Identification and targeting of a SOX2 and SOX9 degradation pathway in glioblastoma

Abstract

Glioblastoma multiforme (GBM) is the most malignant adult brain tumour. A subpopulation of cells, termed glioblastoma stem cells (GSCs), possesses similar phenotypic characteristics to normal neural stem cells (NSCs) and may drive tumorigenesis. GSCs frequently express many essential neurodevelopmental transcription factors (TFs) at high levels. Of particular note are SOX2 and SOX9 genes, which are master regulatory TFs in the forebrain development with reprogramming activity that are essential for GSC self-renewal. Here we explored a set of newly identified small-molecule compounds that was shortlisted following phenotypic screening for loss of SOX2 in GSCs. We showed that a subset of these hits operates via the ubiquitin-proteasome system (UPS), revealing a potentially new pathway controlling SOX2 and SOX9 protein turnover. We demonstrated that these compounds likely act through a class of enzymes called deubiquitylases (DUBs).

The lead compound 1035 was validated across a range of different patient models using two orthogonal reporter lines. We also confirmed that the endogenous protein was degraded via a proteasomal mediated pathway. We explored the selectivity of effects on cell viability and proliferation using an independent panel of non-GBM cell lines, including human umbilical vein endothelial cells, fibroblasts, non-tumour pericyte-like cells as well as human pluripotent stem cells (hPSCs)-derived neuroepithelial. The compound appeared selective, yet had some degree of off-target toxicities at higher doses. However, we did notice a strikingly significant positive correlation between the effect on GSC proliferation and SOX2/SOX9 expression levels. USP36, USP39 and USP42 were determined as potential DUB targets to which 1035 might bind, based on their restricted subcellular localisation to the nuclear/nucleolar compartments, in addition to lower SOX2/SOX9 abundance after individual knockdown. Concurrent deletion of these USPs was also able to phenocopy the compound effects, with diminished cell viability as well as proliferative arrest, alongside a robust SOX2 and SOX9 protein decline.

Next, four structural analogues of 1035 were significantly less effective in overall at reducing endogenous SOX2 and SOX9, implying that the presence of a chiral carbon atom or/and methyl group at the asymmetrical carbon position is/are absolutely critical to retain compound activity. Individually purified R or S enantiomer was unexpectedly less potent than the original 1035, but this was restored in a racemic mixture, consistent with both enantiomers possibly operating together with a concerted synergism. The right-handed stereoisomer R mainly facilitated SOX2 and SOX9 protein degradation via USP42 (and maybe USP36 to a limited extent); meanwhile the left-handed stereoisomer S gave rise to broad cytotoxicity via nucleolar disruption pathway in a non-cell-selective manner. The structure-activity relationship can then be applied for future medicinal chemistry optimisation to improve lead compound potency and on-target specificity, whilst minimising the non-selective killing impact on normal human cells.