Deconvolution of heterogeneity in the Glioblastoma cellular ecosystem for understanding treatment resistance and improving patient stratification

The following interdisciplinary proposal aims at investigating treatment resistance mechanisms in Glioblastoma (GBM). GBM is the most aggressive and incurable brain tumor with a dismissal prognosis of 12-15 months. Recent studies, including ours, have shown that GBM cells display very strong intrinsic plasticity and adapt reversibly to dynamic microenvironmental conditions, forming a very dynamic ecosystem. The role of GBM plasticity in creating resistant states upon treatment is currently less understood. We hypothesize that high plasticity allows GBM cells to adapt towards drug resistant states upon treatment. We posit that treatment can simultaneously modulate cells within tumor microenvironment, leading to an overall resistant GBM ecosystem.
This proposal is based on our complementary expertise in GBM biology and computational methods for deconvolution of the complex biological systems. By combining state-of-the art preclinical models, transcriptomic analyses at the single and bulk cell level and powerful deconvolution methods we aim to unravel mechanisms that shape treatment resistance in GBM. First we will assess transcriptomic changes upon treatment in our GBM patient-derived orthotopic xenografts (PDOXs). We will investigate transcriptomic adaptation at the single cell level within tumor cells and subpopulations forming tumor microenvironment. By comparing longitudinal models derived from patients prior and after treatment we will reveal long-term stable changes. Direct treatment of PDOXs will allow investigation of transcriptomic transitions towards resistant states at the moment of treatment. We will next investigate presence of treatment tolerant and resistant states in GBM patient tumors.
Advanced systems biology methods tailored to single cell transcriptomic datasets will reveal signatures of treatment resistant states and their regulators. We will further apply our consensus Independent Component Analysis (consICA), a reference-free deconvolution method, to assess treatment resistance signatures based on wide published and unpublished single cell and bulk transcriptomic datasets. consICA will further allow linking independent molecular signals from tumor and TME subpopulations to clinical outcomes such as patient treatment history and survival. The identified treatment resistance signatures will be validated at the protein level. Finally we will explore potential drugs targeting treatment resistance states and test their efficacy in 3D organoids ex vivo.
Our analyses will elucidate potential therapeutic targets for innovative combinatory treatment strategies. Assessing tumor composition prior and after treatment may further reveal predictive biomarkers of response at the level of individual genes and biological processes and may lead to improved stratification of patients for personalized therapies. Our scripts will be shared via website interface available to researchers with minimal computational skills. Our methodology will be accessible for a wider use, e.g. analysis transcriptomic datasets obtained during clinical trials for identification of biomarkers of responders and non-responders upon targeted therapies.