Nuclear topology modulates the mutational landscapes of cancer genomes

1000 VOLUME 24 NUMBER 11 NOVEMBER 2017 NAture StructurAL & moLecuLAr bIoLogY Emerging evidence indicates that somatic mutations in cancer genomes are nonrandomly distributed and are influenced by factors such as genomic context and DNA secondary structures, chromatin organization, transcriptional activity, and replication timing1–11. Local variation in the mutation burden stems from variability in processes of DNA damage and/or repair3,5,12,13, and has implications for the identification of potential cancer-driver genes14 and the clinical management of cancer patients, for example, radiosensitivity and immunotherapy15. However, the factors identified to date do not explain the entire extent of regional variation of the mutational burden in cancer genomes, thus suggesting that other factors remain to be identified. Genomic DNA is folded into higher-order domains, which occupy different territories in the three-dimensional architecture of the nucleus16–18, and nuclear-lamina-binding regions are usually at the nuclear periphery16,19,20. Nuclear organization of genetic material plays an important role in DNA replication21 as well as the processes of DNA damage and repair22–24. For instance, the nuclear-lamina-associated regions are refractory to homologous-recombination-mediated repair and use an error-prone alternative end-joining mechanism to repair DNA double-strand breaks25. Oct-1and p53-dependent pathways link lamin functions to the oxidative-stress response26. Indeed, a previous multivariate analysis has suggested that nuclear-lamina association significantly contributes to variations in germline mutation rates27. Furthermore, it has recently been reported that regulatory-domain boundaries are frequently disrupted in cancer28, and in some cases, such boundaries and the chromatin loops that underlie them are associated with unusual mutational spectra29. Here, we hypothesized that the nuclear organization of genomic DNA might modulate the somatic mutational landscape of cancer genomes and that its effects might surpass the variations due to known covariates such as chromatin domains and DNA-replication timing4,6.