MB-84IDENTIFICATION OF MEDULLOBLASTOMA MOLECULAR SUBGROUPS USING METABOLITE PROFILES

Abstract

MB-84. IDENTIFICATION OF MEDULLOBLASTOMA MOLECULAR SUBGROUPS USING METABOLITE PROFILES Sarah Kohe1,2, Simrandip K. Gill1,2, Debbie Hicks3, Ed C. Schwalbe3, Stephen Crosier3, Lisa Storer4, Anbarasu Lourdusamy4, Christopher D. Bennett1,2, Martin Wilson1,2, Simon Bailey3, Daniel Williamson3, Richard G. Grundy4, Steven C. Clifford3, and Andrew C. Peet1,2; Institute of Cancer and Genomic Sciences, University of Birmingham, Birmingham, UK; Birmingham Childrens Hospital, Birmingham, UK; Northern Institute for Cancer Research, Newcastle University, Newcastle, UK; Childrens Brain Tumour Research Centre, Queens Medical Centre, University of Nottingham, Nottingham, UK Identification of the four consensus molecular subgroups of medulloblastoma is becoming increasingly important for determining risk-based treatment strategies. Metabolite profiles can distinguish between brain tumour types, therefore we investigated whether profiling can discriminate the molecular subgroups within medulloblastoma. Metabolite concentrations were determined using high resolution magnetic resonance spectroscopy (MRS) on biopsied tissue from 29 medulloblastomas. Molecular subgroup was determined by Illumina 450K DNA methylation array and consensus clustering. Mean metabolite concentrations showed taurine, characteristic of PNETs, was prominent in all subgroups. Lipid was significantly elevated in high-risk Group 3 tumours (n 1⁄4 5, p , 0.0002) consistent with it being a marker of poor prognosis. The ratio of glutamate to glutamine was significantly higher in SHH (n 1⁄4 6), and lower in Group 4 (n 1⁄4 15), p , 0.03. Decreased creatine was detected in SHH tumours, p , 0.002. Only low risk WNT tumours (n 1⁄4 3) contained GABA suggesting it may be a subgroup specific marker, supported by links between GABA and WNT signalling in the developing cerebellum. Glycine, typically associated with poor prognosis, was also significantly lower in WNT tumours (p , 0.003). Multivariate PLS discriminant analysis found metabolite profiles could discriminate subgroup with a classification accuracy of 85% in this pilot set. Lipid, glutamate, glutamine, taurine, hypotaurine, GABA, myoinositol, glycine and creatine were most discriminatory. Matched 1.5T in-vivo MRS concentrations (n 1⁄4 19) found significant correlations with ex-vivo values for taurine, glutamate, glutamine, glycine, creatine, and myoinositol (Pearson’s r range:0.61-0.67, p , 0.05). Identified profiles will inform non-invasive MRS methods for pre-operative subgroup identification, with potential to guide extent of surgical resection and enhanced disease monitoring. Neuro-Oncology 18:iii97–iii122, 2016. doi:10.1093/neuonc/now076.80 #The Author(s) 2016. Published by Oxford University Press on behalf of the Society for Neuro-Oncology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.