Computation could allow new high-affinity and specific protein receptors and sensors to be designed for any number of small molecules of interest, thanks to researchers in the US. Such artificial receptors could ultimately find a role to play in medical diagnostics, drug design, and sensors.
According to biochemist Homme Hellinga and colleagues at the Duke University Medical Center, Durham, North Carolina, the formation of complexes between proteins and ligands is a fundamental interaction in molecular biology that lies at the heart of countless biological process.
Hellinga points out that manipulating the molecular recognition between ligands and their associated proteins is crucial to basic biological studies. From a technological standpoint though, improved understanding could also allow us to create bespoke enzymes, tailor-made biosensors, genetic circuits, and to carry out chiral separations very effectively. With such rewards in the offing it is not surprising that the systematic manipulation of binding sites is still “a major challenge”, Hellinga emphasises.
The team has taken a novel approach to improving our understanding of protein-ligand interactions. They have devised a structure-based computational method that can be used to redesign protein ligand-binding specificities, which can then be engineered into a microbial genome for fermentation-like protein manufacture. In a commentary on Hellinga’s research, William DeGrado of the University of Pennsylvania School of Medicine, Philadelphia, explains how organisms use many different small molecules that bind to proteins. Receptors, enzymes, and antibodies for instance all interact with small molecules to control cell communication, signalling, and protection against pathogens. Exploitation of these interactions has so far been limited, but diagnostics and new disease therapies could emerge from greater understanding of them.
The researchers have demonstrated how the approach works by constructing new soluble receptors for the explosive TNT (trinitrotoluene), the sugar L-lactate and the medically important hormone serotonin (5-HT). The new receptors have high selectivity and affinity for their ligands and could be used as the sensing component of a detector. Intriguingly, the team has also incorporated their new proteins into a synthetic bacterial signal transduction pathway, which means they can be used to regulate the switching on and off of various genes in response to the presence of TNT or L-lactate in a bacterial culture. “The aim is to create synthetic signal transduction pathways that may allow bacteria to function as biological sentinels to chemical threats and pollutants in the environment by switching on a reporter gene,” Hellinga told us.
They started with a series of bacterial periplasmic binding proteins (PBPs) from Escherichia coli, which DeGrado describes as “Venus-flytrap-like receptors”. These PBPs are composed of two protein domains that snap shut on their ligand, just as the fly-catching plant’s specialist leaves grab their prey. When the ligand binds, a signal is transmitted. “In vivo the signal is binding of the closed form of the protein to a transmembrane receptor that triggers a cytoplasmic phosphorylation cascade that ultimately results in transcriptional activation of a reporter gene,” explains Hellinga. The natural function is the control of chemotaxis or outer membrane protein expression, depending on the system, and the natural ligands include sugars and amino acids. The researchers wanted to redesign the PBP’s trap so that it would bind a range of other small molecules in order to engineer “biological sentinels”. They chose L-lactate, serotonin (5-HT), and TNT as their targets as these compounds demonstrate great molecular diversity structurally and chemically diverse, both from one another and the natural PBP ligands.
A computer model of the PBPs was then investigated by placing a “virtual” version of TNT, 5-HT or lactate in the trap. Their powerful algorithms then mutated the binding site amino acids one at a time and scanned for new protein sequences that had a surface into which the ligand would fit. The results are astounding, with just 12 to 18 amino acids being changed, 10^23 possible sequences are generated, many more than achievable with conventional methods. Moreover, if ligand approach is also considered the combinatorial possibilities rocket to between 10^53 and 10^76.
To screen such a vast array of virtual proteins, Hellinga’s team then used another algorithm – an enhanced version of “dead-end elimination”. The original algorithm was developed by Frank DeSmet of the Catholic University of Leuven, Belgium, but was then enhanced substantially by Hellinga’s team. Further work then allowed them to handle the design of ligand-binding sites needed for their research. The algorithm queries an entry in the library on the basis of hydrogen bonds, van der Waals interactions, electrostatic interactions and atomic solvation. However, rather than scanning each individual entry those library members lower down the diversity tree are pruned off if they don’t fit. The rationale for this being that if a lower member does not fit, then any twiglets further along its branch won’t either. In this way, only the mutant Venus fly traps with a global energy minimum are retained for further investigation. The result – from billions and billions of possibilities, the researchers have pruned down to a top seventeen.
The researchers synthesised these seventeen potential receptors and tested them in vitro against their target small molecules. Fluorescence measurements shed great light on each, revealing them to be highly specific and selective for their respective ligands.
Until now, explains De Grado, the proteins in question have been “developed” either through the generation of large libraries of proteins for testing and improved through evolutionary type methods. However, this is time wasteful and energy consuming. As De Grado points out the Hellinga team has now accomplished the task of creating such a library and screening it by a very rapid computational means.
References
Nature 2003, 423, 185; Loren L. Looger, Mary A. Dwyer, James J. Smith & Homme W. Hellinga
Nature 2003, 423, 132; William F DeGrado.