Small Molecules Can Selectively Inhibit Ephrin Binding to the EphA4 and EphA2 Receptors*

The erythropoietin-producing hepatocellular (Eph) family of receptor tyrosine kinases regulates a multitude of physiological and pathological processes. Despite the numerous possible research and therapeutic applications of agents capable of modulating Eph receptor function, no small molecule inhibitors targeting the extracellular domain of these receptors have been identified. We have performed a high throughput screen to search for small molecules that inhibit ligand binding to the extracellular domain of the EphA4 receptor. This yielded a 2,5-dimethylpyrrolyl benzoic acid derivative able to inhibit the interaction of EphA4 with a peptide ligand as well as the natural ephrin ligands. Evaluation of a series of analogs identified an isomer with similar inhibitory properties and other less potent compounds. The two isomeric compounds act as competitive inhibitors, suggesting that they target the high affinity ligand-binding pocket of EphA4 and inhibit ephrin-A5 binding to EphA4 with Ki values of 7 and 9 μm in enzyme-linked immunosorbent assays. Interestingly, despite the ability of each ephrin ligand to promiscuously bind many Eph receptors, the two compounds selectively target EphA4 and the closely related EphA2 receptor. The compounds also inhibit ephrin-induced phosphorylation of EphA4 and EphA2 in cells, without affecting cell viability or the phosphorylation of other receptor tyrosine kinases. Furthermore, the compounds inhibit EphA4-mediated growth cone collapse in retinal explants and EphA2-dependent retraction of the cell periphery in prostate cancer cells. These data demonstrate that the Eph receptor-ephrin interface can be targeted by inhibitory small molecules and suggest that the two compounds identified will be useful to discriminate the activities of EphA4 and EphA2 from those of other co-expressed Eph receptors that are activated by the same ephrin ligands. Furthermore, the newly identified inhibitors represent possible leads for the development of therapies to treat pathologies in which EphA4 and EphA2 are involved, including nerve injuries and cancer.

The Eph 2 receptors compose a large family of receptor tyrosine kinases that have been extensively studied for their roles in the developing and adult nervous system and in the developing cardiovascular system (1)(2)(3)(4)(5)(6). In recent years the Eph receptors have also been implicated in many different physiological and pathological processes, including the regulation of insulin secretion, bone homeostasis, immune function, blood clotting, pathological forms of angiogenesis, and cancer (7). The ability to modulate the activities of this family of receptors is therefore of critical interest to gain a better understanding of their functions in the physiology of many organs and in various pathological conditions, as well as for medical therapy.
The Eph receptors exert their effects by interacting with ligands, the ephrins, which are also membrane-bound proteins. Eph receptor-ephrin interaction is mediated by two binding sites in the amino-terminal ephrin-binding domain of the receptor as follows: a high affinity site, which includes a hydrophobic cavity that accommodates a protruding loop of the ephrin (the G-H loop), and a separate low affinity site (8). A third molecular interface located in the adjacent cysteine-rich region of the receptor has also been described (9). Despite the presence of several binding interfaces, peptides that target the high affinity site are sufficient to inhibit Eph receptor-ephrin binding (10 -12). Interestingly, unlike the ephrins whose binding is highly promiscuous, a number of the peptides that were identified by phage display selectively bind to only one or a few of the Eph receptors (10,13,14).
Other molecules that modulate Eph-ephrin interactions have also been identified, including antibodies and soluble forms of Eph receptors and ephrins extracellular domains (2,(15)(16)(17). Several small molecule inhibitors of the Eph receptor kinase domain have also been reported (18 -21). These inhibitors occupy the ATP binding pocket of the receptors and are usually broad specificity inhibitors that target different families of tyrosine kinases (18,19). Epigallocatechin gallate, a green tea derivative known to inhibit several tyrosine kinases, has also been shown to inhibit EphA receptor-mediated a human umbilical vein endothelial cell (HUVE) * This work was supported, in whole or in part, by a National Institutes of Health grant (to E. B. P.). This work was also supported by the Department of Defense (to E. B. P.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. migration and capillary-like tube formation, but the mechanism of action of this molecule has not been elucidated (22). Although the size, polarity, and geometry of the high affinity ephrin-binding pocket of the Eph receptors suggest that it might accommodate the binding of a small molecular weight chemical compound (23), no such inhibitors have been identified so far for any of the Eph receptors. The Eph receptors are subdivided in two classes, which in the human genome include nine EphA receptors, which preferentially bind the five ephrin-A ligands, and five EphB receptors, which preferentially bind the three ephrin-B ligands. Binding between receptors and ephrins of the same class is highly promiscuous, and few examples of inter-class binding have also been reported (24). In particular, EphA4 can bind both ephrin-A and ephrin-B ligands and represents the most promiscuous member of the Eph family. This peculiar feature of EphA4 makes its ephrin-binding pocket particularly interesting to target. Furthermore, besides being a well know regulator of neural connectivity during development and of synaptic function in the adult brain (25,26), EphA4 has also been linked to several pathologies, which suggests that this receptor could be a promising new target for drug development. For example, EphA4 has been implicated in the inhibition of spinal cord regeneration after injury, by promoting the formation of the glial scar and inhibiting axon regrowth (27)(28)(29). In addition, EphA4 is expressed on the surface of human platelets, where it promotes thrombus stabilization (30). EphA4 has also been detected in different types of cancer cells (31)(32)(33) as well as in tumor endothelial cells (34,35). Hence, modulation of EphA4-ephrin binding could be useful in the treatment of different pathological conditions.
In this study, we have used a high throughput screening approach to identify small molecular weight compounds that inhibit ligand binding to the EphA4 receptor. This screen identified two isomeric 2,5-dimethylpyrrolyl benzoic acid derivatives that selectively inhibit ephrin binding to EphA4 and EphA2 as well as the functions of these receptors in live cells.

EXPERIMENTAL PROCEDURES
Chemical Library Screening for EphA4 Inhibitors-A 96-well format in vitro assay was used for compound screening. Polystyrene high binding capacity plates (Corning Glass) were incubated for 18 h at room temperature with 2 g/ml streptavidin (Pierce) diluted in borate buffer (0.1 M boric acid, 0.1 M sodium borate, pH 8.7), washed three times with binding buffer (Trisbuffered saline (TBS: 150 mM NaCl, 50 mM Tris-HCl, pH 7.5) with 1 mM CaCl 2 and 0.01% Tween 20), blocked with 0.5% bovine serum albumin in TBS for 1 h at room temperature, washed three times with binding buffer, and then coated by overnight incubation at 4°C with 0.1 M biotinylated KYL peptide (14) in binding buffer. Peptide-coated plates were washed five times with binding buffer, and compounds were added to the wells at a final concentration of 10 g/ml in 1% dimethyl sulfoxide (DMSO) together with EphA4 alkaline phosphatase fusion protein (EphA4 AP) produced from transfected cells. Cell culture medium containing the secreted EphA4 AP was diluted 1:16 in binding buffer. The mixture was incubated for 3 h at room temperature on a plate shaker before washing five times with binding buffer and adding as the substrate 1 mg/ml p-nitrophenyl phosphate (Pierce) in SEAP buffer (105 mM diethanolamine, 0.5 mM MgCl 2 , pH 9.8). After 1 h at room temperature the reaction was stopped by adding 2 N NaOH, and the absorbance at 405 nm was measured using an ELISA plate reader. Alkaline phosphatase activity from wells where AP was added instead of EphA4 AP was subtracted as background. The inhibitory activity of the compounds was calculated by dividing the absorbance observed in the presence of compound and the absorbance from wells where no compound was added. Compounds with inhibitory activity higher than 50% were considered hits. The inhibitory activity of the hits was confirmed by repeating the assay.
ELISAs and K i Determination-Protein A-coated wells (Pierce) were used to immobilize ephrin Fc fusion proteins (R & D Systems, Minneapolis, MN). Compounds at different concentrations were incubated in the wells with EphA4 AP (36) or EphA2 AP (13) for 3 h. Alternatively, Eph receptor Fc fusion proteins were immobilized on protein A-coated wells, and ephrin-A5 AP (37) or ephrin-B2 AP (GeneHunter, Nashville, TN) was added with the compounds. The amount of bound AP fusion protein was quantified using p-nitrophenyl phosphate as the substrate. Alkaline phosphatase activity from wells with Fc only was subtracted as background.
To confirm that the binding of the compounds to EphA4 was reversible, the compounds were removed, and the wells were incubated in binding buffer for 3 h before washing and incubating with ephrin AP fusion proteins. Under these conditions, no inhibition of ephrin binding was observed, as expected for reversible inhibitors. Further control experiments verified that the compounds do not inhibit the activity of alkaline phosphatase in solution and also do not inhibit binding of EphA4 AP to an anti-EphA4 antibody (R & D Systems, Minneapolis, MN) immobilized to protein G-coated plates (Pierce), ruling out nonspecific inhibitory effects.
To calculate the inhibition constant (K i ) values, the binding of ephrin-A5 AP to EphA4 Fc immobilized on protein A-coated wells was measured in the absence and in the presence of the compounds at different concentrations. Each data set was fitted to the Michaelis-Menten equation is the concentration of ephrin AP fusion protein, and K D is the dissociation constant observed in the absence or in the presence of the compound, using nonlinear regression and the program Prism (Graph-Pad Software Inc.). To evaluate whether the inhibition is competitive, noncompetitive, or uncompetitive, the K D and B max values were determined at different compound concentrations. The K i was obtained from the linear regression plot of K D /B max as a function of the concentration of the inhibitor according to the following: K D /B max ϭ (K D [S])/(K i B max ) ϩ K D /B max . Alternatively, K i values were obtained from the dose-response curves, using the Cheng-Prusoff equation: (38). Ephrin-A5 AP concentrations were calculated from alkaline phosphatase activity (39).
For the synthesis of compounds 1, 26, 27, 37, 39, 41, 42, and 54, a 15-ml glass pressure vessel was charged with the appropriate aniline (1.0 mmol), 2,5-hexanedione (1.2 mmol), p-toluenesulfonic acid (0.2 mmol), and toluene (5.0 ml). The mixture was stirred and heated under reflux for 24 h. After evaporation of the toluene, the crude product was purified first by flash chromatography (ethyl acetate/hexanes) and then by reverse phase chromatography. The final products were lyophilized to give solids in yields ranging from 47 to 82%. Final product purities of greater than 95% were confirmed by 1 H NMR or liquid chromatography/mass spectrometry.
For the synthesis of compounds 3, 4, 7, and 19, a 35-ml microwave tube was charged with the appropriate aniline (1.0 mmol), 2,5-hexanedione (1.2 mmol), p-toluenesulfonic acid (0.2 mmol), and ethanol (5.0 ml). The mixture was heated under microwave irradiation at 180°C for 5 min. The solvent was then evaporated, and the residue was subjected to flash chromatography (0 -15% ethyl acetate/hexanes or 0 -10% methanol/dichloromethane) and then reverse phase chromatography if required. The final products were lyophilized to give solids in yields ranging from 30 to 80%. Final product purities of greater than 95% for compounds 4, 7, and 19, and greater than 80% for compound 3 were confirmed by 1 H NMR or liquid chromatography/mass spectrometry.
For the synthesis of compounds 5, 22, 40, and 47, the appropriate aryl halide (0.5 mmol) was mixed with 2,5-dimethylpyrrole (0.7 mmol), CuI (0.1 mmol), N-methylglycine (0.2 mmol), and potassium carbonate (1.5 mmol) in dimethylformamide (5.0 ml). The mixture was placed in a sealed glass vial and irradiated under microwave conditions at 200°C for 20 min. The resulting mixture was cooled, filtered, and concentrated in vacuo. The resulting residue was dissolved in acetonitrile and purified via reverse phase chromatography. After lyophilization, the product pyrroles were furnished as solids with yields ranging from 26 to 57%. Final product purities of greater than 95% were confirmed by 1 H NMR or liquid chromatography/mass spectrometry. The identity and purity of all the synthesized compounds and compound 1 purchased from InterBioScreen was verified by liquid chromatography/ mass spectrometry.
Measurement of Receptor Tyrosine Phosphorylation in Cells-HT22 neuronal cells, which endogenously express EphA4, are derived from immortalized mouse hippocampal neurons (40). COS7 cells, which endogenously express EphA2, EphB2, and the epidermal growth factor (EGF) receptor, were obtained from ATCC. Both cell lines were grown in Dulbecco's modified Eagle's medium (DMEM) (Mediatech, Inc, Herndon, VA) with 10% fetal bovine serum (FBS) (Hyclone, Logan, UT) and penicillin/streptomycin. For EphA4 immunoprecipitations, HT22 cells were serumstarved overnight in 0.5% FBS in DMEM and incubated for 15 min with the compounds or DMSO as a control. The cells were then stimulated with 0.5 g/ml ephrin-A5 Fc, ephrin-A4 Fc, or Fc for 20 min in the continued presence of the compounds. After stimulation the cells were lysed in modified RIPA buffer (1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 20 mM Tris, 150 mM NaCl, 1 mM EDTA) containing 10 M NaF, 1 M sodium pervanadate, and protease inhibitors. Protein concentrations were calculated using the BCA protein assay kit (Pierce). Cell lysates were immunoprecipitated with 4 g of anti-EphA4 antibody (41).
For EphA2 and EphB2 immunoprecipitations, serumstarved COS7 cells were stimulated with 0.1 g/ml ephrin-A1 Fc or 0.5 g/ml ephrin-B2 Fc, respectively. The cells were then lysed and incubated with 2 g of anti-EphA2 antibody (Millipore-Upstate, Inc, Temecula, CA) or 7 g of anti-EphB2 antibody made to a glutathione S-transferase fusion protein of the EphB2 carboxyl-terminal tail (42). To assess EGF receptor phosphorylation, COS7 cells were serum-starved overnight in 0.2% FBS in DMEM. The cells were preincubated with the compounds as described above and then stimulated for 15 min with 0.1 M EGF. PC3 cells were grown in RPMI 1640 medium (Mediatech, Inc, Herndon, VA) with 10% FBS and penicillin/streptomycin. EphA2 was immunoprecipitated from PC3 cells as described above but after stimulation with 0.5 g/ml ephrin-A1 Fc.
To assess inhibition of EphA2 phosphorylation in response to endothelial cell stimulation with tumor necrosis factor-␣ (TNF␣), HUVE cells obtained from Cascade Biologics (Portland, OR) were grown in Medium 200 supplemented with low serum growth supplements (Cascade Biologics), 10% FBS, penicillin/streptomycin, and fungizone. The cells were serumstarved overnight in 2% FBS containing medium before adding 7 nM TNF␣ together with the compound or DMSO for 2 h.
Immunoprecipitates and lysates were probed by immunoblotting with anti-phosphotyrosine antibody (Millipore Inc., Temecula, CA) and reprobed with antibodies to the respective Eph receptors or anti-EGF receptor antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) followed by secondary peroxidase-conjugated antibodies (GE Healthcare). The EphA2 and EphA4 antibodies used for immunoblotting were from Invitrogen/Zymed Laboratories Inc.
MTT Assay-The cytotoxicity of the compounds was measured using the MTT colorimetric assay. Cells were seeded in 96-well plates and treated with compounds or DMSO starting 3, 2, or 1 day before they reached 100% confluency. For the assay, MTT (Sigma) was added at a final concentration of 0.5 g/ml and incubated with the cells for 3 h. The resulting formazan crystals were then solubilized by addition of 100% DMSO. The absorbance in each well was measured at 570 nm using an ELISA plate reader. The results were expressed as the ratio of the absorbance of the cells treated with the compounds or left untreated.
Growth Cone Collapse Assay-Nasal retina explants from embryonic day 6 chicken embryos were cultured on coverslips coated with 200 g/ml poly-L-lysine and 20 g/ml laminin for 12-24 h in DMEM/F-12 culture medium containing 10% FBS and 0.4% methylcellulose. Three hours before adding the Fc fusion proteins, the medium was changed to DMEM/F-12 without methylcellulose. The explants were OCTOBER 24, 2008 • VOLUME 283 • NUMBER 43 incubated for 15 min with the KYL peptide or the compounds and then stimulated for 30 min with 1 g/ml preclustered ephrin-A5 Fc or Fc as a control. The cultures were fixed for 30 min in 4% paraformaldehyde, 4% sucrose in phosphate-buffered saline (PBS), permeabilized in 0.1% Triton X-100 in PBS, and stained with rhodamine-conjugated phalloidin (Invitrogen). Cells were photographed under a fluorescence microscope, and growth cones were scored in a blinded manner as collapsed when no lamellipodia or filopodia were present at the tip of the neurite.

Eph-Ephrin Small Molecule Inhibitors
PC3 Cell Retraction Assay-PC3 cells were plated on glass coverslips, and after 17 h they were starved for 3 h in 0.5% FBS in DMEM and then incubated for 40 min with the compounds or DMSO, before stimulation for 10 min with 0.5 g/ml ephrin-A1 Fc or Fc as a control. The cells were then fixed in 4% formaldehyde in PBS, permeabilized in 0.5% Triton X-100 in TBS, and stained with rhodamine-conjugated phalloidin (Invitrogen) and 4Ј,6-diamidino-2-phenylindole. Cells were FIGURE 1. High throughput screening identifies small molecules that inhibit EphA4 ligand binding. A, results from the screen showing the ELISA plate from which compound 1 was identified. Orange, well containing compound 1; yellow, wells containing compounds from the library that are not inhibitory; red, control wells containing EphA4 AP; blue, control wells containing AP; and white, control wells containing only buffer. B, 2,5-dimethylpyrrolyl benzene derivatives identified in the high throughput screening for EphA4 inhibitors; the names of the compounds, indicated at the left of the structures, correspond to those in Fig. 4. The first value in the % inhibition (% Inhib.) column was obtained in the original screen with 10 g/ml compound; the second value was obtained in a confirmatory repeat of the experiment. IC 50 values were calculated by measuring binding of EphA4 AP to immobilized KYL peptide or ephrin-A5 AP to immobilized EphA4 Fc in the presence of different concentrations of the compounds. photographed under a fluorescence microscope, and cell area was measured in a blinded manner using ImageJ software (rsb.info.nih.gov). Cells having rounded shape and area equal to or below 20% of the area of Fc control-treated cells were considered as retracting.

Chemical Library Screening to Identify Compounds That Inhibit Ligand Binding to the EphA4
Receptor-To identify small molecule inhibitors of ligand binding to the EphA4 receptor, we designed an assay that takes advantage of a peptide ligand previously identified by phage display (14). The peptide, designated KYL, has some sequence similarity with the ephrin-A G-H loop, which mediates high affinity binding to Eph receptors (43). Furthermore, the KYL peptide was shown to competitively inhibit ephrin binding to EphA4, suggesting that it targets the high affinity ligand-binding site of the receptor (14). We considered the peptide more suitable for high throughput screening assays than an ephrin-A Fc ligand because it is less expensive to produce and binds to EphA4 with lower affinity, which should facilitate identification of initial hits.
The biotinylated KYL peptide was immobilized on streptavidincoated ELISA wells, and binding of the extracellular domain of EphA4 fused to alkaline phosphatase (EphA4 AP) was measured in the presence of chemical compounds. We screened 10,000 compounds from the DIVERSet TM library (ChemBridge, Inc.) at 10 g/ml in a 96-well format, and identified 43 compounds that reproducibly inhibited EphA4 AP binding by more than 50% in both the original screen and a rescreen of the hits (Fig. 1A). Four of the compounds shared a 2,5-dimethylpyrrolyl benzene scaffold and inhibited EphA4 AP binding to the KYL peptide with IC 50 values ranging from 3 to 56 M (Fig. 1B). Importantly, compound 1, 2-hydroxy-4-(2,5-dimethyl-1pyrrolyl)benzoic acid, also inhibited binding of ephrin-A5 AP to the EphA4 extracellular domain with an IC 50 value of 13 M (Fig. 2). Control experiments also verified that the compound binds reversibly to EphA4 and does not inhibit alkaline phosphatase activity or protein-protein interac-  tions other than EphA4 ligand binding (data not shown). Thus, compound 1 can inhibit binding of the EphA4 receptor to both a synthetic peptide ligand and a natural ephrin ligand.

Two 2,5-Dimethylpyrrolyl Benzoic Acid Derivatives Selectively Target the EphA4 and EphA2
Receptors-We obtained 49 additional compounds belonging to the same class as compound 1 from ChemBridge and other sources, and we examined them in ELISA experiments for their ability to inhibit EphA4-KYL and EphA4-ephrin-A5 binding. Compound 2, a 1,2-isomer of compound 1, also inhibited binding of ephrin-A5 AP to immobilized EphA4 (Fig. 2). The IC 50 value for inhibition of EphA4-KYL peptide binding by compound 2 was 3 M and for inhibition of EphA4-ephrin-A5 binding was 9 M (Fig. 2). By measuring ephrin-A5 AP-binding curves at different compound concentrations, we found that compounds 1 and 2 competitively inhibit EphA4-ephrin-A5 binding with K i values of 8 and 7 M, respectively (Fig. 2). These data suggest that compounds 1 and 2 target the high affinity ephrin-binding pocket of EphA4, which is consistent with the chemical shift perturbations caused by compounds 1 and 2 in NMR heteronuclear single quantum coherence spectra of the ephrin-binding domain of EphA4 (see accompanying article, Ref. 74). The K i value can also be obtained from the IC 50 value and the dissociation constant (K D ) for receptor-ligand binding, using the Cheng-Prusoff equation (see "Experimental Procedures"). The K i values for compounds 1 and 2 calculated from the inhibition curves shown in Fig. 2 were 10 and 6 M, respectively. K i values calculated from other inhibition curves obtained using different ephrin concentrations ranged from 6 to 10 M for compound 1 and from 6 to 8 M for compound 2 (data not shown).
Interestingly, despite the small size of the compounds and the ability of each ephrin ligand to bind promiscuously to different Eph receptors, compounds 1 and 2 preferentially inhibited ephrin binding to EphA4 and EphA2 among the EphA and EphB receptors examined (Fig. 3A). Assuming that compound 1 and 2 also competitively inhibit ligand binding to the EphA2 recep- EphA2 immunoprecipitates were probed with anti-phosphotyrosine antibody (PTyr) and reprobed for EphA2. F, histogram shows the levels of EphA2 phosphorylation quantified from immunoblots and normalized to the amount of immunoprecipitated EphA2. Error bars represent standard errors from three experiments. Receptor phosphorylation levels in cells treated with compound 1 were compared with those in nontreated samples by nonpaired Student's t test. *, p Ͻ 0.05; ***, p Ͻ 0.001. G, the same protocol described in C was used, except that COS7 cells were stimulated with 0.5 g/ml ephrin-B2 Fc, and the EphB2 receptor was immunoprecipitated. H, COS7 cells pretreated with the indicated concentrations of compounds 1 or 2 were stimulated with EGF (ϩ) or left unstimulated (Ϫ). Lysates were probed with anti-phosphotyrosine antibody (PTyr) and reprobed for the EGF receptor.
tor, the Cheng-Prusoff equation was used to calculate the K i values for inhibition of EphA2-ephrin-A5 binding, which ranged from 11 to 14 M for compound 1 and from 10 to 13 M for compound 2 (data not shown). Both compounds inhibited binding of most ephrin ligands to EphA4, except for ephrin-A4 and ephrin-B2, suggesting differences in how these ephrins bind to EphA4 (we could not examine the effect of the compounds on ephrin-B1 binding because we could not consistently detect binding of this ephrin to EphA4 under the conditions of our assays). Similar selectivity was obtained for EphA2-ephrin-A binding (Fig. 3B), suggesting that ephrin-A4 also interacts with EphA2 differently than other ephrins.
Structure-Activity Relationship Analysis of Small Molecules with a 2,5-Dimethylpyrrolyl Benzene Scaffold and Related Compounds-To obtain information that may help improve the potency of compounds 1 and 2, we measured IC 50 values for structurally related compounds available from commercial sources or that we synthesized ( Fig. 4 and supplemental Fig. 1). Among the 49 analogs initially examined (com- pounds 5, 6, 8 -18, and 20 -55), none detectably inhibited EphA4ephrin-A5 binding. Even small changes to the structures of compounds 1 and 2 abolished the ability to inhibit ephrin binding. For example, the presence of the hydroxyl and carboxylic acid moieties and their position on the benzene ring appears to be crucial for the antagonistic activity of the compounds (supplemental Fig. 1). No inhibition of EphA4-ephrin binding was observed with the methyl ester derivative of compound 1 (compound 21) or when a methoxy group replaced the carboxylic acid group of compound 1 (compound 40), suggesting that the carboxylic acid group may be involved in hydrogen bonding with EphA4. The two methyl groups on the pyrrole ring also appear to be critical, possibly because they modulate the dihedral angle of the benzene and pyrrole rings or because they contribute to favorable lipophilic interactions with the binding site in EphA4.
Although none of the compounds tested showed measurable inhibition of EphA4-ephrin-A5 binding, their IC 50 values for inhibition of EphA4-KYL peptide binding were  used as a guide to design modified versions of compounds 1 and 2 that might have increased potency (Fig.  4). For example, compounds 5, 6, and 8, which have a phenylpropanoic acid, a phenylacetic acid, and phenoxyacetic acid in place of the benzoic acid in compound 25, inhibited EphA4-KYL binding with 10 -40-fold lower IC 50 values than compound 25. This suggested that substituting the carboxylic acid group of compound 1 with these other groups might improve its inhibitory activity. We therefore synthesized compounds 3 and 4. The phenolic hydroxyl group highly improved the ability of compounds 3 and 4 to inhibit ephrin-A5 binding compared with compounds 6 and 5, which lack the hydroxyl and did not show any activity against ephrin-A5. Compounds 3 and 4 are still selective EphA4 and EphA2 inhibitors and show the same differential inhibition of ephrin binding as compounds 1 and 2 (data not shown). However, inhibition of EphA4-KYL binding was not greatly affected by the phenolic hydroxyl group, suggesting that this group is more important for inhibition of ephrin-A5 binding than peptide binding. This was confirmed by the lack of activity with ephrin-A5 of compound 7, a methyl ether of compound 4, which, however, inhibited EphA4-KYL binding with a relatively low IC 50 value. It is also interesting that despite being able to inhibit ephrin-A5 binding to EphA4, compound 3 inhibited EphA4-KYL binding less effectively than compounds 5-7, which do not measurably inhibit ephrin binding. This suggests that somewhat different structural features may be required for inhibition of EphA4 interaction with ephrin-A5 versus the KYL peptide.
The IC 50 values for compounds 10 and 15 were ϳ6and 3-fold lower than those for compounds 25 and 30, which only differ for the absence of a methyl group attached to the benzene ring. This suggested that adding a methyl group to the benzene ring of compounds 1 and 2 may improve their inhibitory activity. We therefore synthesized compound 19, which corresponds to compound 2, with an additional methyl group as a substituent on the benzene ring. However, compound 19 did not inhibit EphA4-ephrin-A5 binding and inhibited EphA4-KYL binding only when present at high concentration. Perhaps the ability of the methyl group to enhance the activity of compounds 10 and 15 depends on its position with respect to the other substituents on the benzene ring. If this is true, the synthesis of alternative compounds carrying the methyl group at different positions may give different results.
Compounds 1 and 2 Selectively Inhibit EphA4 and EphA2 Activation by Ephrin in Cells without Showing Toxicity-Compounds 1 and 2 were the best antagonists in the ELISAs. Therefore, we examined the ability of these two compounds to inhibit ephrin-induced EphA4 and EphA2 tyrosine phosphorylation (indicative of receptor activation) in cultured cells. Both compounds blocked tyrosine phosphorylation of endogenous EphA4 in HT22 neuronal cells stimulated with ephrin-A5 Fc, although the concentrations needed were higher than those effective in the ELISAs (Fig. 5, A and B). The compounds also inhibited tyrosine phosphorylation of endogenous EphA2 in COS7 cells stimulated with ephrin-A1 Fc (Fig. 5, C and D) and in HUVE cells treated with TNF␣ to stimulate expression of endogenous ephrin-A1 (Fig. 5, E and F) (44 -46). The TNF␣dependent increase in EphA2 phosphorylation did not occur in cells treated with the protein synthesis inhibitor cycloheximide, consistent with a TNF␣-mediated up-regulation of ephrin-A1 expression (data not shown). Furthermore, the compounds prevented ephrin-dependent degradation of EphA2 (47), as expected from inhibition of ephrin binding. Consistent with the selectivity observed in the ELISAs, compounds 1 and 2 did not inhibit EphA4 phosphorylation in cells stimulated with ephrin-A4 Fc (data not shown) or phosphorylation of endogenous EphB2 in COS7 cells stimulated with ephrin-B2 Fc (Fig. 5G). Moreover, the compounds did not inhibit phosphorylation of the EGF receptor in COS7 cells stimulated with EGF (Fig. 5H) or overall tyrosine phosphorylation in COS and HT22 cells (data not shown). Assessment of cell viability using the MTT assay did not reveal any toxicity of compounds 1 and 2 at concentrations up to 400 M for several days (Fig. 6).
Compounds 1 and 2 Inhibit EphA4-dependent Growth Cone Collapse in Retinal Neurons-Growth cones are enlarged structures at the leading edge of axons and control the growth of the axons toward their synaptic targets by responding to environmental cues (48,49). The growth cones of chicken retinal neurites are well known to collapse in response to ephrin-A ligand stimulation (50,51). Because EphA4 is homogeneously expressed in different parts of the retina, whereas other EphA receptors are preferentially expressed in the temporal but not the nasal region of the retina (52), EphA4 is the predominant EphA receptor in nasal retinal neurons. Therefore, we used explants from the chicken nasal retinal to examine the ability of compounds 1 and 2 to counteract EphA4-mediated growth cone collapse. Although co-expression of ephrin-A ligands with EphA4 in the nasal retina makes the growth cones less sensitive to the collapsing effects of ephrin-A5 Fc, the growth cones still collapse when exposed to high concentrations of the ephrin (50 -52). The KYL peptide, which has been shown to selectively inhibit EphA4-ephrin binding (14), blocked collapse of nasal growth cones stimulated with ephrin-A5, confirming the requirement for EphA4 activation (Fig. 6, A and B). Compound 1 (Fig. 7, C and D) and compound 2 (Fig. 7, E and F) also blocked the growth cone collapsing effects of ephrin-A5 Fc. Importantly, despite the sensitivity of growth cones to their surrounding environment (48,49), neither the KYL peptide nor the two compounds at concentrations as high as 400 M affected the shape of unstimulated growth cones.
Compounds 1 and 2 Inhibit EphA2-dependent Retraction of the Cell Periphery-EphA2 is known to induce changes in cell morphology when activated by ephrin-A1, including retraction of the cell periphery and cell rounding (53,54). Because EphA2 is the predominant EphA receptor expressed in PC3 prostate cancer cells (55), we used these cells to examine whether compounds 1 and 2 are able to inhibit EphA2-mediated cell retraction. Treatment with the compounds blocked EphA2 activation following stimulation with ephrin-A1 Fc (Fig. 8, A and B) as well as the decrease in cell spreading (Fig. 8, C, D, F, and G) and the increase in the percentage of rounded cells (Fig. 8, C, E, F, and H) caused by ephrin-A1 Fc stimulation. Importantly, the compounds did not affect cell morphology in the absence of ephrin treatment (Fig. 8, C-H).

DISCUSSION
We report here for the first time the identification of small molecules that inhibit the interaction between Eph receptors and ephrins. To isolate small molecule inhibitors of EphA4, we performed a high throughput screening designed to identify compounds that inhibit ligand binding to this receptor. These inhibitors are advantageous compared with standard tyrosine kinase inhibitors because they can act without penetrating inside the cell and can be more selective. Among the many Eph receptors tested, the two 2,5-dimethylpyrrolyl benzoic acid derivatives that we have identified indeed show preferential inhibition of only two Eph receptors, EphA4 and the closely related EphA2. Our results also suggest that the two compounds are competitive inhibitors that target the high affinity ligand binding pocket of the receptors, a conclusion that is supported by NMR studies with EphA4 (74).
Given the small size of the two dimethylpyrrole derivatives compared with the ephrin binding pocket, their selectivity for EphA4 and EphA2 is particularly interesting and suggests that these compounds target a region that is not highly conserved in other Eph receptors. The two dimethylpyrrole derivatives also show selectivity with regard to ephrin binding, because they inhibited association of most ephrins tested except for ephrin-A4 and ephrin-B2, even when used at high concentrations. This suggests that these ephrins bind differently to the receptors compared with other ephrins of the same class. For example, interfaces not involving the ephrin-binding pocket may be of higher affinity with ephrin-A4 and ephrin-B2 than with other ephrins. Alternatively, there may be differences in the binding of ephrin-A4 and ephrin-B2 to the ephrin-binding pocket despite the similarity of the G-H loops of these ephrins with those of other ephrins whose binding is inhibited by the compounds. Structural studies will be necessary to elucidate how different ephrins interact with EphA4 and EphA2. The selectivity of the two dimethylpyrrole derivatives toward different Eph receptors and ephrins was confirmed in cell-based assays, where the addition of the compounds selectively blocked the ephrin-dependent tyrosine phosphorylation of EphA4 and EphA2 but not EphB2. The compounds also had no effect on the EGF-dependent phosphorylation of the EGF receptor, which is instead inhibited by many of the small molecules targeting kinase domains (19) and by epigallocatechin gallate (56).
The two pyrrole derivatives, like the KYL peptide, also blocked EphA4-mediated growth cone collapse in retina explants, suggesting that the compounds and the KYL peptide could promote axon growth. Interestingly, EphA4 has been proposed to play multiple roles in the inhibition of spinal cord regeneration after injury. In mouse and rat models of spinal cord injury, expression of this receptor is up-regulated in both glial cells and neurons near the site of injury (27,28). EphA4 expressed in the reactive glial cells may act as a negative regulator of axon regeneration by favoring the formation of the glial scar and by stimulating ephrin-B reverse signaling in axons. Furthermore, EphA4 expressed in the damaged axons may interact with ephrin-B2 expressed in the surrounding astrocytes and ephrin-B3 expressed in myelin, leading to inhibition of axon sprouting and outgrowth (28,57). The relative importance of these different effects is not yet known; however, some data suggest that inhibiting EphA4 function may be beneficial for the treatment of spinal cord injuries. For example, it has been reported that EphA4 knock-out mice have a significantly reduced glial scar and improved ability to regenerate spinal cord connections after spinal cord injury (27). In addition, a recent study has shown that the KYL peptide protects rat neocortical growth cones from collapsing after ephrin-A5 Fc treatment and that infusion of the peptide (14) into the lesioned spinal cord enhances axon sprouting, reduces cavity formation, and improves behavioral recovery (29). Inhibition of retinal growth cone collapse by the two dimethylpyrrole derivatives is an encouraging result that suggests that similar compounds with higher affinity could be used to enhance axon regrowth after injury. Inhibition of EphA4-ephrin interaction could also be useful in neuropathologies characterized by dendritic spine loss in the brain (14,58), to promote blood clotting (30), and to inhibit some forms of cancer (32,33,35).
The other Eph receptor preferentially targeted by the two dimethylpyrrole derivatives, EphA2, is widely expressed in many types of cancer cells and in the tumor vasculature (15,59,60). The dimethylpyrrole derivatives inhibit EphA2-dependent retraction and rounding of prostate cancer cells stimulated with exogenous ephrin-A1 Fc, suggesting that treatment with the compounds can inhibit the functional effects of EphA2. Interestingly, the compounds completely reverted the effect of ephrin-A1 treatment on cell retraction and rounding at concentrations that only partially inhibited EphA2 tyrosine phosphorylation, suggesting that high levels of EphA2 activation may be required to promote changes in cell adhesion and morphology. Inhibiting EphA2-ephrin binding in cancer cells is expected to be useful in the cases where EphA2 is highly activated and its signaling activity promotes tumorigenesis (61-63), but not in other cases where the tumor cells express low levels of endogenous ephrin-A1 (64). However, the most exciting application of EphA2-targeting molecules is for inhibition of tumor angiogenesis and other forms of pathological angiogenesis (65)(66)(67)(68)(69)(70)(71). Importantly, EphA2 is expressed in adult angiogenic blood vessels, but not in embryonic or adult quiescent blood vessels (60,72), consistent with evidence that targeting the pathological effects of EphA2 does not affect the normal vasculature. Unlike the previously identified EphA2-targeting peptides, which inhibit EphA2-ephrin binding in ELISAs but stimulate EphA2 phosphorylation in cells (13), the dimethylpyrrole derivatives also inhibit EphA2 activation in cells, including endothelial cells treated with the angiogenic factor TNF␣ (45). Thus, this class of compounds may be further developed for inhibition of pathological forms of angiogenesis, similar to the EphA receptor Fc fusion proteins that have been successfully used to inhibit angiogenesis in mouse tumor models and in a rat model of retinopathy of prematurity (65,66,70,73).
Analysis of the structure-activity relationship of many analogs of the dimethylpyrrole derivatives for inhibition of EphA4-KYL peptide binding did not lead to the identification of compounds inhibiting EphA4-ephrin interaction with improved potency. However, the rational design of other analogs with improved potency should now be possible based on the threedimensional structure of EphA4 in complex with compounds 1 and 2, which provides valuable insight into the molecular interactions of the compounds with the receptor (74).
Pharmacological tools to manipulate Eph receptor function will open new avenues of research and therapy. The compounds identified in this study may be used as leads to develop pharmaceuticals for the treatment of pathologies caused by dysregulation of EphA2 and EphA4 function. Importantly, our results provide evidence that the high affinity Eph receptorephrin interface can be successfully targeted by inhibitory small molecules and demonstrate the feasibility of approaches to identify ligand-binding inhibitors for the Eph receptors, which may also have wide application with other families of receptors.