The repeatability of evolutionary trajectories and implications for strain optimization

Synthetic biologists often liken microbes to chassis in which engineered circuits are placed. From this perspective, the genetic background and various functions of the cell are an unfortunate necessity, and often times a hindrance to the desired product of a given genetic circuit. However, cellular fitness is important for many applications, and much can be gained from improving the background itself. One traditional and unbiased way of doing this is to simply evolve microbes in relevant environments for an extended period of time. We have done this for 96 populations of haploid S. cerevisiae, evolved for 300 generations using an array of miniature chemostat bioreactors in glucose, phosphate and sulfate-limited conditions. For this study, we performed inline measurements of fitness every 50 generations by competing the evolving population against a GFP expressing ancestral strain. We found that populations grown in both sulfate and phosphate limited conditions reached increased levels of fitness (avg. 15% increased) after just 50 generations. These populations eventually grew up to 60% more fit than the ancestral strain after 300 hundred generations of growth. However, yeast populations grown in glucose-limited media acquired significantly lower increases in fitness (~10% on avg.) after 300 generations of growth. Thus, a given background will be variably adaptable to different environments – however in some cases (sulfate and phosphate limitation for S. cerevisiae) one can obtain strains with greatly increased fitness in relatively little time. We have also performed whole genome sequencing of these 96 evolved populations. Our sequencing results show that (1) nutrient transporter genes SUL1, PHO84, and HXT6,7 are frequently amplified or otherwise mutated in sulfate, phosphate, and glucose-limited environments. Frequently recurrent large effect mutations such as nutrient transporter amplifications could be used to guide traditional targeted strain engineering approaches. (2) Hundreds of other mutations are observed in these evolution experiments having varied and unknown effects on fitness. These run the gamut from canonical nutrient sensing/signaling genes to more varied and hard to rationalize mutations across most other gene functions. The large number of mutations and continual increases in fitness throughout 300 generations of evolution in a constant environment suggest that multiple mutations seen in our evolution experiments may be adaptive. The experimental evolution approach can help us bypass our lack of understanding of the additive effect of many mutations and arrive at more fit strains in relatively little time.