Enhancement of Zika virus infection by antibodies from West Nile virus seropositive individuals with no history of clinical infection

The study involved recruiting two groups of subjects 1) individuals from the general population in the El Paso area self-reported as never having been diagnosed with WNV infection (N = 388; asymptomatic) and 2) subjects with a previous history of documented WNV infection recruited in collaboration with the public health department (N = 20; symptomatic). The study participants were requested to donate a blood sample and complete a study questionnaire. Serum samples from all the participants were analyzed for WNV Envelope IgG antibodies via ELISA (Fig. 1a). Of a total of 388 participants, 28 individuals (7.2%) were positive for WNV antibodies with an ELISA cut-off value of 1.1 or higher (Fig. 1b). Interestingly, 8 subjects showed borderline sero-positivity (ELISA OD > 0.8 to < 1.1) and could be due to the presence of low titer WNV antibodies or presence of cross reactive antibodies to related flaviviruses. Rest of the subjects showed serum ELISA OD value of 0.8 or lower. Samples were also obtained from individuals with a previous history of a symptomatic WNV infection and/or hospitalization in collaboration with the Public Health Department (City of El Paso). All of the symptomatic subjects showed ELISA O. D values of 1.1 or higher except one who was an OD value of 0.03; the reason for which remains undetermined (Fig. 1a). There was no significant difference (p = 0.09) between the WNV E IgG ELISA OD values of symptomatic versus asymptomatic individuals (Fig. 1b). Subject SMP003 was reported as having a symptomatic WNV infection to the public health department. However, neutralization and ADE data indicated that the subject was positive for ZIKV antibodies which was confirmed by ZIKV NS1 IgG ELISA (EUROIMMUN) (Table S1). Of the WNV seropositive samples, four (N = 4) were also positive for DENV1–4 NS1 IgG in an ELISA (Table S1 and S2).

The demographics of the study population are described in Tables 1 and 2. The healthy volunteer population comprised of 214 females (~ 55%) and 174 males (~ 45%) (Table 1). There was a significantly higher percentage (p = 0.0140) of subjects with age ≥ 50 in symptomatic WNV+ participants confirming that symptomatic WNV infections are often seen in older individuals (Table 2). There was a significantly higher percentage (p = 0.0166) of seropositive males than females (Table 3) which did not correlate with time spent outdoors (Table S3). Interestingly, certain zip codes of residence showed higher WNV sero-positivity for WNV (Fig. S1 and Table S4) which also correlated with occurrence of standing water (Table S5).

K562 cells bearing the Fc gamma receptor have been widely used to study ADE in vitro [23]. The cells are relatively non-permissive to flavivirus infection in the absence of specific binding antibodies. Binding of antibody to virus allows interaction of virus antibody complex with Fc gamma receptor on K562 cells that in turn permit virus uptake via endocytosis resulting in infection. While the K562 cell based ADE assay is widely used by researchers, the process involves infection of cells with live virus, staining cells for viral proteins as a measure of infection and quantifying the number of infected cells via flow cytometry. We wanted to develop a high throughput assay for ADE so larger number of samples could be tested against multiple viruses. Hence, we adapted the classical ADE assay in K562 cells using Reporter Virus Particles (RVPs) for infection (in 96 wells) and quantifying the number of infected cells via automated microscopy. We then compared the read out of ADE assay via automated microscopy and flow cytometry. K562 cells were infected with ZIKV, WNV or DENV-1 (Hawaii) RVPs in the presence of serial dilutions of WNV positive serum obtained from a symptomatic subject. Cells were analyzed 48 h post infection either by flow cytometry or automated microscopy. As demonstrated in Fig. 2, there was a concentration dependent enhancement of ZIKV (Fig. 2a), WNV and DENV RVP infection in the presence of specific sera. As expected, no infection was evident in the absence of enhancing sera (media control, Fig. 2a). Interestingly, results of the ADE assay via flow cytometry were similar to automated microscopy both with regards to peak enhancement as well as serum dilution resulting in maximum enhancement for each RVP (Fig. 2b & c). Correlation analysis of measurement of GFP+ cells by flow cytometry versus automated microscopy was highly correlative with R² = 0.9439 (Fig. 2d). Thus, the automated microscopy based assay can be readily used in a high throughput format to study the phenomenon of ADE.

We further validated the sensitivity of the high throughput RVP based assay to differentiate between antibodies elicited by WNV, DENV and ZIKV infection. Sera specific for each flavivirus was used in either a neutralization or ADE assay (Fig. 3) against the three different RVPs (WNV, ZIKV, DENV-1). Specifically, WNV+ sera was from a symptomatic WNV exposed subject as confirmed by WNV E IgG ELISA, and ZIKV and DENV+ sera were from subjects positive for ZIKV or DENV antibodies confirmed by specific NS1 IgG ELISAs. As demonstrated in Figs. 3 a-c, sera against each virus neutralized the specific virus most efficiently. Similarly, as evident in Figs. 3 d-f, ADE was seen at lower serum concentrations for the specific virus, while the cross-reactive RVPs showed ADE at higher serum concentrations. The clear differences in the IC₅₀ for neutralization and the dilution of peak ADE provided a highly reliable assay to differentiate between these three closely related flaviviruses. The assay was highly reproducible as the error between wells was minimum and differences in peak ADE were higher than neutralization IC50s suggesting that an ADE assay might detect specific virus infection better than neutralization or when cross reactive neutralizing antibodies are present. Thus, in our hands, neutralization and ADE assays conducted simultaneously were capable of differentiating antibodies elicited by ZIKV, DENV and WNV infection.

Having obtained an adequate sample size of WNV seropositive samples, we investigated the effect of WNV sero-positivity on neutralization and enhancement of WNV, ZIKV and DENV infection. As demonstrated in Fig. 4a, serum samples from asymptomatic WNV seropositive individuals effectively neutralized WNV infection even at lower serum concentrations. Some neutralization of DENV infection was also apparent at higher serum concentrations although not as effective as WNV neutralization (Fig. 4c). Surprisingly, there was minimum neutralization of ZIKV infection, with a general trend of no neutralization at all to enhancement of infection in some cases (Fig. 4b). A similar trend was seen for WNV, DENV and ZIKV neutralization from serum samples obtained from symptomatic WNV seropositive subjects (Fig. S2).

We next determined ADE of WNV, DENV and ZIKV infection in the presence of WNV+ sera. In accordance with the neutralization data, enhancement of WNV infection in the presence of WNV seropositive sera was observed only at lower serum concentrations with inhibition of infection seen at higher serum concentrations (Fig. 4d). However, with regards to ZIKV, for most samples significant enhancement of infection was seen at higher serum concentrations (1:10 dilution) with minimum neutralization (Fig. 4e). An intermediate phenomenon was seen with DENV with some inhibition apparent at higher serum concentrations followed by enhancement of infection at lower concentrations (Fig. 4f). A similar trend was seen for WNV, DENV and ZIKV ADE from serum samples obtained from symptomatic WNV seropositive subjects (Fig. S2). These data demonstrate that presence of WNV antibodies in the sera cause effective WNV neutralization but leads to ZIKV enhancement in in vitro assays. Moreover, there is individual variation evident in the phenotype of the sera in terms of neutralization and ADE that may be overlooked/minimized after pooling the data.

We next analyzed pooled data from symptomatic and asymptomatic WNV seropositive subjects for neutralization and ADE of WNV, ZIKV and DENV. As demonstrated in Fig. 5a, the dilution of serum that caused peak ADE for WNV infection was significantly higher than dilution of serum that caused peak ADE for ZIKV infection. DENV ADE followed a similar trend as ZIKV ADE. When comparing neutralization, serum dilution that caused 50% neutralization of ZIKV was significantly lower than the serum dilution that led to 50% neutralization of WNV. Serum dilution for 50% DENV neutralization fell intermediate between ZIKV and WNV (Fig. 5b). A similar trend was seen when segregating the data based on symptomatic and asymptomatic WNV seropositive donors (Fig. S3) with no statistically significant difference between the groups. Along similar lines, we also compared pooled data for ADE and neutralization of infection at a serum dilution of 1:10, a relevant serum concentration in vivo. As demonstrated in Fig. 5c, there was a significant difference in ZIKV ADE when compared to WNV ADE at a 1:10 serum dilution. DENV ADE at 1:10 serum dilution was significantly higher than WNV but less than ZIKV ADE. When comparing neutralization at 1:10 serum dilution, both WNV and DENV infection was neutralized significantly better than ZIKV infection (Fig. 5d). Overall, these data revealed that pre-existing WNV antibodies while capable of neutralizing both WNV and DENV, poorly neutralize ZIKV infection. On the contrary, significant enhancement of ZIKV infection was evident even at higher serum concentrations.

Our data above showed that WNV antibodies in the serum effectively neutralize WNV while enhancing ZIKV infection. We hence sought to investigate if the reverse was true and whether ZIKV+ sera would neutralize ZIKV and enhance WNV. To study this effect, we utilized serum samples from ZIKV+ human subjects (BEI Resources, Table S6) and analyzed their ability to neutralize or enhance ZIKV, WNV and DENV infection. As demonstrated in Fig. 6a, ZIKV+ sera effectively neutralized ZIKV infection. Moreover, peak ZIKV ADE was seen at lower serum concentrations with higher serum concentrations effectively inhibiting infection in K562 cells as well (Fig. 6d & h). Interestingly, for WNV, neutralization was seen but at significantly higher serum concentrations than ZIKV (Fig. 6b & g). With regards to WNV ADE, there was marked enhancement at higher serum concentrations with minimum neutralization (Fig. 6e and h). This phenomenon was similar to that seen in Fig. 5 regarding ZIKV enhancement in the presence of WNV serum (Fig. 6h). With regards to DENV neutralization, effective inhibition was evident at mean serum dilution of ~ 1:100 (Fig. 6 c & g). However, DENV ADE in the presence of ZIKV sera was not as pronounced with some inhibition of infection seen at higher serum concentrations and pronounced ADE at lower concentrations (Fig. 6 f and h). Taken together, these findings suggest that just like WNV+ sera enhances ZIKV, the reverse holds true for ZIKV+ sera. Furthermore, these data validate the sensitivity our RVP based assay for studying the ADE phenomenon.

The phenomenon of ADE is a complex process that remains poorly understood. In our studies above with human serum samples, enhancement of virus infection was apparent even when there was no neutralization of infection. To obtain further insights into the phenomenon, we utilized ZIKV neutralizing and non-neutralizing monoclonal antibodies in our RVP based ADE and neutralization assay. Two well characterized neutralizing antibodies (4G2- a pan-flaviviral mouse anti-dengue E protein and ZV-54, an anti-Zika E protein) and one non-neutralizing antibody (ZV-13, anti-Zika E protein) were used to understand how they behave in terms of enhancing and neutralizing ZIKV infection. As demonstrated in Fig. 7a, and as expected, the neutralizing antibodies 4G2 and ZV-54 effectively neutralized ZIKV infection while the non-neutralizing antibody ZV-13 did not. Interestingly, with regards to ZIKV enhancement, the neutralizing antibodies 4G2 and ZV-54 caused ADE at lower antibody concentrations, with higher concentrations actually inhibiting infection (Fig. 7b). However, the non-neutralizing antibody led to ZIKV enhancement even at higher concentrations with no inhibition of infection. The trend for neutralization and enhancement of ZIKV was strikingly similar to the pooled data for neutralization (Fig. 7c) and enhancement seen with WNV seropositive serum samples (Fig. 7d). Interestingly, our data shows different patterns of ADE curves for neutralizing vs non-neutralizing antibodies. Neutralizing antibodies show a peak ADE followed by decline whereas non- neutralizing antibodies form a plateau with a peak ADE reaching saturation. It is plausible that neutralizing antibodies cause enhancement at suboptimal concentrations unlike non-neutralizing antibodies. Coincidentally, the shape of the curves are reminiscent of what is seen with sera samples obtained from human subjects (Fig. 7c and d).

Our studies with ZIKV monoclonal antibodies showed that non-neutralizing antibodies are potent inducers of ADE of ZIKV infection. Our results with the sera are also indicative of the presence of non- neutralizing antibodies against ZIKV in WNV+ sera. We hence asked whether we could detect anti-E antibodies against ZIKV in the WNV+ samples. Immunoprecipitation of cells lysates expressing the desired proteins provides a comprehensive overview of the repertoire of antibodies against different viral proteins in a sample. Studies have also implicated a role for antibodies other than the Envelope like prM in mediating ADE [24]. To study the cross-reactive antibody repertoire present in the WNV positive serum samples, we performed an immunoprecipitation (IP) analysis of ZIKV CprME expressing cell lysates with the indicated samples. As demonstrated in Fig. 8a and Fig. S4, the predominant antibodies present in the serum samples were against the E protein. We quantified the E protein bands in the IP gels and correlated it with neutralization and ADE of ZIKV. Interestingly, a higher intensity E protein band did not correlate with neutralization of ZIKV infection (Fig. 8b) but correlated significantly with ZIKV ADE (Fig. 8c) in vitro. These data confirm that the predominant antibodies responsible for enhancement of infection in case of ZIKV are directed towards the E protein and are most likely non-neutralizing antibodies.

The study was reviewed and approved by the Texas Tech University Health Sciences Center’s regional Institutional Review Board (IRB). All methods were performed in accordance with the relevant guidelines and regulations. The study design was cross-sectional and the study number recorded as IRB# E17056, approval date 03/16/2017. All participants were provided with written and oral information about the study. Written informed consent of all study participants in accordance with the Institutional policy was documented. All participants were identified by coded numbers to assure anonymity and all patient records kept confidential.

Three hundred and eighty-eight (N = 388) healthy human volunteers were recruited from a WNV endemic region of Texas. Twenty subjects (N = 20) with a previous history of symptomatic WNV infection were also recruited in collaboration with the Department of Public Health, City of El Paso, Texas. As WNV is a notifiable disease, symptomatic patients in the local hospitals reported to the public health department were confirmed for WNV infection by testing for presence of specific IgM antibodies in the serum or cerebrospinal fluid (CSF). Officials from the public health department were requested to contact these subjects to participate in the study. The average duration of time lapse from the date of WNV diagnosis to sample collection was ~ 2.2 years. To determine appropriate sample size for the study, statistical analysis was conducted to calculate required sample size for a sero-prevalence study with a 95% confidence interval. We anticipated the sero-prevalence rate for WNV antibodies in our population to vary from 3 to 10%. Considering those estimates, a sample size of > 217 would have adequate power. Each subject was requested to complete a questionnaire and provide a blood sample in a serum separator tube and the serum was aliquoted and stored at − 70 °C until further analysis. The questionnaire was administered to obtain demographic data and other information pertinent to the study.

Experiments were conducted in BSL2 containment following appropriate safely protocols and guidelines. 293 T and Vero cells were obtained from American Type Culture Collection (ATCC) and cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% Fetal Bovine Serum (FBS). K562 cells were obtained from ATCC and cultured in Roswell Park Memorial Institute (RPMI) medium supplemented with 10% FBS. Transfections were performed using Turbofect reagent (Thermo Fisher) as per the manufacturer’s instructions. The WNV plasmid containing the sub-genomic GFP expressing replicon derived from lineage II strain of WNV-Rep/GFP, WNV CprME and Zika CprME expression vectors have been described previously [49, 50]. The DENV-1(Hawaii) virus was obtained from the Biodefense and Emerging Infections Research Resources Repository (BEI resources). Viral RNA was used to amplify the CprME region using specific primers and RT-PCR followed by cloning into pCDNA3.1 TOPO vector (Invitrogen). The ZIKV immune sera from seropositive individuals was kindly provided by BEI resources. The ZIKV neutralizing and non-neutralizing antibodies 4G2 (neutralizing), ZV-13 (non-neutralizing) and ZV-54 (neutralizing), were obtained from Millipore Sigma (St. Louis, MO). The 4G2 antibody binds to the fusion loop at the extremity of domain II of E protein of different flaviviruses. The ZV-54 antibody binds to the DIII of the envelope protein corresponding to the lateral ridge and the ZV-13 binds to the conserved FL in DII.

The WNV E IgG ELISA, DENV NS1 Type 1–4 IgG ELISA and ZIKV NS1 IgG ELISA kits were obtained from EUROIMMUN (Luebeck, Germany) and used as per the manufacturer’s protocol. The anti-WNV ELISA (IgG) is intended for the detection of IgG antibodies to WNV in human serum and plasma samples. The antigen source is a recombinant detergent extracted glycoprotein E of WNV from the membrane fraction of human cells. The anti-WNV antibodies bind to the antigen coated in microtiter wells followed by addition of anti-human IgG HRP conjugate that results in a calorimetric reaction on addition of TMB substrate. The controls and calibrator included on the kit must be used in each run and results are evaluated by calculating a ratio of OD value of patient sample over OD value of the calibrator. A ratio of ≥1.1 is considered as positive for WNV IgG antibodies.

RVPs were prepared by transfecting 293 T cells with WNV Rep/GFP replicon construct along with WNV or ZIKV or DENV-1 CprME expression constructs. The RVPs were harvested 48 h post transfection, aliquoted and stored for future use. RVPs were titrated in Vero cells plated in 96 well clear bottom black plates at 5000 cells per well. Serial two-fold dilutions of the RVPs were used to infect Vero cells and incubated for 72 h. The plates were fixed using 4% formalin in PBS and images acquired using the Cytation5 imaging system (BioTek, Winooski, VT, USA). Images of whole wells were acquired and the number of GFP+ cells per well was quantified using Gen5 software (BioTek).

For neutralization assays, human sera or antibodies were serially diluted in appropriate media (DMEM or RPMI) and incubated with the RVPs for 1 h at 37 °C. For control wells, an equivalent volume of media was added. Sera samples were heat inactivated at 56 °C for 30 min prior to dilution. Subsequently, the virus:sera mix was added to Vero cells and number of GFP positive cells quantitated 72 h post infection as described above. Percent neutralization was calculated after normalizing data to cells infected in the absence of any sera (media control). Assays for ADE were conducted as described above except FcƴRII bearing K562 cells were used instead of Vero and plates read 48 h post infection. ADE data was normalized to the wells that showed maximum enhancement as detected by number of GFP+ cells.

The protocol for radiolabeling and immunoprecipitation of cell and virus lysates has been described previously [51]. Briefly, the stable cell line Zika-CprME-NS2B3-E2 [52] was washed with RPMI medium lacking Methionine and Cysteine. Thereafter, cells were incubated in RPMI medium supplemented with FBS and [³⁵S]Met/Cys. Cells were lysed with Triton X containing lysis buffer and cell lysates immunoprecipitated with Protein-A beads coated with WNV+ human serum samples. Immunoprecipitated lysates were washed, resolved by SDS-PAGE followed by PhosphorImager analysis.

Statistical analysis was conducted using GraphPad Prism (GraphPad Software, San Diego, CA). Comparison between multiple groups was done using the Kruskal-Wallis test. Spearman’s correlation with linear regression was used for all correlation determination using the GraphPad Prism Software. P values were considered significant at p < 0.05.