ATM Regulation of the Cohesin Complex Is Required for Repression of DNA Replication and Transcription in the Vicinity of DNA Double-Strand Breaks

This article is featured in Highlights of This Issue, p. 187

Cellular DNA is constantly challenged by genotoxic stresses that result in tens of thousands of DNA lesions per day (1, 2). DNA double-strand breaks (DSB) are one of the most cytotoxic DNA lesions as an inability to efficiently respond to DSBs can lead to alteration or loss of genetic information, which can ultimately result in disease or cancer formation. The serine/threonine protein kinase, ataxia-telangiectasia mutated (ATM), coordinates cellular responses to DSBs by phosphorylating numerous effector proteins regulating cell-cycle checkpoint activation, DNA repair pathways, and cell death signaling pathways (3, 4). ATM signaling at DSBs rapidly alters the local chromatin environment surrounding DSBs through posttranslational modification of histones and other chromatin-bound proteins. These ATM-dependent modifications stimulate chromatin remodeling, allowing DNA repair proteins to access the damaged DNA while simultaneously suppressing other DNA metabolic processes such as transcription and origin of replication firing, which could otherwise negatively interfere with the repair process (5–7). One of the chromatin bound proteins regulated by ATM at DSBs previously identified and functionally characterized by our lab and others is SMC1A, a component of the cohesin complex (8–10).

The cohesin complex is composed of SMC1A, SMC3, RAD21, and SA1 or SA2 which forms a ring-like structure extruding DNA through its lumen (11). Cohesin has a dynamic association with chromatin, with its constant loading regulated by NIPBL/MAU2, and unloading regulated by WAPL and PDS5A or PDS5B (12, 13). Cohesin participates in almost every chromatin metabolic process including DNA replication, gene expression, establishment of higher order chromatin architecture, and promoting responses to DNA damage (8–10, 14–19). Following ionizing radiation (IR) exposure, ATM rapidly phosphorylates SMC1A on S957 and S966 and SMC3 on S1083; phosphorylation of these sites is required for chromosomal stability and cell survival after IR exposure and for effective cell-cycle checkpoint activation (8–10, 18).

In addition to facilitating responses to DSBs, ATM has regulatory roles outside of the DNA damage response, functioning in RNA splicing, cell metabolism, and immune responses (20–22). Mutations in ATM result in the disease Ataxia-Telangiectasia (A-T), characterized by cancer predisposition, neurodegeneration, and radiation sensitivity (23). To better understand how ATM regulates these varied biological pathways and how loss of ATM signaling gives rise to A-T, we conducted two complementary unbiased phosphoproteomic and proteomic screens to identify novel ATM substrates. From these screens, we found that ATM phosphorylates the cohesin proteins RAD21, WAPL, NIPBL, PDS5B, and PDS5A in response to DSB formation, in addition to the previously described SMC1A and SMC3 proteins. Introduction of a phospho-silent mutation of the ATM phosphorylation site into PDS5A results in decreased cell survival after IR and increased chromosomal instability, indicating that PDS5A is another critical factor in the ATM-mediated DNA damage response. Furthermore, using a system that introduces site-specific DNA breakage, we find that ATM phosphorylation of cohesin proteins is required for suppression of both RNA transcription and DNA replication within the physical vicinity of a DSB, suggesting that ATM regulation of cohesin represses both of these DNA metabolic processes around a break.

All experiments were repeated at least three times as described in figure legends.

HEK293T, U2OS, and U2OS 2–6-3 (Fok1 reporter) cells were cultured in DMEM + 10% FBS. Plasmid transfections were performed with Lipofectamine 2000. siRNA transfections were performed with RNAi Max. The U2OS 2–6-3 cells were generously provided by Dr. Roger Greenberg (University of Pennsylvania). All cell lines and their derivates (including CRISPR manipulated cells) were authenticated using certified human cell line authentication analysis provided by the Duke University DNA analysis facility and tested monthly for Mycoplasma contamination.

PDS5A cDNA was a gift from Elphege Nora (pCAGGS-3XFLAG-mKate2-(human) PDS5A-bpA Addgene plasmid # 156449; http://n2t.net/addgene:156449; RRID:Addgene_156449; ref. 24). myc-BioID2-MCS was a gift from Kyle Roux (Addgene plasmid # 74223; http://n2t.net/addgene:74223; RRID:Addgene_74223; ref. 25).

The following drugs were used: ATMi (KU60019, 1 μmol/L), ATRi (AZD6738, 1 μmol/L), DNAPKi (M3814, 1 μmol/L), Shield-1 (1 μmol/L), and 4-hydroxytamoxifen (4-OHT; 1 μmol/L).

U2OS 2–6-3 Fok1 D450A cells were made by lentiviral infection and selection for the linked puromycin resistance cassette.

The following antibodies were used: Phospho-ATM/ATR Substrate Motif (pS/TQ; 1:1,000; Cell Signaling Technology, #6966), NIPBL (1:1,000; Santa Cruz Biotechnology, sc-374625), WAPL (1:1,000; Cell Signaling Technology, 77428S), PDS5A (1:1,000; Bethyl, A300–089A), PDS5B/APRIN (1:1,000; Novus, NB100–755), Flag (1:1,000; Sigma, F1804), SMC1A (1:1,000, Bethyl, A300–055A), pSMC1A S957 [1:1,000 (WB), 1:200 (IF); Cell Signaling Technology, 4805S), 53BP1 [1:200 (IF), Bethyl, A300–237A]; γH2AX [1:200 (IF); Millipore 05–636], ATM (1:1,000; Sigma, A1106), pATM S1981 (1:1,000, Rockland, 200–301–400), CHK2 (1:1,000, Cell Signaling Technology, 2662), pCHK2 T68 (1:1,000; Cell Signaling Technology, 2661), DNAPK (1:1,000; Abcam, ab70250), pDNAPK S2056 (1:1,000; Abcam, ab18192), CHK1 (1:1,000, Epitomics, 2865–1), pCHK1 S345 (1:1,000, Epitomics, S0660). For immunoblotting proteins were transferred to nitrocellulose after separation on an acrylamide gel, blocked in 5% milk in TBS/tween. Primary antibodies were diluted at the indicated dilutions above in 5% BSA in TBS/t. Proteins were visualized by staining with Alexfluor 700 or 800 conjugated secondary antibodies on an Odyssey CLx imaging system.

U2OS 2–6-3 cells were treated with 1 μmol/L 4-OHT and 1μmol/L Shield-1 for 1 hour before addition of either 1 μg/mg doxycycline for transcriptional repression assays or 10 μmol/L EdU for DNA replication repression assays. Cells were fixed in 3% paraformaldehyde, permeabilized with 0.5% Triton X-100 in PBS, and blocked with 5% BSA in PBS. To visualize EdU incorporation, EdU was conjugated to Alexa Fluor 488 Azide using click chemistry. For immunofluorescence experiments, cells were incubated in primary antibodies diluted in 1% BSA in PBS at the concentrations described above, then incubated with Alexa Fluor 488 secondary antibodies diluted in 1% BSA in PBS. Slides were imaged on a Zeiss Axio Observer with fixed camera exposure times. Yellow fluorescent protein (YFP), Fok1, EdU, DAPI, pSMC1 S957, and γH2AX intensities were all determined using Cell Profiler software.

HEK293T cells were left untreated or pretreated with 0 or 1 μmol/L ATMi for 30 minutes before irradiation with 10 Gy IR. Cells were collected 1 hour after irradiation, lysed in 8 mol/L urea with 200 mmol/L dithiothreitol and proteins alkylated with 20 mmol/L iodoacetamide before digestion with TPCK-treated trypsin (25:1) overnight. Digested peptides were cleaned up using solid phase extraction and subsequently lyophilized to dryness before immunoprecipitation with ATM/ATR Substrate Motif (pSQ) immunoaffinity beads (Cell Signaling Technology) according to the manufacturer's protocol. Following IP, phosphopeptides were further purified using TiO₂ resin (GL Sciences). Peptides were analyzed using 90-minute LC-MS/MS on Fusion Lumos Mass Spectrometer (Thermo Scientific). The raw data was searched on Mascot v 2.5.1 using the SwissProt database with human taxonomy selected, fixed modification of peptides with carbamidomethylated (C), and variable modification of peptides with deamidated (NQ) and phospho (STY). Semitryptic cleavage rules were allowed to increase the number of peptides found, and up to two missed cleavages were allowed along with tolerances of ± 5 ppm precursor and ± 0.02 Da product.

HEK293T cells were transfected with BioID2-ATM. The following day cells were left untreated or treated with 0 or 1 μmol/L ATMi then irradiated with 10 Gy IR. Cells were labeled with 50 μmol/L biotin for 16 hours before harvesting and lysis in RIPA buffer (150 mmol/L NaCl, 1% Igepal, 0.5% NaDeoxycholate, 0.1% SDS, 50 mmol/L Tris-Cl pH 8.0). A total of 6 mg of biotinolyated lysate was combined with streptavidin dynabeads and rotated at 4°C overnight. Protein-bound beads were washed four times with RIPA lysis buffer and then eluted in elution buffer (200 mmol/L NaCl, 100 mmol/L Tris pH 8.0, 2%SDS, and 1 mmol/L Biotin) for 30 minutes at 60°C.

293T cells were collected, lysed in NP40 lysis buffer (150 mmol/L NaCl, 1% NP40, 50 mmol/L Tris-Cl pH 8.0), and 2 mg of lysate was coincubated with 4 μg of antibody conjugated to protein A or G Dynabeads. Antibody-bound beads were washed four times with NP40 lysis buffer proteins eluted by boiling in 1X NuPAGE LDS sample buffer (Invitrogen). Immunoprecipitates were then separated by SDS-PAGE and proteins detected by immunoblotting.

PDS5A and PDS5A S1278A transfected or siRNA transfected 293T cells were left untreated or irradiated 0.5 Gy IR and allowed to recover for 1 hour before addition of 150 ng/mL colcemid to enrich for mitotic cells. Metaphase spreads were prepared 4 hours after colcemid addition as previously described. Chromosome spreads were mounted and stained in prolong gold mounting medium with DAPI and imaged on a Zeiss Axio Observer. At least 750 chromosomes were counted per condition/per experiment using blinded samples.

Viability assays were performed in U2OS cells transfected with PDS5A or PDS5A S1278A cDNAs or NIPBL, WAPL, PDS5A, and PDS5B siRNAs. Cells were plated the morning after transfection, and irradiated later that day with 0, 2, 4, or 6 Gy IR. Colonies were scored approximately 2 weeks by staining with methylene blue (50% methanol, 48% water, 2% methylene blue).

Raw data for the proteomic screens were generated at Duke Proteomics and Metabolomics core. Derived data supporting the findings of this study are available within the article and its Supplementary Data files.

To identify novel substrates of ATM, we conducted two unbiased proteomic screens. In the first screen, HEK293T cells were pretreated with 0 or 1 μmol/L ATM inhibitor (ATMi, KU60019) and exposed to 0 or 10 Gy IR to induce DSBs (Fig. 1A). Cells were harvested 1 hour after irradiation, lysed, and proteins proteolytically degraded by trypsin. Using an approach similar to previous reports (3, 4), antibodies that recognize a phosphorylated serine or threonine followed by a glutamine (pS/TQ), the specific phosphorylation motif of ATM (26), were used to enrich for phosphopeptides which were subsequently identified and quantified by mass spectrometry (MS). This screen identified 1889 unique phosphosites on 958 different proteins, which can be found in Supplementary Table S1. Motif analysis of the identified phosphosites indeed showed enrichment of the S/TQ motif as 75% of all unique phosphopeptides were phosphorylated on an S/TQ site (Fig. 1B and C). To be defined as a damaged-induced ATM substrate, proteins were selected that had a two-fold or greater increase in peptide count in IR-treated cells compared with untreated cells, and had a less than two-fold difference in peptide count in untreated samples compared with ATMi/IR-treated samples. Out of the 958 phosphoproteins identified from this screen, 213 proteins had phosphorylation sites that met these criteria as IR-induced and ATM-dependent phosphorylation events.

As a novel and complementary approach to identify ATM substrates, we examined proteins that come into close proximity to ATM using proximity-dependent Biotin Identification (BioID), a method which uses a promiscuous BirA biotin ligase domain fused to a protein-of-interest to biotinylate surrounding proteins within a 10-nm radius (ref. 27; Supplementary Fig S1A). The BioID method is effective at capturing both stable and transient protein-protein interactions, such as those between a kinase and its substrate (27–29). To identify putative ATM interactors, 293T cells expressing ATM-BirA were treated with 0 or 1 μmol/L ATMi prior to irradiation with 0 or 10 Gy IR (Fig. 1D). Cells were labeled with 50 μmol/L biotin for 16 hours post irradiation and then harvested, lysed, biotinylated proteins isolated by streptavidin pulldown, and subsequently identified by MS. From this screen, 2,536 unique proteins were identified and the untreated, IR, and IR/ATMi treatments shared a high degree of overlap with 1,876 proteins (74% of all captured proteins) observed in all three conditions with similar relative abundance (Supplementary Fig. S1B; Supplementary Table S2). The proteins that were unique to each treatment condition tended to have a low peptide count and therefore were lower confidence targets. The lack of proteins unique to each treatment condition is likely due to the obligatory 16-hour biotin labeling time which would likely mask transient ATM interactions that occur right after irradiation. However, the consistent results between the three treatment conditions did provide high confidence in the ATM-interactions that were observed. Known ATM interactors, such as MRE11A, RAD50, NBS1 (NBN), and p53 were among the proteins with the highest peptide counts for each treatment condition providing confidence that the identified proteins from all three treatment conditions are indeed bona fide ATM interactors (Supplementary Fig. S1C).

To identify potential biological pathways or protein complexes of interest regulated by ATM, a gene ontology (GO) enrichment analysis was performed on the proteins found enriched in both screens (Fig. 1E). Several biological pathways were significantly enriched that have been previously reported to be dependent on ATM, such as the G2 DNA damage checkpoint, regulation of DNA transcription initiation, and DNA replication (Fig. 1F). Interestingly, one of the most highly enriched GO terms was mitotic sister chromatid cohesion. Our lab and others had previously identified SMC1A and SMC3 as ATM substrates (8–10, 18), and these known phosphorylation sites were also identified in our phosphoproteomics screen as ATM-dependent and IR-induced, further supporting the validity of this approach (Fig. 1G). Furthermore, the two screens identified additional cohesin complex members, including WAPL, PDS5A, PDS5B, NIPBL, RAD21, and SA2, as either novel ATM-substrates or ATM-interactors (Fig. 1G). As such, ATM regulation of cohesin appears to be more expansive than previously appreciated.

To explore whether these putative ATM-interactive cohesin regulatory proteins, NIPBL, WAPL, PDS5A, and PDS5B could be implicated in cellular responses to DNA breakage, we examined if loss of these proteins sensitized cells to IR-induced DSBs. U2OS cells transfected with siRNAs targeting NIPBL, WAPL, PDS5A, or PDS5B were irradiated with 0, 3, and 5 Gy IR and cell survival was determined 2 weeks later by clonogenic assay. Knockdown of NIPBL, PDS5A, and PDS5B all resulted in increased sensitivity to IR, while knockdown of WAPL did not result in a significant decrease in viability compared with cells transfected with a non-targeting (siNT) control (Supplementary Fig. S2A and S2B).

Failure to repair DSBs can also result in chromosomal instability and we previously demonstrated that loss of ATM regulation of SMC1A increases the number of chromosomal aberrations after IR (8, 9). We therefore examined if NIPBL, WAPL, PDS5A, and PDS5B were similarly required for chromosomal stability in 293T cells 1 hour after treatment with 0 or 0.5 Gy IR. Prior to irradiation, cells were pretreated with 0 or 1 μmol/L ATMi, or transfected with siRNAs to NIPBL, WAPL, PDS5A, and PDS5B and metaphase spreads were prepared 1 hour after IR treatment. Cells were scored for the frequency of chromosome aberrations such as chromatid breaks (CTB), chromatid gaps (CTG), chromosome gaps (CHG), and chromosome breaks (CHB). siRNA knockdown of NIPBL, PDS5A, and PDS5B, but not WAPL, resulted in an increase in chromosome aberrations to an extent comparable with those treated with ATM inhibitors, demonstrating that these cohesin proteins are required for effective DSB repair and chromosomal stability (Supplementary Fig. S2C).

To validate that these additional cohesin complex members were ATM substrates, endogenous NIPBL, WAPL, PDS5A, PDS5B, and exogenously expressed Flag-RAD21, were immunoprecipitated from 293T cells that had been treated with 0 or 1 μmol/L ATMi prior to irradiation with 0 or 10 Gy IR. Phosphorylation of cohesin complex members was examined 1 hour after IR by probing immunoprecipitated proteins with the pS/TQ antibody, which recognizes the ATM substrate phosphorylation motif (26). NIPBL, WAPL, PDS5A, PDS5B, and RAD21 were only bound by the pS/TQ antibody following IR treatment and recognition by the pS/TQ antibody was reduced by pretreatment with ATMi prior to irradiation, suggesting that they are damage-induced ATM substrates (Fig. 2A–E). In addition, WAPL protein immunoprecipitated from irradiated 293T ATM-knock out cells was not recognized by the pS/TQ antibody, further demonstrating dependence on ATM for its phosphorylation (Fig. 2B). Collectively these results indicate that NIPBL, WAPL, PDS5A, PDS5B, and RAD21 all have phosphorylated S/TQ sites that are IR-induced and ATM-dependent.

To identify if these ATM-dependent phosphorylation events have functional significance in regulating cellular responses to DNA damage, we first needed to identify and validate a specific S/TQ phosphorylation site within the newly identified phosphorylated cohesin complex proteins. For a variety of reasons, PDS5A proved to be the most straightforward new potential cohesin complex member for validating and functionally characterizing an ATM phosphorylation site. PDS5A was identified in our phosphoproteomic screen to have only one ATM regulated S/TQ site, S1278 (Fig. 1G). To validate phosphorylation of PDS5A on this site, 293T cells were transfected with FLAG-PDS5A or a mutant FLAG-PDS5A where S1278 was mutated to alanine to block phosphorylation by ATM. Cells expressing wild-type (WT) PDS5A or PDS5A-S1278A were pretreated with 0 or 1 μmol/L ATMi before irradiation with 0 or 10 Gy IR, then harvested 1 hour later. PDS5A was immunoprecipitated and phosphorylation was examined by immunoblotting with pS/TQ antibody. As with the endogenous protein, FLAG-PDS5A was recognized by the pS/TQ antibody after IR treatment (Fig. 3A). Significantly, mutation of S1278 in PDS5A or treatment with the ATM inhibitor eliminated binding of the pS/TQ antibody to PDS5A in IR-treated cells, demonstrating that S1278 is an ATM-regulated phosphorylation site in PDS5A (Fig. 3A).

Previously, our lab had shown that phosphorylation of SMC1A by ATM is required for several critical responses to DNA DSBs and that mutation of the ATM phosphorylation sites in SMC1A results in radiosensitivity, impaired checkpoint activation, and increased chromosomal breakage after IR (8, 9). Overexpression of SMC1A phosphomutant cDNAs (SMC1A S957A/S966A) was sufficient to induce these phenotypes, revealing that the phosphomutant SMC1A acts as a dominant-negative blocking normal cohesin function in response to DSBs. To determine if loss of ATM-dependent PDS5A phosphorylation results in these same phenotypes, WT-PDS5A and PDS5A-S1278A were overexpressed in 293T cells (Fig. 3B). Similar to overexpression of SMC1A S957/S966A, overexpression of PDS5A-S1278A reduced cell survival after IR, resulting in a 1.5-fold increase in IR-induced cell death (Fig. 3C). Following DSB formation, ATM signaling pathways activate cell-cycle checkpoints, transiently halting DNA replication in damaged cells. To examine if ATM-dependent phosphorylation of PDS5A is required for cell-cycle checkpoint activation, 293T cells were transfected with WT-PDS5A or PDS5A-S1278A and left untreated or irradiated with 10 Gy IR. Ongoing DNA replication 1 hour after IR treatment was monitored by incorporation of the nucleoside analog bromodeoxyuridine (BrdU). Cells overexpressing PDS5A-S1278A had a significantly larger percentage of cells continuing to actively replicate DNA following IR than cells overexpressing WT PDS5A, demonstrating that phosphorylation of PDS5A on S1278 by ATM is required to establish the S-phase cell-cycle checkpoint following DNA breakage (Fig. 3D).

Because knockdown of PDS5A resulted in increased chromosomal aberrations after IR (Supplementary Fig. S2C), we next examined if phosphorylation of PDS5A-S1278 was required to maintain chromosomal stability. 293T cells overexpressing WT-PDS5A or PDS5A-S1278A were left untreated or irradiated with 0.5 Gy and chromosome spreads were prepared 1 hour after irradiation. While there was no significant difference in the number of chromosome aberrations in untreated cells, cells overexpressing PDS5A-S1278A showed an increase in each type of chromosome aberration after irradiation compared with WT-PDS5A transfected cells (Fig. 3E). These results demonstrate that ATM phosphorylation of PDS5A is required for chromosomal stability following DNA damage, similar to the phenotype seen with the cohesin protein SMC1A. It is noted that in contrast to the effects of PDS5A-S1278A overexpression on these various IR responses, overexpression of WT-PDS5A had no measurable impact.

ATM and the cohesin complex have independently both been reported to be required to repress transcription within the vicinity of DSBs (5, 30). However, it has not been reported whether they are linked in this function; in other words, whether ATM phosphorylation of cohesin is required for transcription repression. To examine if ATM-mediated phosphorylation of SMC1A and PDS5A are required for transcriptional repression, we used a reporter cell line which allows simultaneous visualization of DSB formation and ongoing transcription of a reporter gene (5). The reporter is integrated at a single site on chromosome 1p3.6 in human osteosarcoma U2OS cells and is comprised of 4kb of tetracycline response elements (TRE) which is bound by a doxycycline-inducible transactivator (Fig. 4A). After addition of doxycycline, a minimal cytomegalovirus promoter drives expression of an RNA transcript containing 24 repeats of a MS2 stem loop structure that is specifically bound by YFP tagged MS2, enabling visualization of nascent transcript accumulation (Fig. 4B). DSBs are induced by the nuclease Fok1, which is fused to mCherry and LacI, enabling visualization of Fok1 accumulation to the 256 repeats of the lac operator upstream of the TREs. To enable spatiotemporal regulation of DSB formation, the Fok1 fusion protein is also linked to a destabilization domain (DD) and estrogen receptor (ER) domain. The DD targets Fok1 for proteosomal degradation until addition of a ligand, Shield-1, which stabilizes Fok1 resulting in rapid accumulation in cells. The ER domain sequesters Fok1 in the cytoplasm until addition of 4-OHT which results in translocation of Fok1 into the nucleus. After addition of both Shield-1 and 4-OHT, Fok1 accumulates in the nucleus at the reporter site where it induces DSBs, as denoted by formation of a mCherry focus (Fig. 4B). As previously reported (5), induction of DSBs represses localized transcription in a process dependent on ATM as addition of ATMi prior to DSB formation blocked repression of transcription (Fig. 4B and C). To test if ATM phosphorylation of cohesin was required for this transcriptional repression, reporter cells were transfected with WT-SMC1A, WT-SMC3, or WT-PDS5A or the corresponding phosphomutant proteins (SMC1A-S957A/966A, SMC3-S1067A/1083A, PDS5A-S1278A). Cells overexpressing WT-SMC1A, WT-SMC3, and WT-PDS5A showed a significant decrease in transcription activity following DSB induction, demonstrating effective transcriptional repression. However, cells expressing the cohesin phosphomutant proteins had no significant decrease in transcription following DSB induction, consistent with a defect in transcriptional repression near DSBs dependent on ATM phosphorylation of these cohesin proteins (Fig. 4C).

The processes of DNA replication and RNA transcription have many fundamental similarities, including that both use large protein complexes which require helicases to open DNA, providing access for a polymerase to synthesize complementary nucleotide strands of DNA or RNA. Consequently, both processes would be equally disruptive for repair proteins attempting to access and repair damaged DNA. Because both DNA replication and transcription would have to overcome the same modifications to chromatin surrounding DNA damage sites, we predicted that the same ATM-dependent chromatin alterations surrounding DSBs that repress transcription may simultaneously repress DNA replication.

While we had already demonstrated that ATM phosphorylation of cohesin proteins is required for IR-induced reduction of nucleotide incorporation in S-phase in general, we used the site-specific breakage introduced by the Fok1 system to specifically examine replicative DNA synthesis in the 3-dimensional vicinity of the DSB, examining DNA replication around the Fok1-induced break by labeling cells with 5-ethynyl-2′-deoxyuridine (EdU), a nucleoside analogue which is incorporated into nascent DNA, and visualizing the DSB with the mCherry label. DSBs were induced into reporter cells by addition of 4-OHT/Shield-1 for 90 minutes before labeling with EdU for 2.5 hours (Fig. 5A). EdU positive nuclei with a single Fok1 (mCherry) focus were imaged. Cells with a Fok1 focus exhibited an obvious decrease in EdU staining in and around the Fok1 focus, suggesting that DNA replication is suppressed within the geographic vicinity of DSBs (Fig. 5B). To quantify DNA replication surrounding the DSB site, the median fluorescence intensity of Fok1, EdU, and DAPI was measured starting at the epicenter of the Fok1 focus (radius = 0) and continuously measured outward using concentric circles increasing in radius by one pixel until reaching a final radius of 30 pixels (Fig. 5C). The resulting radial profile plot revealed that at the center of the Fok1 focus, where the Fok1-mCherry fluorescence was the highest (radius = 0–15 pixels), EdU showed the lowest amount of incorporation relative to the surrounding area outside of the Fok1 focus (radius = 20–30 pixels; Fig. 5C).

Binding of LacI to multiple repeats of the Lac operator can act as an impediment to replication fork progression (31). To ensure that DNA replication is repressed by formation of DSBs and not simply by the binding of Fok1 to the Lac operator array, we mutated the nuclease domain of Fok1 (D450A). While both Fok1 and Fok1 D450A localized to the reporter, Fok1-D450A did not induce DSBs as evidenced by the lack of DSB-dependent phosphorylation of SMC1A on S957 and H2AX phosphorylation on S139 (γH2AX) that colocalized with Fok1, nuclease competent foci (Fig. 5B; Supplementary Fig. S3A–S3H). Furthermore, in contrast to WT-Fok1, the nuclease-dead Fok1-D450A did not exhibit reduced EdU incorporation within or surrounding any of its foci, demonstrating that DNA breakage is required for repression of DNA replication (Fig. 5B and D).

We further validated that the repression of DNA replication around DSBs is not unique to the Fok1-DSB reporter by examining DNA replication around IR-induced DSBs. U2OS cells were treated with 0.2 Gy IR and 10 minutes after irradiation, pulsed with EdU for 20 minutes (Supplementary Fig. S4A) and then fixed for immunofluorescence. IR-induced DSB sites were visualized by staining for 53BP1 foci and the EdU intensity surrounding 53BP1 foci was quantified as by measuring the radial profile intensity as previously described for Fok1-induced DSBs. Like Fok1-induced DSBs, IR-induced DSBs displayed reduced EdU intensity within and surrounding 53BP1 foci demonstrating that suppression of DNA replication is not unique to our Fok1 reporter assay (Supplementary Fig S4B and S4C).

In response to DSBs, ATM phosphorylates its downstream effector kinase CHK2, which activates the S-phase checkpoint through phosphorylation of the phosphatase CDC25A (32). Damage-induced phosphorylation of CDC25A by CHK2 causes the rapid degradation of CDC25A and subsequent inhibition of DNA synthesis throughout the nucleus (32, 33). Interestingly, while we observed a significant decrease in EdU incorporation around the Fok1 cutsite (Fig. 5C), we detected no significant difference in total nuclear EdU intensity between cells with DSBs (Fok1 focus positive) compared with cells without DSBs (Supplementary Fig. S5A), suggesting that only localized DNA replication was repressed. To further confirm Fok1-induced DSBs do not cause global repression of DNA synthesis, cells were labeled with BrdU for 4 hours after mock treatment or treatment with 4-OHT/Shield-1 to induce DSBs. In contrast to IR-induced DNA breaks (Fig. 3D), Fok1-induced breakage did not significantly alter DNA synthesis throughout the rest of the nucleus (Supplementary Fig. S5B). The extent to which DNA synthesis is inhibited is dependent on IR-dosage indicating the extent of DNA damage and number of DNA breaks affects global S-phase checkpoint activation (32). Fok1-induced DSBs localized at a single genetic locus may not activate DNA damage signaling to the same extent as 10 Gy IR which generates potentially hundreds of breaks, each at different genomic locations. To examine ATM signaling after DSB induction, we performed a time course examining ATM signaling at different times after addition of 4-OHT/Shield-1 (Supplementary Fig. S5C). Although ATM autophosphorylation 1 hour after Fok1-induced DSB formation is comparable with IR-treated cells, CHK2 phosphorylation is barely detectable even 24 hours after continuous FOK1-induced breakage. The higher sensitivity of ATM activation in response to acute and localized DSBs may allow it to enforce a localized S-phase checkpoint without disrupting global DNA synthesis through activation of the CHK2-CDC25A pathway which is activated in response to more extensive and disperse DSBs (Supplementary Fig. S5D).

Repression of transcription within the vicinity of DSBs is dependent on ATM activity (5), so ATM may similarly repress DNA replication within close spatial proximity to DNA break sites. To examine if repression of DNA replication at DSBs is dependent on ATM activity, Fok1 cells were treated with an ATMi during DSB induction and EdU labeling. Fok1 cells treated with ATM inhibitor did not have decreased EdU incorporation within or around the DSB site (Fig. 5B and E). Untreated Fok1 cells showed a significant decrease in EdU incorporation within the Fok1 foci compared with Fok1 cells treated with an ATMi or to cells that expressed Fok1 D450A, demonstrating that ATM activity is required to repress DNA replication at DSB sites (Fig. 5F).

ATM belongs to a family of kinases called the phosphatidylinositol 3-kinase-related kinases (PIKKS) which also includes DNAPK, another responder to DSBs, and ATR, which regulates responses to DNA replication stress. To examine if either DNAPK or ATR could also regulate repression of DNA replication at DSBs, Fok1 cells were treated with a DNAPK inhibitor (DNAPKi, M3814) or an ATR inhibitor (ATRi, AZD6738) during DSB induction and EdU labeling (Supplementary Fig. S6A–S6E). Although the doses of DNAPKi, ATMi, and ATRi were sufficient to block downstream ATM, DNAPK, or ATR signaling (Supplementary Fig. S6F), only addition of ATMi resulted in loss of repression of DNA replication (Supplementary Fig. S6E). Therefore, the repression of DNA replication around the break after DNA damage is reliant only on ATM activity and not the other PIKKs.

To examine if the cohesin complex is also required for repression of DNA replication in the vicinity of this site-specific DNA break, various cohesin proteins were depleted using siRNAs and examined for DNA replication around the Fok1 cut site. Knockdown of SMC1A, SMC3, NIPBL, WAPL, and PDS5A all blunted the decrease in EdU incorporation at the Fok1 DSB site (Fig. 6A and B; Supplementary Fig. S7A–S7D). Collectively this reveals that the core cohesin complex (SMC1A and SMC3), as well as the proteins that regulate cohesin loading and unloading on chromatin (NIPBL, WAPL, and PDS5A), are all required for repression of DNA replication within the vicinity of a DSB.

Cohesin is required for the formation of higher order chromatin structure by forming loops which organizes the genome into compartments (34). The boundaries of these compartments are determined by another protein called CTCF, which positions cohesin within the genome (34). In addition to its role in establishing chromatin architecture, CTCF is recruited to DSB sites and promotes repair by homologous recombination (35). To determine if CTCF functions with cohesin to repress replication at DSBs, we knocked down CTCF in Fok1-expressing cells and examined EdU incorporation around the Fok1 DSB site. Similar to loss of cohesin proteins, knockdown of CTCF resulted a significant decrease in DNA replication repression at the Fok1 DSB site (Fig. 6C; Supplementary Fig. S7C and S7D). Because both CTCF and cohesin are required for repression of DNA replication at Fok1 cut sites, chromatin architecture around DSBs may influence responses to DNA damage, including DNA replication repression.

Because ATM phosphorylation of cohesin was required for transcriptional repression, we examined if ATM phosphorylation of cohesin proteins was also required for repression of DNA replication at DSB sites. DSB reporter cells were transfected with WT SMC1A, SMC3, or PDS5A or their ATM-phopho-silent mutant counterparts: SMC1A-S957A/966A, SMC3-S1067A/S1083A, or PDS5A-S1278A. DSBs were induced in transfected cells and replication monitored by incorporation of EdU surrounding the Fok1 cut site. Cells overexpressing WT SMC1A, SMC3, or PDS5A all showed significantly reduced EdU incorporation at the DSB sites compared with cells overexpressing their corresponding phosphomutant, SMC1A-S957/966A, SMC3-S1067/1083A, and PDS5A-S1278A (Fig. 6D–F; Supplementary Fig. S7E). Collectively these data demonstrate that direct phosphorylation of multiple cohesin complex proteins by ATM is required for repression of DNA replication in the geographic vicinity of DSB sites.

Cohesin is required for efficient responses to DSBs. Here we show that in addition to the previously identified ATM cohesin substrates, SMC1A and SMC3, the cohesin proteins PDS5A, PDS5B, RAD21, NIPBL, and WAPL are all novel DNA damage-induced ATM substrates. Mutation of the PDS5A ATM phosphorylation site resulted in decreased cell survival after IR, increased chromosomal instability, and decreased checkpoint activation. These same phenotypes are also observed in cells in which ATM phosphorylation of SMC1A and SMC3 is blocked, suggesting that ATM phosphorylation of each of these members of the cohesin complex participates in regulating cellular responses to DSBs. One caveat to these experiments is that overexpression of exogenous proteins may not always emulate the normal function of that protein because it is expressed above endogenous levels. However, because overexpression of WT SMC1A, SMC3, and PDS5A proteins did not appear to disrupt any biological functions like their phosphomutant counterparts, these phenotypes do not seem to be caused by the exogenous proteins being over expressed.

One way that ATM may regulate cohesin function and promote responses to DNA damage is by altering the amount of time cohesin spends on DNA. ATM-dependent phosphorylation of cohesin may promote its recruitment and retention at DSBs because the cohesin loader, NIPBL, and unloaders, WAPL and PDS5A/B, were all identified as ATM substrates. Indeed previous studies have found that cohesin localization to DSBs is at least partially dependent on ATM activity (36, 37). Altering cohesin binding to chromatin causes dramatic alterations to chromatin architecture (17, 38, 39). Our results suggest that ATM phosphorylation of cohesin near DSBs may regulate transcriptional repression and S-phase checkpoint activation through alteration of local chromatin structure, a concept supported by the observation that down-regulation of CTCF also impacts these damage-induced processes.

Cohesin establishes higher order chromatin structures through a mechanism called loop extrusion in which cohesin is loaded onto DNA by NIPBL/MAU2 and then extrudes DNA through its lumen to form loops of increasing size until it encounters CTCF bound to convergently oriented CTCF-binding sites (38–43). These chromatin loops serve a biological function as they bring linearly distant genomic elements, such as promoters and enhancers, into close spatial proximity (15). In addition, these cohesin-dependent chromatin structures organize the genome into topologically associated domains (TAD), which functionally compartmentalizes genomic regions into units with similar histone modifications, transcription levels, and replication timing (Fig. 7; refs. 15, 39, 40, 44). The cohesin-dependent organization of the genome may also compartmentalize functional responses to DSBs, allowing for localized checkpoint activation.

The preexisting chromatin architecture before a DSB directs DSB repair as propagation of ATM-dependent γH2AX spreading and 53BP1 accumulation around a DSB break correlates with the boundaries of the TAD in which the break occurred (6, 37). We expand on these findings and propose that in addition to delineating the extent of ATM-signaling, TADs may also compartmentalize checkpoint responses to DSBs (Fig. 7). Traditional DSB-induced S-phase checkpoint assays monitor global changes in DNA replication after damage formation by IR or other DSB-inducing agents that create numerous DSB sites randomly located throughout the nucleus. Here, we developed a novel assay to examine the localized effects of a single DSB locus on DNA replication using a spatiotemporally regulated nuclease, Fok1. We found that DNA breaks at a single genomic location did not cause a reduction in total nuclear EdU intensity following DSB induction at the Fok1 cut site (Supplementary Fig. S5), but significantly reduced EdU incorporation within the spatial proximity of the DSB. These findings suggests that the S-phase checkpoint can be divided into global and localized responses to DSBs, both of which are dependent on ATM phosphorylation of cohesin proteins (Supplementary Fig. S5D). Extensive and disperse DNA damage activate an ATM signaling cascade that inhibits global DNA synthesis (32). However, for DNA damage that is highly localized (i.e., located within a single TAD), widespread inhibition of transcription and DNA replication may be too drastic a response. In such cases, it may be more beneficial to only repress replication and transcription within the same narrow genomic confines as the DSB, analogous to controlling a fire in a single room with a fire extinguisher as opposed to disturbing the entire building by activating the sprinklers and fire alarm for a single, localized threat (Fig. 7).

This novel localized S-phase checkpoint may be induced by alteration of local chromatin structure surrounding the break caused by ATM phosphorylation of cohesin. Direct phosphorylation of SMC1A, SMC3, and PDS5A by ATM is required for both repression of transcription and DNA replication at DSBs. In addition, siRNA knockdown of CTCF, which is required to position cohesin at TAD boundaries, is also required for local repression of DNA replication. Collectively, these results suggest that ATM-dependent responses to DSBs may be dependent on alteration of chromatin architecture. Cohesin is located at origins of replication and regulates DNA replication through the formation of loops that bring clusters of DNA origins into close proximity called replication factories (16). Relocation of cohesin or alteration of cohesin activity at DSBs by ATM may disrupt and alter local chromatin architecture at DSBs resulting in disruption of nearby replication factories and resulting in a localized repression of DNA replication.

In summary, we conducted two unbiased complementary proteomic screens which putatively identified several novel ATM substrates including PDS5A, NIPBL, WAPL, and RAD21. ATM-dependent phosphorylation of PDS5A on S1278 is required for effective responses to DSBs. In addition, we observed that direct phosphorylation of SMC1A, SMC3, and PDS5A by ATM are all required for repression of transcription within the spatial vicinity of DSBs. Furthermore, we developed a novel “localized” S-phase checkpoint assay and demonstrated that suppression of replication surrounding DNA DSBs is dependent on ATM, cohesin, and CTCF. ATM phosphorylation of the cohesin proteins SMC1A, SMC3, and PDS5A are all required for establishment of the localized S-phase checkpoint, which suggests a functional relationship between chromatin structure and ATM regulation of DNA metabolism.

D.E. Fleenor reports grants from NIH during the conduct of the study. P.E. Burrell reports grants from NIH during the conduct of the study. M.B. Kastan reports grants from NIH during the conduct of the study; other support from XRAD Therapeutics; and other support from Dragon Therapeutics outside the submitted work. No disclosures were reported by the other author.

T.E. Bass: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. D.E. Fleenor: Conceptualization, resources, investigation, methodology, writing–review and editing. P.E. Burrell: Investigation, visualization, writing–review and editing. M.B. Kastan: Conceptualization, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

Financial support by NIH (grants R01CA157216 and P30CA014236). We thank the Duke University School of Medicine for the use of the Proteomics and Metabolomics Shared Resource, which processed and analyzed mass spectrometry samples, and the DNA Analysis Facility for DNA sequencing analysis. We thank Dr. Roger Greenberg (University of Pennsylvania) for generously providing the FokI-expressing U2OS 2–6-3 cells.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.