Figures and data
Overview of the experimental procedures (a) and the variation in parallel gene expression evolution across genes (b,c).
a. Experimental Evolution: ten replicated populations seeded from one common founder population have been evolving for >100 generations in a hot laboratory environment at 18 and 28°C. Common Garden Experiment: at the 103rd generation of the evolution experiment, the ten evolved populations (each with three biological replicates) were maintained together with the reconstituted ancestral population (with five biological replicates) for two generations in the same environment as in the evolution experiment. After two generations in common garden, 50 males from each biological replicates were pooled and subjected to RNA-Seq. b. Replicate frequency spectrum. Number of populations (x-axis) in which a given gene experienced a significant change in gene expression. The y-axis indicates the number of genes in each category. Most of the genes experienced a significant change in gene expression in few evolved populations while much fewer genes were significant in all 10 populations. This pattern suggests that the parallelism in gene expression evolution differs across genes. c. The distribution of gene expression evolution parallelism (log(1/F)) across genes. Larger log(1/F) values indicate more parallel evolution of a gene. The exhibit variation suggests that genes varied in their parallelism of its expression evolution.
Association between the magnitude of evolution parallelism and the strength of pleiotropy.
The distribution of evolution parallelism (log(1/F)) of different genes is shown in boxplots binned by their strength of pleiotropy (1-). Overall, the strength of pleiotropy was positively correlated with evolution parallelism (=0.26, p-value<2.2e-16).
Association between ancestral variation in gene expression with the strength of pleiotropy (1- τ) (a) and with the evolution parallelism (b).
a. The distribution of gene expression variation in the ancestral population (log (BCV2)) of different genes is shown in boxplots binned by their strength of pleiotropy (1-τ). Overall, the strength of pleiotropy was negatively correlated with ancestral variation in gene expression (ρ =-0.34, p-value<2.2e-16). b. The distribution of evolutionary parallelism (log(1/F)) of different genes is shown in boxplots binned by their strength of ancestral variation. The strength of parallelism is negatively correlated with the ancestral variation in gene expression (ρ =-0.22, p-value<2.2e-16).
Computer simulations illustrate the negative effect of ancestral variation on parallel evolution.
Computer simulations assume a fitness related trait determined by the expression of four genes. The expression of these four genes is determined by a different number of genetic loci, reflecting the influence of pleiotropy on ancestral genetic variation. A shift in trait optimum was used to illustrate the connection between ancestral variation in expression (x-axis) and parallelism of expression change across ten evolved populations (y-axis). The negative correlation between them (rho = -0.26, p-value < 1.77e-07) suggests that the gene with less ancestral variation are resulting in more parallel responses.
Schematic figure illustrating the causal models evaluated.
Five possible relationships between pleiotropic effects, ancestral variation, and evolutionary parallelism. L denotes the likelihood of each model given the data. P is the probability or conditional probability of the measurements for each gene; Pa is the parallelism level and A is the level of ancestral variation in gene expression. Pl is the pleiotropic effects. See materials and methods for a more detailed description.
BIC value for Model I-V. The model with the smallest BIC is the one best support by the data.
