Maintenance of phenotypic diversity within a set of virulence encoding genes of the malaria parasite Plasmodium falciparum

Abstract

Infection by the human malaria parasite Plasmodium falciparum results in a broad spectrum of clinical outcomes, ranging from severe and potentially life-threatening malaria to asymptomatic carriage. In a process of naturally acquired immunity, individuals living in malaria-endemic regions build up a level of clinical protection, which attenuates infection severity in an exposure-dependent manner. Underlying this shift in the immunoepidemiology as well as the observed range in malaria pathogenesis is the var multigene family and the phenotypic diversity embedded within. The var gene-encoded surface proteins Plasmodium falciparum erythrocyte membrane protein 1 mediate variant-specific binding of infected red blood cells to a diverse set of host receptors that has been linked to specific disease manifestations, including cerebral and pregnancy-associated malaria. Here, we show that cross-reactive immune responses, which minimize the within-host benefit of each additionally expressed gene during infection, can cause selection for maximum phenotypic diversity at the genome level. We further show that differential functional constraints on protein diversification stably maintain uneven ratios between phenotypic groups, in line with empirical observation. Our results thus suggest that the maintenance of phenotypic diversity within P. falciparum is driven by an evolutionary trade-off that optimizes between within-host parasite fitness and between-host selection pressure.

Keywords: Plasmodium falciparum; evolutionary trade-off; mathematical modelling; phenotypic diversity; var genes.

Figure 1.

Default model behaviour under variant-specific immunity only. (a) Time series of a single model run show the changing prevalence of strains with differently structured repertoires, with strains coloured according to their repertoire structure (A : B). Phenotypically diverse strains are driven to extinction as a pair of strains consisting of only one or the other phenotype (i.e. with repertoire structures (9 : 0) and (0 : 9)) start to dominate. (b) Bar graph of the frequency by which different repertoire structures dominate the population, based on 10 000 model runs, shows how phenotypically less diverse parasites are favoured under variant-specific immunity. Stacked bars are used to visually represent within-repertoire phenotype distribution. Other parameter values: r = 9, N_A = 13, N_B = 13.

Figure 2.

Cross-immunity selects for phenotypically diverse repertoires. (a) Total infection length as a function of the number of antigenic variants presented over the course of infection, c, under variation in the degree of cross-immunity, σ. For high levels of cross-immunity, infection length starts to plateau well before the parasite can exhaust its antigenic repertoire and additionally expressed genes will not contribute to further infection length. (b) Time series of a single model run with temporary cross-immunity (σ = 0.3) showing the competitive exclusion of phenotypically uniform parasites (blue lines) in favour of strains containing partitioned and hence phenotypically diverse repertoires (red and purple lines). (c) Bar graph of the average strain dominance frequency based on 10 000 model runs shows how phenotypically diverse strains gain a selective advantage under phenotype-specific cross-immunity (σ = 0.3). Stacked bars are used to visually represent within-repertoire phenotype distribution. Other parameter values: r = 9, N_A = 13, N_B = 13.

Figure 3.

Pairwise and mean infection length under changes in cross-immunity. (a–d) Matrix plots show the (i,j) pairwise infection lengths, resulting from an infection of strain i in a host previously exposed to strain j. The bar graphs depict the mean infection length of each repertoire structure (represented as stacked bars to represent within-repertoire phenotype distribution), derived by averaging the strains' length of infection against all non-self-strains (i.e. the column mean of the pairwise infection matrix). As cross-immunity is increased ((a–d)), phenotypically diverse repertoires start to gain a competitive advantage. Note that different scales are used to illustrate the qualitative change in repertoire fitness. (a) σ = 0, (b) σ = 0.1, (c) σ = 0.3, (d) σ = 0.7. Other parameter values: r = 9, N_A = 9, N_B = 9.

Figure 4.

Functional constraints on PfEMP1 select for skewed phenotype distributions. Bar graphs show the average strain dominance frequencies based on 10 000 model runs, with stacked bars used to represent within-repertoire phenotype distribution. (a) Considering different antigenic pool sizes of phenotypes A and B (N_A = 36 and N_B = 9) with variant-specific immunity only selects for parasites with mono-phenotypic repertoires. (b) Including cross-immunity (σ = 0.1) restricts the benefit of expressing more variants of the same phenotype, leading to the evolution of phenotypically diverse repertoires skewed towards the more diverse phenotype (A). The inset shows the repertoire distribution of UpsA versus non-UpsA var genes in seven sequenced P. falciparum isolates, shows a skewed and conserved repertoire partitioning. (c) Differences in the degree of cross-reactivity, with σ_A = 0.1 and σ_B = 0.8, has the same effect as different antigenic pool sizes and select for repertoires with uneven numbers of phenotypic variants. Other parameter values: r = 9.