The Actionable Genomic Landscape Of Relapsed Medulloblastoma Is Defined By Maintenance And Acquisition Of Driver Events

Abstract Medulloblastoma relapse (rMB) occurs in 30–40% of patients and is almost universally fatal. Understanding the genomic landscape of rMB, and its relationship to disease characteristics at diagnosis, will be essential to underpin the development of improved therapeutic strategies, delivered at both diagnosis and relapse. Utilising NGS and Illumina DNA methylation arrays, we interrogated the molecular landscape of >100 rMBs, alongside matched diagnostic samples (n>80), encompassing molecular subgroup, novel subtypes, copy number (CNV) and mutational variants. Molecular subgroup and novel subtypes were stable over disease-course. The majority of genomic aberrations were also maintained (total arm-level CNVs at relapse, 60% maintained/40% acquired; deleterious/driver mutations, 75% maintained/25% acquired). Importantly, however, the landscape of alterations differed markedly at relapse, through both selective maintenance and acquisition of specific gene and pathway aberrations. For instance, we observed significant enrichment of subgroup-specific events at relapse, including focal CDK6/CDK14 amplifications (4/26 (15%) of MBGroup4) and CDKN2A/CDKN2B deletions (3/48 (6%) of MBSHH). In contrast, mutations in DNA damage response pathways were commonly enriched across all molecular subgroups, most significantly in MBSHH (~40% of rMBSHH; TP53, 9/36 (25%); ATM, 5/36 (14%)). Driver events in rMB are characterised by both selective maintenance and acquisition across disease-course, and together combine to define its actionable genetic landscape. Evaluation of their clinical and biological significance will be essential to establish their potential (i) as biomarkers to direct disease management and (ii) as a basis for therapeutic strategies targeted against medulloblastoma relapse.