Effects of selection on gene expression variation and mutation accumulation lines

Abstract

Natural selection is one of the dominant evolutionary forces determining genetic and phenotypic diversity within natural populations. Precise quantification of the effect of selection on trait variation and newly occurring mutations is indispensable both for advancing evolutionary theory and for applications that depend on predicting evolution.

However, in certain situations, a prerequisite for studying how selection shapes geno- and phenotypes requires measurements to be made in the absence of selection.

Here, I investigate selection as an explanatory force shaping trait expression and as a confounding variable in experimental design. Specifically, this thesis investigates selection bias in microbial mutation accumulation (MA) lines (Chapter 2), constructs a distribution of fitness effects (DFE) for single spontaneous mutations in Escherichia coli (Chapter 3) and explores the relationship between the strength of stabilising selection and environmentally-induced gene expression variability in E. coli (Chapter 4).

To obtain an unbiased DFE using MA lines, mutations must accumulate spontaneously and be detected with a probability that does not depend on their fitness effect. Microbial MA experiments attempt to do this by repeatedly bottlenecking each population to a single colony, thereby lowering the effective population size to reduce the effect of natural selection. However, recent theoretical studies indicate that substantial selection bias may still persist in MA lines.

These analyses, though, assumed homogeneous, well-mixed population growth, whereas most microbial MA experiments propagate populations as spatially structured colonies. Spatially explicit simulations showed that colony-based MA can either in- or decrease selection bias relative to homogeneous MA cultures, depending on how colony growth is modelled. Given this uncertainty, I looked for empirical signatures of such bias, whereby I compiled mutation data from 29 colony-based MA experiments and detected a significant deficit of non-synonymous changes, a pattern indicative of purifying selection.

Additionally, I propagated 220 E. coli MA lines for 36 days to accumulate 232 spontaneous mutations. Combined with data from three other E. coli MA studies and using a multinomial model which takes sequence context into account, I once again found a depletion of non-synonymous mutations, although this was non-significant. Together, these results indicate that selection bias in colony-propagated MA lines is likely stronger than previously predicted.

Despite the potential selection bias, the mutants generated in the MA experiment were used to assemble an E. coli DFE, where fitness was quantified as growth rate in Luria Broth. To get the entire shape of the DFE, technically only single mutants should be used to construct the DFE. However, I also measured fitness of lines that accumulated no mutations so I could partition the observed fitness variance among lines into genetic versus non-genetic components. I found substantial between-line differences in fitness, but little evidence that these differences were caused by new mutations. Instead, the between-line differences appeared to be mainly caused by inherited environmental effects. Previous microbial MA studies have assumed such effects to be negligible, and may have therefore incorrectly attributed them to the effects of new mutations, leading to biased estimates of the DFE. In my final chapter, I examined how environmentally induced phenotypic variation relates to the strength of stabilising selection.

Nucleotide diversity of promoter regions was employed as an inverse proxy for stabilising selection strength on gene expression and was quantified for 806 E. coli regulatory sequences. Environmentally-induced gene expression variability was taken from a previously published RNA-seq dataset in which clonal E. coli populations were cultured in 408 distinct environments. Promoter nucleotide diversity was positively correlated with gene-expression variance, suggesting that stabilising selection predominantly reduces environmentally-induced transcriptional variation in E. coli. Overall, in this thesis I show that selection likely strongly biases the observation of new mutations in colony growth MA lines, and that environmentally-induced variability in gene expression can, in part, be explained by the amount of stabilising selection acting on gene expression. I also point out other sources of previously unnoticed bias, such as environmental, non-genetic factors influencing fitness measurements.

Together, these results further elucidate the role of selection in different biological contexts, and, importantly, highlight the critical need for continued validation that empirical measurements correctly capture the evolutionary processes being studied.