Chimeric flavivirus enables evaluation of antibodies against dengue virus envelope protein in vitro and in vivo

In a secondary dengue virus (DENV) infection, the presence of non-neutralizing antibodies (Abs), developed during a previous infection with a different DENV serotype, is thought to worsen clinical outcomes by enhancing viral production. This phenomenon is called antibody-dependent enhancement (ADE) of infection, and it has delayed the development of therapeutic Abs and vaccines against DENV, as they must be evaluated for the potential to induce ADE. Unfortunately, limited replication of DENV clinical isolates in vitro and in experimental animals hinders this evaluation process. We have, therefore, constructed a recombinant chimeric flavivirus (DV2ChimV), which carries premembrane (prM) and envelope (E) genes of type 2 DENV (DENV-2) R05-624 clinical (Thai) isolate in a backbone of Japanese encephalitis virus (Nakayama strain). DENV E-protein is the most important viral target, not only for neutralizing Abs, but also for infection-enhancing Abs. In contrast to DENV-2 R05-624, DV2ChimV replicated efficiently in cultured mammalian cells and was lethal in interferon-α/β–γ-receptor double-knockout mice. With DV2ChimV, we were able to perform neutralization assays, in vitro and in vivo ADE assays, and in vivo protection assays. These results suggest that the chimeric virus is a powerful tool for evaluation of Abs against DENV.

vascular permeability, and cytokine storms, but without neurologic disease, in interferon (IFN)-α/β-γ receptor (R) double-knockout (dKO) mice 12 . ADE was accessed in the mouse model infected with D2S10 13 . Thus, there are currently several mouse models available; however, only a limited number of DENV strains actually cause lethal infections with human-like symptoms 7,9,10 . Therefore, it is difficult to test the effect of Abs or antisera against various DENV serotypes, genotypes, and strains. Notably, most of the Abs that are produced in DENV-infected patients target the prM, E, and NS1 proteins 14 , and Abs against E have crucial (but complex and incompletely understood) roles in the control of virus replication via neutralization or ADE 15 . It may be better to have a simpler mouse model in which DENV E and Abs against E can be tested easily. So far, no study has developed alternative mouse models that focus on E and E-specific Abs.
Here, we describe the construction of a recombinant chimeric flavivirus (DV2ChimV), which carried premembrane (prM) and envelope (E) genes of low passage-number DENV-2 R05-624 clinical isolate in a Japanese encephalitis virus (JEV) backbone. Our results demonstrated the potential of DV2ChimV for evaluation of anti-DENV Abs in vitro and in vivo.

Replication of DV2ChimV in cultured cells.
Replication of DENV-2, JEV, and DV2ChimV was examined in cultured cells (Fig. 1a). Vero (African green monkey kidney) cells infected with DENV-2 R05-624 produced a low level of virus (8.0 × 10 2 FFU/mL) at Day 1 post-infection (p.i.) and reached a low peak titer (6.3 × 10 5 FFU/mL) at Day 5 p.i. (Fig. 1b). By contrast, Vero cells infected with DV2ChimV or JEV produced high levels of virus at Day 1 p.i. (4.8 × 10 5 FFU/mL and 6.8 × 10 5 FFU/mL, respectively), and reached peak titers of 8.6 × 10 7 FFU/mL at Day 3 p.i. (DV2ChimV) and 5.0 × 10 7 FFU/mL at Day 5 p.i. (JEV) (Fig. 1b). In mosquito C6/36 cells, JEV replicated efficiently and reached a peak titer of 5.4 × 10 6 FFU/mL at Day 5 p.i. (Fig. 1b). Both DV2ChimV and DENV-2 replicated more slowly than JEV, reaching peak titers of 2.0 × 10 4 FFU/mL at Day 5 p.i. and 1.5 × 10 3 FFU/mL at Day 3 p.i., respectively (Fig. 1c). Next, we examined the replication of each virus in murine B7 cells derived from wild-type BALB/c mice 16 . In these cells, JEV replicated most efficiently at the early stages, reaching a peak titer of 1.8 × 10 7 FFU/mL at Day 2 p.i. (Fig. 1d). DV2ChimV also replicated efficiently, reaching a peak titer of 1.1 × 10 6 FFU/mL. Although production of JEV and DV2ChimV decreased gradually, both maintained production up until at least Day 7 p.i. By contrast, DENV-2 R05-624 started replicating slowly, reaching a peak titer of 1.2 × 10 4 FFU/mL at Day 3 p.i.; however, virus production completely ceased at Day 5 p.i. (Fig. 1d). DENV-2 production was probably suppressed by the innate immunity in B7 cells because IFN-α and -β in infected B7 cells started to increase on Day 2 p.i and reached a maximum level at Day 5 p.i. ( Supplementary  Fig. S1). These results demonstrate that DV2ChimV production is similar to that of JEV Nakayama in cultured cells, particularly mouse cells. . Culture media were collected at the indicated time points post-infection. Viral titers were determined by focus-forming assays and expressed as the logarithm of focus-forming units (FFU) per milliliter. Results are expressed as mean + SD of triplicate experiments. Viral titers of each day were analyzed by one-way ANOVA after log transformation. Significance of the levels was assessed by the Dunnett's Multiple Comparison Test by using viral titer of R05-624 as a control. Statistical differences in viral titers were calculated relative to that of R05-624. *p < 0.01, **p < 0.001.
Neutralization and ADE assays with DV2ChimV in vitro. Neutralization assays with cultured cells are commonly used for evaluation of the ability of Abs to protect against pathogens. We examined the neutralizing activity of two Abs to DENV E-protein: human monoclonal Ab (HuMAb) D23-1G7C2, which was previously derived from DENV-infected patients 17 , and the mouse monoclonal Ab 4G2. In the neutralization assay, various concentrations of these Abs were incubated with DV2ChimV, which was then tested in a focus-forming assay in Vero cells. HuMAb D23-1G7C2 demonstrated strong neutralizing activity to DV2ChimV, with a 50% focus reduction neutralization test (FRNT 50 ) concentration of 0.029 µg/mL (Fig. 3a). The neutralizing activity of 4G2 was nearly tenfold lower (FRNT 50 = 0.278 µg/mL). We next examined whether Abs to DENV E-protein neutralized or enhanced infectivity in mouse macrophages bearing Fc receptors. The 4G2 Ab is known to cause ADE in cultured cells 18 , and D23-1G7C2 has also been shown to have ADE activity in cultured cells 17 . Peritoneal-exudate cells (PECs) were collected from IFN-α/βR-γR dKO mice and infected at a multiplicity-of-infection (MOI) of 0.1 with DV2ChimV, which had been incubated with serially diluted 4G2 or D23-1G7C2. Production of DV2ChimV from infected PECs after 3 days was maximally enhanced (772-fold by D23-1G7C2 and 292-fold by 4G2, compared with viral titers in the absence of Ab treatment) in the presence of 0.025 µg/mL D23-1G7C2 or 2.5 µg/mL 4G2, respectively (Fig. 3b). D23-1G7C20 at > 25 µg/mL reduced viral production, indicating strong neutralization, whereas 4G2 did not neutralize DV2ChimV, even at a concentration of 25 µg/mL, which is consistent with our observation of weaker neutralizing activity than D23-1G7C20 (Fig. 3a). These results demonstrate the potential for the use of DV2ChimV for examination of both neutralizing and enhancing activities of Abs to DENV E-protein.

Protection and ADE assessment with DV2ChimV in vivo.
To determine whether our model system was suitable for evaluation of the ability of Abs to protect against DENV infection, IFN-α/βR-γR dKO mice were  (Fig. 4a). With a 100 µg dose, 60% of the mice were protected at Day 40 p.i. (P = 0.0172) By contrast, 4G2 provided no protection (Fig. 4b).
To test its potential for therapeutic application, D23-1G7C2 was introduced at different time points after infection (Fig. 4c). An Ab against influenza virus (5E4) was used as a negative control. Infected IFN-α/βR-γR dKO mice were administered with 300 µg per animal of D23-1G7C2 at 4 h, 1 day, 2 days, and 3 days p.i. (Fig. 4c). D23-1G7C2 provided complete protection (in terms of survival) when it was introduced 1 day p.i., whereas only 60% of mice survived to day 40 p.i. when the Ab was introduced 4 h p.i. Late treatment with Ab 2 days or 3 days p.i. did not prevent mortality, even though it significantly prolonged survival compared with administration of 5E4 at 4 h p.i. (Fig. 4c).
We assessed a number of previously identified HuMAbs to DENV proteins for neutralizing and ADE activities. D23-1A10H7, D23-1B3B9, D23-1G7C2, and D23-3A10G12 are anti-E Abs, and D25-4D4F10 is an anti-prM Ab 19 . D32-2H8G1 has neutralizing activity to DENVs, but its target protein has not been identified 19 . These Abs showed 50% neutralizing activity against DENV-2 (at concentrations ranging from 1.2 to 5.1 μg/mL) 17 . For each Ab, 300 µg was introduced into DV2ChimV-infected IFN-α/βR-γR dKO mice at 4 h post-infection. D32-2H8G1, D23-1B3B9, and D23-1G7C2 provided significant protection, with ≥ 80% survival 40 days p.i., compared with 100% mortality by day 9 p.i. with 5E4 (all P < 0.01) (Fig. 4d). Notably, D32-2H8G1 protected 100% of mice, which suggests that it targets DENV prM or E. However, we were unable to identify the target protein of D32-2H8G1 because the Ab did not react with recombinant prME on western blots 17 , suggesting that it might recognize a structurally-specific epitope of the DENV viral particle. The D23-1A10H7 anti-E Ab also Ab-virus mixtures were then assessed in focus-forming assays to determine the 50% focus reduction neutralization test (FRNT 50 ) values of the Abs, which were 0.278 µg/mL for 4G2 and 0.029 µg/mL for D23-1G7C2. FRNT 50 was calculated from nonlinear log-dose-response curves using GraphPad Prism software. Results are expressed as mean + SD of triplicate experiments. (b) To assess Ab-mediated enhancement of infection, 4G2 was serially diluted by tenfold dilution from 2.5 × 10 1 µg/mL to 2.5 × 10 −5 µg/mL, and incubated with 4 × 10 4 FFU of DV2ChimV. Peritonealexudate cells were infected with virus-Ab mixtures, at a multiplicity-of-infection of 0.1, and incubated for 3 days. The level of release of viruses into the culture supernatant was determined by titration in Vero cells. Results are expressed as mean + SD of triplicate experiments. Viral titers of each group were analyzed by oneway ANOVA after log transformation. Significance of the levels was calculated relative to Ab untreated by the Dunnett's Multiple Comparison Test. *p < 0.01, **p < 0.001.

Scientific Reports
| (2020) 10:21561 | https://doi.org/10.1038/s41598-020-78639-x www.nature.com/scientificreports/ www.nature.com/scientificreports/ provided significant protection, with 60% survival at Day 40 p.i. By contrast, D23-3A10G12 showed only 40% survival at Day 40 p.i. The anti-prM Ab D25-4D4F10 provided only limited protection relative to 5E4 (P < 0.05), which is consistent with previous findings 18 . Anti-prM Abs do not usually show strong neutralizing or protective activity. Our observations suggest that the mouse model system is adequate to evaluate the protective abilities of anti-DENV prM and E Abs. Next, we used the mouse model to assess ADE. The mouse monoclonal anti-E Ab 4G2 was serially diluted and injected into IFN-α/βR-γR dKO mice 1 day prior to infection with 8.0 × 10 2 FFU per animal of DV2ChimV. Notably, whereas the mice that received PBS rather than 4G2 all died by day 11 p.i., mice inoculated with 8 µg 4G2 died significantly earlier, by day 6 p.i. (P < 0.05), suggesting ADE (Fig. 5a). There were no significant differences in the survival of mice injected with other doses of 4G2 compared with the PBS-treated mice. In addition, IFN-α/βR-γR dKO mice were inoculated with 8 µg 4G2 (8.0 × 10 2 FFU per animal) at 24 h post-infection with DV2ChimV. These mice died significantly earlier (by Day 7 p.i.) than mice inoculated with control IgG (Supplementary Fig. S2). In similar experiments involving various doses of D23-1G7C2, no induction of ADE was observed (data not shown), presumably because D23-1G7C2 has greater neutralizing activity than 4G2 (Fig. 3a). To further investigate ADE in this system, we assessed viral production in the organs of mice inoculated with 8 µg 4G2 or with PBS, 5 days p.i. Notably, with the exception of PEC and liver, viral production was not significantly higher in the organs of 4G2-treated animals than in those of phosphate-buffered saline (PBS)-treated animals, although there was a trend for increased viral production in serum, PEC, thymus, and lung (Fig. 5b).
Elevated cytokine levels have been observed in DENV-infected patients [20][21][22][23] . On Day 5 p.i., TNF-α levels in sera from infected mice treated with 8 μg 4G2 were 6.4-fold higher than those in sera from control infected mice treated with IgG (393.8 pg/mL versus 61.4 pg/mL, respectively), whereas IL-6 levels were 6.9-fold higher (1064.5 pg/mL versus 153.9 pg/mL, respectively) (Fig. 5c). By contrast, the serum level of MCP-1 was 2.4-fold higher in control mice than in mice under ADE conditions (593.4 pg/mL versus 1,425.0 pg/mL, respectively). Induction of IFN-γ was similar in both groups (Fig. 5c). We did not observe detectable levels of IL-12p70 or IL-10 (data not shown). Cytokine levels in mock-infected mice treated with 8 μg 4G2 were measured to establish basal levels. These results suggest that induction of pro-inflammatory cytokines may be important for lethality in this mouse model of ADE.

Discussion
In this study, we demonstrated the potential of a model system for evaluation of Abs to DENV E-protein. In vivo testing of the efficacy of Abs is carried out in mouse models, but only a limited number of DENV strains cause death in these models at an early time point (< 12 days) with human-like symptoms, such as increased vascular permeability 7,9,10 . Although high doses of some other DENV strains cause lethal infections in IFN-α/βR-γR dKO mice, most of these deaths are caused by the spread of virus into the brain at a late stage after clearance of virus from other organs 10,24 , which does not resemble the pathology of severe dengue fever in humans. Notably, our chimeric virus, DV2ChimV caused death at an early time point, and induced vascular permeability, especially in the liver and intestine at the moribund stage (data not shown).
A benefit of our chimeric virus model is that the E gene in DV2ChimV can be readily replaced with any DENV E sequence of interest, such as that of a currently prevalent strain, to test relevant Abs or antisera. DENVs include several genotypes and strains, which can vary in the sequence of the E gene [25][26][27] , thereby possibly affecting Ab interactions 25,28 . In addition, there were differences between the characteristics of the highly passaged laboratory DENV strain and those of the low-passaged DENV strain. For example, autologous patient-derived DENVs, but not highly passaged laboratory virus strains, showed a low level of ADE 25,29 . Furthermore, it has been previously reported that different types of cells show different susceptibilities to infection by low-passage DENVs 29 . These observations suggest that there is difference in envelope protein between low-passaged clinical isolates and highly-passaged laboratory strains, although it is still unclear how much this difference between strains or passage number will affect the development of a vaccine and therapeutics, as well as the study of DENV pathogenesis. Another advantage of DV2ChimV is that the JEV-derived backbone, containing C and NS1-5 genes, resulted in a higher level of replication than was seen with DENV-2 in mouse cells, as well as a high level of virulence in mice. Although several mouse models of DENV infection have been reported, high doses of viruses are required 7,9,10 . In the case of human infection, only a small amount of virus is thought to be required for infection 30,31 . In our model, only a few hundred DV2ChimV virions were required for lethal infection (Fig. 2), suggesting that the replication of this virion is faster than that of DENVs in other mouse models. This property implies a requirement of more strict condition for testing effective therapeutics. Neutralizing antibody must efficiently inactivate viral particles, because otherwise the virus can escape to produce large amounts of virus progeny. The efficient replication of DV2ChimV in our model may mirror the high infectivity of DENV in human. Evaluation of Abs in vivo can provide more information than in vitro assessment. For example, five HuMAbs (D23-1G7C2, D23-1A10H7, D23-1B3B9, D23-3A10G12, and D32-2H8G1), which were previously found to have similar neutralizing activity in Vero cells (FRNT 50 = 1.2 to 5.1 μg/mL to DENV-2) 17 , had considerably different protective effects in our in vivo study (Fig. 4d). In vivo assays could help differentiate between Abs on the basis of characteristics such as stability, effective recycling, and Ab-dependent cell-mediated cytotoxicity. Another potential advantage of our in vivo system was that it enabled observation of the mice over a long period, which made the differences among the Abs more apparent.
By making use of DV2ChimV, we have obtained important information about the target viral protein of D32-2H8G1, which protected 100% of five mice over a 40-day observation period (Fig. 4d). The target protein of D32-2H8G1 had not been previously identified because the Ab did not react with recombinant DENV E-protein 17  www.nature.com/scientificreports/ antibody (5J7) that binds across three surface DENV E proteins 32 . D32-2H8G1 may similarly recognize conformationally specific E-protein epitopes located on the surface of the viral particle, which may explain why D32-2H8G1 does not react with recombinant DENV E-protein.
Our results revealed important insights into the role of timing in the effectiveness of Ab treatment. Overall, treatment with D23-1G7C2 was more effective the earlier it was administered post-infection (Fig. 4c). However, there was a notable exception, in that treatment 4 h p.i. resulted in 60% survival, whereas treatment 1 day p.i. resulted in 100% survival. This result suggests that treatment with Ab at a very early time point might attenuate the therapeutic effect. One possible mechanism for attenuation would be the delay or failure of induction of the host immune response. Further study is needed to determine whether immediate treatment after infection results in a delay of the host immune response.
The ADE phenomenon provides a plausible explanation for the occurrence of severe disease in secondary and serial infections 5 . In vitro results demonstrated that appropriate concentrations of anti-E Abs can result in elevation of viral production in cells bearing Fcγ receptors 25,33,34 . In our study, enhancement of viral production by 4G2 ADE was limited in the in vivo mouse model (Fig. 5b) compared with the in vitro murine PEC system (Fig. 3b). Nevertheless, the in vivo effect of ADE on disease outcome was evident in the early deaths of affected mice (Fig. 5a), and in the induction of production of TNF-α and IL-6 (Fig. 5c). It is not yet possible to say whether ADE affects disease outcomes through small changes in viral production that in turn have greater effects on induction of pro-inflammatory cytokines. Other mechanisms might be involved in pathogenesis, such as stimulation of immune responses and pro-inflammatory cytokine production by Abs or Ab-virus complexes. It may be better to consider ADE of infection and ADE of disease separately 35 , although high viral titer is certainly necessary for severe disease 36 .
Recently, Raut et al., reported that DENV virions produced in human are more infectious and mature than those produced in cultured cells 37 . They suggested that there is a structural difference between human plasma and cell-culture derived virions, that is, virions produced in humans contain less undigested prM protein. We measured the number of infectious virions of DV2ChimV in a focus assay and the number of viral genomes by quantitative RT-PCR. DV2ChimV harvested from C6/36 cells contained more infectious viral particles than R05-624, and DV2ChimV harvested from Vero cells contained similar numbers of infectious viral particles to those of DENV-2 R05-624, although JEV contained higher numbers of infectious viral particles ( Supplementary Fig. S3). The average number of infectious DV2ChimV virons produced in PEC derived from IFN-α/βR-γR dKO mice was higher than that of DENV-2 virions, although the difference was not significant (Supplementary Fig. S3). Besides, there was no significant difference between the number of infectious DV2ChimV virions and that of JEV virions in sera derived from IFN-α/βR-γR dKO mice. We also examined the levels of E and prM proteins by western-blotting analysis with anti-E (D23-1G7C2) and anti-prM (D25-4D4F10) Abs. DV2ChimV, DENV-2, and JEV were produced from C6/36 cells, Vero cells, and PEC derived from IFN-α/βR-γR dKO mice, and collected. An equivalent number of FFU (5 × 10 5 ffu) was precipitated by methanol/chloroform, and subjected to westernblotting analysis without reducing agent. An anti-E (D23-1G7C2) does not react to E protein in the presence of reducing agents. In both C6/36 cells and Vero cells, DV2ChimV contained similar levels of uncleaved prM protein to those of original DENV-2 R05-624 ( Supplementary Fig. S4). Two bands of DV2ChimV and DENV-2 E proteins may be due to the result of different levels of glycosylation or partial cleavage. Besides, dimerized E proteins were found because E protein forms dimer through disulfide bonds. Interestingly, DV2ChimV produced in PEC derived from IFN-α/βR-γR dKO mice contained lower levels of uncleaved prM ( Supplementary Fig. S4), although we failed to detect E or prM in the serum of this model mouse. PEC is a major target cell in this model (data not shown). On the other hand, JEV viral particles appeared to contain less uncleaved prM; however, this may be due to the low reactivity of anti-prM Ab. In terms of maturation, DV2ChimV may produce mature virions at a higher rate than DENV-2 in IFN-α/βR-γR dKO mice. However, it is still unclear whether virion particle heterogeneity influences DENV serotypes, genotypes, or strains 38 . So far, there has been no report about the degree of maturity of DENV in mice. Further study is needed to understand the effect of viral maturation.
There is a question about whether this model using a chimeric dengue virus is applicable to the study of the pathogenesis of severe dengue. DENV NS1 protein is thought to play an important role [39][40][41] ; however, the detailed mechanism is not fully understood. The NS1 gene of DV2ChimV is derived from the JEV genome. However, DV2ChimV infection caused vascular leakage and thrombocytopenia in IFN-α/βR-γR dKO mice (data not shown). Further comparisons of DV2ChimV and DENV would provide more information about pathogenesis.
Our results demonstrated the potential of a chimeric flavivirus for the evaluation of therapeutic Abs to DENV. Although humans and mice differ physiologically, we believe that this system will be useful for early evaluation of potentially therapeutic Abs. It should also be amenable to the assessment of antisera resulting from the use of vaccine candidates. Although we believe that the chimeric flavivirus and other recombinant viruses will be useful for the development of therapeutics, we recommend that experiments using chimeric viruses be conducted under the proper degree of biological containment to avoid accidental release.
Virus production. To  To quantify vRNA in the spleen, liver, kidney, thymus, lung, brain, PECs, and bone marrow, total RNA was extracted from these organs with TRIzol RNA Isolation Reagents (Life Technologies), with final resuspension into 30 µL RNase-free water. The concentration of each extracted RNA was adjusted to 50 µg/mL. The following primers, modified from a previous report 43 , were used for PCR: JEF (5′-AGA GCG GGG AAA AAG GTC AT-3′) and JER#110 (5′-CTT CAC GCT CTT CCT ACA GT-3′). One-step, real-time quantitative RT-PCR amplification with SYBR Green I was performed with the CFX Connect Real-Time System (Bio-Rad, Hercules, CA, USA) and the One-Step SYBR PrimeScript RT-PCR Kit II (Takara). The final concentration of each PCR primer was 0.08 µM, and the concentration of total RNA was 8 µg/mL, with 12.5 µL reaction volumes. The conditions for reverse transcription were 42 °C for 5 min and 95 °C for 10 min. PCR amplification used 45 cycles of 95 °C for 5 s, 55 °C for 30 s, and 72 °C for 30 s. The quantity of vRNA in the initial total RNA was determined by interpolation analysis from a standard curve generated from tenfold serial dilutions of in vitro-transcribed DV2ChimV RNA made with the MEGAscript Kit (Ambion). The limit of detection was ≥ 10 copies. Data were analyzed with CFX Manager ver. 1.6 (Bio-Rad). To quantify vRNA derived from organs, the amounts were normalized to the total RNA from corresponding organs of mock-infected mice.
Data analysis. All data were analyzed with Graphpad Prism software (Graphpad, San Diego, CA, USA).
Received: 21 July 2020; Accepted: 25 November 2020 Scientific Reports | (2020) 10:21561 | https://doi.org/10.1038/s41598-020-78639-x www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.