Structural and Biophysical Characterization of the EphB4·EphrinB2 Protein-Protein Interaction and Receptor Specificity*

Increasing evidence implicates the interaction of the EphB4 receptor with its preferred ligand, ephrinB2, in pathological forms of angiogenesis and in tumorigenesis. To identify the molecular determinants of the unique specificity of EphB4 for ephrinB2, we determined the crystal structure of the ligand binding domain of EphB4 in complex with the extracellular domain of ephrinB2. This structural analysis suggested that one amino acid, Leu-95, plays a particularly important role in defining the structural features that confer the ligand selectivity of EphB4. Indeed, all other Eph receptors, which promiscuously bind many ephrins, have a conserved arginine at the position corresponding to Leu-95 of EphB4. We have also found that amino acid changes in the EphB4 ligand binding cavity, designed based on comparison with the crystal structure of the more promiscuous EphB2 receptor, yield EphB4 variants with altered binding affinity for ephrinB2 and an antagonistic peptide. Isothermal titration calorimetry experiments with an EphB4 Leu-95 to arginine mutant confirmed the importance of this amino acid in conferring high affinity binding to both ephrinB2 and the antagonistic peptide ligand. Isothermal titration calorimetry measurements also revealed an interesting thermodynamic discrepancy between ephrinB2 binding, which is an entropically driven process, and peptide binding, which is an enthalpically driven process. These results provide critical information on the EphB4·ephrinB2 protein interfaces and their mode of interaction, which will facilitate development of small molecule compounds inhibiting the EphB4·ephrinB2 interaction as novel cancer therapeutics.

Increasing evidence implicates the interaction of the EphB4 receptor with its preferred ligand, ephrinB2, in pathological forms of angiogenesis and in tumorigenesis. To identify the molecular determinants of the unique specificity of EphB4 for ephrinB2, we determined the crystal structure of the ligand binding domain of EphB4 in complex with the extracellular domain of ephrinB2. This structural analysis suggested that one amino acid, Leu-95, plays a particularly important role in defining the structural features that confer the ligand selectivity of EphB4. Indeed, all other Eph receptors, which promiscuously bind many ephrins, have a conserved arginine at the position corresponding to Leu-95 of EphB4. We have also found that amino acid changes in the EphB4 ligand binding cavity, designed based on comparison with the crystal structure of the more promiscuous EphB2 receptor, yield EphB4 variants with altered binding affinity for ephrinB2 and an antagonistic peptide. Isothermal titration calorimetry experiments with an EphB4 Leu-95 to arginine mutant confirmed the importance of this amino acid in conferring high affinity binding to both eph-rinB2 and the antagonistic peptide ligand. Isothermal titration calorimetry measurements also revealed an interesting thermodynamic discrepancy between ephrinB2 binding, which is an entropically driven process, and peptide binding, which is an enthalpically driven process. These results provide critical information on the EphB4⅐ephrinB2 protein interfaces and their mode of interaction, which will facilitate development of small molecule compounds inhibiting the EphB4⅐ephrinB2 interaction as novel cancer therapeutics.
The protein-protein interaction between the membranebound Eph receptor tyrosine kinases with the membrane-bound ephrin ligands have now been reported in the overexpression/dysregulation in numerous tumor cell lines (1). First reported for their role in axonal guidance, this group of proteins now has defined roles in regulating several cellular processes including developmental patterning, cell attachment, and motility (2)(3)(4). The importance of these proteins in development is underscored by the fact that deletion of either the EphB4 receptor or the ephrinB2 ligand results in lethality by embryonic day 11 as a result of arrested angiogenesis but not vasculogenesis (5). Understanding the EphB4⅐ephrinB2 interaction and exploring the determinants for the unique specificity of this receptor-ligand complex is at the core of modulating this activity and will allow for a deeper understanding into the basic biology behind this interaction and for the development of novel anti-angiogenesis and anti-tumorigenesis therapeutic approaches.
The EphB4 receptor and the ephrinB2 ligand are capable of transducing a signal bi-directionally into either the EphB4-expressing cell (forward signaling) or the ephrinB2-expressing cell (reverse signaling) (12,13). Therefore, a cellular response is conducted only upon cell-cell contact. The Eph receptors are divided into two subclasses, A and B, based on both sequence conservation and binding preferences with the ephrins (6). The receptors have a modular domain architecture, extracellularly characterized by an N-terminal ligand binding domain, a cysteine-rich domain, and two fibronectin type III-like repeats. Intracellularly, the receptors consist of a juxtamembrane domain, a conserved tyrosine kinase domain, a C-terminal sterile ␣-domain, and a PDZ binding motif. The ephrin ligands are also divided into two subclasses, A and B; ephrinA ligands are anchored to the membrane through a glycosylphosphatidylinositol linker, whereas members of the B-subclass are tethered to the membrane by a transmembrane region and contain a cytoplasmic PDZ domain binding motif. In general, EphA receptors (EphA1-EphA10) are promiscuously activated by ephrinA ligands (ephrinA1-ephrinA6), whereas EphB receptors (EphB1-EphB6) are promiscuously activated by eph-rinB ligands (ephrinB1-ephrinB3) (7). Cross-subclass interactions have been identified specifically between EphB2 and eph-rinA5 (8) and between EphA4 and ephrinB2/B3 (9); however, these interactions are rare and appear to be exceptions to the general rule. Unlike the other promiscuous receptors in this family, however, the EphB4 receptor has a distinctive specificity for a single ligand, ephrinB2 while binding only weakly to both ephrinB1 and ephrinB3. EphrinB2, on the other hand, binds to several receptors within the B-subclass (10). The ligand specificity of the EphB4 receptor is somewhat surprising, since the receptor has the characteristics of a promiscuous receptor, including a large and hydrophobic binding cavity and a flexible "lid" capable of accommodating a range of ligands (11). Here we provide a structural perspective describing the molecular determinants of EphB4 specificity.
Crystal structures have been previously determined for ephrin ligands (ephrinB2, ephrinB1) and Eph receptors (EphB2, EphB4) as well as receptor-ligand complexes (EphB2⅐ephrinB2, EphB2⅐ephrinA5, EphB4-TNYL-RAW (8,10,19). The crystal structure of the homodimeric ephrinB2 ligand reveals a ␤-barrel structure arranged in a deviation of the common Greek Key topology (17). Comparison of the crystal structures of both apo and receptor-bound ephrinB2 revealed that the solvent-exposed and high affinity G-H loop is conserved in position between the two structures. This critical loop shares high sequence homology in both A and B subclasses, respectively, and is involved in receptor-ligand heterodimerization as well as in ephrinB2 homodimerization (10,17). In addition, the crystal structures of the ligand binding domains of EphB2 and EphB4, which share only 44% sequence identity, show a similar jellyroll folding topology, with 13 antiparallel ␤-sheets connected by loops of varying length (18,19). The crystal structure of the EphB4 receptor in complex with the phage display derived TNYL-RAW peptide revealed that the peptide binds to the ephrin binding cavity of the receptor, effectively inhibiting interaction with the ligand (19). Furthermore, the structure of the EphB2⅐ephrinB2 complex showed that the ligand binding channel of the receptor is located at the upper convex surface of EphB2 and is formed by the flexible J-K, G-H, and D-E loops, which become ordered to accommodate the solvent-exposed ephrin G-H loop (10). A low affinity tetramerization interface was also identified at the concave surface of the receptor H-I loop, which interacts with the C-D loop of the ephrin.
Given that the EphB4⅐ephrinB2 interaction is important in angiogenesis and that EphB4 is overexpressed in several tumor types (1, 20 -22), modulating this protein-protein interaction is a potential approach to slowing tumor angiogenesis and tumor growth. In mouse models of breast cancer, high EphB4 expression correlates with increased malignancy and tumor aggressiveness (23)(24)(25). EphB4 expression is also increased in human primary infiltrating ductal breast carcinoma and is correlated to increased malignancy (26). There is evidence that the EphB4 ectodomain stimulates endothelial cell migration and proliferation, suggesting that ephrinB2-expressing endothelial cells interact with the EphB4 ectodomain to promote angiogenesis and tumor progression. Furthermore, a kinase-deficient EphB4 mutant has been shown to increase breast cancer cell growth, indicating that downstream forward kinase signaling is not an absolute requirement for tumorigenesis, at least in breast cancer cells (27). Several groups have more recently demonstrated that the full extracellular domain of EphB4 is indeed a viable therapeutic target. First, the soluble extracellular domain of EphB4 was described to block both forward and reverse signaling, resulting in an inhibition of tumor growth in vivo (28,29). Second, phage dis-play studies have identified a peptide (TNYL-RAW) that antagonizes the EphB4⅐ephrinB2 interaction in the high nanomolar range (30). However, a more complete understanding of the biological role of EphB4⅐ephrinB2 signaling in tumorigenesis and in forms of pathological angiogenesis is now required.
To enable anti-EphB4⅐ephrinB2 therapeutic development and probe EphB4 specificity, we determined the three-dimensional crystal structure of the EphB4⅐ephrinB2 complex to high resolution. In addition, we conducted site-directed mutagenesis and biophysical analyses to investigate the role of several residues within the ligand binding cavity of EphB4 in contributing to the binding of both ephrinB2 and the antagonistic TNYL-RAW peptide. These results will enable the development of predictive models for structure-based drug design of small molecule compounds for use as therapeutics and to probe the biology of EphB4⅐ephrinB2 bi-directional signaling.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The human EphB4 receptor  was expressed and purified in insect cells as described elsewhere (19). The wild type EphB4 construct was used as a template for the generation of site-specific mutants. The human ephrinB2 (extracellular domain, residues 25-187) was designed based on the previously published EphB2⅐ ephrinB2 structure and cloned into a modified pFastBac1 vector containing a GP67 leader peptide. Recombinant baculovirus was generated using the Bac-to-Bac system (Invitrogen). Briefly, large scale expression of ephrinB2 was carried out using Wave Bioreactors on a 5-liter scale at a multiplicity of infection of 5 for 48 h, resulting in ϳ10 mg of ephrinB2/liter of Hi-5 insect cells (Invitrogen). Media containing secreted ephrinB2 protein was concentrated and buffer was exchanged using a Hydrostart Crossflow Filter (Sartorius Edgewood, NY). The ligand was purified by immobilized metal affinity chromatography and cleaved overnight with tobacco etch virus protease. Material was further re-purified by immobilized metal affinity chromatography (S2000) to remove the protease and an N-terminal fragment containing the histidine tag. The EphB4⅐ephrinB2 complex was formed with a 1.5-fold molar excess of ephrinB2 overnight at 4°C in buffer containing 50 mM Tris, pH 7.8, 100 mM NaCl, and 10 mM Imidazole. The complex was purified by immobilized metal affinity chromatography followed by size exclusion chromatography to remove trace aggregates (Phenomenex, Torrance, CA). The final purity of the complex was greater than 95%.
Crystallization, Data Collection, and Structure Solution-The EphB4⅐ephrinB2 complex was concentrated to 20 mg/ml in a buffer containing 25 mM Tris, pH 7.8, 150 mM NaCl, and 5 mM CaCl 2 and crystallized by sitting drop vapor diffusion at 20°C against a precipitant of 2.2 M ammonium sulfate and 100 mM Tris, pH 7.8. EphB4⅐ephrinB2 crystals were cryoprotected in 25% glycerol and flash-cooled. Crystals formed in the P4 1 space group and contained one monomer of receptor and one monomer of ligand in the asymmetric unit. Data were collected at the Advance Photon Source (Argonne, IL) on beamline GM/CA-CAT. Images were processed and reduced using HKL2000 (31). The structure was solved by molecular replacement with MolRep (CCP4i), using the EphB2⅐ephrinB2 struc-ture (Protein Data Bank code 1KGY) as a search model (10,32). The structure was refined by a rigid body refinement followed by model building in O and iterative refinements with Refmac (32,33). The structure exhibits good geometry with no Ramachandran outliers.
Isothermal Titration Calorimetry-All mutants and ligands were dialyzed into buffer containing 50 mM Tris-Cl, pH 7.8, 150 mM NaCl, and 1 mM CaCl 2 before use in isothermal titration calorimetry (ITC) 2 experiments. All experiments were performed with a Microcal MCS ITC at 25°C. Titrations were completed as described (19). EphB4 (wild type or mutant) was present in the sample cell at a concentration of 12-15 M, and the injection syringe contained either 127 M ephrinB2 or 200 M TNYL-RAW. Titrations of TNYL-RAW with the L95R mutant of EphB4 were performed with 2 mM TNYL-RAW in the injection syringe and 15 M EphB4 (L95R) in the sample cell. Data for these titrations were fit assuming a stoichiometry of 1, and at least 60% saturation at the final peptide concentration as described (19,34).
Fluorescent Polarization Assay-Alexa-532 labeled TNYL-RAW peptide was synthesized by Biopetide (San Diego, CA). A serial dilution of EphB4 was prepared in assay buffer (50 mM Tris, pH 7.8, 150 mM NaCl, 1 mM CalCl 2 , 0.1% Pluronic 124). TNYL-RAW-Alexa-532 labeled peptide was prepared as a 100 M stock solution in the assay buffer, and a 300 nM working solution was made fresh before the measurements by dilution in the assay buffer. 5 l of serially diluted EphB4 (9 -2362 nM concentration range) was combined with 5 l of labeled peptide (final concentration 75 nM) in the final volume of 20 l in the absence and presence of 200 M TNYL-RAW as a control for nonspecific binding. The mixture was allowed to equilibrate for 30 min at room temperature, and measurements were performed with a Tecan Genios Pro (Tecan Instruments) using 535-nm excitation and 580-nm emission wavelength. All experimental data were analyzed using Prism 4.0 software (GraphPad Software Inc., San Diego, CA), and K d values were generated by fitting the experimental data using a onesite binding hyperbola nonlinear regression model. The calculated Z-factor for 108 samples, each at 2 different protein concentrations representing upper and lower plateaus of the dose response curve, is 0.715.

RESULTS
Overall Structure-EphB4 and ephrinB2 were co-concentrated to 20 mg/ml and crystallized by sitting drop vapor diffusion against a precipitant of 2.2 M ammonium sulfate and 100 mM Tris, pH 7.8, at 20°C. The co-crystal structure was refined to 2.0-Å resolution with an R-factor of 23% and a free R factor of 29% (Table 1; Fig. 1). Unlike crystals of the of the EphB2⅐ephrinB2 complex, which consisted of a heterotetramer, crystals of the EphB4⅐ephrinB2 complex consist of a heterodimer. Previously, formation of EphB2⅐ephrinB2 heterotetramers was observed for a concentration range around 1 mM using size exclusion chromatography analysis, whereas analytical ultracentrifugation demonstrated that the EphB2⅐ephrinB2 complex was a heterodimer at concentrations in the low micro-molar range (10). Our size exclusion chromatography analysis of the EphB4⅐ephrinB2 complex in a concentration range up to 500 M indicates that the EphB4⅐ephrinB2 complex exists as a heterodimer. 3 The overall structure of the EphB4⅐ephrinB2 complex is similar to that of the EphB2⅐ephrinB2 complex, with a root mean square deviation of 5.0 Å over 316 equivalent C␣ positions. Significant deviation is mostly evident, however, throughout the structure of the receptor loop regions compared with the EphB4-TNYL-RAW and EphB2⅐ephrinB2 structures, with root mean square deviations of 1.8 and 5.3 Å, respectively, in the J-K loop. The ephrinB2 ligand deviates minimally   (10,17). EphB4⅐EphrinB2 Interface-The high affinity EphB4⅐ ephrinB2 heterodimer is formed by insertion of the solventexposed ligand G-H loop into the upper convex and hydrophobic surface of the EphB4 receptor, positioned above receptor strands E and M. Hydrophobic contacts drive receptor-ligand binding in this region. Ligand (L) residues Phe-120L, Pro-122L, Trp-125L, and Leu-127L participate in van der Waals interactions with receptor residues lining the receptor binding cavity in the D-E, G-H, and J-K loops (Fig. 2). Phe-120L forms hydrophobic interactions with Leu-95R (R, receptor; see below), Leu-  (16). The ligand is depicted with all bonds shown, whereas receptor residues (cyan) are drawn schematically.

EphB4⅐EphrinB2 Complex Structure
100R, and Pro-101R, whereas Leu-124L interacts with Thr-147R from the receptor J-K loop. Meanwhile, Trp-125L extends to the surface of the receptor in between the J-K and G-H loops, participating in hydrophobic interactions with residues Leu-48R, Glu-50R, Val-159R, Leu-188R, and Ala-186R. In addition, Pro-122L, similar to all previous crystal structures, maintains its position by participating in a direct interaction with the receptor Cys-61R-Cys-184R disulfide bridge. Few polar contacts are formed at the receptor-ligand dimer interface. Ser-121L forms a side chain side-chain hydrogen bond with Glu-59R as well as a main chain-side chain hydrogen bond with Lys-149R, whereas Asn-123L forms a hydrogen bond with the main chain oxygen of Leu-48R. Additionally, Lys-149R extends to the body of the ephrinB2 G-H loop, forming side chain-side chain hydrogen bonds with Glu-128L, and side chain-main chain hydrogen bonds with Ser-121L and Asn-123L, which are both part of the high affinity ligand FSPN sequence (Fig. 3). The introduction of this new interaction at the EphB4⅐ephrinB2 interface is certain to contribute to the high affinity of this receptor-ligand complex.
Similar to the EphB2⅐ephrinB2 structure, a second portion of the high affinity heterodimerization interface exists immediately adjacent to the ligand binding cavity, formed by ligand strands C, G, and F, and receptor strands B-C, E, and D. This region of the complex deviates minimally from the corresponding EphB2⅐ephrinB2 complex, with a maximum deviation of 2.1 Å from furthest atoms, and is predominantly characterized by backbone-backbone, backbone-side chain, and side chain-side chain hydrogen bonds. In particular, side chain-side chain interactions between Glu-59R (Glu-68 in EphB2)-Gln-118L and Ser-121L, Asp-29R (Glu-40 in EphB2)-Lys-112L, and Glu-43 (Glu-52, EphB2)-Lys-116L provide the binding potential characteristic of this low nanomolar interaction. Side chain-main chain interactions between Ser-55 and Lys-116L, and between Glu-44R and Lys-60R complete the binding network in this region.
Although the overall shape of the EphB4⅐ephrinB2 interaction interface is in good agreement with that previously described in the EphB2⅐ephrinB2 structure, marked differences exist within the receptor loops that frame the ligand binding channel. The EphB4 J-K loop assumes a distinct position compared with previously described crystal structures and is situated directly above the ligand G-H loop and 15 Å from the D-E loop (Fig.  4). The corresponding J-K loop from the EphB2⅐ephrinB2 structure, on the other hand, is positioned only 6.4 Å from the D-E loop and, therefore, maintains a more compact binding cavity. In fact, the J-K loops differ in position by up to 10 Å from furthest positions between the two ephrinB2bound complex structures. Not surprisingly, the J-K loop shows remarkable flexibility in each structure described, also shifting in position by up to 20 Å from furthest positions between the EphB4-TNYL-RAW structure. Furthermore, crystallization trials with the apo form of EphB4 failed to produce crystals, likely because of the inherent flexibility of the J-K and D-E loops. A feature unique to EphB4 is a three-residue insert in the J-K loop, which is absent in all other Eph receptors. It has been speculated that this insert contributes to the ligand binding specificity inherent to the EphB4 receptor (35). Indeed, two (Pro-151, Gly-152) of the three residues (Pro-151, Gly-152, Ala-153) form the tip of the J-K loop and make contacts with the ligand; Pro-151R forms a hydrophobic contact with Phe-120L, whereas Gly-152R makes a main chain to side chain polar contact with Glu-152L. In addition, the G-H and D-E loops, which form two walls of the ligand binding cleft, also shift to accommodate the ligand. The G-H loop is shifted by over 4.5 Å between the EphB4 and EphB2-bound ephrinB2 structures, whereas the D-E loop only deviates by 1.5 Å between the two structures.
EphB4 Specificity-Sequence comparison and structural analysis of the EphB4 and EphB2 receptors suggested that one residue in EphB4 is particularly important in determining the specificity of the EphB4⅐ephrinB2 interaction: Leu-95. The corresponding residue in EphB2, Arg-103, is strictly conserved across both A and B subclasses and deviates only in the EphB4 receptor. Arg-103R participates in hydrogen bonds with residues from the high affinity ephrin G-H loop, including Ser-121L and Glu-128L, and is situated in proximity to Phe-120L, a residue critical for receptor binding. However, superposition of an Arg at position 95 in the EphB4⅐ephrinB2 structure suggests that a steric clash would result between the superposed arginine and Phe-120L. The corresponding Leu-95R, on the other hand, is able to form a 3.2 Å van der Waals interaction with Phe-120L due to its position within the ligand binding cavity. Not surprisingly, Arg-95R would also sterically clash with the phenylalanine from the TNYL-RAW peptide in the EphB4-TNYL-RAW structure (19), whereas the smaller Leu-95R forms favorable contacts with the peptide. Interestingly, the highly conserved Phe-120L is shifted in position by ϳ90°as compared with previous complex structures (8,10,19) and is buried within the hydrophobic cleft of the receptor, unlike its position in the EphB2⅐ephrinB2 complex structure, where it is directed toward the surface (Fig. 5). In addition, the position of Arg-103R requires the J-K loop of the EphB2 receptor to extend away from the ligand G-H loop and toward the receptor D-E loop to FIGURE 3. Stereoview of A weighted 2 F obs ؊ F calc electron density at 2. 0 Å resolution, contoured at 1 for the EphB4⅐ephrinB2 interface. EphB4 is in green, and ephrinB2 is in yellow. Clear density of the interface shows Phe-120 in a novel position with respect to previously described structures in order to interact with Leu-95. avoid steric interference with residues lining the ephrinB2 G-H loop. The smaller Leu-95R together with Phe-120L allows the J-K loop of EphB4 to adopt a novel position directly above the ligand G-H loop.
Biophysical Characterization of EphB4 Specificity; Enthalpic Versus Entropic Contributions-A series of site-specific mutations was generated by changing residues lining the EphB4 G-H and J-K loops to the corresponding residues found in EphB2 ( Table 2). The EphB4 mutants were analyzed based on their binding to fluorescently labeled Alexa-532-TNYL-RAW peptide. Fluorescence polarization (FP) analysis corroborated the prediction that Leu-95 is a critical determinant for binding of the TNYL-RAW peptide because the L95R mutant did not exhibit significant binding of the peptide in our assay. EphB4 mutants T147F, A186S, and K149Q showed an ϳ4 -5-fold reduction in binding affinity of the fluorescently labeled peptide.
Based on the structural information and preliminary binding characterization by fluorescence polarization analysis, two EphB4 mutants, L95R and K149Q, were chosen for a more detailed thermodynamic analysis of their binding to both ephrinB2 and the TNYL-RAW peptide ligand using ITC. As reported previously, EphB4 binds to ephrinB2 with an affinity of 40 nM and a ⌬H obs of 3.3 kcal mol Ϫ1 (19). Mutation of EphB4 Lys-149 to Gln has no effect on the binding affinity or enthalpy of ephrinB2 binding (Table 3). In contrast, mutation of EphB4 Leu-95 to Arg reduces the binding affinity of ephrinB2 by nearly 2 orders of magnitude. Binding of ephrinB2 to all forms of EphB4 is endothermic, and the binding of ephrinB2 is more endothermic with the L95R mutation in EphB4 (5.2 versus 3.3 kcal mol Ϫ1 for wild type EphB4). Preliminary experiments carried out in a buffer with different enthalpy of ionization showed a similar enthalpy change to that reported here. For example, binding of ephrinB2 to EphB4 (K149Q) resulted in a ⌬H obs of 3.9 (Ϯ0.1) kcal mol Ϫ1 in phosphate (⌬H ion ϭ 0.8 kcal molϪ1) compared with the ⌬H obs of 3.7 (Ϯ0.2) kcal mol Ϫ1 value obtained in Tris (⌬H ion ϭ 11.34 kcal mol Ϫ1 ) ( Table 3). Thus, it appears that protonation/deprotonation is not coupled to ephrinB2 binding under the conditions of the ITC experiments.
Binding of the TNYL-RAW peptide to the wild type, K149Q, and L95R forms of EphB4 was also monitored by ITC. TNYL-RAW binds to EphB4 with an affinity of 70 nM and a ⌬H obs of Ϫ14.7 kcal mol Ϫ1 (19). In contrast to the different effects of mutations in EphB4 on the interaction of EphB4 with ephrinB2, mutation of EphB4 of either Lys-149 to Gln or Leu-95 to Arg reduces the affinity of the EphB4-TNYL-RAW interaction (3-and 500-fold, respectively; Table 3). Binding of the TNYL-RAW peptide to all three forms of EphB4 is characterized by an exothermic enthalpy.
Thus, thermodynamic analysis reveals that TNYL-RAW binding to the EphB4 ligand binding domain is an enthalpically driven process, whereas ephrinB2 binding to EphB4 is an entropically driven process. The differences in the binding thermodynamics are consistent with the available structural information. Burial of the hydrophobic ligand G-H loop within the hydrophobic receptor binding cleft should entropically drive the interaction through the release of water,

EphB4⅐EphrinB2 Complex Structure
increasing the solvent entropy. In addition, the ephrinB2 ligand G-H loop is quite rigid, both in apo and receptor-bound structures, minimizing massive conformational entropy losses. The small loss of conformational entropy counteracts the heterodimerization process by ordering the otherwise flexible receptor J-K, D-E, and G-H loops. Unlike ephrinB2, however, the free peptide ligand will lose significant conformational degrees of freedom upon EphB4 binding, resulting in an overall entropy loss. This is compensated by an enthalpy gain due to the formation of favorable interactions, both polar and nonpolar, at the receptor-peptide interface. A complete structurebased thermodynamic analysis is precluded, however, due to our inability to crystallize apo EphB4.
It should be noted that we produced the ephrinB2 extracellular domain in insect cells in a glycosylated form, whereas the ephrins used for previous crystal structure determinations were produced in Escherichia coli and, therefore, not glycosylated. A conserved glycosylation site exists in ephrinB2 at residue Asn-39, in proximity to the low affinity tetramerization interface. Consistent with its possible glycosylation, Asn-39 is located near the surface of the protein, and its side chain extends toward the surface of the complex. Although the carbohydrate was not observed in our electron density map, likely because it was disordered, it is conceivable that a sugar at this location could interfere with the formation of receptor-ligand tetramers in the crystal lattice. However, previous reports have suggested that carbohydrate moieties would play more a favorable than an unfavorable role in tetramerization (36).

DISCUSSION
We have determined the three-dimensional crystal structure of the EphB4 receptor ligand binding domain in complex with the extracellular domain of its ligand, ephrinB2, and identified the determinants for EphB4 specificity. A multiple sequence alignment with members of the EphB subclass reveals that the EphB4 receptor lacks a conserved arginine and instead contains a leucine at position 95. A L95R mutation was previously predicted to result in steric interference with the antagonistic TNYL-RAW peptide ligand (19). Here we report that this mutation would also result in steric clashes with Phe-120L in the G-H loop of ephrinB2 due to the unique positioning of the J-K loop of EphB4. A leucine instead of an arginine at position 95 of the EphB4 receptor indeed appears to be sufficient to cause a substantial structural rearrangement of the receptor J-K loop. Of particular interest is the novel position of the conserved Phe-120L in the high affinity FSPN sequence of the ephrinB2 G-H loop. Although ephrinB2 is conserved in structure in both receptor-bound and apo structures, there appears to be variability within the rigid G-H loop to conform to a specific receptor.
EphrinB2 shares 46 and 40% sequence identity with ephrinB1 and ephrinB3, respectively. EphB4 binds only weakly to both of these ligands while exhibiting high affinity for ephrinB2. Considering the B-subclass ephrin G-H loop (ephrinB1-B3), it is interesting to speculate on why EphB4 preferentially binds ephrinB2 over other B-subclass ligands. EphrinB1 shares significant sequence identity with the high affinity ephrinB2 G-H loop, except at position 124, which is a Tyr in ephrinB1 and a Leu in ephrinB2. Although Leu-124L does not form substantial interactions with EphB4, the small size of the leucine allows tight packing within the receptor binding cavity and maintains the hydrophobic nature of the binding cleft. Superposition of a tyrosine on the ephrinB2 structure would require the rearrangement of the EphB4 J-K loop to accommodate the bulky tyrosine, likely accounting for the reduced affinity of EphB4 for ephrinB1. The ephrinB3 G-H loop is also very similar to the ephrinB2 G-H loop but deviates in the FSPN sequence, which contains a tyrosine instead of the phenylalanine (YSPN). Phe-120L forms critical interactions with residues lining the EphB receptor-ephrinB2 binding cavity in the three complex crystal structures thus far described (8,10,19). In the previous crystal structures Phe-120L extends to the surface of the binding cavity, adjacent to the receptor G-H loop. Thus, superposition of a tyrosine on the EphB2⅐ephrinB2 structure would not affect the dynamics of the ligand binding cavity, and this residue would likely interact with several water molecules on the surface of the complex. However, in our structure the Phe-120L of ephrinB2 is observed in a novel position, buried within the hydrophobic binding cleft and forming interactions with Leu-95R and the Cys-61R-Cys-184R disulfide bridge. Insertion of a tyrosine at this position would, therefore, result in both steric clashes within the receptor binding cavity and a polar redistribution of the active site, consistent with the weak affinity of EphB4 for ephrin-B3.
Insights into the Modulation of EphB4⅐EphrinB2 Interaction-Unlike the heterotetrameric EphB2⅐ephrinB2 crystal structure, EphB4 and ephrinB2 only form heterodimers, both in the cry-  stals and in solution. Interestingly, this is most similar to the EphB2⅐ephrinA5 co-crystal structure than the more related EphB2⅐ephrinB2 structure. Our studies suggest that higher order oligomerization may require regions outside of the ligand binding domain, which is consistent with previous reports (14,37). Mutagenesis studies have demonstrated that biologically active Eph-ephrin multimeric complexes are dependent on three interfaces; the high affinity dimerization interface, the tetramerization interface, and a third interface in the cysteine-rich region, which is immediately adjacent to the ligand binding domain and previously reported to mediate receptor-ligand low affinity interaction. The thermodynamic discrepancies between peptide and ephrin binding should be considered in the design of therapeutics to treat disease related to the Eph receptor family. Iterative rounds of structure-based drug design will require an understanding of the enthalpic and entropic contributions of small molecule compounds. The ephrin ligand, with entropically driven binding, interacts with multiple members of the EphB family. In contrast, the TNYL-RAW peptide, with enthalpically driven binding, is a specific inhibitor of the EphB4⅐ephrinB2 interaction. The specific nature of the enthalpically driven peptide binding is consistent with the observation made by the group of Ernesto Freire concerning the selectivity of binding of human immunodeficiency virus-1 protease inhibitors (15). Further studies are needed to determine whether the correlation between the specificity of the interaction and the contribution of the enthalpic component to the total binding energy is a coincidence or a more general property of the interaction of inhibitors.
We have characterized the interaction between the EphB4 ligand binding domain and the extracellular domain of eph-rinB2 both structurally and biophysically. We now better understand the structural determinants that confer the high affinity and specificity of EphB4 for ephrinB2 and the role of the critical EphB4 residue, Leu-95. Structure-based design of EphB4 and ephrinB2 mutants will provide more precise tools for understanding the mechanisms of tumorigenesis stimulated by their interaction. The characterization of enthalpic versus entropic contributions in EphB4 ligand binding using ITC complements the structure-based approach to compound and ligand design. Our results combined with the information from other related Eph receptor-ephrin complex structures offer the possibility to rationally design small molecule compounds and bio-therapeutics with improved pharmacokinetic properties that would antagonize the EphB4⅐ephrinB2 interaction.