Structural basis of a potent human monoclonal antibody against Zika virus targeting a quaternary epitope

Neutralization and Protection Activity of the ZIKV-195 Antibody.

The ZIKV-195 mAb broadly inhibits ZIKV strains of African (Dakar 41525 and MR 766), Asian (H/PF/2013 and Malaysia P6740), and American (Paraiba 2015) origin, with 50% half maximal inhibitory (IC₅₀) values in a range of 77–600 ng/mL (Fig. 2A). However, ZIKV-195 did not show neutralizing activity against representative strains of the DENV 1–4 serotypes when tested at a concentration of 20 μg/mL. ZIKV-195 showed therapeutic activity in a stringent postexposure protection model in mice against lethal infection. A single 100-μg (5-mg/kg) dose administered at day 5 after ZIKV inoculation significantly reduced rates of lethal infection (Fig. 2B). This level of therapeutic protection is comparable to or better than that observed previously with mAbs binding to the cross-reactive E dimer epitope (EDE) or a ZIKV-specific epitope in domain II at the interdimer interface (22, 31).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Functional activity of ZIKV-195. (A) Breadth of neutralization activity of ZIKV-195. Indicated ZIKV strains were incubated with increasing concentrations of ZIKV-195 for 1 h at 37 °C before inoculation of Vero cell monolayers. Subsequently, a focus reduction neutralization test was performed. The results are representative of two independent experiments, each performed in triplicate. Error bars indicate SEM. (B) Protective, therapeutic activity of ZIKV-195 in adult male mice. Wild-type male 4- to 5-wk-old mice were treated with 1 mg of anti-Ifnar1 mAb followed by s.c. inoculation with 10⁵ FFU of ZIKV-Dakar 41525. Mice were treated with a single 100-μg dose of isotype control mAb (hCHK-152) or ZIKV-195 on D + 5 (n = 10 per group from two independent experiments). Significance was analyzed by the log-rank test (**P < 0.01).

Cryo-EM Structure of ZIKV Complexed with ZIKV-195 Fab Fragments.

To understand the structural basis for its potent activity, purified mature ZIKV strain H/PF/2013 was incubated with excess ZIKV-195 Fab fragments (2 Fab: 1 E molar ratio) at room temperature to form complexes. The virion–Fab complexes were frozen on grids and imaged using a Titan Krios microscope with a Gatan K2 direct electron detector. The particles became spiky in appearance, in contrast to the smooth surface of the native mature unliganded ZIKV, confirming the attachment of the ZIKV-195 Fab fragments to the virions. Similar to most flavivirus preparations for EM studies, some heterogeneity of the particles was observed. After nonreference 2D classification, the classes consisting of broken particles, those lacking icosahedral symmetry, or those lacking an intact lipid layer were discarded. About 21% (a total of 10,687) of the particles were used for 3D reconstruction (SI Appendix, Fig. S1A). The final cryo-EM map of the ZIKV–Fab complex was determined to a resolution of 4.0 Å, according to the 0.143 Fourier shell correlation criterion (SI Appendix, Fig. S1B).

The overall cryo-EM map showed good connectivity of density (Fig. 3A and SI Appendix, Fig. S1C). However, the constant domains of the Fab fragments were not well resolved in the density map, suggesting some flexibility of the distal regions of the Fab molecules upon binding to the virus. There were three Fab fragments bound to the E proteins within each icosahedral asymmetric unit, which extended radially from the viral surface. This finding is equivalent to 180 copies of Fab fragments per virion and complete saturation of all available binding sites. However, different contour levels were required to visualize the bound Fab molecules (SI Appendix, Fig. S2), suggesting different occupancy of the Fab molecules, which was presumably associated with the chemical differences at the three independent binding sites.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Structure of ZIKV in complex with Fab ZIKV-195. (A) Cryo-EM map of the ZIKV-Fab complex at a contour level of 1.5σ. The map is radially colored. The icosahedral asymmetric unit is outlined by a black triangle. (B) Fab variable domain (heavy chain in magenta and light chain in cyan) bound to one E dimer (DI in red, DII in yellow, DIII in blue, and transmembrane domain in tan). Only one Fab of the two Fabs bound to an E dimer is shown for clarity. (C) The binding interface, outlined by a rectangle in B, is enlarged.

The structure of the E protein (35) was fitted independently into each of the three E protein locations within the icosahedral asymmetric unit of the ZIKV–Fab complex cryo-EM density map. The resultant structure was refined against the experimental density map in real space (SI Appendix, Table S1). The superimposition between the E protomers of the complex structure and of the high-resolution native ZIKV structure showed an rmsd of ∼0.9 Å between equivalent Cα atoms. This finding showed that binding of Fab ZIKV-195 to the virus did not induce any significant structural change other than about 10 residues of the i–j loop in the E protein (SI Appendix, Fig. S3).

The density map of the complex had lower density in the Fab region than in the E proteins (SI Appendix, Table S2). Thus, the bound Fabs did not attain full occupancy. The average density heights of the E proteins and the variable domains of ZIKV-195 Fab were determined using sumf in the program EMfit, where sumf is the average density height of the fitted atoms (36). The occupancy of each Fab binding site was then estimated as the ratio of the sumf value of the variable domain of the bound Fab molecule to the average sumf value of the three E monomers in an icosahedral asymmetric unit (SI Appendix, Table S2). The occupancy of the Fab fragments closest to the icosahedral fivefold axes was ∼62%, whereas that of the Fab fragments bound closest to the icosahedral twofold axes or threefold axes was ∼37% or ∼23%, respectively. Because there is no obvious steric hindrance between the bound Fab molecules, the difference of occupancy at the three independent binding sites is most likely accounted for by the structural difference of the environment of these binding sites within the same icosahedral asymmetric unit.

Footprint of the ZIKV-195 Antibody.

The near-atomic resolution of the density map enabled the building of the Fab structure, especially the complementarity determining regions (CDRs) of the Fab variable domains, which principally contribute to epitope recognition. One Fab molecule of the ZIKV-195 mAb was found to interact with the two E protomers that form a dimer within a raft (Fig. 3B). The total buried area of the interface between the E proteins and the ZIKV-195 Fab is ∼1,400 Ų. The footprint of the ZIKV-195 Fab on one ZIKV E dimer consists of the b–d strands, the i–j loop, and the fusion loop in DII within one E protomer (E), as well as the Asn154 glycosylation loop in DI, the k strand, and the hinge region of DII in the adjacent E protomer (E′) (Fig. 3C). Most of the CDR loops of Fab ZIKV-195 interact with ZIKV E proteins. The contact surface between the heavy (H) chain and the E proteins is about 790 Ų, whereas the interface between the light (L) chain and the E proteins is about 575 Ų. Hydrogen bonds and salt bridges are formed between the CDR loops of the H and L chains and the E protomers: (i) the L-CDR3 loop of Fab and the Asn154 glycosylation loop of E′-DI; (ii) the L-CDR2 and H-CDR3 loops of Fab and the b strand of E-DII; (iii) the H-CDR3 loop of Fab and the fusion loop of E-DII; and (iv) the H-CDR3, L-CDR1, and L-CDR2 loops of Fab and the i–j loop of E-DII (SI Appendix, Table S3).

To corroborate these structural studies and define key amino acid residues important for binding, we mapped the epitope of ZIKV-195 using alanine-scanning shotgun mutagenesis of the prM–E genes coupled with mammalian cell expression and flow cytometry (37). Residues D67, M68, R73, and K251 of the E protein were critical for ZIKV-195 recognition as mutation to alanine resulted in substantial loss-of-binding of ZIKV-195 without disrupting the binding of other conformation-dependent MAbs (Fig. 4A and SI Appendix, Table S4). These observations are consistent with the cryo-EM results because these residues are located within the ZIKV-195 Fab footprint (Figs. 4B and 5).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Alanine scanning epitope mapping of mAb ZIKV-195. (A) The mAb reactivity for each alanine mutant of ZIKV prM–E was tested and normalized to reactivity for the wild-type prM–E. Mutants with reactivity less than 30% were identified as critical for mAb binding. The unrelated ZIKV-15, ZIKV-116, and ZIKV-117 mAbs were used as internal controls. Bars were calculated from at least two replicates of independent experiments. (B) Comparison of the cryo-EM and alanine mutagenesis mapping for the ZIKV-195 epitopes. Surface top view of the ZIKV E dimer is shown with DI, DII, DIII, and fusion loop colored in red, yellow, blue, and green, respectively. Epitopes identified in the ZIKV-Fab cryo-EM structure are shown in gray, among which critical residues from alanine mutagenesis epitope mapping are shown in magenta.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Roadmap shows footprint of Fab ZIKV-195 on the ZIKV virion. (A) The projected surfaces of ZIKV E proteins are colored according to the radial distance of the surface from the virus center. The black triangle denotes the boundary of an icosahedral asymmetric unit. The residues in the Fab footprint (defined by any atoms in the virus less than 4 Å from any atom in the bound Fab) are shown in white, among which the critical residues from alanine mutagenesis epitope mapping are shown in magenta. Three E protomers within one icosahedral asymmetric unit are outlined in black lines. (B) The footprint region near the icosahedral fivefold axis, outlined by a yellow rectangle in A, is enlarged. Residue numbers of the E protomers closest to the icosahedral twofold, fivefold, or threefold axes are incremented by 0; 1,000; or 2,000, respectively.

Although the overall structure of the E proteins of ZIKV in complex with Fab ZIKV-195 is nearly identical to that of the E proteins in the native, uncomplexed ZIKV particle, a structural rearrangement occurs for the i–j loop, which is located in the Fab footprint (SI Appendix, Fig. S3). The rmsd between the Cα atoms of the i–j loop in the complex structure and in the native structure is ∼3.6 Å. The quality of the loop density is greatest for the Fab binding sites near the icosahedral fivefold axes and worst for the sites near the icosahedral threefold axes. Furthermore, the quality of the density correlated with the occupancy of the Fab molecules. The i–j loop has large structural differences between some flaviviruses (8, 35). This finding suggests that the i–j loop has greater mobility than other parts of the E protein structure, consistent with its displacement on ZIKV-195 binding. The flexibility of the i–j loop may be used by the ZIKV-195 Fab molecule to create a favorable binding site.

Strand b, the Asn154 glycosylation loop, and the i–j loop of the E protein constitute a large part of the Fab ZIKV-195 epitope. The amino acids in these regions have considerable diversity among flaviviruses but are mostly conserved within different ZIKV strains, providing a basis for the specificity of ZIKV-195 and lack of neutralization of different DENV serotypes (Fig. 2 and SI Appendix, Figs. S4 and S5). Because dengue viruses usually have an N-linked glycosylation site at position 67 of the E protein, this glycan would interfere with binding of ZIKV-195.

The exposed solvent accessible regions of the three nonequivalent Fab binding sites in the icosahedral asymmetric unit have areas of ∼1,608, ∼1,499, or ∼1,414 Ų, for the footprints close to the icosahedral fivefold, twofold, or threefold axes, respectively. The footprints of these three binding sites are composed of similar amino acids but are presented slightly differently, explaining the different occupancies of these sites (Fig. 5).

Structural Insight into the Neutralization Mechanism of ZIKV-195 mAb.

Compared with other ZIKV-neutralizing antibodies with known high-resolution structures of the Fab–ZIKV or Fab–ZIKV E complexes (21, 30, 32), the footprint of ZIKV-195 is fairly unique (SI Appendix, Fig. S6). The footprint of Fab ZIKV-195 includes multiple secondary structural elements on DI and DII within an E dimer, with E-DII providing ∼89% of the interaction surface. Fab ZIKV-195 binds to both E protomers forming dimers within one icosahedral asymmetric unit, thereby stabilizing the E dimers on the mature ZIKV particle and possibly preventing formation of fusogenic trimers. Furthermore, interaction between Fab ZIKV-195 and the fusion loop of E likely would protect against the structural transitions required for membrane fusion. The conformational transition of the E proteins from dimers to fusogenic trimers is an essential step during virus entry into host cells (38). Thus, binding of ZIKV-195 to the mature ZIKV particle might result in inhibition of membrane fusion.

The Asn154 glycosylation loop of the ZIKV E protein contains an N-linked glycosylation site at residue Asn154, which is conserved among ZIKV, DENV, WNV, and JEV. The Asn154 glycan was reported to be associated with ZIKV virulence and pathogenesis (39, 40). Glycosylation of the E protein at various sites was found to contribute to cell attachment of DENV (41) and pathogenesis of WNV (42). Thus, binding of Fab ZIKV-195 to the Asn154 glycosylation loop of the ZIKV E protein may also hamper infectivity and pathogenicity of the virus.

Antibody-dependent enhancement (ADE) is a potential problem for vaccines or antibody-based therapies that combat infection by flaviviruses (43). In this case, antibodies can enhance viral infection by increasing virion uptake through cellular Fcγ-receptors. Therefore, some E protein antibodies that cross-neutralize ZIKV and DENV could increase the risk for ADE (44), if the concentration of antibody were to fall below the stoichiometric threshold required for neutralization (45). ZIKV-195 mAb, because of its specificity, is less likely to enhance infection of heterologous flaviviruses. Thus, the potent neutralizing ZIKV-195 mAb could have therapeutic potential against ZIKV infection with probably low risk of ADE for DENV.