Section: New Results

Evolution of phenotypic noise

Participants: Charles Rocabert, Guillaume Beslon, Carole Knibbe

The phenotype of an organism is a complex non-linear cascade of developmental, physiological and regulatory processes, formalized by the concept of genotype-to-phenotype map. An increasing number of experimental studies demonstrate the existence of phenotypic noise, which can be finely tuned by the genotype-to-phenotype map, and that phenotypic noise can be adaptive.

In stabilizing selection, when the population is at a fitness optimum, phenotypic noise is deleterious and minimized by evolution. Nevertheless, phenotypic noise can be positively selected when the population is exposed to stressful conditions. It was thus suggested that during an adaptation event, phenotypic noise would increase in directional evolution, and then be reduced when the selection becomes stabilizing. In 1930, R.A. Fisher suggested with its so-called Fisher's Geometric Model (FGM) that organisms adapting to a new environment experience a “cost of complexity”, where beneficial mutations become increasingly harder to fix when the number of phenotypic characters increases. Predictions made on the evolution of phenotypic noise are mostly based on single trait observations. Is there also a cost of complexity on the phenotypic noise?

To address this question, we extended the FGM by adding an evolvable phenotypic noise. First, using a simple form of noise, affecting similarly every phenotypic character, we show that a cost of complexity indeed makes phenotypic noise deleterious in directional evolution. Second, we extended the FGM with a fully evolvable noise, allowing evolution on noise amplitudes on each character, as well as on noise correlations between characters. In directional evolution, we show that phenotypic noise evolves towards a flattened shape, with elevated noise in the direction of the optimum, and minimized noise in all other directions. In this case, the noise becomes advantageous again, even with many characters. Non-isotropic phenotypic noise thus facilitates evolution towards the fitness optimum, and significantly reduces the cost of complexity. Our results show that such non-isotropic phenotypic noise could be exploited by evolution, and suggest further experiments to assess the functional nature of phenotypic noise.

This result is currently under review for the Evolution journal. It is the result of an enriching collaboration between the Beagle team (Charles Rocabert, Guillaume Beslon, Carole Knibbe), and the Dracula team (Samuel Bernard). Although the results are grounded in theory and mathematical modeling, they provide stringent conditions for noise to be beneficial, which are experimentally testable. We believe the results to be of wide interest for researchers working on phenotypic evolution. By deciphering the conditions in which phenotypic noise evolves towards specific patterns, our work may also contribute to a better understanding of drug resistance and cancer cells proliferation, and also to the growing field of predictive biology.