Chromatin compartmentalization regulates the response to DNA damage – Nature

Cell culture and treatments

DIvA (AsiSI-ER-U2OS)¹⁹, AID-DIvA (AID-AsiSI-ER-U2OS)²³ and 53BP1-GFP DIvA²⁰ cells were developed in U2OS (ATCC HTB-96) cells and were previously described. Authentication of the U2OS cell line was performed by the provider ATCC, which uses morphology and short tandem repeat profiling to confirm the identity of human cell lines. DIvA, AID-DIvA and 53BP1-GFP-DIvA cells derived from these U2OS cells were not further authenticated. All of the cell lines were regularly tested for absence of mycoplasma contamination using the MycoAlert Mycoplasma (Lonza). All of the cell lines were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% SVF (Invitrogen), antibiotics and either 1 µg ml⁻¹ puromycin (DIvA cells) or 800 µg ml⁻¹ G418 (AID-DIvA and 53BP1-GFP AID-DIvA) at 37 °C under a humidified atmosphere with 5% CO₂. To induce DSBs, U2OS cells were treated with etoposide (Sigma-Aldrich, E1383) for 4 h at 500 nM and DIVA or AID-DIvA cells were treated with 300 nM OHT (Sigma-Aldrich, H7904) for 4 h (supplementary table 1 of ref. ⁵ for positions). For ATM or DNA-PK inhibition, cells were respectively pretreated for 1 h with 20 μM KU-55933 (Sigma-Aldrich, SML1109) or 2 μM NU-7441 (Selleckchem, S2638) and during subsequent OHT treatment. For cell synchronization, cells were incubated for 18 h with 2 mM thymidine (Sigma-Aldrich, T1895), then released during 11 h, followed by a second thymidine treatment for 18 h. S, G2 and G1 cells were then respectively treated with OHT at 0, 6 or 11 h after thymidine release and collected 4 h later. siRNA and plasmid transfections were performed using the 4D-Nucleofector kit and the SE cell line 4D-Nucleofector X L or S kit (Lonza) according to the manufacturer’s instructions, and subsequent treatment(s) were performed 48 h later. siRNA transfections were performed using a control siRNA (siCtrl): CAUGUCAUGUGUCACAUCU; or siRNA targeting SCC1 (siSCC1): GGUGAAAAUGGCAUUACGG; SMC1 (siSMC1): UAGGCUUCCUGGAGGUCACAUUUAA; 53BP1 (si53BP1): GAACGAGGAGACGGUAAUA; SUN2 (siSUN2): CGAGCCTATTCAGACGTTTCA; ARP2 (siARP2): GGCACCGGGUUUGUGAAGU; SETX (siSETX): GAGAGAAUUAUUGCGUACU; and RNASEH1 (siRNASEH1)⁴⁸: CUGUCUUGCUGCCUGUACU. pICE-NLS-mCherry (referred to as pICE-empty in the text; Addgene plasmid, 60364) and pICE-RNase-H1-WT-NLS-mCherry (referred to as pICE-RNase-H1 in the text, Addgene plasmid, 60365) were respectively used as control and for RNase H1 over-expression (gifts from P. Calsou and S. Britton).

Illegitimate rejoining assay

Illegitimate rejoining assays after siRNA transfection were performed at least in triplicate in AID-DIvA cells as described previously³⁸. Illegitimate rejoining assays in synchronized cells were performed in DIvA cells after an OHT treatment (n = 4 biological replicates). The genomic DNA was extracted from 1–5 × 10⁶ cells using the DNeasy Blood & Tissue Kit (Qiagen), according to the manufacturer instructions. Two different illegitimate rejoinings between different AsiSI sites were assessed by qPCR using the following primers: 1Fw, GACTGGCATAAGCGTCTTCG and 1Rev, TCTGAAGTCTGCGCTTTCCA; and 2Fw, GGAAGCCGCCCAGAATAAGA and 2Rev, TCCATCTGTCCCTATCCCCAA. The results were normalized using two control regions, both far from any AsiSI sites and γH2AX domains using the following primers: Ctrl_chr1_82844750_Fw, AGCACATGGGATTTTGCAGG and Ctrl_chr1_82844992_Rev, TTCCCTCCTTTGTGTCACCA; and Ctrl_chr17_9784962_Fw, ACAGTGGGAGACAGAAGAGC and Ctrl_chr17_9785135_Rev, CTCCATCATCGCACCCTTTG. Normalized illegitimate rejoining frequencies were calculated using Bio-Rad CFX Manager v.3.1.

RT–qPCR

RNA was extracted from fresh DIvA cells before and after DSB induction using the RNeasy kit (Qiagen). RNA was then reverse transcribed to cDNA using AMV reverse transcriptase (Promega, M510F). qPCR experiments were performed to assess the levels of cDNA using primers targeting RPLP0 (FW, GGCGACCTGGAAGTCCAACT; REV, CCATCAGCACCACAGCCTTC), RNF19B (FW, CATCAAGCCATGCCCACGAT; REV, GAATGTACAGCCAGAGGGGC), PLK3 (FW, GCCTGCCGCCGGTTT; REV, GTCTGACGTCGGTAGCCCG), GADD45A (FW, ACGATCACTGTCGGGGTGTA; REV, CCACATCTCTGTCGTCGTCC). PPM1D (FW, CTTGTGAATCGAGCATTGGG; REV, AGAACATGGGGAAGGAGTCA), SLC9A1 (FW, TGTTCCTCAGGATTTTGCGG; REV, ATGAAGCAGGCCATCGAGC), LPHN2 (FW, CGATTTGAAGCAACGTGGGA; REV, TGATACTGGTTGGGGAAGGG), UTP18 (FW, TCCTACTGTTGCTCGGATCTC; REV, ATGAAGCAGGCCATCGAGC), IPO9 (FW, CACAGATGCCACTTGTTGCT; REV, TGCTGTACCACGGGAAAGAT) or PARP1 (FW, ATTCTGGACTGGAACACTCTGC; REV, CTGTTCCAGTTTGTTGCTACCG). cDNA levels were then normalized to RPLP0, then expressed as the percentage of the undamaged condition.

Immunofluorescence

DIvA cells were grown on glass coverslips and fixed with 4% paraformaldehyde during 15 min at room temperature. A permeabilization step was performed by treating cells with 0.5% Triton X-100 in PBS for 10 min, then cells were blocked with PBS-BSA 3% for 30 min. Primary antibodies against γH2AX (JBW301, Millipore Sigma, 05-636) were diluted 1:1,000 in PBS-BSA 3% and incubated with cells overnight at 4 °C. After washes in 1× PBS, cells were incubated with anti-mouse secondary antibodies (conjugated to Alexa 594 or Alexa 488, Invitrogen), diluted 1:1000 in PBS-BSA 3%, for 1 h at room temperature. After staining with DAPI, Citifluor (Citifluor, AF-1) was used for coverslip mounting. Images were acquired with the software MetaMorph, using the ×100 objective of a wide-field microscope (Leica, DM6000) equipped with a camera (DR-328G-C01-SIL-505, ANDOR Technology). The spatial analysis (Ripley function) of γH2AX spot distribution was performed as described previously²⁰ using the spot detector plug-in in Icy.

High-content microscopy

Transfected DIvA and U2OS cells were plated in 96-well Cell Carrier plates (Perkin Elmer). Cells were subjected to OHT or etoposide treatments and an immunofluorescence as described above using anti-mouse Alexa 488 secondary antibodies (Invitrogen) for γH2AX detection. Cells were stained with 5 µg ml⁻¹ Hoechst 33342 (Invitrogen) for 30 min at room temperature. γH2AX foci were analysed using the Operetta CLS High-Content Imaging System (Perkin Elmer) and Harmony software (v.4.9). For quantitative image analysis, 81 fields per well were acquired with a ×40 objective lens to visualize around 4,000 cells per well in triplicate. Subsequent analyses were performed with Columbus software (v.2.8.2). Determination of the number (y axis) and size of γH2AX foci (mean area on the x axis) in each nucleus was been used to infer clustering (clustering leads to a low number of large foci) (>8,000 nuclei analysed per sample). Scatter plots are divided into four quadrants on the basis of the median number and size of foci and the percentages of cell populations that are cluster positive (bottom right quadrant) or cluster negative (top left quadrant) are determined.

3D RIM super-resolution acquisition and reconstruction

3D + t live RIM was performed on AID-DIvA cells expressing 53BP1–GFP²⁰ using an upgrade of the system and method described previously⁴⁹. In brief, 60 3D images were acquired during 47 min using an inverted microscope (TEi Nikon) equipped with a ×100 magnification, 1.49 NA objective (CFI SR APO 100XH ON 1.49 NIKON) and SCMOS camera (ORCA-Fusion, Hamamatsu). The temporal resolution of all of the Supplementary Videos is 40 s. A commercial acquisition software (INSCOPER SA) enables a whole-cell single-timepoint 3DRIM acquisition in only 6 s under a low-photobleaching regime (1 W cm⁻²). Fast diode lasers (Oxxius) with the wavelengths centred at 488 nm (LBX-488-200-CSB) were used to produce a TEM00 2.2-mm-diameter beam. The polarization beam was rotated with an angle of 5° before hitting a X4 Beam Expander beam (GBE04-A) and produced a 8.8 mm TEM00 beam. A fast spatial light phase binary modulator (QXGA fourth dimensions) was conjugated to the image plane to create 48 random illumination by each plane as described previously⁴⁹. 3D image reconstruction was then performed as described previously⁴⁹ and at GitHub (https://github.com/teamRIM/tutoRIM).

3D film editing

Bleaching correction was performed after RIM reconstruction using open-source FIJI software (https://imagej.net/software/fiji/) based on exponential FIT from the background signal. The 3D drift correction FIJI plugin was performed for 3D registration (https://github.com/fiji/Correct_3D_Drift). 3D + t video rendering was generated using the VTK library implemented in ICY software (https://icy.bioimageanalysis.org/) from a 3D crop of the area of interest.

FRAP measurements

FRAP experiments were performed on 53BP1–GFP foci using a Zeiss LSM 710 confocal light scanning microscope (Carl Zeiss) equipped with a ×63/1.2 NA oil-immersion objective. Typically, images were acquired at 512 × 256 px at a scan speed corresponding to 200 ms per image, and 300 images were acquired over 2 min, with an interval of 500 ms between subsequent images. Before photobleaching half of the foci, three images were recorded.

For each experiment, a custom R script was used to segment the image, track the bleached foci and retrieve the average intensity of the bleached half (I_B), the non-bleached half (I_NB), the background (I_BG), and the rest of the nucleus (I_REF) at each frame. These intensity values were used to calculate FRAP curves for the bleached half (FRAP_B) and the non-bleached half (FRAP_NB), according to ref. ²⁴:

Here, A represents unwanted bleaching in the non-bleached half. FRAP_B and FRAP_NB were multiplied by the size of their respective regions of interest (ROIs) (N_B and N_NB, respectively) to obtain curves that are proportional to the number of particles in each half

The curves were then normalized with respect to the number of bleached molecules:

Here, t_pre and t_bleach are the acquisition times of the last frame before the bleach and the first frame after the bleach, respectively. The resulting FRAP curves are proportional to the ROI sizes and double-normalized. Finally, an additive offset was applied to the signal in the non-bleached half to normalize to unity before the bleach.

The resulting curves reflect the change in the number of labelled molecules in each half. In the presence of an immobile fraction of molecules that do not move during the course of the experiment because they tightly bind to immobile binding sites, the signal in both halves will not recover to the same level, but there will be an offset between them that corresponds to the immobile fraction X_immobile. To correct for those immobile molecules, the FRAP curves are modified according to:

Finally, to determine the presence or absence of an interfacial barrier in the foci, the curves were compared to half-FRAP experiments performed in a solution of freely diffusing poly-lysine-fluorescein (0.1 mg ml⁻¹) and to half-FRAP experiments of 53BP1-GFP molecules diffusing in the nucleoplasm (adjacent to foci). In this case, the maximum decrease in normalized fluorescence down to a value of 0.90 ± 0.03 indicates the absence of a barrier (horizontal black lines in Fig. 2f,g) while larger decreases indicate the presence of an interfacial barrier.

DNA damage induction by laser irradiation

DIvA cells were transfected with a plasmid encoding the macrodomain of macroH2A1.1 fused to mKate2 (macro–mKate2)⁵⁰ and treated with OHT for 4 h. For microirradiation, a circular spot with a radius of 1 µm was selected in the nucleoplasm where 53BP1–GFP foci were absent. Then, continuous illumination with a 405 nm laser at maximum power was applied for 1 min. The recruitment of 53BP1–GFP and macro–mKate2 was observed during the next 10 min. Half-FRAP experiments at these laser-induced 53BP1–GFP foci were performed as described above after 5 min of their formation.

53BP1 focus fusion quantification

Fusion events were selected from the 3D images acquired by RIM. During the fusion, the foci were segmented and the normalized aspect ratio, or eccentricity, was calculated as AR_norm = (R_max − R_min)/(R_max + R_min), where R_max and R_min are the longest and shortest radius, respectively. The normalized aspect ratio over time was then fitted with an exponential function with an additive offset to obtain the relaxation constant (t_R) and the aspect ratio of the end product after fusion. t_R values obtained from five different experiments were plotted versus the size of the respective foci after the fusion and fitted with a linear function, the slope of which is the inverse capillary velocity.

RNAscope

RNAscope enables the visualization of transcription sites^51,52. DIvA cells were grown on Chambered Cell Culture Slides (Corning Falcon, 08-774-25) and fixed with 4% paraformaldehyde during 15 min at room temperature. The RNAscope assay was performed using the RNAscope Multiplex Fluorescent Kit v2 kit (ACDBio, 323100) according to the manufacturer’s instructions. In brief, fixed cells were pretreated with RNAscope Hydrogen Peroxide for 10 min, then permeabilized with RNAscope Protease III (1:15 diluted) for 10 min. Cells were then incubated with the probes RNAscope Probe-Hs-GADD45A-C1 (ACDBio, 477511) and RNAscope Probe-Hs-CCL2-C2 (ACDBio, 423811-C2) or RNAscope Positive Control Probe-Hs-PPIB-C2 (ACDBio, 313901-C2), or with the intronic probes RNAscope Probe-Hs-PLK3-intron-C1 (ACDBio #1263411) and RNAscope Probe-Hs-CDC42-intron-C2 (ACDBio, 1263101) or RNAscope Probe-Hs-ADGRL2-intron-C2 (ACDBio, 1263111) in a HybEZTM Oven at 40 °C for 2 h. Signal-amplification steps were performed, followed by the development of the HRP-C1 and HRP-C2 signals, using Opal dye 620 and 690 (Akoya Biosciences) diluted to 1:750. Finally, immunofluorescence was performed as described above, without an additional permeabilization step and using γH2AX (JBW301, Millipore) and DAPI staining. Images were acquired with the software Micro-Manager, using the ×40 or ×60 objective of a spinning-disk/high-speed widefield CSU-W1 microscope, equipped with an Andor Zyla sCMOS camera. The colocalization between γH2AX foci and RNA foci was measured using Cell Profiler. Two foci were considered to be colocalizing when part of their areas was overlapping and if their respective centroids were separated by less than 1 μm.

Western blot

AID-DIvA cells were incubated in RIPA buffer (50 mM Tris at pH 8, 150 mM NaCl, 0.5% deoxycholate, 1% NP-40, 0.1% SDS) on ice for 20 min and centrifuged at 13,000 rpm for 10 min. The supernatant, containing soluble protein extracts, was then mixed with SDS loading buffer and reducing agent, resolved on 3–8% NuPAGE Tris-acetate gels (Invitrogen) and transferred onto PVDF membranes (Invitrogen) according to the manufacturer’s instructions. For RNase H1 expression, 0.5 × 10⁶ DIvA cells were incubated 15 min at room temperature with 625 U of GENIUS Nuclease (Santa Cruz Biotechnology, sc-202391) in SDS loading buffer. After adding a reducing agent, the samples were heated at 95 °C for 5 min and loaded onto a NuPAGE 4–12% Bis-Tris gel in MOPS SDS running buffer and transferred onto a PVDF membrane (Invitrogen) with the Trans-Blot Turbo Transfer System according to the manufacturer’s instructions (Bio-Rad). PVDF membranes were incubated in TBS containing 0.1% Tween-20 (Sigma-Aldrich, P1379) and 3% non-fat dry milk for 1 h for blocking, followed by overnight incubation at 4 °C using primary antibodies targeting SUN2 (Abcam, ab124916, 1:1,000), ARP2 (Abcam, ab128934, 1:1,000), 53BP1 (Novus Biologicals, NB100-305, 1:1,000), RNase H1 (Invitrogen, PA5-30974, 1:1,000), SCC1 (Abcam, ab992, 1:500), SMC1 (Abcam, ab75819, 1:1,000), myosin I-β (Sigma-Aldrich, M3567, 1:1,000), GAPDH (Sigma-Aldrich, MAB374, 1:10,000) or α-tubulin (Sigma-Aldrich, T6199, 1:10,000). The corresponding mouse or rabbit horseradish-peroxidase-coupled secondary antibodies were used at 1:10,000 to reveal the proteins (Sigma-Aldrich, A2554 and A0545), using a luminol-based enhanced chemiluminescence HRP substrate (Super Signal West Dura Extended Duration Substrate, Thermo Fisher Scientific). Picture acquisition of the membranes was performed using the ChemiDoc Touch Imaging System and pictures were visualized using Image Lab Touch software.

RNA-seq

RNA-seq in DIvA cells was performed as described previously³⁸. Raw sequencing data were mapped in paired-end to a custom human genome (hg19 merged with ERCC92) using STAR. Count matrices were extracted using htseq-count with union as resolution-mode and reverse-strand mode. Differential expression analysis was performed on the count matrix using edgeR with two replicates per condition (with or without 4 h OHT treatment) and differential genes were determined using log-ratio test. Whole-genome coverage was computed using the bamCoverage command form deeptools to generate bigwig from BAM files (without PCR duplicate suppression). Differential coverage between two conditions was performed using BamCompare from deeptools with setting the binsize parameter at 50 bp. The log₂-transformed fold change was calculated using edgeR in differential expression analysis. Using a cut-off adjusted P value of 0.1 and a log₂-transformed fold change of 0.5 (~41% increase/decrease of expression), we were able to determine 286 upregulated and 125 downregulated genes with 11 of them directly damaged by a DSB. On chromosomes 1, X and 17, n = 35 genes were found to be upregulated and n = 1,839 not regulated (Fig. 3d). On chromosomes 1, 2, 6, 9, 13, 17, 18, 20 and X, n = 86 genes were found upregulated and n = 3,829 not upregulated (Extended Data Fig. 5g). For further classification in the D and non-D compartment, to analyse enough genes, the cut-off for upregulation was set at a log₂-transformed fold change of 0.3 (see below).

qDRIP–seq

qDRIP–seq was adapted from a previous study³⁹. In brief, 2.5 × 10⁶ of trypsinized DIvA cells were mixed with 1.67 × 10⁶ Drosophila S2 cells and lysed overnight at 37 °C in TE buffer containing 0.5% SDS and 800 µg proteinase K (Roche, 03115828001). DNA was extracted by phenol–chloroform extraction using phase-lock tubes (Qiagen, MaxTract, 129065) followed by ethanol precipitation. DNA was resuspended on ice in 130 µl TE buffer before sonication in the Covaris S220 system (microtubes, PN520045) to obtain ~300 bp DNA fragments (Covaris S220, 140 W peak incident power, 10% duty factor, 200 cycles per burst for 80 s). Immunoprecipitation was performed in triplicate by incubating 4 µg of sonicated DNA with 10 µg of S9.6 antibody (Antibodies Incorporation) in 1× binding buffer (10 mM NaPO₄, 140 mM NaCl, 0.05% Triton X-100) overnight and at 4 °C. Agarose A/G beads (Thermo Fisher Scientific, 20421) were added to samples for 2 h at 4 °C. Beads were washed four times in binding buffer followed by incubation with elution buffer (50 mM Tris pH 8, 10 mM EDTA, 0.5% SDS, 0.3 µg µl⁻¹ proteinase K) for 45 min at 55 °C. The samples were subjected to phenol–chloroform extraction and ethanol precipitation, and were resuspended in low EDTA TE. Sequencing libraries were prepared using the Swift ACELL-NGS 1S Plus kit according to the manufacturer’s instructions using 12 PCR cycles. Libraries were pooled at equimolar concentrations and sequenced using the Illumina NextSeq 500 system with 75 paired-end reads.

qDRIP–seq data were processed using a custom pipeline taking into account stranded and spike-in library preparation. In brief, reads were trimmed using Trimmomatic (v.0.39)⁵³ to remove remaining primers from the library. BWA-MEM was used for mapping reads to a custom reference genome merging hg19 and dm6 (spike-in) chromosomes. Samtools was used to generate BAM files with reads based on their mapping location (hg19 or dm6). Strand-specific data were generated using Samtools view and merge with flags filters: 80;160 for reverse fragments, and 96;144 for forward fragments. BAM files were then sorted, indexed and duplicates were removed. Bigwig files were generated on these data, normalized to total read counts (counts per million) or by the number or reads mapped on dm6 (spike-in). Differential coverage between two conditions was performed using BigWigCompare from deeptools with the substract setting⁵⁴ and with setting bin size parameter at 50 bp. Narrow peaks were detected using macs3 callpeak algorithm⁵⁵ on qDRIP bams using -q 0.1, and by taking only good quality peaks with a score (fold-change at peak summit) at least superior to 100.

DRIP–qPCR

DRIP–qPCR was performed as described previously³⁸ using primers for RPL13A (FW, AATGTGGCATTTCCTTCTCG; REV, CCAATTCGGCCAAGACTCTA) and PLK3 (FW, CGGAGCAGAGGAAGAAGTGA; REV, CATGCATGAACAGCCCATCA).

Amplicon–seq

AID-DIvA cells were treated with or without 300 nM OHT for 4 h followed by treatment with indole-3-acetic acid for 14 h to degrade AsiSI²³. The genomic DNA was extracted from 5 × 10⁶ cells using the DNeasy Blood & Tissue Kit (Qiagen) according to the manufacturer’s instructions. Genomic DNA was then used in a multiplex PCR reaction that amplified 25 target sites: 20 AsiSI cut sites and 5 uncut control sites (Extended Data Table 1). Amplicons were size-selected using SPRIselect beads (Beckman, B23318) and processed for DNA library preparation using the NEBNext Ultra II kit (NEB, E7645L). Libraries were pooled at equimolar concentrations and sequenced using the Illumina NextSeq 500 system with paired-end 150 cycles. The data were analysed using our custom tool mProfile, available at GitHub (https://github.com/aldob/mProfile). This identified the genomic primers used in the original genomic PCR reaction to amplify each read in the pair. Translocated reads were therefore identified as those where each read in a pair was amplified by a different primer set, and this was normalized to the total reads that were correctly amplified by these primer sets. The heat map of illegitimate rejoining/translocation events was made between 20 DSBs by computing the ratio between the +DSB and −DSB sample for each pair of DSBs and comparing the log₂-transformed ratio distribution between each condition and control. Significance was computed using nonparametric Wilcoxon tests.

4C-seq

4C-seq experiments performed in synchronized cells, before and after DSB induction, were performed as described previously⁷. In brief, 10–15 × 10⁶ DIvA cells per condition were cross-linked, lysed and digested with MboI (New England Biolabs). DNA ligation was performed using the T4 DNA ligase (HC) (Promega), and ligated DNA was digested again using NlaIII (New England Biolabs). Digested DNA was religated with the T4 DNA ligase (HC) (Promega) before proceeding to 4C-seq library preparation. A total of 16 individual PCR reactions was performed to amplify around 800 ng of 4C-seq template, using inverse primers including the Illumina adaptor sequences and a unique index for each condition (Extended Data Table 2). Libraries were pooled and sent to a NextSeq 500 platform at the I2BC Next Generation Sequencing Core Facility (Gif-sur-Yvette).

4C-seq data were processed as described previously⁷. In brief, BWA-MEM was used for mapping and Samtools was used for sorting and indexing. A custom R script (https://github.com/bbcf/bbcfutils/blob/master/R/smoothData.R) was used to build the coverage file in bedGraph format, to normalize using the average coverage and to exclude the nearest region from each viewpoint. Differential 4C-seq data were computed using BamCompare from deeptools with binsize=50 bp. The average of total trans interactions between viewpoints and DSBs was then computed using a 1 Mb window around the breaks (80 best) and after exclusion of viewpoint–viewpoint (cis) interactions.

Hi-C

Hi-C data obtained before and after DSB induction and after control or SCC1 depletion in DIvA cells were retrieved from a previous study⁷. Hi-C experiments with or without DSB induction and after ATM or DNA-PK inhibition, or after transfection with control or SETX siRNAs were performed in DIvA cells as described previously⁷. In brief, 10⁶ cells were used per condition. Hi-C libraries were generated using the Arima Hi-C kit (Arima Genomics) according to the manufacturer’s instructions. DNA was sheared to an average fragment size of 350–400 bp using the Covaris S220 system and sequencing libraries were prepared on beads using the NEB Next Ultra II DNA Library Prep Kit for Illumina and NEBNext Multiplex Oligos for Illumina (New England Biolabs) according to the instructions of the Arima Hi-C kit.

Hi-C data analyses

Hi-C heat maps

Hi-C reads were mapped to the hg19 genome and processed using Juicer with the default settings (https://github.com/aidenlab/juicer). Hi-C count matrices were generated using Juicer at multiple resolutions: 100 kb, 50 kb, 25 kb, 10 kb and 5 kb. Hi-C heat map screenshots were generated using Juicebox (https://github.com/aidenlab/Juicebox/wiki/Download). Aggregate heat maps were computed on a set of submatrices extracted from the originally observed Hi-C matrices at 50 kb resolution or 100 kb resolution. The region of 5 Mb around DSBs (80 best) was extracted and then averaged. The log₂-transformed ratio was then computed using Hi-C counts (+DSB/−DSB) and plotted as heat maps.

Cis contact quantification

For cis contact quantification, interactions within γH2AX domains (−0.5/+0.5 Mb around the 80 best DSBs) were extracted from the observed Hi-C matrix at 100 kb resolution, and the log₂-transformed ratio was computed on damaged versus undamaged Hi-C counts (+DSB/−DSB). Adjacent windows (−1.5 Mb–0.5 Mb and +0.5 Mb–1.5 Mb around 80 best DSBs) were retrieved to quantify interactions between damaged domains and adjacent undamaged domains.

Trans contact quantification

To determine interaction changes in trans (interchromosomal), we built the whole-genome Hi-C matrix for each experiment by merging together all chromosome–chromosome interaction matrices using Juicer and R. The result is a genome matrix with 33,000 × 33,000 bin interactions for 100 kb resolution. Interactions between bins inside damaged TADs (240 × 240 for 80 DSBs) were extracted and counted for each condition, the log₂-transformed ratio was calculated on normalized counts (counts per million), and plotted as box plots or heat maps. For the box plots, the centre line shows the median; the box limits show the first and third quartiles; the whiskers show the maximum and minimum values without outliers; and the points show the outliers. For the heat maps, each tile corresponds to log₂[+DSB/−DSB] between DSB (100 kb bins within ±1 Mb regions were averaged). They were further sorted on the basis of the 53BP1 ChIP–seq level (Extended Data Fig. 7f), according to previously determined homologous-recombination-prone and non-homologous-end-joining-prone DSBs⁵ (Extended Data Fig. 2d), or to the level of PC1 determined by applying PCA in Hi-C DIvA to identify A/B compartments (Extended Data Fig. 4d).

APA on endogenous breaks

Endogenous breaks were identified by calling peaks on pATM ChIP–seq data without OHT treatment⁷ using macs2 with -q 0.01, giving 1,206 narrow peaks. Of these peaks, only interchromosomal pairs (trans) were retained using a BEDPE format file. Random positions were generated using the gkmsvm package and pairs of interactions were built according to the same procedure. APA was then performed on these trans pairs using juicertools.

TAD cliques

TAD cliques were computed using the igraph R package on an undirected graph representing DSB clustering. This graph was computed on the differential Hi-C matrix (+DSB/−DSB) counts, at 500 kb resolution, considering a change of around 86% of interaction (0.9 in log₂) between two DSBs as a node on the graph. Averaged ChIP–seq signal values (53BP1/γH2AX/H1/Ubiquitin FK2) were then computed for each category of cliques using 500 kb windows around DSBs. For prior RNA polymerase II occupancy, the signal was computed on 10 kb around DSBs.

A/B compartment

To identify the two mains chromosomal compartments (A/B), the extraction of the first eigenvector of the correlation matrix (PC1) was performed on the observed/expected matrix at 500 kb resolution using the juicer eigenvector command. The resulting values were then correlated with the ATAC-seq signal to attribute positive and negative values to the A and B compartment, respectively, on each chromosome. The observed/expected bins were arranged on the basis of the PC1 values and aggregated into 21 percentiles, to visualize A–B interactions on our experiments (saddle plots).

D compartment

To identify the D compartment, we retrieved the first component (PC1) of a PCA made on the differential observed Hi-C matrix \({\log }_{2}\left(\frac{{\rm{damaged}}}{{\rm{undamaged}}}\right)\) at 100 kb resolution. Each matrix was extracted from the .hic files using Juicer and the ratio was computed bin per bin. Pearson correlation matrices were then computed for each chromosome and PCA was applied on each matrix. The first component of each PCA was then extracted and correlated with the positions of DSB. A PC1 showing a positive correlation with DSB was then called the D compartment and PC1 showing a negative correlation with DSBs were multiplied by −1. We were able to extract the D compartment on chromosomes 1, 17 and X for +DSB/−DSB and chromosomes 1, 2, 6, 9, 13, 17, 18, 20 and X for +DSB/−DSB in the DNA-PKi condition. The D compartment (first component of the PCA) was converted into a coverage file using the rtracklayer R package. Using the same package, D compartment value were computed around DSBs and genes at 100 kb resolution.

Determination of D genes

First, all genes embedded in γH2AX domains (that is, −1 Mb/+1 Mb around the 80 DSBs) were removed from the gene set. Under normal conditions (without DNA-PKi), on chromosomes 1, 17 and X, n = 493 genes displaying a positive PC1 value on their entire length in each of the three Hi-C replicates experiments were identified as ‘genes in compartment D’ genes, while n = 346 genes displaying a negative value in each of the three Hi-C replicates experiments were labelled as ‘genes not in compartment D’. In the presence of DNA-PKi, where the D compartment was identified on chromosomes 1, 2, 6, 9, 13, 17, 18, 20 and X, n = 2,161 were found in compartment D, while n = 2,112 were not. The genes were further categorized according to their upregulation after DSB induction (fold change > 0.3, unexpressed genes filtered), giving four categories. Without DNA-PKi, upregulated/compartment D (n = 40); upregulated/no compartment D (n = 32); downregulated and not regulated/compartment D (n = 453); downregulated and not regulated/no compartment D (n = 314).

Transcription factor motif analysis

Transcription-factor-binding motifs were extracted on the promoter regions (−500 bp of the transcription start site (TSS)) of genes with positive value of D compartment (2,161) versus genes with negative value (2,112) identified on chromosomes 1, 2, 6, 9, 13, 17, 18, 20 and X from the PCA analysis of Hi-C (+DSB + DNA-PKi/−DSB) using motifmatchr and TFBSTools R packages on the JASPAR2020 database. Motifs were sorted by significance using Fisher’s exact test and adjusted using the Benjamini–Hochberg procedure between motifs found on gene inside the D compartment versus genes outside D compartment.

Correlation with DRIP–seq or qDRIP datasets

To assess a correlation between R-loop accrual and D-compartment formation, R-loop levels obtained from DRIP–seq experiments performed in DIvA cells³⁸ were computed on extended gene bodies (±2 kb TTS) and plotted as a box plot for categories (Figs. 3f and  4c). Conversely, to establish whether R-loop enriched genes, display higher levels of differential CEV (D compartment) signal (Extended Data Fig. 6c), R-loop levels were computed on all genes of chromosome 1, 17 and X (±1 kb of the TTS) and further categorized into 10 groups (based on percentiles). The D-compartment signal was compared between the highest (n = 180) and lowest (n = 190) groups.

For qDRIP experiments, given that the signal accumulates on narrow peaks within genes (in contrast to DRIP–seq), we used the identified locations of narrow peaks (see the ‘qDRIP–seq’ section) inside the genes in each category (±2 kb of the TTS) (Fig. 3h). We were able to identify 83 peaks on upregulated D genes, 313 peaks on not upregulated D genes, 30 peaks on upregulated non-D genes and 457 peaks on not-upregulated non-D genes.

Correlation with SETX ChIP-seq data

ChIP-seq SETX data in DIvA cells were from a previous study³⁸. To assess SETX accruals at R-loops, BigWig coverage files were used to get SETX ChIP-seq coverage on gene bodies (±2 kb of the TTS) that overlap with qDRIP peaks (see above) (Fig. 3j).

Translocation breakpoints

For translocation breakpoints, data from a previous study⁴² were retrieved, and only breakpoints for interchromosomal structural variants were selected (n = 28,051). Genes reproducibly enriched in the D compartment in the three biological replicates, on chromosomes 1, 17 and X as well as genes not enriched in the D compartment were retrieved. The significance of the overlap between genes and breakpoints was determined using the regioneR package⁵⁶ by using resampling test with PermTest. In brief, we selected 1,000 times a control set of genes, with the same size and on the same chromosome as our original gene set. We tested the overlap between each gene and breakpoints to determine a distribution of the number of overlaps between control set and breakpoints. We further tested whether the overlap between our gene set (D compartment or non-D compartment) and breakpoints was significant by counting the number of times more overlaps occurred in the control set than in our gene set.

GenomicRanges, plyranges, tidyverse, patchwork, ggforce, ggside and ggtext were used to read, manipulate and visualize genomic data in R and produce figures. Bedtools was used to manipulate genomic location and produce bed or bigwig files. Integrated Genome Browser was used to visualize bed and bigwig files. All of the box plots show the median (centre line), first and third quartiles (box limits), maximum and minimum without outliers (whiskers) and outliers (points).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.