NOVEL PROGNOSTIC MOLECULAR SIGNATURES FOR IMPROVED RISK-CLASSIFICATION OF ATYPICAL TERATOID RHABDOID TUMOURS

Abstract Malignant Rhabdoid Tumours (MRT) are aggressive paediatric malignancies seen in the central nervous system (Atypical Teratoid Rhabdoid Tumours (ATRT)), and kidney and other soft tissues (Extra-cranial Rhabdoid Tumours (ECRT)). With current therapies often proving ineffective and a lack of clear prognostic associations with consensus subgroups, we explored the possibility of using prognostic molecular signatures to further identify the biological characteristics of high risk ATRT patients. By employing a cross-validated feature selection method the methylation profiles of 121 MRT patients were analysed with clinical data to obtain meta-CpG signatures associated with prognosis for ATRT, ECRT and MRT. The relationship between these meta-CpG signatures and the consensus subgroups were further explored, along with the correlation of meta-CpGs with gene expression to establish biological significance. By selecting CpGs for their ability to predict survival this method obtained three novel prognostic methylation signatures which predict MRT outcome (ATRT-5, ECRT-14 and MRT-42). These signatures are independent of molecular subgroup and each signature was significantly associated with overall survival (OS) and event free survival (EFS) in their respective cohorts (p<0.001). Both ATRT-5 and MRT-42 maintained their significant association with OS in an independent ATRT cohort (n=64) and each meta-CPG signature is prognostically independent of other major clinical risk factors (e.g. receipt of radiotherapy and presence of metastases). Biologically, individuals with high-risk methylation signatures showed a gene expression profile suggestive of higher proliferative rates and tumours with low-risk scores in ATRT-5 and MRT-42 had an upregulated inflammatory response and increased immune infiltration. Combining these meta-CpGs with other significant clinical risk-factors produced high performing multivariate Cox-models enabling us to propose new stratification models for ATRT and MRT patients. These subgroup-independent prognostic signatures represent a distinct biology in ATRT and, if validated in prospective studies, could progress the use and efficacy of precision-based medicine in this therapeutically challenging disease.