ADAR2-mediated RNA editing of DNA:RNA hybrids is required for DNA double strand break repair

The maintenance of genomic stability requires the coordination of multiple cellular tasks upon the appearance of DNA lesions. RNA editing, the post-transcriptional sequence alteration of RNA, has a profound effect on cell homeostasis, but its implication in the response to DNA damage was not previously explored. Here we show that, in response to DNA breaks, an overall change of the Adenosine-to-Inosine RNA editing is observed, a phenomenon we call the RNA Editing DAmage Response (REDAR). REDAR relies on the checkpoint kinase ATR and the recombination factor CtIP. Moreover, depletion of the RNA editing enzyme ADAR2 renders cells hypersensitive to genotoxic agents, increases genomic instability and hampers homologous recombination by impairing DNA resection. Such a role of ADAR2 in DNA repair goes beyond the recoding of specific transcripts, but depends on ADAR2 editing DNA:RNA hybrids to ease their dissolution. One sentence summary: DNA recombination requires RNA editing of DNA:RNA hybrids to ease their melting and facilitate DNA end resection


INTRODUCTION
Cells are continuously challenged by DNA damage. Among all kinds of insults that a DNA molecule has to deal with, double strand breaks (DSBs) are the most dangerous. Indeed, just one unrepaired DSB is enough to either kill or terminally arrest cells. For these reasons when DSBs are formed, a complex cellular response -the DNA damage response (DDR) -is triggered in order to ensure the proper repair of such a threat to genomic integrity 1 .
There are several pathways that can be used in order to repair a DSB and the choice between them is highly regulated. An eukaryotic cell can repair a DSB either by the simple re-ligation of the DNA ends (a process known as Non-Homologous End-Joining, NHEJ) 2 or by a homology-driven repair event. There are different routes among the repair pathways that use homologous regions for repair, all of which are grouped in a process called homologous recombination (HR) 3 . All HR events share a first biochemical step called DNA resection, which is key to decide the pathway that will be eventually used to repair the DSB 4,5 . This process consists of the nucleolytic degradation of the DNA ends of the break that produces tails of 3´ended single stranded DNA (ssDNA) that are rapidly protected by the RPA protein complex.
In recent years, the importance of RNA and RNA-related factors in DNA repair has become clear [6][7][8][9] . Indeed, many RNA-related proteins have been shown to be targets of the DNA damage-induced post-translational modifications [10][11][12] . Also, direct roles of specific RNA-related factors in DNA repair have been recently reported (for a review see 9 ). Moreover, the RNA molecule itself seems to impact on DNA repair. Several labs have shown the formation of DNA:RNA hybrids around DSBs in different eukaryotes, either dependent on previous transcription 13,14 or upon de novo transcription of the broken chromatin 15,16 . The relevance of such RNA molecules is still under debate, with both pro-and anti-repair effects ascribed to them 9 .
An important co-transcriptional RNA modification that, so far, has not been extensively studied in its putative relationship with DNA repair and the response to DNA damage is RNA editing. This process alters RNA sequences by the action of specific deaminases that convert one base into another. Every mammalian transcript can be subjected to RNA editing [17][18][19] . RNA editing can be classified in several categories 20 , including Adenosine-to-Inosine (A-to-I) deamination, which is accomplished by a family of RNA-Specific Adenosine Deaminase known as ADARs 18,19 . This family is formed by ADAR1, ADAR2 (also known as ADARB1) and ADAR3; however, only ADAR1 and ADAR2 have been shown to present catalytic activity. A-to-I deamination is the most abundant form of RNA-editing in mammals and   d  e  f  e  c  t  s  i  n  t  h  i  s  p  r  o  c  e  s  s are associated with human diseases, such as disorders of the central nervous system 21 or paediatric astrocytomas 22 . Only limited information has been published regarding the connection of A-to-I editing and DNA damage, albeit at least the mRNA of NEIL1, has been shown to be re-coded by ADAR1 to alter its enzymatic properties 23 . Moreover, Ato-I editing has been proposed to be involved in the pathogenesis of cancer 24,25 .
Here we show that the general pattern of ADAR2-mediated A-to-I editing changes upon DSB formation. Such changes depend on the DDR, specifically the ATR kinase and the resection protein CtIP. As a consequence, ADAR2 is required for maintenance of genomic integrity and, specifically for DNA end resection and HR.
Strikingly, mRNAs from either resection-related or recombination-related genes are not affected by ADAR2. Instead, ADAR2 role in resection is related with its ability to edit DNA:RNA hybrids. Not only such structures increase when ADAR2 is depleted, but this protein physically and functionally interacts with the BRCA1-SETX complex for this role.

siRNAs, plasmids and transfections
siRNA duplexes were obtained from Sigma-Aldrich or Dharmacon (Supplementary   Table S2) and were transfected using RNAiMax Lipofectamine Reagent Mix (Life Technologies), according to the manufacturer's instructions. RNWG and RNAG was a gift from Dr. Jantsch´s lab 27 . The GFP-ADAR2 and GFP-ADAR2 mutant (GFP-ADAR2-E/A-) plasmids were previously described 26 . RNaseH1 overexpression was achieved with the pCDNA3-RNAseH1 vector 28 , and pCDNA3 (Invitrogen) was used as a control. Plasmid transfection of U2OS cells was carried out using FuGENE 6 Transfection Reagent (Promega) according to the manufacturer's protocol, with the exception of pCDNA3-RNaseH1 and pCDNA3 plasmids that were transfected using Lipofectamine 3000 (Invitrogen) according to the manufacturer's instructions.

HR and NHEJ analysis
U2OS cells bearing a single copy integration of the reporters DR-GFP (Gene conversion) 29 , SA-GFP (SSA) 30 or EJ5-GFP (NHEJ) 30 were used to analyse the different DSB repair pathways. In all cases, 50,000 cells were plated in 6-well plates in duplicate. One day after seeding, cells were transfected with the indicated siRNA and the medium was replaced with fresh one 24h later.

SDS-PAGE and western blot analysis
Protein extracts were prepared in 2× Laemmli buffer (4% SDS, 20% glycerol, 125 mM Tris-HCl, pH 6.8) and passed 10 times through a 0.5 mm needle-mounted syringe to reduce viscosity. Proteins were resolved by SDS-PAGE and transferred to low fluorescence PVDF membranes (Immobilon-FL, Millipore). Membranes were blocked with Odyssey Blocking Buffer (LI-COR) and blotted with the appropriate primary antibody and infra-red dyed secondary antibodies (LI-COR) (Table S3 and S4).
Antibodies were prepared in blocking buffer supplemented with 0.1% Tween-20.
Membranes were air-dried in the dark and scanned in an Odyssey Infrared Imaging System (LI-COR), and images were analysed with ImageStudio software (LI-COR).

Cell cycle analysis
Cells were fixed with cold 70% ethanol overnight, incubated with 250 μ g ml -1 RNase A (Sigma) and 10 μ g ml -1 propidium iodide (Fluka) at 37ºC for 30 min and analysed with a FACSCalibur (BD). Cell cycle distribution data were further analysed using ModFit LT 3.0 software (Verity Software House Inc).

UV laser micro-irradiation
Cells were micro-irradiated using a wide-field Angström's microscope (Leica) equipped with a Micropoint pulsed dye laser of 365 nm (Photonic instruments, Inc. Immunofluorescence images were acquired using a Leica DM6000 wide-field microscope equipped with a DFC390 camera at x63 magnification using the LAS AF software (Leica). Microscopy data analysis was performed using the Metamorph v7.5.1.0 software (Molecular Probes).

ICGC data retrieval and analysis
Mutations sets were retrieved from the International Cancer Genome Consortium (ICGC) Data Portal MALY-DE and CLLE-ES datasets. ADAR2 expression levels from each donor were obtained using the UCSC Xena web tool. Open-access somatic mutations information from each mutation set was obtained by comparing each mutation set with the latest release of the Aggregated Somatic Mutation VCF file by the ICGC using custom python scripts. The percentage of A to G and T to C mutations was calculated as the quotient between the number of A to G mutations and T to C mutations to the total number of mutations from each donor set.

Statistical analysis
Unless specifically specified, statistical significance was determined with a Student's t-test using PRISM software (Graphpad Software Inc.). Statistically significant differences were labelled with one, two or three asterisks if p < 0.05, p < 0.01 or p < 0.001, respectively.

RNA editing changes after DNA damage
As previously mentioned, a crosstalk between RNA metabolism and DNA repair has been extensively documented 9 , but a connection between DNA repair and RNA editing has not been extensively analysed. Thus, we wanted to study whether the appearance of DNA damage had any effect on RNA editing. In order to explore this possibility, we used a previously published reporter system (RNAG) that measures levels of RNA editing using the accumulation of the fluorescent proteins GFP and RFP 27 . This system bears both the RFP and GFP ORFs in a single transcript, with a stop codon between them ( Figure 1A). So, cells bearing such reporter express RFP constitutively, but GFP is only produced if an RNA editing event changes the A of the stop codon to an I ( Figure 1A) 27 . Therefore, the number of red cells that are also green indicates the efficiency of RNA editing. As a control to discard other effects non-related to editing on this system, we used the RNWG control reporter, in which the stop codon is pre-edited so all cells bearing the construct fluoresce, indeed, in red and green 27 . In U2OS cells stably transfected with the reporter we observed that DNA damage induced by ionizing radiation increased GFP expression by 50% specifically in the RNAG reporter and not in the RNWG control, in agreement with a DNA damage-stimulation of RNA A-to-I editing in this system ( Figure 1B). Similar results were obtained when using the DNA damage inducing drug camptothecin ( Figure 1C), where we could observe a dose-dependent effect on RNA editing stimulation. One possibility is that DNA damage induces the accumulation of the A-to-I editing machinery, namely ADAR1 and ADAR2 enzymes, thus increasing this process. However, neither of these proteins was upregulated, but slightly downregulated, upon exposure to IR (Supplementary Figure S1A). Then, in order to confirm this was a canonical induction of RNA editing, we depleted the A-to-I editing machinery. To choose which member of the ADAR family to downregulate, we revisited the data we obtained in a previous genome-wide screening for factors that unbalance the choice between DSB repair pathways 36 . Interestingly, both ADAR1 and ADAR2, but not the catalytically inactive ADAR3, skewed DSB repair towards end-joining (Supplementary Figure S1B), and this was not due to changes in the cell cycle (Supplementary Figure S1C). However, the effect was more prominent and clearer upon ADAR2 depletion. Indeed, downregulation of ADAR2 severely compromised both the basal and the DNA damage-induced expression of the GFP in the RNAG ( Figure 1D; for ADAR2 depletion efficiency see Supplementary Figure S2A), but, as expected, not in the RNWG control reporter (Supplementary Figure S2B).
To better understand this phenomenon, we decided to look for DDR factors that affected the DNA damage-induced RNA editing. Recently, we have found that CtIP, a core DNA end resection factor that is also required for ATR activation, plays additional roles in DNA damage-induced RNA splicing 37 . Interestingly, we could see that CtIP downregulation specifically eliminated the DNA damage-dependent induction of RNA editing without affecting the basal levels ( Figure 1D). Again, CtIP depletion did not alter GFP levels in the control RNWG system (Supplementary Figure S2B). We could also complement this effect with the expression of siRNA-resistant flag-tagged CtIP in CtIP depleted cells, to the same extent as the control cells transfected with a nontargeting siRNA, even though overexpression of FLAG-CtIP on its own reduced the intensity of this DNA damage-induced phenotype (Supplementary Figure S2C).
The general response to DNA damage is mainly controlled by the activation of two related apical kinases, ATM and ATR 1 . Thus, we also tested if any of them was required for the induction of RNA editing upon irradiation. Interestingly, ATM inhibition did not affect DNA damage-induced RNA editing, while ATR inhibition decreased the DNA damage-induced editing increase ( Figure 1E). This agrees with the notion that ATR and CtIP act on the same branch of the DNA damage checkpoint in response to DSBs 38 . The lack of response with the ATM inhibitor could be explained by a compensation by another member of the PIKK family, most likely DNA-PK. Along those lines, the ATR effect could also be affected by this phenomenon. Thus, we repeated the experiment with ATM, ATR and DNA-PK inhibitors in different combinations ( Figure 1F). As shown, inhibition of ATR suppressed the DNA damageinduction of RNA editing, regardless of the presence of the inhibitors of ATM or DNA-PK. Interestingly, chemical inhibition of DNA-PK showed a limited increase in the basal levels of RNA editing, but importantly the exposure to DNA damage still provoked a hyper-activation of the process. Notably, concomitant inhibition of both DNA-PK and ATM abolished the induction of RNA editing caused by irradiation. Thus, it seems that those two kinases could have an overlapping role in this phenomenon.

ADAR2 depletion causes genomic instability and DNA damage sensitivity
To understand the consequences of reduced A-to-I RNA editing for genomic stability, we decided to globally reduce such RNA modifications. Based on our previous data with ADAR2 ( Figure 1D and Supplementary Figure S1B), we decided to use the downregulation of this protein as a tool to reduce A-to-I RNA editing. Strikingly, and in agreement with a role in maintaining genomic stability, the depletion of ADAR2 impaired DSB repair, measured as the presence of γH2AX foci 24h after irradiation.
Spontaneous DNA damage accumulated in the absence of any exogenous genotoxic agent in ADAR2-depleted cells (Figure 2A). Confirming a DSB repair impairment, the disappearance of γH2AX foci upon exposure to ionizing radiation was delayed ( Figure   2B). Indeed, repair levels of DSBs at 6 and 24 hours after irradiation in ADAR2depleted cells were similar to those observed after downregulation of the critical repair factor CtIP ( Figure 2B). Interestingly, and in agreement with an increased burden of spontaneous DNA damage, in the absence of ADAR2 we observed a significative increase of BRCA1 foci in cells unchallenged with any genotoxic agents ( Figure 2C). A similar effect was observed upon ADAR1 downregulation ( Figure 2D). Furthermore, micronuclei accumulated at high levels in ADAR2-depleted cells, regardless of the exposure to an external source of DNA damage ( Figure 2E). Finally, and confirming a role of ADAR2 in DNA repair and the maintenance of genomic stability, its depletion rendered cells hypersensitive to DSBs-inducing agents such as ionizing radiation or camptothecin ( Figures 2F and 2G).

ADAR2 depletion affects DNA repair pathway choice
Next, we decided to test a possible requirement of A-to-I RNA editing for DSB repair. As mentioned, ADAR2 depletion skewed the balance between HR and NHEJ towards the latter (see reference 36 ; Supplementary Figure S1B To confirm that the observed phenotype in DNA resection was truly due to the reduction of ADAR2 levels and not to an indirect off-target effect, we studied RPA foci formation in cells bearing siRNA-resistant, GFP-tagged variants of ADAR2. Indeed, the resection impairment caused by depletion of ADAR2 was rescued by wild-type GFP-ADAR2 ( Figure 4B and Supplementary Figure S3C). Importantly, this rescue was not observed with the expression of a catalytically-dead version of the protein (GFP-ADAR2E-A) ( Figure 4B), arguing that ADAR2 deaminase activity was required for processive resection. Moreover, we could reproduce the resection defect in U118 cells 26 , a glioblastoma cell line that is defective for ADAR2 expression, when compared with the same cells complemented with a wild-type copy of the gene, but not a catalyticallydead mutant ( Figure 4C).
To validate these observations, we analyzed recruitment of RPA to DSBs by other means. U2OS cells depleted for ADAR2, or CtIP as a control, were laser microirradiated and immunostained with antibodies against RPA and γ H2AX to identify the irradiated areas. The percentage of γ H2AX-positive stripes that were co-stained by RPA was determined ( Figure 4D). In agreement with our previous results, depletion of ADAR2, or CtIP, significantly diminished the presence of RPA at the irradiated areas.
Finally, in order to analyse in more detail the resection defect related to ADAR2 depletion and to investigate whether only resection initiation was impaired or if resection processivity was also compromised, we used SMART, a high resolution technique that measures resection in individual DNA fibres 31, 40 . As seen in Figure 4E, the length of ssDNA fibres formed during the resection process was reduced upon ADAR2 depletion. Again, similar results were observed upon ADAR1 downregulation (Supplementary Figure S3D), reinforcing the connection between A-to-I editing and DNA end resection.
Altogether, these results confirm that RNA editing by ADAR proteins facilitates DNA end resection at DSBs.

The role of ADAR2 on resection does not rely on the recoding of mRNAs that encode resection factors
Next, we wondered how this RNA editing activity might be needed for DNA end processing. We studied the recruitment DSB repair factors, such as 53BP1, BRCA1 and CtIP, to DNA damage foci shortly after DNA damage induction upon ADAR2 depletion. Notably, neither of them was affected in cells exposed to ionizing radiation or laser micro-irradiation (Supplementary Figure S4A-C). Then, we wondered if ADAR2 was specifically editing mRNAs that code for resection factors. To analyze this possibility, we exposed the ADAR2-defective cell line U118 to ionizing radiation or mock treatment, isolated total RNA, and sequenced it. U118 cells complemented with either wild-type ADAR2 or a catalytically-dead mutant were also processed in parallel.
The levels of ADAR2 mRNA in the different samples are shown in Supplementary Figure S4D. As expected, U118 cells complemented with wild-type ADAR2 showed a higher efficiency in editing the coding codons of known ADAR2 targets, expressed as the weighted average over all known recoding sites, known as the Recoding Editing Index (REI) 35 . Instead, non-complemented U118 cells showed little recoding editing, regardless of whether or not exposed to ionizing radiation ( Figure 4F). Equally expected, the expression of a catalytically-dead enzyme also showed almost no recoding of mRNAs ( Figure 4F), despite the fact that such variant was expressed almost 3 times more than the wild-type ADAR2 (Supplementary Figure S4D). Thus, only the expression of catalytically-active enzyme led to the expected ADAR2-dependent recoding due to editing of specific codons ( Figure 4F). Interestingly, although some specific coding codons were edited more efficiently upon irradiation than in mock treated cells, we observed a general decrease in the recoding editing efficiency of known ADAR2 substrates upon exposure to DNA damage ( Figure 4F and Supplementary Table S1). Moreover, we could not observe any change in editing of mRNA from genes that code for recombination or resection factors either in cells exposed to DNA damage or in undamaged cells (Supplementary Table S1). Thus, we

ADAR2 facilitates resection over DNA:RNA hybrids
We hypothesized that the ADAR2-depletion related phenotype in DNA resection could be caused by a direct effect of ADAR2 on an RNA molecule located in the vicinity of the break that would act as a physical barrier for DNA end processing.
The presence of DNA:RNA hybrids close to DSBs has been documented from yeast to mammals, both as pre-existing R-loops, R-loops formed as a consequence of breaks in transcribed regions or as DNA:RNA hybrids resulting from de novo transcription of resected DNA ends 9,13,16,[41][42][43] . The actual effect of such DNA:RNA hybrids in resection is controversial, with both pro-and anti-resection effects described 9,16,44 . Importantly, ADAR2 has been proposed to recognize and edit DNA:RNA hybrids in vitro 45 . In order to define whether ADAR2 involvement in DNA end resection depended on the presence of DNA:RNA hybrids, we repeated the resection assay in the presence or absence of ectopically overexpressed RNaseH1, an enzyme that degrades the RNA moiety of such structures 46 . ADAR2 depletion and RNaseH1 overexpression efficiency are documented in Supplementary Figure S5A. Strikingly, the overexpression of RNaseH1 in U2OS reverted ADAR2 resection phenotype as measured by RPA foci accumulation ( Figure   5A and Supplementary Figure S5B). Indeed, the mere overexpression of RNaseH1 facilitated RPA foci formation even in cells transfected with a control siRNA, arguing that DNA:RNA hybrids act generally as physical barriers for the resection process, and that ADAR2 helps overcome such roadblocks. To confirm this finding, we studied the recruitment of RPA to damaged chromatin in HeLa cells depleted of ADAR2 and overexpressing RNaseH1. Again, such overexpression rescued the resection phenotype of ADAR2 depletion ( Figure 5B). The same was observed, albeit only partially, when laser micro-irradiation experiments were performed ( Figure 5C). In none of those cases, the increase in RPA-positive cells was due to changes in cell cycle profile when the RNaseH1 was overexpressed (Supplementary Figure S5C). Then, to assess whether ADAR2 helped remove DNA:RNA hybrids, we analyzed the accumulation of such structures by immunofluorescence using the DNA:RNA hybrid-specific antibody S9.6 47 .
As shown in Figure 5D, depletion of ADAR2 increases the nuclear signal with that antibody. To rule out the contribution of other nucleic acid structures to the increase in S9.6 signal, we overexpressed RNaseH1 and observed a significant reduction in the staining ( Figure 5D). Of relevance, ADAR2 has been previously found to be a part of the so-called "DNA:RNA hybrid interactome" 48 . Therefore, we decided to test if ADAR2 could interact directly with DNA:RNA hybrids. Indeed, ADAR2 was specifically immunoprecipitated using the S9.6 antibody, in a similar fashion as Senataxin (SETX), a helicase described to dissolve such hybrids and that has been shown to be recruited to DSBs at transcribed regions 13 ( Figure 5E). This immunoprecipitation was specific, as did not occur when a control antibody was used ( Figure 5E).
Taken together, our results suggest ADAR2 effect in resection was caused by DNA:RNA hybrid accumulation. A prediction of this model is that the decrease in homologous recombination caused by ADAR2 downregulation should be suppressed by RNaseH1 overexpression. In fact, recombination was almost completely restored in ADAR2-depleted cells when such enzyme was ectopically expressed ( Figure 5F). Such effect was not due to a general increase of recombination mediated by RNaseH1 overexpression, as it was not observed either in CtIP depleted or control cells.

Increase of DNA:RNA hybrids generally impairs DNA end resection
Due to our observation that DNA:RNA hybrids hampered resection in the absence of ADAR proteins, we wondered if other enzymes involved in the removal of hybrids also affected DNA end resection, therefore this represented a more general phenomenon. We decided to test SETX, due to its aforementioned relationship with DSBs. So, we checked if its loss of also affected RPA foci formation. Indeed, depletion of Senataxin also produced a DNA resection defect measured by RPA foci accumulation ( Figure 6A) and reduced the length of resected DNA ( Figure 6B). As seen for ADAR2 depletion, SETX downregulation also led to an increased burden of spontaneous DNA damage, measured as BRCA1 foci in unchallenged cells ( Figure 6C).

ADAR2 physically interacts with Senataxin and BRCA1
Our data suggested a possible role of ADAR2 in R-loop resolution, which would be key for facilitating DNA end resection and HR. Hence, we wondered if ADAR2 might interact with known players in the homeostasis of R-loops that are also connected with DSB repair. Many different enzymes have been associated with the removal of Rloops. Among them, we decided to focus on the BRCA1-SETX complex. These two proteins have been shown to cooperate in the elimination of R-loops in the 3' end of many transcribed genes 49 . Moreover, both proteins play roles in DNA end resection, as it has been previously established for BRCA1 40 and in this study for Senataxin ( Figure   6A-C). Thus, we tested whether ADAR2 might physically interact with BRCA1 and Senataxin. Using antibodies against BRCA1, we confirmed the previously shown interaction of this protein with Senataxin 49 and also established an interaction with ADAR2 ( Figure 6D). These interactions were not stimulated by the presence of DNA damage ( Figure 6D). Also, by co-immunoprecipitations with different antibodies we could observe the reciprocal interaction with both BRCA1 and SETX ( Figure 6E). The possibility that ADAR2 immunoprecipitation could be bringing down any proteins accumulating at sites of DNA double strand breaks was excluded, since we could not observe an interaction with the DNA-end binding protein Ku80, suggesting that ADAR2 interaction with BRCA1 and SETX was indeed specific. It is worth pointing out that these co-immunoprecipitations were performed in the presence of benzonase, thus they are not bridged by DNA. Moreover, immunoprecipitation using an antibody against SETX also allows the co-precipitation of ADAR2 ( Figure 6F). Therefore, we could confirm that these three proteins interact in a DNA damage-independent fashion, most likely to facilitate the removal of DNA:RNA hybrids globally. propose an RNA-editing DNA damage response (REDAR) that is essential for DNA repair, and specifically for HR, contributing to the maintenance of genomic integrity.

DISCUSSION
Specifically, we postulate that upon the triggering of the DNA damage response (DDR), REDAR is activated by ATR and CtIP. Then, ADAR2 is mobilized from its usual targets to new ones, most likely including DNA:RNA hybrids, in order to help with their dissolution. In this regard, ADAR2 action might have two non-mutuallyexclusive outcomes; on the one hand it can rapidly and transiently relocate to promote RNA editing at the sites of DSBs, aiding in the removal of DNA:RNA hybrids and facilitating DNA end resection; and on the other hand, it can increase the editing of a small fraction of yet undisclosed mRNAs, whose role in the DDR should be clarified in further studies. Interestingly, it has been previously shown that ADAR action can alter the biochemical properties of some DNA repair enzymes 23 . We envision that the transcriptional inhibition induced by the DDR 52 will cause a reduction of the normal cotranscriptional RNA editing 53  Editing of the DNA strand would greatly increase the mutagenesis associated with HR, something that might be deleterious for the cells, but has been observed due to the action of C-to-U deaminases in cancer 55 . Interestingly, the accumulation of N6methyladenosine modification on the RNA at DNA:RNA hybrids has also been shown to ease the dissolution of such structures 51 . As A-to-I editing is negatively influenced by m 6 A modifications 56,57 , the crosstalk between those two RNA post-transcriptional modification during DNA repair and in response to DNA damage will be worth exploring further. To integrate all our observations, we present a model in which ADAR2, BRCA1 and SETX facilitate the removal of hybrids genome-wide even in the absence of DNA damage, but in a way that is stimulated by the DDR during REDAR. When exposed to a source of DSBs, the levels of DNA:RNA hybrids and R-loops increase, as previously described by other authors [13][14][15][16] , and many of them specifically accumulates at break sites. Although some authors argue that these structures might favor the repair process 44 , other favor a view in which they act as roadblocks for repair (for a review see 43 ). This contradiction might simply stem from a differential effect of DNA:RNA hybrids depending on the timing of repair, i.e. very early events might need them but later they have to be eliminated, or the position of the hybrids, as discussed elsewhere 9,61 . Our data agree with the notion that at least some of them represent roadblocks for the progression of the repair machinery that have to be removed prior to repair ( Figure 6G Mechanistically, we suggest that ADAR2-mediated editing of DNA:RNA hybrids, an activity that has been confirmed in vitro 45 , might alter the sequence of the RNA strand creating ribo-Inosine (rI) and the deoxiribo-thymine (dT) mismatches ( Figure 6G-3).
The appearance of those mismatches will loosen up the interaction between the RNA and the DNA strand, facilitating the unwinding of the structure by the helicase activity of SETX ( Figure 6G-3,4), allowing the resection machinery to go through ( Figure 6G is supported by the regional government of Andalucia (Junta de Andalucía).

DATA AVAILABILITY.
All relevant data are included in the manuscript. Raw data will be provided upon request.