A Modular Platform for Single-Fluorescent Protein Biosensors

We demonstrate a method for systematic construction of intensity-based biosensors from circularly permuted fluorescent proteins (cpFP). The method entails analysis of the structure (or model) of a periplasmic binding protein (PBP) to identify points in the amino acid sequence into which a cpFP can be inserted in such a way that the ligand-induced conformational change of the PBP results in a change in the local environment of the FP chromophore. The ideal composition of the residues that comprise the environment of the chromophore is not obvious, and thus libraries of variants at residues near the chromophore (the PBP-cpFP & cpFP-PBP linkers, as well as nearby positions) are screened to identify those variants that have significant changes in fluorescence emission intensity upon saturation with analyte. We have demonstrated that both the ligand-binding properties and the sensor color can be changed, largely independently, by mutation of the residues comprising the binding site or the chromophore, respectively.

We demonstrated the in vivo utility of the engineered sensors to visualize maltose transport across the plasma membrane of individual E. coli bacteria. Further experiments are necessary to determine whether the sensors will work similarly well in mammals at body temperature. All four colors of engineered sensors were sufficiently bright under both 1-photon and 2-photon excitation, and showed ample ∆F/F and SNR under both conditions, to permit preliminary experiments in eukaryotic cells or intact animals. Intriguingly, the performance of some sensors was significantly better under multi-photon excitation.
The bacterial periplasmic binding protein superfamily contains proteins specific for glutamate, glutamine, GABA, glycine, acetylcholine, diverse carbohydrates, peptides, lactate, and Zn2+, among numerous other analytes. Further diversification of the ligand-binding properties of the family by rational design or directed evolution may provide additional specificities (e.g. histamine from histidine-binding protein). Sensors for each of these target molecules would be of great utility in visualizing neurotransmission, cellular metabolism, and disease states in cells and intact organisms. It is likely that the methods developed here for detection of maltose will be generally applicable to other family members, and to other classes of allosteric proteins as well. Fluorescent-protein family diversity may provide for the creation of red-shifted sensors, facilitating deep imaging in biological tissues.

We found it difficult to predict a priori the precise FP insertion sites, linker sequences, and surrounding protein mutations that would give rise to optimal sensors. However, we were able to rank-order potential insertion sites based on a simple geometric estimate and found that highly-ranked positions gave rise to high-SNR sensors, after optimization. A combination of rational design and random screening produced the best results -- the precise effect of mutations on fluorescence readout is dificult to predict, as fluorescence is particularly sensitive to small perturbations. Mutations to the binding and fluorescent proteins were found to be largely independent (i.e. chromophore changes induced the expected colors; grafting literature mutations on the binding pocket conveyed the anticipated binding specificity). However, the precise nature of the binding and fluorescence properties produced a number of unanticipated results (i.e. allosteric hinge mutations did not graft easily; binding pocket mutations produced sensors with lower-than-expected affinity), illustrating that each sensor will likely have to be specifically optimized by rational design and directed evolution.