Widespread Compensatory Evolution Conserves DNA-Encoded Nucleosome Organization in Yeast
We simulated the evolution of a fixed size population of small (20 bp) “genomes” in simple fitness landscapes that depend only on the G+C content of the sequence (Methods). We used a mutational input that favors A/T over G/C, resulting in a neutral stationary G+C content of 30%. A) G+C goal fitness landscape. Shown are fitness landscapes (fitness value as a function of the G+C content) preferring low G+C content (the “low occupancy” regime). We generated regimes with different selection intensities (denoted by η) by changing the slope of the depicted parabola (shown are the landscapes for two different intensities). B) Substitution rates. Shown are substitution rates of A/T loss (red) and A/T gain (blue) measured in populations evolving under different levels of selection intensities (X axis) in the low G+C content fitness landscape. We note the increase in A/T gain rate to values higher than neutral at intermediate selection intensities. C) Stationary G+C content. Shown are the population average G+C contents for population evolving in different selection intensities (X-axis), showing that at intermediate levels of selection where A/T gain rate were elevated (dashed line), the average G/C content is only slightly higher than the optimum. D) Substitution rates for intermediate selection intensity. We summarize the simulation by showing substitution rates for a specific level of selection intensity (marked in dashed lines in B and C). Data is shown for the low G+C fitness landscape and for a high G+C fitness landscape that was defined symmetrically with a preferred G+C content of 40% (Fig S8). The data are generally compatible with our empirical observations on yeast divergence rates (Fig 2C, 4C). E–H) Evolutionary dynamics for a threshold fitness landscape. An analysis similar to the above, but with a threshold-function fitness landscape (E) reveals that an increase in A/T gaining substitution rates can be observed over a wide range of fitness intensities. I) Compensatory evolution explains the increase in A/T gaining substitution rates. According to our model, the evolution of low occupancy sequences is attempting to maintain low G+C content in spite of a flux of slightly deleterious mutations that pushes toward a higher stationary G+C content. Selection may not be sufficiently powerful to purge every deviation from the optimum and mutations that decrease A/T content may persist (even partially) in the population. These mutations trigger adaptive evolution of corrective mutations, which is efficient since it can occur at multiple positions. The schematic shown here assumes evolution in the threshold fitness landscape (E–H), in which A/T gaining substitutions are never deleterious and therefore robustly increased in rate even if selection is intensive. A variation of the same argument shows why A/T gain rate increases in the fitness landscape of A–D.