Neutralizing antibodies can initiate genome release from human enterovirus 71

Significance

Enterovirus 71 (EV71) causes yearly outbreaks of hand, foot, and mouth disease in Southeast Asian countries including China and Malaysia. Some of the infected children develop encephalitis that can be fatal or result in permanent brain damage. There are no anti-EV71 therapeutic agents available. Here it is shown that an antibody that had been generated by using an immature EV71 virus as an antigen induced the release of genome from EV71 virions, rendering the virus noninfectious. The induction of genome release is a mechanism by which antibodies can neutralize viruses. Furthermore, the approach presented in the paper could be used to prepare antibodies with similar properties against related viruses that include significant human pathogens.

Abstract

Antibodies were prepared by immunizing mice with empty, immature particles of human enterovirus 71 (EV71), a picornavirus that causes severe neurological disease in young children. The capsid structure of these empty particles is different from that of the mature virus and is similar to “A” particles encountered when picornaviruses recognize a potential host cell before genome release. The monoclonal antibody E18, generated by this immunization, induced a conformational change when incubated at temperatures between 4 °C and 37 °C with mature virus, transforming infectious virions into A particles. The resultant loss of genome that was observed by cryo-EM and a fluorescent SYBR Green dye assay inactivated the virus, establishing the mechanism by which the virus is inactivated and demonstrating that the E18 antibody has potential as an anti-EV71 therapy. The antibody-mediated virus neutralization by the induction of genome release has not been previously demonstrated. Furthermore, the present results indicate that antibodies with genome-release activity could also be produced for other picornaviruses by immunization with immature particles.

Enterovirus 71 (EV71) is a picornavirus that causes hand, foot, and mouth disease (1). In infants and small children, the infection may proceed to encephalitis that can be fatal or result in permanent brain damage. EV71 virions are nonenveloped with a diameter of approximately 300 Å. The capsid has icosahedral, pseudo-T=3 symmetry with four viral proteins VP1, VP2, VP3, and VP4 in each icosahedral asymmetric unit (2, 3). Subunits VP1, VP2, and VP3 have a jelly-roll fold common to many viruses. VP4 is a small protein attached to the inner face of the capsid. EV71 infections produce fully infectious RNA-filled particles and empty immature particles that lack genome and contain capsid protein VP0, the precursor of VP4 and VP2 (3). These empty particles have approximately 5% larger diameter than the mature virions. Furthermore, the protomer formed by VP0, VP1, and VP3 is rotated by 5.4° relative to the protomer formed by VP1, VP2, VP3, and VP4 in the mature particle with respect to the icosahedral symmetry axes. The empty particles are presumably precursors of the mature infectious virions (3).

Rhino and entero picornaviruses have a depression, called the “canyon,” on the virion surface encircling the icosahedral fivefold axes (4). The canyon is frequently the site of binding of picornavirus receptors (5–8), although some receptor molecules bind to other sites on picornavirus capsids (9, 10). Experimental evidence indicates that binding of a receptor into the canyon results in the expulsion of the “pocket factor” from the hydrophobic cavity within VP1 (11–14). Ejection of the pocket factor leads to destabilization of virions. Such activated “A” particles are characterized by expansion of the capsid, release of VP4, and externalization of the VP1 N-termini (6). The organization of the major capsid proteins in the A particle and in the immature empty particles are similar (3). Transition of the virion to the A state is a prerequisite for the release of the genome (15). Heating of picornavirus particles to nonphysiological temperatures of 50 °C to 60 °C can also induce transformation of virions to the A state in vitro (6, 16, 17).

Here we present an analysis of the interactions of the monoclonal antibodies E18 and E19 with EV71. By using cryo-EM, we show that binding of E18 to EV71 causes the virus to change its conformation to the A state and to eject much of its genome. This was further verified by fluorescence activation when SYBR Green dyes interact with RNA. In contrast, although mAb E19 does neutralize the virus, it has a quite different footprint on the virus surface and does not cause ejection of the genome.

Results and Discussion

The E18 and E19 antibodies were prepared by immunizing mice with empty, immature EV71 particles containing VP0 (18). Both E18 and E19 could neutralize the virus as intact antibodies or as Fab fragments (Fig. 1). Both these mAbs can recognize conformational epitopes on the surface of heat-inactivated EV71 particles by indirect ELISA. However, these antibodies could not recognize linear epitopes by using immunoblot analysis (Fig. 2). The Fab fragments of these mAbs were incubated with EV71 for cryo-EM studies of the mAb–virion complexes. Visual inspection of the cryo-EM micrographs showed that as many as 20% of the EV71 particles that had been incubated with E18 had lost much or all of their RNA genome (Fig. 3 A, D, E, and F). In contrast to the action of E18 Fab, only approximately 1% of the EV71 particles incubated with E19 Fab had lost their genome (Fig. 3 B and C). Thus, E18 but not E19 induced genome release from virions.

Fig. 1.

Neutralization of EV71 by monoclonal antibodies E18 and E19. Whole IgG and Fab fragments of the monoclonal antibodies E18 (A) and E19 (B) were used to inhibit EV71 at different concentrations (x axis) by using a plaque reduction neutralization test. The red symbols represent whole antibody and the blue symbols represent Fab fragments. Inhibition of virus was represented as the percentage of plaques relative to plaques in the control wells. We demonstrated neutralization of EV71 by Fab fragments as well as by whole IgG, with whole IgG being more efficient than Fab fragments.

Fig. 2.

Analysis of binding of E18 and E19 to EV71 viral proteins by immunoblot and indirect ELISA. (A) Mock- (M) and EV71-infected (V) cell lysates were separated by SDS/PAGE and transferred onto membrane. Membrane was probed with R525, E18, or E19. E18 and E19 did not bind denatured viral proteins and, therefore, recognize conformational epitopes. (B) Indirect ELISA was performed by coating wells with recombinant viral proteins or heat-inactivated EV71-infected cell lysates. Various concentrations of mAb were added in duplicates. The mAbs were detected by HRP assay (Materials and Methods). The amount of bound mAbs are presented as average OD₄₅₀ ± SDs.

Fig. 3.

Stability of EV71 virions and their complexes with E18 and E19. (A) Plot of time and temperature dependence of interaction of the fluorescent SYBR Green I and II dyes with the EV71 genomic RNA. Native EV71 virions (blue line) as well as EV71 complexes with E18 (green line) and E19 (red line) were incubated with the fluorescent dyes at 37 °C. The purple line represents a negative control without virus. Subsequently, the complexes were gradually heated to 90 °C. The increase in fluorescence showed that the fluorescent dyes were binding to the genomic RNA. The numbers at 60, 120, and 360 min associated with each of the lines show the percentages of empty particles in cryo-EM images for each sample. Cryo-EM images of (B) EV71 virions and of (C) EV71 complexed with E19 Fab and (D) EV71 complexed with E18 Fab that were incubated at 37 °C for 120 min. (E and F) Also shown are virions in the process of genome release observed in the E18–EV71 mixture. (Scale bars: 50 nm.)

Separate reconstructions were made of the empty immature EV71–E18 complexes and of the full EV71–E18 complexes. The resolutions of these reconstructions were 10 Å and 20 Å, respectively, as judged by the resolution at which the Fourier shell correlation coefficient decreased to less than 0.5 (Fig. S1). The correlation coefficient between the cryo-EM electron density maps of the capsid region of the heat-induced EV71 A particles and that of the corresponding density of the EV71–E18 (full) complex was 0.83, whereas the correlation between the electron density of the native virus capsid and the EV71–E18 complex was only 0.61 (Table 1). Thus, the E18 Fab had induced the conformational change of the EV71 virions to the A state (Fig. 4 A and D), a conformational change similar to what was produced by heating EV71 virions to 56 °C (Fig. 4 D and G) (16). Consistent with these observations, ELISA tests showed that E18 binds better to heat-induced EV71 A particles than to mature virions (Fig. 5). The 3D cryo-EM reconstructions showed that the capsid structures of the RNA-containing EV71–E18 Fab complex and of the empty EV71–E18 Fab complex are nearly identical (correlation coefficient, 0.95; Fig. 4 A, B, D, and E and Table 1). Thus, the conformational change must be the consequence of E18 binding to the virions, which then leads to genome release.

Table 1.

Correlation coefficients comparing electron density distribution in the capsid regions of EV71 virions, A particles, and E18 and E19 antibody complexes

Structure	EV71 mature	EV71 “A” particle	EV71 + E19	EV71 + E18 empty	EV71 + E18 full
EV71 + E18 full	0.61	0.83	0.39	0.95	—
EV71 + E18 empty	0.42	0.57	0.48	—
EV71 + E19	0.81	0.39	—
EV71 “A” particle	0.52	—
EV71 mature	—

Fig. 4.

Cryo-EM reconstructions of EV71–Fab complexes showing also comparisons with native and expanded EV71 virions. (Top) Cryo-EM reconstructions of (A) genome containing EV71–E18, (B) empty EV71–E18, and (C) genome containing EV71–E19 complexes. The reconstructions are rainbow colored according to the distance of the surface from the particle center. (Middle) Center sections of the cryo-EM reconstructions of (D) genome containing EV71–E18, (E) empty EV71–E18, and (F) genome containing EV71–E19 complexes. The sections are rainbow-colored according to the electron density height. (Bottom) Center sections of (G) heat-induced EV71 A particles, (H) empty particles after genome release, and (I) native EV71 capsid.

Fig. 5.

Comparison of binding of E18 (A) and E19 (B) to heat-inactivated EV71 (i.e., A particles) and native EV71. The purified EV71 was stored on ice or incubated at 56 °C for 30 min. Serial dilutions of the samples were added to wells coated with polyclonal antibodies against VP1, and bound viral particles were detected by the addition of E18 or E19 (Materials and Methods). The average OD values indicating amounts of bound mAbs ± SDs are shown.

Although EV71 can be completely inhibited by E18, the electron micrographs show that not all particles had released their genomes (Fig. 3F). However, the virus particles would be inactivated even if only a small part of the genome were released from the virion and degraded by RNAses. Such particles, even though noninfectious, would be evaluated as genome containing in our analysis. Indeed, the electron density corresponding to the genome was lower in the EV71–E18 complex (minimum, −2.2, maximum, 1.1; average, −0.17) than in the EV71–E19 complex (minimum, −1.6; maximum, 2.0; average, −0.15), even though only full-looking particles were used to calculate the (full) EV71–E18 reconstruction.

Alternatively, the neutralization of EV71 by E18 might be achieved not only by inducing genome release but also by other means such as preventing receptor binding. Indeed, the E18 footprint on the virion surface includes Lys-149 of VP2 that has been implicated to have a role in attachment of EV71 to the P-selectin glycoprotein ligand-1 receptor (Fig. 6 and Fig. S2) (19).

Fig. 6.

Antibody footprints on the EV71 surface. The figure shows 2D projections of the EV71 virion surface. Residues of capsid proteins VP1, VP2, and VP3 are outlined in blue, green, and red, respectively. Residues involved in binding (A) E18 and (B) E19 are shown in bright colors. The footprints of E18 and E19 are outlined by yellow lines in A and B, respectively. The border of one VP4–VP2–VP3–VP1 protomer is indicated by a dotted line. Positions of twofold, threefold, and fivefold icosahedral symmetry axes are shown as ovals, triangles, and pentagons, respectively. One icosahedral asymmetric unit is outlined by a triangle.

The E18 binding sites on the EV71 capsids are located between VP4–VP2–VP3–VP1 protomers (Fig. 6A and Table 2). However, the protomer in both A particles (after receptor binding) and empty, immature (before VP0 cleavage) particles is rotated by 5.4° relative to its position in the mature capsid with respect to the icosahedral axes (3, 16). Because the E18 antibody was generated by immunization with empty, immature particles, it is likely that, when E18 binds to mature EV71 particles, there will occur an “induced fit” that requires local rearrangements of the capsid to an immature-like capsid conformation. Thus, possibly the E18 antibody binds to the virus when the capsid temporarily and locally changes structure to be like an A particle because of the natural capsid dynamics. However, mature virions contain VP2 and VP4 instead of VP0. Therefore, the capsid proteins reorganize to resemble A particles that are generated when the virus recognizes a receptor and expels the pocket factors (Fig. 4 A, G, H, and I). As the E18 Fabs bind across the interface between protomers (Fig. 6A), binding of E18 to the mature virion induces the protomers to rotate by 5.4° to the A form. In contrast, E19 binds wholly within a single protomer (Fig. 6B), and therefore its binding does not require any conformational change of the capsid. It has been shown that picornavirus genomes are released through channels at the icosahedral twofold axes that form upon transition of the virions to the A state (3, 20). Therefore, binding of E18 to mature virions results in a conformational change to A-like particles and thus facilitates the release of the genome. Hence, the release of the genome upon E18 binding is achieved by a mechanism that could be similar to that of a receptor binding to a virus. The present study of E18 binding to EV71 may indirectly provide a description of a mechanism by which picornavirus receptors that bind outside of the canyon induce genome release.

Table 2.

EMfit statistics for fitting of Fab fragments into cryo-EM electron density maps.

Fab	Sumf	Clash	−Den
E18	54.9	0.0	5.0
E19	38.6	0.0	17.0

Clash, percentage of atoms in the model that have clashes with symmetry related protein molecules; −Den, percentage of atoms positioned in negative density; Sumf, average value of density at atomic positions normalized by setting the highest density in the map to 100.

The effect of antibody binding on mature EV71 virions was also evaluated with an assay that measured the interaction of the genome with the RNA-binding fluorescent dyes SYBR Green I and II (Fig. 3A and Fig. S3) (20, 21). The results demonstrated that binding of E18 antibodies or Fab fragments to EV71 increased accessibility of the genome to the dyes at temperatures between 4 °C and 37 °C, whereas E19 had limited effect on genome accessibility (Fig. S3). However, these experiments do not differentiate between the genome being released from a virion and then interacting with the dyes outside the virion or the dyes entering the virion and interacting with the genome remaining inside the virion. However, these experiments do confirm that binding of E18 to the virus induces a conformational change that allows communication between the inside of the capsid and the external environment.

The Fab fragment of the E19 mAb binds primarily to the VP3 “knob” (4), a different site than that occupied by E18 (Fig. 6). Binding of E19 Fab to EV71 did not induce any detectable rearrangements of the capsid relative to the native state (Fig. 4 C, F, and I). The correlation coefficient of the capsid of mature EV71 with the capsid of EV71 complexed with E19 was 0.81, whereas its correlation with the A particle capsid (heated mature virus) was only 0.39 (Table 1). In other picornaviruses, binding of receptors to the knob region (Fig. 6) does not induce transition of particles to the A state (9, 10), but require additional coreceptors for successful infection (10). The E19 footprint on the EV71 virion surface does not overlap with any putative receptor binding sites (19, 22). However, it is possible that E19 might neutralize the virus by preventing EV71 binding to an as yet unidentified cellular receptor. The geometry of binding of E18 and E19 Fab fragments indicates that neither of the antibodies could bind to the virus divalently as intact IgG.

Virus capsids evolved to serve as efficient vessels for transport of virus genetic material from one host to another. However, some of the functions that the capsids perform exert conflicting selection pressures on the design of the capsid. For instance, the capsids need to be stable to protect the viral genomes in the extracellular environment, but they also need to release the genome at the right time to initiate infection. Therefore, the capsids are selected for optimal—not too high, not too low—stability. Infection of some picornavirus genera can be limited by small molecules that bind with high affinity into the VP1 pocket in place of the pocket factor (23, 24). These compounds inhibit infection by overstabilizing the virions (14, 23). In contrast, the present results show that antibody binding can promote transition of virions to the A state and deactivate EV71 by inducing untimely genome release. It has been shown here that antibodies capable of causing the release of genome can be generated by immunization with empty particles containing VP0. As an example of this strategy, empty immature virus-like particles (VLPs) were purified on an E18 affinity column and were used to immunize mice. The resultant sera were assayed to determine the neutralizing antibody titres (Fig. S4). Mice immunized with VLPs exhibited higher neutralizing antibody titers against EV71 (geometric mean titer, 153) compared with control mice (geometric mean titer, 55). Pooled serum from VLP-immunized mice inhibited 60% of E18 binding indicating that serum from these mice contains antibodies that recognize E18 epitope (Fig. S4). Thus, VLPs selected for having the E18 epitope can induce neutralizing antibodies in mice. Therefore, therapeutic antibodies with genome-release activity might also be obtainable for other picornaviruses using the approach described here.

Materials and Methods

Preparation of Fab Fragments of Monoclonal Antibodies.

The Fab fragments of the antibodies were prepared with the use of the Pierce Fab Preparation Kit according to the manufacturer’s instructions. Animal care and use was conducted in accordance with the National Animal Welfare Standards and Guidelines of Malaysia under the Animals Act of 2006.

Immunoblot Analysis.

Equal volume of mock- and EV71-infected cell lysates were separated on a 12% (wt/vol) SDS/PAGE, transferred to nitrocellulose membrane, and probed with R525 (polyclonal antibody against EV71 VP1), E18, or E19. Bound antibody was detected by incubation with HRP-conjugated secondary antibodies (Dako) followed by TMB membrane peroxidase substrate (KPL).

ELISA Analyses.

An indirect ELISA was performed by coating Nunc-Immuno plate with recombinant viral proteins or heat-inactivated EV71-infected rhabdomyosarcoma cell lysates as positive control. Nonspecific binding was blocked using 5% skim milk, antibodies were added at various concentrations in duplicate, and bound antibodies were detected by using HRP-conjugated anti-mouse IgG (Dako). SureBlue Reserve TMB microwell peroxidase substrate (KPL) was added for 5 min, 0.5 M HCl was added to stop the enzyme reaction, and wells were read at 450 nm.

A sandwich ELISA was performed whereby the wells were coated with R525 antibody against VP1, and PEG-precipitated EV71 that was untreated or heat-inactivated at 56 °C for 30 min was allowed to bind to the VP1 antibody. The bound particles were detected by the monoclonal antibodies E18 or E19, followed by HRP IgG (Dako) as described earlier.

A competitive ELISA was conducted to examine the presence of antibodies containing E18 epitope in mouse serum. Sera from four mice immunized with VLP that had high plaque reduction neutralization tests were pooled, and sera from four mice immunized with PBS solution were pooled for the competitive ELISA. Wells were coated with R525 followed by equal protein concentrations of mock- and EV71-infected RD cell lysates. Sera pooled from mice (at 1/250 dilution) were added into the wells, and HRP-conjugated E18 was added immediately afterward. Reserve TMB microwell peroxidase substrate (KPL) was added for 5 min, 0.5 M HCl was added to stop the enzyme reaction, and wells were read at 450 nm. Adjusted OD values were obtained by subtracting OD of mock RD cell lysate from OD of EV71-infected RD cells. The relative percentage of binding of E18 was derived by dividing the adjusted OD of samples by the adjusted OD of well containing only HRP-E18, multiplied by 100.

Plaque Reduction Neutralization Test.

Different concentrations of Abs, Fab fragments, or heat-inactivated mouse serum were incubated in 1:1 volume ratios with infectious EV71 strain MY104 (300 pfu/mL) for 1 h at 37 °C. The virus–antibody (or Fab) mixture was inoculated in duplicates over Vero cell monolayers in 24-well plates (Nunc/Thermo-Fisher). The monolayers were prepared with 0.5 mL per well of Vero cells at 3 × 10⁵ per milliliter in DMEM supplemented with 5% FBS and antibiotics (all from Invitrogen) and left to adhere overnight before inoculation. Media was aspirated before inoculation with 200 μL of the antibody (or Fab)–virus mixtures and incubated in a CO₂ incubator at 37 °C for 2 h before 1 mL of overlay was added containing DMEM supplemented with 2% FBS, antibiotics, and 1.5% carboxymethyl cellulose. Plates were incubated at 37 °C with 5% CO₂ for 4 d and stained with naphthalene black. Plaques were counted manually. The percent inhibition was determined relative to controls in which the mean number of plaques in wells in which the virus had been incubated with media alone.

Virus Production and Purification.

EV71 virions were produced and purified as described previously (25).

VLP Production and Purification.

Briefly, EV71 empty immature capsids were produced using a baculovirus expression system in which the complete P1 coding sequence and the protease 3CD of EV71 were recombinantly inserted downstream of the polyhedrin promoter and the recombinant baculovirus was used to infect Sf9 cells at a multiplicity of infection of 0.1. The supernatant harvested on day 4 was clarified and concentrated by using tangential flow filtration (GE Healthcare Lifesciences), and the retentate was run through an affinity column prepared by coupling E18 to a HiTrap NHS-activated HP column (GE Healthcare Lifesciences). The particles bound were eluted by using a glycine buffer at pH 3.0 and immediately neutralized to pH7.2 with 1 M Tris⋅HCl. The particles were transferred to Dulbecco's phosphate-buffered saline buffer (Invitrogen).

Immunization of Mice.

Mice (n = 10 per group) were immunized with two doses of DPBS or 10 µg of VLP in the presence of Imject Alum (Thermo Scientific) 3 wk apart. Serum were inactivated by incubation at 56 °C for 30 min, and stored at −20 °C for further analysis.

CryoEM Data Collection and Reconstruction.

Either the E18 or E19 Fab fragments were incubated with EV71 at 37 °C for 1, 2, or 6 h at a ratio of three Fab fragments per icosahedral asymmetric unit of the virus. Small aliquots (3.5 μL) of this mixture were applied to holey carbon-coated grids, blotted with filter paper, and vitrified by plunging into liquid ethane. Electron micrographs were recorded on Kodak SO-163 film by using a Philips CM200 FEG microscope. Micrographs were digitized with a Nikon Cool-Pix scanner. The final averaged pixel size was 2.48 Å. The program e2boxer.py was used to box 8325 and 13346 particles for the E18 and E19 complexes, respectively (26). The particles were corrected for the contrast transfer function using the programs ctfit and e2projectmanager.py from eman and eman2 packages (26, 27). The defocus ranged from 1.12 to 3.67 μm. The reconstruction was started by combining projections down twofold, threefold and fivefold axes using the program starticos from the eman package (27). The EM reconstruction processes were performed using icosahedral averaging with the same software. The resolution of the resulting maps were estimated by comparing structure factors of the virus shell computed from two independent half data sets (Fig. 3). For the final 3D reconstruction, data were included to the resolution (approximately 16 Å, 9 Å, and 13 Å for the complexes with empty E18, full E18, and full E19, respectively) at which the correlation between the Fourier coefficients of two independent data sets was better than 0.3.

Fitting of Fab Protein Data Bank Models into the Cryo-EM Density.

The program EMfit was used to calibrate the exact magnification of the cryo-EM map of EV71 reconstructions by comparing them with maps derived from the crystallographically determined coordinates of EV71 mature and immature particles [Protein Data Bank (PDB) ID codes 3ZFE and 3VBO]. For the EV71–E19 complex, the mature EV71 structure (PDB ID code 3ZFE) was used to model the capsid. For the EV71–E18 complex, the model of the capsid was based on the immature EV71 particle (PDB ID code 3VBO). “Difference maps” were then calculated by masking out the density of the capsid by setting to zero all grid points within 3 Å from any EV71 capsid protein atom. Modeled Fab fragments (based on PDB model ID number 1QGC) were then fitted into the difference map by using the program EMfit (28) (Table 2).

Buried Surface Area and Residues Forming the Protein–Protein Interface.

The residues forming the virus–Fab interfaces were identified with the Web service Proteins, Interfaces, Structures, and Assemblies at the European Bioinformatics Institute (www.ebi.ac.uk/msd-srv/prot_int/pistart.html) (29) based on buried surface area between the fitted Fab fragments and capsid proteins.