Predicting evolution: using comparative experimental evolution to test the role of mutation, selection and genetic background on repeatable evolution

What do we need to know in order to predict evolution? For a long time we have been able to predict the fate of a known mutation within a population. However, a more difficult task is predicting which mutations are likely to emerge, and the consequences of those mutations within the context of the pre-existing genetic background. We know that there are certain biases that make some mutations more likely to occur than others, and we know that the effect of mutations on an individual's fitness can vary depending on the mutations already carried by that individual. But, we have yet to bring these pieces of information together to enable effective forecasting of likely adaptive mutations in an evolving population.
The ability to forecast adaptive evolution of populations has many important implications. It will improve our ability to predict antibiotic, herbicide and pesticide resistance; grant opportunities to optimise treatment strategies for cancers and infectious diseases; and in a rapidly changing climate, allow strategic manoeuvres to limit detrimental effects to ecosystems and at risk populations.

To address this problem, we will compare the evolutionary trajectories of two strains of bacteria of the same species that show different adaptive routes to the same selective challenge - one repeatable the other variable. My previous work has used real-time evolution of microbes in the laboratory to show that a non-motile bacteria can re-evolve motility within 96 hours. Interestingly, we found the same mutation in 90% of cases. We repeated this experiment in a different strain of bacteria of the same species: they were also able to rescue motility within 96 hours via mutations in the same genes, or within the same network of genes, but never at the same site. These bacteria are closely related enough such that they share most of their genes and these genes carry out the same functions. However, they also carry a number of differences across their genomes. By comparing these two strains under the same selective conditions, we are able to experimentally test what might be causing differences in the evolved mutations conferring motility, and the consequences of these differences in driving predictable evolutionary trajectories.

We will test whether there are factors that might be biasing which mutations are emerging. To do this we will evolve both bacterial strains, which have been genetically engineered to be non-motile, under positive selection and in the absence of selection for motility. We will look for differences in the frequency of mutations that rescue motility across the two bacterial strains. This will inform us as to whether differences in the structure or composition of the genome might be contributing to accessible mutations. Next, we will reciprocally generate mutations that confer motility in either strain and measure the fitness effect of these mutations across strains. This will tell us how differences in the genome translate into different fitness effects of the same mutations in the same genes. Finally, we will maintain selection for motility in the reciprocal motile lines across each strain, and select for faster motile bacteria. We will look to see whether mutations that confer a faster moving bacteria are dependent on the mutations that precede it, and whether certain mutations are more common in one strain compared to the other. This will tell us how accessible certain mutational routes are given differences in fitness effects across different bacteria.

By gaining an in depth understanding of the rules that determine repeatable evolution in a simple and tractable system, we can build the foundation for a more generalised ruleset. The principals discovered here will help improve our understanding of how underlying mutational biases and the wider genetic background contribute to accessible adaptive mutations. With a long term goal of improved ability to forecast adaptive evolution more generally.

The ability to predict adaptive evolution has important implications, for example in evolving resistance profiles and optimising treatment strategies. However, there is currently a major gap in our knowledge limiting our potential: that is accurately predicting the likelihood of emergence of particular mutations within a population, and the likely fitness consequence of mutations within a given genetic background.
My recent work has used experimental evolution to explore the rapid evolutionary re-wiring of the flagellum network via the nitrogen regulation pathway in two engineered non-flagellate strains of Pseudomonas fluorescens (SBW25 and Pf0-1). SBW25 showed remarkably repeatable evolution, with the same SNP rescuing motility in 90% of cases. However Pf0-1 showed strikingly lower levels of repeatability: mutations were in the same genes, or within the same network of genes, but never at the same site. By comparing the evolutionary trajectories of these two strains under the same selective challenges we will investigate the role of biases and genetic background in determining accessible mutational routes.
Firstly, we will look for instances of biases that might skew the frequency at which certain mutations arise in a population by evolving non-motile bacteria from each strain in the presence and absence of selection for motility. Secondly, we will reciprocally generate mutations that confer a motile phenotype in each strain, and assay for phenotypic, pleiotropic and epistatic effects. And lastly, we will empirically test the interplay between mutation and genetic background in realised evolutionary trajectories by comparing the same mutation variations engineered into each bacterial strain under continued selection for motility. This will tell us how accessible certain mutational routes are given differences in fitness effects and biases. The general principals discovered here will help build a foundation for further comparative studies.

This research proposal uses innovative methods to address questions that directly relate to our ability to predict the fate of evolving populations. But how could the ability to forecast adaptive evolution benefit society? For example, we could predict the types of mutation and the fate of mutations in a bacterial population under selection to evolve drug resistance. This tackles one of the BBSRC responsive mode priorities that is combatting antimicrobial resistance. The wide reaching impact aim of this proposal will be to encourage those within and outside the academic community to think about the complexity of knowledge required for accurate evolutionary forecasting and the powerful potential applications such knowledge would provide.

Effective communication of the broad evolutionary principals from the proposed research will be of benefit in many educational circles. At the secondary level there is an opportunity to develop, optimise and dispatch a practical lab schools package that will demonstrate the principals of adaptive evolution and trade-offs associated with evolutionary change. One benefit of this experimental system proposed is that the desired phenotype evolves very quickly (over the course of a weekend) and the phenotype itself is very obvious once evolved (motility). Developed and optimised in the Milner Centre, we will invite teachers to learn a simple lab practical that will allow pupils to experience evolution in action for themselves, in real time. Hands-on learning creates opportunities for deeper experiential learning and better comprehension of subject matter. In addition, it can inspire young scientists to get excited about microbiology and creates opportunity for the discussion of its importance in wider society.

We will engage with the general public via the "Soap Box Science" platform. These nationwide (and increasingly international) annual events raises the profile of women in science and gives an open platform to allow the general public to interact with your science. We will challenge the public to consider why it matters whether we can predict evolutionary outcomes, and the ways such knowledge could be applied to current world problems. If the PDRA hire is female, I will encourage her to take up this opportunity to engage with the wider public. Otherwise, I will put myself forward for this activity.

We hope to organise a symposium at the international 2021 SMBE conference in New Zealand on "What do we need to know to predict evolutionary outcomes?". This will allow us to engage with the wider academic community and create opportunities to forge networks and collaborations during the project.

Finally, career development of the PDRA is an essential goal of the project. The University of Bath and the Milner Centre for Evolution promote career development of research staff through annual appraisals, career advice and training e.g. in science communication and bioinformatics (BaMBU - Bath Milner Bioinformatic Users). Findings will be published in leading international journals, and shared via oral presentations at UK and international research conferences.