Antigenic variation in Plasmodium falciparum malaria involves a highly structured switching pattern

Abstract

Many pathogenic bacteria, fungi, and protozoa achieve chronic infection through an immune evasion strategy known as antigenic variation. In the human malaria parasite Plasmodium falciparum, this involves transcriptional switching among members of the var gene family, causing parasites with different antigenic and phenotypic characteristics to appear at different times within a population. Here we use a genome-wide approach to explore this process in vitro within a set of cloned parasite populations. Our analyses reveal a non-random, highly structured switch pathway where an initially dominant transcript switches via a set of switch-intermediates either to a new dominant transcript, or back to the original. We show that this specific pathway can arise through an evolutionary conflict in which the pathogen has to optimise between safeguarding its limited antigenic repertoire and remaining capable of establishing infections in non-naïve individuals. Our results thus demonstrate a crucial role for structured switching during the early phases of infections and provide a unifying theory of antigenic variation in P. falciparum malaria as a balanced process of parasite-intrinsic switching and immune-mediated selection.

Figure 1. Two patterns of var gene activation.

Parasites were cloned by limiting dilution and grown in continuous culture. From twenty generations post-cloning the expressed var gene repertoire of each clone was measured by quantitative PCR every few generations. Over time the clones show different reproducible patterns of var transcriptional change. The left panel (A, C, E) shows the clones with stable transcriptional hierarchies, while the right panel (B, D, F) shows the clones where we observed a change in the dominant transcript where a second variant replaces the initial dominant variant after 40–50 generations.

Figure 2. Underlying pathways of var gene switching.

Shown are the results of our analysis of three different clones that exhibited either replacement of the dominant transcript over time, 3D7_AS2 (A) and IT_2B2 (B), or stable expression, IT_2F6 (C). The switch matrix in the left panel represents the switch biases, β_ij, where the size of each circle corresponds to the transition probabilities from gene i to gene j (with 0<β_ij<1); similarly for the vector below the matrix where the size corresponds to the off-rate of each individual var gene, ω_i (with 0<ω_i<0.06). In each matrix we can identify a set of genes with a high transcription bias towards the same gene (here variant 2). The switch pathway suggested by the matrix is illustrated in the middle panel where the arrows represent the switch bias (thicker arrows correspond to higher bias). On the right panel the model output for these ‘best fit’ on- and off-rates is compared to the measured transcription profiles.

Figure 3. Network optimisation over two evolutionary traits.

An initially random network (top left) can be evolved to optimise either robustness (top right), i.e. the potential for evading immune responses, resulting in a fully connected network, or path length (bottom left) which minimised repertoire exposure to the immune system and thus results in a ring-like structure with minimum connectivity between nodes. Optimising over both traits simultaneous produces a network consisting of variants which either switch to many other variants or being switched to from many other variants (bottom right). The lattice-like structure is highlighted to the right of the network, indicating ‘sink’ (red) and ‘source’ nodes (blue).

Figure 4. Effects of switching pattern on malaria infection dynamics.

Simulating malaria infections in the naïve host, here shown as parasitaemia levels of the various antigenic variants under two different assumptions about the nature of switching, does not reveal major qualitative differences between random and preferential switching (A and B, respectively). However, a marked increase in infection length, as measured in multiples of a single variant infection, can be observed once switching is more structured (C). Shown are the median (blue bars) and lower and upper quartiles of 500 model realisations.

Figure 5. Effects of switching pattern on malaria infection dynamics during re-infection.

Simulating infection, clearance and re-infection by the same pathogen reveals the vulnerability of the highly ordered one-to-one switching pathway (A). The expansion-contraction process within the sms pathway allows for greater flexibility to overcome the inhibitory responses to find an alternative route of expression (B). This is clearly demonstrated by both robustness, measured as the proportion of runs where secondary infections were successfully established, and the length of secondary infection (C). Shown are the median (blue bars) and lower and upper quartiles of 500 model realisations.