Researchers at Novo Nordisk used variations in crystal structures to identify novel inhibitors of tRNA-guanine transglycosylase (TGT) (43). A previously designed series of pyridazinedione compounds was used in traditional 2D substructure searches of the corporate database. All five compounds selected and screened for TGT inhibition were active; two showed Ks under 10hM. Attempts to dock these compounds into the high-resolution X-ray structure suggested that an important donor to the backbone carbonyl of Leu231 was absent. One of the analogs (20) was subsequently soaked into TGT crystals, and showed an unexpected backbone inversion wherein the NH of Ala232 is oriented into the binding site and the carbonyl of Leu231 is rotated out. A previously unobserved water molecule bridges the inhibitor to Ala232. The investigators devised a Unity database query that includes a donor to Leu231 or a donor/acceptor to the bridging water. Other consistently observed hydrogen bonding features were also included. Spatial tolerances for the features were derived from the results of active site mapping with the DrugScore program. Virtual screening was conducted on a combination of eight databases (totaling over 800,000 compounds), and was performed in stages of increasing complexity. Compounds with more than seven rotatable bonds and molecular weights over 450 were first eliminated. The remaining compounds were then filtered to assure the presence of the hydrogen-bonding features; the spatial requirements were subsequently introduced, and finally combined with excluded volume features to approximate the steric requirements of the active site. The resulting 856 hits comprised six chemical classes and represented the alternate binding features; 726 of the hits utilize the bridging water. Appropriate compounds were selected by inspection for docking using FlexX, followed by minimization. Only nine compounds were submitted for enzyme inhibition assays; five of these exhibited activities under 10^. The most potent of these, 21, has an inhibition constant of 250^tM. Importantly, the predicted binding modes for the inhibitors have been used to rationalize the relative activities and are being applied in an optimization process.
Exploiting multiple subsites - Iwata, et al. opted to use virtual screening to identify novel inhibitors of aldose reductase (AR) (44). The ternary complex of AR with NADPH and glucose-6-phosphate was used to calculate an active site potential energy grid, the output of which was used to search the ACD (ca. 120,000 compounds) with a proprietary program. Of 718 virtual hits, 179 were selected by inspection, and 36 were eventually purchased. Ten of the purchased compounds were considered active; three had IC50s below lO^M, and all comprise novel series. The three most active compounds are illustrated below (22,23, and 24). Details of the proposed binding modes for the three series suggests non-overlapping features that could be merged in future designs. Several analogues were synthesized based on compound 22, with SAR that is consistent with the predicted binding mode and included two analogues with IC50S of 210 and 310 nM.
Virtual screening using homology models - Bcl-2 is a key regulator of apoptosis; in their search for cell-permeable chemical probes, Wang, et al. constructed a homology model of human Bcl-2 based on the X-ray and NMR structures of the related BcI-xl (47% sequence identity), and utilized the features of the essential BakBH3 binding pocket in a virtual screen of the ACD (45). The DOCK program was applied in a rigid ligand-docking mode, and the 1000 highest scoring results were minimized in the active site. Although interaction energies were computed, the investigators relied upon manual inspection for compounds with favorable shape complementarity and hydrogen bonding features. Of 53 compounds that met the selection criteria, 28 diverse compounds were acquired for testing. A diastereomeric mixture of 25 was identified in a fluorescence polarization assay as a competitive inhibitor with an IC50 of ca. 9 i^M. Additional functional experiments demonstrated that 25 also induced cell death In HL-60 tumor cells in a dose dependent manner, with DNA fragmentation patterns characteristic of apoptotic cells. Apoptosis was shown to involve activation of caspases-9 and 3 in a pathway distinct from that of Fas/TNFr, and in a fashion dependent on Apaf-1.
Was this article helpful?