Identification and characterization of adaptive traits

Understanding adaptation is a long-term goal of evolutionary biology, but it is not clear what phenotypes are targeted by natural selection. Hence, for a long-time researchers were making informed guesses on which phenotypes may be selected and studied their geographic distribution and evolutionary dynamics. With the progress of the sequencing technologies, an alternative approach became more popular, which inferred the selected traits based on genes with a genomic selection signature. At least two factors have precluded further advances a large number of outlier loci and pleiotropy of candidate genes. This proposal turns the emphasis from studying genomic signatures of selection back to investigating the traits that are actually selected. It is proposed to measure many phenotypes by means of state of the art omics-techniques in ancestral and evolved populations to identify the selected traits in replicated Drosophila populations that evolved in the same environment and adapted the same traits independently, but used different paths to do so. Once established by this project, it is anticipated that the strategy can be applied to other species and to natural populations and it will therefore have a broad impact on understanding adaptation in the laboratory and in natural populations.

Selection operates on phenotypes (traits), but the distinction between selected and correlated phenotypes is a long-standing problem. Recent progress in high-throughput sequencing has led to a shift in the emphasis of evolutionary research from phenotypes to the signature of selection in the genome. While increasing amounts of sequence data are becoming available and improved statistical methods for the detection of selection signatures are being developed, very rarely have such "reverse ecology" approaches resulted in the identification and characterization of selected traits. In particular, traits with a complex genetic basis are difficult to identify from allele frequency shifts because the associated allele frequency changes are very small and in replicate experiments, different loci may contribute to the selected shift in phenotype. This project aimed to develop a new approach to identify selected traits. The reasoning builds on the idea that traits are organized in a hierarchical manner, with fitness being on top of the hierarchy and sequencing data at the lowest. Using replicate populations the selected trait can be identified by the lowest hierarchical level at which no redundancy is observed-i.e. a fully parallel response in all replicates. The project took advantage of 10 replicate populations, which evolved for 60 generations from the same founder population in a novel hot environment. We collected intermediate molecular phenotypes by quantifying metabolites, RNA and proteins. Consistent with the concept of trait hierarchies, we found that metabolites were more parallel than RNA and proteins, but we did not find evidence for metabolites being the direct target of selection. The comparison of RNA and protein data showed that proteomics provided basically no new insights compared to the RNA-Seq data. For the rest of the project we focussed on the RNA-Seq data. First we evaluated the evolution of phenotypic variance to determine how polygenic gene expression is. Our results showed that for most genes with a shift in mean expression, the variance did not change. Using computer simulations, we show that this pattern is not expected for a very simple trait architecture, suggesting that trans-effects play a key role in shaping the adaptive gene expression evolution. We also showed that for a small subset of genes the variance in gene expression was reduced, but not the mean. We propose that this pattern reflect stronger stabilizing selection under simple laboratory conditions. Finally, we focussed on the influence of pleiotropy on the evolution of gene expression. Our results showed a positive correlation between parallelism and pleiotropy, a pattern we attribute to synergistic pleiotropy, where one gene affects multiple traits and the fitness effects are positively correlated.