(505d) Partial Least Squares Regression Modeling Identifies Combination Treatments That Overcome Sprouty2-Mediated Glioblastoma Resistance to Multiple Classes of Therapeutics

Glioblastoma is a devastating form of brain cancer with a median patient survival time of less than 15 months under the current standard of care. Glioblastoma tumors will often recur due to post-translational protein modifications that drive glioblastoma resistance to therapeutic agents. These resistance mechanisms render current treatments inadequate. A quantitative understanding of broad regulators of resistance is essential to the design of therapeutic treatments that improve patient survival. We previously identified the protein Sprouty2 (SPRY2) as a critical regulator of glioblastoma phenotypes. We found that SPRY2 knockdown reduced colony formation by human glioblastoma cells and antagonized the growth of xenografts in mice, suggesting that SPRY2 promotes glioblastoma progression. These findings were surprising because SPRY2 acts as a tumor suppressor in numerous other cancers (e.g., lung cancer). However, therapeutics that target SPRY2 have not yet been developed. We hypothesized that SPRY2 regulates druggable pathways that can be rationally identified using a network level systems biology approach. We explored the role of SPRY2 in modulating glioblastoma cell response to the DNA damaging agents temozolomide and carboplatin, and inhibitors of the EGFR and MET receptor tyrosine kinases. In parallel with measurements of cell death and cell cycle, we assessed dynamic changes in the phosphorylation or abundance of a network of proteins regulating cell survival, apoptosis, and DNA damage response using a multiplexed microbead immunoassay. We constructed a quantitative data-driven model of the relationships between the measured signaling processes and cellular phenotypes using partial least squares regression (PLSR). Our results show that endogenous expression of SPRY2 regulates cell signaling in a manner that drives cancer cell resistance to DNA damaging agents and tyrosine kinase inhibitors. The PLSR model revealed a core subset of SPRY2-regulated kinases that were, surprisingly, common to the response to both DNA damaging agents and tyrosine kinase inhibitors. The model also identified SPRY2-regulated kinases that were unique to the response to DNA damaging agents or tyrosine kinase inhibitors. Model predictions were validated in multiple glioblastoma cell backgrounds. Therapeutic treatment, in combination with an inhibitor of model-identified-kinases of the DNA damage response pathway, augmented cell death in a manner comparable to SPRY2 knockdown. Our results identify novel combination treatments with the potential to overcome glioblastoma cell resistance to a broad range of therapeutics and demonstrate the utility of data-driven modeling approaches for the rational development of combination therapy in oncology.