GFRα-2 and GFRα-3 Are Two New Receptors for Ligands of the GDNF Family

The receptor for glial cell line-derived neurotrophic factor (GDNF) consists of GFRα-1 and Ret. Neurturin is a GDNF-related neurotrophin whose receptor is presently unknown. Here we report that neurturin can bind to either GFRα-1 or GFRα-2, a novel receptor related to GFRα-1. Both GFRα-1 and GFRα-2 mediate neurturin-induced Ret phosphorylation. GDNF can also bind to either GFRα-1 or GFRα-2, and activate Ret in the presence of either binding receptor. Although both ligands interact with both receptors, cells expressing GFRα-1 bind GDNF more efficiently than neurturin, while cells expressing GFRα-2 bind neurturin preferentially. Cross-linking and Ret activation data also suggest that while there is cross-talk, GFRα-1 is the primary receptor for GDNF and GFRα-2 exhibits a preference for neurturin. We have also cloned a cDNA that apparently codes for a third member of the GFRα receptor family. This putative receptor, designated GFRα-3, is closely related in amino acid sequence and is nearly identical in the spacing of its cysteine residues to both GFRα-1 and GFRα-2. Analysis of the tissue distribution of GFRα-1, GFRα-2, GFRα-3, and Ret by Northern blot reveals overlapping but distinct patterns of expression. Consistent with a role in GDNF function, the GFRαs and Ret are expressed in many of the same tissues, suggesting that GFRαs mediate the action of GDNF family ligands in vivo.

Glial cell line-derived neurotrophic factor (GDNF) 1 is a multipotent neurotrophic factor that has a variety of effects on cells of both the central and peripheral nervous systems (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15). Recently, a novel neurotrophic factor called neurturin that is 42% identical in amino acid sequence to GDNF was reported (16). Both GDNF and neurturin are synthesized in pre-pro forms and their precursor molecules are proteolytically processed to yield mature proteins of about 100 amino acids that assemble into disulfide-linked homodimers (7,16). Seven cysteine residues are present in both GDNF and neurturin and their spacing is identical (16). Although the biological activities of neurturin have not yet been thoroughly investigated, initial results indicate that they are very similar to those of GDNF.
The striking structural and biological similarities between GDNF and neurturin suggest that their action may be mediated by the same or related receptors. The receptor for GDNF consists of a complex of GDNF Family Receptor ␣-1 (GFR␣-1, previously abbreviated as GDNFR-␣) and the Ret protein tyrosine kinase (PTK) (22,23). GFR␣-1 is a glycosyl phosphatidylinositol-anchored cell surface molecule. GFR␣-1 binds to GDNF but cannot signal independently since it lacks a cytoplasmic domain. GDNF signaling is accomplished via association of the complex of GDNF and GFR␣-1 with Ret, resulting in activation of the Ret kinase.
GFR␣-1 mRNA is widely distributed in neuronal and nonneuronal tissues and is expressed throughout embryonic development to adulthood, implying a broad spectrum of biological functions (23,24). The other component of the GDNF receptor complex, Ret, is a receptor type PTK encoded by the ret protooncogene (25). Ret mRNA and protein are highly expressed in the central and peripheral nervous systems, as well as in the kidney (26,27). Various mutations in the ret gene are associated with human inherited diseases, including familial medullary thyroid carcinoma (28,29), multiple endocrine neoplasia type 2A (MEN2A) and 2B (MEN2B) (28 -32), and Hirschsprung's disease (33,34). Targeted disruption of the ret gene in knockout mice results in severe phenotypic defects, including renal agenesis or severe dysgenesis and lack of the entire enteric nervous system (35). These defects are very similar to those caused by GDNF null mutations (36 -38), implying that GDNF-mediated signaling through Ret is required for the development of these tissues. Much less severe defects, however, were detected in a number of neuronal structures in which both GFR␣-1 and Ret were expressed, such as the trigeminal and vestibular ganglia, the facial motor nucleus, the substantia nigra, and the locus coeruleus (35)(36)(37)(38). This suggests that either GDNF signaling is not required for the embryonic development of these structures, or that some unknown signaling molecules similar to GDNF or Ret may exist that can substitute for them. Alternatively, the embryonic development of these tissues may rely completely on other as yet unknown signaling systems.
In this paper we report the cloning of GFR␣-2 and GFR␣-3, two novel receptors related to GFR␣-1, and provide evidence that GFR␣-2 is a receptor for both GDNF and neurturin. Our data also indicate that GFR␣-1 is a receptor for neurturin as well as for GDNF. We describe a related cDNA that codes for a protein, GFR␣-3, that shares significant amino acid homology with both GFR␣-1 and GFR␣-2 and is likely to be a third member to the family of receptors for GDNF-related ligands.

EXPERIMENTAL PROCEDURES
Cloning of GFR␣-2 and GFR␣-3-A search of the GenBank data base for sequences related to GFR␣-1 resulted in the identification of a single related EST, H12981.Gb_Est1. Primers corresponding to nucleotides 47-65 (5Ј-CTGCAAGAAGCTGCGCTCC-3Ј) and 244 -265 (5Ј-CTTGTC-CTCATAGGAGCAGC-3Ј) of H12981.Gb_Est1 were synthesized and used for RT-PCR with human fetal brain mRNA (CLONTECH, catalog number 64019-1) as the template. A 218-nucleotide fragment was amplified, subcloned into pBlueScript (Stratagene, La Jolla, CA), and sequenced. The fragment was then radiolabeled with [ 32 P]dCTP using a Random Prime DNA Labeling Kit (Stratagene) according to the manufacturer's instructions. The oligonucleotide primers described above were also used as primers for PCRs to screen DNAs isolated from 27 pools (1500 clones each) of a rat photoreceptor cDNA library (22). A single positive pool was identified, leading to the identification of an individual rat cDNA clone from this pool by hybridization to the radiolabeled human PCR fragment described above. Filters were prehybridized at 55°C for 3.5 h in 200 ml of 6 ϫ SSC, 1 ϫ Denhardt's, 0.5% SDS, and 50 g/ml salmon sperm DNA. Following the addition of 2 ϫ 10 8 cpm of the radiolabeled probe, hybridization was continued for 18 h. Filters were then washed twice for 30 min each at 55°C in 0.2 ϫ SSC, 0.1% SDS and exposed to x-ray film overnight with an intensifying screen.
Cloning of GFR␣3-The GenBank data base was searched for sequences related to GFR␣-1 and GFR␣-2 using the Wisconsin sequence analysis package (Wisconsin Package version 9.0, Genetics Computer Group, Madison, WI). Oligonucleotide primers corresponding to regions near the ends of an EST AA238748.Gb_New2 were synthesized. Primers corresponding to AA238748.Gb_New2 were used for PCR screening of 83 pools of 1000 clones each from a rat E15 embryonic cDNA library. 2 A single positive pool was identified by this method, and the DNA fragment amplified from this pool was subcloned into a plasmid vector and labeled with [ 32 P]dCTP using a Random Primed DNA Labeling Kit (Stratagene) according to the manufacturer's instructions. Clones from the cDNA library pool that had been identified as positive by PCR were plated on 15-cm agarose plates and replicated on duplicate nitrocellulose filters for screening by hybridization to the radiolabeled insert. Hybridization conditions were the same as those described in the preceding section. Positive clones were sequenced as described below.
DNA Sequencing and Sequence Analysis-DNA sequencing was performed using an automated Applied Biosystems 373A DNA sequencer and Taq DyeDeoxy Terminator cycle sequencing kits (Applied Biosystems, Foster City CA). Comparison of the GFR␣-1 and GFR␣-2 sequences with public data bases was carried out using the FASTA computer algorithm (39). The peptide sequences of GFR␣-1, GFR␣-2, and GFR␣-3 were aligned using the Lineup program. All sequence analysis programs used were included in the Wisconsin sequence analysis package (Wisconsin Package Version 9.0, Genetics Computer Group, Madison, WI).

Binding of [ 125 I]GDNF and [ 125 I]NTN to NGR-38 and NNR-9
Cells-Recombinant human neurturin was expressed in Escherichia coli as insoluble protein sequestered into inclusion bodies. The inclusion bodies were isolated, solubilized, and the neurturin protein was re-folded and purified by ion exchange and hydrophobic interaction chromatography as described previously (7). [ 125 I]NTN (ϳ1800 Ci/mmol) was prepared using the purified E. coli-expressed protein by Amersham (Arlington Heights, IL; custom iodination, catalog number IMQ1057). Recombinant human GDNF was also radioiodinated by Amersham to a similar specific activity (22).
The isolated human GFR␣-2 cDNA was subcloned into an eukaryotic expression plasmid, pBK RSV (Stratagene) to generate the GFR␣-2 expression vector pHGLRSV. 2 NNR-9, a cell line expressing GFR␣-2, was derived from Neuro-2a cells (ATCC catalog number CCL 131) transfected with pHGLRSV. The transfection was accomplished by using the calcium phosphate transfection system (Life Technologies, Inc.) according to the manufacturer's directions. Transfected cells were selected for expression of the plasmid by growth in 400 g/ml G418 antibiotic (Sigma). G418-resistant clones were expanded and assayed for GFR␣-2 expression by Northern blot using the GFR␣-2 cDNA as probe. Expression of GFR␣-2 in individual clones was confirmed by binding to [ 125 I]NTN.
Binding of [ 125 I]NTN and [ 125 I]GDNF to NNR-9 and NGR-38 cells were carried out as described previously (40). Briefly, cells were seeded 1 day before the assay in 24-well Costar tissue culture plates precoated with polyornithine at a density of 3 ϫ 10 4 cells/cm 2 . Cells were placed on ice for 5-10 min, washed once with ice-cold buffer (Dulbecco's modified Eagle's medium containing 25 mM HEPES, pH 7.0), and incubated at 4°C in 0.2 ml of binding buffer (washing buffer containing 2 mg/ml bovine serum albumin) containing 50 pM [ 125 I]NTN or [ 125 I]GDNF in the absence or presence of 500 nM unlabeled ligand for 4 h. Cells were washed 4 times with 0.5 ml of ice-cold washing buffer and lysed with 0.5 ml of 1 M NaOH. The lysates were counted in a 1470 Wizard Automatic Gamma Counter (Wallac Inc., Gaithersburg, MD).
Chemical Cross-linking-The coding regions of the first 455 amino acids of the human GFR␣-1 and the first 451 residues of human GFR␣-2 cDNAs were fused in-frame with a DNA fragment encoding the Fc region of human IgG 1 tagged with 6 histidine residues at the carboxyl terminus (41). These constructs were then inserted into the expression vector pBK RSV as described previously (22). The GFR␣-1/Fc and GFR␣-2/Fc fusion constructs were transfected into 293T cells and con- Immunoblotting Analysis-Ret autophosphorylation was examined by immunoblot analysis as described previously (22). Briefly, cells were seeded 24 h prior to the assay in 6-well tissue culture dishes at a density of 1.5 ϫ 10 6 cells/well. Cells were washed once with binding buffer and treated with various concentrations of neurturin or GDNF (0.5 pM to 50 nM) in binding buffer at 37°C for 10 min. Treated cells and untreated controls were lysed in Triton X-100 lysis buffer (50 mM HEPES, pH 7.5, 1% Triton X-100, 50 mM NaCl, 50 mM sodium fluoride, 10 mM sodium pyrophosphate, 1% aprotinin (Sigma, catalog number A-6279), 1 mM phenylmethylsulfonyl fluoride (Sigma, catalog number P-7626), 0.5 mM Na 3 VO 4 (Fisher catalog number S454-50) and immunoprecipitated with an anti-Ret antibody (Santa Cruz Biotechnology) and protein-A Sepharose as described (22). Immunoprecipitates were fractionated by 7.5% SDS-PAGE and transferred to nitrocellulose membranes as described by Harlow and Lane (42). The membranes were blocked with 5% bovine serum albumin (Sigma) and tyrosine phosphorylation of the Ret receptor was detected by probing with an anti-phosphotyrosine monoclonal antibody 4G10 (UBI, catalog number 05-321) at room temperature for 2 h. The amount of Ret protein in each lane was determined by stripping and re-probing the same membrane with the anti-Ret antibody. Detection was accomplished using sheep anti-mouse secondary antibody or protein-A conjugated to horseradish peroxidase (Amersham, catalog number NA931) in conjunction with chemiluminescence reagents (ECL, Amersham) following the manufacturer's instructions.
Blot Hybridization Analysis-For blot hybridization analysis, the cloned rat GFR␣-1, GFR␣-2, and GFR␣-3 cDNA was labeled using the Random Primed DNA Labeling Kit (Boehringer Mannheim, Indianapolis, IN) according to the manufacturer's instructions. Rat and mouse RNA blots purchased from CLONTECH were hybridized with the probe and washed at high stringency using the reagents of the ExpressHyb Kit (CLONTECH) according to the manufacturer's instructions. Following exposure on x-ray film, the filters were stripped of probe by boiling in 0.5% SDS for 10 min and rehybridized with a ␤-actin probe (CLON-TECH) as a control for total RNA loading.

RESULTS
Cloning and Sequence Analysis of GFR␣-2 and GFR␣-3-A human expressed sequence tag (EST) with significant homology to GFR␣-1 was found by a search of the publicly available nucleic acid sequence data bases. 3 Oligonucleotides corresponding to the ends of this EST were synthesized and used in a reverse transcription-polymerase chain reaction (RT-PCR) with human fetal brain mRNA as the template. A fragment of the expected length was isolated, labeled, and used as a hybridization probe to screen a human fetal brain cDNA library. The longest clone isolated in this manner was sequenced and found to contain an open reading frame coding for a 464-amino acid protein related in sequence to GFR␣-1. We have named this protein GDNF Family Receptor ␣-2 (GFR␣-2). The oligonucleotides described above were also used to screen pools from a rat photoreceptor cDNA library (22) by PCR and a product of the expected length was obtained from a single pool. An indi-vidual cDNA clone from this pool was identified by hybridization to the radiolabeled human GFR␣-1 PCR product and sequenced. This clone contained an open reading frame coding for a 460-amino acid peptide that is nearly identical to human GFR␣-2 and almost certainly represents its rat ortholog.
Publicly available sequence data bases were searched using GFR␣-1 and GFR␣-2 as query sequences and a short EST with homology to both GFR␣-1 and GFR␣-2 was found. 4 Oligonucleotides corresponding to the ends of this EST were used as primers in RT-PCR with total rat embryo RNA as the template. A 225-nucleotide fragment was amplified, cloned into a plasmid vector, and sequenced to verify that it corresponded to the original GFR␣-1/GFR␣-2-related EST. Plasmid DNAs isolated GFR␣-1, GFR␣-2, and GFR␣-3 amino acid sequence comparison. The amino acid sequences of the human GFR␣-1, GFR␣-2, and GFR␣-3 are aligned and a consensus sequence is shown above the three receptor sequences. Uppercase letters in the consensus sequence indicate amino acids that are conserved in all three receptors, lowercase letters indicate that two of the three receptors share that amino acid, and dots indicate all three receptors have a different amino acid at that position. Predicted signal peptide sequences and the hydrophobic COOH-terminal regions of all three receptors are underlined. Conserved cysteines are highlighted. Potential N-glycosylation sites are shown in boldface and are outlined by boxes if conserved between two receptors. from pools of an E15 rat embryo cDNA library were screened by PCR with the same oligonucleotides and a single positive pool was found. Clones from this pool were screened by hybridization to the radiolabeled 225-nucleotide PCR fragment to isolate a single positive clone. Sequence analysis of the 1.8-kb insert from this clone revealed an open reading frame coding for a 397-amino acid peptide related to both GFR␣-1 and GFR␣-2. We have designated this putative receptor GFR␣-3.
An alignment of the amino acid sequences of rat GFR␣-1, GFR␣-2, and GFR␣-3 is shown in Fig. 1. The overall amino acid sequence identity among the three receptors is approximately 30 -50%. GFR␣-1 and GFR␣-2 are somewhat more closely related to each other (48% identity) than they are to GFR-␣3 (35% and 33% identity, respectively). Hydrophobic regions are found at both the amino and carboxyl termini of all three molecules (underlined, Fig. 1) and the amino-terminal hydrophobic regions have the characteristics expected for signal peptide sequences (43). The carboxyl-terminal hydrophobic region of GFR␣-1 is known to be involved in glycosyl phosphatidylinositol linkage to the cell membrane (22,23), and it is likely that the corresponding regions in GFR␣-2 and GFR␣-3 serve the same purpose. The most striking feature of the sequence alignment is the conservation of 28 cysteine residues among all three receptors (highlighted, Fig. 1), indicating that these proteins probably have similar three-dimensional structures. Several potential N-glycosylation sites are present in the GFR␣s (shown in boldface, Fig. 1), but none are found at the same position in all three receptors. GFR␣-1 and GFR␣-2 share sites at positions 365 and 427 that are not found in GFR␣-3, and GFR␣-2 shares a possible site with GFR␣-3 at positions 322-323 (Fig. 1).
Cross-linking of Neurturin and GDNF to GFR␣-1 and GFR␣-2-Our binding experiments suggest that both neurturin and GDNF interact with GFR␣-1 and GFR␣-2. However, lack of an antibody specific for GFR␣-2 has hampered our effort to further study these interactions. To overcome this difficulty, we generated plasmids that transiently express GFR␣-1/Fc and GFR␣-2/Fc fusion proteins when transfected into 293T cells. Conditioned medium (CM) containing either GFR␣-1/Fc or GFR␣-2/Fc fusion proteins was incubated with [ 125 I]NTN or [ 125 I]GDNF, chemically cross-linked, and then precipitated directly using Protein A-Sepharose beads. The immunoprecipitates were then analyzed by reducing SDS-PAGE (Fig. 3). When either [ 125 I]GDNF or [ 125 I]NTN was incubated with media containing GFR␣-1/Fc, broad major bands centered at ϳ110 and ϳ220 kDa were observed (Fig. 3). Similar species were observed when either radiolabeled ligand was incubated with media containing GFR␣-2/Fc (Fig. 3). The ϳ110-kDa bands correspond to the expected molecular weights of monomeric GFR␣-1/Fc or GFR␣-2/Fc associated with either [ 125 I]GDNF or [ 125 I]NTN. The ϳ220-kDa species probably represent a dimeric GFR␣-1/Fc or GFR␣-2/Fc complexed to either [ 125 I]GDNF or [ 125 I]NTN. When CM from mock transfected cells were used, no cross-linked band was precipitated by Protein A-Sepharose (data not shown). In all cases, addition of the corresponding unlabeled ligand to each sample resulted in a reduction in intensity of the major cross-linked bands (Fig. 3).
Neurturin Induces Ret Autophosphorylation in Cells That Express GFR␣-1-The ability of neurturin to associate with GFR␣-1 suggests that neurturin, like GDNF, might be able to activate Ret through GFR␣-1. To examine this possibility, we tested the ability of neurturin to induce Ret autophosphorylation in the GFR␣-1 expressing NGR-38 cells (22). NGR-38 cells were treated with various concentrations of neurturin up to 50 nM, lysed, and the lysates immunoprecipitated with anti-Ret antibody. The immunoprecipitates were analyzed by SDS-PAGE followed by immunoblotting using an anti-phosphotyrosine antibody. A 170-kDa band, corresponding to phosphorylated Ret was observed (Fig. 4, lanes 8 -14 from left) with clear stimulation of Ret phosphorylation detectable at 500 pM neurturin and increasing with higher doses (Fig. 4). In a parallel experiment using GDNF in place of neurturin, an increase in the level of phosphorylation of the 170-kDa Ret band over background could be seen at a GDNF concentration of 5 pM (Fig. 4, lanes 1-7 from left). When the filters were stripped and re-probed with the anti-Ret antibody, the 170-kDa Ret protein band appeared in all lanes with approximately equal intensity (data not shown). Treatment of the parental Neuro-2a cells with even high doses of neurturin or GDNF produced minimal phosphorylation of the 170-kDa band.
Neurturin and GDNF Induce Ret Autophosphorylation in Cells That Express GFR␣-2-We have shown that both neur-turin and GDNF bind to GFR␣-2, and that the Ret PTK can be activated by either neurturin or GDNF through GFR␣-1. These observations suggest that GFR␣-2 may also be able to mediate neurturin and/or GDNF activation of Ret. To assess this possibility, Neuro-2a-derived NNR-9 cells expressing high levels of GFR␣-2 and Ret were treated with various concentrations of neurturin or GDNF and processed for immunoblotting as described above. In these cells, Ret autophosphorylation was stimulated by 0.5 pM neurturin, the lowest dose tested (Fig. 5). In contrast, the dose of GDNF required to produce a comparable stimulation of Ret phosphorylation was approximately 50 -500 pM (Fig. 5), about 100 -1000-fold higher than for neurturin.
Expression of GFR␣-1, GFR␣-2, and GFR␣-3 in Adult Rat-The expression of GFR␣-1, GFR␣-2, and GFR␣-3 mRNAs in adult rat tissues was examined by blot hybridization analysis. GFR␣-1 mRNA is widely expressed, with high levels found in lung, brain, liver, kidney, and spleen (Fig. 6A). Expression is also detectable in heart and among the tissues examined is absent only in muscle and testis. Two distinct size transcripts are observed and their relative amounts vary among the tissues. The 3.6-kb transcript is predominant in liver, lung, heart, and spleen while comparable amounts of the 3.6-and 8.5-kb transcripts are present in brain and kidney. The tissue distribution of GFR␣-2 mRNA is similar to that of GFR␣-1 (Fig. 6B). GFR␣-2 expression is highest in lung, spleen, and brain, with lesser amounts in kidney and heart. The most striking difference is the lack of GFR␣-2 expression in liver. The size of the GFR␣-2 transcripts is approximately 3.6 kb, similar to the smaller of the two GFR␣-1 transcripts. The expression of GFR␣-3 mRNA is highest in kidney and is notably absent in brain (Fig. 6C). Detectable expression of GFR␣-3 is also present in spleen, lung, liver, and heart. The transcript size for GFR␣-3 is somewhat smaller (ϳ2.1 kb) than that observed for GFR␣-1 and GFR␣-2.
Expression of GFR␣-1, GFR␣-2, and GFR␣-3 in Mouse Embryo-Developmental expression of GFR␣-1, GFR␣-2, and GFR␣-3 mRNA was examined in mouse on embryonic days 7, 11, 15, and 17. Expression of the 3.6-kb transcript of GFR␣-1 is first apparent at E11, seems to decrease somewhat at E15, but then increases dramatically by E17 (Fig. 7A). A minor amount of the 8.5-kb GFR␣-1 mRNA can be detected on E11, but no expression of this transcript is detected thereafter. The expression of the 3.6-kb GFR␣-2 transcript is barely detectable at E11, but increases gradually through E17 (Fig. 7B). Expression of the 2.1-kb GFR␣-3 mRNA is not detected at E7, but is quite strong by E11 (Fig. 7C). After E11, expression decreases and remains constant from E15-E17. DISCUSSION The classic mechanism of signal transduction for receptor PTKs involves a direct interaction of the PTK with its ligand that results in receptor dimerization and activation (45). A novel alternative to this process was revealed by the characterization of the receptor for GDNF (22,23). GDNF induces signaling through the Ret PTK by first binding to GFR␣-1, a glycosyl phosphatidylinositol-linked cell surface molecule lacking a transmembrane domain. Only after binding to GFR␣-1 can GDNF interact with Ret and induce activation of the Ret kinase. Here we report the cloning of GFR␣-2, a receptor closely related to GFR␣-1. The 48% amino acid identity between GFR␣-1 and GFR␣-2, along with the nearly complete conservation of cysteine residues suggests that these two molecules have very similar three-dimensional structures and are likely to have similar biological functions.
We have demonstrated that both neurturin and GDNF can bind to either GFR␣-1 or GFR␣-2. Furthermore, binding of GDNF or neurturin to either receptor results in association of the ligand with Ret and consequent activation of the Ret PTK. Although both ligands bind both receptors, the evidence indicates that GDNF is the preferred ligand for GFR␣-1 and that neurturin is preferred by GFR␣-2. GFR␣-1-expressing NGR-38 cells are able to bind more GDNF than neurturin while the binding preference is reversed for GFR␣-2-expressing NNR-9 cells (Fig. 2). Consistent with these results, GDNF cross-links more effectively to GFR␣-1/Fc fusion proteins than to GFR␣-2/Fc fusions, while neurturin's cross-linking preference is reversed (Fig. 3). The relative abilities of GDNF and neurturin to stimulate Ret phosphorylation in conjunction with cells expressing either GFR␣-1 or GFR␣-2 is consistent with both the binding and cross-linking data. In the GFR␣-1 expressing NGR-38 cells, the concentration of GDNF required to stimulate Ret autophosphorylation is approximately 100-fold lower than that observed for neurturin (Fig. 4). Conversely, in the GFR␣-2 expressing NNR-9 cells, neurturin stimulates Ret autophosphorylation at 100-fold lower concentrations than those required by GDNF (Fig. 5). Taken together, the binding, crosslinking, and phosphorylation data described above strongly suggest that although there is some cross-talk, GFR␣-1 is the primary receptor for GDNF while GFR␣-2 is the primary receptor for neurturin. We may also speculate that GFR␣-3 has an as yet undiscovered cognate ligand of its own.
It is interesting that the binding of [ 125 I]GDNF to both GFR␣-1 and GFR␣-2 can be replaced by both unlabeled GDNF and neurturin, but that of [ 125 I]NTN can only be inhibited by unlabeled neurturin (Fig. 2). A possible explanation for this result is the existence of two distinct sites: a neurturin-binding site that does not bind GDNF very effectively and a GDNFbinding site that can harbor either GDNF or neurturin. Human GDNF has a 33-amino acid highly basic motif at its NH 2 terminus that is absent in human neurturin (16). This domain makes the GDNF molecule larger and more positively charged than neurturin and could prevent GDNF from competing for a smaller, less negatively charged neurturin-binding site. However, neurturin might still be able to compete for a larger, less restrictive GDNF site. If this model is correct, truncated forms of GDNF may be able to effectively compete with neurturin for its binding site.
We have presented evidence that both GFR␣-1 and GFR␣-2 utilize the Ret PTK to effect transmembrane signaling. The question of whether other signaling components, possibly Retlike molecules, might interact with the GFR␣s is still open. If all GDNF-like ligands signal exclusively through Ret, the same intracellular pathways should be triggered by all members of the family. While the tissue-specific distribution and the developmental expression of the GFR␣s may determine where and when each ligand should be in action, the final biological readout should not vary. The near identity of the phenotypes resulting from the targeted disruption of the GDNF and Ret genes implies that if other signaling partners for GDNF exist, they are not important during development of these tissues. Further comparison of the biological activities of GDNF and neurturin should provide clues regarding the existence of alternate signaling partners for the GFR␣s.
The properties of GFR␣-1 and GFR␣-2 suggest that they define a new family of ligand binding receptors (the GFR␣s) for GDNF-like molecules. We have also described the isolation of a cDNA clone, GFR␣-3, that has sequence homology to both GFR␣-1 and GFR␣-2 and probably represents a third member of the GFR␣ family. It is not clear whether or not GFR␣-3 as well as any other yet to be discovered members of the GFR␣ receptor family will be capable of binding GDNF and/or neurturin and mediating phosphorylation of Ret. We can speculate, however, that additional members of the GDNF-like family of ligands exist that may bind the GFR␣s. Cross-talk among ligands and receptors has been seen in other families of receptor PTKs, such as the Trk, epidermal growth factor, and fibroblast growth factor families. Single receptors have also been reported to be shared equally by two ligands, such as TrkB by BDNF and NT-4/5. Our data suggest that this is also a feature of the GDNF ligand family and the GFR␣ receptor family of molecules. Such cross-talk among receptors and ligands, however, makes it very difficult to determine the physiological role of each ligand or receptor. Further elucidation of the patterns of expression of the GFR␣ family receptors will help us to advance our knowledge in this field. Targeted disruption of the genes for the GFR␣s and GDNF-like molecules will provide more valuable insights into the biological roles of this receptor ligand family.