Evolving gene expression: from G to E to GxE

Abstract

Analyses of gene expression data sets for multiple individuals and species promise to shed light on the mode of evolution of gene expression. However, complementary complexities challenge this goal. Characterization of the genetic variation underlying gene expression can easily be compromised by lack of environmental control. Conversely, the breadth of conclusions from studies of environmental effects has been limited by the use of single strains. Controlled studies have hinted at extensive genexenvironment interaction. Thus, both genetics and environment are key components in models of the evolution of gene expression. We review the literature on the evolution of gene expression in terms of genetics (G), environmental response (E) and GxE interactions to make this conceptual point.

Figure 1. Gene expression evolution under models of stabilizing selection, directional selection and neutral drift

Grey bell curves represent population genetic variation in the level of expression of a gene in a particular environment. Red curves represent fitness in relation to gene expression level. Black lines represent exemplar mutants that survive or die (red cross) each generation. Identical mutations generated under varying selective conditions produce different final expectations for the mean and variance of distributions of gene expression levels. A) Stabilizing selection: The production of new genetic variants is balanced by selection against variants at the extremes of the distribution of gene expression levels. B) Directional selection: Selection preferentially eliminates genetic variants at one end of the distribution, leading to shifting of the distribution mean until a new balance is achieved or a new selective regime is entered. C) Neutral drift: Under unbounded neutrality, mutational biases and sampling dynamics determine evolutionary trends. If all expression levels are equally fit, expected variance will increase with time. Depending on population sizes, chance sampling effects could lead to occasional loss of variation and consequential distributional shifts. The expected mean of the distribution will remain unchanged.