Phenotypic screening and AI uncovers modulators of YAP and TAZ activity in glioblastoma

Glioblastoma (GBM) is the most common and advanced primary malignant brain tumour accounting for almost 50% of all primary tumour diagnoses; this is also the most lethal brain tumour with an average survival between 14-16 months. The Stupp protocol, representing the standard of care, was introduced in 2005 and consists of surgery by maximal resection, radiotherapy, and concomitant chemotherapy (Temozolomide), but patients still show rapid recurrence of resistant tumours leading to patient fatality with less than 5% surviving beyond 5 years. Treatment and development of therapeutics for GBM is hampered by a highly complex disease that displays vast amounts of inter and intra-patient heterogeneity. This is enhanced by phenotypic plasticity of tumour cells which allows them to differentiate and regress into multiple glial-like cell types with neoplastic traits and enhanced stem cell characteristics contributing to radio- and chemo-resistance. These stem-like cells can propagate recurrent tumours representing parental tumours from single-cells and are a vital target in the development of new therapeutics for GBM. The main Hippo pathway transcriptional coactivators YAP and TAZ are master regulators of up to 70% of the entire glioma stem cell gene signature and so represent key molecular targets in the development of new therapeutic strategies for GBM. However, targeting these transcription cofactors in the context of GBM represents a significant challenge due to the complexity of targeting CNS disorders – which have the lowest success rate of drug approval - partly due to the highly selective blood-brain-barrier impeding efficient drug delivery – and the fact that YAP and TAZ are small disorganised transcriptional coactivators with no bona fide binding pockets for small molecules. Phenotypic drugdiscovery strategies are undergoing a renaissance which is being driven by the constant evolution of cell-based assay technologies including automated high-content imaging which aims to identify subtle phenotypes and obtain detailed mechanistic information. In this thesis, I initially investigated the regulation of the Hippo pathway mediators such as YAP/TAZ at the post-translational level in publicly available clinical datasets of GBM patient samples. Following confirmation of YAP/TAZ pathway activity in GBM tumours and patient-derived GBM stem cell cultures, I established an image-based high-content phenotypic screening assay and bespoke machine learning assisted image analysis pipeline to automatically quantify YAP and TAZ nuclear localisation and cell survival across genetically distinct patient-derived GBM stem cell lines. I subsequently performed a screen of 3080 compounds that represented a diverse panel of approved and preclinical stage compounds. Using a bespoke CellProfiler image analysis pipeline and a combination of Hit Selection using a Neural Network classifier (StratoMiner HC), 32 hit compounds were prioritised for testing in secondary screening assays before further evaluation in multiple functional assays. The most promising compounds consisted of the Src family kinase inhibitors Dasatinib and eCF506 which showed dual reduction in YAP and TAZ nuclear translocation and subsequently reduction in cell proliferation through cell cycle arrest in both 2-dimensional and 3- dimensional spheroid assays. Another validated hit compound identified through screening was the Na+/K+ ATPase inhibitor ouabain which showed selective inhibition of TAZ and potent reduction in GBM cell number by apoptosis, in 2D and 3D tumour spheroid assays. Ouabain, Dasatinib and eCF506 showed enhanced activity when tested as drug combinations across several GBM stem cell lines. Finally, I explored the broader utility of the most active YAP/TAZ inhibitors identified in the GBM phenotypic screen in other cancer cell line models where Hippo pathway modulation and YAP/TAZ activity are prevalent.

Grantor dc:publisher

The University of Edinburgh

Year dc:date.issued

2025

Author dc:creator

• Harvey, Mungo James Bruce

Advisors dc:contributor.advisor

• Carragher, Neil
• Dawson, John
• Frame, Margaret

• Glioblastoma (GBM)
• YAP and TAZ
• Phenotypic screening
• Hippo pathway
• Machine learning

Language dc:language.iso

en

Identifier URI

http://dx.doi.org/10.7488/era/6616

OAI identifier oai:identifier

oai:era.ed.ac.uk:1842/44090