Binding of GDNF and Neurturin to Human GDNF Family Receptor a 1 and 2 INFLUENCE OF cRET AND COOPERATIVE INTERACTIONS*

The members of the glial cell line-derived neurotrophic factor (GDNF) family signal via binding to the glycosyl phosphatidylinositol-anchored membrane proteins, the GDNF family receptors a (GFR a ), and activation of cRET. We performed a detailed analysis of the binding of GDNF and neurturin to their receptors and investigated the influence of cRET on the binding affinities. We show that the rate of dissociation of 125 I-GDNF from GFR a 1 is increased in the presence of 50 n M GDNF, an effect that can be explained by the occurrence of negative cooperativity. Scatchard plots of the ligand concentration binding isotherms reveal a pronounced downward curvature at low 125 I-GDNF concentrations suggesting the presence of positive cooperativity. This effect is observed in the range of GDNF concentrations responsible for biological activity (1–20 p M ) and may have an important role in cRET-independent signaling. A high affinity site with a K D of 11 p M for 125 I-GDNF is detected only when GFR a 1 is co-expressed with cRET at a DNA ratio of 1:3. These results suggest an interaction of GFR a 1 and cRET in the absence of GDNF and demonstrate that the high 48 curve fitting of the Langmuir binding isotherm with the mass transfer to the primary data. Data Calculation— Ligand concentration binding isotherms and com- petition binding isotherms were analyzed via nonlinear regression analysis using Prism 2.01 (GraphPad Prism Software, San Diego, CA) to determine the maximal number of binding sites ( B max ) and the apparent equilibrium dissociation constant ( K D ) or the IC 50 values. Assuming simple competition, the competition data were fitted according to both one- and two-site mass action binding models, and the better model was determined by an extra-sum-of-squares test using Prism. Data from dissociation and association kinetic experiments were ana- lyzed by Prism according to both monoexponential and biexponential models to estimate k d and k ob , the radioligand dissociation and associ- ation apparent rate constants.

The members of the glial cell line-derived neurotrophic factor (GDNF) family signal via binding to the glycosyl phosphatidylinositol-anchored membrane proteins, the GDNF family receptors ␣ (GFR␣), and activation of cRET. We performed a detailed analysis of the binding of GDNF and neurturin to their receptors and investigated the influence of cRET on the binding affinities. We show that the rate of dissociation of 125 I-GDNF from GFR␣1 is increased in the presence of 50 nM GDNF, an effect that can be explained by the occurrence of negative cooperativity. Scatchard plots of the ligand concentration binding isotherms reveal a pronounced downward curvature at low 125 I-GDNF concentrations suggesting the presence of positive cooperativity. This effect is observed in the range of GDNF concentrations responsible for biological activity (1-20 pM) and may have an important role in cRET-independent signaling. A high affinity site with a K D of 11 pM for 125 I-GDNF is detected only when GFR␣1 is co-expressed with cRET at a DNA ratio of 1:3. These results suggest an interaction of GFR␣1 and cRET in the absence of GDNF and demonstrate that the high affinity binding can be measured only when cRET is present.
Glial cell line-derived neurotrophic factor (GDNF), 1 neurturin (NTN), persephin, and artemin are members of the transforming growth factor ␤ superfamily. Their neurotrophic activity has been described in neuronal populations from the central and peripheral nervous systems (1)(2)(3). These neurotrophic factors bind to members of the family of the glycosyl phosphatidylinositol-anchored membrane protein, GDNF family receptor ␣ (GFR␣) (4). This GFR␣-ligand complex, together with the tyrosine kinase receptor (cRET) forms a functional receptor that activates downstream signal transduction pathways (5)(6)(7)(8). The binding of these neurotrophic factors to GFR␣1-4 receptors and activation of cRET has been investigated by a number of groups. These investigations revealed different bind-ing affinities of the natural ligands toward the different GFR␣ receptors (9 -11).
GFR␣1 was initially described as a high affinity receptor for GDNF (9,10). Jing et al. (9) described two affinity sites of 125 I-GDNF binding to the GFR␣1 receptor with K D values of 2.3 pM for the high and 170 pM for the low affinity site. Co-expression of GFR␣1 and cRET in Neuro-2a cells had little influence on the two affinity sites suggesting a small or no effect of cRET on equilibrium binding. Using a cell-free binding assay, Klein et al. (11) observed one high affinity site with a K D of 3 pM. Another single binding site (K D ϭ 63 pM) was described by Treanor et al. (10) using CHO cells stably expressing GFR␣1. The binding of NTN to GFR␣1 has produced the following diverse results. Using Neuro-2a cells, binding of iodinated NTN to the GFR␣1 was demonstrated (12). However, using a cellfree system, Klein et al. (11) was unable to detect specific binding of iodinated NTN to GFR␣1, although NTN was displacing iodinated GDNF from GFR␣1 at concentrations higher than 1 nM. Similarly, in the functional assays, some groups could demonstrate cRET activation mediated by GFR␣1 and NTN in NIH cells (13,14), whereas Buj-Bello et al. (15) presented results suggesting that in neuronal cells, NTN is unable to stimulate cRET via GFR␣1. GFR␣2 was described as a high affinity receptor for NTN (K D ϭ 10 pM) and binding of 125 I-GDNF to this receptor was not observed (11). In contrast, using a cell-free binding assay, Sanicola et al. (16) described the binding of 125 I-GDNF to both GFR␣1 and GFR␣2, but binding to GFR␣2 was only detectable in the presence of cRET (17). Different groups (18 -22) have reported the cloning of the GFR␣3 receptor. The ligand for GFR␣3, artemin, was recently shown to bind to the GFR␣3 receptor and to be a survival factor for different sensory and sympathetic neurons (23). The latest member of this growing family is GFR␣4 (24). It was reported that persephin binds to GFR␣4 and that co-expression of GFR␣4 and cRET results in a receptor complex that can be activated by persephin and not by GDNF and NTN (25).
In this report we describe the first detailed characterization of 125 I-GDNF and 125 I-NTN binding to the GFR␣1 and GFR␣2 receptors. We have established conditions for the binding of GDNF and NTN to the full-length receptors expressed in Chinese hamster ovary (CHO) cells and to recombinant fusion proteins, thus permitting a detailed pharmacological characterization. This study demonstrates that the presence of cold GDNF influences the rate of dissociation of iodinated GDNF. Further, we provide evidence for positive and negative cooperative interactions and propose a binding model that includes the cooperativity and the influence of cRET on binding kinetics. Our investigations demonstrate that the presence of the high affinity site can only be detected if a sufficient amount of cRET is present.
Pioneer sensor chip F1, amine coupling kit, HEPES-buffered saline containing 3.5 mM EDTA and 0.05% P-20 buffer, and acetate buffer (pH 4.5) were from Biacore (Biacore AB, Uppsala, Sweden). Native Taq polymerase, PCR buffer with MgCl 2 , the Expand TM high fidelity PCR system, ampicillin, and all restriction enzymes used were from Roche Molecular Biochemicals. A 10 mM dNTP mixture was purchased from Life Technologies, Inc. The Qiagen plasmid mini-and maxi-DNA purification kits, the Qiaquick gel extraction kit, and the Qiaquick PCR purification kit were purchased from Qiagen GmbH (Dü sseldorf, Germany). All PCR reactions were performed in a GeneAmp PCR system 9600 cycler (Perkin-Elmer). A human substantia nigra cDNA library and placenta Marathon Ready TM cDNA were from CLONTECH Laboratories.
Cloning of Human GFR␣1 and GFR␣2 and Construction of GFR␣-IgG-Fc Fusion Vectors-The human GFR␣1 coding sequence was cloned from a human substantia nigra cDNA library. The translated cDNA sequence obtained is identical to protein sequences found by other groups (GenBank TM accession numbers AF038421 and AF042080).
The human GFR␣2 coding sequence was cloned using an EST clone detected as a GFR␣1 homologue in the EMBL data base (accession number H12981) in combination with 5Ј-rapid amplification of cDNA ends reactions on placenta Marathon Ready TM cDNA. The translated cDNA sequence of the full-length plasmid construct was identical to the GFR␣2 protein sequence present in the SwissProt data base (accession number O00451).
The cDNA regions of GFR␣1 and GFR␣2 (coding for amino acid residues 27-427 and 28 -371, respectively) excluding the sequences coding for the signal peptide and for the COOH-terminal hydrophobic region involved in glycosyl phosphatidylinositol-anchoring were amplified by PCR using primers incorporating HindIII and BamHI restriction sites at the 5Ј-and 3Ј-ends, respectively. The resulting products were cloned in-frame in the expression vector Signal pIg plus. The inserts of all constructs were confirmed by sequence analysis. The resulting proteins expressed from these constructs contain a 17-amino acid NH 2 -terminal CD33 signal peptide, the respective GFR␣ protein region, and a 243-amino acid COOH-terminal human IgG 1 -Fc fusion domain.
Cloning of Synthetic Human Neurturin-Because of the highly GCrich nature of the human NTN DNA sequence (Accession no. U78110) a synthetic gene was prepared covering the coding sequence of mature human NTN using the codon preferences of highly expressed enteric bacteria genes. Primer set pairs were denatured to 90°C in a heating block and slowly cooled to room temperature at a concentration of 0.5 g/ml. The resulting annealed double-stranded fragments were ligated together and amplified by PCR using the Expand High Fidelity TM PCR system according to the manufacturer instructions. The resulting PCR product was cloned into pRSETb, which added a 6-His tag at the amino terminus of the mature NTN protein. NTN expression and purification were performed as described previously (14).
Expression of GFR␣1-2 and Fusion Proteins in CHO Cells-CHO cells were routinely cultured in Dulbecco's modified Eagle's medium/ F12 medium supplemented with 10% heat-inactivated fetal calf serum. Cells were transfected with GFR␣1 cloned into pcDNA 3 using a calcium phosphate method. All other transfections with GFR␣1-2 in pcDNA 3 , cRET in pIREShygro, and GFR␣1-2 in Signal plg plus were performed using an optimized LipofectAMINE PLUS method. For this, a total amount of 6.5 g of DNA was incubated with 17.5 l of PLUS reagent in 750 l of serum-free medium for 15 min at room temperature. LipofectAMINE was diluted 50-fold into serum-free culture medium; 750 l of this mixture was added to the DNA solution. Following a 15-min incubation at room temperature, 3.5 ml of serum-free medium was added, and the mixture was brought onto the cells (in a 100-mm Petri dish). The cells were incubated for 3 h at 37°C in 5% CO 2 , after which a 5-ml culture medium, containing 20% heat-inactivated fetal calf serum, was added. 24 h later, the medium was changed into regular culture medium. Transfection efficiencies using these optimized conditions were typically ϳ50 -60%. For permanent transfections the selection medium contained either 800 g of G418 or 800 g of G418 and 800 g of hygromycin. Antibiotic-resistant clones were expanded and assayed for GFR␣1 expression by binding to 125 I-GDNF. A CHO clone permanently expressing GFR␣1 was used for generation of a GFR␣1/ cRET-expressing clone. Clones were assayed for cRET expression or GFR␣1-2-Fc expression using specific antibodies. For some experiments, transiently transfected cells were used 48 h following a transfection.
For purification of GFR␣-Fc fusion proteins, CHO cells permanently expressing fusion proteins were grown in serum-free medium (HyQ CCM5) that was collected every 3 days. Medium was centrifuged and applied to a recombinant protein A column. Bound protein was eluted with 0.1 M sodium citrate, pH 3.0, and collected into 1 M Tris buffer, pH 8.4 (dilution ratio 1:6). Protein concentration was estimated by absorbance at 280 nm using an extinction coefficient of 1.5.
Radioligand, Cell-based Binding Assays-For cell suspension binding experiments, CHO cells were collected, counted, and transferred to prewetted Multiscreen filter plates (MW 96, Millipore, 0.65 m). Routinely, 100,000 -120,000 cells were used/well. Plates were kept on ice. Cells were incubated with 0.25 ml of binding buffer (Dulbecco's modified Eagle's medium containing 25 mM HEPES and 2 mg/ml bovine serum albumin, pH 7.4) containing 125 I-GDNF or 125 I-NTN with or without cold ligands for 2 h at 4°C. Following the 2-h incubation, plates were transferred on a Multiscreen vacuum filtration manifold, and supernatants were filtered. Cells were washed two times with washing buffer (50 mM Tris-HCl, 120 mM NaCl, pH 7.4), and filters were punched out and counted in a ␥ counter. In some experiments transiently transfected cells were used. Following transient transfection, binding was performed 48 h later as described above.
Binding Kinetics-The rate of association of GDNF to GFR␣1 or GFR␣1/cRET was determined using transiently transfected cells (GFR␣1/cRET DNA transfection ratio 1:1). Whole cells were incubated at 4°C with 50 pM 125 I-GDNF in the presence of a 1000-fold excess of cold GDNF. At different time points after the addition of 125 I-GDNF, samples were filtered on a Multiscreen vacuum filtration manifold. The rate of dissociation was determined at 4°C in two different ways, by adding excess cold ligand and by a large dilution of assay medium. For each method, permanently transfected cells were incubated with 50 pM 125 I-GDNF for 2 h to achieve binding equilibrium. To initiate displacement, a 1000-fold excess of cold GDNF was added at different time points, and samples were filtered on a Multiscreen vacuum filtration manifold. For the dilution method, cells preincubated for 2 h were diluted 100-fold with cold incubation medium and filtered using a single filter filtration flask and 25-mm Durapore filters, 0.65 m at different time points after dilution.
Radioligand Binding, Scintillation Proximity Assay (SPA)-Binding was performed in MW 96 plates. GFR␣-Fc fusion proteins at a concentration of 100 -300 ng/ml were precoated to protein A SPA beads (0.5 mg/ml) overnight. Following precoating, beads were centrifuged, the supernatant was removed, and beads were resuspended in 50 mM Tris, pH 7.4. Precoated SPA beads were then incubated with 125 I-GDNF or 125 I-NTN in the presence or absence of cold ligands. After a 24-h incubation at room temperature, plates were counted in a MicroBeta Trilux scintillation counter (Wallac).
Surface Plasmon Resonance Analyses-Surface plasmon resonance (SPR) analyses were performed at 25°C with a BIAcore 2000 biosensor. The carboxylated matrix of a F sensor chip was first activated with a 1:1 mixture of 400 mM N-ethyl-N-(dimethylaminepropyl)-carbodiimide and 100 mM N-hydroxy-succinimide. Then, recombinant GDNF or NTN was applied on the activated surface in HEPES-buffered saline (150 mM NaCl, 3.5 mM EDTA, 0.05% P-20, 10 mM HEPES, pH 7.4). Unoccupied reactive groups were deactivated with 1 M ethanolamine HCl. For kinetic experiments, soluble fusion proteins were applied at concentrations of 0.8 -25 nM for GFR␣1-Fc and 1.5-100 nM for GFR␣2-Fc in HEPES-buffered saline containing 3.5 mM EDTA and 0.05% P-20 buffer (150 mM NaCl, 3.5 mM EDTA, 0.05% P-20, and 10 mM HEPES, pH 7.4). GFR␣1-Fc and GFR␣2-Fc were perfused over the immobilized ligands at a flow rate of 10 l/min for 2 min. The dissociation was monitored for 3 min followed by regeneration with 5 mM NaOH. The BIAcore evaluation software, 3.0.1 was used to calculate the association rate (k a ), dissociation rate (k d ), and the apparent equilibrium dissociation constant (K D ). Data on equilibrium binding were analyzed by nonlinear curve fitting of the Langmuir binding isotherm with the mass transfer to the primary data.
Data Calculation-Ligand concentration binding isotherms and competition binding isotherms were analyzed via nonlinear regression analysis using Prism 2.01 (GraphPad Prism Software, San Diego, CA) to determine the maximal number of binding sites (B max ) and the apparent equilibrium dissociation constant (K D ) or the IC 50 values. Assuming simple competition, the competition data were fitted according to both one-and two-site mass action binding models, and the better model was determined by an extra-sum-of-squares test using Prism. Data from dissociation and association kinetic experiments were analyzed by Prism according to both monoexponential and biexponential models to estimate k d and k ob , the radioligand dissociation and association apparent rate constants.

RESULTS
Binding Kinetics of GDNF and NTN Binding to GFR␣1 and GFR␣2 Receptors Assayed in Whole Cells-The association and dissociation rates of the GDNF-receptor and NTN-receptor complexes were studied using CHO cells and a filtration assay at 4°C. The association curves are shown in Fig. 1, A-B. The observed rate of association (k ob ) of 125 I-GDNF with GFR␣1 receptors transiently expressed in CHO cells was 8.9 ϫ 10 Ϫ4 s Ϫ1 (Table I). A similar observed rate of association of GDNF was obtained with GFR␣1/cRET receptors. The DNA ratio used for transient expression of GFR␣1 and cRET was 1:1. The association rate of NTN with GFR␣2 and NTN with GFR␣2/ cRET were in the same range (Table I). To study dissociation, intact CHO cells permanently expressing GFR␣1 and GFR␣2 with or without cRET were incubated with a radiolabeled ligand for 2 h at 4°C, and dissociation was induced either by an addition of excess of cold ligand or by dilution. The dissociation curves are shown in Fig. 1, C-F. Dissociation, induced by the addition of unlabeled GDNF (50 nM) was very rapid. The derived rate of dissociation (k d ) of 125 I-GDNF from GFR␣1 receptor was 9.1 ϫ 10 Ϫ3 s Ϫ1 (Table I). In contrast, the rate of dissociation initiated by a 100-fold dilution was 10 times slower. The rate of dissociation obtained with the GFR␣1-cRET receptor complex was in the same range. The dissociation constants obtained with 125 I-NTN and GFR␣2 receptors, measured by the addition of unlabeled NTN (50 nM) were much slower in comparison to 125 I-GDNF dissociation from GFR␣1 receptors (Table I). When the rate of dissociation was determined by dilution, it was also very slow, with a t1 ⁄2 of more than 130 min for both GFR␣2 and GFR␣2-cRET receptor complexes (Table I).
Equilibrium Binding Studies-Equilibrium saturation isotherms of 125 I-GDNF binding to cells expressing either GFR␣1 alone or with cRET were determined using intact cells in suspension at 4°C. Equilibrium saturation binding isotherms are shown in Figs. 2 and 3. When GFR␣1 was expressed alone either transiently or as a permanent clone, only a low affinity site was observed (Table II), with apparent K D values of 613 and 317 pM, respectively. The binding levels (B max values) were 571 and 273 fmol/mg protein, respectively. A higher level of binding was obtained with cells permanently expressing GFR␣1 and cRET (B max ϭ 981 fmol/mg protein) with a K D of 356 pM ( Fig. 3B and Table II). When GFR␣1 was transiently co-expressed with cRET at DNA ratios GFR␣1/cRET 3:1 and 1:3, the affinity was increasing with an increased amount of cRET cDNA (Fig. 2, B-D), whereas the B max declined (Table II). Although no high affinity site with a K D below 20 pM was observed, Scatchard analysis of the binding in the permanently and transiently expressing cell lines showed curvilinear plots at concentrations below 10 pM. These results suggest the presence of positively cooperative interactions at low concentrations of the ligand. Only when the DNA ratio in transiently transfected cells was 1:3, was there an indication of the presence of a small percentage of high affinity sites (Fig. 2D). To confirm the presence of the high affinity site with this DNA ratio, a detailed analysis of binding was performed at concentrations of 125 I-GDNF below 20 pM (Fig. 2, E-F). With a GFR␣1/ cRET DNA ratio of 1:3, the high affinity site with a K D of 11 pM and a B max of 12 fmol/mg protein was revealed. In contrast, a DNA ratio of 3:1 did not result in a saturable binding at these low concentrations of iodinated GDNF (result not shown). These data clearly demonstrate the presence of the high affinity site when cRET is expressed in a sufficient amount. Only when a sufficient number of cRET receptors is present, interaction of cRET and GFR␣1 results in the presentation of a small number of GFR␣1-cRET receptor complex in a high affinity conformation. Control experiments with varying amounts of GFR␣1 alone (1-6.5 g) did not result in a saturable binding at low concentrations of iodinated GDNF, suggesting that changes in GFR␣1 protein levels are not responsible for the presence of the high affinity site. Competition Binding Studies on GFR␣-expressing Cells-Initial equilibrium binding studies using CHO cells transiently expressing GFR␣1 or GFR␣2 demonstrated that 125 I-GDNF binds specifically to GFR␣1 only and that 125 I-NTN binds specifically only to GFR␣2. GDNF binding to GFR␣2 could not be detected even in the presence of cRET (DNA transfection ratio 1:1, data not shown). Fig. 4 shows the binding of 50 pM 125 I-GDNF to CHO cells that permanently express GFR␣1 or GFR␣1/cRET. The binding was inhibited concentration dependently in the range of 0.1 to 100 nM by unlabeled GDNF or NTN. The IC 50 values obtained with cold GDNF and NTN are summarized in Table III. From the inhibition binding curves it can also be seen that the binding of 125 I-GDNF was enhanced for up to 20 -25% as compared with radioligand alone, in the presence of either cold GDNF or cold NTN in a concentration range of 1-100 pM. To investigate the influence of cRET in competition binding studies, CHO cells were transiently transfected with different DNA ratios of GFR␣1 and cRET. Fig. 5 shows the effect of cRET on IC 50 values for GDNF, when co-expressed with GFR␣1 at different DNA ratios. The IC 50 values obtained with cold GDNF and NTN are summarized in Table III. When GFR␣1 was expressed alone or with cRET at a DNA ratio of 3:1, there was an increase of bound 125 I-GDNF in the presence of low concentrations of competitive GDNF. In addition, the co-expression of GFR␣1 with cRET at ratios 1:1 and 1:3 resulted in biphasic inhibition curves. The co-expression ratio of 1:1 revealed the presence of a high affinity site with an IC 50 of 10.6 Ϯ 2.1 pM and low affinity site with an IC 50

I-GDNF binding to GFR␣1 and GFR␣1/cRET receptors transiently expressed CHO cells.
Binding was performed on whole cells with a range of 125 I-GDNF concentrations for 2 h at 4°C as described under "Experimental Procedures." Nonspecific binding was determined in the presence of a 500-fold excess of cold GDNF. CHO cells were transiently transfected with different DNA ratios of GFR␣1 and cRET; GFR␣1/pcDNA3 1:1 (E) (A and B), GFR␣1/cRET 3:1 (q) (A and C), and GFR␣1/cRET 1:3 (f) (A, D-F). The results represent a typical saturation experiment (A) with its derived Scatchard plots (B-D). A separate saturation experiment in the low pM range with its derived Scatchard plot is shown in E and F. Each data point represents the mean of three values. Concentration binding isotherms were best fitted to a one site binding model using nonlinear regression analysis.

FIG. 3. Saturation binding isotherms and Scatchard plots of 125 I-GDNF binding to GFR␣1 and GFR␣1/cRET receptors using permanently transfected CHO cells.
Binding was performed on whole cells with a range of 125 I-GDNF concentrations for 2 h at 4°C as described under "Experimental Procedures." Nonspecific binding was determined in the presence of a 500-fold excess of cold GDNF. The data represent a typical saturation experiment (A) and derived Scatchard plots (B) with CHO cells permanently expressing GFR␣1 (q) and GFR␣1/cRET (E). Each data point represents the mean of three values. Concentration binding isotherms were best fitted to a one binding site model using a nonlinear regression analysis. Competition Studies on GFR␣-Fc Fusion Protein-coated SPA Beads-To further investigate the specificity of 125 I-GDNF and 125 I-NTN binding to GFR␣1 and GFR␣2 receptors, respectively, we used purified GFR␣-Fc fusion proteins coated to SPA beads. Using GFR␣-Fc-coated SPA beads, we detected specific binding of 50 pM 125 I-GDNF to both GFR␣1 and GFR␣2 (Fig. 6). The IC 50 values obtained with cold GDNF and NTN are summarized in Table IV. Specific binding of 125 I-NTN was also detected to both GFR␣1 and GFR␣2 with an important difference ( Fig. 7 and Table IV). Binding of 50 pM 125 I-NTN to GFR␣1 was significantly lower in comparison to GFR␣2. For identical amounts of fusion proteins coated, the difference in specific 125 I-NTN binding was 3-fold. The addition of cold NTN above 100 nM, increased binding of labeled NTN to both GFR␣1 and GFR␣2. In comparison to NTN, GDNF was only able to displace partially 125 I-NTN binding to GFR␣2. Already at concentrations above 10 nM, GDNF increased binding of labeled NTN to GFR␣1 and GFR␣2 receptors.
Surface Plasmon Resonance Binding Studies-Additional binding experiments were performed using a surface plasmon resonance technique. In these experiments a BIAcore F1 chip was coated with GDNF and NTN and superfused with soluble forms of GFR␣1-Fc and GFR␣2-Fc fusion proteins. Table V shows the summary of the derived apparent binding constants. In the SPR experiment where dissociation is initiated by dilution, the rate of dissociation of 125 I-GDNF was very similar to the filtration assay with a dilution initiated dissociation FIG. 4. Inhibition of 125 I-GDNF binding to membrane-anchored GFR␣1 and GFR␣1/cRET receptors using whole cells. Binding was performed on whole cells using 50 pM 125 I-GDNF and CHO cells permanently expressing GFR␣1 or GFR␣1/cRET as described under "Experimental Procedures." A, inhibition of 125 I-GDNF binding to GFR␣1-expressing CHO cells by cold GDNF (f) and NTN (Ⅺ). B, inhibition of 125 I-GDNF binding to GFR␣1/cRET-expressing CHO cells by cold GDNF (f) and NTN (Ⅺ). Each data point represents the mean of three to five independent determination Ϯ S.E. For each experiment the pIC 50 value was derived from curve fitting using a nonlinear regression analysis.  (t1 ⁄2 ϭ 950 s). It was found that the dissociation rate of 125 I-NTN determined by SPR was not in agreement with the rate constant obtained in a cell-based assay. It was much faster, with t1 ⁄2 of 770 s for GFR␣2. The reason for this discrepancy may be a higher sensitivity of NTN binding kinetics to the assay temperature. The association rates obtained with GFR␣1-Fc and GFR␣2-Fc fusion proteins and SPR were much faster in comparison to the cell-based assay, 1.0 ϫ 10 Ϫ2 (s Ϫ1 ) for GFR␣1 to GDNF and 1.8 ϫ 10 Ϫ2 (s Ϫ1 ) for GFR␣2 to NTN. The binding specificities of GFR␣1 and GFR␣2 for GDNF and NTN were also different from whole cell or SPA binding assays. We have found that soluble GFR␣1 binds to both GDNF and NTN with an apparent K D of 627 pM for GDNF and 1.0 nM for NTN. Soluble GFR␣2 binds only to NTN with an apparent K D of 0.9 nM. In contrast to the SPA assay where 125 I-GDNF binding to GFR␣2 was detected, significant binding of GFR␣2 to GDNF could not be detected with SPR. DISCUSSION The current view on the signal transduction mechanism of GDNF is based on an initial binding of GDNF to its receptor, GFR␣1. After the binding of GDNF to GFR␣1, this complex is able to interact with cRET and induces its activation (9, 10). More recently, it has been shown that GFR␣1 and GFR␣2 subunits are able to modulate cRET tyrosine phosphorylation in the absence of neurotrophic factors, thus suggesting that GFR␣ and cRET can interact without the prior binding of GDNF (17). Other authors have also demonstrated that GFR␣1 and cRET can interact in the absence of ligand (16,34). Our own data indicate that GFR␣1-2 receptors are able to decrease the constitutive levels of cRET phosphorylation in CHO cells (data not shown). Here we report a detailed binding characterization of GDNF and NTN to GFR␣1 and GFR␣2. Our results support a signal transduction model in which interactions of cRET and GFR␣ subunits affect the affinity for GDNF and therefore suggest complex formation prior to GDNF binding. Our data also demonstrate that the high affinity sites in the low pM range can be detected only in the presence of sufficient cRET.
The results in this study show complex binding characteristics of GDNF to GFR␣1. Initial experiments measuring binding kinetics showed that the dissociation rate of GDNF from GFR␣1 is accelerated 10 times in the presence of cold GDNF, which can be interpreted as an evidence for negative cooperativity by GDNF for its receptor (Fig. 1). Cold GDNF re-occupy receptor when labeled GDNF dissociates, thus an exchange of unlabeled for labeled GDNF at receptor binding sites occurs. Potential experimental artifacts such as rebinding of labeled GDNF after dilution or isotopic dilution of labeled GDNF by cold GDNF in an unstirred layer at the cell surface cannot be completely excluded (28). However, cRET co-expression with GFR␣1 at a DNA ratio of 1:1 did not affect this negative cooperativity. The presence of negative cooperative interactions have already been described for insulin binding to the insulin-receptor complex and also for NGF and BDNF binding to Trk receptors (26 -29). In contrast, the dissociation experiment with NTN and GFR␣2 showed very slow dissociation in the presence of 50 nM cold NTN. These results demonstrate a substantial difference in binding characteristics between GDNF-GFR␣1 and NTN-GFR␣2 complexes. The association rates of both GDNF and NTN were similar and were not influenced by the presence of cRET, when expressed transiently at a DNA ratio of 1:1.
To determine the binding affinity of GDNF to GFR␣1 and to investigate further the influence of cRET, we have used in our study a whole cell binding assay with CHO cells expressing permanently or transiently GFR␣1 with or without cRET. By using transient transfection, we were able to vary the relative ratio of GFR␣1 and cRET cDNAs and to compare results to permanent clones. Equilibrium saturation isotherms revealed the presence of a high affinity site in the range of 300 -600 pM in CHO cells permanently expressing GFR␣1 (Fig. 3). We have obtained a similar affinity using a permanent clone that ex-  7. Inhibition of 125 I-NTN binding to GFR␣1-Fc and GFR␣2-Fc fusion proteins adsorbed to SPA beads. Binding was performed on fusion protein-coated SPA beads using 50 pM 125 I-NTN as described under "Experimental Procedures." Inhibition of 125 I-NTN binding to GFR␣2-Fc by cold NTN is shown in A and by cold GDNF in B. Inhibition of 125 I-NTN binding to GFR␣1-Fc by cold NTN is shown in C and by cold GDNF in D. Each data point represents the mean of five to six independent determinations Ϯ S.E. For each experiment the pIC 50 value was derived from curve fitting using a nonlinear regression analysis.

TABLE V Apparent ligand binding constants for GFR␣1-Fc and GFR␣2-Fc fusion proteins measured at 25°C using a surface plasmon resonance
The responses of serial dilutions of GFR␣1-Fc (0.8 -25 nM) and GFR␣2-Fc (1.5-100 nM) were fitted to the 1:1 Langmuir binding isotherm with the mass transfer using global fit and BIAcore evaluation software 3.0.1. presses GFR␣1 and cRET. In both permanent clones, we were unable to detect the presence of a high affinity site in the lower pM range (9,10). Moreover, Scatchard analysis of equilibrium saturation isotherms revealed the presence of downward curvilinearity at concentrations below 10 pM. Despite the variability of binding data at this low pM range, this curvilinearity was always reproducible between different experiments. One possible explanation of these results is the presence of positive cooperativity at this low pM range. Such positive cooperative interactions were already described at insulin receptors and TrkA receptors (30,31). It is also possible that much longer times are needed to reach an equilibrium at concentrations below 20 pM. In subsequent experiments we have varied the transient expression level of cRET by using different DNA ratios of GFR␣1 and GFR␣1/cRET of 3:1 and 1:3. Using either GFR␣1 alone or with cRET at a DNA ratio of 3:1, equilibrium saturation isotherms have also revealed the presence of a high affinity site in the range of 100 pM with a downward curvilinearity in the Scatchard plot (Fig. 2, A-C). Surprisingly, using a DNA ratio of 1:3, an increased affinity was obtained, and the presence of a small percentage of high affinity sites was detected (Fig. 2D). More detailed equilibrium saturation isotherm in the low pM range clearly demonstrated the presence of this high affinity site (Fig. 2, E and F). These results suggest that the high affinity site is present only if a sufficient amount of cRET receptors is present on the cell surface. In the same experimental conditions, no positive cooperativity was observed. The mechanism for this observation remains unknown. If GFR␣1 is expressed alone or cRET expression is relatively low, positive cooperative interactions can be seen. These results are not in agreement with earlier studies where low and high affinity sites were reported in cells expressing either GFR␣1 or GFR␣1/cRET (9, 10). Using 293T cells expressing GFR␣1 and a whole cell-based binding assay, Jing et al. (9) reported a high affinity site with a K D value of 2.3 pM and low affinity site with a K D value of 170 pM. These affinities were not influenced by the presence of cRET; K D values of 1.5 and 332 pM were found in Neuro 2A cells expressing both proteins. The authors reported the presence of low levels of cRET mRNA in 293T cells, and we have also found endogenous cRET mRNA in HEK 293 cells (data not shown). These low levels of cRET mRNA may be sufficient to produce enough cRET protein to influence the affinity of iodinated GDNF. It has also been shown that CHO cells under certain experimental conditions can express low levels of neurturin (2). With the experimental conditions used in this study, such as extensive washing and a 2-h incubation on ice, it is very unlikely that endogenous NTN could influence the affinity of iodinated GDNF.
To further investigate the influence of cRET, competition binding studies were performed with CHO cells transiently or permanently expressing GFR␣1 and cRET. Both GDNF and NTN were able to displace 125 I-GDNF binding to CHO cells permanently expressing either GFR␣1 or GFR␣1/cRET, with single low affinity sites (Fig. 4). When GFR␣1 was transiently expressed alone or with cRET at a DNA ratio of 3:1, GDNF also displaced 125 I-GDNF binding with a single affinity site of lower affinity than in permanent clones ( Fig. 5 and Table III). Only with an increased amount of cRET cDNA, either as a DNA ratio of 1:1 or 1:3, GDNF displacement of 125 I-GDNF binding fitted best a two site affinity model. At these conditions a high affinity binding site with a K D of 2-10 pM was observed (Fig. 5, C  and D). These results indicate again that the presence of a high affinity site is dependent on the expression level of cRET. In these competition experiments, an initial increase of specific binding from 100 to 125% at a low pM range of cold ligand supports the presence of positive cooperative interactions between GFR␣1 receptors. With an increased expression of cRET, positive cooperativity was abolished, and cold GDNF displaced iodinated GDNF with two affinities.
Based on our data, we tentatively propose a reaction scheme to be verified by further experiments. The reaction scheme is presented in Fig. 8. Our binding scheme is based on interactions between GDNF, GFR␣1, and cRET. It allows for interactions of GFR␣1 and cRET in the absence of GDNF and gives the possibility to explain cooperative interactions (Fig. 8) (32). In the absence of cRET and the presence of low GDNF concentrations, ligand-induced conformational change results in positive cooperativity and favorable effects on the binding of the subsequent ligand(s) (k2 Ͼ k1, Fig. 8A). When cRET is present in sufficient amount, its preassociation with GFR␣ results in the detection of low (RZ) and high (R*Z) affinity states of receptor complex (Fig. 8B). In the presence or absence of cRET, high concentration of GDNF results in a conformational change that has unfavorable effects on the binding of the subsequent ligand(s), negative cooperativity (k3 Ͼ Ͼ k4; kЈ3 Ͼ Ͼ kЈ4, Fig. 8C). This is a simplified reaction scheme based on our results, and more detailed kinetic studies are needed to confirm these observations.
We further characterized the binding, using fusion GFR␣1-Fc and GFR␣2-Fc proteins coated to SPA beads or using SPR. We compared these results to GDNF binding affinities obtained with full-length GFR␣1 receptors expressed in CHO cells. Although in the cell-based assays, iodinated GDNF bound only to GFR␣1, using SPA assays binding of iodinated GDNF to both GFR␣1-Fc and GFR␣2-Fc could be detected. Similarly, binding of iodinated NTN to both GFR␣1-Fc and GFR␣2-Fc could also be detected. Interestingly, using the SPR, the binding of GFR␣2-Fc to GDNF could not be detected. These results demonstrate that some reported discrepancies in binding specificities of GDNF and NTN could be because of different experimental approaches. It is interesting that in SPA assays, the binding of iodinated GDNF to GFR␣1-Fc and GFR␣2-Fc can be competed by both cold GDNF and NTN, but binding of FIG. 8. Proposed reaction scheme for GDNF binding to GFR␣1 receptors. R, RZ, and R*Z are different binding states of GFR␣1 receptor. Z represents cRET. LL is a GDNF dimer. A, in the absence of cRET and the presence of low GDNF concentrations, ligand-induced conformational change results in positive cooperativity and has favorable effects on the binding of the subsequent ligand(s) (k2 Ͼ k1). B, when cRET is present in sufficient amount, its pre-association with GFR␣ results in the detection of low (RZ) and high (R*Z) affinity states of receptor complex. C, in the presence or absence of cRET, high concentrations of GDNF result in a conformational change of the receptorligand complex that has unfavorable effects on the binding of subsequent ligand(s), negative cooperativity (k3 Ͼ Ͼ k4; kЈ3 Ͼ Ͼ kЈ4). iodinated NTN to GFR␣2-Fc can only be competed by cold NTN. Cold GDNF competes only with a small percentage of 125 I-NTN binding. Similar observations have been reported and suggest a difference in neurotrophic factor binding pockets at GFR␣1 and GFR␣2 receptors (12). In addition, cold NTN and GDNF used at higher concentrations increased the binding of iodinated NTN to both GFR␣1-Fc and GFR␣2-Fc receptors. A similar effect could not be seen with iodinated GDNF.
In summary, we have presented the first detailed binding characterization of GDNF and NTN to GFR␣1 and GFR␣2 receptors. The results presented here demonstrate the complexity of GDNF interaction with its receptor underlined by the presence of positive and negative cooperative interactions. The physiological role of positive cooperativity may be to ensure full occupancy of GDNF receptors at low GDNF concentrations. Positive cooperativity occurs in the range of GDNF concentrations responsible for biological activity (1) and may have an important role in cRET-independent signaling (33). On the other hand, the role of negative cooperativity may be to attenuate the cellular response at higher concentrations of GDNF. A similar binding mechanism has been proposed for the insulin receptor (26). Finally, our model of GDNF binding to GFR␣1 expressed in CHO cells proposes the presence of a high affinity site in the low pM range only in the presence of cRET tyrosine kinase and suggests that interaction between GFR␣1 and cRET in the absence of GDNF is necessary for receptor complex conformation with two affinity sites.