Ciliary Neurotrophic Factor, Cardiotrophin-like Cytokine, and Neuropoietin Share a Conserved Binding Site on the Ciliary Neurotrophic Factor Receptor α Chain*

Ciliary neurotrophic factor, cardiotrophin-like cytokine, and neuropoietin are members of the four-helix bundle cytokine family. These proteins signal through a common tripartite receptor composed of leukemia inhibitory factor receptor, gp130, and ciliary neurotrophic factor receptor α. Binding to ciliary neurotrophic factor receptor α occurs through an interaction site located at the C terminus of the cytokine AB loop and αD helix, known as site 1. In the present study, we have generated a model of neuropoietin and identified a conserved binding site for the three cytokines interacting with ciliary neurotrophic factor receptor α. To identify the counterpart of this site on ciliary neurotrophic factor receptor α, its cytokine binding domain was modeled, and the physicochemical properties of its surface were analyzed. This analysis revealed an area displaying properties complementary to the site 1 of ciliary neurotrophic factor, cardiotrophin-like cytokine, and neuropoietin. Based on our computational predictions, residues were selected for their potential involvement in the ciliary neurotrophic factor receptor α binding epitope, and site-directed mutagenesis was carried out. Biochemical, cell proliferation, and cell signaling analyses showed that Phe172 and Glu286 of ciliary neurotrophic factor receptor α are key interaction residues. Our results demonstrated that ciliary neurotrophic factor, cardiotrophin-like cytokine, and neuropoietin share a conserved binding site on ciliary neurotrophic factor receptor α.

proteins using the same tripartite receptor, and their bioactivities are mainly restricted to the nervous system (7).
In particular, CNTF maintains motor neuron viability in vitro, prevents the degeneration of axotomized motor neurons, and attenuates motor deficits in different strains of mice with neuromuscular deficiencies (8 -13). It is also a protective factor in demyelinating central nervous system disease (14). In addition to its activities in the nervous system, CNTF has trophic effects on denervated skeletal muscle and is a regulator of muscular strength in aging (15,16). Other findings show that CNTF is also a potential agent in the treatment of diabetes as well as obesity (17,18).
CLC, a CNTF paralog, shares 37% similarity with CNTF (19). We previously showed that CLC forms a composite cytokine when it is associated with the soluble receptor cytokine-like factor 1. Cytokine-like factor 1 mainly behaves as a chaperone protein, allowing secretion of CLC (2). Subsequently, we observed that CLC can form a second secreted composite cytokine when associated with the soluble form of CNTFR␣ (20).
We have recently reported inactivating mutations in the CLC gene of a patient suffering from cold-induced sweating syndrome (21). This syndrome is characterized by a profuse sweating after exposure to cold. This patient also displays congenital physical abnormalities, degenerative disease of the cervical and lumbar spine, and a mild sensorimotor peripheral neuropathy.
NP, the third CNTF paralog identified by our group (1), is expressed concomitantly and co-localized with CNTFR␣ in the mouse embryo, suggesting a key role in nervous system development. However, a deletion of 8 nucleotides inactivates the NP gene (NP) in humans. This finding suggests that NP has evolved into a pseudogene in humans.
To exert their biological functions, CNTF, CLC, and NP bind to a tripartite receptor complex, which includes gp130-and LIFR-transducing chains associated with the aforementioned specific ␣-binding chain, CNTFR␣ (1,2,(22)(23)(24)(25). This specific ␣ chain exists either in a soluble or GPI-anchored membrane form, which in both cases contributes to gp130 and LIFR recruitment by CNTF and to its subsequent biological response (26,27). The consequent signaling events involve the recruitment and activation of the Janus kinase/signal transducer and activator of transcription (STAT) pathways (3, 28 -30).
gp130, LIFR, and CNTFR␣ are members of the class I hematopoietin receptor family. This family of receptors is defined by a modular structure and characterized by a cytokine binding domain (CBD) (31,32). The CBD is composed of two "specialized" fibronectin III domains linked by a proline-rich sequence (Fig. 1). The N-terminal fibronectin III domain of the CBD (the D2 domain in Fig. 1) is characterized by two conserved disulfide bridges. The C-terminal fibronectin III domain of the CBD (the D3 domain in figure 1) contains a conserved WSXWS motif. The crystal structure of several type I cytokine receptors and cytokine-receptor complexes has revealed that CBD forms an elbow-shaped module bent at an angle of ϳ90°. The receptor CBD interacts with cytokines through loops located at the boundary of the two fibronectin III domains (33).
The IL-6 family of cytokines belongs to the "long chain" four-␣ helix bundle class, with an up-up down-down topology (7). The different secondary structures are named A-D from the N terminus to C terminus (Fig. 1). Crystallographic structures and site-directed mutagenesis studies have shown that these cytokines interact with their receptor through three different binding sites, numbered 1-3 by analogy with growth hormone (Fig. 1) (34,35). Cytokines requiring an ␣ chain, like CNTF, CLC, NP, IL-6, and IL-11, bind first to the receptor through a site 1 composed of residues from the C-terminal parts of the AB loop and the ␣D helix (Fig. 1).
Site 2 is located on solvent-exposed faces of the ␣A and ␣C helices and is important for binding to gp130. Site 3 is specific to the IL-6 family and corresponds to an additional signaling receptor binding site for LIFR in the case of CNTF, CLC, and NP.
The LIFR binding site is characterized by an FXXK motif located at the N terminus of the ␣D helix (36). The CNTF and CLC binding site 1 have been previously identified by site-directed mutagenesis (37,38). These studies have shown that a conserved tryptophan residue and an RXXXD motif, which are located in the C-terminal parts of the AB loop and the ␣D helix, respectively, are required for CNTF and CLC interaction with CNTFR␣. However, the counterpart of this contact site on CNTFR␣ is still unknown.
The aim of the present study was to determine the location of the binding site of CNTF, CLC, and NP on CNTFR␣. For this purpose, NP was modeled, and its predicted binding site 1 was compared with the binding sites of CNTF and CLC. This analysis led us to identify a conserved binding site 1 among all of the CNTFR␣ ligands. Subsequently, the CNTFR␣ CBD was modeled, and its binding surface was analyzed to identify an area displaying physicochemical properties complementary to those of the CNTF, CLC, and NP binding site 1.

EXPERIMENTAL PROCEDURES
Sequence Alignments-CNTF, NP, CLC, LIF, cardiotrophin-1, oncostatin M, CNTFR␣, IL-6R␣ and IL-12p40 sequences were retrieved from the Uniprot and GenBank TM data bases. Multiple sequence alignments were obtained using the T-Coffee (available on the World Wide Web) and GeneDoc (available on the World Wide Web) programs. CNTFR␣ and NP secondary structures were predicted using Network Prediction Server @nalysis at the Pôle Bioinformatique Lyonnais (Lyon, France) (39). Multiple sequence alignments were also represented with WebLogo (available on the World Wide Web) (40).
Molecular Modeling-NP was modeled by satisfaction of spatial constraints using the Modeler program based on multiple sequence alignments and secondary structure predictions (41). The structural coordinates of human CNTF (42) and LIF (43) were selected as molecular templates for the four-helical core of NP; the coordinates of an additional ␣ helix predicted in the AB loop of these cytokines were computed with the Biopolymer module of InsightII (Accelrys, San Diego, CA) and integrated during the modeling process.
CNTFR␣ was modeled as previously reported (21). Additional refinements of the CNTFR␣ model were carried out by optimizing side chain conformation using SCRWL3 (44). Electrostatic potentials were computed using Adaptive Poission-Boltzmann Solver with the AMBER force field (45,46), and Eisenberg's hydrophobicity scale was used to display the protein surface hydrophobicity (47).
Protein Docking-CNTF, NP, and CNTFR␣ were manually superimposed onto their respective partner in the IL-6⅐IL-6R␣ complex (Protein Data Bank accession number 1P9M) to build preliminary models. These models were further refined with the molecular docking program HEX 5.0 (48) to optimize structural complementarities. The best scoring solutions for each complex were energy-minimized with CHARMM (49) using the 100 steepest descent steps, followed by ABNR steps until a convergence gradient of 0.001 was reached.
Cells and Reagents-The COS-7, T98G glioblastoma, and SK-N-GP neuroblastoma cells were grown in RPMI 1640 medium supplemented with 10% fetal calf serum. The culture medium of Ba/F3 cells, modified to express the functional tripartite receptor for CNTF (BAF GLC), was supplemented with 5 ng/ml human CNTF (2). Anti-CNTFR␣ (AN-B2, AN-C2, and AN-E4) monoclonal antibodies and the polyclonal anti-CNTF antibody were generated in the laboratory. The anti-protein C (HPC4) antibody was purchased from Roche Applied Science. The anti-STAT3 polyclonal antibody and the mAb-specific for phospho-705-STAT3 were from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA) and New England Biolabs (Beverly, MA), respectively. CNTF, LIF, and IL-2 were bought from R&D Systems (Oxon, UK). CLC and mouse NP were produced as previously described (1,2).
Site-directed Mutagenesis and Cell Transfection-The cDNAs encoding NP and the membrane or soluble forms of CNTFR␣ were subcloned in the pcDNA3.1 vector and subjected to site-directed mutagenesis using the QuikChange TM site-directed mutagenesis kit (Stratagene, La Jolla, CA), following the manufacturer's instructions. Mutations were verified by automatic DNA sequencing. cDNAs encoding mutant and wild type CNTFR␣ were transfected by electroporation in T98G glioblastoma cells using the Amaxa Nucleofactor TM technology (Amaxa, Köln, Germany), following the manufacturer's instructions. cDNAs encoding mutant and wild type (WT) NP were transfected in COS-7 cells, as previously described (1). After a 48-h culture period, cells or supernatants were harvested and assayed. Mutant and WT NP were purified from the supernatant, as previously described (1).
Tyrosine Phosphorylation Analysis and Western Blotting-Parental or transfected T98G cells and SK-N-GP neuroblastoma cells were activated for 10 min with the indicated cytokines before being lysed in 10 mM Tris-HCl, pH 7.6, 5 mM EDTA, 50 mM NaCl, 30 mM sodium pyrophosphate, 50 mM sodium fluoride, 1 mM sodium orthovanadate, protease inhibitors (1 g/ml pepstatin, 2 g/ml leupeptin, 5 g/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride), and 1% Nonidet P-40. The lysates were subjected to SDS-PAGE and immunoblot analysis with a mAb specific for the tyrosine-phosphorylated form of STAT3. Membranes were stripped in 0.1 M glycine, pH 2.5, for 15 h and neutralized in 1 M Tris-HCl, pH 7.6, before reblotting with an antibody recognizing all STAT3 isoforms.
Immunoprecipitation-COS-7 cell supernatants containing the wild type or mutated forms of soluble CNTFR␣ were harvested 48 h after transfection. For co-immunoprecipitation experiments, the receptors were incubated at a concentration of 1 nM with 1 nM CNTF. Then proteins were incubated overnight with the AN-C2 anti-CNTFR␣ mAb at a concentration of 10 g/ml. Complexes were isolated using beads coupled to protein A, followed by SDS-PAGE. Western blot analyses were performed with an anti-protein C biotinylated mAb or an anti-CNTF biotinylated polyclonal antibody, and revealed by polystreptavidin peroxidase. The reaction was visualized using an image Master camera from Amersham Biosciences. Membranes were stripped as described above, before reblotting with the biotinylated anti-CNTFR␣ AN-E4 mAb.
Flow Cytometry Analysis-Parental and transfected T98G glioblastoma cells were incubated for 30 min at 4°C with anti-CNTFR␣ monoclonal antibodies (AN-B2 and AN-C2) (10 g/ml) or an isotype control antibody before using a phycoerythrin-conjugated anti-mouse antibody. CLC was biotintagged using the Bir A biotin ligase AviTag peptide substrate as previously described (50). CNTF and IL-2 were biotinylated using NHS-LC-biotin (Pierce). Binding of biotinylated CLC, CNTF, NP, or IL-2 as a control cytokine was revealed using phycoerythrin-labeled polystreptavidin for an additional 30-min incubation step. Fluorescence was subsequently analyzed on a FACScalibur flow cytometer (BD Biosciences).
Biological Assays-BAF GLC cells were seeded in 96-well plates at a concentration of 5 ϫ 10 3 cells/well in RPMI 1640 medium containing 5% fetal calf serum. Serial dilutions of the cytokines tested were performed in triplicate. After a 72-h incubation period, 0.5 Ci of [ 3 H]Tdr was added to each well for the last 4 h of the culture, and the incorporated radioactivity was determined by scintillation counting, as previously described (51).

Analysis of the CNTF, CLC, and NP Binding Site 1 and NP
Modeling-First, we performed a multiple sequence alignment of orthologs and paralogs of the IL-6 family members recruiting LIFR in their receptor complex ( Fig. 2A). A conserved tryptophan residue (Trp 64 and Trp 94 in CNTF and CLC, respectively), located in the AB loop of these cytokines was previously reported to contribute to their binding site 1 (37,38). The alignment shows that Trp 85 in NP was highly conserved with CNTF Trp 64 and CLC Trp 94 and suggests that this residue is also involved in the NP site 1.
Additional residues in the ␣D helix of these cytokines also contribute to CNTF and CLC binding site 1 (37,38). An RXXXD motif (Arg 171 and Asp 175 in CNTF and Arg 197 and Asp 201 in CLC) is conserved in the C-terminal part of this ␣ helix. Interestingly, this RXXXD motif was also present in the C-terminal part of the NP ␣D helix (Arg 190 and Asp 194 ). These observations suggested that, like CNTF and CLC, NP possesses a conserved binding site 1 for CNTFR␣ ( Fig. 2A).
We built a NP molecular model using CNTF and LIF structures as templates. Next, we compared the NP model obtained with the CNTF structure or the CLC model previously reported (38,42) and analyzed their binding site 1 (Fig. 2B). Strikingly, the NP binding surface involved a tryptophan residue located in the AB loop surrounded by positively charged residues, as already observed for CNTF and CLC. This clearly showed that the physicochemical properties of CNTF, CLC, and NP binding site 1 were conserved.
Site-directed Mutagenesis of NP-To confirm the modeling study, we generated the W85A NP mutant. As a control, an irrelevant F82A mutation, located in the vicinity of the area studied, was introduced into the NP sequence. WT and mutant forms of NP were expressed in the COS-7 cell line and purified on a Ni 2ϩ -nitrilotriacetic acid column, as previously described (1). The functionality of WT NP and mutant NP was tested in a proliferation assay using the BAF GLC cell line expressing the tripartite CNTF receptor complex. The results showed a lack of proliferation when cells were grown in the presence of the W85A NP mutant compared with the proliferation observed in the presence of the WT or F82A forms of NP (Fig. 3A).
In addition, the NP mutants were studied for their ability to induce the recruitment of the STAT3 signaling pathway in SK-N-GP, a neuroblastoma cell line expressing the three CNTF receptor chains (Fig. 3B). A robust STAT3 tyrosine phosphorylation was observed in response to WT or F82A OCTOBER 31, 2008 • VOLUME 283 • NUMBER 44 mutant forms of NP as well as to CNTF. In contrast, W85A NP mutant failed to activate STAT3. To conclude, Trp 85 is a crucial residue for NP biological activities and is likely to contribute to the definition of a conserved site 1 among CNTF, CLC, and NP.

Characterization of the CNTF⅐CNTFR␣ Binding Site
Molecular Modeling of CNTFR␣-To identify counterpart(s) of the binding site 1 of CNTF, CLC, and NP on CNTFR␣, we analyzed the CNTFR␣ model that we previously reported (21). For this purpose, we refined this model by optimizing the side chain conformation using SCRWL3.
Cytokine-receptor complexes, solved by x-ray crystallography (34,52), have revealed that several receptor loops contain residues involved in their binding interfaces. Indeed, loops connecting ␤ strands E and F, BЈ and CЈ, and FЈ and GЈ are parts of FIGURE 2. Multiple sequence alignment of cytokines belonging to the IL-6 family and ribbon representation of the CNTF structure and CLC and NP models. A, dark blue-shaded letters indicate conserved or type-conserved amino acids according to the BLOSUM45 matrix. 80% conserved residues are shaded in cyan. The NP sequence is murine, and the other sequences are human. The figure shows the assignment of the four ␣ helices of the cytokines represented as predicted by the secondary structure prediction server Network. Yellow boxes highlight residues implicated in the CNTF, CLC, and NP binding site 1. B, ribbon representation of the NP model displaying charged and aromatic residues in its binding site. Green residues, aromatic amino acids; blue residues, polar and positively charged.
the receptor binding sites (34,52,53). This observation prompted us to carefully analyze these elements in CNTFR␣.
We performed a multiple sequence alignment of CNTFR␣ with other short ␣ chain receptor orthologs and paralogs, including IL-6R␣ and IL-12p40 (Fig. 4A). This alignment showed that ␤ strands were globally conserved between the different receptors as well as apolar residues involved in the hydrophobic core of the proteins. However, connecting loops, including EF, BЈCЈ, and FЈGЈ loops, were not conserved between the different receptors.
Analysis of the CNTFR␣ model indicated that the hinge region connecting the D2 and D3 domains harbored an aromatic cluster composed of five residues (depicted in green in Fig. 4, B and C). These amino acids stabilized the CBD as commonly observed in this family of proteins (34,52). However, we observed a structural divergence between the different ␣ receptor chains when we analyzed the EF, BЈCЈ, and FЈGЈ loops, which prevented identification of key CNTFR␣ interaction residues by homology with IL-6R␣ and IL-12p40. Therefore, we studied the physicochemical properties of the CNTFR␣ surface and focused on the putative area(s) complementary to CNTF, CLC, and NP binding site 1.
Determination of the Putative CNTFR␣ Binding Site-To identify binding site complementary area(s), we computed and then compared the electrostatic potential of the CNTF and CNTFR␣ surfaces (Fig. 5). Indeed, residues Arg 25 and Arg 28 , located in the CNTF ␣A helix, mainly contributed to a positively charged area, as previously reported (37,42). Fig. 5C shows the presence of an extended negatively charged surface centered on CNTFR␣ Glu 236 and Glu 286 , which forms a mirror image of the positively charged area of CNTF.
The hydrophobic potential was mapped onto the CNTF and CNTFR␣ surfaces. In agreement with previous studies (37,42), Trp 64 forms part of a hydrophobic pocket on the CNTF binding site 1 (Fig. 5B). A complementary residue, displaying similar physicochemical properties, was identified in the CBD hinge region of CNTFR␣. It appears that the protruding side chain of Phe 172 forms a hydrophobic patch complementary to the CNTF Trp 64 pocket (Fig. 5D).
Moreover, we analyzed residue conservation in the EF, BЈCЈ, and FЈGЈ interacting loops in different CNTFR␣ orthologs. Phe 172 , Glu 236 , and Glu 286 , located in these loops, were evolutionary conserved or type conserved residues (Fig. 5E). These results further sustained the idea that the three residues identified were potential CNTFR␣ binding hot spots.
We also performed the same analyses for CLC and NP. Strikingly, CLC Trp 94 , in agreement with our previous study (21), and NP Trp 85 contributed to a hydrophobic pocket complementary to CNTFR␣ Phe 172 (data not shown).
Molecular Docking of CNTFR␣ with Its Ligands-To further characterize the interaction of CNTFR␣ with its three ligands, molecular docking experiments were carried out. The IL-6⅐IL-6R␣ complex was used as a template to build preliminary CNTF⅐CNTFR␣ and NP⅐CNTFR␣ models (the CLC⅐CNTFR␣ docking model was previously published (21)). The precise cytokine-receptor orientations were computed using HEX to  The side chains represented correspond to residues involved in the IL-6R␣ binding site or predicted to be involved in the CNTFR␣ binding site. Green residues, aromatic amino acids; red residues, polar and negatively charged.
optimize protein shape complementarities. Complexes with the best scoring solutions were then energy minimized to allow side chain reorientation at the cytokine-receptor interface. The CNTF⅐CNTFR␣ complex model is presented in Fig. 6 (the NP⅐CNTFR␣ complex appears as supplemental Fig. 1).
This model shows that the CNTF Trp 64 and CNTFR␣ Phe 172 are involved in an aromatic-aromatic interaction. In the vicinity of this hydrophobic area, a cluster of polar residues is formed by CNTF Arg 25 and Arg 28 located in the ␣A helix and CNTFR␣ Glu 236 and Glu 286 located in the D3 domain. These residues are predicted to form salt bridges between the two proteins. Additional charged residues (CNTF Arg 177 and CNTFR␣ Asp 234 ) might also contribute to the interaction by long range electrostatic effects. Similarly, CNTFR␣ Glu 236 and Glu 286 were also predicted to be involved in salt bridge or hydrogen bond with polar NP residues.
Site-directed Mutagenesis and Expression of the Mutated Forms of CNTFR␣-Based on the above observations, we selected Phe 172 , Glu 236 , and Glu 286 of CNTFR␣ as important residues for interaction with its different cognate ligands. Mutants were designed to alter the receptor binding capacities, and the following mutations were introduced: F172A, E236A, and E286A. In addition, Phe 199 located in the vicinity of the potentially critical residue Phe 172 was also mutated to alanine and used as irrelevant control for our model.
Corresponding cDNAs were transfected in the T98G human glioblastoma, a cell line expressing gp130 and LIFR, but not CNTFR␣ (1). A strong expression of both WT and mutated forms of CNTFR␣ was detected by FACS analysis (Fig. 7). A similar expression level was observed using either AN-B2 or AN-C2 antibody, two mAbs directed against different CNTFR␣ conformational epitopes (20). This finding indicates that the mutations introduced did not significantly alter CNTFR␣ folding and structure (Fig. 7).
Binding Properties of CNTFR␣ Mutants-The binding properties of CNTFR␣ mutants expressed in the T98G glioblastoma cell line were examined. For this purpose, CNTF and CLC were biotinylated and incubated with the transfected cells. Binding to the cell surface was monitored by fluorescence-activated cell sorting analysis (Fig. 8). Readily detectable signals were measured when either CLC or CNTF was incubated with cells expressing WT CNTFR␣ or the F199A mutant. As expected, no binding could be detected when either CNTF or CLC were incubated with the T98G parental cell line. Importantly, a significant decrease or complete abrogation of CNTF and CLC binding was observed when the cytokines were added to the cells expressing either E286A or F172A forms of CNTFR␣. In contrast, no significant change in cytokine interaction was seen for the E236A mutant (Fig. 8). Similar biotinylation experiments were carried out with NP, but major protein instability was observed during the labeling process. Therefore, we could not perform the binding analyses (data not shown). Together, these results indicate that CNTFR␣ Phe 172 and Glu 286 play an important role in CLC and

. Connolly surfaces of CNTF (A and B) and of the CNTFR␣ CBD (C and D) and residue conservation of CNTFR␣-interacting loops (E).
The electrostatic potentials (A and C) and the hydrophobic index of the exposed residues (B and D) were mapped onto the surfaces of CNTF and CNTFR␣ as described under "Experimental Procedures." The orientations of CNTF and CNTFR␣ have been manually adjusted to provide an "open book" surface representation of the interfaces. E, conservation of residues present in CNTFR␣-interacting loops are represented as logos. Each logo consists of stacks of symbols, one stack for each position in the sequence. The overall height of the stack indicates the sequence conservation at that position, whereas the height of symbols within the stack indicates the relative frequency of each amino acid at that position. Green residues, aromatic amino acids; blue residues, polar and positively charged; red residues, polar and negatively charged; black residues, hydrophobic; orange residues, polar and neutral.
CNTF binding, in agreement with our molecular modeling study.
Mutant CNTFR␣ Association with CNTF and CLC-Previous studies have shown that a soluble form of CNTFR␣ can bind CNTF (20,26). To monitor the interaction of soluble CNTFR␣ with CNTF, co-immunoprecipitation experiments were performed (Fig. 9A). Results showed that the soluble CNTFR␣ F172A and E286A mutants failed to interact with CNTF compared with the WT or F199A forms used as a control.
We previously reported that CLC was secreted as a composite cytokine when associated with a soluble receptor moiety that could be either cytokine-like factor 1 or soluble CNTFR␣ (2,20). We studied the interaction between CLC and soluble CNTFR␣ by co-expressing a tagged form of the cytokine together with the soluble receptor mutants in the COS-7 cell line. Co-immunoprecipitation experiments showed the secretion of a stable composite when CLC was co-expressed either with WT or F199A mutant (Fig. 9B). Interestingly, no stable heterocomplex formation of CLC with the F172A or E286A soluble receptors could be detected. However, a slight secretion of CLC was observed when co-expressed with the E286A CNTFR␣ mutant. This result underlines a transient interaction between the two proteins in agreement with the results presented in Fig. 8. Analysis of cell lysates confirmed the correct expression of both partner proteins. NP interaction with soluble CNTFR␣ mutants could not be assessed, since we previously reported that NP only recognized the membrane form of CNTFR␣ (1). These results indicate that CNTF and CLC interact predominantly with CNTFR␣ through the same residues, Phe 172 and Glu 286 .
Functional Properties of CNTFR␣ Mutants-We examined the functional properties of CNTFR␣ mutants in response to CNTF, CLC, and NP. We analyzed the tyrosine phosphorylation level of STAT3, a major signaling pathway recruited by the tripartite CNTF receptor complex. Experiments were carried out using the T98G glioblastoma cell line transfected with CNTFR␣ mutants. LIF, which mediates its responses through   OCTOBER 31, 2008 • VOLUME 283 • NUMBER 44 the gp130⅐LIFR heterocomplex, was used as a positive control (Fig. 10).

Characterization of the CNTF⅐CNTFR␣ Binding Site
Western blot assays showed that CNTFR␣ is necessary to induce STAT3 phosphorylation in response to CNTF, CLC, or NP. A nearly total extinction of the signal was observed when cells were transfected with F172A or E286A receptor mutants and activated with CNTFR␣ ligands. In the cells expressing the E236A mutant or the F199A mutant, the STAT3 pathway remained sensitive to stimulation with the three ligands. These combined results corroborate the modeling and binding results of the study.

DISCUSSION
CNTFR␣ is able to specifically recognize three different ligands sharing 16 -25% sequence identity. Previous studies have identified binding site 1 of CNTF and CLC (37,38), but its counterpart on CNTFR␣ has not yet been described to date regarding the binding site 1 counterpart on CNTFR␣. We therefore undertook the identification of the molecular determinants implicated in the interaction of CNTFR␣ with its ligands, in order to understand its cross-reactivity.
The resolution of cytokine-receptor complexes by crystallography has determined the molecular assembly of this family of proteins (34,(52)(53)(54). It first revealed that this protein-protein assembly requires complementary shapes and physicochemical properties. Moreover, site-directed mutagenesis experiments have demonstrated that only a few residues, defined as interac- tion hot spots, contribute to the binding free energy (55). Statistical analyses of numerous receptor binding sites have revealed that hot spots are often conserved and are significantly enriched in aromatic amino acids (56). These notions were used to look for the complementary site 1 of CNTFR␣. In the present study, we have shown that CNTF, CLC, and NP share the same binding site 1. The CNTFR␣ putative structural binding epitope was determined by searching a surface area with physicochemical properties complementary to the binding site 1 of CNTF, CLC, and NP. To reinforce the modeling prediction, the residues identified were substituted with alanine, and the mutants obtained were further tested in a panel of biological assays.
Both Panayotatos's group and ours previously reported the involvement of the conserved residues CNTF Trp 64 and CLC Trp 94 for the interaction of these cytokines with CNTFR␣ (37,38). Additional residues also contribute to CNTF and CLC binding site 1. An RXXXD motif (CNTF Arg 171 and Asp 175 and CLC Arg 197 and Asp 201 ) is conserved in their ␣D helix and is essential for the side chain orientation of the tryptophan residue hot spot (37,42). Furthermore, we have identified inactivating mutations in the gene coding for CLC in a patient suffering from cold-induced sweating syndrome (21). One of these mutations, a substitution of Arg 197 , impairs the CLC/CNTFR␣ interaction. This observation confirms the importance of RXXXD motif for the integrity of binding site 1.
In the present study, we characterized the NP binding site 1, which is conserved across the three CNTFR␣ ligands. The NP Trp 85 residue, equivalent to CNTF Trp 64 and CLC Trp 94 , was found to be a hot spot at the interface of the NP⅐CNTFR␣ complex. Interestingly, an RXXXD motif was also conserved in the ␣D helix of NP (Arg 190 and Asp 194 ). The determination of a conserved site 1 for CNTF, CLC, and NP was the starting point for our investigation of the CNTFR␣ binding epitope.
The structural diversity of the receptor-interacting loops led us to compute surface physicochemical properties of CNTFR␣ in order to identify complementarities with CNTF, CLC, and NP binding surfaces. A mirror image of the three cytokines' binding site 1 was found on CNTFR␣. The residues, Phe 172 , Glu 236 , and Glu 286 , were identified as potential components of its binding epitope. Our study clearly established that the F172A and E286A CNTFR␣ mutations impaired or decreased CNTF, CLC, and NP interactions.
The F172A mutation in the EF loop of the CNTFR␣ domain completely abrogated the binding of CNTF, CLC, and NP with this receptor. Moreover, the CNTFR␣ F172A mutant failed to mediate biological responses induced by these three cytokines. These results confirmed the data obtained with CNTF W64A, CLC W94A, and NP W85A mutants (37,38). Our docking study indicated that these tryptophan residues were involved in aromaticaromatic interactions with CNTFR Phe 172 . The highly disruptive effects of corresponding mutations further confirmed that Phe 172 was a hot spot residue on the CNTFR␣ surface.
Similarly, the E286A mutation impaired biological activity mediated through CNTFR␣ in response to CNTF, CLC, and NP. A salt bridge between CNTF Arg 25 and CNTFR␣ Glu 286 was observed in the CNTF⅐CNTFR␣ complex model that we generated. These observations fit with results obtained with the CNTF Arg 25 mutant, which also lost its CNTFR␣ recognition capacity (37).
Analysis of the related cytokine-receptor ␣ chain complexes shows that aromatic-aromatic and salt bridge interactions are globally conserved at these protein interfaces (34,52). In the IL-6⅐IL-6R␣ structure, the IL-6R␣ Phe 248 residue is critical for the interaction with IL-6 Phe 106 located in the cytokine AB loop, as supported by mutational experiments (34,57). A similar situation was observed for IL-12p35⅐IL-12p40, where the IL-12p40 Tyr 246 residue and its paring residue (Tyr 193 ) present in the ␣D helix of p35 were found to be important in the interaction of the IL-12p35 and IL-12p40 subunits (52). Conserved salt bridges between IL-6 Arg 207 and IL-6R␣ Glu 296 or IL-12p35 Arg 239 and IL-12p40 Glu 203 are also important for cytokinereceptor binding.
The present work, as well as previous studies (34,52), suggests that despite structural divergence of receptor-interacting loops, the different cytokines and receptors of the IL-6/IL-12 family have kept similar intermolecular interactions in order to associate with each other. Importantly, conserved pairs of aromatic or charged residues on these protein interfaces seem to be a conserved feature of cytokines and receptors interacting through a binding site 1 in the IL-6/IL-12 family. This finding could help to identify crucial residues involved in the interactions of newly described composite cytokine subunits, such as IL-23, IL-27, or IL-35 (6, 58 -60).
In conclusion, CNTF, CLC, and NP share an overlapping binding epitope located in the CNTFR␣ CBD and involving Phe 172 and Glu 286 residue hot spots. The CNTFR␣ Phe 172 residue is the mirror image of the conserved tryptophan hot spot identified in CNTF, CLC, and NP. These pairs of residues are predicted to form an aromatic-aromatic interaction at the cytokine/receptor interface. The CNTFR␣ Glu 286 residue is a hot spot residue at the CNTF/CNTFR␣ and NP/CNTFR␣ interfaces and plays a major role in CLC⅐CNTFR␣ complex stability.