Gene expression signature to predict radiation response in lower-grade gliomas.

Abstract

2019 Background: Standard of care for lower-grade glioma (LGG) is maximal safe resection and risk-adaptive adjuvant therapy. While patients who benefit the most from adjuvant chemotherapy have been elucidated in prospective randomized studies, comparable insights for adjuvant radiotherapy (RT) are lacking. We sought to identify and validate patterns of gene expression that are associated with differential outcomes among LGG patients treated by RT from two large genomics databases. Methods: Patients from The Cancer Genome Atlas (TCGA) with LGG (WHO grade II–III gliomas) treated by surgery and adjuvant RT were randomized 1:1 to a training set or an internal validation set. Using patients in the training set, association between gene expression from resected tumor and progression-free survival (PFS) as well as overall survival (OS) was evaluated with adjustment for clinicopathologic covariates. A genomic risk score (GRS) was then constructed from the expression levels of top genes also screened for involvement in glioma carcinogenesis. The prognostic value of GRS was subsequently validated in the internal validation set of TCGA and a second distinct database, compiled by the Chinese Glioma Genome Association (CGGA). Results: From TCGA, 289 patients with LGG received adjuvant RT alone (38 grade II, 30 grade III) or chemoradiotherapy (CRT) (51 grade II, 170 grade III) between 2009 and 2015. From CGGA, 178 patients with LGG received adjuvant RT alone (40 grade II, 13 grade III) or CRT (41 grade II, 84 grade III) between 2004 and 2016. The genes comprising GRS are MAP3K15, MAPK10, CCL3, CCL4, and ADAMTS1, involved in MAP kinase activity, T cell chemotaxis, and cell cycle transition. High GRS, defined as having a GRS in the top third, was significantly associated with worse outcomes independent of age, sex, glioma histology, WHO grade, IDH mutation, 1p/19q co-deletion, and chemotherapy status in the training set (OS HR 2.74, P < 0.001; PFS HR 1.61, P = 0.014). These findings were further validated in the internal validation set (OS HR 1.84, P = 0.015; PFS HR 1.58, P = 0.027) and again in the CGGA external validation set (OS HR 1.72, P = 0.001). Association between GRS and outcomes was observed only among patients who received RT (RT alone or CRT), in both TCGA and CGGA. Conclusions: This study successfully identified an expression signature of five genes that stratified outcomes among LGG patients who received adjuvant RT, with two rounds of validation leveraging independent genomics databases. Expression levels of the highlighted genes were associated with survival only among patients whose treatments included RT, but not among those with omission of RT, suggesting that expression of these genes may be predictive of radiation treatment response. While additional prospective studies are warranted, interrogation of these genes to determine high/low GRS may be considered in the multidisciplinary management of LGGs.