Homme Hellinga set out a decade ago to exploit computational approaches to control the lock-and key fit between ligand and receptor known as molecular recognition. This relationship is the “central hallmark of all biological events at the molecular level,” according to Hellinga, an associate professor of biochemistry at Duke University. “At some stage, some molecule has to recognize another molecule to do something,” he said. Taking a computational approach, he reasoned, would lead to understanding of these biological mechanisms at the most fundamental level.
Fifteen years of research have paid off in the form of a set of algorithms capable of designing protein binding sites to fit small molecules of interest. In a paper in this week’s Nature, Hellinga and his colleagues demonstrated the effectiveness of their approach: The team showed that it could computationally redesign receptor sites for five members of the Escherichia coli periplasmic binding protein (PBP) superfamily. The reengineered proteins were shown experiment- ally to bind with trinitrotoluene (TNT), L-lactate, and serotonin in place of wild-type ligands, effectively serving as biosensors for the target molecules.
The team’s set of algorithms is akin to a docking program on steroids: Beginning with high-resolution, 3D structures, “you both dock and mutate the protein simultaneously, looking for the best possible docked ligand with the best possible surface that will bind that new docked ligand,” Hellinga explained. Because the problem is so large in combinatorial terms — with 1053 to 1076 possible choices available for a typical calculation — Hellinga and his colleagues modified a pruning method known as the dead-end elimination theorem to simplify the problem.
The result is a rapid method for calculating the optimal receptor structure, and subsequent amino acid sequence, for a binding region in a protein for a given ligand. In the research described in Nature, the design process required three days on a 20-processor Beowulf cluster. Hellinga said that the approach speeds linearly, so the team has been able to reduce the computation to a single day on its recently expanded 120-processor cluster. The next stage, genetically engineering the new gene, “in theory is no more than a week,” Hellinga said. “Then you have a new receptor.”
Hellinga said he plans to release the software publicly at some point, but right now, he warned, “it’s not straightforward to use. I can’t hand you the code and say ‘good luck.’”
In addition to packaging the algorithms in a more user-friendly format, Hellinga’s team is exploring possible applications of the method on multiple fronts. The three targets described in the Nature paper demonstrate the versatility of the approach, he said: TNT was chosen on the recommendation of the Office of Naval Research, which was seeking an effective biosensor to detect unexploded mines in seawater, while lactate and serotonin were chosen to show that the designed receptors could be useful diagnostic tools. In the case of lactate, the engineered receptors were shown to be chirally selective — to bind to only a single mirror-image form of the ligand. This characteristic should be of particular interest to the pharmaceutical industry, Hellinga said, because it could lead to a low-cost, high-throughput chiral purification process.
“The final aspect of the paper was to demonstrate that we can take these computationally designed proteins and use them not only as biosensors ex vivo, but we can put them back into E. coli where they came from, and they can drive signal transduction responses in E. coli,” Hellinga said. “Ultimately you could imagine that those bacteria could, let’s say, produce green fluorescent protein in response to TNT and then they would become biological sentinels.”
In addition to these promising application areas, Hellinga said the approach opens the door to a much broader range of possibilities for biological research. His team is now testing the method against a new set of challenges, he said, including enzyme design, protein-protein interactions, protein-DNA interactions, and construction of new pathways — a precursor to what Hellinga describes as “synthetic biology,” the control of biological systems at the molecular level.
Short term, Hellinga said his team is improving the technique for chiral purification and is also building receptors that recognize chiral forms of thalidomide, ibuprofen, and other target molecules of interest to the funding agencies who support his work: DARPA, ONR, and NIH. The team has secured enough funding to rapidly expand its efforts in a number of different research directions, Hellinga noted with obvious pleasure. “To be honest, we’re having a blast,” he said.