MCM8-9 complex promotes resection of double-strand break ends by MRE11-RAD50-NBS1 complex

MCM8-9 complex is required for homologous recombination (HR)-mediated repair of double-strand breaks (DSBs). Here we report that MCM8-9 is required for DNA resection by MRN (MRE11-RAD50-NBS1) at DSBs to generate ssDNA. MCM8-9 interacts with MRN and is required for the nuclease activity and stable association of MRN with DSBs. The ATPase motifs of MCM8-9 are required for recruitment of MRE11 to foci of DNA damage. Homozygous deletion of the MCM9 found in various cancers sensitizes a cancer cell line to interstrand-crosslinking (ICL) agents. A cancer-derived point mutation or an SNP on MCM8 associated with premature ovarian failure (POF) diminishes the functional activity of MCM8. Therefore, the MCM8-9 complex facilitates DNA resection by the MRN complex during HR repair, genetic or epigenetic inactivation of MCM8 or MCM9 are seen in human cancers, and genetic inactivation of MCM8 may be the basis of a POF syndrome.

D ouble-strand break (DSB) repair is essential for the maintenance of DNA integrity 1 . Deregulation of this process leads to significant genetic instability, which can result in the development of tumours 2 . DSB repair systems are largely classified into homologous recombination (HR) and non-homologous end joining (NHEJ). In the first step of HR repair, MRN (MRE11-RAD50-NBS1) complex and C-terminal binding protein (CtBP)-interacting protein (CtIP) recognize DNA breaks, resect single-stranded DNA (ssDNA) together with BLM/Dna2 and Exo1, and thus generate a long stretch of 3 0 -overhanging ssDNA [3][4][5][6] . Following DNA resection, RPA proteins are recruited to the ssDNA to stabilize the structure, and mediator proteins, including Rad51, Rad52 and BRCA2, promote the formation of Rad51 filaments 7 . Recent papers show that the endonuclease activity of MRE11 in the MRN complex distinguishes HR from NHEJ 8,9 . Interstrand-crosslinking (ICL)-inducing chemotherapy agents such as cisplatin and mitomycin C produce lesions that are repaired by HR 10,11 . Hence, inactivation of HR makes cells very sensitive to ICL adducts, and cancers defective in HR (for example, those with BRCA1 or BRCA2 mutations) are good targets for treatment with cisplatin or mitomycin C. HR, as an important part of meiosis, is also very important for the generation of germ cells.
Even though MCM8-9 proteins have initially been identified as components of pre-replication complexes, recent findings counter the original suggestion that MCM8-9 is essential for DNA replication. The Xenopus MCM8 clearly shows DNA helicase activity in vitro 12 , and mutants of the Drosophila MCM8 homologue, REC, have meiotic crossover defects 13 . However, mice with homozygous deletions of MCM9 are viable and fertile, albeit with some deficits in the germ cell lineage, a lineage notable for meiosis 14 . Mouse and chicken cell lines with deletions of MCM8 or MCM9 are viable, although more sensitive to cisplatin [15][16][17] . A single-nucleotide polymorphism (SNP; rs16991615) that leads to an amino-acid change from glutamic acid (Glu) to lysine (Lys) in the gene MCM8 was found in genome-wide association analysis to be significantly correlated with age at natural menopause 18 . As a whole, the evidence suggests that MCM8-9 is more important for HR than for DNA replication, but the exact role of MCM8-9 in HR is unclear.
Here we show that MCM8-9 is essential for DNA resection by the MRN complex at DSBs, and is required for proper localization of the MRN complex to the DSBs. In addition, a cancer cell line having homozygous deletion of the MCM9 locus exhibits inefficient HR and high sensitivity to ICL reagents. Inactivation of the MCM8-9 complex is seen in cancers and in a premature ovarian failure (POF). Genetic or epigenetic inactivation of MCM8-9 is a new mechanism by which cancers can blunt the HR repair pathway, and as in cancers with mutations in BRCA1 or BRCA2, cancers that inactivate MCM8-9 are susceptible to chemotherapy agents that target cells with blunted HR.

Results
MCM8-9 is required for ssDNA generation in HR repair. Deficiency of MCM8 or MCM9 impairs the recruitment of Rad51 on chromatin after DNA damage 15,17 . DNA resection by the MRN complex, CtIP, EXO1 and DNA2, produces the ssDNA that has to be coated by RPA before subsequent Rad51 assembly 19,20 . We thus measured the accumulation of RPA at DNA damage sites in MCM8-or MCM9-depleted cells. As reported previously, depletion of MCM8 decreases MCM9, but depletion of MCM9 does not deplete MCM8 (Fig. 1a). Even though levels of RPA70 were unchanged, there was a decrease in cisplatin-induced RPA foci formation on MCM8 or MCM9 depletion (Fig. 1a,b). HeLa DR13-9 cells contain a single I-SceI cut site integrated into their genome that is repaired by HR after cleavage by the I-SceI 21 . Although the expression level of I-SceI was not diminished in MCM8-or MCM9-depleted cells ( Supplementary Fig. 1), RPA binding to the I-SceI cut site was decreased in cells depleted of MCM8 or MCM9 compared with control short interfering RNA (siRNA; siGL2)-transfected cells (Fig. 1c). To eliminate the possibility of off-target effects of siRNAs, we rescued cisplatin-induced RPA foci formation in siMCM8transfected cells by stably expressing siRNA-resistant Flag-tagged MCM8 (Flag-MCM8r; Fig. 1d,e). H2AX phosphorylation on S139, and by inference the number of DNA breaks, was not affected by MCM8-9 depletion, suggesting that MCM8-9 specifically affected a step after DNA break formation.
To directly examine whether MCM8-9 is required for ssDNA generation, the generation of ssDNA at DSB sites was directly visualized by immunostaining with anti-BrdU (5-bromodeoxyuridine) antibody without denaturation of the double-stranded DNA after labelling of genomic DNA with BrdU in a previous cell cycle. Cisplatin increased the number of cells with BrdU signal, but this was decreased by depletion of MCM8 or MCM9 (Fig. 2a). Similarly, cisplatin treatment of mouse embryonic fibroblast (MEF) cells from Mcm9-null (XG/XG) mice produced fewer ssDNA-positive cells compared with wild-type (WT) MEFs (Fig. 2b). Cell cycle profile of the MEF cells from XG/XG mice was not changed by the 4-h cisplatin treatment, suggesting that the decrease in ssDNA formation could not be explained by a change in S phase progression in these cells ( Supplementary  Fig. 2). Next, we utilized a new assay for quantitatively measuring ssDNA at regions adjoining a specific DSB site induced by the restriction enzyme AsiSI in the estrogen receptor (ER)-AsiSI U2OS cells 22 . The formation of ssDNA makes BsrG1 sites in the adjoining DNA resistant to digestion by that enzyme, thus allowing us to measure whether the end resection complex has digested past each of the three BsrGI site. DSB induced by 4-hydroxytamoxifen (4-OHT) treatment resulted in more ssDNA at the BsrG1 site closer to AsiSI cut site, but at both this site and the next, depletion of MCM8 or MCM9 decreased the generation of ssDNA (Fig. 2c). Depletion of MRE11 also affected ssDNA formation at these sites, validating the assay. Therefore MCM8-9 is required for ssDNA formation at DSBs before HR.
MCM8-9 is required for MRN localization to HR repair sites. To screen for the nuclease(s) requiring MCM8-9 activity for ssDNA formation, we first tested which nucleases were required for RPA focus formation after cisplatin treatment. The knockdown of MRE11, but not EXO1 and/or DNA2, suppressed cisplatin-induced RPA foci formation to the same extent as seen after MCM8-9 depletion (Fig. 3a,b and Supplementary Fig. 3a,b). This result suggested that MCM8-9 might work during the initial DNA resection alongside the MRN complex. Therefore, we examined whether MRN and MCM8-9 physically associate with each other. Indeed, MCM9 immunoprecipitates contained not only MCM8 but also MRE11, NBS1 and RAD50 (Fig. 3c, left), and conversely RAD50 immunoprecipitates contained not only its cofactors, NBS1 and MRE11, but also MCM8-9 (Fig. 3c, right). The co-immunoprecipitation was performed in the presence of the DNA intercalating chemical, ethidium bromide (EtBr), suggesting that the MCM8-9 and MRN did not interact via a bridging DNA molecule. Furthermore, MRE11 foci generated by cisplatin treatment co-localized with Flag-MCM8 foci (Fig. 3d). Chromatin immunoprecipitation (ChIP) assays on HeLa DR13-9 cells showed that all of the MRN components normally accumulated on the I-SceI cut site following the expression of I-SceI, but this association was impaired on depletion of MCM8 or MCM9 (Fig. 3e). In a parallel experiment, siMCM8 or siMCM9 decreased the number of cisplatin-induced foci formed by all MRN complex members (Fig. 3f). On the other hand, the recruitment of CtIP to cisplatin induced foci and to I-SceI cut sites was not affected by depletion of MCM8-9 ( Supplementary  Fig. 4a,b). Therefore, in multiple experimental systems, MCM8-9 is required for the proper localization of the MRN complex to DSB sites for DNA resection.
ATPase activity of MCM9 is essential for the function of MRN. MCM8 and MCM9 are related in sequence to all of the components of the MCM2-7 complex, which unwinds duplex DNA during DNA replication 23 . Human MCM8 and MCM9 contain MCM domains marked by the Walker A (WA) and Walker B (WB) motifs important for ATP binding and hydrolysis. Therefore, we tested whether the ATPase activities of MCM8 or MCM9 were essential for DNA resection in HR repair. U2OS cell lines stably expressing WA or WB motif mutants of siRNA-resistant MCM8 or MCM9 were used. On depletion of endogenous MCM8 or MCM9 by siRNA, WT MCM8 and MCM9 (MCM8r-WT and MCM9r-WT) restored the RPA-and MRE11 foci on cisplatin treatment (Fig. 4a,b). The WA-or WB motif mutants of MCM8 or MCM9 were unable to restore these foci.
Because knockdown of MCM8 decreased both MCM8 and MCM9, while knockdown of MCM9 decreases that protein alone, we focused on the MCM9 mutants from here on. The recruitment of the DNA damage response kinase ATR/ATRIP to damaged DNA is mediated through its interaction with RPA, which is necessary for proper checkpoint activation [24][25][26] . Consistent with a role of MCM9 in ssDNA formation and RPA recruitment, phospho-Chk1 was decreased in siMCM9-transfected cells, and phospho-Chk1 was restored following expression of MCM9r-WT, but not WA-or WB mutants of MCM9 ( Supplementary Fig. 5). The WA-or WB mutants of MCM9 may even have a dominant-negative effect on checkpoint activation, perhaps because they interact with endogenous MCM8 and suppress any residual activity from MCM8 alone. The sensitivity of the cells to cisplatin was increased by siMCM9, restored by WT MCM9, but not by WA-or WB mutants of MCM9, suggesting that the ATPase activity of MCM9 is critical for cell survival following treatment with ICL-inducing agents  ( Fig. 4c and Supplementary Fig. 6). WT MCM9 partly restored HR efficiency to HeLa DR13-9 cells depleted of endogenous MCM9, whereas WA-or WB mutants of MCM9 did not (Fig. 4d).
Next, we examined whether the ATPase activity of MCM9 is necessary for MRE11 recruitment to DSBs. In ChIP experiments in HeLa DR13-9 cells, MRE11 recruitment to I-SceI cut sites was not rescued by MCM9 with mutations in WA or B motifs (Fig. 4e). The co-immunoprecipitation of MRE11 with MCM9 was also inhibited by the same mutations (Fig. 4f). Furthermore, in vitro nuclease assay on linearized pUC19 plasmid using purified HA-MCM9 from HeLa DR13-9 cells suggested that only MCM9 WT was associated with a strong nuclease activity (Fig. 4g).
MRE11 endonuclease initiates resection at DSBs before HR 8 . We purified the MRN complex from U2OS cells stably expressing FLAG-NBS1 and tested its endonuclease activity on circular ØX174 ssDNA ( Fig. 5a and Supplementary Fig. 7a). MCM8 knockdown decreased the endonuclease activity of   Fig 7b). Thus, the nuclease activity in the MRN immunoprecipitate was mainly due to MRE11, although we cannot rule out the presence of other contaminating endonucleases. Next, we purified recombinant Xenopus MCM8-MCM9 WT (WT/WT) from baculovirus-infected insect cells to investigate whether MCM8-9 can stimulate MRE11 endonuclease in vitro (Fig. 5b). MCM8 WT MCM9 WA mutant (WT/WA) complexes were to be used as inactive controls. We were thwarted from doing the experiment by a nuclease that was associated with the purified WT/WT MCM8-9 but not with the inactive WT/WA MCM8-9 (Fig. 5c). Since the WA mutation of human MCM9 decreased its association with human MRE11 (Fig. 4f), we wondered whether insect MRE11 was co-purifying with the Xl-WT/WT MCM8-9 complex. Remarkably, a 65-kDa protein, corresponding in size to Drosophila MRE11 and recognized by anti-MRE11 antibody, was associated with WT/WT but not WT/ WA MCM8-9 complex (Fig. 5d, the B lanes). Anti-MRE11 antibody immunodepleted the 65-kDa protein (Fig. 5d, the A and I lanes) and decreased the nuclease activity associated with MCM8-9 (Fig. 5e). Several controls showed that the WA mutation of MCM9 did not decrease MCM8-9 complex formation ( Supplementary Fig. 8), but decreased the ATPase activity of MCM8-9 ( Supplementary Fig. 9), and that the MRE11 antibody decreased the nuclease associated with the recombinant Xl-MCM8-9 without decreasing the amount of MCM8-9 proteins (Supplementary Fig. 10).
These results suggest that ATPase activity of MCM8-9 complex is required for ssDNA generation at DSBs as well as optimal nuclease activity of the MRN. Also, ATP binding and hydrolysis by the MCM9 in the MCM8-9 complex is essential for maximal association with Mre11 and for the recruitment of MRN to DNA damage sites.  Table 1). Several cancer cell lines, including a non-small cell lung cancer cell line, NCI-H2291, also have a homozygous deletion of the MCM9 locus. To test whether the cancer-specific loss of MCM9 affected HR repair, we compared the ability of NCI- (c) WT/WT MCM8-9 has more nuclease activity than WT/WA MCM8-9. The in vitro nuclease assay was performed using indicated amounts of the recombinant MCM8-9 or bovine serum albumin (BSA). (d) Immunodepletion of MRE11, as detected by immunoblot, from purified recombinant MCM8-9. Immunodepletion was done by incubating anti-human MRE11 antibody with indicated recombinant proteins. B, before immunodepletion; A, after immunodepletion; I, the total immunoprecipitate. (e) Reduced nuclease activity of WT/WT MCM8-9 after immunodepletion of MRE11 (A), compared with that before immunodepletion (B). Amount of full-length linear DNA remaining was quantified after 60 min of an in vitro nuclease assay with 5 nM of BSA or purified MCM8-9 before or after immunodepletion of MRE11. The y axis shows the ratio of the residual substrate relative to that at the 0-min point. ***Po0.005, *Po0.05; Student's t-test. All error bars represent s.d. of the mean from triplicates. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8744 ARTICLE H2291 and NCI-H1299, a control non-small cell lung cancer cell line, to cope with cisplatin-induced DNA damage. Although protein levels of MCM8 and other HR repair factors were similar, MCM9 protein was not detected in NCI-H2291 (Fig. 6a). The NCI-H2291 cells showed fewer cells positive for RPA and MRE11 foci following cisplatin treatment (Fig. 6b). Furthermore, NCI-2291 showed significant defect on HR efficiency compared with NCI-1299 in an HR assay based on transient transfection of the HR reporter and I-SceI expressing plasmids (Fig. 6c). Complementation with hemagglutinin (HA)-tagged WT MCM9 (HA-MCM9-WT; Fig. 6d) rescued RPA foci formation (nuclei with HA-MCM9 in Supplementary Fig. 11). Clonogenic assays showed that the deletion of MCM9 in NCI-H2291 cells also made them more sensitive to cisplatin than those complemented with HA-MCM9-WT (Fig. 6e). Thus, deletion of MCM9 produces a functional defect of HR repair in this cancer cell line. In addition, we found that several prostate cancer cell lines do not have deletions in MCM9, but express lower levels of the MCM9 protein compared with HEK293T cells (Fig. 6f). The cisplatin resistance of these prostate cancer cell lines was remarkably correlated to the amount of MCM9 protein expressed ( Fig. 6g and Supplementary  Fig. 12). Therefore, epigenetic suppression of MCM9 could also predispose cancer cells to cisplatin sensitivity.

Functional inactivation of MCM8 by mutation in cancer or POF.
Finally, we turned to a lung squamous cell carcinoma in CBioPortal, with a point mutation that changed proline 456 in the WA motif of MCM8 to alanine (Fig. 7a, the upper panel). Functional studies in a cell line stably expressing siRNA-resistant MCM8 P456A showed that this mutation inactivated MCM8 because there was a decrease of cisplatin-induced RPA70-and MRE11 foci on knockdown of endogenous MCM8 (Fig. 7a, lower  panel). Thus, genetic and epigenetic inactivation of MCM8-9 is seen in diverse human cancers, and such inactivation is associated with sensitivity to therapy by ICL agents.
Genome-wide association studies (GWAS) identified an SNP on MCM8 leading to an amino-acid change (E341K) that is associated with an early age of natural menopause 18 (a POF syndrome). MCM8 E341K was not able to rescue the RPA70-and MRE11 foci formation seen after depletion of endogenous MCM8 (Fig. 7b,c) nor could it fully restore the sensitivity of the MCM8depleted cells to cisplatin (Fig. 7d). Consistent with a previous result 31 , the naturally occurring SNP of MCM8 associated with POF impairs the function of MCM8 in HR repair. Since HR repair is an essential function of meiosis, mutational inactivation of MCM8 may explain the genetic basis of this POF syndrome.

Discussion
DNA end resection is more pronounced in HR than in classical NHEJ 1 . In this study, we show that MCM8-9 is required for ssDNA generation at DSBs using the following three independent approaches: (a) detection of RPA70, (b) exposure of BrdU-labelled DNA and (c) quantitative PCR (qPCR)-based assay to measure ssDNA adjoining a DSB (Figs 1 and 2). Several DNA nucleases have been implicated in the resection: MRN complex and CtIP for the initial resection/EXO1 and DNA2 for the longer stretches of ssDNA 3,32-36 . It has also been suggested that MRE11 endonuclease initiates resection followed by bidirectional exonuclease activity of MRE11 and EXO1 to generate ssDNA for HR 8 . In our hands, EXO1 and DNA2 appear dispensable for RPA focus formation after cisplatin treatment, in contrast to the requirement of MRE11 ( Fig. 3b and Supplementary Fig. 3b), perhaps because the secondary and extensive DNA resection attributed to EXO1 and DNA2 is slower than that caused by MRE11, and is not necessary for forming RPA foci. This does not mean that EXO1 and DNA2 are totally dispensable for resection in vivo. However, the parallel effects of MCM8-9 and MRE11 depletion on RPA focus formation is consistent with our suggestion that MCM8-9 is particularly involved in the initial resection by MRE11.
MCM8-9 recruits or promotes the stable association of MRN to the DSB (Fig. 3), without much effect on CtIP recruitment. This suggests that CtIP can be recruited to DSBs independent of MRN, which is in line with a previous report 37 . Also, since the MRN complex directly binds to DNA ends in vitro 38 , these results suggest either that MCM8-9 is required in cells on top of the direct DNA binding to recruit MRN, or that MCM8-9 prevents the detachment of active MRN after it has been recruited. MCM8-9 has highly conserved ATPase motifs and shows a helicase activity in vitro 12,39 . We show that mutation of its ATPase motif causes loss of interaction with MRN complex, increase of cellular sensitivity to cisplatin as well as decrease of HR efficiency (Figs 4 and 5). If anything, the mutant forms of MCM9 appear to have a dominant-negative effect on checkpoint activation (an indirect measure of ssDNA formation), HR and cell survival after cisplatin treatment ( Supplementary Fig. 5 and Fig. 4c,d). The WA-or WB mutants of MCM9 most likely associate with endogenous MCM8 or other endogenous proteins to inactivate them further than seen after simple depletion of MCM9.
MCM8-9 ATPase could be a new therapeutic target whose inhibition would impair HR and augment the efficacy of ICL-inducing agents during chemotherapy. However, it remains to be elucidated how ATPase activity of MCM8-9 promotes the functional activity of MRN. Considering that DNA helicases utilize ATP hydrolysis as an energy source 40 , MCM8-9 may function as a helicase in its support of MRN. Alternatively, the physical association of MCM9 with MRE11 is impaired by the mutations in the WA or B motifs of MCM9 (Fig. 4f), suggesting that ATP binding or hydrolysis is important for the protein-protein interaction between these two complexes. Although MCM8-9 has been shown to associate weakly with RAD51 (ref. 17), the results we present here argue against that being the primary function of MCM8-9 in the loading of RAD51. Instead, MCM8-9 is clearly involved in preparing the ssDNA substrate that eventually loads RAD51.
Impaired HR makes cancers more sensitive to ICL-inducing agents. We show that a naturally occurring homozygous deletion of the MCM9 sensitizes a cancer cell line to ICL reagents and leads to the HR defect (Fig. 6). The expression level of the MCM9 protein is correlated to the cisplatin resistance of some prostate cancer cell lines (Fig. 6f,g) and a cancer-derived mutation of MCM8 inactivated MCM8 (Fig. 7a). Thus, the deletion, point mutation or reduced expression of MCM8 or MCM9 in cancers that we report here should be tested in patients as a biomarker for predicting a cancer's sensitivity to ICL-inducing agents such as cisplatin or mitomycin C. As a promoter of HR, MCM8 or MCM9 may serve as a bona fide tumour suppressor similar to other genes important for HR, BRCA1 and BRCA2. A cancerderived point mutation and a naturally occurring SNP on MCM8 associated with POF diminish the functional activity of MCM8 (Fig. 7). Together, with a very recent study that appeared while this paper was under review 31 , our results explain how the SNP in MCM8 impairs HR, an essential function in the germ line, and thus could lead to a genetically determined POF syndrome.
Establishment of stable cell lines. The siRNA-resistant WT or mutant of MCM8 and MCM9 was inserted into retrovirus vector pBabe-puro and pLHCX (Clontech), respectively. Retroviral vectors, together with retroviral packaging vector, were transfected into 293T cells using Lipofectamin 2000 (Invitrogen) according to the manufacturer's instruction. At 48 h after transfection, viral culture supernatants were harvested and filtered through a 0.45-mm filter, and added to the U2OS or HeLa DR13-9 cells in the presence of 8 mg ml À 1 polybrene (Sigma-Aldrich). After 48 h of infection, drug selection was carried out with either 100 mg ml À 1 hygromycin B (Sigma-Aldrich) or 2 mg ml À 1 puromycin (Sigma-Aldrich) to select retrovirus-infected cells over a 10-day period.