DENV infection elicits primarily a poor quality immune response directing a high proportion of antibody against non-protective, potentially pathogenic epitopes and only a small proportion against potently neutralizing and protective epitopes. In this report we have shown the manipulation of these potentially pathogenic epitopes as a vaccine strategy  that can reduce ADE in vitro and in vivo. Immunization of mice demonstrated that knocking out immunodominant cross-reactive epitopes in the EDIIFP and EDIII did not significantly effect DENV-2 neutralization, however the removal of these epitopes dramatically altered the vaccine induced antibody repertoire and sera from vaccinated mice shows reduced ADE in vitro and reduced lethal enhancement of DENV in vivo. Such a strategy could be applicable to other DENV vaccine formats, however, it might not be applicable to DENV live-attenuated vaccines because mutations in the EDIIFP can be lethal .
Our findings demonstrate that by introducing targeted amino acid substitutions into immunodominant cross-reactive E protein epitopes of a DENV-2 DNA vaccine that we can significantly reduce the induction of antibodies associated with immune enhancement, that are stimulated from these epitopes . Epitope-specific IgG ELISA revealed that cross-reactivity reduced vaccinated mouse sera contained dramatically reduced proportions of cross-reactive EDIIFP specific antibodies, had reduced potential for ADE in vitro, and when passively transferred into AG129 mice exhibited a significant reduction in hemorrhagic vascular leak associated mortality compared to passively transferred WT vaccinated mouse sera. WT vaccinated mouse sera however, contained EDIIFP specific antibodies in proportions similar that of naturally infected human sera. The capability of EDIIFP recognizing antibodies to induce severe DENV disease similar to DHF via ADE has been demonstrated in the AG129 mouse model [23–25]. Cross-reactivity reduced vaccines with alterations in EDIIFP (RD and RDERR) elicited almost no antibodies which recognize the WT EDIIFP epitope (Table 2), and therefore are less likely to participate in ADE in vivo. Supporting this conclusion, The RD and RDERR vaccine candidates exhibited the least ADE potential in vitro with RDERR vaccinated mouse sera being the only cross-reactivity reduced candidate that completely lacked in vitro ADE across all dilutions and serotypes of DENV.
One objective of this study was to dampen the immunodominance of the EDIIFP through alteration of EDIIFP amino acids. Although studies with immunotoxin and other therapeutic proteins have shown amino acid substitutions can dampen the immunogenicity of B cell epitopes , RD and RDERR vaccines did elicit antibodies which recognize the mutated EDIIFP epitope, in similar proportions as those recognizing the WT EDIIFP from WT vaccinated mice (67% and 73% respectively). These results suggest that although the specificity of this antigenic region was completely altered, the immunodominance was not reduced by the RD substitutions. These data also suggest that instead of dampening the recognition of the modified RD EDIIFP epitope by B cell receptors, the substitution of G106R might allow for anchoring of the B cell receptor, as suggested by studies identifying arginine as a main anchor residue in protein-protein and antibody-antigen interactions [44, 45] Though RD-induced novel antibodies do not recognize WT DENV (Table 2) and therefore would not participate in ADE in a natural infection, the production of novel antibodies is potentially concerning and ongoing studies include mapping of amino acid residues responsible for immunodominance and subsequent alterations in an effort to ablate the antigenicity of this epitope.
Severe DHF is associated with secondary DENV infections in older children and adults, or primary infections in infants. To evaluate and compare the different cross-reactivity reduced vaccine constructs for altered antibody profiles, we measured Fc receptor-dependent ADE of WT and cross-reactivity reduced vaccinated mouse serum in vitro. Sera from mice vaccinated with WT DENV-2 DNA vaccine maximally enhanced DENV-2 infection of K562 cells at a dilution four times higher than the average DENV-2 neutralization titer. These results show that as the antibody concentration decreases the proportion of neutralizing antibodies becomes insufficient to neutralize virus, allowing for ADE . This phenomena is similar to primary DHF in infants where maternal antibody titers correlated to infant age at onset of severe disease  when maternally derived anti-DENV IgG maintained reactivity with DENV virions but could not neutralize virus . None of the cross-reactivity reduced vaccine constructs enhanced homologous DENV-2 replication at any dilution tested.
WT vaccinated mouse serum significantly enhanced both DENV-1 and DENV-3 replication in human K562 cells at 1:2 serum dilution (the lowest dilutions tested due to the availability of testing serum), which is the closest simulation to undiluted sera and the most relevant for ADE from natural exposure. Sera from RD and ERR vaccinated mice were also able to significantly enhance DENV-1 and DENV-3 at some serum dilutions. Conversely, sera from RDERR vaccinated mice did not enhance any DENV serotype at any dilution tested. This finding suggests the possible involvement of the adjacent antigenic regions of EDIIFP and EDIII in the development of enhancing antibodies as concurrent modifications to both regions were necessary to eliminate ADE. Our MAb mapping of cross-reactivity reduced vaccine constructs  points to a possible antibody class to explain this phenomenon. MAb D3-5 C9-1 is a DENV-4 derived, complex cross-reactive, non-neutralizing antibody. D3-5 C9-1 exhibited only minor reductions in reactivity for the RD plasmid derived VLPs, no reductions for ERR VLPs, but greater than 98% reduction in reactivity compared to WT for the RDERR plasmid derived VLPs. This implies that D3-5 C9-1 recognizes an epitope overlapping EDIIFP and EDIII of adjacent monomers within the E dimer. The DENV complex cross-reactivity and lack of any neutralizing capability of this MAb suggest that antibodies recognizing epitopes similar to D3-5 C9-1 could be an important DENV disease enhancing antibody class.
In an in vivo DENV disease AG129 mouse model, we demonstrated that DENV-2 RDERR vaccination reduced mortality from homologous ADE compared to WT vaccine. While 100% of mice receiving WT immune sera succumbed to lethally enhanced disease, only 50% of mice receiving RDERR immune sera developed terminal disease. The apparent decrease in ADE by RDERR immune sera raises the question whether reduced ADE is due to shifting  or blunting of the enhancement curve? Due to the limited availability of sera, we only tested homologous enhancement at two passively transferred serum volumes. Mice receiving RDERR immune sera of either volume had higher survival compared to mice receiving WT immune sera. Mice receiving 100 μl RDERR immune sera had post transfer FRμNT50 titer = 1:10, the same post-transfer titer as mice receiving 50 μl of WT vaccinated serum (Figure 3). Thus, RDERR post serum transfer neutralizing antibody titer fell within that of the maximal enhancing range of WT immune sera and yet maintained significantly higher survival, suggesting RDERR immune sera blunts the enhancement curve. In addition, mice immunized with either WT or RDERR plasmids elicited similar levels of total IgG recognizing DENV-2 E (Table 2), suggesting the distinct differences in the quality of the immune response and reduced potentially pathogenic antibody in RDERR was responsible for differences in ADE.
In a homologous enhancement scenario, we would not expect cross-reactivity reduced immunized mice to exhibit completely reduced ADE because many unaltered antibody classes from RDERR immune sera can still react with the homologous E protein epitopes of the challenge virus. Similar phenomena have been demonstrated with passively transferred, protective and potently neutralizing antibodies [24, 46]. Moreover, the role of flavivirus non-neutralizing type and strain-specific antibodies  in ADE is not well characterized.
Similarly, the residual enhancement observed with passively transferred RDERR vaccinated mouse sera could be due to prM/M targeting antibodies [37, 39]. Recent studies have begun to expand upon the prescient historical deductions of Henchal et al., regarding the potential importance of prM antibodies in the human polyclonal immune response to DENV infection and their ability to enhance the infection of immature virus particles potentially exacerbating secondary DENV disease [18, 26, 37, 39, 49]. DNA vaccines direct the expression of prM/M and E proteins which self-assemble and are secreted as immunogenic VLP. A potential limitation of this study is that we did not characterize the particulate nature of the cross-reactivity reduced DENV-2 vaccine antigens and although they do form a pelletable antigen in vitro, this might not be identical to the VLPs which have been characterized with our WT plasmids . Regardless the physical nature of the antigens secreted by mutant plasmids is identical or not, the critical outcome of immune response is the most important parameter being measured. Previous studies by our group have shown the expression of prM [50, 51] by our DNA plasmids and vaccination with VLPs elicits anti-prM as wells as anti-E antibodies . Although the current report did not evaluate the presence of prM targeting antibodies in the antibody response toward our cross-reactivity reduced vaccines or their ability to enhance DENV infection, the manipulation of the EDIIFP and EDIII epitopes were effective in significantly reducing the ADE capability of serum from vaccinated mice in vitro and in vivo. As the role of prM antibody associated ADE becomes clearer, the same approach used here to ablate immunodominant enhancing E protein epitopes can be applied to prM. Studies to identify and characterize prM epitopes and to examine their inclusion into our DENV cross-reactivity reduced vaccine constructs are currently ongoing.