Binding of Leukemia Inhibitory Factor (LIF) to Mutants of Its Low Affinity Receptor, gp190, Reveals a LIF Binding Site Outside and Interactions between the Two Cytokine Binding Domains*

The gp190 transmembrane protein, the low affinity receptor for the leukemia inhibitory factor (LIF), belongs to the hematopoietin family of receptors characterized by the cytokine binding domain (CBD). gp190 is one of the very few members of this family to contain two such domains. The membrane-proximal CBD (herein called D2) is separated from the membrane-distal one (called D1) by an immunoglobulin-like (Ig) domain and is followed by three fibronectin type III repeats. We used truncated gp190 mutants and a blocking anti-gp190 monoclonal antibody to study the role of these repeats in low affinity receptor function. Our results showed that the D1Ig region was involved in LIF binding, while D2 appeared to be crucial for the proper folding of D1, suggesting functionally important interactions between the two CBDs in the wild-type protein. In addition, a point mutation in the carboxyl terminus of the Ig region strongly impaired ligand binding. These findings suggest that at least two distinct sites, both located within the D1Ig region, are involved in LIF binding to gp190, and more generally, that ligand binding sites on these receptors may well be located outside the canonical CBDs.

The receptor for the cytokine leukemia inhibitory factor (LIF) 1 comprises gp190, a transmembrane protein with low affinity for this cytokine, and the gp130 signal-transducing chain, which is the low affinity receptor for oncostatin M (1), as well as the signal transducer for IL6, IL11, ciliary neurotrophic factor (CNTF), OSM, and cardiotrophin-1 (CT-1) (reviewed in Ref. 2). The last three cytokines also use gp190 as part of their high affinity receptor complex (3)(4)(5), in conjunction with a specific low affinity binding subunit in the case of CNTF (6) and probably CT-1 (7). gp190 belongs to the large and growing family of hematopoietin-binding receptors, which is characterized by the presence of the 200-amino acid-long cytokine binding domain (CBD), which comprises two modules each of around 100 amino acids, containing 4 conserved cysteine residues in the amino-terminal one and a consensus WSXWS motif in the carboxyl-terminal one. gp190 is unusual because it contains two CBDs (8), like a few other receptors, i.e. c-Mpl (9), Ob-R (10), and KH-97 (11), which are, respectively, the thrombopoietin receptor, the leptin (Ob) receptor, and the ␤-common signal transducing chain shared by IL3, IL5, and granulocyte/ monocyte-colony stimulating factor in humans. The murine homologs of KH-97, AIC2A (12) and AIC2B (13), also contain two CBDs. gp190 and other receptors of this family have additional domains in their extracellular regions, such as an immunoglobulin-like (Ig) module of around 100 amino acids situated between the two CBDs in gp190, and a membraneproximal region encompassing 300 amino acids and similar in structure to three repeats of type III fibronectin (FN region). Comparison of the primary structure of gp190 with those of other family members showed that it was most homologous with a group including the alternative OSM receptor OSM-R␤ (14), the signal-transducing chain gp130, G-CSF-R, IL12-R␤, and Ob-R, with the percentage of amino acid identity ranging from 32% for the former to 20% for the latter (14).
For several members of this family of receptors, the extracellular region is composed only of one CBD, as is the case with the receptors for erythropoietin, prolactin (PRL-R), and IL2 (␤ chain, IL2-R␤), for example. Therefore, the CBD was thought to be fully responsible for the cytokine binding. This hypothesis was confirmed by deletion studies and single-point mutagenesis analysis of many members of this family of receptors, including growth hormone receptor (15), IL2-R␤ (16), AIC2A (17), IL6-R gp80 (18), G-CSF-R (19), PRL-R (20), and CNTF-R ␣ chain (21). Because gp190 contains two CBDs and is involved with four different cytokines, one could speculate that these two domains, as well as the Ig region and the FN repeats, do not have the same importance for ligand binding. In this study, the relationships among the different constitutive domains of gp190 and LIF and each other were examined using deletion mutants of the receptor and a panel of anti-gp190 monoclonal antibodies (mAbs).

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis of gp190 -The cDNA encoding human gp190 was obtained from Dr. C. Wood (Genetics Institute, Boston, MA). Soluble gp190 (sgp190) consisting of the extracellular region of the receptor, was obtained by inserting in frame, using polymerase chain reaction, an XbaI restriction site immediately followed by a stop codon at the junction with the transmembrane domain of the molecule, i.e. after nucleotide 2674 or amino acid 832 of the original gp190 sequence * This work was supported by the French Association pour la Recherche sur le Cancer and the Ligue contre le Cancer de Gironde. 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.
Deletion mutants within sgp190 were prepared by site-directed mutagenesis using the pALTER-1 phagemid system (Promega, Charbonnières, France), and following the manufacturer's recommendations. sgp190 digested from pGEM-3Zf(Ϫ) with XhoI and XbaI, was subcloned in pALTER-1 cut with SalI and XbaI. Oligonucleotides were synthesized that allowed creation of an SpeI site or an XbaI site without insertion or deletion, respectively, at the very beginning of the membrane-distal CBD D1 following the signal sequence (oligonucleotide 001), at the junctions between D1 and the Ig region (oligonucleotide 006), between Ig and the membrane-proximal CBD D2 (oligonucleotide 002), and at the junction between D2 and the FN region of the molecule (oligonucleotide 003). They generated mutants sgp190(001), sgp190(002), sgp190(003), and sgp190(006), which were verified by restriction analysis and DNA sequencing. Taking advantage of a unique PstI restriction site in the FN region of gp190 cDNA, all these mutants were subcloned in pEDr by exchanging the PstI fragment from pEDr-sgp190myc, to generate sgp190(001)myc, sgp190(002)myc, sgp190(003)myc, and sgp190(006)myc.
Since all the mutations were made in the same reading frame, this strategy allowed us to easily obtain by subcloning deletion mutants lacking one or several of these domains, all fused to the c-Myc epitope. The following battery of truncated variants of gp190 were directly constructed from these mutants in the pEDr plasmid: FNmyc, D1IgD2myc, D1Igmyc, D1myc, D1IgFNmyc, D2myc, and IgD2FNmyc.
Expression and Metabolic Labeling of gp190 Mutants-Two cell lines were used. Mycoplasma-free simian COS cells and Chinese hamster ovary (CHO) dehydrofolate reductase (DHFR) Ϫ/Ϫ cells were maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc., Les Ulis, France) supplemented with 8% fetal calf serum (Life Technologies, Inc.). Culture medium for CHO cells also contained nucleosides (adenosine, deoxyadenosine, thymidine) at 10 g/ml each (Sigma), to circumvent DHFR deficiency of these cells.
Transfections were performed as follows. COS cells were transiently transfected using the DEAE-dextran method, with 5 g of plasmid DNA. CHO cells were stably transfected by electroporation at 300 V and 900 microfarads using an Easyject Plus apparatus (Eurogentec, Seraing, Belgium), and selection was started the next day in medium without nucleosides to allow emergence of DHFR-producing cells.
Metabolic labeling was performed 72 h after transfection of COS cells, or on a confluent monolayer of CHO cells, 3-4 weeks after transfection (around 10 7 cells/dish). Cells were starved for 2 h in 4 ml of Dulbecco's modified Eagle's medium without methionine and cysteine supplemented with 2 mM glutamine and 5% dialyzed fetal calf serum; 200 Ci of [ 35 S]methionine/cysteine (Tran 35 S-label, ICN, Orsay, France) were added per dish for 12 h. Then, supernatants were harvested and stored at 4°C until use.
Immunoprecipitations-To check for the effective secretion of the recombinant proteins, 1 ml of COS or CHO supernatant was precleared with 0.05 ml of a 50% suspension of protein A-Sepharose beads (Affi-Gel protein A; Bio-Rad, Ivry-sur-Seine, France) for 1 h at 4°C under continuous rolling. Beads were eliminated by centrifugation, and supernatants were incubated with 30 g of the anti-Myc mAb 9E10 for 2 h under similar conditions. Immune complexes bound to protein A were sedimented by rapid centrifugation, and beads were washed three times with 1 ml of washing buffer (50 mM Tris, 1 mM EDTA, 150 mM sodium chloride, 0.2% Nonidet P-40, pH ϭ 8.0). Bead pellets were resuspended in 0.025 ml of sample loading buffer containing 0.1 M dithiothreitol and boiled for 5 min. Proteins were separated by SDS-PAGE on 10% gels and visualized by fluorography.
To study the low affinity complex formation between human LIF and its gp190 receptor or our deletion mutants, supernatants containing the 35 S-labeled receptor component were first incubated for 1 h with 0.4 g of CHO-derived LIF obtained as described previously (24). Then the non-blocking anti-LIF mAb 1F10 (30 g) (25) was added for another 2 h. Complexes were immunoprecipitated with protein A, as described above. An alternative protocol was also used; 35 S-labeled LIF was produced as described above, and 0.3 ml were incubated with 0.65 ml of supernatant containing the non-labeled gp190 mutant. Then the recep-tor was immunoprecipitated with 30 g of the specified mAb in 0.05 ml.
Preparation of Anti-human gp190 Monoclonal Antibodies-The production and the characterization of a first series of anti-human gp190 mAbs have been described elsewhere (26). mAbs 12D3 and 2G3, which have not been described so far, were obtained using the sgp190myc protein as the immunogen. Domains recognized by the anti-gp190 antibodies were analyzed in a flow cytometric assay, using CHO cells expressing membrane-bound glycosylphosphatidylinositol-linked forms of the sgp190 deletion mutants D1IgD2 and D2FN as described previously (26). Distinction among D1Ig, D2, and FN depended on the flow cytometric profile obtained for these two cell lines. For example, mAbs 12D3 and 2G3 bound to both of these deletion mutants, demonstrating that they recognized D2.
Functional Assays in Ba/F3 Cells-Ba/F3 cells coexpressing wildtype gp130 together with chimeric sgp190 or its mutants fused to the transmembrane and intracellular region of gp130 were obtained as follows. Like the strategy used for gp190, gp130 cDNA was first mutated at the junction between the extracellular and the transmembrane regions to create a unique XbaI restriction site, allowing easy generation of chimeric gp190 mutants by one step subcloning in the pEDr vector. The chimeric receptors were cotransfected into Ba/F3 in combination with wild-type gp130 in the pRcglo vector which contains the neomycin phosphotransferase resistance gene. Transfected cells were selected with both LIF and G418 (Life Technologies, Inc.), as described previously (27). The rationale for expressing these chimeric forms of gp190 was the reported inability of the gp190 intracellular region to transduce a proliferative signal in Ba/F3 cells in the absence of the gp130 intracellular region (28). In that report, a chimera expressing the G-CSF extracellular region and the transmembrane and intracellular regions of gp190 did not proliferate upon dimerization in response to G-CSF, in contrast to a G-CSF-R-gp130 chimera. Two benefits were expected from this approach. First, co-transfection of the deleted chimeric receptors fused to transmembrane and intracellular regions of gp130 together with gp130 would not impair the emergence of transfectants expressing the two types of receptors and showing dependence on LIF via dimerization of intracellular gp130. Second, transfection of the deleted chimeric receptor alone would easily allow the emergence of cells spontaneously signaling through homodimerization independently of any cytokine stimulus, which would not be expected with truncated mutants of sgp190 fused to transmembrane and intracellular regions of gp190, since they do not trigger proliferation of Ba/F3 cells. The cell lines raised upon progressive replacement of IL3 with LIF were then tested for their dependence on LIF or OSM (R&D Systems, Indianapolis, IN) and the expression of the receptors by flow cytometry as described elsewhere (26), using the anti-gp190 mAbs raised in the laboratory and the H1 anti-gp130 mAb kindly provided by Dr. J. Brochier (INSERM U291, Montpellier, France).
Radioiodination of LIF-Escherichia coli-derived human LIF (Pep-roTech Inc., Rocky Hill, NJ) was iodinated according to the chloramine-T method (29). LIF was labeled at a specific radioactivity of around 35,000 cpm/fmol. Binding experiments were carried out in PBS containing 0.5% bovine serum albumin (PBS-BSA) as described previously (29). The binding data was subjected to regression analysis using a one-or two-site equilibrium-binding equation (Grafit, Erathicus Software, Staines, United Kingdom). Binding to gp190 mutants was performed either with the Ba/F3 transfectants cultured for 3 days in the presence of IL3 instead of LIF, then washed three times and resuspended in PBS-BSA, or with the mutants of the soluble receptor, as follows. Sgp190myc or a myc-tagged mutant in CHO or COS supernatants (0.5 g, quantified with a sandwich enzyme-linked immunosorbent assay specific for human gp190; Refs. 26 and 30), was incubated with 10 g of anti-Myc mAb 9E10 and 0.01 ml of a 50% suspension of protein A-Sepharose beads, in 0.1 ml of PBS-BSA, for 2 h at 4°C under continuous rolling. Free LIF was separated from LIF bound to beads by centrifugation through a dibutylphthalate cushion at 15,000 rpm for 10 min.

Production of sgp190myc and Its Deletion Mutants-
The extracellular region of the low affinity LIF receptor, sgp190, was subjected to site-directed mutagenesis as described under "Experimental Procedures," to introduce a unique restriction site at the boundaries between the different modules D1, Ig, D2, and FN (Fig. 1A). The mutations induced two amino acid changes at these positions, except for mutation 003, which induced only one change (Table I). The mutants and native sgp190 were fused COOH-terminally and in-frame to the nu-cleotide sequence encoding the c-Myc epitope recognized by mAb 9E10. From these fusions, a panel of secreted gp190myc (sgp190myc) truncated mutants was obtained, which included D1IgD2myc, FNmyc, and mutants lacking either the membrane-proximal CBD D2 (D1IgFNmyc, D1Igmyc, and D1myc) or the membrane-distal CBD D1 (IgD2FNmyc and D2myc).
These constructs were transiently expressed in COS cells, and several of them were stably expressed in CHO cells. Production of the recombinant proteins was assessed using metabolic labeling of the cells, followed by immunoprecipitation with the anti-Myc mAb 9E10. Results of immunoprecipitations from COS cells supernatants are depicted in Fig. 1B. Wild-type sgp190myc and all but the D1Igmyc and the D1myc proteins could be detected in variable amounts in the culture supernatants, with molecular masses corresponding to what was expected from the deletions performed, thereby showing that the recombinant proteins were correctly processed and secreted. D1Igmyc and D1myc constructs were stably transfected into CHO cells but, as for COS cells, these recombinant proteins were not secreted into the supernatant. In this study, sgp190myc, sgp190(002)myc, and D1IgD2myc were derived from CHO cells, and FNmyc, D2myc, IgD2FNmyc, D1IgFNmyc, sgp190(001)myc, sgp190(003)myc, and sgp190-(006)myc were obtained from COS cells. Despite repeated attempts, D1Igmyc and D1myc could never be obtained in these two cell lines, and thus could not be further analyzed.
Irrelevance of the FN Region for the Reconstitution of the LIF-sgp190 Complex in Solution- 35 S-Labeled sgp190myc from CHO cells was incubated with 10 nM 40-kDa CHO-derived human LIF (0.4 g/ml). LIF was then immunoprecipitated using the non-blocking anti-LIF mAb 1F10 (24). SDS-PAGE and autoradiography were carried out to detect the coprecipitation of the low affinity LIF receptor. The radiolabeled sgp190myc was immunoprecipitated by the anti-LIF mAb only in the presence of LIF (Fig. 2), as no specific band was detected with 1F10 when LIF was omitted. We therefore concluded that sgp190 fully retained its binding capacity when produced in CHO cells. The complex was also immunoprecipitated using sgp190myc from COS cells, showing that COS cells were also capable of producing functional sgp190 (data not shown), and that the recombinant proteins produced in this cell line could be used as well. Preliminary experiments showed that LIF binding to sgp190myc was dose-dependent, with a maximum signal obtained with 10 nM LIF. At a higher concentration, an excess of free LIF may have saturated the 1F10 mAb, thereby decreasing the signal. Considering the signal intensity of labeled sgp190myc with as little as 0.2 nM LIF, this system appeared to be suitable for the detection of sgp190 mutants with at least a 50-fold decrease in their affinity for LIF (data not shown).
The role of the FN segment of gp190 in LIF binding was investigated in our immunoprecipitation assay with the truncated D1IgD2myc form of sgp190myc lacking the FN domain, and the isolated FNmyc fragment. The full-length point mutants sgp190(001)myc and sgp190(003)myc, which were used to generate both of these truncated receptors, were also assayed. Fig. 3 shows that sgp190(001)myc and sgp190(003)myc were still able to bind LIF, as the anti-LIF mAb 1F10 immunoprecipitated the labeled protein only after preincubation with LIF. This finding demonstrated that amino acid changes induced by mutations 001 and 003 neither altered the sgp190 conformation nor involved residues implicated in ligand binding, at least to a significant extent. The truncated D1IgD2myc protein was also able to bind LIF, whereas the FNmyc fragment was not (Fig. 3). In a binding experiment with iodinated LIF followed by immunoprecipitation via anti-Myc mAb, the affinity of LIF for D1IgD2myc was measured at 15 Ϯ 7 nM, which was similar to that of sgp190myc in this assay (see Fig. 7C). Therefore, the membrane-proximal FN region of gp190 is not involved in LIF binding, a function that appeared to be accorded to one or the two CBDs D1 and D2 separated by the Ig region.
A Crucial Role for D2 in Maintaining a Functional D1-To determine the relative importance of each of the two CBDs for LIF binding, we attempted to reconstitute the ligand-receptor complex with sgp190 mutants lacking one of them. As described above, the proteins encoded by the D1Igmyc and D1myc constructs were never detected in the supernatants of transfected COS and CHO cells. However, using the anti-Myc mAb, it was possible to immunoprecipitate, from cell lysates, small amounts as several isoforms of different sizes probably corresponding to various maturational steps of these proteins (data not shown). This finding suggested protein instability impairing its intracellular processing and leading to intracellular degradation prior to secretion. Since the FN region was not able by itself to bind the cytokine, the binding function of D1Ig was studied using the D1IgFNmyc mutant whose FN region forced the secretion of fused D1Ig (Fig. 1). The D1IgFNmyc protein was recognized by the anti-Myc mAb 9E10, but it did not bind LIF since it was not precipitated using the anti-LIF mAb in the presence of LIF (Fig. 4A). The D1Ig conformation in the D1IgFNmyc protein was assessed by immunoprecipitating the 35 S-labeled protein using a panel of mAbs specific to the D1Ig region of human gp190 that we recently characterized (26). Fig. 4B shows that none of six mAbs recognizing different epitopes in the D1Ig region could immunoprecipitate radiolabeled D1IgFNmyc, although the anti-Myc mAb did. Among them, four recognized conformation-dependent epitopes (10B2, 1B4, 6G8, and 1C7) because they could not bind to denatured sgp190myc in Western blot, unlike the 12D9 and 6C10 mAbs, which apparently bound to linear epitopes (26). These observations strongly suggested that the D1Ig spatial conformation was profoundly altered, thus explaining why it could no longer bind to the ligand. However, D1IgFNmyc could be immunoprecipitated by a polyclonal anti-D1IgD2 antiserum (26), indicating that the protein was correctly translated in the cells (data not shown). Therefore, as suspected for D1Igmyc and D1myc, the absence of D2 seemed to markedly impair protein conformation, which in turn might have abrogated the binding capacity of these mutants.
The CBD D2 Is Unable to Bind LIF in the Absence of D1-To investigate whether D2 could directly interact with LIF, the truncated IgD2FNmyc receptor lacking D1 was assayed in the receptor-reconstitution assay (Fig. 5A). It was well secreted and recognized by the anti-Myc antibody, but failed to bind LIF, since the anti-LIF mAb 1F10 did not immunoprecipitate the radiolabeled truncated receptor in the presence of LIF. In the binding experiment with iodinated LIF, no affinity of LIF for this mutant could be measured (see Fig. 7C). The spatial conformation of D2 was assessed using the two conformationdependent anti-D2 mAbs we have obtained so far, 8C2 and 2G3, which bind to two different epitopes on D2. Both recognized the IgD2FNmyc deletion mutant in the immunoprecipitation assay (Fig. 5B), whereas the anti-D1Ig 1C7 and 10B2 mAbs did not. This observation suggested that D2 is most probably properly folded. Therefore, the absence of D1 did not seem to substantially modify D2 conformation, in contrast to what was observed for D1 with mutants lacking D2, but the absence of D1 impaired the capacity of the truncated mutant to interact with LIF.
LIF Binding Directly Involves the D1Ig Region-We previously reported that the anti-D1Ig mAb 1C7 specifically and dose-dependently inhibited the LIF-induced proliferation of Ba/F3 cells expressing wild-type human gp130 and gp190 (26). We therefore investigated, using our immunoprecipitation assay, whether the blocking activity of 1C7 was mediated through competition with LIF for the binding to the low affinity receptor. In such a case, mAb 1C7 would interfere with the precipitation of the radiolabeled cytokine bound to the nonlabeled receptor. Fig. 6 shows that the labeled ligand was recognized by anti-LIF mAb 1F10, and that the LIF-sgp190myc could be efficiently immunoprecipitated by the non-blocking anti-gp190 10B2. Conversely, the blocking anti-D1Ig mAb 1C7 did not precipitate any LIF-sgp190 complexes. This failure was not due to a lower ability of 1C7 to immunoprecipitate sgp190myc, since both 10B2 and 1C7 bound equally well to the receptor in this assay (data not shown, and Fig. 7B with an sgp190 mutant). Therefore, the blocking effect of 1C7 was most likely explained by competition with LIF for its receptor. This result suggested that the D1Ig region was directly involved in the interaction with LIF.
A LIF Binding Site in the Ig Region Close to the Junction with D2-Since the deletion mutants lacking D2 had been obtained by subcloning from sgp190(002)myc and sgp190(006)myc point mutants, these latter have also been assayed for their abilities to bind LIF (Fig. 7A). The sgp190(006)myc protein was fully capable of interacting with LIF, indicating that the amino acid changes induced by this mutation at the junction between D1 and the Ig region did not impinge on its function. Unexpectedly, the sgp190(002)myc mutant, whose residues Phe 328 and Ala 329 in the carboxyl terminus of the Ig region were mutated, respectively, to Thr and Ser, was unable to bind LIF in our system. A possible explanation was that mutation 002 disrupted the overall conformation of the molecule, thereby affecting its ability to bind LIF, as demonstrated above for D1IgFNmyc. We then immunoprecipitated the sgp190(002)myc mutant with a panel of conformation-dependent anti-D1Ig mAbs, including the blocking 1C7, and with anti-D2 mAbs. Because all these mAbs recognized the sgp190(002)myc protein (Fig. 7B), we deduced that the conformation of the protein was not significantly altered by the mutation. In the equilibrium binding of radioiodinated LIF to soluble receptor via anti-Myc mAb, no binding of LIF to sgp190(002)myc could be measured (Fig. 7C), while the affinity of LIF for sgp190(006)myc was 15 Ϯ 7 nM, which was similar to that for the wild-type sgp190myc. This also suggested that LIF binding to mutant 002 is of very low affinity. Therefore, the lack of binding that we observed was most probably due to the disruption of a punctual site exquisitely involved to some extent in the interaction with LIF.
The sgp190(002) mutant receptor was fused to the transmembrane and intracellular region of gp130, and transfected in Ba/F3 cells together with wild-type gp130. Double-transfectants were selected based on their capacity to grow in the absence of IL3 but in the presence of LIF, and tested for: 1) dependence on LIF for proliferation, 2) membrane expression of the receptor chains by flow cytometry, and 3) binding characteristics of iodinated LIF. Results were compared with those obtained with a cell line transfected with both a non-mutated gp190-gp130 chimeric construct and wild-type gp130, which was raised simultaneously. Both cell lines proliferated in a dose-dependent manner in the presence of LIF, OSM, or IL3 as control. However, the cells bearing the mutated gp190(002) had dramatically lower capability (80 -100-fold) to grow in the presence of subsaturating concentrations of LIF, whereas no difference could be noted for OSM and IL3 (Fig. 8A). Both cell lines  2. Reconstitution of the LIF-sgp190 complex in solution. 35 S-Labeled sgp190myc as supernatant from transfected CHO cells was incubated with 10 nM cold LIF (lane 1) or without LIF (lanes 2 and 3), and immunoprecipitated with the anti-LIF mAb 1F10 (lanes 1 and 2) or the anti-Myc mAb 9E10 (lane 3). expressed comparable surface levels of gp190 and gp130, as determined in flow cytometry using the anti-gp190 10B2 and the anti-gp130 H1 (31) mAbs (Fig. 8B). Therefore, the different responses of the two cell lines to LIF and OSM could not be explained by a limiting amount of gp190 or gp130 membrane receptors, and the unaltered function of mutant 002 in response to OSM suggested that the intrinsic function of this mutated receptor was not impaired by the mutation.

FIG. 4. The truncated D1IgFNmyc receptor does not bind LIF and does not fold properly. A,
FIG. 6. The anti-gp190 1C7 blocking antibody competes with LIF for binding to gp190. 35 S-Labeled LIF was incubated with (lanes 3 and 5) or without (lanes 1, 2, and 4) cold sgp190myc, and immunoprecipitated with the non-blocking anti-gp190 mAb 10B2 (lanes 2 and 3) or the blocking anti-gp190 mAb 1C7 (lanes 4 and 5). LIF was also precipitated directly with the anti-LIF mAb 1F10 (lane 1). nated LIF (K d impossible to determine accurately), explaining why much higher concentrations of LIF were required to induce the proliferation of this cell line. As a whole, a selective loss of binding affinity for LIF due to the disruption of a LIF binding site at the carboxyl terminus of the Ig domain impaired the function of gp190. DISCUSSION We demonstrated that the 300-amino acid-long membraneproximal FN region was not able by itself to bind LIF and that its deletion did not impair the capacity of the remaining upstream fragment to interact normally with the cytokine. Therefore, the binding site(s) lie(s) within the two CBDs D1 and D2 separated by the Ig-like region. A homologous FN region is also found in the G-CSF-R (33), the IL12-R ␤ chain (34), the Ob-R (10), and the IL6 signal transducer gp130 (35). Similar deletions have also been made in G-CSF-R and gp130, and led to the same conclusions (36,37). Therefore, it appears to be a general feature in this family of receptors that the FN region,  (lanes 1-3) and sgp190(002)myc (lanes 4 -6) were incubated with 10 nM cold LIF (lanes 2 and 5) or without LIF (lanes 1, 3,  4, and 6). The receptor was immunoprecipitated (arrowhead) with the anti-LIF mAb 1F10 (lanes 1, 2, 4, and 5) or the anti-Myc mAb 9E10 (lanes 3 and 6). B, 35 S-labeled sgp190(002)myc was immunoprecipitated (arrowhead) with the anti-Myc mAb (lane 1), an unrelated mAb (lane 2), the anti-D1Ig mAbs 1B4, 10B2, 6G8, and 1C7 (lanes 3-6), and the anti-D2 mAbs 8C2 and 12D3 (lanes 7 and 8). C, iodinated LIF was incubated with 0.5 g of sgp190myc (q), sgp190(006)myc (E), D1IgD2myc (OE), IgD2FNmyc (Ⅺ), or sgp190(002)myc (‚), and immunoprecipitated via anti-myc mAb and protein A beads, before separating bound and free LIF. The curve depicts the average binding of sgp190myc, D1IgD2myc, and sgp190(006)myc. when present, has no direct or indirect function in the binding to the specific ligand.
The respective deletion of either D1 or D2 in the truncated IgD2FNmyc or the D1IgFNmyc receptors abolished ligand binding in our immunoprecipitation systems. This observation suggested either that both missing CBDs were necessary for the interaction between LIF and gp190, or that these truncated molecules had an altered tertiary structure responsible for the incapacity to bind LIF. Indeed, it is well known that the correct folding of the protein is absolutely essential for its exportation outside the cell and its ability to bind to its ligand(s), as has been shown with the growth hormone receptor (15,38), the IL2-R ␤ chain (39), the PRL-R (20), the AIC2A (17), the IL6-R ␣ chain (18), and the erythropoietin receptor (40). The truncated D1myc or D1Igmyc receptors were suspected of folding improperly since they could not be recovered from cell supernatants. Despite being well secreted, the D1IgFNmyc was not recognized by anti-D1Ig mAbs, attesting to the profound alterations in the folding of this part of the molecule. As a consequence, no definitive conclusion could be drawn as to the function of D1Ig in the binding to LIF, at this step. In contrast, the IgD2FNmyc protein was immunoprecipitated by conformationdependent anti-D2 mAbs, but was not capable of binding LIF in our two binding assays. Isolated D2myc could also be recovered from cell supernatants and was recognized by anti-D2 antibodies (data not shown). Although these two mutants strongly suggested that D2 folding was not significantly altered in the absence of D1, it also argued against a direct involvement of D2 in LIF binding. Conversely, the proper folding of D1 seemed to depend on the presence of D2, and the minimal truncated mutant with detectable LIF binding capability was D1IgD2myc, which as expected harbored a correctly folded D1Ig region, as demonstrated with a panel of anti-D1Ig mAbs (data not shown). In experiments not shown, we also replaced D2 by the homologous CBD from human gp130, which does not bind LIF directly, but the chimeric protein produced was also unable to bind the cytokine either, and still bore an improperly folded D1Ig region. Therefore, in the wild-type gp190 receptor, direct interactions seem to exist between the two CBDs, which are crucial for receptor conformation and consequently for ligand binding.
In this regard, the behavior of gp190 appears to be different from that of human KH-97 and its murine counterparts AIC2A and AIC2B. These receptors also contain two CBDs, but they immediately follow each other without an intercalated Ig region. Residues important for cytokine binding have been located in the membrane-proximal CBD of AIC2A and KH-97 (17,41), and a truncated mutant of KH-97 lacking the membrane-distal CBD was correctly expressed on the cell surface, and remained functional and dependent on IL3 for signal transduction (42). This observation demonstrates that, for this particular receptor, one domain involved in ligand binding is sufficient to achieve its own proper folding. This situation contrasts with the results described here for gp190. The Ig domain lying between D1 and D2 could help reconcile these data; its persistence as a hinge would help maintaining biological function of the gp190 protein by allowing interactions between the two CBDs.
The role of the D1Ig region in LIF binding was also studied using the anti-D1Ig mAb 1C7, which inhibits LIF-induced proliferation of Ba/F3 cells coexpressing gp130 and gp190. Blocking mAb 1C7 was able to compete with LIF for sgp190 in our immunoprecipitation assay. Although the 1C7 did not bind to the IgD2FNmyc protein, and in the absence of available isolated D1 for mapping experiments, we cannot exclude that it recognizes a conformational epitope in the Ig region. Overall, these results, in agreement with the immunoprecipitations performed with the truncated receptors lacking D1, emphasize a major role for D1 and/or Ig in LIF binding.
The gp190(002) mutant, which bore a punctual mutation at the junction between the Ig region and the membrane-proximal CBD D2, also shed light on the LIF-gp190 molecular interactions. First, despite a proper folding as assessed with anti-gp190 mAbs, sgp190(002)myc did not bind LIF in the immunoprecipitation assay. This finding also implied that mutation 002 could not be held responsible for the altered conformation of the D1Ig region in the D2-deleted D1IgFNmyc truncation mutant, which was derived from mutant 002. Second, when expressed in Ba/F3 cells, the chimeric gp190(002)-gp130 receptor remained able to trigger LIF-induced proliferation, only in the presence of gp130, which indicates the reconstitution of a high affinity functional tripartite receptor complex. Although the mutant's biological activity was 80 -100 times lower than that of the non-mutated gp190-gp130 chimeric receptor, this experiment demonstrated that it was still able to bind LIF. This difference could not be explained by significantly different amounts of membrane low affinity receptors and high affinity converters, which were similar in both cell lines as assessed by flow cytometry. The intrinsic binding ability and signal transduction capacity of mutant 002 was preserved, since the response to OSM was not altered. Scatchard analysis of LIF binding to Ba/F3 cells expressing gp130 and the mutant, or to soluble sgp190(002)myc, revealed a markedly decreased affinity for LIF. Taking into account that at a concentration of 8.5 nM LIF there is less than 5% binding of LIF to sgp190(002)myc as compared with non-mutated sgp190myc, it can be calculated that the K d of mutant 002 for LIF is higher than 450 nM, corresponding to at least a 30-fold loss in affinity. On Ba/F3 cells surface, a few receptors with higher affinity still remained, which were thought to be responsible for the residual ability to trigger proliferation of Ba/F3 cells. Therefore, mutation 002 was introduced in an area that is directly and selectively involved in the binding to LIF, thereby substantially decreasing the affinity of the receptor for LIF. This could be the consequence of a faster off-rate, explaining why binding is not detectable at the level of the low affinity receptor, i.e. in the absence of gp130, in the experiments performed with sgp190(002)myc. On the contrary, on the Ba/F3 cell surface, the gp130 could stabilize the interaction between LIF and mutant 002, allowing signal transduction to occur and detection of higher affinity receptors, although in small numbers.
The only partial loss of binding and function of mutant 002 could mean that at least one other LIF binding site exists on gp190. Consistent with this hypothesis, previous reports involved two distinct binding sites for LIF on gp190 (28,(43)(44)(45). The anti-D1Ig blocking mAb 1C7 impaired LIF binding to gp190 but still recognized sgp190(002)myc. Therefore, in addition to the site at the carboxyl terminus of the Ig region, the second LIF binding site is most probably located upstream within D1Ig. Moreover, since the IgD2FNmyc deletion mutant, which does not bear the deleterious mutation 002, was unable to bind LIF in the immunoprecipitation assay, this second LIF binding site could well lie within D1, and not in the Ig region. This possibility is supported by other experiments not shown in Ba/F3 cells where the IgD2FN deletion mutant fused to the transmembrane and intracellular region of gp130 and transfected together with gp130 did not allow the emergence of LIF-dependent transfected cell lines. However, a residual activity of the IgD2FN receptor could be expected in Ba/F3 cells, since it does not bear the deleterious mutation 002 and therefore should bind LIF weakly. If, as proposed above, direct interactions occur between D1 and D2, then the Ig region would be expected to lose any mobility toward D2 in the wild-type receptor. Such spatial constraints would stiffen the junction between Ig and D2, and facilitate the interaction with LIF. It is conceivable that the deletion of D1 in the IgD2FN molecule frees the Ig region from its spatial constraints toward D2, thereby masking the LIF binding site and leading to the observed loss of LIF binding.
Mutation 002 lies one amino acid upstream from the 3Ј end of the sequence encoded by exon 7 of the gp190 gene, which encompasses the full Ig region (46). Recently, it has been reported, based on experiments using chimeric receptors between murine and human gp190 domains, that the Ig loop is involved in LIF binding (47). In that study, the downstream boundary of the Ig region was considered to be 7 amino acids downstream from our mutation 002. Therefore, the binding site we describe at mutation 002 position would still remain within Ig as defined by those authors. It is noteworthy that mutation 002 substitutes a phenylalanine with a threonine. Phenylalanine and other hydrophobic aromatic amino acids have often been involved in interactions with ligands, but so far, they have always been located within the CBDs (reviewed in Ref. 48). To the best of our knowledge, no such deleterious mutations have been described in any of the other receptors belonging to this family. Of note, deletion of the Ig region in the IL6-R ␣ chain did not decrease the binding of the cytokine (18), whereas a similar deletion in G-CSF-R only impaired the reconstitution of the high affinity complex (49). For these two receptors, the Ig region was proposed to play a role in the oligomerization of the receptor chains in high affinity complexes (49,50). From the experiments described herein, the Ig region, at least in gp190, could be involved in other or additional roles, first as a LIF binding module, and second as a structurally constrained hinge sandwiched between the two CBDs whose direct interactions seem necessary for receptor conformation and function.