Investigation of the molecular and clinical heterogeneity of medulloblastoma subgroups

Background Medulloblastoma is the most common paediatric malignant brain tumour, with heterogeneous clinico-molecular characteristics and survival outcomes. Current WHO classification distinguishes 3 molecular subgroups: WNT, SHH (named after characteristic activation of the WNT/wingless and Sonic Hedgehog signalling pathways) and non-WNT/non-SHH medulloblastoma. Current risk stratification incorporates molecular and clinico-pathological disease correlates. Survival is associated with high-risk factors such as metastasis, large cell/anaplastic histology, MYC/MYCN amplification and subtotal resection; the presence of one or more of these factors defines high-risk disease. High-risk patients receive more intensive therapies at the cost of severe late effects as survivors. Moreover, 20% of standard-risk patients (defined by absence of all high-risk features) will die of their disease. Tumour profiling with genome-wide DNA methylation arrays is the current gold standard for molecular classification of brain tumours. Methylation arrays are also suitable for identifying DNA copy number (CN) changes, enabling simultaneous genomic and epigenomic characterisation. Hypothesis It was hypothesised that genome-wide Illumina HumanMethylation arrays provide a robust alternative to gold-standard SNP arrays for DNA CN detection and allow for single-platform, integrated genetic and epigenetic assessment , suitable for application to DNA derived from fresh-frozen and formalin-fixed, paraffin embedded tumour materials Aims Confirm usability of methylation arrays and develop methods to detect genomic alterations (aneuploidy and focal oncogene amplifications) Showcase application of methylation arrays as a cost-effective single-platform, integrated approach for improved prognostication within medulloblastoma patients. Methods In this project, methods to detect genomic alterations (aneuploidy and focal oncogene amplifications) using Illumina 450k methylation arrays were developed and validated. These methods were implemented to assess previously published cytogenetic prognostication schemes in medulloblastoma. Next, the GLMnet algorithm was used to identify prognostic methylation loci. These markers were assessed in non-WNT/non-SHH high-risk medulloblastomas and validated in an independent, mixed-risk non-WNT/non-SHH cohort. The previously published cytogenetic prognostic signature for standard-risk, non-WNT/non- SHH medulloblastoma, identified in the PNET4 clinical trial, and its potential for prognostication was assessed in high-risk, non-WNT/non-SHH disease, alone and in conjunction with methylation markers Results Easy to use methods with a low barrier to entry were developed to robustly identify genomic copy number and oncogene amplification. These methods were applied to validate two independent, previously published cytogenetic prognostication schemes within medulloblastoma. Two DNA methylation loci, mapping to MYO7A and TRIM72 genes, were identified as independently prognostic markers. A novel prognostication scheme, that combined DNA methylation markers with the PNET4 cytogenetic signature, was devised for non-WNT/non-SHH medulloblastoma. This scheme outperformed the PNET4 signature in the high-risk cohort, reclassifying 21% of high-risk patients to a favourable-risk category. Conclusion These results demonstrate the potential for routine cytogenetic assessment concurrent with molecular sub-classification using DNA methylation microarrays. Additionally, the integrated genetic and epigenetic stratification from a single platform enabled a more refined prognostication and the identification of a subset of patients, currently classified as high risk, who demonstrate improved outcomes and who may be eligible for reduced intensity treatments that would offer a better quality of life as brain tumour survivors