Navigating parasite antigen genetic diversity in the design of Plasmodium vivax serological exposure markers for malaria

Abstract

Background: Plasmodium vivax poses a major obstacle to malaria elimination because it can lie dormant in the liver for weeks or months before reactivating and causing a relapse of infection. These dormant forms (hypnozoites) cannot be detected using standard diagnostics, but recent P. vivax exposure and by proxy, hypnozoite carriage, can be inferred using antibody-based tests (serological markers). In this study, we examined how genetic variation in P. vivax affects the utility of these antibody markers, and whether redesigned antigens could improve performance.

Methods: We analysed global P. vivax genetic data to assess variation in leading serological markers. Based on this, we produced new antigen versions (haplotypes) that better reflect global sequence diversity, compared to the commonly used reference strain (Sal-1). Antibody responses against these new constructs were then tested using samples from well-characterised cohorts in Brazil and Thailand. Antibody levels were assessed in relation to how recently participants had a qPCR-detectable blood-stage P. vivax infection. We compared the ability of the haplotypes and reference constructs to correctly identify individuals infected within the prior 9-months.

Findings: Extensive genetic diversity was identified in two P. vivax antigens, DBPII and MSP5. Several antigens had large numbers of circulating haplotypes globally, with the percentage with similar sequence identity to the reference Sal-1 ranging from 0.4% (MSP5) to 99% (S16). Two antigens exhibited strong differences in immunogenicity by region and construct (RBP2a and DBPII). However, for most proteins (5 out of 8), these differences had little impact on the accuracy of identifying recent exposure. In cases where performance was affected (e.g. RBP2a), this could be overcome by adding multiple antigens into the classification model.

Interpretation: Even highly diverse antigens can be effective serological exposure markers. Our findings highlight the importance of testing the impact of genetic diversity when designing serological tests and suggest practical strategies, such as using a mix of antigens, to ensure consistent performance across regions.

Figure 1.. Nucleotide (π) and haplotype diversity (Hd) in the expressed regions of P. vivax SEM antigens across diverse populations.

The darker color gradient indicates higher π, and the circle’s size corresponds to Hd, with larger nodes representing higher Hd. Diversity was only assessed within the expressed region of the protein used in the P. vivax SEM tool.

Figure 2.. Global nucleotide diversity along the length of all 11 P. vivax genes.

Nucleotide diversity across the gene sequence. Upper schematic highlights the region of the gene included in the expressed P. vivax protein constructs. Coloured blocks highlight different domains based on previous studies, with grey indicating undefined regions. Nucleotide diversity (π) was calculated in sliding window approach (window size=100; step size=3).

Figure 3.. Haplotype network plots and the distribution of diversity-covering haplotypes for each antigen.

Each node in the plots represents a unique haplotype, with the node size corresponding to haplotype frequency, and the color indicating specific country (refer to the legend). The red arrow denotes the known reference alleles Sal-1 and P01, whereas the blue arrows indicate selected haplotypes that could cover the diversity observed in these antigens. Dotted circles represent identified clusters for highly diverse antigens.

Figure 4.. IgG antibody levels in Thai and Brazil individuals against the P. vivax serological exposure marker proteins comparing the reference strain (Sal-1) and the identified haplotypes.

IgG levels were measured at the final visit of yearlong cohorts in Thailand and Brazil. Monthly bleeds enabled qPCR-detection of Plasmodium infections and calculation of the time since last detected P. vivax blood-stage infection. Participants were stratified by this calculation: current P. vivax infection (n=13–20 for Thai and n=39 for Brazil), P. vivax infection 1–9 months ago (n=21–36 for Thai and n=165 for Brazil), P. vivax infection 9–12 months ago (n=12–19 for Thai and n=31 for Brazil), no detected P. vivax infections (n=501–699 for Thai and n=688 for Brazil). IgG levels were also measured in four panels of negative controls, Australian Red Cross (n=100), Volunteer Biospecimen Donor Registry (n=102), Brazil Red Cross (n=96) and Thai Red Cross (n=72).

Figure 5.. Summary performance for correctly classifying recent exposure to P. vivax when using either the reference or variant haplotypes selected.

Results are presented as area under the curve (AUC) of the ROC curves presented in Figure S8, with 95% confidence intervals. Statistical difference between classification for each antigen when using the reference (Sal-1) or the identified haplotype variants is shown. Statistical significance was assessed using bootstrap resampling (2000 replicates). Results were considered significant at p < 0.05 (*), p < 0.01 (**), and p < 0.001 (***).

Figure 6.. Impact of RBP2a genetic diversity when used in combination with other P. vivax SEMs.

Performance of the top eight P. vivax antigens in classifying seropositive infection within the last nine months was measured by the cross-validated AUC of the ROC curve values generated from the random forest classifier. Results are shown for the instances where the model was trained and tested on (A) the combined dataset, including cohort studies from Brazil and Thailand along with negative control samples, and separately on (C) the Brazilian cohort and (C) the Thai cohort. The analysis was conducted using the RBP2a Sal-1 (pink), SEM 28 (green) and SEM 29 (blue) haplotypes.