Light-directed evolution of dynamic, multi-state, and computational protein functionalities

Evolving dynamic, multi-state, and computational protein functionalities is challenging because it requires selection pressure on all the states of a protein of interest (POI) and the transitions between them. To create a continuous directed evolution paradigm for such properties, we genetically engineered budding yeast for optogenetic input to switch a POI "on" and "off," which, in turn, controls a Cdk1 cyclin that is essential for one cell-cycle stage but detrimental for another. The method, "optovolution," generates dynamic selection pressure on POI cycling at the timescale of tens of minutes. We used it to evolve 19 new variants of the LOV transcription factor El222, including in vivo green-light-responsive variants allowing LOV color-multiplexing. Evolving the PhyB-Pif3 optogenetic system, we discovered that loss of YOR1 makes supplementing phycocyanobilin (PCB) unnecessary. Finally, we demonstrated the generality of the method by evolving a non-light-responsive AND gate (PEST-rtTA). Optovolution makes difficult-to-engineer protein functionalities continuously evolvable.