Virtual Screens for Lead Compounds against Bacterial Toxins

Virtual screens can yield a list of potential “hits” against a drug target (receptor) and the approach has finally evolved where it can provide valuable contributions in "hit-and lead-compound" discovery with a number of recent successes. Our group is now routinely conducting virtual screens using OpenEye software http://www.openeye.net/products/software/ on SHARCNET. Our first virtual screen effort produced 72 hits with 31 compounds selected for experimental testing, resulting in several good mARTT inhibitors. Subsequently, we conducted virtual screens against other mARTT family subclasses including iota toxin (1GIQ:C2-like), C3bot1 toxin (2C8A:C3-like) and cholera toxin (2A5F:CT-PT) (unpubl. results).  We will be testing all these new hits using our protocols according to our recent work.  We are finding that there is “cross-reactivity” in our inhibitor libraries where some of the compounds screened against one mARTT subclass also work well against another subclass.  We anticipate that some of our compounds will function as broad-spectrum antivirulence therapeutics, but it will require SARS/QSARS iterations to second- and third-generation compounds. On the other hand, we are also now realizing (largely based upon our computational MD work) that there are important subsite differences within each mARTT subclass that can be exploited to provide descriminating specificity for inhibition.

Prior to attaining crystallography data for Vis toxin, we conducted a virtual screen against a high-resolution structure of a representative mART family member, iota toxin, in complex with NADH (PDB code: 1GIQ) to identify a set of compounds to use in tests against C2/C3-like mART toxins. This screening was conducted using tools from both the Schrödinger software suite (www.schrodinger.com; Friesner, 2004) and the OpenEye software suite (www.eyesopen.com). Briefly, NAD+ was used to define the iota toxin active site and all the waters were removed. The structure was processed using the Schrödinger Protein Preparation Wizard with default settings. All Ser, Thr and Tyr hydroxyl groups pointing into the active site were treated as rotatable. Also, two H-bond constraints were set for R296. The starting library was the ZINC Drugs which is now a set of ~6.5 million drug-like compounds (zinc.docking.org/subsets/drugs-now). High-throughput virtual screening was conducted using Schrödinger Glide and we kept only the top 1% of the docked compounds for the next step. Then, standard precision docking was conducted using Schrödinger Glide and, again, we kept only the top 1% of the docked compounds (294 in total). These compounds were filtered again and refined to select a set for in vitro and cell-based testing. First, 11 compounds were selected on the basis of the Schrödinger r_i_docking_score alone, 5 compounds were selected on the basis of passing the OpenEye BlockBuster filter and 9 compounds were selected based on an additional round of focused screening using a previously described procedure. Notably, several of the final 25 compounds also scored well in confirmatory screens against the Vis-NAD+ structure, conducted upon completion of the structure presented herein.