Malaria life cycle intensifies both natural selection and random genetic drift

Abstract

Analysis of genome sequences of 159 isolates of Plasmodium falciparum from Senegal yields an extraordinarily high proportion (26.85%) of protein-coding genes with the ratio of nonsynonymous to synonymous polymorphism greater than one. This proportion is much greater than observed in other organisms. Also unusual is that the site-frequency spectra of synonymous and nonsynonymous polymorphisms are virtually indistinguishable. We hypothesized that the complicated life cycle of malaria parasites might lead to qualitatively different population genetics from that predicted from the classical Wright-Fisher (WF) model, which assumes a single random-mating population with a finite and constant population size in an organism with nonoverlapping generations. This paper summarizes simulation studies of random genetic drift and selection in malaria parasites that take into account their unusual life history. Our results show that random genetic drift in the malaria life cycle is more pronounced than under the WF model. Paradoxically, the efficiency of purifying selection in the malaria life cycle is also greater than under WF, and the relative efficiency of positive selection varies according to conditions. Additionally, the site-frequency spectrum under neutrality is also more skewed toward low-frequency alleles than expected with WF. These results highlight the importance of considering the malaria life cycle when applying existing population genetic tools based on the WF model. The same caveat applies to other species with similarly complex life cycles.

Fig. 1.

Simulation diagram. The simulation model is complete inbreeding because it does not allow for any human host to be multiply infected by parasite lineages from different mosquito vectors. The definitions of parameters are shown in the text under Allele-Frequency Change from Generation to Generation and Table S2.

Fig. 2.

Comparison of probability of loss and average number of mutations after one generation between the malaria life cycle and the WF model. (A) Probability of loss is greater in the malaria model. (B) Average number of mutations after one generation is more extreme in the malaria model except when the mutation is neutral.

Fig. 3.

Comparison of longer timescale properties between the malaria life cycle and the WF model. (A) Segregation time is shorter in the malaria model. (B) Time to fixation for beneficial alleles is shorter in the malaria model when the selection coefficient is smaller than threshold value (s = 0.01 under the default settings) and is greater than in the WF model if the selection coefficient exceeds the threshold. (C) Probability of fixation of beneficial alleles in the malaria model is smaller than in the WF model due to greater effects of random genetic drift and stochastic transmission among hosts in the malaria life cycle. (D) Time to loss of deleterious alleles is shorter in the malaria model, suggesting highly efficient purifying selection in the malaria parasite.

Fig. 4.

Allele-frequency spectrum of neutral alleles in the malaria model compared with WF. The allele-frequency spectrum in the malaria model is more skewed toward lower-frequency alleles than in the WF model.