Pleiotropy alleviates the fitness costs associated with resource allocation trade-offs in immune signalling networks

Abstract

Many genes and signalling pathways within plant and animal taxa drive the expression of multiple organismal traits. This form of genetic pleiotropy instigates trade-offs among life-history traits if a mutation in the pleiotropic gene improves the fitness contribution of one trait at the expense of another. Whether or not pleiotropy gives rise to conflict among traits, however, likely depends on the resource costs and timing of trait deployment during organismal development. To investigate factors that could influence the evolutionary maintenance of pleiotropy in gene networks, we developed an agent-based model of co-evolution between parasites and hosts. Hosts comprise signalling networks that must faithfully complete a developmental programme while also defending against parasites, and trait signalling networks could be independent or share a pleiotropic component as they evolved to improve host fitness. We found that hosts with independent developmental and immune networks were significantly more fit than hosts with pleiotropic networks when traits were deployed asynchronously during development. When host genotypes directly competed against each other, however, pleiotropic hosts were victorious regardless of trait synchrony because the pleiotropic networks were more robust to parasite manipulation, potentially explaining the abundance of pleiotropy in immune systems despite its contribution to life history trade-offs.

Keywords: coevolution; evolutionary immunology; host–parasite interactions; life history trade-offs; network robustness.

Figure 1.

Independent effector hosts tend to be significantly more fit than shared effector hosts. Each dot represents the arithmetic mean fitness of the shared (black) or independent effector population (red) in the final generation of a simulation for synchronous (filled dots) or asynchronous (open dots) signalling conditions. Top row = scarce resources, middle = alternating resources, and bottom = plentiful resources. Within a resource condition, independence of samples was determined using a Kruskal–Wallis non-parametric ANOVA and if significant differences were detected, multiple comparisons were carried out using pairwise Signed Rank tests. Groups that share a letter are not significantly different from each other and letters are re-used between resource conditions.

Figure 2.

Shared effector hosts outcompete independent effector hosts in most conditions. The black lines show the predicted probability of shared effector hosts winning a competitive simulation when signalling was synchronous, the grey lines show the predicted probability of shared effector hosts winning a competition when signalling was asynchronous. In both cases, the predicted probability was determined by a quasi-binomial regression on the proportion of the population that had a shared effector in the final generation of a simulation. The y-axis shows the proportion of the host population that was composed of shared effector hosts, with the size of dots corresponding to the number of simulations that ended with that proportion. The x-axis shows the number of generations that were allotted for independent evolution prior to competition beginning.

Figure 3.

Shared effector host immune responses tend to be more robust than independent effector host immune responses. Host signalling network robustness was measured as the mean absolute difference between immune effector activity in intact and knockout hosts, using the most common hosts from the end of independent evolution simulations as the population of intact hosts. Higher values indicate networks were less robust; see §2d for detailed methods. The y-axis shows the mean absolute difference between the knockout immune response and the intact host immune response. The columns indicate the resource availability of the simulations intact hosts evolved in. Black/red lines are mean change in effector abundance following knockout. Within a resource condition, independence of samples was determined using a Kruskal–Wallis non-parametric ANOVA and if significant differences were detected, multiple comparisons were carried out using pairwise Mann–Whitney U tests. Groups that share a letter are not significantly different from each other and letters are re-used between resource conditions.