Epiregulin Recognition Mechanisms by Anti-epiregulin Antibody 9E5

Abstract

Epiregulin (EPR) is a ligand of the epidermal growth factor (EGF) family that upon binding to its epidermal growth factor receptor (EGFR) stimulates proliferative signaling, especially in colon cancer cells. Here, we describe the three-dimensional structure of the EPR antibody (the 9E5(Fab) fragment) in the presence and absence of EPR. Among the six complementarity-determining regions (CDRs), CDR1–3 in the light chain and CDR2 in the heavy chain predominantly recognize EPR. In particular, CDR3 in the heavy chain dramatically moves with cis-trans isomerization of Pro¹⁰³. A molecular dynamics simulation and mutational analyses revealed that Arg⁴⁰ in EPR is a key residue for the specific binding of 9E5 IgG. From isothermal titration calorimetry analysis, the dissociation constant was determined to be 6.5 nm. Surface plasmon resonance analysis revealed that the dissociation rate of 9E5 IgG is extremely slow. The superimposed structure of 9E5(Fab)·EPR on the known complex structure of EGF·EGFR showed that the 9E5(Fab) paratope overlaps with Domains I and III on the EGFR, which reveals that the 9E5(Fab)·EPR complex could not bind to the EGFR. The 9E5 antibody will also be useful in medicine as a neutralizing antibody specific for colon cancer.

Introduction

Recently, antibody therapy has been attracting considerable attention as a possible cure for several types of diseases. For instance, trastuzumab is a humanized IgG1κ monoclonal antibody that is targeted for the human epidermal growth factor (EGF) receptor (EGFR)⁶ 2 (HER2, ErbB-2), which is used in the treatment of metastatic breast cancer (1).

Initially, the EPR precursor protein is expressed as a type I transmembrane protein. A disintegrin and metalloproteinase 17 (ADAM17) catalyzes ectodomain shedding of the EPR precursor protein, which produces mature EPR (2). EPR induces dimerization of EGFR and promotes autophosphorylation in the intracellular kinase domain of EGFR (3). EGFR phosphorylation activates several types of intracellular signaling pathways, such as the mitogen-activated protein kinase (MAPK), phosphoinositide 3-kinase (PI3K)/Akt, and STAT5 pathways (4–6). As a result, proliferation, cell survival, and angiogenesis are induced in the cell.

Although the expression of EPR is suppressed in most adult normal tissues, EPR is overexpressed in human colon, breast, and ovarian cancers (7–10). Therefore, normalization of EGF signaling is expected to cure these cancers. Recently, humanized anti-EPR antibodies with high affinity targeted cytotoxicity have been prepared and characterized (11), and these antibodies have the potential to act as anticancer drugs.

The structure of EPR was first determined by NMR (12). Similar to the other EGF family ligands, EPR (residues Val¹–Leu⁴⁶) is composed of an N-terminal domain (residues Ile³–Glu³³) that has a β-hairpin motif called the core region (residues Gly¹⁷–Cys³²) and a C-terminal domain (residues Val³⁴–Phe⁴⁵). Three disulfide bridges stabilize the entire EPR structure. For the EGF family antibody ligand, the structures of transforming growth factor β complexed with Fab or single chain Fv of fresolimumab have been reported (13).

To design an effective humanized antibody, we investigated the antibody recognition mechanism between mature EPR and the 9E5(Fab) fragment by x-ray structural analysis. In this study, we describe the three-dimensional structure of the 9E5(Fab) fragment with and without EPR. Moreover, a molecular dynamics (MD) simulation, isothermal titration calorimetry (ITC), and surface plasmon resonance (SPR) analysis were performed to clarify the structure-function relationship. These findings are expected to aid in the development of future drugs, especially those that target cancers.

Experimental Procedures

Production and Purification of 9E5(Fab)

The 9E5 monoclonal antibody was produced using a method described previously (11). Hybridoma cells were intraperitoneally implanted in BALB/c nude mice (BALB/cSlc-nu/nu), and ascites were obtained from the mice and examined with a Bio-Scale Mini UNOsphere SUPrA cartridge (Bio-Rad). The peak fractions were injected into a Bio-Scale Mini Bio-Gel P-6 (Bio-Rad).

To prepare 9E5(Fab), the Fc fragments of 9E5 IgG released by papain digestion (9E5 IgG:papain, 100:1) were used. The digested samples were loaded onto a Bio-Scale CHT5-I column (Bio-Rad) and eluted with a linear gradient of 0.5–250 mm sodium phosphate buffer (pH 6.8). The peak fractions were collected and concentrated, and they were then injected onto a HiLoad 16/60 Superdex 75 prep grade column (GE Healthcare), which was developed with 20 mm Tris-HCl (pH 7.5) buffer containing 300 mm NaCl. The peak fractions containing 9E5(Fab) were collected and concentrated to 10 mg ml⁻¹ by ultrafiltration with Vivaspin (10-kDa cutoff; GE Healthcare).

Construction of the EPR Expression Plasmids

We constructed EPR from Homo sapiens (hEPR) and Mus musculus (mmEPR) pro-EPR cDNA (residues 1–46), which is elongated by 24 residues toward the N terminus (residues −23 to 46) to improve its fusibility. The EPR gene was cloned into a modified pET32a vector (Novagen, Billerica, MA), which was in-frame with a hexahistidine tag, thioredoxin, and the HRV3C protease cleavage site at the N terminus.

Site-directed mutagenesis was performed with PCR mutagenesis. In hEPR, the following oligonucleotide primer pairs were used (the mutated sites are underlined): D9A forward, 5′-TCACCAAATGTTCTAGCGCAATGAATGGTTATTGTCT-3′; D9A reverse, 5′-AGACAATAACCATTCATTGCGCTAGAACATTTGGTGA-3′; S26R forward, 5′-GTATCTATCTGGTTGACATGCGTCAGAATTATTGTCGTTGCGA-3′; S26R reverse, 5′-TCGCAACGACAATAATTCTGTGCCATGTCAACCAGATAGATAC-3′; R40A forward,5′-TCGGTTACACCGGCGTCGCATGCGAGCACTTCTTCCT-3′; and R40A reverse, 5′-AAGAAGTGCTCGCATGCGACGCCGGTGTAACCGA-3′. In mmEPR, the following designed primer pairs were used: R26S forward, 5′-TATCTACCTGGTCGATATGTCTGAGAAATTCTGTCGTTGTG-3′; R26S reverse, 5′-CACAACGACAGAATTTCTCAGACATATCGACCAGGTAGATA-3′; E27Q/K28N/F29Y forward, 5′-CTACCTGGTCGATATGCGTCAGAACTACTGTCGTTGTGAGGTTGGTT-3′; and E27Q/K28N/F29Y reverse, 5′-AACCAACCTCACAACGACAGTAGTTCTGACGCATATCGACCAGGTAG-3′.

Expression and Purification of Recombinant EPRs

Escherichia coli SHuffle T7 cells (New England Biolabs, Ipswich, MA) were transformed with the prepared plasmids. The cells were cultured in lysogeny broth containing 100 μg ml⁻¹ ampicillin at 37 °C until the optical density at 600 nm reached 0.6. The temperature was lowered to 15 °C, and then 0.4 mm isopropyl 1-thio-β-d-galactopyranoside was added to induce protein expression. After 24 h of cultivation, the cells were collected and stored at −80 °C until further use.

The cells were thawed and disrupted with an EmulsiFlex-C3 homogenizer (Avestin Inc., Ottawa, Canada) in 20 mm Tris-HCl buffer (pH 8.0) containing 500 mm NaCl, 20 mm imidazole, and 2500 units of Benzonase. After removal of the cell debris by centrifugation, the supernatant was applied to an nickel-nitrilotriacetic acid Superflow (Qiagen, Hilden, Germany) column and eluted with 20 mm Tris-HCl buffer (pH 8.0) containing 500 mm NaCl and 500 mm imidazole. HRV3C protease was added to the eluate, and it was dialyzed against dialysis buffer (20 mm Tris-HCl (pH 7.5) containing 600 mm NaCl). To remove the HRV3C protease and uncleaved fusion proteins, the dialyzed sample was applied to GS Trap and His Trap columns (GE Healthcare), and the flow-through fraction was recovered. The sample was concentrated and loaded onto a gel filtration chromatograph with a Hi-Load 16/60 Superdex 75 prep grade column, which was developed with the dialysis buffer. The fractions containing the EPR protein were buffer-exchanged into 20 mm Tris-HCl (pH 7.5) containing 300 mm NaCl and concentrated to 10 mg ml⁻¹.

X-ray Crystallography

The entire crystallization was performed with the sitting drop vapor diffusion method with a VIORAMO 96-well protein crystallization plate (Azone, Edobori, Osaka, Japan). For the crystallization of 9E5(Fab), 0.5 μl of protein solution (10 mg ml⁻¹ 9E5(Fab), 20 mm Tris-HCl (pH 7.5), and 300 mm NaCl) was mixed with 0.5 μl of reservoir solution (50 mm HEPES-Na (pH 7.3) and 21.5% (v/v) polyethylene glycol (PEG) 4000) and incubated at 20 °C. Crystals of 9E5(Fab) formed within 7 days. For x-ray data collection, a 9E5(Fab) crystal was soaked in cryoprotectant solution (50 mm HEPES-Na (pH 7.3), 24% (v/v) PEG 4000, and 10% (v/v) glycerol) and flash frozen in liquid nitrogen.

For crystallization of the 9E5(Fab)·hEPR complex, 0.5 μl of protein solution (10 mg ml⁻¹ 9E5(Fab)·hEPR, 20 mm Tris-HCl (pH 7.5), and 300 mm NaCl) was mixed with 0.5 μl of reservoir solution (100 mm MES monohydrate (pH 6.0) and 14% (v/v) PEG 4000) at 20 °C. The 9E5(Fab)·hEPR complex crystal formed within 7 days. For data collection, a 9E5(Fab)·hEPR crystal was soaked in cryoprotectant solution (100 mm MES monohydrate (pH 6.0), 17% (w/v) PEG 4000, and 20% (v/v) glycerol) and then flash frozen in liquid nitrogen.

The x-ray diffraction data sets for the 9E5(Fab) and 9E5(Fab)·hEPR complex crystals were collected at Photon Factory BL-5A and SPring-8 BL44XU, respectively. The diffraction data were integrated and scaled with HKL2000 (14). The structure of 9E5(Fab) was determined by the molecular replacement method using 82D6A3, which is an antithrombotic antibody (15) (Protein Data Bank code 2ADF), as the starting model with PHASER (16). To determine the 9E5(Fab)·hEPR complex structure, molecular replacement was performed with PHASER using the refined 9E5(Fab) structure and the NMR structure of hEPR (Protein Data Bank code 1K37) as the search models (12). Model building was performed using Coot (17), and the structure was refined using REFMAC5 (18) and PHENIX (19); 5% of the reflections were set aside for R_free calculations (20). The quality of the models was assessed with Ramachandran plots, and model geometry analyses were conducted with Rampage (21). All of the structural figures were drawn with PyMOL (22). The data collection and refinement statistics are summarized in Table 1.

TABLE 1.

Data collection and refinement statistics for 9E5(Fab) and the 9E5(Fab)·EPR complex

	9E5(Fab)	9E5(Fab)·hEPR complex
Data collection
X-ray source	Photon Factory BL5a	SPring-8 BL44XU
Detector	ADSC Quantum 210r	Rayonix MX-225HE
Wavelength (Å)	1.0000	1.0000
Space group	P21	I222
Unit-cell parameters (Å; °)	a = 41.00, b = 79.83, c = 59.98; β = 92.59	a = 68.60, b = 100.29, c = 187.37
Resolution range (Å)	50.00–1.60 (1.66–1.60)	50.00–2.50 (2.59–2.50)
Total no. of reflections	192,615	100,371
No. of unique reflections	48,127	22,299
I/σ(I)	16.2 (2.8)	28.9 (2.9)
Redundancy	4.0 (3.6)	4.5 (3.0)
Completeness (%)	95.3 (98.8)	97.3 (93.4)
Rmergea (%)	6.9 (34.9)	5.1 (30.9)
Refinement statistics
Resolution (Å)	17.27–1.60	14.91–2.50
No. of reflections	46,334	22,005
Rcrystb/Rfreec (%)	19.2/23.7	18.9/26.3
Mean Wilson B value (Å2)	21.2	59.6
No. of non-H-atoms	3,627	3,660
r.m.s.d. from ideal values
Bond lengths (Å)/angles (°)	0.007/1.181	0.009/1.329
Ramachandran statistics (%)
Favored	97.2	93.5
Allowed	2.8	5.9

^a R_merge = Σ_jΣ_h|I_hj − I_h|/Σ_jΣ_hI_h.

^b R_cryst = Σ||F_o| −|F_c||/Σ|F_o| calculated from 95% of the data, which were used during the course of the refinement.

^c R_free = Σ||F_o| − |F_c||/Σ|F_o| calculated from 5% of the data, which were used during the course of the refinement.

Molecular Dynamics Simulations

All of the simulations were performed with the GROMACS 4.6.1 package (23–25) using the Fuji force field (26) for proteins, AMBER force field for ions, and TIP3P water potential. Na⁺ and Cl⁻ ions were added to produce a neutral solution of 0.15 m. The Nosé-Hoover thermostat (27, 28) with a relaxation time of 1 ps was used to keep the solutions at 298 K. The Parrinello-Rahman scheme (29) was used as a barostat at 1 atm with a relaxation time of 1 ps. The simulation time step was 3 fs, and all of the bond lengths of the proteins were constrained using the LINCS algorithm (30). The leap-frog algorithm was used to integrate the equations of motion, and the particle mesh Ewald method (31) was used to calculate the electrostatic interactions with a real space cutoff of 1.0 nm. The neighbor list cutoff was also set at 1.0 nm. The initial structure was taken from our x-ray crystal structure of the complex. After energy minimization, the heavy atoms of the protein were restrained for 200 ps using a harmonic potential with a force constant of 1000 kJ⁻¹ nm⁻² to relax the water molecules. Four NPT (constant number of particles, pressure, and temperature) simulations were then performed for 1 s with initial random velocities that obeyed a Maxwell-Boltzmann distribution at 298 K.

Isothermal Titration Calorimetry

Thermodynamic analyses of the interaction between EPR and 9E5 IgG were performed with an iTC200 calorimeter (GE Healthcare). In the calorimeter cell experiment, 9E5 IgG was placed in phosphate-buffered saline (10 mm phosphate buffer (pH 7.4), 150 mm NaCl, and 45 mm KCl) at a concentration of 5 μm, and it was titrated with 100–130 μm EPR solution in the same buffer at 25 °C. The EPR solution was injected 25 times. The thermograms were analyzed with Origin 7 software (GE Healthcare) after subtracting a thermogram measured against only the buffer. The enthalpy change (ΔH) and binding constant (K_A) for the interaction were directly obtained from the experimental titration curve fitted to a one-site binding isotherm. The dissociation constant (K_D) was calculated as 1/K_A. The Gibbs free energy change (ΔG = −RTln K_A) and the entropy change (ΔS = (−ΔG + ΔH)/T) for the association were calculated from ΔH and K_A.

SPR Analysis

SPR was carried out to analyze the interaction between 9E5 IgG and hEPR in a Biacore T100 system. Thioredoxin-fused hEPR was immobilized by an amine coupling method at a level of about 124 resonance units on a CM5 sensor chip (GE Healthcare). The binding of 9E5 IgG to hEPR was accomplished by injecting increasing concentration of 9E5 IgG (3.1–50 nm) into the sensor chip under the buffer condition of HEPES-buffered saline with surfactant P20 (pH 7.4) at a flow rate of 30 ml min⁻¹ at 25 °C. The data were corrected by subtracting the responses from a blank flow cell in which an amine coupling reaction was carried out. The kinetic parameters and the binding affinity were calculated using the bivalent analyte model with Biacore T100 evaluation software (GE Healthcare).

Results

Complex Structure of 9E5(Fab)·hEPR

We determined the structure of the complex of 9E5(Fab) with hEPR at 2.5-Å resolution (Fig. 1 and Table 1). The asymmetric unit contained one 9E5(Fab)·hEPR complex in a rectangular box with approximate dimensions of 35 × 45 × 90 Å. The interaction between 9E5(Fab) and hEPR formed a solvent-accessible surface of ∼919 Ų, which is in the typical range of interaction surfaces between antibodies and antigens (32).

FIGURE 1.

Overall structure of the 9E5(Fab)·hEPR complex. A, front view of the 9E5(Fab)·hEPR complex. hEPR and the heavy and light chains of 9E5(Fab) are colored pink, green, and cyan, respectively. B, top view of the complex structure. The black dotted squares in A and B show the locations of interaction-1 (enlarged in C), interaction-2 (enlarged in D), and interaction-3 (enlarged in E). C, interaction between CDR-L1, CDR-L3, and CDR-H2 in 9E5(Fab) and the N-terminal domain of hEPR (interaction-1). D, interaction between CDR-H1 and CDR-H2 in 9E5(Fab) and the C-terminal region of hEPR (interaction-2). E, interaction between CDR-L2 and CDR-H3 in 9E5(Fab) and the core region of hEPR (interaction-3). In C, D, and E, oxygen, nitrogen, and sulfur atoms are shown in red, blue, and yellow, respectively. Hydrogen bonds and salt bridges are shown as black dashed lines.

All six CDRs in 9E5(Fab) (CDR-L1, CDR-L2, and CDR-L3 in the light chain and CDR-H1, CDR-H2, and CDR-H3 in the heavy chain) interacted with hEPR and formed 27 hydrogen bonds or salt bridges as shown in Table 2 with numerous van der Waals interactions. The N-terminal domain of hEPR is recognized by CDR-L1, CDR-L3, and CDR-H2 (Fig. 1C). The C-terminal domain of hEPR is stabilized by CDR-H1 and CDR-H2 (Fig. 1D). The core region of hEPR (Gly^E17–Cys^E32; superscript E refers to epiregulin) interacts with CDR-L2 and CDR-H3 (Fig. 1E).

TABLE 2.

Hydrogen bonds and salt bridges between 9E5(Fab) and hEPR (distance <3.5 Å)

Interaction-1, -2, and -3 correspond to the regions shown by Fig. 1, C, D, and E, respectively.

Interaction part	hEPR		9E5(Fab)			Interaction distance
	Residue	Atom	Residue	CDR	Atom
						Å
Interaction-1	Lys5	O	Tyr32	L1	Oη	3.1
	Cys6	O	Tyr32	L1	Oη	3.5
	Ser7	Oγ	Tyr91	L3	O	2.5
	Ser7	Oγ	Asp92	L3	O	3.1
	Ser8	N	Asp92	L3	O	3.2
	Ser8	Oγ	Asp92	L3	Oδ1	3.2
	Asp9	N	Asp92	L3	O	3.1
	Asp9	Oδ1	Leu94	L3	N	3.2
	Asp9	Oδ2	Arg95	L3	Nϵ	3.2
	Asp9	Oδ2	Arg95	L3	Nη2	3.3
	Asp9	Oδ1	Arg50	H2	Nη1	3.0
	Asp9	Oδ2	Arg50	H2	Nη1	3.5
	Asp9	O	Arg50	H2	Nη1	3.4
	Asp9	O	Arg50	H2	Nη2	3.4
Interaction-2	Arg40	Nη1	Lys30	H1	O	3.3
	Arg40	Nη1	Asp52	H2	Oδ1	2.8
	Arg40	Nη2	Asp52	H2	Oδ2	3.5
Interaction-3	Tyr13	Oη	Gly101	H3	N	3.3
	Tyr21	Oη	Tyr91	L3	Oη	2.6
	Val23	O	His49	L2	Nϵ2	2.7
	Ser26	Oγ	His49	L2	Nδ1	3.5
	Ser26	Oγ	His55	L2	Nδ1	3.1
	Gln27	Nϵ2	Gly101	H3	O	2.4
	Gln27	Nϵ2	Asp102	H3	N	3.4
	Gln27	Nϵ2	Asp102	H3	Oδ1	3.0
	Asn28	N	Gly101	H3	O	2.8

We also solved the crystal structure of the 9E5(Fab) fragment at a resolution of 1.6 Å. The superimposition of 9E5(Fab) of the complex on the 9E5(Fab) structure showed relatively small root mean square deviation (r.m.s.d.) values of 0.9 (Fv domain) and 0.7 Å (C_L and C_H1 domains). The core region of hEPR (residues Gly¹⁷–Cys³²) also superimposed well on the NMR structure (12) (r.m.s.d., 1.0 Å). Although no dynamic movement of the Fv domain and the core region of hEPR occurs, conformational changes occur in the N- and C-terminal domains in hEPR and in CDR-H3 in 9E5(Fab) (Fig. 2 and Table 2).

FIGURE 2.

9E5(Fab) fragment superimposed on the 9E5(Fab)·hEPR complex. A, top view of the 9E5(Fab) fragment superimposed on the 9E5(Fab)·hEPR structure. The 9E5(Fab) fragment is shown in the lighter shade. The hEPR molecule in the 9E5(Fab)·hEPR complex is omitted in this figure. Drastic change of CDR-H3 is highlighted in orange. The r.m.s.d. of the Fv domain (residues 2–117 in the heavy chain and residues 2–14 and 17–104 in the light chain) is 0.9 Å. The noticeable interaction (interaction-3) is indicated by the black dashed square. B, close-up view of the superimposed structures of 9E5(Fab) with (solid colors) and without (faint colors) hEPR. 9E5(Fab) (upper panel) and 9E5(Fab)·hEPR (middle panel) show an electron density map around CDR-H3 and CDR-L2. Each 2F_o − F_c electron density map is contoured at 1 σ. In B, the orientation corresponds to Fig. 1E. Bottom panel, 9E5(Fab) fragment superimposed on the 9E5(Fab)·hEPR complex at interaction-3 region.

Movement of CDR-H3 Induced by hEPR Binding

Interaction-3 is composed of the interaction of CDR-L2 and CDR-H3 with the core region of hEPR (Tyr^E13, Tyr^E21–Val^E23, and Ser^E26–Asn^E28) (Fig. 1E). In interaction-3, little conformational change occurs in hEPR. However, drastic conformational changes occur in CDR-H3 (Fig. 2 and Table 2). The r.m.s.d. value of the Cα atoms of CDR-H3 (Arg^H98–Pro^H103; superscript H refers to the heavy chain) in the presence or absence of hEPR is 2.4 Å, which is about 2.5 times larger than that of the variable region of the heavy chain (0.9 Å). Asp^H102 is originally hydrogen-bonded to His^L49 (superscript L refers to the light chain) in the apo form. However, the hydrogen bond breaks with the insertion of hEPR, resulting in flipping of Asp^H102 and formation of a new salt bridge with Arg^H98. The Cγ carbon in the carboxyl group of Asp^H102 moves more than 10.8 Å, and the Cα atoms of Gly^H101 move more than 6.5 Å. A conformational change from cis-Pro^H103 to trans-Pro^H103 also occurs upon binding with hEPR. Gly^H100, Gly^H101, Asp^H102, and Pro^H103 in CDR-H3 form six hydrogen bonds with Tyr^E13, Gln^E27, and Asn^E28 (Table 2).

Calculation of the Interaction Energy by Molecular Dynamics

The r.m.s.d. values of hEPR and the Fv part of 9E5(Fab) were compared with the x-ray crystal structure, and the block average of the total energy was calculated from four MD simulations (Fig. 3). The block average was calculated within each 1.5-ns period. Because the system seemed to have reached equilibrium after about 700–800 ns (Fig. 3), the binding interactions were analyzed for the trajectories from 900 ns to 1 μs. The interaction energy is defined here as the sum of the short range Lennard-Jones (r < 0.9 nm) and coulombic (r < 1.0 nm) interactions between the residue pairs, which are the dominant contributions to the binding of hEPR to 9E5(Fab).

FIGURE 3.

Results from MD simulations of the 9E5(Fab)·hEPR complex. A, r.m.s.d. of the backbone of hEPR and the Fv part of 9E5(Fab) between the simulated structures and the x-ray crystal structure. The r.m.s.d. was averaged for the four simulations. B, block averages of the total energy averaged for the four simulations. The block averages were calculated within each 1.5-ns period. The error bars indicate the standard errors of the block averages of the four total energies.

Fig. 4 shows the interaction energies of each hEPR residue, which are 100-ns time averages of the equilibrated structures in solution. The solvated structures differed a little from the crystal structure. For example, Table 2 shows that the interaction distance between the carboxyl oxygen atom of Cys^E6 of hEPR and Oη of Tyr^L32 of 9E5(Fab) is 3.5 Å in the crystal structure. Although the MD structure provided the shortest O (Cys^E6)-OH (Tyr^L32) distance of 2.6 Å, the longest and time-averaged distances were calculated to be 4.0 and 6.8 Å, respectively. This explains why the interaction energy of Cys^E6 is small in Fig. 4. Asp^E9 and Arg^E40 interact with several atoms as shown in Table 2, and all of the distances are greater than 3.0 Å except for Nη1 (Arg^E40)-Oδ2 (Asp^H52) (2.8 Å) (Table 2). However, these residues have the strongest and second strongest interaction energies with 9E5(Fab) in Fig. 4: −204.1 kJ/mol for Asp^E9 and −147.0 kJ/mol for Arg^E40. These residues are located in the regions of interaction-1 and interaction-2, respectively. The hEPR residues in the interaction-3 region moderately interact with CDR-H3 of 9E5(Fab) (−20.7 to −60.3 kJ/mol). The strong interactions of Asp^E9 and Arg^E40 cause large conformational changes of hEPR in the interaction-1 and interaction-2 regions.

FIGURE 4.

Calculated short range interaction energies of each hEPR residue with 9E5(Fab). The error bars indicate the standard errors of the mean of the four interaction energies averaged over the respective last 100-ns trajectories.

Table 3 shows the details of the interaction energies of hEPR residues that are greater than −20 kJ/mol in Fig. 4. In the interaction-1 region, Asp^E9 interacts very strongly not only with Arg^H50 but also with Arg^L95. In the crystal structure, Asp^E9 (Oδ2) has interaction distances of 3.0 Å with Arg^H50 (Nη1) and 3.3 Å with Arg^L95 (Nη2). However, in solution, Asp^E9 has more stable hydrogen bonds and salt bridges with Arg^L95 than with Arg^H50. Thus, Arg^L95 has higher interaction energies with Asp^E9 than with Arg^H50 in Table 3. In the interaction-2 region, both Arg^E40 and Glu^E42 interact with CDR-H1 and CDR-H2 and form strong salt bridges. In the interaction-3 region, seven residues of hEPR (Met^E10, Tyr^E13, Tyr^E21, Val^E23, Ser^E26, Gln^E27, and Asn^E28) interact with a total of eight residues of CDR-L2 (His^L49 and Tyr^L50), CDR-H1 (Asp^H31 and Tyr^H33), and CDR-H3 (Gly^H100, Gly^H101, Asp^H102, and Pro^H103) in a relatively weak manner. The total interaction energies in the interaction-1, -2, and -3 regions are 290.0, 214.5, and 195.8 kJ/mol, respectively.

TABLE 3.

Interaction energies of selected residues of EPR with each residue of 9E5(Fab)

The energies are listed for the residues of EPR that are less than −20 kJ/mol in Fig. 4. S.E. indicates the standard error of the mean of four interaction energies averaged over the last 100-ns trajectories.

Interaction part	EPR	9E5(Fab)		Interaction energy (<−4.0 kJ/mol)
		Residue	CDR	Average	S.E.
				kJ/mol
Interaction-1	Lys5	Tyr32	L1	−23.5	4.0
	Ser7	Tyr32	L1	−12.2	0.5
		Tyr91	L3	−30.9	0.2
		Asp92	L3	−12.8	0.6
	Ser8	Asp92	L3	−23.2	5.2
		Asn93	L3	−7.5	0.8
	Asp9	Arg50	H2	−44.0	9.0
		Tyr91	L3	−8.8	1.2
		Asp92	L3	−11.0	1.3
		Leu94	L3	−13.9	4.1
		Arg95	L3	−102.2	8.2
Interaction-3	Met10	Tyr33	H1	−10.1	4.8
		Gly100	H3	−4.2	1.5
		Pro103	H3	−5.6	1.7
	Tyr13	Asp31	H1	−24.3	3.7
		Tyr33	H1	−10.9	0.6
		Gly100	H3	−9.1	1.1
		Gly101	H3	−6.4	1.7
	Tyr21	Pro103	H3	−5.5	0.6
		His49	L2	−7.6	1.5
		Tyr50	L2	−9.7	0.7
		Tyr91	L3	−15.3	4.6
	Val23	His49	L2	−5.7	3.0
		Tyr50	L2	−4.6	1.7
	Ser26	Pro103	H3	−5.0	0.6
		His49	L2	−10.6	2.5
		His55	L2	−8.7	3.0
	Gln27	Gly101	H3	−7.1	3.2
		Asp102	H3	−7.4	4.4
		Pro103	H3	−5.0	0.8
	Asn28	Gly100	H3	−18.3	6.2
		Gly101	H3	−16.9	2.2
		Asp102	H3	−7.6	1.7
		Pro103	H3	−9.2	1.5
Interaction-2	Arg40	Asp52	H2	−99.7	3.8
		Leu54	H2	−3.8	0.2
		Lys30	H1	−23.2	7.2
		Asp31	H1	−14.9	5.0
		Thr32	H1	−7.1	1.6
	Glu42	Lys30	H1	−42.3	11.3
		Leu54	H2	−4.9	1.0

Thermodynamic Analyses

To characterize the binding of the antibody to EPR from a thermodynamic viewpoint, we performed ITC analyses of the interaction of 9E5 IgG with EPR wild type (WT) and hEPR and mmEPR mutants (Table 4 and Fig. 5). The mmEPR triple mutant E27Q/K28N/F29Y (m3) was investigated because of the sequential differences between hEPR and mmEPR (Fig. 6).

TABLE 4.

Thermodynamic parameters of the interaction of 9E5 IgG with EPR WT and mutants at 25 °C

N, stoichiometry; ND, not detected.

Source	Mutant	N	KA	KD	ΔH	−TΔS	ΔG
			m−1	nm	kcal/mol	kcal/mol	kcal/mol
hEPR	WT	1.90 ± 0.09	16.2 ± 4.9 × 107	6.5 ± 2.0	−8.6 ± 0.7	−2.6 ± 1.0	−11.2 ± 0.2
	D9A	2.01 ± 0.04	6.7 ± 0.2 × 107	15.0 ± 0.6	−14.0 ± 0.2	3.4 ± 0.2	−10.7 ± 0.1
	S26R	1.92 ± 0.16	2.2 ± 0.8 × 107	50.6 ± 18.1	−5.9 ± 0.5	−4.1 ± 0.7	−10.0 ± 0.2
	R40A	ND	ND	ND	ND	ND	ND
mmEPR	WT	ND	ND	ND	ND	ND	ND
	R26S	ND	ND	ND	ND	ND	ND
	m3	1.87 ± 0.10	2.3 ± 0.7 × 107	46.4 ± 12.0	−5.0 ± 0.1	−5.0 ± 0.1	−10.0 ± 0.2

FIGURE 5.

Titration calorimetry of the interaction between 9E5 IgG and EPR. A–G, typical calorimetric titration of 9E5 IgG (5 μm) with 100–130 μm EPR at 25 °C (top) and integration plot of the data calculated from the raw data (bottom). The solid line corresponds to the best fit curve obtained by least square deconvolution. A, hEPR WT; B, hEPR D9A; C, hEPR S26R; D, hEPR R40A; E, mmEPR WT; F, mmEPR R26S; and G, mmEPR m3.

FIGURE 6.

Multiple sequence alignment of the EPRs. An asterisk (*) indicates fully conserved residues. A colon (:) indicates strongly similar residues. A period (.) indicates weakly similar residues. In mammals, pEPR, rEPR, and mpEPR indicate Pan troglodytes (chimpanzee), Rattus norvegicus (rat), and Mustela putorius furo (European domestic ferret) EPR, respectively. In avian, cEPR indicates Gallus gallus (chicken) EPR. In amphibian, xtEPR indicates Xenopus tropicalis (western clawed frog) EPR. In fish, xmEPR indicates Xiphophorus maculatus (southern platyfish) EPR. The UniProt accession numbers are as follows: hEPR, O14944; pEPR, H2QPP3; mmEPR, Q61521; rEPR, Q9Z0L5; mpEPR, M3YCI3; cEPR, P13387; xtEPR, Q28BU9; and xmEPR, D1MGM2. The alignment and figure drawing were performed using the ClustalΩ and ClustalX programs (39).

hEPR WT showed an exothermic profile, and its binding enthalpy, ΔH, is −8.6 ± 0.7 kcal/mol. The interaction has a strong binding affinity (K_D = 6.5 nm). The hEPR mutant R40A showed no heat in the ITC analysis, indicating that Arg^E40 is one of the hot spot residues in the interaction between hEPR and 9E5 IgG. The other Ala mutant, hEPR D9A, has a lower binding affinity with a large unfavorable entropy change. The binding energy of hEPR S26R is 1.2 kcal/mol higher than that of hEPR WT. These results indicate that steric hindrance or electric repulsion reduces the binding affinity.

As expected, no heat was detected for mmEPR WT. In contrast, mmEPR m3 has a similar binding affinity to S26R hEPR. mmEPR R26S exothermically binds to 9E5 IgG, but the dissociation constant could not be determined because of the weak binding.

SPR Analysis

Kinetic analysis of the interaction between 9E5 IgG and hEPR was carried out by SPR. The sensorgram showed that the dissociation rate of 9E5 IgG is slow (Fig. 7). The kinetic parameters (association rate constant k_on and dissociation rate constant k_off) were calculated with the bivalent analyte model. The results show that the binding affinity is dominated by the high k_on (k_on1 = 1.15 × 10⁶ m⁻¹ s⁻¹, k_off1 = 9.83 × 10⁻⁴ s⁻¹). The K_D value (=k_off1/k_on1) was calculated to be 0.86 nm.

FIGURE 7.

Surface plasmon analysis of the interaction between 9E5 IgG and hEPR. Thioredoxin-fused hEPR was immobilized by an amine coupling method on a CM5 sensor chip. The analyses were performed by injecting various concentration of 9E5 IgG (3.1–50 nm) into the sensor chip under the buffer condition of HEPES-buffered saline with surfactant P20 (pH 7.4) at a flow rate of 30 ml min⁻¹ at 25 °C. Black dashed lines show the fitted curves. RU, resonance units.

Discussion

In this study, we describe the crystal structures of 9E5(Fab) in the presence and absence of its antigen hEPR. To investigate the recognition mechanism of hEPR by 9E5(Fab), we solved the x-ray structure of 9E5(Fab) with and without hEPR. To bind to hEPR, CDR-H3 undergoes the following three characteristic structural changes (Fig. 2). First is the formation of Asp^H102-Arg^H98 salt bridges. Asp^H102 in 9E5(Fab) without hEPR forms a hydrogen bond with His^L49, thereby contributing to the interaction with CDR-H3 and CDR-L2. The binding of hEPR induces rearrangement of the hydrogen bonds so that Asp^H102 forms a salt bridge with Arg^H98, which was originally exposed to the solvent region, and His^L49 forms a hydrogen bond with Ser^E26. Second are the conformational changes in the Gly^H99–Gly^H101 loop. As described above, Gly^H101 moves more than 6.5 Å upon binding of hEPR. All of the residues between Arg^H98 and Asp^H102 are glycine, and thus proper contact with hEPR is possible because of the flexibility. The third change is cis-trans Pro^H103 isomerization. In the structure of 9E5(Fab), Pro^H103 in CDR-H3 is stabilized by hydrophobic interactions with His^L49 in CDR-L2 and a couple of hydrophobic residues. Although the difference in the energy level between the cis and trans forms is only 2 kJ/mol, the activation energy of cis-trans isomerization is 80–90 kJ/mol (33), meaning that cis-trans isomerization of proline is an energy-requiring reaction.

From the results of the MD simulations, the interaction energy for interaction-3 is relatively small (Fig. 4). However, it is predicted that Asp^E9 and Arg^E40 energetically contribute to interaction-1 and interaction-2, respectively. The ITC analysis clearly indicates that D9A hEPR has a comparable binding affinity with hEPR WT, suggesting loss of entropic energy in D9A and the existence of water molecules around Arg^H50 in the counterpart of 9E5(Fab). It also indicates that Asp^E9 does not contribute to complex formation. Conversely, the R40A mutant of hEPR does not bind to 9E5 IgG, suggesting that Arg^E40 is one of the hot spots for 9E5 IgG (Table 4). This interaction energy may contribute to cis-trans isomerization. The formation of a salt bridge between Asp^H102 and Arg^H98 may also contribute to cis-trans isomerization. Once these conformational changes have occurred, it may not be able to return to the structure of CDR-H3, suggesting that the 9E5(Fab)·hEPR complex is difficult to dissociate without some type of energy, such as thermal energy. In fact, SPR analysis indicates that the rate of dissociation is extremely slow (Fig. 7). It is concluded that 9E5(Fab) is an effective antibody against hEPR because 9E5(Fab) strongly binds to hEPR and cannot easily dissociate.

9E5(Fab) can only recognize hEPR, and it can be called a human trap antibody. We will now discuss the specific recognition by 9E5(Fab) from the viewpoint of the amino acid alignment of hEPR (Fig. 6). Ser^E26–Tyr^E29 in hEPR interacts- with CDR-H3 in 9E5(Fab), corresponding to Arg^E26–Phe^E29 in mmEPR. The results of ITC analysis indicate that the K_D value of the S26R mutant of hEPR is about 7 times higher than that of WT (Table 4). In contrast, the binding affinity of mmEPR m3 is on the order of 10⁻⁸ m. These results suggest that all of the Ser^E26–Phe^E29 sequence in hEPR is essential for the specific recognition of 9E5(Fab).

EPR specifically binds to the homodimers of EGFR, ErbB-1, and ErbB-4 (34, 35). To date, three structures of ligands of the EGF family complexed with the EGFR ectodomain have been reported: EGF·ErbB-1 (Protein Data Bank code 1IVO), transforming growth factor α (TGFα)·ErbB-1 (Protein Data Bank code 1MOX), and neuregulin1β·ErbB-4 (Protein Data Bank code 3U7U) (36–38). To investigate how to accomplish binding of 9E5(Fab)·hEPR to EGFR, we superimposed 9E5(Fab)·hEPR on the EGF·ErbB-1 ectodomain (Fig. 8). hEPR in EPR-9E5(Fab) superimposed well on EGF in EGF·ErbB-1, and the average r.m.s.d. between 40 Cα atom pairs was 0.9 Å. The light chain of 9E5(Fab) does not interact with ErbB-1 and EGF. However, the heavy chain of the N-terminal region (Glu¹–Gln³), CDR-H1 (Asn^H28–Lys^H30 and Tyr^H33), CDR-H2 (Arg^H50–Lys^H59), and the region from the β7 sheet to the η3 3¹⁰ helix (Thr^H71–Asn^H77) in 9E5(Fab) interact with Domain I (Tyr⁸⁸ and Asn⁹¹–Ser⁹²) and Domain III (Ile³¹⁸–Leu³²⁵, Asn³²⁸, Thr³³⁰, Asp³⁵⁵, and Phe³⁵⁷–Pro³⁶¹) in ErbB-1. For this reason, the interactions of the heavy chain of 9E5(Fab) prevent binding of the complex of 9E5(Fab)·hEPR to ErbB-1. This tendency is almost the same as TGFα·ErbB-1 and NRG1β·ErbB-4 complexes (37, 38). Therefore, 9E5(Fab)-captured hEPR could not bind to ErbB-1 and ErbB-4.

FIGURE 8.

Superimposed structures of 9E5(Fab)·hEPR and EGF·EGFR complexes. The upper panel shows hEPR in the 9E5(Fab)·hEPR complex (colored) superimposed on EGF in the EGF-EGFR complex (gray; Protein Data Bank code 1IVO). The lower panel shows a close-up view around the binding site of EGFR indicated by the arrow in the upper panel. EGF is wheat-colored. The r.m.s.d. between hEPR and EGF is 0.9 Å. The EGFR residues within 2 Å of the hEPR residues are shown in red. The 9E5(Fab) residues within 2 Å of EGFR are shown in yellow.

From the viewpoint of kinetics, the K_D values of 9E5 IgG and hEPR WT are 0.86–6.5 nm, which were observed by ITC and SPR analysis (Fig. 7 and Table 4). hEPR is a much weaker antagonist of the ErbB-1 and ErbB-4 receptors with IC₅₀ values of 2800 nm and >5 μm, respectively (34), indicating that 9E5 IgG binds to hEPR more strongly than ErbB-1 and ErbB-4. According to previous studies, mutational analysis and chemical regeneration suggest that the guanidinium group of Arg^E40 in hEPR is essential for binding of the ErbB receptor (40, 41). These results support that 9E5(Fab) acts as not only the simple capturer of EPR but also the competitive neutralization antibody against EGFR with inhibition of the functional residue Arg^E40.

In conclusion, 9E5(Fab) binds to only hEPR with rearrangement of the hydrogen bonding network along with cis-trans isomerization of Pro^H103 and shows high affinity and slow dissociation. MD simulation and ITC analyses uncovered that Arg^E40 acts as a hot spot in the interaction between hEPR and 9E5 IgG. Antibody drugs based on the structure of 9E5 with the conservation of the human trap recognition mechanism are expected, especially for colon cancer.

Author Contributions

M. I., Y.-H. L., and Y. S. prepared 9E5(Fab); Y. K., A. S., T. M., and T. N. constructed EPR expression plasmids; Y. K., E. M., T. N., and T. Y. purified EPR or EPR-9E5(Fab) complex; Y. K., E. M., T. Y., H. M., and T. I. performed X-ray crystallography; S. N., T. M., and K. T. performed thermodynamic analyses; K. S., H. D., and H. F. performed molecular dynamics simulations; T. Kawamura determined 9E5 sequence by proteomics analysis; Y. K. wrote the manuscript with input from all the coauthors; E. M., T. I., and T. Kodama conceived and designed the study; and T. Kodama supervised the Molecular Dynamics for Antibody Drug Development project.