Characterising ligand binding sites in PI3Kα using a molecular docking-based fragment hotspot detection method

Abstract

Phosphatidylinositol-3-kinases (PI3Ks) are a family of proteins that phosphorylate the inositol head group on membrane embedded phospholipids. Phosphoinositide kinases have an essential role in many cell-signaling pathways. The work in this thesis focuses on the PI3Kα isoform, responsible for the conversion of phosphatidylinositol-4,5-bisphosphate to phosphatidylinositol- 3,4,5-trisphosphate in response to cell surface receptor tyrosine kinase (RTK) activation. PI3Kα is mutated in many cancers and drug discovery programmes have identified many compounds that bind to its ATP binding site, as well as to sites outside of this. Unpublished data from the Flanagan lab showed that the ATP site blocker GSK2126458 was also able to prevent the PI3Kα enzyme from binding lipid bilayers, indicating potential for a dual mechanism of action. The aim of this thesis is to model inhibitor binding to a new, non- ATP binding site. Chapter 3 describes the adaption of a literature fragment-based hotspot mapping method to use molecular-docking. The MolDock workflow was tested against both the ATP-binding site and alternate drug-like and fragment-like binding sites known in PI3Kα. It was able to predict some of the interactions used by the drug-like compounds, identifying them as possible hotspots for the drug binding. The PI3Kα fragment compounds did not use the predicted interactions, but the method was able to predict some fragment-like interactions using the Ras protein. Chapter 4 describes PI3Ka labelling experiments that confirm the presence of second non-ATP binding site able to bind reactive analogues of GSK2126458 and uses the rare oncogenic PI3Kα mutant G106V. Protection experiments trialled to explore this possibility of GSK2126458 binding to the same site, but these data were inconclusive. Chapter 5 describes application of the MolDock workflow to the identification of possible drug binding hotspots in the new binding site, and the use of ligand-specific fragment sets to find the points that might be accessible to compounds characterized in Chapter 4, and in the selection of possible binding models. This resulted in a binding mode for both SN38762 and GSK2126458 that can be used in future studies.