Population Genomics of Plasmodium malariae from Four African Countries

Abstract

Plasmodium malariae is geographically widespread but neglected and may become more prevalent as P. falciparum declines. We completed the largest genomic study of African P. malariae to-date by performing hybrid capture and sequencing of 77 isolates from Cameroon (n=7), the Democratic Republic of the Congo (n=16), Nigeria (n=4), and Tanzania (n=50) collected between 2015 and 2021. There is no evidence of geographic population structure. Nucleotide diversity was significantly lower than in co-localized P. falciparum isolates, while linkage disequilibrium was significantly higher. Genome-wide selection scans identified no erythrocyte invasion ligands or antimalarial resistance orthologs as top hits; however, targeted analyses of these loci revealed evidence of selective sweeps around four erythrocyte invasion ligands and six antimalarial resistance orthologs. Demographic inference modeling suggests that African P. malariae is recovering from a bottleneck. Altogether, these results suggest that P. malariae is genomically atypical among human Plasmodium spp. and panmictic in Africa.

Figure 1 –. Complexity of infection in A) P. malariae by country and B) geographically matched P. falciparum and P. malariae isolates overall.

COI values are significantly lower in P. malariae than P. falciparum (ANOVA F = 12.5, p < 0.001, df = 1), but there is no significant variation by country within species.

Figure 2 –. A) Nucleotide diversity (π) of orthologous genes among P. malariae and P. falciparum isolates.

A log-transformed boxplot of π for each gene is shown for each of the 1,377 orthologs retained after masking. 68 orthologs where missing data precluded π calculation for P. malariae are not shown. Boxes represent the 25^th, 50^th, and 75^th percentiles, with outliers represented by dots. The difference in π between the species is highly significant (t = −113, p < 0.001, df = 5340). B) Linkage disequilibrium (LD) decay in P. falciparum and P. malariae. R² values were calculated in PLINK for each distance. LD is higher for P. malariae compared to P. falciparum, with rapid decay of LD in P. falciparum over short distances. The LD decay difference is highly significant (t = −108, p < 0.001, df = 73,920). The insert shows LD from 2 to 10bp, demonstrating that linkage in P. falciparum is higher than in P. malariae only at very short distances (<8 bp).

Figure 3 –. Principal component analysis of monoclonal P. malariae isolates.

71 monoclonal P. malariae isolates and 178,036 biallelic SNPs are included. The first two principal components (percent of total variation explained) are depicted with isolates colored by country of origin (Cameroon n = 6, DRC n = 16, Nigeria n = 3, Tanzania n = 45).

Figure 4 –. Schematics of demographic models tested for goodness of fit.

Plots are organized horizontally, with time in generations on the X-axis and width of colored shape corresponding to effective population size (N_e). T_A indicates ancient population while T_C indicates contemporary population. A) Standard neutral model for one population, with no change in population size. B) Growth model where population growth begins at time T. C) “Bottlegrowth” model where an instantaneous size change is followed by exponential time growth at time T. D) Two epoch model where an instantaneous size change occurs at time T followed by a constant N_e. E) Three epoch model where an instantaneous size change occurs prior to time T, with T_B corresponding to the length of the bottleneck, and bottleneck recovery begins after time T, with the time since bottleneck recovery represented by T_F. Among the five models tested, the three epoch model was the best fit (LL = −514, CL-AIC = −59,663).