High Levels of Necrosis in TCGA Glioblastoma Samples Are Associated with the Mesenchymal Gene Expression Class and Characterized by Enhanced Expression of Master Transcriptional Regulators of Mesenchymal Transition.
Daniel J Brat, David A Gutman, Lee AD Cooper, Jun Kong, Candace Chisolm, Erwin G Van Meir, Joel H Saltz, Carlos S Moreno. Emory University School of Medicine, Atlanta, GA
Background: The Cancer Genome Atlas (TCGA) project defined four clinically relevant subsets of glioblastoma (GBM) based on gene expression: mesenchymal, classical, proneural and neural. Genomic alterations were identified for each class, yet complete correlation with expression was not evident, suggesting tumor micro-environment may contribute. We investigated if the degree of necrosis within GBM samples used for TCGA studies was associated with a specific expression class and examined which genes were most tightly correlated with extent of necrosis.
Design: We utilitzed publicly available digitized frozen section slides from 79 GBMs used for quality assurance before molecular testing by TCGA. Necrotic foci in each slide were outlined using a machine-human interface. Extent of necrosis (% tissue area) was correlated with expression class. Using the Significance Analysis of Microarray (SAM) survival module, a Cox regression analysis was performed using TCGA Affymetrix U133 expression data to identify genes correlated with %-necrosis. The SAM survival module used a permutation algorithm to correct for multiple hypothesis testing and identify correlated probes with a false discovery rate cutoff of < 4%. 1026 probes were positively associated with %-necrosis and 18 were negatively associated. These 1044 probes were used as input for Ingenuity Pathway Analysis (IPA) to identify enriched networks.
Results: The mesenchymal expression class was enriched with high %-necrosis GBMs (33% with > 25% necrosis) compared to the other 3 classes (2% with > 25% necrosis). Average %-necrosis was higher in mesenchymal GBMs (21.2%) than the other 3 classes (7.4%; one-way ANOVA p = 0.0022). Regression analysis of genes correlated with %-necrosis revealed strong upregulation of transcription factors identified as master regulators of mesenchymal transition (Carro, et al. Nature 263: 318-25, 2010), including C/EBP-B, C/EBP-D, FOSL2, STAT3 (p < 0.005) and RUNX1 (p < 0.01). Increased expression of hypoxia-inducible genes, including VEGF, validated the analysis. FOSL2 and C/EBP-B were strongly correlated with extent of necrosis within the mesenchymal class (p < 0.00001).
Conclusions: Our results suggest that the degree of necrosis in GBM is closely related to the mesenchymal gene expression class.
Category: Special Category - Pan-genomic/Pan-proteomic approaches to Cancer
Monday, February 28, 2011 1:00 PM
Poster Session II # 241, Monday Afternoon