Signaling of Human Ciliary Neurotrophic Factor (CNTF) Revisited AS AN (cid:1) FOR CNTF*

THE INTERLEUKIN-6 RECEPTOR CAN SERVE -RECEPTOR Human ciliary neurotrophic factor (CNTF) is a neurotrophic cytokine that exerts a neuroprotective effect in multiple sclerosis and amyotrophic lateral sclerosis. Clinical application of human CNTF, however, was pre-vented by high toxicity at higher dosages. Human CNTF elicits cellular responses by induction of a receptor complex consisting of the CNTF (cid:1) -receptor (CNTFR), which is not involved in signal transduction, and the (cid:2) -recep-tors gp130 and leukemia inhibitory factor receptor (LIFR). Previous studies with rat CNTF demonstrated that rat CNTF is unable to interact with the human interleukin-6 (cid:1) -receptor, whereas at high concentrations, it can directly induce a signaling heterodimer of human gp130 and human LIFR in the absence of the CNTF receptor. Here, we demonstrate that human CNTF cannot directly induce a heterodimer of human gp130 and LIFR. However, human CNTF can use both the membrane-bound and the soluble human IL-6R as a substitute for its cognate (cid:1) -receptor and thus widen the target spectrum of human CNTF. Engineering a CNTFR-specific human CNTF variant may therefore be a pre-requisite

The expression of CNTF is restricted to Schwann cells in the peripheral and astrocytes in the central nervous system (CNS) (14,15). Although involved in neuronal differentiation processes in vitro (16), CNTF is primarily considered as a lesion factor because of the lack of a signal peptide in the CNTF gene and its immunohistochemically determined cytosolic localization (1). Several studies showed that CNTF prevents degeneration of axotomized peripheral motor neurons and also retrograde cell death of neurons in thalamic nuclei after dissection of intracerebral neuronal circuits (17,18). Intraperitoneal implantation of a CNTF-producing cell line improved survival in a mouse model of motor neuropathy (19), and in different models of HuntingtonЈs disease, CNTF exerted a neuroprotective effect on striatal cells (20 -22). CNTF also enhanced the survival of sensory, hippocampal, and cerebellar neurons (23)(24)(25) and increased survival of retinal photoreceptors in animal models of retinal degeneration (26 -28). On denervated rat skeletal muscles, CNTF reduced denervation-induced atrophy and increased muscular strength (29,30).
CNTF knock-out (Ϫ/Ϫ) mice display a mild phenotype showing progressive loss of motor neurons and reduced muscular strength only in the adult, not in childhood (31). Cross-breeding of CNTF Ϫ/Ϫ mice with LIF Ϫ/Ϫ mice resulted in a much more prominent phenotype, suggesting that LIF also exerts a physiological neurotrophic effect (32). In contrast to CNTF Ϫ/Ϫ mice, CNTFR Ϫ/Ϫ mice died within 24 h postnatally because of a suckling defect caused by severe motor neuron deficits (33). This indicated the existence of a second ligand for the CNTFR. Surprisingly, the knock-out of an orphan receptor, NR6, resulted in a phenotype almost identical to that of CNTFR Ϫ/Ϫ mice (34). This enigma was resolved by the discovery, that CLC is the second ligand to the CNTFR and that NR6 is identical to another orphan receptor, CLF, which is essential for transport to the cell surface and secretion of CLC (9,35,36).
In contrast to mice, a null mutation of the CNTF gene found in high prevalence in the Japanese population was not causally related to neurological diseases (37). However, recent studies strongly suggest that the absence of CNTF modifies the course of some important neurological diseases. Postmortem spinal cord samples from patients with amyotrophic lateral sclerosis (ALS) showed decreased levels of CNTF (38), and in patients with ALS due to a mutation of the superoxide dismutase gene, CNTF deficiency was associated with an earlier onset and a more severe course of the disease (39). CNTF-deficient patients also displayed a significantly earlier onset of multiple sclerosis (MS) and more severe symptoms (40). Apparently, CNTF exerts a protective effect in demyelinating disease by preventing apoptosis of oligodendrocytes (41). CNTF also exerts an antiinflammatory effect in the central nervous system by inhibition of TNF␣ production and activation of the hypothalamic pituitary axis (42). The association of CNTF deficiency with clinically more severe MS in humans is all the more surprising, since a model disease of MS in mice, experimental autoimmune encephalitis, cannot be elicited in IL-6 Ϫ/Ϫ mice (43)(44)(45)(46).
In recent mouse studies, CNTF activated hypothalamic satiety centers independently of leptin and reduced obesity and diabetes associated with overweight or leptin resistance (47,48). At concentrations sufficient for weight reduction, no intolerable side effects occurred. It was suggested that CNTF could therefore be the basis for new strategies to combat human obesity provided the chemical structure of CNTF could be redesigned to reduce or eliminate side effects (49). In a toxicity trial of CNTF in humans, cachexia, aseptical meningitis, reactivation of virus infections, and respiratory failure were observed at high doses (Ͼ5 g/kg/day) after subcutaneous injection of CNTF (50). In a placebo-controlled trial of CNTF in patients with ALS, the dosage was therefore limited to 5 g/ kg/day, where, however, no beneficial clinical effects were observed compared with placebo (51).
In experiments with human HepG2 hepatoma cells, rat CNTF, which has an 85% amino acid identity to human CNTF, induced expression of acute phase proteins (52,53). It was explicitly shown that the rat CNTF activity on human HepG2 cells was not relayed via the human IL-6R (52), although apparently, rat CNTF stimulates the acute phase response on primary hepatocytes after binding to the rat IL-6R (54). At high concentrations, rat CNTF induces an active heterodimer of human gp130 and human LIFR even in the absence of the CNTFR␣ (11). We have recently observed activity of recombinant human CNTF at high concentrations on HepG2 cells, which do not express the CNTFR (53), but did not observe a similar effect on murine BaF/3 cells stably transfected with human gp130 and human LIFR. The discrepancy prompted the present study, which demonstrates that in contrast to rat CNTF, human CNTF is unable to elicit a heterodimer of human gp130 and LIFR in the absence of an ␣-receptor. However, the human IL-6R can serve as a substitute ␣-receptor for human CNTF in the assembly of a functional receptor complex. Molecular modeling revealed the molecular basis for the difference between human and rat CNTF in binding to the human IL-6R.  (55). The murine monoclonal antibody (mAb) ab6276 (anti-␤-actin) was bought from abcam (Cambridge, UK), the polyclonal rabbit antibody anti-phospho-STAT3 (Tyr-705) was from Cell Signaling Technology (Beverly, CA), polyclonal horseradish peroxidase-coupled goat anti-rabbit and antimouse antibodies were obtained from Pierce.

Cells and Reagents-Human
The neutralizing anti-IL-6 receptor mAb PM-1 has been described before (56). Soluble IL-6R and CNTFR were obtained from R&D Systems (Wiesbaden, Germany). [ 3 H]thymidine was from Amersham Biosciences. Brij-96 and all other reagents were bought from Sigma-Aldrich. OSM was cloned into the pET-14b bacterial expression vector (Novagen, Schwalbach, Germany) and purified with a Ni-chelate column (Amersham Biosciences). All other human or designer proteins used in this study were produced in Escherichia coli bacteria as described before (55,57,58).
Bioassays-Proliferation of the transfected BAF/3-[gp130] cell lines in response to human IL-6, human CNTF, human LIF, and the designer cytokine IC7 was measured in 96-well microtiter plates. The cells were exposed to test samples for 72 h and subsequently pulse-labeled with [ 3 H]thymidine for 4 h. Proliferation rates were measured by harvesting the cells on glass filters and determination of the incorporated radioactivity by scintillation counting. Each proliferation assay was performed at least three times in triplicates.
Analysis of STAT3 Phosphorylation-A lysis buffer consisting of 50 mM Tris, pH 7.5, 100 mM NaCl, 50 mM sodium fluoride, 3 mM sodium orthovanadate, 1% Brij-96, and proteinase inhibitors (1 mM phenylmethylsulfonyl fluoride, one tablet of the Roche proteinase inhibitor mixture) was prepared. Transfected BaF/3-[gp130,LIFR,IL-6R] cells were starved for 4 h in serum-free Dulbecco's minimal essential medium supplemented with penicillin and streptomycin. After stimulation with cytokines (50 ng/ml) for 10 min at 37°C, cells were pelleted and resuspended in 1 ml of lysis buffer. An aliquot of 40 l was subjected to SDS-PAGE and blotted to a polyvinylidene difluoride membrane (Amersham Biosciences). The membranes were incubated with the antiphospho-STAT3 antibody before being labeled with a secondary antibody coupled to peroxidase. Subsequently, the membranes were developed using the Amersham Biosciences ECL chemiluminescence kit. HepG2 cells were grown on 24-well plates (TPP, Biochrom, Berlin, Germany) starved overnight in serum-free medium before stimulation for 10 min with the indicated cytokines. The supernatants were removed, and the cells lysed by addition of Laemmli buffer (2ϫ concentrated). The lysate was subsequently Western blotted as above. Loading was controlled by a Western blot against ␤-actin.
CD Spectra-CD spectra of all cytokines produced were recorded with a Jasco J-720 spectropolarimeter (Japan Spectroscopic Co., Ltd., Tokyo, Japan) to check correct refolding. The instrument was calibrated according to Chen and Yang (73). The spectral bandwidth was 2 nm. The measurements were carried out at a temperature of 23°C, the solvent was phosphate-buffered saline, pH 7.4, throughout.
Determination of the Affinity Constants of IL-6 and CNTF Binding to the IL-6R by Plasmon Resonance-IL-6R-Fc was covalently immobilized to a carboxymethyl dextran matrix (Fisons, Loughborough, UK) at 28.0 g/ml for 5 min in 10 mM sodium acetate buffer, pH 5.0, as recommended by the manufacturer. Binding experiments were performed at controlled temperature (25°C) with different concentrations of purified IL-6 and CNTF protein using the IASYS TM optical biosensor (Affinity Sensors, Cambridge, UK). Association was monitored for at least 2 min, the sample was replaced by phosphate-buffered saline/ 0.05% Tween 20, dissociation was monitored, and the cuvette was equilibrated again in phosphate-buffered saline/0.05% Tween 20. Association and dissociation affinograms were analyzed by nonlinear regression with the FAST fit (Fisons) software, which uses the Marquardt-Levenburg algorithm for iterative data fitting.
Molecular Modeling-The model of the CNTF/IL-6R complex was built using the x-ray structure of CNTF (Ref. 59, PDB accession code: 1cnt) and the model of the IL-6/IL-6R complex (60) as a template. In a first step the CNTF molecule was fitted onto the IL-6 model (using only the C␣ positions of helices A and D). The next steps were performed in an iterative manner. First, the interaction area was inspected for overlapping side chains. Unfavorable contacts were then eliminated by rotating them properly. Second, the accessible surface was calculated for this complex to find cavities in the interaction area. If possible these cavities were filled by adjustment of side chains from their neighborhood. These complexes were then energy-minimized using the steepest descent algorithm implemented in the GROMOS force field (61) and again analyzed for unfavorable contacts and cavities in the interaction area. This procedure was repeated until a low energy conformation of the complex was reached. Accessible surfaces were calculated using the algorithm implemented in the software package WHATIF (62). For graphical representation the Ribbons program was used (63). All programs were run on a Silicon Graphics Indy workstation.

RESULTS
Murine BaF/3-cells do not express IL-6-type cytokine receptors, but proliferate in response to human IL-6 type cytokines upon transfection of appropriate human receptors. To analyze the interaction of human CNTF with different receptors of the IL-6 family, we made use of a set of murine BaF/3 cells stably transfected with different combinations of human IL-6-type receptors (Fig. 1). gp130 and LIFR are expressed at equal levels in these cell lines (55). Human CNTF-stimulated proliferation of BaF/3-cells stably transfected with gp130, LIFR, and CNTFR (Fig. 1A), but even at very high concentrations up to 10 g/ml, it was inactive on BaF/3-[gp130,LIFR] cells that do not express the CNTFR (Fig. 1B), suggesting that the CNTFR is an absolute requirement for activity of human CNTF. Similarly, CNTF was inactive on BaF/3-[gp130,IL-6R] cells, which lack the LIFR and the CNTFR (Fig. 1C). Unexpectedly, however, CNTF showed activity on BaF/3-[gp130,LIFR,IL-6R] cells that expressed the human IL-6R instead of the CNTFR (Fig. 1D). Maximal activity of CNTF was achieved at concentrations exceeding 250 ng/ml, but a clear effect of CNTF was already discernible at 10 ng/ml. LIF, IL-6, and the designer cytokine IC7, which induces a gp130/LIFR heterodimer after binding to the IL-6R (55), stimulated proliferation with ED 50 of 0.6, 0.9, and 0.5 ng/ml, whereas an about 42-fold higher CNTF concentration (ED 50 ϭ 30 ng/ml) was required for the same stimulatory effect.
To further substantiate binding of CNTF to the IL-6R, we tested the ability of the neutralizing anti-human IL-6R antibody PM-1 to block binding of CNTF to the IL-6R. PM-1 binds to the cytokine binding epitope of the IL-6R and thus inhibits binding of IL-6 to its cognate ␣-receptor (56). PM-1 did not inhibit LIF-induced proliferation of BaF/3-[gp130,LIFR,IL-6R] cells ( Fig. 2A), nor phosphorylation of STAT3 induced by LIF (Fig. 3, A and B). In contrast, there was a clear competitive inhibition of BaF/3-[gp130,LIFR,IL-6R] cell proliferation by PM-1 after stimulation with IL-6, IC7 and importantly also CNTF (Fig. 2, B-D). Consistent with the competitive inhibition by PM-1 of IL-6, IC7-and CNTF-dependent proliferation of BaF/3-[gp130,LIFR,IL-6R] cells, phosphorylation of STAT3 after stimulation with CNTF was almost completely inhibited by PM-1 (Fig. 3, A and B). Phosphorylation of STAT3 in BaF/3-[gp130,LIFR,IL-6R] cells was also inhibited by PM-1 after stimulation with 50 ng/ml IL-6 or IC7, but not after stimulation with LIF. The inhibitory effect of PM-1 on CNTF-induced proliferation of BaF/3-[gp130,LIFR,IL-6R] cells, however, was abolished when these cells were stimulated by the combination of CNTF and the soluble CNTFR (Fig. 2D), demonstrating that inhibition of CNTF activity by PM-1 on BaF/3-[gp130,LIFR,IL-6R] cells was specific for interaction of CNTF with the IL-6R.
Because we and others (53,55) have shown, that CNTF is biologically active on HepG2 cells at high concentrations, we wanted to ascertain whether this effect also depended on binding of CNTF to the IL-6R. We therefore stimulated HepG2 cells with increasing amounts of IL-6 and CNTF as well as OSM, which, like LIF, does not require binding to an ␣-receptor to induce phosphorylation of STAT3 via a gp130/LIFR or gp130/ OSMR heterodimer. Similar to our results with BaF/3-[gp130,LIFR,IL-6R] cells, PM-1 competitively inhibited phosphorylation of STAT3 after stimulation with IL-6 and CNTF, but had no effect on the cellular response to OSM (Fig. 4, A and  B).
We next wanted to analyze whether the ability of CNTF to induce cellular responses via binding to the membrane-bound IL-6R also pertained to the soluble IL-6R. Maximal proliferative responses of BaF/3-[gp130,LIFR] cells to the combination IL-6/soluble IL-6R were achieved with concentrations of sIL-6R exceeding 80 ng/ml (Fig. 5A) g/ml in the presence of 100 or 500 ng/ml sIL-6R (Fig. 5C). As in the previous experiments, maximal responses to CNTF, which were similar to those observed after 50 ng/ml LIF (Fig.  5B), were observed at around 250 ng/ml CNTF. A marked proliferative response was already discernible at 50 ng/ml CNTF. There was no difference between cellular responses to CNTF in the presence of 100 and 500 ng of sIL-6R.
To demonstrate direct interaction between CNTF and the IL-6R, CNTF was incubated with IL-6R-Fc, a recently constructed fusion protein of the IL-6R and the Fc part of human IgG1, in the presence of 1% Brij 96 (57). Following precipitation with protein A-Sepharose and SDS-PAGE, CNTF could be detected in the precipitate with a Western blot against human CNTF (Fig. 6, lanes 4 -7). IL-6R-Fc alone did not react in the Western blot (lane 2). The polyclonal antibody specifically detected CNTF (lane 1) and did not cross-react with human IL-6 (lane 3). To quantify real-time binding of IL-6 and CNTF to the IL-6R, IL-6R-Fc was immobilized to a IASYS TM cuvette, and the association and dissociation rates for binding of CNTF and IL-6 to IL-6R-Fc determined (Fig. 7, A and B). A K D of 184 nM was measured for binding of IL-6 (k on ϭ 3.71 ϫ 10 4 M Ϫ1 s Ϫ1 , k off ϭ 6.83 ϫ 10 Ϫ3 s Ϫ1) , whereas the K D for CNTF was 9.0 M (k on ϭ 2.68 ϫ 10 2 M Ϫ1 s Ϫ1 , k off ϭ 2.44 ϫ 10 Ϫ3 s Ϫ1 ). For comparison, the reported dissociation constants for IL-6 binding to the complete IL-6R (64) and the membrane-proximal domain of the IL-6R alone (65) are 5-fold smaller and 2-fold higher, respectively, than the ones measured here for IL-6R-Fc. These results suggest that the 42-fold difference in CNTF activity after bind- ing to the IL-6R compared with IL-6 and IC7 (Fig. 1D) most probably reflects the roughly 50-fold lower affinity of human CNTF binding to the human IL-6R.
To understand the difference in binding to the human IL-6R between human and rat CNTF on a molecular level, we generated a three-dimensional model of the human CNTF/human IL-6R complex (Fig. 8A). The x-ray structure of the human CNTF and a model of the human IL-6/ILR-6 complex were used as templates as described under "Experimental Procedures." The human IL-6R is bound to site I of human CNTF, and therefore we analyzed which amino acid residues involved in the interacting epitope are conserved in human and rat CNTF. Interestingly, we found that only one residue, Gln-63 in human CNTF, is replaced by an arginine in the rat molecule. Fig. 8B shows a magnified view of the interaction area between the two human molecules. Gln-63 is deeply buried in the interface, closely surrounded by amino acid side chains from the CNTF (Asp-62, Trp-64, His-174) and IL-6R (Pro-126, Leu-127, Tyr-  188, Ser-246, Phe-248, Gln-300). The fact that rat CNTF does not bind to the human IL-6R, therefore, results from the central involvement of the Gln-63 side chain in the binding interface between CNTF and the IL-6R. Compared with Gln, the side chain of arginine is substantially more extended and can thus not be accommodated in the interaction area. The increased size (Gln, 143.8 Å 3 ; Arg, 173.4 Å 3 ) and additional charge of the arginine in rat CNTF, compared with the Gln-63 in human CNTF, sterically prevents complex formation with the human IL-6R. Since rat CNTF in contrast to human CNTF is able to directly induce an active heterodimer of human gp130 and human LIFR, we examined the structure of human CNTF and a newly created model of rat CNTF (not shown) for differences in site II and site III that would explain this effect. However, comparisons of these epitopes with regard to charge distribution and hydrophobic patch extension did not reveal an obvious cause for an increased affinity of rat CNTF site II/III to gp130 or the LIFR. DISCUSSION The results presented here clearly demonstrate binding of human CNTF to the human IL-6R with an affinity roughly 50-fold lower than that of IL-6 binding to the IL-6R. Human CNTF is capable of eliciting a functional gp130/LIFR heterodimer after binding to both, the membrane-bound or the soluble IL-6R. Even at high concentrations, human CNTF is unable to directly induce a human gp130/LIFR heterodimer. Thus, human CNTF differs on both accounts from rat CNTF that can directly induce an active heterodimer of human gp130 and LIFR (11) and does not bind to the human IL-6R (52). The ability of human CNTF to induce the acute-phase response on human liver cells, which do not express the CNTFR (53) is therefore due to the formation of an active receptor complex consisting of IL-6R, gp130, and LIFR (Fig. 4). Conspicuously, although rat CNTF behaves differently than human CNTF on human cells, it can also stimulate rat hepatocytes via the IL-6R (54). A certain degree of caution might therefore be applied to receptor-ligand binding studies using proteins from different species, since the results obtained may be artifactual and reflect a situation foreign to both species as in the case of CNTF.
The finding that the new IL-6 family member CLC, which signals via a gp130/LIFR heterodimer, also uses the CNTFR as an ␣-receptor (9, 36) already violated the concept that ␣-receptors primarily confer specificity to cytokine-dependent receptor complex assembly. Our results demonstrate that specificity of ␣-receptors is not even restricted to the same subgroup of IL-6 type cytokines, i.e. those signaling via a gp130 homo-or a gp130/LIFR heterodimer. Specificity of ␣ -receptors for certain cytokines may therefore not be an absolute property of the ␣-receptor, but a quantitative one. An important result of this study is that human CNTF can induce cellular responses after binding to the soluble IL-6R. This contrasts with the requirement of CLC for the membrane-bound CNTFR to elicit biological responses (66). Transignaling via soluble receptors of the IL-6 family significantly widens the array of potential target cells (67), and recent evidence suggests that this process plays a critical role in a variety of different diseases (68 -70).
Concentrations of 10 Ϫ100 ng/ml CNTF are sufficient to induce marked biological responses after binding to both, the membrane-bound and the soluble IL-6R, with maximal responses observed beyond 250 ng/ml. Average serum levels of sIL-6R are around 80 ng/ml. Intolerable toxicity of human CNTF was observed at doses greater than 5 g/kg of body weight in a toxicity trial (50) with the strongest and most dangerous effects occurring at the doses between 20 and 100 g/kg of body weight. Given that the extracellular water represents 40% of total body weight, this would result in an initial concentration of between 50 and 250 ng/ml CNTF in this trial, compared with usually undetectable levels of CNTF in the serum (71). Since serum levels of sIL-6R vary between 80 and 320 ng/ml (70), effective CNTF and sIL-6R concentration in the trial are in the same range as studied here. Thus, the clinical side effects observed which included aseptic meningitis, respiratory failure, and reactivation of viral infections, may not only be due to the interaction of CNTF with its cognate membranebound ␣-receptor, but probably due to activation of target cells hitherto considered unresponsive to CNTF via the soluble or membrane-bound IL-6R. In support of such mechanisms, an analysis of the STAT3 phosphorylation in the rat retina after intravitreal application of the CNTF analogue axokine demonstrated that the activation of STAT3 is not restricted to retinal neurons, which express CNTFR␣ predominantly, but is also found in surrounding glial cells (28). CNTF exerts an immunosuppressive effect in the CNS (42), and a CNTF gene defect aggravates multiple sclerosis in humans (40), whereas IL-6deficient mice are resistant to experimental autoimmune encephalitis, an animal model of multiple sclerosis (43)(44)(45)(46). Considering the very similar signaling pathways activated by CNTF and IL-6 (6), these different in vivo effects may reflect different target cells activated by these cytokines.
The side effects of CNTF in the toxicity study limited the CNTF concentration in a trial of CNTF for the treatment of amyotrophic lateral sclerosis to 5 g/kg of body weight, where no positive effects compared with placebo were noted (51). Construction of a human CNTF variant that specifically interacts with the human CNTFR may improve the safety profile of CNTF and thus allow to reinvestigate clinical benefits of the application of higher doses of the agent in amyotrophic lateral sclerosis. In the CNTF analogue axokine, Gln-63 of human CNTF was substituted by Arg-63 of rat CNTF (47), which is known to increase the specific activity of human CNTF (72). Since rat CNTF does not bind to the human IL-6R (52), it is conceivable that axokine does not bind to the human IL-6R. Our molecular model of the human CNTF/human IL-6R complex provides a rationale to understanding this difference, since it shows that sterically, Arg-63 cannot be accommodated in the interaction area between human CNTF and the human IL-6R. Axokine may thus be both, more active and more specific than natural human CNTF, which appears particularly desirable in view of the therapeutic potential of CNTF and may even allow the reinvestigation of the clinical effects of CNTF signaling in ALS at higher dosages.