CCRG-03. DETECTION OF HETEROGENEOUS CELL CYCLE ALTERATIONS UPON TMZ CHEMOTHERAPY TREATMENT THROUGH ADVANCED MACHINE LEARNING

Abstract Glioblastoma (GB), the most aggressive central nervous system tumor, is hallmarked by its frequent relapse under alkylating chemotherapy. GB cells frequently exhibit changes in cell cycle phases during TMZ treatment, with a reduction in the G0-G1 phase and an increase in the S and G2 phases. This indicate both tumor cell re-entry into the cell cycle and a TMZ-induced arrest, contributing to resistance and progression. In this study, we leverage dynamic protein abundance of Ki-67, a proliferation marker peaking during S, G2, and M phases, to analyse cellular responses to TMZ in various cell lines and tissue samples. We designed an image-based machine learning approach for pattern recognition to reconstruct and analyse the cell cycle at the single-cell level. Six primary cell lines were characterized through whole-genome DNA, methylation and RNA sequencing. Functional read-outs were performed with large-scale, time- and concentration-dependent TMZ treatments in cell culture and GB-tissue slice models. Our model successfully reconstructed detailed cell cycle stages, showing that untreated cells were predominantly in the G0-G1 phase. TMZ treatment significantly shifted cells toward the S/G2 phase (p < 0.02) in 3 of 6 cell lines and the pseudo-time reconstruction highlighted dose-dependent cell cycle changes. Cell lines with enhanced stem-like features, defined based on epigenetic and transcriptional data signature, showed no cell cycle changes at high TMZ doses (>150µM), regardless the MGMT promoter methylation status. This heterogeneity in cell cycle response emphasizes the complexity of GB and the need for personalized treatments. Our study provides insights into GB MGMT cell cycle alterations and resistance mechanisms that are independent of MGMT methylation status, a known TMZ response biomarker. Therefore, the observed TMZ-induced cell cycle changes may serve as novel biomarkers for assessing treatment response, paving the way for personalized therapeutic approaches.