Flaviviruses are plus-sense RNA viruses transmitted between infected humans by mosquitoes or ticks. This group includes the recently prominent Zika virus and such well-known human pathogens as dengue virus and yellow fever virus (after which the group is rather oddly named). Two collaborative papers, one by Screaton and colleagues in this issue of Nature Immunology1 and another in Nature2, and at least three other reports published concurrently3,4,5 show an unexpected antigenic relationship between Zika and dengue viruses, with implications for vaccines and perhaps for therapeutics. Other related studies are undoubtedly in the press.

Infectious flavivirus particles are compact, icosahedrally symmetric assemblies, with a lipid bilayer membrane smoothly coated with 180 copies of a three-domain envelope (E) protein (Fig. 1). E protein has three functions associated with viral entry and host defense: cell attachment, membrane fusion and antigenicity. Antibodies that recognize E protein, when present in sufficient titer, neutralize infection in cell culture and protect against disease in human hosts. Dengue and West Nile viruses can proliferate in cells of the immune system, many of which bear FcγR receptors. Sub-neutralizing concentrations of antibody can promote infection in such cells by facilitating Fc-receptor-mediated entry, a phenomenon called 'antibody-dependent enhancement' (ADE)6.

Figure 1: Antibodies that neutralize both dengue virus and Zika virus.
figure 1

In the arrangement of 180 E-protein subunits on the surface of a flavivirus particle (top left), the subunits associate as dimers, packed in a 'herringbone' pattern14; a single E-protein dimer (far left) is oriented to show how it relates to the whole (right). An expanded view of the ectodomain dimer of E protein (PDB accession code, 1OAN (dengue virus type 2)) (bottom left) shows the three domains (I (red), II (yellow) and III (blue)) and the fusion loop at the tip of domain II (orange), plus glycans and disulfides ('sticks') and the amino (N) and carboxyl (C) termini of each subunit. A single-chain Fv from an antibody to EDEI (isolated from cells of a patient infected with dengue virus (top right): dark green, heavy chain; gray, light chain) is shown bound to a Zika virus E-protein ectodomain dimer (top right; 'ribbons' in colors as at left); outlined area indicates the Fv–E protein interface and shows how the epitope spans a dimer contact. The surface of the Zika virus E dimer (bottom right), presented as if viewed from above at top right, shows the footprint of Fv (green outline) and conservation of the exposed residues. Right half of figure is from ref. 2.

The four distinct and (in recent evolution) relatively stable serotypes of dengue virus have made vaccine development much more complicated than it was for the vaccine against yellow fever—still one of the most successful vaccines ever made. The antigenic differences among the four types are great enough that robust immunity to one conferred by recovery from infection does not confer immunity to the other three. Instead, previous exposure to one serotype increases the risk of severe disease after infection by virus of a second serotype. It was proposed many years ago that ADE, mediated by low titers of cross-reactive antibodies from the primary infection, might account for the enhanced severity of the secondary infection7. Studies of infants who had passively acquired antibodies to dengue virus from their mothers and results obtained with subsequently developed mouse models support that hypothesis6. Thus, to avoid risk of enhancement, a safe vaccine against dengue virus will need to confer protective immunity against all four serotypes.

In principle, conservation across serotypes of certain surface patches on E protein suggests that it should be possible to find broadly cross-reactive and cross-protective antibodies and hence that a goal of vaccine research should be to elicit these. One such patch is the so-called 'fusion loop', critical for viral entry8. This hydrophobic loop, at the tip of domain II of E protein (Fig. 1), interacts with the endosomal membrane when low pH triggers a fusogenic conformational change in E protein. On a mature virus at a neutral pH (its state when an antibody might bind), the interface with the dimer partner subunit buries the fusion loop, making it inaccessible. In 2015, a collaboration between the groups of Screaton (at Imperial College) and Rey (at Institut Pasteur) showed that recently identified human antibodies recognize another conserved patch on the surface of the same interface that buries the fusion loop9,10. These antibodies neutralize all four serotypes quite effectively. Structures of the antibodies bound to E protein from each of the four serotypes show that the antibodies fall into two classes, with overlapping contacts. The researchers called these contact regions EDE1 ('E-dimer epitope 1') and EDE2, because they bridge the dimer interface (Fig. 1).

In the two papers now published, the same two groups have now shown that the EDE1-binding antibodies also neutralize Zika virus1,2. They illustrate by phylogenetic analysis based on the amino-acid sequence of E protein that for this protein, Zika virus is closer to dengue virus than to any of the other flaviviruses and indeed is almost close enough to think of it as a fifth serotype2, although the rest of the genome places Zika virus closer to some of the other mosquito-borne flaviruses. A structure of one of these antibodies in complex with Zika virus E protein shows the reason for the cross-neutralization: nearly all the contact residues are the same in Zika virus and in all four serotypes of dengue virus (Fig. 1). The EDE2-binding antibodies do not neutralize Zika virus at levels that would protect in vivo, but because they do bind weakly, they instead promote ADE in a cell-based assay, as the paper in this issue of Nature Immunology shows1. Two other papers confirm these results3,4.

The new data clearly influence potential strategies for vaccines against dengue and Zika viruses. The ADE observed in vitro does not show that immunity to dengue virus can enhance the risk of infection with Zika virus. That conclusion would be provided only by epidemiological analyses, together with studies in validated animal models. Moreover, it remains unknown whether enhanced infection of Fc-expressing cells would influence the course of infection with Zika virus in humans. Nonetheless, Screaton and colleagues1, along with others3,4,5, provide clear evidence of immunogenic cross-talk between Zika and dengue viruses. Four of the seven hospitalized, dengue-virus-infected patients from whom blood was obtained for the initial study had plasmablasts that encoded EDE-reactive antibodies at their rearranged immunoglobulin loci. Thus, B cells that make these antibodies are relatively common in patients with secondary infection. Moreover, plasma samples collected from patients 6 months following acute secondary infection with dengue virus all showed the presence of antibodies that bound Zika virus (as well as dengue virus) but neutralized Zika virus either marginally or not at all. Instead, most of the plasma from convalescent patients enhanced infection in an assay for ADE.

Given the current data, the epitopes of the Zika-virus-reactive antibodies in plasma from convalescent patients remain unknown. Antibodies directed against EDE1 and EDE2 might account for some of the activity, but antibodies directed against the fusion-loop epitope (FLE) can also enhance infection through a mechanism that depends on the presence of a second surface protein: precursor of M (prM)11,12. Flaviruses assemble by budding into the endoplasmic reticulum, where their core structures (genome plus a nucleocapsid protein) acquire both a lipid-bilayer membrane and 180 copies each of E protein and prM. The precursor domain of a prM subunit covers the fusion loop of each E-protein subunit, which projects outward rather than lying flat on the membrane surface13. This domain, cleaved from the rest of the protein by furin in the trans-Golgi network, dissociates from the virus when it emerges from the cell, leaving only a residual, membrane embedded segment called 'M' (Fig. 1) and allowing E protein to form the 'flat' dimers found on mature virions. The way the precursor domain of prM associates with the fusion-loop tip permits attachment of fusion-loop-directed antibodies. Because many viral particles produced during infection have incomplete cleavage of prM, antibodies to FLE can bind these partly immature particles and enhance the infection of Fc-receptor-expressing cells. Exposure of FLE on partly immature particles might also account for the prevalence of antibodies to FLE elicited by infection with dengue virus2.

Each of the five papers1,2,3,4,5 outlines some of the next steps for understanding how the potential cross-reactivity of human immune responses to dengue and Zika viruses can affect plans for the development of vaccines or antibody-based therapeutics. Clearly, it is essential to know whether there is any epidemiological evidence that immunity to one virus influences the risk of infection after exposure to the other. Cross-neutralization of Zika virus and all four serotypes of dengue virus by EDE1-directed antibodies also suggests the value of more extensive cross-protection studies in animal models to determine how worthwhile it would be to pursue proposals for vaccine immunogen design and especially to suggest whether antibody-based therapeutics are a sensible objective. As the inconclusive outcome of studies during the 2015 outbreak of Ebola virus illustrates, analysis of the efficacy (or lack thereof) of passive transfer after infection or after diagnosis should occur before an emergency, not once the need becomes acute.

These papers collectively offer both 'good news' and 'bad news'1,2,3,4,5. They show that certain members of a group of well-characterized antibodies that neutralize all four serotypes of dengue virus can neutralize Zika virus as well. These antibodies might indeed be useful for treatment or prophylaxis. The papers also show, however, that in the plasma of convalescent patients infected with dengue virus, these cross-neutralizing antibodies are present at a titer too low to neutralize Zika virus. In any case, the close antigenic relationship between Zika and dengue viruses that these papers illustrate adds a new dimension to the study of each.