Revolutionising the design and analysis of protein engineering experiments using fractional factorial design

Protein engineering is one of the foundations of biotechnology, used to increase protein stability, re-assign the catalytic properties of enzymes or increase the interaction affinity between antibody and target. To date, strategies for protein engineering have focussed on systematic, random or computational methods for introducing new mutations. Here, we introduce the statistical approach of fractional factorial design as a convenient and powerful tool for the design and analysis of protein mutations, allowing sampling of a large mutational space whilst minimising the tests to be done. Our test case is the integral membrane protein, Acridine resistance subunit B (AcrB), part of the AcrAB-TolC multi-protein complex, a multi-drug efflux pump of Gram-negative bacteria. E. coli AcrB is naturally histidine-rich, meaning that it is a common contaminant in the purification of recombinantly expressed, histidine-tagged membrane proteins. Coupled with the ability of AcrB to crystallise from picogram quantities causing false positives in 2-D and 3-D crystallisation screening, AcrB contamination represents a significant hindrance to the determination of new membrane protein structures. Here, we demonstrate the use of fractional factorial design for protein engineering, identifying the most important residues involved in the interaction between AcrB and nickel resin. We demonstrate that a combination of spatially close, but sequentially distant histidine residues are important for nickel binding, which were different from those predicted a priori. Fractional factorial methodology has the ability to decrease the time and material costs associated with protein engineering whilst expanding the depth of mutational space explored; a revolutionary concept.\n\nSignificance statementProtein engineering is important for the production of enzymes for bio-manufacturing, stabilised protein for research and production of therapeutic antibodies against human diseases. Here, we introduce a statistical method that can reduce the time and cost required to perform protein engineering. We validate our approach experimentally using the multi-drug efflux pump AcrB, a target for understanding drug-resistance in pathogenic bacteria, but also a persistent contaminant in the purification of membrane proteins from E. coli. This provides a general method for increasing the efficiency of protein engineering.