Structural basis of a flavivirus recognized by its neutralizing antibody: solution structure of the domain III of the Japanese encephalitis virus envelope protein.

The flavivirus envelope protein is the dominant antigen in eliciting neutralizing antibodies and plays an important role in inducing immunologic responses in the infected host. We have determined the solution structure of the major antigenic domain (domain III) of the Japanese encephalitis virus (JEV) envelope protein. The JEV domain III forms a beta-barrel type structure composed of six antiparallel beta-strands resembling the immunoglobulin constant domain. We have also identified epitopes of the JEV domain III to its neutralizing antibody by chemical shift perturbation measurements. Site-directed mutagenesis experiments are performed to confirm the NMR results. Our study provides a structural basis for understanding the mechanism of immunologic protection and for rational design of vaccines effective against flaviviruses.

three structural proteins, a 12-kDa nucleocapsid or core protein, a 8-kDa non-glycosylated membrane protein, and a 53-kDa glycosylated envelope protein (E), as well as seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, NS5) (2). The E protein is the dominant antigen in eliciting neutralizing antibodies and plays an important role in inducing immunologic responses in the infected host (3). Structural elements of the E protein are also involved in viral attachment, fusion, penetration, hemagglutination, host range and cell tropism, and virus virulence and attenuation (1).
The three-dimensional structure of the TBEV E protein has been determined by x-ray crystallography (4). The x-ray structure reveals that the E protein forms a head-to-tail dimer with 170 Å in length. Each monomer of the E protein is composed of three discernible domains. Domain III containing the C-terminal 100 amino acids forms an Ig-like ␤-barrel structure composed of seven antiparallel ␤-strands. This domain is connected by a flexible region to domain I in the center, which folds into an eight-stranded antiparallel ␤-barrel. Domain I contains about 120 residues in three segments (residues 1-51, 137-189, and 285-302). The two long loops between these three segments form the dimerization domain II.
Recently, the structure of the E protein dimer of TBEV has been used to fit into the outer layer of density in the cryoelectron microscopy reconstruction of dengue virus (5). The structure suggests that the fusion mechanism of flaviviruses involves the insertion of the distal ␤-barrels of domain II of the E protein into the cellular membrane. On the other hand, the mechanism of antibody-mediated neutralization of the flavivirus remains a matter of speculation. Mutations on domain III of the E protein leading to the escape from antibody neutralization have been reported for JEV (6 -9), louping ill virus (10), TBEV (11,12), yellow fever virus (13), and dengue virus (14 -16). Synthetic peptides have also been used to map the linear epitopes of domain III (17). However, none of these studies provide a detailed spatial configuration of the epitopes of domain III. To date, no structure has been reported for the E protein complexed with neutralizing antibodies. In this study, we have determined the solution structure of the domain III of the JEV E protein. In addition, we have identified the neutralizing epitopes of the JEV domain III to its monoclonal antibody (mAb E3.3) by comparing NMR chemical shifts of the free (14 kDa) and the antibody-bound (178 kDa) forms. Our results provide a structural basis for understanding the mechanism of immunologic protection and for rational design of vaccines effective against flaviviruses.

EXPERIMENTAL PROCEDURES
Protein Preparation and Purification-The DNA fragments encoding residues 292-402 of the domain III of the JEV (CH2195LA) E protein were amplified by PCR. The amplified PCR products were subcloned into pET-22b vector. Plasmids encoding JEV domain III were introduced into Escherichia coli strains BL21(DE3), and transformants were cultured overnight at 37°C in SOB medium supplemented with ampicillin (100 g/ml). The uniform 15 N-labeled and/or 15 N/ 13 C-labeled samples were expressed in cells grown in M9 minimal media containing 15 NH 4 Cl and/or (U-13 C) glucose. For preparation of the (U-2 H, 15 N)domain III for NMR spectroscopy, the protein was expressed in the M9 minimal media with 99% D 2 O in place of H 2 O and 15 NH 4 Cl as a sole nitrogen source (18). The cell lysates were centrifuged, and the supernatant from the centrifugation was applied to a nickel-nitrilotriacetic acid column according to the protocols provided by the manufacturer (Qiagen Inc.) at 4°C. Fractions containing the JEV domain III were pooled, concentrated, and added to a 1.6 ϫ 60.0 cm Sephacryl S-100 column (Amersham Biosciences). Finally, the protein was eluted with NMR buffer containing 50 mM K 2 HPO 4 (pH 6.5), 100 mM NaCl, 10 mM NaN 3 , and 0.1 mM EDTA. Purity of the protein was confirmed by high pressure liquid chromatography and mass spectroscopy. The final protein concentration for NMR study was 2 mM in 90% H 2 O, 10% D 2 O. To prepare D 2 O samples, concentrated protein solutions were lyophilized and redissolved in D 2 O.
Monoclonal Antibody Preparation-The hybridoma cell line producing mAb E3.3 was prepared from BALB/c mice immunized with JEV Beijing-1 strain. The specificity of mAb E3.3 was demonstrated previously on JEV E protein (8,19), showing Ͼ95% plaque neutralization against Beijing-1 strain and blocking viral hemagglutination at the reciprocal titer of 320.
NMR Spectroscopy-All NMR data used in the structural analysis were acquired with Bruker Avance 600 or 500 MHz spectrometers equipped with triple resonance probes at 25°C. TROSY spectra were performed on a Bruker Avance 800 MHz spectrometer with a sample (JEV domain III/mAb E3.3) concentration of 100 M at 25°C. 1 H NMR data were referenced to the 1 H resonance frequency of 2,2-dimethyl-2silapentanesulfonic acid; 13  Experimental Restraints-NOE-based distance restraints were collected from analysis of three-dimensional 15 N-edited NOESY-HSQC and 13 C-edited NOESY-HSQC spectra recorded with mixing times of 150 and 80 ms, respectively. NOEs were broadly categorized as "strong," "medium," and "weak," corresponding to distance limits of 1.8 -2.8, 1.8 -3.4, and 1.8 -5.0 Å, respectively. NOE restraints involving methyl group and pseudoatoms were corrected by relaxing 0.3 Å to take into account equivalent atoms.
The and angles were obtained based on 3 J HNH␣ coupling constants derived from the HNHA experiment and predicted from TALOS (23). Hydrogen bond restraints were included in calculations only if the amide protons were slowly exchanging and if the ␤-strand inter-strand NOE cross-peaks were observed. Each hydrogen bond was enforced by two distance restraints of 1.8 -2.3 Å (amide proton to carbonyl oxygen) and 2.8 -3.3 Å (amide nitrogen to carbonyl oxygen).
Structure Calculation-The structure calculations were carried out using the X-PLOR 3.851 (24) program on a SuSE Linux 7.3 PC equipped with an Intel Pentium 4 1.8 GHz CPU and 512 MB DDR-SDRAM. The structures were calculated and refined with ab initio simulated annealing protocol, "sa.inp," and SA refinement protocol, "refine.inp" in X-PLOR 3.851. All force constants and molecular parameters, etc. were set to their default values, as in the original sa.inp and refine.inp protocols, with the exception of the timestep (ps), which was decreased to 0.001 ps throughout the calculations. Simulated annealing was performed using 9,000 steps at 1,000 K and 10,000 steps by gradually cooling to 100 K. Finally, these structures were energy-minimized using 500 steps of Powell energy minimization. The best 15 lowest energy structures were further analyzed with MOLMOL (25) and PRO-CHECK-NMR (26). These final structures contained no distance constraint violations greater than 0.3 Å and no dihedral angle constraint violations greater than 5°. The coordinates of both the representative structure and the family of structures have been deposited at the Brookhaven Protein Data Bank (access number 1PJW).

RESULTS AND DISCUSSION
Structure Determination-The JEV domain III is well behaved at millimolar concentrations and well suited for solution structural studies using modern NMR techniques. The backbone 1 H, 15 N, 13 C, and 13 CO assignments of the 111 residue JEV domain III are essentially complete. More than 85% of the side chain 1 H resonances are assigned. A total of 763 NOEderived distance constraints, including 205 intraresidue, 263 sequential, 43 medium (1 Ͻ i Ϫ j Ͻ 5), 206 long range (i Ϫ j Ն 5), and 46 hydrogen bond distance restraints in conjunction with 194 backbone dihedral angles are used in the structure calculations. The overlay of the backbone atoms for the 15 lowest energy structures of the JEV domain III is shown in Fig.  1A. Fig. 1B shows the ribbon diagram of the lowest energy structure of the JEV domain III. The energetic and structural statistics are listed in Table I. The root mean square deviation calculated from the averaged coordinates for the JEV domain III is 0.68 Å for the backbone heavy atoms (N, C, and C␣) and 1.43 Å for all heavy atoms. Analysis of and backbone angles is facilitated by using the program Procheck_NMR (26). Only 0.1% of the residues are in disallowed regions, and more than 98% of the residues are within most favored or additional allowed regions of the Ramachandran plot.
Structure Description and Homology-The JEV domain III forms a ␤-barrel type structure composed of six antiparallel ␤-strands resembling the immunoglobulin constant domain (Fig. 1B). The six ␤-strands consist of residues 308 -316, 322-328, 339 -344, 369 -374, 381-386, and 391-396. A conserved disulfide bridge is found between residues Cys-304 and Cys-335. The top of the ␤-barrel is closed by the N-terminal residues and the loops between the ␤2-␤3, ␤3-␤4, and ␤5-␤6 strands. The amino acid sequence of the JEV domain III is 37.1% identical and 65.0% similar to that of the TBEV domain III, and the global fold of the JEV domain III is similar to that of the TBEV domain III (4). In the JEV domain III, strands ␤1, ␤2, ␤3, ␤4, ␤5, and ␤6 correspond to strands A, B, C, E, F, and G of the TBEV domain III (4). However, there are localized differences between these two proteins. Formation of extra strands (Ax, Cx, Dx, and D in Fig. 2b of Rey et al. (4)) has been observed in the TBEV domain III. These extra strands, except strand D, are all located in the loop regions on the top of the ␤-barrel. Strand D is found in the long loop region between strands ␤3 and ␤4. Since the x-ray structure of the TBEV E protein has been determined to be a head-to-tail dimer, contacts between domain III and domains I and II from the other subunit may force the loop regions to form more ordered conformations. We have used reduced spectral density functions to analyze the NMR relaxation rates and 1 H-15 N heteronuclear NOEs of the JEV domain III (data not shown) (27,28). Our results indicate that the loop regions, especially the ␤3-␤4 loop, undergo picosecond fast motions, as evident from the larger J(H) values. Thus, the ordered conformations are less favored in these loop regions of the JEV domain III. However, more ordered conformations could not be ruled out in the full-length JEV E protein.
Epitope Mapping by NMR-Antibody-antigen interactions play important roles in the immune system. Identification of an epitope is the first step toward the elucidation of the molecular mechanism underlying antibody-antigen interactions. Series of short peptides encompassing the sequences of a protein antigen can be used to locate the antibody binding sites. However, exhaustive mutagenesis experiments are needed to identify the epitope if an epitope is comprised of discontinuous fragments. Alternatively, NMR can be used to map an epitope (29,30). Chemical shifts are highly sensitive to the local magnetic environment and are excellent probes for conformational change. Previously, the chemical shift perturbation method has been shown to accurately identify the epitopes of several antibodybound antigens (29,31,32). JEV domain III gives a well resolved 1 H-15 N HSQC spectrum in which backbone amide peaks for most of the residues can be identified ( Fig. 2A). However, low sample concentrations due to the limited availability of mAb E3.3 and the size of the JEV domain III/mAb E3.3 com-  (Fig. 2A). Comparison of the 1 H-15 N cross-peaks in the free and the antibody-bound forms show that complex formation causes significant chemical shift changes for a discrete set of resonances including residues 302-312, 322-339, 360 -372, and 385-392 (Fig. 2B). The residues with significant chemical shift changes upon binding with mAb E3.3 are mapped onto the structure of the JEV domain III, as shown in Fig. 3A. Basically, the mapped epitopes are located on the top portion of the ␤-barrel including residues in the N termini, the ␤2-␤3, ␤3-␤4, and the ␤5-␤6 loops. Residues close to the top portion of the ␤-barrel on the six ␤-strands are also found to have large chemical shift changes upon mAb E3.3 binding. The mapped epitopes of the JEV domain III are presented in the three-dimensional structure of the TBEV E protein to demonstrate the relative positions of the antibody binding sites of the domain III to domains I and II of the E protein (Fig. 3B). It has been suggested that the CFG strands of the TBEV domain III, which form the outer lateral surface of the dimer (Fig. 3B), may be the antigenic sites (4). However, our studies of the JEV domain III show that the top portion of the ␤-barrel, including residues in the N termini, the ␤2-␤3, ␤3-␤4, ␤5-␤6 loops, and the adjacent residues on the six ␤-strands, are the major neutralizing antibody binding sites. Previously, mutations that lead to the escape from antibody neutralization have been reported for the JEV domain III (Fig. 4A) (6-9,17). These mutation sites are all located in the N termini, the ␤2-␤3, ␤3-␤4, and the ␤5-␤6 loops on the top of the ␤-barrel covered by the NMR mapped epitopes.  3. A, the location of residues whose 1 H-15 N cross peak positions are affected by the formation of the JEV domain III/mAb E3.3 complex. The color scheme used is: orange, a residue whose peak shows chemical shift change ⌬␦ Ն 0.15 ppm; red, a residue whose peak exhibits a significantly reduced intensity at the same position as in the free spectrum and has been reported to be the mutation site. Mutations leading to the escape from antibody neutralization or neutralizing epitopes mapped by phage-displayed peptide libraries are represented as dots. B, chemical shift changes of the JEV domain III/mAb E3.3 complex mapped onto the three-dimensional structure of the TBEV E protein (Protein Data Bank access number 1SVB). The TBEV domain III is shown in yellow, and the residues analogous to the JEV domain III, which have significant chemical shift changes upon antibody binding, are shown in orange. Structural elements (strands A, B, C, D, E, F, G, Ax, Cx, Dx) are labeled according to Fig. 2 of Rey et al. (4). and 360 are located in the ␤2-␤3 and ␤3-␤4 loops close to the top of the ␤-barrel but have not been reported to be involved in the neutralizing antibody binding. Residue 387, which has been suggested to be the antigenic site (17), is located in the ␤5-␤6 loop. These three residues are chosen because they all exhibit significant chemical shift changes upon mAb E3.3 binding, and they are located in the three major antigenic loops ( Fig. 2A). Although residue 352 is located in the ␤3-␤4 loop, it does not show any chemical shift change upon mAb E3.3 binding. Thus, mutation of residue 352 is expected to have little influence on the antibody binding. Based on the Western blot analysis, the I337A, F360A, and R387A mutations all exhibit significant reductions of binding to the neutralizing antibody mAb E3.3 (Fig. 4B) (10), residues 384 and 386 of TBEV (11,12), residue 305 of yellow fever virus (13), and residues 307, 333-351, and 383-389 of the dengue viruses (14 -16). Fig. 4A shows mutation sites that abolish binding of the corresponding neutralizing antibodies of various flaviviruses. Not surprisingly, most of the mutation sites are covered by the NMR mapped epitopes of the JEV domain III (Fig. 4A). Because of the sequence homology and the similar global fold between the JEV and TBEV domain III, a common three-dimensional structure of the domain III for all flaviviruses can be predicted. Moreover, because the mutation sites that abolish binding of the corresponding neutralizing antibodies of various flaviviruses are covered by the present NMR mapped epitopes, it can be anticipated that the neutralizing epitopes on the domain III of different flaviviruses are located within the similar regions. To further understand the mechanism of antibody-mediated neutralization of the flaviviruses, NMR structural studies and epitope mapping of the domain III of the dengue virus are currently under way in our laboratory.
In conclusion, NMR structural studies and site-directed mutagenesis were performed to identify and characterize the neutralizing epitopes of the JEV domain III. Our results provide a structural basis for understanding the mechanism of immunologic protection and for rational design of vaccines effective against flaviviruses.