Rational design and preliminary validation of BRD9 receptor antagonists based on the BI-7273 scaffold

Abstract

Bromodomain protein 9 (BRD9) has been identified as being overexpressed in Acute Myeloid Leukaemia (AML) blasts as compared to normal progenitors. Del Gaudio et al. established that depletion of BRD9 receptors induces myeloid cell apoptosis through Capase8 activation. Thus, inhibition of BRD9 receptors can be regarded as a promising therapeutic alternative in the management of AML. BI-7273 was identified as a novel lead-like molecule for AML via its antagonistic effect on the target, BRD9 receptor. Its efficiency was investigated by Martin et al. showing decreased proliferation of murine AML RN2 cell line following treatment with BI-7273. This study aimed to use the BI7273 scaffold to model the critical interactions to probe the BRD9 receptor’s ligand binding pocket and, identify superior high affinity lead-like molecules. Protein Data Bank crystallographic deposition 5EU1 describing the bound co-ordinates of BRD9 receptor and the chemical probe BI-7273 was used as a template for this study. Two different computational techniques were employed, the virtual screening and de novo design. This allowed for the generation of novel BRD9 receptor antagonists through different but complementary methodologies. All molecular modelling throughout the virtual screening approach was performed in SYBYL-X® v1.1 and LigandScout® software. The chemical probe BI-7273 was extracted from its cognate receptor and a pharmacophore was created for high-throughput screening in chemical libraries. Hit molecules were then docked into a generated protomol and affinities were calculated. The de novo approach involved the rational design of seed structures based on recognition of favourable and unfavourable interactions of BI-7273 within the BRD9 receptor’s ligand binding pocket. These novel user-driven pharmacophores created, retained the favourable moieties for binding at the BRD9 receptor’s ligand binding pocket. LigBuilder® v1.2 was used to generate novel molecules from these pharmacophores through ‘GROW’ and ‘LINK’ algorithms. Novel molecules generated from both virtual screening approach and de novo design were ranked in order of affinity. A total of 364 novel structures were identified of which 298 originated through the virtual screening and 66 through the de novo approach. All the de novo generated molecules exhibited a higher ligand binding affinity (> 5.88) than the original lead molecule.