation curves are shown in Fig. 4B for mutant Q251E/N321D, and the control cell lines expressing mLIFR or hgp190/130. The EC 50 concentrations are given for all the mutants in Table III. The Ba/F3 cell line bearing the hgp190/ 130 chimera was as responsive to hLIF (EC 50 of 0.18 Ϯ 0.05 ng/ml) as the one bearing mLIFR (EC 50 of 0.13 Ϯ 0.03 ng/ml); was 40-fold more responsive to hOSM; and was 140-fold less sensitive to mLIF (EC 50 of 0.26 Ϯ 0.08 ng/ml versus EC 50 of 37.00 Ϯ 5.60 ng/ml) (Table III and Fig. 4B). All the mutations performed conferred dose-dependent proliferation with the three cytokines, and all the cell lines responded identically to mIL-3 (data not shown). None of the mutations performed altered the response to any of the three cytokines, and the minimal differences that can be seen on the figure are not biologically significant (p Ͼ 0.05 for all the cell lines tested in comparison to the hgp190/130 receptor, see Table III), especially for the mutants with a higher affinity for hLIF in the ELISA (EC 50 around 0.10 ng/ml for Q251E, N321D, and Q251E/N321D; p Ͼ 0.1 for all three).
The effect of the Q251E/N321D mutation on the high and low affinity hLIF binding components was measured by the method of Scatchard, in comparison with the non-mutated hgp190/130 chimeric receptor or the mLIFR (Table II). In the context of low affinity binding (gp190 component), the K d for hgp190/130 was around 15-fold lower than for mLIFR (K d of 2.4 Ϯ 0.80 nM for hgp190 versus 0.16 Ϯ 0.02 nM for mLIFR). The dramatic increase in hLIF binding brought by the double mutation Q251E/ N321D, as determined in the ELISA, corresponded to a 3-fold increase in the affinity of the receptor for hLIF (K d of 0.77 Ϯ 0.16 nM). In the context of high affinity binding (in the presence of hgp130), hLIF displayed a slightly higher affinity for the mLIFR/hgp130 complex than for the human receptor complex. However, this could not be explained by the Q251E/N321D mutation, which did not increase the affinity of hLIF for the human receptor complex (Table II).
Surface Plasmon Resonance Analysis of the Binding of hLIF to the Q251E/N321D Mutant of hgp190 -The kinetic association and dissociation constants k on and k off for mutant Q251E/ N321D in a soluble form were determined by surface plasmon resonance toward immobilized hLIF (Fig. 5). As a negative control, we used mutant 002, which did not bind hLIF at all. Both shgp190myc and mutant Q251E/N321D bound hLIF with a similar association rate (3.0 ϫ 10 5 M Ϫ1 s Ϫ1 versus 2.3 ϫ 10 5 M Ϫ1 s Ϫ1 ). However, the dissociation rate was around three times lower for the mutated receptor (1.3 ϫ 10 Ϫ3 s Ϫ1 versus 4.2 ϫ 10 Ϫ3 s Ϫ1 ), leading to a 2.5 times lower K d (5.7 nM versus 14.1 nM for shgp190myc). In addition to confirming the result obtained by the Scatchard method, this experiment demonstrated that mutations Q251E/N321D increased the affinity of the receptor for hLIF via an improvement of the stability of the ligand/receptor complex.
Enhancement of Membrane Receptor Turnover by Mutations Q251E/N321D-The failure to observe an improved hLIF responsiveness for mutant Q251E/N321D despite a significantly enhanced binding to hLIF prompted us to analyze whether the improvement of the binding could come along with a faster turnover of the receptor from the cell surface (Fig. 6), as already demonstrated for hgp190 and hgp130 (23,24). Ba/F3 cells expressing hgp130 and either hgp190/130, mutant 002 or mutant Q251E/N321D in their transmembrane form were incubated with hLIF (100 ng/ml) or in the absence of any added cytokine for up to 320 min. At the indicated times, the level of membrane receptors was determined by flow cytometry and expressed as percentage of the amount measured at the beginning of the incubation period. In the absence of hLIF, we observed that the amount of each of the membrane receptors gradually increased until 160 min, before starting to decrease. In the presence of hLIF, mutant 002 slightly increased, whereas hgp190/130 and mutant Q251E/N321D were dimin- ished as early as 80 min after the addition of hLIF. However, at 320 min, only mutant Q251E/N321D was still decreasing whereas hgp190/130 returned to the basal level.
Enhancement of the Blocking Activity of shgp190 on hLIFdependent Cell Proliferation by Mutations Q251E/N321D-Since soluble gp190 is an inhibitor of LIF action by competing with the membrane-bound receptor for the binding to LIF, we hypothesized that the mutant Q251E/N321D in a soluble form should be a more potent inhibitor of LIF due to its higher affinity for the cytokine. We therefore analyzed its ability to impair hLIF-induced proliferation of the Ba/F3 cell line expressing wild-type hgp190 and hgp130, with hLIF used at a saturating concentration of 1 ng/ml (Fig. 7, hatched histograms). Soluble hgp190myc had only a marginal inhibitory effect at high concentrations (15% at 1.67 g/ml), whereas mutant 002 did not impair hLIF-induced proliferation at all, as could be expected due to its very low affinity for hLIF. We also used normal mouse serum as a source of mLIFR since it is known to contain about 1-2 g/ml of this soluble receptor (11,25). Mouse serum strongly inhibited the proliferation of the Ba/F3 cells, although it exerted a nonspecific trophic effect at low dilutions (1:50 and 1:100, see Fig. 7 and data not shown), which counterbalanced the blocking effect. Half-maximal blockade was obtained with a serum dilution between 1:200 and 1:400, which grossly corresponds to 5-10 ng/ml of soluble mLIFR. Mutant Q251E/N321D displayed a stronger inhibitory activity than shgp190myc on cell proliferation (30% inhibition versus 15%, respectively, at 1.67 g/ml), but was far less efficient than mouse serum. ELISA measurement of soluble gp190 in culture medium maintained at 37°C showed that 40 -50% of the shgp190myc and mutant Q251E/N321D is no more detectable as early as day 1, suggesting that it is unstable in the conditions of the proliferation assay (data not shown). Therefore, in order to try to improve the effect of the soluble human receptors, we added in the cell culture medium the shgp190myc   a Cytokine concentration giving 50% maximal proliferative response, mean Ϯ S.D. of three experiments. No statistically significant differences were seen when comparing the Ba/F3 cell line expressing the non-mutated chimeric gp190/130 and all the Ba/F3 cell lines expressing the human-to-mouse mutants using a Student's t-test (p Ͼ 0.05).

FIG. 5. Analysis of the association and dissociation rates of hLIF to the double mutant Q251E/N321D in a soluble form.
CHO-derived shgp190myc, soluble mutant 002, and soluble mutant Q251E/N321D were purified by affinity chromatography, and the characteristics of their binding to immobilized CHO-derived hLIF were assessed with surface plasmon resonance. The association and dissociation (k on and k off ) rates and the dissociation constant (K d ) were determined for shgp190myc and soluble mutant Q251E/N321D. and mutant Q251E/N321D every day over the 3-day-long proliferation assay (Fig. 7, filled histograms). This treatment significantly improved the activity of the Q251E/N321D mutant while it only had a weak effect on shgp190myc (50% inhibition versus 20%, respectively).
Enhancement of the Blocking Activity of shgp190 on STAT3 Phosphorylation by Mutations Q251E/N321D-Because of the quick disappearance of the added receptor in solution, we performed a short term assay that employed hLIF-induced STAT3 phosphorylation in the Ba/F3 hgp190ϩhgp130 cell line, in order to compare the ability of shgp190myc and mutant Q251E/ N321D to inhibit hLIF signaling, with mutant 002 and mouse serum as controls. For this purpose, the soluble receptors at three different concentrations were preincubated with a saturating concentration of hLIF, before adding the cells. After 5 min, the cells were lysed, and the phosphorylation status of STAT3 was analyzed by immunoblotting with an antibody specific to the tyrosine-phosphorylated form of STAT3 (Fig. 8). Phosphorylated STAT3 was not detected in the absence of hLIF, but was strongly induced by the cytokine. A partial blockade of STAT3 phosphorylation was observed with 10 g/ml of shgp190myc. This inhibition was dose-dependent and specific since it was not observed with mutant 002 at the same concentration. In contrast, mutant Q251E/N321D almost completely blocked STAT3 phosphorylation at a concentration as low as 2 g/ml, demonstrating that it was a much stronger inhibitor of hLIF action than shgp190myc. In this assay, mouse serum was the most efficient since it completely blocked STAT3 phosphorylation at a 1:400 dilution (i.e. at an estimated concentration of 5 ng/ml).
Model of the Three-dimensional Structure of the Ig-like Domain-The three-dimensional structure of the Ig-like domain of the hgp190 has not yet been described. We therefore constructed a structural model of this domain based upon the coordinates of the Ig-like domain of hgp130 (5). Placing the location of the mutations described above on this model reveals that they are all predicted to be clustered on a single face of the molecule (Fig. 9), which corresponds in location to the site of interaction between the Ig-like domain of hgp130 and the cognate ligand vIL-6 (5). This region corresponds to the carboxylterminal G strand of the ␤ sandwich and some elements of the physically adjacent A strand including Gln-251. DISCUSSION Our alanine-scanning analysis of the carboxyl terminus of the Ig-like domain of hgp190 showed that mutation to alanine of 6 of 8 residues impaired hLIF binding in our ELISA assay, whereas in contrast hLIF binding was not affected by the mutation of any of the 4 residues at the very amino terminus of D2. Among the 6 residues thus identified in ELISA five were hydrophobic (Phe-323, Val-326, Ile-327, Phe-328, Gly-330) and one hydrophilic (Thr-325). However, in the Ba/F3 proliferation assay, only mutation F328A impaired receptor function. Scatchard determination of the affinity for hLIF confirmed that mutant F328A had a dramatically decreased ability to bind hLIF, whereas mutant G330A displayed only a slight affinity decrease (3-fold). To explain the apparent discrepancy between the ELISA and the Scatchard determination, one should consider that the former does not measure binding at equilibrium in contrast to the latter. In the ELISA, all the steps occurring after the removal of the excess hLIF are performed in the absence of the ligand. Part of the hLIF/hgp190 complexes can dissociate during this 1.5 h time lapse and the loss in the ELISA signal as compared with equilibrium conditions will be exponentially related to the kinetic dissociation constant of these complexes. Nevertheless, the ELISA helped to discriminate among the mutants with a strongly decreased ability to bind hLIF, those with a residual binding ability (i.e. Phe-323, Thr-325, Val-326, Ile-327, and Gly-330) at high hLIF concentrations, and mutant F328A for which no binding to hLIF could be detected even at the highest concentration of ligand. Therefore, the ELISA screening assay is more sensitive to minute variations in binding ability in comparison with the Scatchard determination. Residue Phe-328 is conserved in the mLIFR protein sequence, and we found that it was also important for mLIF-induced proliferation of Ba/F3 cell lines bearing the mutated human receptor. Therefore, this aromatic amino acid is crucial for ligand binding, as was previously reported for such residues in other cytokine receptors (7,26,27).
The importance of the carboxyl terminus of the Ig-like region of hgp190 in hLIF binding was confirmed with the analysis of the human-to-mouse mutants. Interestingly, the stretch of twelve residues that was analyzed by alanine-scanning presents a higher degree of interspecies homology than the overall Ig-like domain, since only four amino acids are different. Three of the four mutants generated displayed a strongly impaired ability to bind hLIF in the ELISA (I322V, F323Y, and I327V), as was found with the corresponding alanine mutations for two of them (Phe-323 and Ile-327). In contrast, mutation N321D increased hLIF binding 14-fold in the ELISA, while mutation to alanine had no effect. The amino terminus of the Ig-like domain also displays a high homology between both species, and we identified mutation Q251E in this region, which increased hLIF binding by 7-fold. Combining mutations Q251E and N321D led to a 35-fold increase in hLIF binding by ELISA. Scatchard and surface plasmon resonance analysis showed that this increased ability to bind hLIF corresponded to a 3-fold increase in affinity for hLIF, which could be fully attributed to a slower dissociation rate of the ligand/receptor complex. This is consistent with our interpretation of the higher sensitivity of the ELISA assay in detecting changes in hLIF binding ability, as discussed earlier. In addition, the 6G8 and 8C2 anti-hgp190 mAbs used in ELISA to measure the soluble receptor or the hLIF/receptor complexes, bound with comparable affinities to the gp190/130 chimera and to the membrane forms of the F328A and Q251E/N321D mutants, as determined by Scatchard analysis (data not shown). It becomes therefore evident that the results obtained with the ELISA cannot be attributed to fluctuations of the affinities of the anti-gp190 mAbs used toward the different mutated receptors.
Our results also show that the affinity for LIF of the mutants as determined by the Scatchard analysis does not strictly correlate with their potency to trigger cell proliferation. Indeed, only mutant F328A displayed an impaired ability to trigger cell proliferation. We suggest two possible explanations for this phenomenon. First, in the Ba/F3 proliferation assay, hgp130 may compensate at least partially for a loss of affinity of hLIF for hgp190, and this may occur via the stabilization of labile hLIF/hgp190 complexes. Indeed, we previously reported using a Ba/F3 cell line co-expressing hgp130 and mutant 002 that a few receptors with a high affinity still formed in the presence of hLIF, probably accounting for the residual proliferation effect (8). Second, a very low number of functional receptors is sufficient to trigger the activation signal, since it has been reported previously for the LIF-responsive murine DA1a that the occupancy of less than 5% of all 50 membrane receptors was sufficient to trigger half-maximal proliferation (17), similar results having also been obtained with the murine M1 cell line (28). Therefore, it may be rather difficult to alter the response to hLIF, unless the affinity for hgp190 is drastically diminished as it is in the case for mutant F328A. For the mutants with moderately decreased affinity for hLIF, no consequences on the function of the receptor should be expected. The same reasoning may be applied to mutant Q251E/N321D, i.e. the increase in affinity obtained may not be sufficient to have functional consequences. However, our results also show that the Ba/F3 cell line expressing mLIFR does not display a higher sensitivity to hLIF since the EC 50 are comparable (see Table III), despite the much higher affinity for hLIF of the mLIFR. Therefore, it is possible that the signal triggered by the activation of the receptor cannot be improved beyond that given by the wild-type receptor simply via an increase in affinity for the ligand. Alternatively, the mLIFR may not allow us to form a definitive conclusion, because we used it in combination with the hgp130 chain, and this heterodimer may not behave like the human high affinity receptor complex in this respect. As another explanation for the lack of functional effect of mutant Q251E/ N321D, we noticed that the enhanced binding to hLIF caused a faster turnover of this mutant in comparison to the non-mutated hgp190/130 receptor. Therefore, such a decrease in the pool of membrane receptor could also explain the failure to improve hLIF responsiveness, via a decrease in cellular hLIF binding capacity and/or a shortening in the duration of signal transduction through the activated receptors.
Although transmembrane mutant Q251E/N321D did not enhance the sensitivity to hLIF in comparison to the wild-type receptor, the soluble form was a more potent inhibitor of hLIF action on hLIF-sensitive cells than shgp190myc. This was more striking with the short term STAT3 phosphorylation assay than with the 3-day-long MTT proliferation assay, during which the integrity (via degradation) or activity of the soluble receptor was deeply affected. These findings demonstrate that FIG. 9. Model of the three-dimensional structure of the Ig-like domain. A structural model of the Ig-like domain of gp190 was generated and the surface rendered in GRASP. The location of the core glycine 324 is colored in red, and residues identified in this study to have an impact on ligand recognition are colored in blue. Residues whose side chains are predicted to exhibit solvent exposure in this model include Phe-323, Phe-328, and Gln-251. The amino and carboxyl termini of the Ig-like domain are indicated. an increase in hLIF binding ability may indeed have functional consequences, as shown with the soluble receptor, and reinforce our hypothesis that the lack of enhancing effect of the membrane form may be intrinsic to the functioning of the high affinity receptor complex.
Crystallographic and mutagenesis studies have shown that the A to D four-helix cytokine LIF interacts with gp190 via two epitope sites called site I and site III (10,29,30,37). Site III consists of the B-C loop, the C-D loop, and the second half of the D helix, and contributes the majority of the free energy for the binding (10,29,30). The cognate recognition epitopes on the gp190 ligands OSM and CNTF have also been defined. For these three cytokines, the most prominent feature of the site III epitope is a conserved solvent-exposed tryptophan residue in the amino terminus of the D helix (reviewed in Refs. 2 and 31). In the case of ligands such as IL-11, which interact with gp130 via site III, the interaction involves an analogous solvent-exposed non-polar phenylalanine residue (32). Inspection of the structure of the complex between vIL-6 and hgp130 reveals that this residue interacts with a glycine residue and the preceding phenylalanine residue via a pi-stacking mechanism (3). This glycine residue is conserved in the Ig-like domains of both hgp190 and OSM-specific receptor chain, as is the presence of an adjacent non-polar residue (Phe-323 in the case of gp190, Fig. 9). We therefore suggest that the site III interaction between LIF, OSM and CNTF with the hgp190 involves burying the exposed tryptophan of the ligand on the glycine core (Gly-324) with additional interactions occurring with the side chain of Phe-323. This arrangement is essentially similar to the site II interaction between OSM and hgp130 (31) except in this case the donor glycine core is on the ligand, and exposed hydrophobic residue is provided by the receptor. Following the principles established for other examples of this recognition strategy (2) we would expect that additional specificity would be conferred by the identity of residues surrounding the glycine core. The identity of such residues is also suggested from the mutagenesis studies and the model. Thus Phe-328 is predicted to form an exposed surface for interaction with ligand as is Thr-325 adjacent to the glycine core. Collectively these results suggest that the interaction between hLIF and the Ig-like domain of hgp190 conforms to a common pattern of ligand/receptor recognition: a non-polar core interaction is modified by residues in the immediate vicinity. In this respect we can therefore account for the increase in affinity observed in the Q251E and N321D mutants: the predicted fold of the Ig-like domain places these residues in the vicinity of the core (Fig. 9) and the consequence of the mutation is to introduce, directly or indirectly, additional interaction points with the ligand. Mutant Q251E/N321D did not increase the response of the Ba/F3 cells to mLIF. This suggests that by themselves these residues are not sufficient to improve mLIF binding, although mutant F328A showed that mLIF binds in the vicinity.
Interestingly, hLIF site III contains three positively charged lysine residues in a row in the C-D loop, whereas the carboxyl terminus of the Ig-like domain of mLIFR contains three negatively charged aspartic acid residues in a row, to which correspond one glutamic acid, one aspartic acid, and one positively charged asparagine (Asn-321) in the human receptor at positions 319 -321. Therefore, it is tempting to speculate that these acidic amino acids in the Ig-like domain could participate in the binding of gp190 to hLIF site III by engaging electrostatic interactions. If so, the higher affinity of mLIFR for hLIF could be explained by the difference at position 321, since mutation N321D could increase the affinity of hgp190 for hLIF via the creation of a stretch of three consecutive negatively charged amino acids as found in mLIFR.
This work illustrates a new approach for modulating the action of cytokines. Mutations of cytokines have been made that either enhanced their affinity for a low affinity binding receptor, or impaired their binding to the high affinity transducing chain, providing superagonists or antagonists, respectively (33)(34)(35). Here for the first time, we bring evidence that the ability to bind a cytokine can also be improved by mutating a very limited number of residues in the receptor sequence.