Genomic sequencing of Plasmodium falciparum malaria parasites from Senegal reveals the demographic history of the population

Abstract

Malaria is a deadly disease that causes nearly one million deaths each year. To develop methods to control and eradicate malaria, it is important to understand the genetic basis of Plasmodium falciparum adaptations to antimalarial treatments and the human immune system while taking into account its demographic history. To study the demographic history and identify genes under selection more efficiently, we sequenced the complete genomes of 25 culture-adapted P. falciparum isolates from three sites in Senegal. We show that there is no significant population structure among these Senegal sampling sites. By fitting demographic models to the synonymous allele-frequency spectrum, we also estimated a major 60-fold population expansion of this parasite population ∼20,000-40,000 years ago. Using inferred demographic history as a null model for coalescent simulation, we identified candidate genes under selection, including genes identified before, such as pfcrt and PfAMA1, as well as new candidate genes. Interestingly, we also found selection against G/C to A/T changes that offsets the large mutational bias toward A/T, and two unusual patterns: similar synonymous and nonsynonymous allele-frequency spectra, and 18% of genes having a nonsynonymous-to-synonymous polymorphism ratio >1.

Fig. 1.

Folded allele-frequency spectrum. Allele-frequency spectra of different classes of nucleotide sites all show excess of rare alleles. The neutral allele-frequency spectrum was obtained assuming a constant population size. The excess of rare alleles in the empirical spectrum indicates a population expansion in the past.

Fig. 2.

Demographic models. In the exponential growth model, population size starts increasing earlier than in the two-epoch model, and the current population size is larger.

Fig. 3.

Tajima’s D versus log(π_N/π_S). Tajima’s D and π_N/π_S do not always have the same pattern. By considering both together, genes under selection can more readily be identified. Orange dots show genes with both high Tajima’s D (P value < 0.05) and π_N/π_S (top 5%), red dots show genes with only high Tajima’s D, and blue dots represent genes with only high π_N/π_S.

Fig. 4.

Manhattan plot of iHS P-values on a negative log₁₀ scale. The dashed line indicates a Bonferroni threshold for significance. Dots are colored by chromosome, and their sizes are scaled according to P value. Several regions of recent positive selection are suggested by the iHS statistic, including areas near the known drug resistance loci pfcrt and pfmdr1.

Fig. 5.

Derived site-frequency spectrum. The unfolded site-frequency spectrum was generated by using P. reichenowi as an outgroup. The G/C to A/T spectrum is more skewed toward low frequencies than the A/T to G/C spectrum, suggesting positive selection favoring C/G nucleotides or purifying selection against A/T nucleotides.