Computer-designed ricin toxin inhibitors
November 2, 2018 § Leave a comment
To date, there is no known antidote for ricin poisoning.
Ricin is a cytotoxic protein extracted from mashed castor beans during castor oil production. It was tested in the 1940s as a chemical warfare agent by the United States military and is believed to have been used in the 1980s by some terrorist organizations. In 2014, a student at Georgetown University was arrested for manufacturing the poison in his dorm room.
Gerd Rocha, an associate professor at the Federal University of Paraiba in Brazil, is attempting to identify novel ricin antidotes. Rocha and colleagues are employing computational chemistry techniques to measure small-molecule binding affinities for ricin toxin. Their findings are published in the Journal of Chemical Information and Modeling.
Ricin consists of two subunits, ricin toxin A (RTA) and ricin toxin B (RTB), connected by a disulfide bond. RTB causes the protein to be absorbed into the endoplasmic reticulum of a cell, where an isomerase cleaves the disulfide bond, releasing RTA, the subunit responsible for toxicity. Once inside, RTA binds ribosomal RNA and depurinates adenine from the ribosome’s GAGA motif at the sarcin-ricin loop. Ultimately, this prevents protein synthesis and causes cell death.
RTA’s active site is a challenging target for small molecule drugs because of its large cavity and polar characteristics, which cause it to form favorable hydrogen bonding networks with water. Drug binding disrupts these networks, resulting in unfavorable energetic penalties.
“Ricin intoxication involves extra- and intracellular events, so competitive inhibitors for the RTA active site is only one of the strategies for inhibiting this potent toxin,” Rocha said. “However, with the increase of the screening libraries, such as ZINC15, which presents approximately 1 billion compounds, searching for new scaffolds for the RTA active site can still be a promising alternative, which needs to be exhaustively evaluated.”
Computational chemists use databases, like ZINC15, to explore structure-activity relationships between chemical scaffolds and protein targets through virtual screening. Virtual screening is like traditional high-throughput screening except that a computer model predicts activity instead of a binding assay. This approach allows scientists to rapidly identify interesting scaffolds that are then prioritized for experimental testing. Virtual screening’s goal is to increase efficiency of high-throughput screening for finding active compounds.
But before they can virtually screen databases like ZINC15 for new ricin antidote scaffolds, chemists have to build models capable of distinguishing ricin binders from non-binders. Rocha and colleagues designed a two-tiered model combining molecular docking and steered molecular dynamics to rank a small set of RTA inhibitors.
Molecular docking measures the interactions between a flexible ligand and rigid protein. These interactions include hydrogen bonding, hydrophobic contacts, pi-pi stacking between aromatic groups, and penalization terms (for solvent exposure or conformational strain) and are reported as a “docking score.” This score represents a molecule’s predicted binding affinity. This study used the free docking program, AutoDock Vina.
“We attempted, as a method validation, to use the most potent (ricin) inhibitors previously described and available in the Protein Data Bank,” Rocha said. “Unfortunately, the low availability of structures of RTA-ligand complexes in the PDB is a limiting feature of our study.”
Rocha and colleagues identified six pterin scaffold ricin inhibitors. Three inhibitors were considered active with binding affinities below 70 µM; the remaining inhibitors were inactive.
“In our results, we verified that Vina score function can efficiently predict the bioactive poses of the ligands in the RTA active site,” Rocha said. However, he said, the Vina score function alone did not correctly rank ligands by their ricin binding affinity. This can generate false-positive results in a large-scale virtual screening study.
False-positives are detrimental to any virtual screening analysis and occur when an inactive compound is predicted active. Computational chemists can minimize false-positives by modifying scoring functions, a task easier said than done, or including a molecular docking post-processing technique.
Where molecular docking is considered a “fast” approach, post-processing approaches are not. These techniques employ more sophisticated functions to measure ligand binding affinities and often include “flexible” techniques that incorporate protein movement. This study used steered molecular dynamics (SMD) for post-processing.
Molecular dynamics is a time-dependent calculation enabling researchers to visualize protein and ligand interactions. However, the time scale for these calculations is too small to replicate all biological activities, like ligand binding or unbinding. Steered molecular dynamics dictates simulations toward desired biological events by applying an external force. This external force is like two tug-of-war teams. The protein is attempting to “hold” the ligand in place, while the computer model tries to pull the pterin inhibitor from the RTA binding pocket. The strength, or force, needed to separate the ligand and protein is the ligand binding affinity.
Using SMD, the researchers ranked the six ligands in terms of binding affinity and showcased an important post-docking processing tool to prioritize ligands for experimental testing. However, the model does not guarantee a ligand will have druggable characteristics in organisms, and SMD should not be used as a stand-alone screening approach for large libraries of compounds. Instead, Rocha suggests adapting a two-tiered approach that combines molecular docking and SMD.
Rocha plans to expand the model for a large-scale virtual screening study and experimentally validating the top compounds. And maybe his model will help expand the field of known small-molecule ricin binders, ultimately giving way to an inhibitor.
George Van Den Driessche (firstname.lastname@example.org) is a graduate research assistant in the Fourches lab at North Carolina State University.