IDDF2018-ABS-0100 Deciphering molecular properties of hypermutated colorectal and gastric cancer

    loading  Checking for direct PDF access through Ovid



It is well-known that tumour is caused by somatic mutations. However, great mutational heterogeneity is observed both across cancer types (>1000 fold) and with a given cancer type, with a fraction of them harbour >10 mutations per Mb, thus termed hypermutation. Hypermutated patients are suitable for PD-1 blockade with favourable prognosis. Nevertheless, other omics such as transcriptome and methylome in hypermutated samples remain poorly understood. Here, we try to determine the genome-wide effects of high mutation loads on transcriptome and methylome across two cancer types, namely colorectal cancer (CRC) and stomach adenocarcinoma (STAD).


All tumour mRNA expression datasets (RNASeqV2) and DNA methylation data (HumanMethylation450) were obtained from The Cancer Genome Atlas. Known batch effects were corrected using the ComBat function implemented in the Bioconductor sva package. Differentially expressed gene (DEGs, false discovery rate (FDR) adjusted P-value<0.05 and fold change >2 ) analysis between hypermutated and non-hypermutated was performed by DEGSeq package for R/Bioconductor. Significantly differentially methylated site (DMS, FDR adjusted P-value<1E-20 and beta value change >0.2) was performed by limma package for R/Bioconductor.


Most hypermutated cases were driven by microsatellite instability-high (MSI-H). Meanwhile, 935 and 1,047 DEGs and 1604 and 53 DMSs were identified in CRC and STAD (hypermutated vs. non-hypermutated), respectively. A well-conceived five-gene signature (DNAI2, EPHA5, GAS2L1, RNH1, TAGLN3) was identified that could predict prognosis of STAD hypermutated patients with high performance (C-index: 0.84). Additionally, hierarchical clustering of expression and methylation profiles show that majority of CRC and STAD hypermutated samples were mixed and separated from their respective non-hypermutated samples, exceeding the boundary of tissue-specificity. Further in-detail exploration uncovered that the underlying molecular mechanism was related to perturbation of chromatin remodelling genes (ARID1A, NCOR1, and MLL1 ~4) both in CRC and STAD hypermutated samples.


We thus concluded that MSI-H->chromatin remodelling genes inactivation->DNA methylation/expression variation axis shared in hypermutated samples was responsible for the consistent expression and methylation patterns across cancer types.

Related Topics

    loading  Loading Related Articles