Many drugs are small molecules that are able to block the active site of a protein, usually by binding to a small pocket. But for many proteins, especially signaling proteins overexpressed in cancer and other diseases, we would like to target their ability to make binding interactions with other proteins. To inhibit these so-called protein-protein interactions (PPIs), researchers have been trying to find good ways to design molecules that disrupt protein-protein binding interactions.
One way to find a good PPI disrupter is to design a mimic of the binding partner that binds better than the original. For example, so-called “stapled helix” mimics have been designed that mimic helix regions of particular protein binding partners. These mimics are small peptides that contain a hydrocarbon “staple” keeping them well structured in solution. Stapled helices bind very strongly because it takes less work for them to fold before they bind.
A stapling strategy could work well for designing beta-hairpin mimics too, but the synthetic chemistry is more difficult, and it is unclear exactly what kind of chemical “staples” would work best to use to produce a well-folded hairpin in solution.
In a new study from the Voelz Lab, we set out to see if computer simulation could help us pick the best designs. In collaboration with the Wuest Lab at Temple University, we simulated the folding of several different synthetic “stapled hairpins” designed to disrupt a PPI important in bacterial biofilm formation. Biofilms are dense aggregates of bacteria (like plaque on your teeth) that are hard for antibiotics to penetrate.
We evaluated each design using initial simulations on our computers at Temple, and then took the most promising designs to Folding@home, so we could get a very clear picture of how they are folding in solution. We found some surprising things that will help future design efforts. First, unlike helices, stapled hairpins can turn inside-out and fold all sorts of ways! This means it’s very important to have accurate models and lots of computing power to predict what shapes they assume in solution. Second, we found that even small chemical changes (like adding a –CH3 group) can have a big effect on conformational populations. In the future, it may be possible to use Markov State Models (MSMs) of potential designs to figure out which chemical changes will help them fold the best. We think this is a really exciting direction for molecular simulation, and look forward to experimentally testing a new round of designs with the Wuest Lab. Ultimately, this work may lead to new chemical probes of the proteins involved in bacterial biofilm formation, as well as new classes of antibiotics.
Our results are described in a new paper (here).