The DNA damage response acts as a safeguardagainst harmful DNA–RNA hybrids ofdifferent origins

Article The DNA damage response acts as a safeguard against harmful DNA–RNA hybrids of different origins Sonia Barroso ,EmiliaHerrera-Moyano , Sergio Muñoz, María García-Rubio, Belén Gómez-González &AndrésAguilera * Abstract Despite playing physiological roles in specific situations, DNA–RNA hybrids threat genome integrity. To investigate how cells do counteract spontaneous DNA–RNA hybrids, here we screen an siRNA library covering 240 human DNA damage response (DDR) genes and select siRNAs causing DNA–RNA hybrid accumulation and a significant increase in hybrid-dependent DNA breakage. We identify post-replicative repair and DNA damage checkpoint factors, including those of the ATM/CHK2and ATR/CHK1pathways. Thus, spontaneous DNA–RNA hybrids are likely a major source of replication stress, but they can also accumulate and menace genome integrity as a consequence of unrepaired DSBs and post-replica- tive ssDNA gaps in normal cells. We show that DNA–RNA hybrid accumulation correlates with increased DNA damage and chromatin compaction marks. Our results suggest that different mechanisms can lead to DNA–RNA hybrids with distinct consequences for replication and DNA dynamics at each cell cycle stage and support the conclusion that DNA–RNA hybrids are a common source of spontaneous DNA damage that remains unsolved under a deficient DDR. Keywords ATM; ATR; DNA damage response; post-replicative repair; R loop Subject Category DNA Replication & Repair DOI 10.15252/embr.201847250 | Received 16 October 2018 | Revised 28 June 2019 | Accepted 1July 2019 | Published online 24 July 2019 EMBO Reports (2019)20:e47250 Introduction R loops, nucleic acid structures composed of a DNA–RNA hybrid and the displaced non-template DNA strand, are central to a number of cellular processes, including mitochondrial DNA replication and immunoglobulin diversification as well as some cases of transcription regulation [1]. Despite their positive role in cell physiology, an increasing number of reports have shown that R loops can be pathological [2,3]. Given the potential deleterious consequences of R loops, cells bear a number of factors to prevent and resolve DNA–RNA hybrids [2,4]. As first discovered in yeast mutants of the THO complex [5], the optimal co-transcriptional assembly of the messenger ribonucleoprotein particle (mRNP) is crucial to prevent that the nascent RNA hybridizes back with the DNA template. The accumulation of R loops has been subsequently demonstrated in human cells depleted of the splicing factor SRSF1 [6], the THO complex [7], and several other mRNP processing factors, including THSC/TREX-2 [8], the mRNA cleavage and polyadenylation protein FIP1L1 [9], or the XRN2 exoribonuclease [10]. Hyper-acetylated chromatin also facilitates DNA–RNA hybrid accumulation [11]. Although R loops can be removed by the action of RNase H, which specifically cleaves the RNA moiety of the DNA–RNA hybrid, emerging evidence suggests the existence of a varied number of different helicases that can unwind DNA–RNA hybrids, including senataxin (SETX) [12], aquarius (AQR) [13], DHX9 [14], DDX1 [15], DDX19 [16], DDX23 [17], and DDX21 [18]. A number of reports indicate that replication impairment is a major mechanism of DNA–RNA hybrid-mediated genetic instability [11,19–24]. Along this line, R loop and R loop-mediated DNA damage accumulation was reported in cells depleted of the doublestrand break (DSB) repair and tumor suppressor genes BRCA1 and BRCA2, the Fanconi anemia (FA) factors involved in inter-strand crosslink (ICL) repair or the FACT chromatin reorganizer complex, required for the progression of replication fork (RF) through chromatin [25–29]. Since all these factors are involved in replication fork progression, these findings also support that R loops constitute a potential obstacle to replication [30]. Interestingly, R loops seem not to be harmful by themselves. Instead, certain chromatin modifications may be required for R loop-induced genome instability, as recently shown in budding yeast for histone H3 Ser10 phosphorylation (H3S10-P) [31]. The connection between DNA–RNA hybrids and specific chromatin marks, such as H3S10-P or H3K9me2, has been observed also in human cells and C. elegans depleted of the THO complex or SETX or at specific fragile sites, like those of Friedreich ataxia (FRDA) and Fragile X syndrome (FXS) [32–34]. Also, genome-wide mapping has Centro Andaluz de Biología Molecular y Medicina Regenerativa-CABIMER, Universidad de Sevilla-CSIC-Universidad Pablo de Olavide, Seville, Spain *Corresponding author. Tel: +34 95 4468372; E-mail: [email protected] ª2019 The Authors EMBO reports 20:e47250 |2019 1of 15 shown a correlation between spontaneous R loops and a set of histone modifications [35,36]. Although the cause–effect relationship between these chromatin marks and R loops is yet to be understood, the accumulated evidence suggests that DNA–RNA hybrids can modulate chromatin remodeling and vice versa [4,30]. To prevent the potential harmful effect of DNA damage and genetic instability in cell physiology and development, cells have evolved a coordinated cellular response, the DNA damage response (DDR), which is an intrinsic barrier to the early phases of human tumorigenesis [37,38]. The DDR includes a varied number of mechanisms that detect and enable the repair of different types of DNA damage [39,40]. Among them, DSBs activate the ATM/CHK2 DNA damage checkpoint (DDC) pathway [39,40]. By contrast, the RPA-coated single-stranded DNA (ssDNA) generated by stalled replication forks triggers the activation of the 9-1-1/ATR/CHK1 DDC pathway, which would mainly protect stalled forks from breakage. In addition, replication can restart downstream of the DNA lesion, leaving ssDNA gaps behind the fork that would be subsequently repaired by post-replicative repair (PRR). Under the premise that R loops may be a natural source of DNA damage, we wondered whether DDR factors might be important for spontaneous R loop detection and/or dissolution to protect genome integrity. To search for DDR factors involved in R loop homeostasis, we screened a 240 siRNA library of DDR genes using immunofluorescence (IF) with the S9.6 anti-DNA–RNA antibody. After a first selection of candidates based on the IF-positive signals, the presence of DNA–RNA hybrids was confirmed by RNase H-sensitive IF and DNA–RNA immunoprecipitation (DRIP). These included the DDC sensors RAD17 and the 9-1-1 complex as well as the DDC kinases, ATR, ATM, CHK1, and CHK2 and the PRR factors UBE2B and RAD18. Cells depleted of each of these DDR factors increased H3S10-P and H3K9me2 chromatin marks and DNA damage partially in a DNA–RNA hybrid-dependent manner. However, R loop-depen- dent replication fork stalling was detected in cells depleted of ATR or CHK1 but not in cells depleted of ATM, CHK2, or PRR factors. We propose a model in which in the absence of ATR/CHK1, harmful DNA–RNA hybrids accumulate leading to fork stalling, but in the absence of the ATM/CHK2 and PRR machineries they accumulate in association with unrepaired DSBs and post-replicative ssDNA gaps, respectively. Altogether, these results show that, in addition to the previously reported role of replication-associated repair factors such as FACT or the Fanconi anemia pathway [25–28,41], the 9-1-1/ ATR/CHK1 and ATM/CHK2 checkpoints and the post-replicative repair pathways are a safeguard against the accumulation of DNA– RNA hybrids as a common source of DNA damage. Results Screening for DDR factors involved in R loop homeostasis To search for DDR factors that could be involved in the protection against potentially harmful R loops, we performed an siRNA screening in HeLa cells using an arrayed collection of siRNAs targeting 240 human genes involved in DDR (Dharmacon Human ON-TARGET- plus DDR siRNA library). We prepared 96-well plates containing a duplicate of each four-siRNA pool for every DDR gene, and a nontargeted siRNA used as negative control (Fig 1A). Since THOC1 depletion was previously shown to cause accumulation of DNA– RNA hybrids [7], an siRNA pool targeting THOC1 was included as a positive control. For the screening, cells were immunostained with the S9.6 monoclonal antibody that recognizes DNA–RNA hybrids and with the anti-nucleolin antibody to exclude any nucleolar signal [42]. Experiments were repeated twice and all S9.6 nuclear intensity data analyzed excluding the nucleolar signal. We selected factors whose depletion led to a 10% higher S9.6 average nuclear intensity with respect to the negative control in both experiments as those to be considered further. For those siRNAs for which only one of the experiments was positive, the analysis was repeated up to 6 times to obtain a more reliable median value. With these criteria, we selected 16 candidate factors including four genes involved in the DNA damage checkpoint (DDC) (RAD1, RAD9A, TOPBP1, and MDC1), two post-replicative repair (PRR) factors (UBE2B and RAD18), two nucleotide excision repair (NER) factors (GTF2H5 and DDB2), the tyrosyl-DNA phosphodiesterase 2 (TDP2), the APEX1 gene of base excision repair (BER), the PMS1 gene of mismatch repair (MMR), the superoxide dismutase (SOD1) gene, and two Fanconi anemia (FA) genes, FANCD2 and FANCA (Fig 1A). In agreement with the reported role for FA in R loop resolution [25,28], all FA siRNAs present in the library, except FANCB and FANCJ, conferred an S9.6 signal above the control (Fig 1A). Given that RAD18 is involved in FANCD2 ubiquitination in addition to PRR [43,44], we assayed the possibility that the observed accumulation of hybrids in PRR-deficient cells was due to a lack of FA function. We observed a significant decrease in the percentage of cells with FANCD2 foci when RAD18 was depleted (Fig EV1A). By contrast, siATM, siATR, and siUBE2B cells had no significant defect in FANCD2 foci formation. This is likely due to the redundant role of UBE2A, with which UBE2B shares 95% identity, both proteins being the homologs of the yeast PRR protein Rad6 [45], since the double depletion of UBE2B and UBE2A is reported to cause a defect in FANCD2 ubiquitination [46]. By contrast, depletion of UBE2B alone is reported to be impaired in PRR [47]. Therefore, we support that the DNA–RNA hybrid accumulation phenotype observed in RAD18 and UBE2B-depleted cells was caused by a defect in PRR rather than in FA. To validate the top-hit candidates, we performed S9.6 IF in coverslips with the 14 selected candidates not involved in FA and with FANCG, as a representative of the FA pathway (Fig 1B and C). For this purpose, we used in each case a pool of two to four siRNAs from the original pool, which we previously validated to confer RNA silencing by RT–qPCR (Fig EV1B). As shown in Fig 1B and C, 6 out of the 14 candidates showed a significant increase in S9.6 nuclear intensity. The most representative functional group was the DNA damage checkpoint (DDC), with three genes: the RAD1 and RAD9A components of the 9-1-1 complex and the MDC1 mediator. Consequently, we extended our analysis to other relevant DDC genes such as the remaining 9-1-1 complex member HUS1, the clamp loader RAD17, and the main DDC kinases CHK1, ATR, CHK2, and ATM. Although these additional candidates, with the exception of siCHK2, had scored above the siC control in our screening (Fig 1A), a significant increase in S9.6 signal was only validated after depletion of CHK1, ATR, and CHK2. We next confirmed the effect of ATR depletion on S9.6 signal accumulation with four different siRNAs (Fig EV1C). Furthermore, treating cells with the 2of 15 EMBO reports 20:e47250 |2019 ª2019 The Authors EMBO reports Sonia Barroso et al 0.8 0.9 1.0 1.1 1.2 S9.6 signal intensity per nucleus (relative values) siC siUBE2B siRAD18 siGTF2H5 siDDB2 siTDP2 siAPEX1 siPMS1 siSOD1 siHMGB2 siRAD52 siFANCG siRAD1 siRAD9A siTOPBP1 siMDC1 siHUS1 siRAD17 siCHK1 siATR siCHK2 siATM S9.6 signal intensity per nucleus (relative values) *** *** *** *** *** AB C 0 1 2 3 4 *** Top hit selected candidates siFA genes siTHOC1 DDC additional candidates 1 TOPBP1 2 TDP2 3 PMS1 4 GTF2H5 5 HMGB2 6 APEX1 7 MDC1 8 RAD1 9 DDB2 10 SOD1 11 UBE2B 12 RAD18 13 RAD9A 14 RAD52 S9.6 Merge (S9.6 +nucleolin +DAPI) S9.6 Merge (S9.6 +nucleolin +DAPI) siC siRAD1 siRAD9A siTOPBP1 siMDC1 siHUS1 siRAD17 siCHK1 siATR siATM siUBE2B siRAD18 siCHK2 S9.6 Merge (S9.6 +nucleolin +DAPI) 25 µm Figure 1. Screening for DDR factors involved in R loop homeostasis. A Top-hit selected candidates are listed. The plot shows relative S9.6nuclear intensity values for cells transformed with each of the indicated pool of siRNAs. B Representative images of HeLa cells immunostained with S9.6and nucleolin antibodies after transfection with at least a pool of two siRNAs of each original pool. C Relative S9.6signal intensity per nucleus after nucleolus signal removal in HeLa cells transfected with the indicated siRNAs. More than 250 total cells from three independent experiments were considered. The median of each population is shown. Boxes and whiskers indicate 25–75 and 10–90 percentiles, respectively. ***P<0.001 (Mann–Whitney U-test). ª2019 The Authors EMBO reports 20:e47250 |2019 3of 15 Sonia Barroso et al EMBO reports siC siFANCD2 siHUS1 siRAD1 siRAD9A siRAD17 siTOPBP1 siMDC1 siCHK1 siATR siCHK2 siATM siUBE2B siRAD18 - + - + - + - + - + - + - + - + - + - + - + - + - + - + RNH1 **** * siC siFANCD2 siHUS1 siRAD1 siRAD9A siRAD17 siTOPBP1 siMDC1 siCHK1 siATR siCHK2 siATM siUBE2B siRAD18 - + - + - + - + - + - + - + - + - + - + - + - + - + - + RNH1 ** *** siC siATR - + - + siATM siUBE2B - + - + RNH1 * * A BC siC siATR - + - + siATM siUBE2B - + - + RNH1 S9.6 nucleolin Merge + DAPI siC siATR *** *** *** ****** *** *** MIB2 APOE RPL13A * S9.6 IP (relative values) S9.6 IP (relative values) S9.6 IP (relative values) siC siATR - + - + siATM siUBE2B - + - + RNH1 * * * * RHOT2 * S9.6 IP (relative values) 0 1 2 3 0 1 2 3 4 S9.6 signal intensity per nucleus after CE and RNAseIII (relative values) 0 1 2 3 4 *** **** ** ** ** * * ** 0 1 2 3 4 *** ** *** * * ** ** * ** * *** ** 0 2 4 6 8 25 µm Figure 2. 4of 15 EMBO reports 20:e47250 |2019 ª2019 The Authors EMBO reports Sonia Barroso et al phosphatidylinositol 3-kinase-like kinases (PIKK) inhibitors caffeine and ETP-46464 had a similar effect (Fig EV1D), which indicates that the increase in DNA–RNA hybrids is unlikely due to siRNA offtargets. We therefore decided to focus on DDC and PRR pathways as potentially involved in the protection against DNA–RNA hybrids. DNA–RNA hybrids accumulate in DDC- and PRR-deficient cells To confirm further that the DDC- and PRR-deficient cells accumulate DNA–RNA hybrids, we performed DNA–RNA immunoprecipitation (DRIP) with or without in vitro RNase H treatment, followed by qPCR at APOE and RPL13A genes, previously identified as R loopprone regions and used as model human genes for these studies [8,25,26,35]. The SNRPN gene was used as a negative control region at which low levels of detection correspond to background (Fig EV2A). As shown in Fig 2A, depletion of most of the DDC- and PRR-selected genes, including both ATM/CHK2 and ATR/CHK1 branches, increased the DRIP signal in the RPL13A and APOE genes to similar levels than FANCD2-depleted cells, which were used as positive control [25]. Importantly, the DRIP signals were significantly reduced by RNase H treatment, implying that DNA–RNA hybrids do indeed accumulate in DDC- or PRR-defective conditions. This is unlikely related to altered gene expression since, although slightly increased in siATM cells, the RNA levels of RPL13A were not significantly changed in siATR or siUBE2B cells (Fig EV2B). We next confirmed DNA–RNA hybrids at two other genes, MIB2 and RHOT2, when each of the three selected pathways was depleted (siATM, siATR, and siUBE2B, Fig 2A). Interestingly, some depletions (such as siATM) increased the RNase H-sensitive DRIP signal at some of the genes studied, but showed no effect in the S9.6 IFs used for validation of the screening results. However, despite the high reactivity with DNA–RNA hybrids, the S9.6 antibody also recognizes dsRNA [48,49]. Whereas DRIP is a highly specific method of detection of DNA–RNA hybrids, given that putative dsRNA molecules are not amplified by qPCR and signals are considered positive when sensitive to in vitro treatment with RNase H, which only removes RNA–DNA hybrids [50], dsRNA could be masking our initial validation by IF. Consequently, we repeated the IF analysis after in vitro treatment with RNase III, which degrades dsRNA and after pre-extraction of the cytoplasm, to avoid any cytoplasmic interference [50]. As shown in Fig 2B and C, we observed an increased S9.6 signal when each of the three selected DDR pathways was inactivated (siATR, siATM, and siUBE2B). These signals were sensitive to RNase H (Fig 2C), further confirming that they correspond to DNA–RNA hybrids. To assess whether these hybrids were transcription-dependent, we performed S9.6 IF in cytoplasm preextracted cells treated with the transcription inhibitors 5,6-dichloro- 1-b-D-ribofuranosylbenzimidazole (DRB) and cordycepin (Cord). As shown in Fig EV2C and D, both compounds significantly reduced the S9.6 signal observed after depletion of ATR or UBE2B. Since the 9-1-1/ATR/CHK1 DDC branch responds to RPA-coated singlestranded DNA (ssDNA) such as that generated by stalled forks, the PRR machinery acts after replication, and ATM is activated by DSBs [39,40], these results suggest that DNA–RNA hybrids are a common source for both replication stress and DSB-mediated cellular responses. We conclude that DNA–RNA hybrids accumulate in the absence of a proper response to either replication stress or DSBs. DNA–RNA hybrids are a source of DNA breaks in specific DDR-deficient cells Next, we reasoned that if DNA–RNA hybrids are a frequent source of DNA damage, part of the spontaneous damage potentially accumulated in these DDC- and PRR-defective cells should be suppressed by RNase H overexpression. To directly analyze DNA damage, we performed single-cell alkaline gel electrophoresis (comet assay) in cells depleted of the DDR-selected factors with or without overexpression of RNase H (Fig 3A). The comet tail moment (a measurement of DNA breaks) in cells depleted of all the DDC and PRR factors tested increased with respect to cells transfected with a non-targeting siC. This damage was likely a consequence of both spontaneous metabolism and the stress induced by the double transfection since single siATR- and siUBE2B-transfected cells showed a much modest increase (Fig EV2E). Importantly, this damage was partially dependent on transcription as it was reduced by cordycepin treatment (Fig EV2E). Furthermore, although RNase H overexpression caused DNA damage by itself in siC-treated cells, in agreement with previous reports [11], it caused no further damage in ATR-depleted cells. By contrast, RNase H overexpression caused a slight decrease in the occurrence of DNA breaks in cells depleted of specific factors of the 9-1-1/ATR/CHK1 pathway, such as siRAD9A-, siRAD17-, and siCHK1-treated cells (Fig 3B), even though they did not show statistical significance. These results suggest that, whereas the DNA breaks accumulated in the absence of the ATM/CHK2 pathway arise independently on DNA–RNA hybrids, part of the DNA breaks observed in the absence of the 9-1-1/ATR/CHK1 pathway are due to DNA– RNA hybrids. Notably, the tail moment increase observed in siUBE2B and siRAD18 cells was significantly reduced by RNase H overexpression, implying that a defect in the PRR machinery causes DNA damage due to the formation of DNA–RNA hybrids. DDR inactivation causes R loop-dependent H3S10-P and H3K9me2 Given the emerging connection between DNA–RNA hybrids and specific chromatin marks such as H3S10-P and H3K9me2 [31–34] ◀Figure 2. R loop accumulation after depletion of the selected DNA Damage Checkpoint (DDC) and Post-replicative Repair (PRR) candidates. A Relative DRIP–qPCR signal values at RPL13A, APOE,MIB2,and RHOT2genes in HeLa cells transfected with the indicated siRNAs and treated in vitro with RNase H pre-immunoprecipitation where indicated. The mean SEM from at least three independent experiments is shown. *P<0.05,**P<0.01, ***P<0.001 (one-tailed paired t-test). B Representative images of immunostaining with S9.6and anti-nucleolin antibodies in HeLa cells transfected with the indicated siRNAs. C Relative S9.6signal intensity per nucleus after nucleolus signal removal in HeLa cells after cytoplasm pre-extraction (CE) and treated in vitro with RNase III and RNase H where indicated. More than 500 total cells from three independent experiments were considered. The median of each population is shown. Boxes and whiskers indicate 25–75 and 10–90 percentiles, respectively. ***P<0.001 (Mann–Whitney U-test). Data information: Black stars denote significant increases, whereas red stars denote significant decreases. ª2019 The Authors EMBO reports 20:e47250 |2019 5of 15 Sonia Barroso et al EMBO reports and the recent observation in yeast that certain chromatin alterations are required for R loop-driven genetic instability [31], we determined by IF whether DNA–RNA hybrids accumulated in DDR-deficient cells were also accompanied by these chromatin marks. As can be seen in Fig 4A, the proportion of cells with elevated H3S10-P foci significantly increased after depletion of most of the factors analyzed. Importantly, such an increase was reversed by RNase H overexpression in at least five of them siC siHUS1 siRAD1 siRAD9A siRAD17 siTOPBP1 siMDC1 siCHK1 siATR siCHK2 siATM siUBE2B siRAD18 - + - + - + - + - + - + - + - + - + - + - + - +- + RNH1 *** *** * ** *** *** *** *** ** *** ***** Comet tail moment (a.u.) ** siC siRAD1 siRAD9A siRAD17 siCHK1 siATR siATM siUBE2B siRAD18siCHK2 +RNH1 -RNH1 +RNH1 -RNH1 A B ** 0 5 10 15 20 75 µm Figure 3. DNA–RNA hybrids are a source of DNA breaks in DDR-deficient cells. A Representative images of single-cell alkaline gel electrophoresis (comet assay) of HeLa cells transfected with the indicated siRNAs and either pEGFP-C1(RNH1  )or pEGFP-M27 (RNH1 + ). B Comet tail moment from single-cell alkaline gel electrophoresis (comet assay) of HeLa cells transfected with the indicated siRNAs and either pEGFP-C1(RNH1  )or pEGFP-M27 (RNH1 + ). More than 250 total cells were considered. The mean SEM of the median from at least three independent experiments is shown, except for RAD1, RAD17, and UBE2B(n=2). *P<0.05,**P<0.01, ***P<0.001 (one-tailed unpaired t-test). Black stars denote significant increases, whereas red stars denote significant decreases. 6of 15 EMBO reports 20:e47250 |2019 ª2019 The Authors EMBO reports Sonia Barroso et al (siRAD9A, siRAD17, siCHK2, siATM, and siUBE2B), indicating a tight link between DNA–RNA hybrids and H3S10-P. An increase in H3K9me2 nuclear intensity that was significantly reduced by RNH1 overexpression was also observed in cells depleted of the three selected DDR pathways (siATR, siATM, and siUBE2B) (Fig 4B). Importantly, no major changes were observed in the cell cycle distribution in cells depleted of the selected DDR factors (Fig EV3A). These results indicate that the DNA–RNA hybrid accumulation observed is linked to chromatin alterations. Differential effects on replication fork progression upon depletion of DDR factors Compelling evidence supports that R loop accumulation perturbs DNA replication and that this is a major cause of R loop-induced genetic instability [2,51–53]. Thus, we analyzed replication fork dynamics by DNA combing in the selected DDC- and PRR-deficient cells. In agreement with the reported role for ATR/CHK1 in fork progression [54], a decrease in fork velocity and track length was observed after depletion of factors of the ATR/CHK1 pathway but not after depletion of ATM/CHK2 or the PRR factors (Figs 5A and EV4). However, this decrease was not suppressed by RNase H overexpression. Therefore, we analyzed fork asymmetry as a measurement for R loop-mediated replication stalling. As shown in Fig 5B, we observed a significant increase in fork asymmetry after depletion of RAD1, RAD9A, RAD17, and CHK1. Although not significant, a similar tendency was observed after ATR depletion in agreement with increased fork stalling in the absence of a proper ATR/CHK1 checkpoint response. This increase was lost after RNase H overexpression in siRAD1-, siRAD9A-, and siRAD17-treated cells, supporting that although DNA–RNA hybrids are obstacles to replication fork progression, they are not the only kind of spontaneously occurring obstacles, consistent with our actual knowledge [55]. By *** ** ********* siC siRAD1 - + - + siRAD9A siCHK1 - + - + siATR siATM - + - + siUBE2B siRAD18 - + - + RNH1 AB H3K9me2 nuclear intensity (relative values) Cells with more than 5 H3S10-P foci (%) 0 5 10 15 ** ** * ** ** * * *** ** RNH1 siC siRAD1 siRAD9A siRAD17 siCHK1 siATR siCHK2 siATM siUBE2B siRAD18 -+ -+ -+ -+ -+ -+ -+ -+-+ -+ H3K9me2 Merge +DAPI + GFP Merge +DAPI + GFP siC -RNH1 siC +RNH1 siATR -RNH1 siATR +RNH1 siC -RNH1 siC +RNH1 siATM -RNH1 siATM +RNH1 H3S10-P 0 200 400 600 800 *** *** *** *** *** *** 25 µm 25 µm ** * Figure 4. DDR deficiencies lead to DNA–RNA hybrid-dependent accumulation of histone H3S10-P and H3K9me2chromatin marks. A Representative images and percentage of HeLa cells with more than 5H3S10-P foci after transfection with the indicated siRNAs and either pEGFP-C1(RNH1  )or pEGFP-M27 (RNH1 + ). Mitotic cells were excluded for the analysis by DAPI staining. More than 300 total cells were considered. Data represent mean SEM from three independent experiments. B Representative images and H3K9me2nuclear signal intensity of HeLa cells transfected with the indicated siRNAs and either pEGFP-C1(RNH1  ) or pEGFP-M27 (RNH1 + ). At least three experiments were performed. The median of each population in a representative experiment with at least 100 cells per condition is shown. Boxes and whiskers indicate 25–75 and 10–90 percentiles, respectively. Data information: *P<0.05,**P<0.01, ***P<0.001 (one-tailed paired t-test). Black stars denote significant increases, whereas red stars denote significant decreases. ª2019 The Authors EMBO reports 20:e47250 |2019 7of 15 Sonia Barroso et al EMBO reports *** *** *** *** *** *** * *** * ** *** * 0 2 4 6 8 10 Velocity (kb/min) ** *** *** *** * Asymmetry (longest/shortest) 1 10 100 A B siC siRAD1 siRAD9A siRAD17 siCHK1 siATR siCHK2 siATM siUBE2B siRAD18 - + - + - + - + - + - + - + - + - + - + RNH1 siC siRAD1 siRAD9A siRAD17 siCHK1 siATR siCHK2 siATM siUBE2B siRAD18 - + - + - + - + - + - + - + - + - + - + RNH1 50Kb 50Kb Figure 5. 8of 15 EMBO reports 20:e47250 |2019 ª2019 The Authors EMBO reports Sonia Barroso et al contrast, the fork asymmetry observed after depletion of ATM, CHK2, or the PRR factors RAD18 or UBE2B was similar to that of the control cells, which implies an important difference between the impact on replication of the DNA–RNA hybrids accumulated by depletion of the two groups of DDR factors, ATR/CHK1 and 9-1-1, on the one hand and ATM/CHK2 and PRR on the other. Our results suggest that DNA–RNA hybrids formed before replication in the absence of DDC factors ATR/CHK1 and the 9-1-1 complex, thus potentially causing replication fork stalling. However, DNA–RNA hybridization might also be promoted by unrepaired DSBs (siATM and siCHK2 cells), in which case no additional effect is expected regarding replication fork stalling, or at post-replicative ssDNA gaps (those accumulated in siUBE2B and siRAD18) after replication fork passage. During replication, spontaneous DNA obstacles including DNA– RNA hybrids should be frequently encountered by replisomes, potentially causing fork stalling or even breakage. In principle, if DNA–RNA hybrids are stimulated by breaks [56] they could be enriched at such broken forks. However, fork stalling can also lead to post-replicative DNA gaps, which are potential substrates for RNA hybridization. Therefore, we hypothesized that DNA–RNA hybrids might be differently formed along the cell cycle. To test this hypothesis, we analyzed DNA–RNA hybrid accumulation in whole cells in the different cell cycle phases. We measured the S9.6 IF intensity of whole cells treated with RNase III by flow cytometry and observed a significant increase from G1 to S and from S to G2 phases in all samples (siC-, siATR-, siATM-, and siUBE2B-treated cells) (Fig 6A and B). This tendency was more noticeable when DDC or PRR pathways were compromised (Fig 6C). Similar results were obtained with cytoplasm preextracted cells (Fig EV3B) as well as after sorting these preextracted cells in two populations (before and after replication) and subjecting them to S9.6 IF (Fig EV3C). These results agree with spontaneous DNA–RNA hybrids being stimulated after DNA replication, thus creating a specific target for the ATR and PRR- mediated DDR response, and are consistent with the idea that ssDNA gaps could efficiently hybridize with RNA. Discussion By screening an siRNA library targeting DDR genes, we have uncovered three DDR pathways that are important for the prevention of DNA–RNA hybrid accumulation in human cells. These are, the 9-1- 1/ATR/CHK1 pathway, known to respond to ssDNA accumulation; the ATM/CHK2 pathway, which responds to DSBs; and the PRR pathway, which triggers the filling of ssDNA gaps left in the daughter strands after replication bypass of damaged DNA. Despite its mixed specificity for DNA–RNA hybrids [48], the S9.6 antibody has enabled us to detect new functions in the DDR whose depletion leads to DNA–RNA hybrid accumulation and genetic instability by focusing on the top candidates of our screening (Fig 1). Importantly, we have reproducibly seen that the depletion of FA genes leads to an increase in S9.6 signal, in agreement with previous reports [25,28]. The fact that 9-1-1/ATR/CHK1 depletion led to DNA–RNA hybrid accumulation and that RNase H partially reduced the damage observed (Fig 7A) indicates that DNA–RNA hybrids are a frequent source of replication-associated DNA damage, likely through transcription–replication conflicts [57]. Indeed, DNA–RNA hybrids have been shown to accumulate at human common fragile sites (CFSs) [34,58,59], which suggest that they can also promote fragility at these sites. Our results suggest that persistent DNA–RNA hybrids themselves or some derivative intermediates could be sensed during replication. In agreement, R loop-accumulating yeast cells have been shown to activate the S-phase checkpoint and to require it for survival [60] and head-on transcription–replication collisions enhanced by R loops have been recently shown to activate ATR in human cells [21]. Although R loop-driven replication fork stalling is likely the reason behind ATR activation, the displaced ssDNA of R loops in principle could also promote local ATR activation as it has been reported at centromeres in mitotic cells [61]. Together with our recent data [11], our results support that DNA–RNA hybrids not necessarily impact replication fork velocity. Indeed, so far, both faster and slower forks have been detected in conditions of transcription–replication conflicts. Thus, slower fork speed was detected in cells depleted for FACT [26], histone H1 [62,63], ASF/SF2, and TOP1 [23], whereas faster fork speed was reported in SIN3A- and THOC1-depleted cells [7,11]. Instead, obstacles in the DNA should cause replication fork stalling. Indeed, fork asymmetry has been reported in all cases of cells accumulating R loops and analyzed, regardless of whether forks move faster (such as THOC1- or SIN3- depleted cells) or slower (such as in histone H1-, ASF/SF2-, or TOP1-depleted cells). In support of the conclusion that R loops are a frequent source of spontaneous fork stalling, we observed RNase H-sensitive fork stalling after 9-1-1/ATR/CHK1 depletion (Fig 7A). ATM/CHK2 might also have a role in R loop resolution. However, the accumulation of DNA–RNA hybrids observed in ATM/CHK2-depleted cells does not seem to be a major problem for replication fork progression, as we were not able to detect any increase in fork asymmetry (Fig 7A). Also, the fact that RNase H overexpression had no effect on the number of breaks induced by ATM/CHK2 depletion (Fig 7A) suggests that most breaks occurring in the absence of ATM/CHK2 are independent on DNA–RNA hybrids. Given the views and recent observations supporting that DNA breakage, whether single- or double-stranded, is a driving force for DNA–RNA hybrid formation [15,56,64–66], the accumulation of DNA–RNA hybrids in ATM/CHK2-depleted cells might rather be a consequence of such unrepaired DSBs, which would not imply any additional consequences in fork progression. ◀Figure 5. Differential effects on replication fork progression upon depletion of DDR factors. A Fork velocity as measured by DNA combing assay in HeLa cells transfected with the indicated siRNAs and either pEGFP-C1(RNH1  ) or pEGFP-M27 (RNH1 + ). More than 200 tracks were considered except for RAD1+ RNH1(n=125), ATR +RNH1(n=182), and RAD18 + RNH1(n=146). B Fork asymmetry as measured by DNA combing assay in HeLa cells transfected with the indicated siRNAs and either pEGFP-C1(RNH1  ) or pEGFP-M27 (RNH1 + ). From 40 to 200 total measurements were considered for each candidate. Data information: Median values are indicated. Boxes and whiskers indicate 25–75 and 10–90 percentiles, respectively. *P<0.05,**P<0.01, ***P<0.001 (Mann– Whitney U-test). Black stars denote significant increases, whereas red stars denote significant decreases. ª2019 The Authors EMBO reports 20:e47250 |2019 9of 15 Sonia Barroso et al EMBO reports