Protective envelope dimer epitope-like antibodies are elicited against dengue virus in children after infection and vaccination

Abstract

Cross-reactive antibodies to epitopes that span envelope proteins on the virion surface are hypothesized to protect against dengue virus (DENV) infection and disease. Here, we measured antibodies targeting a quaternary epitope called the envelope dimer epitope (EDE) as well as neutralizing and binding antibodies and evaluated their association with DENV infection, vaccine response, and disease outcome in dengue-vaccinated (n = 164) and dengue-unvaccinated children (n = 88) within a longitudinal cohort in Cebu, Philippines (n = 2996). Antibodies targeting EDE were prevalent and associated with broad neutralization of mature DENV1 to DENV4 virions in those with evidence of at least two prior DENV infections but were mostly absent in those with only one prior infection. Subsequent infection and vaccination boosted titers of EDE-like antibodies, neutralizing antibodies, and DENV-binding antibodies. EDE-like antibodies were associated with reduced risk of symptomatic dengue and more severe dengue and statistically explained the protective effect of binding and neutralizing antibodies on dengue. Thus, antibodies targeting quaternary epitopes help explain the broad cross-protection observed in those with multiple prior DENV exposures, making them useful for evaluation and development of future vaccines and therapeutics.

Figure 1.. EDE-like Abs are part of the serum repertoire of secondary immune children in the Philippines cohort at baseline.

(A) Heatmap showing log₁₀ serum neutralizing Ab IC₅₀ titers to each mature DENV1–4 and the GMT for each participant. Pie chart shows the percent of participants by number of mature serotypes neutralized, with the total number of study participants shown in the center of the circle. Titers ≥1:20 were defined as positive. (B) Bar plot showing the percent of secondary immune children with EDE-like Ab titers (IC₅₀ titer ≥10) to the DENV2 fusion loop (FL) mutant, DENV2, and DENV3 stabilized E dimers. (C) Heatmaps of Log₁₀ EDE-like Ab IC₅₀ titers and Log₁₀ binding IgG EC₅₀ Ab titers to each stabilized E dimer. Binding IgG Abs to WT DENV1–4 E proteins and the GMT are reported as optical density (OD, 405nm) values at a 1:100 serum dilution. Individuals are ordered from lowest to highest EDE-like Ab titers to the DENV2 fusion loop mutant. DV, Dengue virus; FL, fusion loop; Mut, mutant; E, envelope; GMT, geometric mean titer; WT, wild-type.

Figure 2.. EDE-like Ab potency is associated with the magnitude and the breadth of mature virus neutralizing and binding Abs at baseline.

(A) Pearson’s correlation coefficients (r, indicated by color in the top right triangle and number in the bottom left) among Ab measures for secondary immune children (n=252) at baseline, including: EDE-like and binding Ab titers to the DENV2 fusion loop mutant, DENV2, and DENV3 stabilized E dimers; binding OD values to WT DENV1–4 E proteins and the GMT; and neutralizing Ab titers to mature DENV1–4 and the GMT. *p <0.05, **p <0.01, ***p <0.001. (B) Scatterplots showing the association of EDE-like Abs for primary (n=50) versus secondary immune individuals (n=252), who are further stratified by the number (#) of mature serotypes neutralized. (C) EDE-like Abs for primary (n=50) and secondary immune (n=56) individuals who only neutralized one mature serotype, stratified by the inferred primary infecting serotype, as defined in the legend. Bar plots show means with standard errors of the mean. Significance in (B) and (C) was determined by analysis of variance (ANOVA, p-values are shown above the top bracket in each panel) with Tukey’s correction for multiple comparisons (p-values are shown above the bracket for each comparison). For (B), only comparisons for stepwise increases in immune breadth are shown.

Figure 3.. Infection and vaccination boost Abs in secondary DENV-immune children.

(A) Scatterplots comparing Ab measures for unvaccinated (n=88) and vaccinated (n=164) individuals at baseline and follow up. Ab measures include the GMT of neutralizing Abs to DENV1–4 (log IC₅₀); EDE-like Ab titers to the DENV2 fusion loop mutant, DENV2, and DENV3 stabilized E dimers (log IC₅₀) and percent of EDE-like Abs attributable to the fusion loop (%); the GMT of binding OD values to WT DENV1–4 (OD 405nm); and binding Abs to the DENV2 fusion loop mutant, DENV2, and DENV3 stabilized E dimers (log EC₅₀) and percent of binding Abs attributable to the fusion loop (%). (B and C) Pie charts show the proportions of vaccinated and unvaccinated individuals who experienced a boost in mature virus neutralizing Abs (defined as a seroconversion or ≥4-fold rise in neutralization to ≥2 mature serotypes; B) or had evidence of vaccine replication (positivity to YFV NS1 at follow up; C). Numbers in the center of each chart indicate the N for each group. (D) Scatterplots comparing Ab measures for unvaccinated individuals and vaccinated individuals without (YFV NS1−) or with (YFV NS1+) evidence of vaccine replication. A p-value <0.05 was considered significant and calculated using ANOVA followed by Tukey’s correction for multiple comparisons. Bar plots show means with standard errors of the mean. Vax, vaccinated; Unvax, unvaccinated; YFV, Yellow fever virus; NS1, nonstructural protein 1.

Figure 4.. Baseline mature virus neutralizing Abs were modestly associated with reduced DENV infection and replication of the YFV-based vaccine.

Bar plots are shown comparing baseline Ab measures between unvaccinated (n=88) and vaccinated (n=164) individuals who did or did not exhibit a boost between baseline and follow up. Baseline Ab measures include: the GMT of neutralizing Abs to DENV1–4 (log IC₅₀); EDE-like Ab titers to the DENV2 fusion loop mutant, DENV2, and DENV3 stabilized E dimers (log IC₅₀) and percent of EDE-like Abs attributable to the fusion loop (%); the GMT of binding OD values to WT DENV1–4 E protein (OD 405nm); and binding Abs to the DENV2 fusion loop mutant, DENV2, and DENV3 stabilized E dimers (log EC₅₀) and percent of these binding Abs attributable to the fusion loop (%). The boost was defined as a seroconversion or ≥4-fold rise in mature virus neutralization to ≥2 serotypes. Bar plots are presented as means and standard error of mean. T-tests were used for the comparison between groups. A p-value <0.05 is defined as significant.

Figure 5.. Mature virus neutralizing, EDE-like, and binding Abs at follow up are associated with protection against subsequent dengue.

(A) The bar plots show Abs measured at follow up across all assays for those with (n=43) and without (n=209) subsequent dengue. Follow up Ab measures include: the GMT of neutralizing Abs to DENV1–4 (log IC₅₀); EDE-like Ab titers to the DENV2 fusion loop mutant, DENV2, and DENV3 stabilized E dimers (log IC₅₀) and percent of EDE-like Abs attributable to the fusion loop (%); the GMT of binding OD values to WT DENV1–4 E protein (OD 405nm); and binding Abs to the DENV2 fusion loop mutant, DENV2, and DENV3 stabilized E dimers (log EC₅₀) and percent of these binding Abs attributable to the fusion loop (%). Data are presented as means and standard errors of the mean. T-tests were used for the comparisons between groups. (B) Heat map showing the relationship between each Ab measure at the follow up time point and the odds of subsequent dengue disease using unadjusted logistic regression. All Ab measures were standardized as z-scores before analysis to enable comparison across assays. Columns correspond to Ab measures and rows indicate the dengue outcome: DEN (dengue caused by any serotype), DWWS (Dengue with warning signs), HOSP. (Hospitalized dengue), and dengue caused by each serotype separately: DENV1, DENV2, DENV3, and DENV4. The scaled color represents the odds ratio for each predictor variable. Asterisks represent the statistical significance: *p <0.05, **p <0.01, ***p <0.001, ****p <0.0001. P-values <0.05 were considered significant. (C) Least Absolute Shrinkage and Selection Operator (LASSO) analysis plot showing the selection of predictor variables within the logistic model. The left y-axis represents the regression coefficients, the bottom x-axis represents Log Lambda (regularization parameter that determines how much penalty is imposed on the model), and the top y-axis represents the degrees of freedom, which described how many non-zero variables are present in the model at a specific Log Lambda. Variables with non-zero coefficients at higher log lambda values are retained in the model as predictors, with those kept in the model at the highest log lambda (e.g., −4 and above) demonstrating the strongest association with dengue outcome. Each colored curve corresponds to the predictor of the same color.

Figure 6.. EDE-like Abs mediate the association between the GMT of mature virus neutralizing and WT E binding Abs on dengue disease.

(A) The path diagram shows the effect of EDE-like Abs to each E dimer (tested as mediators) on the relationship between the GMT of mature virus neutralizing Abs (independent variable) and dengue disease (outcome) in logistic and linear models adjusted for age, sex, study site, and vaccination status. For the model evaluating the association between the independent variable and mediator, the regression coefficients and p-values are shown. For the combined regression model including the independent variable, mediator, and outcome, the regression coefficients and p-values for the independent and mediator variables are shown above their respective arrows. Thick arrows show significant pathways and thin arrows shown non-significant pathways. (B) Same as (A) but testing the effect of EDE-like Abs on the relationship between the GMT of WT E binding Abs (independent variable) and dengue disease (outcome) in adjusted models. (C and D) Shown are models testing the mediation effect of EDE-like Abs to each E dimer on the association between DENV2 mature virus neutralizing Abs and DENV2 disease (C) and DENV3 mature virus neutralizing Abs and DENV3 disease (D). In all models, p-value <0.05 was considered significant.