Shared and Unique Determinants of the Erythropoietin (EPO) Receptor Are Important for Binding EPO and EPO Mimetic Peptide*

We have shown previously that Phe93 in the extracellular domain of the erythropoietin (EPO) receptor (EPOR) is crucial for binding EPO. Substitution of Phe93 with alanine resulted in a dramatic decrease in EPO binding to the Escherichia coli-expressed extracellular domain of the EPOR (EPO-binding protein or EBP) and no detectable binding to full-length mutant receptor expressed in COS cells. Remarkably, Phe93 forms extensive contacts with a peptide ligand in the crystal structure of the EBP bound to an EPO-mimetic peptide (EMP1), suggesting that Phe93 is also important for EMP1 binding. We used alanine substitution of EBP residues that contact EMP1 in the crystal structure to investigate the function of these residues in both EMP1 and EPO binding. The three largest hydrophobic contacts at Phe93, Met150, and Phe205 and a hydrogen bonding interaction at Thr151 were examined. Our results indicate that Phe93 and Phe205 are important for both EPO and EMP1 binding, Met150 is not important for EPO binding but is critical for EMP1 binding, and Thr151 is not important for binding either ligand. Thus, Phe93 and Phe205 are important binding determinants for both EPO and EMP1, even though these ligands share no sequence or structural homology, suggesting that these residues may represent a minimum epitope on the EPOR for productive ligand binding.

Erythropoietin (EPO), 1 the primary cytokine involved in the regulation of red blood cell production, functions by binding and signaling through a cell surface receptor (EPOR) on red blood cell precursors (1)(2)(3)(4)(5). Signaling through the EPOR includes the activation of a receptor-associated tyrosine kinase, JAK2 (3), which leads to the subsequent phosphorylation and activation of STAT5 (4,5), a member of the signal transducer and activator of transcription (STAT) family of transcriptional activators (reviewed in Ref. 6). Activated STAT5 moves to the nucleus where it promotes transcription of target genes by binding to specific STAT5 response sequences. In binding studies, the EPOR has been detected on purified primary human erythroid progenitor cells with one affinity (K d ϭ 0.1 nM) or with two affinities (K d ϭ 0.1 nM and 0.57 nM) for 125 I-EPO (7,8). The significance of two affinities of the EPOR for EPO is not clear. Evidence suggests that there is an accessory component of the EPOR that increases EPO binding affinity (9 -11), but this component(s) has yet to be identified. Alternatively, two affinities of the EPOR for EPO may arise from two nonequivalent receptor binding sites on EPO, one of high affinity (ϳ1 nM) and one of low affinity (ϳ1 M), that have been reported for the extracellular domain of the EPOR in solution (12). Structurally, the human EPOR is a 484-amino acid glycoprotein with a single transmembrane segment located between extracellular and intracellular domains each of nearly equal size (13). EPOR is a member of a large family of cytokine and growth factor receptors whose ligand binding domains contain homologous sequences and are predicted to be structurally related (14). These ligand binding domains consist of approximately 200 -250 amino acid residues and contain two subdomains, each predicted to consist of seven ␤-strands and to be structurally related to fibronectin type III (FNIII) domains (14,15). The amino-terminal FNIII-like domain contains a pair of spatially conserved cysteine bridges, while the carboxyl-terminal FNIIIlike domain contains a conserved ␤-strand F and a highly conserved WSXWS motif that are hallmarks of receptors of the cytokine receptor family (14).
Previously, we reported the characterization of an EPO-binding protein (EBP), consisting of amino acids 1-225 of the extracellular domain of the mature human EPOR, that was expressed in E. coli in quantity and purity sufficient for sitespecific mutagenesis and crystallographic studies (16). Recently, the crystal structures of the EBP bound to agonist peptide (EPO-mimetic peptide, EMP1) and an antagonist peptide (EMP33) have been determined (17,18). EMP1 is a 20amino acid peptide that is one of a series of related peptides discovered by phage display methodology (19). EMP1 exhibits no homology to EPO, yet binds specifically to the EPOR and mimics the biological effects of EPO both in vitro and in vivo (19). The structure of the EBP in the EMP1⅐EBP complex * This work was supported in part by National Institutes of Health Grant GM49497 (to I. A. W.) and Grant GM54351 (to M. A. G.).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  consists of two domains, each containing seven ␤-strands arranged in a FNIII-like topology (17). This topology for the EBP was also observed in the recently described crystal structure of EBP (EPObp) bound to EPO (20). The structure of the EBP shows close homology with the structures of ligand binding domains of GHR, prolactin receptor, granulocyte colony-stimulating factor receptor, and GP130 (21)(22)(23)(24), all members of the cytokine receptor family. The EBP also shows structural homology with the extracellular domains of TF (25,26) and ligand bound IFN␥R␣ (27), and even a natural killer cell inhibitory receptor (28). Tissue factor and IFN␥R␣ are "class 2" cytokine receptors (14) in which the seven-␤-strand FNIII-like fold is maintained but the relative orientation of the two domains differs.
The interaction between EMP1 and EBP consists of two molecules of EMP1 bound by two molecules of EBP to form a two-fold symmetrical assembly, where each peptide interacts with its peptide partner and both receptor molecules (17). As a result of this binding symmetry, EBP residues that interact with EMP1 are practically identical for each EBP in the complex. These interactions include hydrophobic contacts with the side chains of Phe 93 , Met 150 , and Phe 205 and a hydrogen-bonding interaction with the side chain of Thr 151 of the EBP. The interactions with Phe 93 , Met 150 , and Phe 205 are among the most significant non-polar contacts in the binding interface in terms of the area of molecular surface buried by each of these residues (17). Prior to the determination of the EMP1⅐EBP crystal structure, we investigated what role Phe 93 and Met 150 might have in EPO binding, based on their predicted secondary structural homologies to the positions of ligand binding determinants in other receptors of the cytokine receptor family (29). Phe 93 was found to be critical for EPO binding, while Met 150 was found to be relatively unimportant. In the present study, we used the alanine substitution mutants at Phe 93 and Met 150 to determine if these residues contribute significantly to EMP1 binding. In addition, further alanine substitution mutants were created at EBP residues Thr 151 and Phe 205 . These residues were selected for mutagenesis, based on their contacts with EMP1 in the EMP1⅐EBP crystal structure and on predicted secondary structural homologies to the positions of ligand binding determinants in other receptors of the cytokine receptor family. Furthermore, these residues contact EPO in the recently reported EPO⅐EPObp crystal structure (20). Our results indicate that Phe 93 and Phe 205 are important for both EPO and EMP1 binding, while Thr 151 appears to be unimportant for the binding of these ligands. In contrast, Met 150 , which is relatively unimportant for EPO binding, is critical for EMP1 binding. Thus, some of the same residues on the EPOR are important for binding two very different ligands, EPO and EMP1, suggesting that these residues may be required for productive ligand binding to the EPOR.

EXPERIMENTAL PROCEDURES
Expression and Mutagenesis of the EBP-The bacterial expression, purification, and characterization of the 225-amino acid extracellular domain of the EPOR (referred to here as an EBP) has been described in detail elsewhere (16). Biophysical analysis indicated that the purified EBP contained the expected amino terminus, disulfide-bridging pattern, and molecular mass. The EBP exhibited a low nanomolar binding affinity (K d ϭ 5 nM) for EPO. Mutagenesis of the EBP was performed as described previously (29), and the concentrations of purified wild-type and mutant EBP were estimated using the experimentally determined extinction coefficient of the wild-type EBP of 2.3 absorbance units/ mg/ml at 280 nm (16). Gross structural characterization of the F205A-EBP by circular dichroism (CD) was performed as described previously (29).
Assay of the Binding Activity of Wild-type and Mutant EBP-Competition binding assays in which wild-type or mutant EBP compete for 125 I-EPO binding with EPOR on the surface of TF-1 cells were per-formed as described (30). Equilibrium binding experiments on the wildtype and mutant EBP were carried out using a surface plasmon resonance assay on a BIAcore 2000 instrument. Proteins were coupled to the carboxymethylated dextran surface using amine coupling chemistry. The derivatization levels were ϳ3,000 resonance units. Different concentrations of either EPO or EMP1 were injected over all four flow cells sequentially at a flow rate of 5 l/min. The surfaces were regenerated with a 1-min pulse of 10 mM sodium acetate, pH 4.0. Each chip had one negative control surface and one surface containing wild-type EBP. To calculate the dissociation constants, the change in the equilibrium amount of ligand bound as a function of the concentration of ligand was fit to the equation for a simple 1:1 binding model, or to a model for two independent binding sites where R is the response, R max is the maximum response, and K d is the dissociation constant. EMP1-mediated Dimerization of the EBP-The ability of EMP1 to mediate the dimerization of wild-type and mutant EBP was evaluated in chemical cross-linking assays using the sulfhydryl-reactive crosslinking reagent (1,4-di-(3Ј-(2Ј-pyridyldithio) propionamido)butane (DP-DPB; Pierce). The assay was performed as described previously (17,30), except that the reaction products were detected and quantified using high performance-size exclusion chromatography (HP-SEC). Briefly, wild-type or mutant EBP (11 M) were incubated with a variable concentration of EMP1 (stock prepared in 0.1% trifluoroacetic acid) and 0.5 mM DPDPB (stock prepared in dimethyl sulfoxide) in 75 l of phosphate-buffered saline, pH 7.4, for 4 h at room temperature then overnight at 4°C. All reactions and controls contained a final concentration of 4.4% Me 2 SO and 0.007% trifluoroacetic acid to improve the solubility of the cross-linker. The samples were analyzed by HP-SEC on a Waters 625 HPLC system equipped with a Waters 996 detector. Separations were performed at room temperature on a 7.8 ϫ 300-mm G3000 SWXL column (Supelco, Bellfonte, PA). The column was equilibrated in 10 mM Na 2 PO 4 , pH 7.2, 150 mM NaCl at a flow rate of 1 ml/min and was monitored at 220 nm. Under these conditions, EBP eluted at ϳ9.6 min, while the dimer product eluted earlier at ϳ8.9 min. These elution times correspond to M r ϳ25,000 for EBP monomer and M r ϳ52,000 for the dimer product based on calibration of the column using a commercial HP-SEC standard mixture (16). The percentage of dimer product formed was calculated by adding the integrator determined peak area for the monomer and dimer protein peaks to determine total protein peak area. The dimer area value was divided by the total protein peak area and multiplied by 100 to yield the percentage of dimer observed in each reaction mixture. The reported value for each concentration is the average over three different chromatographic separations, and the error bars are the standard deviation value for the three experiments.
STAT Activation Assay-For expression in COS7 cells, the various mutants were subcloned from the EBP bacterial expression plasmid pSAM3 (16) to a mammalian expression plasmid encoding the fulllength human EPOR in pSG5 (Stratagene, La Jolla, CA). Transient transfection of COS7 cells (ATCC, Rockville, MD) was performed using LipofectAMINE (Life Technologies, Inc.) per the manufacturer's instructions. Cells were cotransfected with expression plasmids encoding the wild-type or mutant (F93A, M150A, or F205A) forms of the human EPOR along with a plasmid encoding STAT5a (pME18S-STAT5a, kindly provided by Dr. Alice Mui, DNAX Research Institute). Stimulations were performed for 10 min at 37°C with either EPO or EMP1 at the stated concentrations. Electrophoretic mobility shift assays were performed using nuclear extracts prepared from resting or stimulated cells and an oligonucleotide probe containing the Fc␥RI STAT response sequence, as described previously (31).

RESULTS
Previously, we expressed the extracellular domain of the human EPOR as a soluble EBP in E. coli (16). The crystal structure of the EBP bound to an EPO-mimetic peptide (EMP1) revealed several EBP residues that form significant contacts with the peptide (17). These include major nonpolar interactions with the side chains of Phe 93 , Met 150 , and Phe 205 and a hydrogen bond with the side-chain hydroxyl group of Thr 151 ( Fig. 1). Phe 93 , Met 150 , and Phe 205 of the EBP along with Tyr P4 , Phe P8 , Trp P13 , and Cys P15 of the peptide form a hydrophobic core at the heart of the interaction between EBP and EMP1 (Ref. 17; Fig. 1). The hydrogen bond with the hydroxyl group of Thr 151 is part of a network of mostly main-chain hydrogen bonds formed with main-chain atoms of the type I ␤-turn of the peptide (Fig. 1).
To determine if Thr 151 and Phe 205 contribute significantly to EPO binding, these residues were replaced with alanine and the resultant mutant proteins were purified and tested for their ability to compete for 125 I-EPO binding with EPOR on the surface of TF-1 cells. Alanine was chosen as the replacement residue, since it is likely to result in a loss of interactions from the original side chain without introducing new interactions or gross structural changes (32,33). The T151A-EBP exhibited an IC 50 of 6.5 nM in this assay, nearly identical to the wild-type value of 3.5 nM, indicating that Thr 151 does not contribute significantly to EPO binding (Fig. 2). In contrast, the F205A-EBP does not compete for EPO in this assay except at very high concentrations. About 45% inhibition of 125 I-EPO binding to the cells was observed at the highest concentration of F205A-EBP tested (10 M; Fig. 2). Thus, the F205A-EBP exhibits an IC 50 Ͼ 10 M in this assay compared with the wild-type value of 3.5 nM, suggesting that this mutation results in a dramatic decrease in the ability to bind EPO. Competition binding analyses were also performed on the F93A-and M150A-EBP. Similar to the F205A-EBP, the F93A-EBP exhibited about 45% inhibition of 125 I-EPO binding at 10 M (Fig. 2) and, thus, also has an IC 50 Ͼ 10 M in this assay. The M150A-EBP bound EPO as well as wild-type, exhibiting an IC 50 of 3.5 nM. These results suggest that Phe 93 and Phe 205 are critical for EPO binding, while Met 150 and Thr 151 are relatively unimportant for EPO binding.
We have previously shown that the dramatic reduction in EPO binding in the F93A-EBP is not a result of gross structural changes caused by the mutation (29). To investigate the possible structural consequences of the F205A mutation, CD was used to evaluate the overall secondary structure and thermal denaturation of the F205A-EBP relative to the wild type. The F205A-EBP produced a CD spectrum similar to wild type except for reduced intensity in the 200 -220-nm region (Fig. 3A) and resulted in a small (ϳ5°C) decrease in thermal stability relative to the wild-type EBP (Fig. 3B). Taken together, the limited differences in CD spectrum and decrease in T m of the F205A-EBP indicate that this mutation does not cause significant changes in the global secondary structure or the stability of the EBP.
To determine an EPO binding affinity (K d ) for the EBP mutants, equilibrium binding analyses were performed on the wild-type and mutant EBP using surface plasmon resonance. As can be seen in Fig. 4A, the EPO binding curves for the M150A-and T151A-EBP were similar to wild type. The data for the wild-type and T151A-EBP fit best to a two-site model (see "Experimental Procedures"), yielding a high affinity K d of ϳ5 nM and a low affinity K d of ϳ900 nM for EPO. For the M150A-EBP, however, the two-site fit of the data gave no improvement over the 1:1 model. This mutant exhibited a single K d for EPO of 36 nM, only slightly increased relative to the high affinity (5 FIG. 1. EBP residues contacting EMP1 in the EMP1⅐EBP crystal structure. EMP1 binds EBP with a 2:2 stoichiometry in the crystal structure (17). Each peptide (blue) forms symmetrical contacts with both receptors (gray) in the complex. Phe 93 , Met 150 , and Phe 205 of the EBP interact with Tyr P4 , Phe P8 , Trp P13 , and Cys P15 of the peptide to form the hydrophobic core of the interaction between EBP and EMP1. The hydrogen bond with the hydroxyl group of Thr 151 is part of a network of mostly main-chain hydrogen bonds formed with main-chain atoms of the type I ␤-turn of the peptide. The six loops containing binding determinants in receptors or the cytokine receptor family are labeled L1-L6. For clarity, the side chains of Phe P8 and Cys P15 are not shown.
FIG. 2. EPO binding activity of the EBP mutants in competition binding assays. The wild-type and mutant EBP were tested for their ability to compete for 125 I-EPO binding with full-length EPOR on the surface of TF-1 cells. The percentage of 125 I-EPO bound was calculated by dividing the counts/min at each data point by the counts/min with no added EBP and multiplying by 100. The IC 50 values derived from these assays are 3.5 nM for both the wild-type and M150A-EBP, 6.5 nM for the T151A-EBP, and Ͼ10 M both the F205A-and F93A-EBP, indicating that the F205A and F93A mutations result in a large decrease in the binding of 125 I-EPO. nM) K d of the wild-type EBP. In contrast, the F205A-EBP exhibited a dramatically increased K d of 1 M relative to wildtype EBP (Fig. 4A). The difference between the 1 M K d and the Ͼ10 M IC 50 for EPO observed for the F205A-EBP (see above) might be expected, since competitive binding data (IC 50 ) are not directly comparable to equilibrium binding data (K d ). Furthermore, in the competition binding format, the EBP mutants compete for EPO binding with cell surface wild-type EPOR exhibiting a K d for EPO of 400 pM (34). Less active mutants might compete poorly if at all with the high affinity EPOR, leading to increased IC 50 values relative to K d values from a direct binding assay. In any case, the 1 M K d of the F205-EBP represents a decrease in binding affinity of about 200-fold from the 5 nM K d of the wild type, confirming the results of the competition binding experiments that Phe 205 is an important EPO binding determinant. The F93A-EBP showed no binding to EPO over the concentration range studied (1 nM to 10 M; Fig. 4A), confirming that Phe 93 is also an important EPO binding determinant of the EPOR (29).
Surface plasmon resonance was also used to examine EMP1 binding affinities of the wild-type and mutant EBP. The sensorgrams for EMP1 binding at a concentration of 1 M are shown in Fig. 4B. The F93A-, M150A-, and F205A-EBP did not show specific binding to EMP1 above the background seen for an underivatized surface or surfaces coated with unrelated proteins (leptin and leptin receptor). Only the T151A-EBP bound EMP1, showing a response similar to that of wild type (Fig. 4B). In binding curves generated at various concentrations of EMP1 (0.1-10 M), the wild-type and T151A-EBP had comparable affinities for EMP1 of ϳ350 nM (data not shown). This value is similar to the half-maximal response (EC 50 ) of 400 nM reported for EMP1 in cell proliferation assays (19). The F93A-, M150A-, and F205A-EBP showed slight binding only at M concentrations of EMP1 (from 6 to 10 M); however, the response was not high enough above controls to be attributed with certainty to specific binding (data not shown). Therefore, if the peptide binds at all to these mutants, it is with a much lower affinity (Ͼ10 M) than for wild type (0.35 M). These data indicate that Phe 93 , Phe 205 , and Met 150 are all critical binding determinants for EMP1.
The ability of EPO and EMP1 to bind and signal through full-length wild-type, F93A-, M150A-, and F205A-EPOR was tested in a STAT activation assay. The T151A mutation was not tested in this assay, since this mutation did not appear to have a significant effect on either EPO or EMP1 binding. COS cells cotransfected with full-length receptor and STAT5a constructs were treated with EPO or EMP1, and active STAT5a dimer bound to an oligonucleotide containing a STAT-responsive element was detected in gel shift assays. Initial experiments were performed with concentrations of ligand that elicited a good response for the wild-type EPOR. At these concentrations (0.24 or 1.23 nM EPO and 5 or 20 M EMP1), only the M150A-EPOR showed a response to EPO and none of the mutants showed a response to EMP1 (Fig. 5A). A faint band was observed for the wild-type EPOR at 0.24 nM EPO, but no band was detectable at this concentration for the M150A-EPOR, suggesting that this mutant may be slightly less sensitive to EPO than wild type. The wild-type and mutant EPOR were also tested at the relatively high concentrations of 1.23 M EPO and 200 M EMP1 (Fig. 5B). At these concentrations, the F205A-EPOR gave a faint response to EPO but still did not respond to EMP1. The F93A-EPOR, however, exhibited the opposite pattern. It still did not respond to EPO but demonstrated a significant response to EMP1. Finally, the M150A-EPOR did not show a response to EPO or EMP1 at the high CD spectra (A) were generated at 25°C and 0.365 mg/ml protein in phosphate buffered saline. Overall, the CD spectrum of the F205A-EBP is similar to that of the wild-type EBP except for reduced intensity from 200 to 220 nm. The F93A mutation results in essentially no change in the CD spectrum of the EBP (29). Thermal denaturation curves (B) were generated by plotting the ellipticity at 228 nm derived from the CD spectra at each temperature using 0.365 mg/ml protein in phosphate-buffered saline. The F205A mutation causes a small decrease in the T m of the EBP. The T m of the F93A-EBP is identical to that of the wild-type EBP (29).

FIG. 4. Analysis of EPO and EMP1 binding to the wild-type and mutant EBP by surface plasmon resonance.
A, equilibrium binding analyses of EPO binding to the wild-type and mutant EBP. The lines are the best fit of the data to Equations 1 or 2 (see "Experimental Procedures" and "Results"), and the symbols and vertical bars are the average and standard error, respectively, of three independent experiments. The dissociation constants (K d ) for EPO derived from these data are 5 nM (high affinity) and 900 nM (low affinity) for both the wild-type and T151A-EBP, and 36 nM (single affinity) for the M150A-EBP. The concentrations. The loss of a response to EPO at 1.23 M for the M150A-EPOR is probably due to supersaturating concentrations of ligand occupying the receptor binding site at a 1:1 ratio, thereby driving the receptor to monomerization and reducing signaling. This effect was seen for the wild-type EPOR at 1.23 M EPO and at 20 and 200 M EMP1 (Fig. 5, A and B). Evidently, the M150A-EPOR binds EPO well enough to exhibit this effect. To be sure that supersaturating concentrations of ligand were not inhibiting the signaling of the other mutants, intermediate concentrations of 49 nM EPO and 100 M EMP1 were tested. Under these conditions, none of the mutants showed a response to EPO and only the F93A-EPOR showed a response to EMP1, which was less than that observed at 200 M EMP1 (data not shown).
To investigate further the effects of the EBP mutations on the interaction with peptide, an EBP cross-linking assay was used to evaluate the ability of EMP1 to mediate the dimerization of the EBP mutants (17,30). Wild-type and mutant EBP were incubated with various concentrations of EMP1, and the reaction products were detected by HP-SEC. The percentage of dimer formed, calculated from the areas under the peaks, was plotted at each concentration of peptide (Fig. 6). The maximum amount of dimer formed by the T151A-EBP was about 80% at 100 M EMP1, similar to the wild-type EBP, which showed a maximum of about 95% dimer at 100 M peptide. The F205A-EBP also exhibited maximal dimerization of about 80%, although at a higher concentration of EMP1 (200 M) than wild type. Thus, the T151A and F205A mutations do not appear to have a large effect on EMP1-mediated dimerization of the EBP. In the case of the F93A-EBP, peak dimerization occurred at 100 M peptide, similar to wild type, but the maximal amount of dimerization achieved was only around 35% (Fig. 6). More dramatically, dimerization of the M150A-EBP was almost undetectable at 100 M peptide and only about 5% dimer was formed at the highest concentration of EMP1 tested (400 M).
These data indicate that the F93A mutation and especially the M150A mutation result in a large decrease in the ability of EMP1 to facilitate the dimerization of the EBP. Taken together with the results of the equilibrium binding and STAT activation assays, these results indicate that Phe 93 , Phe 205 , and Met 150 are all important for the binding and/or agonist activity of EMP1.

DISCUSSION
In this study, we investigated the effects of alanine substitution of EPOR residues Phe 93 , Phe 205 , Met 150 , and Thr 151 on binding to both EPO and an EPO-mimetic peptide (EMP1). These residues were chosen for investigation based on several criteria. First, all of these residues interact with EMP1 in the crystal structure of EMP1 bound to the EBP (17), suggesting that they may contribute significantly to EMP1 binding. Second, Phe 93 has been shown to be critical for EPO binding (29), suggesting that residues that interact with peptide in the EMP1⅐EBP crystal structure might also be involved in EPO binding. Third, many cytokine receptors contain ligand binding determinants in positions homologous to Phe 93 , Met 150 , and Thr 151 (29) and Phe 205 (35)(36)(37)(38)(39)(40)(41)(42), further suggesting that these residues may be involved in ligand binding to the EPOR. In addition, all these residues are buried in the interface with EPO in the EPO⅐EPObp crystal structure (20).
Although both Phe 93 and Phe 205 are important binding determinants for EPO, their relative contributions to the binding of this ligand differ. For example, the F205A-EPOR exhibits slight activity at 1.23 M EPO in the STAT activation assay, while the F93A-EPOR does not. The activity seen for the F205A-EPOR at 1.23 M EPO might be expected, based on the 1 M K d for EPO of the F205A-EBP in the equilibrium binding assay. In the same vein, the F93A-EPOR might not be expected to respond to 1.23 M EPO in the STAT activation assay, since the F93A-EBP did not exhibit detectable binding to concentrations of EPO as high as 10 M in the equilibrium binding assay. It is unlikely that the lack of activity of the F93A-EPOR in the STAT assay results from poor expression on the surface of the COS cells used in the assay, since this mutant exhibited expression on COS cells comparable to wild type (29). Combined, these results suggest that Phe 93 is more important for EPO binding than Phe 205 . The 200-fold increased K d for EPO of the F205A-EBP compared with the 1,000-fold increased IC 50 for EPO reported for the F93A-EBP (29) adds further support to this hypothesis.
The data presented here indicate that Phe 93 , Phe 205 , and Met 150 are all important binding determinants for EMP1, since none of the respective mutant proteins exhibit detectable bind- ing to concentrations of EMP1 up to 10 M in the equilibrium binding assay, while Thr 151 is relatively unimportant for EMP1 binding in this assay. In addition, the F205A-and M150A-EPOR do not respond to peptide in the STAT activation assay, while the F93A-EPOR did show a response, but only at high concentrations of EMP1 (100 and 200 M). These effects could be the result of the reduced binding of EMP1 or aberrant dimerization and/or signaling of the mutant receptors. In any case, the results of the STAT activation assay indicate that Phe 93 is not as important as Phe 205 and Met 150 for activity with EMP1. The results of the EMP1-mediated dimerization (crosslinking) assay suggest that Phe 93 and especially Met 150 are important for the dimerization of the EBP. Phe 205 appears to be relatively less important for dimerization, since the F205A-EBP exhibits significant dimerization in the presence of peptide, albeit at higher concentrations. Apparently, the ability of EMP1 to mediate the dimerization of the F205A-EBP does not translate to activity in the STAT assay. The presence of the cross-linker capturing a transient dimerization of this mutant at the high concentrations of EBP used in the cross-linking assay might explain this discrepancy. The relatively small effect of the F205A mutation in the cross-linking assay does not mean that Phe 205 is unimportant for EMP1 binding, since the amount of dimerization observed for the EBP mutants may not be related to relative binding affinities due to the high concentrations of protein used in the assay. In addition, the presence of the cross-linking reagent complicates interpretation of the results, since different modes of binding and/or dimerization resulting from the mutations may affect the cross-linking reaction.
Of the 45 amino acid residues (20%) of the EBP that we have examined by alanine substitution to date (29,43), only Phe 93 and Phe 205 were found to be critical for EPO binding. The remaining residues had relatively little or no role in EPO binding or were important for structure. These results are consistent with the suggestion that only a few residues make significant contributions to binding in protein-protein interactions (functional epitope), even though a large number of residues may be involved in the contact interface (structural epitope; Refs. 44 -46). In the GH-GHR interaction, two tryptophans (Trp 104 and Trp 169 ) accounted for the majority of the binding free energy of the interaction, constituting a hot spot for GH binding to GHR (44). Alanine substitution of Trp 104 and Trp 169 resulted in the largest decreases in binding affinity by far, estimated to be greater than 2,500-fold, relative to wildtype GHbp (44,47). Our results suggest that Phe 93 and Phe 205 of the EPOR may be functionally analogous to Trp 104 and Trp 169 of the GHR in providing the majority of the binding energy for the interaction with ligand. When the EPO⅐EPObp crystal structure became available (20), we found we had already examined the function of 40% of the residues involved in H-bonds or salt bridges and nearly 50% of the residues involved in non-polar interactions with EPO (29,43). The data presented here, combined with the fact that Phe 93 and Phe 205 dominate the non-polar contacts with EPO site 1 and site 2 (20), indicate that these residues are an important component of the hot spot for EPO binding.
In the EPO⅐EPObp crystal structure, Met 150 contacts only a single residue (Phe 48 ) in binding site 1 on EPO and is more buried in site 2, contacting three EPO residues (20). Our results indicate that the single contact with site 1 is unimportant for EPO binding and the more extensive contacts with site 2 are apparently also relatively unimportant for EPO binding, since the response of the M150A-EPOR to EPO in the STAT activation assay was similar to that of wild type. In the case of EMP1 binding to the EPOR, Met 150 buries significant surface area with EMP1 (ϳ60 Å 2 ; Ref. 17) and is critical for EMP1 binding. However, Thr 151 also forms hydrophobic interactions with peptide in the EMP1⅐EBP crystal structure. In addition, the sidechain hydroxyl of Thr 151 forms a hydrogen bond with peptide (see Fig. 1). Despite these interactions, Thr 151 does not contribute significantly to EMP1 binding. Apparently, the deletion of a single hydrogen bond in the network of hydrogen bonds with peptide in the region around Thr 151 (see Fig. 1) is not sufficient to disrupt peptide binding. These results support the conclusion that the amount of surface area buried by a contact residue does not always correlate with binding affinity (44 -46) and emphasize the importance of mutagenesis studies to verify structural based predictions of the function of specific residues.
In conclusion, the fact that residues important for binding EPO (Phe 93 and Phe 205 ) are also important for binding the unrelated peptide, EMP1, suggests that these residues may represent a minimum epitope for productive ligand binding to the EPOR and has important implications for the design of small molecule mimetics. In addition, our results suggest that, as is the case for GHR, only a few residues of the EPOR contribute significantly to ligand binding (Phe 93 and Phe 205 for EPO binding and Phe 93 , Phe 205 , and Met 150 for EMP1 binding). Thus, our results provide support for the notion that the functional epitope for ligand binding is relatively small compared with the structural epitope. Indeed, this may account for the existence of a relatively small peptide mimetic of EPO (EMP1) in the first place and indicates that it may be possible to discover non-peptide small molecule mimetics of EPO and other biologically important cytokines.