RRP6/EXOSC10 is required for the repair of DNA double-strand breaks by homologous recombination

The exosome is a multiprotein complex with ribonucleolytic activity that plays important roles in RNA metabolism, including processing of RNA precursors and surveillance of mRNA biogenesis (reviewed in Houseley et al., 2006; Chlebowski et al., 2013; Eberle and Visa, 2014). The eukaryotic exosome is composed of a core that is catalytically inactive (reviewed in Lorentzen et al., 2008; Januszyk and Lima, 2014). The RNases RRP6 (also known as EXOSC10 in mammals) and DIS3 (also known as RRP44) associate with this core and provide catalytic activity (reviewed by Lykke-Andersen et al., 2011). RRP6 and DIS3, however, have other interaction partners than the exosome core and can act independently of the exosome (Synowsky et al., 2009; Graham et al., 2009; Kiss and Andrulis, 2011).

The activity of the mRNA surveillance machinery of Saccharomyces cerevisiae has been linked to the DNA damage response (Hieronymus et al., 2004; Manfrini et al., 2014), and the human exosome, when complexed with the RNA and DNA helicase senataxin, is targeted to sites of transcription-induced DNA damage in human cells (Richard et al., 2013). However, the contribution of the exosome to the repair of DNA lesions has not been investigated.

The DNA-damage response (DDR) integrates signaling cascades that lead to the arrest of cell proliferation and the activation of the cellular DNA repair machineries (reviewed by Polo and Jackson, 2011). Double-strand breaks in the DNA (DSBs) trigger the recruitment and activation of the sensor protein kinases ATM or ATR to the damaged sites. These kinases phosphorylate a number of specialized substrates, including the histone variant H2AX in humans or its functional homolog H2Av in D. melanogaster (Madigan et al., 2002). The local phosphorylation of H2AX/H2Av in the vicinity of DSBs generates binding sites for adaptor proteins and DDR effectors. Two major DNA repair mechanisms, non-homologous end-joining (NHEJ) and homologous recombination, are responsible for the repair of DSBs. NHEJ is the predominant mechanism in mammalian cells, whereas homologous recombination depends on the availability of homologous sequences and is more frequent in Drosophila (Preston et al., 2006; Lieber, 2008). Regardless of the repair mechanisms used, the sequential recruitment of adaptor proteins and effector repair factors to the DSBs leads to the formation of cytologically detectable foci at the damaged DNA sites. These foci are known as ‘DDR foci’.

Short non-coding RNAs are required for the assembly of DDR foci and for DNA repair by homologous recombination (Francia et al., 2012; Wei et al., 2012; Gao et al., 2014). The fact that the exosome acts on virtually all types of RNA in eukaryotic cells led us to investigate whether the catalytic subunit of the nuclear exosome, RRP6/EXOSC10, plays a role in DSB repair. We show that RRP6/EXOSC10 interacts, directly or indirectly, with the homologous recombination factor RAD51 and is required for the proper assembly of DDR foci in fly and human cells. Moreover, experiments with a catalytically inactive RRP6 mutant suggested that the ribonucleolytic activity of RRP6 is required for this process. In summary, our results identify RRP6/EXOSC10 as a new player in the DDR and link RRP6/EXOSC10 to the homologous recombination pathway.

We used a Drosophila S2 cell line (S2-RRP6–V5) that expresses a V5-tagged version of RRP6 (Hessle et al., 2009) to analyze whether the interaction network of RRP6 was affected in response to DNA damage. The RRP6–V5 protein is imported into the nucleus, is able to engage in functional interactions and can be immunoprecipitated by an anti-V5 antibody (Hessle et al., 2009). We exposed S2-RRP6–V5 cells to gamma irradiation (γIR) and immunoprecipitated RRP6–V5 30 min after the irradiation. We identified 49 proteins that co-immunoprecipitated with RRP6–V5 in non-irradiated control cells (supplementary material Tables S1, S2). Only 23 of these 49 interactors were detected in the irradiated cells, which also revealed two unique interactions with the linker histone H1 and the histone variant H2Av (Fig. 1A). These results suggest that γIR rapidly remodels the interaction network of RRP6.

We then analyzed the distribution of RRP6–V5 by immunofluorescence using antibodies against the V5 tag and against the phosphorylated form of H2Av (γH2Av) as a marker for DDR foci (Fig. 1B,C). Exposure to γIR increased the percentage of cells with γH2Av-rich foci to almost 100% 10 min after the irradiation. At the same time, the distribution of RRP6–V5 became more granular, and 66% of the cells displayed distinct RRP6–V5-rich foci (Fig. 1B,C). The same relocalization was observed with an antibody directed against the endogenous RRP6 protein. Many γH2Av-rich foci overlapped (58%) or were juxtaposed (27%) with the RRP6–V5-rich foci (Fig. 1D,E). The distributions of γH2Av and RRP6–V5 in γIR-treated cells did not overlap completely, which is not surprising considering the many roles of RRP6 in RNA metabolism. Similar immunofluorescence experiments were performed with S2 cells that expressed RRP4–V5, a subunit of the exosome core (Hessle et al., 2009). The distribution of RRP4–V5 was not affected by the γIR treatment (Fig. 1C).

In another series of experiments, we asked whether the γIR-induced relocalization of RRP6 was dependent on the phosphorylation of H2Av. S2 cells were treated with CGK733, an inhibitor of ATM and ATR, and stained with antibodies against γH2Av or endogenous RRP6. Phosphorylation of H2Av was impaired in CGK733-treated cells, as expected. Furthermore, the formation of RRP6-rich foci was drastically reduced by the CGK733 treatment (Fig. 1F; supplementary material Fig. S1).

In summary, γIR induces a rapid and transient relocalization of RRP6, but not of RRP4, to sites of radiation-induced DNA damage. This relocalization is dependent on the activity of the DDR kinases ATM and ATR.

We next analyzed whether the kinetics of assembly and clearance of DDR foci were affected by depletion of RRP6. S2 cells were treated with doubled-stranded RNA (dsRNA) to deplete RRP6 by RNA interference (RNAi), and control cells were treated in parallel with dsRNA complementary to GFP (Fig. 2A). The γIR treatment caused a significant increase in the percentage of cells with DDR foci, and there were no differences between RRP6-depleted and control cells 1 h after irradiation (Fig. 2B,C). However, 20 h after the irradiation the fraction of cells that displayed prominent γH2Av-rich foci was significantly higher in the cells depleted of RRP6 (Fig. 2B,C), which suggests that RRP6 is required for efficient DNA repair.

We also analyzed whether the depletion of RRP6 affected the sensitivity of S2 cells to γIR, which would be expected if RRP6 were necessary for DNA repair. The growth rate of the RRP6-depleted cells irradiated with 10 Gy was almost 50% lower than that of GFP control cells (Fig. 2D), and RRP6-depleted cells irradiated with doses as high as 50 Gy did not recover (Fig. 2E). These experiments demonstrate that physiological levels of RRP6 are required for the efficient repair of DSBs, and that depletion of RRP6 renders S2 cells more sensitive to γIR.

The recruitment of the ATM and ATR kinases to the DSBs and the local phosphorylation of H2Av trigger the assembly of either the NHEJ or homologous recombination machineries, depending on the nature of the damage and the stage in the cell cycle. We hypothesized that RRP6 is involved in the homologous recombination pathway because homologous recombination is the predominant repair mechanism in D. melanogaster (Preston et al., 2002), because homologous recombination requires short RNAs (Gao et al., 2014) and because the homologous recombination replication factor RPA1 (also known as RFA1 or RpA-70 in Drosophila) is a putative component of the exosome interaction network in both S2 cells (supplementary material Table S2) and human cells (Lubas et al., 2011). We thus analyzed the recruitment of the homologous recombination machinery to DDR foci using an antibody against RAD51 (reviewed by Krejci et al., 2012) as a marker for homologous recombination. At 1 h after irradiation, 81% of the GFP control cells showed RAD51-rich foci whereas the frequency of RAD51-rich foci did not increase in RRP6-depleted cells (Fig. 3A,B). This effect was specific for RRP6, because the formation of RAD51 foci in cells depleted of RRP4 was the same as in the GFP controls. It is important to point out that the depletion of RRP6 did not reduce the levels of RAD51 protein (Fig. 3C). These observations led us to conclude that RRP6 is required for the efficient assembly of the homologous recombination machinery at the DSBs.

We hypothesized that RRP6 interacts physically with components of the DDR machinery based on the relocalization of RRP6 to DDR foci and on the fact that RRP6 is required for the recruitment of homologous recombination factors. To test this possibility, we carried out co-immunoprecipitation experiments in RRP6–V5-expressing S2 cells, and we searched for interactions of RRP6 with proteins detected in the mass spectrometry analysis, including histone H1, histone H2Av and RPA1/RFA1 (supplementary material Tables S1, S2). We could not demonstrate interactions of RRP6 with any of these proteins (data not shown). We found instead that RRP6 co-immunoprecipitates specifically with RAD51 (Fig. 3D). The association of RRP6 with RAD51 was further confirmed using a proximity ligation assay (PLA) to reveal protein–protein interactions in situ (Fig. 3E). The RAD51–RRP6 interaction was detected in both control and irradiated cells, but quantification of PLA data revealed a significant increase of RRP6–RAD51 association after exposure to γIR (right panel in Fig. 3E).

In summary, the experiments reported above support the conclusion that RRP6 interacts with RAD51 and is required for the recruitment of RAD51 to DDR foci.

We used a catalytically inactive RRP6 mutant to analyze whether the ribonucleolytic activity of RRP6 is necessary for homologous recombination. This mutant, RRP6-Y361A–V5, carries a single amino acid substitution in the active site (Phillips and Butler, 2003; Fig. 4A). In a first series of experiments, we overexpressed either RRP6–V5 or RRP6-Y361A–V5 in S2 cells (Fig. 4B), and analyzed the assembly and clearance of γH2Av foci (Fig. 4C). The expression of RRP6-Y361A–V5 did not affect the γIR-induced phosphorylation of γH2Av compared to the expression of the wild-type RRP6 protein, but 20 h after the irradiation the fraction of cells that displayed prominent γH2Av-rich foci was significantly higher in cells expressing the catalytically inactive RRP6-Y361A–V5 than in cells that expressed the wild-type RRP6–V5 (Fig. 4C). This result suggests that the activity of RRP6 is required for efficient DNA repair.

We also analyzed the effect of overexpressing RRP6-Y361A–V5 on the recruitment of RAD51 to DSBs. The formation of RAD51 foci was inhibited (from 64% to 27%) in cells that overexpressed the inactive RRP6-Y361A–V5 mutant (Fig. 4D). Moreover, overexpression of RRP6-Y361A–V5 significantly reduced the growth rate of the cells after irradiation with 10 Gy (Fig. 4E). Importantly, the overexpression of RRP6–V5 or RRP6-Y361A–V5 did not affect the abundance of the RAD51 protein (Fig. 4B), and the Y361A mutation did not inhibit the interaction of RRP6 with RAD51 (Fig. 4F). These results link the assembly of RAD51 foci to the catalytic activity of RRP6, and suggest that a 3′→5′ RNA degradation event is required for homologous recombination. The fact that RRP6 interacts with RAD51 and is recruited to DSBs supports the conclusion that RRP6 contributes to local RNA degradation at or near the DSBs.

Given the structural and functional conservation of the exosome, we hypothesized that the role of RRP6 in DNA repair is conserved in human cells. In a first series of experiments, we asked whether EXOSC10, the human RRP6 ortholog, is recruited to sites of radiation-induced DNA damage in HeLa cells. The use of homologous recombination is limited to the S and G2 phases of the cell cycle in mammalian cells. For this reason, and in order to be able to study homologous-recombination-related events, we studied the localization of EXOSC10 in HeLa cells synchronized in G2 using an antibody against the endogenous protein. EXOSC10 was predominantly concentrated in the nucleoli of non-irradiated cells (Fig. 5A), and γIR induced a progressive relocation of EXOSC10 from the nucleoli to the nucleoplasm. This relocation could be detected 15 min after irradiation and was very prominent after 45 min (Fig. 5A,B). EXOSC10 did not accumulate in prominent foci, nor was EXOSC10 concentrated to areas irradiated by ultraviolet laser irradiation (data not shown). However, PLA experiments with antibodies against the endogenous EXOSC10 and RAD51 proteins revealed that a significant increase in the interaction was induced by γIR (Fig. 5C). These results demonstrate that EXOSC10 is relocalized to the nucleoplasm in response to γIR and suggest that EXOSC10 interacts with RAD51.

We then asked whether EXOSC10 depletion by small interfering RNA (siRNA) in human cells (Fig. 6A) had any effect on the assembly and disassembly of DDR foci. The percentage of cells with γH2AX-positive foci was almost 100% at 1 h after γIR, regardless of the siRNA treatment (Fig. 6B,C). The γH2AX frequency recovered to background levels within 24 h in control cells treated with scrambled (Scr) siRNA, but was still high in EXOSC10-depleted cells, which implies a delay in the clearance of the DDR foci in the depleted cells. Furthermore, the RAD51-rich foci that formed in EXOSC10-depleted cells were fewer and did not colocalize with the γH2AX-rich foci (Fig. 6B,D). Importantly, depletion of EXOSC10 did not affect the levels of RAD51 (Fig. 6E) and did not cause arrest in G1 (data not shown), which would in itself suppress homologous recombination. These results support the conclusion that EXOSC10 is required for the recruitment of RAD51 to DSBs in human cells. In agreement with a defect in DNA repair, depletion of EXOSC10 increased the sensitivity of the cells to γIR (Fig. 6F). Similar immunofluorescence experiments were carried out with antibodies against 53BP1, a NHEJ factor. Depletion of EXOSC10 did not affect the recruitment of 53BP1 to sites of radiation-induced DNA damage (Fig. 6G), which suggests that the role of EXOSC10 in DNA repair is restricted to the homologous recombination pathway.

Cells depleted of EXOSC10 presumably contain large amounts of aberrant RNA that could engage in DNA–RNA hybrid formation and hijack RAD51 to DNA–RNA hybrids (Wahba et al., 2013). However, the depletion of EXOSC10 per se did not induce the appearance of RAD51-rich foci (Fig. 6D), which argues against indirect effects due to competition between DNA–RNA hybrids and DSBs.

Given the recently reported role of AGO2 in homologous recombination (Gao et al., 2014), we also considered the possibility that AGO2 is mislocalized in EXOSC10-depleted cells owing to increased levels of dsRNA, for example pre-rRNA processing intermediates, which could indirectly cause homologous recombination defects. However, immunofluorescence experiments did not reveal any effect of EXOSC10 depletion on the distribution of AGO2 (data not shown).

We transfected HeLa cells with a plasmid coding for the homing endonuclease I-PpoI to generate site-specific DSBs and further analyze the requirement of EXOSC10 for the recruitment of RAD51 to sites of DNA damage. I-PpoI has a cleavage site in the 28S rDNA (Monnat et al., 1999) and cells transfected with the I-PpoI expression plasmid often showed γH2AX-rich foci associated with the nucleolus (Fig. 7A). We used the I-PpoI system to determine whether EXOSC10 is recruited to DSBs, as suggested by the PLA experiment reported in Fig. 5C. HeLa cells were transfected with the I-PpoI expression plasmid and analyzed by chromatin immunoprecipitation followed by real-time quantitative PCR (ChIP-qPCR) using the anti-EXOSC10 antibody and PCR primers designed to amplify two regions located at each side of the I-PpoI-induced DSBs. The level of EXOSC10 detected at the DSBs in cells transfected with I-PpoI was significantly higher than that of mock-transfected cells. This increase was specific for the I-PpoI target site as shown by the fact that an unrelated genomic region was unaffected (Fig. 7B). In summary, these results show that EXOSC10 is recruited to I-PpoI-induced DSBs in HeLa cells.

Next we depleted EXOSC10 by siRNA as in Fig. 6 and analyzed the phosphorylation of H2AX and the recruitment of RAD51 to the I-PpoI-induced foci by immunofluorescence. In 63% of the control (Scr) cells, RAD51 was recruited to the I-PpoI-induced DSBs, whereas only 27% of the EXOSC10-depleted cells showed association of RAD51 with γH2AX-rich foci (Fig. 7C,D). We also analyzed the recruitment of RAD51 to I-PpoI-induced DSBs by ChIP-qPCR. In control (Scr) cells, the levels of RAD51 detected in the vicinity of the DSBs were between two and five times higher in cells transfected with I-PpoI than in mock-transfected cells, whereas depletion of EXOSC10 abolished the recruitment of RAD51 to the I-PpoI-induced DSBs (Fig. 7E). These results confirmed that EXOSC10 is required for the recruitment of RAD51 to DSBs.

RRP6 and EXOSC10 are predominantly nuclear proteins that participate in a plethora of RNA metabolic processes to which they contribute with their 3′→5′ ribonucleolytic activity (reviewed by Chlebowski et al., 2013). We show that RRP6 and EXOSC10 are redistributed in response to DNA damage cues in insect and mammalian cells, respectively, and that a fraction of RRP6/EXOSC10 is rapidly recruited to DSBs. This finding is consistent with data from proteomics studies in which it has been shown that exosome components interact physically with factors involved in the DDR. A systematic analysis of protein–protein interactions involving BRCT-domain-containing proteins, for example, has identified exosome subunits among the interaction partners of mediator of DNA-damage checkpoint 1 (MDC1) (Woods et al., 2012), a protein that interacts with γH2AX near DSBs to promote the recruitment of the repair machinery (Stewart et al., 2003). The results presented here are consistent also with data from a genome-wide siRNA screen in which depletion of EXOSC10 inhibited homologous recombination in a human cell line (Adamson et al., 2012).

We have focused our analysis on RRP6/EXOSC10, but the multisubunit structure of the exosome makes it difficult to establish whether the entire exosome has a role in DNA repair. We could not detect any relocalization of the core exosome subunit RRP4 in response to γIR, and RRP4 depletion did not affect the recruitment of RAD51 to γIR-induced DSBs in S2 cells. These observations suggest that, at least in Drosophila, the involvement of RRP6 in homologous recombination does not involve the exosome core. This possibility would be compatible with previous reports on the core-independent function of RRP6 in cell cycle progression and error-free mitosis in Drosophila (Graham et al., 2009).

RRP6 and EXOSC10 interact with RAD51, as shown by co-immunoprecipitation and PLA experiments. Moreover, depletion of RRP6 in S2 cells or EXOSC10 in HeLa cells prevents the recruitment of RAD51 to DSBs, regardless of whether the damage is induced by radiation or by endonuclease cleavage. By contrast, depletion of EXOSC10 in HeLa cells did not affect the recruitment of the NHEJ factor 53BP1 to the damaged sites. These findings link RRP6/EXOSC10 specifically to the homologous recombination pathway.

The most interesting issue that arises from the results presented here is the molecular mechanism by which RRP6/EXOSC10 contributes to the recruitment of RAD51 to DSBs. RRP6 and EXOSC10 are distributive RNase D-like exonucleases with 3′→5′ activity (reviewed by Houseley et al., 2006; Lykke-Andersen et al., 2011). We constructed a catalytically inactive RRP6 mutant, RRP6-Y361A–V5, and we show that overexpression of RRP6-Y361A–V5 in S2 cells reproduces the effects of RRP6 depletion: delayed repair of γIR-induced DSBs, as inferred from clearance of γH2Av foci; impaired recruitment of RAD51 to DSBs; and increased sensitivity to γIR. Overexpression of RRP6-Y361A–V5 did not affect the cellular level of RAD51 and did not disrupt the interaction of RRP6 with RAD51. Taken together, these observations strongly support the conclusion that the 3′→5′ ribonucleolytic activity of RRP6 is needed for the recruitment of RAD51 to DSBs.

DNA damage induces bidirectional transcription and formation of short RNAs in the vicinity of the damaged sites (reviewed by d'Adda di Fagagna, 2014; Sharma and Misteli, 2013). These short RNAs, referred to as DDR RNAs (‘DDRNAs’) or ‘diRNAs’ (Francia et al., 2012; Wei et al., 2012), are DROSHA- and DICER-dependent RNA products that act together with AGO2 to facilitate the recruitment of RAD51 to DSBs (Gao et al., 2014). The prevalent roles of RRP6/EXOSC10 in the processing of other non-coding RNAs suggest that RRP6 and EXOSC10 might participate in (DDRNA) biogenesis, which would link RRP6/EXOSC10 to DSB repair. However, inhibition of DDRNA biogenesis in mammalian cells affects not only the homologous recombination pathway but also the recruitment of early NHEJ factors such as 53BP1 to DSBs (Francia et al., 2012), whereas our experiments with EXOSC10 depletion failed to reveal defects in the NHEJ pathway. This difference argues against a mechanism in which EXOSC10 is necessary for the biogenesis of DDRNAs.

Recent studies have shown that accumulation of dsRNAs in the chromatin has deleterious effects and leads to genomic instability (Bhatia et al., 2014; White et al., 2014). In the DNA repair context, it is easy to envision that the presence of complementary RNA strands derived from bidirectional transcription near the DSBs could compromise the assembly of RAD51 nucleofilaments on the DNA. Our results are consistent with a hypothetical mechanism in which the 3′→5′ ribonucleolytic activity of RRP6/EXOSC10 is required in the vicinity of the DSBs to degrade RNA molecules that could inhibit the recruitment of homologous recombination factors. This hypothesis is compatible with recent observations on the role of Rrp6 and Trf4 in the loading of RPA to single-stranded DNA (ssDNA) tails generated at DSBs and replication forks in S. cerevisiae (Manfrini et al., 2014).

Culture of D. melanogaster S2 cells and induction of protein expression with CuSO₄ were carried out according to the instructions of the Drosophila Expression System manual from Invitrogen. S2-RRP6–V5 cells were characterized by Hessle et al. (Hessle et al., 2009). HeLa cells were grown at 37°C in Dulbecco's modified Eagle's medium (HyClone, Thermo Scientific) supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin in a humidified incubator at 10% CO₂. DSBs were induced in HeLa cells using a ¹³⁷Cs Gammacell 1000 radiation source (Nordion) at a photon dose rate of 6.4 Gy/min, and in S2 cells using a Scanditronix chamber (Radiation Products Design) at a photon dose rate of 0.4 Gy/min. S2 cells and HeLa cells were irradiated with 10 Gy and 5 Gy, respectively. In some experiments, S2 cells were treated overnight with 25 µg/ml CGK733 (Sigma Aldrich) before irradiation.

S2 cells were knocked down using dsRNA against RRP6 or GFP as a control, exposed to 10 Gy of γIR, left in culture for 72 h, and counted. Non-irradiated samples were cultured in parallel as the control. The cell growth rate in each plate was calculated by dividing the number of cells in the plate at 72 h by the number of cells at 0 h. The growth rate for each condition (RRP6-KD and GFP-KD) was obtained by dividing the growth rate of the irradiated samples by that of the non-irradiated samples. For cell cycle analysis, HeLa cells were stained with 7-aminoactinomycin D and analyzed in a FACSCalibur (BD Biosciences).

A single amino-acid mutation in RRP6-V5 was made by oligonucleotide-directed site-specific mutagenesis using the site-directed mutagenesis kit (Invitrogen) on the pMT-Rrp6 plasmid (Hessle et al., 2009). The resulting protein was given the name RRP6-Y361A–V5. The sequences of the oligonucleotides used were 5′-CAGCTGGTGGACGCCGCTCGTCAGGATAC-3′ (oligonucleotide Ae125) and 5′-GTATCCTGACGAGCGGCGTCCACCAGCTG-3′ (oligonucleotide Ae126). The mutation was verified by DNA sequence analysis. The plasmids for the expression of RRP6–V5 and RRP6-Y361A–V5 were co-transfected into S2 cells together with a hygromycin selection plasmid (Invitrogen), and stable transfectants were selected with 0.3 mg/ml hygromycin. ‘Empty’ control cells were transfected with the hygromycin resistance plasmid alone.

A rabbit anti-V5 antibody (ab9116, Abcam) was used for immunoprecipitation and a mouse anti-V5 (p/n 46-0705, Invitrogen) was used for immunofluorescence and western blotting. The anti-γH2Av antibody was from Rockland, and the anti-AGO2 and anti-RPA were from Abcam (ab57113 and ab97338, respectively). The rabbit antibodies against Drosophila RAD51 (Chiolo et al., 2011) and histone H1 were kindly provided by James Kadonaga (Division of Biological Sciences, University of California San Diego, CA). The anti-RRP6 antibody was a serum from rabbits immunized with the peptide CLVQMSTRSKDYIFDT. Immunization was carried out at Agrisera AB. For detection of NONA in co-immunoprecipitation experiments we used the monoclonal antibody Bj6, kindly provided by Harald Saumweber (Institute of Biology, Humboldt University Berlin, Germany). The following antibodies were used against human proteins: anti-γH2AX from Cell Signaling Technology, mouse and rabbit anti-human-RAD51 from Abcam (ab88572 and ab63801, respectively), anti-RAD51 (ChIP grade; ab76458), anti-EXOSC10 from Santa Cruz Biotechnology (sc-374595), anti-53BP1 from Novus Biologicals (NB100-904), and anti-MYBBP1 from Abcam (ab89121). The non-specific control antibody used for ChIP is an anti-IgG (ab46540). Secondary antibodies conjugated to FITC and Alexa Fluor 594 were from Jackson ImmunoResearch. Horseradish peroxidase (HRP)-conjugated antibodies (Dako) were used for western blotting.

Cells were homogenized in PBS containing 0.2% NP-40 substitute (Sigma), Protease Inhibitor-Complete EDTA Free (Roche), 25 mM β-glycerophosphate (Sigma) and Phostop-Easy Pack (Roche) and the homogenate was centrifuged at 1500 g. The pellet was resuspended in detergent-free buffer, sonicated and supplemented with 0.1% NP-40 substitute (Roche) and 0.1 mg/ml RNase A (Sigma). The sample was centrifuged at 16,000 g and the supernatant was used as input to the immunoprecipitation. Primary antibodies were used at 1–3 µg/ml and the bound proteins were isolated on Sepharose beads.

Proteins were separated by SDS-PAGE using a Mini-Protean II system (BioRad), and transferred to polyvinylidene fluoride membranes (Millipore) in Tris-glycine buffer with 0.02% SDS and 4 M urea using a semi-dry electrophoretic transfer cell (BioRad). The membranes were probed with antibodies following standard procedures. Chemiluminiscent detection was carried out using the ECL kit (GE Healthcare).

The liquid chromatography and mass spectrometry analysis was carried out at the Proteomics Karolinska (PK/KI) facility. Selected lanes from SDS-PAGE gels were cut out, and the proteins were in-gel reduced, alkylated and digested with modified sequence-grade trypsin (Promega). The resulting peptides were extracted, resolved in a Nano-Easy-HPLC (Thermo Scientific) and analyzed using a Q-Exactive instrument (Thermo Scientific) modified with a nanoelectrospray ion source (Thermo Scientific). The proteins contained in each sample were identified against the SwissProt database (Concat_2012.062). The proteins with a coverage value above 2% were selected and are presented in supplementary material Tables S1 and S2.

Long double-stranded RNAs (dsRNAs) were used for RNAi in S2 cells as described by Hessle et al. (Hessle et al., 2009). HeLa cells were seeded in 60-mm plates 1 day before transfection, with a confluence of 30–50%, in 4 ml DMEM supplemented with 10% FBS without penicillin-streptomycin. Transfection was performed with 8 µl of Lipofectamine RNAiMAX (Invitrogen) and 10 nM of siRNA in 1 ml Opti-MEM (Gibco) per plate. The siRNA oligonucleotides used were 5′-GUGCGAGGGGGUUGUAAUCTT-3′ (Scrambled control) and 5′-GCUGCAGCAGAAGAGGCCATT-3′ (EXOSC10-KD).

S2 cells were plated onto polylysine-coated slides (Menzel-Glässer) and allowed to attach before irradiation. HeLa cells were seeded on coverslips. The cells were fixed with 3.7% formaldehyde in PBS for 10 min at room temperature, permeabilized in 0.1% Triton X-100 for 16 min, and washed with PBS before incubation in a solution of 5% BSA in PBS for 40 min. Immunostaining was carried out following standard procedures. The slides were mounted with Vectashield containing 1.5 µg/ml DAPI (Vector Laboratories) and analyzed either in an epifluorescence (Axioplan2, Carl Zeiss) or confocal microscope (LSM780, Carl Zeiss). Confocal images were acquired using a 63× oil immersion objective. The lasers used were 405 nm for DAPI staining, 488 nm argon laser for FITC and 561 nm for Alexa Fluor 594. The thickness of the optical sections was 0.9 µm for Hela and 1.1 µm for S2 cells. Quantitative analyses of the number of cells with foci or the number of foci per cell were carried out in random areas of the preparations. Statistical significance was assessed using paired Student's t-tests. The ‘Colocalization finder’ and ‘Plot profile’ functions of ImageJ 1.42q (Schneider et al., 2012) were used for colocalization analysis and measurement of profile intensities, respectively.

Cells were fixed 10 min after irradiation with 3.7% formaldehyde in PBS for 10 min and permeabilized with 0.1% Triton X-100 for 15 min at room temperature. A blocking solution of 5% BSA was added for 30 min followed by 1 h incubation of the primary antibodies diluted in 1% BSA. The proximity ligation assay (PLA) was carried out using the Duolink PLA in situ kit (Olink) following the manufacturer's protocol. PLA experiments in S2-RRP6–V5 cells were carried out with antibodies against V5 and RAD51. PLA experiments in HeLa cells were carried out with antibodies against endogenous EXOSC10 and RAD51. The preparations were analyzed in an LSM 780 confocal microscope (Carl Zeiss). The ‘In cell’ module of Imaris v.7.7.2 was used for automated quantification of PLA results.

A cDNA corresponding to the open reading frame (ORF) of endonuclease I-PpoI was amplified by PCR using as a template the pICE-HA-NLS-I-PpoI plasmid (Addgene) and oligonucleotides 5′-ATAGGTACCTATGTACCCCTACGACGTGCCC-3′ and 5′-TATCTCGAGTGTTGCACCACGAAGTGGGAGC-3′, and cloned between the KpnI and XhoI sites of the pOPRSVI/MCS vector (Agilent Technologies). The resulting pOPRSVI/MCS-IPpoI plasmid was transfected to HeLa cells using a Lipofectamine 3000 transfection kit (Invitrogen) and the cells were harvested 24 h after the transfection.

HeLa cells were transfected with the pOPRSVI/MCS-IPpoI plasmid and harvested 24 h after transfection. Chromatin was extracted, sonicated in a Bioruptor (Diagenode) and immunoprecipitated with an antibody against EXOSC10 or RAD51 essentially as described by Eberle et al. (Eberle et al., 2010). The DNA in input and immunoprecipitated samples was purified using Zymo-ChIP Clean & Concentrator (Zymo Research) and analyzed by qPCR in a RotorGene (Qiagen) using Power SYBR Green PCR Master Mix (Kapa Biosystems). Two primer-pairs were used to amplify sequences in the proximity of I-PpoI cleavage sites: 5′-AATCAGCGGGGAAAGAAGAC-3′ and 5′-GGGGCCTCCCACTTATTCTA-3′ amplified a sequence located at −128 to −43 bp relative to a I-PpoI cleavage site in chromosome 21, 5′-GTCTTCTTTCCCCGCTGATT-3′ and 5′-CGAGATTCCCACTGTCCCTA-3′ amplified a sequence located at +108 to +195 bp from a I-PpoI cleavage site in chromosome 8. An unrelated region (Rad50) was analyzed in parallel as a control using primers 5′-CACACCCTGTGAGAAACAC-3′ and 5′-CAAGCCACTGCCATCTCTAA-3′. ARPP primers 5′-GCACTGGAAGTCCAACTACT-3′ and 5′-TGAGGTCCTCCTTGGTGAACAC-3′ were used to normalize the samples. Data from two technical replicates of two independent experiments were analyzed. The data from each sample was related to the input of that same experiment and normalized to ARPP.

We thank J. Kadonaga for providing antibodies, S. Böhm and S. Haghdoost for technical help, A. K. Östlund Farrants for constructive discussions and G. W. Farrants for language editing. We are also grateful to D. Rutishauser (PKKI), A.S. Höglund (IFSU) and the Severinson's laboratory (MBW, Stockholm University) for assistance with proteomics, imaging, and FACS, respectively.