ChIP-Sequencing

Data Generation Protocols Data Analysis Protocols
CaCxIcon for HTMCP- Cervical Cancer Project CaCxIcon for HTMCP- Cervical Cancer Project

Data Generation Protocols

The data generation protocols for the HTMCP-Cervical Cancer project were acquired from the following manuscript.

Gagliardi A, Porter VL, Zong Z, et al. Analysis of Ugandan cervical carcinomas identifies human papillomavirus clade-specific epigenome and transcriptome landscapes. Nat Genet. 2020;52(8):800-810. (PMID: 32747824)

Native chromatin immunoprecipitation (ChIP) sequencing

Fifty-two tumor samples were lysed in 0.1% Triton X-100, 0.1% Deoxycholate buffer plus protease inhibitors (PI).  Extracted chromatin was digested with micrococcal nuclease (MNase) enzyme (NEB) and the reaction quenched using 250 µM of EDTA. 1% Triton X-100 and 1% Deoxycholate were mixed and added to the samples on ice. 4% of digested chromatin was used as input control, the remaining was pre-cleared with Protein A/G Dynabeads (Invitrogen) in IP buffer (20 mM Tris-HCl [pH7.5], 2 mM EDTA, 150 mM NaCl, 0.1% Triton X-100, 0.1% Deoxycholate, PI) at 4oC for 1.5 hours. Supernatants were transferred to a 96-well plate containing the antibody-bead complex, and incubated overnight at 4oC with agitation. Immunoprecipitated samples were washed twice with low salt buffer (20 mM Tris-HCl [pH 8.0], 0.1% SDS, 1% Triton X-100, 2 mM EDTA, 150 mM NaCl) and twice with high salt buffer (same, but with 500 mM NaCl). DNA-antibody complexes were eluted in Elution Buffer (100 mM NaHCO3, 1% SDS), at 65°C for 1.5 hours with mixing (1350 rpm). Qiagen Protease was used to digest protein in the eluted DNA at 50°C for 30 minutes with mixing (600 rpm). ChIP DNA was purified using Sera-Mag beads (Fisher Scientific) with 30% PEG before library construction as described for custom capture.

Amplified libraries were purified as described above (ALINE Biosciences) and the DNA quality and quantity determined using Caliper LabChip GX DNA High Sensitivity assay (PerkinElmer) and the Quant-iT dsDNA high sensitivity assay (ThermoFisher Scientific).

Experimental protocols

To request more information or approval regarding the following protocols, please contact BC Cancer at labqa@bcgsc.ca

qPCR of Native ChIP Libraries

Native ChIP Using 100,000 Cells

Data Analysis Protocol

The data analysis protocols for the HTMCP- Cervical Cancer project were acquired from the following manuscript.

Gagliardi A, Porter VL, Zong Z, et al. Analysis of Ugandan cervical carcinomas identifies human papillomavirus clade-specific epigenome and transcriptome landscapes. Nat Genet. 2020;52(8):800-810. (PMID: 32747824)

ChIP-Seq alignment and peak calling

ChIP sequence reads (75nt) were aligned to the human reference genome (hg19) with BWA-MEM (v0.7.6a, parameters: -M). Read duplicates were marked using sambamba (v0.5.5). Forty-seven samples had all 6 histone marks (4 broad: H3K4me1, H3K9me3, H3K27me3, H3K36me3 and 2 narrow: H3K4me3 and H3K27ac), and 5 had a subset of these.

Peaks were called using MACS2 (v2.1.1) with default parameters, comparing each mark to its control. Bedgraph output files were converted to the library size-normalized bigWig format for manual inspection using the UCSC and IGV genome browsers.

ChIP-seq data quality was assessed using Encode guidelines. Samples had a minimum of 50 million sequenced reads for narrow marks and 100 million for broad marks. The percentage of uniquely mapped reads was above 70%, and the percentage of duplicated reads varied between 1 and 10%. The non-redundant fraction, fraction of reads in peaks (FRIP) and sequencing saturation using preseq v2.0.2 (https://github.com/smithlabcode/preseq) were also assessed.

ChIP clustering analyses

The union of peaks for each histone mark was found by concatenating peak files and merging overlapping regions using bedtools v2.27.1. The normalized coverage of each sample in the peak union was counted using deeptools (v3.0.2). For each mark, the top 1% most variable peaks were clustered using the ConsensusClusterPlus (v1.38.0, R) using the ‘pearson’ distance and ‘complete’ clustering method with 1000 iterations for k=2-10 clusters. The 54 consensus clustering solutions (6 marks x 9 solutions) were then analysed using a Cluster of Clusters Analysis (COCA). For active marks, pairwise probabilities were generated for 27 solutions (3 marks x 9 solutions). For marks in which some samples had missing data, pairwise comparisons were normalized to exclude samples in those marks. Matrices of probabilities (54x52, 27x52) were clustered using pheatmap (v1.0.10, R) with the ‘pearson’ distance and ‘complete’ clustering method.

H3K4me3, H3K27ac and H3K4me1 peaks differentially present between HPV clades (A7 vs. A9) were determined using DiffBind with DESeq2 method (v2.2.12, R, FDR<0.01, fold-change>2). Coverage at peaks was counted 500bp around the centre of the peak, and a multifactorial experimental design was performed to normalize histology differences (referred to as blocking factor). Significantly differential peaks were intersected using bedtools (v2.27.1). Associated genes were identified using the nearest transcription start site (TSS) to the differential H3K4me1 and intersected H3K4me3/H3K27ac regions, identified using bedtools (v2.27.1) to RefSeq’s hg19 annotation.

HPV integration events and ChIP

HPV integration sites were determined using chimeric reads mapping to both human and HPV genomes. Within each sample, integration sites were merged into a single integration event (n=257) if they were <500 kb apart. HPV integration hotspots were determined by counting the number of events that fell within a 500 kb bin across the genome.

ChIP-seq alterations at HPV events were clustered using the log2(fold-change) of normalized coverage (RPM) of the integrated sample versus the mean RPM of the unintegrated samples using the ‘pheatmap’ (v1.0.10, R) with a ‘ward.D2’ clustering method. Events <20 kb were extended to 20 kb to obtain adequate coverage of the region.

For each mark per event (6 modifications in 99 events), a control peakset was made by randomly selecting 1,000 peak regions of the same mark on the same chromosome as the event, and extending the peaks from the center to the same size as the event. Normalized ChIP-seq coverage of the histone modification at these 1,000 random peaks was counted in the 52 samples, and the log2(fold change) of coverage was calculated for the integrated sample. A p-value was calculated for the fold change of the integration event based on the distribution of fold changes in these control peaks. Benjamini-Hochberg adjusted p-values<0.05 was regarded as significant.

Last updated: August 07, 2020