Poster Presentation & Flash Talk Presentation 29th Lorne Cancer Conference 2017

CpG methylation accounts for genome-wide C>T mutation variation and cancer driver formation across different cancer types (#11)

Rebecca C. Poulos 1 , Jake Olivier 2 , Jason W. H. Wong 1
  1. Prince of Wales Clinical School and Lowy Cancer Research Centre, UNSW Australia, Sydney, NSW, Australia
  2. School of Mathematics and Statistics, The Red Centre, UNSW Australia, Sydney, NSW, Australia

Cancer develops when somatic mutations accumulate within cells, but the origin of many mutations remains unknown. Methylation of cytosine nucleotides (5mC) is an important epigenetic mark, involved particularly in gene transcription and silencing. However, 5mC nucleotides are more likely to undergo spontaneous deamination than unmethylated cytosines, resulting in increased mutation rates at these sites, and selection against CpG dinucleotides across evolutionary history. In this study, we perform a genome-wide analysis of the influence of methylation on mutation accumulation across 11 cancer-types. By analysing over 900 cancer genomes, together with replication-timing and tissue-specific methylation and heterochromatin data, we reveal differential associations between these factors and mutation accumulation. We describe a novel finding in colorectal cancers whereby samples with microsatellite instability (MSI) and Polymerase Epsilon exonuclease-domain deficiency (Pol epsilon exo-/-) exhibit incredibly high rates of mutations at methylated CpG sites. We identify that hotspot coding mutations occurring almost exclusively in Pol epsilon exo-/- colorectal cancers occur at methylated CpG dinucleotides. Many such mutations are in key cancer driver genes such as APC and TP53, highlighting the potential impact that this association with methylation can have on oncogenesis in this cancer subtype. We also demonstrate that methylation partially underlies the observation of the loss of regional variation in mutation rates in MSI colorectal cancers. Incorporating additional cancer types and subtypes, we develop statistical models accounting for the differential influence of replication timing, methylation and heterochromatin. These models reveal cancer samples with mutation signatures characterised by APOBEC/AID cytosine deaminases to exhibit unique methylation-mutation profiles. Finally, we show that while mutations in most cancer-types are positively correlated with methylation at low to intermediate levels, in melanoma, pancreatic and liver cancer, mutation rates are reduced at highly methylated regions. This is the first study to systematically analyse the genome-wide influence of methylation on mutation rate across cancer-types, revealing significant associations which are vital for accurately mapping regional variation in mutation density and pinpointing driver mutations.