Genome sequencing of Plasmodium malariae identifies continental segregation and mutations associated with reduced pyrimethamine susceptibility

Abstract

Plasmodium malariae parasites are widely observed across the tropics and sub-tropics. This slow-growing species, known to maintain chronic asymptomatic infections, has been associated with reduced antimalarial susceptibility. We analyse 251 P. malariae genomes from 28 countries, and leveraging 131,601 high-quality SNPs, demonstrate segregation of African and Asian isolates. Signals of recent evolutionary selection were identified in genes encoding putative surface proteins (pmmsp1) and putative erythrocyte invasion proteins (pmdpap3, pmrbp2, pmnif4). Amino acid substitutions were identified in orthologs of genes associated with antimalarial susceptibility including 2 amino acid substitutions in pmdhfr aligning with pyrimethamine resistance mutations in P. falciparum. Additionally, we characterise pmdhfr mutation F57L and demonstrate its involvement in reduced susceptibility to pyrimethamine in an in vitro parasite assay. We validate CRISPR-Cas9 mediated ortholog replacement in P. knowlesi parasites to determine the function of pmdhfr mutations and demonstrate that circulating pmdhfr genotypes are less susceptible to pyrimethamine.

Fig. 1. Population structure of P. malariae isolates (n = 157) using 131,601 high-quality filtered SNPs.

A Maximum-likelihood tree calculated using IQTREE with ModelFinder (GTR + F + R4 model), ultrafast bootstrap and sh-aLRT test both to 1000 replicates. Bootstrap scores >50 annotated with a black circle, and branches coloured by isolate continent. The region (separating the African continent into subregions) is annotated using the outer circular track where each isolate has a corresponding bar on the outer circle coloured to the region of origin. C Africa = Central Africa, E&S Africa = Eastern and Southern Africa, N Africa = Northern Africa, SAM = South America, W Africa = Western Africa. B Principal component analysis (PCA) of 157 isolates based on a Manhattan distance matrix, isolates coloured by region. C Admixture prediction of ancestral populations (K) per sample, visualised using 2 ancestral populations, which had the lowest cross-validation (CV) error score, and using 4 ancestral populations. The CV error score for K = 4 was higher, however this is used only to visualise preliminary segregation of African isolates and is not the most likely ancestral pattern. Isolates ordered according to the country of origin and grouped into regional population groups. Admixture analysis used only with countries that had > 5 isolates, (141 total isolates). Each bar represents an isolate which is coloured to reflect the percentage accordance with each predicted ancestral population.

Fig. 2. Structure of DHFR and pmdhfr-ts variants.

A The amino acid sequence of pmdhfr-ts (PmUG01_05034700) was obtained from PlasmoDB with domains annotated using Interpro domains (dihydrofolate reductase enzyme, DHFR: IPR001796; thymidylate synthase, TS: IPR000398). The number of nonsynonymous mutations (NS) found in each domain in the P. malariae database is annotated. B Amino acid sequences were aligned using Clustalo and visualised in Jalview. Conserved amino acids between P. malariae and P. falciparum highlighted in grey, with mutations identified in P. malariae highlighted in blue and amino acid positions associated with pyrimethamine resistance in P. falciparum highlighted in red. C The crystal structure of PfDHFR-TS (with pyrimethamine bound, 3QGT) was downloaded from RCSB Protein Data Bank (grey) and the structure of PmDHFR-TS was predicted using I-TASSER (pink). Protein structures were aligned and visualised in UCSF Chimera, highlighting the overlapping mutations around the pyrimethamine binding site, with the structure for pyrimethamine in black. P. malariae variant positions are annotated in green (49, 57, 58, 114) and P. falciparum variants associated with pyrimethamine resistance are annotated in red (51, 59, 108, 164).

Fig. 3. Pyrimethamine susceptibility measured via growth inhibition assays of parasite lines.

Pkdhfr^WT (Parental parasite line, yellow dot), Pkdhfr^OR (recodonised P. knowlesi DHFR, transfection control line, green dot), Pfdhfr^OR_PfS (P. falciparum sensitive DHFR genotype: N51, C59, S108, purple dot), Pmdhfr^OR_PmS (P. malariae hypothetical sensitive sequence with two mutations: 58S and 114S, navy dot), Pmdhfr^OR_PmUG01 (P. malariae reference sequence, with suspected pyrimethamine resistance: R58 and N114, red dot), Pmdhfr^OR_PmM5 (P. malariae reference sequence: R58 and N114, with additional mutation F57L, pink dot), Pfdhfr^OR_PfR (P. falciparum resistant IRN sequence: 51I, 59 R and 108 N, aqua dot). All data points in both plots represent an independent biological replicate, with the overall mean EC50 in nM given in the summary table. For all lines, a minimum of 3 independent biological replicates were performed. For each independent biological replicate, 3 technical replicates were performed, and the mean EC₅₀ value across the technical replicates is visualised as an individual data point in each graph. In the table, the mean EC50 value across all the biological replicates is shown (Mean (nM)), with the standard deviation (SD) given. A Susceptibility to pyrimethamine using serial dilutions. A log₁₀ scale is used on the plot, with mean EC₅₀ values in nM in the table (standard deviations, SD). Regression analysis was used to investigate the differences in pyrimethamine susceptibility and asterisks used to demonstrate levels of significance (p = 0 − 0.001, ***; p = 0.001 − 0.01, **; p = 0.01-0-.1, *, p values). Regression analysis is a two-sided statistical test, and there is no adjustment for multiple comparisons as the testing is nested within the same model. Pkdhfr^WT (n = 6 biological replicates, minima 0.629, maxima 0.851, median 0.673, Q1 0.653, Q3 0.734), Pkdhfr^OR (n = 4, 0.691, maxima 0.836, median 0.828, Q1 0.792, Q3 0.831), Pfdhfr^OR_PfS (n = 3, minima 0.905, maxima 1.49, median 1.33, Q1 1.12, Q3 1.41), Pmdhfr^OR_PmS (n = 3, minima 1.06, maxima 1.23, median 1.18, Q1 1.12, Q3 1.20), Pmdhfr^OR_PmUG01 (n = 5, minima 3.44, maxima 3.89, median 3.57,Q1 3.51, Q3 3.87), Pmdhfr^OR_PmM5 (n = 3, minima 4.39, maxima 4.56, median 4.39, Q1 4.39, Q3 4.47), Pfdhfr^OR_PfR (n = 3, minima 4.66, maxima 4.92, median 4.80, Q1 4.73, Q3 4.86). P-values not included in the table include pfdhfr^OR_PfS 0.000768, pmdhfr^OR_PmUG01 2e-16, pmdhfr^OR_PmM5 2e-16, pfdhfr^OR_PfR 2e-16. B Dihydroartemisinin (DHA) was used as a control drug, with EC₅₀ values below 7 nM for all parasite lines. Pkdhfr^WT (n = 4 biological replicates, minima 0.231, maxima 0.466, median 0.319, Q1 0.313, Q3 0.466), Pkdhfr^OR (n = 3, minima 0.485, maxima 0.555, median 0.512, Q1 0.499, Q3 0. 533), Pfdhfr^OR_PfS (n = 4, minima 0.424, maxima 0.742, median 0.697, Q1 610, Q3 727), Pmdhfr^OR_PmS (n = 5, minima 0.071, maxima 0.569, median 0.214, Q1 0.180, Q3 0.254), Pmdhfr^OR_PmUG01 (n = 4, minima 0.437, maxima 0.872, median 0.557, Q1 0.522, Q3 0.641), Pmdhfr^OR_PmM5 (n = 3, minima 0.372, maxima 0.820, median 0.805, Q1 0.589, Q3 0.812), Pfdhfr^OR_PfR (n = 5, minima 0.474, maxima 0.668, median 0.638, Q1 0.489, Q3 0.638). In both assays (for pyrimethamine and DHA), pkdhfr^OR is first compared to the parental parasite line (pkdhfr^WT) to confirm that the transfection process itself has no impact on drug susceptibility. Following this, all other parasite lines are compared to pkdhfr^OR to determine the change in drug susceptibility.