In this study, we present the generation and characterization of ZIKV E-specific polyclonal antibodies, for which we compared the binding sites and the neutralizing activity with the previously described ZIKV neutralizing antibodies. For antibody production, we immunized rabbits with different domains of ZIKV E protein, which were produced in
E. coli. The production of ZIKV E domains in
E. coli and the further immunization was already described in other studies in the literature (examples are References [
24], [
10], [
25], [
13]). Aiming to produce domain-specific antibodies, we used each of the truncated ZIKV E protein (aa 1–409), ZIKV E domains 1+2 (ZIKV ED1+2: aa 1–295) and the truncated ZIKV E Domain 3 (ZIKV ED3: aa 296–409) in their denatured form for immunization. Truncation (deletion of the C-terminal stem region and the transmembrane domain) was performed to ensure efficient production of the recombinant proteins in
E. coli. Since the antigens were used in their denatured form for immunization, the produced antibodies were expected to recognize sequential epitopes. Indeed, all our produced and purified antibodies were able to recognize the denatured ZIKV E protein from infected cells using Western blot (
Figure 4). Interestingly, the molecular weight of the detected E protein from cells infected with the Asian strain was slightly higher compared to the African strain (
Figure 4). This might be partially explained by the single glycosylation, which occurs on N154 in the Asian strain E protein, and is absent in the African strain [
26]. Additionally, four residues in E protein of the Asian strain, which are missing in the African one, might contribute to the observed shift. Importantly, to purify the antibodies from the sera, we used purified ZIKV E protein domains in their native form; therefore, all purified antibodies were expected to recognize sequential epitopes, which are surface-accessible on the soluble, folded ZIKV E protein domains. ELISA data (
Figure 3b), Western blot data (
Figure 3c), and SPR data (
Figure 8) showed indeed that the purified antibodies bind, in a domain-specific pattern, the soluble ZIKV E protein domains. Moreover, ELISA data in
Figure 5 show the capacity of our antibodies to bind the surface of intact immobilized ZIKV particles. Interestingly, the measured binding to the Asian strain-particles was stronger than the binding to the African strain-derived particles. This difference might be partially explained by the difference in the amino acid sequence between the two strains. The sequence between aa 145 and 183 in the Asian strain-E protein contains four residues, N154, D155, T156, and G157, which are missing in the Uganda strain-E protein. This sequence belongs to Domain 1 and is recognized from the purified antibodies isolated from the sera K87, K88, K89, and K90. In addition, aa 337 to 399, which are recognized from the ED3-specific antibodies (K45 and K48) and the E-specific antibody K89, contain three substitutions between the two strains: A343V, V341I, and E393D. These differences might explain, to a limited extent, the higher binding capacity to the French Polynesia strain particles. However, since the remaining defined binding sites are mainly conserved, the observed difference is unlikely to be mainly strain-related. Other possibilities, such as lower immobilizing efficiency and lower stability of the African strain purified particles, might be causative for the observed results. Binding of our antibodies to the surface of ZIKV particles (at least to the Asian strain) indicates that the recognized epitopes are not only surface-exposed in the soluble form of ZIKV E proteins, but also in the context of the intact virus. Additional methods, such as co-immunoprecipitation, can also be further applied to support these findings. Peptide array results allowed us to identify seven binding sites of our antibodies in the ZIKV E protein (
Figure 6). These sites are distributed among the three E domains, and are, in some cases, recognized to different extents from the different antibodies. Mapping of these sites on the resolved structure of ZIKV revealed that some of these sites seem to be partially or not surface-exposed on the virus surface (
Figure 7). The site P64–P71 is located partially in the interface between two E monomers, and the site P100–P101 faces the inner side of the virion (
Figure 7). These sites do not seem to be fully surface-exposed; however, the conformational dynamics of E protein (flavivirus E protein in general) might play a role in exposing such cryptic epitopes on the surface of ZIKV particles [
27]. Interestingly, the PRNT data suggest that our antibodies are not able to potently neutralize ZIKV infection in vitro under the applied conditions (
Figure 9). These results cannot be explained by low affinity binding, since SPR data showed high binding affinity with
KD in the range of nM to the corresponding ZIKV E protein domains. We suppose that the determined binding sites of our antibodies might contain non-/weak-neutralizing epitopes in the ZIKV E protein. Indeed, comparison between the binding sites, which were defined in this study, and the clearly defined epitopes of ZIKV neutralizing antibodies revealed only partial overlap, which may support our assumption. This indicates that the neutralizing epitopes are exactly defined. Therefore, our data contribute to increase our knowledge about the distribution pattern of neutralizing and non-/weak-neutralizing epitopes. Even the knowledge about non-/weak-neutralizing epitopes will be helpful for the development of robust test systems for the detection of protective antibodies. In light of the frequent fatal outcome of ZIKV infection during pregnancy, the development of test systems to determine the existence/absence of protective immune status is of major relevance. Cross-reactivity antibodies raised against ZIKV with other
Flaviviruses, such as Dengue virus and WNV, is one of the difficulties in the development of ZIKV-specific diagnostic tools. The presented data suggested that the antibodies generated and characterized in this study do not react with WNV E protein in Western blot analysis and immunofluorescence microscopy (
Figure 4) and exhibit weak/no neutralizing activity against WNV (
Figure 9). These results argue against a general cross-reactivity of these antibodies with other
Flaviviruses as we have excluded the reactivity of our antibodies against one of ZIKV-related
Flaviviruses, as WNV, therefore, the presented information might be helpful to improve specificity of ZIKV-detection tools. Testing the reactivity of the antibodies (K45–K90) against other related
Flaviviruses, such as Dengue virus, should be considered in further work to extend the information about ZIKV-specificity of these antibodies.
The fatal outcome of ZIKV infection during pregnancy requires a detailed knowledge about neutralizing and non-neutralizing epitopes for the development of robust detection systems that are able to discriminate between just cross-reactive and protective antibodies. The identification of epitopes that are not the base of a neutralizing immune response helps to draw a map showing the distribution of neutralizing and non-neutralizing epitopes. This could contribute to the development of assay systems that selectively determine protective antibodies.
In summary, we describe in this work the production of ZIKV E domain-specific antibodies, which bind, with high affinity, to sites distributed among the three E domains. These sites overlap only partially with the previously defined neutralizing epitopes of ZIKV which might explain why the purified antibodies have weak/no neutralizing activity. The information presented in this study helps to extend knowledge of the antigenicity of ZIKV E protein and the relationship between epitopes and the neutralizing activity to improve the detection assays and the selection of new candidates for vaccine development.