Cross-linking identifies leukemia inhibitory factor-binding protein as a ciliary neurotrophic factor receptor component.

Ciliary neurotrophic factor (CNTF) and leukemia inhibitory factor (LIF) are cytokines that give rise to an identical set of tyrosine-phosphorylated proteins upon addition to responsive cells. One of these proteins is the interleukin-6 signal-transducing molecule gp130, which is required for signal transduction by both CNTF and LIF. Here we identify another prominent tyrosine-phosphorylated protein as LIF receptor (LIFR) beta, which was originally cloned as a LIF-binding protein. Cross-linking experiments with iodinated factors were carried out on a cell line responsive to CNTF and LIF, as well as on COS cells that were cotransfected with various combinations of gp130, LIFR beta, and CNTF receptor (CNTFR) alpha, the previously cloned CNTF-binding protein. These experiments reveal that LIF cross-links to LIFR beta alone, as well as to gp130 when it is coexpressed with LIFR beta. However, cross-linking of CNTF to LIFR beta and gp130 is only observed in the presence of CNTFR alpha. These and other data show that the two known LIF receptor components are recruited by CNTF and CNTFR alpha to form a trimeric CNTF receptor complex.

* 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.
Candidate CNTF receptor components were discovered through comparative studies of signal transduction following addition of CNTF, LIF, or IL-6 to responsive cell lines ( 7 ) . A distinct set of proteins called CLIPS (CNTF and &IF-induced phosphoproteins) became rapidly phosphorylated on tyrosine after stimulation by CNTF and LIF. Two of these phosphoproteins, CLIPl and CLIP2, were transmembrane proteins that could be specifically recovered in a complex with biotinylated CNTF, suggesting that they constituted receptor components ( 7 ) . CLIP2 has been identified as gp130 using monoclonal antibodies, and we have suggested that CLIPl is LIFRP (a protein first identified as a LIF-binding protein; Ref. 8) based on its molecular weight and the identity of the phosphorylation pattern induced by LIF and CNTF. We therefore proposed that a functional CNTF receptor is composed of CNTFRa, gp130, and LIFRp ( 7 ) . Since functional LIF receptors appear to require only gp130 and LIFRP (8,9), a consequence of this model is that a functional LIF receptor can be converted into a functional CNTF receptor by the presence of CNTFRa (4, 7 , 10, 11). Thus in a sense, the LIF signaltransducing machinery, which occurs with widespread distribution, would be appropriated for use by CNTF in the restricted set of cells that also express CNTFRa (11). Recently, we have shown that CNTFRa can function when present in a soluble form (10). The detection of soluble CNTFRa in cerebrospinal fluid, as well as its release from muscle after nerve injury, enhances the possibility that CNTF could act at LIF receptors through the regulated release of soluble CNTFRa in vivo (10).
In order to test our conjecture that CLIPl is LIFRP and to determine whether CLIPl and CLIP2 directly interact with CNTF, we employed specific antisera and cross-linking approaches on cells bearing functional receptors, as well as on COS cells transfected with various combinations of CNTFRa, gp130, and LIFRP. The results of these experiments confirm that gp130 and LIFRP are receptor components for CNTF, and along with CNTFRa, appear to directly contact CNTF.

EXPERIMENTAL PROCEDURES
Reagents-Recombinant rat CNTF was prepared as described (12). LIF and anti-phosphotyrosine (4G10) conjugated to agarose were purchased from Upstate Biotechnology, Inc. The human LIFRp open reading frame was cloned by polymerase chain reaction methodology using primers derived from the published sequence,' while the clones for gp130 (6) and CNTFRa (3) were described previously. Antisera recognizing LIFRp was raised in rabbits against a COOH-terminal peptide with the sequence KDEDSPKSNGGGWSFTNFFQNKPND, which was synthesized with an NH'-terminal extension of CGG to facilitate coupling to purified protein derivative of tuberculin (Parke-Davis) using m-maleimidobenzoyl-N-hydroxysuccinimide. Iodination of CNTF was carried out with either Bolton-Hunter reagent (Du Pont-New England Nuclear) or soluble lactoperoxidase and sodium iodide (13); only the latter procedure was used for LIF. lZ5I-LIF was purified from unincorporated iodine on an Econo-Pac lODG desalting column (Bio-Rad). '"I-CNTF was purified by chromatography on a Superdex HR-75 column (Pharmacia LKB Biotechnology, Inc.) to allow separation of monomer from dimer; the monomeric fraction was used in all experiments. Specific activities of the iodinated probes ranged from 650 to 2000 Ci/mmol. Iodination on tyrosine was reported not to affect the bioactivity of LIF (14); '"I-CNTF generally showed full retention of bioactivity on ciliary ganglia. IARC-EW-1 (hereafter referred to as EW-1) cells were cultured as previously described (7). Construction of Myc tags on CNTFRa and LIFRP, insertion of the coding regions in the COS expression vector CMX, and transfection of COS cells is described elsewhere.' Cross-linking and Immunoprecipitation-Six-well plates of the indicated cells were incubated with 1 nM iodinated factors for 30 min at room temperature in a buffer consisting of phosphate-buffered saline + 1 mg/ml bovine serum albumin. Where indicated, either 1 pg/ml CNTF (44 nM) or 0.5 pg/ml LIF (22 nM) was added as an unlabeled competitor. Experiments employing precipitation with a-Tyr(P) instead used 1 nM '"I-CNTF or 2 nM '2sI-LIF f 100 nM nonradioactive competitor in a 10-min incubation. For cross-linking, plates were placed on ice and incubated for 20 min with 0.2-0.5 mM disuccinimidyl suberate taken from a 100 mM stock freshly made in dimethyl sulfoxide. Following 2 washes with cold TBS (20 mM Tris hydrochloride, pH 7.8, 0.15 M sodium chloride), the cells were lysed in TBSN (TBS containing 1% Nonidet P-40, 1 mM EDTA, 5 pg/ml leupeptin, 0.14 units/ml aprotinin, 1 mM sodium vanadate, and 1 mM phenylmethylsulfonyl fluoride) for 15-30 min on ice. After centrifugation for 15 min at top speed in a microcentrifuge, the lysate was either subjected to SDS-PAGE directly or incubated overnight at 4 "C in the indicated antibody, then 1 h in goat anti-mouse IgG conjugated to agarose (for a-gpl30; Sigma) or protein A-Sepharose (for a-LIFRS; Pharmacia). The beads were washed 3 times in 1 ml of lysis buffer, then boiled in sample buffer and subjected to SDSpolyacrylamide gel electrophoresis on gels containing 7% acrylamide. The gels were dried before autoradiography.
Anti-pbspbtyrosim Blots-Immunoblots probed with a-Tyr(P) were visualized by enhanced chemiluminescence (ECL; Amersham Corp.). Following electrotransfer to polyvinylidene difluoride, blots were blocked with 10% BSA in TBS for 3 h, then incubated for 3 h with the a-Tyr(P) monoclonal 4G10 at a dilution of 1:4000 in a buffer containing TBST (TBS + 0.1% Tween 20) and 4% BSA. Blots were then washed 3 X 10 min in TBST, and incubated for 1 h in goat antimouse IgG conjugated to horseradish peroxidase (Promega) diluted 1:20,000 in TBST containing 2.5% BSA. The blots were then washed 3 X 10 min with TBST, 3 X 5 min with TBS + 0.3% Tween 20, followed by a 30-min wash in TBS. The ECL reaction was then carried out according to the manufacturer's recommendations.

RESULTS
Specific Antisera Verify That CLIPl Is LIFRP-Previous experiments revealed that a human Ewing sarcoma (EW-1) cell line gives a robust tyrosine phosphorylation of CLIPl and CLIP2 in response to either CNTF or LIF. Monoclonal antibodies were used to demonstrate that CLIP2 is gp130, t h e IL-6 signal transducer (7). In order to test whether CLIPl corresponds to LIFRP, we employed polyclonal antisera raised against a carboxyl-terminal peptide of human LIFRP. This antisera specifically immunoprecipitated CLIPl and depleted CLIPl from lysates of EW-1 cells stimulated with either C N T F or LIF (Fig. lA); recognition was inhibited by the presence of the cognate peptide. Coprecipitation of CLIP2 is not observed in this experiment in which the lysates were heated to 95 "C before incubation with the antisera; coprecipitation of CLIP2 is observed if the heating step is omitted (data not shown). These results demonstrate that LIFRP corresponds to CLIPl and becomes tyrosine-phosphorylated in response to C N T F or LIF.
Cross-linking of CNTF and LIF to Endogenous Receptor Components-We next employed cross-linking of iodinated factors to determine whether CNTF directly contacts gp130 and LIFRB. EW-1 cells were cross-linked to lZ5I-CNTF or lZ5I-LIF, followed by immunoprecipitation with antibodies to phosphotyrosine, gp130, or LIFRP (Fig. 1B). All three antibodies precipitated a common set of radioactive bands after  cross-linking to lZ5I-CNTF, consistent with the formation of a complex containing CNTF together with CNTFRa and the tyrosine-phosphorylated forms of gp130 and LIFRP. As expected, the apparent molecular weights of two labeled bands (indicated by * and + in Fig. lB, lune 3 ) correspond to those predicted for the cross-linked products of lZ5I-CNTF (22 kDa) with CNTFRa (80 kDa) and LIFRP/CLIPl (190 kDa). Although the major remaining cross-linked product is larger than would be expected for gpl3O/CLIP2 (145 kDa) crosslinked to CNTF (165 kDa; in Fig.   lB), experiments using a transfected version of gp130 verify that this product migrates aberrantly (see below). Higher molecular weight species observed in some cases most likely contain multiple receptor components. Bands corresponding to LIFRPICLIPl and gpl3O/CLIP2 are also observed cross-linked to '251-LIF (Fig.  1B, lune I ) and may be equivalent to species previously observed after LIF cross-linking (14). Although the LIF-crosslinked products migrate slightly faster relative to the corresponding products resulting from cross-linking to CNTF, this difference is also observed with transfected versions of these receptor components (see below). These cross-linking results assert that gp130 and LIFRP directly contact CNTF in the receptor complex, consistent with their roles as signal-transducing receptor components.

Cross-linking of C N T F and LIF to Transfected Receptor
Components-We next turned to a COS cell expression system to explore the combinations of receptor components required to reconstruct a receptor complex mimicking that A. CotlilnSfectbn with LIFRP-myc and gp130 found in EW-1 cells. The LIFRP and CNTFRa coding sequences were modified to include a 10-amino acid "tag" that corresponds to a well-defined peptide epitope from c-Myc. This epitope is recognized by monoclonal antibody 9E10, allowing for unequivocal identification of the Myc-tagged proteins (3). 2 Expression of LIFRB-Myc alone gives good cross-linking to '"I-LIF as expected since this protein was identified as a LIFbinding protein (8), while cross-linking to '251-CNTF is not observed (Fig. 2, left panel). Expression of gp130 alone gives very faint cross-linking to either factor (Fig. 2, center panel). Coexpression of CNTFRa-Myc with gp130 not only allows for cross-linking of CNTF to CNTFRa (lower band in Fig. 2,  right panel) but also increases slightly the amount of lZ5I-CNTF cross-linked to gp130 (upper band in Fig. 2, right  panel); this result is consistent with independent evidence proving that CNTF can form a relatively stable interaction with CNTFRa and gp130 in the absence of LIFRP.' The identities of the receptor components in the various crosslinked products were verified by immunoprecipitation with 0-gp130 and a-Myc (not shown).
Cotransfection of gp130 and LIFRP-Myc in the absence of CNTFRa allowed for prominent cross-linking of lZ5I-LIF to both gp130 and LIFRP, while no cross-linking to '251-CNTF was observed (Fig. 3A, left panel). Specific precipitation of the '251-LIF-containing species was achieved by a-gpl30 or a-Myc (Fig. 3A, right panel). These two cross-linked products reproduce those observed in EW-1 cells cross-linked to '251-LIF; this result is consistent with the notion that LIFRP and gp130 are sufficient to form a functional LIF receptor complex, while CNTFRa is also required to form a functional CNTF receptor complex (4, 7, 10, 11).
COS cells cotransfected with CNTFRa-Myc, gp130, and LIFRP-Myc now reveal three main products upon crosslinking to lZ5I-CNTF (Fig. 3B, left panel) pattern observed in EW-1-cells. Immunoprecipitation with athose resulting from CNTF cross-linking, reproducing the recovers a complex containing all three products (Fig. 3B). iments show that while LIFRP and gp130 expression are and LIFRP in these transfected COS cells; as above, the CNTFRa is required to form a ternary receptor complex and @130 primarily recovers the product, a"yc behavior observed in EW-1 cells. These reconstruction exper-SuTrisingly* LIF cross-1inks to CNTFRf-Y as we11 as ~1 3 0 sufficient to observe cross-linking to LIF, coexpression of identities Of the cross-1inked products were confirmed by allow cross-linking of CNTF to all three receptor components. immunoprecipitation (Fig-3B)-Although LIF cross-linking This result corroborates previous findings that physiologic to CNTFRa is easily observed when all three components are levels of CNTF could not induce transduction events over-expressed in COS cells, a minor product of similar size in cells bearing gp130 and LIFRP in the absence of CNTFRa is also observed in EW-1 cells (Fig. 1B). The products result-(7, ing from cross-linking to LIF migrate slightly faster than Cross-competition between CNTF ad LIF-one prediction of our models for the LIF and CNTF receptors is that the factors should cross-compete for cross-linking to shared oncostatin M and LIF (9), as well as between human IL-3 3B shows that cotransfection of CNTFRa, gp130, and LIFRP gives rise to a situation in which a 22-fold excess of LIF does excess of CNTF does compete for cross-linking of '251-LIF to all three components. As would be expected, competition of CNTF for 1251-LIF cross-linking is dependent on the expres- sion of CNTFRa (Fig. 3A). The observed inability of LIF to CNTRamyc compete for CNTF cross-linking could be explained by the FIG. 2. Cross-linking of 1261 trations of LIF did in fact effectively compete for cross-linking of CNTF in EW-1 cells (data not shown). 3) Finally, since CNTF cannot compete for LIF binding to LIFRP in the absence of CNTFRa, competition by CNTF of 1251-LIF binding to LIFRP upon cotransfection of all components (Fig. 3B) implies that both gp130 and CNTFRa must be in excess of LIFRP in this experiment. Differences in cross-competition between IL-3 and IL-5 resulting from different ratios of receptor component expression have been noted previously (15).

DISCUSSION
The results presented here demonstrate that CLIP1, the 190-kDa protein that becomes tyrosine-phosphorylated after addition of CNTF and LIF to responsive cells, is indeed LIFRP. The ability to cross-link CNTF and LIF to both LIFRP and gp130 indicates that these factors come in close proximity to, and probably contact, both LIFRp and gp130 in their functional, signal-transducing receptor complexes. Although the cross-linking approaches described here are only qualitative and do not reveal the stoichiometry of each component, other results suggest a 1:l ratio of LIFRP and gp130 in the functional complex.2 The finding that cross-linking of CNTF to gp130 and LIFRB occurs only in the presence of CNTFRa supports our earlier proposal (4, 7, 10, 11) that CNTFRa is required to convert a heterodimeric LIF receptor complex (ie. LIFRP and gp130) into a tripartite CNTF receptor complex. Other experiments using epitope-tagged versions of gp130, LIFRP, and CNTFRa expressed in COS cells revealed that complex formation between CNTFRa, gp130, and LIFRj3 does not occur in the absence of added CNTF, and also showed that CNTF can form a stable, non-functional complex with CNTFRa and gp130 in the absence of expressed LIFR/3.2 These experiments demonstrate that addition of CNTF drives complex formation from three separate, previously non-interacting receptor components, which may well proceed in an ordered fashion through intermediates in which CNTF first binds to CNTFRa, followed by gp130, then LIFRB. Complex formation then presumably initiates the signal transduction process (see below).
A surprising result was the observation that 12'I-LIF could cross-link to CNTFRa in the presence of LIFRP and gp130. CNTFRa is not required to initiate signal transduction by LIF (7, lo), although there is some evidence that the presence of CNTFRa may augment the responsiveness of cells to LIF (10). Other results indicate that the interaction of CNTFRa with the heterodimeric LIF receptor complex is not stable since CNTFRa is not coprecipitated with gp130 and LIFRP when the complex is formed in the presence of LIF, while it is recovered if the complex contains CNTF.2 It is conceivable that cross-linking of LIF to CNTFRa results not only from overexpression of CNTFRa, but also reflects direct receptorreceptor contacts made between CNTFRa and the heterodimer consisting of gp130 and LIFRP, which brings CNTFRa into the proximity of the bound LIF. We find no clear evidence for any LIF-specific a receptor components, although faint cross-linked products of 80-120 kDa are occasionally observed in cells lacking CNTFRa (e.g. Fig. 2, left and middle panels).
Cytokine receptors have no identifiable catalytic domain, and the mechanism by which they initiate signal transduction remains unclear. However, a 61-amino acid cytoplasmic domain in gp130 that has been identified as critical for function (17) is conserved in a variety of cytokine receptor / 3 subunits including LIFRj3 (8), erythropoietin receptor, KH97 (a GM-CSF receptor component), G-CSF receptor, and the IL-3 receptor (17). It is conceivable that the function ofthis domain is to interact with downstream signal-transducing molecules, such as a cytoplasmic tyrosine kinase. In fact, all cytokines appear to induce tyrosine phosphorylation in responsive cells, and tyrosine kinase activity has been found to coprecipitate with functional signal-transducing complexes for the erythropoietin receptor (18,23), as well as for the CNTF and LIF receptor complexe~.~ We have previously proposed that dimerization of "/3" type cytokine receptor subunits initiates subsequent signaling events (4, 7, 10, 11). The rapid tyrosine phosphorylation of LIFRp and the presence of the conserved cytoplasmic region mentioned above indicates that LIFRP is likely to function as a signal-transducing partner to gp130. Thus for CNTF and LIF, heterodimerization due to apposition of gp130 and LIFRP may be analogous to the homodimerization of the P-type receptors for erythropoietin (19) , G-CSF (20), human growth hormone (21,24), and perhaps IL-6 (4, 7). Thus it is possible that a cytoplasmic tyrosine kinase is non-covalently associated with the cytoplasmic domain of each /3 receptor, and that either heteroor homodimerization of two P subunits would allow for kinase transactivation in a manner analogous to receptor tyrosine kinases (22).