ID #897 Patterns of chromosomal instability and their clinical correlates in pediatric central nervous system tumors
Lara Pohl, Nahal Saifar, Maximilian Leitheiser, Michael BockmayrAbstract
Background
Copy number alterations (CNAs) have been identified as prognostic factors in pediatric CNS tumors. While certain entities are known to exhibit a higher number of alterations, a comprehensive characterization of chromosomal instability (CIN) is still lacking. Previously, 17 signatures of CIN corresponding to distinct biological mechanisms were defined in a pan-cancer single nucleotide polymorphism (SNP) array cohort. While SNP profiles remain the gold standard for CNA detection, they are largely unavailable for pediatric CNS tumors, where DNA methylation arrays are routinely used.
Methods
We defined optimal parameters for CIN signature quantification from DNA methylation arrays by analyzing 25 TCGA lung adenocarcinoma cases with available Affymetrix SNP 6.0, Illumina 450k, and EPIC v2 array data. Comparisons were performed at both the segment and signature levels. The optimal parameter combination was applied to a pan-brain cancer cohort to quantify 17 CIN signatures (CX1–CX17).
Results
We achieved a high correlation between the signature activities on SNP and EPIC arrays in the TCGA cohort used for parameter optimization (Pearson: R = 0.93, Spearman: R = 0.75). CIN signatures were quantified in a pan-brain cancer cohort including 3,369 tumors (EPIC arrays) representing 61 molecular types, including 670 medulloblastomas (MB) and 510 ependymomas (EPN). All analyzed entities showed a high frequency of signatures indicating long-sized copy number changes (CX1, CX3, CX7, and CX15). EPN frequently harbored CX2, CX5, and CX6 while exhibiting high heterogeneity between molecular types. MB SHH-3 showed a high frequency of signatures related to replication stress (CX8, CX11, and CX13), unlike other MB subtypes. Medulloblastoma cases with CX11 and CX13 showed an inferior progression-free survival (p < 0.001).
Conclusion
Signatures of CIN were successfully measured in a large pan-brain cancer cohort using array-based DNA methylation data. Heterogeneous patterns of CIN were identified in pediatric CNS tumors, with prognostic relevance in MB.