ChIP-seq dataset for epigenomics landscape of colorectal cancer

Cancer cells utilize genetic and epigenetic aberrations for their excessive growth. Although we have sufficient understanding of the genomic alterations in colorectal cancer, we have incomplete knowledge of epigenomic aberrations and their impact on tumor growth. In order to comprehensively define the epigenetic patterns specific to colorectal cancer, we generated profiles for 6 histone modification marks, including H3K4me1 (enhancer), H3K27Ac (active enhancer), H3K9me3 (heterochromatin), H3K27me3 (polycomb repression), H3K79me2 (transcription) and H3K4me3 (promoter), using a high-throughput ChIP-Seq methodology developed in house applicable to frozen tumors. Chromatin state transitions specifically pointed to drastic changes in enhancer patterns, consistent with some prior studies. Furthermore, we identified the best combinatorial chromatin states that could most efficiently distinguish and predict CRC from normal colon. In a more detailed investigation into patterns of active enhancers using normal colon, adenomas and colorectal cancers, we identified specific changes in enhancers from normal tissue to these neoplastic lesions. Importantly, we noted gains of enhancers in a large number of genomic loci in colon cancer compared to adjacent normal tissues. In summary, we have identified aberrant enhancer gains as a major feature of colorectal cancer and propose this to be utilized as a therapeutic approach. ChIP-seq samples across six different marks -- H3K4me1 (n = 32; 16 Normal and 16 Tumor), H3K9me3 (n = 30 samples; 14 Normal and 16 Tumor), H3K27ac (n = 68 samples; 15 Normal, 34 Tumor, 4 FAPP Normal, 4 Adenoma and 11 CCLE cell lines), H3K79me2 (n = 31 samples; 17 normal and 14 Tumor), H3K4me3 (n = 33 samples; 17 Normal and 16 Tumor) and H3K27me3 (n = 29; 14 normal and 15 Tumor). Raw fastq reads for all ChIP-seq experiments were processed by using a snakemake-based pipeline (https://zenodo.org/record/819971). The raw reads were first processed by using FastQC, and uniquely mapped reads were aligned to the hg19 reference genome by using Bowtie Ver. 1.1.2. Duplicate reads were removed by using SAMBLASTER(7) before compression to bam files. To directly compare ChIP-seq samples, uniquely mapped reads for each mark were downsampled per condition to 15 million, sorted, and indexed using samtools ver. 1.5. To visualize ChIP-seq libraries on the IGV genome browser, we used deepTools Version 2a.4.0, which generated bigWig files by scaling the bam files to reads per kilobase per million (RPKM). Super ChIP-seq tracks were generated by merging BAM files from each phenotype, sorting and indexing using SAMtools, and scaling to RPKM using deepTools. We used the Model-based analysis of ChIP-seq (MACS) version 1.4.2(10) peak calling algorithm to identify peaks for the sharp marks H3K4me1, H3K27ac, and H3K4me3; we used MACS version 2.1.0 to identify peaks for broad marks H3K9me3, H3K79me2, and H3K27me3. The peak enrichment was calculated over a whole genome "input" background with a p-value of < 1e-5. Super-enhancers were identified using the ranking of super-enhancers (ROSE) algorithm based on H3K27ac ChIP-seq data.