We leveraged a large prospective, observational cohort to assess various antibody measures as potential correlates of risk for dengue. Total binding antibodies and both GMT and serotype-specific nAb measured with standard and mature viruses were associated with a decreased dengue risk. Standard and mature serotype-specific nAb titers had the strongest protective effects against the matched serotype, supporting the dogma that exposure to one DENV serotype induces excellent immunity to that type32. However, when compared to the standard serotype-specific titers, the mature titers had bigger effect sizes against matched serotypes and decreased cross-reactivity among titers, suggesting that these viruses bind more type-specific antibodies. Comparative PRNTs indicated that both maturation state and strain impact titer levels, but titers measured against the clinical mature strains had lower, more consistent thresholds associated with protection. Thus, in population and vaccine studies, the use of clinical and/or mature strains to measure titers may be preferable as correlates. In contrast, reference standard strains are more easily neutralized, potentially making them preferable for identifying any previous DENV exposure.
We found that high total IgG and GMT measured by standard and mature viruses decreased the risk of dengue. We did not observe any enhancing effects at low to medium total binding antibodies and GMT. This may be because our cohort had few cases overall and almost no severe dengue cases. Previous work demonstrating enhancement at lower titers evaluated larger numbers of cases and identified the strongest enhancement signal for severe disease manifestations33. GMT as measured with standard and mature viruses were also associated with reduced odds of inapparent infection, which could be important for examining viral outbreak potential. The protective effects observed with high total binding and nAbs are consistent with findings from inhibition ELISA and vaccine-specific PRNT assays14, 34.
The type of virus used in the assay seems to impact the accuracy of the correlate. Specifically, although GMT > 200 as measured using standard reference strains significantly decreased dengue risk, 38% of cases occurred among individuals with this antibody level. Additionally, 52% of cases occurred among those with a standard GMT > 100. This is aligned with previous work where standard vaccine strains were used to measure GMT, and GMT > 100 at month 13 post Dengvaxia was associated with 50% vaccine efficacy14. Conversely, antibodies as measured with the mature clinical strains had a lower threshold for protection and only 2% of cases occurred among those with GMT > 100. Notably, 78% of our cohort had multitypic immunity, and GMT is likely a more accurate correlate in individuals with this immune history. For monotypics, the GMT is the average of a positive titer against one serotype, and low titers against the other three. Thus, the GMT would be accurate except against the strain that induced monotypic immunity. In sum, GMT is likely most useful as a correlate when measured with mature clinical strains in more multitypic populations.
Among all the antibodies and cases assessed, the mature clinical serotype-specific nAb titer against the matched serotype had the strongest effect sizes, and correlation analyses suggest that mature clinical virions bind more type-specific antibodies than reference standard strains. The strong protective effects of matched serotype-specific nAb are consistent with previous reports that type-specific antibodies correlate with protection21. Additionally, the highest effect sizes with the mature clinical strains suggest that these virions bind more potent antibodies. Interestingly, heterotypic titers were not associated with protection against DENV2 or DENV3 except that standard titers against DENV1 did decrease DENV3 disease risk. This protective effect and the tight correlation between the standard DENV1 and DENV3 titers suggest that these virions are likely binding some of the same antibodies. Moreover, all serotype-specific standard nAb were strongly correlated with GMT (r ≥ 0.79), while mature serotype-specific nAb had weaker correlations with GMT (r < 0.79). Tight associations between serotype-specific nAb and GMT are likely due to standard virions binding cross-reactive antibodies. In contrast, weaker correlations among clinical mature serotype-specific titers likely reflect more selective antibody binding. We hypothesize that some of this discrimination is due to the structure of mature virions with their less accessible fusion loop, but the use of clinical strains may also contribute. We also evaluated whether the magnitude of antibodies binding to certain antigens that are thought to be the targets of type-specific protection, including EDIII and NS1, were associated with dengue disease and found they were not strong predictors. This finding suggests that measuring a broader antibody pool may be important for dengue immune correlates.
Neutralization assays assessing the impact of strain and maturation state on titers revealed that the reference standard viruses had the highest titers, except against DENV3, where they were similar to the reference mature titers. Conversely, the reference mature virus induced the lowest titers against DENV2. Clinical standard and mature viruses induced similar titer levels, except against DENV2, where the clinical mature titers were lower. High titers against reference standard strains are consistent with previous work indicating that highly laboratory passaged DENVs are the most sensitive to neutralization21. The significant differences in titers measured among all DENV2 strains may reflect increased differences in epitope exposure and sensitivity to neutralization due to viral breathing35. DENV2 aside, the consistent titers among clinical strains suggests that these may bind primarily higher quality antibodies in either maturation state. Thus, maturation state may have less influence on titer and virus neutralization when non-DENV2 clinical strains are used. Although these conclusions require confirmation with larger datasets examining more serotypes, this work suggests that clinical and mature strains are less sensitive to neutralization.
The decreased sensitivity to neutralization may result in the identification of higher quality antibodies measured by mature and clinical strains, thereby lowering the threshold for detecting disease reduction. Specifically, while an ELISA value of 4.5 and a standard GMT > 200 were associated with 70% disease reduction, mature GMT of 84 provided that level of protection. Additionally, only titers measured with the mature clinical strains were associated with 90% disease reduction. However, reaching 90% disease reduction required very high titers suggesting that titers measured with the mature clinical strains may still be an unsatisfying correlate. Moreover, these are time intensive assays requiring specialized labor, which also may limit their utility for large studies.
Our study has several limitations. Given the difficulties of performing PRNTs, our data are limited by the use of random subsets and two-dilution rather than full dilution assays for the standard reference strain. The two-dilution PRNT allowed us to estimate titers on more individuals, has been validated in a separate cohort36, and we found strong correlations between full titers and estimated two dilution titers when assessed with the mature clinical data (Supplemental Fig. 1). Inverse probability weighting was used to complete the datasets, and this was validated by comparing the estimates obtained using this method versus those from bootstrapping and from the full dataset using the ELISA measures (Supplemental Fig. 2). Additionally, our cohort included no severe dengue cases, a high percentage of individuals with multitypic immunity, and infections mostly with DENV2 and DENV3. Thus, nAb thresholds associated with protection may differ in other populations and with the virus strain chosen for the assay. Despite their limitations, these data provide compelling support for the use of mature and clinical isolates in PRNTs and for titers obtained from these assays to serve as correlates of risk.
Overall, an accurate and reproducible correlate of protection would be a valuable contribution to the development of dengue vaccines. After adjusting for age, sex, and enrollment location, we demonstrate that ELISA IgG, GMT, and serotype-specific nAb titers measured by the standard and mature assays serve as correlates of risk for inapparent and symptomatic DENV infection. Mature and clinical strains likely identify higher quality nAbs given the virion structure and similarities with infecting strains as evidenced by a lower, more consistent threshold associated with protection. Thus, nAb titers measured with mature and clinical isolates may be a helpful though exacting correlate of protection, and this should be further validated in other studies.