Antagonistic selection and pleiotropy constrain the evolution of plant chemical defenses

Abstract

When pleiotropy is present, genetic correlations may constrain the evolution of ecologically important traits. We used a quantitative genetics approach to investigate constraints on the evolution of secondary metabolites in a wild mustard, Boechera stricta. Much of the genetic variation in chemical composition of glucosinolates in B. stricta is controlled by a single locus, BCMA1/3. In a large-scale common garden experiment under natural conditions, we quantified fitness and glucosinolate profile in two leaf types and in fruits. We estimated genetic variances and covariances (the G-matrix) and selection on chemical profile in each tissue. Chemical composition of defenses was strongly genetically correlated between tissues. We found antagonistic selection between defense composition in leaves and fruits: compounds that were favored in leaves were disadvantageous in fruits. The positive genetic correlations and antagonistic selection led to strong constraints on the evolution of defenses in leaves and fruits. In a hypothetical population with no genetic variation at BCMA1/3, we found no evidence for genetic constraints, indicating that pleiotropy affecting chemical profile in multiple tissues drives constraints on the evolution of secondary metabolites.

Keywords: Boechera stricta; G-matrix; Plant-insect interactions; evolutionary constraints; glucosinolates; selection analysis.

Figure 1. Spatial pattern of glucosinolate profile.

Family mean proportions of methionine-derived glucosinolate compounds (in purple) and branch-chain amino acid-derived compounds (in yellow) in cauline leaves. The white square indicates the location of the common garden. Charts generated by PhyloGeoViz (Tsai 2011). Terrain image copyright by Google Earth.

Figure 2. Selection against the direction of the genetic covariance for standardized BC-ratio in cauline leaves and fruits in the observed population and the BC subpopulation.

Points represent genotype means, scaled to a mean of zero. Panel A shows the full dataset. The genetic covariance between the two traits was significant (P < 0.0001). Both traits had significant selection gradients (P < 0.05). Panel B shows the hypothetical BC population, the subset of the population that produced BC-GS. In the BC population, the genetic correlation between the two traits was not significant. The selection gradient was significant for BC-ratio in the fruits, but not for the cauline leaves. In both panels, the red arrow indicates β, the vector of selection gradients. The blue arrow indicates $\mathbf{\Delta }{\overline{z}}_i$, the response to selection with the covariances included. The black arrow indicates $\mathbf{\Delta }{\overline{\mathit{z}}}_{\text{nc}i}$, the response to selection when the covariances are computationally set to zero. The length of each of the arrows has been increased by a factor of four for improved visibility.

Figure 3. Selection gradients and predicted responses to selection with and without genetic covariances.

For each trait, we show the selection gradient (blue arrow), $\mathbf{\Delta }{\overline{z}}_i$, the response to selection with covariances (black dotted line), and $\mathbf{\Delta }{\overline{\mathit{z}}}_{\text{nc}i}$, the response to selection when covariances are eliminated (solid orange line). Traits in bold had a significant difference between $\mathbf{\Delta }{\overline{z}}_i$ and $\mathbf{\Delta }{\overline{\mathit{z}}}_{\text{nc}i}$.