Gene expression evolution is predicted by stronger indirect selection at more pleiotropic genes

Changes in gene expression play a fundamental role in the process of adaptation and can provide insight into the genetic basis of adaptation. We utilized transcriptome-wide variation in gene expression as a means to uncover genes under selection for expression changes during adaptation to heat and drought stress, and to understand the nature of selection on gene expression traits of the red flour beetle Tribolium castaneum . We showed that estimates of genetic selection on transcript abundance were predictive of evolutionary changes in gene expression after 20 generations of adaptation in seven independent experimental lines. Having measured the genetic covariance between gene expression and relative fitness and among expression traits, we showed that evolutionary changes were driven more by indirect selection acting on genetically correlated partners rather than by direct selection acting on isolated genes. Consequently, genes with central positions in gene co-expression networks experienced stronger selection and exhibited larger evolutionary changes in expression. Our genomic analysis revealed that selection on expression levels is associated with parallel allele frequency changes (AFCs) in the respective genes. More pleiotropic genes and those carrying expression quantitative trait loci (eQTLs) showed a higher degree of parallel evolution. More generally, the stronger the parallelism of AFCs in a gene, the stronger its genetic selection. Contrary to previous evidence of constrained evolution at more connected genes, adaptation was driven by selection acting disproportionately on genes central to co-expression gene networks. We demonstrated that measures of selection at the transcriptome level can provide accurate evolutionary predictions and critical information on the molecular basis of rapid adaptation.