Epithelioid glioblastomas stratify into established diagnostic subsets upon integrated molecular analysis

Abstract

Epithelioid glioblastoma (eGBM) is a newly defined and rare GBM variant in the current WHO 2016 classification. BRAF V600E mutation is overrepresented in these tumors and there is known some morphological overlap with anaplastic epithelioid PXA (ePXA). In order to further elucidate this diagnostic category, we molecularly characterized 64 pediatric and adult examples initially diagnosed as "eGBM." Tumors were analyzed using array based methylation and direct sequencing of the BRAF and TERT genes. Our results demonstrated considerable molecular and clinical heterogeneity among eGBM cohort. Methylation patterns, copy number alterations, and mutational analysis data, in combination with clinical findings disclosed three different, well established tumor subtypes: (i) PXA-like tumors with favorable prognosis, predominantly in children and young adults (38), (ii) IDHwt GBM-like tumors with poor prognosis, mainly occurring in older adults, albeit with more frequent BRAF mutations (17), and (iii) RTK1 pediatric GBM-like neoplasms of intermediate prognosis in children and young adults, associated with chromothripsis and frequent PDGFRA amplifications (9). We conclude that the histopathologically defined eGBM do not represent a single diagnostic entity, but rather at least three molecularly and biologically distinct categories. Therefore, additional molecular testing through genome-wide molecular profiling is recommended to further stratify these rare cases.

Keywords: cytogenetic prognostic; epithelioid; glioblastoma; methylation; pleomorphic xanthoastrocytoma; subgroup; survival.

Figure 1

A. Unsupervised hierarchical clustering analysis of eGBM samples (based on the 5000 most variably methylated probes) together with epigenetic profiles generated for various CNS tumors. eGBM (black points) did not form any separate cluster but they were clustered either with “canonical” PXA grade II and III (yellow points), IDHwt GBM (blue and green points) or with pediatric RTK1 GBM (red points). B. Grouping of tumor methylation profiles according to t‐SNE confirms distribution of eGBM among either PXA, adult IDHwt GBM or pediatric RTK1 GBM.

Figure 2

Histopathological (A, C, E) and cytogenetic (B, D, F) patterns of eGBM. All tumors showed similar histopathological “epithelioid” patterns—collections of large melanoma‐like tumor cells with abundant cytoplasm eccentric nuclei. However, these tumors disclosed distinct cytogenetic and epigenetic profiles. A, B. eGBM from “PXA cluster” shows no amplifications, numerous chromosomal gains and CDKN2A homozygous deletion (arrow). C, D. eGBM from “adult IDHwt cluster” discloses amplification of CDK4/MDM2 (arrow), gain of 7 and monosomy 10. E, F. eGBM from “pediatric GBM RTK1 cluster” with amplification of PDGFRA (arrow) and patterns of chromothripsis.

Figure 3

Overall survival for the molecular subtypes of eGBM grouped as 3 separate cohorts. Red line—eGBM clustered with pediatric RTK1 GBM, blue line—eGBM clustered with adult IDHwt GBM, and yellow line—tumors clustered with canonical PXA. All inter‐groups differences are statistically significant (log rank test; P < 0.0001).