Epigenomic evolution during tumorigenesis in colorectal cancer

Cancer cells exhibit distinct features compared with normal cells, such as genomic instability and phenotypic plasticity. Growing evidence suggests that epigenomic reprogramming during tumorigenesis, in addition to genomic mutations, plays a critical role in establishing cancer hallmarks. However, the genome-wide landscape of epigenomic alterations and the interplay among the genome, epigenome, and transcriptome during tumorigenesis remain poorly understood, particularly at single-cell resolution. Here, we investigated the genome, DNA methylome, and transcriptome of 129 single-cell-derived cancer clones and 142 single-crypt-derived matched normal clones from 13 patients with colorectal cancer. Compared with normal clones, in which ~75% of CpG sites were fully methylated, cancer clones exhibited pervasive global demethylation, with only ~50% of CpGs remaining fully methylated. We estimated the somatic epimutations using the consensus methylation states of normal clones and found that, on average, 7.5 million CpG sites (~25% of all CpGs) were demethylated in cancer clones. The burden of somatic demethylation increased linearly with the endogenous somatic SNV burden in cancer, suggesting that widespread demethylation primarily results from imperfect DNMT1-mediated maintenance of DNA methylation. The estimated timing of chromosomal gains inferred from genomic mutations and CpG demethylation was largely consistent, supporting the clock-like accumulation of demethylation across broad genomic regions during tumorigenesis. Interestingly, promoter CpG methylation was slightly increased in cancer compared with normal clones, despite pervasive demethylation. We identified genes with differentially methylated promoters, including hypermethylated tumor-suppressor genes and hypomethylated oncogenes, and correlated these methylation changes with gene-expression profiles, supporting their role as driver epimutations. In addition, despite the activation of L1 elements during tumorigenesis, demethylation rates in L1HS promoters were lower than in the rest of the genome in both normal and cancer clones, suggesting that L1 activation in cancer is a passive consequence of global demethylation rather than an active regulatory process. Overall, our single-cell multi-omics analysis provides a comprehensive view of epigenomic reprogramming during tumorigenesis and offers valuable insights into how epigenomic alterations contribute to cancer hallmarks.