A High-Throughput Yellow Fever Neutralization Assay

Abstract

Quick and accurate detection of neutralizing antibodies (nAbs) against yellow fever is essential in serodiagnosis during outbreaks for surveillance and to evaluate vaccine efficacy in population-wide studies. All of this requires serological assays that can process a large number of samples in a highly standardized format. Albeit being laborious, time-consuming, and limited in throughput, the classical plaque reduction neutralization test (PRNT) is still considered the gold standard for the detection and quantification of nAbs due to its sensitivity and specificity. Here, we report the development of an alternative fluorescence-based serological assay (SNT^FLUO) with an equally high sensitivity and specificity that is fit for high-throughput testing with the potential for automation. Finally, our novel SNT^FLUO was cross-validated in several reference laboratories and against international WHO standards, showing its potential to be implemented in clinical use. SNT^FLUO assays with similar performance are available for the Japanese encephalitis, Zika, and dengue viruses amenable to differential diagnostics. IMPORTANCE Fast and accurate detection of neutralizing antibodies (nAbs) against yellow fever virus (YFV) is key in yellow fever serodiagnosis, outbreak surveillance, and monitoring of vaccine efficacy. Although classical PRNT remains the gold standard for measuring YFV nAbs, this methodology suffers from inherent limitations such as low throughput and overall high labor intensity. We present a novel fluorescence-based serum neutralization test (SNT^FLUO) with equally high sensitivity and specificity that is fit for processing a large number of samples in a highly standardized manner and has the potential to be implemented for clinical use. In addition, we present SNT^FLUO assays with similar performance for Japanese encephalitis, Zika, and dengue viruses, opening new avenues for differential diagnostics.

Keywords: Japanese encephalitis virus; PRNT; Zika virus; dengue virus; high-throughput; neutralization assay; reporter virus; serodiagnosis; yellow fever virus.

FIG 1

A high-throughput fluorescence-based seroneutralization assay for YFV diagnostics. (a) Schematic representation of YFV-mCherry. The reporter mCherry gene was inserted immediately downstream of codon 21 of the YF17D C gene and flanked by a Thosea asigna virus self-cleaving 2A peptide at the 3′ end, followed by a repeat of C gene codons with an alternative sequence (C2-C21*). (b) Assay flowchart of SNT^FLUO. A detailed step-by-step bench protocol is provided as an extended data file. YFV-mCherry was coincubated for 1 h with serially diluted sera in triplicate prior to infecting preseeded BHK-21J cells in a 96-well plate. Three days postinfection, cells were fixed, and the fluorescent spots of infected cells were quantified and analyzed using a CTL ImmunoSpot reader and Genedata Screener software tool, respectively. (c and d) Representative image and neutralization curves of a counted plate containing 2 positive and 1 negative serum samples. (e) Neutralization curves obtained by SNT^FLUO and PRNT on the WHO reference sample. Data are means ± standard deviations of three (d) or six (e) replicates.

FIG 2

Benchmarking against other seroneutralization assays. (a and b) Correlation analysis of PRNT and SNT^FLUO (a) and SNT^CPE and SNT^FLUO (b). Left panels show linear regression analysis to calculate correlation coefficients. The Pearson correlation efficient, R², represents number of tested sera (n), and P values are indicated. Perfect correlation is indicated by the red dashed line, whereas correlation between SNT₅₀^FLUO and PRNT₅₀ or SNT₅₀^FLUO and SNT₅₀^CPE is indicated by a black solid line. Ninety-five percent confidence intervals are indicated by gray shaded areas. LLOQ, lower limits of quantification. Right panels show Bland-Altman analysis to estimate the degree of agreement between assays. Differences between PRNT₅₀ and SNT₅₀^FLUO and SNT₅₀^CPE and SNT₅₀^FLUO values are compared with their average log₁₀ neutralization titer. The lines of no bias (dotted line), average bias (solid line), and 95% confidence intervals (i.e., lower and upper limits of agreements [LOA; dashed lines]) are shown. Data are means from three replicates.

FIG 3

SNT^FLUO cross-validation by reference laboratories. YF17D-vaccinated (n = 6) and unvaccinated (n = 4) monkey sera were serially 1:10 diluted (1:10¹ to 1:10⁴) and assessed for the presence of neutralizing antibodies by the three different laboratories using their respective assays. The EC₅₀ values for each serum sample between the upper and lower quantification limits were used for correlation analysis (see Table S1 in the supplemental material). Data were analyzed by linear regression to calculate correlation coefficients. The Pearson correlation efficient, R², indicates number of tested sera (n), and P values are indicated. Perfect correlation is indicated by a red dashed line, whereas correlation between the different assays is indicated by a black solid line. Ninety-five percent confidence intervals are indicated by gray shaded areas. LLOQ, lower limits of quantification. Data are the means of six (ITM and our laboratory) or two (Sciensano) replicates.

FIG 4

Specificity of SNT^FLUO against other flaviviruses. (a) Schematic representation of mice vaccination scheme for YF17D (n = 8), JEV (n = 8), ZIKV (n = 9), and DENV2 (n = 5), and day of serum collection (total of 30 serum samples). (b) Schematic representation of YFV-mCherry, JEV-eGFP, ZIKV-mCherry, and DENV2-mCherry. Reporter viruses were generated in a similar way as for YFV-mCherry, with the exception of extended and alternative codons of the C gene, respectively, flanking both ends of the fluorescent tag sequence (34 for JEV and ZIKV and 35 for DENV2). (c) Specificity heatmap of neutralizing titers (SNT₅₀^FLUO) of the 30 serum samples tested, using the 4 reporter flaviviruses. Data are the means of three replicates. Heatmap gradient indicates log₁₀ SNT₅₀^FLUO values.

FIG 5

High-throughput performance of SNT^FLUO. (a) We analyzed 600 assay plates containing approximately 2,000 serum samples for assay robustness by calculating Z′ values (left), percent coefficients of variation (CV%; middle), and signal-to-background ratios (S/B; right). Gray area in the left panel indicates Z′ score between 0 and 0.5. Plates with a Z′ score between 0.5 and 0.1 were subjected to visual inspection by fluorescence microscopy (orange circles). Plates with a Z′ score of <0.1 were rejected (red circles). (b) YF17D vaccination efficiency in a larger population-wide study prior to (n = 249) and postvaccination (n = 478). Blood was collected on day 14 and day 28 postvaccination and assessed for its neutralizing activity. Data are presented as log₁₀ SNT₅₀^FLUO (left y axis) or transformed to log₁₀ mIU/mL (right y axis). Data are the means of three replicates. LLOD, lower limits of detection.