Long non-coding RNA expression profiles predict treatment outcome in glioma

Abstract

BACKGROUND: Histopathologic diagnosis of gliomas based on current WHO criteria offers a valuable but insufficient prediction for clinical outcome. Defining glioma subtypes based on molecular signatures that can capture the heterogeneity within each WHO diagnostic category is necessary. Long non-coding RNAs (lncRNAs) are newly emerging non-coding gene regulators in cancer, the aberrant expressions of which have been indicated in cancer biogenesis and prognosis. Here, we investigated whether lncRNA expression signatures could predict clinical outcome in glioma. MATERIALS AND METHODS: Using a lncRNA-mining approach, we performed comprehensive lncRNA profiling analysis on two large cohorts of glioma specimens which contain comprehensive WHO grades and histology cell types. We analyzed the differential lncRNA expression patterns between tumor malignancy grades and histological cell types. We also assessed the associations between lncRNA signatures and treatment outcome in glioblastoma (GBM). RESULTS: We identified sets of deregulated lncRNAs that were unique to different malignancy grade and histological subtypes in glioma. These lncRNA sets may reflect the malignancy progression and differentiation in glioma. In GBM, a six-lncRNA signature was significantly associated with the overall survival. It could reliably classify GBM patients into two subgroups with different prognosis in the training set (n=107), testing set (n=106) and two independent validation sets (n=101, and 68, respectively). Moreover, we found that the predictive power of this six-lncRNA signature was independent of MGMT methylation status, and could add to its prognostic value even after the patients were stratified. CONCLUSION: LncRNAs may play important role in glioma biogenesis and prognosis. The identification of prognostic lncRNAs may have clinical implications in development of novel biomarkers and targeted therapy.