The Amino Acid Exchange R28E in Ciliary Neurotrophic Factor (CNTF) Abrogates Interleukin-6 Receptor-dependent but Retains CNTF Receptor-dependent Signaling via Glycoprotein 130 (gp130)/Leukemia Inhibitory Factor Receptor (LIFR)*

Background: CNTF signaling is mediated by CNTFR or IL-6R in complex with gp130 and LIFR. Results: The CNTFR variant CV-1 is CNTFR-selective. Conclusion: The single amino acid exchange R28E within CNTF abrogated IL-6R binding. Significance: CV-1 allows discrimination between CNTFR- and IL-6R-mediated effects in vivo. Ciliary neurotrophic factor (CNTF) is a neurotrophic factor with therapeutic potential for neurodegenerative diseases. Moreover, therapeutic application of CNTF reduced body weight in mice and humans. CNTF binds to high or low affinity receptor complexes consisting of CNTFR·gp130·LIFR or IL-6R·gp130·LIFR, respectively. Clinical studies of the CNTF derivative Axokine revealed intolerance at higher concentrations, which may rely on the low-affinity binding of CNTF to the IL-6R. Here, we aimed to generate a CNTFR-selective CNTF variant (CV). CV-1 contained the single amino acid exchange R28E. Arg28 is in close proximity to the CNTFR binding site. Using molecular modeling, we hypothesized that Arg28 might contribute to IL-6R/CNTFR plasticity of CNTF. CV-2 to CV-5 were generated by transferring parts of the CNTFR-binding site from cardiotrophin-like cytokine to CNTF. Cardiotrophin-like cytokine selectively signals via the CNTFR·gp130·LIFR complex, albeit with a much lower affinity compared with CNTF. As shown by immunoprecipitation, all CNTF variants retained the ability to bind to CNTFR. CV-1, CV-2, and CV-5, however, lost the ability to bind to IL-6R. Although all variants induced cytokine-dependent cellular proliferation and STAT3 phosphorylation via CNTFR·gp130·LIFR, only CV-3 induced STAT3 phosphorylation via IL-6R·gp130·LIFR. Quantification of CNTF-dependent proliferation of CNTFR·gp130·LIFR expressing cells indicated that only CV-1 was as biologically active as CNTF. Thus, the CNTFR-selective CV-1 will allow discriminating between CNTFR- and IL-6R-mediated effects in vivo.

The interleukin (IL)-6 type cytokine family consists of cardiotrophin-1, cardiotrophin-like cytokine (CLC), 3 CNTF, IL-6, IL-11, IL-27, IL-30, IL-31, leukemia inhibitory factor (LIF), and oncostatin M (OSM) (1). Except for IL-31, all IL-6 type cytokines bind to the ␤-receptor gp130, demonstrating a high degree of binding site plasticity within gp130 toward the respective cytokines. Whereas IL-6 and IL-11 bind to homodimeric gp130 receptor complexes, all other cytokines of this family bind to heterodimeric signal-transducing ␤-receptor complexes, e.g. CNTF and CLC to gp130⅐LIFR (1). IL-6 type cytokine signaling activates the intracellular signaling cascades JAK/STAT, MAPK, and PI3K. For some IL-6 type cytokines, binding to their ␤-receptors is dependent on the initial complex formation with an ␣-receptor. The following high-affinity cytokine/␣-receptor pairs exist: CLC/CNTFR, CNTF/CNTFR, IL-6/IL-6R, IL-11/IL-11R, and p28/Epstein-Barr virs-induced gene 3 (1). Moreover, the IL-6R was described as alternative low affinity ␣-receptor for CNTF and p28 (2,3), highlighting receptor plasticity also at the ␣-receptor/cytokine binding interface. IL-6 type cytokines have a typical four-helix bundle fold that links the four ␣-helices (A-D) by two long loops (A-B and C-D) and one short loop (B-C) placing the four helices in an up-updown-down orientation. Here, site I of the cytokine binds to the ␣-receptor, whereas site II and site III are needed for the interaction with the ␤-receptors. Site I is constituted by the C-terminal A-B loop (site Ia) and the C-terminal D helix (site Ib), site II by parts of the A and C helix and site III by the C-terminal A helix and the N-terminal A-B loop (site IIIa), the B-C loop with adjacent parts of the B and C helix (site IIIb), and the C-terminal C-D loop with the adjoining NH 2 -terminal D helix (site IIIc).
Binding of IL-6 type cytokines to their receptors is mainly mediated by ionic and hydrophobic interactions (4). Close inspection of the electrostatic surface potential of site I, II, and III of IL-6 type cytokines, however, reveals almost no similarity. Cross-reactivity is likely entropy driven by water exclusion from the interacting surfaces (5).
Interestingly, species-specific differences of cytokine/receptor interactions within the IL-6 type cytokine family were described. This is exemplified by differences for human and murine IL-6 and OSM. Human IL-6 binds to human and murine IL-6R⅐gp130 receptor complexes, but murine IL-6 exclusively binds to murine IL-6R⅐gp130 receptor complexes (2). Furthermore, human OSM bind to both human LIFR and OSM receptor, but murine OSM exclusively interacts with murine OSM receptor (6,7). Rat CNTF is able to engage signaling via a heterodimer of gp130/LIFR in the absence of CNTFR (8), which has not been observed for murine or human CNTF.
Furthermore, transfer of the binding site III between IL-6 type cytokines led to chimeric cytokines with novel and unique receptor specificities (9). IC-7 is a chimera of IL-6 and CNTF, in which site III of IL-6 is exchanged by site III of CNTF. Following binding to the IL-6R via site I, IC-7 recruits gp130 to site II and the LIFR by site III. On cells expressing the IL-6R, gp130, and the LIFR, IC-7 displays biological activities comparable with that of LIF or IL-6, but it is inactive on cells that do not express the IL-6R (9). Viral IL-6, from the human herpesvirus 8 has ϳ25% sequence homology with human IL-6. Viral IL-6 signals via a gp130 receptor homodimer but in contrast to IL-6 without the need of the IL-6R␣ (10). IV-9 is a chimera of IL-6 and viral IL-6, in which an extended site III of IL-6 is exchanged by site III of viral IL-6. Consequently, IV-9 signaling via a gp130 homodimer was independent of the IL-6R (11).
The transfer of site I or II has, however, not been described as of yet. Moreover, abrogation of promiscuous cytokine/receptor usage by introduction of single point mutations within the cytokine or cytokine receptor has also not been achieved. Generation of CNTF or p28 variants specific for a single receptor would pave the way to analyze the function of CNTF/CNTFRversus CNTF/IL-6R-induced or p28/EBI3 versus p28/IL-6R-induced signal transduction in vivo.
A CNTFR-selective CNTF might be achieved by transfer of site I from the CNTFR-selective cytokine CLC to CNTF or by introduction of specific amino acid exchanges in CNTF. Importantly, because even though the binding to the low affinity ␣-receptor will be abrogated, this must not result in a major reduction of binding affinity of CNTF to the high-affinity CNTFR.
Here, we describe the development the human CNTF variant (CV)-1, which binds with high affinity to the CNTFR and induces signal transduction via LIFR and gp130. Due to the single point mutation R28E, interaction of CV-1 with the IL-6R was abrogated. CV-1 might be used to discriminate between IL-6R and CNTFR-mediated effects of CNTF in vivo.
Proliferation Assays-The different transduced Ba/F3 cell lines were washed three times with sterile PBS and suspended in DMEM containing 10% FBS at a final concentration of 5 ϫ 10 3 cells per well in a 96-well microtiter plate. Cells were incubated for 72 h as indicated with the cytokines and cytokine receptors in a final volume of 100 l. After 72 h, cell growth was measured using the CellTiter Blue cell viability assay reagent (Promega, Karlsruhe, Germany) following the manufacturer's instructions. The extinction was measured using a Lambda FLUORO 320 microplate fluorescence reader (excitation, filter 530/25; emission, filter 590/35; Software KC4, Bio-Tek Instruments). Normalization of relative light units was achieved by subtraction of negative control values. One representative example of two to three experiments was chosen for the figures. For clarity, no standard deviations were included in the diagrams; the S.D. were, as common for Ba/F3 cells, only minimally and usually below 5%.
Immunoprecipitation with sCNTFR-Fc and sIL-6R-Fc-For co-precipitation, supernatants of p409DC-IL-6R-Fc and pDC-CNTFR-Fc transiently transfected COS-7 were collected. For co-precipitation supernatants (transfected or co-transfected as indicated) were mixed with recombinant proteins (1 g) incubated 2 h at 4°C under gentle agitation. As control, a further tube containing the corresponding cytokine was incubated without receptor containing supernatant. 150 l of protein A-agarose (Roche Diagnostics GmbH, Mannheim, Germany) was added, and the mixture was incubated at 4°C for 2 h under gentle agitation. The samples were washed six times with PBS, and proteins were eluted by adding 50 l of Laemmli buffer and incubated for 10 min at 95°C. The resulting supernatants were subjected to Western blot analysis.
Western Blotting-For Western blotting, ϳ2 ϫ 10 7 cells per experiment were washed three times with sterile PBS. The cells were distributed to 2-ml tubes and starved in FBS-free medium for 4 -6 h (Ba/F3 cell lines) or overnight (HepG2 cells) at 37°C and CO 2 saturation. Cells were stimulated with the indicated cytokines for 10 min followed by centrifugation at 4°C and 2,000 rpm for 10 min. 150 l of 5ϫ Laemmli buffer (312 mM Tris-HCl, pH 6.8, 50% glycerol, 10% sodium dodecyl sulfate, 5% ␤-mercaptoethanol, and 0.13% bromphenol blue) were added to each tube, and cells were lysed by sonification (Sonopuls HD2200, Bandelin, Berlin, Germany) for 10 s and boiling at 95°C for 10 min. Proteins were separated by SDS-PAGE and transferred to a PVDF membrane using a Trans-Blot SD semi-dry transfer cell (Bio-Rad). The membrane was blocked in 5% low fat milk in TBS-T (10 mM Tris-HCl, pH 7.6, 150 mM NaCl, and 1% Tween 20) and probed with the primary antibody in 1% low fat milk in TBS-T (STAT3-mAb) or 5% BSA (pSTAT3-mAb) at 4°C overnight. The membranes were washed and incubated with the secondary peroxidase-conjugated antibody for 1 h before applying the ECL-plus peroxidase substrate (GE Healthcare). The Fluor ChemQ system (Cell Biosciences, Santa Clara, CA) was used for signal detection according to the manufacturer's instructions. The membranes were stripped with stripping buffer (20 ml of 10% SDS, 12.5 ml of 0.5 M Tris-HCl, pH 6.8, 67.5 ml of ultra-pure water, 0.8 ml of ␤-mercaptoethanol), blocked again, and probed with another primary antibody. The STAT3 phosphorylation assays were reproduced three times with one representative experiment shown.

RESULTS
Cloning, Bacterial Expression, Refolding, and Purification of CNTF and the CNTF Variants CV-1 to -5-Binding of CNTF to its ␣-receptors CNTFR or IL-6R is mediated via site I of CNTF and the cytokine binding module of the ␣-receptors. CNTF has a higher affinity to the CNTFR than to the IL-6R (3). Also, CLC needs the CNTFR for signaling via gp130 and LIFR, but cannot bind to the IL-6R. CLC has, however, a lower efficiency to signal via CNTFR⅐gp130⅐LIFR complexes as compared with CNTF (16). Fig. 1A illustrates the binding site I of CLC and CNTF, respectively. In this study, we aimed to generate a CNTFRselective CNTF variant without influencing the affinity of CNTF to the CNTFR⅐gp130⅐LIFR complex. In the past, successful transfer of the binding site III of CNTF to IL-6 led to the chimeric IL-6 variant IC-7, which signals via the artificial IL-6R⅐gp130⅐LIFR complex (9). Here, this strategy was also applied to site I of CNTF. We generated potential CNTFRselective CNTF variants by transfer of amino acid residues of site I from CLC to CNTF (Fig. 1A). Alternatively, using molecular modeling of CLC⅐CNTFR and CNTF⅐CNTFR complexes, we predicted that the amino acid Glu-26 in CLC might be involved in CNTFR selectivity because it is in close contact to the CNTFR (Fig. 1B). The respective amino acid in CNTF was identified to be Arg 28 .
As a starting point, we generated a shortened CNTF variant to improve recombinant CNTF protein expression in E. coli and to support the refolding efficiency from CNTF inclusion bodies. Therefore, the cDNA coding for CNTF was modified to code for a N-terminal deletion of the amino acids 1-14 and a C-terminal deletion of the amino acids 186 -200 (17). Moreover, two point mutations coding for the amino acid exchanges C17A and Q63R were introduced (Fig. 1C). With the exception of the C-terminal deletion, all other modifications were described previously for the CNTF variant Axokine. Axokine was shown to have an improved biological activity as compared with CNTF (18). A sequence coding for a C-terminal c-Myc and His tag was introduced in all cytokine-encoding cDNAs, including CNTF, CLC, IL-6, LIF, and IC-7. The cDNA coding for CNTF served as a platform for the transfer of amino acids from CLC to CNTF in CV-1 to CV5. CV-1 contained the single amino acid substitution R28E, located in close proximity to site I, CV-2 contained the extended site Ia (CNTF-His 41 -Leu 67 to CLC-Gly 41 -Leu 72 ) and CV-3 the minimal site Ia (CNTF-Ala 59 -Leu 67 to CLC-Ala 62 -Leu 72 ), CV-4 contained the site Ib (CNTF-Thr 169 -Gly 185 to CLC-Leu 170 -Pro 186 ) and CV-5 contained all exchanges, the extended site Ia and site Ib and the amino acid substitution R28E (CNTF-His 41 -Leu 67 to CLC-Gly 41 -Leu 72 /CNTF-Thr 169 -Gly 185 to CLC-Leu 170 -Pro 186 / R28E) (Fig. 1C).
CLC, CNTF, CV-1, CV-2, CV-3, CV-4, CV-5, IL-6, and IC-7 were expressed in E. coli as inclusion bodies, refolded, and purified. Purification was completed by nickel-nitrilotriacetic acid affinity chromatography and size-exclusion chromatography. All cytokines were produced as monomers ( Fig. 2A and  supplemental Fig. 1) and correctly folded as verified by circular dichroism revealing the typical spectra of ␣ helical proteins ( Fig.  2B and supplemental Fig. 1). Purity and identity of correctly folded and monomeric CLC, CNTF, CV-1 to CV-5, IL-6, and IC-7 was shown by Coomassie staining of SDS-PAGE gels and Western blotting against the C-terminal cytokine Myc tag, respectively (Fig. 2, C and D, and supplemental Fig. 1, A-H).
The CNTF Variants (CV-1, CV-2, CV-3, CV-4, CV-5) Bind to CNTFR, but only CV-3 and CV-4 Bind to IL-6R-First, we tested whether CNTF, CLC, IL-6, and the CNTF variants CV-1 to CV-5 were able to bind to CNTFR or IL-6R by co-precipitation. To this end, cDNAs coding for sCNTFR and sIL-6R C-terminally fused to an Fc part of an IgG1 antibody were subcloned and transiently transfected into COS-7 cells. sIL-6R-Fc or sCNTFR-Fc conditioned cell culture supernatants were used for Protein A-mediated co-precipitation of recombinant CNTF, CLC, IL-6, and CV-1 to CV-5. Precipitated cytokines were detected by Western blotting using a mAb against the C-terminal c-Myc tag. Non-precipitated input served as positive control, precipitation without sCNTFR-Fc or sIL-6R-Fc as negative control. As expected, co-precipitation of CLC and CNTF but not of IL-6 was mediated by sCNTFR-Fc (Fig. 3A). In the CNTF analog Axokine and in CNTF, Gln-63 of human CNTF was exchanged to Arg 63 of rat CNTF, which increased the specific activity of human CNTF (18). Because rat CNTF does not need to bind to the human IL-6R (8), it remained open, whether this mutation also abolished the binding of human CNTF to signal via CNTFR or IL-6R. Molecular modeling, however, indicated that Arg 63 is not involved in the interaction area between human CNTF and the human IL-6R or CNTFR. Here, our co-precipitation experiments showed that the Q63R did not abolish binding of CNTF to the IL-6R and the CNTFR (Fig. 3, A and B).

DISCUSSION
CNTF prevents degeneration of axotomized peripheral motor neurons and retrograde cell death of neurons in thalamic nuclei after dissection of intracerebral neuronal circuits (20,21). Intraperitoneal implantation of a CNTF-producing cell line improved survival in a mouse model of motor neuropathy (22). In models of Huntington disease, CNTF has neuroprotective effects on striatal cells (23)(24)(25). CNTF also enhanced the survival of sensory, hippocampal, and cerebellar neurons (26 -28) and increased survival of retinal photoreceptors in animal models of retinal degeneration (29 -31). On denervated rat skeletal muscles, CNTF reduced denervation-induced atrophy and increased muscular strength (32,33).
Moreover, CNTF is still unique in its ability to overcome leptin resistance and reduce food intake and body weight in obese, leptin-resistant humans and rodents. Importantly, weight loss was sustained after discontinuation of the drug (34). Because leptin resistance is still a major problem in obesity, CNTF was suggested as therapeutic principle to combat human obesity (35). It was hypothesized that CNTF has leptin-like actions in the hypothalamus. Activation of either leptin receptor or CNTFR in the hypothalamus leads to local STAT3 and mTOR activation (36 -38). Both cytokines also reduce AMPactivated kinase phosphorylation (39,40). Additionally, CNTFR and leptin receptor expression is overlapping in the hypothalamus (41), including the paraventricular nucleus and arcuate nucleus (42)(43)(44). However, a recent report indicates that despite anatomical overlap of CNTFR and leptin receptor expression, CNTF and leptin act within distinct neuronal populations to elicit anorectic effects (45).
Here, we describe the development of the CNTFR-selective CNTF variant CV-1, which is characterized by the single amino acid exchange R28E from CLC to CNTF. Biological activity of CV-1 was comparable with CNTF on Ba/F3-CNTFR-gp130-LIFR cells, indicating that the affinity of CV-1 to CNTFR⅐ gp130⅐LIFR complex was unaffected.
CV-4 contained the exchange of site Ib. Although this variant appeared to bind to the IL-6R, proliferation assays revealed that CV-4 was CNTFR-selective. Biological activity of CV-4 was, however, ϳ100-fold reduced as compared with CNTF and appeared to be similar to CLC. We have, however, no explanation why CV-4 bind to IL-6R but did not induce proliferation of Ba/F3-IL-6R/gp130/LIFR cells. CV-2 and CV-3 contained the extended or the minimal site Ia from CLC, respectively. Exchange of the minimal site Ia did not affect receptor selectivity and biological activity of CV-3 toward CNTFR/gp130/LIFR but, surprisingly, increased the biological activity toward IL-6R⅐gp130⅐LIFR complexes. Exchange of the extended site Ia in CV-2 abrogated interaction with IL-6R, but at the highest concentration applied, CV-2 still induced cytokine-dependent proliferation via the IL-6R⅐gp130⅐LIFR receptor complex, albeit no detectable STAT3 phosphorylation. Moreover, the biological activity toward CNTFR⅐gp130⅐LIFR complexes was reduced by a factor of 10 for CV-2. From these data, we concluded that simple combination of site Ia and site Ib would result in CNTFR-selectivity but with overall reduced biological activity toward CNTFR⅐gp130⅐LIFR complexes. Therefore, we decided to combine all site exchanges, including the amino acid substitution R28E in CV-5. Albeit CV-5 was CNTFR-selective, exchange of the complete site Ia and Ib from CLC to CNTF abolished the positive affinity effect of R28E observed in CV-1.
All CNTF variants were generated using the human CNTF cDNA. Murine CNTF has about 82% sequence homology with human CNTF. In CNTF, Arg 28 and the surrounding amino acids are identical between mice and humans; therefore, we assume that either receptor selectivity will be transferrable to murine CNTF and/or CV-1 will maintain its receptor selectivity also on murine receptor complexes. This must, however, to be studied before CV-1 will be tested in in vivo studies.
The CNTFR-selective CV-1 may be of particular interest in view of the therapeutic potential of CNTF in neurodegenerative diseases and obesity and may even allow the reinvestigation of the therapeutic effects of CNTF at higher dosages.