Intratumoral heterogeneity and longitudinal changes in gene expression predict differential drug sensitivity in newly diagnosed and recurrent glioblastoma

We investigated longitudinal transcriptomic patterns associated with glioblastoma (GB) recurrence by integrative approach utilizing multisampling strategy, RNA sequencing and complementary investigations of tumor tissues and GB stem cells. In total, 128 tissue samples of 44 tumors including 23 first diagnosed, 19 recurrent and 2 secondary recurrent GBs were analyzed in parallel with 27 primary cultures of GB stem cells. We found that intratumoral transcriptomic heterogeneity is an intrinsic characteristic that is conserved in newly diagnosed and recurrent GBs and captured in GB stem cells. We revealed a high degree of concordance between GB tumor tissues and stem cells in the longitudinal transcriptomic changes associated with tumor recurrence. We used gene expression data to model the efficacy of 130 anti-cancer drugs. For the recurrent tumor tissues and isolated GB stem cells we showed dramatically reduced simulated sensitivities to the first line chemotherapeutic temozolomide. In turn, several therapeutics including immune checkpoint inhibitors were predicted to be more effective in the recurrence setting. Our results strongly suggest that the spectrum of potentially effective drugs may differ between newly diagnosed and recurrent glioblastomas and provides a transcriptional rationale for the lack of significant therapeutic benefit from temozolomide in patients with recurrent GB. Overall design: Primary and recurrent GB tissues and GB-derived stem cells cultures were RNAseq-profiled. Gene expression values and molecular pathway activation scores were calculated for each sample. The efficacy of 130 anti-cancer drugs was modeled based on gene expression and molecular pathway activation data.