Design of a senolytic proteolysis targeting chimera against the B-cell lymphoma – extra large protein using in silico approaches

Abstract

Senescent cells have been linked to age-related diseases and frailty. Their pathogenic effects are mediated by a group of cellular secretions called the senescence-associated secretory phenotype (SASP). Removal of senescent cells has been associated with the delay of the onset of age-related diseases as well as improved physical function. Senescent cells have upregulated pro-survival pathways which confer resistance to apoptosis. The anti-apoptotic BCLXL (B-cell lymphoma-extra large) protein is involved in the survival of senescent cells. Inhibition of BCLXL’s activity has been found to induce apoptosis in senescent human umbilical vein endothelial cells (HUVECs). The von Hippel-Lindau tumor suppressor protein (pVHL) is the substrate recognition subunit of an E3 ubiquitin ligase. The pVHL protein binds to substrate proteins, recruiting them to the E3 ligase for ubiquitination which will result in their degradation. A Proteolysis Targeting Chimera (PROTAC), is a bivalent molecule that promotes interaction between an arbitrary protein target and an E3 substrate recognition subunit. A PROTAC, therefore, forces the ubiquitination and subsequent degradation of a target protein. Computational methods were used to design PROTACs capable of recruiting BCLXL to the pVHL protein to trigger its degradation and therefore specifically induce apoptosis in senescent HUVECs. The techniques used included pharmacophore-based screening, structural screening, molecular docking and molecular dynamics (MD) simulation. A PROTAP (Proteolysis Targeting Peptide) targeting the same proteins was obtained from literature and recreated in silico to serve as a comparison and proof of concept. The structure of a PROTAC that interacted with pVHL and BRD4 was obtained from the PDB as an additional comparison. MD simulations were run on ligand-bound BCLXL and pVHL proteins, consequently relatively stable conformations were identified and five pharmacophore models for each protein were developed based on different structures. The BCLXL pharmacophores placed more importance on aromatic rings while the pVHL pharmacophores preferred hydrogen bond donors and acceptors. Pharmacophores were validated by decoy set validation, using BEDROC, accuracy, Güner-Henry score and area under curve as metrics, yielding barely passable results. A starting library of 2 million molecules from the Mcule database was screened, identifying 33005 and 221481 hits for BCLXL and pVHL respectively. Structural screening and molecular docking were used to search for molecules with good binding characteristics to BCLXL and pVHL, retaining 33 and 216 molecules respectively. Two potential BCLXL ligands and two potential pVHL ligand were joined in different combinations with different glycine linker lengths, yielding eight potential PROTACs. Ternary protein-PROTAC complexes were assembled and used in 100 ns long MD simulations. From the MD trajectories, the RMSD of the proteins and PROTACs, as well as the number of bonds/interactions between them, were examined. Comparisons were made to the RMSD and interactions of the reference PROTAC and PROTAP. Three attractive PROTACs were identified, with one having superior ADMET properties. The binding energy calculations did not find fault with any of the three candidate PROTACs. While there is currently no satisfactory way to evaluate the performance of a PROTAC, the procedures employed identified molecules with the potential to act as a PROTAC.