Artificial allosteric protein switches with machine-learning-designed receptors

Abstract Protein allostery underlies most information and energy processing in biology and the development of artificial allosteric proteins is a key objective of synthetic biology and biotechnology. We show that machine-learning-engineered minimal ligand-binding domains act as efficient receptors in single-component allosteric switches, despite lacking global conformational change. Such colorimetric, luminescent and electrochemical biosensors of small molecules, peptides and proteins can be compiled into intramolecular YES and AND logic gates. Furthermore, we report fully synthetic allosteric switches composed of artificial receptor and reporter domains. Hydrogen/deuterium exchange mass spectrometry and 19 F nuclear magnetic resonance analyses suggest that ligand binding reduces the conformation entropy of the system, increasing the catalytic activity of the reporter domain. The potential practical utility of this approach is demonstrated by engineering Escherichia coli cells with steroid-dependent antibiotic resistance and by developing bioelectronic devices capable of quantifying steroid hormones.