GENE EXPRESSION SIGNATURE ASSOCIATED WITH AGGRESSIVE GLIOBLASTOMA GROWTH IS ENRICHED IN CHROMATIN MODIFICATION AND STEMNESS TRANSCRIPTIONAL REGULATION PROGRAMS

Abstract Prognosis of patients diagnosed with IDHwt glioblastoma is influenced by known clinical and demographic factors, and likely by physiological characteristics. Our goal is to determine tumor-intrinsic gene expression signatures associated with aggressive tumor growth in patient-derived xenografts (PDXs). Cancer stem cell (CSC) lines established from 10 IDHwt glioblastoma tumors were implanted orthotopically in cohorts of 10 to 15 nude mice (3x10E5 viable cells/mouse), for development of PDX under uniform conditions. Mice were monitored and sacrificed when symptomatic. Five PDX lines, presenting median survival of 29 to 59 days were classified as short (S) survivors, and 5 lines with median survival between 111 and 134 days as long (L) survivors. RNA was isolated from terminal PDX tumors (n=3/line) and sequenced using Illumina HiSeq 2000. Differential gene expression analysis between tumors in S and L survival groups was conducted using the lmFit and eBayes, and genes were ranked by Benjamini-Hochberg adjusted P-values, set to adj.p< 0.05, resulting in 1663 genes upregulated and 1539 genes downregulated in the aggressive S group. Gene ontology analysis was performed using Metacore (Clarivate Analytics) and Metascape (http://metascape.org). Chromatin modification was significantly enriched in the aggressive tumor group (Metacore, p= 1x 10–12). Remarkably, 40% (654/1663) of the genes upregulated in the aggressive PDX tumors were co-expressed with an epigenetic master regulator in the TCGA glioblastoma RNAseq dataset (cBio Portal, q< 10E-15), and this subset was also highly enriched in chromatin modification, stemness transcriptional regulation and DNA repair (q=10E-45 to q=10E-13). These results indicate that novel host-independent prognostic gene expression signatures can be derived from the PDX models and underline the potential of epigenetic regulators as therapeutic target for aggressive glioblastomas. Our results further indicate that these models are suitable for testing a new generation of epigenetic drugs currently in pre-clinical and clinical development.