Computational design of small molecule biosensors

Optical biosensors can allow us to visualize the spatial, and temporal, distribution of various small molecules in plants, animals and microbes. However, the construction of effective sensors is not trivial. In this work we have used computational design to automate the selection of sites at which to attach fluorescent dyes to allow the straightforward conversion of solute binding proteins into biosensors.

HFSP Young Investigator Grant holders Colin Jackson, Harald Janovjak and Christian Henneberger and colleagues
authored on Tue, 07 February 2017

Optical biosensors that exploit Förster resonance energy transfer (FRET) between two fluorophores allow small molecule analytes within physiological environments to be detected and quantified with excellent spatiotemporal precision. However, such sensors require a solute binding protein that undergoes some conformational change upon ligand binding, and the ability to identify sites at which to attach fluorophores that will result in a detectable change in FRET. The identification of sites for fluorescent labelling is not easy and often involves trial and error or intuition.

Figure: The transition between open (purple) and closed (blue) forms of maltose binding proteins. The sampled positions of the fused fluorescent protein are shown in cyan, with an estimated average position shown in red. This sampling was used to predict the optimum position to label the binding protein with a fluorescent dye to convert it to a FRET sensor. 

In this study, we describe a method that greatly simplifies the design of fluorescent biosensors and create sensors for maltose, arginine, and sialic acid to demonstrate its utility. The process involves a single round of computational screening with no experimental optimization, making use of crystal structures, or homology models, of the solute binding protein of interest. Estimated distances between the expected position of a fused fluorescent protein, and every surface-exposed amino acid (for labelling with a fluorescent dye) are calculated for both the bound and unbound states. Sites that result in the largest expected change in FRET efficiency upon ligand binding are selected for labelling.

This process resulted in correct predictions for all sites, and led to the construction of a ratiometric sensor for maltose binding protein with a dynamic range ~5-fold greater than other maltose sensors, as well as the development of first reported sialic acid sensor, which will now be available for the detection and study of sialic acid, which has been implicated in the early human neurodevelopment. The algorithm, which we call “Rangefinder”, runs in seconds on a modern personal computer and is freely available to the scientific community. It is designed for use with proteins of the solute binding protein superfamily, which includes binding proteins for hundreds of different small molecules, but its simplicity will allow it to be adapted for virtually any structural fold that undergoes a conformational change on ligand binding.

Reference

Rangefinder: A Semisynthetic FRET Sensor Design Algorithm. Joshua A. Mitchell, Jason H. Whitfield, William H. Zhang, Christian Henneberger, Harald Janovjak, Megan L. O’Mara, Colin J. Jackson. ACS Sens., 2016, 1 (11), pp 1286–1290. DOI: 10.1021/acssensors.6b00576.

Link to article