The gold-standard method for quantifying neutralizing antibody responses to many viruses, including dengue virus (DENV), is the plaque reduction neutralization test (PRNT, also called the immunofocus reduction neutralization test). The PRNT conducted on 96-well plates is high-throughput and requires a smaller volume of antiserum than on 6- or 24-well plates, but manual plaque counting is challenging and existing automated plaque counters are expensive or difficult to optimize. We have developed Viridot (Viridot package), a program for R with a user interface in shiny, that counts viral plaques of a variety of phenotypes, estimates neutralizing antibody titers, and performs other calculations of use to virologists. The Viridot plaque counter includes an automatic parameter identification mode (misses <10 plaques/well for 87% of diverse DENV strains [n = 1521]) and a mode that allows the user to fine-tune the parameters used for counting plaques. We compared standardized manual and Viridot plaque counting methods applied to the same wells by two analyses and found that Viridot plaque counts were as similar to the same analyst's manual count (Lin's concordance correlation coefficient, ρc = 0.99 [95% confidence interval: 0.99-1.00]) as manual counts between analysts (ρc = 0.99 [95% CI: 0.98-0.99]). The average ratio of neutralizing antibody titers based on manual counted plaques to Viridot counted plaques was 1.05 (95% CI: 0.98-1.14), similar to the average ratio of antibody titers based on manual plaque counts by the two analysts (1.06 [95% CI: 0.84-1.34]). Across diverse DENV and ZIKV strains (n = 14), manual and Viridot plaque counts were mostly consistent (range of ρc = 0.74 to 1.00) and the average ratio of antibody titers based on manual and Viridot counted plaques was close to 1 (0.94 [0.86-1.02]). Thus, Viridot can be used for plaque counting and neutralizing antibody titer estimation of diverse DENV strains and potentially other viruses on 96-well plates as well as for formalization of plaque-counting rules for standardization across experiments and analysts.

Neutralizing antibodies are important correlates of protection against dengue. Yet, determinants of variation in neutralization across strains within the four dengue virus serotypes (DENV1-4) is imperfectly understood. Studies focus on structural DENV proteins, especially the envelope (E), the primary target of anti-DENV antibodies. Although changes in immune recognition (antigenicity) are often attributed to variation in epitope residues, viral processes influencing conformation and epitope accessibility also affect neutralizability, suggesting possible modulating roles of nonstructural proteins. We estimated effects of residue changes in all 10 DENV proteins on antigenic distances between 348 DENV collected from individuals living in Bangkok, Thailand (1994-2014). Antigenic distances were derived from response of each virus to a panel of twenty non-human primate antisera. Across 100 estimations, excluding 10% of virus pairs each time, 77 of 295 positions with residue variability in E consistently conferred antigenic effects; 52 were within ±3 sites of known binding sites of neutralizing human monoclonal antibodies, exceeding expectations from random assignments of effects to sites (p = 0.037). Effects were also identified for 16 sites on the stem/anchor of E which were only recently shown to become exposed under physiological conditions. For all proteins, except nonstructural protein 2A (NS2A), root-mean-squared-error (RMSE) in predicting distances between pairs held out in each estimation did not outperform sequences of equal length derived from all proteins or E, suggesting that antigenic signals present were likely through linkage with E. Adjusted for E, we identified 62/219 sites embedding the excess signals in NS2A. Concatenating these sites to E additionally explained 3.4% to 4.0% of observed variance in antigenic distances compared to E alone (50.5% to 50.8%); RMSE outperformed concatenating E with sites from any protein of the virus (ΔRMSE, 95%IQR: 0.01, 0.05). Our results support examining antigenic determinants beyond the DENV surface.