Bypassing Local Maxima in Protein Directed Evolution through Negative Selection

Directed evolution strategies often use increasing stringency of positive selection to engineer improved proteins. However, proteins frequently exhibit a high degree of sign epistasis, implying a rugged fitness landscape containing many local maxima. Positive selection is constrained by this topology since the number of accessible evolutionary pathways that improve fitness is limited. Present approaches for bypassing local maxima, such as generation of library diversity through neutral drift or genetic recombination, rely largely on chance or require a connected pathway through equally fit sequences. Instead, we suggest that traversing a fitness valley using negative selection may efficaciously provide pathways to new maxima with higher fitness. As a model system, we compared four directed evolution strategies for increasing the antibiotic resistance conferred by β-lactamase: positive selection, neutral selection, negative selection, and oscillating selection. We used a synthetic gene circuit that kills cells possessing β-lactamase activity above or below a desired threshold to perform the neutral and negative selections. All four strategies ended with positive selection steps to climb to fitness peaks. Interestingly, the strategy that included negative selection yielded more highly active variants of β-lactamase than the other three selection strategies. We performed deep sequencing on the intermediate stages of the negative selection libraries to describe and reconstruct evolutionary pathways leading to these improved mutants. The ability of negative selection to provide access to novel fitness peaks has important implications for applied directed evolution as well as the natural evolutionary mechanisms, particularly of antibiotic resistance.