Computationally Guided Rational Design Of Loo-Gfp Biosensors And Biosensor Materials

Split fluorescent proteins have been engineered for various purposes, in each case signaling their spontaneous reconstitution by fluorescence. By combining split protein reconstitution and computational protein design, we have constructed a circularly permuted and truncated variant of green fluorescent protein (GFP) in which the seventh beta strand has been left out and the sites around it computationally designed to accommodate a peptide from influenza hemagglutinin. We call this a “leave-one-out” GFP biosensor (LOO-GFP). A LOO-GFP was designed using DEEdesign, selected by plate screening a bacterial library in the presence of the influenza peptide target, and was found to have seven point mutations. But binding was weak (9μM) and was at the expense of stability. The weakened, partially folded protein aggregated in the absence of its target. Furthermore, the aggregated biosensor fluoresced more than its monomeric peptide-bound form. In this work, the LOO-GFP was rationally redesigned to fold more robustly and bind the target tighter. Modeling of the GFP folding pathway suggested that one of the seven mutations, F83W, interfered with the closing of the beta barrel. Mutating this residue back to a F indeed, along with several other rationally justified changes followed by re-screening, produced several biosensor sequences with slower unfolding rates, a positive binding signal, and higher chromophore maturation efficiency. We also observed a blue-shift in the excitation spectrum, and lower background fluorescence in the unbound state. Kd was unchanged. In parallel experiments, LOO-GFP biosensors were genetically fused to fibers formed by the Drosophila protein ultrabithorax (Ubx), and were found to be absent any background fluorescence in the unbound state, but recovered fluorescence when exposed to the target peptide. Implications for the design of biosensing materials are discussed.