Epigenetic feedback on noisy expression boosts evolvability

Abstract

Adapting organisms often face fitness valleys, i.e. barriers imposed by ubiquitous genetic interactions, while optimizing functions. Elucidating mechanisms that facilitate fitness valley traversals is integral to understanding evolution. Therefore, we investigated how protein expression noise, mechanistically decomposed into instant variation and epigenetic inheritance of optimal protein dosage ( transgenerational feedback), shapes the fitness landscape. For this purpose, we combined a minimal model for expression noise with diverse data of Saccharomyces cerevisiae from literature on e.g. expression and fitness to representatively simulate mutational fitness effects. For our proxy of point mutations, which are very often near-neutral, instant dosage variation by expression noise typically incurs a 8.7% fitness loss (17% in essential genes) for non-neutral point mutations. However, dosage feedback mitigates most of this deleterious effect, and additionally extends the time until extinction when essential gene products are underexpressed. Taken together, we consider dosage feedback as a relevant example of Waddingtons canalization: a mechanism which temporarily drives phenotypes towards the optimum upon a genetic mismatch, thereby promoting fitness valley traversal and evolvability. Author summaryGene products frequently interact to generate unexpected phenotypes. This universal phenomenon is known as epistasis, and complicates step-wise evolution to an optimum. Attempts to understand and/or predict how the optimum is found are further compromised by the countless combinations of mutations that are considered by nature, and necessitate the formulation of general rules on how the obstacles that epistasis presents are bridged. To make such a rule as insightful as possible, we reduced cell division to a generation-based model focusing on one protein at a time for reproductive success. Importantly, protein production between divisions is stochastic and we show how the resulting expression noise affects epistasis. After validating the model on experimental fitness landscapes, we combine high-throughput data of budding yeast from multiple sources to make our model predictions on mutational effects on fitness as representative as possible. We find different effects per mutation type: gene duplications have little effect, as genes in our simulated pool are rarely toxic, loss-of-function mutations decrease mutational gains as adaptation progresses, and point mutations permit expression noise to unlock its roles in adaptation. For non-neutral point mutations, noise imposes a sizeable fitness penalty or even induces extinction, which is alleviated by an epigenetic, transgenerational feedback on protein dosage which is never deleterious. Particularly for essential genes, we predict that this effect reduces the obstacles of epistasis and hence significantly increases evolvability, adding to the general rules of evolution.