Mapping local fitness landscapes over 50,000 generations of evolution

Abstract

Evolution is short-sighted; as a population adapts, natural selection can only act on existing genetic variation, a key tenet of Darwinian evolutionary theory. For asexually evolving organisms such as bacteria, only those genetic variants that arise in the mutational neighborhood of the starting population are visible to selection. Because effects of mutations are context dependent, accumulation of mutations during adaptation can modulate the access to and effects of yet-to-occur mutations. Therefore, the local mutational neighborhood can begin to look different over long periods of time. With the advent of modern genetic engineering and genomics tools, and ability to sequence at scale, we can create and measure effects of thousands of mutations and address a key question: \textit{how different are the local fitness landscapes before and after thousands of generations of adaptation to an environment?}

In this dissertation, I describe my work exploring how the properties of the local fitness landscape of loss of function mutations changes over the long-term evolution experiment (LTEE) in \textit{E. coli}, and computational and experimental approaches to optimize fitness measurements from sequencing based fitness assays. In Chapter 2, I detail research where I construct highly saturated transposon insertion mutagenesis libraries in ancestral and evolved strains from the LTEE, finding that while the statistical properties of the fitness landscape do not change systematically over 50,000 generations of evolution, access to particular evolutionary paths changes consistently across the replicate evolving populations. In Chapter 3, I discuss limitations of pooled sequencing-based fitness assays, tradeoffs in resolving deleterious and near-neutral fitness effects in a single experiment given fixed total sequencing, and suggest best practices to tune design of fitness assay experiments for the effect sizes relevant to the specific biological question. And in Chapter 4, I describe the development of UMI-TnSeq, a unique molecular identifier enabled sequencing library preparation protocol, and apply this method to show that PCR amplification bias is not consequential for transposon insertion sequencing experiments. Lastly, in Chapter 5, I review my findings and speculate on promising future directions stemming from this research.