Genome-wide analysis of human global and transcription-coupled excision repair of UV damage at single-nucleotide resolution

Abstract

We developed a method for genome-wide mapping of DNA excision repair named XR-seq (excision repair sequencing). Human nucleotide excision repair generates two incisions surrounding the site of damage, creating an ∼30-mer. In XR-seq, this fragment is isolated and subjected to high-throughput sequencing. We used XR-seq to produce stranded, nucleotide-resolution maps of repair of two UV-induced DNA damages in human cells: cyclobutane pyrimidine dimers (CPDs) and (6-4) pyrimidine-pyrimidone photoproducts [(6-4)PPs]. In wild-type cells, CPD repair was highly associated with transcription, specifically with the template strand. Experiments in cells defective in either transcription-coupled excision repair or general excision repair isolated the contribution of each pathway to the overall repair pattern and showed that transcription-coupled repair of both photoproducts occurs exclusively on the template strand. XR-seq maps capture transcription-coupled repair at sites of divergent gene promoters and bidirectional enhancer RNA (eRNA) production at enhancers. XR-seq data also uncovered the repair characteristics and novel sequence preferences of CPDs and (6-4)PPs. XR-seq and the resulting repair maps will facilitate studies of the effects of genomic location, chromatin context, transcription, and replication on DNA repair in human cells.

Keywords: UV damage; divergent transcription; enhancer; genome-wide; nucleotide excision repair; transcription-coupled repair.

Figure 1.

The XR-seq method. (A) Schematic of the procedure to isolate the nominal 30-mer generated by nucleotide excision repair. UV-induced photoproducts are removed from the genome by dual incisions, releasing the primary excision product in complex with TFIIH. The primary product is degraded with a half-life of ∼2 h to ∼20-nt-long fragments that are bound to RPA. For XR-seq, the primary products are isolated by TFIIH immunoprecipitation. (B) Excision patterns of photoproducts in NHF1 (wild-type), XP-C (deficient in global repair), and CS-B (deficient in transcription-coupled repair) cells. The excised oligonucleotides were immunoprecipitated with either anti-(6-4)PP antibodies or anti-CPD antibodies, and then the indicated fraction of purified DNAs was radiolabeled at the 3′end with ³²P-cordycepin and analyzed on sequencing gels. (C) Procedure for preparation of the dsDNA library for the Illumina HiSeq 2000 platform. (D) Analysis of dsDNA libraries of the excised nominal 30-mer by polyacrylamide gel electrophoresis. One percent of the ligation products were PCR-amplified with the indicated cycles.

Figure 2.

Genome-wide maps of CPD and (6-4)PP excision repair in NHF1 wild-type cells. (A) Distribution of the XR-seq signal, separated by strand, for CPD and (6-4)PP over the entire chromosome 7 (chr7; top) or focused on a 1.5-Mb region (bottom). ENCODE total stranded RNA-seq tracks in black are plotted above the XR-seq tracks for comparison. Arrows on the bottom depict the direction and length of annotated genes. (B) Distribution of the aligned reads between annotated genes (UCSC hg19 genes; green), the 5000 bp upstream of the gene (light green), the 5000 bp downstream from the gene (gray), and intergenic regions (blue). For comparison are the average results of 50 permutations of a random set of 26mers from the hg19 genome. (C) Spearman's correlation coefficient ρ calculated between biological replicates (experiments conducted in two independently UV-treated populations of cells) and between CPD and (6-4)PP XR-seq in NHF1 cells. Samples are ordered by hierarchal clustering. Darker box shades indicate higher correlation. (D) Average profile of CPD XR-seq and (6-4)PP XR-seq over all University of California at Santa Cruz (UCSC) reference genes. Genes and XR-seq signal were separated based on their direction to allow differentiation between template (purple) and nontemplate (turquoise) strand repair. Signals over the gene body were normalized to a 3000-bp window to allow for comparison.

Figure 3.

Mapping transcription-coupled and global excision repair. (A) Distribution of the XR-seq signal, separated by strand, for CPD (top) and (6-4)PP (bottom) over a 1.5-Mb region of chromosome 3. Shown is signal from NHF1 wild-type cells (green), XP-C cells that are proficient in only transcription-coupled repair (purple), and CS-B cells that are proficient only in global excision repair (blue). ENCODE total stranded RNA-seq tracks in black are plotted above the XR-seq tracks for comparison. Arrows on the bottom depict the direction of annotated genes. (B) Genomic distribution of aligned reads between genes (green), the 5000 bp upstream of the gene (light green), the 5000 bp downstream from the genes (gray), and intergenic regions (blue) shows that XR-seq signal for both damages is highly enriched over genes in XP-C cells but is more evenly distributed in CS-B cells. (C,D) Average profile of transcription-coupled (C) and global (D) excision repair of CPD over all UCSC reference genes. Genes and XR-seq signal were separated based on their direction to allow differentiation between template (purple) and nontemplate (turquoise) strand repair. Signals over gene body were normalized to a 3000-bp window to allow for comparison. (E) Focusing in on the intergenic region highlighted in yellow in A, transcription-coupled repair of CPD and (6-4)PP in XP-C cells occurs on both strands and overlaps DNase-seq (DNase sequencing) and H3K27ac ChIP-seq (chromatin immunoprecipitation [ChIP] coupled with deep sequencing) signal and chromHMM strong enhancer prediction (yellow/orange). (F) Average plus strand (dark lines) and minus strand (lighter lines) CPD transcription-coupled repair in XP-C cells around the center of intergenic DNase peaks. DNase peaks were divided into “enhancer peaks,” which overlap H3K27ac peaks or chromHMM strong enhancers (golds), and “intergenic peaks,” which do not (blues).

Figure 4.

Strong association of transcription-coupled repair with RNA levels. CPD repair profile around the TSS is plotted for the template strand (A) or nontemplate strand (B). (Top row)Average profile for five gene groups. Genes were divided into five groups based on expression level and include nonexpressed (black), lowest (red), low (orange), high (purple), and highest (blue) based on the calculated FPKM from RNA-seq in NHDF cells. (Bottom row) Corresponding heat map of repair over all expressed genes, which are ordered by ascending FPKM.

Figure 5.

Single-nucleotide resolution of excision repair in NHF1 wild-type cells. (A) Distribution of excised oligonucleotide sequence lengths, calculated after removal of flanking adapter sequences from sequenced 50-nt reads. (B) Analysis of the frequency of each of the possible dipyrimidines along reads of 26-nt length shows enrichment 5–7 nt and 6–7 nt from the 3′ end for CPD XR-seq and (6-4)PP XR-seq, respectively. (C) Analysis of the nucleotides flanking the putative damaged pyrimidines at position 19–20 of the 26-nt-long excised fragments reveals sequence context preferences. Depicted are TT for CPD XR-seq and TC for (6-4)PP XR-seq. For comparison, the expected frequencies from an average of 50 random permutations of 26mers in the hg19 reference genomes are shown. (D) The oligonucleotides containing (6-4)PP were first repaired by photolyase and then digested by RecJ exonuclease. Only repaired or undamaged DNA could be completely degraded by RecJ, which is blocked by (6-4)PP. Locations of completely degraded products (repaired) and partially degraded products (unrepaired) are indicated by brackets. (E) Quantification of D. Values are the average of three independent experiments and are shown with SD.