Transcriptome analyses reveal molecular mechanisms underlying phenotypic differences among transcriptional subtypes of glioblastoma.

Yuan‐Bo Pan,Siqi Wang,Zhenqi Jiang,Biao Yang,Cameron Lenahan,Jianhua Wang,Anwen Shao,Jianmin Zhang

doi:10.1111/jcmm.14976

Abstract

Using molecular signatures, previous studies have defined glioblastoma (GBM) subtypes with different phenotypes, such as the proneural (PN), neural (NL), mesenchymal (MES) and classical (CL) subtypes. However, the gene programmes underlying the phenotypes of these subtypes were less known. We applied weighted gene co‐expression network analysis to establish gene modules corresponding to various subtypes. RNA‐seq and immunohistochemical data were used to validate the expression of identified genes. We identified seven molecular subtype‐specific modules and several candidate signature genes for different subtypes. Next, we revealed, for the first time, that radioresistant/chemoresistant gene signatures exist only in the PN subtype, as described by Verhaak et al, but do not exist in the PN subtype described by Phillips et al PN subtype. Moreover, we revealed that the tumour cells in the MES subtype GBMs are under ER stress and that angiogenesis and the immune inflammatory response are both significantly elevated in this subtype. The molecular basis of these biological processes was also uncovered. Genes associated with alternative RNA splicing are up‐regulated in the CL subtype GBMs, and genes pertaining to energy synthesis are elevated in the NL subtype GBMs. In addition, we identified several survival‐associated genes that positively correlated with glioma grades. The identified intrinsic characteristics of different GBM subtypes can offer a potential clue to the pathogenesis and possible therapeutic targets for various subtypes.

Highlights

GBMs are the most common malignant primary brain tumour found in adults
We revealed that the radioresistant/chemoresistant gene signatures only exist in the Verhaak et al PN subtype, but not in the PN subtype that was described by Phillips et al Thirdly, we revealed that the tumour cells in the MES subtype are under ER stress and that angiogenesis and the immune inflammatory response are both significantly elevated in this subtype
Gene set enrichment analysis (GSEA) of the top 3 hub genes (SRPX2, ITGA3 and FAM20C) was performed (Figure 4C and Table S4). These results showed that the tumour cells in the MES subtype GBM were exposed to stress conditions, such as ischaemia or hypoxia, which lead to ER stress (ERS), resulting in the unfolded protein response (UPR), regulation of cell apoptosis and autophagy levels.[15]

Summary

| INTRODUCTION

GBMs are the most common malignant primary brain tumour found in adults. The primary treatment for GBM involves maximal surgical resection, followed by radiotherapy and alkylating chemotherapy with temozolomide.[1,2] The median survival of optimally treated GBM patients is 14 months, with a 26% 2-year survival rate.[1]. In order to categorize the tumours, Phillips et al identified three distinct subtypes of highgrade astrocytomas by performing gene expression profiling These subtypes are described as: proneural (PN), mesenchymal (MES) and proliferative (Prolif).[4]. The networks that function within specific molecular subtypes of GBMs were identified based on module-trait correlations, allowing us to identify the biological processes and molecular programmes behind specific phenotypes of different subtypes. We used WGCNA to identify several gene modules associated with four different molecular subtypes, which are involved in major biological processes. We revealed that the radioresistant/chemoresistant gene signatures only exist in the Verhaak et al PN subtype, but not in the PN subtype that was described by Phillips et al Thirdly, we revealed that the tumour cells in the MES subtype are under ER stress and that angiogenesis and the immune inflammatory response are both significantly elevated in this subtype. We identified several survival-associated genes that were positively correlated with glioma grades, using the RNA-seq and immunohistochemical data

| MATERIALS AND METHODS

Findings

| DISCUSSION