Unravelling the features that give rise to specific and high-affinity binding of ligands by proteins is crucial for both advancing the basic understanding of molecular biological interactions, and for a number of applications in biotechnology and medicine. Reporting in the 9 May issue of Nature, Looger et al. describe a new computational approach for designing receptor and sensor proteins with specificity and affinity for desired small molecules.

Looger et al. started with a series of bacterial periplasmic binding proteins (PBPs) — Venus-flytrap-like receptors comprising two domains that snap shut on binding their ligand. This ligand binding transmits a signal through the receptor that activates a pathway resulting in changes in bacterial gene expression. The researchers set out to redesign the binding site of these PBPs so that they would specifically bind a range of small molecules: the explosive trinitrotoluene (TNT), the sugar L-lactate and the hormone serotonin (5-HT). These ligands are structurally and chemically diverse, both from one another and the natural ligands of the PBPs.

The first step in the design process was to use an algorithm to identify the PBP binding site, and to place, in a computer model, a 'virtual' ligand in the binding site. Then, amino acids making up the binding site were sequentially mutated to find sequences that provided complementary surfaces to the desired ligand. This creates 1023 possible sequences, a number much greater than can be screend in vitro. In addition, the orientation of the ligand, and the conformations that the amino-acid side-chains could potentially take, were factored in, thereby creating further diversity. In order to search through these truly vast numbers of proteins, Looger et al. used an enhanced version of a search algorithm, which identified 17 virtually designed receptors that were selected and synthesized for experimental testing.

The designed PBPs were put through their paces in two experimental set-ups. In the first, the engineered PBPs were attached to a dye molecule that fluoresced on ligand binding by the PBP. By measuring levels of fluorescence, novel PBPs that bound TNT, L-lactate and 5-HT with affinities up to those typical of antibody–ligand interactions were detected. In the other set of experiments, the selected PBPs were inserted into Escherichia coli such that the receptors were incorporated into a synthetic signalling pathway leading to the production of a reporter enzyme.

This approach is an important advance because of its generality, and the scope it provides for redesigning protein binding sites for ligands very dissimilar to the natural ligand of the protein. This was achieved in this study by taking into account both the orientation of the ligand and the myriad conformations that amino-acid side-chains within the binding site can adopt. Looking ahead, this method could have a wide range of applications, from improving chiral separations to developing molecular sensors for the presence of small-molecule biomarkers in clinical diagnostics, and perhaps ultimately in the design of catalytic proteins.