Selection in a growing bacterial/yeast colony biases results of mutation accumulation experiments

Abstract

Mutations provide the raw material for natural selection to act. Therefore, understanding the variety and relative frequency of different type of mutations is critical to understanding the nature of genetic diversity in a population. Mutation accumulation (MA) experiments have been used in this context to estimate parameters defining mutation rates, distribution of fitness effects (DFE), and spectrum of mutations. MA experiments performed with organisms such as Drosophila have an effective population size of one. However, in MA experiments with bacteria and yeast, a single founder is allowed to grow to a size of a colony (~108). The effective population size in these experiments is of the order of 10. In this scenario, while it is assumed that natural selection plays a minimal role in dictating the dynamics of colony growth and therefore, the MA experiment; this effect has not been tested explicitly. In this work, we simulate colony growth and perform an MA experiment, and demonstrate that selection ensures that, in an MA experiment, fraction of all mutations that are beneficial is over represented by a factor greater than two. The DFE of beneficial and deleterious mutations are accurately captured in an MA experiment. We show that the effect of selection in a growing colony varies non-monotonically and that, in the face of natural selection dictating an MA experiment, estimates of mutation rate of an organism is not trivial. We perform experiments with 160 MA lines of E. coli, and demonstrate that rate of change of mean fitness is a non-monotonic function of the colony size, and that selection acts differently in different sectors of a growing colony. Overall, we demonstrate that the results of MA experiments need to be revisited taking into account the action of selection in a growing colony.

Selection bias in mutation accumulation

“…The correction yields excellent results both in correcting the shape of the DFE, and in correcting summary statistics such as the beneficial fraction of the DFE, which are likewise strongly biased by selection during the MA. Similar simulation results have recently been reported for an empirically-relevant parameter set by Mahilkar et al (2021), where the effects of selection at different spatial positions in a growing colony are also investigated. In that model, the beneficial fraction of the DFE was found to be over-represented by a factor of two or more, consistent with our predictions.…”

“…2019, Mahilkar et al. 2021). For this second case, we take $$\begin{equation*} p(s) = {\begin{cases} (1-\beta ) \alpha _d e^{\alpha _d s} & s\le 0 \\ \beta \alpha _b e^{-\alpha _b s} & s> 0.$$…”

“…A recent simulation study also demonstrates that selection can substantially favor the observation of beneficial mutations during MA (Mahilkar et al. 2021). Although many studies acknowledge the potential selection bias during MA, it is relatively rarely quantified and corrected.…”

“…Such biases can lead, for example, to an underestimation of the deleterious mutation rate (Kibota and Lynch 1996, Gordo and Dionisio 2005, Trindade et al 2010. A recent simulation study also demonstrates that selection can substantially favor the observation of beneficial mutations during MA (Mahilkar et al 2021).…”

“…As a result of the correction, the maximum likelihood estimate of the fraction of beneficial mutations was reduced from as high as 38% to 5.75%. In general, these correction methods adjust estimates of underlying mutation rates, or the mean deleterious effect of mutations, for negative selection; but the effects of positive selection have typically been assumed negligible or explored only in simulations (Trindade et al 2010, Mahilkar et al 2021. None of these approaches allows a correction to the DFE itself.…”