Development of novel inhibitors for Aldo-Keto Reductase 1C3 (AKR1C3)

Abstract

Breast and prostate cancers are the second leading cause of cancer-related death around the world. Although there are various risk factors associated with induction of these cancers, one of the causes is believed to be overexpression of hormones such as estrogen. Aldo-keto reductase 1C3 (AKR1C3) is a human enzyme that is involved in trans-activation of androgen and estrogen receptors. It catalyses NADPH-dependent reduction of carbonyl moieties on substrates such as 4-androstene-3,17-dione forming testosterone (a potent androgen), and on estrone forming 17β-estradiol (a potent estrogen). AKR1C3 also catalyses prostaglandin biosynthesis which leads to proliferative cellular signalling and tumour growth. Overexpression of AKR1C3 protein has been detected in tissues of breast and prostate cancer patients suggesting that AKR1C3 is a source of proliferative signals in hormone-dependent and hormone-independent breast and prostate cancer. Therefore this research was on the discovery and development of novel inhibitors of AKR1C3 with unique and diverse structures as potential anti-cancer therapeutics. Virtual screening was conducted to discover potential inhibitors from a 78,000 compound library. Twelve compounds were studied further by differential scanning fluorimetry as an initial screen for binding, followed by spectrophotometric and fluorometric inhibition assays. Another part of this project was to study how virtual screening of a fragment library correlates with the experimental fragment screening data of 492 fragments from the MayBridge Ro3 500 Library; differential scanning fluorimetry of all 492 compounds as experimental data was compared with computational screening data performed by the GOLD program and using the Goldscore fitness function. The Goldscore fitness function was shown to produce false positives and poor correlation was observed between the rankings of fragments on virtual screening and the experimental data, indicating the limitations of our virtual screening experiments to accurately predict potential ligands of AKR1C3. A final part of this project was to find a reliable and accurate docking method to allow us to predict the true binding modes of drug-like compound. Eighteen AKR1C3 ligands and four scoring functions were used in docking and compared to the previously determined crystal structures. Chemscore and ChemPLP scoring functions were shown to predict the true binding pose more accurately than other docking scoring functions. Keywords: AKR1C3, breast cancer, prostate cancer, anticancer drug, overexpression of AKR1C3, crystallisation, docking, virtual screening, inhibition