Abstract

The rapid spread of Zika virus (ZIKV), which causes microcephaly and Guillain-Barré syndrome, signals an urgency to identify therapeutics. Recent efforts to rescreen dengue virus human antibodies for ZIKV cross-neutralization activity showed antibody C10 as one of the most potent. To investigate the ability of the antibody to block fusion, we determined the cryoEM structures of the C10-ZIKV complex at pH levels mimicking the extracellular (pH8.0), early (pH6.5) and late endosomal (pH5.0) environments. The 4.0 Å resolution pH8.0 complex structure shows that the antibody binds to E proteins residues at the intra-dimer interface, and the virus quaternary structure-dependent inter-dimer and inter-raft interfaces. At pH6.5, antibody C10 locks all virus surface E proteins, and at pH5.0, it locks the E protein raft structure, suggesting that it prevents the structural rearrangement of the E proteins during the fusion event—a vital step for infection. This suggests antibody C10 could be a good therapeutic candidate.

 Introduction

Zika virus1 (ZIKV) is a member of the flavivirus genus that includes dengue virus (DENV) and West Nile virus (WNV). ZIKV cryoEM structures2,3 show its surface proteins (envelope (E) and membrane (M) proteins) are organized similar to DENV4 except with a tighter packing, making the virus more thermally stable2.

The virus surface consists of 180 copies of E protein2 arranged in icosahedral symmetry with 60 asymmetric units. In each asymmetric unit, there are three individual E proteins – molecules A, B and C. The E proteins exist as dimers; three dimers lie parallel to each other forming a raft containing two asymmetric units. There are in total 30 rafts arranged in a herringbone pattern on the virus surface.

An E protein contains three domains—DI, DII and DIII5. It is known for other flaviviruses that DIII contains the receptor-binding site and plays an important role in fusion of the virus with the endosomal membrane during cell entry6,7. The tip of DII contains a fusion loop that interacts with the endosomal membrane. DI is the central domain linking DII and DIII together. The DI-DII hinge is highly flexible allowing DII to expose its fusion loop during the fusion event. The DI-DIII hinge was thought to be more rigid but it was observed to change in conformation in the post-fusion E protein trimeric structure6,7. The fusion event is hypothesized to occur in this sequence: (1) virus E protein binds to cell receptors, (2) it is endocytosed, (3) the low pH environment of the endosome causes the E proteins to flip up exposing their fusion loops, allowing them to interact with the endosomal membrane, (4) the E proteins rearrange to trimeric structures, (5) the DIIIs of the E protein trimers change in conformation twisting the trimers leading to the fusion of viral membrane with the endosomal membrane, before the release of the viral genome into cell cytosol.

The recent explosion of the number of ZIKV cases, together with the association of ZIKV with the development of microcephaly in fetuses8 and Guillian-Barré syndrome in adults9, ignite a pressing need for the development of therapeutics. Currently there are no published human monoclonal antibodies (HMAb) generated against ZIKV. To hasten the process of therapeutics development, DENV HMAbs were rescreened10,11,12 for those that cross-neutralize ZIKV. One group of antibodies has recently been shown to be highly neutralizing to ZIKV—the envelope dimer epitope binding antibodies10,11. Of these HMAbs, C10 is one of the most potent plaque reduction neutralisation test (PRNT50=0.024 μg ml−1), as demonstrated recently in ZIKV infected cell culture11,13 and mouse model13. In addition, it can prevent antibody dependent enhancement (ADE) of ZIKV infection in myeloid cells induced by dengue human sera10. In this ADE model, the myeloid cells are mostly resistant to direct ZIKV infection, suggesting that its specific receptor is lacking. When sub-neutralizing concentrations of dengue human serum was added to ZIKV, cell infection was enhanced. This is because antibodies, which are attached to ZIKV, bind to the Fc receptor on myeloid cells thus bypassing the need for ZIKV to directly interact with its specific receptor. When HMAb C10 is added to this mixture, it neutralizes the ADE effect. Since HMAb C10 is also an antibody that would likely facilitate attachment to Fc receptor on myeloid cells, it likely neutralizes the virus at a post-attachment step of infection. We investigated the ability of Fab C10 to prevent virus surface protein rearrangement during fusion. We observed Fab C10 is able to lock the entire virus surface at pH6.5, and at pH5.0, the E protein raft thereby preventing structural rearrangement necessary for fusion.

Results

Effect of Fab C10 on ZIKV particles at different pHs

We solved the cryoEM structures of Fab C10 complexed with ZIKV at pH8.0, pH6.5 and pH5.0 mimicking the extracellular, early and late endosomal conditions, respectively, and compared them to the cryoEM maps of the uncomplexed ZIKV controls at pH8.0 (ref. 2), pH6.5 (Supplementary Fig. 1b) and the two-dimensional (2D)-class average of pH5.0 particles (Fig. 1).

Figure 1: CryoEM micrographs of the uncomplexed ZIKV control and the Fab C10-ZIKV complex samples at various pH levels.

The deformed particles and aggregates are indicated with red and black arrows, respectively. The right upper corner inset shows a quarter of a 2D class average of the round particles. The E protein layer is indicated with a green arrow, the outer and inner leaflets of the bilayer lipid membrane with cyan arrows. In the pH5.0 uncomplexed ZIKV control, the E protein layer is missing in the 2D class average. Bottom right inset in the pH5.0 uncomplexed ZIKV control is a median filtered (5 × 5 pixel) image that showed particles with hair-like protrusions (blue arrow), which are likely the E proteins flopping on the virus surface. Scale bar is 500 Å.

Micrographs of the uncomplexed ZIKV control at pH8.0 sample show mostly smooth surfaced spherical particles (Fig. 1). In the pH6.5 control sample (Fig. 1), some virus particles aggregated, others become deformed, but there are also spherical particles present. 2D class average of the pH6.5 spherical particles (Fig. 1 inset), as well as its low resolution cryoEM map (Supplementary Fig. 1a and b), show the outer E protein layer has moved to a slightly larger radius compared to the pH8.0 control virus. This suggests that the E protein layer has loosened. Micrographs of the pH5.0 control sample (Fig. 1) show aggregation of some particles, while others appear to be smaller in diameter with hair-like densities protruding from the virus surface. The 2D class average of these small particles (Fig. 1 inset) showed the absence of the E protein compact layer, which was present in the pH8.0 and pH6.5 control samples. This suggests the E proteins are likely ‘flopping’ on the virus surface. The ZIKV controls demonstrate some of the structural transformation stages of the virus particles during fusion, from the compact structure at pH8.0 to a slightly expanded structure at pH6.5 and finally to the E proteins loosening and extending out from the virus lipid membrane at pH5.0.

Micrographs of the ZIKV-C10 complexes at all pH conditions show spiky looking particles, due to the Fab molecules bound to virus surface (Fig. 1). The 2D class average of the pH6.5 complex particles shows the E protein layer to remain at the same radius as the pH8.0 control (Fig. 1 inset), unlike its pH6.5 ZIKV control. The 2D class average of the pH5.0 complex particles in contrast to its pH5.0 control shows the E protein layer is still present (Fig. 1 inset).

CryoEM structures of ZIKV-C10 at different pHs

The cryoEM structures of ZIKV-C10 complex at pH8.0, pH6.5 and pH5.0 are determined to 4.0, 4.4 and 12 Å resolution, respectively (Fig. 2, Supplementary Fig. 2). In each of these structures, there are 180 copies of Fab C10 bound to the virus surface. The pH8.0 and pH6.5 complex structures are very similar to each other (Supplementary Fig. 3a) and their cryoEM maps correlate to 4.5 Å resolution at FSC 0.143 cutoff (Supplementary Fig. 3b). Therefore, only the higher resolution pH8.0 complex structure will be described. A comparison of the E proteins of the pH8.0 complex structure with that of the previously solved uncomplexed ZIKV2 shows that they are largely the same (Supplementary Fig. 4a). Only molecule A of the E protein in the asymmetric unit on the pH8.0 complex structure shows clear densities for the ‘150 glycan loop’ (Supplementary Fig. 4b). The glycan loop changed in conformation when compared to the uncomplexed virus (Supplementary Fig. 4a, inset), likely due to its interaction with the Fab molecule (Supplementary Table 1). This glycan loop on ZIKV is five residues longer than in DENV. The previously solved crystal structure of DENV-C10 (ref. 14) did not show densities corresponding to the glycan loop; therefore, it is not known if this region interacts directly with the Fab. However, mutational studies15 indicate that the residue 153 glycosylation site on DENV is not important for HMAb C10 binding.

Figure 2: CryoEM maps of the Fab C10-ZIKV complex.

Structures at (a) pH8.0, (b) pH6.5, (c) pH5.0, determined to 4.0, 4.4 and 12 Å resolution, respectively. Left panels show the surface of the cryoEM maps. Densities corresponding to the E protein layer and Fabs are coloured in yellow and magenta, respectively. Black triangle indicates an asymmetric unit and the 5-, 3-, 2-fold vertices are labelled. Right panels show zoom-in views of the fitted molecules into the density maps. (a, right panel) The 4.0 Å resolution pH8.0 complex show well-resolved bulky side chain densities (grey mesh). The Cα backbone, the nitrogen and oxygen atoms are coloured in green, blue and red, respectively. (b, right panel) The 4.4 Å resolution pH6.5 complex map showed density (grey transparent surface) separation between the β strands. DII of E protein is coloured in yellow. (c, right panel) Densities of the 12 Å resolution pH5.0 complex showed clear borders and shapes corresponding to the Fab C10-E protein dimeric structures. The variable region of the Fab molecule, DI, DII and DIII of the E protein are coloured in green, red, yellow and blue, respectively.

In the 4.0 Å resolution pH8.0 complex cryoEM map (Fig. 2a), the likely interacting residues that form the epitope were identified, by using a cutoff of 5 Å distance16 (hydrogen bonds/electrostatic interaction: 4 Å and hydrophobic interactions: 5 Å) between side chains of the Fab and E proteins (Fig. 3, Supplementary Fig. 5d, Supplementary Table 1). We also presented the epitope identified with a cutoff of 8 Å distance17 between the Cα chains of the Fab and E proteins (Supplementary Fig. 5c). Each end of the E protein dimer has a Fab molecule attached (Fig. 3a). The Fabs bind across the E proteins at the intra-dimer interface (Fig. 3a, Supplementary Fig. 5b). The Fab bound near the five-fold vertex end of the A-C′ dimer also likely interacts with residues from the adjacent E protein at the inter-raft interface, whereas the Fab molecule at the other end is also involved in inter-dimer interactions within the raft (Fig. 3a, Supplementary Fig. 5b). The epitopes recognized by the Fab molecules that bind to B-B′ dimer also span across the inter-dimer E protein interfaces. The ability of Fab C10 to bind E proteins at the intra-dimer interface together with the virus quaternary structure-dependent sites—the inter-dimer and inter-raft interfaces, suggests that the entire E protein layer is locked. This is consistent with the cryoEM structure (Supplementary Fig. 3a) and the 2D class average (Fig. 1 inset) of the ZIKV-C10 complex at pH6.5 showing the E protein layer remains at a similar radius as the uncomplexed ZIKV pH8.0 control, unlike its pH6.5 control.

Figure 3: The C10 epitopes on the pH8.0 ZIKV-C10 complex structure.

The C10 epitopes (circled by green dots) in an E protein raft identified by using a distance cutoff of 5 Å between the side chains of Fab and the E protein. The Fab molecules bind to both ends of each E protein dimer. The DI, DII and DIII of the E proteins in one raft are coloured in red, yellow and blue, respectively, those in neighbouring rafts are in grey. The three individual E proteins in an asymmetric unit are labelled as A, B and C molecules and those in the neighbouring asymmetric unit within the raft as A′, B′ and C′. The epitope residues within the intra-dimer interface are shown as light blue sph​eres, those at the inter-dimer and inter-raft interfaces as red and dark blue spheres, respectively. (b) The epitope within the intra-dimer interface on B-B′ dimer. The ZIKV c10 epitope residues that are conserved (similar charges or hydrophobicity) and non-conserved when compared to DENV are shown as green and magenta spheres, respectively. (c) Charge complementarity of the C10 intra-dimer epitope with the Fab paratope. Positive, negative and neutral

References

1. Baronti, C. et al. Complete coding sequence of Zika virus from a French polynesia outbreak in 2013. Genome Announc. 2, e00500–e00514 (2014).

2. Kostyuchenko, V. A. et al. Structure of the thermally stable Zika virus. Nature 533, 425–428 (2016).

3. Sirohi, D. et al. The 3.8Å resolution cryo-EM structure of Zika virus. Science 352, 467–470 (2016).

4. Kostyuchenko, V. A., Chew, P. L., Ng, T. S. & Lok, S. M. Near-atomic resolution cryo-electron microscopic structure of dengue serotype 4 virus. J. Virol. 88, 477–482 (2014).

5. Rey, F. A., Heinz, F. X., Mandl, C., Kunz, C. & Harrison, S. C. The envelope glycoprotein from tick-borne encephalitis virus at 2A resolution. Nature 375, 291–298 (1995).

6. Modis, Y., Ogata, S., Clements, D. & Harrison, S. C. Structure of the dengue virus envelope protein after membrane fusion. Nature 427, 313–319 (2004).

7. Bressanelli, S. et al. Structure of a flavivirus envelope glycoprotein in its low-pH-induced membrane fusion conformation. EMBO J. 23, 728–738 (2004).

8. Calvet, G. et al. Detection and sequencing of Zika virus from amniotic fluid of fetuses with microcephaly in Brazil: a case study. Lancet Infect. Dis. 16, 653–660 (2016).

9. Cao-Lormeau, V. M. et al. Guillain-Barre Syndrome outbreak associated with Zika virus infection in French Polynesia: a case-control study. Lancet 387, 1531–1539 (2016).

10. Dejnirattisai, W. et al. Dengue virus sero-cross-reactivity drives antibody-dependent enhancement of infection with zika virus. Nat. Immunol. 17, 1102–1108 (2016).

11. Barba-Spaeth, G. et al. Structural basis of potent Zika-dengue virus antibody cross-neutralization. Nature 536, 48–53 (2016).

12. Dai, L. et al. Structures of the Zika virus envelope protein and its complex with a flavivirus broadly protective antibody. Cell Host Microbe 19, 696–704 (2016).

13. Swanstrom, J. A. et al. Dengue virus Envelope Dimer Epitope (EDE) monoclonal antibodies isolated from dengue patients are protective against Zika virus. MBio 7, e01123–01116 (2016).

14. Rouvinski, A. et al. Recognition determinants of broadly neutralizing human antibodies against dengue viruses. Nature 520, 109–113 (2015).

15. Dejnirattisai, W. et al. A new class of highly potent, broadly neutralizing antibodies isolated from viremic patients infected with dengue virus. Nat. Immunol. 16, 170–177 (2015).

16. Chung-Jung Tsai, S. L. L. & Wolfson, H. J. Ruth Nussinov. A dataset of protein-protein interfaces generated with a sequence-order-independent comparison technique. J. Mol. Biol. 260, 604–620 (1996).

17.Mukherjee, S. & Zhang, Y. MM-align: a quick algorithm for aligning multiple-chain protein complex structures using iterative dynamic programming. Nucleic acids res. 37, e83