DNA methylation-based profiling for paediatric CNS tumour diagnosis and treatment: a population-based study.

BACKGROUND: Marked variation exists in the use of genomic data in tumour diagnosis, and optimal integration with conventional diagnostic technology remains uncertain despite several studies reporting improved diagnostic accuracy, selection for targeted treatments, and stratification for trials. Our aim was to assess the added value of molecular profiling in routine clinical practice and the impact on conventional and experimental treatments. METHODS: This population-based study assessed the diagnostic and clinical use of DNA methylation-based profiling in childhood CNS tumours using two large national cohorts in the UK. In the diagnostic cohort-which included routinely diagnosed CNS tumours between Sept 1, 2016, and Sept 1, 2018-we assessed how the methylation profile altered or refined diagnosis in routine clinical practice and estimated how this would affect standard patient management. For the archival cohort of diagnostically difficult cases, we established how many cases could be solved using modern standard pathology, how many could only be solved using the methylation profile, and how many remained unsolvable. FINDINGS: Of 484 patients younger than 20 years with CNS tumours, 306 had DNA methylation arrays requested by the neuropathologist and were included in the diagnostic cohort. Molecular profiling added a unique contribution to clinical diagnosis in 107 (35%; 95% CI 30-40) of 306 cases in routine diagnostic practice-providing additional molecular subtyping data in 99 cases, amended the final diagnosis in five cases, and making potentially significant predictions in three cases. We estimated that it could change conventional management in 11 (4%; 95% CI 2-6) of 306 patients. Among 195 historically difficult-to-diagnose tumours in the archival cohort, 99 (51%) could be diagnosed using standard methods, with the addition of methylation profiling solving a further 34 (17%) cases. The remaining 62 (32%) cases were unresolved despite specialist pathology and methylation profiling. INTERPRETATION: Together, these data provide estimates of the impact that could be expected from routine implementation of genomic profiling into clinical practice, and indicate limitations where additional techniques will be required. We conclude that DNA methylation arrays are a useful diagnostic adjunct for childhood CNS tumours. FUNDING: The Brain Tumour Charity, Children with Cancer UK, Great Ormond Street Hospital Children's Charity, Olivia Hodson Cancer Fund, Cancer Research UK, and the National Institute of Health Research.

Comprehensive molecular characterisation of epilepsy-associated glioneuronal tumours.

Glioneuronal tumours are an important cause of treatment-resistant epilepsy. Subtypes of tumour are often poorly discriminated by histological features and may be difficult to diagnose due to a lack of robust diagnostic tools. This is illustrated by marked variability in the reported frequencies across different epilepsy surgical series. To address this, we used DNA methylation arrays and RNA sequencing to assay the methylation and expression profiles within a large cohort of glioneuronal tumours. By adopting a class discovery approach, we were able to identify two distinct groups of glioneuronal tumour, which only partially corresponded to the existing histological classification. Furthermore, by additional molecular analyses, we were able to identify pathogenic mutations in BRAF and FGFR1, specific to each group, in a high proportion of cases. Finally, by interrogating our expression data, we were able to show that each molecular group possessed expression phenotypes suggesting different cellular differentiation: astrocytic in one group and oligodendroglial in the second. Informed by this, we were able to identify CCND1, CSPG4, and PDGFRA as immunohistochemical targets which could distinguish between molecular groups. Our data suggest that the current histological classification of glioneuronal tumours does not adequately represent their underlying biology. Instead, we show that there are two molecular groups within glioneuronal tumours. The first of these displays astrocytic differentiation and is driven by BRAF mutations, while the second displays oligodendroglial differentiation and is driven by FGFR1 mutations.