Biological mechanisms of disease relapse in childhood medulloblastoma

Abstract

Over 30% of patients diagnosed with a medulloblastoma experience disease recurrence. Relapse is almost universally fatal, only infants who receive delayed radiotherapy at disease recurrence typically survive long term. Consequently relapse is the single leading cause of mortality disease-wide. Improved understanding of medulloblastoma at diagnosis has led to the identification of four distinct molecular subgroups with differing biology and outcome. These comprise of medulloblastomas associated with WNT and SHH pathway disruption (MBWNT and MBSHH respectively), and Group 3 and Group 4 tumours (MBGroup3 and MBGroup4). In contrast, very little is understood about the disease at recurrence, and at present there are only two published studies interrogating the biology of relapsed medulloblastoma. However, improved understanding of the biology at relapse is critical to improving treatment. Events at disease relapse could be explored as therapeutic targets or, if predictive of disease recurrence, provide an opportunity to escalate upfront therapy with the aim of preventing relapse. This study compiled a cohort of medulloblastoma tumours sampled at relapse (n=29), paired with their diagnostic counterparts. All clinicopathological and molecular features, with an established relationship to disease prognosis at diagnosis, were interrogated in this paired relapse cohort. With the exception of molecular subgroup, all features investigated displayed evidence of alteration and predominantly acquisition at recurrence. Most strikingly, the emergence of combined p53-MYC defects was commonly observed at relapse and these features were associated with locally aggressive, rapidly progressive disease following relapse. Through collaborative work, this discovery was explored further, with the development of a novel GTML/Trp53 KI/KI mouse model which faithfully recapitulated the clinicopathological and molecular features of the p53-MYC human tumours, and demonstrated the dependency of tumourigenesis and maintenance on this genetic interaction. Moreover, therapeutic inhibition of Aurora A kinase using MLN8237 in these mouse tumours led to degradation of MYCN, tumour reduction and prolonged survival. v A novel genome-wide DNA methylation analysis was next undertaken in the paired relapse cohort, focusing on MBGroup4 tumours, to interrogate maintained and acquired DNA methylation events between diagnosis and relapse, which may play a role in tumour development. Individual CpG sites on the Infinium DNA methylation 450K array were assessed for changes in their DNA methylation status between diagnosis and relapse. Fifteen candidate genes demonstrated tumour-specific methylation states that emerged at relapse and correlated with gene expression. The T-box and Homeobox gene families accounted for 8/15 (53%) candidates identified. Both these families are reportedly important for tumour development in other cancers. In addition, several studies suggest that epigenetic mechanisms, such as DNA methylation, play a regulatory role in their gene expression. Finally, a large cohort of medulloblastoma tumours (n=206), sampled at diagnosis, from patients who are known to go on and recur, was assembled to investigate any subgroup-specific patterns and timings of relapse. MBWNT rarely relapsed, whereas MBSHH frequently relapsed at both local and distant sites, but were the tumour subgroup most readily salvaged by radiotherapy in patients who were not treated with craniospinal irradiation (CSI) at diagnosis (8/12, 67%). Both MBGroup3 and MBGroup4 were widely metastatic at recurrence (34/41 (83%) and 52/61 (85%)) but contrastingly MBGroup3 relapsed quickly (p=0.0022), whereas MBGroup4 relapsed more slowly (p=0.0008). In patients who did not receive upfront CSI, MYC amplification at diagnosis was associated with rapid disease progression after relapse (p=0.0003). No diagnostic feature was significantly associated with time to death following relapse in the cohort of patients who received upfront CSI. This finding was supported by data from the paired relapse cohort where, in patients who received upfront CSI, it was the biological features of the tumour at relapse and not diagnosis, which were associated with disease course. In summary, this study has discovered emergent combined p53-MYC defects at medulloblastoma relapse which are associated with disease behaviour, identified potentially epigenetically regulated candidate genes in relapsed MBGroup4 tumours, and shown that the patterns of disease relapse are associated with radiotherapy and molecular subgroup. Together these findings demonstrate that medulloblastoma tumour biology is significantly different at relapse and that the timings and location of vi disease recurrence should be considered in the context of molecular subgroup and treatment. Biopsy at disease recurrence is now essential to validate and expand on these novel findings, interrogate all molecular subgroups at disease recurrence, and translate these discoveries into improved outcomes for the patients suffering from this devastating diagnosis.