After its introduction into Brazil in 2015, Zika virus (ZIKV) has spread rapidly, and in February 2016, the World Health Organization (WHO) declared it a Public Health Emergency of International Concern (
1–
3). The main route of ZIKV infection is through bites by
Aedes mosquitos, but the virus may also be sexually (
4) and vertically transmitted (
5). Although most of the ZIKV infections are asymptomatic or cause only mild symptoms, there is evidence that ZIKV infection can lead to neurological complications, such as Guillain-Barré syndrome in adults (
6) and congenital birth defects, including microcephaly in the developing fetus (
5,
7,
8), likely through its ability to infect human neural progenitor cells (
9).
Whereas flavivirus envelope (E) proteins mediate fusion and are the main target of neutralizing antibodies, the nonstructural protein 1 (NS1) is secreted by infected cells and is involved in immune evasion and pathogenesis (
10). Two recent studies showed a high level of structural similarity between the E protein of ZIKV and that of other flaviviruses—such as dengue virus (DENV), yellow fever virus (YFV), and West Nile virus (WNV)—but also revealed distinct features that may be related to the ZIKV neurotropism (
11,
12). Similarly, the structural analysis of ZIKV NS1 revealed conserved features with NS1 of other flaviviruses, although with different electrostatic characteristics (
13).
A phenomenon that is characteristic of certain flaviviruses is the disease-enhancing activity of cross-reactive antibodies elicited by previous infections by heterologous viruses, termed antibody-dependent enhancement (ADE). In the case of DENV, for which four serotypes are known, there is epidemiological evidence that a primary infection protects from reinfection with the same serotype but represents a risk factor for the development of severe disease upon reinfection with a different serotype (
14). The exacerbated disease is triggered by E- and prM-specific antibodies that fail to neutralize the incoming virus but instead enhance its capture by Fc receptor–expressing (FcR
+) cells, leading to enhanced viral replication and activation of cross-reactive memory T cells. The resulting cytokine storm is thought to be the basis of the most severe form of disease, known as dengue hemorragic fever/dengue shock syndrome (
15,
16). The role of antibodies in severe dengue is supported by studies showing that waning levels of maternal antibodies in infants represent a higher risk for development of severe dengue disease (
16–
19). Whether individuals with antibodies induced by previous DENV infections can develop a more severe ZIKV infection or have higher risk of fetal transmission is unknown. Similarly, it is unclear whether ZIKV antibodies may affect subsequent DENV infection. Therefore, it is important to dissect the level of cross-reactive immunity at the B cell and T cell level in response to ZIKV infection.
To understand the role of antibodies in ZIKV neutralization and heterologous enhancement of flavivirus infection, we isolated a panel of 119 monoclonal antibodies (mAbs) from four ZIKV-infected donors from the current epidemic, of which two were DENV-naïve (ZA and ZD) and two had serological records of DENV infection (ZB and ZC) (table S1) (
20). These mAbs were selected from Epstein-Barr virus–immortalized memory B cells on the basis of their binding to ZIKV NS1 or E proteins and for their ability to neutralize ZIKV infection, and were compared, side by side, with a panel of mAbs previously isolated from DENV-infected donors (
21). The mAbs were primarily immunoglobulin G (IgG), were highly polyclonal, and carried a lower level of somatic mutations as compared with that of other acute or recurrent infections (fig. S1 and table S2).
Out of the 119 ZIKV-reactive mAbs isolated, 41 bound to NS1. ZIKV and DENV NS1 share 51 to 53% of amino acid identity (
Fig. 1, A and B, and fig. S2). The mAbs isolated from ZIKV-infected, DENV-naïve donors were to a large extent ZIKV-specific and did not cross-react to DENV NS1, whereas those isolated from ZIKV-infected, DENV-immune donors showed, as expected, a higher degree of cross-reactivity (
Fig. 1C, fig. S3, and table S3). The limited cross-reactivity of NS1-reactive mAbs from ZIKV-infected, DENV-naïve donors was corroborated by the lack of reactivity to DENV1-4 NS1 proteins of plasma from ZIKV-infected, DENV-naïve donors (fig. S4). Conversely, the plasma of the ZIKV-infected, DENV-immune donors reacted strongly to DENV1-4 NS1 proteins, which is consistent with a recall response to NS1 dominated by cross-reactive antibodies. Reciprocally, NS1-reactive mAbs isolated from DENV-immune, ZIKV-naïve donors showed variable patterns of cross-reactivity between DENV serotypes but, with the exception of a single mAb, did not cross-react with ZIKV NS1 (
Fig. 1D, fig. S5, and table S4).
Given the interest in developing a diagnostic tool that would discriminate ZIKV-specific from DENV-specific antibodies, we identified antigenic sites (figs. S1 and S2) on NS1 targeted by ZIKV-specific but not by cross-reactive mAbs (fig. S6, A to D). MAbs reacting to two distinct sites on ZIKV NS1 were used in a two-determinant immunoassay so as to detect circulating NS1 protein in body fluids during acute infection (fig. S6E). In addition, mAb ZKA35 binding to the antigenic site S2 was used as a probe in a blockade-of-binding assay (
22) so as to detect ZIKV-specific, but not DENV-specific, serum antibodies to NS1 (
Fig. 1E) and has a potential to be developed in a serological assay in order to detect clinical and subclinical ZIKV infections at the population level.
Given the role of T cells in antibody production and in immunopathology, we also investigated the specificity and cross-reactivity of CD4
+ memory T cells from the same donors using the T cell library method (
Fig. 1, F and G) (
23). NS1-specific memory CD4
+ T cells were present in the CXCR3
+ T helper 1 (T
H1) compartment and, with variable frequencies, in the CXCR3
– T
H compartment. With a few exceptions, these T
H cells were specific for either ZIKV or DENV NS1 (
Fig. 1, F and G, and fig. S7A), which is consistent with a low level of T cell cross-reactivity, even in subjects who were already exposed to DENV.
The E protein is formed by three domains: EDI, which is involved in the conformational changes required for viral entry; EDII, which contains the fusion loop; and EDIII, which may be involved in binding to cellular receptors (
24). DI, DII, and DIII of ZIKV and DENV share 35, 51, and 29% amino acid identity, respectively (
Fig. 2A). According to the structure of the ZIKV E protein dimer (
11,
12), the majority of the ZIKV- and DENV-conserved solvent-accessible residues are located in EDII, particularly in the fusion loop and the neighboring region (
Fig. 2B and fig. S8). Strikingly, the majority of ZIKV EDI/II–specific mAbs isolated from the four ZIKV donors (65%; 24 of 37) cross-reacted with the E protein of all four DENV serotypes (
Fig. 2C, fig. S9, and table S5). Similarly, a large proportion (67%; 31 of 46) of DENV EDI/II–reactive mAbs isolated from DENV donors were also cross-reactive with ZIKV E protein (
Fig. 2D, fig. S10, and table S6). These data indicate that the human antibody response to the E protein and in particular to EDI/II is cross-reactive between ZIKV and DENV, which is consistent with previous reports on mAbs from mice immunized with flavivirus antigens (
11,
25). ZIKV-immune plasma was also found to cross-react with ZIKV and DENV1-4 E proteins (fig. S11), possibly because of the abundant EDI/II cross-reactive antibody response (
21,
26–
28).
In contrast, most of the EDIII-reactive mAbs (90%; 27 of 30) isolated from ZIKV or DENV donors were specific for either ZIKV or DENV E protein (
Fig. 2, C and D; figs. S9 and S10; and tables S5 and S6). In addition, whereas the EDI/II mAbs showed only modest or even no neutralizing activity against ZIKV infection, the EDIII-reactive antibodies were overall more highly neutralizing than EDI/II–reactive antibodies (table S5). By screening for ZIKV neutralization, we also isolated several highly potent neutralizing antibodies that failed to bind to ZIKV E and EDIII proteins (
Fig. 2C). These mAbs, that we define as neutralizing non-E-binding (NNB), could recognize quaternary epitopes that are displayed on the infectious virions but not on soluble proteins, as recently demonstrated for DENV (
29–
31). E-specific CD4
+ T cells were primarily detected in the CXCR3
+ T
H cell subset and, with only a few exceptions, were specific for ZIKV E protein (
Fig. 2, E and F, and fig. S7B). Furthermore, E-reactive T cells directly isolated from ex vivo cultures of memory CD4
+ T cells of ZIKV- or DENV-infected donors (fig. S12) did not cross-react upon secondary stimulation with heterologous E proteins (
Fig. 2G). These findings indicate an overall low level of T cell cross-reactivity between ZIKV and DENV E proteins.
To address the biological properties of ZIKV and DENV antibodies, we used a selected mAb panel and measured in parallel their binding (
Fig. 3A), neutralizing, and ADE activity (
Fig. 3, B and C). This panel also contained Fc mutant versions of mAbs that do not bind to FcγR and complement [LALA mutants; leucine (L) to alanine (A) substitution at the position 234 and 235] (
21,
32). The EDIII-specific mAbs ZKA64 and ZKA190 and the NNB mAb ZKA230 were highly potent in ZIKV neutralization, with median inhibitory concentration (IC
50) values of 93, 9, and 10 ng/ml, respectively. Furthermore, all these mAbs enhanced infection of ZIKV in the nonpermissive K562 human erythroleukemic cell line at a broad range of concentrations, including those that fully neutralized ZIKV infection on Vero cell line. Whereas EDIII mAbs ZKA64 and ZKA190 did not enhance ZIKV infections of K562 cells above 1 μg/ml, the NNB mAb ZKA230 failed to do so, a result that might be due to the different mechanisms of neutralization of free viruses versus FcγR-internalized viruses. In contrast, the cross-reactive EDI/II–specific mAbs ZKA3 and ZKA78 only partially neutralized ZIKV infectivity, while effectively enhancing infection of K562 cells. Consistent with their cross-reactivity, these EDI/II–specific mAbs also neutralized and enhanced DENV1 infection.
The above results suggest that cross-reactive antibodies elicited by either ZIKV or DENV infection and primarily directed to EDI/II can mediate heterologous ADE. We therefore asked whether ADE could be also induced with immune sera and whether this could be blocked by neutralizing mAbs delivered in a LALA format (
Fig. 3, D and E). In a homologous setting, four ZIKV-immune plasma collected from convalescent patients showed similar capacity to enhance ZIKV infection of K562 cells, and this ADE effect was completely blocked by the EDIII-specific ZKA64-LALA mAb, but only partially inhibited by the cross-reactive EDI/II DV82-LALA mAb (
Fig. 3D). In a heterologous setting, the four ZIKV-immune plasma strongly enhanced infection by DENV1 to levels comparable with those observed with DENV3 plasma. The enhancement of DENV1 infection by both ZIKV and DENV3 plasma was completely inhibited by DV82-LALA, but not by ZKA64-LALA because of its lack of cross-reactivity to DENV1 (
Fig. 3E). The ADE-blocking ability of a single EDIII-specific LALA mAb could be related not only to its capacity to out-compete serum-enhancing antibodies but also to neutralize virus once internalized into endosomes, similarly to what has been reported for the WNV-neutralizing antibody E16 (
33). Conversely, the lack of ADE-blocking ability of DV82-LALA might be explained by the inability of antibodies to EDI/II to effectively neutralize endocytosed virus.
The above results suggest that previous ZIKV and DENV immunity may pose a risk of a more severe disease upon exposure to heterologous virus through ADE. To address this possibility in vivo, we selected two EDI/II cross-reactive mAbs elicited by ZIKV (ZKA78) or DENV (DV82) and tested them for their capacity to enhance infection by DENV or ZIKV in animal models. Preadministration of ZKA78 or DV82, but not their LALA versions, to DENV2-infected AG129 mice led to severe symptoms and lethality by day 5 (
Fig. 4A), suggesting that ZIKV immunity could predispose to enhanced DENV pathology in vivo. Given the high lethality of ZIKV infection in immunodeficient mice (
34–
36), we investigated a possible ZIKV disease–enhancing activity of the DENV cross-reactive mAb DV82 in 129Sv/ev immunocompetent mice that are not permissive for productive ZIKV infection. However, in this mouse model we did not observe signs of enhanced disease or infection (fig. S13).
To develop a therapeutic approach to ZIKV infection, we tested the LALA form of the potently neutralizing EDIII-specific mAb ZKA64 in prophylactic and therapeutic settings. ZKA64-LALA completely protected A129 mice challenged with a lethal dose of ZIKV from body weight loss and lethality when given 1 day before or 1 day after ZIKV challenge (
Fig. 4B).
This study reports the first characterization of the human immune response to ZIKV infection. The highly cross-reactive antibodies to EDI/II elicited by ZIKV or DENV infection that are poorly neutralizing but potently enhancing may pose a risk for heterologous ADE. We have shown in vitro and in a relevant animal model that an EDI/II cross-reactive mAb raised by ZIKV can induce lethal DENV infection. However, enhancement of ZIKV infection by DENV-elicited cross-reactive antibodies could be observed in vitro but not in vivo, a finding that may be due to the tropism of the virus or to the limitation of the mouse model. It will be important to address this issue in clinical and epidemiological studies, which may be facilitated by the development of serological diagnostics, such as the blockade-of-binding assay described in this study.
T cells in flavivirus infections play a complex role in both protection and pathogenesis (
15). The low degree of CD4
+ T cell cross-reactivity between DENV and ZIKV, which we observed even in individuals who are immune to both viruses, suggests that in ZIKV infection original antigenic sin may not play a pathogenic role. Our findings also suggest that the risk of cytokine storm and consequent severe disease after enhanced heterologous infection of DENV by antibodies to ZIKV may be mitigated by the poor T cell cross-reactivity.
Our study describes two classes of potent neutralizing antibodies that are specific for ZIKV and directed either against EDIII or quaternary epitopes present on infectious virus, the latter being particularly frequent, similar to what has been described in DENV (
30). Given the high potency and in vivo efficacy shown in this study, these antibodies, developed in wild type or their LALA versions produced in wild-type version or in LALA version so as to avoid possible enhancement, could be used in prophylactic or therapeutic settings to prevent congenital ZIKV infection in pregnant women living in high-risk areas (
37).
Acknowledgments
The data presented in this manuscript are tabulated in the main paper and in the supplementary materials. The views expressed are those of the authors and not necessarily those of the funding bodies. This work was supported by grants from the European Union (HEALTH-F3-2011-281803-Project IDAMS, “International Research Consortium on Dengue Risk Assessment, Management, and Surveillance” to F.S.), the European Research Council (grant 323183 PREDICT to F.S.), the Swiss National Science Foundation (grant 160279 to A.L.), the U.S. National Institutes of Health (grant R01 AI24493 to E.H.), and the Ministero della Salute, Fondazione IRCCS Policlinico San Matteo, Ricerca Corrente grant 8020615 (to E.P. and F.B.). A.L. is supported by the Helmut Horten Foundation. A.L. is the scientific founder of Humabs BioMed. D.C., A.L., and F.S. hold shares in Humabs BioMed. F.S. and F.B. are a scientific advisor and a consultant at Humabs BioMed, respectively. The MAbs described in this study are available under a materials transfer agreement with Humabs BioMed SA. L.V. is grateful for financial support by SNF grant 310030_166445 and Lions Club Monteceneri. A.L. is an inventor on a patent related to DENV mAbs described in this study: PCT/lB2009/007372 (Dengue virus neutralizing antibodies and uses thereof). D.C. is an inventor on a patent application held by Humabs BioMed SA related to ZIKV mAbs described in this study: PCT/EP2016/066684 (Novel antibodies specifically binding to Zika virus epitopes and uses thereof). Antibody nucleotide and amino acid sequences for the 68 mAbs described in table S2 have been deposited in GenBank (accession nos. KX496807 to KX496874).